AdaCluster: Adaptive Query-Key Clustering for Sparse Attention in Video Generation

TL;DR

AdaCluster introduces an adaptive clustering framework for sparse attention in video diffusion transformers, significantly reducing inference latency while maintaining high accuracy.

Contribution

It presents a training-free, adaptive clustering method for query and key vectors that improves efficiency in DiTs without sacrificing performance.

Findings

Achieves up to 4.31x speedup in video generation.

Maintains negligible quality degradation across multiple datasets.

Operates efficiently on a single GPU.

Abstract

Video diffusion transformers (DiTs) suffer from prohibitive inference latency due to quadratic attention complexity. Existing sparse attention methods either overlook semantic similarity or fail to adapt to heterogeneous token distributions across layers, leading to model performance degradation. We propose AdaCluster, a training-free adaptive clustering framework that accelerates the generation of DiTs while preserving accuracy. AdaCluster applies an angle-similarity-preserving clustering method to query vectors for higher compression, and designs a euclidean-similarity-preserving clustering method for keys, covering cluster number assignment, threshold-wise adaptive clustering, and efficient critical cluster selection. Experiments on CogVideoX-2B, HunyuanVideo, and Wan-2.1 on one A40 GPU demonstrate up to 1.67-4.31x speedup with negligible quality degradation.