haoran wang comments

Results 4 comments of


                                            haoran wang

InternVL2_5-4B-MPO lora微调

你好，我尝试用2.5的lora微调脚本来进行MPO的lora微调，但我遇到了warning: shape mismatch: value tensor of shape [4608, 4096] cannot be broadcast to indexing result of shape [1098, 4096], i nput_embeds[selected].shape=torch.Size([1098, 4096]), vit_embeds.shape=torch.Size([4608, 4096])这样的问题，请问你遇到了吗，或者你成功使用lora微调了吗，感谢

[Bug] 在跑8B-MPO finetune完成后保存模型出现UnboundLocalError: cannot access local variable 'default_config' where it is not associated with a value

> @JackeyHRan transformers 4.45.1 遇到了同样的问题，降级到4.40.0似乎可以正常保存权重感谢，我去尝试一下

[Bug] 在跑8B-MPO finetune完成后保存模型出现UnboundLocalError: cannot access local variable 'default_config' where it is not associated with a value

> > @JackeyHRan transformers 4.45.1 遇到了同样的问题，降级到4.40.0似乎可以正常保存权重 > > 请问你遇到这个问题了吗？File "/opt/conda/lib/python3.8/site-packages/transformers/modeling_utils.py", line 1705, in get_input_embeddings raise NotImplementedError 应该怎么解决啊，感谢虽然我没遇到过这个问题，但我觉得大概率是transformers版本的问题

[Bug] 在跑8B-MPO finetune完成后保存模型出现UnboundLocalError: cannot access local variable 'default_config' where it is not associated with a value

> > > > @JackeyHRan transformers 4.45.1 遇到了同样的问题，降级到4.40.0似乎可以正常保存权重 > > > > > > > > > 请问你遇到这个问题了吗？File "/opt/conda/lib/python3.8/site-packages/transformers/modeling_utils.py", line 1705, in get_input_embeddings raise NotImplementedError 应该怎么解决啊，感谢 > > > >...