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
cadc6a97
P
Paddle
项目概览
PaddlePaddle
/
Paddle
11 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
cadc6a97
编写于
10月 14, 2019
作者:
W
WangXi
提交者:
gongweibao
10月 15, 2019
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
fix dgc test and bug when not set trainers_endpoints_, test=develop (#20617)
上级
46797f53
变更
3
显示空白变更内容
内联
并排
Showing
3 changed file
with
8 addition
and
4 deletion
+8
-4
paddle/fluid/framework/ir/multi_devices_graph_pass/multi_devices_graph_pass.cc
...k/ir/multi_devices_graph_pass/multi_devices_graph_pass.cc
+1
-2
python/paddle/fluid/incubate/fleet/collective/__init__.py
python/paddle/fluid/incubate/fleet/collective/__init__.py
+0
-1
python/paddle/fluid/tests/unittests/test_dist_base.py
python/paddle/fluid/tests/unittests/test_dist_base.py
+7
-1
未找到文件。
paddle/fluid/framework/ir/multi_devices_graph_pass/multi_devices_graph_pass.cc
浏览文件 @
cadc6a97
...
...
@@ -465,8 +465,7 @@ void MultiDevSSAGraphBuilderBase::CreateAllReduceOp(ir::Graph *result,
new
details
::
SparseAllReduceOpHandle
(
result
->
CreateEmptyNode
(
"allreduce"
,
ir
::
Node
::
Type
::
kOperation
),
scopes
,
places
,
multi_nccl_ctxs_
,
is_encoded
,
static_cast
<
int
>
(
strategy_
.
trainers_endpoints_
.
size
())
*
places_
.
size
()));
strategy_
.
num_trainers_
*
places_
.
size
()));
}
else
{
result
->
Get
<
GraphOps
>
(
kGraphOps
).
emplace_back
(
new
details
::
AllReduceOpHandle
(
...
...
python/paddle/fluid/incubate/fleet/collective/__init__.py
浏览文件 @
cadc6a97
...
...
@@ -271,7 +271,6 @@ class CollectiveOptimizer(DistributedOptimizer):
node_num
=
self
.
_node_num
()
assert
node_num
>=
1
,
"nccl2 node_num must >= 1, now:{}"
%
node_num
self
.
_strategy
.
fuse_all_reduce_ops
=
True
exec_strategy
=
self
.
_strategy
.
exec_strategy
if
node_num
<=
1
:
...
...
python/paddle/fluid/tests/unittests/test_dist_base.py
浏览文件 @
cadc6a97
...
...
@@ -291,6 +291,10 @@ class TestDistRunnerBase(object):
build_stra
.
num_trainers
=
1
build_stra
.
trainer_id
=
0
if
args
.
use_dgc
:
# fuse_all_reduce_ops require that gradients should not be sparse types
build_stra
.
fuse_all_reduce_ops
=
False
print_to_err
(
type
(
self
).
__name__
,
"begin to compile with data parallel"
)
binary
=
compiler
.
CompiledProgram
(
trainer_prog
).
with_data_parallel
(
loss_name
=
avg_cost
.
name
,
...
...
@@ -852,7 +856,9 @@ class TestDistBase(unittest.TestCase):
if
check_error_log
:
required_envs
[
"GLOG_vmodule"
]
=
\
"fused_all_reduce_op_handle=10,all_reduce_op_handle=10,alloc_continuous_space_op=10,fuse_all_reduce_op_pass=10,alloc_continuous_space_for_grad_pass=10,fast_threaded_ssa_graph_executor=10,executor=10,operator=10"
"fused_all_reduce_op_handle=10,all_reduce_op_handle=10,alloc_continuous_space_op=10,fuse_all_reduce_op_pass=10,"
\
"alloc_continuous_space_for_grad_pass=10,fast_threaded_ssa_graph_executor=10,executor=10,operator=10,"
\
"sparse_all_reduce_op_handle=10"
required_envs
[
"GLOG_logtostderr"
]
=
"1"
local_losses
\
...
...
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