Amazon SageMaker Model Monitor prebuilt container
SageMaker provides a built-in image called sagemaker-model-monitor-analyzer
that
provides you with a range of model monitoring capabilities, including constraint suggestion,
statistics generation, constraint validation against a baseline, and emitting Amazon CloudWatch
metrics. This image is based on Spark version 3.3.0 and is built with Deequ
Note
You can not pull the built-in sagemaker-model-monitor-analyzer
image
directly. You can use the sagemaker-model-monitor-analyzer
image when you
submit a baseline processing or monitoring job using one of the AWS SDKs.
Use the SageMaker Python SDK (see image_uris.retrieve
in the SageMaker Python
SDK reference guide
<ACCOUNT_ID>
.dkr.ecr.<REGION_NAME>
.amazonaws.com/sagemaker-model-monitor-analyzer
For example:
159807026194.dkr.ecr.us-west-2.amazonaws.com/sagemaker-model-monitor-analyzer
If you are in an AWS region in China, the prebuilt images for SageMaker Model Monitor can be accessed as follows:
<ACCOUNT_ID>
.dkr.ecr.<REGION_NAME>
.amazonaws.com.rproxy.goskope.com.cn/sagemaker-model-monitor-analyzer
For account IDs and AWS Region names, see Docker Registry Paths and Example Code.
To write your own analysis container, see the container contract described in Custom monitoring schedules.