Hongyi Chen*, Yilun Du*, Yiye Chen*, Joshua B. Tenenbaum, Patricio A. Vela

*equal contribution

A link to our paper can be found on OpenReview. A link to our project website can be found on here.

Please cite our paper as:

@inproceedings{chenplanning,
  title={Planning with Sequence Models through Iterative Energy Minimization},
  author={Chen, Hongyi and Du, Yilun and Chen, Yiye and Tenenbaum, Joshua B and Vela, Patricio A},
  booktitle={International Conference on Learning Representations},
  year={2023}
}

In this repo, we provide training and evaluation code for Multistep Energy-Minimization Planner (LEAP) model in Babyai environment.

Babyai simulator

cd babyai
pip3 install --editable .

LEAP training and testing in our environment (should be good in other close version)

python=3.7.12
pytorch=1.11.0+cu113

To reproduce the babyai result

bash test_run.sh

In case of any confusion, please email hongyic@andrew.cmu.edu.

The codebase is derived from decision transformer repo.