nvalchemi.csp.sample_space_groups#

nvalchemi.csp.sample_space_groups(num_samples, num_operations, *, probabilities=None, crystal_system=None, sohncke_only=False, seed=None, device=None)[source]#

Sample compatible International space-group numbers with replacement.

A supplied probabilities mapping replaces the default relative sampling weights. Missing group entries have zero weight. After removing incompatible groups, the remaining weights determine relative sampling probabilities.

Parameters:
  • num_samples (int) – Number of samples to draw; must be positive.

  • num_operations (int) – Required number of symmetry operations.

  • probabilities (Mapping[int, float], optional) – Replacement relative weights keyed by group number; values must be finite and nonnegative.

  • crystal_system (CrystalSystem, optional) – Restrict the compatible groups to one crystal system.

  • sohncke_only (bool, default=False) – Use only groups whose symmetry-generated copies preserve molecular handedness, such as for chiral molecules.

  • seed (int, optional) – Nonnegative seed for a local CPU torch.Generator.

  • device (torch.device or str, optional) – Device for the returned int32 tensor.

Returns:

space_groups – Sampled International numbers.

Return type:

torch.Tensor, shape [num_samples], dtype=torch.int32

Raises:
  • TypeError – If the prior, filters, or integer arguments have invalid types.

  • ValueError – If no compatible group has positive weight or seed is negative.

Examples

>>> groups = sample_space_groups(4, 1, seed=7)
>>> groups.dtype
torch.int32