Repository navigation
Getting Started
Welcome to TokenPrint! This section provides everything you need to get the platform running locally on your machine and to execute your very first live generation.
Because TokenPrint connects to a real PyTorch backend and renders highly complex 3D scenes in the browser, having a correctly configured environment is crucial. A flawless setup ensures that what you see on the screen exactly matches the raw tensor data generated during the forward pass.
TokenPrint is split into two lightweight, decoupled services:
- The Backend: A FastAPI Python application running PyTorch that loads the model into memory and executes inference.
- The Frontend: A Next.js application with React Three Fiber (R3F) that connects to the backend via REST and WebSockets to render the UI.
This separation means you can run the backend on a powerful machine with a GPU while opening the frontend on a lightweight laptop browser.
flowchart TD
User([User])
subgraph Frontend [Next.js Frontend]
UI[React UI]
WebGL[React Three Fiber Canvas]
end
subgraph Backend [FastAPI Backend]
API[REST & WS Endpoints]
Torch[PyTorch Engine]
end
User -->|Opens Browser| UI
UI -->|Requests| WebGL
UI <-->|WebSocket Stream| API
API <-->|Forward Pass| Torch
Follow these pages in order to get up and running:
- Installation: Step-by-step instructions for installing dependencies for both the frontend and backend.
- Quick Start: How to boot both servers and verify they are communicating.
- Running your first visualization: A walkthrough of executing your first live inference or loading your first GGUF file.
Tip If you encounter issues during installation, especially relating to Python packages like PyTorch, ensure you have the appropriate CUDA or MPS tools installed for your operating system.
| Previous | Home | Next |
|---|---|---|
| Home | Home | Installation |
TokenPrint Wiki | Home | Getting Started | User Guide | Concepts | Architecture | API | GitHub
TokenPrint is an interactive 3D visualization platform for exploring transformer architectures, tensors, and real-time LLM inference.
- User Guide
- Architecture Explorer
- Live Inference
- Tensor Inspector
- Camera Controls
- HUD
- Timeline
- Scene Navigation
- Transformer Concepts
- Tokenization
- Embeddings
- Positional Encoding
- RoPE
- Multi-Head Attention
- KV Cache
- LayerNorm
- Feed Forward Network
- Residual Connections
- Logits
- Softmax
- Sampling
- Autoregressive Generation
- Visualization System
- Scene Graph
- Geometry
- Materials
- Animation System
- Color Mapping
- Camera System
- Interaction Model
- Architecture
- Overall Architecture
- Frontend
- Backend
- WebSocket Protocol
- Data Pipeline
- Renderer
- Event System