A Closer Look at Time Steps is Worthy of Triple Speed-Up for Diffusion Model Training

TL;DR

This paper presents SpeeD, a simple, architecture-agnostic method that accelerates diffusion model training by focusing on the importance of different time steps, achieving threefold speed-up with minimal overhead.

Contribution

The paper introduces a novel asymmetric sampling and weighting strategy based on time step analysis, significantly improving training efficiency for diffusion models.

Findings

Achieves 3x acceleration across various architectures and datasets

Identifies imbalance in time step importance during diffusion training

Reduces training costs with minimal additional overhead

Abstract

Training diffusion models is always a computation-intensive task. In this paper, we introduce a novel speed-up method for diffusion model training, called, which is based on a closer look at time steps. Our key findings are: i) Time steps can be empirically divided into acceleration, deceleration, and convergence areas based on the process increment. ii) These time steps are imbalanced, with many concentrated in the convergence area. iii) The concentrated steps provide limited benefits for diffusion training. To address this, we design an asymmetric sampling strategy that reduces the frequency of steps from the convergence area while increasing the sampling probability for steps from other areas. Additionally, we propose a weighting strategy to emphasize the importance of time steps with rapid-change process increments. As a plug-and-play and architecture-agnostic approach, SpeeD consistently achieves 3-times acceleration across various diffusion architectures, datasets, and tasks. Notably, due to its simple design, our approach significantly reduces the cost of diffusion model training with minimal overhead. Our research enables more researchers to train diffusion models at a lower cost.