espnet2.asr.state_spaces.pool.upsample

source

espnet2.asr.state_spaces.pool.upsample(x, stride=1, expand=1, transposed=False)

Upsample a tensor by applying stride and expansion.

This function increases the dimensions of the input tensor x by repeating its elements based on the specified stride and expand parameters. The function can also handle transposed operations.

• Parameters:
  • x (torch.Tensor) – The input tensor to upsample. It should be a 3D or higher-dimensional tensor.
  • stride (int , optional) – The factor by which to increase the sequence length. Default is 1.
  • expand (int , optional) – The factor by which to increase the feature dimension. Default is 1.
  • transposed (bool , optional) – If True, the operation will be treated as a transposed operation. Default is False.
• Returns: The upsampled tensor.
• Return type: torch.Tensor

Examples

>>> import torch
>>> x = torch.tensor([[1, 2], [3, 4]])
>>> upsample(x, stride=2, expand=2)
tensor([[1, 1, 2, 2],
        [3, 3, 4, 4]])

>>> upsample(x, stride=1, expand=3)
tensor([[1, 2, 2, 2],
        [3, 4, 4, 4]])

>>> upsample(x, stride=2, expand=1, transposed=True)
tensor([[1, 2],
        [1, 2],
        [3, 4],
        [3, 4]])

• Raises:AssertionError – If the input tensor is None.