This release is a pre-release and may not be stable for production use.
Fast Rust kernels for computational materials science, with Python and WebAssembly bindings.
- Structures and chemistry: compositions with exact rational amounts, periodic structures, CIF and extxyz I/O, symmetry and protostructure labels, Niggli and Minkowski reduction with integer transforms, primitive cells that respect magnetic and constraint order, structure matching, MP2020 corrections and convex hulls (also finite-temperature)
- Relaxation, NEB and MD: batched geometry optimization, nudged elastic bands, NVT and NPT molecular dynamics, equations of state and elastic constants
- Phonons: harmonic and anharmonic lattice dynamics, QHA, thermal conductivity (with four-phonon scattering), force-constant fitting, TDEP and SSCHA
- docs/alloys.md: solid solutions, derivative-structure enumeration, SQS and lattice Monte Carlo
- docs/free-energy.md: absolute free energies and melting points by nonequilibrium thermodynamic integration
- docs/calphad.md: CALPHAD Gibbs models, phase equilibria, binary phase diagrams, TDB I/O and fitting
- docs/mace.md: native MACE inference on CPU, Metal, CUDA and ROCm
- docs/wasm.md: the WebAssembly bindings, their API coverage and numerics
Design
- One model interface. Anything with
compute(structures, properties) -> list[dict]is a model. Energy, forces and stress have fixed slots; every other head (magnetic moments, charges, ...) is a named array, so new physics never changes the interface. Adapters wrap existing calculators. - Native MACE.
matscikit.MACEruns MACE foundation models (MACE-MH-1, all heads) without torch on CPU (f64/f32), Apple GPUs (Metal), NVIDIA GPUs (CUDA feature) and AMD GPUs (ROCm,hipfeature), with f64 parity to mace-torch at 1e-13 eV/atom on the materials heads (≤ 1e-15 relative on every head) for all 89 supported elements. It loads mace-torch.modelcheckpoints directly (a native zip and pickle reader, no torch at any point), is thread-safe, and itsmetadatarecords the checkpoint sha256 andmatscikit.BUILD_ID(a hash of the source the extension was built from); see docs/mace.md. - Batched by default. Relaxation, MD and thermodynamic integration advance many structures (replicas, phases, temperatures) in lock step with one model call per step.
Install
uv pip install matscikit # or `uv add matscikit` in a uv project
From source (needs a Rust toolchain):
uv pip install -e .
The type stub bindings/python/matscikit/_core.pyi is generated from the PyO3 signatures: after changing src/python/, regenerate it with uvx maturin@1.16.0 generate-stubs -q -F stubs --out bindings/python (the pre-push pyo3-stubs hook does this, and CI fails when it is stale).
Python
Anything with compute(structures, properties) is a model, so the same workflows run with EMT, EAM, MACE or an adapted ASE calculator:
import matscikit as msk
cu = msk.Structure([[0, 1.8, 1.8], [1.8, 0, 1.8], [1.8, 1.8, 0]], ["Cu"], [[0, 0, 0]])
print(msk.symmetry_dataset(cu)["hm_short"]) # F m -3 m
# batched relaxation: one model call per step for all structures
[result] = msk.relax([cu.make_supercell(2)], msk.EMT(), f_max=0.01)
print(result["converged"], result["energy"]) # eV
# harmonic phonons
phonons = msk.Phonons(cu, [4, 4, 4])
phonons.generate_displacements(distance=0.01)
phonons.produce_fc2(msk.EMT()) # or forces as an (n_sets, n_atoms, 3) array
print(phonons.frequencies([[0, 0, 0], [0.5, 0, 0.5]])) # THz, (n_q, n_bands)
print(phonons.thermal_properties([16, 16, 16], t_max_k=1000)["free_energy_kj_per_mol"][-1])
# the same calls with a MACE foundation model, read straight from the mace-torch checkpoint
mace = msk.MACE("mace-mh-1.model", head="omat_pbe", device="metal", dtype="float32")
[result] = msk.relax([cu.make_supercell(2)], mace, f_max=0.01)
print(result["converged"], result["energy"])
JavaScript / WebAssembly
The same kernels run in browsers and Node through WebAssembly bindings with TypeScript types (src/wasm/, cargo feature wasm): everything except the Metal, CUDA and HIP engines, the ASE adapters and the few workflows docs/wasm.md lists as Python-only, MACE-MH-1 on CPU included. Results match the Python bindings bitwise or to round-off (MACE float64 forces within 1e-14 eV/Å).
uv run scripts/build_wasm.py # bindings/wasm/pkg: the npm package matscikit, with license files
import init, { EMT, Structure, relax } from './pkg/matscikit.js'
await init()
const cu = new Structure([[0, 1.8, 1.8], [1.8, 0, 1.8], [1.8, 1.8, 0]], [`Cu`], [[0, 0, 0]])
const [result] = relax([cu.make_supercell(2)], new EMT(), { f_max: 0.01 })
See docs/wasm.md for usage per area, the API coverage table and measured numerics.
Source
The Rust source by module, areas proportional to non-blank lines (inline tests included); regenerate with uv run --no-project docs/source_treemap.py.
Acknowledgements
matscikit builds on, and is tested against, these projects:
- phonopy by Atsushi Togo and contributors (Togo, Chaput, Tadano and Tanaka, J. Phys.: Condens. Matter 35, 353001 (2023); Togo, J. Phys. Soc. Jpn. 92, 012001 (2023); Togo and Tanaka, Scr. Mater. 108, 1 (2015)): harmonic phonons, force constants and their conventions
- phono3py by Atsushi Togo and contributors (Togo, Chaput and Tanaka, Phys. Rev. B 91, 094306 (2015)): lattice thermal conductivity from third-order force constants
- pymatgen (Ong et al., Comput. Mater. Sci. 68, 314 (2013)): compositions, structure matching, phase diagrams, MP2020 corrections and the structure dict format
- ASE (Larsen et al., J. Phys.: Condens. Matter 29, 273002 (2017)): optimizers, cell filters, MD integrators, NEB and the extxyz format
- moyo by Kohei Shinohara: the symmetry engine behind every space-group operation
- MACE (Batatia et al., NeurIPS 2022): the model architecture the native engine runs, checked against mace-torch
- SeeK-path (Hinuma et al., Comput. Mater. Sci. 128, 140 (2017)): the band-path convention
- hiPhive (Eriksson, Fransson and Erhart, Adv. Theory Simul. 2, 1800184 (2019)): cluster-space force-constant fitting
- pycalphad (Otis and Liu, J. Open Res. Softw. 5, 1 (2017)): the reference for the CALPHAD models and TDB handling
License
MIT (license). Bundled data and fixtures: docs/third-party-notices.md. Linked Rust crates: docs/crate-licenses.md.
Metadata
Release files for matscikit 0.1.0b2
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 |
|---|---|---|---|---|
| matscikit-0.1.0b2.tar.gz | 2.5 MB | Without provenance | Details for matscikit-0.1.0b2.tar.gz |
Built distributions (wheels)
| File | Actions Reset | |||
|---|---|---|---|---|
| matscikit-0.1.0b2-cp311-abi3-win_amd64.whl | CPython 3.11 | abi3 | Windows x86-64 | Details for matscikit-0.1.0b2-cp311-abi3-win_amd64.whl |
| matscikit-0.1.0b2-cp311-abi3-manylinux_2_28_x86_64.whl | CPython 3.11 | abi3 | Linux glibc 2.28+ x86-64 | Details for matscikit-0.1.0b2-cp311-abi3-manylinux_2_28_x86_64.whl |
| matscikit-0.1.0b2-cp311-abi3-manylinux_2_28_aarch64.whl | CPython 3.11 | abi3 | Linux glibc 2.28+ ARM64 | Details for matscikit-0.1.0b2-cp311-abi3-manylinux_2_28_aarch64.whl |
| matscikit-0.1.0b2-cp311-abi3-macosx_11_0_arm64.whl | CPython 3.11 | abi3 | macOS 11.0+ ARM64 | Details for matscikit-0.1.0b2-cp311-abi3-macosx_11_0_arm64.whl |
Total release size: 37.6 MB
Release files / matscikit-0.1.0b2.tar.gz
| Download URL | matscikit-0.1.0b2.tar.gz |
|---|---|
| Size | 2.5 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
bf470890ab701102fb13fb01f72e0652bcf8ef467160ec12eca64ceb532881f8
|
|
BLAKE2b-256 checksum How to use checksums |
d6418a8741191fd91d8472e32cbe67fbe0bee2015c27324dc5a8aa20709f93cf
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.13.5 {"installer":{"name":"uv","version":"0.13.5","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
|
Release files / matscikit-0.1.0b2-cp311-abi3-win_amd64.whl
| Download URL | matscikit-0.1.0b2-cp311-abi3-win_amd64.whl |
|---|---|
| Size | 9.6 MB |
| Tags | CPython 3.11 Windows x86-64 abi3 |
|
SHA-256 checksum How to use checksums |
f590002c985113d911197cf323b84ac8386b4905d35f7ebedac19503942dfc8b
|
|
BLAKE2b-256 checksum How to use checksums |
df83f0cf458a5930ef6f514ab93fad68409c54a9cdf537d8d3f94db766ff2bc3
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.13.5 {"installer":{"name":"uv","version":"0.13.5","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
|
Release files / matscikit-0.1.0b2-cp311-abi3-manylinux_2_28_x86_64.whl
| Download URL | matscikit-0.1.0b2-cp311-abi3-manylinux_2_28_x86_64.whl |
|---|---|
| Size | 9.3 MB |
| Tags | CPython 3.11 Linux glibc 2.28+ x86-64 abi3 |
|
SHA-256 checksum How to use checksums |
bfd78bf871b7dde608ab58a9d599a4ec0e3df8fd0cedd98eafb011987c266243
|
|
BLAKE2b-256 checksum How to use checksums |
915857b631e9c2a13de60b6b9e7bab6a021ffcafdc0bc0849a476632bd176632
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.13.5 {"installer":{"name":"uv","version":"0.13.5","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
|
Release files / matscikit-0.1.0b2-cp311-abi3-manylinux_2_28_aarch64.whl
| Download URL | matscikit-0.1.0b2-cp311-abi3-manylinux_2_28_aarch64.whl |
|---|---|
| Size | 8.2 MB |
| Tags | CPython 3.11 Linux glibc 2.28+ ARM64 abi3 |
|
SHA-256 checksum How to use checksums |
055ac4a5b8db95bf87ebf36906dfc0ff5b9e1e06f30b9516b34706c30a83e299
|
|
BLAKE2b-256 checksum How to use checksums |
451c0e7d3ce06680467b284ec4972cbddd9652b0cd6c77541899c2002833573f
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.13.5 {"installer":{"name":"uv","version":"0.13.5","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
|
Release files / matscikit-0.1.0b2-cp311-abi3-macosx_11_0_arm64.whl
| Download URL | matscikit-0.1.0b2-cp311-abi3-macosx_11_0_arm64.whl |
|---|---|
| Size | 7.9 MB |
| Tags | CPython 3.11 abi3 macOS 11.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
0986fc0e409e59b8bb3cc13b6408d9ada3c283550e2ffbac5b3374a6fa211b28
|
|
BLAKE2b-256 checksum How to use checksums |
a8be6a41da0cd687698f04b9a8847e08b4b16e41c8891c2b0a7cd227fcd95f41
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.13.5 {"installer":{"name":"uv","version":"0.13.5","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
|
Security for matscikit 0.1.0b2
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.