News

By quantizing weights with different precision for different parts of a network, mixed-precision quantization promises to reduce the hardware cost and improve the speed of deep neural network (DNN) ...