llama.cpp/ggml-cuda.h

#ifdef  __cplusplus
extern "C" {
#endif

void dequantize_row_q4_0_cuda(const void * vx, float * y, int k, cudaStream_t stream);
void dequantize_row_q4_1_cuda(const void * vx, float * y, int k, cudaStream_t stream);
void dequantize_row_q4_2_cuda(const void * vx, float * y, int k, cudaStream_t stream);
void dequantize_row_q4_3_cuda(const void * vx, float * y, int k, cudaStream_t stream);

#ifdef  __cplusplus
}
#endif