llama.cpp/ggml-cuda/diagmask.cuh

6 lines
150 B
Plaintext
Raw Normal View History

#include "common.cuh"
#define CUDA_DIAG_MASK_INF_BLOCK_SIZE 32
void ggml_cuda_op_diag_mask_inf(ggml_backend_cuda_context & ctx, ggml_tensor * dst);