Flappy-Bird-Double-DQN-Pytorch

Train an agent to play flappy bird game using double DQN model and implement it with pytorch.

Installations (suggest using virtualenv)

• Pygame

$ pip install pygame

• ple

$ git clone https://github.com/ntasfi/PyGame-Learning-Environment.git
$ cd PyGame-Learning-Environment/
$ pip install -e .

• gym

$ pip install gym

• gym-ple

$ pip install gym-ple

How to run

• Train

# If cuda is available, add --cuda Y.
# Add --ckpt [ckpt_file] to train from this checkpoint. 
$ python main.py --mode train 

• Evaluation

# If cuda is available, add --cuda Y.
$ python main.py --mode eval --ckpt [checkpoint.pth.tar]