whisper.cpp/ggml-cuda/quantize.cuh

6 lines
188 B
Plaintext
Raw Normal View History

#include "common.cuh"
#define CUDA_QUANTIZE_BLOCK_SIZE 256
void quantize_row_q8_1_cuda(const float * x, void * vy, const int kx, const int ky, const int kx_padded, cudaStream_t stream);