nequix

library0.4.5pypypiunverified

A PyTorch-based library for building and training equivariant neural network potentials, supporting E(3)-equivariant message passing architectures. Current version 0.4.5, with monthly release cadence.

ai-ml

pip install nequix

github ↗homepage ↗

INSTALL

IMPORT

SIG · NEQUIX

nequix

ai-mlpythonv0.4.5

harness data pending

Install & Compatibility

Where this runs

No compatibility data collected yet for this library.

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

NequIPDataset

✓ from nequix import NequIPDataset

✗ from nequip.data import NequIPDataset

Minimal training loop. Requires data in supported format (e.g., .xyz). See https://docs.nequip.org for full example.

import torch from nequip.data import AtomicData, dataset_from_config from nequip.train import Trainer # Build a simple dataset (assuming you have data) # dataset = dataset_from_config('config.yaml') # Initialize and train # trainer = Trainer(model_cfg=..., data_cfg=..., verbosity='low') # trainer.train() print('NequIP quickstart example (incomplete without data files).')

Debug

Known issues

breakingv0.4.0 introduced a config restructuring: `model_builders` replaced `model` in config files. Old configs will raise KeyError.

fix

Update config files to use new 'model_builders' key. Refer to migration guide in GitHub release notes.

affects: <0.4.0

gotchaDataset loading requires explicit `root` path; relative paths can cause confusing FileNotFoundError if the working directory is not as expected.

fix

Always use absolute paths or set the working directory before script execution.

affects: all

deprecatedThe `nequip.scripts` module is deprecated; use the CLI via `nequip-train` and `nequip-evaluate` commands instead.

fix

Replace `from nequip.scripts.train import main` with command-line `nequip-train`. See docs for CLI usage.

affects: >=0.4.0

Upgrade

Version history

0.4.5latest on PyPI · released Mar 17, 2026

Audit

Dependencies

torchrequiredCore dependency for tensor operations and neural network modules. Version >=1.10 required.

e3nnrequiredProvides equivariant neural network layers (o3, irreps). Version >=0.5.0 required.

Agent activity

16 hits · last 30 days

node

Resources

githubhttps://github.com/atomicarchitects/nequix ↗

homepagehttps://github.com/atomicarchitects/nequix ↗