Demo: Publish a regression model service

Overview

In this article, we will show you how to use Luna service to publish an AI service to train regression model and use it to do batch inference or real time scoring.

Deploy Luna Service

You can deploy Luna service by following the instruction here. If you want to use our pre-deployed demo environment, please contact xiwu@microsoft.com

Clone Luna project template

Clone the template from here

Update AML environment

Update AML workspace info

Update the value of "Scope" property in ./.cloud/.azureml/test_workspace.json with your AML workspace Azure resource Id

Update AML compte targets

Update the value of "target" property in ./.cloud/.azureml/aml_run_config.yml with the compte cluster name. The compute cluster should be already created and registered in your AML workspace.

Update your deployment targets

Update the value of "deployment_target" in ./.cloud/.azureml/compute.yml. Right now we support aci (Azure Container Instance) or aks (Azure Kubernetes Service). If you choose AKS as deployment target, you also need to update the value of "aks_cluster" to the name of AKS cluster you created and registerd in AML.