Prelude

W

This tutorial shows how to write machine learning code in a productive and enjoyable way, through the journey of building three architectures: a linear regression model, a CNN, and a transformer.

We do this using types and abstraction, which offload low-level mechanics to PyPie. Types give clarity to our programs and catch bugs before they run. Abstraction hides tedious details in recurring patterns.

It requires little background to learn these principles: high school math is enough; first-time programmers are welcome. We only ask for curiosity and hands-on practice. Each chapter is a dialog between a teacher and a student, filled with questions, answers, and code. To get the most out of it, type and run the programs in PyPie.

Our teacher is waiting. Shall we begin?