HelioSync (A Federated Learning Platform)

Presenting HelioSync, a comprehensive platform for running federated learning projects across multiple devices while preserving data privacy.

Screenshots

Here are a few snapshots of the platform in action:

Features

• Privacy Preserving Learning: Train machine learning models without sharing raw data
• Multi-Framework Support: Use TensorFlow or PyTorch for model development
• Organization Management: Create and manage organizations with multiple users
• Client Management: Register and monitor client devices participating in training
• Project Dashboard: Track project progress, metrics, and model performance
• Model Deployment: Deploy trained models as APIs or download for local use
• Real-time Metrics: Monitor training progress with visualizations
• API Access: Programmatic access for client integration

Architecture

The platform consists of the following components:

• Web Interface: Flask-based web application for managing federated learning projects
• Federated Server: Coordinates the training process and aggregates model updates
• Federated Clients: Run on devices with local data, train models without sharing data
• Examples: Pre-built examples for common datasets (MNIST, CIFAR-10, Sentiment Analysis)

Installation

Prerequisites

• Python 3.8+
• pip (Python package manager)
• Virtual environment tool (recommended)

Setup

1. Clone the repository:

git clone https://github.com/yourusername/federated-learning-platform.git
cd federated-learning-platform

2. Create and activate a virtual environment:

python -m venv <name_of_env>
source <name_of_env>/bin/activate  # On Windows: venv\Scripts\activate

3. Install the dependencies:

pip install -r requirements.txt

4. Initialize the database:

python init_db.py

Follow the prompts to create an admin user and organization. Make sure to save the API key that is displayed after initialization.

Running the Application

Development Mode

To run the application in development mode:

python run.py --config development --debug

The application will be available at http://localhost:5000.

Production Mode

For production deployment:

python run.py --config production --host 0.0.0.0 --port 5000

Using the Platform

1. Login: Access the web interface and login with the admin credentials created during setup
2. Create Project: Set up a new federated learning project with the desired dataset and parameters
3. Register Clients: Use the API key to register client devices that will participate in training
4. Start Training: Launch the federated training process from the project dashboard
5. Monitor Progress: Track metrics and model performance in real-time
6. Deploy Model: Once training is complete, deploy the model as an API or download it

Client Configuration

To configure a client to participate in federated learning:

1. Ensure you have the required dependencies:

pip install requests

2. Run the federated client with your API key:

python -m fl_client.run_client --client_id client1_saurabh --api_key API_KEY    --server_url http://localhost:5000 --batch_size 32  --epochs 5     --data_split 0.5

3. Command line options:

--server        Server URL (default: http://127.0.0.1:5000)
--api_key       API key for authentication
--name          Client name (defaults to hostname)
--verbose       Enable verbose logging

4. Or import the client in your own code:

from federated_client import FederatedClient

# Initialize client
client = FederatedClient(
    server_url="http://your-server-url:5000",
    api_key="your-api-key",
    client_name="my-device"
)

# Start and run the client
if client.start():
    client.run()

Example Projects (TODO)

MNIST Handwritten Digits

The platform includes an example federated learning project for MNIST digit recognition:

cd examples/mnist
python run_server.py  # Start the federated server
python run_client.py --client-id 1  # Start a client instance

CIFAR-10 Image Classification

For CIFAR-10 image classification:

cd examples/cifar10
python run_server.py
python run_client.py --client-id 1

Sentiment Analysis

For text sentiment analysis:

cd examples/sentiment
python run_server.py
python run_client.py --client-id 1

API Documentation

The platform provides a RESTful API for client integration:

Authentication

All API requests require a valid API key, which should be included in the X-API-Key header.

Client Registration

• /api/register_client: Register a new client (primary endpoint)
• /api/client/register: Alias for client registration
• /api/clients/register: Alias for client registration
• /api/register: Alias for client registration

Client Heartbeat

• /api/client/heartbeat: Update client status with heartbeat
• /api/clients/heartbeat: Alias for client heartbeat
• /api/clients/<client_id>/heartbeat: Update status for a specific client ID

Tasks and Projects

• /api/clients/tasks: Get available tasks for a client
• /api/projects: Get available projects
• /api/projects/<project_id>/server_info: Get server information for a project
• /api/projects/<project_id>/update_status: Update project status

Health Check

• /api/health: Check server status

For detailed API usage examples, see the documentation in the web interface.

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.