layer.py File Reference

Neural network layers. More...

Functions
def	nabu.neuralnetworks.components.layer.blstm (inputs, sequence_length, num_units, layer_norm=False, scope=None)
	a BLSTM layer More...
def	nabu.neuralnetworks.components.layer.pblstm (inputs, sequence_length, num_units, num_steps=2, layer_norm=False, scope=None)
	a Pyramidal BLSTM layer More...
def	nabu.neuralnetworks.components.layer.projected_subsampling (inputs, input_seq_lengths, num_steps, name=None)

Detailed Description

Neural network layers.

Function Documentation

blstm()

def nabu.neuralnetworks.components.layer.blstm	(		inputs,
			sequence_length,
			num_units,
			layer_norm = False,
			scope = None
	)

a BLSTM layer

Parameters

inputs	the input to the layer as a [batch_size, max_length, dim] tensor
sequence_length	the length of the input sequences as a [batch_size] tensor
num_units	The number of units in the one directon
layer_norm	whether layer normalization should be applied
scope	The variable scope sets the namespace under which the variables created during this call will be stored.

Returns

the blstm outputs

pblstm()

def nabu.neuralnetworks.components.layer.pblstm	(		inputs,
			sequence_length,
			num_units,
			num_steps = 2,
			layer_norm = False,
			scope = None
	)

a Pyramidal BLSTM layer

Parameters

inputs	the input to the layer as a [batch_size, max_length, dim] tensor
sequence_length	the length of the input sequences as a [batch_size] tensor
num_units	The number of units in the one directon
num_steps	the number of time steps to concatenate
layer_norm	whether layer normalization should be applied
scope	The variable scope sets the namespace under which the variables created during this call will be stored.

Returns

• the PBLSTM outputs
• the new sequence lengths