AsyncDiff: Parallelizing Diffusion Models by Asynchronous Denoising
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866914957491175424 |
|---|---|
| author | Chen, Zigeng Ma, Xinyin Fang, Gongfan Tan, Zhenxiong Wang, Xinchao |
| author_facet | Chen, Zigeng Ma, Xinyin Fang, Gongfan Tan, Zhenxiong Wang, Xinchao |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_06911 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | AsyncDiff: Parallelizing Diffusion Models by Asynchronous Denoising Chen, Zigeng Ma, Xinyin Fang, Gongfan Tan, Zhenxiong Wang, Xinchao Computer Vision and Pattern Recognition Artificial Intelligence 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. |
| title | AsyncDiff: Parallelizing Diffusion Models by Asynchronous Denoising |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2406.06911 |