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
s920243400
PaddleDetection
提交
46e14bbc
P
PaddleDetection
项目概览
s920243400
/
PaddleDetection
与 Fork 源项目一致
Fork自
PaddlePaddle / PaddleDetection
通知
2
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
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
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
提交
46e14bbc
编写于
4月 03, 2018
作者:
M
mozga-intel
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Enforce: 2 and 4 dims, remove information about out in format
上级
32f8ac7d
变更
2
隐藏空白更改
内联
并排
Showing
2 changed file
with
9 addition
and
18 deletion
+9
-18
paddle/fluid/operators/fc_mkldnn_op.cc
paddle/fluid/operators/fc_mkldnn_op.cc
+3
-3
paddle/fluid/operators/fc_op.cc
paddle/fluid/operators/fc_op.cc
+6
-15
未找到文件。
paddle/fluid/operators/fc_mkldnn_op.cc
浏览文件 @
46e14bbc
...
...
@@ -125,10 +125,10 @@ class FCMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
auto
input
=
ctx
.
Input
<
Tensor
>
(
"Input"
);
auto
w
=
ctx
.
Input
<
Tensor
>
(
"W"
);
PADDLE_ENFORCE
(
input
->
dims
().
size
()
==
4
||
input
->
dims
().
size
()
==
2
,
PADDLE_ENFORCE
(
input
->
dims
().
size
()
==
2
||
input
->
dims
().
size
()
==
4
,
"Input must be with 2 or 4 dimensions, i.e. NCHW"
);
PADDLE_ENFORCE
(
w
->
dims
().
size
()
==
2
,
"Weights must be with 2
dimensions, i.e. NC
"
);
PADDLE_ENFORCE
(
w
->
dims
().
size
()
==
2
||
w
->
dims
().
size
()
==
4
,
"Weights must be with 2
or 4 dimensions, i.e. OI or OIHW
"
);
bool
with_bias
=
ctx
.
Attr
<
bool
>
(
"bias_attr"
);
MKLDNNMD
<
Tensor
>
md
(
input
,
w
,
with_bias
);
...
...
paddle/fluid/operators/fc_op.cc
浏览文件 @
46e14bbc
...
...
@@ -13,6 +13,7 @@ See the License for the specific language governing permissions and
limitations under the License. */
#include "paddle/fluid/operators/fc_op.h"
#include <vector>
namespace
paddle
{
namespace
operators
{
...
...
@@ -29,11 +30,11 @@ void FCOp::InferShape(framework::InferShapeContext* ctx) const {
auto
w_dims
=
ctx
->
GetInputDim
(
"W"
);
std
::
vector
<
int64_t
>
output_shape
({
in_dims
[
0
],
w_dims
[
1
]});
PADDLE_ENFORCE
(
in_dims
.
size
()
==
4
||
in_dims
.
size
()
==
2
,
PADDLE_ENFORCE
(
in_dims
.
size
()
==
2
||
in_dims
.
size
()
==
4
,
"Fully Connected input should be 2-D or 4-D tensor."
);
PADDLE_ENFORCE
(
w_dims
.
size
()
==
2
,
"Fully Connected input should be 2-D tensor."
);
PADDLE_ENFORCE
(
w_dims
.
size
()
==
2
||
w_dims
.
size
()
==
4
,
"Fully Connected input should be 2-D
or 4-D
tensor."
);
ctx
->
SetOutputDim
(
"Out"
,
framework
::
make_ddim
(
output_shape
));
ctx
->
ShareLoD
(
"Input"
,
"Out"
);
...
...
@@ -73,19 +74,9 @@ framework::OpKernelType FCOpGrad::GetExpectedKernelType(
FCOpMaker
::
FCOpMaker
(
OpProto
*
proto
,
OpAttrChecker
*
op_checker
)
:
OpProtoAndCheckerMaker
(
proto
,
op_checker
)
{
AddInput
(
"Input"
,
"(Tensor) The input tensor of fully connected operator. "
"The format of input tensor is NCHW, where N is batch size, C is the "
"number of channels, H is the height of the feature, "
"and W is the width of the feature."
);
AddInput
(
"Input"
,
"(Tensor) The input tensor of fully connected operator. "
);
AddInput
(
"W"
,
"(Tensor), The second input tensor of fc op."
);
AddOutput
(
"Out"
,
"(Tensor) The output tensor of fully connected operator. "
"The format of output tensor is also NCHW, "
"where N is batch size, C is the number of channels, "
"H is the height of the feature, "
"and W is the width of the feature."
);
AddOutput
(
"Out"
,
"(Tensor) The output tensor of fully connected operator. "
);
AddAttr
<
bool
>
(
"use_mkldnn"
,
"(bool, default false) Only used in mkldnn kernel"
)
.
SetDefault
(
false
);
...
...
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