complextorch.nn.modules.casting

Drop-in torch.nn.Module adapters that convert between real-tensor layouts and complex tensors, so they compose inside torch.nn.Sequential.

These layouts come up in real-to-complex pipelines:

• Interleaved: [..., 2*D] with real and imaginary parts interleaved: (re_0, im_0, re_1, im_1, ...).

• Concatenated: [..., 2*D] with all real parts followed by all imaginary parts: (re_0, ..., re_{D-1}, im_0, ..., im_{D-1}).

If your data is in (..., 2) final-dim layout (one slot for real, one for imag), use torch.view_as_complex() / torch.view_as_real() directly — no wrapper is needed.

Classes

ComplexToConcatenated	Complex → Concatenated Real Layout
ComplexToInterleaved	Complex → Interleaved Real Layout
ConcatenatedToComplex	Concatenated Real Layout → Complex
InterleavedToComplex	Interleaved Real Layout → Complex
RealToComplex	Real → Complex (Zero Imaginary)

Module Contents

class complextorch.nn.modules.casting.ComplexToConcatenated(*args: Any, **kwargs: Any)

Bases: torch.nn.Module

Complex → Concatenated Real Layout

Inverse of ConcatenatedToComplex. Maps complex [..., D] to real [..., 2D] with all real parts followed by all imaginary parts.

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

forward(input: torch.Tensor) → torch.Tensor

class complextorch.nn.modules.casting.ComplexToInterleaved(*args: Any, **kwargs: Any)

Bases: torch.nn.Module

Complex → Interleaved Real Layout

Inverse of InterleavedToComplex. Maps complex [..., D] to real [..., 2D] with real/imag interleaved.

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

forward(input: torch.Tensor) → torch.Tensor

class complextorch.nn.modules.casting.ConcatenatedToComplex(*args: Any, **kwargs: Any)

Bases: torch.nn.Module

Concatenated Real Layout → Complex

Maps [..., 2D] with the first D slots being real parts and the last D slots being imaginary parts to a complex tensor of shape [..., D].

Input: (re_0, ..., re_{D-1}, im_0, ..., im_{D-1}).

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

forward(input: torch.Tensor) → torch.Tensor

class complextorch.nn.modules.casting.InterleavedToComplex(*args: Any, **kwargs: Any)

Bases: torch.nn.Module

Interleaved Real Layout → Complex

Maps [..., 2D] with real/imag interleaved along the last dim to a complex tensor of shape [..., D].

Input: (re_0, im_0, re_1, im_1, ...).

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

forward(input: torch.Tensor) → torch.Tensor

class complextorch.nn.modules.casting.RealToComplex(dtype: torch.dtype = torch.cfloat)

Bases: torch.nn.Module

Real → Complex (Zero Imaginary)

Lifts a real tensor into a complex tensor by setting the imaginary part to zero. Useful as the first layer of a network whose input is a real signal but whose internal representations are complex.

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

extra_repr() → str

Return the extra representation of the module.

To print customized extra information, you should re-implement this method in your own modules. Both single-line and multi-line strings are acceptable.

forward(input: torch.Tensor) → torch.Tensor

dtype = Ellipsis