espnet2.gan_tts.hifigan.loss.GeneratorAdversarialLoss

source

class espnet2.gan_tts.hifigan.loss.GeneratorAdversarialLoss(average_by_discriminators: bool = True, loss_type: str = 'mse')

Bases: Module

Generator adversarial loss module.

This module computes the adversarial loss for the generator in a GAN setting. The loss can be calculated using either Mean Squared Error (MSE) or Hinge loss based on the specified configuration during initialization.

average_by_discriminators

Whether to average the loss by the number of discriminators.

• Type: bool

criterion

Loss function used for computing the adversarial loss, either MSE or Hinge.

• Type: callable

• Parameters:

  • average_by_discriminators (bool) – Whether to average the loss by the number of discriminators.
  • loss_type (str) – Loss type, either “mse” or “hinge”.

• Returns: Generator adversarial loss value.

• Return type: Tensor

####### Examples

>>> loss_fn = GeneratorAdversarialLoss(loss_type="mse")
>>> outputs = [torch.tensor([0.5]), torch.tensor([0.8])]
>>> loss = loss_fn(outputs)
>>> print(loss)

• Raises:AssertionError – If an unsupported loss_type is provided.

Initialize GeneratorAversarialLoss module.

• Parameters:
  • average_by_discriminators (bool) – Whether to average the loss by the number of discriminators.
  • loss_type (str) – Loss type, “mse” or “hinge”.

forward(outputs: List[List[Tensor]] | List[Tensor] | Tensor) → Tensor

Calculate generator adversarial loss.

This method computes the adversarial loss for the generator based on the outputs received from the discriminator(s). The loss is calculated using either mean squared error (MSE) or hinge loss, depending on the specified loss type during initialization.

• Parameters:outputs (Union *[*List *[*List *[*Tensor ] ] , List *[*Tensor ] , Tensor ]) – Discriminator outputs, which can be provided as a list of discriminator outputs, a list of lists of discriminator outputs, or a single tensor.
• Returns: The calculated generator adversarial loss value.
• Return type: Tensor

####### Examples

>>> loss_fn = GeneratorAdversarialLoss(loss_type="mse")
>>> outputs = [torch.tensor([0.5]), torch.tensor([0.3])]
>>> loss = loss_fn(outputs)
>>> print(loss)  # Outputs the loss value