TensorX
返回文献探索

Paper · arXiv 2607.24027

Sol-Attn: Accelerating Video Generation Inference via On-the-Fly Attention Sparsification

Haopeng Li, Yitong Li, Junsong Chen, Tian Ye, Haozhe Liu, Jincheng Yu, Duomin Wang, Ruihua Zhang, Zeke Xie, Enze Xie, Song Han

38 upvotesJuly 27, 2026arXiv 预印本
AI 摘要

Sol-Attn improves training-free sparse attention for diffusion transformers by combining dynamic block routing, sparse computation, and approximation correction in a single online pass to accelerate video generation without sacrificing quality.

diffusion transformerssparse attentiononline softmaxdynamic routingproxy-score reuseblock thresholdingtraining-free sparse attentionvideo generation

Abstract

Diffusion transformers are essential for high-fidelity video generation, but long token sequences make attention a dominant inference bottleneck. Training-free dynamic sparse attention alleviates this bottleneck by computing only selected key-value blocks, yet existing methods struggle to sparsify attention both efficiently and accurately for two reasons: (1) Rigid, unpredictable, and costly routing: selecting a fixed fraction of top-ranked blocks by proxy score imposes fixed budgets, whereas retaining blocks to reach a target cumulative proxy probability mass yields dynamic but potentially imbalanced budgets; both incur non-negligible overhead from computing and materializing proxy scores. (2) Lossy keep-or-drop sparsification: unselected blocks are discarded entirely, degrading accuracy under aggressive sparsity. These limitations motivate cheaper dynamic-budget routing while limiting accuracy degradation. In this paper, we introduce training-free Sol-Attn (Sparsifying online attention), which unifies dynamic routing, sparse computation, and approximation correction in a single online-softmax pass, achieving a better accuracy-efficiency trade-off in sparse attention. The core of Sol-Attn is on-the-fly block thresholding with proxy-score reuse, which selects critical blocks by comparing block proxy scores against a threshold during online softmax. This design enables dynamic yet controllable block budgets without materializing the proxy map, while directly reusing the proxy scores of unselected blocks to approximate their contribution. Experiments across image and video generation tasks show that Sol-Attn advances the quality-efficiency frontier of training-free sparse attention, delivering 2.1 times and 2.3 times end-to-end speedups for video generation and editing, respectively, while preserving visual quality.

北京市昌平区探索星信息技术及软件开发工作室

京ICP备2026059466号
Sol-Attn: Accelerating Video Generation Inference via On-the-Fly Attention Sparsification | TensorX