lygeng0427 / baku

Code for BAKU: An Efficient Transformer for Multi-Task Policy Learning

BAKU: An Efficient Transformer for Multi-Task Policy Learning

This is a repository containing the code for the paper BAKU: An Efficient Transformer for Multi-Task Policy Learning.

Installation Instructions

In order to install the required dependencies, please follow the instructions provide here.

Access to Datasets

We have added the instructions for running BAKU on the LIBERO benchmark here. For access to the datasets for Meta-World, DMControl, and the real world xArm Kitchen, please send an email to the [email protected].

Bibtex

If you find this work useful, please cite the paper using the following bibtex:

@article{haldar2024baku,
  title={BAKU: An Efficient Transformer for Multi-Task Policy Learning},
  author={Haldar, Siddhant and Peng, Zhuorang and Pinto, Lerrel},
  journal={arXiv preprint arXiv:2406.07539},
  year={2024}
}