v2.2.0

What's Changed

• Update default_trainer.yaml by @stefaanhessmann in #664
• update tensorboard logdir and formatting by @ken-sc01 in #665
• Sh/elementwise statistics by @stefaanhessmann in #653
• Lammps PBC fix by @ken-sc01 in #670
• Emphasize where strain needs to be in inputs by @WardLT in #673
• fix dataset split into train/val/test by @Maltimore in #675
• cli.py: fix error message with correct recommendation by @Maltimore in #676
• Remove redundant SQLite connections by @lkkmpn in #686
• Minor tutorial fixes by @ken-sc01 in #683
• Sa/bugfixes by @sundusaijaz in #687
• [Bug] GroupSplit missing in QM7X dataset by @jamesETsmith in #689
• Ks/qm7x data config by @ken-sc01 in #691
• Added new API support for nextgen Materials Project by @sundusaijaz in #692
• enable changing wd during run time by @jnsLs in #694
• Jl/schnetpack requirements by @jnsLs in #697
• weights_only=False by default by @jnsLs in #701
• key_value_pairs as list by @jnsLs in #707
• fix bug in python version by @stefaanhessmann in #711
• delete verbose from lr_scheduler by @sundusaijaz in #710
• spkcalculator takes either model path or model torch.nn.Module by @jnsLs in #709
• [nit] : typo in base.py by @yu2C in #714
• Jl/split mismatch error by @jnsLs in #708
• self.conn is now defined before it is called by @jnsLs in #716
• Update getstarted.rst by @jnsLs in #720
• make spkcalc faster by @jnsLs in #719
• enabled jupyter black and jupyter notebook output is automatically removed by @jnsLs in #717
• [nit] typo in utils/__init__.py by @yu2C in #721
• SA/ Ensembler Calculator by @sundusaijaz in #713
• removed old painn and schnet models by @jnsLs in #725
• merge master -> dev by @stefaanhessmann in #726
• fixed dtype bug in calculator by @jnsLs in #727
• Jl/fixing tutorials by @jnsLs in #734
• Dev by @jnsLs in #735
• Bump protobuf from 3.20.2 to 4.25.8 in /docs by @dependabot[bot] in #736
• Sh/petnames by @stefaanhessmann in #738
• typo by @jnsLs in #740
• typo error by @Prism194 in #748
• Dev by @jnsLs in #769
• Update version to 2.2.0 by @stefaanhessmann in #770
• updated load function by @stefaanhessmann in #771
• Jl/update docs by @jnsLs in #772

New Contributors

• @ken-sc01 made their first contribution in #665
• @lkkmpn made their first contribution in #686
• @jamesETsmith made their first contribution in #689
• @yu2C made their first contribution in #714
• @Prism194 made their first contribution in #748

Full Changelog: v2.1.1...v2.2.0

v2.1.1

What's Changed

• fixes a bug: due to new python versions, the hydra config_dir was not found

Full Changelog: v2.1.0...v2.1.1

v2.1.0

What's Changed

• Update README.md by @jnsLs in #574
• windows issue with index tensors by @jnsLs in #579
• all tutorials on cpu by @jnsLs in #580
• fix ISO17 database by @stefaanhessmann in #582
• doc: Readme: various minor fixes by @Maltimore in #586
• fix typo: propery -> property by @Maltimore in #584
• requirements: add pytest-benchmark by @Maltimore in #585
• requirements.txt is linked to setup.py by @jnsLs in #588
• [WIP] README: remove outdated or less important information by @Maltimore in #595
• torch.testing: update deprecated assert_allclose by @Maltimore in #594
• hydra: add self to defaults list to avoid warning by @Maltimore in #593
• doc: put link to docs earlier in the readme, and add tests README by @Maltimore in #597
• provide batch size to self.log() of lightning module to prevent warning by @Maltimore in #598
• replace setup.py with pyproject.toml by @Maltimore in #596
• Unpin dependencies and cleanup by @Maltimore in #599
• Ignore model hyperparameters when saving because they are saved at checkpointing by @Maltimore in #600
• random split is default by @jnsLs in #606
• Implemented a stratified sampler by @jnsLs in #539
• added wandb logger by @jnsLs in #610
• quick model save patch by @jnsLs in #614
• Adding the functionality to define matmul precision for pytorch in configs by @khaledkah in #611
• Embedding spin multiplicity and charge based on SpookyNet implementation by @epens94 in #608
• fix backwards compatibility and clean up by @stefaanhessmann in #617
• fixed device (cpu) in qm9 tutorial by @jnsLs in #621
• Add tmqm dataset by @sgugler in #623
• Kk/qm7 x by @khaledkah in #616
• batchwise optimizer is deprecated by @jnsLs in #638
• test rtd by @jnsLs in #640
• if no seed is specified, it is chosen randomly by @jnsLs in #636
• Update spkdeploy by @jnsLs in #644
• restructure embeddings to avoid issues with torch jit by @stefaanhessmann in #634
• Jl/upgrade black by @jnsLs in #645
• added black workflow by @stefaanhessmann in #648
• Delete src/scripts/test_ase.py by @stefaanhessmann in #652
• md seed is initialized randomly if not specified by @jnsLs in #651
• drafting the load_model method with conversion between spk versions by @jnsLs in #646
• upgrade dependencies to numpy 2 by @stefaanhessmann in #654
• load model and upgrade version in spkpredict by @jnsLs in #655
• fixes bug in RemoveOffsets for intensive properties by @stefaanhessmann in #656
• update versions for release by @stefaanhessmann in #659
• fixes bug and updates to new simpy verison by @stefaanhessmann in #658

New Contributors

• @Maltimore made their first contribution in #586
• @khaledkah made their first contribution in #611
• @sgugler made their first contribution in #623

Full Changelog: v2.0.4...v2.1.0

v2.0.4

Updated documentation and requirements, minor bug fixes.

What's Changed

• Update citations of software papers by @NiklasGebauer in #528
• fixed md22 downloading method by @jnsLs in #530
• Jl/fix md22 by @jnsLs in #532
• Sh/batchwise optimizer by @Stefaanhess in #531
• raise Exception if array is passed for energy in calculator + typehint by @Stefaanhess in #533
• device can be specified when deploying models by @jnsLs in #544
• fix atomic mass for aspirine lammps example by @Stefaanhess in #545
• old field schnet data can be converted to new input dimension by @jnsLs in #550
• fixed calculator stress bug by @jnsLs in #552
• fixed training parameters by @jnsLs in #557
• url for uracil data has changed by @jnsLs in #558
• fixed device bug in tutorial 03 - clean by @jnsLs in #559
• updated pytorch-lightning requirement by @jnsLs in #560
• updated tutorial 1 by @Stefaanhess in #562
• data dimensions are updated in materials tutorial by @jnsLs in #564
• Update requirements.txt by @jnsLs in #566
• Update setup.py by @Stefaanhess in #572
• Update setup.py by @Stefaanhess in #573

Full Changelog: v2.0.3...v2.0.4

v2.0.3

What's Changed

• bug fix: per_atom_output_key in Atomwise by @Stefaanhess in #527

Full Changelog: v2.0.2...v2.0.3

v2.0.2

Updated pytorch-lightning, updated Lammps interface, enhanced batchwise optimization, general small bug fixes.

What's Changed

• improved batchwise optimization by @jnsLs in #508
• num_val_workers and num_test_workers can now be set to 0 by @NiklasGebauer in #515
• adapted to new pytorch-lightning version by @jnsLs in #517
• fix per_atom_output_key usage in Atomwise by @Vosatorp in #523
• data.load_properties can now be set to an empty list by @NiklasGebauer in #524
• allow to also use higher lammps versions by @Stefaanhess in #522
• updated batchwise calculator by @jnsLs in #525

New Contributors

• @Vosatorp made their first contribution in #523

Full Changelog: v2.0.1...v2.0.2

v2.0.1

Updated SchNetPack 2.0 release, which fixes a series of bugs and adds an interface to LAMMPS.

What's Changed

• Added materials tutorial to docs index by @mgastegger in #479
• fix bug in OMDB dataset by @Stefaanhess in #481
• Updated MD17 download urls by @mgastegger in #483
• Update default_run.yaml by @NiklasGebauer in #486
• Update README.md by @Stefaanhess in #489
• fix: LightningLoggerBase was removed in lightning 1.9 by @Stefaanhess in #491
• fix: add support for force models in spkpredict by @Stefaanhess in #492
• Sh/spkpredict by @Stefaanhess in #493
• fixed dtype of materiels project dataset by @Stefaanhess in #498
• check for legacy API-key in materials project by @Stefaanhess in #499
• reviewed the lammps doc files by @jnsLs in #500
• Sh jl/lammps by @Stefaanhess in #501
• Niklas gebauer docstring fixes by @NiklasGebauer in #502
• Fix bugs in MD module by @mgastegger in #503
• Removed deprecated arguments … by @NiklasGebauer in #504
• added schnetpack-gschnet extension to readme… by @NiklasGebauer in #505
• Updated README and fixed deprecated numpy dtype by @mgastegger in #506

Full Changelog: v2.0.0...v2.0.1

v2.0.0

Notes

This is the first release of SchNetPack 2.0 which uses the Hydra configuration framework, Pytorch Lightning and a new indexing scheme.
It also includes an improved data pipeline, modules for equivariant neural networks and a PyTorch implementation of molecular dynamics.

What's Changed

• Inital commit for v1 rewrite by @ktschuett in #267
• center transformations by @Stefaanhess in #271
• Mg/torch env by @mgastegger in #274
• Kts/qm9datamodule by @ktschuett in #275
• Training script driven by Hydra+Lightning by @ktschuett in #277
• Implement training of potential energy surfaces by @ktschuett in #278
• Add split file by @ktschuett in #282
• PaiNN representation by @ktschuett in #284
• Postprocessors and TorchScript by @ktschuett in #285
• Mg/symfuncs by @mgastegger in #287
• Initial API docs by @ktschuett in #293
• Fix problem with transforms in new lightning version by @ktschuett in #294
• stress and custom experiment by @Stefaanhess in #292
• Update docs (and fix postprocess bug) by @ktschuett in #295
• Mg/calculators by @mgastegger in #296
• Refactor models by @ktschuett in #297
• fix small bug by @Stefaanhess in #298
• Sh/datasets by @Stefaanhess in #300
• Proposal for MultiPropertyModel by @Stefaanhess in #299
• add epsilon to painn by @Stefaanhess in #301
• Some minor updates by @ktschuett in #305
• Fix API docs by @ktschuett in #306
• Unify model classes & refactor configs by @ktschuett in #309
• Fix ModelOutput package by @ktschuett in #310
• add bessel representation for future painn usage by @Divide-By-0 in #311
• Update examples & add data workdir by @ktschuett in #314
• Dipole moment & polarizability by @ktschuett in #316
• Dynamics caching neighborlist by @ktschuett in #320
• Dev by @jnsLs in #322
• Add long range cutoff by @ktschuett in #325
• Mg/md by @mgastegger in #315
• Refactor configs and add predict script by @ktschuett in #334
• Fix install bug by @ktschuett in #335
• Add automatic position derivatives by @ktschuett in #341
• Update QM9 tutorial by @ktschuett in #347
• ZBL Potential, Electrostatics (+Ewald summation) and stress tensor fixes by @mgastegger in #349
• Fixes to some MD desfaults and torchscript issues by @mgastegger in #350
• Disable automatic use of torchscript in MD calculators by @mgastegger in #351
• Response properties and field representations by @mgastegger in #339
• Fixed derivative graph settings for basic response properties by @mgastegger in #352
• Fixed sign for shift type cutoff by @mgastegger in #353
• Refactor AtomisticModel by @ktschuett in #354
• Fix bug when using ddp with set run.id by @ktschuett in #359
• Fix mixing bias by @ktschuett in #365
• Fix mixing residual by @ktschuett in #366
• fixed aggregation_mode bug for avg pooling by @jnsLs in #361
• Sh/ep device by @Stefaanhess in #360
• Fix AtomsDataSubset for use inside ConcatAtomsData by @chgaul in #357
• Add learning rate warmup & SGD config by @ktschuett in #370
• Fixed creation of subset in BaseAtomsData by @NiklasGebauer in #369
• Fix DDP training by @ktschuett in #372
• Updated for new yaml behavior by @mgastegger in #375
• Nwag/comment-update by @NiklasGebauer in #374
• Fix OMDB _convert dataset preparation by @bartolsthoorn in #373
• Fix pin_memory and some deprecation warnings by @ktschuett in #378
• Update tutorials and filter outputs by @ktschuett in #377
• Add conversion script for old datasets by @ktschuett in #380
• Fixed bug in map_properties by @mgastegger in #381
• Fix testing bc lightning API changed by @ktschuett in #382
• Updated MD docstrings and tutorial by @mgastegger in #383
• Wrapping of atom positions under PBC by @mgastegger in #385
• Fixed energy logging for multiple molecules by @mgastegger in #386
• Some refactoring and cleanup by @ktschuett in #387
• Retrained ethanol model for new postprocessing convention by @mgastegger in #388
• Added on_step_finalize in MD simulation hooks by @mgastegger in #390
• Updated weight init in FieldSchNet representation by @mgastegger in #391
• Added tmpdir functionality for MDs, fixed calculator bug by @mgastegger in #395
• Improved config loading for MD by @mgastegger in #396
• Fixed criterion for recomputing MD neighborlists by @mgastegger in #398
• Fixed using subset in ASEAtomsData.iter_properties by @NiklasGebauer in #393
• update deprecated code to new torch version by @Stefaanhess in #399
• Fix strain input module by @ktschuett in #403
• Add resolver for tmp directory by @NiklasGebauer in #406
• Fix #401 by @ktschuett in #404
• Added tempfile import for custom tmpdir resolver by @mgastegger in #408
• Added routine to NeighborListMD to properly filter out pairs due to the buffer region by @mgastegger in #409
• Removed n_out argument from DipoleMoment and Polarizability layer docstrings by @mgastegger in #411
• Fix PyTorch Lightning deprecations by @ktschuett in #414
• Consider only a selection of atomic forces in training, validation and testing, and ASE neighborlist with skin implemented by @jnsLs in #405
• Added a few classes util for structure relaxations (in particular MOMONANO) by @jnsLs in https://g...

v2.0.0 pre

What's Changed

• Inital commit for v1 rewrite by @ktschuett in #267
• center transformations by @Stefaanhess in #271
• Mg/torch env by @mgastegger in #274
• Kts/qm9datamodule by @ktschuett in #275
• Training script driven by Hydra+Lightning by @ktschuett in #277
• Implement training of potential energy surfaces by @ktschuett in #278
• Add split file by @ktschuett in #282
• PaiNN representation by @ktschuett in #284
• Postprocessors and TorchScript by @ktschuett in #285
• Mg/symfuncs by @mgastegger in #287
• Initial API docs by @ktschuett in #293
• Fix problem with transforms in new lightning version by @ktschuett in #294
• stress and custom experiment by @Stefaanhess in #292
• Update docs (and fix postprocess bug) by @ktschuett in #295
• Mg/calculators by @mgastegger in #296
• Refactor models by @ktschuett in #297
• fix small bug by @Stefaanhess in #298
• Sh/datasets by @Stefaanhess in #300
• Proposal for MultiPropertyModel by @Stefaanhess in #299
• add epsilon to painn by @Stefaanhess in #301
• Some minor updates by @ktschuett in #305
• Fix API docs by @ktschuett in #306
• Unify model classes & refactor configs by @ktschuett in #309
• Fix ModelOutput package by @ktschuett in #310
• add bessel representation for future painn usage by @Divide-By-0 in #311
• Update examples & add data workdir by @ktschuett in #314
• Dipole moment & polarizability by @ktschuett in #316
• Dynamics caching neighborlist by @ktschuett in #320
• Dev by @jnsLs in #322
• Add long range cutoff by @ktschuett in #325
• Mg/md by @mgastegger in #315
• Refactor configs and add predict script by @ktschuett in #334
• Fix install bug by @ktschuett in #335
• Add automatic position derivatives by @ktschuett in #341
• Update QM9 tutorial by @ktschuett in #347
• ZBL Potential, Electrostatics (+Ewald summation) and stress tensor fixes by @mgastegger in #349
• Fixes to some MD desfaults and torchscript issues by @mgastegger in #350
• Disable automatic use of torchscript in MD calculators by @mgastegger in #351
• Response properties and field representations by @mgastegger in #339
• Fixed derivative graph settings for basic response properties by @mgastegger in #352
• Fixed sign for shift type cutoff by @mgastegger in #353
• Refactor AtomisticModel by @ktschuett in #354
• Fix bug when using ddp with set run.id by @ktschuett in #359
• Fix mixing bias by @ktschuett in #365
• Fix mixing residual by @ktschuett in #366
• fixed aggregation_mode bug for avg pooling by @jnsLs in #361
• Sh/ep device by @Stefaanhess in #360
• Fix AtomsDataSubset for use inside ConcatAtomsData by @chgaul in #357
• Add learning rate warmup & SGD config by @ktschuett in #370
• Fixed creation of subset in BaseAtomsData by @NiklasGebauer in #369
• Fix DDP training by @ktschuett in #372
• Updated for new yaml behavior by @mgastegger in #375
• Nwag/comment-update by @NiklasGebauer in #374
• Fix OMDB _convert dataset preparation by @bartolsthoorn in #373
• Fix pin_memory and some deprecation warnings by @ktschuett in #378
• Update tutorials and filter outputs by @ktschuett in #377
• Add conversion script for old datasets by @ktschuett in #380
• Fixed bug in map_properties by @mgastegger in #381
• Fix testing bc lightning API changed by @ktschuett in #382
• Updated MD docstrings and tutorial by @mgastegger in #383
• Wrapping of atom positions under PBC by @mgastegger in #385
• Fixed energy logging for multiple molecules by @mgastegger in #386
• Some refactoring and cleanup by @ktschuett in #387
• Retrained ethanol model for new postprocessing convention by @mgastegger in #388
• Added on_step_finalize in MD simulation hooks by @mgastegger in #390
• Updated weight init in FieldSchNet representation by @mgastegger in #391
• Added tmpdir functionality for MDs, fixed calculator bug by @mgastegger in #395
• Improved config loading for MD by @mgastegger in #396
• Fixed criterion for recomputing MD neighborlists by @mgastegger in #398
• Fixed using subset in ASEAtomsData.iter_properties by @NiklasGebauer in #393
• update deprecated code to new torch version by @Stefaanhess in #399
• Fix strain input module by @ktschuett in #403
• Add resolver for tmp directory by @NiklasGebauer in #406
• Fix #401 by @ktschuett in #404
• Added tempfile import for custom tmpdir resolver by @mgastegger in #408
• Added routine to NeighborListMD to properly filter out pairs due to the buffer region by @mgastegger in #409
• Removed n_out argument from DipoleMoment and Polarizability layer docstrings by @mgastegger in #411
• Fix PyTorch Lightning deprecations by @ktschuett in #414
• Consider only a selection of atomic forces in training, validation and testing, and ASE neighborlist with skin implemented by @jnsLs in #405
• Added a few classes util for structure relaxations (in particular MOMONANO) by @jnsLs in #415
• New datasets and fixes by @ktschuett in #417
• Fixed inverted grad context for calculator by @mgastegger in https://github.com/atomistic-machine-learning/schnetpack/...

v1.0.1

What's Changed

• Sh/ep device by @Stefaanhess in #360
• Fix AtomsDataSubset for use inside ConcatAtomsData by @chgaul in #357
• Updated for new yaml behavior by @mgastegger in #375
• Fix OMDB _convert dataset preparation by @bartolsthoorn in #373
• update deprecated code to new torch version by @Stefaanhess in #399

New Contributors

• @chgaul made their first contribution in #357

Full Changelog: v1.0.0...v1.0.1