Removing the dependency on Markov library in Sagemaker

aws-deepracer-community / deepracer-sagemaker-container

A Deepracer enabled Sagemaker Container for local training

Apache License 2.0

13 stars 9 forks source link

Removing the dependency on Markov library in Sagemaker #1

Closed larsll closed 4 years ago

larsll commented 4 years ago

This is possible as the coach script can upload the entire code-base 'on the fly'

estimator = RLEstimator(entry_point="training_worker.py",
                        source_dir='markov',
                        dependencies=["common/sagemaker_rl","markov"],

It would be a breaking change (the dependencies need to be there) but would remove dependency from sagemaker to robomaker.