espnet2.uasr.generator.abs_generator.AbsGenerator

source

class espnet2.uasr.generator.abs_generator.AbsGenerator(*args, **kwargs)

Bases: Module, ABC

Abstract base class for generators in the ESPnet2 UASR framework.

This class defines the essential methods that must be implemented by any concrete generator subclass. Generators are responsible for producing outputs based on input tensors, and this class provides a standardized interface.

None

output_size() → int

Returns the size of the output tensor produced by the generator.

forward(xs_pad

torch.Tensor, ilens: torch.Tensor) -> Tuple[torch.Tensor,

torch.Tensor, Optional[torch.Tensor]]

Takes padded input tensors and their lengths, processes them, and returns the output tensor along with additional information.

• Raises:NotImplementedError – If a subclass does not implement the abstract methods.

######### Examples

class MyGenerator(AbsGenerator): : def output_size(self) -> int: : return 256 <br/> def forward(self, xs_pad: torch.Tensor, ilens: torch.Tensor) -> Tuple[ : torch.Tensor, torch.Tensor, Optional[torch.Tensor]]: # Implementation goes here pass

NOTE

This class should not be instantiated directly. Instead, subclass it to create a specific generator implementation.

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

abstract forward(xs_pad: Tensor, ilens: Tensor) → Tuple[Tensor, Tensor, Tensor | None]

Performs the forward pass of the generator model.

This method takes padded input tensors and their corresponding lengths, and produces the output tensors as per the generator’s architecture.

• Parameters:
  • xs_pad (torch.Tensor) – A tensor containing the padded input sequences.
  • ilens (torch.Tensor) – A tensor containing the lengths of the input sequences before padding.
• Returns:
  • A tensor representing the output of the generator.
  • A tensor representing the hidden states.
  • An optional tensor representing any additional outputs, if applicable.
• Return type: Tuple[torch.Tensor, torch.Tensor, Optional[torch.Tensor]]
• Raises:NotImplementedError – If this method is not overridden in a subclass.

######### Examples

>>> generator = MyGenerator()  # MyGenerator is a subclass of AbsGenerator
>>> xs_pad = torch.tensor([[1, 2, 3], [4, 5, 0]])
>>> ilens = torch.tensor([3, 2])
>>> output, hidden, additional = generator.forward(xs_pad, ilens)

NOTE

The implementation of this method must handle the specific logic for generating outputs based on the provided input sequences and their lengths.

abstract output_size() → int

Returns the size of the output from the generator.

This property is expected to be implemented by subclasses of AbsGenerator to provide the specific output size of the generator.

• Returns: The size of the output produced by the generator.
• Return type: int
• Raises:NotImplementedError – If the method is not implemented by the subclass.

######### Examples

class MyGenerator(AbsGenerator): : def output_size(self) -> int: : return 256

gen = MyGenerator() print(gen.output_size()) # Output: 256