SLURM Job Script Generator

Build a ready-to-submit sbatch script for the Caltech HPC cluster. The output updates as you type — copy it, save it as job.sh, and run sbatch job.sh.

Basics Job name Account / Group (optional, your PI's group) Email notifications (optional) When to email

Job type

Serial (1 CPU) Multi-threaded (OpenMP) MPI (multi-node) GPU Array (parameter sweep)

Resources

Nodes Tasks per node CPUs per task

Memory Unit

Days Hours Minutes

Walltime cap is 7 days. Aim for ~20% over your estimate.

Partition & QoS

Partition (GPU jobs are pinned to gpu; CPU jobs leave blank) QoS

What to run Modules to load (comma-separated) Conda environment (optional) Command(s) to run

Output Working directory (optional, defaults to submit dir) Send stderr to a separate file Reserve the whole node (--exclusive) Use strict bash (set -euo pipefail)

Generated script

Submit it

scp job.sh hpc:~/
ssh hpc 'sbatch job.sh'