Skip to content

fix: fix the flash attention with qwenimage. - #38

Merged
Hermit-w merged 1 commit into
THU-MIG:mainfrom
Hermit-w:fix/qwen-fused-attn-kv-scale
Jul 17, 2026
Merged

Hermit-w merged 1 commit into
THU-MIG:mainfrom
Hermit-w:fix/qwen-fused-attn-kv-scale

Conversation

@Hermit-w

Copy link
Copy Markdown
Collaborator

Summary

Describe the change and the affected model/backend/API surface.

Validation

  • CPU build or tests
  • CUDA build or tests, if relevant
  • Python tests, if relevant
  • Correctness or image/video quality check, if generation behavior changed
  • Benchmark numbers, if performance changed

Notes

Include hardware, model paths in generic form, and any known limitations. Do not
include secrets, private paths, model weights, or generated large artifacts.

@Hermit-w
Hermit-w force-pushed the fix/qwen-fused-attn-kv-scale branch from 1e74f4c to 4670cb1 Compare July 17, 2026 16:17
@Hermit-w
Hermit-w merged commit 65f54d5 into THU-MIG:main Jul 17, 2026
2 checks passed
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