The Cluster's GPU Nodes are initially going to use the SLURM Scheduler. Cluster users should expect PBS Scheduler to be replaced with SLURM on all nodes of the cluster later this year.
Rosetta Stone of Workload Managers as a good starting point to learn about SLURM:
SLURM's Quickstart User Guide provides some information about how SLURM is architected. The example commands and output should be useful:
A couple useful comands:
In order to create an interactive job with two GPUs, you would run something like this from the head node:
srun -N1 --gres=gpu:2 --pty $SHELL
Adding something like this to your submit script will request a GPU with the job:
There are a variety of sample SLURM job submission scripts made available by other institutions using SLURM. Many of these should be fairly portable though they may contain some bits specific to their environments.
-- David Pane - 2017-01-12