Emerging applications of artificial generative intelligence are addressing crucial needs for rapid processing, interpretation and summarization of complex biomedical data. We develop Oncointerpreter.ai built on the Mistral-7B and Llama-2 7B large language models trained on a local, curated corpus to offer personalized summarization of clinical cancer genomic and pathological data in real time via a graphical interface.

To get Oncointerpreter.ai up and running on your local machine, follow these simple steps:

python3 page.py

You must have CUDA configured on your machine to run this application

All the libraries needed for the application can simply be install via requirements.txt

pip install -r requirements.txt

• llama2.py: Contains the code for the Llama2 portion, integrating Llama2 models to enhance the analytical capabilities of the application.
• mistral.py: Houses the Mistral-7B model implementation, contributing to the application's ability to understand and interpret complex medical texts.
• gpt-neo.py: Incorporates the GPT-Neo model, adding to the app's depth in generating human-like text and responses.
• page.py: The main UI of the app. This script initializes and runs the web interface, serving as the primary point of interaction for users.