[Quickfix] Add the missing apply_router_weight_on_input in FusedMoE init

[MengqingCao]

What this PR does / why we need it?

Add the missing apply_router_weight_on_input in FusedMoE init
Quick fix on #2268 (comment)

Does this PR introduce any user-facing change?

N/A

How was this patch tested?

CI passed with existing test.

• vLLM version: v0.10.0
• vLLM main: vllm-project/vllm@6807af8

[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.

[gemini-code-assist]

Code Review

This pull request aims to fix an issue where the apply_router_weight_on_input parameter was not being passed to the FusedMoE base class constructor. The change correctly adds this parameter to the super().__init__ call.

However, my review found that this change by itself is not sufficient to enable the feature, as the parameter is not subsequently used in the AscendFusedMoE's logic. A follow-up change is required in AscendUnquantizedFusedMoEMethod.apply to pass this parameter to the fused_experts function. I've added a comment with more details.

Also, please note that the pull request title seems to contain a typo; it mentions e_score_correction_bias while the change is for apply_router_weight_on_input.

[gemini-code-assist]

Passing apply_router_weight_on_input to the superclass constructor is the correct first step. However, for this parameter to have any effect, it needs to be utilized by the MoE implementation.

Currently, the apply method in AscendUnquantizedFusedMoEMethod does not pass this parameter down to the fused_experts function where it is actually used. This means the feature controlled by apply_router_weight_on_input will remain inactive despite this change.

A follow-up change will be needed in vllm_ascend/ops/fused_moe.py to make this functional. For example, in AscendUnquantizedFusedMoEMethod.apply:

# around line 1170
return fused_experts(
    ...
    expert_map=expert_map,
    apply_router_weight_on_input=layer.apply_router_weight_on_input
)

[ApsarasX]

title: e_score_correction_bias -> apply_router_weight_on_input

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 75.74%. Comparing base (992271b) to head (ee6b871).
⚠️ Report is 3 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2348   +/-   ##
=======================================
  Coverage   75.74%   75.74%           
=======================================
  Files         118      118           
  Lines       13525    13525           
=======================================
  Hits        10245    10245           
  Misses       3280     3280           

Flag	Coverage Δ
unittests	75.74% <ø> (ø)

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.