We are happy to announce that we have reprovisioned two older Nvidia Tesla K80 GPUs and added one of them to each of the two nodes that are forming the notchpeak-shared-short partition.
This in effect means that users can use these GPUs (4 in total as one K80 card consists of two GPUs) for short interactive jobs, either via the terminal or via Open OnDemand web portal.
We are hoping this will be helpful for those in need of interactive GPU work, such as deep learning.
To access the GPUs via a interactive job, run the following:
srun -N 1 -n 2 --gres=gpu:k80:2 -t 1:00:00 -A notchpeak-shared-short -p notchpeak-shared-short --pty /bin/bash -l
Access via Open OnDemand is shown in the picture below.
As an added bonus, if you use the modules listed on the top of the picture you'll get Jupyter Notebook with Tensorflow, which we installed following instructions at https://www.chpc.utah.edu/documentation/software/python-anaconda.php
If you have any questions on issues accessing the GPUs, please let us know at firstname.lastname@example.org.