A deep reinforcement learning Snake game built with PyTorch and Pygame. The agent learns to play Snake using Q-learning and neural networks. Built by Krystian Filipek
- Classic Snake Gameplay with AI training
- Deep Q-Network with experience replay
- Live training graphs showing progress
- Manual play mode for testing
git clone https://github.com/kfilipekk/Pytorch_snake.git
cd Pytorch_snake
pip install .
python main.py- State: 11D vector (danger, direction, food location)
- Actions: [straight, right, left]
- Rewards: +10 (food), -10 (collision)
- Model: 2-layer neural network with experience replay
Developed by kfilipekk