Target-Driven Distillation: Consistency Distillation with Target Timestep Selection and Decoupled Guidance

TL;DR

This paper introduces Target-Driven Distillation (TDD), a novel approach for diffusion models that improves image quality and efficiency by strategic timestep selection and flexible guidance, achieving state-of-the-art results in few-step generation.

Contribution

TDD proposes a new target timestep selection strategy and decoupled guidance, enhancing efficiency and flexibility in consistency distillation for diffusion models.

Findings

TDD achieves state-of-the-art performance in few-step image generation.

TDD reduces blurs and detail loss compared to previous methods.

Flexible sampling and guidance tuning improve image quality.

Abstract

Consistency distillation methods have demonstrated significant success in accelerating generative tasks of diffusion models. However, since previous consistency distillation methods use simple and straightforward strategies in selecting target timesteps, they usually struggle with blurs and detail losses in generated images. To address these limitations, we introduce Target-Driven Distillation (TDD), which (1) adopts a delicate selection strategy of target timesteps, increasing the training efficiency; (2) utilizes decoupled guidances during training, making TDD open to post-tuning on guidance scale during inference periods; (3) can be optionally equipped with non-equidistant sampling and x0 clipping, enabling a more flexible and accurate way for image sampling. Experiments verify that TDD achieves state-of-the-art performance in few-step generation, offering a better choice among consistency distillation models.