neppo
neppo trains neuroevolution potential (NEP) models in PyTorch and writes them in the nep.txt format that GPUMD runs. It implements the following models of GPUMD.
NEP4 potentials, with type-dependent cutoffs, the three forms of ZBL repulsion and an optional dispersion term.
Temperature-dependent NEP models, whose networks take the temperature as an input.
qNEP charge models, with atomic charges that depend on the environment and their electrostatic energy.
TNEP models of dipoles and polarizabilities.
Models are trained with AdamW on the GPU, with the whole training step replayed from a CUDA graph. The loss is a weighted sum of terms for energies, forces, virials or stresses, charges, Born effective charges, dipoles or polarizabilities, and the regularization of the parameters, and its weights can change between stages of the training. A run directory of the nep executable of GPUMD, with nep.in and train.xyz, can be trained directly with neppo train. The user guide describes the functionality and holds tutorials on training NEP, qNEP and TNEP models, the parameters of a model, Pareto-optimal models, the learning rate and the batch size, training in stages, the construction of the loss and its weights, and intermolecular interactions. neppo and its development are hosted on GitLab.
Installation
neppo can be installed via pip:
pip install neppo
neppo requires Python 3.10 or newer and builds on ASE, calorine, NumPy, pandas, and PyTorch. Please consult the installation section of the user guide for more information.
Credits
neppo is developed at the Department of Physics and Astronomy of Chalmers University of Technology. The NEP potentials were introduced in Physical Review B 104, 104309 (2021) and extended in Journal of Chemical Physics 157, 114801 (2022) and Nature Communications 15, 10208 (2024). The qNEP models were introduced in Journal of Chemical Theory and Computation 22, 4787 (2026), and the TNEP models in Journal of Chemical Theory and Computation 20, 3273 (2024).
Please consult the Credits page of the documentation for additional references.
1 maintainer
Metadata
1 maintainer
Release files for neppo 1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| neppo-1.0.tar.gz | 87.9 kB | Without provenance | Details for neppo-1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| neppo-1.0-py3-none-any.whl | Python 3 | none | any | Details for neppo-1.0-py3-none-any.whl |
Total release size: 186.0 kB
Release files / neppo-1.0.tar.gz
| Download URL | neppo-1.0.tar.gz |
|---|---|
| Size | 87.9 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
22eac222655cc09a86e80df7dd072c5749f09f019d04be3d4c3f1070a82e9a52
|
|
BLAKE2b-256 checksum How to use checksums |
4be9066e9258d51aaa5ab0edd151f5621394539b9d12ecb278870db5527223c0
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.12.15
|
Release files / neppo-1.0-py3-none-any.whl
| Download URL | neppo-1.0-py3-none-any.whl |
|---|---|
| Size | 98.1 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
c08177b6778f905fa217eb02f90fcf643bc4cbf010bffa51b898b93e00b80b18
|
|
BLAKE2b-256 checksum How to use checksums |
d6547aba2cbae56dad416b33e262897ab26b5abd164068e85f3a81348ed1cc92
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.12.15
|
Security for neppo 1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.
