Adapt autoround format (#2038)

Signed-off-by: Kaihui-intel <[email protected]>

Author: Kaihui-intel

neural_compressor/transformers/models/modeling_auto.py

@@ -325,7 +325,7 @@ def load_low_bit(cls, pretrained_model_name_or_path, *model_args, **kwargs):
             quantization_config = TeqConfig.from_dict(quantization_config)
         elif quantization_config["quant_method"] == "gptq":
             quantization_config = GPTQConfig.from_dict(quantization_config)
-        elif quantization_config["quant_method"] == "autoround":
+        elif quantization_config["quant_method"] in ["autoround", "intel/auto-round"]:
             quantization_config = AutoRoundConfig.from_dict(quantization_config)
 
         assert quantization_config is not None, "Detect this model is not a low-bit model."