std :: ml::gpu_cuda
Function std::cuda_gemm_relu
int cuda_gemm_relu(
struct CudaBuf a,
int ta,
struct CudaBuf b,
int tb,
struct CudaBuf c,
unsigned long m,
unsigned long n,
unsigned long k,
)c = relu(op(a) * op(b)) in one kernel (cuBLASLt ReLU epilogue).