Amazon Nova Forge
Amazon Nova Forge is a first-of-its-kind service that offers organizations the easiest and most cost-effective way to build their own frontier models using Nova.
Amazon Nova Forge introduces the concept of “open training" models, which give organizations access to a variety of early model checkpoints and the ability to blend proprietary data with Amazon-curated data sets at every stage of model training. This allows the models to maximize learning from proprietary data while minimizing risk of forgetting foundational skills like reasoning.
Nova Forge provides the following key capabilities:
-
Access checkpoints across all phases of model development, and leverage new Nova models before they are widely available
-
Blend your proprietary data with Amazon Nova-curated training data
-
Perform reinforcement learning with reward functions in your environment
-
Use push-button recipes that are optimized to build with Nova through visual workflows or a command line interface
-
Use the built-in responsible AI toolkit to implement custom safety guardrails
Prerequisites
Subscribe to Nova Forge
To request access to the Nova Forge service, first step is to add the following key-value pair as tags to your role: forge-subscription. After you've added these tags to your role, please go to SageMaker AI Console > Model training and customization and click on Nova Forge. On this page, you'll find details about the service, pricing information and the capabilities. You can request subscription and then manage your subscription from this page.
-
The role should have permission to call api
ListAttachedRolePolicy, and the response should include eitherAdministratorAccessorAmazonSageMakerFullAccesspolicy. -
The sign-in role should have permission to call api
ListRoleTags, and the response tags should includetag.key=forge-subscription.
Other prerequisites
Also ensure the following prerequisites are complete:
-
Additional steps for SageMaker HyperPod users: Add Restricted Instance Group (RIG) to your HyperPod cluster (to complete follow steps here)
Initial HyperPod setup
Follow the steps in the AWS documentation to connect
the HyperPod CLI to your cluster. Ensure you use nova-lite-2.0-release
branch instead of release_v2 for testing. This topic covers:
-
Setting up AWS IAM permissions and policies
-
Configuring the HyperPod CLI
-
Creating your cluster with the necessary instance groups
-
Connecting to your cluster
Verification: After completing the setup, confirm you
can successfully run hyperpod connect-cluster and access your
cluster.
Setting up permissions for Nova Forge
To enable access to Nova Forge models and data mixes, grant your HyperPod cluster's execution role permissions to access Nova Forge S3 buckets.
-
Locate your execution role.
Your execution role follows the naming convention:
arn:aws:iam::<AWS_ACCOUNT_ID>:role/<your-cluster-name>-9610a1d3ExecRoleTo find your execution role:
-
Navigate to IAM → Roles in the AWS Management Console
-
Search for your cluster name (e.g.,
9610a1d3Exec) -
Locate the role matching the pattern above
-
Alternately, if the cluster is already created, the execution role can be viewed in the AWS Management Console under SageMaker AI > Cluster management > [Your Cluster Name] > Instances > Restricted Instance Groups > Execution Role ARN
-
-
Apply Changes to Existing RIG (if applicable)
Important
If you granted this permission after creating your RIG, you must update the cluster for changes to take effect. See the Troubleshooting section below for instructions on updating your cluster.
Note
Without this permission, your RIG will not have access to Nova datasets or model checkpoints required for Forge features.