- This repo contains implementations of the following, built on top of rllab:
- generative adversarial imitation learning (GAIL)
- InfoGAIL
- Hierarchical GAIL
- The main goal of these implementations is for them to be easy to use with rllab environments
- The repo is called
hgailbecause the original purpose was implementing the hierarchical GAIL variant, though it has since then primarily been used for the GAIL implementation
- install rllab
- clone this repo, and run setup.py in develop mode
git clone https://github.com/sisl/hgail.git
cd hgail
python setup.py develop- after installing, run the tests in the
testsdirectory to check that everything is correctly installed
cd tests
python runtests.py- for an artificial example on cartpole, see
scripts/README.md - for an actual application, see this repo on learning human driver models with gail