llama.cpp/ggml/src/ggml-cuda/roll.cuh

6 lines
133 B
Plaintext

#include "common.cuh"
#define CUDA_ROLL_BLOCK_SIZE 256
void ggml_cuda_op_roll(ggml_backend_cuda_context & ctx, ggml_tensor * dst);