Florence API Service

A FastAPI-based service that leverages the Florence-2 model for advanced image and text processing. This project provides a containerized API service that can be easily deployed using Docker.

🚀 Features

• FastAPI-based REST API
• Florence-2 model integration
• GPU acceleration support
• Containerized deployment
• Environment variable configuration

📋 Prerequisites

• Docker
• NVIDIA GPU with CUDA support
• NVIDIA Container Toolkit installed

🛠️ Project Structure

florence_api/
├── app/                    # Application code
├── model/                  # Florence model files
├── docker/                 # Docker configuration
│   ├── Dockerfile
│   └── docker-compose.yml
├── main.py                # FastAPI application entry point
├── requirements.txt       # Python dependencies
└── .env                   # Environment variables

🔧 Configuration

Create a .env file in the root directory with the following variables:

API_HOST=0.0.0.0
API_PORT=8000
MODEL_PATH=/app/model

🚀 Running with Docker

1. Build the Docker image

   cd docker
   docker compose build --no-cache

2. Start the service

   docker compose up

   The API will be available at http://localhost:8000

3. Stop the service

   docker compose down

🔍 API Documentation

Once the service is running, you can access:

• API documentation: http://localhost:8000/docs
• Alternative documentation: http://localhost:8000/redoc

🛟 Troubleshooting

CUDA Version Issues

Make sure your NVIDIA driver supports the CUDA version specified in the Dockerfile. You can check your supported CUDA version with:

nvidia-smi

GPU Access Issues

Ensure the NVIDIA Container Toolkit is properly installed:

nvidia-container-cli info

📝 License

[Your License Here]

🤝 Contributing

[Your Contributing Guidelines Here]