Amazon SageMaker Studio introduces GPU capacity reservation via Flexible Training Plans
Amazon SageMaker Studio now supports GPU capacity reservations through Flexible Training Plans, offering up to 65% cost savings and seamless resource management. Users can easily procure and manage instances without handling infrastructure.
Amazon has announced that its SageMaker Studio Integrated Development Environments (IDEs), including JupyterLab and Code Editor, now offer support for GPU capacity reservations through SageMaker Flexible Training Plans (FTP). This new feature is designed to provide users with reliable access to high-demand, high-performance computational resources while allowing them to stay within their budget constraints. By utilizing FTP, users can benefit from cost savings of up to 65% compared to using On-Demand instances for running machine learning workflows in JupyterLab or Code Editor.
The FTP service offers a straightforward, self-service procurement process. Users can begin by navigating to the SageMaker FTP console to select their desired instance type, reservation duration, and start date for their Studio IDE workload. After reviewing their order, they can complete the purchase and wait for the plan to be activated. Once the plan is active, users can create a Studio app from the SageMaker Studio UI and select their purchased plan from the Instance dropdown menu. SageMaker will automatically provision the instance, eliminating the need for users to manage infrastructure.
As the expiration of a plan approaches, the IDE will notify users in advance, allowing them ample time to save their work before the reservation concludes. For more detailed information on utilizing FTP capacity reservation capability with Studio IDEs, users can refer to the ‘Using Training Plans with Studio IDEs’ guide. Additionally, those interested in launching JupyterLab and Code Editor applications in SageMaker Studio can consult the ‘Studio Spaces’ documentation.