Run vLLM on gfx1151

launch-vllm.sh

#!/bin/env bash
# Launch VLLM on a gfx1151 AMD Strix Halo 395+ 
# As of August 27, 2025
podman run -it --rm --ipc=host --network=host --privileged \
       --cap-add=CAP_SYS_ADMIN  \
       --device=/dev/kfd --device=/dev/dri --device=/dev/mem \
       --group-add render \
       --cap-add=SYS_PTRACE \
       --security-opt seccomp=unconfined \         
       -e HSA_OVERRIDE_GFX_VERSION=11.0.0 \ 
       -v ~/.cache/huggingface:/root/.cache/huggingface  \
       rocm/vllm-dev:main bash -c "pip install --upgrade transformers && vllm serve Qwen/Qwen3-4B-Instruct-2507"