[Isotr0py]

Reverts #19319

[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 @Isotr0py, 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!

I've created this pull request to revert a previous change that introduced FP32 support to the V1 engine. This involves removing the specific logic and tests related to FP32 handling and restricting the V1 engine's supported data types to FP16 and BF16.

Highlights

• Revert FP32 Support for V1 Engine: This pull request reverts the changes introduced in PR #19319, specifically removing the added support for FP32 dtype within the V1 engine.
• Restrict V1 Engine Dtypes: The V1 engine is now explicitly restricted to only support FP16 and BF16 dtypes, disallowing others like FP32.
• Remove FP32 Attention Backend Fallback: The logic that selected the Flex Attention backend for non-FP16/BF16 dtypes (like FP32) in the V1 CUDA platform has been removed.
• Remove FP32 Test Cases: A test case specifically designed to verify the FP32 fallback behavior in the attention backend selector has been removed.

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 configureGemini 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 reverts the changes from PR #19319, which added fp32 support to the V1 engine through FlexAttention. The changes in this revert correctly undo the previous modifications:

• Test Code: The fp32-specific test (test_fp32_fallback) and related test setups (like torch.set_default_dtype(torch.float16) and the import torch if it was solely for these) are removed from tests/kernels/attention/test_attention_selector.py and tests/v1/worker/test_gpu_model_runner.py.
• Argument Handling: The dtype check in vllm/engine/arg_utils.py is re-instated, restricting the V1 engine to fp16 and bf16 data types.
• Backend Selection: The logic for selecting FlexAttentionBackend for fp32 in vllm/platforms/cuda.py is removed.

The revert appears to be clean and consistent with the goal of removing fp32 support for the V1 engine as introduced by the reverted PR. No issues of medium, high, or critical severity were identified in the diff.