AsyncDiff: Parallelizing Diffusion Models by Asynchronous Denoising

TL;DR

AsyncDiff introduces an asynchronous, parallel computation scheme for diffusion models, significantly reducing inference latency by dividing the model across devices and exploiting high similarity between diffusion steps.

Contribution

It presents a universal, plug-and-play acceleration method that enables model parallelism in diffusion models by transforming sequential denoising into an asynchronous process.

Findings

Achieves 2.7x speedup on Stable Diffusion v2.1 with negligible quality loss.

Attains 4.0x speedup with only 0.38 CLIP Score reduction.

Successfully applied to video diffusion models with promising results.

Abstract

Diffusion models have garnered significant interest from the community for their great generative ability across various applications. However, their typical multi-step sequential-denoising nature gives rise to high cumulative latency, thereby precluding the possibilities of parallel computation. To address this, we introduce AsyncDiff, a universal and plug-and-play acceleration scheme that enables model parallelism across multiple devices. Our approach divides the cumbersome noise prediction model into multiple components, assigning each to a different device. To break the dependency chain between these components, it transforms the conventional sequential denoising into an asynchronous process by exploiting the high similarity between hidden states in consecutive diffusion steps. Consequently, each component is facilitated to compute in parallel on separate devices. The proposed strategy significantly reduces inference latency while minimally impacting the generative quality. Specifically, for the Stable Diffusion v2.1, AsyncDiff achieves a 2.7x speedup with negligible degradation and a 4.0x speedup with only a slight reduction of 0.38 in CLIP Score, on four NVIDIA A5000 GPUs. Our experiments also demonstrate that AsyncDiff can be readily applied to video diffusion models with encouraging performances. The code is available at https://github.com/czg1225/AsyncDiff.