This repository implements reinforcement learning (RL) approaches to play the Atari game Surround. The goal is to train agents that can effectively learn and play the game using various RL algorithms.
📦
├─ Atari.ipynb
├─ LunarLander.ipynb
├─ agents
│ ├─ agent.py
│ └─ reinforce.py
├─ models
│ ├─ cnns.py
│ └─ mlps.py
├─ setup.sh
└─ wrappers
└─ gym_wrappers.py
- agents/: Contains the implementation of RL algs, including REINFORCE.
- models/: Defines neural network architectures used by the agents.
- wrappers/: Custom Gym wrappers for preprocessing and reward shaping.
- Atari.ipynb: Jupyter Notebook for training and evaluating agents on Atari Surround.
- LunarLander.ipynb: Jupyter Notebook for training and evaluating agents on Lunar Lander.
- setup.sh: Script to set up the required environment and dependencies.
video and checkpoint directory will also be created automatically to record episodes and model weights.
Set up the environment:
./setup.sh