All news
2 articles- paperHugging Face Daily Papers·21 Oct 20263.9
Sparsity Should Be ALLOTted by Output: Training-Free Allocation of Block-Sparse Attention for Video Diffusion Transformers
Training-free output-driven allocation of block-sparse attention in video diffusion transformers. - paperHugging Face Daily Papers·20 Oct 20263.9Stratified sparse attention narrows the gap to dense attention at extreme sparsity for long contexts.