| |
|---|
Updates from Molssi | BP: QCF v0.15.8 released. BP: lots of operations in the next iteration moved server-side BP – Could use feedback from users on new portal interface (General) – Could loop in Josh Horton, Willa Wang, Heejune Park, Yuanqing Wang, DD – Let’s do this during the current meeting time slot. Do we want to do it weekly or every-other-weekly? BP – Weekly DD – PB, could you include the folks above on the invite for both meetings? Then we’ll prioritize reviewing the new QCPortal interface in these meetings. PB – Yes, I will add those people to the invite for future weeks. JW – I’ll tag those people in a post on the OpenFF #qcfractal channel.
BP: was looking at the base result table, but it is one of the largest in the DB, because it holds extras, and extras holds QCVARS, and QCVARS has lots of stuff e.g. mayer indices these get duplicated currently as mayer_indices and mayer indices for backward-compatibility, but want to find out PB – Should we also only store them only for the final geometries? For optimizations we don’t consume wBOs at every step. JW – Could we keep forces but not store WBOs? Or is it all-or-nothing? BP – It’s all or nothing. But if I know which one of the meyer_indices labels you use, I can delete the other Â
BP: Acquiring new server with additional storage JW: I proposed to the OpenFF governing board to share the cost of storage, they didn’t accept or deny right away, so it would great to get a quote so I can begin a more formal proposal for the new server. BP – I’ve become aware of a grant proposal for an equipment supplement. It’d be due in March. Not sure if we should pursue that either in combination with speccing a server, or just do one or the other. Timeline is uncertain since if the proposal goes in in March, we may not know if we got it until May or later.
Â
  |
New submissions | |
Throughput status | OpenFF ESP Industry Benchmark Set v1.0: 6748/56054 ~12% from last week Openff dipeptides torsiondrives v2.0: 23/26 TD complete from 19/26 last week SPICE DES370K Single Points Dataset v1.0: 323K+ from 234K+ last week ~ 93% from 67% DD – Per PE’s comments, PRP is dying due to memory issues on a lot of these jobs. So we’re relabeling/retagging them so they only get sent to higher memory workers. PB – I’m running jobs on UCI cluster, but they’re pre-emptible, and like 80% of them are killed before they hit the three-day mark. PB – With the new QCF release, am I clear to crete more pre-emptible workers and avoid previous issues (like scratch space fillup?)
SPICE Dipeptides Single Points Dataset v1.2: 12552/33K ~ 37% done
|
 | |
 |  |