NVIDIA Hopper Tensor Core GPU

We are glad to announce that, with generous funding support from the department, SCRP now offers access to NVIDIA Hopper datacenter GPUs.

For accounting reasons, Hopper GPUs are also named a100 on the cluster. To specifically request Hopper GPUs, use the --constraint option with compute/srun/sbatch commands:

compute/srun examples:

# compute
compute --gpus-per-task=a100 --constraint=hopper -c 16 --mem=160G python my_script.py

# srun
srun -p a100 --gpus-per-task=1 --constraint=hopper -c 16 --mem=160G python my_script.py

sbatch example:

#!/bin/bash
#SBATCH --job-name=my_job
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=16
#SBATCH -p a100
#SBATCH --gpus-per-task=1
#SBATCH --constraint=hopper

...