Distribution Backtracking Builds A Faster Convergence Trajectory for Diffusion Distillation

TL;DR

This paper introduces Distribution Backtracking Distillation (DisBack), a novel method that leverages the entire convergence trajectory of teacher diffusion models to accelerate and improve the distillation process for faster image generation.

Contribution

DisBack extends score distillation by incorporating the convergence trajectory, enabling faster and more effective diffusion model distillation with comparable quality.

Findings

DisBack achieves faster convergence than existing methods.

DisBack attains a low FID score of 1.38 on ImageNet 64x64.

The method is easy to implement and generalizes well to other distillation approaches.

Abstract

Accelerating the sampling speed of diffusion models remains a significant challenge. Recent score distillation methods distill a heavy teacher model into a student generator to achieve one-step generation, which is optimized by calculating the difference between the two score functions on the samples generated by the student model. However, there is a score mismatch issue in the early stage of the distillation process, because existing methods mainly focus on using the endpoint of pre-trained diffusion models as teacher models, overlooking the importance of the convergence trajectory between the student generator and the teacher model. To address this issue, we extend the score distillation process by introducing the entire convergence trajectory of teacher models and propose Distribution Backtracking Distillation (DisBack). DisBask is composed of two stages: Degradation Recording and Distribution Backtracking. Degradation Recording is designed to obtain the convergence trajectory of the teacher model, which records the degradation path from the trained teacher model to the untrained initial student generator. The degradation path implicitly represents the teacher model's intermediate distributions, and its reverse can be viewed as the convergence trajectory from the student generator to the teacher model. Then Distribution Backtracking trains a student generator to backtrack the intermediate distributions along the path to approximate the convergence trajectory of teacher models. Extensive experiments show that DisBack achieves faster and better convergence than the existing distillation method and accomplishes comparable generation performance, with FID score of 1.38 on ImageNet 64x64 dataset. Notably, DisBack is easy to implement and can be generalized to existing distillation methods to boost performance. Our code is publicly available on https://github.com/SYZhang0805/DisBack.