36
25

Accelerating Convergence of Score-Based Diffusion Models, Provably

Abstract

Score-based diffusion models, while achieving remarkable empirical performance, often suffer from low sampling speed, due to extensive function evaluations needed during the sampling phase. Despite a flurry of recent activities towards speeding up diffusion generative modeling in practice, theoretical underpinnings for acceleration techniques remain severely limited. In this paper, we design novel training-free algorithms to accelerate popular deterministic (i.e., DDIM) and stochastic (i.e., DDPM) samplers. Our accelerated deterministic sampler converges at a rate O(1/T2)O(1/{T}^2) with TT the number of steps, improving upon the O(1/T)O(1/T) rate for the DDIM sampler; and our accelerated stochastic sampler converges at a rate O(1/T)O(1/T), outperforming the rate O(1/T)O(1/\sqrt{T}) for the DDPM sampler. The design of our algorithms leverages insights from higher-order approximation, and shares similar intuitions as popular high-order ODE solvers like the DPM-Solver-2. Our theory accommodates 2\ell_2-accurate score estimates, and does not require log-concavity or smoothness on the target distribution.

View on arXiv
Comments on this paper