Bering.training.EarlyStopper

class Bering.training.EarlyStopper(patience=6, min_delta=0.025)[source]

Early Stop implementation for loss values

Paramemters

patience: Maximal number of consecutive times allowed to have loss greater than min_loss + min_delta before stopping
min_delta: Minimal gap of a new loss and minimal loss for adding to a count to counter

Methods

early_stop(validation_loss)

Check if the training should be stopped