Unified Continuous Generative Models

TL;DR

This paper introduces a unified framework for training and sampling continuous generative models, achieving state-of-the-art results across multi-step and few-step methods, simplifying the paradigm and improving efficiency.

Contribution

The authors present UCGM, a unified training and sampling framework that consolidates diffusion, flow-matching, and consistency models, enabling improved performance and efficiency.

Findings

UCGM achieves 1.30 FID in 20 steps on ImageNet 256x256.

UCGM reduces FID to 1.06 in 40 steps with pre-trained models.

Unified approach simplifies training and sampling of continuous generative models.

Abstract

Recent advances in continuous generative models, including multi-step approaches like diffusion and flow-matching (typically requiring 8-1000 sampling steps) and few-step methods such as consistency models (typically 1-8 steps), have demonstrated impressive generative performance. However, existing work often treats these approaches as distinct paradigms, resulting in separate training and sampling methodologies. We introduce a unified framework for training, sampling, and analyzing these models. Our implementation, the Unified Continuous Generative Models Trainer and Sampler (UCGM-{T,S}), achieves state-of-the-art (SOTA) performance. For example, on ImageNet 256x256 using a 675M diffusion transformer, UCGM-T trains a multi-step model achieving 1.30 FID in 20 steps and a few-step model reaching 1.42 FID in just 2 steps. Additionally, applying UCGM-S to a pre-trained model (previously 1.26 FID at 250 steps) improves performance to 1.06 FID in only 40 steps. Code is available at: https://github.com/LINs-lab/UCGM.