2021-08-13 QC meeting notes

Participants

  • Ben Pritchard

  • @Joshua Horton

  • @Pavan Behara

  • @Chapin Cavender

  • @David Dotson

Goals

  • Updates from MolSSI

  • User questions/issues, new submissions

  • Science support needs

  • Infrastructure needs / advances

Discussion topics

Item

Presenter

Notes

Item

Presenter

Notes

Updates from MolSSI

Ben

  • No updates to share, ongoing education commitments; not working on QCArchive until after next week or so

  • Did install a bandwidth monitor; gathering data on ingress, egress in order to gauge future hosting

    • so far seeing about 20-30GiB egress a day, 10-15GiB ingress

      • ingress from managers mainly

      • egress likely from users

  • New QCArchive version will track how big individual requests are

Production QCA Managers

David

  • Now running with psi4 1.4, latest openff-forcefields and toolkit in production

Psi4 prod testing

David

  • qca-dataset-submission#221 introduces testing of prod testing of psi4

  • Trigger is broken, needs to trigger on PR

  • JH: perhaps expand to all prod envs; qcsubmit test suite will test with what it has, do many combinations of e.g. forcefields

Production on PRP

David

  • Been getting choked out due to competition from other deployments on PRP; may have slowed compute progress on our datasets

New submissions

Pavan

  • PB: will do a new submission next week

    • want a single point dataset as in #221, made an OptimizationDataset with maxiter 0

    • DD: qca-ds-sub#153 is an example of a basic dataset from @Simon Boothroyd

New submissions

Chapin

  • Working on TorsionDrive dataset for dipeptides

  • CC: would like to have separate torsiondrives of the same molecule, but they are getting combined by qcsubmit based on identity

    • would like them to stay separate

    • JH: best way would be to take an empty TorsiondriveDataset, fill it in manually; avoid the TorsiondriveDatasetFactory for this use case

Science support needs

Josh, David

  • Torsiondrive executor in QCEngine

    • DD: will ping Ben again for feedback there since it is blocking, will also weigh in; I think QCFractal TorsionDrive models make sense migrated to QCElemental, but Ben may have insights on what will work and what won’t from the QCFractal side

    • want to hear approval from him to advance

  • GridOptimization support in QCSubmit

    • JH: not familiar with GridOptimization dataset, never used it

    • DD: not clear to me if it’s closer to a TorsionDrive dataset or an OptimizationDataset in execution on server-manager

    • DD: will schedule a time to meet with JH to scope out what’s needed for GridOptimization dataset support in qcsubmit

      • JH: Tuesdays work well!

 

Action items

@David Dotson will expand prod env CI to all prod envs, use qcsubmit tests as coverage
@Chapin Cavender will try building a 2D TorsiondriveDataset manually for dipeptides, avoid use of TorsiondriveDatasetFactory
@Pavan Behara will try building a Dataset of single point calculations based on OpenFF Theory Benchmarking Constrained Optimization Set MP2 heavy-aug-cc-pVTZ v1.1
@Joshua Horton and @David Dotson will meet to scope needs for GridOptimization dataset support in openff-qcsubmit

Decisions

Â