NaNs

LW:do we know the source of NaNs from benchmarks?

Finlay in the past found tons of NaNs, but these are hacks in place to avoid inflated DDE statistics so they’re not necessarily alarming (could be distinct from LW’s question)

• ​https://openforcefieldgroup.slack.com/archives/C03T3LLVC1J/p1757505091008569?thread_ts=1757451600.564919&cid=C03T3LLVC1J

LW: Unclear if NaNs are intentional or not

• FC mentions some of these come from ICRMSDs, mostly from Impropers. I’m wondering why

• MT – Urgency?

• LW – Medium to long term. Been using non-yammbs scripts for 2.3.0.

• MT – Ok, I’ll coordinate with Finlay. Will try to figure out how many are genuine NaNs.

• 
  

• 
  

•