微调合并的模型 不能用transformers加载

[flowermlh]

Reminder

• I have read the README and searched the existing issues.

System Info

华为昇腾910B npu

Reproduction

使用lora微调之后，然后合并模型。
之后用AutoTokenizer加载模型失败。

from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained('/home/jovyan/lama-3-chinese-8b-lora-merge', local_files_only=True)

报错：
Exception: data did not match any variant of untagged enum ModelWrapper at line 1251003 column 3

Expected behavior

No response

Others

No response

[hiyouga]

把原来的 tokenizer 复制并覆盖到新的目录

[flowermlh]

把原来的 tokenizer 复制并覆盖到新的目录

请问是这三个文件吗?
tokenizer.json
tokenizer_config.json
special_tokens_map.json
谢谢!