Formalizing the ML development process

Let us first understand what constitutes a model and how we define MLOps. It is important to emphasize that this chapter is not about the nitty gritty details of creating a model, but rather about the data aspects of creating a good model and the process of continuously refining it to keep it relevant and useful to the business.

What is a model?

A model is an artifact that has several inputs and outputs. Let's list them so we have a firm idea of what a model encompasses. The following diagram captures our definition of an ML asset and its refinement zones:

Figure 9.7 – What is a model?

The inputs to this process include the following elements:

• One or more datasets
• One or more libraries used to create the model
• The source code used to create the model with a given architecture
• The distinct values for the various hyperparameters used to train the model
• Additional metadata such as...