/PESMan

PESMan - a program to manage global PES calculations

Primary LanguagePython

PESMan

A program to manage global ab initio Potential Energy Surfae (PES) calculations using MOLPRO1 quantum chemistry software.

Basic commands:

Every features of this package is controlled through the different subcommands of the PESMan.py code,

  • Use PESMan export to export job.
  • Use PESMan import to import job after successful run.
  • Use PESMan zip/unzip to archive or extract job folders.
  • Use PESMan delete to delete any geometry/calculation from database
  • Use PESMan status anytime to check current status of calculation

Initial Set up:

  1. Create Molpro template (*.template) files for jobs.
  2. Modify pesman.config file with necessary information and template names.
  3. Modify geometry.py file according to the system details and create the object geomObj.
  4. Modify CreateNewDbs.py corresponding to the system details and grid required and run to create the databases relevant to the system type.
  5. Modify ReadResults.py file for reading results from databases.
  6. Run python PESMan.py addcalc to add the calculation defined in the pesman.config.
  7. Now we are ready to start the ab intio jobs. For first time run python PESMan.py export -cid 1 -gid 1 -sid 0 to export 1st geometry for calcid 1 without any starting geometry (-sid 0). Run the job file (RunJob*.py) created inside the ExpDir. Upon successful run import it with PESMan.py import -e <path to export.dat>
  8. Now the RunManager script can do the calculations for all the geometries sequentially. For first geometries in multi calculation it has to be done one-by-one sequentially to accumulate enough seed geometry, after that it can be run in bunches. Be careful about the depth value in the script. Tweak the different parameters in RunManager to get better performance. Note, RunManager also calls the ReadResults script to periodically to extract the results from database and store it in a datafile.
  9. Manual export/import can also be done. Run ./PESMan.py export -cid 2 -j 1000 -n 3 -par to parallel (3 process) export 1000 for calc id 2 > Run > import as PESMan.py import -e <export.dat file> -ig wfu -n 3 -zip -del. -ig wfu ignores wavefunction files during import. But its recommended to use the RunManager script instead.
  10. Run ./PESMan.py status at any moment to check the current status of the calculation.

Notes & Tips:

  1. Molpro calculation of Calc Ids other than 1 usually doesn't depend on other geometries so, its more efficient to run those jobs in parallel. Specify the option -par during export. Number of parallel processes is taken from the pesman.config
  2. If no -par flag is provided or the RunManager is used in serial mode, then proc value in pesman.config is used to run the Molpro itself in parallel using its internal MPI implementation.
  3. Use -zip preferably while importing to archive the data while keeping in GeomData, this will slow down the import/export process, but it heavily reduces file number and sizes. If you don't want to slow down the export/import, you can do the archiving altogether at the end too, using ./PESMan.py zip -all.
  4. Though the neighbor database is created to keep the distances between neighbors, the database itself is never used in any of the calculation.
  5. RunManager saves the mcscf iteration numbers in IterMultiJobs.dat, and only imports if its below 39 iteration.
  6. The RunManager.py is used to automatically export/run/import one after other. This can be done with a single job at a time or a bunch of jobs per iteration. As the mcscf is run using information from neighboring geometries, it is advisable to run those, for initial few jobs in serial. other wise this script can be invoked to run all jobs in bunches of jobs in parallel
  7. The RunManager depends on the successful run of the neighboring geometries to export a particular job. If a large number of jobs are failed or the database doesn't have enough neighbor information the RunManager sequence can break. In that case, RunManager searches globally for the closest neighbor pair of incomplete jobs, and performs a brute-force export to keep the job going. If that fails, a new job has to be exported manually.
  8. If some calculation has to be removed, Run the ./PESMan.py delete with calcId and geomId to remove that from the database.
  9. To add some new geometries to an existing data base, re-run the CreateNewDbs.py by providing the new complete list of full geometries. The script will check for existing geometries in the data base and update it with new additional geometries. This will modify all the neighbor geometry information, but won't interfere with any existing results in database.
  10. If some calc is semi successful i.e. if like mrci energy and ddr are done in a single calc and energy is successful but nact is failed, then run the collectFailed.py by providing the export.dat as a command-line argument to store the semi successful jobs in a new table in database. This script will also store the result in datafile.
  11. Running python PESMan.py addcalc for first time will add the calc names and template to the database. If you run the same again with same calc names, PESMan instead of inserting a new calcinfo, will update the existing info template

Footnotes

  1. https://www.molpro.net/ ↩