7B就行lora微调过程报错

[lichengyang666]

训练7B的过程中，将原始的lora参数调整之后，训练到一半时报错
Exception: Current loss scale already at minimum - cannot decrease scale anymore. Exiting run. ，应该如何解决？

尝试将fp16修改为bf16就行尝试，然后又报错
File "/home/jupyter/.cache/huggingface/modules/transformers_modules/TeleChat-7B/modeling_telechat.py", line 185, in forward
assert all((i.dtype in [torch.float16, torch.bfloat16] for i in (q, k, v)))

期待您的回复~