Azure Machine Learning is a cloud-based environment where you can use to train, deploy, automate, manage, and track ML models. Whether you prefer to write Python or R code or work with no-code/low-code options in the studio, you can build, train, and track machine learning and deep-learning models in an Azure Machine Learning Workspace. Data scientists and developers can access the assets of an Azure ML Workspace not only with the web portal (Azure Machine Learning studio) but also with Azure ML Python SDK.
In this article, you will learn how to authenticate to an Azure Machine Learning Workspace with Python, whether it is in your default tenant or not. There are several different methods to access an ML Workspace in a subscription in a defined (not default) tenant. Here we will discuss it with two scenarios: local computer and compute instance in ML Workspace.
Access to ML Workspace in a Default Tenant
Interactive authentication is the default mode when using Azure ML SDK. When you connect to your workspace using workspace.from_config, you will get an interactive login dialog. The config.json file can be downloaded in ML Workspace in Azure Portal. It can also be edited manually with the following template
{
"subscription_id": "my-subscription-id",
"resource_group": "my-ml-rg",
"workspace_name": "my-ml-workspace"
}
from azureml.core import Workspace
ws = Workspace.from_config()
Also, you can use the following code with the subscription ID, resource group, and workspace name as input arguments.
ws = Workspace(subscription_id="my-subscription-id",
resource_group="my-ml-rg",
workspace_name="my-ml-workspace")
Access to ML Workspace in a Defined Tenant on a Local Computer
With the above-mentioned method, if you get an error message like: All the subscriptions that you have access to = []
In such a case, the ML Workspace you want to access might not in your default tenant. You may have to name the tenant ID of the Azure Active Directory you’re using. You specify the tenant by explicitly instantiating InteractiveLoginAuthentication with Tenant ID as an argument. The Tenant ID can be found, for example, from Azure Portal under Azure Active Directory, Properties as Directory ID.
from azureml.core.authentication import InteractiveLoginAuthentication
interactive_auth = InteractiveLoginAuthentication(tenant_id="my-tenant-id")
ws = Workspace(subscription_id="my-subscription-id",
resource_group="my-ml-rg",
workspace_name="my-ml-workspace",
auth=interactive_auth)
Access to ML Workspace in a Defined Tenant on Compute Instance
If you want to access the ML Workspace not in your default tenant on a compute instance (Notebooks) in ML Workspace, the InteractiveLoginAuthenticationmethod also does not always work. Through personal experience, the following methods with AzureCliAuthentication can solve it.
First type following code in a terminal (opened with your Jupyter Notebook) or in a Jupyter Notebook cell with!magic. If you don’t know which tenant you can access, you can also type az login only, you can then see all the information of your tenants after you logged in. Select the right tenant and type the following code again.
az login --tenant MY_TENANT
After you logged in with Azure CLI, you can run the following code in your notebook.
from azureml.core.authentication import AzureCliAuthentication
cli_auth = AzureCliAuthentication()
ws = Workspace(subscription_id="my-subscription-id",
resource_group="my-ml-rg",
workspace_name="my-ml-workspace",
auth=cli_auth)
print("Found workspace {} at location {}".format(ws.name, ws.location))
This method also works on a local computer if you have installed azure-cli package.
Hope this article can help you. There are other ways of authentication, such as Managed Service Identity (MSI) authentication service principal authentication, token authentication. You can find more detail in this notebook.