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
夏有凉风,冬有雪
Paddle
提交
6612068e
P
Paddle
项目概览
夏有凉风,冬有雪
/
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
体验新版 GitCode,发现更多精彩内容 >>
未验证
提交
6612068e
编写于
2月 08, 2018
作者:
W
whs
提交者:
GitHub
2月 08, 2018
浏览文件
操作
浏览文件
下载
差异文件
Merge pull request #8114 from wanghaoshuang/fix_ctc_align
Make CTC align op support for empty output
上级
e80255ce
3aae7815
变更
4
隐藏空白更改
内联
并排
Showing
4 changed file
with
47 addition
and
20 deletion
+47
-20
paddle/operators/ctc_align_op.cu
paddle/operators/ctc_align_op.cu
+8
-0
paddle/operators/ctc_align_op.h
paddle/operators/ctc_align_op.h
+8
-1
python/paddle/v2/fluid/layers/nn.py
python/paddle/v2/fluid/layers/nn.py
+20
-19
python/paddle/v2/fluid/tests/test_ctc_align.py
python/paddle/v2/fluid/tests/test_ctc_align.py
+11
-0
未找到文件。
paddle/operators/ctc_align_op.cu
浏览文件 @
6612068e
...
@@ -80,6 +80,14 @@ class CTCAlignOpCUDAKernel : public framework::OpKernel<T> {
...
@@ -80,6 +80,14 @@ class CTCAlignOpCUDAKernel : public framework::OpKernel<T> {
// resize output dims
// resize output dims
output
->
Resize
({
static_cast
<
int64_t
>
(
host_out_lod0
.
back
()),
1
});
output
->
Resize
({
static_cast
<
int64_t
>
(
host_out_lod0
.
back
()),
1
});
if
(
host_out_lod0
.
back
()
==
0
)
{
output
->
Resize
({
1
,
1
});
output
->
mutable_data
<
T
>
(
ctx
.
GetPlace
());
math
::
SetConstant
<
platform
::
CUDADeviceContext
,
T
>
set_constant
;
set_constant
(
ctx
.
template
device_context
<
platform
::
CUDADeviceContext
>(),
output
,
-
1
);
}
}
}
};
};
...
...
paddle/operators/ctc_align_op.h
浏览文件 @
6612068e
...
@@ -16,6 +16,8 @@ limitations under the License. */
...
@@ -16,6 +16,8 @@ limitations under the License. */
#include <string.h>
#include <string.h>
#include "paddle/framework/op_registry.h"
#include "paddle/framework/op_registry.h"
#include "paddle/operators/math/math_function.h"
namespace
paddle
{
namespace
paddle
{
namespace
operators
{
namespace
operators
{
...
@@ -65,9 +67,14 @@ class CTCAlignKernel : public framework::OpKernel<T> {
...
@@ -65,9 +67,14 @@ class CTCAlignKernel : public framework::OpKernel<T> {
framework
::
LoD
output_lod
;
framework
::
LoD
output_lod
;
output_lod
.
push_back
(
output_lod0
);
output_lod
.
push_back
(
output_lod0
);
output
->
set_lod
(
output_lod
);
output
->
set_lod
(
output_lod
);
// resize output dims
// resize output dims
output
->
Resize
({
static_cast
<
int64_t
>
(
output_lod0
.
back
()),
1
});
output
->
Resize
({
static_cast
<
int64_t
>
(
output_lod0
.
back
()),
1
});
// for empty sequence
if
(
output_lod0
.
back
()
==
0
)
{
output
->
Resize
({
1
,
1
});
output_data
=
output
->
mutable_data
<
T
>
(
ctx
.
GetPlace
());
output_data
[
0
]
=
-
1
;
}
}
}
};
};
...
...
python/paddle/v2/fluid/layers/nn.py
浏览文件 @
6612068e
...
@@ -410,12 +410,12 @@ def dynamic_lstmp(input,
...
@@ -410,12 +410,12 @@ def dynamic_lstmp(input,
"""
"""
**Dynamic LSTMP Layer**
**Dynamic LSTMP Layer**
LSTMP (LSTM with recurrent projection) layer has a separate projection
LSTMP (LSTM with recurrent projection) layer has a separate projection
layer after the LSTM layer, projecting the original hidden state to a
layer after the LSTM layer, projecting the original hidden state to a
lower-dimensional one, which is proposed to reduce the number of total
lower-dimensional one, which is proposed to reduce the number of total
parameters and furthermore computational complexity for the LSTM,
parameters and furthermore computational complexity for the LSTM,
espeacially for the case that the size of output units is relative
espeacially for the case that the size of output units is relative
large (https://research.google.com/pubs/archive/43905.pdf).
large (https://research.google.com/pubs/archive/43905.pdf).
The formula is as follows:
The formula is as follows:
...
@@ -441,27 +441,27 @@ def dynamic_lstmp(input,
...
@@ -441,27 +441,27 @@ def dynamic_lstmp(input,
the matrix of weights from the input gate to the input).
the matrix of weights from the input gate to the input).
* :math:`W_{ic}`, :math:`W_{fc}`, :math:`W_{oc}`: Diagonal weight
\
* :math:`W_{ic}`, :math:`W_{fc}`, :math:`W_{oc}`: Diagonal weight
\
matrices for peephole connections. In our implementation,
\
matrices for peephole connections. In our implementation,
\
we use vectors to reprenset these diagonal weight matrices.
we use vectors to reprenset these diagonal weight matrices.
* :math:`b`: Denotes bias vectors (e.g. :math:`b_i` is the input gate
\
* :math:`b`: Denotes bias vectors (e.g. :math:`b_i` is the input gate
\
bias vector).
bias vector).
* :math:`\sigma`: The activation, such as logistic sigmoid function.
* :math:`\sigma`: The activation, such as logistic sigmoid function.
* :math:`i, f, o` and :math:`c`: The input gate, forget gate, output
\
* :math:`i, f, o` and :math:`c`: The input gate, forget gate, output
\
gate, and cell activation vectors, respectively, all of which have
\
gate, and cell activation vectors, respectively, all of which have
\
the same size as the cell output activation vector :math:`h`.
the same size as the cell output activation vector :math:`h`.
* :math:`h`: The hidden state.
* :math:`h`: The hidden state.
* :math:`r`: The recurrent projection of the hidden state.
* :math:`r`: The recurrent projection of the hidden state.
* :math:`
\\
tilde{c_t}`: The candidate hidden state, whose
\
* :math:`
\\
tilde{c_t}`: The candidate hidden state, whose
\
computation is based on the current input and previous hidden state.
computation is based on the current input and previous hidden state.
* :math:`\odot`: The element-wise product of the vectors.
* :math:`\odot`: The element-wise product of the vectors.
* :math:`act_g` and :math:`act_h`: The cell input and cell output
\
* :math:`act_g` and :math:`act_h`: The cell input and cell output
\
activation functions and `tanh` is usually used for them.
activation functions and `tanh` is usually used for them.
* :math:`\overline{act_h}`: The activation function for the projection
\
* :math:`\overline{act_h}`: The activation function for the projection
\
output, usually using `identity` or same as :math:`act_h`.
output, usually using `identity` or same as :math:`act_h`.
Set `use_peepholes` to `False` to disable peephole connection. The formula
Set `use_peepholes` to `False` to disable peephole connection. The formula
is omitted here, please refer to the paper
is omitted here, please refer to the paper
http://www.bioinf.jku.at/publications/older/2604.pdf for details.
http://www.bioinf.jku.at/publications/older/2604.pdf for details.
Note that these :math:`W_{xi}x_{t}, W_{xf}x_{t}, W_{xc}x_{t}, W_{xo}x_{t}`
Note that these :math:`W_{xi}x_{t}, W_{xf}x_{t}, W_{xc}x_{t}, W_{xo}x_{t}`
operations on the input :math:`x_{t}` are NOT included in this operator.
operations on the input :math:`x_{t}` are NOT included in this operator.
Users can choose to use fully-connected layer before LSTMP layer.
Users can choose to use fully-connected layer before LSTMP layer.
...
@@ -479,8 +479,8 @@ def dynamic_lstmp(input,
...
@@ -479,8 +479,8 @@ def dynamic_lstmp(input,
- Hidden-hidden weight = {:math:`W_{ch}, W_{ih},
\
- Hidden-hidden weight = {:math:`W_{ch}, W_{ih},
\
W_{fh}, W_{oh}`}.
W_{fh}, W_{oh}`}.
- The shape of hidden-hidden weight is (P x 4D),
- The shape of hidden-hidden weight is (P x 4D),
where P is the projection size and D the hidden
where P is the projection size and D the hidden
size.
size.
- Projection weight = {:math:`W_{rh}`}.
- Projection weight = {:math:`W_{rh}`}.
- The shape of projection weight is (D x P).
- The shape of projection weight is (D x P).
...
@@ -525,9 +525,9 @@ def dynamic_lstmp(input,
...
@@ -525,9 +525,9 @@ def dynamic_lstmp(input,
hidden_dim, proj_dim = 512, 256
hidden_dim, proj_dim = 512, 256
fc_out = fluid.layers.fc(input=input_seq, size=hidden_dim * 4,
fc_out = fluid.layers.fc(input=input_seq, size=hidden_dim * 4,
act=None, bias_attr=None)
act=None, bias_attr=None)
proj_out, _ = fluid.layers.dynamic_lstmp(input=fc_out,
proj_out, _ = fluid.layers.dynamic_lstmp(input=fc_out,
size=hidden_dim * 4,
size=hidden_dim * 4,
proj_size=proj_dim,
proj_size=proj_dim,
use_peepholes=False,
use_peepholes=False,
is_reverse=True,
is_reverse=True,
cell_activation="tanh",
cell_activation="tanh",
...
@@ -2525,7 +2525,8 @@ def ctc_greedy_decoder(input, blank, name=None):
...
@@ -2525,7 +2525,8 @@ def ctc_greedy_decoder(input, blank, name=None):
interval [0, num_classes + 1).
interval [0, num_classes + 1).
Returns:
Returns:
Variable: CTC greedy decode result.
Variable: CTC greedy decode result. If all the sequences in result were
empty, the result LoDTensor will be [-1] with LoD [[0]] and dims [1, 1].
Examples:
Examples:
.. code-block:: python
.. code-block:: python
...
...
python/paddle/v2/fluid/tests/test_ctc_align.py
浏览文件 @
6612068e
...
@@ -31,6 +31,8 @@ def CTCAlign(input, lod, blank, merge_repeated):
...
@@ -31,6 +31,8 @@ def CTCAlign(input, lod, blank, merge_repeated):
result
.
append
(
token
)
result
.
append
(
token
)
prev_token
=
token
prev_token
=
token
result
=
np
.
array
(
result
).
reshape
([
len
(
result
),
1
]).
astype
(
"int32"
)
result
=
np
.
array
(
result
).
reshape
([
len
(
result
),
1
]).
astype
(
"int32"
)
if
len
(
result
)
==
0
:
result
=
np
.
array
([
-
1
])
return
result
return
result
...
@@ -72,5 +74,14 @@ class TestCTCAlignOpCase1(TestCTCAlignOp):
...
@@ -72,5 +74,14 @@ class TestCTCAlignOpCase1(TestCTCAlignOp):
[
19
,
1
]).
astype
(
"int32"
)
[
19
,
1
]).
astype
(
"int32"
)
class
TestCTCAlignOpCase2
(
TestCTCAlignOp
):
def
config
(
self
):
self
.
op_type
=
"ctc_align"
self
.
input_lod
=
[[
0
,
4
]]
self
.
blank
=
0
self
.
merge_repeated
=
True
self
.
input
=
np
.
array
([
0
,
0
,
0
,
0
]).
reshape
([
4
,
1
]).
astype
(
"int32"
)
if
__name__
==
"__main__"
:
if
__name__
==
"__main__"
:
unittest
.
main
()
unittest
.
main
()
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