HGAIL

• 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 hgail because the original purpose was implementing the hierarchical GAIL variant, though it has since then primarily been used for the GAIL implementation

Install

• 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 tests directory to check that everything is correctly installed

cd tests
python runtests.py

Examples

• for an artificial example on cartpole, see scripts/README.md
• for an actual application, see this repo on learning human driver models with gail