Skip to content

Add log_multimodal param to GRPOConfig and RLOOConfig to control image logging#5408

Open
apardyl wants to merge 3 commits intohuggingface:mainfrom
apardyl:logmultimodal
Open

Add log_multimodal param to GRPOConfig and RLOOConfig to control image logging#5408
apardyl wants to merge 3 commits intohuggingface:mainfrom
apardyl:logmultimodal

Conversation

@apardyl
Copy link
Copy Markdown
Contributor

@apardyl apardyl commented Mar 30, 2026

What does this PR do?

This PR introduces a log_multimodal option to GRPOConfig and RLOOConfig, which controls whether multimodal data is logged alongside completions.

When multiple high-resolution images are used during training, the log can quickly grow in size. While viewing images together with completions is very useful for debugging, disabling this feature for longer training runs can significantly reduce (cloud) storage usage.

The log_multimodal option defaults to True, preserving the current behavior.

Before submitting

  • This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
  • Did you read the contributor guideline, Pull Request section?
  • Was this discussed/approved via a GitHub issue? Please add a link to it if that's the case.
  • (autodoc) Did you make sure to update the documentation with your changes?
  • Did you write any new necessary tests?

AI writing disclosure

We welcome the use of AI tools to help with contributions. For transparency and to help us improve our review process, please indicate the level of AI involvement in this PR.

  • No AI usage: the PR was written entirely by a human.
  • AI-assisted: some parts were suggested or improved by AI, but the PR was written and reviewed by a human.
  • AI-generated: the PR was mostly or fully generated by an AI tool.

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR.


Note

Low Risk
Low risk: gated logging-only behavior change with default preserving existing output, affecting only whether image data is gathered/stored in completion logs.

Overview
Adds a new log_multimodal option to GRPOConfig and RLOOConfig (default True) to control whether images/multimodal payloads are included in completion logging.

Updates GRPOTrainer and RLOOTrainer to honor this flag by skipping the images gather/append step when disabled, reducing log size for high-resolution multimodal training runs.

Written by Cursor Bugbot for commit 9dbbd3c. This will update automatically on new commits. Configure here.

@apardyl apardyl changed the title Add log_multimodal param to GRPOConfig to control image logging Add log_multimodal param to GRPOConfig and RLOOConfig to control image logging Mar 30, 2026
Copy link
Copy Markdown

@cursor cursor bot left a comment

Choose a reason for hiding this comment

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

Cursor Bugbot has reviewed your changes and found 1 potential issue.

Fix All in Cursor

Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.

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.

1 participant