rl-ws22-ex01

RL Training Demo with Popular Libraries.

We train and optimize a BipedalWalker.

For more info see the notebook.

Setup

Install the open-source-distribution anaconda.
Create a new environment and activate it.
Inside this environment, install pip with conda install pip.
Install requirements with pip install -r requirements.txt.
Open jupyter notebook with jupyter-lab in this directory (activate your conda env first).

For training, activate your conda env and run the following in this dir:

python train.py

Check out the folder configs for possible parameters. You can read here how to set parameters via the commandline.

tensorboard --logdir .

See the notebook.