Quickstart: partition gpu-2

The gpu-2 partition

In the gpu-2 partition, each A100 is divided into two MIG instances of ~20 GB (3g.20gb). Use this partition for large GPU jobs.

Why choose gpu-2?

• If we have already gone through the gpu-3 partition and have experience using GPUs.
• If our model needs ~20 GB of VRAM.
• If we are not running tests or concepts.

Examples

Quick interactive example:

salloc --partition=gpu-2 -n 1 --cpus-per-task=8 --mem=32G --gres=gpu:1

Batch template (copy, paste and adjust parameters):

#!/bin/bash
#SBATCH -p gpu-2
#SBATCH -n 1
#SBATCH --cpus-per-task=8
#SBATCH --mem=32G
#SBATCH --gres=gpu:1
#SBATCH --time=04:00:00

module purge
module load CUDA/12.0.0

nvidia-smi
# here run your GPU code, e.g. python train_large.py

Notes - QOS rules may limit total CPU/memory/GPU per user; if a submission is rejected, reduce requested resources or check the QOS configuration. - Confirm the GRES labels on the node with scontrol show nodes | egrep "gres" if in doubt.