This tool can aid you in setting up your HPC submit scripts. Simply enter the parameters for your jobs, and copy/paste the output script.
For advanced parameter settings, you may need to manaually tune settings after using this tool. If you experience any issues or have a feature request, let us know!
Create a name for your job (alphanumeric, dashes, and spaces allowed)
Enter the program you wish to run for your job. If you pipe input or pass arguments, include those.
This control allows you to separate out the results of your job run from any verbose output from the code itself including errors. This results in two distinct files, one for results and one for verbose output and errors. Please click the checkbox if you want to do this then enter a filename.
Please select the type of email notifications about your job you would like to receive.
Select the number of processor cores your job will run on.
Select the number of compute nodes your job will run on. Note that adjusting this does not guarantee
that processes will be evenly distributed across all nodes. The default is "No Preference" and the Number of Processes, #SBATCH -n are adjusted instead.
If Number of Nodes is set to "No Preference", the #SBATCH -N line will not appear in the script.
Enter the maximum runtime that you wish to allow your job to run for. This runtime must be equal to or less than the maximum runtime for your selected partition.
This control displays the amount of runtime you are requesting to be allocated for your job. By default, the default job time for your chosen partition is displayed. If you want to adjust this manually, just click the checkbox next to "Manually Adjust Runtime" and this will allow you to adjust the controls.
Enter the amount of memory you wish to allocate to each core.
This control allows you to manually adjust the amount of memory you want to allocate to the cores in your job. If you leave this blank, your job will run with the default memory allocation for your chosen partition.
If your job needs to be run on specific hardware (either make or model year), then you can enforce those constraints using this control. To find out
what hardware is available on your chosen partition run sinfo -p PARTITION_NAME -o %f in the terminal. This will print a list of available
features.
GPUs are available only on the backfill2 partition and certain owner partitions. Please note, if you are using backfill2:
If you have specified the
#SBATCH -N parameter for number of Nodes, you CANNOT request GPUs. You must either set -N to 0 (default) and then set the number of GPUs you require
or you must specify the number of nodes you require and do not specify GPUs.
Select any modules that your job needs to run. Refer to the documentation for the software package that you wish to run for details.
It is possible to supress the output warnings that SLURM gives. This is an advanced option and not recommended for beginning users
It is possible to set the minimum number of threads you want to run on each core included in your job.
{[{ error }]}
NOTE: Please be aware that imposing hardware constraints on your job may increase your wait time between the time you submit the job and the time your job runs.
#!/bin/bash#SBATCH --job-name={[{ jobname_final }]}
#SBATCH -e {[{ errorfile_name_final }]} #SBATCH --mail-type={[{ email_types }]}#SBATCH -n {[{ n_procs }]}
#SBATCH -N {[{ nodes_final }]} #SBATCH -p {[{ selected_partition.attributes.name }]}#SBATCH -t {[{ runtime }]}
#SBATCH --mem-per-cpu={[{ mem_per_proc_final }]} #SBATCH -C {[{ constraints_final }]} #SBATCH --gres=gpu:{[{n_GPU}]} #SBATCH -Q #SBATCH --profile={[{profile_types}]} #SBATCH --threads-per-core={[{n_thred_per_core}]} module load {[{ module }]}
## Replace PROGRAM, INFILE, OUTFILE and OPTIONS with the program, I/O files and parameters needed for your program run. ## Submit Script Generator automatically added "srun" to your script srun {[{ program_call }]} {[{ program_call }]}
To use this script:
vi my_submit_script.sh)sbatch my_submit_script.sh