[yeahdongcn]

PLEASE FILL IN THE PR DESCRIPTION HERE.

Purpose

Allow SwiGLU on MUSA to use the fused op.

Test Plan

vLLM Version:

vLLM-Omni Commit:

Test Result

Fair A/B Comparison: Baseline → Final

Version	TP8 Median	Throughput Improvement
Baseline	105.351 s	—
#6281 only	101.492 s	+3.802%
#6283 + #6364	104.801 s	+0.525%
Final combined	101.811 s	+3.477%

BEFORE SUBMITTING: read CONTRIBUTING.md and run the precheck-pr skill with the code agent for a self-check against project conventions.
(anything written below this line will be removed by GitHub Actions)

[chatgpt-codex-connector]

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Credits must be used to enable repository wide code reviews.

[yeahdongcn]

This is a follow-up to #6283
Please @gcanlin take a look. Thanks!

[vllm-omni-review-bot]

This PR appears to belong to: docs/design/module/diffusion/diffusion_model_integration.md.

Module owners: @fhfuih @Bounty-hunter @wtomin @RuixiangMa

@yeahdongcn, please review your own changes and leave a short self-review comment describing what you checked. PRs without author self-review may not be assigned a reviewer.

Please take a look when you have a chance. If you would like an automated review, mention @vllm-omni-review-bot in a comment.

[Bounty-hunter]

need to add forward_hip, It seem that self.op only init when current_omni_platform.is_cuda() or current_omni_platform.is_rocm() or current_omni_platform.is_musa() in __init__

[yeahdongcn]

need to add forward_hip, It seem that self.op only init when current_omni_platform.is_cuda() or current_omni_platform.is_rocm() or current_omni_platform.is_musa() in __init__

rocm uses forward_hip (which falls back to forward_cuda by default), so leaving it as is should work.

[Bounty-hunter]

LGTM