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
31815010
P
Paddle
项目概览
PaddlePaddle
/
Paddle
12 个月 前同步成功
通知
2292
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
体验新版 GitCode,发现更多精彩内容 >>
提交
31815010
编写于
3月 21, 2018
作者:
Y
Yu Yang
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Rerange code
上级
f28ae6e4
变更
5
隐藏空白更改
内联
并排
Showing
5 changed file
with
84 addition
and
28 deletion
+84
-28
paddle/fluid/framework/CMakeLists.txt
paddle/fluid/framework/CMakeLists.txt
+2
-2
paddle/fluid/framework/details/CMakeLists.txt
paddle/fluid/framework/details/CMakeLists.txt
+1
-0
paddle/fluid/framework/details/computation_op_handle.cc
paddle/fluid/framework/details/computation_op_handle.cc
+40
-0
paddle/fluid/framework/details/computation_op_handle.h
paddle/fluid/framework/details/computation_op_handle.h
+39
-0
paddle/fluid/framework/parallel_executor.cc
paddle/fluid/framework/parallel_executor.cc
+2
-26
未找到文件。
paddle/fluid/framework/CMakeLists.txt
浏览文件 @
31815010
...
...
@@ -94,8 +94,8 @@ else()
set
(
parallel_executor_cuda_deps
)
endif
()
cc_library
(
parallel_executor SRCS parallel_executor.cc DEPS op_registry device_context scope
framework_proto
backward glog lod_rank_table simple_threadpool scale_loss_grad_op_handle
fetch_op_handle
${
parallel_executor_cuda_deps
}
)
backward glog lod_rank_table simple_threadpool scale_loss_grad_op_handle
fetch_op_handle
computation_op_handle
${
parallel_executor_cuda_deps
}
)
cc_library
(
prune SRCS prune.cc DEPS framework_proto
)
cc_test
(
prune_test SRCS prune_test.cc DEPS op_info prune recurrent_op device_context
)
...
...
paddle/fluid/framework/details/CMakeLists.txt
浏览文件 @
31815010
...
...
@@ -4,3 +4,4 @@ cc_library(scale_loss_grad_op_handle SRCS scale_loss_grad_op_handle.cc DEPS op_h
cc_library
(
fetch_op_handle SRCS fetch_op_handle.cc DEPS op_handle_base scope lod_tensor ddim memory
)
nv_library
(
nccl_all_reduce_op_handle SRCS nccl_all_reduce_op_handle.cc DEPS op_handle_base scope lod_tensor ddim memory
dynload_cuda
)
cc_library
(
computation_op_handle SRCS computation_op_handle.cc DEPS framework_proto scope place operator op_registry
)
paddle/fluid/framework/details/computation_op_handle.cc
0 → 100644
浏览文件 @
31815010
// Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "paddle/fluid/framework/details/computation_op_handle.h"
namespace
paddle
{
namespace
framework
{
namespace
details
{
ComputationOpHandle
::
ComputationOpHandle
(
const
OpDesc
&
op_desc
,
Scope
*
scope
,
platform
::
Place
place
)
:
op_
(
framework
::
OpRegistry
::
CreateOp
(
op_desc
)),
scope_
(
scope
),
place_
(
place
)
{}
void
ComputationOpHandle
::
RunImpl
()
{
auto
*
cur_ctx
=
dev_ctx_
[
place_
];
for
(
auto
*
in
:
inputs_
)
{
bool
need_wait
=
in
->
generated_op_
&&
in
->
generated_op_
->
dev_ctx_
[
place_
]
!=
cur_ctx
;
if
(
need_wait
)
{
in
->
generated_op_
->
Wait
(
cur_ctx
);
}
}
op_
->
Run
(
*
scope_
,
place_
);
}
}
// namespace details
}
// namespace framework
}
// namespace paddle
paddle/fluid/framework/details/computation_op_handle.h
0 → 100644
浏览文件 @
31815010
// Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#pragma once
#include "paddle/fluid/framework/details/op_handle_base.h"
#include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/framework/operator.h"
#include "paddle/fluid/framework/scope.h"
#include "paddle/fluid/platform/device_context.h"
namespace
paddle
{
namespace
framework
{
namespace
details
{
struct
ComputationOpHandle
:
public
OpHandleBase
{
std
::
unique_ptr
<
OperatorBase
>
op_
;
Scope
*
scope_
;
platform
::
Place
place_
;
ComputationOpHandle
(
const
OpDesc
&
op_desc
,
Scope
*
scope
,
platform
::
Place
place
);
protected:
void
RunImpl
()
override
;
};
}
// namespace details
}
// namespace framework
}
// namespace paddle
paddle/fluid/framework/parallel_executor.cc
浏览文件 @
31815010
...
...
@@ -17,6 +17,7 @@ limitations under the License. */
#include "lod_tensor.h"
#include "lod_tensor_array.h"
#include "op_registry.h"
#include "paddle/fluid/framework/details/computation_op_handle.h"
#include "paddle/fluid/framework/details/fetch_op_handle.h"
#include "paddle/fluid/framework/details/nccl_all_reduce_op_handle.h"
#include "paddle/fluid/framework/details/op_handle_base.h"
...
...
@@ -34,6 +35,7 @@ using details::OpHandleBase;
using
details
::
ScaleLossGradOpHandle
;
using
details
::
VarHandle
;
using
details
::
VarHandleBase
;
using
details
::
ComputationOpHandle
;
class
ParallelExecutorPrivate
{
public:
...
...
@@ -127,32 +129,6 @@ class ParallelExecutorPrivate {
}
};
struct
ComputationOpHandle
:
public
OpHandleBase
{
std
::
unique_ptr
<
OperatorBase
>
op_
;
Scope
*
scope_
;
platform
::
Place
place_
;
explicit
ComputationOpHandle
(
const
OpDesc
&
op_desc
,
Scope
*
scope
,
platform
::
Place
place
)
:
op_
(
framework
::
OpRegistry
::
CreateOp
(
op_desc
)),
scope_
(
scope
),
place_
(
place
)
{}
protected:
void
RunImpl
()
override
{
auto
*
cur_ctx
=
dev_ctx_
[
place_
];
for
(
auto
*
in
:
inputs_
)
{
bool
need_wait
=
in
->
generated_op_
&&
in
->
generated_op_
->
dev_ctx_
[
place_
]
!=
cur_ctx
;
if
(
need_wait
)
{
in
->
generated_op_
->
Wait
(
cur_ctx
);
}
}
op_
->
Run
(
*
scope_
,
place_
);
}
};
ParallelExecutor
::
ParallelExecutor
(
size_t
num_threads
,
const
std
::
vector
<
platform
::
Place
>
&
places
,
const
std
::
unordered_set
<
std
::
string
>
&
params
,
...
...
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