According to the Microsoft Azure AI Fundamentals (AI-900) Official Study Guide and the Microsoft Learn module "Explore Azure Machine Learning," registering a model is the correct way to track multiple versions of models in Azure Machine Learning.
When you train models in Azure Machine Learning, each trained version can be registered in the workspace's Model Registry. Registration stores the model's metadata, including version, training environment, parameters, and lineage. Each registration automatically increments the version number, enabling you to manage, deploy, and compare multiple model iterations efficiently.
The other options are incorrect:
* A. Provision an inference cluster - Used for model deployment, not version tracking.
* B. Explain the model - Provides interpretability but does not track versions.
* D. Register the training data - Registers data assets, not models.