Do I have to deploy a model before doing inference?

[furtheraway]

Hi all,

I have trained a model successfully in Azure machine leaning workspace, and now I am working on how to make inference from that model. I followed the following link: https://github.com/microsoft/InnerEye-Inference

But I don't know where to get the following two parameters:

export CUSTOMCONNSTR_AZUREML_SERVICE_PRINCIPAL_SECRET=
export CUSTOMCONNSTR_API_AUTH_SECRET=

I guess I need to deploy my model in Azure first. Is there any guidence on how to do that? There seems a lot options to proceed:

[ant0nsc]

There is no need to deploy the model manually. If you have run training successfully in AzureML, the job overview page ("Details") will have an entry for "Registered model" - click on that link.
About the two environment variables:

• CUSTOMCONNSTR_AZUREML_SERVICE_PRINCIPAL_SECRET is the password for an Azure Service Principal (think of that as a machine account) that can talk to your AzureML workspace. You need to create this service principal first, and give it access to your AzureML workspace.
• CUSTOMCONNSTR_API_AUTH_SECRET is a random string that you can choose, you can use a GUID for example. This is the magic secret that you put into the API call to authenticate yourself to the inference app.

[furtheraway]

Thank you @ant0nsc !

Looks like the "Models" is located on the left pane now.
Thank you for explaining the two configuration parameters!