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
机器未来
Paddle
提交
efd2d335
P
Paddle
项目概览
机器未来
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
1
Star
1
Fork
0
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1
列表
看板
标记
里程碑
合并请求
0
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
Issue
1
列表
看板
标记
里程碑
合并请求
0
合并请求
0
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
提交
efd2d335
编写于
5月 29, 2017
作者:
W
wwhu
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
add multiplex layer for python API
上级
06850e98
变更
4
显示空白变更内容
内联
并排
Showing
4 changed file
with
129 addition
and
1 deletion
+129
-1
python/paddle/trainer_config_helpers/layers.py
python/paddle/trainer_config_helpers/layers.py
+53
-0
python/paddle/trainer_config_helpers/tests/configs/file_list.sh
.../paddle/trainer_config_helpers/tests/configs/file_list.sh
+1
-1
python/paddle/trainer_config_helpers/tests/configs/protostr/test_multiplex_layer.protostr
...pers/tests/configs/protostr/test_multiplex_layer.protostr
+63
-0
python/paddle/trainer_config_helpers/tests/configs/test_multiplex_layer.py
...iner_config_helpers/tests/configs/test_multiplex_layer.py
+12
-0
未找到文件。
python/paddle/trainer_config_helpers/layers.py
浏览文件 @
efd2d335
...
...
@@ -119,6 +119,7 @@ __all__ = [
'eos_layer'
,
'smooth_l1_cost'
,
'layer_support'
,
'multiplex_layer'
,
]
...
...
@@ -185,6 +186,7 @@ class LayerType(object):
MAXOUT
=
"maxout"
SPP_LAYER
=
"spp"
PAD_LAYER
=
"pad"
MULTIPLEX_LAYER
=
"multiplex"
PRINT_LAYER
=
"print"
PRIORBOX_LAYER
=
"priorbox"
...
...
@@ -5465,3 +5467,54 @@ def smooth_l1_cost(input, label, name=None, coeff=1.0, layer_attr=None):
**
ExtraLayerAttribute
.
to_kwargs
(
layer_attr
))
return
LayerOutput
(
name
,
LayerType
.
SMOOTH_L1
,
parents
=
[
input
,
label
],
size
=
1
)
@
wrap_name_default
()
def
multiplex_layer
(
input
,
name
=
None
,
layer_attr
=
None
):
"""
This layer multiplex multiple layers according to the index,
which is provided by the first input layer.
inputs[0]: the index of the layer to output of size batchSize.
inputs[1:N]; the candidate output data.
For each index i from 0 to batchSize -1, the output is the i-th row of the
(index[i] + 1)-th layer.
For each i-th row of output:
.. math::
y[i][j] = x_{x_{0}[i] + 1}[i][j], j = 0,1, ... , (x_{1}.width - 1)
where, y is output. :math:`x_{k}` is the k-th input layer and
:math:`k = x_{0}[i] + 1`.
.. code-block:: python
maxid = multiplex_layer(input=layers)
:param input: Input layers.
:type input: list of LayerOutput
:param name: Layer name.
:type name: basestring
:param layer_attr: extra layer attributes.
:type layer_attr: ExtraLayerAttribute.
:return: LayerOutput object.
:rtype: LayerOutput
"""
assert
isinstance
(
input
,
collections
.
Sequence
)
assert
len
(
input
)
>
2
,
'multiplex_layer should have more than 2 inputs'
for
i
in
range
(
1
,
len
(
input
)):
assert
isinstance
(
input
[
i
],
LayerOutput
)
assert
input
[
i
].
size
==
input
[
1
].
size
,
\
"All the input layers except the first one should have the same size"
l
=
Layer
(
name
=
name
,
type
=
'multiplex'
,
inputs
=
[
x
.
name
for
x
in
input
],
size
=
input
[
1
].
size
,
**
ExtraLayerAttribute
.
to_kwargs
(
layer_attr
))
return
LayerOutput
(
name
=
name
,
layer_type
=
LayerType
.
MULTIPLEX_LAYER
,
parents
=
input
,
size
=
l
.
config
.
size
)
python/paddle/trainer_config_helpers/tests/configs/file_list.sh
浏览文件 @
efd2d335
...
...
@@ -5,6 +5,6 @@ last_first_seq test_expand_layer test_ntm_layers test_hsigmoid
img_layers img_trans_layers util_layers simple_rnn_layers unused_layers test_cost_layers
test_rnn_group shared_fc shared_lstm shared_gru test_cost_layers_with_weight
test_spp_layer test_bilinear_interp test_maxout test_bi_grumemory math_ops
test_seq_concat_reshape test_pad test_smooth_l1
)
test_seq_concat_reshape test_pad test_smooth_l1
test_multiplex_layer
)
export
whole_configs
=(
test_split_datasource
)
python/paddle/trainer_config_helpers/tests/configs/protostr/test_multiplex_layer.protostr
0 → 100644
浏览文件 @
efd2d335
type: "nn"
layers {
name: "index"
type: "data"
size: 1
active_type: ""
}
layers {
name: "data1"
type: "data"
size: 30
active_type: ""
}
layers {
name: "data2"
type: "data"
size: 30
active_type: ""
}
layers {
name: "data3"
type: "data"
size: 30
active_type: ""
}
layers {
name: "__multiplex_layer_0__"
type: "multiplex"
size: 30
active_type: ""
inputs {
input_layer_name: "index"
}
inputs {
input_layer_name: "data1"
}
inputs {
input_layer_name: "data2"
}
inputs {
input_layer_name: "data3"
}
}
input_layer_names: "index"
input_layer_names: "data1"
input_layer_names: "data2"
input_layer_names: "data3"
output_layer_names: "__multiplex_layer_0__"
sub_models {
name: "root"
layer_names: "index"
layer_names: "data1"
layer_names: "data2"
layer_names: "data3"
layer_names: "__multiplex_layer_0__"
input_layer_names: "index"
input_layer_names: "data1"
input_layer_names: "data2"
input_layer_names: "data3"
output_layer_names: "__multiplex_layer_0__"
is_recurrent_layer_group: false
}
python/paddle/trainer_config_helpers/tests/configs/test_multiplex_layer.py
0 → 100644
浏览文件 @
efd2d335
from
paddle.trainer_config_helpers
import
*
settings
(
batch_size
=
1000
,
learning_rate
=
1e-5
)
index
=
data_layer
(
name
=
'index'
,
size
=
1
)
din1
=
data_layer
(
name
=
'data1'
,
size
=
30
)
din2
=
data_layer
(
name
=
'data2'
,
size
=
30
)
din3
=
data_layer
(
name
=
'data3'
,
size
=
30
)
dout
=
multiplex_layer
([
index
,
din1
,
din2
,
din3
])
outputs
(
dout
)
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