Detox CoT

Welcome to Detox CoT repository! Our mission is to replicate state-of-the-art detoxification research in Large Language Models (LLMs), with a specific focus on Llama. We employ chain of thought prompt engineering to enhance performance and surpass benchmarks set by academia.

Prerequisites

Before running the script, ensure you have the following:

1. Download and set up Llama 2 models from Llama 2 recipes repository.
2. In a conda env with PyTorch/ CUDA available clone and download this repository.
3. Register for the Perspective API by Google

Note

• Our experiments were conducted using 8 A100 GPUs with 40GB memory for fine-tuning and data generation.
• Access the complete dataset, including finetuning and evaluation data, from AI2's RealToxicityPrompts
• Check out our presentation slide deck summarising our research experience - https://docs.google.com/presentation/d/1AUWSJLxICEr5BcfJ1b5wLceU3S-hWQKmUHckSc-Vp-k/edit?usp=sharing !