This repo contains various notebooks and scripts where I experiment with MLflow regarding experiment tracking, model evaluation, and registry, while using a docker container emulating MLflow server.

The following is a short description of what can be found in each folder.

A quickstart tutorial, using MLflow 2.8.0, where learn how to start a MLflow Tracking Server and the MLflow UI Server locally. We create a MLflow experiment with a unique name and identifying tags. We connect to the Tracking Server with the MLflow Client and search for experiments, while using relevant identifying tag values. Then we train a model using the wine dataset and log the trained model, metrics, parameters, and artifacts.

Another quickstart tutorial, based on an older version of the MLflow docs, similar to the previous folder, where we train a model using the wine dataset and log the trained model, metrics, parameters, and artifacts.

This folder contains an MLflow project that trains a RandomForestRegressor scikit model on the UC Irvine Wine Quality Dataset. The project is an example of how we can share ML code and improve reproducibility. It uses a Docker image to capture the dependencies needed to run training code. Running a project in a Docker environment allows for capturing non-Python dependencies, as opposed to a conda environment.

Launching MLflow Tracking Server:

• mlflow server --host 127.0.0.1 --port 8080

With the Tracking Server operational, it’s time to start the MLflow UI. Launch it from a new command prompt. As with the Tracking Server, ensure this window remains open:

• mlflow ui --host 127.0.0.1 --port 8090