[docker] fix flashinfer-cubin version mismatch#629
[docker] fix flashinfer-cubin version mismatch#629yueming-yuan wants to merge 2 commits intomainfrom
Conversation
Summary of ChangesHello @yueming-yuan, 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 resolves a version mismatch issue within the Docker development environment by precisely pinning the versions of Highlights
Changelog
Using Gemini Code AssistThe 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 by creating a comment using either
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 Limitations & Feedback Gemini Code Assist 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. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. 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
|
There was a problem hiding this comment.
Code Review
The pull request addresses a version mismatch between flashinfer_python and flashinfer-cubin by pinning both to version 0.6.2. This is a necessary fix for stability when using pre-built kernels. However, using the exact equality operator (==) in the Dockerfile can lead to build failures because CUDA-specific wheels often include local version suffixes (e.g., +cu121) which are not ignored by the == operator. I have suggested using a version range to ensure compatibility with these wheels while still enforcing the 0.6.2 release version.
| # Current sglang requires "flashinfer_python>=0.6.2" | ||
| RUN pip uninstall flashinfer-jit-cache -y | ||
| RUN pip install "flashinfer_python>=0.6.2" --no-deps | ||
| RUN pip install "flashinfer_python==0.6.2" "flashinfer-cubin==0.6.2" --no-deps |
There was a problem hiding this comment.
Using an exact version match (==0.6.2) for CUDA-dependent packages like flashinfer is risky. Pip's == operator requires an exact string match, including any local version identifiers like +cu121 or +cu124. Since flashinfer wheels are typically tagged with the CUDA version, pip install "flashinfer_python==0.6.2" will fail if only 0.6.2+cu121 is available.
Switching to a version range like >=0.6.2,<0.6.3 allows pip to match the 0.6.2 release regardless of the local CUDA suffix, while still preventing an accidental upgrade to 0.6.3.
RUN pip install "flashinfer_python>=0.6.2,<0.6.3" "flashinfer-cubin>=0.6.2,<0.6.3" --no-deps
No description provided.