aws-step-functions-data-science-sdk-python icon indicating copy to clipboard operation
aws-step-functions-data-science-sdk-python copied to clipboard

Published 20 hours ago • verified publisher icon aws

Problem with CheckpointConfig s3uri in Tensorflow estimator

Open roshsuta opened this issue 5 years ago • 1 comments
trafficstars

Values from estimator not passing to the step function definition (JSON). Had to manually set it in json by editing the step function from console to make the training job work.

"CheckpointConfig": { "S3Uri": "s3://bucket_name/location/for/checkpoint/" }, S3Uri is taken from checkpoint_s3_uri right? but in my case this is not working. CheckpointConfig dictionary is not generated.

Thank you for taking a look at it.

roshsuta avatar Sep 16 '20 11:09 roshsuta

Hello @roshsuta, Sorry for the late response! I was able to repro the issue by passing a Tensorflow Estimator with CheckpointConfig to TrainingStep constructor and creating a workflow with that step. The workflow definition is indeed missing the CheckpointConfig property.

Thank you for bringing this to our attention - tagging this as a bug.

ca-nguyen avatar Oct 01 '21 07:10 ca-nguyen