Thesa: A Therapy Chatbot 👩🏻‍⚕️

Thesa is an experimental project of a therapy chatbot trained on mental health data and fine-tuned with the Zephyr GPTQ model that uses quantization to decrease high computatinal and storage costs.

Table of Contents

1. Dataset(s)
2. Training the model
3. How to use
4. Reproduce results
5. Results

Dataset(s)

At the moment, Thesa is trained with two datasets:

• CounselChat - extracted from HuggingFace
• Mental Health Conversational Data - extracted from Kaggle

Training the model

Hardware

Thesa has been trained on Google Colab Pro, using a V100 GPU.

Model architecture

• Model type: fine-tuned from TheBloke/zephyr-7B-alpha-GPTQ on various mental health datasets
• Language(s): English
• License: MIT

Intended uses & limitations

This model is purely experimental and should not be used as substitute for a mental health professional.

Training loss

How to use

Thesa can be tested by running playground.py. Open it and edit the example and prompt variables as desired, then run the code.

Requirements: transformers installed and inference.py downloaded, though it's recommended to clone this entire repository.

Reproduce results

1. Clone the repository: git clone https://github.com/johnhandleyd/thesa
2. Install dependencies via pip install -r requirements
3. Open thesa.py on your favourite code editor and run it!

Results

To see a few samples, check out results.md.