Skip to yearly menu bar Skip to main content


Poster

Plug-and-Play Diffusion Distillation

Yi-Ting Hsiao · Siavash Khodadadeh · Kevin Duarte · Wei-An Lin · Hui Qu · Mingi Kwon · Ratheesh Kalarot

Arch 4A-E Poster #396
[ ] [ Project Page ] [ Paper PDF ]
[ Slides [ Poster
Thu 20 Jun 10:30 a.m. PDT — noon PDT

Abstract:

Diffusion models have shown tremendous results in image generation. However, due to the iterative nature of the diffusion process and its reliance on classifier-free guidance, inference times are slow. In this paper, we propose a new distillation approach for guided diffusion models in which an external lightweight guide model is trained while the original text-to-image model remains frozen.We show that our method reduces the inference computation of classifier-free guided latent-space diffusion models by almost half, and only requires 1% trainable parameters of the base model. Furthermore, once trained, our guide model can be applied to various fine-tuned, domain-specific versions of the base diffusion model without the need for additional training: this "plug-and-play" functionality drastically improves inference computation while maintaining the visual fidelity of generated images. Empirically, we show that our approach is able to produce visually appealing results and achieve a comparable FID score to the teacher with as few as 8 to 16 steps.

Chat is not available.