FHL1998 comments

Results 18 comments of


                                            FHL1998

The loss curve exhibits a stair-step pattern of descent.

Same issue here.Have you solved it?

TypeError: getattr(): attribute name must be string

Someone says it is the issue related to diffusers version, anynoe know the correct version of the package?

TypeError: getattr(): attribute name must be string

> Thx! I get rid of the issue now.

accelerate does not support multi-gpu load int-8bit

same issue here.

Diverging evaluation loss using finetuning scripts Guanaco 7b

Same issue here, wonder if it is led by the token_id? I tried "self-instruct" and "alpaca", but still get the same problem, the performance even worse than using GPT2.

Diverging evaluation loss using finetuning scripts Guanaco 7b

> > wonder if it is led by the token_id? > > Glad to hear I'm not the only one. But what do you mean exactly by this comment? Basically,...

ValueError: You can't train a model that has been loaded in 8-bit precision on a different device than the one you're training on.

Try to update the accelerate version.

ValueError: You can't train a model that has been loaded in 8-bit precision on a different device than the one you're training on.

Have you installed `einops` ? Yeah, I can run the code but the results is poor. The GPU I use is A100-80GB.

ValueError: You can't train a model that has been loaded in 8-bit precision on a different device than the one you're training on.

> > einops > > Already installed. You can successfully run the code in A100-80GB, and I get the same GPU, then the only difference should be that I got...

[Question] Overfitting in my finetune experiment using my custom data

What's the current inference performance? Do you think LLava is suitable for this kind of object detection task?