Over the summer we integrated the new LLaMA2 Large Language Model (LLM) into the SURF Research Cloud environment, a cloud computing solution offered by SURF. The explicit goal was to create a low-threshold experimentation playground for researchers to tinker with LLMs. To that end the integration features many example scripts for the running, chatting, and finetuning of the LLaMA2 model (on custom datasets). Furthermore, we will be releasing video tutorials to make it easier to get started, and are also very excited to be hosting a workshop for researchers interested in the technology to guide you through the proces of setting up the models in the SURF Research Cloud and how to start experimenting with it.

This workshop is specifically meant for researchers interested in tinkering with LLMs. The environment has been set up to be easy to use also for infrequent coders, but a bit of python should not scare you. If you’re an advanced coder there’s enough room for you to tinker more deeply with the model, the workshop will then be more of an introduction in how to easily use the HuggingFace infrastructure to set up LLMs (and specifically LLaMA2 models) within the SURF Research Cloud in a collaborative workspace.

Program SURF LLaMA2 workshop

09:00 – 09:30 Walk-in

09:30 – 09:45 Introduction

09:45 -10:10 SURF Research Cloud explanation

10:10 – 10:30 BREAK

10:30 – 11:00 SURF Research Cloud set-up

11:00 – 11:45 LLaMA2 overview and implementation

11:45 – 12:45 LLaMA2 practical

12:45 – 13:00 Closing

Location
Hogeschool Utrecht, Padualaan 99, 
PL99-0.011 ’the Connector’

Date & Time
12th of October 2023
09.00 – 13.00

Register
There are only 35 seats available, please register for the event here.

Organisers
Hogeschool Utrecht en SURF

Hosts
Fabian Kok – Researcher Lectoraat AI – HU
Rins Rutgers – Research Engineer – HU