Marllib 버전비교

'''

• marllib: for_pettingzoo_1.23.0
• marllib_1.12: for_pettingzoo_1.12(original)
• marllib_bean: for_pettingzoo_bean(코드 업데이트 진행중) '''

pettingzoo 버전비교

'''

• pettingzoo: afrl용도의 pettingzoo
• pettingzoo_pet: 기존 pettingzoo_1.23버전
• pettingzoo_1.12: 알아서 설치(굳이 할 필요 없음) '''

Details_1.12(다운 하고 싶다면)

conda create -n marllib python=3.8
conda activate marllib
git clone https://github.com/MaDoKaLiF/AFRL_MARLlib.git
pip install --upgrade pip
pip install -r requirements.txt

pip install gym>=0.22.0
cd MARLLIB
python marllib/patch/add_patch.py -y

update 1.23(다운 하고 싶다면)

pip install pettingzoo==1.23.1
pip install supersuit==3.9.0
pip install pygame==2.3.0

Expected Errors

• gym=0.20.0 설치 과정에서 에러 발생
  setuptools=65.5.0으로 downgrade

  pip install setuptools==65.5.0 "wheel<0.40.0"
  

• (vessl 사용 시) import torch 과정에서 AttributeError 발생
  global environment의 torch를 삭제하고 진행

MARLlib documentation (아래의 MARLlib 아키텍처를 바탕으로 수정한 것임.)

https://marllib.readthedocs.io/en/latest/index.html

@article{hu2022marllib, author = {Siyi Hu and Yifan Zhong and Minquan Gao and Weixun Wang and Hao Dong and Xiaodan Liang and Zhihui Li and Xiaojun Chang and Yaodong Yang}, title = {MARLlib: A Scalable and Efficient Multi-agent Reinforcement Learning Library}, journal = {Journal of Machine Learning Research}, year = {2023}, }