:py:mod:`anomalib.models.ganomaly.loss`
=======================================

.. py:module:: anomalib.models.ganomaly.loss

.. autoapi-nested-parse::

   Loss function for the GANomaly Model Implementation.



Module Contents
---------------

Classes
~~~~~~~

.. autoapisummary::

   anomalib.models.ganomaly.loss.GeneratorLoss
   anomalib.models.ganomaly.loss.DiscriminatorLoss




.. py:class:: GeneratorLoss(wadv=1, wcon=50, wenc=1)

   Bases: :py:obj:`torch.nn.Module`

   Generator loss for the GANomaly model.

   :param wadv: Weight for adversarial loss. Defaults to 1.
   :type wadv: int, optional
   :param wcon: Image regeneration weight. Defaults to 50.
   :type wcon: int, optional
   :param wenc: Latent vector encoder weight. Defaults to 1.
   :type wenc: int, optional

   .. py:method:: forward(latent_i: torch.Tensor, latent_o: torch.Tensor, images: torch.Tensor, fake: torch.Tensor, pred_real: torch.Tensor, pred_fake: torch.Tensor) -> torch.Tensor

      Compute the loss for a batch.

      :param latent_i: Latent features of the first encoder.
      :type latent_i: Tensor
      :param latent_o: Latent features of the second encoder.
      :type latent_o: Tensor
      :param images: Real image that served as input of the generator.
      :type images: Tensor
      :param fake: Generated image.
      :type fake: Tensor
      :param pred_real: Discriminator predictions for the real image.
      :type pred_real: Tensor
      :param pred_fake: Discriminator predictions for the fake image.
      :type pred_fake: Tensor

      :returns: The computed generator loss.
      :rtype: Tensor



.. py:class:: DiscriminatorLoss

   Bases: :py:obj:`torch.nn.Module`

   Discriminator loss for the GANomaly model.

   .. py:method:: forward(pred_real, pred_fake)

      Compye the loss for a predicted batch.

      :param pred_real: Discriminator predictions for the real image.
      :type pred_real: Tensor
      :param pred_fake: Discriminator predictions for the fake image.
      :type pred_fake: Tensor

      :returns: The computed discriminator loss.
      :rtype: Tensor