Page Comparison

Project updates

BW

• Past Next week

  • More benchmarks

    https://openforcefield.github.io/yammbs-dataset-submission/

    • v3 besmarts

    • null-looser-priors protein ff

    • null-0.0.3

  • QCA workers

    • 2x 1x Hessian datasets finished2x datasets still going (protomers)

    • lipid-maps-medium

    • lipid-maps-smalllarge (on virginia tech)

  • Organometallics dataset

    • JW gave go-ahead to submit mols that made it through the toolkit. Brainstormed how to handle CIF file

    • Brainstormed ways to bypass the toolkit and get data to QCArchive

  • yammbs/YDS torsions

    • working with MT to get torsion benchmarks running

    • MT question: where should we get a benchmarking TD dataset?

    Protein fits

  • Protein fits

    • check weights and resubmit if they aren’t equal

    • LW: to paste weights-checking script

  • nagl2

    • combine opt datasets into one

    • basic dataset from combined dataset

      • same props as mlpepper

      • save wavefunction

• Next week

  • More benchmarks

  • QCA workers

  • Second lipidmaps dataset

  • Organometallics dataset

  • yammbs/YDS torsions

  • Protein fits

Project

updates

AMI

• Past week

  • NAGL2 training optimization set

  • Clean up HPC3

  • Integrate S data fit into Sage 2.3.0 repo

  • Protein fit (Maybe approach 3?)

    • Approach III: training to FF data only

    • Looks like 9 iterations / 3 days or so

      • 0.005 for the opt geo, and 5.0 for TD

• Next week

  • Finish offboarding stuff

  • Hopefully get something done with NAGL2 benchmark set

    • Github link macro
      link https://github.com/openforcefield/proteinbenchmark/blob/main/proteinbenchmark/benchmark_targets.py

    • LW: in benchmark_targets.py – would recommend setting traj_length to 5 microseconds (in a new target) instead of 10, for purposes of speed. We should be able to extend the trajectory fairly easily later

      • paid GPU queue on HPC3, request a week or two

      • gpu

        • dmobley_lab_gpu

        • LW: to send BW a script

      • free-gpu

    • 3 replicates to start with

LW

• Past week:

  • Protein FF fits – looks like ~1 iter/day

    TDs: 10, OptGeo 0.01

    Mostly off sick

• Next week:

  • Evaluator on Kubernetes

    • Trying to optimize GPU/CPU allocations

  • Interchange validation

    • Liquid boxes

  Next week:

  • protein FF fits

  • Evaluator on Kubernetes– maybe passing off to Matt

  • Liquid simulation validation

    • thermostats are not regulating temperatures