General updates | JW – Following up from openmmforcefields meeting - Would you be up to expand capabilities of interchange.combine assuming we’re NOT committing to entire validation effort. MT – combine is not the critical thing - from_openmm is also major player - lots of development possible independent of using openmmforcefields. I’m basically ok with anything related to those but openmmforcefields has a huge range of outputs and asking whether we can import and combine … If people have feature requests with from_openmm and combine, send them to OpenFF and it’s our job. If they have issues that are unrelated to those or can’t be reproduced then that’s less clear. JW – Would OpenFE sending us lots of well-formed bug reports along these lines be exciting or draining? MT – Neither, it’d be work, but it would end up getting done. Very unexcited about anything involving polarizability - we’re far from having any even initial work in that direction. MT – OMMFFs is not good - it’s good it’s not part of our infrastructure - I understand why OFE uses it but I think it’s a big liability in their stack. Not good that we’re at risk of inheriting their problem. JW – I’ll mention this as a shared view between you and I at lead team meeting. I think it’s bad that we’re inheriting this responsibility but it’s the PMs place to make this call. Â
Any objection to supervision/career growth meeting next Weds? I’ll share template (and any instructions) Friday. AmberTools stuff - I think you’re threading the needle perfectly (how long did emails to DCase take to write, out of curiosity?). My assessment is that we can stay on AT23 for some time longer if they don’t get 24 working, though will limit our python version range. For AT24 package builds, we can assist as able but they are primarily responsible. MT – First email took 20-30 mins, second was very brief. One thought is to get a bunch of engineers together to train DCase to use conda-forge again but not optimistic about outcome there. After my second reply I intended to be done with conversation. MT – Agree with putting aside AT24 altogether (we don’t care about new GAFF versions). there are 3 problems to getting AT23 working macos 13 compiler thing. Smoking duck guy and I messaged up to conda forge core, and they haven’t made progress. Source code of feedstock didn’t change, just something upstream. DSlochower identified an issue with antechamber a while ago - something with diagonsization - Issue is still open, may be related to builds In a tangent of a conversation, DSwenson asked what it might take to make a slimmed down AM1BCC package - I said 10-20 hours. I made no commitments about this, ball’s in their court now.
Â
Intel mac support - Lead team discussed - we’re interested in supporting this as far as our partners find it valuable. I’ll poll them at next ad board meeting. MT – Build/pacaging/env resolution issues are, to my understanding, related to AmberTools and DGL, but only for very particular permutations of platform/major package versions. I anticipate that workarounds will be easy. MT – I agree with this approach, won’t give significant pushback on anything the ad board says. I’d 3 or 4/10 like to drop intel macs.
Â
MT
|