牛宇霖

Results 5 comments of 牛宇霖

[Question]: 请问用paddlenlp微调llama2后如何将模型转换回torch的权重呢？

请问是无法转换吗？

DPO训练Lora后，模型的生成结果是乱码

@hiyouga 好的，谢谢您的回复。如果我先merge lora后再训练，那我的model_name_or_path和adapter_name_or_path这两个参数，都是指向同一个合并后的模型吗？ 还是说我脚本中直接删除adapter_name_or_path只保留model_name_or_path？ 非常感谢！

DPO训练Lora后，模型的生成结果是乱码

@hiyouga 谢谢您！我按您说的方式试了一下，DPO训练后模型输出正常了。非常感谢！

```RuntimeError: Bad StatusOr access: UNKNOWN: TPU initialization failed: Invalid --2a886c8_slice_builder_worker_addresses specified. Expected 4 worker addresses, got 1.``` when using kaggle tpu

I encountered the same error on kaggle's TPU VM v3-8 when using lit-gpt project's example finetuning code today. is there any progress on this issue?

checkpoint's size is increasing everytime.

Sure. I sent you them by email. Please check it. Thank you so much!