-n: run the job in a sub-directory -r: try to requeue when job exited due to insufficient time --dir=<job_dir>: job directory
name: job display name account: account name for job submission walltime: job execution time nnodes: number of nodes to run job cpus_per_node: overwrite cluster CPU configuration gpus_per_node: overwrite cluster GPU configuration