Python SDK to train, package and deploy models
summary
Shows an end-to-end MLOps workflow using the Azure ML Python SDK: it connects a workspace to a dataset, runs experiments on a compute target, uses HyperDrive for parallel hyperparameter search, registers the best model, builds a container image and deploys a REST scoring endpoint (demo includes a live scoring URL). Also demonstrates assembling steps into Azure ML pipelines for scheduled retrain/deploy and exposing model explanations using SHAP/eli5-style techniques.
AI-written summary · about
chapters (9)
- 0:00Introduction to MLOps
- 3:28Industrializing ML concepts
- 7:33Azure machine learning overview
- 10:55The Python SDK for Azure ML
- 15:26Building and training pipelines
- 20:52Model deployment strategies
- 24:16Advanced workflow techniques
- 30:32Automated machine learning
- 37:02Future directions and closing
related talks
48mStreamlining your Machine Learning projects with Azure Machine Learning
43mA lap around Azure Machine Learning services
10mGet Started with Azure Machine Learning
57mDevOps for Machine Learning
21mMLOps for R with Azure Machine Learning
24mFrom Model to Production: Automating the ML Lifecycle with CI/CD
52mDevOps for AI in the Microsoft World
45mMLOps and ML-Flops: going from dev to prod