Build and deploy PyTorch models with Azure Machine Learning
summary
Uses Azure Machine Learning workspace, NC6 GPU cluster, and the Python SDK to train a PyTorch model image classifier on a 10‑class Simpsons Lego dataset (10 characters × 50 images per class plus a test set), snapshotting code and data, saving outputs to an AKS endpoint. Training runs take about three minutes, the model registers (version 45), shows high validation confidences (examples: 97%, 88%), and packages into a Docker image that deploys to AKS for real‑time scoring.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:04What is machine learning
- 3:08Azure AI services overview
- 6:18Azure Machine Learning explained
- 9:28Setting up the environment
- 15:01Creating compute targets
- 19:18Data preparation and registration
- 21:26Experimenting with the model
- 27:48Submitting the training job
- 30:23Registering the trained model
- 31:50Testing the model locally
- 34:44Deploying the model to AKS
- 38:57Testing the deployed endpoint
- 41:09Recap and resources
- 42:13Q&A
related talks
59mBuild and deploy production ready PyTorch models
44mGetting started with Azure Machine Learning services
42mPython SDK to train, package and deploy models
10mGet Started with Azure Machine Learning
01h 00mA gentle introduction to Deep Learning
22mQuickly deploy NVIDIA optimized generative AI on Azure AI Studio
43mA lap around Azure Machine Learning services
51mSpeeding up deployment of machine learning models