Roundtable Updates | DH – Implementing changes requested in MT’s PLBenchmarks PR review. Working on data for benchmark set. Starting to write up benchmarking results, formulate conclusions. JW – I should review PLBenchmarks PR JW – Would like to learn more about GROMACS development / maybe join developers meeting. DH – I have my automation for running PMX – I’d prefer to leave the infrastructure for this under the GROMACS developers' jurisdiction.
MT – Worked on/wrapped up a very large number of PRs (Largely stuff from JW’s plate). Many issue/PR discussions (like GLOBAL_TOOLKIT_REGISTRY). Lots merged last week, a few more this week. Didn’t spent a ton of time on vsites. Unsure about timing for 0.8.0/vsites merge – Would be OK with doing an 0.7.2. JW – Would like to keep light pressure on getting VSites done, so TG can move forward with clear priorities. Pushing for VSites open for review at end of the week. MT – Exposed aromaticity model, changed TAGNAME, exposed toolkit version, change to return_topology, utils.structure removed, new error for antechamber not found. DD – Make minconda installer builder GH action? Python conda package builder? Python conda package builder = Update version umber to newest release → . build_local.sh Miniconda installer = After above, run conda constructor
MT – Regarding released package tests → We’re having trouble enuring that latest package is the one that’s installed, cloned, and tested. I could work this out using a bash script, but am having trouble translating it to GHA. Could I do this in a BASH script and have Jaime or David show me how to convert stuff over? MT – Connectivity has been good on Google Fi for the most part. Outage this morning, not sure why.
DD – ANI2x jobs submitted to QCA. Some infrastructure issues, need ot debug whether they’re specific to certain compute resources or general. Lifecycle handling can now handle dataset expansions (eg new basis sets or job types). Docker deployments and image building are automated – different environments for different QCF compute workers (permutations of dependencies installed for MM/QM/ML). Now focusing on improving reliability – Would like to reduce error rates, will work closely with BP on areas adjacent to this in QCEngine+QCF. JH is doing exploratory work for implicit solvent (PCM) and ESPs/wavefunction storage (in conjunction with HJ). TG is drafting dataset standards/ policy – Will help us handle datasets and operations consistently. JW – Any thoughts on feasibility of Fractal-izing OpenFF Evaluator? DD – Won’t be possible though QCEngine’s OpenMMHarness. DD – Using a similar architecture will be hard because of large required storage space. DD – To prioritize project infrastructure development, we could make a diagram like the “data quality” diagram of how different pieces fit into fitting/benchmarking. This new diagram would include the “quality” of each segment’s contribution to the final FF fit, and we could also consider how hard improvements to each part of the pipeline are. This would give us a good way to set 6-month/1-year goals for infrastructure development.
JW – Expecting large molecule refactor coming for biopolymer support. That’s why I opened large aromaticity PR on Friday. Looking at other refactors soon – Think that the “iron is hot” on the refactor during this period. Concerned about lack of cheminformatics knowledge (General) – We should figure out if there’s a good written curriculum for cheminformatics, and who we can consult if not. JW – MMOS will be a good place to ping eg. VMD and BSS developers and ask how their parsers treat different formats and the information they contain MT – Can we make a roadmap for molecule/biopolymer/general infrastructure refactors? JW – Working with PELE/Guallar group on OFF integration JW – Working on Amber FF port (last step in backbone terms) JW – Roadmap refactor. Will try to add due dates for pharma-guarnateed tasks
|