s-nandi / bball-sim

A testbed for exploring reinforcement learning & basketball strategy

bball-sim

Dependencies

• Python 3
  • Pygame
  • Pymunk
  • Pytorch
  • Stable Baselines 3
  • Tqdm
  • Pylint
  • Pytest
  • Black

Setup

1. Install pytorch
2. Run

   python -m pip install pygame pymunk stable-baselines3 pylint mypy pytest black tqdm
   

Linting

Run python -m pylint src experiment tests && python -m mypy src tests experiment

Formatting

Run python -m black src tests experiment

Testing

Run python -m pytest

ML Agent

Training

Run

python -m src.neural.basic_offense learn 10000000000 10000000 output/basic_offense

or

python -m src.neural.basic_offense learn 10000000000 10000000 output/basic_offense --epoch ${EPOCH}

to continue training from a specific epoch

Run

tensorboard --logdir output/basic_offense

to visualize training progress

Testing

Run

python -m src.neural.basic_offense load 20 output/basic_offense

to visualize 20 runs of the latest model or

python -m src.neural.basic_offense load 20 output/basic_offense --epoch ${EPOCH}

to visualize a specific epoch