Templated deployment method for machine learning model and custom operator

Abstract

The invention relates to a templated deployment method for a machine learning model and a custom operator. The method comprises the following steps: A, setting a universal template according to the prediction process of the machine learning model; B, matching the framework name input by the user with the model name of the machine learning model in a configuration file, and replacing substituters in the universal template with corresponding fields in the configuration file; C, generating an executable file according to the universal template; D, running the universal template and the machine learning model uploaded by the user in the form of a container in a server, and adding the instantiated machine learning model into the service sequence; and E, when the user makes an HTTP request, searching a machine learning model instance in the service sequence according to the model ID or the custom operator ID input by the user, calling a prediction method in the model instance for calculation, and returning a result. According to the method, the deployment time of the machine learning model can be greatly shortened, and the personnel development cost and the time cost in the deployment link are remarkably reduced.

Description

technical field [0001] The invention relates to a method for deploying a machine learning model and a self-defined operator, in particular to a templated deployment method for a machine learning model and a self-defined operator. Background technique [0002] In recent years, machine learning technology has been widely used, and technologies in many industries have also undergone changes, such as financial risk control technology, image recognition technology, and autonomous driving technology. The application of machine learning technology is mainly divided into two steps: (1) training the model: fitting data through an algorithm, and persisting a reusable model; (2) deploying the model: deploying the model as an API (application programming interface) for other Apply system calls. The former is mainly studied in depth by academia, while the latter is an important and complex link in industrial production applications. Its main challenge is that there are many commonly us...

AI Technical Summary

Problems solved by technology

[0005] The disadvantage of this method is: the definition of MLeap Bundle is not standard and not compatible with all models, and the original purpose of MLeap is to solve the problem of mutual conversion between spark MLlib and sciket-learn models, and the support for other frameworks is weak

[0007] The disadvantage of this method is: as a framework that fully supports the Tensorflow model, it contains too many redundant functions for model deployment, which are relatively "cumbersome", such as node resource management in the training step, etc.

The problem to be solved by workflow is mainly to integrate a large number of unrelated computer programs and make them related. This prior art and the deployment of machine learning models belong to two different application fields.

Images