PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度
Enhance tokenizer saving (#1057)
Wang, Mengni4ec50d1ca7a6
diff --git a/auto_round/compressors/base.py b/auto_round/compressors/base.pyindex 7c90ed7ac..8936b1e75 100644--- a/auto_round/compressors/base.py+++ b/auto_round/compressors/base.py@@ -3006,7 +3006,7 @@ class BaseCompressor(object):if format == "fake" or format == "qdq": # TODO fix act quantization laterself.model = self.model.to("cpu")self.model.save_pretrained(output_dir)- if self.tokenizer is not None:+ if self.tokenizer is not None and hasattr(self.tokenizer, "save_pretrained"):self.tokenizer.save_pretrained(output_dir)processor = kwargs.get("processor", None)if processor is not None: