2020-12-15 Meeting notes

QCSubmit update

• PB – I saw that JH fixed qcsubmit. When will this be pushed to production?

  • JH – By the end of the week

Excluding “light” torsions

• JH + HJ – Automated way to filter out methyl rotation? Should this go in QCSubmit or OFFTK?

  • Possible SMARTS could be [*:1]~[*:2]!@-[*:3]~[!#1:4]

  • JH + JW – Should go in qcsubmit, this could be nother item in the list for torsions_to_ignore in Molecule.find_rotatable_bonds. QCSubmit could have a flag for TorsionDriveDatasetFactory to ignore_hydrogen_torsions which would add something like the pattern above to the ignorelist.

  • JH – If I make more of these patterns, should they go in QCSubmit or OFFTK?

  • JW – Let’s keep this in QCSubmit itself for now. Though someday I’d like to have lists of functional group SMARTS in OFFTK.
    
    

Use of seasonal basis set

• HJ – Do I need ot wait for a QCSubmit release to use the seasonal basis set?

  • JH – Yes, only “heavy” basis sets are available now. Seasonal should be out later this week.

Following guidelines V3

• HJ – Working on a new dataset. QCSubmit doesn’t yet fill in the fields for standards v3. Should I submit new datasets as “v1”?

  • JH – These fields will eventually be auto-populated by QCSubmit. But for now let’s submit it as v1.

  • TG – Current standards don’t say what the field names should be, so JH, you have some freedom in where you store the data

  • TG – Should be labeled as v3 with the “noncompliant” flag, since we intend for this to eventually be compliant with v3

  • (General) – The molecule metadata isn’t mutable, so intiially labelling as “V3” will mean that we can’t change the existing fields.

    • (General) – SOME things are mutable, like metadata, but you need a QCA login. Or we need to build this functionality into qca-dataset-submission since it has a bot with a authentication token for QCA.

    • (General) – some things are IMMUTABLE on a databse-structure level (like, SQL immutability), other things are “KINDA IMMUTABLE”, like pydantic-derived objects, where they can be changed in some ways but not others, other things like metadata dicts are close to fully mutable.

Compute queue

• DD – TG: Could you spin up QCA workers at UCI again?

  • TG – Yes.