Hugging Face's logo

rpanchum
/
jais-13b-int4-sym-ov

Text Generation
OpenVINO
English
Arabic
jais
optimized
int4
awq
bilingual
arabic
english
custom_code
Model card Files Community
rpanchum commited on
Commit
b84811c
·
verified ·
1 Parent(s): 142ce80

Upload folder using huggingface_hub

Browse files
Files changed (13) hide show
  1. README.md +103 -0
  2. config.json +40 -0
  3. generation_config.json +7 -0
  4. openvino_config.json +28 -0
  5. openvino_detokenizer.bin +3 -0
  6. openvino_detokenizer.xml +335 -0
  7. openvino_model.bin +3 -0
  8. openvino_model.xml +0 -0
  9. openvino_tokenizer.bin +3 -0
  10. openvino_tokenizer.xml +648 -0
  11. special_tokens_map.json +23 -0
  12. tokenizer.json +0 -0
  13. tokenizer_config.json +19 -0
README.md ADDED
@@ -0,0 +1,103 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Jais-13B OpenVINO INT4
2
+
3
+ This repository contains the [inceptionai/jais-13b](https://huggingface.co/inceptionai/jais-13b) model optimized for inference with Intel's OpenVINO runtime. The model has been quantized to INT4 using the AWQ quantization scheme for improved performance while maintaining quality.
4
+
5
+ ## Model Details
6
+
7
+ * **Original Model**: [inceptionai/jais-13b](https://huggingface.co/inceptionai/jais-13b)
8
+ * **Model Type**: Bilingual (Arabic-English) Large Language Model
9
+ * **Parameters**: 13B
10
+ * **OpenVINO Version**: 2024.0+
11
+ * **Quantization**: INT4 Symmetric AWQ (Activation-aware Weight Quantization)
12
+ * **Group Size**: -1 (per-channel quantization)
13
+
14
+ Jais-13B is a bilingual model that supports both Arabic and English text generation. The model can:
15
+ - Generate fluent text in both Arabic and English
16
+ - Respond to prompts in either language
17
+ - Handle code-switching between the two languages
18
+
19
+ ## Optimization Details
20
+
21
+ This model was converted from the original Hugging Face model to OpenVINO format using the Optimum Intel library. The following optimization command was used:
22
+
23
+ ```bash
24
+ optimum-cli export openvino \
25
+ -m inceptionai/jais-13b \
26
+ --weight-format int4 \
27
+ --sym \
28
+ --dataset auto \
29
+ --awq \
30
+ --group-size -1 \
31
+ --trust-remote-code \
32
+ jais-13b-int4-sym-ov
33
+ ```
34
+
35
+ ### Optimization Parameters:
36
+ - **INT4 Quantization**: Weights compressed to 4-bit integers
37
+ - **Symmetric Quantization**: Using symmetric quantization for better accuracy
38
+ - **AWQ**: Activation-aware Weight Quantization to preserve model quality
39
+ - **Auto Dataset**: Used automatic dataset sampling for calibration
40
+ - **Group Size**: -1 (quantize each output channel independently)
41
+ - **Trust Remote Code**: Enabled to support custom model code
42
+
43
+
44
+ ## Usage
45
+
46
+ ### Prerequisites
47
+ - OpenVINO 2024.0 or newer
48
+ - optimum-intel
49
+ - transformers
50
+
51
+ ### Sample Inference code with Optimum Intel
52
+
53
+ ```python
54
+ from optimum.intel import OVModelForCausalLM
55
+ from transformers import AutoTokenizer
56
+
57
+ # Load tokenizer and model
58
+ model_id = "rpanchum/jais-13b-int4-sym-ov"
59
+ tokenizer = AutoTokenizer.from_pretrained(model_id)
60
+ model = OVModelForCausalLM.from_pretrained(model_id)
61
+
62
+ # Generate text
63
+ prompt = "Write a short story about a robot learning to paint:"
64
+ input_ids = tokenizer(prompt, return_tensors="pt")
65
+ output = model.generate(
66
+ **input_ids,
67
+ max_new_tokens=512,
68
+ temperature=0.7,
69
+ top_p=0.9,
70
+ )
71
+ response = tokenizer.decode(output[0], skip_special_tokens=True)
72
+ print(response)
73
+ ```
74
+
75
+ ### Alternative: Using OpenVINO GenAI
76
+
77
+ 1. Install packages required for using OpenVINO GenAI.
78
+ ```bash
79
+ pip install openvino-genai huggingface_hub
80
+ ```
81
+
82
+ 2. Download model and run inference.
83
+
84
+ ```python
85
+ import huggingface_hub as hf_hub
86
+
87
+ model_id = "rpanchum/jais-13b-int4-sym-ov"
88
+ model_path = "jais-13b-int4-sym-ov"
89
+
90
+ hf_hub.snapshot_download(model_id, local_dir=model_path)
91
+
92
+ import openvino_genai as ov_genai
93
+
94
+ device = "CPU"
95
+ pipe = ov_genai.LLMPipeline(model_path, device)
96
+ print(pipe.generate("ما هو الذكاء الاصطناعي؟", max_length=200)) # "What is AI?" in Arabic
97
+ print(pipe.generate("What is artificial intelligence?", max_length=200))
98
+ ```
99
+
100
+
101
+ ## License
102
+
103
+ This model inherits the license of the original [inceptionai/jais-13b](https://huggingface.co/inceptionai/jais-13b) model.
config.json ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "inceptionai/jais-13b",
3
+ "activation_function": "swiglu",
4
+ "architectures": [
5
+ "JAISLMHeadModel"
6
+ ],
7
+ "attn_pdrop": 0.0,
8
+ "auto_map": {
9
+ "AutoConfig": "inceptionai/jais-13b--configuration_jais.JAISConfig",
10
+ "AutoModel": "inceptionai/jais-13b--modeling_jais.JAISModel",
11
+ "AutoModelForCausalLM": "inceptionai/jais-13b--modeling_jais.JAISLMHeadModel",
12
+ "AutoModelForQuestionAnswering": "inceptionai/jais-13b--modeling_jais.JAISForQuestionAnswering",
13
+ "AutoModelForSequenceClassification": "inceptionai/jais-13b--modeling_jais.JAISForSequenceClassification",
14
+ "AutoModelForTokenClassification": "inceptionai/jais-13b--modeling_jais.JAISForTokenClassification"
15
+ },
16
+ "bos_token_id": 0,
17
+ "embd_pdrop": 0.0,
18
+ "embeddings_scale": 14.6,
19
+ "eos_token_id": 0,
20
+ "initializer_range": 0.02,
21
+ "layer_norm_epsilon": 1e-05,
22
+ "model_type": "jais",
23
+ "n_embd": 5120,
24
+ "n_head": 40,
25
+ "n_inner": 13653,
26
+ "n_layer": 40,
27
+ "n_positions": 2048,
28
+ "pad_token_id": 0,
29
+ "position_embedding_type": "alibi",
30
+ "reorder_and_upcast_attn": false,
31
+ "resid_pdrop": 0.0,
32
+ "scale_attn_by_inverse_layer_idx": false,
33
+ "scale_attn_weights": true,
34
+ "scale_qk_dot_by_d": true,
35
+ "torch_dtype": "float32",
36
+ "transformers_version": "4.48.3",
37
+ "use_cache": true,
38
+ "vocab_size": 84992,
39
+ "width_scale": 0.11100000000000002
40
+ }
generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 0,
4
+ "eos_token_id": 0,
5
+ "pad_token_id": 0,
6
+ "transformers_version": "4.48.3"
7
+ }
openvino_config.json ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "compression": null,
3
+ "dtype": "int4",
4
+ "input_info": null,
5
+ "optimum_version": "1.24.0",
6
+ "quantization_config": {
7
+ "all_layers": null,
8
+ "backup_precision": null,
9
+ "bits": 4,
10
+ "dataset": "auto",
11
+ "gptq": null,
12
+ "group_size": -1,
13
+ "ignored_scope": null,
14
+ "lora_correction": null,
15
+ "num_samples": null,
16
+ "processor": null,
17
+ "quant_method": "awq",
18
+ "ratio": 1.0,
19
+ "scale_estimation": null,
20
+ "sensitivity_metric": null,
21
+ "sym": true,
22
+ "tokenizer": null,
23
+ "trust_remote_code": true,
24
+ "weight_format": "int4"
25
+ },
26
+ "save_onnx_model": false,
27
+ "transformers_version": "4.48.3"
28
+ }
openvino_detokenizer.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:01dd38c159cba9b1a2ad588bddce13bdc30b746e6b9e77ca8f0a03de824e8575
3
+ size 1068648
openvino_detokenizer.xml ADDED
@@ -0,0 +1,335 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0"?>
2
+ <net name="detokenizer" version="11">
3
+ <layers>
4
+ <layer id="0" name="Parameter_1440876" type="Parameter" version="opset1">
5
+ <data shape="?,?" element_type="i64" />
6
+ <output>
7
+ <port id="0" precision="I64" names="Parameter_1440876">
8
+ <dim>-1</dim>
9
+ <dim>-1</dim>
10
+ </port>
11
+ </output>
12
+ </layer>
13
+ <layer id="1" name="Convert_1440897" type="Convert" version="opset1">
14
+ <data destination_type="i32" />
15
+ <input>
16
+ <port id="0" precision="I64">
17
+ <dim>-1</dim>
18
+ <dim>-1</dim>
19
+ </port>
20
+ </input>
21
+ <output>
22
+ <port id="1" precision="I32">
23
+ <dim>-1</dim>
24
+ <dim>-1</dim>
25
+ </port>
26
+ </output>
27
+ </layer>
28
+ <layer id="2" name="Constant_1440855" type="Const" version="opset1">
29
+ <data element_type="u8" shape="1068575" offset="0" size="1068575" />
30
+ <output>
31
+ <port id="0" precision="U8">
32
+ <dim>1068575</dim>
33
+ </port>
34
+ </output>
35
+ </layer>
36
+ <layer id="3" name="StringTensorUnpack_1440856" type="StringTensorUnpack" version="extension">
37
+ <data mode="begins_ends" />
38
+ <input>
39
+ <port id="0" precision="U8">
40
+ <dim>1068575</dim>
41
+ </port>
42
+ </input>
43
+ <output>
44
+ <port id="1" precision="I32">
45
+ <dim>-1</dim>
46
+ </port>
47
+ <port id="2" precision="I32">
48
+ <dim>-1</dim>
49
+ </port>
50
+ <port id="3" precision="U8">
51
+ <dim>-1</dim>
52
+ </port>
53
+ </output>
54
+ </layer>
55
+ <layer id="4" name="Constant_1440880" type="Const" version="opset1">
56
+ <data element_type="i32" shape="1" offset="1068575" size="4" />
57
+ <output>
58
+ <port id="0" precision="I32">
59
+ <dim>1</dim>
60
+ </port>
61
+ </output>
62
+ </layer>
63
+ <layer id="5" name="Constant_1440878" type="Const" version="opset1">
64
+ <data element_type="i32" shape="1" offset="1068575" size="4" />
65
+ <output>
66
+ <port id="0" precision="I32">
67
+ <dim>1</dim>
68
+ </port>
69
+ </output>
70
+ </layer>
71
+ <layer id="6" name="Constant_1440877" type="Const" version="opset1">
72
+ <data element_type="i32" shape="1" offset="1068579" size="4" />
73
+ <output>
74
+ <port id="0" precision="I32">
75
+ <dim>1</dim>
76
+ </port>
77
+ </output>
78
+ </layer>
79
+ <layer id="7" name="Constant_1440879" type="Const" version="opset1">
80
+ <data element_type="i32" shape="1" offset="1068583" size="4" />
81
+ <output>
82
+ <port id="0" precision="I32">
83
+ <dim>1</dim>
84
+ </port>
85
+ </output>
86
+ </layer>
87
+ <layer id="8" name="Constant_1440882" type="Const" version="opset1">
88
+ <data element_type="i64" shape="1" offset="1068587" size="8" />
89
+ <output>
90
+ <port id="0" precision="I64">
91
+ <dim>1</dim>
92
+ </port>
93
+ </output>
94
+ </layer>
95
+ <layer id="9" name="Slice_1440881" type="Slice" version="opset8">
96
+ <input>
97
+ <port id="0" precision="I32">
98
+ <dim>1</dim>
99
+ </port>
100
+ <port id="1" precision="I32">
101
+ <dim>1</dim>
102
+ </port>
103
+ <port id="2" precision="I32">
104
+ <dim>1</dim>
105
+ </port>
106
+ <port id="3" precision="I32">
107
+ <dim>1</dim>
108
+ </port>
109
+ <port id="4" precision="I64">
110
+ <dim>1</dim>
111
+ </port>
112
+ </input>
113
+ <output>
114
+ <port id="5" precision="I32">
115
+ <dim>1</dim>
116
+ </port>
117
+ </output>
118
+ </layer>
119
+ <layer id="10" name="VocabDecoder_1440883" type="VocabDecoder" version="extension">
120
+ <data skip_tokens="" />
121
+ <input>
122
+ <port id="0" precision="I32">
123
+ <dim>-1</dim>
124
+ <dim>-1</dim>
125
+ </port>
126
+ <port id="1" precision="I32">
127
+ <dim>-1</dim>
128
+ </port>
129
+ <port id="2" precision="I32">
130
+ <dim>-1</dim>
131
+ </port>
132
+ <port id="3" precision="U8">
133
+ <dim>-1</dim>
134
+ </port>
135
+ <port id="4" precision="I32">
136
+ <dim>1</dim>
137
+ </port>
138
+ </input>
139
+ <output>
140
+ <port id="5" precision="I32">
141
+ <dim>-1</dim>
142
+ </port>
143
+ <port id="6" precision="I32">
144
+ <dim>-1</dim>
145
+ </port>
146
+ <port id="7" precision="I32">
147
+ <dim>-1</dim>
148
+ </port>
149
+ <port id="8" precision="I32">
150
+ <dim>-1</dim>
151
+ </port>
152
+ <port id="9" precision="U8">
153
+ <dim>-1</dim>
154
+ </port>
155
+ </output>
156
+ </layer>
157
+ <layer id="11" name="FuzeRagged_1440884" type="FuzeRagged" version="extension">
158
+ <input>
159
+ <port id="0" precision="I32">
160
+ <dim>-1</dim>
161
+ </port>
162
+ <port id="1" precision="I32">
163
+ <dim>-1</dim>
164
+ </port>
165
+ <port id="2" precision="I32">
166
+ <dim>-1</dim>
167
+ </port>
168
+ <port id="3" precision="I32">
169
+ <dim>-1</dim>
170
+ </port>
171
+ </input>
172
+ <output>
173
+ <port id="4" precision="I32">
174
+ <dim>-1</dim>
175
+ </port>
176
+ <port id="5" precision="I32">
177
+ <dim>-1</dim>
178
+ </port>
179
+ </output>
180
+ </layer>
181
+ <layer id="12" name="UTF8Validate_1440885" type="UTF8Validate" version="extension">
182
+ <data replace_mode="true" />
183
+ <input>
184
+ <port id="0" precision="I32">
185
+ <dim>-1</dim>
186
+ </port>
187
+ <port id="1" precision="I32">
188
+ <dim>-1</dim>
189
+ </port>
190
+ <port id="2" precision="U8">
191
+ <dim>-1</dim>
192
+ </port>
193
+ </input>
194
+ <output>
195
+ <port id="3" precision="I32">
196
+ <dim>-1</dim>
197
+ </port>
198
+ <port id="4" precision="I32">
199
+ <dim>-1</dim>
200
+ </port>
201
+ <port id="5" precision="U8">
202
+ <dim>-1</dim>
203
+ </port>
204
+ </output>
205
+ </layer>
206
+ <layer id="13" name="Constant_1440887" type="Const" version="opset1">
207
+ <data element_type="u8" shape="51" offset="1068595" size="51" />
208
+ <output>
209
+ <port id="0" precision="U8">
210
+ <dim>51</dim>
211
+ </port>
212
+ </output>
213
+ </layer>
214
+ <layer id="14" name="Constant_1440889" type="Const" version="opset1">
215
+ <data element_type="u8" shape="2" offset="1068646" size="2" />
216
+ <output>
217
+ <port id="0" precision="U8">
218
+ <dim>2</dim>
219
+ </port>
220
+ </output>
221
+ </layer>
222
+ <layer id="15" name="RegexNormalization_1440890" type="RegexNormalization" version="extension">
223
+ <data global_replace="true" />
224
+ <input>
225
+ <port id="0" precision="I32">
226
+ <dim>-1</dim>
227
+ </port>
228
+ <port id="1" precision="I32">
229
+ <dim>-1</dim>
230
+ </port>
231
+ <port id="2" precision="U8">
232
+ <dim>-1</dim>
233
+ </port>
234
+ <port id="3" precision="U8">
235
+ <dim>51</dim>
236
+ </port>
237
+ <port id="4" precision="U8">
238
+ <dim>2</dim>
239
+ </port>
240
+ </input>
241
+ <output>
242
+ <port id="5" precision="I32">
243
+ <dim>-1</dim>
244
+ </port>
245
+ <port id="6" precision="I32">
246
+ <dim>-1</dim>
247
+ </port>
248
+ <port id="7" precision="U8">
249
+ <dim>-1</dim>
250
+ </port>
251
+ </output>
252
+ </layer>
253
+ <layer id="16" name="StringTensorPack_1440891" type="StringTensorPack" version="extension">
254
+ <data mode="begins_ends" />
255
+ <input>
256
+ <port id="0" precision="I32">
257
+ <dim>-1</dim>
258
+ </port>
259
+ <port id="1" precision="I32">
260
+ <dim>-1</dim>
261
+ </port>
262
+ <port id="2" precision="U8">
263
+ <dim>-1</dim>
264
+ </port>
265
+ </input>
266
+ <output>
267
+ <port id="3" precision="STRING" names="string_output">
268
+ <dim>-1</dim>
269
+ </port>
270
+ </output>
271
+ </layer>
272
+ <layer id="17" name="Result_1440892" type="Result" version="opset1">
273
+ <input>
274
+ <port id="0" precision="STRING">
275
+ <dim>-1</dim>
276
+ </port>
277
+ </input>
278
+ </layer>
279
+ </layers>
280
+ <edges>
281
+ <edge from-layer="0" from-port="0" to-layer="1" to-port="0" />
282
+ <edge from-layer="1" from-port="1" to-layer="10" to-port="0" />
283
+ <edge from-layer="2" from-port="0" to-layer="3" to-port="0" />
284
+ <edge from-layer="3" from-port="1" to-layer="10" to-port="1" />
285
+ <edge from-layer="3" from-port="2" to-layer="10" to-port="2" />
286
+ <edge from-layer="3" from-port="3" to-layer="10" to-port="3" />
287
+ <edge from-layer="4" from-port="0" to-layer="9" to-port="0" />
288
+ <edge from-layer="5" from-port="0" to-layer="9" to-port="1" />
289
+ <edge from-layer="6" from-port="0" to-layer="9" to-port="2" />
290
+ <edge from-layer="7" from-port="0" to-layer="9" to-port="3" />
291
+ <edge from-layer="8" from-port="0" to-layer="9" to-port="4" />
292
+ <edge from-layer="9" from-port="5" to-layer="10" to-port="4" />
293
+ <edge from-layer="10" from-port="7" to-layer="11" to-port="2" />
294
+ <edge from-layer="10" from-port="9" to-layer="12" to-port="2" />
295
+ <edge from-layer="10" from-port="8" to-layer="11" to-port="3" />
296
+ <edge from-layer="10" from-port="6" to-layer="11" to-port="1" />
297
+ <edge from-layer="10" from-port="5" to-layer="11" to-port="0" />
298
+ <edge from-layer="11" from-port="4" to-layer="12" to-port="0" />
299
+ <edge from-layer="11" from-port="5" to-layer="12" to-port="1" />
300
+ <edge from-layer="12" from-port="3" to-layer="15" to-port="0" />
301
+ <edge from-layer="12" from-port="4" to-layer="15" to-port="1" />
302
+ <edge from-layer="12" from-port="5" to-layer="15" to-port="2" />
303
+ <edge from-layer="13" from-port="0" to-layer="15" to-port="3" />
304
+ <edge from-layer="14" from-port="0" to-layer="15" to-port="4" />
305
+ <edge from-layer="15" from-port="5" to-layer="16" to-port="0" />
306
+ <edge from-layer="15" from-port="6" to-layer="16" to-port="1" />
307
+ <edge from-layer="15" from-port="7" to-layer="16" to-port="2" />
308
+ <edge from-layer="16" from-port="3" to-layer="17" to-port="0" />
309
+ </edges>
310
+ <rt_info>
311
+ <add_attention_mask value="True" />
312
+ <add_prefix_space />
313
+ <add_special_tokens value="True" />
314
+ <clean_up_tokenization_spaces />
315
+ <detokenizer_input_type value="i64" />
316
+ <eos_token_id value="0" />
317
+ <handle_special_tokens_with_re />
318
+ <number_of_inputs value="1" />
319
+ <openvino_tokenizers_version value="2025.0.0.0" />
320
+ <openvino_version value="2025.0.0" />
321
+ <original_tokenizer_class value="&lt;class 'transformers.tokenization_utils_fast.PreTrainedTokenizerFast'>" />
322
+ <pad_token_id value="0" />
323
+ <sentencepiece_version value="0.2.0" />
324
+ <skip_special_tokens value="True" />
325
+ <streaming_detokenizer value="False" />
326
+ <tiktoken_version value="0.9.0" />
327
+ <tokenizer_output_type value="i64" />
328
+ <tokenizers_version value="0.21.0" />
329
+ <transformers_version value="4.48.3" />
330
+ <use_max_padding value="False" />
331
+ <use_sentencepiece_backend value="False" />
332
+ <utf8_replace_mode value="replace" />
333
+ <with_detokenizer value="True" />
334
+ </rt_info>
335
+ </net>
openvino_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c49f586d3f6dcc279b8246ddeaa062c987189c7b10ab3ec2f76bd8a8d23b8d8b
3
+ size 7182306708
openvino_model.xml ADDED
The diff for this file is too large to render. See raw diff
 
openvino_tokenizer.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e39228abb37f0d292daaf8637af5d9ecfbe6a1860908c330cbfcfb5753adccf6
3
+ size 2474485
openvino_tokenizer.xml ADDED
@@ -0,0 +1,648 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0"?>
2
+ <net name="tokenizer" version="11">
3
+ <layers>
4
+ <layer id="0" name="Parameter_1440774" type="Parameter" version="opset1">
5
+ <data shape="?" element_type="string" />
6
+ <output>
7
+ <port id="0" precision="STRING" names="Parameter_1440774">
8
+ <dim>-1</dim>
9
+ </port>
10
+ </output>
11
+ </layer>
12
+ <layer id="1" name="Constant_1440780" type="Const" version="opset1">
13
+ <data element_type="i64" shape="" offset="0" size="8" />
14
+ <output>
15
+ <port id="0" precision="I64" />
16
+ </output>
17
+ </layer>
18
+ <layer id="2" name="StringTensorUnpack_1440775" type="StringTensorUnpack" version="extension">
19
+ <data mode="begins_ends" />
20
+ <input>
21
+ <port id="0" precision="STRING">
22
+ <dim>-1</dim>
23
+ </port>
24
+ </input>
25
+ <output>
26
+ <port id="1" precision="I32">
27
+ <dim>-1</dim>
28
+ </port>
29
+ <port id="2" precision="I32">
30
+ <dim>-1</dim>
31
+ </port>
32
+ <port id="3" precision="U8">
33
+ <dim>-1</dim>
34
+ </port>
35
+ </output>
36
+ </layer>
37
+ <layer id="3" name="ShapeOf_1440776" type="ShapeOf" version="opset3">
38
+ <data output_type="i64" />
39
+ <input>
40
+ <port id="0" precision="I32">
41
+ <dim>-1</dim>
42
+ </port>
43
+ </input>
44
+ <output>
45
+ <port id="1" precision="I64">
46
+ <dim>1</dim>
47
+ </port>
48
+ </output>
49
+ </layer>
50
+ <layer id="4" name="Constant_1440777" type="Const" version="opset1">
51
+ <data element_type="i64" shape="" offset="0" size="8" />
52
+ <output>
53
+ <port id="0" precision="I64" />
54
+ </output>
55
+ </layer>
56
+ <layer id="5" name="Constant_1440778" type="Const" version="opset1">
57
+ <data element_type="i64" shape="" offset="0" size="8" />
58
+ <output>
59
+ <port id="0" precision="I64" />
60
+ </output>
61
+ </layer>
62
+ <layer id="6" name="Gather_1440779" type="Gather" version="opset8">
63
+ <data batch_dims="0" />
64
+ <input>
65
+ <port id="0" precision="I64">
66
+ <dim>1</dim>
67
+ </port>
68
+ <port id="1" precision="I64" />
69
+ <port id="2" precision="I64" />
70
+ </input>
71
+ <output>
72
+ <port id="3" precision="I64" />
73
+ </output>
74
+ </layer>
75
+ <layer id="7" name="Constant_1440781" type="Const" version="opset1">
76
+ <data element_type="i64" shape="" offset="8" size="8" />
77
+ <output>
78
+ <port id="0" precision="I64" />
79
+ </output>
80
+ </layer>
81
+ <layer id="8" name="Range_1440782" type="Range" version="opset4">
82
+ <data output_type="i32" />
83
+ <input>
84
+ <port id="0" precision="I64" />
85
+ <port id="1" precision="I64" />
86
+ <port id="2" precision="I64" />
87
+ </input>
88
+ <output>
89
+ <port id="3" precision="I32">
90
+ <dim>-1</dim>
91
+ </port>
92
+ </output>
93
+ </layer>
94
+ <layer id="9" name="Constant_1440783" type="Const" version="opset1">
95
+ <data element_type="i64" shape="" offset="8" size="8" />
96
+ <output>
97
+ <port id="0" precision="I64" />
98
+ </output>
99
+ </layer>
100
+ <layer id="10" name="Constant_1440784" type="Const" version="opset1">
101
+ <data element_type="i64" shape="" offset="8" size="8" />
102
+ <output>
103
+ <port id="0" precision="I64" />
104
+ </output>
105
+ </layer>
106
+ <layer id="11" name="Add_1440785" type="Add" version="opset1">
107
+ <data auto_broadcast="numpy" />
108
+ <input>
109
+ <port id="0" precision="I64" />
110
+ <port id="1" precision="I64" />
111
+ </input>
112
+ <output>
113
+ <port id="2" precision="I64" />
114
+ </output>
115
+ </layer>
116
+ <layer id="12" name="Constant_1440786" type="Const" version="opset1">
117
+ <data element_type="i64" shape="" offset="8" size="8" />
118
+ <output>
119
+ <port id="0" precision="I64" />
120
+ </output>
121
+ </layer>
122
+ <layer id="13" name="Range_1440787" type="Range" version="opset4">
123
+ <data output_type="i32" />
124
+ <input>
125
+ <port id="0" precision="I64" />
126
+ <port id="1" precision="I64" />
127
+ <port id="2" precision="I64" />
128
+ </input>
129
+ <output>
130
+ <port id="3" precision="I32">
131
+ <dim>-1</dim>
132
+ </port>
133
+ </output>
134
+ </layer>
135
+ <layer id="14" name="Constant_1440849" type="Const" version="opset1">
136
+ <data element_type="u8" shape="19" offset="16" size="19" />
137
+ <output>
138
+ <port id="0" precision="U8">
139
+ <dim>19</dim>
140
+ </port>
141
+ </output>
142
+ </layer>
143
+ <layer id="15" name="SpecialTokensSplit_1440850" type="SpecialTokensSplit" version="extension">
144
+ <input>
145
+ <port id="0" precision="I32">
146
+ <dim>-1</dim>
147
+ </port>
148
+ <port id="1" precision="I32">
149
+ <dim>-1</dim>
150
+ </port>
151
+ <port id="2" precision="I32">
152
+ <dim>-1</dim>
153
+ </port>
154
+ <port id="3" precision="I32">
155
+ <dim>-1</dim>
156
+ </port>
157
+ <port id="4" precision="U8">
158
+ <dim>-1</dim>
159
+ </port>
160
+ <port id="5" precision="U8">
161
+ <dim>19</dim>
162
+ </port>
163
+ </input>
164
+ <output>
165
+ <port id="6" precision="I32">
166
+ <dim>-1</dim>
167
+ </port>
168
+ <port id="7" precision="I32">
169
+ <dim>-1</dim>
170
+ </port>
171
+ <port id="8" precision="I32">
172
+ <dim>-1</dim>
173
+ </port>
174
+ <port id="9" precision="I32">
175
+ <dim>-1</dim>
176
+ </port>
177
+ <port id="10" precision="U8">
178
+ <dim>-1</dim>
179
+ </port>
180
+ <port id="11" precision="BOOL">
181
+ <dim>-1</dim>
182
+ </port>
183
+ </output>
184
+ </layer>
185
+ <layer id="16" name="Constant_1440852" type="Const" version="opset1">
186
+ <data element_type="u8" shape="64" offset="35" size="64" />
187
+ <output>
188
+ <port id="0" precision="U8">
189
+ <dim>64</dim>
190
+ </port>
191
+ </output>
192
+ </layer>
193
+ <layer id="17" name="RegexSplit_1440853" type="RegexSplit" version="extension">
194
+ <data behaviour="isolate" invert="false" max_splits="-1" />
195
+ <input>
196
+ <port id="0" precision="I32">
197
+ <dim>-1</dim>
198
+ </port>
199
+ <port id="1" precision="I32">
200
+ <dim>-1</dim>
201
+ </port>
202
+ <port id="2" precision="I32">
203
+ <dim>-1</dim>
204
+ </port>
205
+ <port id="3" precision="I32">
206
+ <dim>-1</dim>
207
+ </port>
208
+ <port id="4" precision="U8">
209
+ <dim>-1</dim>
210
+ </port>
211
+ <port id="5" precision="BOOL">
212
+ <dim>-1</dim>
213
+ </port>
214
+ <port id="6" precision="U8">
215
+ <dim>64</dim>
216
+ </port>
217
+ </input>
218
+ <output>
219
+ <port id="7" precision="I32">
220
+ <dim>-1</dim>
221
+ </port>
222
+ <port id="8" precision="I32">
223
+ <dim>-1</dim>
224
+ </port>
225
+ <port id="9" precision="I32">
226
+ <dim>-1</dim>
227
+ </port>
228
+ <port id="10" precision="I32">
229
+ <dim>-1</dim>
230
+ </port>
231
+ <port id="11" precision="U8">
232
+ <dim>-1</dim>
233
+ </port>
234
+ <port id="12" precision="BOOL">
235
+ <dim>-1</dim>
236
+ </port>
237
+ </output>
238
+ </layer>
239
+ <layer id="18" name="Constant_1440855" type="Const" version="opset1">
240
+ <data element_type="u8" shape="1068575" offset="99" size="1068575" />
241
+ <output>
242
+ <port id="0" precision="U8">
243
+ <dim>1068575</dim>
244
+ </port>
245
+ </output>
246
+ </layer>
247
+ <layer id="19" name="StringTensorUnpack_1440856" type="StringTensorUnpack" version="extension">
248
+ <data mode="begins_ends" />
249
+ <input>
250
+ <port id="0" precision="U8">
251
+ <dim>1068575</dim>
252
+ </port>
253
+ </input>
254
+ <output>
255
+ <port id="1" precision="I32">
256
+ <dim>-1</dim>
257
+ </port>
258
+ <port id="2" precision="I32">
259
+ <dim>-1</dim>
260
+ </port>
261
+ <port id="3" precision="U8">
262
+ <dim>-1</dim>
263
+ </port>
264
+ </output>
265
+ </layer>
266
+ <layer id="20" name="Constant_1440858" type="Const" version="opset1">
267
+ <data element_type="u8" shape="732676" offset="1068674" size="732676" />
268
+ <output>
269
+ <port id="0" precision="U8">
270
+ <dim>732676</dim>
271
+ </port>
272
+ </output>
273
+ </layer>
274
+ <layer id="21" name="StringTensorUnpack_1440859" type="StringTensorUnpack" version="extension">
275
+ <data mode="begins_ends" />
276
+ <input>
277
+ <port id="0" precision="U8">
278
+ <dim>732676</dim>
279
+ </port>
280
+ </input>
281
+ <output>
282
+ <port id="1" precision="I32">
283
+ <dim>-1</dim>
284
+ </port>
285
+ <port id="2" precision="I32">
286
+ <dim>-1</dim>
287
+ </port>
288
+ <port id="3" precision="U8">
289
+ <dim>-1</dim>
290
+ </port>
291
+ </output>
292
+ </layer>
293
+ <layer id="22" name="Constant_1440861" type="Const" version="opset1">
294
+ <data element_type="u8" shape="673127" offset="1801350" size="673127" />
295
+ <output>
296
+ <port id="0" precision="U8">
297
+ <dim>673127</dim>
298
+ </port>
299
+ </output>
300
+ </layer>
301
+ <layer id="23" name="StringTensorUnpack_1440862" type="StringTensorUnpack" version="extension">
302
+ <data mode="begins_ends" />
303
+ <input>
304
+ <port id="0" precision="U8">
305
+ <dim>673127</dim>
306
+ </port>
307
+ </input>
308
+ <output>
309
+ <port id="1" precision="I32">
310
+ <dim>-1</dim>
311
+ </port>
312
+ <port id="2" precision="I32">
313
+ <dim>-1</dim>
314
+ </port>
315
+ <port id="3" precision="U8">
316
+ <dim>-1</dim>
317
+ </port>
318
+ </output>
319
+ </layer>
320
+ <layer id="24" name="BPETokenizer_1440863" type="BPETokenizer" version="extension">
321
+ <data unk_token="" fuse_unk="false" suffix_indicator="" end_suffix="" byte_fallback="false" cache_capacity="20000" />
322
+ <input>
323
+ <port id="0" precision="I32">
324
+ <dim>-1</dim>
325
+ </port>
326
+ <port id="1" precision="I32">
327
+ <dim>-1</dim>
328
+ </port>
329
+ <port id="2" precision="I32">
330
+ <dim>-1</dim>
331
+ </port>
332
+ <port id="3" precision="I32">
333
+ <dim>-1</dim>
334
+ </port>
335
+ <port id="4" precision="U8">
336
+ <dim>-1</dim>
337
+ </port>
338
+ <port id="5" precision="I32">
339
+ <dim>-1</dim>
340
+ </port>
341
+ <port id="6" precision="I32">
342
+ <dim>-1</dim>
343
+ </port>
344
+ <port id="7" precision="U8">
345
+ <dim>-1</dim>
346
+ </port>
347
+ <port id="8" precision="I32">
348
+ <dim>-1</dim>
349
+ </port>
350
+ <port id="9" precision="I32">
351
+ <dim>-1</dim>
352
+ </port>
353
+ <port id="10" precision="U8">
354
+ <dim>-1</dim>
355
+ </port>
356
+ <port id="11" precision="I32">
357
+ <dim>-1</dim>
358
+ </port>
359
+ <port id="12" precision="I32">
360
+ <dim>-1</dim>
361
+ </port>
362
+ <port id="13" precision="U8">
363
+ <dim>-1</dim>
364
+ </port>
365
+ </input>
366
+ <output>
367
+ <port id="14" precision="I32">
368
+ <dim>-1</dim>
369
+ </port>
370
+ <port id="15" precision="I32">
371
+ <dim>-1</dim>
372
+ </port>
373
+ <port id="16" precision="I32">
374
+ <dim>-1</dim>
375
+ </port>
376
+ </output>
377
+ </layer>
378
+ <layer id="25" name="Subtract_1440864" type="Subtract" version="opset1">
379
+ <data auto_broadcast="numpy" />
380
+ <input>
381
+ <port id="0" precision="I32">
382
+ <dim>-1</dim>
383
+ </port>
384
+ <port id="1" precision="I32">
385
+ <dim>-1</dim>
386
+ </port>
387
+ </input>
388
+ <output>
389
+ <port id="2" precision="I32">
390
+ <dim>-1</dim>
391
+ </port>
392
+ </output>
393
+ </layer>
394
+ <layer id="26" name="Constant_1440865" type="Const" version="opset1">
395
+ <data element_type="i32" shape="" offset="2474477" size="4" />
396
+ <output>
397
+ <port id="0" precision="I32" />
398
+ </output>
399
+ </layer>
400
+ <layer id="27" name="Minimum_1440866" type="Minimum" version="opset1">
401
+ <data auto_broadcast="numpy" />
402
+ <input>
403
+ <port id="0" precision="I32">
404
+ <dim>-1</dim>
405
+ </port>
406
+ <port id="1" precision="I32" />
407
+ </input>
408
+ <output>
409
+ <port id="2" precision="I32">
410
+ <dim>-1</dim>
411
+ </port>
412
+ </output>
413
+ </layer>
414
+ <layer id="28" name="Subtract_1440867" type="Subtract" version="opset1">
415
+ <data auto_broadcast="numpy" />
416
+ <input>
417
+ <port id="0" precision="I32">
418
+ <dim>-1</dim>
419
+ </port>
420
+ <port id="1" precision="I32">
421
+ <dim>-1</dim>
422
+ </port>
423
+ </input>
424
+ <output>
425
+ <port id="2" precision="I32">
426
+ <dim>-1</dim>
427
+ </port>
428
+ </output>
429
+ </layer>
430
+ <layer id="29" name="Subtract_1440868" type="Subtract" version="opset1">
431
+ <data auto_broadcast="numpy" />
432
+ <input>
433
+ <port id="0" precision="I32">
434
+ <dim>-1</dim>
435
+ </port>
436
+ <port id="1" precision="I32">
437
+ <dim>-1</dim>
438
+ </port>
439
+ </input>
440
+ <output>
441
+ <port id="2" precision="I32">
442
+ <dim>-1</dim>
443
+ </port>
444
+ </output>
445
+ </layer>
446
+ <layer id="30" name="Constant_1440869" type="Const" version="opset1">
447
+ <data element_type="i32" shape="" offset="2474481" size="4" />
448
+ <output>
449
+ <port id="0" precision="I32" />
450
+ </output>
451
+ </layer>
452
+ <layer id="31" name="ReduceMax_1440870" type="ReduceMax" version="opset1">
453
+ <data keep_dims="false" />
454
+ <input>
455
+ <port id="0" precision="I32">
456
+ <dim>-1</dim>
457
+ </port>
458
+ <port id="1" precision="I32" />
459
+ </input>
460
+ <output>
461
+ <port id="2" precision="I32" />
462
+ </output>
463
+ </layer>
464
+ <layer id="32" name="Constant_1440871" type="Const" version="opset1">
465
+ <data element_type="i32" shape="" offset="2474481" size="4" />
466
+ <output>
467
+ <port id="0" precision="I32" />
468
+ </output>
469
+ </layer>
470
+ <layer id="33" name="RaggedToDense_1440872" type="RaggedToDense" version="extension">
471
+ <data pad_right="false" />
472
+ <input>
473
+ <port id="0" precision="I32">
474
+ <dim>-1</dim>
475
+ </port>
476
+ <port id="1" precision="I32">
477
+ <dim>-1</dim>
478
+ </port>
479
+ <port id="2" precision="I32">
480
+ <dim>-1</dim>
481
+ </port>
482
+ <port id="3" precision="I32" />
483
+ <port id="4" precision="I32" />
484
+ </input>
485
+ <output>
486
+ <port id="5" precision="I32">
487
+ <dim>-1</dim>
488
+ <dim>-1</dim>
489
+ </port>
490
+ <port id="6" precision="BOOL">
491
+ <dim>-1</dim>
492
+ <dim>-1</dim>
493
+ </port>
494
+ </output>
495
+ </layer>
496
+ <layer id="34" name="Convert_1440873" type="Convert" version="opset1">
497
+ <data destination_type="i32" />
498
+ <input>
499
+ <port id="0" precision="BOOL">
500
+ <dim>-1</dim>
501
+ <dim>-1</dim>
502
+ </port>
503
+ </input>
504
+ <output>
505
+ <port id="1" precision="I32">
506
+ <dim>-1</dim>
507
+ <dim>-1</dim>
508
+ </port>
509
+ </output>
510
+ </layer>
511
+ <layer id="35" name="Convert_1440873.0" type="Convert" version="opset1">
512
+ <data destination_type="i64" />
513
+ <input>
514
+ <port id="0" precision="I32">
515
+ <dim>-1</dim>
516
+ <dim>-1</dim>
517
+ </port>
518
+ </input>
519
+ <output>
520
+ <port id="1" precision="I64" names="attention_mask">
521
+ <dim>-1</dim>
522
+ <dim>-1</dim>
523
+ </port>
524
+ </output>
525
+ </layer>
526
+ <layer id="37" name="RaggedToDense_1440872.0" type="Convert" version="opset1">
527
+ <data destination_type="i64" />
528
+ <input>
529
+ <port id="0" precision="I32">
530
+ <dim>-1</dim>
531
+ <dim>-1</dim>
532
+ </port>
533
+ </input>
534
+ <output>
535
+ <port id="1" precision="I64" names="input_ids">
536
+ <dim>-1</dim>
537
+ <dim>-1</dim>
538
+ </port>
539
+ </output>
540
+ </layer>
541
+ <layer id="38" name="Result_1440874" type="Result" version="opset1">
542
+ <input>
543
+ <port id="0" precision="I64">
544
+ <dim>-1</dim>
545
+ <dim>-1</dim>
546
+ </port>
547
+ </input>
548
+ </layer>
549
+ <layer id="36" name="Result_1440875" type="Result" version="opset1">
550
+ <input>
551
+ <port id="0" precision="I64">
552
+ <dim>-1</dim>
553
+ <dim>-1</dim>
554
+ </port>
555
+ </input>
556
+ </layer>
557
+ </layers>
558
+ <edges>
559
+ <edge from-layer="0" from-port="0" to-layer="2" to-port="0" />
560
+ <edge from-layer="1" from-port="0" to-layer="8" to-port="0" />
561
+ <edge from-layer="2" from-port="1" to-layer="3" to-port="0" />
562
+ <edge from-layer="2" from-port="3" to-layer="15" to-port="4" />
563
+ <edge from-layer="2" from-port="2" to-layer="15" to-port="3" />
564
+ <edge from-layer="2" from-port="1" to-layer="15" to-port="2" />
565
+ <edge from-layer="3" from-port="1" to-layer="6" to-port="0" />
566
+ <edge from-layer="4" from-port="0" to-layer="6" to-port="1" />
567
+ <edge from-layer="5" from-port="0" to-layer="6" to-port="2" />
568
+ <edge from-layer="6" from-port="3" to-layer="11" to-port="0" />
569
+ <edge from-layer="6" from-port="3" to-layer="8" to-port="1" />
570
+ <edge from-layer="7" from-port="0" to-layer="8" to-port="2" />
571
+ <edge from-layer="8" from-port="3" to-layer="15" to-port="0" />
572
+ <edge from-layer="9" from-port="0" to-layer="13" to-port="0" />
573
+ <edge from-layer="10" from-port="0" to-layer="11" to-port="1" />
574
+ <edge from-layer="11" from-port="2" to-layer="13" to-port="1" />
575
+ <edge from-layer="12" from-port="0" to-layer="13" to-port="2" />
576
+ <edge from-layer="13" from-port="3" to-layer="15" to-port="1" />
577
+ <edge from-layer="14" from-port="0" to-layer="15" to-port="5" />
578
+ <edge from-layer="15" from-port="6" to-layer="17" to-port="0" />
579
+ <edge from-layer="15" from-port="7" to-layer="17" to-port="1" />
580
+ <edge from-layer="15" from-port="8" to-layer="17" to-port="2" />
581
+ <edge from-layer="15" from-port="9" to-layer="17" to-port="3" />
582
+ <edge from-layer="15" from-port="10" to-layer="17" to-port="4" />
583
+ <edge from-layer="15" from-port="11" to-layer="17" to-port="5" />
584
+ <edge from-layer="16" from-port="0" to-layer="17" to-port="6" />
585
+ <edge from-layer="17" from-port="11" to-layer="24" to-port="4" />
586
+ <edge from-layer="17" from-port="10" to-layer="24" to-port="3" />
587
+ <edge from-layer="17" from-port="9" to-layer="24" to-port="2" />
588
+ <edge from-layer="17" from-port="8" to-layer="24" to-port="1" />
589
+ <edge from-layer="17" from-port="7" to-layer="24" to-port="0" />
590
+ <edge from-layer="18" from-port="0" to-layer="19" to-port="0" />
591
+ <edge from-layer="19" from-port="1" to-layer="24" to-port="5" />
592
+ <edge from-layer="19" from-port="2" to-layer="24" to-port="6" />
593
+ <edge from-layer="19" from-port="3" to-layer="24" to-port="7" />
594
+ <edge from-layer="20" from-port="0" to-layer="21" to-port="0" />
595
+ <edge from-layer="21" from-port="1" to-layer="24" to-port="8" />
596
+ <edge from-layer="21" from-port="2" to-layer="24" to-port="9" />
597
+ <edge from-layer="21" from-port="3" to-layer="24" to-port="10" />
598
+ <edge from-layer="22" from-port="0" to-layer="23" to-port="0" />
599
+ <edge from-layer="23" from-port="1" to-layer="24" to-port="11" />
600
+ <edge from-layer="23" from-port="3" to-layer="24" to-port="13" />
601
+ <edge from-layer="23" from-port="2" to-layer="24" to-port="12" />
602
+ <edge from-layer="24" from-port="15" to-layer="25" to-port="0" />
603
+ <edge from-layer="24" from-port="14" to-layer="25" to-port="1" />
604
+ <edge from-layer="24" from-port="15" to-layer="28" to-port="0" />
605
+ <edge from-layer="24" from-port="16" to-layer="33" to-port="2" />
606
+ <edge from-layer="24" from-port="15" to-layer="29" to-port="0" />
607
+ <edge from-layer="24" from-port="15" to-layer="33" to-port="1" />
608
+ <edge from-layer="25" from-port="2" to-layer="27" to-port="0" />
609
+ <edge from-layer="26" from-port="0" to-layer="27" to-port="1" />
610
+ <edge from-layer="27" from-port="2" to-layer="28" to-port="1" />
611
+ <edge from-layer="28" from-port="2" to-layer="33" to-port="0" />
612
+ <edge from-layer="28" from-port="2" to-layer="29" to-port="1" />
613
+ <edge from-layer="29" from-port="2" to-layer="31" to-port="0" />
614
+ <edge from-layer="30" from-port="0" to-layer="31" to-port="1" />
615
+ <edge from-layer="31" from-port="2" to-layer="33" to-port="3" />
616
+ <edge from-layer="32" from-port="0" to-layer="33" to-port="4" />
617
+ <edge from-layer="33" from-port="6" to-layer="34" to-port="0" />
618
+ <edge from-layer="33" from-port="5" to-layer="37" to-port="0" />
619
+ <edge from-layer="34" from-port="1" to-layer="35" to-port="0" />
620
+ <edge from-layer="35" from-port="1" to-layer="36" to-port="0" />
621
+ <edge from-layer="37" from-port="1" to-layer="38" to-port="0" />
622
+ </edges>
623
+ <rt_info>
624
+ <add_attention_mask value="True" />
625
+ <add_prefix_space />
626
+ <add_special_tokens value="True" />
627
+ <clean_up_tokenization_spaces />
628
+ <detokenizer_input_type value="i64" />
629
+ <eos_token_id value="0" />
630
+ <handle_special_tokens_with_re />
631
+ <number_of_inputs value="1" />
632
+ <openvino_tokenizers_version value="2025.0.0.0" />
633
+ <openvino_version value="2025.0.0" />
634
+ <original_tokenizer_class value="&lt;class 'transformers.tokenization_utils_fast.PreTrainedTokenizerFast'>" />
635
+ <pad_token_id value="0" />
636
+ <sentencepiece_version value="0.2.0" />
637
+ <skip_special_tokens value="True" />
638
+ <streaming_detokenizer value="False" />
639
+ <tiktoken_version value="0.9.0" />
640
+ <tokenizer_output_type value="i64" />
641
+ <tokenizers_version value="0.21.0" />
642
+ <transformers_version value="4.48.3" />
643
+ <use_max_padding value="False" />
644
+ <use_sentencepiece_backend value="False" />
645
+ <utf8_replace_mode value="replace" />
646
+ <with_detokenizer value="True" />
647
+ </rt_info>
648
+ </net>
special_tokens_map.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "eos_token": {
3
+ "content": "<|endoftext|>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "pad_token": {
10
+ "content": "<|endoftext|>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "unk_token": {
17
+ "content": "<|endoftext|>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ }
23
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "0": {
4
+ "content": "<|endoftext|>",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false,
9
+ "special": true
10
+ }
11
+ },
12
+ "clean_up_tokenization_spaces": true,
13
+ "eos_token": "<|endoftext|>",
14
+ "extra_special_tokens": {},
15
+ "model_max_length": 2048,
16
+ "pad_token": "<|endoftext|>",
17
+ "tokenizer_class": "PreTrainedTokenizerFast",
18
+ "unk_token": "<|endoftext|>"
19
+ }