[Fix] Allow kernel compilation for CUDA capability 8.7 by conroy-cheers · Pull Request #19328 · vllm-project/vllm (original) (raw)

Hello @conroy-cheers, 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!

Summary of Changes

Hello! Gemini here, providing a summary of this pull request to help everyone get up to speed quickly.

The primary goal of this PR, as indicated by the title and description, is to enable the compilation of kernels for CUDA compute capability 8.7. This addresses an issue where kernels built only for 8.0 would not run on hardware like the Jetson Orin (which uses 8.7), even though they might work on 8.6 or 8.9. The change involves adding '8.7' to the list of supported architectures in the build configuration for specific kernel types (Marlin and Scaled Dot Product Attention).

Additionally, this PR includes changes to remove the opentelemetry-semantic-conventions-ai dependency from the common requirements file and the OpenTelemetry example README. This seems unrelated to the CUDA capability fix but is part of the changes presented.

Highlights

CUDA 8.7 Support: Adds support for compiling kernels for CUDA compute capability 8.7, specifically targeting devices like the Jetson Orin where existing 8.0 kernels were not compatible.
Dependency Removal: Removes the opentelemetry-semantic-conventions-ai dependency from the project's common requirements and the OpenTelemetry example documentation.

Changelog

CMakeLists.txt
- Modified line 311: Added '8.7' to the list of architectures for MARLIN_ARCHS.
- Modified line 457: Added '8.7' to the list of architectures for SCALED_MM_2X_ARCHS.
- Modified line 687: Added '8.7' to the list of architectures for MARLIN_MOE_ARCHS.
examples/online_serving/opentelemetry/README.md
- Removed line 10: Deleted the line specifying the opentelemetry-semantic-conventions-ai dependency in the installation instructions.
requirements/common.txt
- Removed line 50: Deleted the line specifying the opentelemetry-semantic-conventions-ai dependency. 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.

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. ↩