/CCWM_code

Python code base for the paper "Learning State Representations via Retracing in Reinforcement Learning" accepted in ICLR 2022, by Changmin Yu, Dong Li, Jianye Hao, Jun Wang, and Neil Burgess.

Primary LanguagePython

Learning State Representations via Retracing in Reinforcement Learning

Python code base for the Cycle-Consistency World Model agent in our paper.

The code is developed based on the TF2 code of Dreamer v1.

If you find the code and our paper useful, please cite us in the following format:

@inproceedings{yu2022learning,
    title={Learning State Representations via Retracing in Reinforcement Learning},
    author={Changmin Yu and Dong Li and Jianye Hao and Jun Wang and Neil Burgess},
    booktitle={International Conference on Learning Representations},
    year={2022},
    url={https://openreview.net/forum?id=CLpxpXqqBV}
}

Installation and Dependencies

The code is based on python 3.7, the necessary dependencies can be installed by running:

conda env create -f environment.yml

Training

CUDA_VIDIBLE_DEVICES=0 python ccwm.py --task dmc_walker_walk

Contact

Please feel free to use/extend this code for your own research. Please send any enquiry to changmin.yu.19@ucl.ac.uk or simply open an issue.