Running on a cluster

Running on a cluster#

On a supercomputer cluster, it is a common practice to submit a job script to a batch queue system such as e.g. SLURM.

Job script#

In the job script, one specifies the amount of resources to be used for the calculation as well as a line launching the application across the allocated nodes using e.g. srun or mpirun. In this scenario, VeloxChem is run with an input file in the form of either a Python script or a text file. In the example job script given below, both alternatives are provided and you activate the one that you prefer using.

#!/bin/bash

#SBATCH -A project_name
#SBATCH -p partition_name
#SBATCH -J job_name
#SBATCH -t 01:00:00

#SBATCH --nodes=2
#SBATCH --ntasks-per-node=8
#SBATCH --cpus-per-task=16

module load veloxchem/1.0

export OMP_NUM_THREADS=16
export OMP_PLACES=cores
export SRUN_CPUS_PER_TASK=$SLURM_CPUS_PER_TASK

# using a Python script type of input file
srun python myjob.py > myjob.out
# using a text type of input file
#srun vlx myjob.inp myjob.out

# end of job script

This script will start a job on two nodes with eight MPI ranks per node (one per NUMA domain), each with 16 OpenMP threads.

Input file#

Python script

An example Python script type of input file named myjob.py in the job script above and and which is running an SCF optimization takes the following form.

import veloxchem as vlx

xyz_string = """
3
water
O    0.0000000    0.0000000   -0.1653507
H    0.7493682    0.0000000    0.4424329
H   -0.7493682    0.0000000    0.4424329
"""

molecule = vlx.Molecule.read_xyz_string(xyz_string)
basis = vlx.MolecularBasis.read(molecule, "def2-svp")

scf_drv = vlx.ScfRestrictedDriver()

scf_drv.xcfun = "b3lyp"
scf_drv.filename = "vlx_results_hdf5"

scf_results = scfdrv.compute(molecule, basis)

The results of the calculation are stored in an HDF5 file with a user specified name. This file can be directly read and analyzed with VIAMD.

Text file

The same SCF calculation can be performed with use of an input file in text format named myjob.inp in the in the job script above. Such an input file consists of multiple groups marked with @group name and @end, see the listing of text file input keywords.

@jobs
task: scf
@end

@method settings
xcfun: b3lyp
basis: def2-svp
@end

@molecule
charge: 0
multiplicity: 1
xyz:
O    0.0000000    0.0000000   -0.1653507
H    0.7493682    0.0000000    0.4424329
H   -0.7493682    0.0000000    0.4424329
@end
Contents