Setting up A/B testing on multiple models with production variants

When dealing with production deployments, note that multiple models may be deployed and tested at the same time. This helps data scientists and machine learning engineers compare the performance of models when dealing with data that these models have not seen before. One of the standard ways to manage and test multiple models in production is through the use of A/B testing in inference endpoints. What's A/B testing? It is an experiment that involves randomly selecting a model from a list of deployed models to perform predictions. It helps identify the better (or best) performing model in production before completely replacing a deployed model.

In this recipe, we will deploy two pre-trained XGBoost models within a single endpoint using the multi-model endpoint support of SageMaker. We will configure and set up this endpoint to allow A/B testing of the pre-trained models that have been deployed in this endpoint...