Fix FA2 fallback for Blackwell V1

[mgoin]

elif was not the right choice, as we want FA2 to be a constant fallback on V1 so we can use it for Blackwell when FlashInfer isn't installed

[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 @mgoin, 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!

My aim with this pull request is to solidify FlashAttention 2 (FA2) as the constant fallback attention backend for the V1 engine on GPUs with SM 8.0+ capability, including Blackwell. This change ensures FA2 is always prioritized for these advanced architectures, providing a reliable and performant default even when alternative optimized libraries like FlashInfer are not present.

Highlights

• Attention Backend Selection Logic: I've modified the get_attn_backend_cls function in vllm/platforms/cuda.py. The elif condition for selecting the FlashAttention backend based on SM 8.0+ device capability has been changed to an if condition. This ensures FlashAttention is consistently considered as the default backend for these GPUs on the V1 engine, even if FlashInfer is not installed.

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 fixes a logic error in the attention backend selection for the V1 engine. The change from elif to if for the FlashAttention backend check ensures that FlashAttention (FA2) can correctly serve as a fallback when FlashInfer is selected but not available, particularly for newer GPU architectures like Blackwell. This aligns with the intended behavior of having FA2 as a constant fallback in such scenarios. The change is well-targeted and improves the robustness of backend selection.

[robertgshaw2-redhat]

Do you think we should hard fail if we do not have FI on B200?

[mgoin]

Do you think we should hard fail if we do not have FI on B200?

I think we need FA2 for feature support. Currently there will be a warning printed if FI cannot be installed, so I think this is sufficient to let the user know the caveat

                    logger.info_once(
                        "FlashInfer failed to import for V1 engine on "
                        "Blackwell (SM 10.0) GPUs; it is recommended to "
                        "install FlashInfer for better performance.")

[houseroad]

Our CI seems really unstable.