-
Notifications
You must be signed in to change notification settings - Fork 31.4k
fix qwen25-vl grad acc #40333
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
fix qwen25-vl grad acc #40333
Conversation
|
cc @SunMarc |
|
Hi, I'd appreciate it if you could help review this PR. Thank you. @SunMarc |
|
@bot /style |
|
Style fix runs successfully without any file modified. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for that ! We indeed need to clean a bit this. I started the work here and hopefully can finish it soon. #38432
|
can you run |
Head branch was pushed to by a user without write access
|
[For maintainers] Suggested jobs to run (before merge) run-slow: glm4v, glm4v_moe, qwen2_5_vl, qwen2_vl |
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
What does this PR do?
Currently, there are still many models that do not accept loss_kwargs (such as Qwen25-VL). Therefore, this PR is proposed with reference to the gemma3 PR. Now, the training loss and grad_norm are normal.
reference
#37208