Winograd convolution splitting method for convolutional neural network accelerator

Abstract

The invention discloses a Winograd convolution splitting method for a convolutional neural network accelerator. The method comprises the following steps: 1) reading an input and a convolution kernel of any size from a cache of the convolutional neural network accelerator; 2) judging whether convolution splitting is carried out or not according to the convolution kernel size and the input size, andif convolution splitting needs to be carried out, carrying out the next step; 3) splitting the convolution kernel according to the size and the step length of the convolution kernel, and splitting the input according to the size and the step length of the input; 4) combining and zero-filling the split elements according to the size of the convolution kernel, and combining and zero-filling the split elements according to the input size; 5) performing Winograd convolution on each pair of split input and convolution kernels; 6) accumulating the Winograd convolution results of each input and convolution kernel, and 7) storing the accumulation results in a cache of the convolutional neural network accelerator, so that the convolutional neural network accelerator can support convolution of various different shapes by adopting one Winograd acceleration unit.

Description

technical field [0001] The invention belongs to the field of convolutional neural network algorithms, in particular to a Winograd convolution splitting method for convolutional neural network accelerators. Background technique [0002] Convolutional neural network (CNN) is being widely used in computer vision tasks such as object detection and image classification. However, with the continuous development of network models and the continuous improvement of recognition accuracy, it brings a huge amount of computation and data. Therefore, hardware devices with high performance and low power consumption are required, and at the same time, the flexibility of hardware devices must be guaranteed to meet various network models. [0003] Convolutional neural network accelerators are widely used to accelerate convolutional neural network algorithms on mobile and server. In order to improve its performance, the Winograd algorithm is used to reduce the hardware multiplier introduced b...

AI Technical Summary

Problems solved by technology

In order to expand the flexibility of the accelerator, Winograd computing units with various parameters need to be designed, which increases the resource consumption and power consumption of the accelerator

Secondly, the accelerator obtains data streams of different shapes from the Winograd operation unit with different parameters, which reduces the utilization rate of the accelerator operation unit and seriously reduces the performance of the accelerator.

Images