tensorclouds.nn

Submodules

Classes

TensorCloud

A TensorCloud is a collection of tensors with associated coordinates and masks.

Package Contents

class tensorclouds.nn.TensorCloud

A TensorCloud is a collection of tensors with associated coordinates and masks. It is used to represent a cloud of tensors in a 3D space, where each tensor can have different irreducible representations (irreps).

irreps_array

The array of tensors with their irreducible representations.

Type:

e3nn.IrrepsArray

mask_irreps_array

A boolean mask indicating which irreps are present.

Type:

jax.Array

coord

The coordinates of the tensors in 3D space.

Type:

jax.Array

mask_coord

A boolean mask indicating which coordinates are valid.

Type:

jax.Array

label

An optional label for the TensorCloud, e.g., for classification tasks.

Type:

jax.Array, optional

irreps_array: e3nn_jax.IrrepsArray
mask_irreps_array: jax.Array
coord: jax.Array
mask_coord: jax.Array
label: jax.Array = None
property shape
classmethod empty(irreps: e3nn_jax.Irreps)
classmethod ones_like(tc)
classmethod zeros_like(tc)
property irreps: e3nn_jax.Irreps
property mask: jax.Array
replace(**kwargs)
classmethod concatenate(clouds: List[TensorCloud]) TensorCloud
classmethod zeros(irreps: e3nn_jax.Irreps, shape: tuple) TensorCloud
classmethod stack(clouds: List[TensorCloud]) TensorCloud
split(piece_sizes: List[int]) List[TensorCloud]
centralize()
plot(view=None, viewer=None, colors=None, radius=0.04, mid=0.95)
dot(other)
norm()

Compute the norm of the irreps_array and coord of the TensorCloud. :returns: A tuple containing the norms of the irreps_array and coord. :rtype: Tuple[jax.Array, jax.Array]