2021-06-08 SB/JW/DM Workshop prep

Benchmarking figures?

• JW – Do we want the private set results?

  • (General) – No

One slide deck or two?

One – Easier to assign DOI. Can switch presenters midway trhough.

On Thursday, we’ll append the slide decks and assign numbering

Intro topics

• Workflow diagram –> forcebalance and evaluator

  • JW – Do it make explicit reference to openforcefield-forcebalance? This has been replaced by bespokefit for sage

  • DM will grab most up-to-date version of this figure

• Sage release candidate is available

  • (General) – We should make a conda package for 2.0.0rc1, which would also include 1.3.1.

  • How much should we publicize the 1.3.1 release?

    • Not much, we could just mention it in the talk.

    • SB will make a post on #general when the 1.3.1 release package is available

  • How should we publicize 2.0.0rc1?

    • SB – I’d like to write a big blog post (or series) about it.

    • DM – PB could be a good writer for Sage paper or other FF fitting

• 1.3.1 release if it’s been cut

• System announcement

  • Say it’s coming online in a few months, and we want to drive most traffic through it.

• Emphasis on collaboration – Working with founding labs, and cole group, and roitberg group, and della corte group, Venkat (cresset?), and participating companies (how could we find their individual sites?).

• Emphasize nonlinear/nonserial view of FF development – Many ideas are incubating, some of them get “merged” back into the main line of FFS, though many take longer or die off.

  • DM – narratives like modified seminario and bespokefit illustrate how different efforts can get merged into our operations, even when they’re not strictly intended for that.

How to show benchmark?

• JW – It’s a bit strange to talk about the benchmarking infrastructure in the infrastructure part, and not talk about results until later

  • SB – My plots didn’t use benchmarking infra

  • DM – Might be good for JW to talk mostly about benchmarking infra, and then quickly show a plot. Emphasize the ability to DO this sort of benchmarking, rather than any particular results.

•  

Mention internal adoption of our tools

• Venkat (Cresset) is using fragmenter, YTZ is using recharge, OpenEye using OpenFF FFs, Genentech is doing further benchmarks with our QC drivers.

  • JW will mention this in infrastructure slides.

PL benchmarking slide

• DM – If you don’t think the performance different is significant, then it’s fine to keep showing the slide, and just not fraw attention to the results in the slide text.