A basic Flappy Bird game made in Python with reinforcement learning
I took the assets from https://github.com/zhaolingzhi/FlapPyBird-master | Credits to him :D
You can install all dependencies with requirements.txt This file includes every library used for this program These dependencies are used:
os torch pygame openAI Gymnasium multiprocessing numpy random time collections
training a model is done inside the agent.py. Run agent.py and the terminal will ask you for the best process and model.
If no models exist it will start training a new model. u can use "count" variable to define how many models exists in each process.
The models are found inside the models folder. like this:
process_0_model_0
model 0 does NOT mean no models. It just counts from 0
Playing the game is done inside loadModel.py
This works the same way. Terminal will ask for best process and model, after input the model will start playing the game with the things it has learned from training
It's a simple model, so I'd be very grateful if you could help me to improve the agent