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
wmsofts
Paddle
提交
f4a4a4cb
P
Paddle
项目概览
wmsofts
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
1
Star
0
Fork
0
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
0
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
0
Issue
0
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
体验新版 GitCode,发现更多精彩内容 >>
提交
f4a4a4cb
编写于
8月 22, 2018
作者:
F
fengjiayi
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
add op comment and python layer
上级
ce182d90
变更
3
隐藏空白更改
内联
并排
Showing
3 changed file
with
94 addition
and
0 deletion
+94
-0
paddle/fluid/operators/math/sequence_padding.cu
paddle/fluid/operators/math/sequence_padding.cu
+3
-0
paddle/fluid/operators/sequence_pad_op.cc
paddle/fluid/operators/sequence_pad_op.cc
+46
-0
python/paddle/fluid/layers/nn.py
python/paddle/fluid/layers/nn.py
+45
-0
未找到文件。
paddle/fluid/operators/math/sequence_padding.cu
浏览文件 @
f4a4a4cb
...
...
@@ -66,6 +66,9 @@ class PaddingLoDTensorFunctor<platform::CUDADeviceContext, T> {
if
(
pad_seq_len
==
-
1
)
{
pad_seq_len
=
max_seq_len
;
}
PADDLE_ENFORCE_GE
(
pad_seq_len
,
max_seq_len
,
"The pad_seq_len must be equal to or greater than the "
"original max sequence length."
);
int
step_width
=
seq_tensor
.
numel
()
/
seq_tensor_dims
[
0
];
int
seq_num
=
seq_offsets
.
size
()
-
1
;
...
...
paddle/fluid/operators/sequence_pad_op.cc
浏览文件 @
f4a4a4cb
...
...
@@ -101,6 +101,52 @@ class SequencePadOpMaker : public framework::OpProtoAndCheckerMaker {
"sequence."
)
.
SetDefault
(
-
1
);
AddComment
(
R"DOC(
Sequence Pad Operator
This operator pads sequences in a same batch to a consistent length.
The length is specified by attribute 'padded_length'. New elements,
whose values are specified by input 'PadValue', will be appended to
the end of each sequence, to make their final lengths consistent.
Following are cases to better explain how this works:
Case 1:
Given a 1-level LoDTensor input(X):
X.lod = [[0, 2, 5]]
X.data = [a, b, c, d, e]
and Input(PadValue):
PadValue.data = [0]
and attribite 'padded_length' = 4,
then we get 1-level LoDTensor:
Out.lod = [[0, 4, 8]]
Out.data = [a, b, 0, 0, c, d, e, 0]
Case 2:
Given a 1-level LoDTensor input(X):
X.lod = [[0, 2, 5]]
X.data = [[a1, a2], [b1, b2], [c1, c2], [d1, d2], [e1, e2]]
and Input(PadValue):
PadValue.data = [0]
and attribite 'padded_length' = -1, which mean using the length
of longest input sequence(3 in this case),
then we get 1-level LoDTensor:
Out.lod = [[0, 3, 6]]
Out.data = [[a1, a2], [b1, b2], [0, 0], [c1, c2], [d1, d2], [e1, e2]]
Case 3:
Given a 1-level LoDTensor input(X):
X.lod = [[0, 2, 5]]
X.data = [[a1, a2], [b1, b2], [c1, c2], [d1, d2], [e1, e2]]
and Input(PadValue):
PadValue.data = [p1, p2]
and attribite 'padded_length' = -1, which mean using the length
of longest input sequence(3 in this case),
then we get 1-level LoDTensor:
Out.lod = [[0, 3, 6]]
Out.data = [[a1, a2], [b1, b2], [p1, p2], [c1, c2], [d1, d2], [e1, e2]]
)DOC"
);
}
...
...
python/paddle/fluid/layers/nn.py
浏览文件 @
f4a4a4cb
...
...
@@ -2662,6 +2662,51 @@ def sequence_expand(x, y, ref_level=-1, name=None):
return
tmp
@
templatedoc
()
def
sequence_pad
(
x
,
pad_value
,
maxlen
=
None
):
"""
${comment}
Args:
x(Variable): Input variable which should contain lod information.
pad_value(Variable): The Variable that holds values that will be fill
into padded steps. It can be a scalar or a tensor whose shape
equals to time steps in sequences. If it's a scalar, it will be
automatically broadcasted to the shape of time step.
maxlen(int, default None): The length of padded sequences. It can be
None or any positive int. When it is None, all sequences will be
padded up to the length of the longest one among them; when it a
certain positive value, it must be greater than the length of the
longest original sequence."
Returns:
Variable: The padded sequence batch. All sequences has the same length.
Examples:
.. code-block:: python
import numpy
x = fluid.layers.data(name='y', shape=[10, 5],
dtype='float32', lod_level=1)
pad_value = fluid.layers.assign(input=numpy.array([0]))
out = fluid.layers.sequence_pad(x=x, pad_value=pad_value)
"""
helper
=
LayerHelper
(
'sequence_pad'
,
input
=
x
,
**
locals
())
dtype
=
helper
.
input_dtype
()
out
=
helper
.
create_tmp_variable
(
dtype
)
if
maxlen
is
None
:
maxlen
=
-
1
helper
.
append_op
(
type
=
'sequence_pad'
,
inputs
=
{
'X'
:
x
,
'PadValue'
:
pad_value
},
outputs
=
{
'Out'
:
out
},
attrs
=
{
'padded_length'
:
maxlen
})
return
out
def
beam_search
(
pre_ids
,
pre_scores
,
ids
,
...
...
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