Please login through the login node first, reserve an interactive session in Slurm, then should be able to ssh to the GPU node.

Login to node using ssh

Then run sinfo to get a list of partition and nodelist names, as below.

$ sinfo

PARTITION         AVAIL         TIMELIMIT     NODES       STATE           NODELIST

common*            up             90-00:00:0          1             mix             dkucc-core-01

common*            up             90-00:00:0          7             idle             dkucc-core-[02-08]

common-gpu      up             7-00:00:00          1             alloc           dkucc-core-gpu-01

common-gpu      up             7-00:00:00          1             idle             dkucc-core-gpu-02

The interactive session for a GPU node can be done on the DKUCC with the command:

					srun -p common-gpu --gres=gpu:1 --exclusive --pty bash -i
Table of Contents