atompack-db 0.4.0

Fast, compressed storage for atomic structures with properties.

Atompack

Fast, compressed storage for atomic structures with properties.

Atompack is part of the open-source LeMaterial effort for large-scale materials and molecular ML datasets.

Install

uv pip install atompack-db

Import the package as atompack after installation.

The Hugging Face Hub upload/download helpers are included in the main package.

Quick start

import atompack
import numpy as np

positions = np.array([[0.0, 0.0, 0.0], [1.0, 0.0, 0.0]], dtype=np.float32)
atomic_numbers = np.array([6, 8], dtype=np.uint8)
mol = atompack.Molecule.from_arrays(positions, atomic_numbers)
mol.energy = -123.456
mol.forces = np.array([[0.1, 0.2, 0.3], [0.4, 0.5, 0.6]], dtype=np.float32)

db = atompack.Database("data.atp", overwrite=True)
db.add_molecule(mol)
db.flush()

Reopen with mmap for read-only access, or disable mmap when you want to append:

db = atompack.Database.open("data.atp")             # read-only, mmap=True by default
db = atompack.Database.open("data.atp", mmap=False)  # writable append mode

Hugging Face Hub

import atompack

db = atompack.hub.open("LeMaterial/Atompack", "omat/train")
print(db[0].energy)
db.close()

Remote datasets such as omat/train and omol/train can be opened directly from LeMaterial/Atompack.

More

Full documentation and benchmark notes live at https://entalpic-atompack.readthedocs-hosted.com/en/latest/.

If you are working on the Rust crate directly, the maintained Rust entrypoints are:

• cargo run -p atompack --example basic_usage
• cargo run -p atompack --release --bin atompack-bench -- --help

License

Apache-2.0