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
PaddleRec
提交
fbbc7134
P
PaddleRec
项目概览
PaddlePaddle
/
PaddleRec
通知
68
Star
12
Fork
5
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
27
列表
看板
标记
里程碑
合并请求
10
Wiki
1
Wiki
分析
仓库
DevOps
项目成员
Pages
P
PaddleRec
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
27
Issue
27
列表
看板
标记
里程碑
合并请求
10
合并请求
10
Pages
分析
分析
仓库分析
DevOps
Wiki
1
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
fbbc7134
编写于
4月 16, 2020
作者:
T
tangwei
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
fix import
上级
fd5e7f94
变更
3
隐藏空白更改
内联
并排
Showing
3 changed file
with
23 addition
and
18 deletion
+23
-18
fleetrec/core/factory.py
fleetrec/core/factory.py
+3
-4
fleetrec/core/trainers/ctr_modul_trainer.py
fleetrec/core/trainers/ctr_modul_trainer.py
+0
-0
fleetrec/run.py
fleetrec/run.py
+20
-14
未找到文件。
fleetrec/core/factory.py
浏览文件 @
fbbc7134
...
...
@@ -17,10 +17,6 @@ import sys
import
yaml
from
fleetrec.core.trainers.single_trainer
import
SingleTrainer
from
fleetrec.core.trainers.cluster_trainer
import
ClusterTrainer
from
fleetrec.core.trainers.ctr_trainer
import
CtrPaddleTrainer
from
fleetrec.core.utils
import
envs
...
...
@@ -38,10 +34,13 @@ class TrainerFactory(object):
train_mode
=
envs
.
get_runtime_envion
(
"train.trainer"
)
if
train_mode
==
"SingleTraining"
:
from
fleetrec.core.trainers.single_trainer
import
SingleTrainer
trainer
=
SingleTrainer
(
yaml_path
)
elif
train_mode
==
"ClusterTraining"
:
from
fleetrec.core.trainers.cluster_trainer
import
ClusterTrainer
trainer
=
ClusterTrainer
(
yaml_path
)
elif
train_mode
==
"CtrTraining"
:
from
fleetrec.core.trainers.ctr_modul_trainer
import
CtrPaddleTrainer
trainer
=
CtrPaddleTrainer
(
config
)
elif
train_mode
==
"UserDefineTraining"
:
train_location
=
envs
.
get_global_env
(
"train.location"
)
...
...
fleetrec/core/trainers/ctr_trainer.py
→
fleetrec/core/trainers/ctr_
modul_
trainer.py
浏览文件 @
fbbc7134
文件已移动
fleetrec/run.py
浏览文件 @
fbbc7134
...
...
@@ -8,8 +8,6 @@ from paddle.fluid.incubate.fleet.parameter_server import version
from
fleetrec.core.factory
import
TrainerFactory
from
fleetrec.core.utils
import
envs
from
fleetrec.core.utils
import
util
from
fleetrec.core.engine.local_cluster_engine
import
LocalClusterEngine
from
fleetrec.core.engine.local_mpi_engine
import
LocalMPIEngine
def
run
(
model_yaml
):
...
...
@@ -25,22 +23,16 @@ def single_engine(single_envs, model_yaml):
def
local_cluster_engine
(
cluster_envs
,
model_yaml
):
from
fleetrec.core.engine.local_cluster_engine
import
LocalClusterEngine
print
(
envs
.
pretty_print_envs
(
cluster_envs
,
(
"Local Cluster Envs"
,
"Value"
)))
envs
.
set_runtime_envions
(
cluster_envs
)
launch
=
LocalClusterEngine
(
cluster_envs
,
model_yaml
)
launch
.
run
()
def
local_mpi_engine
(
model_yaml
):
print
(
"use 1X1 MPI ClusterTraining at localhost to run model: {}"
.
format
(
args
.
model
))
mpi_path
=
util
.
run_which
(
"mpirun"
)
if
not
mpi_path
:
raise
RuntimeError
(
"can not find mpirun, please check environment"
)
cluster_envs
=
{
"mpirun"
:
mpi_path
,
"train.trainer"
:
"CtrTraining"
}
def
local_mpi_engine
(
cluster_envs
,
model_yaml
):
from
fleetrec.core.engine.local_mpi_engine
import
LocalMPIEngine
print
(
envs
.
pretty_print_envs
(
cluster_envs
,
(
"Local MPI Cluster Envs"
,
"Value"
)))
envs
.
set_runtime_envions
(
cluster_envs
)
...
...
@@ -81,7 +73,14 @@ if __name__ == "__main__":
single_envs
=
{
"train.trainer"
:
"SingleTraining"
}
single_engine
(
single_envs
,
args
.
model
)
else
:
local_mpi_engine
(
args
.
model
)
print
(
"use 1X1 MPI ClusterTraining at localhost to run model: {}"
.
format
(
args
.
model
))
mpi_path
=
util
.
run_which
(
"mpirun"
)
if
not
mpi_path
:
raise
RuntimeError
(
"can not find mpirun, please check environment"
)
cluster_envs
=
{
"mpirun"
:
mpi_path
,
"train.trainer"
:
"CtrTraining"
}
local_mpi_engine
(
cluster_envs
,
args
.
model
)
elif
args
.
engine
.
upper
()
==
"LOCAL_CLUSTER"
:
print
(
"use 1X1 ClusterTraining at localhost to run model: {}"
.
format
(
args
.
model
))
if
version
.
is_transpiler
():
...
...
@@ -95,7 +94,14 @@ if __name__ == "__main__":
local_cluster_engine
(
cluster_envs
,
args
.
model
)
else
:
local_mpi_engine
(
args
.
model
)
print
(
"use 1X1 MPI ClusterTraining at localhost to run model: {}"
.
format
(
args
.
model
))
mpi_path
=
util
.
run_which
(
"mpirun"
)
if
not
mpi_path
:
raise
RuntimeError
(
"can not find mpirun, please check environment"
)
cluster_envs
=
{
"mpirun"
:
mpi_path
,
"train.trainer"
:
"CtrTraining"
}
local_mpi_engine
(
cluster_envs
,
args
.
model
)
elif
args
.
engine
.
upper
()
==
"CLUSTER"
:
print
(
"launch ClusterTraining with cluster to run model: {}"
.
format
(
args
.
model
))
run
(
args
.
model
)
...
...
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