Amazon SageMaker JumpStart is a machine learning hub that offers pre-trained models and solutions. It allows access to hundreds of foundation models (FMs) and includes a private hub feature for sharing models and notebooks within an organization. Enterprise admins can now configure granular access control over the FMs available in SageMaker JumpStart to restrict user access to specific models based on their requirements.
By using the private hub feature, admins can create repositories tailored to different teams, use cases, or license requirements. They can set up multiple private hubs with different lists of discoverable models for different user groups, providing centralized control over model access. This feature enables organizations to consume the latest in AI development while enforcing governance guardrails. Private hubs can be shared across multiple AWS accounts using AWS Resource Access Manager (AWS RAM) for collaborative model management.
The process for admins to configure granular access control of models in SageMaker JumpStart using a private hub involves updating the SageMaker Python SDK, configuring the private hub, setting up Boto3 client for SageMaker, and managing IAM permissions for users. Admins can then create the private hub, curate models, and configure access control to manage user permissions within their organization.
Users can interact with allowlisted models in SageMaker JumpStart by accessing the SageMaker Studio or using the SageMaker Python SDK. Within SageMaker Studio, users can choose model hubs they have access to, view model details, deploy models, and modify configurations. Using the SageMaker Python SDK, users can list available models in their private hub, get information about specific models, deploy models to endpoints, invoke endpoints with payloads, and delete model endpoints.
Private hubs in SageMaker JumpStart support cross-account sharing, allowing organizations to share curated model repositories across different AWS accounts. This feature enables collaboration and consistent model governance within an organization. Admins can create resource shares and grant access to private hubs to other accounts using AWS RAM, ensuring secure sharing of resources. By extending private hubs across accounts, organizations can promote collaboration and maintain model governance throughout their organization.
Overall, SageMaker JumpStart enables enterprises to adopt foundation models with granular control over model access and usage. By creating private hubs, organizations can manage and share approved models within their organization, aligning their AI initiatives with corporate policies and regulations. The private hub feature simplifies model curation and consumption, allowing admins to manage model inventory while data scientists focus on developing AI solutions. The post provides detailed steps for setting up and using private hubs in SageMaker JumpStart, empowering organizations to leverage AI models effectively.
Article Source
https://aws.amazon.com/blogs/machine-learning/manage-amazon-sagemaker-jumpstart-foundation-model-access-with-private-hubs/
