不训练即可加速视频生成，内存占用更低且速度更快。
LightCache: Memory-Efficient, Training-Free Acceleration for Video Generation
- 分阶段优化缓存策略，异步交换、分块处理、切片解码
- 相比基线提速显著，内存消耗降低，质量下降可控
- 适合需要低内存高效率推理的视频生成应用
基于扩散模型的视频生成中，无训练加速已成为前沿方向。扩散模型推理过程中的潜在表示冗余为加速提供了天然切入点。本文将推理过程分解为编码、去噪和解码三个阶段，发现基于缓存的加速方法常导致后两个阶段出现显著内存激增。为此，我们分析各阶段特性，提出针对性的内存优化策略：1）异步缓存交换；2）特征分块；3）切片解码。同时确保三种策略引入的时间开销低于其带来的加速收益。与基线相比，本方法在保持质量损失可接受的前提下，实现更快的推理速度和更低的内存占用。代码已开源：https://github.com/NKUShaw/LightCache。
原文摘要 · Abstract (English)
Training-free acceleration has emerged as an advanced research area in video generation based on diffusion models. The redundancy of latents in diffusion model inference provides a natural entry point for acceleration. In this paper, we decompose the inference process into the encoding, denoising, and decoding stages, and observe that cache-based acceleration methods often lead to substantial memory surges in the latter two stages. To address this problem, we analyze the characteristics of inference across different stages and propose stage-specific strategies for reducing memory consumption: 1) Asynchronous Cache Swapping. 2) Feature chunk. 3) Slicing latents to decode. At the same time, we ensure that the time overhead introduced by these three strategies remains lower than the acceleration gains themselves. Compared with the baseline, our approach achieves faster inference speed and lower memory usage, while maintaining quality degradation within an acceptable range. The Code is available at https://github.com/NKUShaw/LightCache .
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。