[BugFix] Fix flash comm v1 bugs of Qwen3.5 series on A2 platform

[wangbj127]

What this PR does / why we need it?

Does this PR introduce any user-facing change?

How was this patch tested?

• vLLM version: v0.18.0
• vLLM main: vllm-project/vllm@35141a7

[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 addresses specific bugs affecting the Qwen3.5 series models when running on the A2 platform. The changes primarily refine the conditional logic for tensor communication and reduction operations, ensuring that multimodal and vision-language models correctly interact with flash_comm_v1 optimizations and distributed communication primitives under various operational modes.

Highlights

• Conditional Flash Communication Bypass: Adjusted the condition for bypassing flash_comm_v1_enabled in _maybe_pad_and_reduce_impl to correctly handle cases where a draft Vision-Language (VL) model is not involved in endpoint communication, ensuring proper tensor reduction behavior.
• Multimodal Model Tensor Reduction Logic: Modified the maybe_pad_and_reduce function in eagle_proposer.py to ensure that multimodal models correctly utilize flash_comm_v1_enabled for tensor operations, resolving an issue where they might have incorrectly bypassed these operations within the 'mtp' method.

🧠 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.

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 refines the conditions for applying tensor parallel communication and padding/reduction operations within the vllm_ascend framework. It modifies vllm_ascend/ops/register_custom_ops.py to include is_ep_comm in the bypass condition for draft and VL models when flash_comm_v1 is disabled. Additionally, in vllm_ascend/spec_decode/eagle_proposer.py, it adjusts the maybe_pad_and_reduce function to prevent multimodal models from undergoing padding and reduction when flash_comm_v1 is enabled, by removing an early return and adding a not self.is_multimodal_model check to the reduction logic. These changes aim to ensure correct behavior and potentially optimize performance for multimodal and other specific model configurations under different communication settings. There are no review comments to address.

[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.