services:
  minimax-m3:
    build: .
    image: minimax-m3-sglang:dev-cu13-minimax-m3-patched
    container_name: minimax-m3-sglang
    network_mode: host
    ipc: host
    gpus: all
    ulimits:
      memlock: -1
      stack: 67108864
    environment:
      PYTORCH_CUDA_ALLOC_CONF: expandable_segments:True
    volumes:
      - /mnt/llm_models:/mnt/llm_models:ro
    command:
      - --model-path
      - /mnt/llm_models/MiniMax-M3-NVFP4
      - --tokenizer-path
      - /mnt/llm_models/MiniMax-M3-NVFP4
      - --trust-remote-code
      - --host
      - 0.0.0.0
      - --port
      - "8000"
      - --served-model-name
      - minimax-m3
      - --tp-size
      - "4"
      - --context-length
      - "450000"
      - --max-total-tokens
      - "450000"
      - --kv-cache-dtype
      - bfloat16
      - --quantization
      - modelopt_fp4
      - --reasoning-parser
      - minimax-m3
      - --tool-call-parser
      - minimax-m3
      - --moe-runner-backend
      - flashinfer_cutlass
      - --cuda-graph-backend-decode
      - full
      - --cuda-graph-backend-prefill
      - disabled
      - --max-running-requests
      - "4"
      - --chunked-prefill-size
      - "4096"
      - --max-prefill-tokens
      - "16384"
      - --mem-fraction-static
      - "0.95"