Posted On: Nov 19, 2021
Amazon SageMaker Model Registry, the purpose-built service which enables customers to catalogue their ML models, now supports cross account registration of model versions.
SageMaker Model Registry catalogs customer’s models in a logical group (a.k.a. model package group) and stores incremental versions of models as model package versions. Model package version primarily stores model artifact and metadata. In addition, Model Registry allows customers to approve/reject/deploy model package version using Sagemaker Studio and SDK.
Many customers have multiple accounts set up where they configure different AWS accounts to - train, register and deploy models. Typically, customers want to perform an operation from source AWS account in target AWS account. Previously, users in source AWS account could approve, reject or deploy model package versions located in a target AWS account. However, cross account registration of new model package version was not supported. Now, customers can create cross-account resource policy with Describe/Create/Update/List/Delete actions and attach this policy to a model package group. Once configured, customers can perform cross account registration and deployment of model versions. This will allow customers to organize their model development and registration in separate AWS accounts.
This feature is available in all AWS regions where Amazon SageMaker is available. To get started, create a new SageMaker Model Package Group from the Amazon SageMaker SDK or Studio and visit our documentation page on cross-account model registration.