[eagle][cp] fix eagle_cp enable bug2

[lilinsiman]

What this PR does / why we need it?

Fix acceptance and high-concurrency bug in eagle3 and cp enabled

Does this PR introduce any user-facing change?

no

How was this patch tested?

tests and ut

• vLLM version: v0.16.0
• vLLM main: vllm-project/vllm@4034c3d

[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 and fixes acceptance and high-concurrency bugs within the eagle3 and context parallel (cp) features. The changes primarily involve refining the internal logic for processing query lengths and batch masks in the attention mechanism, and correcting how auxiliary hidden states are aggregated in the model runner, leading to improved stability and correctness in these components.

Highlights

• Context Parallel Attention Logic: Corrected the variable used for slicing query lengths and removed redundant batch mask processing in the context parallel attention mechanism.
• Model Runner Hidden State Processing: Adjusted the concatenation of auxiliary hidden states in the model runner to ensure correct aggregation without unnecessary slicing.

🧠 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.

Changelog

• vllm_ascend/attention/context_parallel/attention_cp.py
  • Corrected the variable used for slicing query_lens from num_decode_tokens to num_decodes.
  • Removed a redundant torch.repeat_interleave operation for batch_chunk_seq_mask.
  • Updated the _build_chunked_context_metadata function call to no longer pass the batch_chunk_seq_mask parameter.
• vllm_ascend/worker/model_runner_v1.py
  • Adjusted the concatenation of aux_hidden_states by removing an extraneous slice [:num_scheduled_tokens].

Activity

• No specific activity (comments, reviews, or progress updates) has been recorded for this pull request yet.

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.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

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 aims to fix bugs related to the Eagle speculative decoding method when context parallelism is enabled. The changes involve correcting an indexing bug in attention_cp.py, removing what appears to be dead code, and modifying how auxiliary hidden states are handled in model_runner_v1.py. While most changes seem correct, the modification in model_runner_v1.py appears to introduce a new issue by removing a necessary slice, and it misses an existing bug where hidden_states is not sliced. I've provided a critical comment with a suggested fix to ensure correctness and consistency.

[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.