Diffusion Models are Secretly Exchangeable: Parallelizing DDPMs via Autospeculation
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911095103422464 |
|---|---|
| author | Hu, Hengyuan Das, Aniket Sadigh, Dorsa Anari, Nima |
| author_facet | Hu, Hengyuan Das, Aniket Sadigh, Dorsa Anari, Nima |
| contents | Denoising Diffusion Probabilistic Models (DDPMs) have emerged as powerful tools for generative modeling. However, their sequential computation requirements lead to significant inference-time bottlenecks. In this work, we utilize the connection between DDPMs and Stochastic Localization to prove that, under an appropriate reparametrization, the increments of DDPM satisfy an exchangeability property. This general insight enables near-black-box adaptation of various performance optimization techniques from autoregressive models to the diffusion setting. To demonstrate this, we introduce \emph{Autospeculative Decoding} (ASD), an extension of the widely used speculative decoding algorithm to DDPMs that does not require any auxiliary draft models. Our theoretical analysis shows that ASD achieves a $\tilde{O} (K^{\frac{1}{3}})$ parallel runtime speedup over the $K$ step sequential DDPM. We also demonstrate that a practical implementation of autospeculative decoding accelerates DDPM inference significantly in various domains. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_03983 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Diffusion Models are Secretly Exchangeable: Parallelizing DDPMs via Autospeculation Hu, Hengyuan Das, Aniket Sadigh, Dorsa Anari, Nima Machine Learning Artificial Intelligence Denoising Diffusion Probabilistic Models (DDPMs) have emerged as powerful tools for generative modeling. However, their sequential computation requirements lead to significant inference-time bottlenecks. In this work, we utilize the connection between DDPMs and Stochastic Localization to prove that, under an appropriate reparametrization, the increments of DDPM satisfy an exchangeability property. This general insight enables near-black-box adaptation of various performance optimization techniques from autoregressive models to the diffusion setting. To demonstrate this, we introduce \emph{Autospeculative Decoding} (ASD), an extension of the widely used speculative decoding algorithm to DDPMs that does not require any auxiliary draft models. Our theoretical analysis shows that ASD achieves a $\tilde{O} (K^{\frac{1}{3}})$ parallel runtime speedup over the $K$ step sequential DDPM. We also demonstrate that a practical implementation of autospeculative decoding accelerates DDPM inference significantly in various domains. |
| title | Diffusion Models are Secretly Exchangeable: Parallelizing DDPMs via Autospeculation |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2505.03983 |