Sending "Environment" while calling "create_training_job" [sagemaker] results in a 500

Boto3 version: 1.17.46 Python: 3.8.8 Botocore: 1.20.50 MacOS Catalina

Pretty much the subject has all the information I have. The actual output is:

  File "sagemaker.py", line 17, in create_training_job
    response = client.create_training_job(
  File "/Users/my_user/miniconda3/envs/jinn/lib/python3.8/site-packages/botocore/client.py", line 357, in _api_call
    return self._make_api_call(operation_name, kwargs)
  File "/Users/my_user/miniconda3/envs/jinn/lib/python3.8/site-packages/botocore/client.py", line 676, in _make_api_call
    raise error_class(parsed_response, operation_name)
botocore.exceptions.ClientError: An error occurred (500) when calling the CreateTrainingJob operation (reached max retries: 4): Internal Server Error

The test code includes:

        Environment={
            'the_environment': 'some_string'
        }

removing that, the job can be created.

This is a container based on a pytorch image if that makes any difference.

[stobrien89]

Hi @flaker,

Thanks for the update! I just double-checked and it looks like the fix has been merged. Closing for now, but let us know if anything else comes up.

[flaker]

Hi @stobrien89

Thanks! ok, then I will keep my workaround for the moment (image instead of algorithmarn) and come back to this later.