Chapter 6: Model to Production and Beyond
In the last chapter, we discussed model training and prediction for online and batch models with Feast. For the exercise, we used the Feast infrastructure that was deployed to the AWS cloud during the exercises in Chapter 4, Adding Feature Stores to ML Models. During these exercises, we looked at how Feast decouples feature engineering from model training and model prediction. We also learned how to use offline and online stores during batch and online prediction.
In this chapter, we will reuse the feature engineering pipeline and the model built in Chapter 4, Adding Feature Stores to ML Models, and Chapter 5, Model Training and Inference, to productionize the machine learning (ML) pipeline. The goal of this chapter is to reuse everything that we have built in the previous chapters, such as Feast infrastructure on AWS, feature engineering, model training, and model-scoring notebooks, to productionize the ML model. As we go through the exercises...