Title: Diffusion Meets Flow Matching
Description: Flow matching and diffusion models are two popular frameworks in generative modeling. Despite seeming similar, there is some confusion in the community about their exact connection. In this post, we aim to clear up this confusion and show that diffusion models and Gaussian flow matching are the same, although different model specifications can lead to different network outputs and sampling schedules. This is great news, it means you can use the two frameworks interchangeably.
Keywords:
Domain: diffusionflow.github.io
| None | IE=edge |
| author |
Links:
| Diffusion Meets Flow Matching | https://diffusionflow.github.io/ |
| Overview | https://diffusionflow.github.io/#overview |
| Sampling | https://diffusionflow.github.io/#sampling |
| Training | https://diffusionflow.github.io/#training |
| Diving deeper into samplers | https://diffusionflow.github.io/#diving-deeper-into-samplers |
| SDE and ODE perspective | https://diffusionflow.github.io/#sde-and-ode-perspective |
| Closing takeaways | https://diffusionflow.github.io/#closing-takeaways |
| Google Colab | https://colab.research.google.com/drive/13lAveB3qwjkgyILWW-9qiOOSHG0U5_O6?usp=sharing |
| This blog post | https://sander.ai/2024/06/14/noise-schedules.html#adaptive |
Viewport: width=device-width, initial-scale=1, shrink-to-fit=no