analogvnn.nn.module.Layer#

Module Contents#

Classes#

Layer

Base class for analog neural network modules.
class analogvnn.nn.module.Layer.Layer[source]#

Bases: torch.nn.Module

Base class for analog neural network modules.

Variables:

  • _inputs (Union[None, ArgsKwargs]) – Inputs of the layer.

  • _outputs (Union[None, Tensor, Sequence[Tensor]]) – Outputs of the layer.

  • _backward_module (Optional[BackwardModule]) – Backward module of the layer.

  • _use_autograd_graph (bool) – If True, the autograd graph is used to calculate the gradients.

  • call_super_init (bool) – If True, the super class __init__ of nn.Module is called

  • https – //github.com/pytorch/pytorch/pull/91819

property use_autograd_graph: bool[source]#

If True, the autograd graph is used to calculate the gradients.

Returns:

use_autograd_graph.

Return type:

bool

property inputs: analogvnn.graph.ArgsKwargs.ArgsKwargsOutput[source]#

Inputs of the layer.

Returns:

inputs.

Return type:

ArgsKwargsOutput

property outputs: Union[None, torch.Tensor, Sequence[torch.Tensor]][source]#

Outputs of the layer.

Returns:

outputs.

Return type:

Union[None, Tensor, Sequence[Tensor]]

property backward_function: Union[None, Callable, analogvnn.backward.BackwardModule.BackwardModule][source]#

Backward module of the layer.

Returns:

backward_function.

Return type:

Union[None, Callable, BackwardModule]

_inputs: Union[None, analogvnn.graph.ArgsKwargs.ArgsKwargs][source]#
_outputs: Union[None, torch.Tensor, Sequence[torch.Tensor]][source]#
_backward_module: Optional[analogvnn.backward.BackwardModule.BackwardModule][source]#
_use_autograd_graph: bool[source]#
call_super_init: bool = True[source]#
__call__(*inputs, **kwargs)[source]#

Calls the forward pass of neural network layer.

Parameters:

  • *inputs – Inputs of the forward pass.

  • **kwargs – Keyword arguments of the forward pass.

set_backward_function(backward_class: Union[Callable, analogvnn.backward.BackwardModule.BackwardModule, Type[analogvnn.backward.BackwardModule.BackwardModule]]) → Layer[source]#

Sets the backward_function attribute.

Parameters:

backward_class (Union[Callable, BackwardModule, Type[BackwardModule]]) – backward_function.

Returns:

self.

Return type:

Layer

Raises:

TypeError – If backward_class is not a callable or BackwardModule.

named_registered_children(memo: Optional[Set[torch.nn.Module]] = None) → Iterator[Tuple[str, torch.nn.Module]][source]#

Returns an iterator over immediate registered children modules.

Parameters:

memo – a memo to store the set of modules already added to the result

Yields:

(str, Module) – Tuple containing a name and child module

Note

Duplicate modules are returned only once. In the following example, l will be returned only once.

registered_children() → Iterator[torch.nn.Module][source]#

Returns an iterator over immediate registered children modules.

Yields:

nn.Module – a module in the network

Note

Duplicate modules are returned only once. In the following example, l will be returned only once.

_forward_wrapper(function: Callable) → Callable[source]#

Wrapper for the forward function.

Parameters:

function (Callable) – Forward function.

Returns:

Wrapped function.

Return type:

Callable

_call_impl_forward(*args: torch.Tensor, **kwargs: torch.Tensor) → analogvnn.utils.common_types.TENSORS[source]#

Calls the forward pass of the layer.

Parameters:

  • *args – Inputs of the forward pass.

  • **kwargs – Keyword arguments of the forward pass.

Returns:

Outputs of the forward pass.

Return type:

TENSORS