Improve perf tuning docs #7071
Improve perf tuning docs #7071
Conversation
There was a problem hiding this comment.
Summary of Changes
Hello @merrymercy, 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 aims to enhance the clarity and coverage of the project's documentation, particularly focusing on performance tuning and troubleshooting. It introduces new sections in the hyperparameter tuning guide to explain the impact and tuning strategies for --mem-fraction-static and --cuda-graph-max-bs. The information on torch.compile caching has been relocated to a dedicated reference page, and the troubleshooting guide now offers more detailed steps for addressing CUDA Out of Memory errors, including a specific scenario involving input logprobs.
Highlights
- Documentation Expansion: Added detailed guidance on tuning
--mem-fraction-staticfor KV cache capacity and--cuda-graph-max-bsfor CUDA graph usage in the hyperparameter tuning guide. - Documentation Refactoring: Moved the documentation regarding
torch.compilecaching to a new, dedicated file (docs/references/torch_compile_cache.md). - Troubleshooting Improvements: Enhanced the troubleshooting guide with more specific advice for CUDA Out of Memory errors, including a new tip for handling long prompts with input logprobs.
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
-
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. ↩
There was a problem hiding this comment.
Code Review
This pull request significantly improves the performance tuning documentation by reorganizing content, adding more detailed explanations for various parameters, and introducing new sections for mem-fraction-static and cuda-graph-max-bs. The torch.compile caching information has been moved to a dedicated, more focused page. The troubleshooting guide for OOM errors has also been enhanced with more specific advice, including a new tip related to logprob_start_len. Overall, these changes make the documentation clearer, more comprehensive, and easier for users to navigate when tuning performance and troubleshooting issues. A few minor improvements are suggested for consistency and clarity in the markdown files.
No description provided.