Experiments for the ICML 2021 submission "Self-Paced Context Evaluations for Contextual Reinforcement Learning".
To run the experiments, you need to install the dependencies:
pip install -r requirementsThe included notebooks for plotting and generating new instance also require jupyter to be installed:
pip install jupyterTo train a SPaCE agent on our provided instances, you can use the provided slurm scripts in the "runscripts" directory. Our own training results are included in this repository. To plot the data, you can use the provided jupyter notebook.