Update rocm7.2 Dockerfile to install amdsmi for QuickReduce Initialization#19091
Update rocm7.2 Dockerfile to install amdsmi for QuickReduce Initialization#19091HaiShaw merged 2 commits intosgl-project:mainfrom
Conversation
Added installation of AMD SMI Python package from ROCm distribution so that QuickReduce can initialize.
clintg6
requested review from
Fridge003,
HaiShaw,
ishandhanani and
ispobock
as code owners
Summary of ChangesHello @clintg6, 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 dependency issue within the Highlights
🧠 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
Using Gemini Code AssistThe 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
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 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
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request addresses an issue where QuickReduce fails to initialize in rocm7.2 Docker images due to a missing amdsmi package. The fix involves adding a RUN instruction to install this package from the ROCm distribution within the Dockerfile, and also adds vim to the list of installed packages. The changes are clear and directly address the problem. I have a couple of suggestions to further improve the Dockerfile based on best practices: using --no-cache-dir with pip install to optimize the Docker image size, and reconsidering the inclusion of vim in the image, as it can increase image size and attack surface, which is generally avoided for production environments.
docker/rocm720.Dockerfile
Outdated
| RUN apt-get purge -y sccache; python -m pip uninstall -y sccache; rm -f "$(which sccache)" | ||
|
|
||
| # Install AMD SMI Python package from ROCm distribution | ||
| RUN cd /opt/rocm/share/amd_smi && python3 -m pip install . |
There was a problem hiding this comment.
To optimize Docker image size by preventing pip from storing downloaded packages in its cache, it's a good practice to use the --no-cache-dir flag with pip install.
RUN cd /opt/rocm/share/amd_smi && python3 -m pip install --no-cache-dir .
| libprotobuf-dev protobuf-compiler libgflags-dev libsqlite3-dev \ | ||
| python3 python3-dev python3-setuptools python3-pip python3-apt \ | ||
| gcc libtinfo-dev zlib1g-dev libedit-dev libxml2-dev \ | ||
| gcc libtinfo-dev zlib1g-dev libedit-dev libxml2-dev vim \ |
There was a problem hiding this comment.
Adding vim is useful for debugging, but it increases the image size and attack surface. For production images, it's generally recommended to keep them as minimal as possible. If vim is only needed for development or debugging, consider creating a separate development Dockerfile or installing it on-the-fly when needed inside a running container.
There was a problem hiding this comment.
vim is standard in the rocm7.0 SGLang docker images
Updated pip install command for AMD SMI package to use --no-cache-dir.
…o xverse_moe * 'xverse_moe' of https://github.com/xiaobaicxy/sglang: (275 commits) fix: add missing blank line after docstring in serving_transcription.py (sgl-project#19206) Whisper model support & `/v1/audio/transcriptions` endpoint & benchmark (sgl-project#16983) fix: patch docker image fixes (sgl-project#19100) [PD-Disagg] Unify prefill info data transition flow, all with `PrefillServerInfo` (sgl-project#19195) [CI] Tiny enhance the dp attention load blance benchmark (sgl-project#19194) add new ci user (sgl-project#19133) [CI] fix the teardown output of disaggregation test (sgl-project#19193) [PD-Disagg] Support query dp rank from bootstrap server. (sgl-project#19168) [Kernel Slimming] Migrate AWQ marlin repack kernel to JIT (sgl-project#18949) [Diffusion] Match rotary_embedding module name style (sgl-project#19179) [Refactor] Split rotary_embedding.py into a modular package (sgl-project#19144) [NPU] bump sgl-kernel-npu to 2026.02.01.post2 (sgl-project#19178) Use single mma warp group for short q_len in FA to optimize decoding performance (sgl-project#18985) Reorganize topk logic to clean up code and expose logical experts (sgl-project#16945) [ROCm] Use unreg path for custom all-reduce during CUDA graph capture (sgl-project#19162) [diffusion] feat: detect Flux2 custom VAE path from component_paths (sgl-project#19170) [AMD] ENV flags tuning and cleanup (sgl-project#19176) Fix bench_one_batch_server by moving the print statements (sgl-project#19175) Update rocm7.2 Dockerfile to install amdsmi for QuickReduce Initialization (sgl-project#19091) Revert "Refactor graph input buffers (sgl-project#18991)" (sgl-project#19173) ...
2 participants
Description:
This update modifies the
rocm720.Dockerfileto include the AMD SMI Python package from the installed ROCm distribution. This enables QuickReduce to initialize properly when using therocm7.2Docker images.Motivation:
The
rocm7.2rocm/sgl-devimages currently lack the AMD SMI package, preventing users from utilizing features like quantized allreduce (e.g., QuickReduce). This results in no improvements in latency or throughput for related tasks.Current Issue:
The issue stems from using the base image
rocm/pytorch:rocm7.2_ubuntu22.04_py3.10_pytorch_release_2.9.1, which is missing theamdsmipackage. Without this package, QuickReduce cannot be initialized as intended.Example Screenshot:
Change Summary:
Before:
After:
Notes:
amdsmipackage is present and accessible for users working with AMD hardware.Why This is Important:
This update ensures that users relying on QuickReduce for efficient allreduce will now have the necessary packages available, improving the usability of the ROCm 7.2 images in production and testing environments.