This article is sort or addendum to the Microsoft Learn articles for setting up Demand Forecasting with Azure Machine Learning. I got the request from one of my colleagues to set this up in out demo environment and as I went through the step-by-step guide I notices that some of the steps were unclear so I thought I would write down my own experiences:

1. To start I downloaded the setup files from GitHub and when I started setting them up I noticed some quirks. First of all you need to install Azure CLI and the correct ML extension. You get Azure CLI from here. To install the azure-cli-ml run the following command:

az extension add --name azure-cli-ml

Note that there are two different Azure ML extensions and the install script has a verification that checks that azure-cli-ml is installed… so the other one will not work (azure-cli-ml is the older version and ml is the newer version). The other thing to note is that the azure-cli-ml and ml is incompatible. If you have installed ml you need to uninstall it dy running:

az extension remove -n ml

I have not tested if you can just change the install script to use the new one instead.

2. The second issue I noticed was that there is a parameter hardcoded in the PowerShell Script which is also documented in GitHub. The solution is the the following line in quick_setup.ps1

$computeInstance_Name = "notebookScryptExecutor"

Change the variable value to something else. (Note that there is a maximum field length of 24).

3. The install script will create the following resources

– Azure AD App registration
– Azure ML Workspace
– Azure Storage Account with some containers

In order to do this, the correct permissions in Azure are needed. Otherwise, the script will fail.

Links
Set up the Azure Machine Learning Service
Demand forecasting setup
GitHub – microsoft/Templates-For-Dynamics-365-Supply-Chain-Management-Demand-Forecasting-With-Azure-Machine-Learning: Samples, templates and setup guides in order to run demand forecasting in Azure Machine Learning Service and integrate with Dynamics 365 SCM
Compute instance name hardcoded in quick_setup.ps1 · Issue #4 · microsoft/Templates-For-Dynamics-365-Supply-Chain-Management-Demand-Forecasting-With-Azure-Machine-Learning · GitHub