Calculation device, related product and calculation method for executing artificial neural network model

Abstract

The invention relates to a computing device and a related product, and a computing method for executing an artificial neural network model. The computing device can be used for executing self-learningoperation of a multi-layer artificial neural network, and the structure of the computing device is different from that of a general processor and an image processor, so that compared with the generalprocessor and the image processor, the computing device can improve the operation speed of the self-learning operation, save the operation time and further reduce the power consumption.

Description

technical field [0001] The present application relates to the technical field of artificial neural network, in particular to a computing device and related products, and a computing method for executing an artificial neural network model. 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 favored by academic circles and researchers due to their high recognition accuracy and good parallelism. industry is getting more and more attention. [0003] The self-learning operation of multi-layer artificial neural network is a very important link in the operation process of multi-layer artificial neural network. At present, the self-learning pre-training of multi-layer artificial neural networks adopts a layer-by-layer training method. For each layer, the pre-training is divided into...

AI Technical Summary

Problems solved by technology

Since the GPU is a device specially used to perform graphics and image calculations and scientific calculations, without special support for multi-layer artificial neural network operations, it still requires a lot of front-end decoding work to perform multi-layer artificial neural network operations, which brings a lot of problems. additional cost

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.

In addition, the GPU has only 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. The off-chip bandwidth has become the main performance bottleneck, and it has brought huge power consumption overhead.

[0013] Therefore, using the above-mentioned traditional technology to perform self-learning operations on multi-layer artificial neural networks, the power consumption overhead is relatively high

Images