Knowledge Graph — Coursera Notes › Organizations › Microsoft › Azure › Azure Kubernetes Service (AKS)

AKS Deployment Steps

feature · part of Azure Kubernetes Service (AKS)

• Register the trained model (e.g., sklearn-regression-model.pkl).
• Create an environment specifying conda and pip packages (e.g., numpy, scikit-learn, scipy, azureml-defaults, inference-schema).
• Write an entry script (score.py) with init() (loads model via joblib) and run() (parses JSON input, runs prediction, returns list).
• Create an inference config using the environment and entry script.
• Optionally profile the model to estimate CPU/memory needs (can take up to 25 minutes).
• Provision the AKS cluster (requires sufficient vCPU quota; default requests 12 vCPUs).
• Deploy the web service using akswebservice.deployConfiguration and model.deploy.
• Test the deployed service using the run method or raw HTTP request.
• Clean up by deleting the service and model.

This is the text view of an interactive 3D knowledge graph — open this page with JavaScript enabled to explore it visually.