# Training Monitor

Real-time monitoring system for ML training processes, specialized for diffusion model training visualization.

## Features

- Real-time training progress visualization
- Training metrics monitoring (loss, learning rate, ETA)
- Sample image gallery with timeline view
- Remote SFTP data source integration
- Docker-based deployment

## Project Structure

```
training-monitor/
├── frontend/                 # React frontend application
│   ├── src/
│   ├── public/
│   ├── Dockerfile
│   └── package.json
├── backend/                  # FastAPI backend application
│   ├── app/
│   │   ├── api/             # API routes
│   │   ├── core/            # Core functionality
│   │   ├── models/          # Data models
│   │   └── services/        # Business logic
│   ├── tests/
│   └── Dockerfile
├── docker/
│   └── docker-compose.yml
└── scripts/                  # Utility scripts
```

## Tech Stack

- **Frontend**: React (v18+)
    - Tailwind CSS for styling
    - React Query for data fetching
    - React Router for navigation

- **Backend**: FastAPI
    - Paramiko for SFTP operations
    - Pydantic for data validation
    - SQLAlchemy for database (if needed)

## Setup Instructions

1. Clone the repository:
```bash
git clone https://github.com/yourusername/training-monitor.git
cd training-monitor
```

2. Start the development environment:
```bash
docker-compose up --build
```

The application will be available at:
- Frontend: http://localhost:3000
- Backend API: http://localhost:8000
- API Documentation: http://localhost:8000/docs

## Development

### Frontend Development
```bash
cd frontend
npm install
npm run dev
```

### Backend Development
```bash
cd backend
python -m venv venv
source venv/bin/activate  # or `venv\Scripts\activate` on Windows
pip install -r requirements.txt
uvicorn app.main:app --reload
```

## Configuration

Create a `.env` file in the root directory:

```env
# Backend
SFTP_HOST=42.42.42.42
SFTP_USER=username
SFTP_PASSWORD=your_password
SFTP_PATH=/data/samples

# Frontend
REACT_APP_API_URL=http://localhost:8000
```

## Contributing

1. Create a feature branch
2. Commit your changes
3. Push to the branch
4. Create a Pull Request

## License

This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.