BaryCache：基于重心外推的扩散 Transformer 内存高效加速方法

AI 摘要

本文提出了一种名为 BaryCache 的训练免费加速方法，用于扩散 Transformer（DiT）的采样过程。该方法利用重心外推器进行逐步特征预测，通过稳定的权重构造和在线解耦技术，在减少内存占用和保持生成质量的同时，实现了高达 3.30 倍的端到端加速。实验在图像和视频生成任务上验证了其有效性。

正文节选

Memory-Efficient Training-Free Acceleration of Diffusion Transformers with BaryCache Abstract Abstract Diffusion Transformers achieve high-fidelity image and video generation, but their iterative sampling remains expensive, for each denoising step requires large matrix operations. Existing cache-based acceleration reduces redundant computation yet increases the VRAM footprint by storing intermediate states, which can directly constrain inference batch size. In this work, we propose a training-fr

发布时间：2026-09-01 12:00

抓取时间：2026-09-01 12:46

来源机构：arXiv