std :: ml::gpu_cuda
Function std::cuda_to_half_many
int cuda_to_half_many(const &Vec<struct CudaBuf> src, const &Vec<struct CudaHalfBuf> dst)Every src[i] -> dst[i] (lengths dst[i].len / src[i].len as in cuda_to_half / cuda_from_half), 32 buffers per launch: a model's parameters in one or a few launches instead of one per tensor.