Evaluation of resources used by deep learning application

Abstract

According to the implementation of the invention, a method and a device for evaluating resources used by a deep learning application, and a computer program product are provided. According to one embodiment, the method comprises steps of obtaining a performance benchmark database, wherein the performance benchmark database includes at least structural data of one or more deep neural network models, time performance data of a plurality of deep learning applications based on the one or more deep neural network models, and computing resource consumption data; extracting a training data set basedon the performance benchmark database, the training data set having a plurality of parameter dimensions, the plurality of parameter dimensions including structures of deep neural network models of a plurality of deep learning applications, resource allocation of the plurality of deep learning applications, and training time of the plurality of deep learning applications; and establishing a corresponding relationship between the parameter dimensions of the training data set so as to construct an evaluation model for evaluating the resources used by the deep learning application. By means of thescheme, the resources used by the deep learning application defined by the user can be effectively evaluated, and on the basis, an optimized resource utilization scheme can be provided for the user.

Description

technical field [0001] Embodiments of the present disclosure relate to deep learning, and more particularly, to methods, apparatus, and computer program products for controlling write requests in a storage system. Background technique [0002] Deep learning applications are widely used in many fields such as image classification, machine translation, speech recognition, etc. For example, on the basis of big data and powerful computing resources, it is possible to train deep neural networks (Deep Neural Networks, DNN) with multiple layers and multiple parameters, also known as deep learning networks. Training deep learning applications based on deep neural networks is usually a very difficult task, which requires a lot of computing resources and time resources to process large-scale training data sets and iterate various parameters of deep neural networks. Usually, users cannot accurately evaluate the consumption of computing and time resources for a defined deep learning ap...

AI Technical Summary

Problems solved by technology

Training deep learning applications based on deep neural networks is usually a very difficult task, requiring a lot of computing resources and time resources to process large-scale training data sets and iterate various parameters of deep neural networks

Usually, users cannot accurately evaluate the consumption of computing and time resources for a defined deep learning application when developing a deep learning application. Therefore, the scheduling of processor resources is arbitrary or attempts to use computing resources in a maximized manner will lead to system failure. , energy consumption, waste of time

Images