3. Part I Framework Design

In Part 1, we present a top-down approach to designing a machine learning framework. We begin by introducing the design of programming models for machine learning frameworks, followed by a discussion on representing a machine learning program as a computational graph. The machine learning program undergoes compilation by an AI compiler, which employs a range of frontend and backend techniques. Additionally, we will delve into the system components within a machine learning framework that facilitate data processing, model deployment, and distributed training.