[Misc][main2main] Adapt vllm-ascend to vLLM 0420

[Meihan-chen]

Commit range: 6f786f2..d886c26

1. Fix 'DPMetadata' object has no attribute 'max_tokens_across_dp_cpu' by [MoE Refactor] Remove MoE DP chunking vllm#39107
2. Fix 'Indexer' object has no attribute 'wk' by [Bugfix][Perf] Indexer upcast WK to BF16 for fusion vllm#38928
3. Fix 'float' object has no attribute 'language_model' by Measure encoder compile time seperate from llm backbone vllm#39240

What this PR does / why we need it?

Does this PR introduce any user-facing change?

How was this patch tested?

• vLLM version: v0.19.0
• vLLM main: vllm-project/vllm@6f786f2

[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]

Summary of Changes

Hello, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request aligns the vllm-ascend extension with recent upstream changes in vLLM, specifically focusing on FusedMoE router compatibility and updating the tracked vLLM commit hash. These changes ensure that the Ascend-specific operations remain functional with the latest core vLLM developments.

Highlights

• Zero-expert logical expert boundary: Updated Ascend FusedMoE implementations to utilize logical_num_experts, ensuring compatibility with upstream vLLM FusedMoE router changes.
• Main-branch synchronization: Bumped the vLLM main-branch commit reference to d886c26d4d4fef7d079696beb4ece1cfb4b008a8 across documentation and CI configuration files.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Ignored Files

• Ignored by pattern: .github/workflows/** (3)
  • .github/workflows/dockerfiles/Dockerfile.lint
  • .github/workflows/pr_test_full.yaml
  • .github/workflows/pr_test_light.yaml

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request updates the vLLM commit hash in the documentation and introduces a _get_num_logical_experts helper function across several MoE implementation files to correctly handle expert counts in zero_experts_compute. Feedback indicates that using global_num_experts as a fallback is incorrect because it includes redundant padding experts, which can disable zero-expert logic. Additionally, the logic should be adjusted to account for shared experts when mix_placement is enabled.

Suggested PR Title:

[Ops][Misc] Update vLLM commit and refine logical expert count in MoE

Suggested PR Summary:

### What this PR does / why we need it?
This PR updates the vLLM commit hash in the documentation and introduces a `_get_num_logical_experts` helper function across several MoE implementation files to correctly handle expert counts in `zero_experts_compute`. Feedback indicates that using `global_num_experts` as a fallback is incorrect because it includes redundant padding experts, which can disable zero-expert logic. Additionally, the logic should be adjusted to account for shared experts when `mix_placement` is enabled.

### Does this PR introduce _any_ user-facing change?
No.

### How was this patch tested?
CI passed with existing tests.

[gemini-code-assist]

The num_experts parameter in zero_experts_compute defines the boundary between real experts and padding (zero) experts. Using global_num_experts as the fallback is incorrect because it includes global_redundant_expert_num (padding), which effectively disables the zero-expert logic for those padding experts. The boundary should be the total number of non-padding experts (routed + shared).

Suggested change

	num_experts=_get_num_logical_experts(layer, global_num_experts),
	num_experts=_get_num_logical_experts(layer, global_num_experts - global_redundant_expert_num),

[gemini-code-assist]

Using global_num_experts as the fallback for num_experts in zero_experts_compute fails to identify padding experts when global_redundant_expert_num > 0.

Additionally, when mix_placement is active, logical_num_experts (which typically represents the number of routed experts) should be incremented by n_shared_experts to ensure shared experts are not incorrectly treated as zero experts. Since valid_global_expert_num is already calculated as the routed expert count, it should be used as the fallback for the routed portion.

Suggested change

	num_experts=_get_num_logical_experts(layer, global_num_experts),
	num_experts=_get_num_logical_experts(layer, valid_global_expert_num) + (n_shared_experts if mix_placement else 0),