data.calculate_stats

data.calculate_stats(dataloader: AtomsLoader, divide_by_atoms: Dict[str, bool], atomref: Dict[str, torch.Tensor] | None = None) → Dict[str, Tuple[torch.Tensor, torch.Tensor]]

Use the incremental Welford algorithm described in [h1] to accumulate the mean and standard deviation over a set of samples.

References:

[h1]

https://en.wikipedia.org/wiki/Algorithms_for_calculating_variance

param dataset:

atoms data set

param divide_by_atoms:

dict from property name to bool: If True, divide property by number of atoms before calculating statistics.

param atomref:

reference values for single atoms to be removed before calculating stats

Returns: