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
BaiXuePrincess
Paddle
提交
d5ee580c
P
Paddle
项目概览
BaiXuePrincess
/
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
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
提交
d5ee580c
编写于
4月 09, 2019
作者:
X
xjqbest
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
move split filelist from trainer.py to fleet & fix error
test=develop
上级
126d2a2f
变更
2
显示空白变更内容
内联
并排
Showing
2 changed file
with
31 addition
and
1 deletion
+31
-1
python/paddle/fluid/incubate/fleet/base/role_maker.py
python/paddle/fluid/incubate/fleet/base/role_maker.py
+1
-1
python/paddle/fluid/incubate/fleet/parameter_server/__init__.py
.../paddle/fluid/incubate/fleet/parameter_server/__init__.py
+30
-0
未找到文件。
python/paddle/fluid/incubate/fleet/base/role_maker.py
浏览文件 @
d5ee580c
...
@@ -128,7 +128,7 @@ class MPIRoleMaker(RoleMakerBase):
...
@@ -128,7 +128,7 @@ class MPIRoleMaker(RoleMakerBase):
"""
"""
finalize the current MPI instance.
finalize the current MPI instance.
"""
"""
self
.
comm_
.
finalize
()
pass
class
MPISymetricRoleMaker
(
MPIRoleMaker
):
class
MPISymetricRoleMaker
(
MPIRoleMaker
):
...
...
python/paddle/fluid/incubate/fleet/parameter_server/__init__.py
浏览文件 @
d5ee580c
...
@@ -241,6 +241,35 @@ class Fleet(object):
...
@@ -241,6 +241,35 @@ class Fleet(object):
"""
"""
self
.
_fleet_ptr
.
save_model
(
save_path
)
self
.
_fleet_ptr
.
save_model
(
save_path
)
def
split_filelist
(
self
,
filelist
):
"""
split filelist before distributed training,
for example, filelist is [a, b, c ,d, e] and trainer_num = 2,
then trainer 0 gets [a, b, c] and trainer 1 gets [d, e]
Args:
filelist(list): list of filename, can be local or hdfs/afs.
Returns: list of filename which belongs to this trainer.
"""
file_num
=
len
(
filelist
)
trainer_id
=
self
.
get_worker_index
()
trainer_num
=
self
.
get_worker_num
()
if
trainer_num
>
file_num
:
raise
ValueError
(
"trainer_num should be <= file_num : "
"%s > %s"
%
(
trainer_num
,
file_num
)
)
# get interval of filelist, it's [ )
start
=
0
end
=
0
for
i
in
range
(
0
,
trainer_id
+
1
):
length
=
file_num
/
trainer_num
+
(
i
<
(
file_num
%
trainer_num
))
start
=
end
end
+=
length
myfilelist
=
filelist
[
start
:
end
]
return
myfilelist
def
_set_opt_info
(
self
,
opt_info
):
def
_set_opt_info
(
self
,
opt_info
):
"""
"""
this function saves the result from DistributedOptimizer.minimize()
this function saves the result from DistributedOptimizer.minimize()
...
@@ -337,3 +366,4 @@ save_pserver_model = fleet_instance.save_pserver_model
...
@@ -337,3 +366,4 @@ save_pserver_model = fleet_instance.save_pserver_model
worker_num
=
fleet_instance
.
get_worker_num
worker_num
=
fleet_instance
.
get_worker_num
server_num
=
fleet_instance
.
get_server_num
server_num
=
fleet_instance
.
get_server_num
worker_index
=
fleet_instance
.
get_worker_index
worker_index
=
fleet_instance
.
get_worker_index
split_filelist
=
fleet_instance
.
split_filelist
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