llama.cpp AMD Vulkan Docker/Podman compose

amd-vulkan-compose.yml

services:
  llama-cpp:
    image: ghcr.io/ggml-org/llama.cpp:server-vulkan
    command: "-m /models/Qwen3-8B-Q8_0.gguf --host 0.0.0.0 --port 8000 --ctx-size 16000 --context-shift"
    devices:
      - "/dev/kfd:/dev/kfd"
      - "/dev/dri:/dev/dri"
    ports:
      - "8000:8000"
    restart: "unless-stopped"
    security_opt: 
      - label=type:container_runtime_t    
    volumes:
      - /path/to/models:/models