TW Sparsity
TW Sparsity [SC'20] is a dense-GEMM-compatible sparsity pattern that can be used to accelerate DNN models on existing GEMM accelerator like NVIDIA's Tensor Core. The GitHub repo is available here.
TW Sparsity [SC'20] is a dense-GEMM-compatible sparsity pattern that can be used to accelerate DNN models on existing GEMM accelerator like NVIDIA's Tensor Core. The GitHub repo is available here.