espnet2.tts2.feats_extract.identity.IdentityFeatureExtract

source

class espnet2.tts2.feats_extract.identity.IdentityFeatureExtract

Bases: AbsFeatsExtractDiscrete

IdentityFeatureExtract is a feature extraction class that keeps the input discrete sequence unchanged. It is designed for use in text-to-speech (TTS) systems within the ESPnet framework. This class inherits from AbsFeatsExtractDiscrete and overrides the forward method to validate and return the input data.

None

• Parameters:None
• Returns:
  • A tensor containing the input converted to long type.
  • A tensor containing the input lengths.
• Return type: Tuple[torch.Tensor, torch.Tensor]
• Raises:
  • AssertionError – If the input tensor is complex, floating point, or
  • boolean**, or** if it does not have 2 dimensions**, or** if the number of –
  • input sequences does not match the number of input lengths. –

####### Examples

>>> extractor = IdentityFeatureExtract()
>>> input_tensor = torch.tensor([[1, 2], [3, 4]])
>>> input_lengths = torch.tensor([2, 2])
>>> output, lengths = extractor.forward(input_tensor, input_lengths)
>>> print(output)
tensor([[1, 2],
        [3, 4]])
>>> print(lengths)
tensor([2, 2])

NOTE

This class is primarily intended for use where the input sequence needs to be passed through without modification.

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

forward(input: Tensor, input_lengths: Tensor) → Tuple[Any, Dict]

Forward pass of the IdentityFeatureExtract class.

This method processes the input tensor and returns it along with its lengths. It ensures that the input tensor meets certain criteria, specifically that it is a 2-dimensional tensor of integer type.

• Parameters:
  • input (torch.Tensor) – A 2D tensor containing the discrete input sequence.
  • input_lengths (torch.Tensor) – A 1D tensor containing the lengths of the input sequences. Its size must match the first dimension of the input.
• Returns: A tuple containing: : - The input tensor converted to long type.
  • The input lengths tensor.
• Return type: Tuple[Any, Dict]
• Raises:AssertionError – If the input tensor is complex, floating point, or boolean, or if the input tensor does not have 2 dimensions, or if the number of input sequences does not match the number of lengths.

####### Examples

>>> extractor = IdentityFeatureExtract()
>>> input_tensor = torch.tensor([[1, 2, 3], [4, 5, 6]])
>>> input_lengths = torch.tensor([3, 3])
>>> output, lengths = extractor.forward(input_tensor, input_lengths)
>>> print(output)
tensor([[1, 2, 3],
        [4, 5, 6]])
>>> print(lengths)
tensor([3, 3])

NOTE

This class is designed to keep the input discrete sequence unchanged.