Docker environment with MLflow tracking server.
To run MLflow tracking server you can use prepared docker-compose.yml file or Dockerfile.
This option additionally creates container with PostgreSQL database. You should
set environment variables in variables section or use .env file. List of
required variable can be found in .env.default file.
If you don't need PostgreSQL database, it will be good option to choose Dockerfile option. Just build and run. Don't forget about required variables – pass it as arguments.
| Name | Required | Description |
|---|---|---|
| ARTIFACT_ROOT | yes | Value for MLflow server --default-artifact-root param |
| HOST | yes | Value for MLflow server --host param |
| BACKEND_STORE_URI | yes | Value for MLflow server --backend-store-uri param |
| GOOGLE_APPLICATION_CREDENTIALS | no | Credentials for Google Cloud |
| POSTGRES_USER | yes – for docker-compose run option | PostgreSQL username |
| POSTGRES_PASSWORD | yes – for docker-compose run option | PostgreSQL password |
| POSTGRES_DB | yes – for docker-compose run option | PostgreSQL database name |