Note

This release updates espaloma to support the newest version (1.1.2) of DGL. There are no API breaking changes or updates to the espaloma model. We have attached espaloma-0.3.2.pt to this release for convince, but this model is identical to espaloma-0.3.1.pt. espaloma-latest.pt is also the same as espaloma-0.3.1.pt. Please see https://github.com/choderalab/espaloma/releases/tag/0.3.1 for more information about the espaloma 0.3.x family of models.

What's Changed

Support latest (1.1.2) DGL version

Bug fixes, packaging, and documentation updates

Installation Guide Update by @mikemhenry in #175
Add docker and apptainer image building on demand by @mikemhenry in #176
Update to new dgl version (1.1.1) by @mikemhenry in #179
Add Instructions on loading local model by @ijpulidos in #183
Test & Update to newest DGL Version (1.1.2) by @mikemhenry in #186
Update readme and docs for version 0.3.2 @mikemhenry in #188

Full Changelog: 0.3.1...0.3.2

Contributors

ijpulidos and mikemhenry

Assets 4

14 Jul 21:48

kntkb

0.3.1

7165038

Version 0.3.1

What's Changed

Fix CI -- updated by @ijpulidos in #152
Support newer versions of OpenFF Toolkit by @mattwthompson in #155
Support for modern openff-toolkit versions (>=0.11) by @ijpulidos in #153
Support different improper torsion multiplicity by @kntkb in #157
Update default reference force field when creating a system from graph (fix #160) by @kntkb in #161
Add comments to why some of the input features were commented out by @kntkb in #163
Improved deployment testing with different charge-method options by @ijpulidos in #164
Changed default charge method to nn by @madilynpaul in #142
Fix loading molecules serialized with older openff-toolkit versions (<=0.11) by @ijpulidos in #166
Use rtd for docs by @mikemhenry in #167
Speed up and improve CI and update readme by @mikemhenry in #169
First pass at model fetch by @mikemhenry in #172
Use dgl.to_homogeneous instead of the old deprecated name by @ijpulidos in #173

New Contributors

@ijpulidos made their first contribution in #152
@mattwthompson made their first contribution in #155
@madilynpaul made their first contribution in #142

Full Changelog: 0.3.0...0.3.1

Model updates

Note that espaloma-0.3.0.pt is also deployed which is compatible with espaloma-0.3.1.pt. espaloma-0.3.0.pt was renamed to espaloma-0.3.1.ptto maintain consistency with the release version.
espaloma-0.3.1.pt (espaloma-0.3.0.pt) is a new espaloma model (see notes_for_espaloma-0.3.1.pdf for more details) with the improved features:
- Trained with curated quantum chemical dataset covering small molecules, peptides, and nucleic acids, containing over 15,000 unique molecules and 1M conformations
- Improved training strategy that uses both energies and forces, as well as L2 regularization to proper and improper torsion force constants to reduce the model complexity and variance
- Restrict improper torsion multiplicity to n=1,2

Notes

espaloma-0.3.1.pt (espaloma-0.3.0.pt) was trained and created using a dev version (commit: 4c6155b) of v0.3.1 release based on openff-toolkit 0.10.6
A manuscript describing the complete fitting process and assessment is forthcoming
Please warn that v0.3.0 is no longer compatible with v0.3.1, and espaloma-0.3.1.pt is not supported by v0.3.0

Contributors

ijpulidos, mattwthompson, and 3 other contributors

Assets 6

26 Apr 17:31

kntkb

0.3.0

c623ca1

Version 0.3.0

What's Changed

Get rid of large storage by @yuanqing-wang in #133
Fix nonbonded force subtraction by @kntkb in #134, #140
Exclude resonance variant input features by @yuanqing-wang in #148

Full Changelog: 0.2.4...0.3.0

Model Update

espaloma_0.3.0rc1.pt is deployed as a release candidate for espaloma_0.3.0 (see notes_for_espaloma-0.3.0rc1.pdf for more details).
Please note that notes_for_espaloma-0.3.0rc1.pdf are temporal and manuscript describing the complete fitting process and assessment is forthcoming.
Please warn that espaloma_0.3.0 is no longer compatible with the 0.2.x models, and vice versa.