Managing the ML Lifecycle using MLflow

The ML lifecycle covers the following stages of a ML engineering project:

• data preparation
• training
• deployment

MLflow is an open-source framework that supports the ML lifecycle by tracking training metrics, storing trained models and model deployment.

Demo Objectives

• How to setup and configure MLflow.
• How to track metrics.
• How to track metrics during hyper-parameter optimisation.
• How to save a model to the MLflow registry.
• How to retrieve the latest model tagged for production.
• How to serve model predictions via a REST API.

Running the Demo

This demo is contained within a single Jupyter notebook - demos/mlflow/mlflow_basics.ipynb. Make sure you have the necessary Python package requirements installed into a Jupyter kernel for it to run successfully.