Hojin Yang comments

Results 9 comments of


                                            Hojin Yang

ImportError: Using `load_in_8bit=True` requires Accelerate: `pip install accelerate` and the latest version of bitsandbytes `pip install -i https://test.pypi.org/simple/ bitsandbytes` or pip install bitsandbytes` when in reality it's a torch issue

same issue... is there any updates?

Getting this right after I login and hit enter

Same error with this lol I think this is based on the chromium security with google account, but not sure

While training GRPO, I noticed that my model crashes. Its loss is 0, its grad_norm and kl are both Nan, and it completes as “!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!””

Did you solve the issue?

While training GRPO, I noticed that my model crashes. Its loss is 0, its grad_norm and kl are both Nan, and it completes as “!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!””

I don't understand how did you change initial reward from 0 to 1e-4.. Can you give more context with this? In addition, can you provide which versions of libraries you...

While training GRPO, I noticed that my model crashes. Its loss is 0, its grad_norm and kl are both Nan, and it completes as “!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!””

cc. @Jintao-Huang Can you tag some people that can help with this issue?

Remove Vision FA warning

So, if I understood this issue correctly, I can use flash_attn backend with multimodal model with original flash attention 2, correct?

[WIP] Add diffllama

I wonder this pr is still working in progress? Or, most of the implementation has been finalized and waiting for the test coverage review?

[WIP] Add diffllama

Are you still working on this PR, @weak-kajuma ?

[WIP] Add diffllama

@Cyrilvallez any plannings to review this pr?