nvalchemi.training.distillation.evaluation.RadialDistribution#

class nvalchemi.training.distillation.evaluation.RadialDistribution(edges, g_r, counts, num_frames, num_atoms, r_max, pair=None)[source]#

Radial distribution function accumulated over one or more frames.

The radial distribution g(r) is the pair correlation function: how often pairs of atoms sit at separation r, normalized so an ideal gas gives 1.

Parameters:
  • edges (Tensor)

  • g_r (Tensor)

  • counts (Tensor)

  • num_frames (int)

  • num_atoms (int)

  • r_max (float)

  • pair (tuple[int, int] | None)

edges#

Bin edges from 0 to r_max, in A.

Type:

Float[torch.Tensor, “num_bins+1”]

g_r#

Pair correlation function, normalized so an ideal gas gives 1.

Type:

Float[torch.Tensor, “num_bins”]

counts#

Ordered-pair counts summed over every graph and frame. Each pair is apportioned between the two bins whose centers bracket its distance.

Type:

Float[torch.Tensor, “num_bins”]

num_frames#

Number of graphs the histogram was accumulated over.

Type:

int

num_atoms#

Number of atoms summed over the same graphs, whatever the pair filter.

Type:

int

r_max#

Cutoff the pairs were collected within.

Type:

float

pair#

Atomic numbers the pairs were restricted to, or None for the total g(r) over every species at once.

Type:

tuple[int, int] | None

to_dict()[source]#

Return the scalar fields and the curve as lists.

Return type:

dict[str, Any]