README.md

Phi-3.5 Vision

This project provides a graphical user interface (GUI) for interacting with the Phi-3.5 Vision model. It allows users to input prompts and image URLs, send requests to a Flask server running the model, and view the responses along with image previews.

Features

Text input for prompts
Text input for image URLs (multiple URLs supported)
Submit button to send requests to the server
Response display area
Image preview of the first entered URL

Prerequisites

Before you begin, ensure you have met the following requirements:

Python 3.7 or higher
pip (Python package installer)

Installation

Clone this repository:

git clone https://github.com/yourusername/phi-3.5-vision-gui.git
cd phi-3.5-vision-gui

Install the required dependencies:
```
pip install -r requirements.txt
```

Usage

Start the Flask server (ensure you have the server.py file set up and running):
```
python server.py
```
In a new terminal, run the GUI:
```
python gui_pyqt6.py
```
Use the GUI to enter your prompt and image URLs, then click “Submit” to send the request to the server.

Project Structure

gui_pyqt6.py: The main GUI application using PyQt6
server.py: Flask server that interfaces with the Phi-3.5 Vision model
requirements.txt: List of Python dependencies

Configuration

By default, the GUI connects to a Flask server running on http://localhost:5000. If your server is running on a different address or port, modify the submit_request method in the Phi35VisionGUI class in gui_pyqt6.py.

Contributing

Contributions to this project are welcome. Please fork the repository and submit a pull request with your changes.

License

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

Acknowledgments

This project uses the Phi-3.5 Vision model developed by Microsoft.
GUI built using PyQt6.

Contact

If you have any questions or feedback, please open an issue in the GitHub repository.

Install

Flash-attn Install

pip install flash-attn --no-build-isolation

Torch Install, use Cuda12.2

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

For 8bit

pip install -U bitsandbytes