可不可以/如何在服务器集群上并行执行pyALPS用以加快计算速度？Can we /How to run the pyALPS code parallelly in the HPC cluster for speeding up the calculation?

[Catch1123]

I'm trying to write scripts to run the alps code parallelly, but it may not work. How can i do? Thanks!

[Ooolab]

To run alps codes parallel, please install the alps code from source by following the instructions page:
https://alps.comp-phys.org/documentation/install/source/
If you run into any problems, please let us know.

[Catch1123]

To run alps codes parallel, please install the alps code from source by following the instructions page: https://alps.comp-phys.org/documentation/install/source/ If you run into any problems, please let us know.

Oooo! Thanks for your reply so quickly. if I have to use pyALPS, any suggestions for me?

like:https://alps.comp-phys.org/documentation/intro/runalps/usepython/.

Because I'm already familiar with serial via python in my personal PC and all my code is based on python. And recently, I must use the HPC cluster to simulate more big system and to get more accurate result. When I modify my code according to the website above
replacing:
pyalps.runApplication('mps_optim',self.mps_optim_params())
to
pyalps.runApplication('mps_optim',self.mps_optim_params()，MPI=self.mpi)
where self.mps_optim_params() is my function to generate the input file, and self.mpi is a number = 4,
and submit my .slurm file:
`#!/bin/bash
#SBATCH -J SpinChain_L8_MPI_test
#SBATCH -N 1
#SBATCH -n 4
#SBATCH --ntasks-per-node=8
#SBATCH --cpus-per-task=1
#SBATCH --partition=normal,normal1
#SBATCH -t 10:59:59
#SBATCH -o %j.out
#SBATCH -e %j.err
#SBATCH --exclusive
#SBATCH --hint=nomultithread

module purge
module load openmpi/4.1.1

export OMP_NUM_THREADS=1
export OPENBLAS_NUM_THREADS=1
export MKL_NUM_THREADS=1
export NUMEXPR_NUM_THREADS=1

cd /public1/home/kzyin/SpinChainProject/Scripts/7.1
srun -n 4 python L8.py`

Then I will get the output as attachment files. if you have any suggestions regarding this issue, it would be very much appreciated!^_^
6619496err.txt
6619496out.txt

[Catch1123]

` def model_params(self):

    model = OrderedDict()
    model['LATTICE_LIBRARY'                ] = '../lattice/my_lattice_{}ops.xml'.format(self.L)
    model['LATTICE'                        ] = 'chain lattice with special bond and site  '
    model['MODEL_LIBRARY'                  ] = '../model/Heisenberg_Chain.xml'
    model['MODEL'                          ] = 'spin'
    model['CONSERVED_QUANTUMNUMBERS'       ] = self.conservation
    model['Sz_total'                       ] = self.Sztotal
    model['MAXSTATES'                      ] = self.Maxstates
    model['COMPLEX'                        ] = 1
    model['local_S'                        ] = '2.0'
    model['local_S0'                       ] = '2.0'
    model['J0'                             ] = 1.0


    model['D'                              ] = self.D
    model['alpha'                          ] = self.alpha
    model['Gamma'                          ] = self.gamma
    model['J2'                             ] = self.J2

    
    return model
  

def mps_optim_params(self):

    ground_parms = deepcopy(self.model_params())
    ground_parms['J1'                           ] = 1.0
    ground_parms['SWEEPS'                       ] = self.Sweep

    #measurements
    ground_parms['MEASURE[Entropy]'             ] = 1
    ground_parms['NUMBER_EIGENVALUES'           ] = self.NumerEigen
    ground_parms['MEASURE_LOCAL[local magnetization]'] = 'Sz'
    ground_parms['MEASURE[Renyi2]'                ] = 1
    ground_parms['entanglement_spectra'           ] = str(self.L/2)
    ground_parms['ALWAYS_MEASURE'                 ] = 'Entropy,Renyi2,entanglement_spectra'

   

    auto_dir(self.path+'/log.txt')
    input_file = pyalps.writeInputFiles(self.path+'/ground',[ground_parms])
    return input_file

def run_mps_optim(self):

    if self.mpi == 0:

        pyalps.runApplication('mps_optim',self.mps_optim_params())

    else:
        pyalps.runApplication('mps_optim',self.mps_optim_params(),MPI = self.mpi)


    data_ground = pyalps.loadEigenstateMeasurements(

        pyalps.getResultFiles(prefix=self.path+'/ground'),

        what = ['Entropy','local magnetization']
        )
    
    return data_ground`

This is a part of my main code. Thanks!!!!!!

[Ooolab]

The binary version pyalps was compiled with a single CPU support, so it does not support the MPI or openMPI run. If you try to run it with MPI, it will simply run the same calculation N times (N is the number of CPUs you allocated for the job), which is not the intended behavior of the parallel version of the ALPS. So please try to compile the ALPS source code with the MPI or openMPI flags in your cluster. If you could consult it with your cluster administrator, they could potentially help you with the installation. Please direct them to our source installation instructions page.

[Catch1123]

Ok! I see. Thank you very much. ^_^! 👍

[Ooolab]

Please keep us posted if the source installation gives you any trouble.