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
62a0fe08
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
未验证
提交
62a0fe08
编写于
10月 31, 2018
作者:
Y
Yan Chunwei
提交者:
GitHub
10月 31, 2018
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
fix tensor array bug (#14166)
remove the optimized but buggy implementation
上级
ed087f82
变更
1
显示空白变更内容
内联
并排
Showing
1 changed file
with
0 addition
and
74 deletion
+0
-74
paddle/fluid/framework/lod_tensor_array.h
paddle/fluid/framework/lod_tensor_array.h
+0
-74
未找到文件。
paddle/fluid/framework/lod_tensor_array.h
浏览文件 @
62a0fe08
...
...
@@ -19,81 +19,7 @@ limitations under the License. */
namespace
paddle
{
namespace
framework
{
// NOTE The vector<LoDTensor> can't be replaced with the class LoDTensorArray
// directly, because there are many vector<LoDTensor> used accross the project,
// and some of them are treated as LoDTensorArray.
#if !defined(PADDLE_ON_INFERENCE)
using
LoDTensorArray
=
std
::
vector
<
LoDTensor
>
;
#else // !PADDLE_ON_INFERENCE
#pragma message "LoDTensorArray is replaced with the inference one."
/*
* A LoDTensorArray which will not deallocate buffer when resized, fix the data
* diff in inference, and more performance friendly in the concurrency
* scenerios.
*/
class
LoDTensorArray
{
public:
LoDTensorArray
()
=
default
;
using
iterator
=
std
::
vector
<
LoDTensor
>::
iterator
;
using
const_iterator
=
std
::
vector
<
LoDTensor
>::
const_iterator
;
const_iterator
begin
()
const
{
return
array_
.
begin
();
}
const_iterator
end
()
const
{
return
array_
.
begin
()
+
size_
;
}
iterator
begin
()
{
return
array_
.
begin
();
}
iterator
end
()
{
return
array_
.
begin
()
+
size_
;
}
void
push_back
(
const
LoDTensor
&
x
)
{
if
(
size_
<
array_
.
size
())
{
array_
[
size_
++
]
=
x
;
}
else
{
array_
.
push_back
(
x
);
++
size_
;
}
}
void
resize
(
size_t
size
)
{
if
(
array_
.
size
()
<
size
)
{
array_
.
resize
(
size
);
}
size_
=
size
;
}
void
emplace_back
()
{
array_
.
emplace_back
();
}
void
emplace_back
(
LoDTensor
&&
x
)
{
array_
.
emplace_back
(
std
::
move
(
x
));
}
LoDTensor
&
back
()
{
return
array_
.
back
();
}
size_t
space
()
const
{
return
array_
.
size
();
}
void
reserve
(
size_t
size
)
{
// Naive warning to tell user this array might be to large. The memory and
// buffer used by this TensorArray will not be deleted during the training
// and inference phase, so attention not to make it expand too long.
if
(
size
>
800UL
)
{
LOG
(
WARNING
)
<<
"TensorArray has more than 800 items"
;
}
array_
.
reserve
(
size
);
}
bool
empty
()
const
{
return
size_
==
0UL
;
}
void
clear
()
{
size_
=
0UL
;
}
LoDTensor
&
operator
[](
size_t
id
)
{
return
array_
[
id
];
}
const
LoDTensor
&
operator
[](
size_t
id
)
const
{
return
array_
[
id
];
}
LoDTensor
&
at
(
size_t
id
)
{
return
array_
.
at
(
id
);
}
const
LoDTensor
&
at
(
size_t
id
)
const
{
return
array_
.
at
(
id
);
}
size_t
size
()
const
{
return
size_
;
}
private:
size_t
size_
{
0
};
std
::
vector
<
LoDTensor
>
array_
;
};
#endif // !PADDLE_ON_INFERENCE
}
// namespace framework
}
// namespace paddle
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