test_attention compat with coming xformers change

[bottler]

Summary:
Update a test in vllm to not depend on internal details of the LowerTriangularMaskWithTensorBias object in xformers. This makes it work both with and without an expected upcoming change to xformers.

Testing: this is a small change to a vllm test which leaves it passing.

Context: the xformers library is expecting to change bias types back to not inherit tensors - basically facebookresearch/xformers@96fd480 will be reverted. That inheritance is no longer needed for torch.compile, and in fact makes life harder for torch.export.

Test Plan:
See CI on D75689669

Rollback Plan:

Differential Revision: D77789850

[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.

🚀

[facebook-github-bot]

This pull request was exported from Phabricator. Differential Revision: D77789850

[gemini-code-assist]

Summary of Changes

Hello @bottler, 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 focuses on future-proofing a specific test within the vllm codebase. The change ensures that the test for multi-query/kv attention remains functional and passes even after an anticipated breaking change in the underlying xformers library, by making the test less dependent on internal implementation details of xformers bias objects.

Highlights

• Test Compatibility Update: Updated the test_multi_query_kv_attention function in tests/kernels/attention/test_attention.py to ensure compatibility with an upcoming change in the xformers library. The materialize call for xformers.AttentionBias objects now explicitly constructs the target shape (num_query_heads, i, i) instead of relying on the internal b.shape attribute, making the test more robust to internal xformers API changes.

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 updates a test for multi-query attention to ensure compatibility with an upcoming change in the xformers library. Explicitly constructing the shape passed to the materialize method improves robustness.