Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
PaddlePaddle
Paddle
提交
7d8e8d90
P
Paddle
项目概览
PaddlePaddle
/
Paddle
大约 1 年 前同步成功
通知
2297
Star
20931
Fork
5422
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1423
列表
看板
标记
里程碑
合并请求
543
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
Paddle
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
1,423
Issue
1,423
列表
看板
标记
里程碑
合并请求
543
合并请求
543
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
体验新版 GitCode,发现更多精彩内容 >>
未验证
提交
7d8e8d90
编写于
12月 23, 2017
作者:
C
Cao Ying
提交者:
GitHub
12月 23, 2017
浏览文件
操作
浏览文件
下载
差异文件
Merge pull request #6654 from ranqiu92/doc
Update annotations of layers.py.
上级
68f2a962
ddd41582
变更
1
显示空白变更内容
内联
并排
Showing
1 changed file
with
124 addition
and
113 deletion
+124
-113
python/paddle/trainer_config_helpers/layers.py
python/paddle/trainer_config_helpers/layers.py
+124
-113
未找到文件。
python/paddle/trainer_config_helpers/layers.py
浏览文件 @
7d8e8d90
...
...
@@ -270,7 +270,7 @@ class LayerType(object):
@
staticmethod
def
is_layer_type
(
type_name
):
"""
If
type_name is a layer type.
Whether
type_name is a layer type.
:param type_name: layer type name. Because layer type enumerations are
strings.
...
...
@@ -441,7 +441,7 @@ def full_matrix_projection(input, size=0, param_attr=None):
with mixed_layer(size=100) as m:
m += full_matrix_projection(input=layer)
2. When used as an independ
a
nt object like this, you must set the size:
2. When used as an independ
e
nt object like this, you must set the size:
.. code-block:: python
...
...
@@ -451,11 +451,11 @@ def full_matrix_projection(input, size=0, param_attr=None):
:param input: The input of this layer.
:type input: LayerOutput
:param size: The
parameter size. Means the width of paramet
er.
:param size: The
dimension of this lay
er.
:type size: int
:param param_attr:
Parameter config, None if use default
.
:param param_attr:
The parameter attribute. See ParameterAttribute for details
.
:type param_attr: ParameterAttribute
:return:
A
FullMatrixProjection Object.
:return: FullMatrixProjection Object.
:rtype: FullMatrixProjection
"""
proj
=
FullMatrixProjection
(
...
...
@@ -468,12 +468,12 @@ def full_matrix_projection(input, size=0, param_attr=None):
def
trans_full_matrix_projection
(
input
,
size
=
0
,
param_attr
=
None
):
"""
Different from full_matrix_projection, this projection performs matrix
multiplication, using transpose of weight.
multiplication, using t
he t
ranspose of weight.
.. math::
out.row[i] += in.row[i] * w^\mathrm{T}
:math:`w^\mathrm{T}` means transpose of weight.
:math:`w^\mathrm{T}` means t
he t
ranspose of weight.
The simply usage is:
.. code-block:: python
...
...
@@ -489,9 +489,9 @@ def trans_full_matrix_projection(input, size=0, param_attr=None):
:type input: LayerOutput
:param size: The parameter size. Means the width of parameter.
:type size: int
:param param_attr:
Parameter config, None if use default
.
:param param_attr:
The parameter attribute. See ParameterAttribute for details
.
:type param_attr: ParameterAttribute
:return:
A
TransposedFullMatrixProjection Object.
:return: TransposedFullMatrixProjection Object.
:rtype: TransposedFullMatrixProjection
"""
proj
=
TransposedFullMatrixProjection
(
...
...
@@ -521,7 +521,7 @@ def table_projection(input, size=0, param_attr=None):
with mixed_layer(size=100) as m:
m += table_projection(input=layer)
2. When used as an independ
a
nt object like this, you must set the size:
2. When used as an independ
e
nt object like this, you must set the size:
.. code-block:: python
...
...
@@ -532,11 +532,11 @@ def table_projection(input, size=0, param_attr=None):
:param input: The input of this layer, which must contains id fields.
:type input: LayerOutput
:param size: The
parameter size. Means the width of parameter
.
:param size: The
dimension of the output
.
:type size: int
:param param_attr:
Parameter config, None if use default
.
:param param_attr:
The parameter attribute. See ParameterAttribute for details
.
:type param_attr: ParameterAttribute
:return:
A
TableProjection Object.
:return: TableProjection Object.
:rtype: TableProjection
"""
proj
=
TableProjection
(
...
...
@@ -547,7 +547,7 @@ def table_projection(input, size=0, param_attr=None):
def
identity_projection
(
input
,
offset
=
None
,
size
=
None
):
"""
1. I
dentityProjection if offset=None. It perform
s:
1. I
f offset=None, it performs IdentityProjection as follow
s:
.. math::
out.row[i] += in.row[i]
...
...
@@ -559,9 +559,8 @@ def identity_projection(input, offset=None, size=None):
proj = identity_projection(input=layer)
2. IdentityOffsetProjection if offset!=None. It likes IdentityProjection,
but layer size may be smaller than input size.
It select dimesions [offset, offset+layer_size) from input:
2. If offset!=None, It executes IdentityOffsetProjection and takes the
elements of the input in the range [offset, offset+size) as output.
.. math::
out.row[i] += in.row[i +
\\
textrm{offset}]
...
...
@@ -573,14 +572,20 @@ def identity_projection(input, offset=None, size=None):
proj = identity_projection(input=layer,
offset=10)
Note that
both of two projections should not have any
parameter.
Note that
neither of the projections have trainable
parameter.
:param input: The input of this layer.
:type input: LayerOutput
:param offset: Offset, None if use default.
:param offset: The offset from the start of the input. The input's
elements in the range [offset, offset+size) will be
taken as output. If this parameter is not set or set
to None, the output will be the same as the input.
:type offset: int
:return: A IdentityProjection or IdentityOffsetProjection object
:rtype: IdentityProjection or IdentityOffsetProjection
:param size: The dimension of this layer. It will be neglected
when offset is None or not set.
:type size: int
:return: IdentityProjection or IdentityOffsetProjection object
:rtype: IdentityProjection | IdentityOffsetProjection
"""
if
offset
is
None
:
proj
=
IdentityProjection
(
input_layer_name
=
input
.
name
)
...
...
@@ -596,8 +601,8 @@ def identity_projection(input, offset=None, size=None):
def
slice_projection
(
input
,
slices
):
"""
slice_projection
can slice
the input value into multiple parts,
and then select some of them to merge
into a new output.
slice_projection
slices
the input value into multiple parts,
then selects and merges some of them
into a new output.
.. math::
output = [input.slices()]
...
...
@@ -608,15 +613,13 @@ def slice_projection(input, slices):
proj = slice_projection(input=layer, slices=[(0, 10), (20, 30)])
Note that slice_projection
should not have any
parameter.
Note that slice_projection
has no trainable
parameter.
:param input: The input of this layer.
:type input: LayerOutput
:param slices: An array of slice parameters.
Each slice contains the start and end offsets based
on the input.
:type slices: pair of int
:return: A SliceProjection object
:param slices: A list of start and end offsets of each slice.
:type slices: list of tuple
:return: SliceProjection object.
:rtype: SliceProjection
"""
assert
len
(
slices
)
>=
1
...
...
@@ -636,8 +639,7 @@ def slice_projection(input, slices):
@
wrap_param_attr_default
()
def
scaling_projection
(
input
,
param_attr
=
None
):
"""
scaling_projection multiplies the input with a scalar parameter and add to
the output.
scaling_projection multiplies the input with a scalar parameter.
.. math::
out += w * in
...
...
@@ -650,9 +652,9 @@ def scaling_projection(input, param_attr=None):
:param input: The input of this layer.
:type input: LayerOutput
:param param_attr:
Parameter config, None if use default
.
:param param_attr:
The parameter attribute. See ParameterAttribute for details
.
:type param_attr: ParameterAttribute
:return:
A ScalingProjection object
:return:
ScalingProjection object.
:rtype: ScalingProjection
"""
proj
=
ScalingProjection
(
input_layer_name
=
input
.
name
,
**
param_attr
.
attr
)
...
...
@@ -663,8 +665,8 @@ def scaling_projection(input, param_attr=None):
@
wrap_param_attr_default
()
def
dotmul_projection
(
input
,
param_attr
=
None
):
"""
DotMulProjection
with a layer as input.
It performs
element-wise multiplication with weight.
DotMulProjection
takes a layer as input and performs
element-wise multiplication with weight.
.. math::
out.row[i] += in.row[i] .* weight
...
...
@@ -679,9 +681,9 @@ def dotmul_projection(input, param_attr=None):
:param input: The input of this layer.
:type input: LayerOutput
:param param_attr:
Parameter config, None if use default
.
:param param_attr:
The parameter attribute. See ParameterAttribute for details
.
:type param_attr: ParameterAttribute
:return:
A DotMulProjection O
bject.
:return:
DotMulProjection o
bject.
:rtype: DotMulProjection
"""
proj
=
DotMulProjection
(
...
...
@@ -698,7 +700,7 @@ def dotmul_operator(a=None, b=None, scale=1, **kwargs):
out.row[i] += scale * (a.row[i] .* b.row[i])
where :math:`.*` means element-wise multiplication, and
scale is a config scalar, its default value is
one
.
scale is a config scalar, its default value is
1
.
The example usage is:
...
...
@@ -706,13 +708,13 @@ def dotmul_operator(a=None, b=None, scale=1, **kwargs):
op = dotmul_operator(a=layer1, b=layer2, scale=0.5)
:param a:
Input layer1
:param a:
The first input of this layer.
:type a: LayerOutput
:param b:
Input layer2
:param b:
The second input of this layer.
:type b: LayerOutput
:param scale:
config scalar, default value is one
.
:param scale:
A scalar to scale the product. Its default value is 1
.
:type scale: float
:return:
A DotMulOperator O
bject.
:return:
DotMulOperator o
bject.
:rtype: DotMulOperator
"""
if
'x'
in
kwargs
or
'y'
in
kwargs
:
...
...
@@ -738,28 +740,29 @@ def context_projection(input,
"""
Context Projection.
It just
simply reorganizes input sequence, combines "context_len" sequenc
e
to one context from context_start. "context_start" will be set to
-(context_len - 1) / 2 by default.
If context position
out of sequence
It just
reorganizes input sequence, combines "context_len" elements of th
e
sequence
to one context from context_start. "context_start" will be set to
-(context_len - 1) / 2 by default.
When context position is
out of sequence
length, padding will be filled as zero if padding_attr = False, otherwise
it is trainable.
For example, origin sequence is [A B C D E F G], context len is 3,
then
after context projection and not set padding_attr
, sequence will
For example, origin sequence is [A B C D E F G], context len is 3,
padding_attr
is not set, then after context projection
, sequence will
be [ 0AB ABC BCD CDE DEF EFG FG0 ].
:param input: The input of this layer, which should be a sequence.
:type input: LayerOutput
:param context_len:
context length
.
:param context_len:
The length of the context
.
:type context_len: int
:param context_start:
context start position. Default
is
:param context_start:
The start position of the context. The default value
is
-(context_len - 1)/2
:type context_start: int
:param padding_attr: Padding Parameter Attribute. If false, it means padding
always be zero. Otherwise Padding is learnable, and
parameter attribute is set by this parameter.
:param padding_attr: Parameter attribute of the padding. If the parameter is
set to False, padding will be zero. In other cases, the
padding is trainable, and its parameter attribute is set
by this parameter.
:type padding_attr: bool | ParameterAttribute
:return: Projection
:return: Projection
object.
:rtype: Projection
"""
context_start
=
-
(
...
...
@@ -791,10 +794,9 @@ class MixedLayerType(LayerOutput):
def
__init__
(
self
,
name
,
size
,
act
,
bias_attr
,
layer_attr
,
parents
=
None
):
"""
Ctor.
:param name: layer name.
:param name: The name of this layer.
:type name: basestring
:param size:
layer size
.
:param size:
The dimension of this layer
.
:type size: int
:param act: Activation type.
:type act: BaseActivation
...
...
@@ -802,8 +804,9 @@ class MixedLayerType(LayerOutput):
whose type is not ParameterAttribute, no bias is defined. If the
parameter is set to True, the bias is initialized to zero.
:type bias_attr: ParameterAttribute | None | bool | Any
:param layer_attr: Extra Layer Attribute.
:type layer_attr: ExtraLayerAttribute or None
:param layer_attr: The extra layer attribute. See ExtraLayerAttribute for
details.
:type layer_attr: ExtraLayerAttribute | None
"""
LayerOutput
.
__init__
(
self
,
...
...
@@ -868,12 +871,12 @@ def mixed_layer(size=0,
bias_attr
=
False
,
layer_attr
=
None
):
"""
Mixed Layer. A mixed layer will add all inputs together, then activate.
Each input
s
is a projection or operator.
Mixed Layer. A mixed layer will add all inputs together, then activate
the sum
.
Each input is a projection or operator.
There are two styles of usages.
1. When
not set inputs parameter
, use mixed_layer like this:
1. When
the parameter input is not set
, use mixed_layer like this:
.. code-block:: python
...
...
@@ -889,21 +892,21 @@ def mixed_layer(size=0,
input=[full_matrix_projection(input=layer1),
full_matrix_projection(input=layer2)])
:param name:
mixed layer name. Can be referenced by other layer
.
:param name:
The name of this layer. It is optional
.
:type name: basestring
:param size:
layer size
.
:param size:
The dimension of this layer
.
:type size: int
:param input: The input of this layer. It is an optional parameter. If set,
then this function will just return layer's name.
:param input: The input of this layer. It is an optional parameter.
:param act: Activation Type. LinearActivation is the default activation.
:type act: BaseActivation
:param bias_attr: The bias attribute. If the parameter is set to False or an object
whose type is not ParameterAttribute, no bias is defined. If the
parameter is set to True, the bias is initialized to zero.
:type bias_attr: ParameterAttribute | None | bool | Any
:param layer_attr: The extra layer config. Default is None.
:param layer_attr: The extra layer attribute. See ExtraLayerAttribute for
details.
:type layer_attr: ExtraLayerAttribute
:return: MixedLayerType object
can add inputs or layer name
.
:return: MixedLayerType object.
:rtype: MixedLayerType
"""
...
...
@@ -938,14 +941,15 @@ def data_layer(name, size, depth=None, height=None, width=None,
:param name: The name of this layer.
:type name: basestring
:param size:
Size
of this data layer.
:param size:
The dimension
of this data layer.
:type size: int
:param height:
Height of this data layer, used for image
:param height:
The height of the input image data.
:type height: int | None
:param width:
Width of this data layer, used for image
:param width:
The width of the input image data.
:type width: int | None
:param layer_attr: Extra Layer Attribute.
:type layer_attr: ExtraLayerAttribute.
:param layer_attr: The extra layer attribute. See ExtraLayerAttribute for
details.
:type layer_attr: ExtraLayerAttribute
:return: LayerOutput object.
:rtype: LayerOutput
"""
...
...
@@ -978,14 +982,15 @@ def embedding_layer(input, size, name=None, param_attr=None, layer_attr=None):
:param name: The name of this layer. It is optional.
:type name: basestring
:param input: The input of this layer, wh
ich
must be Index Data.
:param input: The input of this layer, wh
ose type
must be Index Data.
:type input: LayerOutput
:param size: The
embedding dimension
.
:param size: The
dimension of the embedding vector
.
:type size: int
:param param_attr: The embedding parameter attribute. See ParameterAttribute
for details.
:type param_attr: ParameterAttribute | None
:param layer_attr: Extra layer Config. Default is None.
:type param_attr: ParameterAttribute
:param layer_attr: The extra layer attribute. See ExtraLayerAttribute for
details.
:type layer_attr: ExtraLayerAttribute | None
:return: LayerOutput object.
:rtype: LayerOutput
...
...
@@ -1013,7 +1018,7 @@ def fc_layer(input,
bias_attr
=
None
,
layer_attr
=
None
):
"""
Helper for declar
e fully connected layer.
Th
e fully connected layer.
The example usage is:
...
...
@@ -1035,17 +1040,18 @@ def fc_layer(input,
:type name: basestring
:param input: The input of this layer.
:type input: LayerOutput | list | tuple
:param size: The
layer dimension
.
:param size: The
dimension of this layer
.
:type size: int
:param act: Activation Type. TanhActivation is the default activation.
:type act: BaseActivation
:param param_attr: The
Parameter Attribute|list
.
:param param_attr: The
parameter attribute. See ParameterAttribute for details
.
:type param_attr: ParameterAttribute
:param bias_attr: The bias attribute. If the parameter is set to False or an object
whose type is not ParameterAttribute, no bias is defined. If the
parameter is set to True, the bias is initialized to zero.
:type bias_attr: ParameterAttribute | None | bool | Any
:param layer_attr: Extra Layer config.
:param layer_attr: The extra layer attribute. See ExtraLayerAttribute for
details.
:type layer_attr: ExtraLayerAttribute | None
:return: LayerOutput object.
:rtype: LayerOutput
...
...
@@ -1086,13 +1092,15 @@ def fc_layer(input,
@
wrap_name_default
(
"print"
)
def
printer_layer
(
input
,
format
=
None
,
name
=
None
):
"""
Print the output value of input layers. This layer is useful for debugging.
Print the output value of the layers specified by the parameter input.
This layer is useful for debugging.
:param name: The name of this layer. It is optional.
:type name: basestring
:param input: The input of this layer.
:type input: LayerOutput | list | tuple
:return: LayerOutput
:return: LayerOutput object.
:rtype: LayerOutput
"""
if
isinstance
(
input
,
LayerOutput
):
input
=
[
input
]
...
...
@@ -1135,11 +1143,12 @@ def priorbox_layer(input,
:param aspect_ratio: The aspect ratio.
:type aspect_ratio: list
:param variance: The bounding box variance.
:type min_size: The min size of the priorbox width/height.
:type min_size: The min
imum
size of the priorbox width/height.
:param min_size: list
:type max_size: The max
size of the priorbox width/height. C
ould be NULL.
:type max_size: The max
imum size of the priorbox width/height. It c
ould be NULL.
:param max_size: list
:return: LayerOutput
:return: LayerOutput object.
:rtype: LayerOutput
"""
# plus one for ratio 1.
num_filters
=
(
len
(
aspect_ratio
)
*
2
+
1
+
len
(
max_size
))
*
4
...
...
@@ -1177,7 +1186,7 @@ def multibox_loss_layer(input_loc,
:param name: The name of this layer. It is optional.
:type name: basestring
:param input_loc: The input predict locations.
:param input_loc: The input predict
ed
locations.
:type input_loc: LayerOutput | List of LayerOutput
:param input_conf: The input priorbox confidence.
:type input_conf: LayerOutput | List of LayerOutput
...
...
@@ -1189,13 +1198,15 @@ def multibox_loss_layer(input_loc,
:type num_classes: int
:param overlap_threshold: The threshold of the overlap.
:type overlap_threshold: float
:param neg_pos_ratio: The ratio of the negative bbox to the positive bbox.
:param neg_pos_ratio: The ratio of the negative bounding box to
the positive bounding box.
:type neg_pos_ratio: float
:param neg_overlap: The negative bbox overlap threshold.
:param neg_overlap: The negative b
ounding
box overlap threshold.
:type neg_overlap: float
:param background_id: The background class index.
:type background_id: int
:return: LayerOutput
:return: LayerOutput object.
:rtype: LayerOutput
"""
if
isinstance
(
input_loc
,
LayerOutput
):
input_loc
=
[
input_loc
]
...
...
@@ -1258,19 +1269,20 @@ def detection_output_layer(input_loc,
:type input_conf: LayerOutput | List of LayerOutput.
:param priorbox: The input priorbox location and the variance.
:type priorbox: LayerOutput
:param num_classes: The number of the class
ification
.
:param num_classes: The number of the class
es
.
:type num_classes: int
:param nms_threshold: The Non-maximum suppression threshold.
:type nms_threshold: float
:param nms_top_k: The b
box number kept of the NMS's output
:param nms_top_k: The b
ounding boxes number kept of the NMS's output.
:type nms_top_k: int
:param keep_top_k: The b
box number kept of the layer's output
:param keep_top_k: The b
ounding boxes number kept of the layer's output.
:type keep_top_k: int
:param confidence_threshold: The classification confidence threshold
:param confidence_threshold: The classification confidence threshold
.
:type confidence_threshold: float
:param background_id: The background class index.
:type background_id: int
:return: LayerOutput
:return: LayerOutput object.
:rtype: LayerOutput
"""
if
isinstance
(
input_loc
,
LayerOutput
):
input_loc
=
[
input_loc
]
...
...
@@ -1326,7 +1338,7 @@ def roi_pool_layer(input,
A layer used by Fast R-CNN to extract feature maps of ROIs from the last
feature map.
:param name: The
Layer Name
.
:param name: The
name of this layer. It is optional
.
:type name: basestring
:param input: The input layer.
:type input: LayerOutput.
...
...
@@ -1338,9 +1350,10 @@ def roi_pool_layer(input,
:type pooled_height: int
:param spatial_scale: The spatial scale between the image and feature map.
:type spatial_scale: float
:param num_channels:
number of input channel
.
:param num_channels:
The number of the input channels
.
:type num_channels: int
:return: LayerOutput
:return: LayerOutput object.
:rtype: LayerOutput
"""
if
num_channels
is
None
:
assert
input
.
num_filters
is
not
None
...
...
@@ -1361,18 +1374,19 @@ def roi_pool_layer(input,
@
wrap_name_default
(
"cross_channel_norm"
)
def
cross_channel_norm_layer
(
input
,
name
=
None
,
param_attr
=
None
):
"""
Normalize a layer's output. This layer is necessary for ssd.
This layer applys normalize
across the channels of each sample to
a conv
layer's output and scale the output by a group of trainable
factors which
dimensions equal to the channel's number.
Normalize a layer's output. This layer is necessary for ssd.
This
layer applys normalization
across the channels of each sample to
a conv
olutional layer's output and scales the output by a group of
trainable factors whose
dimensions equal to the channel's number.
:param name: The name of this layer. It is optional.
:type name: basestring
:param input: The input of this layer.
:type input: LayerOutput
:param param_attr: The
Parameter Attribute|list
.
:param param_attr: The
parameter attribute. See ParameterAttribute for details
.
:type param_attr: ParameterAttribute
:return: LayerOutput
:return: LayerOutput object.
:rtype: LayerOutput
"""
assert
input
.
num_filters
is
not
None
Layer
(
...
...
@@ -1413,12 +1427,9 @@ def pooling_layer(input,
Pooling layer for sequence inputs, not used for Image.
If stride > 0, this layer slides a window whose size is determined by stride,
and return the pooling value of the window as the output. Thus, a long sequence
will be shorten.
The parameter stride specifies the intervals at which to apply the pooling
operation. Note that for sequence with sub-sequence, the default value
of stride is -1.
and returns the pooling value of the sequence in the window as the output. Thus,
a long sequence will be shortened. Note that for sequence with sub-sequence, the
default value of stride is -1.
The example usage is:
...
...
@@ -1435,16 +1446,16 @@ def pooling_layer(input,
:type name: basestring
:param input: The input of this layer.
:type input: LayerOutput
:param pooling_type: Type of pooling, MaxPooling(default), AvgPooling,
SumPooling, SquareRootNPooling.
:param pooling_type: Type of pooling. MaxPooling is the default pooling.
:type pooling_type: BasePoolingType | None
:param stride: The step size between successive pooling regions.
:type stride:
I
nt
:type stride:
i
nt
:param bias_attr: The bias attribute. If the parameter is set to False or an object
whose type is not ParameterAttribute, no bias is defined. If the
parameter is set to True, the bias is initialized to zero.
:type bias_attr: ParameterAttribute | None | bool | Any
:param layer_attr: The Extra Attributes for layer, such as dropout.
:param layer_attr: The extra layer attribute. See ExtraLayerAttribute for
details.
:type layer_attr: ExtraLayerAttribute | None
:return: LayerOutput object.
:rtype: LayerOutput
...
...
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录