use_inference_chat_template: False # Tokenization is performed turn by turn and the resulting token ids are concatenated to form the full conversation. # To ensure this matches the result of ...

Reminder I have read the above rules and searched the existing issues. System Info 使用0.9.3版本dpo lora微调Qwen2.5模型，但是保存的模型里tokenizer_config.json里面缺少了chat_template Reproduction model_name_or_path: outp ...