Changelog¶

Unreleased¶

v0.6.0¶

This release is shaped by three major pushes:

External model posture. Model integrations are increasingly maintained in the model’s own package rather than in-tree. Orb (#456), SevenNet (#467), and Metatomic (#523) are migrated to upstream-maintained torch-sim interfaces, Nequix lands as a new model via the same pattern (#511), and testing utilities for this external-model posture are added (#433, released in v0.5.2).
Extensible extras framework. A new scheme for attaching extra atom/system properties to SimState (#464) unlocks long-range physics on top of a common base: Electrostatics (#532), D3 Dispersion (#531), and a SumModel interface for composing force fields (#530, #537).
Physical validation. A major effort from @thomasloux introduced an integration-test suite that checks physical observables across the integrators (#521). This surfaced and drove fixes for a string of long-standing bugs in the NPT Nose-Hoover and NPT Langevin integrators (#498, #504, #505, #520, #528), materially improving MD correctness.

Other highlights include a native PyTorch batched cell list, a general pair-potential framework, a first-class PRNG on SimState with reproducibility docs, basic logging configuration, and duecredit support.

🎉 New Features¶

Introduce more extensible scheme for extra properties by @CompRhys in #464
Add Electrostatics using extensible extras by @CompRhys in #532
Add D3 Dispersion Model by @CompRhys in #531
Add SumModel interface by @CompRhys in #530
Compose models for allegro-pol simulation by @CompRhys in #537
Migrate Metatomic model to use upstream metatomic-torchsim integration by @CompRhys in #523
Add Nequix model via upstream nequix integration by @teddykoker in #511
Migrate Orb model to use upstream orb-models integration by @CompRhys in #456
Migrate SevenNet model to use upstream sevenn integration by @CompRhys in #467
Implement general pair-potential and pair-force by @CompRhys in #483
Add batched Lennard-Jones model by @abhijeetgangan in #474
Add NPTState that carries a stress variable by @thomasloux in #505
Introduce PRNG to SimState and add reproducibility docs by @CompRhys in #460
Configure basic logging by @CompRhys in #484
Implement first pass at duecredit by @CompRhys in #459

🚀 Performance¶

Batch the torch cell list by @CompRhys in #534
Eliminate for-loop in build_naive_neighborhood by @CompRhys in #470
Reduce thread count to prevent overhead slowdown for LJ on CPU by @thomasloux in #506
Remove unnecessary stress evaluation in NPTLangevin by @thomasloux in #528

🛠 Enhancements¶

Add pbc_wrap_batched_and_get_lattice_shifts to handle unwrapped positions in neighbor lists by @CompRhys in #519
Expose angle_tolerance in symmetry pipeline by @danielzuegner in #527
Forward max_memory_padding to _chunked_apply in optimize() by @niklashoelter in #513
Enforce compatible constraints in DOF helper by @janosh in #480
Check that models return detached tensors by @CompRhys in #494
Add log line clarifying BinningAutoBatcher is initialization-only in optimize by @orionarcher in #489
Add n_edges memory metric to auto-batching by @orionarcher in #481
Standardize parameter typing to float | torch.Tensor where appropriate by @CompRhys in #466
Remove StateDict type and ensure_sim_state helper by @orionarcher in #487
Add single-atom motif edge case to the validate function by @CompRhys in #518
Use sequential split instead of iterative __getitem__ for slicing by @thomasloux in #546

🐛 Bug Fixes¶

Fix state.to() not propagating dtype/device to constraints by @danielzuegner in #527
Fix FixSymmetry constraint system_idx remapping on reordered state slicing by @danielzuegner in #509
Fix NPT Nose-Hoover integrator by @alphalm4 in #520
Update momenta before integration in NPT Nose-Hoover by @alphalm4 in #498
Update sources for NPT Nose-Hoover by @thomasloux in #504
Clamp Frechet log-space deformation gradient to prevent NaN from matrix_exp overflow by @janosh in #450
Fix FIRE NaN velocity sentinel corrupting retained systems during autobatcher swaps by @janosh in #490
Fix GPU memory leak from missing .detach() in model wrappers by @reillyosadchey in #491
Adjust test tolerances to resolve false positive test failures by @CompRhys in #507
Update Phonopy test fixture after removal of dummy pbc kwarg in 3.2.0 by @CompRhys in #515
Hotfix: max pin on tables due to broken 3.11 PyTables release by @CompRhys in #461

🧹 House-Keeping¶

Remove graph-pes model integration by @craigxchen in #538
General maintenance & reap FairChem v1 by @CompRhys in #522
Add integration tests checking physical observables across integrators by @thomasloux in #521
Add example benchmarking scripts by @CompRhys in #516
Maintenance to restore CI to all passing by @CompRhys in #533
Nits rollup by @CompRhys in #501
Batched pair-function cleanup by @CompRhys in #488
Pin orb<0.6 by @CompRhys in #493
Consolidate branch protection rules by @CompRhys in #492
Bump tables version and use py3.14 for highest tests by @CompRhys in #486
Run ty in lint CI by @janosh in #339
Restore non-conflicting changes from #339 (ty linting setup) by @CompRhys in #476
Restore #460 seed-removal semantics after #339 by @janosh in #477
Set root path in link check action by @CompRhys in #468

v0.5.2¶

This release adds batched L-BFGS and BFGS optimizers, a FixSymmetry constraint, performance improvements to ts.static and ts.optimize through improved batching, and an MD progress bar. It also includes several bug fixes and quality-of-life improvements.

🎉 New Features¶

Implement batched L-BFGS and BFGS optimizers by @abhijeetgangan in #365
Add FixSymmetry constraint by @danielzuegner in #438
Enable MD progress bar by @danielzuegner in #448
Implement SimState.wrap_positions using pbc_wrap_batched by @Praagnya in #350

🚀 Performance¶

Accelerated ts.static + added scaling scripts by @falletta in #427
Accelerated ts.optimize by batching Frechet Cell Filter by @falletta in #439
Accelerated TrajectoryReporter when trajectory output is disabled by @falletta in #441
Vectorize atom index computation in L-BFGS optimizer by @falletta in #444

🛠 Enhancements¶

Add cumulative strain guard to FixSymmetry by @janosh in #447
Move SPS to a helper method and add documentation by @CompRhys in #455
Testing utils for external model posture by @CompRhys in #433

🐛 Bug Fixes¶

Fix constraint atom_idx not remapped on reordered slice by @falletta in #457
Fix L-BFGS history consistency for state updates by @falletta in #443
Handle wrapping correctly for default neighborlist by @abhijeetgangan in #436

🧹 House-Keeping¶

Bump pre-commit hooks and fix ruff lint errors by @janosh in #442
Make GraphPES tests less flaky by @CompRhys in #446

v0.5.1¶

This release adds several new features including constraints support for molecular dynamics and optimization, trajectory appending capabilities, and a batch cell list implementation. It also includes improvements to integrator initialization, temperature handling, and numerous bug fixes.

🎉 New Features¶

Constraints support for molecular dynamics and optimization by @thomasloux in #294
- Added FixAtoms constraint to fix specific atoms in place
- Added FixCom constraint to prevent center of mass drift
- Constraints automatically adjust degrees of freedom for accurate temperature calculations
- Full support across all integrators (NVE, NVT, NPT) and optimizers (FIRE, Gradient Descent)
- Constraints preserved during state manipulation (slicing, splitting, concatenation)
Scale atoms when changing cell by @thomasloux in #344
Allow different temperatures in ts.integrate by @thomasloux in #367
Add batch cell list by @abhijeetgangan in #388
Enable appending to trajectory when using ts.optimize/ts.integrate by @danielzuegner in #361
Enable user to save initial state of trajectory by @danielzuegner in #415

🛠 Enhancements¶

Better default force convergence function by @orionarcher in #404
Add systemwise max_force as a default property for reporter_dict by @orionarcher in #410
Add charge and spin to common_args dicts by @orionarcher in #413
Replace manual initialization with from_state across integrators and optimizers by @orionarcher in #420
Update energy description to ‘Potential energy’ by @danielzuegner in #408
Use upstream NequipTorchSimModel by @CompRhys in #400

🐛 Bug Fixes¶

Fix cuequivariance MACE by @thomasloux in #391
Fix SevenNet tests by @YutackPark in #393
Fix tutorials dependencies by @thomasloux in #396
Fix offsets in merge_constraints by @falletta in #402
Fix memory scaling calculation for non-periodic boundary conditions by @orionarcher in #412
Download NequIP model from Zenodo instead of nequip.net by @orionarcher in #418

📖 Documentation¶

Update metatrain version in metatomic tutorial by @Luthaf in #395

🧹 House-Keeping¶

Add close stale bot by @CompRhys in #411
Significantly consolidate scripts to speed up testing by @orionarcher in #385
Use validate_model_outputs in testing by @CompRhys in #401
Reduce test wall time by @CompRhys in #403
Pin scipy for fairchem tests by @CompRhys in #405
Disable NequIP tests for Python 3.13 by @curtischong in #421

v0.5.0¶

This release focuses on improving batch processing capabilities across TorchSim. The neighbor list module has been completely refactored to support batched calculations with multiple backend implementations, elastic tensor calculations now leverage batched operations for improved performance, and a bug fix ensures Monte Carlo swaps work correctly with ragged (different-sized) systems.

🎉 New Features¶

Refactor neighbor list module with batched support and multiple backends by @abhijeetgangan in #348
- New unified torchsim_nl function with automatic backend selection
- Multiple implementations: Alchemiops (NVIDIA CUDA), Vesin, torch_nl, and pure PyTorch fallback
- Support for both single-system and batched (multi-system) calculations
- Automatic selection of best available implementation based on installed packages

🛠 Enhancements¶

Batch elastic operations by @orionarcher in #384
- calculate_elastic_tensor now uses ts.static runner for batched calculations
- Added autobatcher parameter for memory-efficient processing of deformations
- Added pbar parameter for progress bar support

🐛 Bug Fixes¶

Fix Monte Carlo swap for ragged systems by @curtischong in #380
- Fixed generate_swaps calculation of system start indices for systems with different atom counts

v0.4.2¶

Thank you to everyone who contributed to this release! This release includes important bug fixes and new features. @thomasloux, @orionarcher, @WillEngler, @RishikeshMagar, @nh-univie, @andrewrm98, @danielzuegner, and others made valuable contributions. 🚀

🎉 New Features¶

Add CSVR / V-Rescale thermostat and anisotropic C rescale barostat by @thomasloux in #326
Support for electrostatics by @orionarcher in #373
Add support for AMD GPUs (consumer/datacenter) by @amacbride in #347

🐛 Bug Fixes¶

Fix: add init_kwargs to ts.integrate by @danielzuegner in #360
Fix PBC extraction to CPU fairchem model by @nh-univie in #368
Handle tensor PBC input in FairChemV1Model by @WillEngler in #372
Fix Comments Issue#309 by @RishikeshMagar in #378
Fix fairchem-legacy tests by removing explicit Hugging Face login by @WillEngler in #369

🛠 Enhancements¶

Consolidate model and model_name args in FairchemModel by @orionarcher in #377

New Contributors¶

@amacbride made their first contribution in #347
@danielzuegner made their first contribution in #360
@RishikeshMagar made their first contribution in #378

v0.4.1¶

Thank you to everyone who contributed to this release! This release includes important bug fixes, new features, and API improvements. @thomasloux, @curtischong, @CompRhys, @orionarcher, @WillEngler, @samanvya10, @hn-yu, @wendymak8, @chuin-wei, @pragnya17, and many others made valuable contributions. 🚀

💥 Breaking Changes¶

Standardize argument order to (state, model) by @pragnya17 in #341
Deprecate pbc_wrap_general by @curtischong in #305

🎉 New Features¶

Mixed PBC Support by @curtischong in #320
Feature: Batched NVT Nose-Hoover by @thomasloux in #265
Add degrees of freedom (dof) in state and temperature computation by @thomasloux in #328
Add sources npt langevin by @thomasloux in #298

🐛 Bug Fixes¶

MACE: transfer atomic numbers to CPU before converting to numpy by @t-reents in #289
Fixed max atoms memory estimation by @nh-univie in #279
Fixing model loading logic (names and cache dir) for fairchem models by @nh-univie in #278
Fix fairchem and neighbors tests by @WillEngler in #317
Fix #293: State to device side effects by @samanvya10 in #297
Fix graph-pes key issue by @jla-gardner in #303
Fix calculate static state after relax by @curtischong in #338
Misc fixes by @orionarcher in #336
Fix cell to cellpar by @thomasloux in #342
Fix failing Docs build by @CompRhys in #296
Fix: correct speedup plot image path in README by @Joaqland in #333

🛠 Enhancements¶

Put SimState Init logic into post_init (and enforce kw_args=true for SimState children) by @curtischong in #335
Replace vars(state) with state.attributes by @orionarcher in #329
Rename and reorder some variables by @orionarcher in #316
Add SevenNet path, str types for model arg by @YutackPark in #322
Add version attribute by @thomasloux in #311
Convert cell_to_cellpar from ase’s numpy implementation to pytorch by @wendymak8 in #306
Autobatch OOM handling by @chuin-wei in #337
Allow Mace to be loaded from a model path by @orionarcher in #349
NPTLangevinState inherits from MDState by @hn-yu in #299

📖 Documentation¶

Add integrators in docs by @thomasloux in #290

🏷️ Type Hints¶

Create py.typed by @arosen93 in #287

🧹 House-Keeping¶

Remove unused imports orb and updates class type check by @thomasloux in #292
Skip FairChem tests on forks due to HF secret by @CompRhys in #295

📦 Dependencies¶

Loosen numpy dependency by @chuin-wei in #321

New Contributors¶

@arosen93 made their first contribution in #287
@nh-univie made their first contribution in #278
@samanvya10 made their first contribution in #297
@wendymak8 made their first contribution in #306
@pragnya17 made their first contribution in #341
@chuin-wei made their first contribution in #321
@hn-yu made their first contribution in #299
@Joaqland made their first contribution in #333

v0.4.0¶

Thank you to everyone who contributed to this release! This release includes significant API improvements and breaking changes. @janosh led a major API redesign to improve usability. @stefanbringuier added heat flux calculations. @curtischong continued improving type safety across the codebase. @CompRhys, @orionarcher, @WillEngler, and @thomasloux all made valuable contributions. 🚀

What’s Changed¶

💥 Breaking Changes¶

Fairchem v2 support by @janosh, @CompRhys, @abhijeetgangan, @orionarcher in #211
Big breaking API redesign by @janosh in #264
Rename Flavors to more descriptive names by @orionarcher in #282

🎉 New Features¶

Enhancement: Heat Flux Function by @stefanbringuier in #127

🐛 Bug Fixes¶

Fix: orb squeeze provides incorrect shape for energy tensor by @thomasloux in #257
Fix docs build by @WillEngler in #271

🛠 Enhancements¶

Fairchem legacy support by @CompRhys in #270

📖 Documentation¶

Update citation in README.md by @orionarcher in #240
Add GOVERNANCE.md and remove Contributor’s Certification checkbox and language by @WillEngler in #272
Remove Contributor License Agreement (CLA) in favor of certification in contributing.md by @WillEngler in #267
Small update to README and CHANGELOG by @orionarcher in #283

🏷️ Type Hints¶

mypy type math.py and test_math.py by @curtischong in #242
Type test_io, neighbors, and transforms by @curtischong in #243
Type trajectory by @curtischong in #244

🧹 House-Keeping¶

MAINT: update pins in MACE phonons example. Remove misleading ty from PR template by @CompRhys in #239

New Contributors¶

@thomasloux made their first contribution in #257
@WillEngler made their first contribution in #267

Full Changelog: https://github.com/TorchSim/torch-sim/compare/v0.3.0…v0.4.0

v0.3.0¶

Thank you to everyone who contributed to this release! @t-reents, @curtischong, and @CompRhys did great work squashing an issue with SimState concatenation. @curtischong continued his crusade to type and improve the TorchSim API. @orionarcher, @kianpu34593, and @janosh all made contributions that continue to improve package quality and usability. 🚀

What’s Changed¶

🛠 Enhancements¶

Define attribute scopes in SimStates by @curtischong, @CompRhys, @orionarcher in #228
Improve typing of ModelInterface by @curtischong, @CompRhys in #215
Make system_idx non-optional in SimState by @curtischong in #231
Add new states when the max_memory_scaler is updated by @kianpu34593 in #222
Rename batch to system by @curtischong in #217, #233

🐛 Bug Fixes¶

Initial fix for concatenation of states in InFlightAutoBatcher by @t-reents in #219
Finish fix for SimState concatenation by @t-reents and @curtischong in #232
Fix broken code block in low-level tutorial by @CompRhys in #226
Update metatomic checkpoint to fix tests by @curtischong in #223
Fix memory scaling in determine_max_batch_size by @t-reents, @janosh in #212

📖 Documentation¶

Update README plot with more models by @orionarcher in #236, #237
Update citation.cff by @CompRhys in #225

Full Changelog: https://github.com/TorchSim/torch-sim/compare/v0.2.2…v0.3.0

v0.2.2¶

What’s Changed¶

💥 Breaking Changes¶

Remove higher level model imports by @CompRhys in https://github.com/TorchSim/torch-sim/pull/179

🛠 Enhancements¶

Add per atom energies and stresses for batched LJ by @abhijeetgangan in https://github.com/TorchSim/torch-sim/pull/144
throw error if autobatcher type is wrong by @orionarcher in https://github.com/TorchSim/torch-sim/pull/167

🐛 Bug Fixes¶

Mattersim fix tensors on wrong device (CPU->GPU) by @orionarcher in https://github.com/TorchSim/torch-sim/pull/154
fix npt_langevin by @jla-gardner in https://github.com/TorchSim/torch-sim/pull/153
Make sure to move data to CPU before calling vesin by @Luthaf in https://github.com/TorchSim/torch-sim/pull/156
Fix virial calculations in optimizers and integrators by @janosh in https://github.com/TorchSim/torch-sim/pull/163
Pad memory estimation by @orionarcher in https://github.com/TorchSim/torch-sim/pull/160
Refactor sevennet model by @YutackPark in https://github.com/TorchSim/torch-sim/pull/172
io optional dependencies in pyproject.toml by @curtischong in https://github.com/TorchSim/torch-sim/pull/185
Fix column->row cell vector mismatch in integrators by @CompRhys in https://github.com/TorchSim/torch-sim/pull/175

📖 Documentation¶

(tiny) add graph-pes to README by @jla-gardner in https://github.com/TorchSim/torch-sim/pull/149
Better module fig by @janosh in https://github.com/TorchSim/torch-sim/pull/168

🚀 Performance¶

More efficient Orb state_to_atoms_graph calculation by @AdeeshKolluru in https://github.com/TorchSim/torch-sim/pull/165

🚧 CI¶

Refactor test_math.py and test_transforms.py by @janosh in https://github.com/TorchSim/torch-sim/pull/151

🏥 Package Health¶

Try out hatchling for build vs setuptools by @CompRhys in https://github.com/TorchSim/torch-sim/pull/177

📦 Dependencies¶

Bump mace-torch to v0.3.12 by @janosh in https://github.com/TorchSim/torch-sim/pull/170
Update metatrain dependency by @Luthaf in https://github.com/TorchSim/torch-sim/pull/186

🏷️ Type Hints¶

Add torch_sim/typing.py by @janosh in https://github.com/TorchSim/torch-sim/pull/157

New Contributors¶

@Luthaf made their first contribution in https://github.com/TorchSim/torch-sim/pull/156
@YutackPark made their first contribution in https://github.com/TorchSim/torch-sim/pull/172
@curtischong made their first contribution in https://github.com/TorchSim/torch-sim/pull/185

Full Changelog: https://github.com/TorchSim/torch-sim/compare/v0.2.0…v0.2.1

v0.2.1¶

What’s Changed¶

💥 Breaking Changes¶

Remove higher level model imports by @CompRhys in #179

🛠 Enhancements¶

Add per atom energies and stresses for batched LJ by @abhijeetgangan in #144
throw error if autobatcher type is wrong by @orionarcher in #167

🐛 Bug Fixes¶

Fix column->row cell vector mismatch in integrators by @CompRhys in #175
Mattersim fix tensors on wrong device (CPU->GPU) by @orionarcher in #154
fix npt_langevin by @jla-gardner in #153
Make sure to move data to CPU before calling vesin by @Luthaf in #156
Fix virial calculations in optimizers and integrators by @janosh in #163
Pad memory estimation by @orionarcher in #160
Refactor sevennet model by @YutackPark in #172
io optional dependencies in pyproject.toml by @curtischong in #185

📖 Documentation¶

(tiny) add graph-pes to README by @jla-gardner in #149
Better module fig by @janosh in #168

🚀 Performance¶

More efficient Orb state_to_atoms_graph calculation by @AdeeshKolluru in #165

🚧 CI¶

Refactor test_math.py and test_transforms.py by @janosh in #151

🏥 Package Health¶

Try out hatchling for build vs setuptools by @CompRhys in #177

🏷️ Type Hints¶

Add torch-sim/typing.py by @janosh in #157

📦 Dependencies¶

Bump mace-torch to v0.3.12 by @janosh in #170
Update metatrain dependency by @Luthaf in #186

New Contributors¶

@Luthaf made their first contribution in #156
@YutackPark made their first contribution in #172
@curtischong made their first contribution in #185

Full Changelog: https://github.com/torchsim/torch-sim/compare/v0.2.0…v0.2.1

v0.2.0¶

Bug Fixes 🐛¶

Fix integrate reporting kwarg to arg error, #113 (raised by @hn-yu)
Allow runners to take large initial batches, #128 (raised by @YutackPark)
Add Fairchem model support for PBC, #111 (raised by @ryanliu30)

Enhancements 🛠¶

breaking Rename HotSwappingAutobatcher to InFlightAutobatcher and ChunkingAutoBatcher to BinningAutoBatcher, #143 @orionarcher
Support for Orbv3, #140, @AdeeshKolluru
Support metatensor models, #141, @frostedoyter @Luthaf
Support for graph-pes models, #118 @jla-gardner
Support MatterSim and fix ASE cell convention issues, #112 @CompRhys
Implement positions only FIRE optimization, #139 @abhijeetgangan
Allow different temperatures in batches, #123 @orionarcher
FairChem model updates: PBC handling, test on OMat24 e-trained model, #126 @AdeeshKolluru
FairChem model from_data_list support, #138 @ryanliu30
New correlation function module, #115 @stefanbringuier

Documentation 📖¶

Improved model documentation, #121 @orionarcher
Plot of TorchSim module graph in docs, #132 @janosh

House-Keeping 🧹¶

Only install HF for fairchem tests, #134 @CompRhys
Don’t download MBD in CI, #135 @orionarcher
Tighten graph-pes test bounds, #143 @orionarcher

v0.1.0¶

Initial release.