You can train and evaluate your model in an isolated Docker container. This keeps your RL environment separate from your local environment, which make reproducibility more robust.
The Dockerfile_CPU
builds a standalone local Docker image that both
hosts the Jupyter notebook server and also can train and evaluate
your submission from the command line.
Build the image and start the notebook server:
make
Train model from command line:
make train
Evaluate most recent submission
make evaluate
You need to have Docker installed in your workstation.