Docker Compose with Ollama

docker-compose.ollama.yaml

x-shared-logs: &shared-logs
  logging:
    driver: "json-file"
    options:
      max-size: "10k"

services:

  ollama:
    image: ollama/ollama:0.9.2
    restart: unless-stopped
    volumes:
      - ./ollama_data:/root
    environment:
      OLLAMA_ORIGINS: "*"
      OLLAMA_KEEP_ALIVE: 60m
      OLLAMA_FLASH_ATTENTION: 1
      OLLAMA_MAX_LOADED_MODELS: 1
      OLLAMA_MAX_QUEUE: 1
      OLLAMA_NUM_PARALLEL: 10
      OLLAMA_GPU_OVERHEAD: 0
      OLLAMA_SCHED_SPREAD: 1
      OLLAMA_KV_CACHE_TYPE: q4_0
    ports:
      - "11434:11434"
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
            device_ids: [ '1' ]
            capabilities: [ gpu ]
    <<: *shared-logs