大纲

In this notebook we're going to illustrate one way to add conditioning information to a diffusion model. Specifically, we'll train a class-conditioned diffusion model on MNIST following on from the

'from-scratch' example in Unit 1

, where we can specify which digit we'd like the model to generate at inference time.

【10分类的MNIST】

步骤

Setup and Data Prep

Creating a Class-Conditioned UNet

Training and Sampling