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
afeixing77
Paddle
提交
2ce5c9d4
P
Paddle
项目概览
afeixing77
/
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
2ce5c9d4
编写于
1月 18, 2018
作者:
Y
Yancey1989
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
fix gpu implement
上级
07e87ff1
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
12 addition
and
10 deletion
+12
-10
paddle/operators/math/sequence_padding.cu
paddle/operators/math/sequence_padding.cu
+12
-10
未找到文件。
paddle/operators/math/sequence_padding.cu
浏览文件 @
2ce5c9d4
...
...
@@ -71,7 +71,8 @@ class PaddingLoDTensorFunctor<platform::CUDADeviceContext, T> {
framework
::
LoD
abs_offset_lod
=
framework
::
ToAbsOffset
(
lod
);
auto
seq_dims
=
seq
.
dims
();
PADDLE_ENFORCE_EQ
(
seq_dims
[
0
],
abs_offset_lod
[
level
].
back
(),
PADDLE_ENFORCE_EQ
(
seq_dims
[
0
],
static_cast
<
int64_t
>
(
abs_offset_lod
[
level
].
back
()),
"The first dimension of LoDTensor seq should be "
"equal to the sum of all sequences's length."
);
...
...
@@ -80,17 +81,17 @@ class PaddingLoDTensorFunctor<platform::CUDADeviceContext, T> {
"The input padding should be a 3-D Tensor of shape "
"[max_sequence_length, num_sequences, sequence_width]."
);
size
_t
max_sequence_length
=
MaximumSequenceLength
(
lod
,
level
);
int64
_t
max_sequence_length
=
MaximumSequenceLength
(
lod
,
level
);
PADDLE_ENFORCE_EQ
(
padding_dims
[
0
],
max_sequence_length
,
"The first dimension of Tensor padding should be the "
"maximum length of all sequences in LoDTensor seq."
);
const
size
_t
num_sequences
=
abs_offset_lod
[
level
].
size
()
-
1
;
const
int64
_t
num_sequences
=
abs_offset_lod
[
level
].
size
()
-
1
;
PADDLE_ENFORCE_EQ
(
padding_dims
[
1
],
num_sequences
,
"The second dimension of Tensor padding should be the "
"number of sequences in LoDTensor seq."
);
const
size
_t
sequence_width
=
seq
.
numel
()
/
seq_dims
[
0
];
const
int64
_t
sequence_width
=
seq
.
numel
()
/
seq_dims
[
0
];
PADDLE_ENFORCE_EQ
(
padding_dims
[
2
],
sequence_width
,
"The third dimension of Tensor padding should be the "
"width of sequence in LoDTensor seq."
);
...
...
@@ -101,7 +102,7 @@ class PaddingLoDTensorFunctor<platform::CUDADeviceContext, T> {
return
;
}
const
size
_t
kBlockSize
=
512
;
const
int64
_t
kBlockSize
=
512
;
/* At least use 32 threads to copy sequence_width elements,
* and at least 8 elements for each thread.
...
...
@@ -143,7 +144,8 @@ class UnpaddingLoDTensorFunctor<platform::CUDADeviceContext, T> {
framework
::
LoD
abs_offset_lod
=
framework
::
ToAbsOffset
(
lod
);
auto
seq_dims
=
seq
.
dims
();
PADDLE_ENFORCE_EQ
(
seq_dims
[
0
],
abs_offset_lod
[
level
].
back
(),
PADDLE_ENFORCE_EQ
(
seq_dims
[
0
],
static_cast
<
int64_t
>
(
abs_offset_lod
[
level
].
back
()),
"The first dimension of LoDTensor seq should be "
"equal to the sum of all sequences's length."
);
...
...
@@ -152,17 +154,17 @@ class UnpaddingLoDTensorFunctor<platform::CUDADeviceContext, T> {
"The input padding should be a 3-D Tensor of shape "
"[max_sequnece_length, num_sequences, sequence_width]."
);
size
_t
max_sequence_length
=
MaximumSequenceLength
(
lod
,
level
);
int64
_t
max_sequence_length
=
MaximumSequenceLength
(
lod
,
level
);
PADDLE_ENFORCE_EQ
(
padding_dims
[
0
],
max_sequence_length
,
"The first dimension of Tensor padding should be "
"the maximum length of all sequences in LoDTensor seq."
);
const
size
_t
num_sequences
=
abs_offset_lod
[
level
].
size
()
-
1
;
const
int64
_t
num_sequences
=
abs_offset_lod
[
level
].
size
()
-
1
;
PADDLE_ENFORCE_EQ
(
padding_dims
[
1
],
num_sequences
,
"The second dimension of Tensor padding should be "
"the number of sequences in LoDTensor seq."
);
const
size
_t
sequence_width
=
seq
.
numel
()
/
seq_dims
[
0
];
const
int64
_t
sequence_width
=
seq
.
numel
()
/
seq_dims
[
0
];
PADDLE_ENFORCE_EQ
(
padding_dims
[
2
],
sequence_width
,
"The third dimension of Tensor padding should be the "
"width of sequence in LoDTensor seq."
);
...
...
@@ -173,7 +175,7 @@ class UnpaddingLoDTensorFunctor<platform::CUDADeviceContext, T> {
return
;
}
const
size
_t
kBlockSize
=
512
;
const
int64
_t
kBlockSize
=
512
;
/* At least use 32 threads to copy sequence_width elements,
* and at least 8 elements for each thread.
...
...
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