Fix different device type adjustment in PP #7760
Conversation
There was a problem hiding this comment.
Summary of Changes
Hello @Qiaolin-Yu, 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 refactors device management within the distributed communication utilities, moving away from hardcoded CUDA device references to a more flexible and generalized approach. The primary goal is to ensure that tensor operations and data transfers correctly adapt to the specific device configured for the system, improving compatibility and maintainability for different hardware environments.
Highlights
- Standardized Device Handling: Replaced direct references to
torch.cuda.current_device()with a more generalizedself.deviceattribute within theParallelStateclass'ssend_objectandrecv_objectmethods. This ensures that tensor creation and movement (object_tensor,size_tensor) are consistently tied to the configured device of the parallel state. - Generalized Tensor Movement: Updated tensor movement operations from
.cuda()to the more versatile.to()method inparallel_state.pyandutils.py. This change, coupled with using a dynamicdevicevariable (obtained viaget_device()) inpoint_to_point_pyobj, allows for better compatibility with different device types beyond just CUDA GPUs. - Improved Point-to-Point Communication: Ensured that all tensors involved in point-to-point object and data transfers (
point_to_point_pyobjfunction) are correctly allocated and moved to the appropriate device, enhancing the robustness of distributed communication across potentially varied hardware setups.
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 configure Gemini 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 refactors device handling to be more generic, replacing CUDA-specific calls with device-agnostic PyTorch functions. The changes in parallel_state.py and utils.py are well-aligned with this goal, improving code portability. I've suggested one minor optimization to use torch.empty for a receive buffer to avoid unnecessary memory initialization, consistent with other parts of the codebase.
|
@zhyncs would you mind merging this? |
This reverts commit 3bc43c6.
Motivation
Fix https://github.com/sgl-project/sglang/pull/7673/files#r2183519896
Modifications
Checklist