SageMaker's FensorFlow Training Job - how to change script archive path?

Question

Maybe someone has encountered such a problem: when creating a Training Job in SageMaker by the sagemaker.tensorflow.TensorFlow class, I need to set a parameter that is responsible for saving the training script code in S3, for its path to be precise: sagemaker_submit_directory or maybe module_dir - I am not sure. By default it looks like this:

s3: //bucket/job-name/source/sourcedir.tar.gz. 

I need to put something between bucket name and job-name. How can I do it? I tried to add such parameter to hyperparameters, but then my Training Job was crushing (unfortunately it was impossible to read the whole error message because the stack trace was larger than a limit and I could not see it fully)?

Any idea?

Answer 1

OK, it seams I was not patient enough in searching the documentation. Here is the answer, parameter "code_location":

https://sagemaker.readthedocs.io/en/stable/api/training/estimators.html?highlight=code_location#sagemaker.estimator.Framework