Skip to content

IncreaseKLDOnEpoch

latent.callbacks.IncreaseKLDOnEpoch

Increase Kullback-Leibler Divergence loss of VAEs during training.

__init__(self, factor=1.5, max_val=1.0, **kwargs) special

Parameters:

Name Type Description Default
factor float

Positive float. Factor by which the KLD will be increased each epoch.

1.5
max_val float

Positive float. Maximum value of KLD.

1.0
**kwargs

Other parameters passed to keras.callbacks.Callback.

{}