This repository implements the algorithms presented in the paper (published in ICML 2019)
- We advise the reader to use virtualenv so that installing dependencies is easy
- Install the gym package given with the code. To do this, activate your virtual env and go inside
gym/
and runpip install -e .
- Install tensorflow (version 1.12.0)
- Install the OpenAI baselines package given with the code. To do this, activate your virtual env and go inside
baselines/
and runpip install -e .
- Install other dependencies. To do this, activate your virtual env and go inside
FAIL/
and runpip install -r requirements.txt
- Install Mujoco-py:
pip install -U 'mujoco-py<1.50.2,>=1.50.1'
- We expect the reader to have a mujoco license to run the mujoco experiments in the paper
To download datasets, go to the directory FAIL/
- run
python download_datasets.py
To run the experiments, go to the directory FAIL/
and for the environment
- Fetchreach, run
./scripts/fetchreachexperiments.sh
- Swimmer, run
./scripts/swimmerdiscreteexperiments.sh
- Reacher, run
./scripts/reacherdiscreteexperiments.sh
The results are generated and stored in the FAIL/data
folder (for all the 10 random seeds in each experiment for every method)