0.15.0 release?

[mikemhenry]

After we get #391 and #392 merged in, how do we feel about making a new release @epretti @mattwthompson ?

[epretti]

I have also been thinking that it might be good to do a new release soon. Given that several force field-related updates have happened since the last release (we now distribute ff19SB, and CHARMM has been updated to the 2024 release), it might be worth calling it 0.15.0.

A question remains of whether or not we want to wait until we can include Lipid21 (#390). It should basically be ready but we decided we want to do more validation on it (and the other Amber force fields as well). This week I am working on #393 which should give us more confidence in the conversion. It shouldn't take too long, although from experience with these force field conversions, I am likely to find additional problems that had previously gone undetected, delaying things by an unknown amount.

[mikemhenry]

I agree that we should call it 0.15.0. Do you think we can do the release tomorrow? I can click all the buttons and get it onto conda-forge. I am happy to cut a 0.15.1 as soon as #390 and/or #393 get merged in. I want to get the release out the door sooner than later since #392 has a fix needed for espaloma to work.

[epretti]

Talking to @peastman yesterday, we thought we might want to wait until #393 is done, but that could be a while. If you want to get a release with the espaloma fix out, I don't know if we should do one that includes the generator-related fixes and somehow doesn't include the updated Amber force fields, or just do a release as is (including the updated Amber force fields but not the extended validation). @peastman may have an opinion on this; I am fine either way.

Before that, to get #391 in, I need to get #397 in, but discovered a bug with espaloma that I am preparing yet another PR for. Running the generator concurrently in multiple processes can cause failures because the code that downloads espaloma models to a temporary directory isn't careful to avoid some race conditions. Stand by for that PR coming soon.

[mattwthompson]

I think #391 is more than enough improvement to justify a release. In an ideal world, #393 would be done overnight, but that is potentially ballooning task that is not worth holding up a release IMHO

[mikemhenry]

Okay so to summarize:

PRs & Issues blocking release

• Remove DummySystemGenerator #397
• Fixes for issues and assorted cleanups to generators #391
• Intermittent failures with espaloma #399

PRs & Issues nice to have

• Validate Amber force fields against sander energies/forces #393
• Add Lipid21 #390

Is this accurate?

[epretti]

Yes, that sounds right. Once the #397 tests pass, I can merge it in, and then merge that back into #391.

I am working on #393, among other things, but @mattwthompson's prediction of it ballooning into something more complicated is correct, so I would say that if we wanted to do a release soon, let's not wait for #393 or #390.

[mikemhenry]

Perfect! Thanks everyone!

[epretti]

OK, #397 and #391 are in. I think I know what's going on with #399, and it's something that should really only be a problem for CI, but I'd like to fix that. If my hypothesis is correct, it should be easy to work around.

[mikemhenry]

Okay! Looks like we got everything done, I can get the release done tomorrow!