Using MLFlow as an experiment tracking system
In this section, you will see how the MLflow library allows you to record your experiments with the MLflow server. The custom notebook image, which you saw in the first part of this chapter, already has MLflow libraries packaged into the container. Please refer to the chapter6/requirements.txt file for the exact version of the MLflow library.
Before we start this activity, it is important to understand two main concepts: experiment and run.
An experiment is a logical name under which MLflow records and groups the metadata, for example, an experiment could be the name of your project. Let's say you are working on building a model for predicting credit card fraud for your retail customer. This could become the experiment name.
A run is a single execution of an experiment that is tracked in MLflow. A run belongs to an experiment. Each run may have a slightly different configuration, different hyperparameters, and sometimes, different...
