AI Platformer Prodigy

A reinforcement learning framework for training AI agents to play Super Mario Bros

---

Overview

AI Platformer Prodigy is a sophisticated AI training framework built for Super Mario Bros, leveraging reinforcement learning techniques to create, train, and evaluate intelligent agents within a dynamic game environment. The project uses state-of-the-art algorithms like PPO and DQN, combined with preprocessing pipelines to significantly optimize training efficiency.

---

Features

• Advanced RL Algorithms — Implements PPO and DQN for optimal agent performance
• Realistic Simulation — Full game environment via gym_super_mario_bros for comprehensive training
• Optimized Preprocessing — Frame stacking and grayscaling to reduce training duration
• Performance Monitoring — Integrated TensorBoard logging for real-time training visualization
• Modular Notebooks — Separate notebooks for training from scratch and incremental training

---

Tech Stack

Layer	Tools
Simulation Environment	OpenAI Gym, gym_super_mario_bros, nes_py
Reinforcement Learning	stable-baselines3 (PPO, DQN)
Deep Learning	PyTorch, torchvision, torchaudio
Visualization	Matplotlib, Pygame, TensorBoard
Data Handling	NumPy, Pandas

---

Project Structure

AI-Platformer-Prodigy/
├── images/
│   ├── sc1.png                         # Training screenshot 1
│   ├── sc2.png                         # Training screenshot 2
│   └── sc3.gif                         # Demo gameplay GIF
├── MarioTrainingFromScratch.ipynb      # Train an agent from a clean slate
├── MarioAIIncrementalTraining.ipynb    # Continue training from a checkpoint
├── logs/
│   └── PPO/                            # TensorBoard training logs
├── .python-version
├── .gitignore
├── LICENSE
├── requirements.txt
└── Readme.md

---

Getting Started

Prerequisites

• Python 3.8+
• CUDA-compatible GPU (recommended for training)
• Jupyter Notebook

Installation

1. Clone the repository

   git clone https://github.com/silentwraith03/ai-platformer-prodigy.git
   cd AI-Platformer-Prodigy

2. Install core dependencies

   pip install -r requirements.txt

3. Install the Mario environment

   pip install gym_super_mario_bros==7.3.0 nes_py

4. Install PyTorch with CUDA support

   pip install torch==1.10.1+cu113 torchvision==0.11.2+cu113 torchaudio==0.10.1+cu113 \
     -f https://download.pytorch.org/whl/cu113/torch_stable.html

5. Install Stable Baselines3

   pip install stable-baselines3[extra]

6. Install Jupyter (skip if already installed)

   pip install jupyter

---

Usage

Running the Notebooks

Launch Jupyter and open either notebook based on your use case:

jupyter notebook

Notebook	Purpose
MarioTrainingFromScratch.ipynb	Train an agent from a clean slate
MarioAIIncrementalTraining.ipynb	Continue training from a saved checkpoint

Monitoring Training

Track rewards, loss, and other metrics in real time using TensorBoard:

cd logs/PPO
tensorboard --logdir=.

Then open http://localhost:6006 in your browser.

---

Screenshots

---

Demo

---

License

Distributed under the MIT License. See LICENSE for details.