Benchmarks – need to still re-filter for, e.g. 3-membered rings
Some benchmarking notes --
IB could use more docs
Dask parallelization doesn’t have helpful logs, so crashing as part of Interchange took a while to debug. Interchange creation uses partial charge generation
Ab Initio work
Absolute energies trigger FB errors – moving to relative energies, working on atom reordering
LW: are you converting just opt geo targets to AI or TD as well?
AMI: just opt-geo for now
LW: can add on TD data to opt-geo if you need
Optimization geometry tolerance – investigating threshold that resulted in YAMMBS
Putting together hypervalent sulfur dataset
OptimizationDataset
Might be easier to create your own smaller molecule if molecules are too large