PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度
fix regression (#1135)
Xin He8874530d3301
diff --git a/auto_round/export/export_to_autoround/export_to_nvfp_mxfp.py b/auto_round/export/export_to_autoround/export_to_nvfp_mxfp.pyindex cbb45d90e..de3991c7f 100644--- a/auto_round/export/export_to_autoround/export_to_nvfp_mxfp.py+++ b/auto_round/export/export_to_autoround/export_to_nvfp_mxfp.py@@ -204,8 +204,8 @@ def save_quantized_as_fp(output_dir, inplace=True, **kwargs):if len(neq_keys) > 0:extra_config[layer_name] = {}for key in scheme_keys:- if cfg[key] is not None:- extra_config[layer_name][key] = cfg[key]+ if cfg.get(key, None) is not None:+ extra_config[layer_name][key] = cfg.get(key, None)regex_config = quantization_config.pop("regex_config")if regex_config is not None: