D2Cache：面向更高视频扩散加速的二阶差分缓存
D2Cache: Second-Order Delta Caching for Higher Video Diffusion Acceleration
D2Cache 提出一种免训练的二阶残差缓存机制，通过利用一阶残差的时序差分（二阶残差）来更准确地预测未来时间步的输出，并基于时间步嵌入估计误差来自适应缩放二阶残差。该技术可直接用于加速视频扩散模型的推理，在 Latte、Open-Sora、LTX-video 和 Wan2.1 上均优于 TeaCache，且在高加速比下增益更大。
原文依据 we propose D2Cache, a training-free method that leverages the smoothness of second-order residual delta
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论视频扩散模型能够实现令人印象深刻的视觉保真度，但由于其顺序去噪过程，在实时或交互式生成中仍然计算代价高昂。近期的缓存方法通过跨时间步复用输出来加速推理，通常从一阶残差（即相邻模型预测之间的差异）估计每个新输出。为了减轻缓存方法中的累积误差，我们提出了 D2Cache，这是一种免训练的方法，利用二阶残差差分（即连续一阶残差之间的时序差异）的平滑性来更准确地预测未来时间步。我们从理论上证明了这种二阶校正提高了预测精度并有效抑制了累积误差。此外，D2Cache 使用从时间步嵌入导出的误差估计来自适应地缩放二阶差分，从而在不同的缓存间隔下保持准确性。实验上，D2Cache 在四个视频扩散模型（Latte、Open-Sora、LTX-video 和 Wan2.1）上以相当的加速率优于最优的 TeaCache，在更高的加速设置下表现出更大的增益。
英文原文摘要 1195 字符
Video diffusion models achieve impressive visual fidelity but remain computationally prohibitive for real-time or interactive generation due to their sequential denoising process. Recent caching methods accelerate inference by reusing outputs across timesteps, typically estimating each new output from the first-order residual, which is the difference between adjacent model predictions.To mitigate the accumulated error in caching methods, we propose D2Cache, a training-free method that leverages the smoothness of second-order residual delta, which is temporal differences between consecutive first-order residuals, to predict future timesteps more accurately. We theoretically show that this second-order correction improves prediction accuracy and effectively suppresses cumulative errors. Moreover, D2Cache adaptively scales second-order deltas using error estimates derived from timestep embeddings, maintaining accuracy across varying cache intervals.Empirically, D2Cache outperforms the state-of-the-art TeaCache across four video diffusion models (Latte, Open-Sora, LTX-video, and Wan2.1) at comparable acceleration rates, showing even larger gains under higher acceleration settings.