Update Mistral-Large-3.md for AMD GPU

[jiacao-amd]

…model

[gemini-code-assist]

Summary of Changes

Hello @jiacao-amd, 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 enhances the documentation for the Mistral-Large-3 model by adding a dedicated section for AMD GPU support. It provides a complete, step-by-step guide for users to deploy and benchmark the model on AMD MI300X and MI355 GPUs, covering everything from Docker setup to vLLM server configuration and performance testing. This addition aims to make the model more accessible and performant for users with AMD hardware.

Highlights

• AMD GPU Support Documentation: A new section has been added to Mistral-Large-3.md providing comprehensive instructions for installing and running Mistral-Large-3-675B-Instruct-2512 models on AMD MI300X and MI355 GPUs.
• Docker Environment Setup: Detailed steps are included for preparing the Docker environment, specifically pulling the vllm/vllm-openai-rocm:v0.14.1 image and launching a container with necessary ROCm-specific configurations.
• vLLM Server Configuration: Instructions cover how to log in to Hugging Face and start the vLLM online serving with specific environment variables and command-line arguments tailored for AMD GPU performance.
• Benchmarking Guide: A guide for running benchmark scripts within the configured Docker environment is provided, allowing users to test the model's performance.

🧠 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.

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 documentation for running the Mistral-Large-3 model on AMD GPUs. The instructions are clear, but I've identified a significant security concern in the provided docker run command, which mounts the entire host filesystem into the container. I've also suggested a couple of minor text changes to improve clarity and consistency regarding the model name and Docker image versioning. Addressing the security issue is highly recommended.

[gemini-code-assist]

Mounting the host's root directory (/) into the container with -v /:/work is a significant security risk. It grants the container unrestricted access to the entire host filesystem, which could lead to a full system compromise if the container is breached. It is strongly recommended to only mount specific directories that are necessary. For example, you could mount the current directory with -v $(pwd):/work or a specific data directory with -v /path/to/your/data:/work.

Suggested change

	-v /:/work \
	-v $(pwd):/work \

[gemini-code-assist]

For clarity and consistency with the commands that follow, it's better to use the full model name including the Hugging Face organization, mistralai/Mistral-Large-3-675B-Instruct-2512.

Suggested change

	Please follow the steps here to install and run Mistral-Large-3-675B-Instruct-2512 models on AMD MI300X and MI355 GPU.
	Please follow the steps here to install and run mistralai/Mistral-Large-3-675B-Instruct-2512 models on AMD MI300X and MI355 GPU.

[gemini-code-assist]

The instruction says to pull the 'latest' vLLM docker image, but the command that follows specifies a fixed version (v0.14.1). This is contradictory and could be confusing. It would be clearer to state that a specific version is required for these instructions.

Suggested change

	Pull the latest vllm docker:
	Pull the required vLLM docker image for ROCm:

[tjtanaa]

@jiacao-amd Please revise all of your command to up-to-date with the latest vLLM version. There is environment variables that have been deprecated many versions ago.

For this documentation, we don't need to attach the benchmark script.

Please merge the ROCm installation step and server command with the main context. Open subheaders to split between two hardware similar to the suggestion mentioned here https://github.com/vllm-project/recipes/pull/219/changes#r2804656457

[jiacao-amd]

merge the ROCm installation step and server command with the main context. Open subheaders to split between two hardware

Done!