2023-02-28 Protein-ligand benchmarks meeting notes

DD : current sprint - ends 3/6

• architecture overview : PL Benchmarks on FAH - Architecture v6.drawio

  • JC – Are colors meaningful?

    • DD – Green is done, yellow is in progress, red is not yet started.

    • DD – (See recording starting around min 6)

  • JW – What’s minimum requirement for first runs? I figure that will let us start getting information about the next stuff to fix.

    • DD – All the items under release 0.1.0. Only a few items left there.

• alchemiscale 0.1.0 milestone

  • MH – Alchemiscale 93 – I’d forgotten to add an error if AWS region isn’t set, now I’ve added but need to do a little more testing.

    • DD – Sounds good. Please merge when you’re satisfied.

  • DD – Alchemiscale 95 – HMO is working on task lifecycle. We did an initial review of this yesterday, he’s actively working on some of these pieces. It’s in good shape, is implementing state statnsitions that are allowed throughout the stack. I’ll review once I’m satisfied with completeness.

  • IP – Perses 1066 – I worked on adding logging capabilities. On top of that, I added a basic yaml formatter for it (shared screen, see 14 minutes into recording). Log entries will have start time, duration, performance info.

    • JC – Is this yaml? Can’t you just capture the information as a data structure as a dict? Also estimated performance seems off.

    • DS – This isn’t valid yaml. Also, generally, for logging formatting it’s best to not force a format on people.

    • IP – I had a plaintext logger, but we had discussed it and I was told that it would be better to do a structured logger.

    • JC – It could be handy to have a performance data log that’s machine-readable.

    • DS – That could be a separate logger - So it could be a separate handler that only logs performance data in a structured way.

    • IP – Yeah, I can do both.

    • DD – Also, we have access to datetime started, datetime stopped, and estimated values. Could we just pull those out and put those into returned dictionary?

    • IP – I do have those returned

    • JC – I think this should return time and performance in ns/day.

    • (…)

    • JC – I’d suggest nesting things in this return dict, so there could be a “performance” subdict. Then the quantities in there could have units implicit in their names, and then be stored as ints or floats.

    • DD – Anything else you want feedback on?

    • IP – I need to do a bit of testing, will bring this back up when I need eyes on it.

    • JC – Target is to get Perses bugfix release out on Monday, which will include Open Force Field support.

  • JC – Perses 1128 – I went through this with RG on Thursday, seems fixable. There are some differences between RDKit and OpenEye that may make things incompatible/the backend switch a bit more complex.

  • DD – Alchemiscale 34 – Working on this this week. I’ve been doing a lot of reviews in the past few days but getting the synchronous compute service into the state we want will be my priority this week. This dovetails into HMO’s work on compute service identities, heartbeats, and having the compute service API have correct claim expiration.

  • RG – ExampleNotebooks 36 – I don’t need anything here. This isn’t urgent.

  • DD – Alchemiscale 88 – HMO is working on scope enforcement

  • DD – GUFE 143 – Relates to whether we should do FF assignment outside of settings, I’ll be exploring this.

  • DD – GUFE 144 – Should protocols not hardcode assumptions about component labels and instead use a mapping as the source of truth for correspendances between mols/atoms.

• coordination board : alchemiscale : Phase 1 - MVP

• updates on In Review, In Progress, and Available cards

JC – DD, tomorrow I’ll be talking to HMO and JS about how we’ll feed in alchemical jobs to feed ASAP. It’ll be midnight Phoenix time so I don’t expect you to join.

• DD – Please loop me in as an optional invite, I’ll join if I happen to be awake.

JC – Also, HMO’s other engagement will be ending, so my lab will be taking him on 100%.