Detect AutoMLJob completion in EventBridge

0

I'm using the Python SDK to create an AutoML experiment with boto3's create_auto_ml_job() function.

I want to be informed when the job finishes. I can see SageMaker events in EventBridge that fire when the individual training jobs finish (SageMaker Training Job State Change/SageMaker Processing Job State Change), but not when the controlling (AutoML) job finishes. Is there an event that fires for that? Am I missing something (a better way than eventbridge)?

Hobo
已提問 3 個月前檢視次數 122 次
2 個答案
0

Hello.

I assume that you have set up S3 to store artifacts from AutoML jobs, but how about using an S3 event trigger to notify you when artifacts are stored in S3?
This setting allows you to be notified when an object is uploaded to S3.
https://docs.aws.amazon.com/AmazonS3/latest/userguide/EventNotifications.html

profile picture
專家
已回答 3 個月前
  • Thanks Riku. That might work, but it seems a strange way to detect job completion. Isn't there anything more direct (like an actual event)? What if the job fails before producing any output?

0

Hi,

Did you try to capture the event emitted by Sagemaker for Processing job state change. See https://docs.aws.amazon.com/sagemaker/latest/dg/automating-sagemaker-with-eventbridge.html#processing-job-state

It is also very probably the one emitted by AutoML jobs when they complete.

Best,

Didier

profile pictureAWS
專家
已回答 3 個月前
  • Hi Didier. Yes, I did. Unless I'm missing something, that event only seems to fire for the runs within an experiment (the processing and training jobs), not for the experiment itself (the controlling automl job). I'm pretty sure there are no sagemaker events that do what I need, but as I say, I might be missing something.

您尚未登入。 登入 去張貼答案。

一個好的回答可以清楚地回答問題並提供建設性的意見回饋,同時有助於提問者的專業成長。

回答問題指南