This repo contains code for our paper State-only Imitation with Transition Dynamics Mismatch published at ICLR 2020.
The code heavily uses the RL machinery from this awesome repository with RL algorithms implemented in PyTorch. We also use some functionality from OpenAI baselines. The code was tested with the following packages:
- python 3.6.6
- pytorch 0.4.1
- gym 0.10.8
To run MuJoCo experiments, use the command below. Update the path to the directory with the expert demonstrations. The demonstrations used in our experiments can be downloaded from this Google drive link. Edit default_config.yaml to change the hyperparameters.
python main.py --env-name "Hopper-v2" --config-file "default_config.yaml" --experts-dir <add-path-here> --seed=$RANDOM