MLFLOW V2 - How to run the wine Project?

This directory contains an MLflow project that trains a linear regression model on the UC Irvine Wine Quality Dataset. The project uses a Docker image to capture the dependencies needed to run training code. Running a project in a Docker environment (as opposed to Conda) allows for capturing non-Python dependencies, e.g. Java libraries.

Steps to reproduce the example:

1. Put a .env file:

POSTGRES_HOST=postgres-db
POSTGRES_PORT=6667
POSTGRES_USER=wine
POSTGRES_PASSWORD=wine
IP_SUBNET=189.230.0.0
MLFLOW_POSTGRES_DB=mlflowdb
MLFLOW_ARTIFACT_ROOT=./mlflow/resources/
MLFLOW_TRACKING_USER=mlflow
MLFLOW_TRACKING_PASSWORD=mlflow
MLFLOW_SERVER_HOST=mlflow-server
MLFLOW_UI_HOST=mlflow-ui
MLFLOW_UI_PORT=89
MLFLOW_SERVER_PORT=6001
JUPYTER_PORT=8877

2. Build the image:

docker-compose build

3. Setup the project containers

docker-compose up

Check the status of each container with docker ps command.

4. In order to execute a run for the specified experiment in train.py run:

docker exec -it base_v2 poetry run python train.py --alpha=2 --l1-ratio=0.2

You can try other values for alpha and l1-ratio parameters to interact with MLFlow service.

5. Go to http://localhost:89/ to see your new experiment and your runs.

If you want to run new experiments on JupyterLab, just go to http://localhost:8877/