A device and method for performing neural network operations

Abstract

The present disclosure provides a device and method for performing neural network operations. The device includes an on-chip interconnection module and a plurality of neural network processing modules communicatively connected to the on-chip interconnection module. Read and write data in the network processing module. In the multi-core multi-layer artificial neural network operation, it is necessary to divide the neural network operation of each layer, and then perform the operation by multiple neural network processing modules to obtain their own operation result data. Result data for data exchange.

Description

Technical field [0001] The present disclosure belongs to the field of neural network operations, and in particular relates to a device and method for performing neural network operations. Background technique [0002] Multi-layer artificial neural networks are widely used in the fields of pattern recognition, image processing, function approximation and optimization calculations. In recent years, multi-layer artificial networks have been widely accepted by the academic community due to their higher recognition accuracy and better parallelism. Industry is getting more and more attention. Artificial neural networks involve a variety of algorithms. Among them, multi-core and multilayer neural network processors are used to perform neural network operations, and they are widely used in various artificial neural network models and various scenarios where neural networks are used. [0003] One known method to support multi-core and multilayer neural network operations is to use a genera...

AI Technical Summary

Problems solved by technology

Since the GPU is a device specially used to perform graphics and image operations and scientific computing, without special support for artificial neural network operations, a large amount of front-end decoding work is still required to perform multi-layer artificial neural network operations, which brings a lot of additional overhead.

In addition, the GPU only has a small on-chip cache, and the model data (weights) of the multi-layer artificial neural network need to be repeatedly moved from off-chip, and the off-chip bandwidth has become the main performance bottleneck.

Images