Multistep Distillation of Diffusion Models via Moment Matching
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866913380212670464 |
|---|---|
| author | Salimans, Tim Mensink, Thomas Heek, Jonathan Hoogeboom, Emiel |
| author_facet | Salimans, Tim Mensink, Thomas Heek, Jonathan Hoogeboom, Emiel |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_04103 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Multistep Distillation of Diffusion Models via Moment Matching Salimans, Tim Mensink, Thomas Heek, Jonathan Hoogeboom, Emiel Machine Learning Artificial Intelligence Computer Vision and Pattern Recognition Neural and Evolutionary Computing 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. |
| title | Multistep Distillation of Diffusion Models via Moment Matching |
| topic | Machine Learning Artificial Intelligence Computer Vision and Pattern Recognition Neural and Evolutionary Computing |
| url | https://arxiv.org/abs/2406.04103 |