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
wk1234123
Paddle
提交
aca9180a
P
Paddle
项目概览
wk1234123
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
4
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
aca9180a
编写于
3月 25, 2018
作者:
C
chengduoZH
浏览文件
操作
浏览文件
下载
差异文件
Merge branch 'develop' of
https://github.com/PaddlePaddle/Paddle
into feature/fix_concat
上级
750aff10
cffe1a91
变更
7
隐藏空白更改
内联
并排
Showing
7 changed file
with
23 addition
and
26 deletion
+23
-26
paddle/fluid/operators/detail/grpc_server.h
paddle/fluid/operators/detail/grpc_server.h
+2
-1
paddle/fluid/operators/detail/test_serde.cc
paddle/fluid/operators/detail/test_serde.cc
+5
-5
paddle/fluid/operators/listen_and_serv_op.cc
paddle/fluid/operators/listen_and_serv_op.cc
+7
-15
paddle/fluid/operators/math/concat.cc
paddle/fluid/operators/math/concat.cc
+2
-2
paddle/fluid/operators/math/concat.cu
paddle/fluid/operators/math/concat.cu
+2
-2
paddle/fluid/operators/send_op.cc
paddle/fluid/operators/send_op.cc
+4
-0
paddle/fluid/platform/CMakeLists.txt
paddle/fluid/platform/CMakeLists.txt
+1
-1
未找到文件。
paddle/fluid/operators/detail/grpc_server.h
浏览文件 @
aca9180a
...
...
@@ -22,9 +22,10 @@ limitations under the License. */
#include "paddle/fluid/framework/selected_rows.h"
#include "paddle/fluid/framework/var_type.h"
#include "paddle/fluid/operators/detail/grpc_service.h"
#include "paddle/fluid/operators/detail/grpc_service.h"
#include "paddle/fluid/operators/detail/send_recv.grpc.pb.h"
#include "paddle/fluid/operators/detail/send_recv.pb.h"
#include "paddle/fluid/operators/detail/sendrecvop_utils.h"
#include "paddle/fluid/operators/detail/simple_block_queue.h"
namespace
paddle
{
namespace
operators
{
...
...
paddle/fluid/operators/detail/test_serde.cc
浏览文件 @
aca9180a
...
...
@@ -199,9 +199,9 @@ TEST(LodTensor, Run) {
RunTestLodTensor
(
place
);
RunTestLodTensor
(
place
,
1
);
#ifdef PADDLE_WITH_CUDA
platform
::
CUDAPlace
place
;
RunTestLodTensor
(
place
);
RunTestLodTensor
(
place
,
1
);
platform
::
CUDAPlace
gpu
(
0
)
;
RunTestLodTensor
(
gpu
);
RunTestLodTensor
(
gpu
,
1
);
#endif
}
...
...
@@ -210,7 +210,7 @@ TEST(SelectedRows, Run) {
RunSerdeTestSelectedRows
(
place
);
#ifdef PADDLE_WITH_CUDA
platform
::
CUDAPlace
place
;
RunSerdeTestSelectedRows
(
place
);
platform
::
CUDAPlace
gpu
;
RunSerdeTestSelectedRows
(
gpu
);
#endif
}
paddle/fluid/operators/listen_and_serv_op.cc
浏览文件 @
aca9180a
...
...
@@ -93,12 +93,6 @@ class ListenAndServOp : public framework::OperatorBase {
"server program should have at least 2 blocks"
);
framework
::
Executor
executor
(
dev_place
);
std
::
vector
<
framework
::
ExecutorPrepareContext
*>
blk_ctx_list
;
blk_ctx_list
.
push_back
(
nullptr
);
// block0 is not used.
for
(
int
blkid
=
1
;
blkid
<
num_blocks
;
++
blkid
)
{
auto
*
exe_ctx
=
executor
.
Prepare
(
*
program
,
blkid
);
blk_ctx_list
.
push_back
(
exe_ctx
);
}
// TODO(typhoonzero): change this to a while_op for every cluster-batch.
bool
exit_flag
=
false
;
...
...
@@ -149,12 +143,11 @@ class ListenAndServOp : public framework::OperatorBase {
std
::
vector
<
std
::
future
<
void
>>
fs
;
// block0 contains only listen_and_serv op, start run from block1.
for
(
int
blkid
=
1
;
blkid
<
num_blocks
-
1
;
++
blkid
)
{
fs
.
push_back
(
framework
::
Async
(
[
&
executor
,
&
program
,
&
recv_scope
,
&
blk_ctx_list
,
blkid
]()
{
fs
.
push_back
(
framework
::
Async
([
&
executor
,
&
program
,
&
recv_scope
,
blkid
]()
{
int
run_block
=
blkid
;
// thread local
try
{
executor
.
RunPreparedContext
(
blk_ctx_list
[
run_block
],
&
recv_scope
,
false
,
false
);
executor
.
Run
(
*
program
,
&
recv_scope
,
run_block
,
false
,
false
);
}
catch
(
std
::
exception
&
e
)
{
LOG
(
ERROR
)
<<
"run sub program error "
<<
e
.
what
();
}
...
...
@@ -164,8 +157,7 @@ class ListenAndServOp : public framework::OperatorBase {
// Run global block at final step, or block1 if there are only 2 blocks
if
(
num_blocks
>=
2
)
{
try
{
executor
.
RunPreparedContext
(
blk_ctx_list
[
num_blocks
-
1
],
&
recv_scope
,
false
,
false
);
executor
.
Run
(
*
program
,
&
recv_scope
,
num_blocks
-
1
,
false
,
false
);
}
catch
(
std
::
exception
&
e
)
{
LOG
(
ERROR
)
<<
"run sub program error "
<<
e
.
what
();
}
...
...
@@ -185,9 +177,9 @@ class ListenAndServOp : public framework::OperatorBase {
sparse_vars
.
clear
();
}
// while(true)
for
(
int
i
=
0
;
i
<
num_blocks
;
++
i
)
{
delete
blk_ctx_list
[
i
];
}
//
for (int i = 0; i < num_blocks; ++i) {
//
delete blk_ctx_list[i];
//
}
}
protected:
...
...
paddle/fluid/operators/math/concat.cc
浏览文件 @
aca9180a
...
...
@@ -20,7 +20,7 @@ namespace math {
/*
* All tensors' dimension should be the same and the values of
* each dimension
ar
e the same, except the axis dimension.
* each dimension
must b
e the same, except the axis dimension.
*/
template
<
typename
T
>
class
ConcatFunctor
<
platform
::
CPUDeviceContext
,
T
>
{
...
...
@@ -63,7 +63,7 @@ class ConcatFunctor<platform::CPUDeviceContext, T> {
/*
* All tensors' dimension should be the same and the values of
* each dimension
ar
e the same, except the axis dimension.
* each dimension
must b
e the same, except the axis dimension.
*/
template
<
typename
T
>
class
ConcatGradFunctor
<
platform
::
CPUDeviceContext
,
T
>
{
...
...
paddle/fluid/operators/math/concat.cu
浏览文件 @
aca9180a
...
...
@@ -125,7 +125,7 @@ __global__ void KernelConcatGrad(const T* input_data, const int in_row,
/*
* All tensors' dimension should be the same and the values of
* each dimension
ar
e the same, except the axis dimension.
* each dimension
must b
e the same, except the axis dimension.
*/
template
<
typename
T
>
class
ConcatFunctor
<
platform
::
CUDADeviceContext
,
T
>
{
...
...
@@ -195,7 +195,7 @@ class ConcatFunctor<platform::CUDADeviceContext, T> {
/*
* All tensors' dimension should be the same and the values of
* each dimension
ar
e the same, except the axis dimension.
* each dimension
must b
e the same, except the axis dimension.
*/
template
<
typename
T
>
class
ConcatGradFunctor
<
platform
::
CUDADeviceContext
,
T
>
{
...
...
paddle/fluid/operators/send_op.cc
浏览文件 @
aca9180a
...
...
@@ -21,6 +21,7 @@ limitations under the License. */
#include <future>
#include "paddle/fluid/operators/detail/grpc_client.h"
#include "paddle/fluid/platform/profiler.h"
namespace
paddle
{
namespace
operators
{
...
...
@@ -59,6 +60,9 @@ class SendOp : public framework::OperatorBase {
platform
::
DeviceContextPool
&
pool
=
platform
::
DeviceContextPool
::
Instance
();
auto
&
ctx
=
*
pool
.
Get
(
place
);
// For profiling
platform
::
RecordEvent
record_event
(
Type
(),
&
ctx
);
auto
client_var_name
=
Output
(
"RPCClient"
);
PADDLE_ENFORCE_NOT_NULL
(
scope
.
FindVar
(
client_var_name
),
"Can not find variable '%s' in the scope."
,
...
...
paddle/fluid/platform/CMakeLists.txt
浏览文件 @
aca9180a
...
...
@@ -49,7 +49,7 @@ nv_test(device_context_test SRCS device_context_test.cu DEPS device_context gpu_
nv_test
(
cudnn_helper_test SRCS cudnn_helper_test.cc DEPS dynload_cuda
)
nv_test
(
transform_test SRCS transform_test.cu DEPS paddle_memory place device_context
)
cc_library
(
device_tracer SRCS device_tracer.cc DEPS profiler_proto
${
GPU_CTX_DEPS
}
)
cc_library
(
device_tracer SRCS device_tracer.cc DEPS
boost
profiler_proto
${
GPU_CTX_DEPS
}
)
cc_library
(
profiler SRCS profiler.cc DEPS device_context device_tracer
)
cc_test
(
profiler_test SRCS profiler_test.cc DEPS profiler
)
...
...
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