Soft error detection method and device

Abstract

The invention discloses a soft error detection method and device, and is applied to the technical field of computers. The method comprises the following steps: determining a redundancy factor of a target layer according to parameters of a systolic array and the total number of filters of the target layer of a target convolutional neural network; determining a fragile filter for redundancy execution in each layer of network according to elastic difference information between filters of the target convolutional neural network and each redundancy factor; in the weight loading process of the target convolutional neural network, each fragile filter is repeatedly loaded into the pulse array; according to the method, the output information of the current period of the pulse array is sent to the cache region for numerical verification, and error detection is performed on the output information of the repetition filter of the next period of the pulse array, so that the energy overhead is remarkably reduced on the basis of ensuring the error detection capability.

Description

technical field [0001] The present application relates to the field of computer technology, and in particular, to a soft error detection method and device. Background technique [0002] In recent years, convolutional neural networks have been widely deployed in IoT smart terminals, autonomous driving and other application systems due to their superior performance. Relying on a large amount of training data, network algorithms can be applied to many technical fields such as image classification and detection, human-computer games, and natural language processing. Typically, a convolutional neural network consists of convolutional layers, pooling layers, and fully connected layers, and most of the computation occurs in the convolutional layer. Each convolutional layer consists of many filters, which are used to capture various features. The pooling layer is used to divide the feature map into several parts, and after processing through specific rules such as max pooling and ...

AI Technical Summary

Problems solved by technology

In addition, when the size of the specific filtering layer cannot be evenly distributed by the width of the systolic array, multiple filters in the convolutional layer will also show elastic differences when errors occur due to their functional tendency. According to statistics, the same convolutional layer Only a small number of filters in the network appear to be error-sensitive and ultimately affect the results of convolutional neural network calculations

Therefore, the dual-mode redundancy method performs poorly in terms of power

Images