How is convergence checked in VOTCA and how to parallelize simulations?

[Satyen Dhamankar]

Hello,

I am running some very simple SPCE water simulations in VOTCA. I am testing out IBI. 

I have to primary questions:

1. How do I check the extent of convergence of the IBI?

This my settings.xml file: 

<cg>
  <!-- example for a non-bonded interaction entry -->
  <non-bonded>
    <!-- name of the interaction -->
    <name>CG-CG</name>
    <!-- types involved in this interaction -->
    <type1>Bead1</type1>
    <type2>Bead1</type2>
    <!-- dimension + grid spacing of tables for calculations -->
    <min>0.0</min>
    <max>0.9</max>
    <step>0.01</step>
    <inverse>
      <!-- target distribution (rdf)  -->
      <target>CG-CG.dist.tgt</target>
      <lammps>
        <!-- name of the table for lammps run -->
        <table>CG_CG.pot</table>
        <!-- nm to angstrom -->
        <scale>10</scale>
        <!-- kJ/mol to kcal/mol -->
        <y_scale>0.239006</y_scale>
        <table_begin>0.20</table_begin>
        <table_bins>0.002</table_bins>
      </lammps>
    </inverse>
  </non-bonded>
 
  <!-- general options for inverse script -->
  <inverse>
    <!-- votca units 300*0.00831451 kJ/mol -->
    <kBT>2.47772398</kBT>
    <!-- use lammps as simulation program -->
    <program>lammps</program>
    <!-- lammps specific options -->
    <lammps>
      <!-- lammps script to run !-->
      <script>cg.spce.in</script>
      <!-- topology to be used by  csg_stat !-->
      <topol>cg.spce.data</topol>
      <!-- traj file created by lammps !-->
      <traj>traj.dump</traj>
    </lammps>
    <initial_configuration>maindir</initial_configuration>
    <!-- these files are copied for each new run -->
    <filelist>cg.spce.data cg.spce.in</filelist>
    <!-- do so many iterations -->
    <iterations_max>300</iterations_max>
    <convergence_check>
        limit>0.0001</limit>
    </convergence_check>
    <!-- ibi: inverse biltzmann imc: inverse monte carlo -->
    <method>ibi</method>
  </inverse>
</cg>

In <inverse> I have plugged <convergence_check> and put a limit of 1e-4, although, I have no idea how it is being processed by the system. 

2. How do I make sure that LAMMPS is multithreading the process? 

I am running these CG simulations with LAMMPS, and I want to make sure that I am using all the threads that are available to me. For a usual lammps simulation, I would do 

srun --ntasks=96 lmp_exe -in in.simulation

But how do I translate the above resource allocation to csg_inverse or csg_fmatch? 

Thank you for taking the time to read my question! 

[Christoph Junghans]

You will need to add the "convergence" task in the post_add list of each interaction you want to consider, that is missing in your listing above.

2. How do I make sure that LAMMPS is multithreading the process? 

I am running these CG simulations with LAMMPS, and I want to make sure that I am using all the threads that are available to me. For a usual lammps simulation, I would do 

srun --ntasks=96 lmp_exe -in in.simulation

But how do I translate the above resource allocation to csg_inverse or csg_fmatch?

csg_fmatch has a --nt option and for csg_inverse it is usually automatic, but can overwrite the number of threads to use in the settings file. Additional you can tell Votca to run lammps with a specific command.

For srun submission you need to allocate the resources, but then run csg_inverse serially and then override the number of threads in the XML settings file and use a special lammps command.

Christoph 

 

Thank you for taking the time to read my question! 

--
Join us on Slack: https://join.slack.com/t/votca/signup
---
You received this message because you are subscribed to the Google Groups "votca" group.
To unsubscribe from this group and stop receiving emails from it, send an email to votca+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/votca/40cc6de3-c363-4f8a-91ff-4a484c31f983n%40googlegroups.com.