Diffusion Model Implementation

This notebook implements a diffusion model from scratch, inspired by this guide. The goal of this project is to demonstrate the process of building and training a basic diffusion model on the MNIST dataset without any pre-trained models or specialized purposes.

Contents

• Introduction: A diffusion model is a type of generative model that learns to reverse a noising process applied to data, here demonstrated on MNIST images.
• Architecture: This implementation follows the architecture and methodology described in the linked guide, providing insight into how diffusion models work on a fundamental level.

Getting Started

Prerequisites

Ensure you have the following libraries installed:

• tensorflow
• numpy
• tqdm

You can install these using:

pip install tensorflow numpy tqdm