Training a model is an iterative process: you try different algorithms, settings, and datasets, and the best result is easy to lose if nobody wrote down how it was produced. Experiment tracking is the practice of logging the parameters, metrics, code and data versions, and output files of every run, so you can compare results and reproduce the winner later.
You'll start with the core ideas of tracking and the main features of MLflow, then see how its model registry stores versions of trained models and manages them as they move toward production. The module also covers Weights & Biases, a popular hosted alternative for tracking and visualizing experiments, so you can recognize and use either tool.
Once a model is tracked and registered, it is ready to be packaged and run anywhere. That's the focus of the next module, Containers & Orchestration.
Explains the basic terminology of experiment tracking and introduces MLflow and its main features.
A beginner tutorial on using MLflow to manage experiments, track metrics, and organize models in the registry.
A lesson on Weights & Biases (W&B), a tool for tracking and visualizing machine learning experiments.
A beginner coding tutorial that goes through the key features of W&B for tracking your experiments.