Open Tara-Lakshmipathy opened 7 months ago
@jan-janssen
@Tara-Lakshmipathy To install LAMMPS with openMPI you can use conda install -c conda-forge lammps=*=*openmpi*
or alternatively for MPICH conda install -c conda-forge lammps=*=*mpich*
. The after the first equal sign the version number is specified and after the second equal sign the build string can be set, that is what we use to install the MPI parallel version.
To install the pyiron resources from conda use conda install -c conda-forge pyiron-data
.
@jan-janssen Thanks for the quick response Jan! I successfully installed the pyiron resources according to your instruction.
I also installed 2023.11.21 openmpi version of LAMMPS from conda-forge according to your instruction. However, there seems to be an issue with this. LAMMPS seems to default to one core with the following error message:
openmpi seems to launch correctly in my environment:
Would you have an idea about this issue? Do you think this is an openmpi issue, or would you suggest that I try with an earlier version of LAMMPS?
To me this sounds like a mismatch of the MPI interface, which version of MPI is commonly used on the BAM cluster? Did you try using srun to start your job?
This is not on the BAM cluster. I am trying to get a local installation running. I do have superuser privileges on the local workstation, so I'm open to any alternative versions of MPI that you would like me try.
Can you post the first 20 lines of your log.lammps? Above you post that you are using one openmp thread but it does not say anything about the number of MPI cores
Well, this is awkward. mpirun seems to work now. I'm not getting the error message from the other cores:
The first 20 lines also verify that 8 cores are being used:
I'm not sure what I did differently though... I just deleted existing job in pyiron and ran it again, and it worked. Thanks for the help till now. If I'm able to reproduce the error, I will let you know.
The conda installation of LAMMPS does not seem have the mpi executable version by default. Another issue is that conda installation of pyiron did not come with the resources folder. I had to manually install it from within the notebook using pyiron.install()