Method for constructing online machine learning project and machine learning system

Abstract

An embodiment of the invention provides a method for constructing an online machine learning project and a machine learning system. The method for constructing the online machine learning project comprises the following steps of: receiving a service request from a user; allocating resources of a user service cluster to a service according to the service request, and constructing an online operation environment, wherein the online operation environment is used for executing a model development service or a model training service of machine learning; receiving an operation from the user in the online running environment; and providing operation information to the user based on the operation, wherein the operation information comprises an operation result or an operation state of the model development service or the model training service. According to the method of the invention, the user locally compiles codes to realize model development or locally uses cloud resources to realize modeltraining, and returns a code development running result and a model training result executed at the cloud to a user browser, so that the effect of the user for performing machine learning projects isimproved.

Description

technical field [0001] The present application relates to the field of machine learning, and specifically, the embodiments of the present application relate to a method for constructing an online machine learning project and a machine learning system. Background technique [0002] At present, the common machine learning (for example, deep learning) solution in the industry is to use a physical machine or virtual machine with a high-performance graphics card to download the data set and model to the local disk, and then write the training code through the command line, run the code, Get the training results of the code execution in the standard output or log file for analysis. [0003] The inventors of the present application have discovered that the above related solutions include at least the following four defects. The first aspect is poor flexibility, because it is difficult to easily change computing resources due to the use of a single machine; second aspect, the data ...

AI Technical Summary

Problems solved by technology

The first aspect is poor flexibility, because it is difficult to easily change computing resources due to the use of a single machine; second aspect, the data flexibility and use threshold are high, and the specifications of the data sets vary greatly. Frequent uploading and downloading of data sets will Brings a huge time cost; thirdly, the lack of management of models and data sets will lead to the inability to manage the data on the machine after a long time; fourthly, writing and running code through the command line cannot achieve friendly saving and running experiments Status and data, let alone analyzing data and sharing experiments

Images