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
PaddlePaddle
PaddleDetection
提交
371c53f8
P
PaddleDetection
项目概览
PaddlePaddle
/
PaddleDetection
12 个月 前同步成功
通知
692
Star
11112
Fork
2696
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
184
列表
看板
标记
里程碑
合并请求
40
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
PaddleDetection
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
184
Issue
184
列表
看板
标记
里程碑
合并请求
40
合并请求
40
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
371c53f8
编写于
3月 16, 2018
作者:
L
Liu Yiqun
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Add profiling event in feed, fetch and load op.
上级
253ba667
变更
4
隐藏空白更改
内联
并排
Showing
4 changed file
with
24 addition
and
18 deletion
+24
-18
paddle/fluid/operators/conv_op.cc
paddle/fluid/operators/conv_op.cc
+8
-8
paddle/fluid/operators/feed_op.cc
paddle/fluid/operators/feed_op.cc
+6
-5
paddle/fluid/operators/fetch_op.cc
paddle/fluid/operators/fetch_op.cc
+4
-1
paddle/fluid/operators/load_op.cc
paddle/fluid/operators/load_op.cc
+6
-4
未找到文件。
paddle/fluid/operators/conv_op.cc
浏览文件 @
371c53f8
...
...
@@ -70,16 +70,16 @@ void ConvOp::InferShape(framework::InferShapeContext* ctx) const {
framework
::
OpKernelType
ConvOp
::
GetExpectedKernelType
(
const
framework
::
ExecutionContext
&
ctx
)
const
{
framework
::
LibraryType
library
_
{
framework
::
LibraryType
::
kPlain
};
framework
::
LibraryType
library
{
framework
::
LibraryType
::
kPlain
};
#ifdef PADDLE_WITH_CUDA
if
(
platform
::
CanCUDNNBeUsed
(
ctx
))
{
library
_
=
framework
::
LibraryType
::
kCUDNN
;
library
=
framework
::
LibraryType
::
kCUDNN
;
}
#endif
#ifdef PADDLE_WITH_MKLDNN
if
(
library
_
==
framework
::
LibraryType
::
kPlain
&&
if
(
library
==
framework
::
LibraryType
::
kPlain
&&
platform
::
CanMKLDNNBeUsed
(
ctx
))
{
library
_
=
framework
::
LibraryType
::
kMKLDNN
;
library
=
framework
::
LibraryType
::
kMKLDNN
;
}
#endif
...
...
@@ -91,15 +91,15 @@ framework::OpKernelType ConvOp::GetExpectedKernelType(
"input and filter data type should be consistent"
);
if
(
input_data_type
==
framework
::
proto
::
VarType
::
FP16
)
{
PADDLE_ENFORCE_EQ
(
library
_
,
framework
::
LibraryType
::
kCUDNN
,
PADDLE_ENFORCE_EQ
(
library
,
framework
::
LibraryType
::
kCUDNN
,
"float16 can only be used when CUDNN is used"
);
}
std
::
string
data_format
=
ctx
.
Attr
<
std
::
string
>
(
"data_format"
);
// TODO(pzelazko-intel): enable MKLDNN layout when it's ready
framework
::
DataLayout
layout
_
=
framework
::
StringToDataLayout
(
data_format
);
return
framework
::
OpKernelType
(
input_data_type
,
ctx
.
GetPlace
(),
layout
_
,
library
_
);
framework
::
DataLayout
layout
=
framework
::
StringToDataLayout
(
data_format
);
return
framework
::
OpKernelType
(
input_data_type
,
ctx
.
GetPlace
(),
layout
,
library
);
}
Conv2DOpMaker
::
Conv2DOpMaker
(
OpProto
*
proto
,
OpAttrChecker
*
op_checker
)
...
...
paddle/fluid/operators/feed_op.cc
浏览文件 @
371c53f8
...
...
@@ -15,6 +15,7 @@ limitations under the License. */
#include "paddle/fluid/framework/feed_fetch_type.h"
#include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/framework/operator.h"
#include "paddle/fluid/platform/profiler.h"
namespace
paddle
{
namespace
operators
{
...
...
@@ -28,6 +29,10 @@ class FeedOp : public framework::OperatorBase {
private:
void
RunImpl
(
const
framework
::
Scope
&
scope
,
const
platform
::
Place
&
place
)
const
override
{
// get device context from pool
auto
*
dev_ctx
=
platform
::
DeviceContextPool
::
Instance
().
Get
(
place
);
platform
::
RecordEvent
record_event
(
Type
(),
dev_ctx
);
auto
feed_var_name
=
Input
(
"X"
);
auto
*
feed_var
=
scope
.
FindVar
(
feed_var_name
);
...
...
@@ -50,14 +55,10 @@ class FeedOp : public framework::OperatorBase {
auto
&
feed_item
=
feed_list
.
at
(
static_cast
<
size_t
>
(
col
));
auto
*
out_item
=
out_var
->
GetMutable
<
framework
::
FeedFetchType
>
();
// get device context from pool
platform
::
DeviceContextPool
&
pool
=
platform
::
DeviceContextPool
::
Instance
();
auto
&
dev_ctx
=
*
pool
.
Get
(
place
);
if
(
platform
::
is_same_place
(
feed_item
.
place
(),
place
))
{
out_item
->
ShareDataWith
(
feed_item
);
}
else
{
framework
::
TensorCopy
(
feed_item
,
place
,
dev_ctx
,
out_item
);
framework
::
TensorCopy
(
feed_item
,
place
,
*
dev_ctx
,
out_item
);
}
out_item
->
set_lod
(
feed_item
.
lod
());
}
...
...
paddle/fluid/operators/fetch_op.cc
浏览文件 @
371c53f8
...
...
@@ -15,6 +15,7 @@ limitations under the License. */
#include "paddle/fluid/framework/feed_fetch_type.h"
#include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/platform/device_context.h"
#include "paddle/fluid/platform/profiler.h"
namespace
paddle
{
namespace
operators
{
...
...
@@ -29,6 +30,9 @@ class FetchOp : public framework::OperatorBase {
private:
void
RunImpl
(
const
framework
::
Scope
&
scope
,
const
platform
::
Place
&
place
)
const
override
{
platform
::
DeviceContextPool
&
pool
=
platform
::
DeviceContextPool
::
Instance
();
platform
::
RecordEvent
record_event
(
Type
(),
pool
.
Get
(
place
));
auto
fetch_var_name
=
Input
(
"X"
);
auto
*
fetch_var
=
scope
.
FindVar
(
fetch_var_name
);
PADDLE_ENFORCE
(
fetch_var
!=
nullptr
,
...
...
@@ -53,7 +57,6 @@ class FetchOp : public framework::OperatorBase {
// FIXME(yuyang18): Should we assume the fetch operator always generate
// CPU outputs?
platform
::
DeviceContextPool
&
pool
=
platform
::
DeviceContextPool
::
Instance
();
auto
&
dev_ctx
=
*
pool
.
Get
(
src_item
.
place
());
TensorCopy
(
src_item
,
platform
::
CPUPlace
(),
dev_ctx
,
&
dst_item
);
...
...
paddle/fluid/operators/load_op.cc
浏览文件 @
371c53f8
...
...
@@ -15,6 +15,7 @@ limitations under the License. */
#include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/platform/device_context.h"
#include "paddle/fluid/platform/profiler.h"
namespace
paddle
{
namespace
operators
{
...
...
@@ -29,6 +30,9 @@ class LoadOp : public framework::OperatorBase {
private:
void
RunImpl
(
const
framework
::
Scope
&
scope
,
const
platform
::
Place
&
place
)
const
override
{
auto
*
dev_ctx
=
platform
::
DeviceContextPool
::
Instance
().
Get
(
place
);
platform
::
RecordEvent
record_event
(
Type
(),
dev_ctx
);
auto
filename
=
Attr
<
std
::
string
>
(
"file_path"
);
std
::
ifstream
fin
(
filename
);
PADDLE_ENFORCE
(
static_cast
<
bool
>
(
fin
),
"Cannot open file %s for load op"
,
...
...
@@ -41,9 +45,7 @@ class LoadOp : public framework::OperatorBase {
auto
*
tensor
=
out_var
->
GetMutable
<
framework
::
LoDTensor
>
();
platform
::
DeviceContextPool
&
pool
=
platform
::
DeviceContextPool
::
Instance
();
auto
&
dev_ctx
=
*
pool
.
Get
(
place
);
DeserializeFromStream
(
fin
,
tensor
,
dev_ctx
);
DeserializeFromStream
(
fin
,
tensor
,
*
dev_ctx
);
if
(
platform
::
is_gpu_place
(
place
))
{
// copy CPU to GPU
...
...
@@ -55,7 +57,7 @@ class LoadOp : public framework::OperatorBase {
out_var
->
Clear
();
tensor
=
out_var
->
GetMutable
<
framework
::
LoDTensor
>
();
tensor
->
set_lod
(
cpu_tensor
.
lod
());
TensorCopy
(
cpu_tensor
,
place
,
dev_ctx
,
tensor
);
TensorCopy
(
cpu_tensor
,
place
,
*
dev_ctx
,
tensor
);
}
}
};
...
...
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