fix gpu_id_step

[TomQuartz]

Motivation

There appears to be a bug in computing the gpu_id when gpu_id_step>1.

Example: TP=2 PP=2 gpu_id_step=2

• expect 0,2,4,6
• however, the first inner loop of tp_rank_range yields 0,2, the second yields 2,4

Modifications

compute the gpu_offset separately and multiply by gpu_id_step

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 @TomQuartz, 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 bug in the calculation of GPU IDs within a distributed computing environment, specifically when a gpu_id_step greater than one is configured. The change ensures that GPU assignments are correctly derived by applying the gpu_id_step to the combined offset from both pipeline and tensor parallelism ranks, preventing misconfigurations and enabling proper scaling across multiple GPUs.

Highlights

• Corrected GPU ID Calculation: The core logic for calculating gpu_id within the _launch_subprocesses function in engine.py has been revised. Previously, the server_args.gpu_id_step was only applied to the tensor parallelism (TP) rank component of the GPU offset. The fix now correctly computes a combined gpu_offset by summing the contributions from both pipeline parallelism (PP) and TP ranks, and then multiplies this total offset by server_args.gpu_id_step. This ensures accurate and consistent GPU assignment in distributed setups, particularly when gpu_id_step is greater than 1, resolving the reported bug.
• Minor Formatting Adjustment: A minor whitespace adjustment was made in an error message string related to data_parallel_rank within the async_generate function (line 266).

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

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

The pull request fixes a bug in computing the gpu_id when gpu_id_step > 1. The fix involves computing the gpu_offset separately and then multiplying it by gpu_id_step to get the correct gpu_id. There's a potential issue in the calculation of gpu_offset where the multiplication by server_args.gpu_id_step might be missing, which could lead to incorrect gpu_id assignment.

[TomQuartz]

Hi @zhaochenyang20 ! Would you please take some time to review this PR? Thanks a lot.