https://www.fast.ai/posts/part2-2022.html

fast.ai

  * About
  *  
  *  

From Deep Learning Foundations to Stable Diffusion

Practical Deep Learning for Coders part 2, 2022
Author

Jeremy Howard

Published

September 16, 2022

Three years ago we pioneered Deep Learning from the Foundations, an
in depth course that started right from the foundations--implementing
and GPU-optimising matrix multiplications and initialisations--and
covered from scratch implementations of all the key applications of
the fastai library.

This year, we're going "from the foundations" again, but this time
we're going further. Much further! This time, we're going all the way
through to implementing the astounding Stable Diffusion algorithm.
That's the killer app that made the internet freak out, and caused
the media to say "you may never believe what you see online again".

[diffusion]

Stable diffusion generated image

Stable diffusion, and diffusion methods in general, are a great
learning goal for many reasons. For one thing, of course, you can
create amazing stuff with these algorithms! To really take the
technique to the next level, and create things that no-one has seen
before, you need to really deeply understand what's happening under
the hood. With this understanding, you can craft your own loss
functions, initialization methods, multi-model mixups, and more, to
create totally new applications that have never been seen before.

Just as important: it's a great learning goal because nearly every
key technique in modern deep learning comes together in these
methods. Contrastive learning, transformer models, auto-encoders,
CLIP embeddings, latent variables, u-nets, resnets, and much more are
involved in creating a single image.

This is all cutting-edge stuff, so to ensure we bring the latest
techniques to you, we're teaming up with the folks that brought
stable diffusion to the world: stability.ai. stability.ai are, in
many ways, kindred spirits to fast.ai. They are, like us, a
self-funded research lab. And like us, their focus is smashing down
any gates that make cutting edge AI less accessible. So it makes
sense for us to team up on this audacious goal of teaching stable
diffusion from the foundations.

The course will be available for free online from early 2023. But if
you want to join the course right as it's made, along with hundreds
of the world's leading deep learning practitioners, then you can
register to join the virtual live course through our official
academic partner, the University of Queensland (UQ). UQ will have
registrations open in the next few days, so keep an eye on the link
above.

During the live course, we'll be learning to read and implement the
latest papers, with lots of opportunity to practice and get feedback.
Many past participants in fast.ai's live courses have described it as
a "life changing" experience... and it's our sincere hope that this
course will be our best ever.

To get the most out of this course, you should be a reasonably
confident deep learning practitioner. If you've finished fast.ai's
Practical Deep Learning course then you'll be ready! If you haven't
done that course, but are comfortable with building an SGD training
loop from scratch in Python, being competitive in Kaggle
competitions, using modern NLP and computer vision algorithms for
practical problems, and working with PyTorch and fastai, then you
will be ready to start the course. (If you're not sure, then I
strongly recommend getting starting with Practical Deep Learning
now--if you push, you'll be done before the new course starts!)

If you're an alumnus of Deep Learning for Coders, you'll know that
course sets you up to be an effective deep learning practitioner.
This new course will take you to the next level, creating novel
applications that bring multiple techniques together, and
understanding and implementing research papers. Alumni of previous
versions of fast.ai's "part 2" courses have even gone on to publish
deep learning papers in top conferences and journals, and have joined
highly regarded research labs and startups.