Submitting job scripts on LUMI
To run Quantum Package (QP) and CHAMP codes properly, some environment
variables should be set. To set up the environment, you should source
the /project/project_465000321/environment.sh file in your shell
when you log in:
source /project/project_465000321/environment.sh
This file contains the following commands:
module load cray-python
source /project/project_465000321/qp2/quantum_package.rc # source QP environment
When working with SLURM, you should use sbatch to submit your QP and CHAMP job
scripts. For example, to submit a QP job script named job_qp.sh, you would use
the command:
sbatch job_qp.sh
1 Example of a QP job script on LUMI
An example of a QP job script would be:
#!/bin/bash #SBATCH --account=project_465000321 #SBATCH --reservation=enccs_training #SBATCH --time=00:20:00 #SBATCH --nodes=1 #SBATCH --ntasks=1 #SBATCH --cpus-per-task=128 #SBATCH --mem=0 #SBATCH --partition=standard source /project/project_465000321/environment.sh qp_srun fci COH2.ezfio > COH2.fci.out
Note that QP uses shared memory parallelism with OpenMP. It will be efficient using all possible threads on the node.
2 Example of a CHAMP job script
For CHAMP, you should request one or many nodes with a certain number of processes (as there are a certain number of cores per node). There each MPI process uses a single thread.
An example of a CHAMP Variational Monte Carlo job script would be:
#!/bin/bash #SBATCH --job-name=champ # Job name #SBATCH --output=champ.o%j # Name of stdout output file #SBATCH --error=champ.e%j # Name of stderr error file #SBATCH --partition=standard # Partition (queue) name #SBATCH --nodes=1 # Total number of nodes #SBATCH --ntasks=128 # Total number of mpi tasks #SBATCH --mem=0 # Allocate all the memory on the node #SBATCH --time=0:20:00 # Run time (d-hh:mm:ss) #SBATCH --mail-type=all # Send email at begin and end of job #SBATCH --account=project_465000321 # Project for billing #SBATCH --reservation=enccs_training # Reservation for training source /project/project_465000321/environment.sh export PMI_NO_PREINITIALIZE=y # CHANGE THE FILE NAME INPUT=vmc_input.inp OUTPUT=${INPUT%.inp}.out # Check that the file exists if [[ ! -f $INPUT ]] ; then echo Error: $INPUT does not exist. > $OUTPUT exit -1 fi srun /project/project_465000321/champ/bin/vmc.mov1 -i $INPUT -o $OUTPUT -e error
An example of a CHAMP Diffusion Monte Carlo job script would be (along with the VMC calculation):
#!/bin/bash #SBATCH --job-name=champ # Job name #SBATCH --output=champ.o%j # Name of stdout output file #SBATCH --error=champ.e%j # Name of stderr error file #SBATCH --partition=standard # Partition (queue) name #SBATCH --nodes=1 # Total number of nodes #SBATCH --ntasks=128 # Total number of mpi tasks #SBATCH --mem=0 # Allocate all the memory on the node #SBATCH --time=0:20:00 # Run time (d-hh:mm:ss) #SBATCH --mail-type=all # Send email at begin and end of job #SBATCH --account=project_465000321 # Project for billing #SBATCH --reservation=enccs_training # Reservation for training source /project/project_465000321/environment.sh set -e # the script will exit if a command below fails export PMI_NO_PREINITIALIZE=y # CHANGE THE FILE NAME VMCINPUT=vmc_input.inp VMCOUTPUT=${VMCINPUT%.inp}.out # CHANGE THE FILE NAME DMCINPUT=dmc_input.inp DMCOUTPUT=${DMCINPUT%.inp}.out # Check that the files exist if [[ ! -f $VMCINPUT ]] ; then echo Error: $VMCINPUT does not exist. > $VMCOUTPUT exit -1 fi if [[ ! -f $DMCINPUT ]] ; then echo Error: $DMCINPUT does not exist. > $DMCOUTPUT exit -1 fi # Launch MPI code srun /project/project_465000321/champ/bin/vmc.mov1 -i $VMCINPUT -o $VMCOUTPUT -e error cat mc_configs_new* >> mc_configs rm mc_configs_new* srun /project/project_465000321/champ/bin/dmc.mov1 -i $DMCINPUT -o $DMCOUTPUT -e error rm problem* walkalize* rm mc_configs_new*