espnet2.enh.layers.complex_utils.to_float

source

espnet2.enh.layers.complex_utils.to_float(c)

Convert a tensor to a float tensor.

This function converts a complex tensor or a native complex PyTorch tensor to a float tensor. If the input is already a float tensor or does not meet the conditions for conversion, it returns the input unchanged.

• Parameters:
  • c (Union *[*torch.Tensor , ComplexTensor ]) – The input tensor which can be
  • tensor. (either a ComplexTensor or a PyTorch)
• Returns: A float tensor if the input is a complex tensor, otherwise returns the input tensor unchanged.
• Return type: Union[torch.Tensor, ComplexTensor]

Examples

>>> c_complex = torch.tensor([1+2j, 3+4j], dtype=torch.complex64)
>>> to_float(c_complex)
tensor([1., 2., 3., 4.])

>>> c_float = torch.tensor([1.0, 2.0, 3.0])
>>> to_float(c_float)
tensor([1.0, 2.0, 3.0])