Atoms

class nfflr.Atoms(cell: Tensor, positions: Tensor, numbers: Tensor, *, batch_num_atoms: Iterable[int] | None = None)

Atoms: basic atomistic data structure supporting batching

The goal is to make it easy to work with a common representation, and transform it as needed to other formats.

• DGLGraph

• ALIGNNGraphTuple

• PyG format

• …

Parameters:

celltorch.Tensor

cell matrix / lattice parameters

positionstorch.Tensor

Cartesian coordinates

numberstorch.Tensor

atomic numbers

Other Parameters:

batch_num_atomsIterable[int], optional

batch size for struct-of-arrays batched Atoms

Attributes:

batch_num_atoms

Methods

batched()	Determine if multiple structures are batched together.
scaled_positions()	Convert cartesian coordinates to fractional coordinates.
to(device[, non_blocking])	Transfer atoms data to compute device.

batched() → bool

Determine if multiple structures are batched together.

scaled_positions() → Tensor

Convert cartesian coordinates to fractional coordinates.

Returns:

scaled_positionstorch.Tensor

to(device, non_blocking: bool = False)

Transfer atoms data to compute device.

Parameters:

devicetorch.device

compute device, e.g. “cpu”, “cuda”, “mps”

non_blockingbool, optional

attempt asynchronous transfer.