LinCa：基于可学习分解特征缓存的扩散模型加速方法
LinCa: Accelerating Diffusion Models via Learnable Decomposed Feature Caching
浏览论文内容
中文总结 AI 辅助
本文提出LinCa框架，通过可学习可逆网络分解缓存特征并差异化预测，仅需少量额外参数即可在5-7倍加速下使扩散模型保持近无损质量，性能优于现有方法。
中文摘要 AI 辅助
扩散模型在图像和视频生成领域已取得显著成果，但迭代采样的高计算成本仍是其实际部署的关键瓶颈。特征缓存作为一种有前景的加速范式，通过跨时间步复用或预测中间特征实现加速。然而，现有无训练方法采用统一预测策略，无法适配异质特征动态性，在高加速比下会导致明显的质量下降。本文提出LinCa，一种基于可学习可逆网络的特征缓存框架：LinCa通过轻量可逆网络将缓存特征分解为具有不同连续性属性的子组件，并为每个组件应用匹配的差异化预测顺序；严格的可逆性保证可无损重构回原始特征空间，形成统一的分解-预测-重构流程。通过为不同模型和时间步段训练独立预测器，LinCa可适配异质特征动态性。在FLUX、Qwen-Image和HunyuanVideo上的实验表明，LinCa仅需不到0.2%的额外参数，即可显著优于现有方法，并在5-7倍加速下保持近无损质量。代码：this https URL
英文摘要
Diffusion models have achieved remarkable success in image and video generation, yet the high computational cost of iterative sampling remains a critical bottleneck for practical deployment. Feature caching has emerged as a promising acceleration paradigm by reusing or predicting intermediate features across timesteps. However, existing training-free methods apply uniform prediction strategies that cannot adapt to the heterogeneous feature dynamics, causing significant quality degradation under high acceleration ratios. We propose LinCa, a feature caching framework based on learnable invertible networks. LinCa decomposes cached features into sub-components with distinct continuity properties via a lightweight invertible network and applies differentiated prediction orders matched to each component. The strict invertibility guarantees lossless reconstruction back to the original feature space, forming a unified Decompose-Predict-Reconstruct pipeline. By training separate predictors for different models and timestep segments, LinCa adapts to heterogeneous feature dynamics. Experiments on FLUX, Qwen-Image, and HunyuanVideo demonstrate that LinCa, with less than 0.2% additional parameters, significantly outperforms existing methods and maintains near-lossless quality at 5-7x speedup. Code: https://github.com/QHR69/LinCa
发表机构
- Shanghai Jiao Tong University（上海交通大学）
- Shandong University（山东大学）
- Terminal Intelligent Computing Division, Alibaba Cloud（阿里云终端智能计算事业部）
- South China University of Technology（华南理工大学）
- Jilin University（吉林大学）
机构由 AI 辅助整理，请以论文原文为准。
