OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Developer Tools & SDKs

Token Print

Token Print is an interactive web application that visualizes transformer architectures, tensors and language-model inference in 3D, with views tied to model data and a real forward pass.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Token Print — OpenViST project profile in AI Developer Tools & SDKs.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Token Print when a static architecture diagram is not enough and you want an interactive view of model structure or inference, with data provenance visible alongside the visualization.

Who it is for

  • Learners exploring transformer internals and inference
  • Developers inspecting model architecture, tensors and activations
  • Researchers or educators demonstrating attention, RoPE and residual-stream behavior

Documented uses

  • Inspect attention heads, RoPE rotations and residual-stream activity during inference.
  • Explore model architecture and tensor values through a 3D interface.
  • Generate text, follow a guided walkthrough or use debugger breakpoints and activation ablation.
  • Load supported model families or inspect a local GGUF model.
Before you choose

Requirements & limits

The repository describes a local web setup and provides an MIT license.

  • The documented local setup requires Python 3.11 or 3.12 and Node 20.9 or newer.
  • The live Qwen view needs no model file; native GGUF execution has separate dependencies.
  • GGUF files are parsed client-side and are not uploaded, according to the repository.
  • Per-layer activations are unavailable for quantized GGUF inference through llama.cpp.
  • Attention weights are signals, not causal proof; MPS/CUDA per-layer timings are approximate.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up Token Print locally

Prepare the documented local backend and frontend for the browser-based model inspector.

Before you start

  • Python 3.11 or 3.12
  • Node 20.9 or newer
  • A local copy of the Token Print repository

Steps

  1. Run the repository launcher with the command below; it installs dependencies, downloads the model with progress and checks both services.

    Documented command
    python3 scripts/start.py

  2. Alternatively, follow the README quickstart to create and activate the Python virtual environment, install backend requirements and start the backend with Uvicorn.

  3. Install frontend dependencies with the command below in the frontend directory and start it with the command below.

    Documented command
    npm install
    Documented command
    npm run dev

Expected outcome

The documented local backend and frontend are started for access through the browser.

Guide limitations

  • The live Qwen view does not require a separate model file.
  • Native GGUF execution uses a separate requirements file.
  • Quantized GGUF inference does not expose per-layer activations through llama.cpp, so layer lighting is disabled.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
13 October 2026
Original website title
Sudharsanselvaraj/Token-Print · Official homepage title · retained in its source language
How we review sourcesSuggest a correction