🧠 NanoGPT Deployment & Smoke Test

📖 Introduction

This repository is a lightweight deployment and verification implementation based on Andrej Karpathy's nanoGPT.

The primary goal of this project is to demonstrate a complete end-to-end LLM lifecycle—from data generation to model training and inference—verified within a minimal compute environment (e.g., Google Colab T4 or local CPU/GPU).

We utilize a synthetic "Tick Tock" pattern dataset (or arithmetic logic) to conduct a "Smoke Test," ensuring that the model architecture, optimizer, and training loop are functioning correctly before scaling up to larger datasets like OpenWebText.

🚀 Features

• Automated Data Pipeline: Scripts to generate synthetic training data instantly.
• Minimalist Configuration: A tuned smoke_test.py config optimized for speed (trains in <30 seconds).
• Deployment Ready: Verified on Cloud (Colab) and Local environments.
• Inference Verification: Includes scripts to validate if the model has "learned" the pattern.

📂 Project Structure

nanoGPT/
├── config/
│   ├── smoke_test.py      # <--- Custom config for rapid testing
│   └── train_shakespeare.py
├── data/
│   └── smoke_test/        # Generated binary data (excluded from git)
├── model.py               # GPT Model Definition
├── train.py               # Training Script
├── sample.py              # Inference/Sampling Script
├── test_deploy.py         # Unit tests for environment checks
└── README.md