A very useful guide about how diffusion models work and implementation: https://keras.io/examples/generative/ddim/

I find the explanation in this article very intuitive.

So anyone up for a discussion on this paper at: https://www.alphaxiv.org/abs/2403.18103

Does something similar exist for LLM/GPT?

Edit to add: I'm mostly interested in this aspect:

"The target audience of this tutorial includes [those] who are interested in [...] applying these models to solve other problems."

Andrej Karpathy has a youtube playlist:

https://www.youtube.com/playlist?list=PLAqhIrjkxbuWI23v9cThs...

He is building new learning materials under his new company "Eureka Labs":

https://eurekalabs.ai

Sebastian Raschka's book "Build a Large Language Model (From Scratch) just released:

https://www.manning.com/books/build-a-large-language-model-f...

All of these resources are excellent.

Andrej Karpathy has very good video tutorials on how to write your own GPT: https://www.youtube.com/watch?v=kCc8FmEb1nY

This is excellent

3Blue1Brown has a pretty great video series walking through Transformers:

https://www.youtube.com/watch?v=wjZofJX0v4M

This paper is much clearer and succinct than the original papers from the field.

Trying to build a protein diffusion model from scratch right now.

The math explainer is quite helpful

The first diffusion model for text generation was published less than a year ago. Should add something about that

> see tutorial on diffusion > get excited > it's all math in latex > despair

Latex is great in this case, because the equations are all written out clearly.

If you want to understand diffusion, it's a little difficult to avoid math.

You may find the huggingface course more approachable

https://huggingface.co/learn/diffusion-course/en/unit0/1

Also this blog post https://yang-song.net/blog/2021/score/

Maybe check out the fast ai course. Jeremy Howard has a way of explaining stuff.

the math is learnable and with gpt nowadays, extremely so

You are right! LLMs accept the formulas and may explain them...

It will be more difficult to tell when they are wrong, though - when you cannot verify directly. But it will be a device for people to get acquainted with the math.

Diffusion is math. It’s difficult to avoid math if you want to understand math.

Be that as it may, this isn't what I'd call a "tutorial," in the sense that you'd better already have a strong command of the subject matter or you won't get much out of it.