🧠 AI Hackathon 2025

Welcome to the AI Hackathon 2025! This project focuses on fine-tuning and prompt engineering using state-of-the-art models in a containerized environment.

You can find a ready-to-use copy of this repository in iridis-X (/scratch/aazd1f17/shared_space/AI_hackathon25)

cp -r /scratch/aazd1f17/shared_space/AI_hackathon25 .

---

🚀 Setup Instructions

1. Prepare the Container Environment

You can either:

• Build the container yourself, or
• Request access to a pre-built .sif container from the hackathon organizers.

To build the container:

cd helical
./build_helical_container.sh

2. Install Required Dependencies

Once the container is ready, install the necessary Python packages:

cd ../
./install_and_download_dependencies.sh

---

📓 Running Jupyter Notebooks

1. Login to Open OnDemand (University of Southampton):
   • Open https://iridisondemand.soton.ac.uk/pun/sys/dashboard in your browser.
2. Launch a Desktop session:
   • Select Desktop, adjust the number of hours and cores (we recommend 32 cores), and launch the session.
3. Open a terminal emulator inside the desktop.
4. Navigate to the project folder and start Jupyter:

cd AI_hackathon25
./run_jupyter.py

5. Copy the Jupyter URL from the terminal and open it in your browser.
6. Open an existing notebook or create a new one to start coding.

---

🧪 Enabling Command-Line Tools in Jupyter

To access command-line tools like ontogpt inside Jupyter notebooks, run this command in the first cell (if you're in the notebooks/ directory):

%env PATH=../.local/bin:$PATH

---

🤖 Using OntoGPT with Ollama (Local LLM Inference)

To run OntoGPT with a locally hosted Ollama model:

1. Open a new terminal.
2. Navigate to the project directory:

cd AI_hackathon25

3. Load the Ollama module and start the server:

module load ollama
ollama serve

4. (Optional) Download a model (e.g., gemma:7b):

ollama pull gemma:7b

Example Usage from a Notebook:

!ontogpt complete -m ollama/gemma:7b "Why did the squid cross the coral reef?"

---