Quantization and fixed-point fusion method and device of neural network

Abstract

The application discloses a neural network quantification and fixed-point fusion method, device, electronic equipment and storage medium, and relates to the field of artificial intelligence, especially the field of automatic driving (including autonomous parking). The specific implementation plan is: quantify the input data and weights of the current layer of the neural network; in the current layer, use the quantized weights to perform calculation operations on the quantized input data to obtain calculation results; The processing parameters are fixed-point processing; the preset processing parameters after the fixed-point processing are used to perform post-processing operations on the calculation operation results to obtain the output result of the current layer. Through the integration of quantization processing and fixed-point processing, the embodiment of the present application significantly reduces the bandwidth requirements for data transmission between operators, effectively reduces the calculation amount of the acceleration unit, and fully utilizes the advantages of the fixed-point calculation of the acceleration unit. The resource requirements for computing are reduced, and the computing efficiency is improved while saving resources.

Description

technical field [0001] The present application relates to the field of information technology, in particular to the field of artificial intelligence, especially the field of automatic driving (including autonomous parking). Background technique [0002] The calculation of traditional neural network is based on high-bit (bit) floating-point operations, which causes a lot of waste of computing resources, and is prone to overfitting, which reduces the generalization ability of the model. In the traditional acceleration method of neural network, even if a low-bit floating-point or integer budget is used, precision will be wasted in the process of processing intermediate floating-point operations, resulting in the final result being truncated before the subsequent process is used. operate. This approach wastes both precision and computational power. [0003] For example, the current neural network reasoning in embedded platform has the following solutions: (1) Select a small ne...

AI Technical Summary

Problems solved by technology

For example, in the application scenario of autonomous parking, traditional neural network calculations are based on high-bit floating-point operations, resulting in a large waste of computing resources

Or, in the traditional neural network acceleration method, even if a low-bit floating-point or integer budget is used, precision is wasted in the process of processing intermediate floating-point operations, which not only wastes precision but also reduces computing power.

Images