Open-Ended-Reinforcement-Learning-with-Neural-Reward-Functions

Robert Meier* and Asier Mujika*

_{^{*Equal contribution}}

NeurIPS 2022, [arXiv]

Each folder contains the code for a set of experiments from Section 4 of the paper. You can run the following commands from the corresponding folder to reproduce the results.

2d Navigation

python3 main.py {experiment_name} where {experiment_name} is the folder that will be used to store the experiment results.

Robotic Environments

python3 run_experiment.py {gpu_id} where {gpu_id} is a single gpu id to tell CUDA which device to use.

Montezumas Revenge

python3 main.py