Skip to content

Fix some bugs#169

Merged
tastelikefeet merged 6 commits intomodelscope:mainfrom
tastelikefeet:fix/0418-4
Apr 18, 2026
Merged

Fix some bugs#169
tastelikefeet merged 6 commits intomodelscope:mainfrom
tastelikefeet:fix/0418-4

Conversation

@tastelikefeet
Copy link
Copy Markdown
Collaborator

PR type

  • Bug Fix
  • New Feature
  • Document Updates
  • More Models or Datasets Support

PR information

Write the detail information belongs to this PR.

Experiment results

Paste your experiment result here(if needed).

Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

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

Code Review

This pull request introduces several bug fixes and feature enhancements across the library, including improved numerical stability for ORPO loss, robust LaTeX extraction for math rewards, and support for additional masks in data processing. Key updates include refined gradient accumulation handling, optimized logprob retrieval in vLLM sampling, and expanded unit normalization for evaluation. Review feedback suggests using idiomatic list comprehensions for prompt expansion in GRPO scripts to improve readability.

Comment thread cookbook/rl/grpo.py
Comment thread cookbook/rl/grpo_mm.py
@tastelikefeet tastelikefeet merged commit cdf6bad into modelscope:main Apr 18, 2026
1 of 4 checks passed
tastelikefeet added a commit that referenced this pull request Apr 18, 2026
(cherry picked from commit cdf6bad)
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