def testRNNCell(self):
cell = rnn.make_rnn_cell(3, 10, dropout=0.1, residual_connections=True)
inputs = tf.random.uniform([4, 5])
states = cell.get_initial_state(inputs=inputs)
outputs, states = cell(inputs, states, training=True)
self.assertEqual(len(states), 3)
self.assertListEqual(outputs.shape.as_list(), [4, 10])
def __init__(self,
num_layers,
num_units,
bridge_class=None,
cell_class=None,
dropout=0.3,
residual_connections=False,
**kwargs):
"""Initializes the decoder parameters.
Args:
num_layers: The number of layers.
num_units: The number of units in each layer.
bridge_class: A :class:`opennmt.layers.Bridge` class to pass the
encoder state to the decoder. Default to
:class:`opennmt.layers.ZeroBridge`.
cell_class: The inner cell class or a callable taking :obj:`num_units` as
argument and returning a cell. Defaults to a LSTM cell.
dropout: The probability to drop units in each layer output.
residual_connections: If ``True``, each layer input will be added to its
output.
**kwargs: Additional layer arguments.
"""
super(RNNDecoder, self).__init__(**kwargs)
self.dropout = dropout
self.cell = rnn.make_rnn_cell(
num_layers,
num_units,
dropout=dropout,
residual_connections=residual_connections,
cell_class=cell_class)
if bridge_class is None:
bridge_class = bridge.ZeroBridge
self.bridge = bridge_class()
def __init__(self,
num_layers,
num_units,
bidirectional=False,
residual_connections=False,
dropout=0.3,
reducer=ConcatReducer(),
cell_class=None,
**kwargs):
"""Initializes the parameters of the encoder.
Args:
num_layers: The number of layers.
num_units: The number of units in each layer.
bidirectional: Use a bidirectional RNN.
residual_connections: If ``True``, each layer input will be added to its
output.
reducer: A :class:`opennmt.layers.reducer.Reducer` instance to merge
bidirectional state and outputs.
dropout: The probability to drop units in each layer output.
cell_class: The inner cell class or a callable taking :obj:`num_units` as
argument and returning a cell. Defaults to a LSTM cell.
**kwargs: Additional layer arguments.
"""
super(RNNEncoderV2, self).__init__(**kwargs)
cell = rnn.make_rnn_cell(num_layers,
num_units,
dropout=dropout,
residual_connections=residual_connections,
cell_class=cell_class)
self.rnn = rnn.RNN(cell, bidirectional=bidirectional, reducer=reducer)
def __init__(self,
num_layers,
num_units,
cell_class=None,
dropout=0.3,
residual_connections=False,
**kwargs):
"""Initializes the parameters of the encoder.
Args:
num_layers: The number of layers.
num_units: The number of units in each layer.
cell_class: The inner cell class or a callable taking :obj:`num_units` as
argument and returning a cell. Defaults to a LSTM cell.
dropout: The probability to drop units in each layer output.
residual_connections: If ``True``, each layer input will be added to its
output.
kwargs: Additional layer arguments.
"""
super(UnidirectionalRNNEncoderV2, self).__init__(**kwargs)
cell = rnn.make_rnn_cell(
num_layers,
num_units,
dropout=dropout,
residual_connections=residual_connections,
cell_class=cell_class)
self.rnn = rnn.RNN(cell)
def testRNN(self):
cell = rnn.make_rnn_cell(3, 10, dropout=0.1, residual_connections=True)
rnn_layer = rnn.RNN(cell)
inputs = tf.random.uniform([4, 5, 5])
outputs, states = rnn_layer(inputs, training=True)
self.assertListEqual(outputs.shape.as_list(), [4, 5, 10])
self.assertIsInstance(states, tuple)
self.assertEqual(len(states), 3)
def testBRNN(self):
cell = rnn.make_rnn_cell(3, 10, dropout=0.1, residual_connections=True)
rnn_layer = rnn.RNN(cell,
bidirectional=True,
reducer=reducer.ConcatReducer())
inputs = tf.random.uniform([4, 5, 5])
outputs, states = rnn_layer(inputs, training=True)
self.assertListEqual(outputs.shape.as_list(), [4, 5, 20])
self.assertIsInstance(states, tuple)
self.assertEqual(len(states), 3)
self.assertEqual(len(states[0]), 2)
self.assertListEqual(states[0][0].shape.as_list(), [4, 20])
def __init__(self,
num_layers,
num_units,
reducer=SumReducer(),
cell_class=None,
dropout=0.3,
residual_connections=False,
**kwargs):
"""Initializes the parameters of the encoder.
Args:
num_layers: The number of layers.
num_units: The number of units in each layer.
reducer: A :class:`opennmt.layers.reducer.Reducer` instance to merge
bidirectional state and outputs.
cell_class: The inner cell class or a callable taking :obj:`num_units` as
argument and returning a cell. Defaults to a LSTM cell.
dropout: The probability to drop units in each layer output.
residual_connections: If ``True``, each layer input will be added to its
output.
Raises:
ValueError: when using :class:`opennmt.layers.reducer.ConcatReducer` and
:obj:`num_units` is not divisible by 2.
"""
if isinstance(reducer, ConcatReducer):
if num_units % 2 != 0:
raise ValueError("num_units must be divisible by 2 to use the ConcatReducer.")
num_units /= 2
super(BidirectionalRNNEncoderV2, self).__init__(**kwargs)
cell = rnn.make_rnn_cell(
num_layers,
num_units,
dropout=dropout,
residual_connections=residual_connections,
cell_class=cell_class)
self.rnn = rnn.RNN(cell, bidirectional=True, reducer=reducer)