openforcefield / openff-benchmark

Comparison benchmarks between public force fields and Open Force Field Initiative force fields

MIT License

10 stars 2 forks source link

Prod envs now use openforcefield 0.8.4 #66

Closed dotsdl closed 3 years ago

dotsdl commented 3 years ago

Description

Version bump for the openforcefield toolkit 0.8.4rc1 -> 0.8.4.

Shouldn't be much different than 0.8.4rc1, so shouldn't be dramatic.

Status

[x] Ready to go

dotsdl commented 3 years ago

Something about 0.8.4 appears to have broken the tests. Based on the differences between 0.8.4 and 0.8.4rc1, is it obvious what may be causing this @j-wags?

dotsdl commented 3 years ago

CI runs on master appear to work fine still: https://github.com/openforcefield/openff-benchmark/actions/runs/647641024

j-wags commented 3 years ago

Looking into this now. Apologies for the delay.

j-wags commented 3 years ago

Probably an interaction with this change. Setting up a dev environment to resolve it.

j-wags commented 3 years ago

Ok, I found and fixed two problems:

OFFTK 0.8.4 pre-populates the mapped SMILES in the extras dict when it runs offmol.to_qcschema. openff-benchmark then tries to set it again using the same mapped SMILES, which causes an error. So I added a try/except that should make it work for 0.8.4rc1 (which doesn't pre-populate the mapped SMILES) AND 0.8.4 (which does pre-populate it). This fix seems to work on my local machine. @dotsdl, could you sanity-check this?
SQLAlchemy 1.4 came out a few hours ago and is causing tests to fail. I don't know much about our use of that package. So I've pinned to 1.3.

codecov-io commented 3 years ago

Codecov Report

Merging #66 (a87a8ea) into master (6c37648) will increase coverage by 3.45%. The diff coverage is 100.00%.

dotsdl commented 3 years ago

Ah, thank you for this @j-wags! Reading your resolution it's clear I should have dove into it further myself, since I was directly involved in those changes.