Create and share assets with Amazon SageMaker Assets - Amazon SageMaker
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Create and share assets with Amazon SageMaker Assets

Use Amazon SageMaker Assets to provide controlled and regulated access to assets, models or data tables, belonging to your organization. Within SageMaker Assets, users from different Amazon accounts can create and share assets related to specific business problems without additional administrator overhead. Instead of having permissions being statically tied to their identity, users can provide permissions to assets that they’re using for their active workflows.

Assets are ML assets or data assets. ML assets are metadata that point to Amazon SageMaker Feature Store feature groups or SageMaker Model Registry Model Groups. Data assets are metadata that point to Amazon Redshift tables or Amazon Glue tables.

For example, the asset for a model group contains the model group name and the Amazon Resource Name (ARN) for the model package group. The asset points to the underlying collection of models. The asset itself can be shared between users.

Users can create assets for their own projects. They can make them visible to users who aren't members of those projects. The users who aren't project members can search through the assets and read their metadata. They can use the metadata to determine whether they want to access to the underlying source of data.

To understand the SageMaker Assets workflow better, imagine that you have two groups of users in your organization, Group A and Group B. The users in Group A are looking to predict home prices. They’re looking to collaborate with the users in Group B who are in a different Amazon account. They have housing data stored in Amazon Glue tables. They also have different models saved as model packages within a model group. With SageMaker Assets, the users in Group A can share their Amazon Glue tables and model packages with the users in Group B in a few clicks. Without administrator intervention, the users in Group A provided precisely scoped permissions to the users in Group B.

Users can create assets and publish them to make them visible throughout the organization. Other users can request access to those assets.

Topics