提出新型采样调度策略，10步生成高质量图像
Analyzing and Improving Fast Sampling of Text-to-Image Diffusion Models
- 基于微分几何设计均匀变化的采样路径
- 10步采样在Flux.1-Dev和SD 3.5上生成高质量图像
- 无需训练，适配多种模型与参数配置
文本到图像扩散模型虽取得突破性进展，但在有限采样预算下仍难生成高质量结果。现有无训练加速方法独立发展，缺乏整体性能与兼容性研究。本文系统分析设计空间，发现采样时间调度是关键因素。基于Frenet-Serret公式揭示的扩散模型几何特性，提出恒定总旋转调度（TORS），确保采样轨迹上几何变化均匀。TORS在不依赖训练的情况下，仅用10步采样即在Flux.1-Dev和Stable Diffusion 3.5上生成高质量图像，且对未见模型、超参数及下游任务具有强适应性。代码已开源。
原文摘要 · Abstract (English)
Text-to-image diffusion models have achieved unprecedented success but still struggle to produce high-quality results under limited sampling budgets. Existing training-free sampling acceleration methods are typically developed independently, leaving the overall performance and compatibility among these methods unexplored. In this paper, we bridge this gap by systematically elucidating the design space, and our comprehensive experiments identify the sampling time schedule as the most pivotal factor. Inspired by the geometric properties of diffusion models revealed through the Frenet-Serret formulas, we propose constant total rotation schedule (TORS), a scheduling strategy that ensures uniform geometric variation along the sampling trajectory. TORS outperforms previous training-free acceleration methods and produces high-quality images with 10 sampling steps on Flux.1-Dev and Stable Diffusion 3.5. Extensive experiments underscore the adaptability of our method to unseen models, hyperparameters, and downstream applications. Code is available at https://github.com/zju-pi/TORS.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。