Fix torch compile cache dir conflicts when enforce_eager=False
#368
Conversation
Summary of ChangesHello @pan-x-c, 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 addresses a critical issue where multiple vLLM instances running simultaneously with Highlights
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
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request aims to fix torch.compile cache directory conflicts when multiple vLLM instances are started with enforce_eager=False. It does this by setting the VLLM_CACHE_ROOT environment variable to a path that includes config.bundle_indices, making the cache path unique for different instances. My review identifies a potential issue where the fix might not be effective if config.bundle_indices is empty, as multiple instances would still share the same cache directory. I've suggested a more robust implementation using the process ID as a fallback to ensure a unique cache path in all relevant scenarios.
|
/unittest-module-explorer |
Summary
Skipped
Tests
Github Test Reporter by CTRF 💚 |
2 participants
Description
Fix compile cache conflicts when multiple vLLM instances are started at the same time.
Please refer to the following for details.
vllm-project/vllm#24601
Checklist
Please check the following items before code is ready to be reviewed.