SpargeAttn: Accurate Sparse Attention Accelerating Any Model Inference

Open in new window