CUDA matrix multiplication, reduction, and softmax kernels optimized for my RTX 4070 in C++17
CUDA Kernels
Build with cmake in cuda/ directory.
