Skip to content

Simplify context-parallel seq KL helper#11

Merged
cklxx merged 1 commit intocodex/optimize-training-time-for-context-parallelismfrom
fix-design-issues-in-loss-and-data-source
Dec 22, 2025
Merged

Simplify context-parallel seq KL helper#11
cklxx merged 1 commit intocodex/optimize-training-time-for-context-parallelismfrom
fix-design-issues-in-loss-and-data-source

Conversation

@cklxx
Copy link
Owner

@cklxx cklxx commented Dec 22, 2025

Summary

  • streamline the context-parallel sequence KL helper to rely on the caller-provided process group without redundant fallbacks

Testing

  • python -m ruff check .
  • python -m pytest (fails: missing dependencies such as flash_attn and test_qwen modules)

Codex Task

@cklxx cklxx merged commit 8687913 into codex/optimize-training-time-for-context-parallelism Dec 22, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant