Function quantizeBlockwise_fp32_nf4

csrc/pythonInterface.cpp:165–167

}

void quantizeBlockwise_fp32_nf4(float* code, float* A, float* absmax, unsigned char* out, int blocksize, const int n) {
    quantizeBlockwise<float, 0, NF4>(nullptr, A, absmax, out, nullptr, 0, blocksize, n);
}

void dequantizeBlockwise_fp16(
    float* code, unsigned char* A, float* absmax, half* out, int blocksize, const int n, cudaStream_t stream