[アップデート] SageMakerの推論エンドポイントを0までスケールダウンできるようになりました #AWSreInvent

# 以下のノートブックより2024/12/4に引用
# https://github.com/aws-samples/sagemaker-genai-hosting-examples/blob/main/scale-to-zero-endpoint/llama3-8b-scale-to-zero-autoscaling.ipynb

variant_name = "AllTraffic"
instance_type = "ml.g5.12xlarge"
model_data_download_timeout_in_seconds = 3600
container_startup_health_check_timeout_in_seconds = 3600

min_instance_count = 0 # Minimum instance must be set to 0
max_instance_count = 3

sagemaker_client.create_endpoint_config(
    EndpointConfigName=endpoint_config_name,
    ExecutionRoleArn=role,
    ProductionVariants=[
        {
            "VariantName": variant_name,
            "InstanceType": instance_type,
            "InitialInstanceCount": 1,
            "ModelDataDownloadTimeoutInSeconds": model_data_download_timeout_in_seconds,
            "ContainerStartupHealthCheckTimeoutInSeconds": container_startup_health_check_timeout_in_seconds,
            "ManagedInstanceScaling": {
                "Status": "ENABLED",
                "MinInstanceCount": min_instance_count,
                "MaxInstanceCount": max_instance_count,
            },
            "RoutingConfig": {"RoutingStrategy": "LEAST_OUTSTANDING_REQUESTS"},
        }
    ],
)