JW – Reviewed some PRs, lots of meetings. Wrapped up QuanSight contracting. Basically making sure that all permutations of OS/processor architecture/CUDA versions were available for builds. Ran into additional issues on Windows with nvcc availability (requires SDK install). Scopatz made most images that we need. Jaime is monkey-patching windows images to install full SDK and have nvcc. MT – IF there’s ana rea where I could meaningfully contribute, I’d like to work with Jaime. But I think OpenMM is domain-specific enough that I’d just get in the way. JW – Can help with benchmarking project if needed. Also will review on conformer generation CLI and propose changes/alpha release process.
PB – Working on WBO barrier height correlations. Looking this week at doing fits using ForceBalance. This week will look more closely at fitting/benchmarking datasets and whether they can be split into different series. Would like to touch base with Josh on charged molecules for fitting. Also looking at basis set benchmarking. DD – Drafted software approach for industry benchmarking project. November is implementation month, will reach out to JW as I start building. Also preparing a bunch of new datasets to QCA. Aiming to add more user tooling to QCSubmit – Like SMILES-based querying for torsiondrives. Would like TG to open PR with optimization-debugger to QCSubmit. Doing some election-related volunteering, will be out Tuesday afternoon and Wednesday. JW – I’m happy to fill effort gaps if JH/TG aren’t available to do it DD – I’m aware that they may not be fully available, will ping JW/BP/PB as needed.
MT – Spent more time coding than usual. Lots of Toolkit PRs, many were merged. Spent some more time on CLI, both on conformer generator and on energy-calculator. Worked on System object, figured out many of the easier things, thinking more about electrostatics/nonbonded. Working on different forms between valence and nonbonded terms, and how many interactions they describe Partial charges can be a sum of more than one thing, not sure how to represent. AM1BCC outputs aren’t derived from numbers in FF, considering how to handle that. Posted some code snippets to MMOSS – Would like more feedback on those. Preparing for unit discussions on Tuesday/Thursday
JW – Looking at getting a windows computer with nvidia GPU to be able to use WSL and test CUDA support. Ideas?
|