The Workshop

This material is designed to help you get started with Diffusion models.

Although the material is designed to be followed in order, you can jump to any section you like. We usually run this event over a day, with a mix of lecture and group discussion. Here is a rough outline of the day:

Time	Content
9:00am - 9:30am	Registration and welcome refreshments
9:30am - 10:00am	Introduction Introduction to Accelerate and overview of the workshop - What do you know about them - How do they work? - What can you use them for? - What are the ethical issues? - What do you intend to use them for? - Brief history of generative models
10:00am - 10:30am	Inpsiration from Physics - Charge distributions - Probability distributions - First look at the diffusion algorithm
10:30am - 11:00am	Break
11:00am - 11:30pm	Autoencoders and VAEs
11:30am - 12:00pm	UNet
12:00pm - 12:30pm	CLIP
12:30pm - 1:30pm	Lunch
1:30pm - 2:00pm	The Diffusion Algorithm - Training - Sampling - Forward - Reverse - Classifier-free diffusion - (Un)conditional diffusion
2:00pm - 2:30pm	History of Generative Models - GANs
2:30pm - 3:00pm	Ethics An overview and discussion of some generative AI case studies
3:00pm - 3:30pm	Break
3:30pm - 4:00pm	Application - Some practical applications of Diffusion models, including in scientific research. - Hugging Face diffusers - Stable diffusion via the Hugging Face Hub
4:00pm - 4:30pm	No Code Options
4:30pm - 5:00pm	Closing