OpenMM-ML 1.6

This release takes advantage of a new feature introduced in OpenMM 8.5 that greatly expands what models we can support. Improvements include

• Support for FeNNol models, including the pretrained FeNNix-Bio1 models.
• Support for the pretrained Orb models.
• Support for the pretrained MACE-OMOL-0 model.
• Support for arbitrary models implemented with ASE calculators.
• Support for periodic boundary conditions with AIMNet2.
• Support for periodic boundary conditions with TorchMD-Net models.

What's Changed

• Converted MACE to use PythonForce by @peastman in #114
• Fixed device selection in MACEPotential by @peastman in #116
• Convert AIMNet2 to use PythonForce by @peastman in #113
• Support ASE Calculators by @peastman in #118
• Converted NequIP to use PythonForce by @peastman in #120
• Add support for FeNNix models by @epretti in #121
• Convert TorchMDNetPotential to use PythonForce by @peastman in #117
• Use numpy array for indices by @peastman in #124
• Skip ANI tests if torchani is not installed by @peastman in #125
• Enable CI by @peastman in #126
• Replace deprecated function in building docs by @peastman in #127
• Remove incorrect sorting of atom list by @epretti in #131
• Add support for Orb models by @epretti in #132
• Workaround for caching bug in torch.compile by @peastman in #134
• Converted ANIPotential to PythonForce by @peastman in #133
• Update README for 1.6 by @peastman in #122
• Update version number to 1.6 by @peastman in #135

New Contributors

• @epretti made their first contribution in #121

Full Changelog: 1.5...1.6

OpenMM-ML 1.5

This release adds several new pretrained models: AceFF (versions 1.0, 1.1, and 2.0), MACE-OFF24(M), MACE-MPA-0, MACE-OMAT-0(S), and MACE-OMAT-0(M). It also supports user provided models created with TorchMD-Net.

What's Changed

• implement TorchMD-Net potential by @sef43 in #107
• Support more MACE foundation models by @peastman in #109
• Default to TorchANI instead of NNPOps by @peastman in #110
• add aceff-2.0 by @sef43 in #111
• Update version number to 1.5 by @peastman in #112

Full Changelog: 1.4...1.5

OpenMM-ML 1.4

This release adds support for AIMNet2 and DeePMD potentials.

What's Changed

• Add deepmdpotential to openmm-ml by @dingye18 in #99
• Fix issue #100 by @lohedges in #101
• Better error message if torchani is not installed by @peastman in #102
• Improved documentation by @peastman in #103
• Fixed errors in documentation by @peastman in #104
• Support AIMNet2 by @peastman in #64
• Prepare for 1.4 release by @peastman in #105

New Contributors

• @dingye18 made their first contribution in #99
• @lohedges made their first contribution in #101

Full Changelog: 1.3...1.4

OpenMM-ML 1.3

This release adds support for NequIP and Allegro models.

What's Changed

• Update inputDict of MACEForce by @JMorado in #86
• Add support for NequIP models by @sef43 in #60
• Fix the workflow for rendering the website by @JMorado in #92
• Updated version number to 1.3 by @peastman in #93

Full Changelog: 1.2...1.3

OpenMM-ML 1.2

This release adds support for the pretrained MACE-OFF23 models, and for loading custom MACE models. It also adds several smaller features, such as the ability to create mixed systems that involve a CMAPTorsionForce, to add a CMMotionRemover to systems, and to automatically register potentials defined in external packages.

What's Changed

• Included CMAPTorsionForce removal from ML Region by @JMorado in #69
• Add Documentation by @sef43 in #65
• Remove file saving in ANIPotentialImp by @JMorado in #71
• Add documentation link to README.md by @RaulPPelaez in #70
• Add MACE model support by @sef43 in #61
• Documentation fixes and README.md update for MACE by @JMorado in #72
• Fix sphinx docs build issue by @JMorado in #74
• Created entry point for registering potentials by @peastman in #75
• Fixed error in periodic boundary conditions when using TorchANI by @peastman in #78
• Add removeCMMotion argument to createSystem by @JMorado in #80
• Updated version number to 1.2 by @peastman in #83

New Contributors

• @JMorado made their first contribution in #69
• @sef43 made their first contribution in #65
• @RaulPPelaez made their first contribution in #70

Full Changelog: 1.1...1.2

OpenMM-ML 1.1

This release fixes a bug in creating mixed ML/MM systems, and adds an option to use a single model out of the ensemble of ANI models.

What's Changed

• Added installation instructions to README by @peastman in #49
• Added modelIndex option by @peastman in #51
• Fixed indexing error creating mixed system by @peastman in #56
• Updated version number to 1.1 by @peastman in #58

Full Changelog: 1.0...1.1

OpenMM-ML 1.0

This is the first official release.