未验证 提交 c2ad3815 编写于 作者: Y Yu Yang 提交者: GitHub

Merge pull request #11560 from JiayiFeng/doc_non_layer_api

Doc of non layer api
......@@ -36,6 +36,25 @@ def _is_number_or_matrix_(var):
class WeightedAverage(object):
"""
Calculate weighted average.
The average calculating is accomplished via Python totally.
They do not change Paddle's Program, nor do anything to
modify NN model's configuration. They are completely
wrappers of Python functions.
Examples:
.. code-block:: python
avg = fluid.average.WeightedAverage()
avg.add(value=2.0, weight=1)
avg.add(value=4.0, weight=2)
avg.eval()
# The result is 3.333333333.
# For (2.0 * 1 + 4.0 * 2) / (1 + 2) = 3.333333333
"""
def __init__(self):
warnings.warn(
"The %s is deprecated, please use fluid.metrics.Accuracy instead." %
......
......@@ -435,18 +435,65 @@ def _get_stop_gradients_(program):
def append_backward(loss, parameter_list=None, no_grad_set=None,
callbacks=None):
"""
Append backward part to main_program
Append backward part to main_program.
Args:
loss(Variable): The variable generated by cost function.
parameter_list(list[string]): Parameters that need to be updated by
optimizer. If None, it means all parameters need to be updated.
no_grad_set(set): Variables that have no gradients in Block 0.
All variables with `step_gradient=True` from all blocks will be
automatically added.
A complete neural network training is made up of forward and backward
propagation. However, when we configure a network, we only need to
specify its forwrd part. The backward part is generated automatically
according to the forward part by this function.
Return:
(list[(Variable,Variable)]): list of (parameter, gradient) pair.
In most cases, users do not need to invoke this function manually. It
will be automatically invoked by the optimizer's `minimize` function.
Args:
loss(Variable): The loss variable of the network.
parameter_list(list[string]|None): Names of parameters that need
to be updated by optimizers.
If it is None, all parameters
will be updated.
Default: None
no_grad_set(set|None): Variables in the Block 0 whose gradients
should be ignored. All variables with
`step_gradient=True` from all blocks will
be automatically added into this set.
Default: None
callbacks(list[callable object]|None): The callbacks are used for
doing some custom jobs during
backward part building. All
callable objects in it will
be invoked once each time a
new gradient operator is added
into the program. The callable
object must has two input
parameters: 'block' and 'context'.
The 'block' is the block which
the new gradient operator will
be added to. The 'context' is a
map, whose keys are gradient
variable names and values are
corresponding original variables.
In addition to this, the 'context'
has another special key-value pair:
the key is string '__current_op_desc__'
and the value is the op_desc of the
gradient operator who has just
triggered the callable object.
Returns:
list[(Variable,Variable)]: Pairs of parameter and its
corresponding gradients. The key is the parameter and the
value is gradient variable.
Raises:
AssertionError: If `loss` is not an instance of Variable.
Examples:
.. code-block:: python
# network configuration code
# ...
avg_loss = fluid.layers.mean(loss)
param_grad_list = fluid.backward.append_backward(loss=avg_loss)
"""
assert isinstance(loss, framework.Variable)
......
此差异已折叠。
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