Skip to content
A framework for reproducible reinforcement learning research
Branch: master
Clone or download
Type Name Latest commit message Commit time
Failed to load latest commit information.
docker Copy local files at the end of docker-compose Mar 6, 2019
docs Move NumPy-based modules to `` (#604) Mar 26, 2019
examples Update links in jupyter notebooks (#617) Apr 24, 2019
garage Add advantage normalization for recurrent policies (#626) Apr 18, 2019
scripts Benchmark readme (#619) Apr 17, 2019
tests Benchmark readme (#619) Apr 17, 2019
.editorconfig Fix local package names in flake8 config (#77) Jun 14, 2018
.pre-commit-config.yaml Add advantage normalization for recurrent policies (#626) Apr 18, 2019
.travis.yml Build docs inside the CI (#558) Mar 5, 2019 Update CHANGELOG for the 2019.02.0 release (#550) Mar 1, 2019
CODEOWNERS Update CODEOWNERS Nov 2, 2018 Update (#629) Apr 22, 2019
LICENSE Update LICENSE Mar 2, 2019
Makefile Replace config_personal with environment variables (#585) Mar 19, 2019 Remove theano (#477) Jan 29, 2019
environment.yml Move viskit to rlworkgroup/viskit Feb 26, 2019
setup.cfg Refactor *MLPPolicy with model (#574) Mar 26, 2019 New Logger (#464) Mar 23, 2019

Docs Build Status License


garage is a framework for developing and evaluating reinforcement learning algorithms. It includes a wide range of continuous control tasks plus implementations of algorithms.

garage is fully compatible with OpenAI Gym. All garage environments implement gym.Env, so all garage components can also be used with any environment implementing gym.Env.

garage only officially supports Python 3.5+.

garage comes with support for running reinforcement learning experiments on an EC2 cluster, and tools for visualizing the results. See the documentation for details.

garage supports TensorFlow for neural network frameworks. TensorFlow modules can be found under garage/tf.


Documentation is available online at

Citing garage

If you use garage for academic research, you are highly encouraged to cite the following paper on the original rllab implementation:


garage is based on a predecessor project called rllab. The garage project is grateful for the contributions of the original rllab authors, and hopes to continue advancing the state of reproducibility in RL research in the same spirit.

rllab was originally developed by Rocky Duan (UC Berkeley/OpenAI), Peter Chen (UC Berkeley), Rein Houthooft (UC Berkeley/OpenAI), John Schulman (UC Berkeley/OpenAI), and Pieter Abbeel (UC Berkeley/OpenAI).

You can’t perform that action at this time.