This project provides an interactive 3D visualization of a GPT-style Large Language Model (LLM) during inference. It's designed for researchers and developers interested in understanding the internal workings and topology of transformer-based neural networks. The visualization allows users to explore a functional model, demonstrating how it processes data, starting with a small example from Andrej Karpathy's minGPT.

How It Works

The visualization employs a 3D rendering approach to represent the complex network architecture of GPT-style models. It dynamically displays the flow of information and computations as the model performs inference, offering an intuitive way to grasp the model's structure and behavior. The project is built using JavaScript utilities, facilitating easy deployment and sharing of its interactive components.

Quick Start & Requirements

• Install dependencies: yarn install
• Start the dev server: yarn dev
• Requires Node.js and Yarn.

Highlighted Details

• Interactive 3D visualization of GPT-style LLM network topology.
• Demonstrates inference on a small list-sorting model (based on minGPT).
• Supports visualization of arbitrary-sized networks, including gpt2 (weights not included).

Maintenance & Community

This repository is maintained by Brendan Bycroft and serves as a central hub for his projects. Further community or maintenance details are not specified in the README.

Licensing & Compatibility

The README does not explicitly state a license. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The CPU Simulation project is marked as Work In Progress (WIP) and is not yet exposed. The visualization of larger models like gpt2 requires significant weight downloads, which are not included.