services:
  vllm:
    #image: atl.vultrcr.com/vllm/vllm-dsv4-nvfp4:latest
    build:
      context: .
    pull_policy: always
    ports:
      - "8000:8000"
    environment:
      - HF_TOKEN=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
      - OMP_NUM_THREADS=128
      - VLLM_USE_FLASHINFER_MOE_FP4=1
    command:
      - /model
      - --trust-remote-code
      - --kv-cache-dtype=fp8
      #- --block-size=256
      - --enable-expert-parallel
      - --tensor-parallel-size=8
      #- --compilation-config={"cudagraph_mode":"FULL_AND_PIECEWISE","custom_ops":["all"]}
      #- --attention_config.use_fp4_indexer_cache=True
      - --tokenizer-mode=deepseek_v4
      #- --speculative_config={"method":"mtp","num_speculative_tokens":2}
      - --host=0.0.0.0
      - --port=8000
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    volumes:
      - /root/nvidia-meeting/DeepSeek-V4-Pro-NVFP4:/model:ro