The final interesting piece of Azure Machine Learning is the deployment tooling that is included with it. The deployment tooling allows you to take a model from the model registry and deploy it to production.
Before you can deploy a model to production, you need to have an image that includes the model and a scoring script. The image is a Docker image that includes a web server, which will invoke the scoring script when a request is made against it. The scoring script accepts input in the form of a JSON payload, and uses it to make a prediction using the model. The scoring script for our iris classification model looks like this:
import osimport jsonimport numpy as npfrom azureml.core.model import Model ...
