ExperimentHub

🚀 A Modern Machine Learning Experiment Management Platform

Track, compare, and optimize your ML models with ease

System Architecture

The ExperimentHub platform uses a modern containerized architecture with the following components:

• Frontend Container:

  • Next.js Application (UI) for the main interface
  • Job Management UI for monitoring training progress
  • Communicates with backend via REST API and WebSocket for real-time updates

• Backend Container:

  • FastAPI Server handling API requests and WebSocket connections
  • ML Model Training Modules for executing experiments
  • Integrates with SQLite for job/experiment tracking
  • Manages MNIST dataset access for training

• Data & Persistence:

  • SQLite Database for storing experiment metadata and results
  • MNIST Dataset for model training

The system uses Docker Compose for orchestration, making it easy to deploy and scale. The architecture ensures real-time communication between components through WebSocket connections for live training updates and REST APIs for general operations.

Features

• Experiment Management: Create, organize, and track machine learning experiments
• Model Training: Train MLP, CNN, and RNN models with customizable hyperparameters
• Real-time Monitoring: Track metrics and progress during training via WebSockets
• Interactive Visualizations: Compare performance metrics across models and experiments
• Job Queue: Manage multiple training jobs simultaneously

Project Structure

The project is organized into two main components:

• Frontend: Next.js-based web interface (React 19, Tailwind CSS)
• Backend: FastAPI server with PyTorch model training

Quick Start

Using Docker

The easiest way to run the entire application stack is with Docker Compose:

# Start all services
docker-compose up

# Access the frontend at http://localhost:3000
# Access the backend API at http://localhost:8000

Manual Setup

For development, you can set up the frontend and backend separately:

1. Backend:

   cd backend
   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   pip install -e ".[dev]"
   uvicorn app.main:app --reload

2. Frontend:

   cd experimenthub
   npm install
   npm run dev

Development

See the individual README files in the frontend and backend directories for detailed development instructions.

License

MIT