espnet2.enh.layers.complexnn.complex_cat

source

espnet2.enh.layers.complexnn.complex_cat(inputs, axis)

Concatenate complex-valued tensors along a specified axis.

This function takes a list of complex-valued tensors, splits each tensor into its real and imaginary parts, concatenates these parts along the specified axis, and returns a single complex-valued tensor.

espnet2.enh.layers.complexnn.inputs

A list of complex-valued tensors, where each tensor is expected to have an even number of channels (real and imaginary parts).

• Type: list

espnet2.enh.layers.complexnn.axis

The axis along which to concatenate the real and imaginary parts.

• Type: int

• Parameters:

  • inputs (list of torch.Tensor) – A list containing complex-valued tensors.
  • axis (int) – The axis along which to concatenate the real and imaginary parts.

• Returns: A single complex-valued tensor formed by concatenating the : real and imaginary parts of the input tensors along the specified axis.

• Return type: torch.Tensor

Examples

>>> import torch
>>> a = torch.tensor([[1.0, 2.0], [3.0, 4.0]])  # Real part
>>> b = torch.tensor([[5.0, 6.0], [7.0, 8.0]])  # Imaginary part
>>> complex_tensor = torch.cat([a, b], dim=-1)  # Combine real and imag
>>> result = complex_cat([complex_tensor, complex_tensor], axis=0)
>>> print(result)
tensor([[1., 2., 5., 6.],
        [3., 4., 7., 8.],
        [1., 2., 5., 6.],
        [3., 4., 7., 8.]])

NOTE

The input tensors must have the same shape along all dimensions except for the specified axis.