docker-compose.yaml

services:
  # Available environment variables can be found at
  # https://github.com/kibae/onnxruntime-server#run-the-server

  onnxruntime_server_simple:
    # After the docker container is up, you can use the REST API (http://localhost:8080).
    # API documentation will be available at http://localhost:8080/api-docs.
    image: kibaes/onnxruntime-server:1.20.1-linux-cuda12
    ports:
      - "8080:80" # for http backend
    volumes:
      # for model files
      # https://github.com/kibae/onnxruntime-server#run-the-server
      - /your_model_dir:/app/models

      # for log files
      - /your_log_dir:/app/logs
    environment:
      # for swagger(optional)
      - ONNX_SERVER_SWAGGER_URL_PATH=/api-docs
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [ gpu ]

  onnxruntime_server_advanced:
    # After the docker container is up, you can use the REST API (http://localhost, https://localhost).
    # API documentation will be available at http://localhost/api-docs.
    image: kibaes/onnxruntime-server:1.20.1-linux-cuda12
    ports:
      - "80:80" # for http backend
      - "443:443" # for https backend
      - "8001:8001" # for tcp backend. binary protocol
    volumes:
      # for model files
      # https://github.com/kibae/onnxruntime-server#run-the-server
      - /your_model_dir:/app/models

      # for log files
      - /your_log_dir:/app/logs

      # for cert files
      - /your_cert_dir:/app/certs
    environment:
      # for https backend
      - ONNX_SERVER_HTTPS_PORT=443

      # https backend needs cert, key files
      - ONNX_SERVER_HTTPS_CERT=/app/certs/cert.pem
      - ONNX_SERVER_HTTPS_KEY=/app/certs/key.pem

      # for onnx session preparation
      - ONNX_SERVER_PREPARE_MODEL="model1:v1(cuda=true) model1:v2(cuda=0) model2:v2"

      # for swagger(optional)
      - ONNX_SERVER_SWAGGER_URL_PATH=/api-docs
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [ gpu ]