A web application that visualizes GPT model token generation and probabilities in real-time. Built with React frontend and Python backend.

• Interactive text generation from prompts
• Visual token nodes graph with interaction
• Token probability analysis
• Real-time decoding visualization
• Customizable generation parameters

Gpt2-token-prob/
├──gpt_visualizer/
                  ├──frontend/           # React application
                  └── backend/           # Python Flask API

• Node.js and npm
• Python 3.x
• Virtual environment (venv)

1. Backend Setup

cd backend
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt

2. Frontend Setup

cd frontend
npm install

1. Start the backend server:

cd backend
source venv/bin/activate  # On Windows: venv\Scripts\activate
./run.sh  # On Windows: run.bat

2. Start the frontend development server:

cd frontend
npm start

The application will be available at http://localhost:3000

• Frontend is built with React and uses Material-UI for components
• Backend uses Flask for API endpoints
• Token visualization implemented using React-based graph visualization

This project is licensed under the MIT License.