[3/3] Optimize: Cache CUDA device to reduce redundant calls during tensor l…

[GeLee-Q]

Motivation

• 1/3: [1/3] Optimize Slime Update Weights: Remove QWen3MOE Load Weight Overhead #8751 （23s => 22s)
• 2/3: [2/3] Optimize Slime Update Weights: Avoid GPU-to-CPU Device Sync when update expert weights #8753 (22s => 12s)
• [Discarded]3/3: [3/3] Optimize Slime Update Weights: Load Weight in non-blocking mode #8754 (12s => 10s)
• 3/3 [3/3] Optimize: Cache CUDA device to reduce redundant calls during tensor l… #8996 (10s => 8s)

Optimize the reinforcement learning weight update process by caching the CUDA device

This PR speeds up the weight update for the Qwen3-30B-a3b model in slime by about 1 second on average.

This PR maps to the red line

Modifications

Accuracy Test

Benchmark & Profiling

Checklist

• Format your code according to the Code Formatting with Pre-Commit.
• Add unit tests as outlined in the Running Unit Tests.
• Update documentation / docstrings / example tutorials as needed, according to Writing Documentation.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to Benchmark and Profiling and Accuracy Results.
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.
• Please feel free to join our Slack channel at https://slack.sglang.ai to discuss your PR.

[gemini-code-assist]

Summary of Changes

Hello @GeLee-Q, 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 optimize the reinforcement learning weight update process by caching the CUDA device. This change prevents redundant calls to "torch.cuda.current_device()", which can improve performance during tensor loading and processing.

Highlights

• CUDA Device Caching: The "update_weights_from_tensor" function now explicitly caches the "torch.cuda.current_device()" after applying "monkey_patch_torch_reductions()". This ensures the correct device is obtained and reused.
• Optimized Tensor Unwrapping: The "_unwrap_tensor" helper function has been modified to accept the "infered_device" as an argument. This eliminates repeated calls to "torch.cuda.current_device()" within "_unwrap_tensor" when processing multiple tensors, leading to more efficient device assignment.

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 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 or fill out our survey 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

1. 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. ↩

[gemini-code-assist]

Code Review

This pull request optimizes the update_weights_from_tensor function by caching the CUDA device, which avoids redundant calls to torch.cuda.current_device() within a loop. The change is correct and achieves its goal. I've pointed out a minor redundancy where monkey_patch_torch_reductions() is called in _unwrap_tensor even though its caller now handles it, which can be removed for better code clarity.

[pkking]

The failed test is affected by a connection issue which is fixed now, please rerun the tests, i will keep eyes on the Ascend runner cluster 👀