2021-01-29 Benchmarking deployment check in

Tech support Qs

• Swope – s99F case sensitivity

  • The previous fix will have handled the optimize execute pathway, but the server-based pathway will hit trouble in qcsubmit. JH will make a new release of this ASAP.

  • DD reproduced issue locally

  • JH is making a qcsubmit release for this now. This process will also require some changes to openff-benchmark, which DD will review and merge.

• TF – Missing TFD entries

  • No follow up when files were requested

  • DD tried to reproduce locally, failed.

  • JW sees this issue when he does ethanol with one conformer

    • Is this because there’s just one conformer, or because there are no 4-heavy-atom torsions?

      • TF’s problem isn’t not the 4-heavy-atom thing

      • DD has some single-conformer molecules from his set, but tehy didn’t show this issue

      • We are unable to explain TF’s error and will need his files to debug.

• Schindler QM keyword mismatch

  • Case-sensitive?

  • DD is checking

  • Solved – Needed to have matched_ in the method name

Feature requests

• XL – Use ANI?

  • DD – No. Would substantially increase infrastructure complexity.

  • DD will reply on Slack

• Coverage report segmentation?

  • JW – I am strongly against this. GT and DH are the “scientific owners” for the project, so they should make the call.

• IC – Final report format?

Deployment decisions

• Should we make a new release from master immediately?

  • Is b3lyp-d3bj case correct in all steps?

• Change step numbering?