CNN acceleration calculation method and system based on low-precision floating-point number data representation form

Abstract

The invention discloses a CNN acceleration calculation method and system based on a low-precision floating-point number data representation form, and relates to the field of CNN acceleration calculation. The acceleration calculation method comprises the following steps: a floating-point number function module receives an input activation value and a weight from a storage system according to a control signal, and distributes the input activation value and the weight to different processing units PE for convolution calculation to complete CNN acceleration calculation; wherein the convolution calculation comprises forward calculation of a convolution layer completed by performing dot product calculation on MaEb floating-point numbers quantized through a low-precision floating-point number representation form. By calculating and using a low-precision floating-point number representation form MaEb, the accuracy of the quantized CNN is ensured under the condition that retraining is not needed; Nm MaEb floating-point multipliers are realized through low-precision floating-point multiplication and a DSP (Digital Signal Processor), so that the acceleration performance of a customized circuit or a non-customized circuit is greatly improved under the condition of ensuring the accuracy, the customized circuit is ASIC (Application Specific Integrated Circuit) or SOC (System On Chip), and the non-customized circuit comprises an FPGA (Field Programmable Gate Array).

Description

technical field [0001] The invention relates to the field of deep convolutional neural network quantization, in particular to a CNN acceleration calculation method and system based on low-precision floating-point number data representation. Background technique [0002] In recent years, the application of AI (Artificial Intelligence, artificial intelligence) has penetrated into many aspects, such as face recognition, game battle, image processing, simulation, etc., although the processing accuracy has been improved, but because the neural network contains many layers and A large number of parameters require a very large calculation cost and storage space. In this regard, technicians have proposed a neural network compression processing scheme, that is, by changing the network structure or using quantization and approximation methods to reduce network parameters or storage space, and reduce network cost and storage space without greatly affecting neural network performance. ...

AI Technical Summary

Problems solved by technology

[0004] Although the current technology improves the quantization and improves the quantization accuracy, there are still several limitations: 1) For quantized deep convolutional neural networks (the number of convolutional layers / fully connected layers exceeds 100 layers), retraining is required to Guaranteed accuracy; 2) Quantization needs to use 16-bit floating-point numbers or 8-bit specific points to ensure accuracy; 3) Under the premise of not using retraining and ensuring accuracy, the current technology can only achieve two at most in one DSP. multiplication operations, resulting in lower acceleration performance on the FPGA

Images