AMD MI300X/MI325X/MI355X support for Qwen3.5 Model

[ajith-sirra-amd]

Summary

Add comprehensive AMD MI300X/MI325X/MI355X GPU support for Qwen3.5 deployment, including installation methods, Docker setup, and deployment configurations.

Changes

Installation Instructions

• Added pip install method for AMD ROCm with Python 3.12, ROCm 7.0, and glibc requirements
• Provides clear guidance on environment prerequisites for AMD GPU users

Docker Support

• Added AMD-specific Docker deployment instructions with proper device mapping and security settings
• Separated Docker sections by vendor (NVIDIA vs AMD) for better organization and clarity

GPU Verification

• Updated documentation to note testing on 8x MI300X/MI355X GPUs
• Validates compatibility across both NVIDIA and AMD hardware platforms

MI355X Deployment

• Added 2-GPU deployment configuration for MI355X nodes with distributed parallelism
• Enables efficient deployment on newer AMD GPU architecture

AMD-specific Notes

• Added note that MTP-1 speculative decoding for AMD GPUs is under development

[gemini-code-assist]

Summary of Changes

Hello, 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 significantly expands the support for AMD GPUs (MI300X, MI325X, MI355X) when deploying the Qwen3.5 model. It integrates comprehensive instructions for both pip and Docker-based installations, updates verification notes to include AMD hardware, and provides specific deployment configurations for MI355X nodes. The changes aim to make the Qwen3.5 model more accessible and performant on AMD's latest GPU architectures.

Highlights

• AMD ROCm Installation: Added detailed pip installation instructions for AMD ROCm, specifying Python 3.12, ROCm 7.0, and glibc >= 2.35 requirements for MI300X, MI325X, and MI355X GPUs.
• AMD Docker Support: Introduced dedicated Docker deployment instructions for AMD MI300X/MI325X/MI355X GPUs, including necessary device mappings and security configurations, and separated Docker sections by vendor.
• Expanded GPU Verification: Updated the documentation to reflect that configurations have been verified on 8x MI300X/MI355X GPUs, in addition to existing H200 GPU verification.
• MI355X Deployment Configuration: Provided a specific 2-GPU deployment configuration for MI355X nodes, enabling efficient distributed parallelism for the Qwen3.5 model.
• AMD-specific Notes: Added a note indicating that MTP-1 speculative decoding for AMD GPUs is currently under development.

🧠 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

• Qwen/Qwen3.5.md
  • Added pip installation instructions for AMD ROCm, including version requirements and supported GPUs.
  • Refactored Docker sections to differentiate between NVIDIA and AMD-specific instructions.
  • Included detailed Docker commands for deploying Qwen3.5 on AMD MI300X/MI325X/MI355X GPUs.
  • Updated the GPU verification statement to confirm testing on AMD MI300X/MI355X GPUs.
  • Added a note regarding the development status of MTP-1 speculative decoding for AMD GPUs.
  • Introduced a new section with deployment configurations for MI355X nodes, specifically for 2-GPU setups.

Activity

• No human activity has occurred on this pull request yet.

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. ↩

[gemini-code-assist]

Code Review

This pull request adds valuable documentation for supporting Qwen3.5 on AMD MI300 series GPUs, including new installation and deployment instructions. The changes are well-structured, separating instructions for different hardware vendors. I have identified a couple of areas for improvement. The Docker command for AMD is incomplete, as it drops the user into a shell without further guidance; I've suggested a more user-friendly, complete command. Additionally, I found a typo in a command-line flag (-dp instead of -tp) in the MI355X deployment example, which would prevent the command from running correctly, and I have provided a fix.

_{Note: Security Review has been skipped due to the limited scope of the PR.}