Skip to content

Feat/qwen image rollout engines#34

Draft
CjhHa1 wants to merge 2 commits into
Tencent-Hunyuan:mainfrom
CjhHa1:feat/qwen-image-rollout-engines
Draft

Feat/qwen image rollout engines#34
CjhHa1 wants to merge 2 commits into
Tencent-Hunyuan:mainfrom
CjhHa1:feat/qwen-image-rollout-engines

Conversation

@CjhHa1

@CjhHa1 CjhHa1 commented Jun 11, 2026

Copy link
Copy Markdown
Collaborator

Summary

Related Issue

Test Plan

Compatibility / Risk

Reviewer Notes

Checklist

  • I reviewed the changed code and removed unrelated/generated artifacts.
  • I updated tests, docs, and configs where needed, or explained why not.

aimicahchen added 2 commits June 11, 2026 14:40
Add the Qwen-Image SGLang recipe and wire the qwen_image family through the SGLang response path so packed rollout trajectories are unpacked for trainer replay.
Add the qwen_image_t2i vLLM-Omni modality, RL pipeline subclass, request/response translators, stage config, and LoRA-separate recipe for CFG-off Qwen-Image rollout.
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