Amazon SageMaker Feature Store enhances support with SageMaker Python SDK V3

Amazon SageMaker Feature Store has expanded its capabilities by incorporating support for the SageMaker Python SDK v3, which now includes enhanced features for Lake Formation access controls and configuration of Apache Iceberg table properties. The Feature Store serves as a fully managed repository designed to store, share, and manage features utilized in machine learning models.

With the integration of the modern and modular SDK v3 interfaces, data scientists are now empowered to manage feature groups with precise access control and optimized offline storage. The streamlined workflows and reduced boilerplate provided by the SageMaker Python SDK v3 facilitate improved management of feature groups.

Notably, the integration with Lake Formation allows data scientists to implement column-level and row-level access controls on offline store data. This is achieved through an opt-in setting during the creation of feature groups. Additionally, the support for Iceberg properties enables data scientists to directly configure additional table properties, such as compaction and snapshot expiration, via the SDK. This configuration is aimed at enhancing storage and query performance.

These advancements in capabilities allow data scientists to streamline governance of feature data access and optimize offline store performance using a single SDK, thereby eliminating the need to manage separate tools. The enhanced features are accessible across all AWS Regions where the Amazon SageMaker Feature Store is available.

To begin utilizing these new features, users are encouraged to install the SageMaker Python SDK version 3.8.0 or later. Further details can be found in the Lake Formation access controls and Iceberg metadata management documentation.