The Product

The project aims to develop a Python command-line script for running large language models (LLMs) on the Leiden DS Labs server, using Hugging Face open source Large Language Models. The script will enable researchers to load and interact with various LLMs, allow for adjusting settings like temperature and system prompts and perform batch processing with multiple prompts from a text file. Importantly, the project does not involve creating or training new LLMs. It will only use existing models available on Hugging Face. Besides the command-line interface we also made a Web-interface that lets users interact with the models in a ChatGPT like fashion. Where the users can chat with these models, have multiple histories and delete or redo prompts.

The Customer

Our client is Suzan Verberne. Suzan is a professor of Natural Language Processing (NLP) at the Leiden Institute of Advanced Computer Science at Leiden University. She is also a member of the interdisciplinary research programme Society, Artificial Intelligence and Life Sciences (SAILS)

  • "The greatest problem in communication is the illusion that it has been accomplished." –George Bernard Shaw
  • "Debugging is like being the detective in a crime movie where you are also the murderer."
The Team

We have split the team into 2 groups where the first group mainly focussed on the command-line interface and to transfer this script on the DS labs server so that it is always reachable for researchers and students of leiden. The other group focused on the web-interface so that we had another deliverable since the command-line interface alone was not sufficient for the Software Engineering course. We had biweekly meetings with the client and held our sprints around the meeting with the client. So that we could distribute the work evenly over all the team members.

The Technologies