[AMD] fix: hip rotary fallback avoiding CUDA JIT

[alphabetc1]

Motivation

to fix #18812

ROCm CI fails in test/registered/test_srt_backend.py during runtime initialization.

On HIP, RotaryEmbedding dispatches to forward_cuda via MultiPlatformOp.
RotaryEmbedding fallback is selected on HIP and, after #17934, imports
sglang.jit_kernel.pos_enc, which JIT-compiles with cuda_files.
In ROCm-only environments, this triggers CUDA toolchain discovery and fails with:

RuntimeError: Could not find CUDA installation. Please set CUDA_HOME environment variable.
see: https://github.com/sgl-project/sglang/actions/runs/22000523146/job/63571754356?pr=18602

Root Cause

Regression introduced by 7d4ae057e ([Kernel] Add JIT rotary_embedding_kernel),
which switched HIP/CUDA fallback import from prebuilt sgl_kernel.rotary_embedding
to JIT sglang.jit_kernel.pos_enc.

Fix

Override RotaryEmbedding.forward_hip() to call forward_native() directly,
bypassing the CUDA JIT fallback path on HIP.

Modifications

Accuracy Tests

Benchmarking and Profiling

Checklist

• Format your code according to the Format code with pre-commit.
• Add unit tests according to the Run and add unit tests.
• Update documentation according to Write documentations.
• Provide accuracy and speed benchmark results according to Test the accuracy and Benchmark the speed.
• Follow the SGLang code style guidance.

Review Process

1. Ping Merge Oncalls to start the PR flow. See the PR Merge Process.
2. Get approvals from CODEOWNERS and other reviewers.
3. Trigger CI tests with comments or contact authorized users to do so.
   • /tag-run-ci-label, /rerun-failed-ci, /tag-and-rerun-ci
4. After green CI and required approvals, ask Merge Oncalls to merge.

[gemini-code-assist]

Summary of Changes

Hello @alphabetc1, 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 in ROCm environments where the RotaryEmbedding operation was inadvertently triggering CUDA-specific JIT compilation, leading to runtime errors. The change ensures that HIP-based systems correctly utilize their native implementation, thereby stabilizing CI pipelines and improving platform compatibility by preventing unintended CUDA toolchain discovery.

Highlights

• ROCm CI Fix: Resolved a CI failure in ROCm environments where RotaryEmbedding on HIP incorrectly attempted to use a CUDA JIT fallback, causing CUDA_HOME errors due to missing CUDA toolchain discovery.
• HIP Dispatch Logic: Modified the RotaryEmbedding class to explicitly override forward_hip(), directing it to call forward_native() and bypass the problematic CUDA JIT compilation path.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Changelog

• python/sglang/srt/layers/rotary_embedding.py
  • Added a forward_hip method to prevent unintended CUDA JIT fallback.

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 by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

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 pull request 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. 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

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. ↩

[alphabetc1]

/rerun-stage stage-a-test-1-amd

[github-actions]

✅ Triggered stage-a-test-1-amd to run independently (skipping dependencies).

[github-actions]

🔗 View workflow run

[gemini-code-assist]

Code Review

This pull request addresses a critical runtime error for ROCm-only environments within the RotaryEmbedding layer. The issue stemmed from the default dispatch mechanism falling back to a CUDA-specific JIT compilation path on HIP devices, leading to a crash. The fix is to introduce a forward_hip method that explicitly calls the generic forward_native implementation. This correctly bypasses the problematic CUDA path. The change is well-targeted, clean, and includes a helpful comment explaining the reasoning. I approve this fix.

[alphabetc1]

ci passed: https://github.com/sgl-project/sglang/actions/runs/22001696442/job/63575711455

[alphabetc1]

/tag-and-rerun-ci

[alphabetc1]

/rerun-failed-ci 1

[alphabetc1]

/rerun-failed-ci 1

[alphabetc1]

fixed by #18920