frederikschubert/hyperbolic-rl

Hyperbolic reinforcement learning minimal repository

Install dependencies via conda:

conda env create -f environment.yml
conda activate procgen_hyper

Replicating the results

Run experiments by executing main_hydra.py and override the appropriate arguments (see hydra for details), e.g. to run PPO with hyperbolic representations on starpilot with generalization configurations:

python main_hydra.py agent=onpolicy/hyperbolic/ppo env=gen/starpilot