Step Functions for managing the deployment of SageMaker endpoints.
Create a new state machine with the contents of definition.json
.
When executing, enter the required variables according to the contents of execution_input.json
.
ECRPath
: URI of the container image used in the modelModelPath
: S3 path of model file (ex.models.tar.gz
)ModelName
: SageMaker model nameEndpointName
: SageMaker endpoint nameEndpointConfigNameCanary
: The name of the SageMaker endpoint config to use when canary releaseEndpointConfigName
: The name of the final SageMaker endpoint config to releaseExecutionRoleArn
: ARN of IAM role to use when running SageMaker endpointSecurityGroupIds
: List of security group IDs used in SageMaker endpointSubnets
: List of subnet IDs used in SageMaker endpointInstanceCount
: Number of instances to runInstanceType
: Instance type to runMinCapacity
: Minimum number of autoscale capacitiesMaxCapacity
: Maximum number of autoscale capacitiesTargetValue
: Autoscale threshold (number of requests per minute per instance)
Read the official docs for a detailed description of the variables.