upload auto_round format

Signed-off-by: n1ck-guo <[email protected]>

config.json

@@ -25,12 +25,11 @@
   "quantization_config": {
     "amp": true,
     "autoround_version": "0.4.4",
+    "backend": "auto_round:gptq:exllamav2",
     "batch_size": 8,
     "bits": 4,
-    "block_name_to_quantize": "model.transformer.blocks",
-    "damp_percent": 0.01,
     "data_type": "int",
-    "desc_act": false,
+    "dataset": "NeelNanda/pile-10k",
     "enable_minmax_tuning": true,
     "enable_norm_bias_tuning": false,
     "enable_quanted_input": true,
@@ -41,11 +40,11 @@
     "lr": 0.001,
     "minmax_lr": 0.001,
     "nsamples": 512,
-    "quant_method": "gptq",
+    "quant_method": "intel/auto-round",
     "scale_dtype": "torch.float16",
     "seqlen": 2048,
     "sym": true,
-    "true_sequential": false
+    "to_quant_block_names": "model.transformer.blocks"
   },
   "rope_theta": 1000000.0,
   "tie_word_embeddings": false,

quantize_config.json → quantization_config.json

@@ -15,11 +15,10 @@
   "amp": true,
   "nsamples": 512,
   "low_gpu_mem_usage": false,
+  "to_quant_block_names": "model.transformer.blocks",
   "enable_norm_bias_tuning": false,
+  "dataset": "NeelNanda/pile-10k",
   "autoround_version": "0.4.4",
-  "block_name_to_quantize": "model.transformer.blocks",
-  "quant_method": "gptq",
-  "desc_act": false,
-  "true_sequential": false,
-  "damp_percent": 0.01
+  "quant_method": "intel/auto-round",
+  "backend": "auto_round:gptq:exllamav2"
 }