Deep learning model management method and system

Abstract

The invention relates to a deep learning model management method and system, and the method comprises the steps: selecting a deep learning model framework from a server, and modifying the hyper-parameters of a deep learning network according to the actual demands; making a data set, taking the data set as a training sample, and inputting the training sample into the modified deep learning networkfor training; renaming the trained deep learning model and training data set according to a given naming rule, and uploading the renamed deep learning model and training data set to a server for storage; and selecting a specified deep learning model from the server to evaluate each index of the prediction sample. The model and the training model file are associated, so that confusion of the training model file and the model is avoided; the model is renamed according to a certain naming mode, so that a user can quickly understand a framework used for training the model, a targeted target, training parameters, various evaluation indexes of the model and the like, the user can conveniently and quickly select a proper model, the time for re-writing codes and debugging is shortened, and the development process is accelerated.

Description

Technical field [0001] The present invention relates to the field of information technology communication, in particular to a management method and system for a deep learning model. Background technique [0002] Recently, deep learning has set off a new wave of artificial intelligence, which is widely used in various fields such as unmanned driving, medicine, face recognition, and speech understanding and translation. Deep learning training requires input of a large number of samples. During the training process, the model parameters of each stage will be saved to the machine. This method of saving model files will take up a lot of computer memory, the naming method is relatively simple, and the management method is relatively cumbersome. And inefficient. [0003] Through the analysis and data collection of a series of deep learning in the early stage, we understand that, so far, most deep learning researchers manage the generated models and the data sets of each training model th...

AI Technical Summary

Problems solved by technology

In the face of a large number of experiments, a large number of models, and a large number of data sets, deep learning researchers should face many difficulties when performing model analysis. For example, researchers need to name folders according to certain rules. When the naming rules become cumbersome, it is inconvenient to find model data, and the data model cannot be displayed intuitively when performing model analysis and comparing the similarities and differences between models. Therefore, researching a flexible, convenient and effective deep learning model management system can reduce the time cost of researchers managing model data and improve the efficiency of model optimization, which is a research with important research and application value

Images