[1/N][refactor] torchair fused_moe refactor

[hust17yixuan]

What this PR does / why we need it?

Move torchair related fused_moe section into torchair_fused_moe to make the code clear. Next step we'll remove all torchair related code outside of torchair_fused_moe .

Does this PR introduce any user-facing change?

No

How was this patch tested?

vLLM version: v0.10.0
vLLM main: vllm-project/vllm@08d5f71

• vLLM version: v0.10.1.1
• vLLM main: vllm-project/vllm@170e8ea

[gemini-code-assist]

Code Review

This pull request refactors the fused_moe logic for torchair by moving it into a new, dedicated file, vllm_ascend/torchair/ops/torchair_fused_moe.py. This is a good step towards better code organization. The changes in torchair_deepseek_v2.py correctly adopt the new refactored module. However, I've identified a critical issue in the newly created torchair_fused_moe.py file which appears to be a copy-paste error from the refactoring. This will cause a NameError at runtime and needs to be addressed.

[gemini-code-assist]

AscendUnquantizedFusedMoEMethod is not defined or imported in this file, which will lead to a NameError at runtime. Based on the refactoring, it seems the check should be against TorchairAscendUnquantizedFusedMoEMethod, which is defined in this file.

Suggested change

	if envs_ascend.VLLM_ASCEND_ENABLE_MOE_ALL2ALL_SEQ and isinstance(
	self.quant_method, AscendUnquantizedFusedMoEMethod):
	if envs_ascend.VLLM_ASCEND_ENABLE_MOE_ALL2ALL_SEQ and isinstance(
	self.quant_method, TorchairAscendUnquantizedFusedMoEMethod):

[github-actions]

👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:‌‌

• A PR should do only one thing, smaller PRs enable faster reviews.
• Every PR should include unit tests and end-to-end tests ‌to ensure it works and is not broken by other future PRs.
• Write the commit message by fulfilling the PR description to help reviewer and future developers understand.

If CI fails, you can run linting and testing checks locally according Contributing and Testing.

[codecov]

Codecov Report

❌ Patch coverage is 65.55091% with 247 lines in your changes missing coverage. Please review.
✅ Project coverage is 77.98%. Comparing base (3629bc4) to head (57ceed3).
⚠️ Report is 9 commits behind head on main.

Files with missing lines	Patch %	Lines
vllm_ascend/torchair/ops/torchair_fused_moe.py	55.12%	245 Missing ⚠️
tests/ut/torchair/ops/test_torchair_fused_moe.py	98.80%	2 Missing ⚠️

❌ Your patch status has failed because the patch coverage (65.55%) is below the target coverage (80.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2438      +/-   ##
==========================================
- Coverage   78.04%   77.98%   -0.06%     
==========================================
  Files         132      134       +2     
  Lines       17557    18519     +962     
==========================================
+ Hits        13702    14442     +740     
- Misses       3855     4077     +222     

Flag	Coverage Δ
unittests	77.98% <65.55%> (-0.06%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[wangxiyuan]

@wxsIcey please note that fused moe in torchair module is changed to this. So for your Fused Moe custom ops refactor work, you can just ignore torchair case.