The goal of this project is to build a closed-source chatbot on a CPU using the quantized Llama2 model (7B parameters).
The resulting application will be evaluated based on it's ability as a tool of convenience for retrieving information from research papers. More specifically, it will evaluated by the quality of it's responses, the run time, and the memory expenditure.
-
Clone this repository using the command:
git clone https://github.com/anair123/Llama2-Powered-QA-Chatbot-For-Research-Papers.git
-
Download a quantized Llama2 model (pick any one) from the following link: https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML/tree/main
-
Store the model in the "models" directory
-
Create a virtual environment and enter it
python -m venv <name_of_venv>
venv/Scripts/Activate
-
Install the dependencies with the command:
pip install -r requirements.txt
-
Add all the pdf documents you want to interact with in the "data" folder.
-
Run the Streamlit web app with the command:
streamlit run app.py
Aashish Nair
LinkedIn: www.linkedin.com/in/aashish-nair
Medium: https://medium.com/@aashishnair