docker-compose.yml

services:
  excoder:
    build:
      context: .
      args:
        BASE_IMAGE: nvidia/cuda:12.4.0-devel-ubuntu22.04
        USE_CUDA: "true"
    volumes:
      - .:/workspace
      - downloads:/workspace/_work/downloads
      - gguf_models:/home/etest/gguf_models
      - ~/.cache/huggingface:/home/etest/.cache/huggingface
    working_dir: /workspace
    tty: true
    stdin_open: true
    network_mode: "host"
    environment:
      - AZURE_END_POINT=${AZURE_END_POINT}
      - AZURE_API_KEY=${AZURE_API_KEY}
    deploy:
      resources:
        reservations:
          devices:
            - driver: cdi
              device_ids:
                - nvidia.com/gpu=all
              capabilities:
                - gpu
volumes:
  # Where the GGUF weights land ($GGUF_DIR in the image); the container runs as
  # the unprivileged `etest` user, so the volume is mounted under its home.
  # run_experiment_pipeline.sh --download-models fills it.
  gguf_models: {}
  downloads: {}