espnet2.spk.projector.abs_projector.AbsProjector

source

class espnet2.spk.projector.abs_projector.AbsProjector(*args, **kwargs)

Bases: Module, ABC

Abstract base class for a projector module in a speaker processing system.

This class serves as a template for implementing various types of projector modules. A projector transforms an input tensor (typically representing speaker embeddings) into an output tensor of a specified size. Subclasses must define the output_size method to specify the size of the output tensor and the forward method to implement the transformation logic.

None

• Parameters:None
• Returns: None
• Yields: None
• Raises:
  • NotImplementedError – If output_size or forward methods are not
  • implemented in a subclass. –

######### Examples

Here is an example of how a subclass might implement this abstract class:

``

`

python class MyProjector(AbsProjector):

def output_size(self) -> int: : return 128

def forward(self, utt_embd: torch.Tensor) -> torch.Tensor: : return torch.relu(torch.nn.Linear(utt_embd.size(1), self.output_size())(utt_embd))

``

`

NOTE

This class inherits from torch.nn.Module, so any subclass must also comply with the requirements of PyTorch modules.

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

abstract forward(utt_embd: Tensor) → Tensor

Abstract base class for a projector in a neural network.

This class defines the interface for projectors that take an input tensor (utt_embd) and produce an output tensor. The output size can be defined by subclasses implementing the output_size method.

None

• Parameters:
  • utt_embd (torch.Tensor) – The input tensor representing the utterance
  • embedding.
• Returns: The projected output tensor.
• Return type: torch.Tensor
• Raises:NotImplementedError – If the method is not implemented in a subclass.

######### Examples

class MyProjector(AbsProjector): : def output_size(self) -> int: : return 128 <br/> def forward(self, utt_embd: torch.Tensor) -> torch.Tensor: : # Implementation of forward pass return torch.nn.functional.relu(utt_embd)

projector = MyProjector() output = projector.forward(torch.randn(1, 256)) # Example input tensor

NOTE

This class is intended to be subclassed and should not be instantiated directly.

abstract output_size() → int

Abstract property that defines the output size of the projector.

This property must be implemented by subclasses to specify the expected output size when processing input tensors through the projector.

• Returns: The size of the output tensor after projection.
• Return type: int
• Raises:NotImplementedError – If the method is not implemented in a subclass.

######### Examples

class MyProjector(AbsProjector): : def output_size(self) -> int: : return 128

projector = MyProjector() print(projector.output_size()) # Output: 128