Skip to content

Conversation

@dsikka
Copy link
Collaborator

@dsikka dsikka commented Oct 30, 2025

SUMMARY:

  • Update the qwen3_next_moe definition to use the moe context
  • Update example to use the correct arguement
  • Add tests
  • Update qwen3_moe model definition to implement restore

Testing:

  • All modeling tests pass

@github-actions
Copy link

👋 Hi! Thank you for contributing to llm-compressor. Please add the ready label when the PR is ready for review.

Note: This is required to complete the testing suite, please only add the label once the PR is code complete and local testing has been performed.

@dsikka dsikka added the ready When a PR is ready for review label Oct 30, 2025
@dsikka dsikka added qwen For any PR / issue related to Qwen support nvfp4 For any PR / issue related to NVFP4 support labels Oct 30, 2025
Copy link
Collaborator

@kylesayrs kylesayrs left a comment

Choose a reason for hiding this comment

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

Tests?

@dsikka
Copy link
Collaborator Author

dsikka commented Oct 30, 2025

Tests?

I'll add. Running basic evals on vLLM to start

@dsikka dsikka marked this pull request as ready for review October 31, 2025 15:56
@dsikka dsikka requested a review from kylesayrs October 31, 2025 16:03
@dsikka dsikka enabled auto-merge (squash) November 1, 2025 05:08
@HDCharles HDCharles self-requested a review November 3, 2025 19:00
Copy link
Collaborator

@HDCharles HDCharles left a comment

Choose a reason for hiding this comment

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

otherwise looks good

@dsikka dsikka merged commit ee755a4 into main Nov 3, 2025
9 checks passed
@dsikka dsikka deleted the fix_qwen3 branch November 3, 2025 23:44
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

nvfp4 For any PR / issue related to NVFP4 support qwen For any PR / issue related to Qwen support ready When a PR is ready for review

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants