Deploy Model to Endpoint
A 4-step recipe with Amazon SageMaker. We can help you get access.
Recipe · 4 steps
Deploy Model to Endpoint
with Amazon SageMaker · Artificial Intelligence Management
Register a model and deploy it to a hosted real-time endpoint.
- Register a model from an inference container image and model artifacts so it can be referenced by an endpoint configuration.
- Define an endpoint configuration that places the newly created model on a single production variant with the requested instance type and count.
- Create a hosted endpoint from the endpoint configuration so SageMaker begins provisioning resources and deploying the model.
- Describe the endpoint to read its status. Repeat this step while the endpoint remains in the Creating state and continue once it reaches a terminal state such as InService or Failed.
What you'll need to have ready
- modelName: A unique name for the model.
- image: The registry path of the Docker image that contains the inference code.
- modelDataUrl: The S3 path where the model artifacts are stored.
- executionRoleArn: The ARN of the IAM role SageMaker can assume to access model artifacts.
- endpointConfigName: A unique name for the endpoint configuration.
- variantName: The name of the production variant.
- initialInstanceCount: The initial number of instances to launch for the variant.
- instanceType: The ML compute instance type to deploy for the variant.
- endpointName: A unique name for the endpoint.
For your developer
Arazzo workflow amazon-sagemaker:deploy-model-to-endpoint, 4 operations: CreateModel → CreateEndpointConfig → CreateEndpoint → DescribeEndpoint. Specs on APIs.io.