Running Jobs¶
The cluster uses Slurm to manage jobs. You write a script, submit it, and Slurm runs it on a compute node when resources are available.
Quick test — CPU job¶
sbatch --partition=cpu \
--wrap="echo Hello from $(hostname)" \
--output=hello-%j.out
squeue --me # check status
cat hello-*.out # read output
Quick test — GPU job (Tesla T4)¶
sbatch --partition=gpu --gres=gpu:1 \
--wrap="nvidia-smi -L" \
--output=gpu-%j.out
squeue --me
cat gpu-*.out
Writing a job script¶
Create a file called my-job.sh:
#!/bin/bash
#SBATCH --job-name=my-experiment
#SBATCH --partition=cpu
#SBATCH --cpus-per-task=4
#SBATCH --mem=4G
#SBATCH --time=01:00:00
#SBATCH --output=logs/%x-%j.out
#SBATCH --error=logs/%x-%j.err
mkdir -p logs
source /projects/YOUR_PROJECT/venv/bin/activate
python train.py --epochs 50
Submit: sbatch my-job.sh
Cancel: scancel JOB_ID
GPU job script¶
#!/bin/bash
#SBATCH --job-name=gpu-train
#SBATCH --partition=gpu
#SBATCH --gres=gpu:1
#SBATCH --cpus-per-task=4
#SBATCH --mem=16G
#SBATCH --time=04:00:00
#SBATCH --output=logs/%x-%j.out
#SBATCH --error=logs/%x-%j.err
mkdir -p logs
source /projects/YOUR_PROJECT/venv/bin/activate
python train.py --device cuda