GNN-based parameter assignment

• JH – Getting espaloma-style stuff working with evaluator.

• JW – We could work on getting this into smirnoff-plugins. Could have a working session where we pair together on this, and it could go into smirnoff-plugins. There may be bugs in this initial implementation and I wouldn’t guarantee that I’d be active in hunting and fixing them - you’d need to own this until it becomes mature enough to merge into the toolkit.

• DC – This would be a good direction.

• DC – Re: SB’s plans - Should I be looping him into discussions? Would he take feature requests?

  • JW – Yes to discussion, no to feature requests.

Bespokefit handoff stages

• Immediate - Acting as backup for user support

• Immediate - Reviewing/direct-merging "obvious maintenance" (like dependency update) PRs without your/Simon's involvement

• May 13 - Acting as primary for user support (but pinging you pretty frequently for hard questions)

• May 27 - Reviewing/direct-merging "bugfix" PRs without your/Simon's involvement

• June 27 - Reviewing/direct-merging "API extension" PRs without your/Simon's involvement

• July 15 - Making releases without your/Simon's involvement

• July 30 - Reviewing/direct-merging "API breaking and behavior changing" PRs without your/Simon's involvement

Release announcement drafting

• JW – Single file installers - I’ll try to get these available ASAP, but I don’t think they’re a precondition for a release announcement.

• JH – Two things I’m working one:

  • Nonessential: Split specifications for reference generation (opt using one method and then single point using another)

  • Weird bug with running torsiondrives in parallel, leading to a crash. This is triggered by asking for multiple torsiondrive workers. Root cause is something like making a child worker on a child worker. So this would be a single torsiondrive running multiple opts in parallel.

    • JH – I’m trying to track this down but I don’t know that it’s in my scope (it’s quite possible that the root cause is a software engineering thing)

    • JW – Gotcha. Since there’s a lot of uncertainty about what hte problem is, and because it may be in a different package, let’s disable the multi-worker behavior for now. If we fix it later then that will be good news for users.

  • JW – Will disabling this require a new release?

    • JH – No, we can handle this in docs

  • JH – I’d also like to merge

    • JH – Not clear who will review, now that SB has left OpenFF.

    • JW – I’ll task JMitchell with reviewing this.

  • DC – From the science advertising side, it would be good to say “here’s the release, and here’s the results you can expect to get”

  • JH – I’m working on a blog post that will show some results, expect to have that done soon/tomorrow.

    • JW – Diego or I can help review this/get it into PR form for the website repo.

• JW – I will draft release announcement, and keep a space for linking the blog post. We can iterate on this in the #bespoke channel.