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
吃玉米的猫
Paddle
提交
1204d9f3
P
Paddle
项目概览
吃玉米的猫
/
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
1204d9f3
编写于
4月 12, 2018
作者:
D
Dang Qingqing
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Refine batch_norm_op.
上级
f605f647
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
15 addition
and
12 deletion
+15
-12
paddle/fluid/operators/batch_norm_op.cu.cc
paddle/fluid/operators/batch_norm_op.cu.cc
+15
-12
未找到文件。
paddle/fluid/operators/batch_norm_op.cu.cc
浏览文件 @
1204d9f3
...
...
@@ -114,23 +114,11 @@ class BatchNormKernel<platform::CUDADeviceContext, T>
const
auto
*
bias
=
ctx
.
Input
<
Tensor
>
(
"Bias"
);
auto
*
y
=
ctx
.
Output
<
Tensor
>
(
"Y"
);
auto
*
mean_out
=
ctx
.
Output
<
Tensor
>
(
"MeanOut"
);
auto
*
variance_out
=
ctx
.
Output
<
Tensor
>
(
"VarianceOut"
);
auto
*
saved_mean
=
ctx
.
Output
<
Tensor
>
(
"SavedMean"
);
auto
*
saved_variance
=
ctx
.
Output
<
Tensor
>
(
"SavedVariance"
);
// alloc memory
y
->
mutable_data
<
T
>
(
ctx
.
GetPlace
());
mean_out
->
mutable_data
<
BatchNormParamType
<
T
>>
(
ctx
.
GetPlace
());
variance_out
->
mutable_data
<
BatchNormParamType
<
T
>>
(
ctx
.
GetPlace
());
saved_mean
->
mutable_data
<
BatchNormParamType
<
T
>>
(
ctx
.
GetPlace
());
saved_variance
->
mutable_data
<
BatchNormParamType
<
T
>>
(
ctx
.
GetPlace
());
auto
&
dev_ctx
=
ctx
.
template
device_context
<
platform
::
CUDADeviceContext
>();
math
::
SetConstant
<
platform
::
CUDADeviceContext
,
BatchNormParamType
<
T
>>
functor
;
functor
(
dev_ctx
,
saved_mean
,
static_cast
<
BatchNormParamType
<
T
>>
(
0
));
functor
(
dev_ctx
,
saved_variance
,
static_cast
<
BatchNormParamType
<
T
>>
(
0
));
auto
handle
=
dev_ctx
.
cudnn_handle
();
...
...
@@ -159,6 +147,21 @@ class BatchNormKernel<platform::CUDADeviceContext, T>
// Run training mode.
// obtain running mean and running inv var, and see if we need to
// initialize them.
auto
*
mean_out
=
ctx
.
Output
<
Tensor
>
(
"MeanOut"
);
auto
*
variance_out
=
ctx
.
Output
<
Tensor
>
(
"VarianceOut"
);
mean_out
->
mutable_data
<
BatchNormParamType
<
T
>>
(
ctx
.
GetPlace
());
variance_out
->
mutable_data
<
BatchNormParamType
<
T
>>
(
ctx
.
GetPlace
());
auto
*
saved_mean
=
ctx
.
Output
<
Tensor
>
(
"SavedMean"
);
auto
*
saved_variance
=
ctx
.
Output
<
Tensor
>
(
"SavedVariance"
);
saved_mean
->
mutable_data
<
BatchNormParamType
<
T
>>
(
ctx
.
GetPlace
());
saved_variance
->
mutable_data
<
BatchNormParamType
<
T
>>
(
ctx
.
GetPlace
());
math
::
SetConstant
<
platform
::
CUDADeviceContext
,
BatchNormParamType
<
T
>>
functor
;
functor
(
dev_ctx
,
saved_mean
,
static_cast
<
BatchNormParamType
<
T
>>
(
0
));
functor
(
dev_ctx
,
saved_variance
,
static_cast
<
BatchNormParamType
<
T
>>
(
0
));
double
this_factor
=
1.
-
momentum
;
CUDNN_ENFORCE
(
platform
::
dynload
::
cudnnBatchNormalizationForwardTraining
(
...
...
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