To run any script/notebook on your workspace, we need to create a compute instance in the workspace. And this compute instance needs to have the user_assigned managed identity assigned as we deal with authenticating a lot of azure services.
Creating a compute instance
Once you launch to your machine learning studio workspace, you will see ‘compute’ under manage. Create cpu/gpu based on your requirement.
Once it is created, open your compute instance and assign the managed identity you created previously.
Creating a custom environment
The AISDK must be installed as a part of the custom environment, the model packaging/delivery will be based on this environment. Upload the SDK wheel files to the custom environment (Latest SDK version is recommended).
Assets->Environment: Docker environment source.
The environment will be built with a version ID.
Azure Notebooks
The Azure ML pipeline we build manages the packaging of the model as well as the delivery to your AI Model Manager instance.
The image below shows the pipeline as you'll see it in the Azure ML portal once submitted.
pack_component.ipynb notebook interactively orchestrates this above workflow. The notebook orchestrates the entire workflow, acting as a central point to invoke the packaging, registration, and delivery processes interactively. Running this notebook delivers the package to your model manager app.
We are using DefaultAzureCredential to get access to workspace. DefaultAzureCredential should be capable of handling most Azure SDK authentication scenarios.
In the next cell, enter your Subscription ID, Resource Group name and Workspace name. To find your Subscription ID:
In the upper right Azure Machine Learning Studio toolbar, select your workspace name.
At the bottom, select View all properties in Azure Portal
Copy the value from Azure Portal into the code**.**
The result is a handler to the workspace that you'll use to manage other resources and jobs.
The packaging & delivery component environment and the steps are discussed below.