get_embedding_size#

pytorch_forecasting.utils._utils.get_embedding_size(n: int, max_size: int = 100) int[source]#

Determine empirically good embedding sizes (formula taken from fastai).

Parameters:
  • n (int) – number of classes

  • max_size (int, optional) – maximum embedding size. Defaults to 100.

Returns:

embedding size

Return type:

int