Question

Your organization runs three separate Azure Machine Learning workspaces (dev, test, prod) in different resource groups. The MLOps team has built a custom training environment with a specific Docker base image and conda dependencies. They want this exact environment, including version history, to be reusable and consistent across all three workspaces without rebuilding or re-registering it in each one. Which approach should you use?

Accepted Answer

Create an Azure Machine Learning registry and register the environment asset in the registry, then reference it from each workspace.

Answer

Register the environment separately in each workspace using the same YAML definition file.

Answer

Store the environment's Dockerfile in a shared Azure Blob container and mount it as a datastore in each workspace.

Answer

Export the environment from the dev workspace and import it into an Azure Container Registry shared by all workspaces.