NVGEMM

An NVIDIA-provided GEMM optimization component that integrates CUTLASS kernels generated by CuTeDSL into the PyTorch Inductor compilation backend, designed to improve the execution performance of matrix multiplication operations within deep learning frameworks.