Skip to content

Latest commit

 

History

History
141 lines (96 loc) · 4.54 KB

README.md

File metadata and controls

141 lines (96 loc) · 4.54 KB
marimo LLMs from scratch banner

Marimo LLMs From Scratch

Implement a ChatGPT-like LLM in PyTorch from scratch, step by step with marimo.

Marimo Python Language License

About • Getting Started • Chapters • Why Marimo • License

📖 About The Project

This repository provides interactive marimo notebook implementations of the best-selling book "Build a Large Language Model From Scratch" by Sebastian Raschka.

We focus on providing a seamless experience for readers of the Traditional Chinese Edition (讓 AI 好好說話！從頭打造 LLM 實戰秘笈), featuring:

  • Dual Language Support: Code comments and explanations in both English and Traditional Chinese.
  • Interactive Learning: Visualizing Attention mechanisms and Transformers using Marimo's UI.

Original Source: LLMs-from-scratch by Sebastian Raschka

🚀 Getting Started

Prerequisites

  • Python 3.12+
  • pip

Installation

  1. Clone the repository:

    git clone https://github.com/thliang01/marimo-LLMs-from-scratch.git
cd marimo-LLMs-from-scratch

  2. Install dependencies:

    pip install -r requirements.txt

🐱 Twinkle's Tip: Running Marimo

"Wait! Don't use python filename.py!"

Marimo files look like standard Python scripts, but to see the magic (graphs, sliders, and interactivity), you need to run them with the marimo editor.

To open a notebook in edit mode:

marimo edit ch3/marimo_ch03.py           # English version
marimo edit ch3/marimo_ch03_zh_tw.py     # Traditional Chinese version

To run a notebook as an app:

marimo run ch3/marimo_ch03.py

📂 Repository Structure

marimo-LLMs-from-scratch/
├── ch3/                           # Chapter 3: Coding Attention Mechanisms
│   ├── ch03.ipynb                # Original Jupyter notebook (English)
│   ├── marimo_ch03.py            # Marimo notebook (English)
│   ├── marimo_ch03_zh_tw.py      # Marimo notebook (繁體中文)
│   └── README.md                 # Chapter 3 documentation
├── requirements.txt              # Python dependencies
└── README.md                     # This file

📚 Chapters

Chapter 3: Coding Attention Mechanisms

Covers the implementation of attention mechanisms, the core engine of LLMs:

  • Simple self-attention without trainable weights
  • Scaled dot-product attention with Q, K, V matrices
  • Causal attention with masking
  • Multi-head attention

📁 See ch3/README.md for detailed documentation

💡 Why Marimo?

Marimo is a next-generation Python notebook that offers several advantages:

  • ✅ Reactive: Cells automatically update when dependencies change
  • ✅ Reproducible: No hidden state, deterministic execution order
  • ✅ Git-friendly: Notebooks are stored as .py files
  • ✅ Interactive: Rich UI elements and real-time feedback
  • ✅ Executable: Can be run as both notebooks and Python scripts

🤝 Contributing

Contributions are welcome! Please feel free to submit issues or pull requests.

📄 License

This project is licensed under the Apache License 2.0. It is based on the original work from LLMs-from-scratch by Sebastian Raschka.

📖 References

🙏 Acknowledgments

Special thanks to Sebastian Raschka for creating the original LLMs-from-scratch materials and book.