DRiffusion: Draft-and-Refine Process Parallelizes Diffusion Models with Ease

TL;DR

DRiffusion introduces a parallel sampling framework for diffusion models that significantly accelerates inference with minimal quality loss, enabling faster high-fidelity content generation.

Contribution

The paper presents DRiffusion, a novel draft-and-refine parallel sampling method that accelerates diffusion inference by a factor of up to 3.7x while maintaining quality.

Findings

Achieves 1.4x to 3.7x speedup across multiple models.

Maintains comparable FID and CLIP scores to original models.

Shows only minor drops in PickScore and HPSv2.1 metrics.

Abstract

Diffusion models have achieved remarkable success in generating high-fidelity content but suffer from slow, iterative sampling, resulting in high latency that limits their use in interactive applications. We introduce DRiffusion, a parallel sampling framework that parallelizes diffusion inference through a draft-and-refine process. DRiffusion employs skip transitions to generate multiple draft states for future timesteps and computes their corresponding noises in parallel, which are then used in the standard denoising process to produce refined results. Theoretically, our method achieves an acceleration rate of $\tfrac{1}{n}$ or $\tfrac{2}{n+1}$, depending on whether the conservative or aggressive mode is used, where $n$ denotes the number of devices. Empirically, DRiffusion attains 1.4$\times$-3.7$\times$ speedup across multiple diffusion models while incur minimal degradation in generation quality: on MS-COCO dataset, both FID and CLIP remain largely on par with those of the original model, while PickScore and HPSv2.1 show only minor average drops of 0.17 and 0.43, respectively. These results verify that DRiffusion delivers substantial acceleration and preserves perceptual quality.