Posted On: Jul 4, 2023
Today, we are excited to announce the integration of Amazon SageMaker Model Cards and Amazon SageMaker Model Registry. This integration allows you to associate a SageMaker Model Card with a specific model version in SageMaker Model Registry, so you can easily track and audit your models.
Amazon SageMaker Model Registry helps you centrally manage your machine learning (ML) models, allowing you to review and approve models for deployment. When you register a model, a model package is created in SageMaker Model Registry, storing all subsequent versions under a single model package group. Amazon SageMaker Model Cards enables you to document critical details about your ML models in a single place for streamlined governance and reporting. You can document information such as the model's intended uses, ethical considerations, your risk rating, and performance goals. Furthermore, available model metadata is auto-populated in the associated SageMaker Model Card. For example, when you associate a SageMaker Model Card with a registered model version, the system automatically pulls information such as training details and metrics, evaluation results, source algorithms, inference specification, model package group, and model approval status from SageMaker Model Registry and surfaces it in the SageMaker Model Card.
The integration of Amazon SageMaker Model Cards with SageMaker Model Registry enables you to establish a single source of truth for your registered model versions, with comprehensive, centralized, and standardized documentation across all stages of the model’s journey on SageMaker. It facilitates discoverability and promotes governance, compliance, and accountability throughout the model lifecycle.
Amazon SageMaker Model Cards is available in all Amazon Web Services regions where Amazon SageMaker is currently available, including in Amazon Web Services China (Beijing) Region, operated by Sinnet, and the Amazon Web Services China (Ningxia) Region, operated by NWCD.
To get started, create a new or map an existing model card to a model package version via the Amazon SageMaker Python SDK. Visit the Amazon SageMaker developer guide for additional information on SageMaker Model Cards.