General updates | JW – QCSubmit 0.53.0 out. Fixes a bug in how datasets are filtered. I’m not sure if you hold on to previously fetched+filtered datasets, but you might need to update those if they’re supposed to represent the current state of retrieval. Made some progress on converting ForceBalance to use modern GROMACS + pseudo-vacuum. BW is working on some sort of qca-dataset-submission-like GH runner for YAMMBS. Seems to be making very rapid progress. We’ve asked him to introduce it at today’s internal benchmarking meeting.
MT – Evaluator release - Updates for pint compatibility (evaluator uses pint NOT via openff-units, so there was an update needed there). Fixes a packaging quirk (a new build of AmberTools changed how packmol was selected, sets run_constrained: packmol=999 , which broke evaluator since it depends directly on packmol, so I removed evaluator’s packmol dep. Users should get an informative error message if they fall through the cracks).
Re: evaluator on AWS - I was able to get integration tests started, but they were failing due to what is probably changes in upstreams. Also there are YANK protocols that are now broken, which I removed. Still some issue with client/server architecture in Evaluator. I don’t estimate it’ll be an easy fix. And I didn’t assess that this was blocking to the release (since it’s probably been broken a long time). Could I take Sep 2, 5, and 6 off?
|