未验证 提交 89bbc4f6 编写于 作者: Y Yang yaming 提交者: GitHub

Merge pull request #7157 from pkuyym/fix-7156

Doc fix and enhancement for lstm_unit python wrapper.
......@@ -1168,25 +1168,26 @@ def lstm_unit(x_t,
.. math::
i_t & = \sigma(W_{x_i}x_{t} + W_{h_i}h_{t-1} + W_{c_i}c_{t-1} + b_i)
i_t & = \sigma(W_{x_i}x_{t} + W_{h_i}h_{t-1} + b_i)
f_t & = \sigma(W_{x_f}x_{t} + W_{h_f}h_{t-1} + W_{c_f}c_{t-1} + b_f)
f_t & = \sigma(W_{x_f}x_{t} + W_{h_f}h_{t-1} + b_f)
c_t & = f_tc_{t-1} + i_t tanh (W_{x_c}x_t+W_{h_c}h_{t-1} + b_c)
c_t & = f_tc_{t-1} + i_t tanh (W_{x_c}x_t + W_{h_c}h_{t-1} + b_c)
o_t & = \sigma(W_{x_o}x_{t} + W_{h_o}h_{t-1} + W_{c_o}c_t + b_o)
o_t & = \sigma(W_{x_o}x_{t} + W_{h_o}h_{t-1} + b_o)
h_t & = o_t tanh(c_t)
The inputs of lstm unit includes :math:`x_t`, :math:`h_{t-1}` and
:math:`c_{t-1}`. The implementation separates the linear transformation
and non-linear transformation apart. Here, we take :math:`i_t` as an
example. The linear transformation is applied by calling a `fc` layer and
the equation is:
The inputs of lstm unit include :math:`x_t`, :math:`h_{t-1}` and
:math:`c_{t-1}`. The 2nd dimensions of :math:`h_{t-1}` and :math:`c_{t-1}`
should be same. The implementation separates the linear transformation and
non-linear transformation apart. Here, we take :math:`i_t` as an example.
The linear transformation is applied by calling a `fc` layer and the
equation is:
.. math::
L_{i_t} = W_{x_i}x_{t} + W_{h_i}h_{t-1} + W_{c_i}c_{t-1} + b_i
L_{i_t} = W_{x_i}x_{t} + W_{h_i}h_{t-1} + b_i
The non-linear transformation is applied by calling `lstm_unit_op` and the
equation is:
......@@ -1198,9 +1199,12 @@ def lstm_unit(x_t,
This layer has two outputs including :math:`h_t` and :math:`o_t`.
Args:
x_t (Variable): The input value of current step.
hidden_t_prev (Variable): The hidden value of lstm unit.
cell_t_prev (Variable): The cell value of lstm unit.
x_t (Variable): The input value of current step, a 2-D tensor with shape
M x N, M for batch size and N for input size.
hidden_t_prev (Variable): The hidden value of lstm unit, a 2-D tensor
with shape M x S, M for batch size and S for size of lstm unit.
cell_t_prev (Variable): The cell value of lstm unit, a 2-D tensor with
shape M x S, M for batch size and S for size of lstm unit.
forget_bias (float): The forget bias of lstm unit.
param_attr (ParamAttr): The attributes of parameter weights, used to set
initializer, name etc.
......@@ -1213,14 +1217,15 @@ def lstm_unit(x_t,
Raises:
ValueError: The ranks of **x_t**, **hidden_t_prev** and **cell_t_prev**\
not be 2 or the 1st dimensions of **x_t**, **hidden_t_prev** \
and **cell_t_prev** not be the same.
and **cell_t_prev** not be the same or the 2nd dimensions of \
**hidden_t_prev** and **cell_t_prev** not be the same.
Examples:
.. code-block:: python
x_t = fluid.layers.fc(input=x_t_data, size=10)
prev_hidden = fluid.layers.fc(input=prev_hidden_data, size=20)
prev_hidden = fluid.layers.fc(input=prev_hidden_data, size=30)
prev_cell = fluid.layers.fc(input=prev_cell_data, size=30)
hidden_value, cell_value = fluid.layers.lstm_unit(x_t=x_t,
hidden_t_prev=prev_hidden,
......@@ -1239,7 +1244,11 @@ def lstm_unit(x_t,
if x_t.shape[0] != hidden_t_prev.shape[0] or x_t.shape[
0] != cell_t_prev.shape[0]:
raise ValueError("The 1s dimension of x_t, hidden_t_prev and "
raise ValueError("The 1st dimensions of x_t, hidden_t_prev and "
"cell_t_prev must be the same.")
if hidden_t_prev.shape[1] != cell_t_prev.shape[1]:
raise ValueError("The 2nd dimensions of hidden_t_prev and "
"cell_t_prev must be the same.")
if bias_attr is None:
......
......@@ -177,8 +177,8 @@ class TestBook(unittest.TestCase):
name='x_t_data', shape=[10, 10], dtype='float32')
x_t = layers.fc(input=x_t_data, size=10)
prev_hidden_data = layers.data(
name='prev_hidden_data', shape=[10, 20], dtype='float32')
prev_hidden = layers.fc(input=prev_hidden_data, size=20)
name='prev_hidden_data', shape=[10, 30], dtype='float32')
prev_hidden = layers.fc(input=prev_hidden_data, size=30)
prev_cell_data = layers.data(
name='prev_cell', shape=[10, 30], dtype='float32')
prev_cell = layers.fc(input=prev_cell_data, size=30)
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册