Documentation references non-existent Docker image tag: rocm/vllm:instinct_main

Issue Description

The documentation for running vLLM on MI300X GPUs references a Docker image tag that does not exist on Docker Hub.

Documentation URL: https://github.com/ROCm/amdgpu-docs/blob/develop/docs/gpu-partitioning/mi300x/run-vllm.rst

Issue: The documentation instructs users to pull the image rocm/vllm:instinct_main:

sudo docker pull rocm/vllm:instinct_main

Problem: This tag does not exist in the rocm/vllm Docker Hub repository.

Available Tags

The repository currently has 20 tags total, including:

• latest
• v0.14.0_amd_dev
• rocm7.0.0_vllm_0.11.2_20251210
• rocm6.4.1_vllm_0.10.1_20250909
• rocm6.3.1_instinct_vllm0.8.3_20250415
• etc.

Suggested Fix

The documentation should be updated to reference a valid tag. Options include:

1. Use latest for the most recent version
2. Use a specific versioned tag like rocm7.0.0_vllm_0.11.2_20251210
3. Use one of the instinct-specific tags like rocm6.3.1_instinct_vllm0.8.3_20250415

Impact

Users following the documentation will encounter an error when attempting to pull the image, preventing them from running the vLLM workload as documented.

Open docs/gpu-partitioning/mi300x/run-vllm.rst and locate the docker pull command using rocm/vllm:instinct_main. Check the current rocm/vllm tags on Docker Hub, select a valid tag consistent with the documented MI300X setup, and update the command. Done means the documented image tag exists and can be pulled successfully.