Educational Quiz Question Generator

Description

This is an educational quiz question generator developed for the IBM – Mixed Reality Board and Toy Games with Granite 3 SLMs project.

The Generator utilizes:

• Quantized IBM granite-3.0-8b-instruct for question generation
• YOLO for object detection

It consists of two main components:

• A GUI desktop application for generating questions
• A FastAPI server backend for handling API requests and AI model processing

---

Prerequisites

• Nvidia GPU with CUDA support (required)
• Python 3.8 or later
• pip package manager
• CUDA Toolkit properly installed
• Compatible NVIDIA drivers installed and updated

---

Installation

1. Clone the project

   git clone https://github.com/dcloud347/EducationalQuestionGenerator.git
   cd EducationalQuestionGenerator

2. Install PyTorch (CUDA version)

   Install the appropriate torch and torchvision versions matching your CUDA version:

   Find your installation command here

3. Install llama-cpp-python with CUDA or Metal support

   For CUDA (Linux/Windows):

   CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python

   For MacOS (Metal support):

   CMAKE_ARGS="-DGGML_METAL=on" pip install llama-cpp-python

4. Install dependencies

   pip install -r requirements.txt

---

Usage

GUI Application

Run the desktop GUI application:

python ./main.py

Features:

• Select number of questions (1-5)
• Choose subject area
• Add specific topics (optional)
• Set target age group
• Generate and save questions to output.json

---

FastAPI AI Server Backend

The FastAPI server handles:

• AI Model Integration: For question generation and image recognition
• Data processing and storage

Start the server:

uvicorn app.main:app --reload

Access API documentation:

http://127.0.0.1:8000/docs

---

Docker Deployment

Deploying via Docker is also supported:

1. Build the Docker image

   docker build -t educational-quiz-generator .

2. Run the Docker container

   docker run --gpus all -p 8000:8000 educational-quiz-generator

Access the API at http://127.0.0.1:8000.

---

Front-End Description

The project includes a front-end UI built with teacher-ui, providing educators an intuitive interface to interact with the quiz generator.

Features:

• Intuitive Dashboard: Easily generate and manage quiz questions
• Responsive Design: Optimized for desktop and mobile
• Real-Time Interaction: Connects seamlessly with the FastAPI server for live feedback
• Customizable Settings: Choose subject area, age group, and topics with ease

For customization and deployment details, visit the teacher-ui GitHub repository.

---

Testing

Always test before pushing!

Run tests:

coverage run --source=app -m pytest
coverage report --show-missing
coverage html --title "${@-coverage}"

---

Troubleshooting

• Ensure CUDA Toolkit and compatible NVIDIA drivers are correctly installed
• Verify PyTorch version matches your CUDA version
• Confirm llama-cpp-python is installed with GGML_CUDA=on (or GGML_METAL=on for MacOS)
• For MacOS, ensure Metal support is available and configured
• GPU not recognized?
  • Check NVIDIA driver installation
  • Check CUDA version compatibility
  • Ensure environment variables like GGML_CUDA=on are set before installation

---

Repository Links

• Main Repository: Educational Question Generator
• Front-End UI: teacher-ui

---

Author

Yue Pan