I have been developing a ML model using Azure ML workspace and the azure-ml SDK. The development of the model and deployment of the model works using the SDK. I have confirmed multiple times that after the endpoint is created I can call it with the correct payload and the webservice will return the correct results.
However, once I deploy a model to ACI and stop-start or restart the container instance resource in Azure. The endpoint stops working and I get a timeout error when calling the webservice.
Have any of you run into the same issues when deploying an Azure ML model to ACI?