Meta Reinforcement Learning with Task Embedding and Shared Policy

Implementation of TESP in Pytorch.

Usage

You can use the main.py script in order to run meta reinforcement learning experiments with TESP. This script was tested with Python 3.7.

python main.py --env-name HalfCheetahDir-v1 --output-folder TESP-halfcheetah-dir --device cuda

References

This repository are based on the paper

Lan L, Li Z, Guan X, et al. Meta Reinforcement Learning with Task Embedding and Shared Policy[J]. arXiv preprint arXiv:1905.06527, 2019. [ArXiv]

If you want to cite this paper

@article{lan2019meta,
  title={Meta Reinforcement Learning with Task Embedding and Shared Policy},
  author={Lan, Lin and Li, Zhenguo and Guan, Xiaohong and Wang, Pinghui},
  journal={arXiv preprint arXiv:1905.06527},
  year={2019}
}