PsychIT

This is a Psychiatrist Chatbot model that has been fine tuned using QLoRA and Quantized using GGUF.

Prerequisites

To running this project, please ensure that you have the following installed:

• Python 3.8 or Higher (3.10 is recommended)
• Pytorch 2.3.1 (CUDA)
• Cuda toolkit 11.8 or 12.1

Installation

1. Clone this repository:

git clone https://github.com/Raditya-P/PsychIT-AI-Therapist.git

2. Navigate into this project directory:

cd PsychIT

3. Create virtual environment:

   • Using venv:

   python3 -m venv psychIT-env
   

   • Using conda :

   conda create -n psychIT-env python=3.10.*
   

4. Activate the virtual environment

   • Using venv:

   source psychIT-env/bin/activate
   

   • Using conda:

   conda activate psychIT-env
   

5. Install the project dependencies

pip install -r requirements.txt

6. Run the project

   • Text classifier (download the model below):

   flask run
   

   • Text generator

   python3 usage.py
   

7. If you are done, you can exit from the virtual environment with:

deactivate

Additions

If you want to use ollama, you can download Quantized model below, copy and paste quantized model in the same directory as Modelfile, and run ollama using:

ollama create psychIT -f Modelfile
ollama run psychIT

Models

• QLoRA model : https://huggingface.co/radityap/QLoRA-PsychIT
• Quantized model : https://huggingface.co/radityap/PsychIT-7B-GGUF
• Text Classification model : https://huggingface.co/radityap/PsychIT-mental-detection

How We Train The Models

https://colab.research.google.com/drive/1g8S3xOBGuzM5eL2l0cpVBSDTTI_I0b1R?usp=sharing https://colab.research.google.com/drive/1Xe5UXD1FAiY11t6w78LOk8sAlJ7XwAQU?usp=sharing https://colab.research.google.com/drive/17l7GE4sfaIVF7PwHxnBwCov5k_EUN0CV?usp=sharing https://colab.research.google.com/drive/1q43nobwY2zFSEkYGFsiLTSGySHpehzZ-?usp=sharing https://drive.google.com/drive/folders/1XqC4SyGVkIVCjC4qWzOWAOZ9gFAziKlZ?usp=sharing