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
PaddleDetection
提交
0fad63a3
P
PaddleDetection
项目概览
PaddlePaddle
/
PaddleDetection
大约 1 年 前同步成功
通知
695
Star
11112
Fork
2696
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
184
列表
看板
标记
里程碑
合并请求
40
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
PaddleDetection
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
184
Issue
184
列表
看板
标记
里程碑
合并请求
40
合并请求
40
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
0fad63a3
编写于
6月 25, 2019
作者:
T
tangwei12
提交者:
guru4elephant
6月 25, 2019
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
fix paddle cloud role maker bug (#18269) (#18311)
* fix paddle cloud role maker bug
上级
f6432604
变更
1
显示空白变更内容
内联
并排
Showing
1 changed file
with
31 addition
and
12 deletion
+31
-12
python/paddle/fluid/incubate/fleet/base/role_maker.py
python/paddle/fluid/incubate/fleet/base/role_maker.py
+31
-12
未找到文件。
python/paddle/fluid/incubate/fleet/base/role_maker.py
浏览文件 @
0fad63a3
...
@@ -19,6 +19,8 @@ __all__ = [
...
@@ -19,6 +19,8 @@ __all__ = [
'UserDefinedCollectiveRoleMaker'
,
'PaddleCloudRoleMaker'
'UserDefinedCollectiveRoleMaker'
,
'PaddleCloudRoleMaker'
]
]
import
os
class
Role
:
class
Role
:
WORKER
=
1
WORKER
=
1
...
@@ -295,45 +297,62 @@ class MPISymetricRoleMaker(MPIRoleMaker):
...
@@ -295,45 +297,62 @@ class MPISymetricRoleMaker(MPIRoleMaker):
class
PaddleCloudRoleMaker
(
RoleMakerBase
):
class
PaddleCloudRoleMaker
(
RoleMakerBase
):
def
__init__
(
self
):
def
__init__
(
self
):
super
(
PaddleCloudRoleMaker
,
self
).
__init__
()
super
(
PaddleCloudRoleMaker
,
self
).
__init__
()
self
.
_role_is_generated
=
False
def
generate_role
(
self
):
def
generate_role
(
self
):
if
not
self
.
_role_is_generated
:
if
not
self
.
_role_is_generated
:
self
.
port
=
os
.
getenv
(
"PADDLE_PORT"
,
"6174"
)
self
.
port
=
os
.
getenv
(
"PADDLE_PORT"
,
"6174"
)
self
.
pserver_ips
=
os
.
getenv
(
"PADDLE_PSERVERS"
,
""
)
self
.
pserver_ips
=
os
.
getenv
(
"PADDLE_PSERVERS"
,
""
)
eplist
=
[]
eplist
=
[]
for
ip
in
pserver_ips
.
split
(
","
):
for
ip
in
self
.
pserver_ips
.
split
(
","
):
eplist
.
append
(
':'
.
join
([
ip
,
port
]))
eplist
.
append
(
':'
.
join
([
ip
,
self
.
port
]))
self
.
endpoints
=
","
.
join
(
eplist
)
self
.
endpoints
=
","
.
join
(
eplist
)
self
.
trainers
=
int
(
os
.
getenv
(
"PADDLE_TRAINERS_NUM"
,
"1"
))
self
.
_
trainers
=
int
(
os
.
getenv
(
"PADDLE_TRAINERS_NUM"
,
"1"
))
self
.
current_endpoint
=
os
.
getenv
(
"POD_IP"
,
self
.
current_endpoint
=
os
.
getenv
(
"POD_IP"
,
"localhost"
)
+
":"
+
port
"localhost"
)
+
":"
+
self
.
port
self
.
role
=
os
.
getenv
(
"TRAINING_ROLE"
,
"TRAINER"
)
self
.
role
=
os
.
getenv
(
"TRAINING_ROLE"
,
"TRAINER"
)
self
.
trainer_id
=
int
(
os
.
getenv
(
"PADDLE_TRAINER_ID"
,
"0"
))
self
.
trainer_id
=
int
(
os
.
getenv
(
"PADDLE_TRAINER_ID"
,
"0"
))
self
.
eplist
=
eplist
self
.
eplist
=
eplist
print
(
"PaddleCloudRoleMaker() endpoints: %s"
%
self
.
endpoints
)
self
.
endpoints
=
self
.
endpoints
.
split
(
","
)
self
.
endpoints
=
self
.
endpoints
.
split
(
","
)
self
.
_server_endpoints
=
self
.
endpoints
if
self
.
role
.
upper
()
==
"PSERVER"
:
if
self
.
role
.
upper
()
==
"PSERVER"
:
self
.
current_id
=
self
.
endpoints
.
index
(
self
.
current_endpoint
)
self
.
_current_id
=
self
.
endpoints
.
index
(
self
.
current_endpoint
)
self
.
_role
=
Role
.
SERVER
else
:
else
:
self
.
current_id
=
self
.
trainer_id
self
.
_current_id
=
self
.
trainer_id
self
.
_role
=
Role
.
WORKER
self
.
_role_is_generated
=
True
self
.
_role_is_generated
=
True
def
is_wokrer
(
self
):
def
is_worker
(
self
):
if
not
self
.
_role_is_generated
:
self
.
generate_role
()
return
self
.
_role
==
Role
.
WORKER
return
self
.
_role
==
Role
.
WORKER
def
is_server
(
self
):
def
is_server
(
self
):
if
not
self
.
_role_is_generated
:
self
.
generate_role
()
return
self
.
_role
==
Role
.
SERVER
return
self
.
_role
==
Role
.
SERVER
def
is_first_worker
(
self
):
def
is_first_worker
(
self
):
if
not
self
.
_role_is_generated
:
self
.
generate_role
()
return
self
.
_role
==
Role
.
WORKER
and
self
.
_current_id
==
0
return
self
.
_role
==
Role
.
WORKER
and
self
.
_current_id
==
0
def
worker_index
(
self
):
def
worker_index
(
self
):
if
not
self
.
_role_is_generated
:
self
.
generate_role
()
return
self
.
_current_id
return
self
.
_current_id
def
server_index
(
self
):
def
server_index
(
self
):
if
not
self
.
_role_is_generated
:
self
.
generate_role
()
return
self
.
_current_id
return
self
.
_current_id
def
worker_num
(
self
):
def
worker_num
(
self
):
return
self
.
_worker_num
if
not
self
.
_role_is_generated
:
self
.
generate_role
()
return
self
.
_trainers
class
UserDefinedRoleMaker
(
RoleMakerBase
):
class
UserDefinedRoleMaker
(
RoleMakerBase
):
...
...
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