[premise-users] New features on the Premise HPC cluster: AI, LLMs, and JupyterLab

RCC Support rcc.support at unh.edu
Thu Sep 26 12:22:09 EDT 2024


Hi Everyone -

We're excited to announce some powerful new features now available on the USNH Premise HPC cluster for existing users. In addition to the standard SSH/Slurm command-line access method, users can now interact with Premise over the web! By connecting to https://premise.sr.unh.edu:8443 (VPN required if connecting from off-campus) using your Premise credentials, users can log into a JupyterLab environment customized and optimized for Premise. For those new to Jupyter, it provides a web-based environment for developing, running, and sharing Python and R based analyses. The Premise JupyterLab environment offers even more, providing the following features:


  *   A customized landing page with information for programmers and non-programmers alike
  *   Access to create and run Python and R Jupyter notebooks with automatic integration with the Slurm job scheduler and access to your Premise files
  *   An AI/LLM library of pre-made routines that run locally on Premise (more coming soon)! This library includes curated and custom LLM prompts for analyzing, writing, and processing documents. It uses popular open-source models like Llama 3.1
  *   Pre-made routines are provided as easy-to-use web forms. No scripting required - though programmers have access to the underlying Python code to use in their own scripts and notebooks
  *   The Python environment on Premise now provides full support for Pytorch, the Hugging Face libraries, and other LLM related features. A range of open source LLM models are also pre-installed. This functionality can be used from the command line environment or JupyterLab

As this is a new service, there may be some bumps along the way, so please let us know if you have any feedback or questions at rcc.support at unh.edu<mailto:rcc.support at unh.edu>. Thanks,

Toni Westbrook, PhD
Deputy Director, Research Computing Center
Managing Director, CIBBR Data Science Core
USNH Enterprise Technology & Services
University of New Hampshire
Office: 106B Service Building

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.sr.unh.edu/pipermail/premise-users/attachments/20240926/68d966cb/attachment.htm>


More information about the premise-users mailing list