StephenMaaa / ChatSense

ChatSense - Llama 2 + Code Llama + CLIP based Chatbot

ChatSense - Llama 2 + Code Llama + CLIP based Chatbot

Demo

Features

• Chat completions powered by Llama 2
• Code completions powered by Code Llama
• Text-image/image-image search chat powered by CLIP (with FAISS)
• Side bar window for multiple chat histories for three models
• ChatGPT-style GUI
• Store and load user's chat history in local storage

To-Do

• Integrate CLIP model into the app (front-end, backend)
• Integrate CodeLlama (front-end, backend)
• App header with dropdown model select
• Scrolling sidebars (chat container and side bar)
• Multiple chats (side bar) for Llama 2
• User settings (theme, logout, delete, profile)
• Multiple chats for Code Llama
• Multiple chats for CLIP
• Data import/export [PENDING]
• Refine minor UI [PENDING]

Setup

1. Clone the repository

git clone https://github.com/StephenMaaa/GPT.git

2. Install client dependencies

cd GPT
pip install -r requirements.txt

3. Download weights of Llama 2 and Code Llama from Huggingface

URL [Llama 2]: https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGUF

llama-2-7b-chat.Q4_K_M.gguf 

URL [Code Llama]: https://huggingface.co/TheBloke/CodeLlama-34B-GGUF

codellama-34b.Q5_K_M.gguf 

1. Choose and download an image dataset for CLIP

MS COCO URL: https://cocodataset.org/#download

Usage

1. Start the app

python manage.py runserver

License