CMU Pokerbots Engine

CMU Pokerbots engine and skeleton bots in Python and other languages coming soon.

This is the reference implementation of the engine for playing vanilla Texas hold'em. Do not update this repo to implement the yearly game variant! Instead, create a new repo within this organization called cmudsc/poker-engine-yyyy.

Improvements which help the engine generalize to new variants, run faster, handle bot subprocesses more safely, etc. should be incorporated into this repo.

Your job

Edit python_skeleton/player.py

To run

As subprocesses

python -m venv .venv
source .venv/bin/activate
pip install -r ./engine/requirements.txt
pip install -r ./python_skeleton/requirements.txt
python run.py

With containers

(Requires docker installed):
Linux
Mac
Brew: brew cask install docker
Windows

./scripts/run_docker.sh

To visualize

Use the deployed app

https://cmu-poker-ai-2024.streamlit.app/

Run locally

Make sure you have the environment set up as in the previous section.

python -m venv .venv
source .venv/bin/activate
pip install -r ./engine/requirements.txt
pip install -r ./python_skeleton/requirements.txt

Then install streamlit, pillow

pip install streamlit==1.32.2
pip install pillow==10.2.0
streamlit run visualize.py

If you ever use any additional packages, be sure to regenerate requirements.txt like so:

pip freeze > python_skeleton/requirements.txt

Gym env

Run

pip install gymnasium

Refer to test_gym_env.py and engine/gym_env.py for more details.

With multiple bots

from engine.gym_env import PokerEnv 

env = PokerEnv(num_rounds=1000)
(obs1, obs2), info = env.reset()
bot1, bot2 = random_bot, random_bot

done = False
while not done:
    if obs1["is_my_turn"]:
        action = bot1(obs1)
    else:
        action = bot2(obs2)
    (obs1, obs2), (reward1, reward2), done, trunc, info = env.step(action)

With a single bot (enemy bot fixed)

env = PokerEnv(num_rounds=10, opp_bot=random_bot)
(obs1, obs2) = env.reset()
bot = random_bot

done = False
while not done:
    action = bot(obs)
    obs, reward, done, trunc, info = env.step(action)