Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

TeachMyAgent: a Benchmark for Automatic Curriculum Learning in Deep RL

TeachMyAgent is a testbed platform for Automatic Curriculum Learning methods. We leverage Box2D procedurally generated environments to assess the performance of teacher algorithms in continuous task spaces. Our repository provides:

  • Two parametric Box2D environments: Stumps Tracks and Parkour
  • Multiple embodiments with different locomotion skills (e.g. bipedal walker, spider, climbing chimpanzee, fish)
  • Two Deep RL students: SAC and PPO
  • Several ACL algorithms: ADR, ALP-GMM, Covar-GMM, SPDL, GoalGAN, Setter-Solver, RIAC
  • Two benchmark experiments using elements above: Skill-specific comparison and global performance assessment
  • Three notebooks for systematic analysis of results using statistical tests along with visualization tools (plots, videos...) allowing to reproduce our figures

See our documentation for an exhaustive list.


Using this, we performed a benchmark of the previously mentioned ACL methods which can be seen in our paper. We also provide additional visualization on our website.


1- Get the repository

git clone
cd TeachMyAgent/

2- Install it, using Conda for example (use Python >= 3.6)

conda create --name teachMyAgent python=3.6
conda activate teachMyAgent
pip install -e .

Note: For Windows users, add -f to the pip install -e . command.

Import baseline results from our paper

In order to benchmark methods against the ones we evaluated in our paper you must download our results:

  1. Go to the notebooks folder
  2. Make the script executable: chmod +x
  3. Download results: ./

WARNING: This will download a zip weighting approximayely 4.5GB. Then, our script will extract the zip file in TeachMyAgent/data. Once extracted, results will weight approximately 15GB.


See our documentation for details on how to use our platform to benchmark ACL methods.


See for details.


If you use TeachMyAgent in your work, please cite the accompanying paper:

  author    = {Cl{\'{e}}ment Romac and
               R{\'{e}}my Portelas and
               Katja Hofmann and
               Pierre{-}Yves Oudeyer},
  title     = {TeachMyAgent: a Benchmark for Automatic Curriculum Learning in Deep
  booktitle = {Proceedings of the 38th International Conference on Machine Learning,
               {ICML} 2021, 18-24 July 2021, Virtual Event},
  series    = {Proceedings of Machine Learning Research},
  volume    = {139},
  pages     = {9052--9063},
  publisher = {{PMLR}},
  year      = {2021}


TeachMyAgent is a testbed platform for Automatic Curriculum Learning methods in Deep RL.







No releases published


No packages published