ollama/llama/ggml-cuda/quantize.cuh
2024-09-03 21:15:12 -04:00

6 lines
200 B
Plaintext
Vendored

#include "common.cuh"
#define CUDA_QUANTIZE_BLOCK_SIZE 256
void quantize_row_q8_1_cuda(const float * x, void * vy, const int64_t kx, const int64_t ky, const int64_t kx_padded, cudaStream_t stream);