globus-labs / mof-generation-at-scale / 8114677821

48%

push

github

Port Colmena application to Polaris (#52)

* Write results to disk as gzipped files

* Make the HPC configuration an option

* Add runscript for Polaris

* Adjust the number of workers accordingly

* Check for duplicate MOFs before accepting

* Add utilization logger

* Handle if NVML is not installed

* Launch logging during run

* Fix typo, add logging message

* Switch to using a non-mpi version of LAMMPS

* Clean up logging messages

* Switch to LAMMPS w/o MPI

We need to explicitly compile with out MPI to run
multiple LAMMPS tasks per node

* Move HPC into its own module

Avoid triggering imports from other components

* Save failed runs, use "-results" for colmena results

* Don't compress logs so we can watch during run

Also use the user-provided name of the HPC system
in the run dir name

* Increase the scale of the test

* Add scripts which plot utilization

* Add analyses for component tests

71 of 93 new or added lines in 2 files covered. (76.34%)

4063 of 9858 relevant lines covered (41.22%)

0.41 hits per line