be67557d5e | ||
---|---|---|
imgs | ||
leaderboard | ||
public | ||
server | ||
src | ||
.gitignore | ||
README.md | ||
package.json | ||
requirements.txt | ||
yarn.lock |
README.md
NOTE: This project is under final tesing. The one in the repo only supports the visualization of some sampled data. The full version will be available soon!
Frontend
Frontend Setup
RLCard-Showdown has separated frontend and backend. The frontend part is built with React and the backend is with Django.
Prerequisite
To set up the frontend part, 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
Install Dependencies
The dependencies can be easily installed with the help of NPM:
git clone https://github.com/datamllab/rlcard-showdown.git
cd rlcard-showdown
npm install
Run Frontend in Development Mode
Run the following command under the project folder to start frontend in development mode.
npm start
The frontend would be started in port 3000 in localhost by default, you can view it by entering http://127.0.0.1:3000/ in the browser.
Config
You will see blank tables on the frontend if you haven't started the backend Django server, or the backend server URL is configured incorrectly.
The default configuration for backend server URL is http://127.0.0.1:8000/, you can also change it in /src/utils/config.js
.
Screenshots
Django Server
Server Setup
Install dependencies:
pip install -r requirements.txt
Migrate the databases:
cd server
python manage.py makemigrations
python manage.py migrate
Run server:
python manage.py runserver
The default URL is http://127.0.0.1:8000/
REST API
The definitions of the fields are as follows:
eval_num
: Integer. The number of evaluation times.name
: String. The name of the environment.agent0
: String. Model name.agent1
: String. Model name.win
: Boolean. True if model in the first seat wins.payoff
: Float. The payoff of the agent in the first seat.index
: Integer. The index of the game of the same environent and same agent. It is in the range [0, eval_num-1]
type | Resource | Parameters | Description |
---|---|---|---|
GET | tournament/launch | eval_num , name |
Launch tournment on the game. Each pair of models will play eval_num times. Results will be saved in database. |
GET | tournament/query_game | name , index , agent0 , agent1 , win , payoff , elements_every_page , page_index |
Query the games with the given parameters |
GET | tournament/query_payoff | name , agent0 , agent1 , payoff |
Query the payoffs with the given parameters |
GET | tournament/query_agent_payoff | name , elements_every_page , page_index , |
Query the payoffs of all the agents |
GET | tournament/replay | name , agent0 , agent1 , index |
Return the replay data |
POST | tournament/upload_agent | model (Python file), name , game , entry |
Upload a model file. name is model ID, entry is the class name of the model |
GET | tournament/delete_agent | name |
Delete the agent of the given name |
GET | tournament/list_uploaded__agents | game |
list all the uploaded agents |
GET | tournament/list_baseline__agents | game |
list all the baseline agents |
Example API
Registered Models
Some models have been pre-registered as baselines
Model | Game | Description |
---|---|---|
leduc-holdem-random | leduc-holdem | A random model |
leduc-holdem-cfr | leduc-holdem | Pre-trained CFR model on Leduc Holdem |
leduc-holdem-rule-v1 | leduc-holdem | A rule model that plays greedily |
doudizhu-random | doudizhu | A random model |
doudizhu-rule-v1 | doudizhu | Dou Dizhu rule model |
Example of uploading a new model
A example model file is prepared:
cd server/upload_test
Upload the model with curl
:
curl -F 'model=@example_model.py' -F "name=leduc-new" -F "entry=LeducHoldemRuleModelV2" -F "game=leduc-holdem" http://127.0.0.1:8000/tournament/upload_agent
Launch the tounament with:
curl 'http://127.0.0.1:8000/tournament/launch?eval_num=200&name=leduc-holdem'
We list the uploaded agent with
curl http://127.0.0.1:8000/tournament/list_uploaded_agents
We can delete the agent with
curl 'http://127.0.0.1:8000/tournament/delete_agent?name=leduc-new'