- Python >= 3.9
- (Optional) CUDA 11.7
Install PyTorch as described here. neroRL is tested with PyTorch version 2.0.1
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117
pip3 install neroRL
pip3 install neroRL[ml-agents]
pip3 install neroRL[obstacle-tower]
pip3 install neroRL[procgen]
pip3 install neroRL[random-maze]
git clone https://github.com/MarcoMeter/neroRL
cd neroRL
pip3 install -e .
Unity environments, like Obstacle Tower, are provided as executables. Obstacle Tower can be build form source using the Unity engine or you can download builds from these links:
For checksums on these files, see this.