Speed up the inference of diffusion models via shortcut MCMC sampling

TL;DR

This paper introduces a shortcut MCMC sampling method to accelerate diffusion model inference, maintaining high image quality while reducing the number of steps needed for image generation.

Contribution

The paper proposes a novel shortcut MCMC sampling algorithm that improves inference speed of diffusion models without sacrificing output quality.

Findings

Promising initial experimental results

Effective global fidelity constraint implementation

Potential for faster diffusion model inference

Abstract

Diffusion probabilistic models have generated high quality image synthesis recently. However, one pain point is the notorious inference to gradually obtain clear images with thousands of steps, which is time consuming compared to other generative models. In this paper, we present a shortcut MCMC sampling algorithm, which balances training and inference, while keeping the generated data's quality. In particular, we add the global fidelity constraint with shortcut MCMC sampling to combat the local fitting from diffusion models. We do some initial experiments and show very promising results. Our implementation is available at https://github.com//vividitytech/diffusion-mcmc.git.