Convolutional neural network calculation method, system on chip and electronic equipment

Abstract

The invention relates to the field of neural networks, and discloses a convolutional neural network calculation method, a system on chip and electronic equipment. The calculation method of the convolutional neural network comprises the following steps: determining a segmentation mode of a convolution kernel according to the size of a parameter cache on a deep learning processing chip; segmenting each convolution kernel in the plurality of convolution kernels according to the segmentation mode so as to divide each convolution kernel into N partial convolution kernels; grouping the plurality of convolution kernels to obtain N convolution kernel groups; loading the first to Nth convolution kernel groups to a parameter cache, and performing convolution operation on the input data and the convolution kernel groups loaded to the parameter cache to obtain N convolution operation results; and combining the N convolution operation results. Therefore, when the storage space of the parameter cache is smaller than the storage space required by the plurality of convolution kernels, the data of the convolution kernels does not need to be loaded frequently, and the efficiency of convolution operation can be improved.

Description

technical field [0001] The present application relates to the field of neural networks, and in particular to a calculation method of a convolutional neural network, a system on a chip and electronic equipment. Background technique [0002] In recent years, with the rapid development of artificial intelligence (AI) technology, AI-supported unmanned vehicles, unmanned aerial vehicles, and smart terminals have become more and more widely used. AI uses neural network technology to process the data input by various sensors in real time to realize the perception of the external environment. Usually, in order to improve the processing performance of AI application terminals, dedicated hardware platforms are used to implement specific operations. For example, convolution operations involved in AI application terminals originally transplanted with convolutional neural network models are implemented through dedicated hardware platforms. At the same time, in order to store the input d...

AI Technical Summary

Problems solved by technology

[0004] The embodiment of this application provides a calculation method of convolutional neural network, a system on chip and electronic equipment, which are used to solve the problem that in the existing solution, all buffers cannot be used for volume When multiple convolution kernels are used in the product operation, the convolution kernel data needs to be loaded frequently, which leads to the low efficiency of the convolution operation.

Images