slurm: chip-gpu

Users and Accounts

Within these slurm clusters, every user is associated with a slurm user and slurm account. These mirror the UNIX user and group. E.g., user student belonging to UNIX group pi_professor is associated with a slurm user student and a slurm account pi_professor.

Partitions and QOSs

Within chip-gpu, there are no partitions or QOSs.

Running a Job

There are three required slurm flags, one for specifying the runtime of the job, one for specifying the CPU memory for the job, and one for specifying the GPUs. These are demonstrated in a few scenarios below. In each case, srun is used as only for illustrative purposes, sbatch and salloc remain perfectly valid across chip.

`srun –cluster=chip-gpu –time=5 –mem=5 –gres=gpu:1  <bash executable>`

`srun –cluster=chip-gpu –time=5 –mem=5 –gres=gpu:1 –constraint=L40S <bash executable>`

`srun –cluster=chip-gpu –time=5 –mem=5 –gres=gpu:2 –constraint=H100 <bash executable>`

 

To see the feature list available on chip-gpu, run `sinfo -o “%15N %50f”` at the CLI.

 

Please see the “User Support” drop-down menu for additional resources.