MoDM: Efficient Serving for Image Generation via Mixture-of-Diffusion Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xia, Yuchen, Sharma, Divyam, Yuan, Yichao, Kundu, Souvik, Talati, Nishil
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912515461480448
author Xia, Yuchen
Sharma, Divyam
Yuan, Yichao
Kundu, Souvik
Talati, Nishil
author_facet Xia, Yuchen
Sharma, Divyam
Yuan, Yichao
Kundu, Souvik
Talati, Nishil
contents Diffusion-based text-to-image generation models trade latency for quality: small models are fast but generate lower-quality images, while large models produce better images but are slow. We present MoDM, a novel caching-based serving system for diffusion models that dynamically balances latency and quality through a mixture of diffusion models. Unlike prior approaches that rely on model-specific internal features, MoDM caches final images, allowing seamless retrieval and reuse across multiple diffusion model families. This design enables adaptive serving by dynamically balancing latency and image quality: using smaller models for cache-hit requests to reduce latency while reserving larger models for cache-miss requests to maintain quality. Small model image quality is preserved using retrieved cached images. We design a global monitor that optimally allocates GPU resources and balances inference workload, ensuring high throughput while meeting service-level objectives under varying request rates. Our evaluations show that MoDM significantly reduces average serving time by 2.5x while retaining image quality, making it a practical solution for scalable and resource-efficient model deployment.
format Preprint
id arxiv_https___arxiv_org_abs_2503_11972
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MoDM: Efficient Serving for Image Generation via Mixture-of-Diffusion Models
Xia, Yuchen
Sharma, Divyam
Yuan, Yichao
Kundu, Souvik
Talati, Nishil
Distributed, Parallel, and Cluster Computing
Diffusion-based text-to-image generation models trade latency for quality: small models are fast but generate lower-quality images, while large models produce better images but are slow. We present MoDM, a novel caching-based serving system for diffusion models that dynamically balances latency and quality through a mixture of diffusion models. Unlike prior approaches that rely on model-specific internal features, MoDM caches final images, allowing seamless retrieval and reuse across multiple diffusion model families. This design enables adaptive serving by dynamically balancing latency and image quality: using smaller models for cache-hit requests to reduce latency while reserving larger models for cache-miss requests to maintain quality. Small model image quality is preserved using retrieved cached images. We design a global monitor that optimally allocates GPU resources and balances inference workload, ensuring high throughput while meeting service-level objectives under varying request rates. Our evaluations show that MoDM significantly reduces average serving time by 2.5x while retaining image quality, making it a practical solution for scalable and resource-efficient model deployment.
title MoDM: Efficient Serving for Image Generation via Mixture-of-Diffusion Models
topic Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2503.11972