NeuralPaper.ai

NeuralPaper.ai is a platform that integrates Neural DSL with NeuralDbg to create interactive, annotated neural network models with visualization features similar to nn.labml.ai. It provides a web-based interface for creating, visualizing, and sharing neural network models with detailed explanations and interactive elements.

Features

• Annotated Models: Explore neural network architectures with detailed annotations explaining each component
• Interactive Visualization: Visualize model architecture, shape propagation, and computation flow in real-time
• Live Debugging: Debug models with NeuralDbg to analyze gradients, activations, and performance
• DSL Playground: Experiment with Neural DSL to create and modify models with instant feedback
• Educational Resources: Learn about neural network concepts with interactive examples and tutorials

Project Structure

neuralpaper/
├── backend/
│   ├── api/                  # FastAPI server for Neural DSL execution
│   ├── models/               # Annotated model implementations
│   └── integrations/         # Connectors to Neural and NeuralDbg
├── frontend/
│   ├── components/           # React components
│   │   ├── ModelViewer/      # Interactive model visualization
│   │   ├── CodeAnnotation/   # Side-by-side code and annotations
│   │   ├── DSLPlayground/    # Interactive DSL editor
│   │   └── DebugPanel/       # NeuralDbg integration
│   ├── pages/                # Next.js pages
│   │   ├── models/           # Model showcase pages
│   │   ├── playground/       # DSL playground page
│   │   └── blog/             # Blog section
│   └── public/               # Static assets
└── shared/                   # Shared types and utilities

Getting Started

Prerequisites

• Node.js 14+ and npm
• Python 3.8+
• Neural DSL and NeuralDbg installed

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/neuralpaper.git
   cd neuralpaper

2. Install backend dependencies:

   cd backend
   pip install -r requirements.txt

3. Install frontend dependencies:

   cd ../frontend
   npm install

Running the Application

1. Initialize the database with sample models:

   cd backend/scripts
   python init_db.py

2. Start the backend server:

   cd backend
   uvicorn api.main:app --reload

3. Start the frontend development server:

   cd frontend
   npm run dev

4. Open your browser and navigate to http://localhost:3000

Alternatively, you can use the provided start.sh script to start both the backend and frontend servers:

cd neuralpaper
./start.sh

Adding New Models

To add a new annotated model:

1. Create a Neural DSL file in backend/models/ (e.g., mymodel.neural)
2. Create an annotations file in backend/models/ (e.g., mymodel.annotations.json) with the following structure:

   {
     "name": "My Model",
     "description": "Description of my model",
     "category": "Computer Vision",
     "complexity": "Medium",
     "sections": [
       {
         "id": "section1",
         "lineStart": 1,
         "lineEnd": 5,
         "annotation": "Explanation of this section"
       }
     ]
   }

3. The model will be automatically available in the frontend

Making the Implementation Fully Real

The current implementation includes some mock data and placeholders. To make it fully real:

1. Complete the WebSocket Integration:

   • Enhance the WebSocket endpoints in backend/api/main.py to stream real data from NeuralDbg
   • Update the neural_connector.py to capture and forward NeuralDbg output

2. Implement Real-Time Debugging:

   • Add methods to query the running NeuralDbg instance for real-time data
   • Replace the mock trace data with real data from your running NeuralDbg instance

3. Add Database Integration:

   • Replace the file-based storage with a proper database (SQLite, PostgreSQL, etc.)
   • Implement proper model management with CRUD operations

4. Enhance the Frontend:

   • Add more interactive visualizations
   • Implement user authentication
   • Add a blog section for tutorials and articles

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Neural DSL - The domain-specific language for neural networks
• NeuralDbg - The neural network debugger
• nn.labml.ai - Inspiration for the annotated model format