MLOps tools
So far, we’ve talked about the goals and benefits of MLOps, but how can we actually achieve these goals and benefits in the real world? Many tools have been developed to facilitate various aspects of MLOps, from data versioning to model deployment and monitoring. In this section, we discuss the tools that make MLOps a reality.
Pipeline orchestration tools
When we want to standardize the steps in a process and run them automatically, we typically configure them as a set of sequential actions. This sequential set of actions is often referred to as a pipeline. However, simply configuring the order of the steps is not enough; we also need some kind of system to “orchestrate” (i.e., execute and manage) the pipeline, and we can use some popular workflow orchestration tools for this purpose, such as Kubeflow, Airflow, and TensorFlow Extended (TFX). We already covered Airflow in Chapter 6, but let’s take a look at Kubeflow and TFX in more detail...
