DataInt.get_activation_outhead DataInt.get_activation_outhead() → str | torch.nn.Module | None[source] Returns the loss function for approximating this feature.