mlflow_quick_start

A quick start to launch MLflow service.

To run it locally, change the .env.example file to .env, and modify the values within accordingly.

The backend database used is postgres, ran as a local container. One also needs to set up the artifact store. Currently, we choose AWS s3, therefore, one needs to have the local ~/.aws/credentials file configurated properly.

Once the configuration is, run:

$ docker-compose up -d

and the MLflow UI will be ready at localhost:5000.

To test the tracking API, run:

python mlflow_tracking.py

and the experiment should show up in the MLflow UI.