Saved in:
Bibliographic Details
Main Authors: Liang, Yuchen, Tian, Yuchuan, Yu, Lei, Tang, Huao, Hu, Jie, Fang, Xiangzhong, Chen, Hanting
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2409.17487
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914957690404864
author Liang, Yuchen
Tian, Yuchuan
Yu, Lei
Tang, Huao
Hu, Jie
Fang, Xiangzhong
Chen, Hanting
author_facet Liang, Yuchen
Tian, Yuchuan
Yu, Lei
Tang, Huao
Hu, Jie
Fang, Xiangzhong
Chen, Hanting
contents The curvature of ODE trajectories in diffusion models hinders their ability to generate high-quality images in a few number of function evaluations (NFE). In this paper, we propose a novel and effective approach to reduce trajectory curvature by utilizing adaptive conditions. By employing a extremely light-weight quantized encoder, our method incurs only an additional 1% of training parameters, eliminates the need for extra regularization terms, yet achieves significantly better sample quality. Our approach accelerates ODE sampling while preserving the downstream task image editing capabilities of SDE techniques. Extensive experiments verify that our method can generate high quality results under extremely limited sampling costs. With only 6 NFE, we achieve 5.14 FID on CIFAR-10, 6.91 FID on FFHQ 64x64 and 3.10 FID on AFHQv2.
format Preprint
id arxiv_https___arxiv_org_abs_2409_17487
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Learning Quantized Adaptive Conditions for Diffusion Models
Liang, Yuchen
Tian, Yuchuan
Yu, Lei
Tang, Huao
Hu, Jie
Fang, Xiangzhong
Chen, Hanting
Computer Vision and Pattern Recognition
The curvature of ODE trajectories in diffusion models hinders their ability to generate high-quality images in a few number of function evaluations (NFE). In this paper, we propose a novel and effective approach to reduce trajectory curvature by utilizing adaptive conditions. By employing a extremely light-weight quantized encoder, our method incurs only an additional 1% of training parameters, eliminates the need for extra regularization terms, yet achieves significantly better sample quality. Our approach accelerates ODE sampling while preserving the downstream task image editing capabilities of SDE techniques. Extensive experiments verify that our method can generate high quality results under extremely limited sampling costs. With only 6 NFE, we achieve 5.14 FID on CIFAR-10, 6.91 FID on FFHQ 64x64 and 3.10 FID on AFHQv2.
title Learning Quantized Adaptive Conditions for Diffusion Models
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2409.17487