neuralmonkey.model.sequence_split module

Split temporal states such that the sequence is n-times longer.

class neuralmonkey.model.sequence_split.SequenceSplitter(name: str, parent: neuralmonkey.model.stateful.TemporalStateful, factor: int, projection_size: int = None, projection_activation: Callable[[tensorflow.python.framework.ops.Tensor], tensorflow.python.framework.ops.Tensor] = None) → None

Bases: neuralmonkey.model.stateful.TemporalStateful, neuralmonkey.model.model_part.ModelPart

__init__(name: str, parent: neuralmonkey.model.stateful.TemporalStateful, factor: int, projection_size: int = None, projection_activation: Callable[[tensorflow.python.framework.ops.Tensor], tensorflow.python.framework.ops.Tensor] = None) → None

Initialize SentenceSplitter.

Parameters:

• parent – TemporalStateful whose states will be split.
• factor – Factor by which the states will be split - the resulting sequence will be longer by this factor.
• projection_size – If not None, specifies dimensionality of a projection before state splitting.
• projection_activation – Non-linearity function for the optional projection.

dependencies

Return a list of attribute names regarded as dependents.

feed_dict(dataset: neuralmonkey.dataset.Dataset, train: bool = True) → Dict[tensorflow.python.framework.ops.Tensor, Any]

Return a feed dictionary for the given feedable object.

Parameters:

• dataset – A dataset instance from which to get the data.
• train – Boolean indicating whether the model runs in training mode.

Returns:

A FeedDict dictionary object.

temporal_mask

Return mask for the temporal_states.

A 2D Tensor of shape (batch, time) of type float32 which masks the temporal states so each sequence can have a different length. It should only contain ones or zeros.

temporal_states

Return object states in time.

A 3D Tensor of shape (batch, time, state_size) which contains the states of the object in time (e.g. hidden states of a recurrent encoder.

neuralmonkey.model.sequence_split.split_by_factor(tensor_3d: tensorflow.python.framework.ops.Tensor, batch_size: tensorflow.python.framework.ops.Tensor, factor: int) → tensorflow.python.framework.ops.Tensor