Neural network quantification method and device, electronic equipment and medium

• Main Authors: QU HUANYU, ZHANG WEIHAO
• Format: Patent
• Language: Chinese; English
• Published: 16.08.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention provides a neural network quantization method and device, electronic equipment and a medium, and the method comprises the steps: carrying out the quantization of a to-be-quantized neural network unit in a neural network, and obtaining a target neural network unit; aiming at any to-be-quantized neural network unit, expanding the neural network unit in a time dimension, and determining static data of a plurality of time steps of the neural network unit; quantizing the static data of the plurality of time steps to obtain corresponding quantized data; compressing the quantized data of the plurality of time steps to obtain shared data of the plurality of time steps; wherein the data volume of the shared data is less than that of the static data; and taking the neural network unit corresponding to the shared data as a target neural network unit. According to the embodiment of the invention, the quantification difficulty can be reduced, and the precision of the neural network is improved. 本公开提供了一种神经网络量