vllm/docker-compose.vllm.yml

services:
  glm52-vllm:
    image: vllm/vllm-openai:glm52
    command:
      - zai-org/GLM-5.2-FP8
      - --host
      - 0.0.0.0
      - --port
      - "8000"
      - --served-model-name
      - glm-5.2
      - --kv-cache-dtype
      - fp8
      - --tensor-parallel-size
      - "${TENSOR_PARALLEL_SIZE:-8}"
      - --speculative-config.method
      - mtp
      - --speculative-config.num_speculative_tokens
      - "5"
      - --tool-call-parser
      - glm47
      - --reasoning-parser
      - glm45
      - --enable-auto-tool-choice
    ports:
      - "8001:8000"
    ipc: host
    volumes:
      - "${HOME}/.cache/huggingface:/root/.cache/huggingface"
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]