PriceProphet is a GenAI application designed to predict the price of a product based on its description. This project aims to combine state-of-the-art natural language processing (NLP) techniques with deep learning to build a robust pricing model. Leveraging the fine tuned Llama model from Hugging Face's library and deploying it via an intuitive frontend interface.

Key Features:

1. Data Collection and Preprocessing:

   • Fetch product descriptions and price data from Hugging Face or other external sources.
   • Perform text cleaning, tokenization, and formatting for model training.

2. Model Training and Fine-Tuning:

   • Fine-tune the pre-trained Llama model on custom pricing data.
   • Use regression-based approaches to predict continuous price values from text descriptions.

3. Evaluation and Validation:

   • Measure model performance using metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE).
   • Conduct thorough evaluation to ensure accuracy and generalizability.

4. Deployment with Gradio:

   • Provide a user-friendly interface for product price prediction.
   • Allow users to input product descriptions and receive instant price estimates.

5. Integration with Hugging Face:

   • Leverage Hugging Face's ecosystem for data management, model hosting, and deployment.
   • Upload the trained model back to Hugging Face for public sharing and accessibility.

---

Objectives:

• Build a comprehensive understanding of the NLP model fine-tuning process.
• Explore regression tasks using LLMs in a practical, business-oriented context.
• Gain hands-on experience with Hugging Face tools and APIs for data and model management.
• Develop a functional, interactive Gradio application to showcase the model's capabilities.

---

Technologies Used:

• Backend:

  • Hugging Face Transformers (Llama model)
  • PyTorch/TensorFlow
  • Flask

• Frontend:

  • Gradio for the interactive interface

• Infrastructure:

  • Hugging Face Datasets and Model Hub
  • Docker for containerization

---

Use Cases:

• E-commerce platforms can leverage this model to estimate product prices based on user-submitted descriptions.
• Market researchers can analyze trends by predicting prices from textual data.
• Small businesses can quickly determine price ranges for products without extensive market research.

---

Why This Project?

PriceProphet is an excellent project for anyone seeking to enhance their machine learning and NLP skills while solving a real-world problem. It provides hands-on experience in data processing, model training, evaluation, and deployment, making it a perfect side project for learning and skill development.

Requirements

1- A system with nvidia graphic card 2- Docker and docker compose

How To Run?

1- Create an .env file inside the app directory and add your hugging face token as HF_TOKEN=. 2- Run docker-compose up --build.