News

Fused3S is a CUDA kernel library that accelerates sparse attention by fusing Sampled Dense-Dense Matrix Multiplication (SDDMM), Softmax, and Sparse Matrix Multiplication (SpMM) into a single optimized ...
2023-08-31 Improving Mandarin Prosodic Structure Prediction with Multi-level Contextual Information Jie Chen et.al. 2308.16577 null 2023-08-31 LightGrad: Lightweight Diffusion Probabilistic Model for ...