1 2 3 4 5
#pragma once #include "common.cuh" #define CUDA_CONV2D_BLOCK_SIZE 256 void ggml_cuda_op_conv2d(ggml_backend_cuda_context & ctx, ggml_tensor * dst);