Skip to content

Conversation

@yuzho-amd
Copy link
Collaborator

Feat: Qwen2.5 VIT hipblaslt swizzle

@yuzho-amd yuzho-amd requested a review from LLLLKKKK as a code owner February 9, 2026 05:56
Copilot AI review requested due to automatic review settings February 9, 2026 05:56
Copy link
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

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

Pull request overview

Adds ROCm/hipBLASLt “swizzle” support for Qwen2.5-VL vision attention weights and swaps vision attention Linear layers to ROCm-optimized implementations after weight load.

Changes:

  • Override multimodal weight loading to optionally swizzle attention weights based on an env flag.
  • Add _replace_with_rocm_linear() to vision attention modules and invoke it post-load on ROCm.
  • Introduce ROCm platform check (is_hip) to gate the ROCm-specific path.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 6 comments.

File Description
rtp_llm/models/qwen2_5_vl/qwen2_5_vl.py Adds swizzle-aware multimodal weight loading and post-load replacement of attention Linear layers.
rtp_llm/models/qwen2_5_vl/modeling_qwen2_5_vl.py Adds ROCm Linear replacement helpers to vision attention classes and ROCm gating via is_hip().

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

@wht21
Copy link
Collaborator

wht21 commented Feb 9, 2026

internal source has been updated, please review the changes!

@zhaoan12-prc zhaoan12-prc force-pushed the zhaoan/vit_hipblaslt_swizzle branch from 5a13105 to 2f7a4e3 Compare February 11, 2026 10:24
@wht21
Copy link
Collaborator

wht21 commented Feb 11, 2026

internal source has been updated, please review the changes!

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.

3 participants