This repository is based on the research lightning framework.
First, create the environment using the .yaml
files.
For details instructions on launching jobs, see the research lightning repo. Below are basic instructions for training. To run a job, simply run the following command:
python scripts/train.py --config configs/<path>/<to>/<config> --path <path>/<to>/<save>/<logs>
The different parameter configurations for running the experiments can be found in the .json
files. Switching between XQL and standard Q-Learning is done by simply changing alg_kwargs.loss
to gumbel_resacle
from mse
.