基于云原生服务网格的AI模型部署方案

人工智能模型训练完成后并不适合直接运行在互联网生产环境,需要对其进行一定的封装和部署。探讨基于容器和云原生服务网格对AI模型进行自动封装和部署,以微服务形式对外提供Restful API,能够提高模型的兼容性和部署灵活性。封装过程采用适用多框架的通用AI模型代码打包方法,解决AI模型运行环境配置问题;通过部署在云原生平台由服务网格进行管理,实现AI模型从实验室开发到互联网生产环境高效应用的自动化实时流转。

Cloud Native Service Grid Based AI Model Deployment Scheme

After the training of the Artificial Intelligence model,it is not suitable to run directly in the Internet production environment,so it needs to be packaged and deployed.It discusses the automatic encapsulation and deployment of AI model based on container and cloud native service grid,and restful API is provided in the form of microservices to improve the compatibility and deployment flexibility of the model.The encapsulation process uses a general AI model code packaging method applicable to a variety of frameworks,and solves the AI model running environment configuration problem.Microservices are deployed on the Cloud platform and managed by the service mesh,so the AI model can be efficiently applied from the laboratory to the Internet production environment.