This is a really nice blogpost by
@RuiqiGao and team that I enjoyed being a part of. My favorite key learnings are:
- DDIM sampler == flow matching sampling
- (Not) straight?
- SD3 weighting (Esser, Rombach, et al) is very similar to the EDM weighting (Karras, et al).
👇

Thanks for the great summary of our post! 😊

Yup yup that's our goal for this post, making it more accessible :).

haha thanks! Credit to Flux.

With amazing colleagues: @emielhoogeboom.bsky.social, Jonathan Heek, @vdebortoli.bsky.social, @sirbayes.bsky.social, @timsalimans.bsky.social

We hope this helps practitioners understand the true degrees of freedom when tuning the algorithm. For example, a flow matching sampler doesn’t have to be deterministic.

Blog post link: diffusionflow.github.io/

Despite seeming similar, there is some confusion in the community about the exact connection between the two frameworks. We aim to clear up the confusion by showing how to convert one framework to another, for both training and sampling.

A common question nowadays: Which is better, diffusion or flow matching? 🤔

Our answer: They’re two sides of the same coin. We wrote a blog post to show how diffusion models and Gaussian flow matching are equivalent. That’s great: It means you can use them interchangeably.