In the AzureML SDK you are able to specify when attaching an AKS Compute Instance to the workspace whether the purpose is DEVTEST or FASTPROD.

For development environments you can specify DEV_TEST so that the minimum 3 Nodes are not imposed on the cluster (keep cost down on test environmets).

I noticed that this functionality does not seemed to be exposed when using ARM Templates (https://docs.microsoft.com/en-us/azure/templates/microsoft.machinelearningservices/2020-01-01/workspaces/computes#AKS).

This means that if you are trying to easily spin up test environments using ARM templates it will set the purpose to FAST_PROD which means that you will need a minimum 3 nodes otherwise the deployment will fail.

This would be nice to have that way we don't have to create custom scripts to attach the AKS Compute Cluster to the ML Workspace.