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
Paddle
提交
4059c9ca
P
Paddle
项目概览
PaddlePaddle
/
Paddle
大约 1 年 前同步成功
通知
2299
Star
20931
Fork
5422
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1423
列表
看板
标记
里程碑
合并请求
543
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
1,423
Issue
1,423
列表
看板
标记
里程碑
合并请求
543
合并请求
543
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
提交
4059c9ca
编写于
1月 10, 2018
作者:
Y
Yang Yu
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Polish GetPlacesOp
上级
fe341bac
变更
2
显示空白变更内容
内联
并排
Showing
2 changed file
with
28 addition
and
22 deletion
+28
-22
paddle/operators/get_places_op.cc
paddle/operators/get_places_op.cc
+18
-15
python/paddle/v2/fluid/layers/device.py
python/paddle/v2/fluid/layers/device.py
+10
-7
未找到文件。
paddle/operators/get_places_op.cc
浏览文件 @
4059c9ca
...
...
@@ -39,17 +39,19 @@ class GetPlacesOp : public framework::OperatorBase {
:
OperatorBase
(
type
,
inputs
,
outputs
,
attrs
)
{}
void
Run
(
const
framework
::
Scope
&
scope
,
const
platform
::
Place
&
place
)
const
override
{
std
::
string
device_type
=
Attr
<
std
::
string
>
(
"device_type"
);
bool
is_gpu
;
if
(
Attr
<
std
::
string
>
(
"device_type"
)
==
"AUTO"
)
{
is_gpu
=
platform
::
is_gpu_place
(
place
);
}
else
{
is_gpu
=
Attr
<
std
::
string
>
(
"device_type"
)
==
"CUDA"
;
}
auto
device_count
=
static_cast
<
size_t
>
(
Attr
<
int
>
(
"device_count"
));
if
(
device_count
==
0
)
{
if
(
device_type
==
"CUDA"
)
{
device_count
=
CUDADevCount
();
}
else
if
(
device_type
==
"CPU"
)
{
device_count
=
std
::
thread
::
hardware_concurrency
();
}
device_count
=
is_gpu
?
CUDADevCount
()
:
std
::
thread
::
hardware_concurrency
();
}
PADDLE_ENFORCE_NE
(
device_count
,
0
,
"Cannot indicate %s device count"
,
device_type
);
is_gpu
?
"GPU"
:
"CPU"
);
auto
out_var_name
=
Output
(
"Out"
);
auto
&
places
=
...
...
@@ -57,14 +59,14 @@ class GetPlacesOp : public framework::OperatorBase {
"Output variable %s cannot be found"
,
out_var_name
)
.
GetMutable
<
platform
::
PlaceList
>
());
places
.
reserve
(
device_count
);
if
(
device_type
==
"CUDA"
)
{
if
(
is_gpu
)
{
PADDLE_ENFORCE_LE
(
device_count
,
CUDADevCount
(),
"Only %d CUDA devices found, cannot set to %d"
,
CUDADevCount
(),
device_count
);
for
(
size_t
i
=
0
;
i
<
device_count
;
++
i
)
{
places
.
emplace_back
(
platform
::
CUDAPlace
(
i
));
places
.
emplace_back
(
platform
::
CUDAPlace
(
static_cast
<
int
>
(
i
)
));
}
}
else
if
(
device_type
==
"CPU"
)
{
}
else
{
for
(
size_t
i
=
0
;
i
<
device_count
;
++
i
)
{
places
.
emplace_back
(
platform
::
CPUPlace
());
}
...
...
@@ -77,10 +79,10 @@ class GetPlacesOpProtoMaker : public framework::OpProtoAndCheckerMaker {
GetPlacesOpProtoMaker
(
OpProto
*
proto
,
OpAttrChecker
*
op_checker
)
:
OpProtoAndCheckerMaker
(
proto
,
op_checker
)
{
AddOutput
(
"Out"
,
"vector of Place"
);
AddAttr
<
int
>
(
"device_count"
,
"device count"
).
SetDefault
(
1
);
AddAttr
<
std
::
string
>
(
"device_type"
,
R"(device type must be in ["CPU", "CUDA"])"
)
.
InEnum
({
"CPU"
,
"CUDA"
}
);
AddAttr
<
int
>
(
"device_count"
,
"device count"
).
SetDefault
(
0
);
AddAttr
<
std
::
string
>
(
"device_type"
,
"device type"
)
.
InEnum
({
"CUDA"
,
"CPU"
,
"AUTO"
}
)
.
SetDefault
(
"AUTO"
);
AddComment
(
R"DOC(
Returns a list of places based on flags. The list will be used for parallel
execution.
...
...
@@ -111,4 +113,5 @@ class GetPlacesInferShape : public framework::InferShapeBase {
namespace
ops
=
paddle
::
operators
;
REGISTER_OPERATOR
(
get_places
,
ops
::
GetPlacesOp
,
ops
::
GetPlacesOpProtoMaker
,
ops
::
GetPlacesInferVarType
,
ops
::
GetPlacesInferShape
);
ops
::
GetPlacesInferVarType
,
ops
::
GetPlacesInferShape
,
paddle
::
framework
::
EmptyGradOpMaker
);
python/paddle/v2/fluid/layers/device.py
浏览文件 @
4059c9ca
...
...
@@ -4,19 +4,22 @@ All util layers.
from
..layer_helper
import
LayerHelper
from
..framework
import
unique_name
from
..registry
import
autodoc
__all__
=
[
'get_places'
]
def
get_places
(
device_count
=
0
,
device_type
=
"CPU"
):
@
autodoc
def
get_places
(
device_count
=
None
,
device_type
=
None
):
helper
=
LayerHelper
(
'get_places'
,
**
locals
())
out_places
=
helper
.
create_variable
(
name
=
unique_name
(
helper
.
name
+
".out"
))
attrs
=
dict
()
if
device_count
is
not
None
:
attrs
[
'device_count'
]
=
int
(
device_count
)
if
device_type
is
not
None
:
attrs
[
'device_type'
]
=
str
(
device_type
)
helper
.
append_op
(
type
=
'get_places'
,
outputs
=
{
"Out"
:
[
out_places
]},
attrs
=
{
"device_type"
:
device_type
,
'device_count'
:
device_count
,
})
type
=
'get_places'
,
outputs
=
{
"Out"
:
[
out_places
]},
attrs
=
attrs
)
return
out_places
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