Individual updates | DN Worked most of the time on the annual report. Expect to be done on Friday. Will check in with KCJ on Thursday but she’s offline this week so I don’t know whether that will happen. Thinking about what we can do as a project plan - JW and LW made “subway plots”, wondering whether these should be shown on the website instead of/in addition to- current roadmaps. CC – I like the idea of showing the subway plots on the website. What’s the pathway to making that change? DN – I think we should show the subway plots without a lot of details, since those will be fluid. So the WBSes and more detailed planning docs should have a lot of details but not be pushed publicly, and the subway plots should be high-level and be very public-facing. MT – I’m wondering about maintainability of this roadmap/subway plot - We have an issue with our public-facing planning. I think it looks confusing to people on the outside, even if we know internally what we’re doing. We lose a lot of time to re-expaining things that we know internally to external stakeholders, or correcting information that we’d previously posted but which is no longer correct. So I’d like to make sure that we commit to maintain these documents. DN – Agree. Keeping these things up to date is expensive. But I don’t think that the public-facing plans should be maintained super precisely (which is why I want to avoid detail in that version). In previous roles I’ve had my team make videos about projects plans, or have central explanations of big topics.
MT Was at scipy most of last week (Tues-Fri). Met DD, RGowers, CStern. Gave a talk, slides were longer than I had time for. Showed a slide on the WBO stuff, which was very interesting since Chaya was the session chair. Talk went pretty well, gave a high-level overview of the project. Questions: Someone asked about polarizability, said that our plans are fuxxy and far from produciton Question about similarity with OpenKim, I said that at a high level we look similar but deep down we’re in different spaces. Jan (from Los Alamos who maintains 400-500 conda-forge packages) asked about weighting between benchmarks and QM or phys prop data. Talked with DD and JW and I think we weight benchmarks mostly towards QM, though each of our benchmarking efforts has been a bit different.
Generally it’s a cool conference because industry people like QuanSight are major contributors/organizers of the conference. Enthought is being replaced as the main host by NumFocus. Anaconda is in a weird-ish position in this group, starting to become a bit more of an incubator rather than a conda maintainer/shop.
JW – OE license update. ~1 day of downtime on ~weds Working on annual report - New team, taking longer than before (in the words of DM “If you want I could write it this afternoon”) but important that we get this right Fixing some small biopolymer stuff - OE PDB writing, etc. F@H/protein dataset contingency planning Worked with JMitchell on bespokefit connectivity-rearrangement-finder logic. Booked for RDKit UGM - Berlin in October
CC Offline most of last week and all of the week before. Most of my work time was monitoring jobs and answering emails. Tripeptide torsiondrives are going well. One torsiondrive actually finished. Seems like each torsiondrive is nearly done and there are just a few more optimizations to finish in each. This is similar to what we saw with the other torsiondrives. DD – Agree that we’re into the tail end of these jobs. CC – Yeah, and we’re not getting the same errors every day. DD – It seems like we probably have optimizations for each scan point, even if the convergence criteria aren’t met. JW – Should we come up with some sort of automation/cicsistent criteria for incomplete torsiondrives? CC – LPW pointed out that many of the non-convergences (the rotating molecule issue) are a unique problem in psi4. LPW has a discussion on a thread in ff-biopolymers where he suggests fixes to GeomeTRIC that could work here, they’re just not implemented yet.
DD
|