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
a5b32637
P
Paddle
项目概览
PaddlePaddle
/
Paddle
1 年多 前同步成功
通知
2302
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
未验证
提交
a5b32637
编写于
9月 25, 2020
作者:
L
Leo Chen
提交者:
GitHub
9月 25, 2020
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Refine error msg in paddle/fluid/imperative (#27521)
* refine err msg * follow comments
上级
09f19532
变更
3
显示空白变更内容
内联
并排
Showing
3 changed file
with
58 addition
and
26 deletion
+58
-26
paddle/fluid/imperative/gradient_accumulator.cc
paddle/fluid/imperative/gradient_accumulator.cc
+9
-3
paddle/fluid/imperative/jit/program_desc_tracer.cc
paddle/fluid/imperative/jit/program_desc_tracer.cc
+9
-4
paddle/fluid/imperative/nccl_context.cc
paddle/fluid/imperative/nccl_context.cc
+40
-19
未找到文件。
paddle/fluid/imperative/gradient_accumulator.cc
浏览文件 @
a5b32637
...
...
@@ -13,9 +13,11 @@
// limitations under the License.
#include "paddle/fluid/imperative/gradient_accumulator.h"
#include <algorithm>
#include <memory>
#include <utility>
#include "paddle/fluid/framework/framework.pb.h"
#include "paddle/fluid/framework/lod_tensor.h"
#include "paddle/fluid/framework/selected_rows.h"
...
...
@@ -136,9 +138,13 @@ void TensorAdd(const framework::Variable& src, framework::Variable* dst) {
return
;
}
PADDLE_ENFORCE_EQ
(
dst_tensor
->
numel
()
==
numel
,
true
,
"dst_numel %d vs. src_numel %d"
,
dst_tensor
->
numel
(),
numel
);
PADDLE_ENFORCE_EQ
(
dst_tensor
->
numel
(),
numel
,
platform
::
errors
::
PreconditionNotMet
(
"The number of elements of source tensor and destination tensor "
"should be equal, but got the number of elements of source tensor is "
"%zu and the number of elements of destination tensor is %zu."
,
numel
,
dst_tensor
->
numel
()));
auto
data_type
=
src_tensor
.
type
();
auto
place
=
src_tensor
.
place
();
...
...
paddle/fluid/imperative/jit/program_desc_tracer.cc
浏览文件 @
a5b32637
...
...
@@ -13,6 +13,7 @@
// limitations under the License.
#include "paddle/fluid/imperative/jit/program_desc_tracer.h"
#include <unordered_map>
#include <unordered_set>
...
...
@@ -203,7 +204,8 @@ TracedProgramTuple ProgramDescTracer::CreateProgramDesc(
void
ProgramDescTracer
::
InsertVarIfNotExist
(
const
std
::
shared_ptr
<
VarBase
>
&
new_var
,
bool
is_input
)
{
PADDLE_ENFORCE_NOT_NULL
(
new_var
);
PADDLE_ENFORCE_NOT_NULL
(
new_var
,
platform
::
errors
::
InvalidArgument
(
"The variable to insert is NULL."
));
if
(
vars_
.
count
(
new_var
)
!=
0
)
return
;
auto
new_var_desc
=
new
framework
::
VarDesc
(
""
);
...
...
@@ -220,7 +222,9 @@ void ProgramDescTracer::InsertVarIfNotExist(
}
const
auto
&
inner_var
=
new_var
->
Var
();
PADDLE_ENFORCE_EQ
(
inner_var
.
IsInitialized
(),
true
);
PADDLE_ENFORCE_EQ
(
inner_var
.
IsInitialized
(),
true
,
platform
::
errors
::
InvalidArgument
(
"The variable to insert is not initialized."
));
if
(
inner_var
.
IsType
<
framework
::
LoDTensor
>
())
{
const
auto
&
tensor
=
inner_var
.
Get
<
framework
::
LoDTensor
>
();
new_var_desc
->
SetType
(
framework
::
proto
::
VarType
::
LOD_TENSOR
);
...
...
@@ -232,8 +236,9 @@ void ProgramDescTracer::InsertVarIfNotExist(
new_var_desc
->
SetDataType
(
framework
::
proto
::
VarType
::
FP32
);
}
}
else
{
PADDLE_THROW
(
"Not support variable type %s"
,
framework
::
ToTypeName
(
inner_var
.
Type
()));
PADDLE_THROW
(
platform
::
errors
::
InvalidArgument
(
"Not support variable type %s."
,
framework
::
ToTypeName
(
inner_var
.
Type
())));
}
}
...
...
paddle/fluid/imperative/nccl_context.cc
浏览文件 @
a5b32637
...
...
@@ -13,6 +13,7 @@
// limitations under the License.
#include "paddle/fluid/imperative/nccl_context.h"
#include "paddle/fluid/platform/collective_helper.h"
namespace
paddle
{
...
...
@@ -21,8 +22,10 @@ namespace imperative {
void
NCCLParallelContext
::
RecvNCCLID
(
const
std
::
string
&
ep
,
ncclUniqueId
*
nccl_id
)
{
auto
addr
=
paddle
::
string
::
Split
(
ep
,
':'
);
PADDLE_ENFORCE_EQ
(
addr
.
size
(),
2UL
,
"The endpoint should contain host and port: %s"
,
ep
);
PADDLE_ENFORCE_EQ
(
addr
.
size
(),
2UL
,
platform
::
errors
::
InvalidArgument
(
"The endpoint should contain host and port, but got %s."
,
ep
));
std
::
string
host
=
addr
[
0
];
int
port
=
std
::
stoi
(
addr
[
1
]);
...
...
@@ -32,27 +35,41 @@ void NCCLParallelContext::RecvNCCLID(const std::string &ep,
char
buffer
[
1024
]
=
{
0
};
int
opt
=
0
;
// creating socket fd
if
((
server_fd
=
socket
(
AF_INET
,
SOCK_STREAM
,
0
))
==
0
)
PADDLE_THROW
(
"create server fd failed"
);
if
(
setsockopt
(
server_fd
,
SOL_SOCKET
,
SO_REUSEADDR
,
&
opt
,
sizeof
(
opt
)))
PADDLE_THROW
(
"set socket opt failed"
);
if
((
server_fd
=
socket
(
AF_INET
,
SOCK_STREAM
,
0
))
==
0
)
{
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Create server file descriptor failed."
));
}
if
(
setsockopt
(
server_fd
,
SOL_SOCKET
,
SO_REUSEADDR
,
&
opt
,
sizeof
(
opt
)))
{
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Set socket options failed."
));
}
address
.
sin_family
=
AF_INET
;
address
.
sin_addr
.
s_addr
=
INADDR_ANY
;
address
.
sin_port
=
htons
(
port
);
if
(
bind
(
server_fd
,
(
struct
sockaddr
*
)
&
address
,
sizeof
(
address
))
<
0
)
PADDLE_THROW
(
"binding failed on ep: %s"
,
ep
);
if
(
bind
(
server_fd
,
(
struct
sockaddr
*
)
&
address
,
sizeof
(
address
))
<
0
)
{
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Bind on endpoint %s failed."
,
ep
));
}
VLOG
(
3
)
<<
"listening on: "
<<
ep
;
if
(
listen
(
server_fd
,
3
)
<
0
)
PADDLE_THROW
(
"listen on server fd failed"
);
if
(
listen
(
server_fd
,
3
)
<
0
)
{
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Listen on server file descriptor failed."
));
}
if
((
new_socket
=
accept
(
server_fd
,
reinterpret_cast
<
struct
sockaddr
*>
(
&
address
),
reinterpret_cast
<
socklen_t
*>
(
&
addrlen
)))
<
0
)
PADDLE_THROW
(
"accept the new socket fd failed"
);
reinterpret_cast
<
socklen_t
*>
(
&
addrlen
)))
<
0
)
{
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Accept the new socket file descriptor failed."
));
}
if
(
read
(
new_socket
,
buffer
,
1024
)
<
0
)
{
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Read from socket failed."
));
}
if
(
read
(
new_socket
,
buffer
,
1024
)
<
0
)
PADDLE_THROW
(
"reading the ncclUniqueId from socket failed"
);
VLOG
(
3
)
<<
"recevived the ncclUniqueId"
;
memcpy
(
nccl_id
,
buffer
,
NCCL_UNIQUE_ID_BYTES
);
...
...
@@ -63,8 +80,10 @@ void NCCLParallelContext::RecvNCCLID(const std::string &ep,
void
NCCLParallelContext
::
SendNCCLID
(
const
std
::
string
&
ep
,
ncclUniqueId
*
nccl_id
)
{
auto
addr
=
paddle
::
string
::
Split
(
ep
,
':'
);
PADDLE_ENFORCE_EQ
(
addr
.
size
(),
2UL
,
"The endpoint should contain host and port: %s"
,
ep
);
PADDLE_ENFORCE_EQ
(
addr
.
size
(),
2UL
,
platform
::
errors
::
InvalidArgument
(
"The endpoint should contain host and port, but got %s."
,
ep
));
std
::
string
host
=
addr
[
0
];
int
port
=
std
::
stoi
(
addr
[
1
]);
// struct sockaddr_in address;
...
...
@@ -73,15 +92,17 @@ void NCCLParallelContext::SendNCCLID(const std::string &ep,
char
buffer
[
1024
]
=
{
0
};
memcpy
(
buffer
,
nccl_id
,
NCCL_UNIQUE_ID_BYTES
);
if
((
sock
=
socket
(
AF_INET
,
SOCK_STREAM
,
0
))
<
0
)
PADDLE_THROW
(
"create socket failed"
);
if
((
sock
=
socket
(
AF_INET
,
SOCK_STREAM
,
0
))
<
0
)
{
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Create socket failed."
));
}
memset
(
&
serv_addr
,
'0'
,
sizeof
(
serv_addr
));
serv_addr
.
sin_family
=
AF_INET
;
serv_addr
.
sin_port
=
htons
(
port
);
if
(
inet_pton
(
AF_INET
,
host
.
c_str
(),
&
serv_addr
.
sin_addr
)
<=
0
)
PADDLE_THROW
(
"invalied address: %s"
,
ep
);
if
(
inet_pton
(
AF_INET
,
host
.
c_str
(),
&
serv_addr
.
sin_addr
)
<=
0
)
{
PADDLE_THROW
(
platform
::
errors
::
Unavailable
(
"Open address %s failed."
,
ep
));
}
int
try_times
=
0
;
while
(
true
)
{
...
...
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