Can you consistently reproduce showcase charge assignment bug? I’m losing my sanity.
JM – Was 100% reproducible on my local machine - Did it like a dozen times. I think it was reproducible in CI too but I’m not 100% sure.
JW – Starting at this from a high level - This is important until NAGL becomes the primary charge assignment method, which will hopefully be in the next few months. So this can be kind of a band aid solution.
…
We’ll try using OMP_NUM_THREADS=2. JM will open a PR to docs to try CI with this change. JW will update the bandaid-1842 branch of OFFTK to have this change.
We should add “making sure openff-docs CI is green” a top priority for you
JM – Agree
JM –
NAGL can’t charge proteins - Memory usage explodes and it never finishes
JW – Hm, you should open an issue for this. Probably due to tautomer enumeration
JM – Will do. I’ll also find a smaller protein
RNA example working from PDB - Now uses pre-prepared solvated PDB
JW – Looks great
JM – Could load using custom substructures API?
…
JW – I wouldn’t recommend it. This is all pretty half baked right now.