Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Zhao, Yang, Xu, Yanwu, Xiao, Zhisheng, Jia, Haolin, Hou, Tingbo
Format: Preprint
Veröffentlicht: 2023
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2311.16567
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866916283280261120
author Zhao, Yang
Xu, Yanwu
Xiao, Zhisheng
Jia, Haolin
Hou, Tingbo
author_facet Zhao, Yang
Xu, Yanwu
Xiao, Zhisheng
Jia, Haolin
Hou, Tingbo
contents The deployment of large-scale text-to-image diffusion models on mobile devices is impeded by their substantial model size and slow inference speed. In this paper, we propose \textbf{MobileDiffusion}, a highly efficient text-to-image diffusion model obtained through extensive optimizations in both architecture and sampling techniques. We conduct a comprehensive examination of model architecture design to reduce redundancy, enhance computational efficiency, and minimize model's parameter count, while preserving image generation quality. Additionally, we employ distillation and diffusion-GAN finetuning techniques on MobileDiffusion to achieve 8-step and 1-step inference respectively. Empirical studies, conducted both quantitatively and qualitatively, demonstrate the effectiveness of our proposed techniques. MobileDiffusion achieves a remarkable \textbf{sub-second} inference speed for generating a $512\times512$ image on mobile devices, establishing a new state of the art.
format Preprint
id arxiv_https___arxiv_org_abs_2311_16567
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle MobileDiffusion: Instant Text-to-Image Generation on Mobile Devices
Zhao, Yang
Xu, Yanwu
Xiao, Zhisheng
Jia, Haolin
Hou, Tingbo
Computer Vision and Pattern Recognition
The deployment of large-scale text-to-image diffusion models on mobile devices is impeded by their substantial model size and slow inference speed. In this paper, we propose \textbf{MobileDiffusion}, a highly efficient text-to-image diffusion model obtained through extensive optimizations in both architecture and sampling techniques. We conduct a comprehensive examination of model architecture design to reduce redundancy, enhance computational efficiency, and minimize model's parameter count, while preserving image generation quality. Additionally, we employ distillation and diffusion-GAN finetuning techniques on MobileDiffusion to achieve 8-step and 1-step inference respectively. Empirical studies, conducted both quantitatively and qualitatively, demonstrate the effectiveness of our proposed techniques. MobileDiffusion achieves a remarkable \textbf{sub-second} inference speed for generating a $512\times512$ image on mobile devices, establishing a new state of the art.
title MobileDiffusion: Instant Text-to-Image Generation on Mobile Devices
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2311.16567