NeuroChain

A learning journey from perceptrons to GPT-style transformers

📌 Overview

Core2GPT is an educational machine learning framework that bridges classical algorithms and modern deep learning. It’s designed to help you implement, train, and understand models ranging from perceptrons and regression all the way to transformer-based GPT models — using real datasets like MNIST and a custom language ID task.

This project provides a hands-on path to see how simple ideas in machine learning grow into the foundations of today’s large language models.

🛠️ Features

• Classical Models: Perceptron, Regression, Digit Classification
• Deep Learning with PyTorch: Fully connected networks and training pipelines
• Transformers & GPT: Character-level GPT (chargpt.py) and GPT-like transformer (gpt_model.py)
• Datasets:
  • mnist.npz → handwritten digit recognition
  • lang_id.npz → language identification
• Testing Tools: autograder.py for automated checks

📂 Project Structure

│── backend.py # Utilities and data handling
│── models.py # Core ML models (Perceptron, Regression, etc.)
│── gpt_model.py # Transformer-based GPT model
│── chargpt.py # Character-level GPT implementation
│── autograder.py # Automated grading/testing
│── data/
│ ├── mnist.npz # Digit classification dataset
│ └── lang_id.npz # Language ID dataset
│── input.txt # Example input for GPT

🚀 Installation

Clone the repository and install dependencies:  
```bash
git clone https://github.com/<your-username>/Core2GPT.git
cd Core2GPT
pip install torch numpy

🏃 Running tests

python autograder.py