- 最新
- 投票最多
- 评论最多
Hi,
Thank you for using AWS Sagemaker.
I understand that while doing the data quality check between baseline and current run data, it’s showing data type check in Constraints when Amazon SageMaker Model Monitor compared this run against the baseline and detected these constraint violations.
"Expected data type: String, Expected match: 100.0%. Observed: Only 73.44% of data is String."
What you are experiencing is known as data drift which might be occurring due to the fact that the serving data has shifted from the original expected data distribution that was used for training.
I would like to mention here that if there are any kind of violations compared to the baseline, they get generated in the violations report and you can use the following code to list the violations.
violations = my_default_monitor.latest_monitoring_constraint_violations() pd.set_option('display.max_colwidth', -1) constraints_df = pd.io.json.json_normalize(violations.body_dict["violations"])
constraints_df.head(10)
Here is an open-source notebook which demonstrates this as well:
See 'Violations report' section here - https://github.com/aws-samples/amazon-sagemaker-immersion-day/blob/master/SageMaker-ModelMonitoring.ipynb
This information will help in identifying which columns you can check for troubleshooting the violations at your end.
Hope this information helps.
To further understand the issue more in depth as I have limited visibility on your setup, I'd recommend you to reach to AWS Support by creating a support case[2] so that the engineer can investigate further and help you overcome the issue.
References: [1] Violations Report - https://docs.aws.amazon.com/sagemaker/latest/dg/model-monitor-interpreting-results.html#w570aac27c18c20c31c11
[2] Open a support case with AWS using the link: https://console.aws.amazon.com/support/home?#/case/create
相关内容
- AWS 官方已更新 1 年前
- AWS 官方已更新 1 年前