A Turkish question-answering application using a fine-tuned T5 model with a FastAPI backend and a React frontend.
📎 Download model from Google Drive: https://drive.google.com/file/d/19ScqfRWVmJ6nBxTBRalqMwiAOTWWNG30/view?usp=sharing
T5 Model - React & FastAPI Application This project is a Question-Answering Application powered by a fine-tuned T5 model trained on the Turkish language. The frontend is built with React.js, and the backend is developed using FastAPI.
🚀 Technologies Used 🔮 T5 (Text-To-Text Transfer Transformer) – Fine-tuned for the Turkish language
⚡ FastAPI – A modern, high-performance web framework for serving the model
⚛️ React.js – For collecting user questions and displaying answers
🧠 HuggingFace Transformers – For loading the model and generating responses
🔗 Google Drive – Used to store the large model file externally
✍️ How to Use Type a question into the input box on the page.
Click the "Submit" button.
Instantly view the response generated by the model.
🔀 Example Question Input: Product: Ravla Zeytinyağı Questions: Soğuk sıkım mı?
Model Response: Yes, this product is cold-pressed olive oil.
📌 Notes The model file is ignored via .gitignore because it exceeds 500 MB.
git-lfs could be used as an alternative, but Google Drive is preferred for simplicity.
🖥️ How to Run the Project
- Backend 📦 Install dependencies:
bash
Kopyala
Düzenle
pip install fastapi uvicorn transformers torch sentencepiece safetensors
bash
Kopyala
Düzenle
cd backend
uvicorn main:app --reload
📎 Download model from Google Drive: https://drive.google.com/file/d/19ScqfRWVmJ6nBxTBRalqMwiAOTWWNG30/view?usp=sharing
👨💻 Developer Mahmut Can Boran 🖥️| AI Engineer & NLP Enthusiast 📧 Email: mahmutcanboran@gmail.com 🔗 GitHub: @mahmutcanborann