Official implementation for OPIRL: Sample Efficient Off-Policy Inverse Reinforcement Learning via Distribution Matching.
Accepted to IEEE International Conference on Robotics and Automation (ICRA) 2022.
Run the following command to install all Python dependencies:
$ pip install -e .
$ pip install -r requirements.txt
Other dependencies:
-
Python 3.8+
-
TensorFlow 2.4+
-
CUDA=11.0
-
cuDNN=8.0
-
Experts/reward functions are provided on Google Drive
First, unzip the expert/reward files from Google Drive.
Then, to simply run experiments on MuJoCo tasks, run the bash scripts in /scripts
directory.
E.g.
$ sh ./scripts/run_halfcheetah.sh