aihwkit.nn.modules.rnn.layers module

Analog RNN layers

class aihwkit.nn.modules.rnn.layers.AnalogBidirRNNLayer(cell, *cell_args)[source]

Bases: aihwkit.nn.modules.container.AnalogSequential

Bi-directional analog RNN layer.

Parameters

  • cell (Type) – RNNCell type (AnalogLSTMCell/AnalogGRUCell/AnalogVanillaRNNCell)

  • cell_args (Any) – arguments to RNNCell (e.g. input_size, hidden_size, rpu_configs)

forward(input_, states)[source]

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

Parameters

  • input_ (torch.Tensor) –

  • states (List[Union[Tuple[torch.Tensor, torch.Tensor], torch.Tensor]]) –

Return type

Tuple[torch.Tensor, List[Union[Tuple[torch.Tensor, torch.Tensor], torch.Tensor]]]

get_zero_state(batch_size)[source]

Returns a zeroed state.

Parameters

batch_size (int) – batch size of the input

Returns

Zeroed state tensor

Return type

torch.Tensor

class aihwkit.nn.modules.rnn.layers.AnalogRNNLayer(cell, *cell_args)[source]

Bases: aihwkit.nn.modules.container.AnalogSequential

Analog RNN Layer.

Parameters

  • cell (Type) – RNNCell type (AnalogLSTMCell/AnalogGRUCell/AnalogVanillaRNNCell)

  • cell_args (Any) – arguments to RNNCell (e.g. input_size, hidden_size, rpu_configs)

forward(input_, state)[source]

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

Parameters

  • input_ (torch.Tensor) –

  • state (Union[Tuple[torch.Tensor, torch.Tensor], torch.Tensor]) –

Return type

Tuple[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]]

get_zero_state(batch_size)[source]

Returns a zeroed state.

Parameters

batch_size (int) – batch size of the input

Returns

Zeroed state tensor

Return type

torch.Tensor

class aihwkit.nn.modules.rnn.layers.AnalogReverseRNNLayer(cell, *cell_args)[source]

Bases: aihwkit.nn.modules.container.AnalogSequential

Analog RNN layer for direction.

Parameters

  • cell (Type) – RNNCell type (AnalogLSTMCell/AnalogGRUCell/AnalogVanillaRNNCell)

  • cell_args (Any) – arguments to RNNCell (e.g. input_size, hidden_size, rpu_configs)

forward(input_, state)[source]

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

Parameters

  • input_ (torch.Tensor) –

  • state (Union[Tuple[torch.Tensor, torch.Tensor], torch.Tensor]) –

Return type

Tuple[torch.Tensor, Union[Tuple[torch.Tensor, torch.Tensor], torch.Tensor]]

get_zero_state(batch_size)[source]

Returns a zeroed state.

Parameters

batch_size (int) – batch size of the input

Returns

Zeroed state tensor

Return type

torch.Tensor

static reverse(lst)[source]

Reverses the list of input tensors.

Parameters

lst (List[torch.Tensor]) –

Return type

List[torch.Tensor]

aihwkit.nn.modules.rnn.layers.stack(tensors, dim=0, *, out=None) → Tensor

Concatenates a sequence of tensors along a new dimension.

All tensors need to be of the same size.

Parameters

  • tensors (sequence of Tensors) – sequence of tensors to concatenate

  • dim (int) – dimension to insert. Has to be between 0 and the number of dimensions of concatenated tensors (inclusive)

Keyword Arguments

out (Tensor, optional) – the output tensor.