[Misc] Refactor AllReduceFusionPass. Remove parameter

[ilmarkov]

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.

Purpose

Removes unnecessary parameter from AllReduceFusionPass class, the parameter can be taken from config.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[gemini-code-assist]

Summary of Changes

Hello @ilmarkov, 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 streamlines the AllReduceFusionPass by removing a redundant constructor parameter. The max_token_num value, previously passed explicitly, is now consistently sourced from the VllmConfig object, simplifying the class's interface and its instantiation across the codebase. This change improves code cleanliness and maintainability by centralizing configuration access.

Highlights

• Refactoring AllReduceFusionPass Constructor: The AllReduceFusionPass class constructor (__init__) has been refactored to remove the max_token_num parameter. This parameter is now consistently retrieved directly from the VllmConfig object, which is already passed to the constructor.
• Updated Parameter Sourcing: Internal calls within AllReduceFusionPass that previously used the max_token_num parameter (specifically, calls to flashinfer_comm.trtllm_create_ipc_workspace_for_all_reduce_fusion) now fetch this value from config.compilation_config.pass_config.fi_allreduce_fusion_max_token_num.
• Simplified Instantiation: All instantiation sites for AllReduceFusionPass, including in PassManager and test files (tests/compile/test_fusion_all_reduce.py), have been updated to reflect the simplified constructor signature, passing only the VllmConfig object.

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 in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

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 issue 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 is currently in preview and 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 to provide feedback.

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 successfully refactors the AllReduceFusionPass class by removing the redundant max_token_num parameter from its constructor. The value is now correctly sourced from the VllmConfig object, which simplifies the class's interface and improves encapsulation. The changes in the test and pass manager files are consistent with this refactoring. I have one suggestion to further improve the readability of the implementation.

[gemini-code-assist]

While this change is functionally correct, this long access path is used multiple times within the __init__ method (here and a few lines below when creating FlashInferFusedAllReduceParams).

To improve readability and avoid repetition (DRY principle), consider extracting this value into a local variable at the beginning of the method.

For example:

def __init__(self, config: VllmConfig):
    # ...
    max_token_num = (
        config.compilation_config.pass_config.fi_allreduce_fusion_max_token_num
    )
    
    self.ipc_handles, workspace_tensor = (
        flashinfer_comm.trtllm_create_ipc_workspace_for_all_reduce_fusion(
            # ...
            max_token_num=max_token_num,
            # ...
        ))

    # ...
    self.allreduce_params = FlashInferFusedAllReduceParams(
        # ...
        max_token_num=max_token_num,
    )
    # ...

This would make the code cleaner and easier to maintain.

[mgoin]

LGTM, nice improvement