Data format conversion method and device and computer readable storage medium

Abstract

The invention provides a data format conversion method and device and a computer readable storage medium. The data format conversion method comprises the following steps of: respectively describing aninitial data format of target data transmitted to a target neural network layer and an effective data format supported by the target neural network layer by adopting standard format description, whenthe initial data format does not accord with the effective data format, obtaining a difference attribute between the initial data format and the effective data format, and converting the initial dataformat through a format conversion strategy corresponding to the difference attribute. Through the implementation of the scheme of the invention, the data format input to the neural network layer andthe data format supported by the neural network layer are described by adopting unified format description, and data format conversion is performed based on the difference between the two formats, sothat the applicability of format conversion is effectively expanded, and the effectiveness of format conversion is ensured.

Description

technical field [0001] The present application relates to the field of electronic technology, and in particular to a data format conversion method, device, and computer-readable storage medium. Background technique [0002] Artificial neural networks or deep learning methods based on artificial neural networks have made great progress and have been successfully applied to many fields such as image processing and language processing. [0003] In various neural network acceleration hardware, when data in a standard format is accelerated by a certain neural network layer, it may become a non-standard data format, and this non-standard data format may not be accepted by the next The neural network layer is used directly. In this case, the format needs to be transformed. At present, when performing format conversion, the conversion between several specific formats is usually performed, for example, only format A is supported to be converted to format B. However, in practical app...

AI Technical Summary

Problems solved by technology

[0004] The embodiment of the present application provides a data format conversion method, device, and computer-readable storage medium, which can at least solve the limitation of format conversion caused by related technologies that can only support format conversion between fixed formats in neural network accelerated computing. Stronger, poorer applicability and effectiveness

Images