softmax.cuh 142 Bytes
Newer Older
mashun1's avatar
v1  
mashun1 committed
1
2
3
4
5
#include "common.cuh"

#define CUDA_SOFT_MAX_BLOCK_SIZE 1024

void ggml_cuda_op_soft_max(ggml_backend_cuda_context & ctx, ggml_tensor * dst);