Go to file
Songyi Huang c663537383 add functionality for modifying evaluation time; adjust download link style 2021-06-03 23:19:07 -07:00
docs Upload PvE 2021-05-28 11:25:02 -05:00
public update website icon; update upload dialog screenshot 2020-12-18 10:46:42 -08:00
pve_server Fix bugs in legal actions 2021-05-30 09:22:28 -05:00
server Add agent info 2021-06-03 11:53:25 -05:00
src add functionality for modifying evaluation time; adjust download link style 2021-06-03 23:19:07 -07:00
.eslintrc add config for eslint, prettier; implementing doudizhu demo... 2021-04-14 23:04:21 -07:00
.gitignore Add RLCard PvE 2021-05-28 15:43:44 -05:00
.prettierignore add config for eslint, prettier; implementing doudizhu demo... 2021-04-14 23:04:21 -07:00
.prettierrc add config for eslint, prettier; implementing doudizhu demo... 2021-04-14 23:04:21 -07:00
README.md Add RLCard PvE 2021-05-28 15:43:44 -05:00
package.json set up i18n plugin; add locale switch functionality for doudizhu demo 2021-05-25 23:15:16 -07:00
requirements.txt Update req 2021-05-30 11:06:55 -05:00
yarn.lock set up i18n plugin; add locale switch functionality for doudizhu demo 2021-05-25 23:15:16 -07:00

README.md

RLCard Showdown

This is the GUI support for the RLCard project and DouZero project. RLCard-Showdown provides evaluation and visualization tools to help understand the performance of the agents. It includes a replay module, where you can analyze the replays, and a PvE module, where you can play with the AI interactively. Currently, we only support Leduc Hold'em and Dou Dizhu. The frontend is developed with React. The backend is based on Django and Flask. Have fun!

Cite this work

Zha, Daochen, Kwei-Herng Lai, Songyi Huang, Yuanpu Cao, Keerthana Reddy, Juan Vargas, Alex Nguyen, Ruzhe Wei, Junyu Guo, and Xia Hu. "RLCard: A Platform for Reinforcement Learning in Card Games." In IJCAI. 2020.

@inproceedings{zha2020rlcard,
  title={RLCard: A Platform for Reinforcement Learning in Card Games},
  author={Zha, Daochen and Lai, Kwei-Herng and Huang, Songyi and Cao, Yuanpu and Reddy, Keerthana and Vargas, Juan and Nguyen, Alex and Wei, Ruzhe and Guo, Junyu and Hu, Xia},
  booktitle={IJCAI},
  year={2020}
}

Installation

RLCard-Showdown has separated frontend and backend. The frontend is built with React and the backend is based on Django and Flask.

Prerequisite

To set up the frontend, you should make sure you have Node.js and NPM installed. Normally you just need to manually install Node.js, and the NPM package would be automatically installed together with Node.js for you. Please refer to its official website for installation of Node.js.

You can run the following commands to verify the installation

node -v
npm -v

For backend, make sure that you have Python 3.6+ and pip installed.

Install Frontend and Backend

The frontend can be installed with the help of NPM:

git clone -b master --single-branch --depth=1 https://github.com/datamllab/rlcard-showdown.git
cd rlcard-showdown
npm install

The backend of leaderboard can be installed with

pip3 install -r requirements.txt
cd server
python3 manage.py migrate
cd ..

Run RLCard-Showdown

  1. Launch the backend of leaderboard with
cd server
python3 manage.py runserver
  1. Download the pre-trained models in Google Drive or 百度网盘 提取码: qh6s. Extract it in pve_server/pretrained.

In a new terminal, start the PvE server (i.e., human vs AI) of DouZero with

cd pve_server
python3 run_douzero.py

Alternatively, you can start the PvE server interfaced with RLCard:

cd pve_server
python3 run_dmc.py

They are conceptually the same with minor differences in state representation and training time of the pre-trained models (DouZero is fully trained with more than a month, while DMC in RLCard is only trained for hours).

  1. Run the following command in another new terminal under the project folder to start frontend:
npm start

You can view leaderboard at http://127.0.0.1:3000/ and PvE demo of Dou Dizhu at http://127.0.0.1:3000/pve/doudizhu-demo. The backend of leaderboard will run in http://127.0.0.1:8000/. The PvE backend will run in http://127.0.0.1:5000/.

Demos

leaderboards upload doudizhu-replay leduc-replay

Contact Us

If you have any questions or feedback, feel free to drop an email to Songyi Huang for the frontend or Daochen Zha for backend.

Acknowledgements

We would like to thank JJ World Network Technology Co., LTD for the generous support, Chieh-An Tsai for user interface design, and Lei Pan for the help in visualizations.