Summary
In this chapter, we learned how to take a trained model and deploy it as a managed service in Azure through a few simple lines of code. To do so, we learned how to prepare a model for deployment and looked into Azure Machine Learning auto-deployments and customized deployments.
We then took an NLP sentiment analysis model and deployed it as a real-time scoring service to ACI and AKS. We also learned how to define the service schema and how to roll out new versions effectively using endpoints and blue-green deployments. Finally, we learned how to integrate a model in a pipeline for asynchronous batch scoring.
In the last section, we learned about monitoring and operating your models using Azure Machine Learning services. We proposed to monitor CPU, memory, and GPU metrics as well as telemetry data. We also learned how to measure the data drift of your service by collecting user input and model output over time. Detecting data drift is an important metric that allows you...