Understanding when Dynamics-Invariant Data Augmentations Benefit Model-Free Reinforcement Learning Updates (ICLR 2024)

This is the source code accompanying the paper Understanding when Dynamics-Invariant Data Augmentations Benefit Model-Free Reinforcement Learning Updates (ICLR 2024) by Nicholas E. Corrado and Josiah P. Hanna.

Install

git clone 
cd RL-augment
pip install -e .
conda install -c conda-forge stable-baselines3
conda install pyyaml

Install MuJoCo for Apple M1 Mac (see openai/mujoco-py#662)

pip install mujoco_py
export MUJOCO_PY_MUJOCO_PATH=~/.mujoco/mujoco210/


mkdir -p $HOME/.mujoco/mujoco210         # Remove existing installation if any
ln -sf /Applications/MuJoCo.app/Contents/Frameworks/MuJoCo.framework/Versions/Current/Headers/ $HOME/.mujoco/mujoco210/include
mkdir -p $HOME/.mujoco/mujoco210/bin
ln -sf /Applications/MuJoCo.app/Contents/Frameworks/MuJoCo.framework/Versions/Current/libmujoco.2.*.dylib $HOME/.mujoco/mujoco210/bin/libmujoco210.dylib
ln -sf /Applications/MuJoCo.app/Contents/Frameworks/MuJoCo.framework/Versions/Current/libmujoco.2.*.dylib /usr/local/lib/

# For M1 (arm64) mac users:
# The released binary doesn't ship glfw3, so need to install on your own
brew install glfw
ln -sf /opt/homebrew/lib/libglfw.3.dylib $HOME/.mujoco/mujoco210/bin

# Please make sure /opt/homebrew/bin/gcc-11  exists: install gcc if you haven't already
# brew install gcc
export CC=/opt/homebrew/bin/gcc-11         # see https://github.com/openai/mujoco-py/issues/605

pip install mujoco-py && python -c 'import mujoco_py'

Citation

If you found any part of our work useful, please consider citing our paper:

@inproceedings{corrado_understanding_2024,
  title = {Understanding when Dynamics-Invariant Data Augmentations Benefit Model-Free Reinforcement Learning Updates},
  url = {https://arxiv.org/abs/2310.17786},,
  booktitle = {Proceedings of the International Conference on Learning Representations ({ICLR})},
  author = {Nicholas E. Corrado, Josiah P. Hanna},
  month = May,
  year = {2024},
}