How to Visualize a custom metrics for a ProcessingStep on Sagemaker Pipeline UI

0

When selecting a step from a running Sagemaker pipeline, the right-side panel has an "Output" tab. We know that if the step is created from an Estimator, we can display all metrics from specifying metric_definitions parameter. However there is no equivalent parameter for Processor or other types of steps. How can we display metrics for a Sagemaker ProcessingStep then? Enter image description here

已提问 1 年前287 查看次数
1 回答
0

Hi there,

Thanks for reaching out and for the question.

Unlike the Estimator subclass which takes 'metric_definitions ' as a parameter, the ProcessingStep does not take on this parameter. An alternative is to make use of the the property_files parameter within the ProcessongStep (see link [1.]) to store information from the output of a processing step which can later be processed using the 'JsonGet' (see link [2.]). At best this will allow you to persist the metrics you'd like to capture and you can fetch the json file from the s3 output location, however this will not be published to the UI like your training jobs. Link [3.] provides an example of how this propertyfile can be defined and passed in the ProcessingStep.

[1.] https://sagemaker.readthedocs.io/en/stable/workflows/pipelines/sagemaker.workflow.pipelines.html#sagemaker.workflow.steps.ProcessingStep [2.] https://docs.aws.amazon.com/sagemaker/latest/dg/build-and-manage-propertyfile.html [3.] https://github.com/aws/amazon-sagemaker-examples/blob/main/sagemaker-pipelines/tabular/customizing_build_train_deploy_project/modelbuild/pipelines/customer_churn/pipeline.py#L183

AWS
已回答 1 年前

您未登录。 登录 发布回答。

一个好的回答可以清楚地解答问题和提供建设性反馈,并能促进提问者的职业发展。

回答问题的准则