Load-balanced sparse convolutional neural network accelerator and acceleration method thereof

Abstract

The invention discloses a load-balanced sparse convolutional neural network accelerator and an acceleration method thereof. The accelerator comprises a main controller, a data distribution module, a convolution operation calculation array, an output result caching module, a linear activation function unit, a pooling unit, an online coding unit and an off-chip dynamic memory. According to the scheme provided by the invention, the high-efficiency operation of the convolution operation calculation array can be realized under the condition of few storage resources, the high multiplexing rate of input excitation and weight data is ensured, and the load balance and high utilization rate of the calculation array are ensured; meanwhile, the calculation array supports convolution operations of different sizes and different scales and parallel scheduling of two layers between rows and columns and between different feature maps through a static configuration mode, and the method has very good applicability and expansibility.

Description

technical field [0001] The invention relates to a load-balanced sparse convolutional neural network accelerator and an acceleration method thereof, belonging to the technical field of deep learning algorithms. Background technique [0002] In recent years, deep learning algorithms have been widely used and achieved excellent results in computer vision, natural language processing and speech recognition, and convolutional neural network (CNN) is one of the most important algorithms. The higher accuracy of the convolutional neural network model often means deeper network layers, as well as more network parameters and calculations, of which 90% of the calculations are concentrated on the convolutional layer, so in order to better in the embedded To run convolutional neural networks efficiently on the system, it is imperative to optimize the energy efficiency ratio of convolution operations. [0003] There are two main characteristics of convolutional neural network CNN convolu...

AI Technical Summary

Problems solved by technology

[0004] Due to the irregular distribution of non-zero elements in the sparse convolutional neural network, invalid calculations will be added during the calculation process, resulting in a high idle rate of computing resources

Images