-
Notifications
You must be signed in to change notification settings - Fork 2.3k
Support completion bootstrap for VLM in GRPO/RLOO #4452
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
Conversation
albertvillanova
left a comment
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.
I think you mixed 2 contributions in this PR:
- Fix bug in GRPOTrainer
- Add experimental RTPO
I would suggest you split your contributions in 2 PRs, and keep here only the bug fix.
Co-authored-by: Albert Villanova del Moral <[email protected]>
Co-authored-by: Albert Villanova del Moral <[email protected]>
|
Thank you for your suggestion |
The problem is I don't know how to split it into 2 PR2 properly. |
qgallouedec
left a comment
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.
LGTM thanks!
|
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. |
|
Thank you guys! @qgallouedec @albertvillanova |
1. Fix a little bug in GRPOTrainer.
When inputs are conversational, bootstrap is a list not str, causing a str concat error.
I add a bootstrap type check to fix the bug while maintain the original program