NLP Group Project with Russell - COMP 482

This project focuses on Natural Language Processing (NLP) applications, utilizing various GPT models, a custom model which was fine-tuned using FAQ data from gigabytes website and a custom GUI. It was developed as part of the COMP 482 course.

---

Installation

Clone the repository

git clone https://github.com/BrianAtkinson93/COMP_482_NLP_Group_Project.git

Create Virtual Environment

Windows

python3.exe -m venv venv

Linux

python3 -m venv venv

Install required dependencies

# Source the virtual environment then:
pip install -r requirements.txt;

---

Usage

---

To run the GUI simply run main.py:

WINDOWS

# For help message and expected arguments
python.exe main.py -h

#usage: main.py [-h] [-a-] [-m {1,2,3,4,5}]
#
#options:
#  -h, --help            show this help message and exit
#  -a-, --api            Use this flag to run with the OpenAI API model.
#  -m {1,2,3,4,5}, --model {1,2,3,4,5}
#                        Choose a model number for the local GPT4All model. (Default is 5)

# Main program execution

python.exe main.py

# Optional model selection
python.exe main.py --model <1-5>

# For use with an API to GPT4
python3 main.py --api

LINUX

# For help message and expected arguments
python3 main.py -h

#usage: main.py [-h] [-a-] [-m {1,2,3,4,5}]
#
#options:
#  -h, --help            show this help message and exit
#  -a-, --api            Use this flag to run with the OpenAI API model.
#  -m {1,2,3,4,5}, --model {1,2,3,4,5}
#                        Choose a model number for the local GPT4All model. (Default is 5)

# Main program execution

python3 main.py

# Optional model selection
python3 main.py --model <1-5>

# For use with an API to GPT4
python3 main.py --api

Contributors

• Brian Atkinson - Main Infrastructure, GUI, Scrapper
• Mason Leitch - Custom model training, Fine-Tuning, Quantization
• Owen Gossen - Testing, file conversion, Presentation

HuggingFace Model Link

The fine-tuned model is automatically downloaded from HuggingFace, but here is the URL for reference:

https://huggingface.co/masonym/gigabyte-1k-q4_0-GGUF.

License

This project includes third-party components under their respective licenses:

• Selenium: Licensed under the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0). For more details, see CC BY-SA 4.0 License.

• GPT-4 Models (GGUF models) from gpt4all.io: The licensing terms for these models should be reviewed at gpt4all.io.

• QLoRA: QLoRA finetuning is licensed under MIT license. QLoRA github repository.

• Llama.cpp: Llama.cpp is licensed under MIT license. Llama.cpp github repository.

• WizardLM 7B: WizardLM 7B is non commercially licensed. WizardLM-7B-V1.0.