services:
  vllm:
    image: ${IMAGE:-quay.io/ascend/vllm-ascend:v0.23.0-310p}
    entrypoint:
      - bash
      - -lc
    container_name: ${CONTAINER_NAME}
    restart: always
    networks:
      - 1panel-network
    devices:
      - /dev/davinci0:/dev/davinci0
      - /dev/davinci1:/dev/davinci1
      - /dev/davinci2:/dev/davinci2
      - /dev/davinci3:/dev/davinci3
      - /dev/davinci4:/dev/davinci4
      - /dev/davinci5:/dev/davinci5
      - /dev/davinci6:/dev/davinci6
      - /dev/davinci7:/dev/davinci7
      - /dev/davinci_manager:/dev/davinci_manager
      - /dev/devmm_svm:/dev/devmm_svm
      - /dev/hisi_hdc:/dev/hisi_hdc
    volumes:
      - ${MODEL_DIR}/:/models/${MODEL_NAME}
      - /root/.cache:/root/.cache
      - /usr/local/dcmi:/usr/local/dcmi
      - /usr/local/bin/npu-smi:/usr/local/bin/npu-smi
      - /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/
      - /usr/local/Ascend/driver/version.info:/usr/local/Ascend/driver/version.info
      - /etc/ascend_install.info:/etc/ascend_install.info
    environment:
      - ASCEND_RT_VISIBLE_DEVICES=${ASCEND_RT_VISIBLE_DEVICES}
      - VLLM_USE_MODELSCOPE=True
      - PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
      - HCCL_BUFFSIZE=512
      - OMP_PROC_BIND=false
      - OMP_NUM_THREADS=1
      - TASK_QUEUE_ENABLE=1
      - HCCL_OP_EXPANSION_MODE=AIV
    ports:
      - ${PANEL_APP_PORT_HTTP}:8000
    privileged: true
    shm_size: 10g
    command:
      - |
        rm -rf /usr/local/python3.12.13/lib/python3.12/site-packages/triton*
        exec vllm serve ${START_COMMAND}
networks:
  1panel-network:
    external: true