JW – JM implied that there were less issues than previously thought in his check in this AM.
MT – Would be good to work through some details of changes he’s working on/advocating. I’m concerned about the number of lines changed in an “notebook refresh” PR.
JW – I’m thinking that the topology positions will be confusing initially, since interchange also has positons. But in the future we can drop one hand just have both APIs reference the same array.
MT – A possible alternative: What if Topology got moved to Interchange? That would leave Toolkit to focus on molecule representation and chemical perception.
JW – I’m broadly supportive, and we don’t have to pursue this immediately.
MT – Agree, I’ll want to think more about pros/cons.
Release prep
JW – I think the remaining things are:
icode handling
Showcase refresh (Will work with JMitchell)
PR/Issue clearance
Action on torsion handedness?
MT – I’ll open up an issue on openff-forcefields
JW – I still don’t understand how 90/270 phases are intended to work. But if they were useless I’d expect the corresponding k values to go to 0 in training.
MT – I’d be interested to see what happens if we relax the constraints on all of the starting parameters. MGilson mentioned doing the same thing two weekends ago.
Add Comment