CSRSPMM is a high-performance library for multiplying Compressed Sparse Row (CSR) matrices with dense matrices on NVIDIA GPUs. It includes a generic CUDA backend, and a PyTorch extension for easy ...
Abstract: Deep learning models rely heavily on matrix multiplication, which is computationally expensive and memory-intensive. Sparse matrices, which contain a high proportion of zero elements, offer ...
Abstract: The performance of sparse matrix vector multiplication (SpMV) is important to computational scientists. Compressed sparse row (CSR) is the most frequently used format to store sparse ...