espnet2.uasr.loss.abs_loss.AbsUASRLoss

source

class espnet2.uasr.loss.abs_loss.AbsUASRLoss(*args, **kwargs)

Bases: Module, ABC

Base class for all Diarization loss modules.

This class serves as an abstract base for creating various types of loss functions used in Unsupervised Audio Source Separation tasks. It defines the essential interface that any derived loss class must implement.

name

The name of the loss function, which will be used as a key

• Type: str

in the reporter.

forward()

This method must be implemented by derived classes to compute

the loss based on the model's predictions and the target values.

• Raises:
  • NotImplementedError – If the forward method is not implemented in a
  • derived class. –

####### Examples

To create a custom loss class, inherit from AbsUASRLoss and implement the name property and forward method as follows:

``

`

python class CustomLoss(AbsUASRLoss):

@property def name(self) -> str:

return “custom_loss”

def forward(self, predictions, targets) -> torch.Tensor: : # Implement the custom loss computation loss = torch.mean((predictions - targets) ** 2) return loss

``

`

Initialize internal Module state, shared by both nn.Module and ScriptModule.

abstract forward() → Tensor

Compute the forward pass of the loss function.

This method must be implemented by any subclass of AbsUASRLoss. The return tensor should have the shape of (batch), representing the computed loss values for each sample in the batch.

• Returns: A tensor containing the computed loss values for the input batch.
• Return type: torch.Tensor
• Raises:NotImplementedError – If the method is not implemented in a subclass.

####### Examples

>>> class MyLoss(AbsUASRLoss):
...     def forward(self) -> torch.Tensor:
...         return torch.tensor([1.0, 2.0, 3.0])
>>> loss = MyLoss()
>>> output = loss.forward()
>>> print(output)
tensor([1.0, 2.0, 3.0])

property name : str