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
Crayon鑫
Paddle
提交
8ebfc153
P
Paddle
项目概览
Crayon鑫
/
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
提交
8ebfc153
编写于
3月 16, 2018
作者:
K
Kexin Zhao
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
update
上级
bfbc25bd
变更
2
隐藏空白更改
内联
并排
Showing
2 changed file
with
8 addition
and
10 deletion
+8
-10
paddle/fluid/operators/conv_cudnn_op.cu.cc
paddle/fluid/operators/conv_cudnn_op.cu.cc
+4
-4
paddle/fluid/operators/pool_cudnn_op.cu.cc
paddle/fluid/operators/pool_cudnn_op.cu.cc
+4
-6
未找到文件。
paddle/fluid/operators/conv_cudnn_op.cu.cc
浏览文件 @
8ebfc153
...
...
@@ -28,6 +28,8 @@ using ScopedTensorDescriptor = platform::ScopedTensorDescriptor;
using
ScopedFilterDescriptor
=
platform
::
ScopedFilterDescriptor
;
using
ScopedConvolutionDescriptor
=
platform
::
ScopedConvolutionDescriptor
;
using
DataLayout
=
platform
::
DataLayout
;
template
<
typename
T
>
using
ScalingParamType
=
typename
platform
::
CudnnDataType
<
T
>::
ScalingParamType
;
static
constexpr
size_t
kCONV_CUDNN_WORKSPACE_LIMIT_BYTES
=
static_cast
<
size_t
>
(
1024
)
*
1024
*
1024
;
...
...
@@ -134,8 +136,7 @@ class CUDNNConvOpKernel : public framework::OpKernel<T> {
platform
::
CUDAPlace
gpu
=
boost
::
get
<
platform
::
CUDAPlace
>
(
ctx
.
GetPlace
());
cudnn_workspace
=
paddle
::
memory
::
Alloc
(
gpu
,
workspace_size_in_bytes
);
// ------------------- cudnn conv forward ---------------------
typename
platform
::
CudnnDataType
<
T
>::
ScalingParamType
alpha
=
1.0
f
,
beta
=
0.0
f
;
ScalingParamType
<
T
>
alpha
=
1.0
f
,
beta
=
0.0
f
;
for
(
int
i
=
0
;
i
<
groups
;
i
++
)
{
PADDLE_ENFORCE
(
platform
::
dynload
::
cudnnConvolutionForward
(
handle
,
&
alpha
,
cudnn_input_desc
,
input_data
+
i
*
group_offset_in
,
...
...
@@ -282,8 +283,7 @@ class CUDNNConvGradOpKernel : public framework::OpKernel<T> {
platform
::
CUDAPlace
gpu
=
boost
::
get
<
platform
::
CUDAPlace
>
(
ctx
.
GetPlace
());
cudnn_workspace
=
paddle
::
memory
::
Alloc
(
gpu
,
workspace_size_in_bytes
);
// ------------------- cudnn conv backward data ---------------------
typename
platform
::
CudnnDataType
<
T
>::
ScalingParamType
alpha
=
1.0
f
,
beta
=
0.0
f
;
ScalingParamType
<
T
>
alpha
=
1.0
f
,
beta
=
0.0
f
;
if
(
input_grad
)
{
T
*
input_grad_data
=
input_grad
->
mutable_data
<
T
>
(
ctx
.
GetPlace
());
// Because beta is zero, it is unnecessary to reset input_grad.
...
...
paddle/fluid/operators/pool_cudnn_op.cu.cc
浏览文件 @
8ebfc153
...
...
@@ -24,6 +24,8 @@ using ScopedTensorDescriptor = platform::ScopedTensorDescriptor;
using
ScopedPoolingDescriptor
=
platform
::
ScopedPoolingDescriptor
;
using
DataLayout
=
platform
::
DataLayout
;
using
PoolingMode
=
platform
::
PoolingMode
;
template
<
typename
T
>
using
ScalingParamType
=
typename
platform
::
CudnnDataType
<
T
>::
ScalingParamType
;
template
<
typename
T
>
class
PoolCUDNNOpKernel
:
public
framework
::
OpKernel
<
T
>
{
...
...
@@ -78,9 +80,7 @@ class PoolCUDNNOpKernel : public framework::OpKernel<T> {
// ------------------- cudnn pool algorithm ---------------------
auto
handle
=
ctx
.
cuda_device_context
().
cudnn_handle
();
typename
platform
::
CudnnDataType
<
T
>::
ScalingParamType
alpha
=
1.0
f
,
beta
=
0.0
f
;
ScalingParamType
<
T
>
alpha
=
1.0
f
,
beta
=
0.0
f
;
PADDLE_ENFORCE
(
platform
::
dynload
::
cudnnPoolingForward
(
handle
,
cudnn_pool_desc
,
&
alpha
,
cudnn_input_desc
,
input_data
,
&
beta
,
cudnn_output_desc
,
output_data
));
...
...
@@ -145,9 +145,7 @@ class PoolCUDNNGradOpKernel : public framework::OpKernel<T> {
// ------------------- cudnn pool algorithm ---------------------
auto
handle
=
ctx
.
cuda_device_context
().
cudnn_handle
();
typename
platform
::
CudnnDataType
<
T
>::
ScalingParamType
alpha
=
1.0
f
,
beta
=
0.0
f
;
ScalingParamType
<
T
>
alpha
=
1.0
f
,
beta
=
0.0
f
;
if
(
input_grad
)
{
T
*
input_grad_data
=
input_grad
->
mutable_data
<
T
>
(
ctx
.
GetPlace
());
// Because beta is zero, it is unnecessary to reset input_grad.
...
...
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