https://github.com/mir-group/allegro - v0.7.1

[0.7.1]

Fixed

• nequip-package will now always pick up Triton (even if packaging without GPUs) and CuEq files (even if CuEq is not installed at package-time)

Changed

• bound nequip version >=0.13.0

- Python
Published by cw-tan 7 months ago

https://github.com/mir-group/allegro - v0.7.0: CuEquivariance Support

We are proud to announce the integration of NVIDIA's CuEquivariance to accelerate Allegro models. Information on how to use it can be found in the CuEquivariance Acceleration page.

Training

CuEquivariance can be used to accelerate training by using nequip.model.modify in the config file: yaml training_module: _target_: allegro.train.EMALightningModule # ... other training module configurations ... model: _target_: nequip.model.modify modifiers: - modifier: enable_CuEquivarianceContracter model: _target_: allegro.model.AllegroModel # ... your standard Allegro model configuration ... compile_mode: compile # ^ CuEquivariance composes with train-time compilation

Inference

CuEquivariance can be used to accelerate TorchScript and AOTInductor inference in ASE by doing nequip-compile with --modifiers enable_CuEquivarianceContracter.

AOT Inductor Compilation

bash nequip-compile \ path/to/model.ckpt \ path/to/compiled_model.nequip.pt2 \ --device cuda \ --mode aotinductor \ --target ase \ --modifiers enable_CuEquivarianceContracter

TorchScript Compilation

nequip-compile \ path/to/model.ckpt \ path/to/compiled_model.nequip.pth \ --device cuda \ --mode torchscript \ --target ase \ --modifiers enable_CuEquivarianceContracter

- Python
Published by cw-tan 8 months ago

https://github.com/mir-group/allegro - v0.6.3

[0.6.3]

Fixed

• enable_TritonContracter restores weights correctly now

- Python
Published by cw-tan 8 months ago

https://github.com/mir-group/allegro - v0.6.2

[0.6.2]

Added

• Exposed AllegroEnergyModel in docs

- Python
Published by cw-tan 8 months ago

https://github.com/mir-group/allegro - v0.6.1

[0.6.1]

Removed unused code and revamped docs.

Removed

• scatter_features option (undocumented feature)

- Python
Published by cw-tan 9 months ago

https://github.com/mir-group/allegro - v0.6.0

[0.6.0]

Changed

• bound nequip version >=0.8.0
• [Breaking] mechanism for using custom Triton TP kernel

- Python
Published by cw-tan 9 months ago

https://github.com/mir-group/allegro - v0.5.1

[0.5.1]

Changed

• bound nequip version under 0.8.0

- Python
Published by cw-tan 9 months ago

https://github.com/mir-group/allegro - v0.5.0

Breaking changes to simplify the user interface for using Allegro models, including refactoring modules and removing hyperparameter options.

Added

• Allegro model docs

Changed

• [Breaking] refactored scalar MLP out of Bessel embedding
• [Breaking] Allegro hyperparameter defaults

Removed

• [Breaking] so3 argument for parity_setting (only o3_full and o3_restricted allowed now)
• [Breaking] turn parity_setting into a bool called parity (true corresponds to o3_full; false corresponds to o3_restricted)
• [Breaking] scalar_embed_output_dim is no longer a hyperparameter (and is fixed as num_scalar_features in the Allegro model)

- Python
Published by cw-tan 10 months ago

https://github.com/mir-group/allegro - v0.4.0 - major update

A major update to the NequIP framework and Allegro model discussed in this preprint.

The NequIP docs contain information on the new user interface.

Please note that this update is not backwards compatible.

- Python
Published by cw-tan 10 months ago

https://github.com/mir-group/allegro - v0.3.0

[0.3.0]

Added

• Hyperparameters tensors_mixing_mode, weight_individual_irreps, typexbasis_mode, tensor_track_weight_init, self_edge_tensor_product
• New Bessel basis with hyperparameters num_bessels_per_basis, bessel_frequency_cutoff, per_edge_type_cutoff

Changed

• Resnet normalization scheme
• [Breaking] latent_resnet: false enables skip connections
• [Breaking] Default for norm_basis_mean_shift changed from True -> False
• [Breaking] Rename env_embed_mul -> num_tensor_features
• [Breaking] Replaced nonscalars_include_parity with tensor_track_allowed_irreps

- Python
Published by cw-tan 11 months ago