experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Update docstring for keras.layers.SimpleRNN and SimpleRNNCell.

Browse Source 
PiperOrigin-RevId: 270153457
This commit is contained in:
Scott Zhu 2019-09-19 16:41:16 -07:00 committed by TensorFlower Gardener
parent 9414a67fc4
commit e627965279
1 changed files with 90 additions and 46 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

136
tensorflow/python/keras/layers/recurrent.py
View File

@ -1131,41 +1131,66 @@ class DropoutRNNCellMixin(object):
class SimpleRNNCell(DropoutRNNCellMixin, Layer): class SimpleRNNCell(DropoutRNNCellMixin, Layer):
"""Cell class for SimpleRNN. """Cell class for SimpleRNN.
This class processes one step within the whole time sequence input, whereas
`tf.keras.layer.SimpleRNN` processes the whole sequence.
Arguments: Arguments:
units: Positive integer, dimensionality of the output space. units: Positive integer, dimensionality of the output space.
activation: Activation function to use. activation: Activation function to use.
Default: hyperbolic tangent (`tanh`). Default: hyperbolic tangent (`tanh`).
If you pass `None`, no activation is applied If you pass `None`, no activation is applied
(ie. "linear" activation: `a(x) = x`). (ie. "linear" activation: `a(x) = x`).
use_bias: Boolean, whether the layer uses a bias vector. use_bias: Boolean, (default `True`), whether the layer uses a bias vector.
kernel_initializer: Initializer for the `kernel` weights matrix, kernel_initializer: Initializer for the `kernel` weights matrix,
used for the linear transformation of the inputs. used for the linear transformation of the inputs. Default:
`glorot_uniform`.
recurrent_initializer: Initializer for the `recurrent_kernel` recurrent_initializer: Initializer for the `recurrent_kernel`
weights matrix, used for the linear transformation of the recurrent state. weights matrix, used for the linear transformation of the recurrent state.
bias_initializer: Initializer for the bias vector. Default: `orthogonal`.
kernel_regularizer: Regularizer function applied to bias_initializer: Initializer for the bias vector. Default: `zeros`.
the `kernel` weights matrix. kernel_regularizer: Regularizer function applied to the `kernel` weights
recurrent_regularizer: Regularizer function applied to matrix. Default: `None`.
the `recurrent_kernel` weights matrix. recurrent_regularizer: Regularizer function applied to the
bias_regularizer: Regularizer function applied to the bias vector. `recurrent_kernel` weights matrix. Default: `None`.
kernel_constraint: Constraint function applied to bias_regularizer: Regularizer function applied to the bias vector. Default:
the `kernel` weights matrix. `None`.
recurrent_constraint: Constraint function applied to kernel_constraint: Constraint function applied to the `kernel` weights
the `recurrent_kernel` weights matrix. matrix. Default: `None`.
bias_constraint: Constraint function applied to the bias vector. recurrent_constraint: Constraint function applied to the `recurrent_kernel`
dropout: Float between 0 and 1. weights matrix. Default: `None`.
Fraction of the units to drop for bias_constraint: Constraint function applied to the bias vector. Default:
the linear transformation of the inputs. `None`.
recurrent_dropout: Float between 0 and 1. dropout: Float between 0 and 1. Fraction of the units to drop for the linear
Fraction of the units to drop for transformation of the inputs. Default: 0.
the linear transformation of the recurrent state. recurrent_dropout: Float between 0 and 1. Fraction of the units to drop for
the linear transformation of the recurrent state. Default: 0.
Call arguments: Call arguments:
inputs: A 2D tensor. inputs: A 2D tensor, with shape of `[batch, feature]`.
states: List of state tensors corresponding to the previous timestep. states: A 2D tensor with shape of `[batch, units]`, which is the state from
the previous time step. For timestep 0, the initial state provided by user
will be feed to cell.
training: Python boolean indicating whether the layer should behave in training: Python boolean indicating whether the layer should behave in
training mode or in inference mode. Only relevant when `dropout` or training mode or in inference mode. Only relevant when `dropout` or
`recurrent_dropout` is used. `recurrent_dropout` is used.
Examples:
```python
inputs = np.random.random([32, 10, 8]).astype(np.float32)
rnn = tf.keras.layers.RNN(tf.keras.layers.SimpleRNNCell(4))
output = rnn(inputs) # The output has shape `[32, 4]`.
rnn = tf.keras.layers.RNN(
tf.keras.layers.SimpleRNNCell(4),
return_sequences=True,
return_state=True)
# whole_sequence_output has shape `[32, 10, 4]`.
# final_state has shape `[32, 4]`.
whole_sequence_output, final_state = rnn(inputs)
```
""" """
def __init__(self, def __init__(self,
@ -1300,35 +1325,38 @@ class SimpleRNN(RNN):
Default: hyperbolic tangent (`tanh`). Default: hyperbolic tangent (`tanh`).
If you pass None, no activation is applied If you pass None, no activation is applied
(ie. "linear" activation: `a(x) = x`). (ie. "linear" activation: `a(x) = x`).
use_bias: Boolean, whether the layer uses a bias vector. use_bias: Boolean, (default `True`), whether the layer uses a bias vector.
kernel_initializer: Initializer for the `kernel` weights matrix, kernel_initializer: Initializer for the `kernel` weights matrix,
used for the linear transformation of the inputs. used for the linear transformation of the inputs. Default:
`glorot_uniform`.
recurrent_initializer: Initializer for the `recurrent_kernel` recurrent_initializer: Initializer for the `recurrent_kernel`
weights matrix, weights matrix, used for the linear transformation of the recurrent state.
used for the linear transformation of the recurrent state. Default: `orthogonal`.
bias_initializer: Initializer for the bias vector. bias_initializer: Initializer for the bias vector. Default: `zeros`.
kernel_regularizer: Regularizer function applied to kernel_regularizer: Regularizer function applied to the `kernel` weights
the `kernel` weights matrix. matrix. Default: `None`.
recurrent_regularizer: Regularizer function applied to recurrent_regularizer: Regularizer function applied to the
the `recurrent_kernel` weights matrix. `recurrent_kernel` weights matrix. Default: `None`.
bias_regularizer: Regularizer function applied to the bias vector. bias_regularizer: Regularizer function applied to the bias vector. Default:
activity_regularizer: Regularizer function applied to `None`.
the output of the layer (its "activation").. activity_regularizer: Regularizer function applied to the output of the
kernel_constraint: Constraint function applied to layer (its "activation"). Default: `None`.
the `kernel` weights matrix. kernel_constraint: Constraint function applied to the `kernel` weights
recurrent_constraint: Constraint function applied to matrix. Default: `None`.
the `recurrent_kernel` weights matrix. recurrent_constraint: Constraint function applied to the `recurrent_kernel`
bias_constraint: Constraint function applied to the bias vector. weights matrix. Default: `None`.
bias_constraint: Constraint function applied to the bias vector. Default:
`None`.
dropout: Float between 0 and 1. dropout: Float between 0 and 1.
Fraction of the units to drop for Fraction of the units to drop for the linear transformation of the inputs.
the linear transformation of the inputs. Default: 0.
recurrent_dropout: Float between 0 and 1. recurrent_dropout: Float between 0 and 1.
Fraction of the units to drop for Fraction of the units to drop for the linear transformation of the
the linear transformation of the recurrent state. recurrent state. Default: 0.
return_sequences: Boolean. Whether to return the last output return_sequences: Boolean. Whether to return the last output
in the output sequence, or the full sequence. in the output sequence, or the full sequence. Default: `False`.
return_state: Boolean. Whether to return the last state return_state: Boolean. Whether to return the last state
in addition to the output. in addition to the output. Default: `False`
go_backwards: Boolean (default False). go_backwards: Boolean (default False).
If True, process the input sequence backwards and return the If True, process the input sequence backwards and return the
reversed sequence. reversed sequence.
@ -1343,8 +1371,8 @@ class SimpleRNN(RNN):
Unrolling is only suitable for short sequences. Unrolling is only suitable for short sequences.
Call arguments: Call arguments:
inputs: A 3D tensor. inputs: A 3D tensor, with shape `[batch, timesteps, feature]`.
mask: Binary tensor of shape `(samples, timesteps)` indicating whether mask: Binary tensor of shape `[batch, timesteps]` indicating whether
a given timestep should be masked. a given timestep should be masked.
training: Python boolean indicating whether the layer should behave in training: Python boolean indicating whether the layer should behave in
training mode or in inference mode. This argument is passed to the cell training mode or in inference mode. This argument is passed to the cell
@ -1352,6 +1380,22 @@ class SimpleRNN(RNN):
`recurrent_dropout` is used. `recurrent_dropout` is used.
initial_state: List of initial state tensors to be passed to the first initial_state: List of initial state tensors to be passed to the first
call of the cell. call of the cell.
Examples:
```python
inputs = np.random.random([32, 10, 8]).astype(np.float32)
simple_rnn = tf.keras.layers.SimpleRNN(4)
output = simple_rnn(inputs) # The output has shape `[32, 4]`.
simple_rnn = tf.keras.layers.SimpleRNN(
4, return_sequences=True, return_state=True)
# whole_sequence_output has shape `[32, 10, 4]`.
# final_state has shape `[32, 4]`.
whole_sequence_output, final_state = simple_rnn(inputs)
```
""" """
def __init__(self, def __init__(self,