espnet2.layers.inversible_interface.InversibleInterface

source

class espnet2.layers.inversible_interface.InversibleInterface

Bases: ABC

InversibleInterface is an abstract base class that defines the interface for

inversible layers. Any concrete implementation of this interface must provide an implementation for the inverse method, which is responsible for computing the inverse of a given input tensor.

None

• Parameters:
  • input (torch.Tensor) – The input tensor to be inverted.
  • input_lengths (torch.Tensor , optional) – The lengths of the input tensor. Defaults to None.
• Returns: A tuple containing the inverted output tensor and the corresponding output lengths.
• Return type: Tuple[torch.Tensor, torch.Tensor]
• Raises:
  • NotImplementedError – If the inverse method is called without an
  • implementation in a subclass. –

####### Examples

class MyInversibleLayer(InversibleInterface): : def inverse(self, input: torch.Tensor, input_lengths: torch.Tensor = None) -> Tuple[torch.Tensor, torch.Tensor]: : # Implementation of the inverse operation output = input.flip(dims=[-1]) # Example operation output_lengths = input_lengths # Example lengths return output, output_lengths

layer = MyInversibleLayer() input_tensor = torch.randn(10, 5) # Example input input_lengths = torch.tensor([5] * 10) # Example lengths output_tensor, output_lengths = layer.inverse(input_tensor, input_lengths)

abstract inverse(input: Tensor, input_lengths: Tensor | None = None) → Tuple[Tensor, Tensor]

Computes the inverse of the given input tensor.

This method must be implemented by any subclass of the InversibleInterface. The implementation should define how to compute the inverse transformation of the input tensor, along with the lengths of the output.

• Parameters:
  • input (torch.Tensor) – The input tensor to be inverted.
  • input_lengths (torch.Tensor , optional) – The lengths of the input sequences. Defaults to None.
• Returns: A tuple containing the output tensor and the output lengths. The output tensor is the result of the inverse transformation applied to the input tensor.
• Return type: Tuple[torch.Tensor, torch.Tensor]
• Raises:NotImplementedError – If the method is not implemented in a subclass.

####### Examples

class MyInversible(InversibleInterface): : def inverse(self, input: torch.Tensor, : > input_lengths: torch.Tensor = None) -> Tuple[torch.Tensor, torch.Tensor]: <br/> # Example implementation of the inverse method output = input.flip(dims=[-1]) # Flipping the tensor for demo output_lengths = input_lengths # Assuming lengths remain the same return output, output_lengths

my_inverse = MyInversible() input_tensor = torch.tensor([[1, 2], [3, 4]]) lengths = torch.tensor([2, 2]) output_tensor, output_lengths = my_inverse.inverse(input_tensor, lengths)