Skip to content
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

Update sample GRPO script to validate num_generations #1509

Merged
merged 4 commits into from
Mar 4, 2025

Conversation

nikg4
Copy link
Collaborator

@nikg4 nikg4 commented Mar 4, 2025

Description

-- Check if global batch size is divisible by num_generations . Otherwise, training may fail (depending on trl version)

Related issues

Fixes OPE-993

Before submitting

  • This PR only changes documentation. (You can ignore the following checks in that case)
  • Did you read the contributor guideline Pull Request guidelines?
  • Did you link the issue(s) related to this PR in the section above?
  • Did you add / update tests where needed?

Reviewers

At least one review from a member of oumi-ai/oumi-staff is required.

@nikg4 nikg4 requested review from optas, taenin, wizeng23 and jgreer013 March 4, 2025 20:20
@nikg4 nikg4 marked this pull request as ready for review March 4, 2025 20:20
Copy link
Contributor

@wizeng23 wizeng23 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do we want to move this logic into our code? This logic likely won't be copied into every future GRPO config. IMO not necessary though, as the current HF error for this is descriptive.

@nikg4
Copy link
Collaborator Author

nikg4 commented Mar 4, 2025

Do we want to move this logic into our code? This logic likely won't be copied into every future GRPO config. IMO not necessary though, as the current HF error for this is descriptive.

Oumi has this logic, but it's currently treated as a warning

logger.warning(

Let's consider upgrading it to error if it fails for new trl versions.

@nikg4 nikg4 merged commit ba65772 into main Mar 4, 2025
1 check passed
@nikg4 nikg4 deleted the nikg4/comp-multiple branch March 4, 2025 21:42
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants