2024-04-23 QCSubmit dataset submission meeting

QCSubmit 0.51.0 released

JW

• Compatibility with OFFTK 0.16 (and some fixes for tests that were broken by QCF 0.54, but those are just limited to tests)

• OFFTK used to allow a molecule to have no conformers either by setting an empty list, or None. It should be None, was broken, now fixed in OFFTK 0.16 (or at least, editable using public API so no crimes can occur)

• LW – If hypothetically I edit it using private API, what should i do

  • JW--set it to none, I’l let you know if/when we put the change to [] in the toolkit

Moving old datasets to “End of Life”

JW

• Many old datasets that haven’t been active for years--any objection to just moving them all to EOL?

  • No, we’ll move them

New write credentials

JW

• QCA write credentials are in LastPass--Lily and Jeff should both have access in case Jeff gets hit by a bus

  • LW--can see credential exists, but doesn’t have permission to use it

    • JW and LW fixed this later in meeting

  • Not the same as the worker credential--access to write and modify QCA itself

  • Need this credential to re-tag datasets, can’t use worker credential for that

• JW and LW stayed at the end of the meeting to debug XFF opt set tags. Turns out JW is stupid and the opt set was running just fine, and is in fact now done.

Moving Brent’s datasets to End of Life

JW/BW

• Bot moved one to archived because it fully finished--high resource node fixed GeomeTRIC seg faults

  • PB (slack) suggested maybe a bad node is causing seg faults? runs locally

  • Have hundreds of workers running, shouldn’t be possible for a “bad node” to be causing seg faults, since it keeps getting resubmitted to new workers. Could theoretically happen, but unlikely. Seems like memory was the fix, even though there wasn’t a memory error

• Submitted TD supplement to high-resource node in case there were memory issues, not hopeful that it will finish

  • Will this be useful without all the results? 875/878

  • BW--defer to Lily

  • LW--these errors won’t be helped by more resources, issue is linear angle errors. just EOL it