reference deployment

NVIDIA Clara Train SDK on AWS

Application-development framework for medical-imaging research and AI

Clara Train is NVIDIA’s domain-optimized application-development framework for medical-imaging researchers and artificial intelligence (AI) developers. Clara Train SDK, which you deploy in a highly available (HA) configuration on the AWS Cloud, includes an AI Assisted Annotation developer toolkit that can be integrated into existing medical viewers, accelerating the creation of AI-ready, annotated medical-imaging datasets. Clara Train also provides a TensorFlow-based training framework with domain-specific pretrained models that accelerate AI development with techniques like transfer learning, federated learning, and automated machine learning. Models trained with Clara Train are packaged as Medical Model Archives (MMARs), which provide a standardized format for training workflows and collaborations.

This Quick Start is for not only researchers and developers but also health IT infrastructure architects, administrators, and DevOps professionals who are planning to implement or extend their Clara Train SDK workloads to the AWS Cloud.

This deployment provides scalable access to NVIDIA V100 Tensor Core graphics processing units (GPUs) and the Amazon Elastic Compute Cloud (Amazon EC2) P3 instance type, with pay-as-you-go pricing. This deployment is based on Amazon Elastic Container Service (Amazon ECS) and Amazon EC2. Amazon Elastic File System (Amazon EFS) is used for storage shared between containers.

This Quick Start was created by NVIDIA in collaboration with AWS. NVIDIA is an APN Partner.

AWS Service Catalog administrators can add this architecture to their own catalog.  

  •  What you'll build
  • The Quick Start sets up the following:

    • A highly available architecture that spans two Availability Zones.*
    • A virtual private cloud (VPC) configured with public and private subnets, according to AWS best practices, to provide you with your own virtual network on AWS.*
    • In the public subnets:
      • Managed network address translation (NAT) gateways to allow outbound internet access for resources in the private subnets.*
      • A Linux bastion host in an Auto Scaling group to allow inbound Secure Shell (SSH) access to Amazon EC2 host instances in the private subnets.*
    • In the private subnets:
      • One instance of the NVIDIA Clara Train SDK container deployed with the Amazon EC2 launch type, on a GPU-enabled Amazon EC2 instance, in an Auto Scaling group.
      • One P3 Amazon EC2 instance.
    • Amazon EFS, a fully managed elastic network file system (NFS) to persist and share data across container instances.
    • Amazon ECS, a fully managed container orchestration service for running and managing Docker containers on a cluster.
    • An Application Load Balancer to route traffic to the NVIDIA Clara Train APIs and over HTTPS.

    * The template that deploys the Quick Start into an existing VPC skips the components marked by asterisks and prompts you for your existing VPC configuration.

  •  How to deploy
  • To deploy the NVIDIA Clara Train SDK, follow the instructions in the deployment guide. The deployment process, which takes about 30 minutes, includes these steps:

    1. If you don't already have an AWS account, sign up at, and sign in to your account.
    2. Launch the Quick Start. You can choose from the following two options:
    3. Test the deployment.

    Amazon may share user-deployment information with the AWS Partner that collaborated with AWS on this solution.  

  •  Cost and licenses
  • You are responsible for the cost of the AWS services used while running this Quick Start reference deployment. There are no additional costs for using the NVIDIA Clara Train SDK or for using the Quick Start.

    The AWS CloudFormation template for this Quick Start includes configuration parameters that you can customize. Some of these settings, such as instance type, affect the cost of deployment. For cost estimates, see the pricing pages for each AWS service you will use. Prices are subject to change.

    Tip: After you deploy the Quick Start, we recommend that you enable the AWS Cost and Usage Report. This report delivers billing metrics to an Amazon Simple Storage Service (Amazon S3) bucket in your account. It provides cost estimates based on usage throughout each month and finalizes the data at the end of the month. For more information about the report, see the AWS documentation.

    By using this Quick Start, and by using the Clara Train SDK container and download models, you accept the terms and conditions of the included licenses. Licenses are available with the NVIDIA Clara Train SDK documentation and along with the model application .zip files.