We present a new method for making diffusion models faster to sample. The method distills many-step diffusion models into few-step models by matching conditional expectations of the clean data given noisy data along the sampling trajectory. Our approach extends recently proposed one-step methods to the multi-step case, and provides a new perspective by interpreting these approaches in terms of moment matching. By using up to 8 sampling steps, we obtain distilled models that outperform not only their one-step versions but also their original many-step teacher models, obtaining new state-of-the-art results on the Imagenet dataset. We also show promising results on a large text-to-image model where we achieve fast generation of high resolution images directly in image space, without needing autoencoders or upsamplers.
@article{arxiv.2406.04103,
title = {Multistep Distillation of Diffusion Models via Moment Matching},
author = {Tim Salimans and Thomas Mensink and Jonathan Heek and Emiel Hoogeboom},
journal= {arXiv preprint arXiv:2406.04103},
year = {2024}
}