SLURM Script Generator

PERUN Slurm Script Generator

Interactive generator with automatic scratch support

Slurm · PERUN · HPC

Job Configuration

Fill in the form and the Slurm script will be generated below.

JOB NAME

ACCOUNTProject account

QOSQuality of Service

SCRIPT TYPE

PARTITIONCPU / GPU on PERUN

RUNTIME (HH:MM:SS)

CPUS PER TASK

GPUs PER NODEGPU jobs only

NODESFor MPI jobs

TASKS / MPI RANKS

OUTPUT FILE

COMMAND

Enable automatic scratch

Force specific node

Append output

Set OMP_NUM_THREADS

Email Notifications

Select when to receive email notifications about your job.

MAIL USEREmail address for notifications

Shortcuts

BEGIN,END,FAIL

ALL

NONE (clear)

Job states

BEGIN

END

FAIL

REQUEUE

ALL

Time limits

TIME_LIMIT (100%)

TIME_LIMIT_90

TIME_LIMIT_80

TIME_LIMIT_50

Other

ARRAY_TASKS

STAGE_OUT

INVALID_DEPEND

Automatic Scratch (Recommended)

When enabled, your job automatically runs in fast Lustre scratch with 40x faster I/O. Just add source .activate_scratch and everything else is handled by prolog/epilog. Results are synced to ~/results_job_XXXXX/ after completion.

GPU Jobs

For GPU jobs, always use the GPU partition and request GPUs via --gres=gpu:X.

Email Notifications

ALL = BEGIN + END + FAIL + REQUEUE + STAGE_OUT. Options can be combined, e.g. END,TIME_LIMIT_80. ARRAY_TASKS sends per-task notifications for job arrays (otherwise only summary).

Node Selection

Avoid forcing a specific node in production. Let Slurm schedule your job optimally.

Generated Slurm Script

How It Works

PROLOG (automatic): Creates scratch, copies files
YOUR JOB: Runs in fast scratch with source .activate_scratch
EPILOG (automatic): Syncs results back, cleanup

Submit & Monitor

sbatch job.sh · squeue -u $USER · tail -f ~/results_job_*/output.log