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
0c7f3575
P
Paddle
项目概览
PaddlePaddle
/
Paddle
接近 1 年 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 >>
未验证
提交
0c7f3575
编写于
12月 06, 2022
作者:
L
LiYuRio
提交者:
GitHub
12月 06, 2022
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
fix global gather (#48736)
上级
c5a45cc6
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
18 addition
and
15 deletion
+18
-15
python/paddle/fluid/tests/unittests/collective/collective_global_gather.py
...id/tests/unittests/collective/collective_global_gather.py
+18
-15
未找到文件。
python/paddle/fluid/tests/unittests/collective/collective_global_gather.py
浏览文件 @
0c7f3575
...
...
@@ -22,7 +22,6 @@ from test_collective_api_base import TestCollectiveAPIRunnerBase, runtime_main
import
paddle
import
paddle.distributed.utils.moe_utils
as
moe_utils
import
paddle.fluid
as
fluid
from
paddle.fluid.framework
import
_enable_legacy_dygraph
paddle
.
enable_static
()
...
...
@@ -61,8 +60,8 @@ class TestCollectiveGlobalGatherAPI(TestCollectiveAPIRunnerBase):
endpoints
=
args
[
"endpoints"
].
split
(
","
)
rank
=
args
[
"trainerid"
]
current_endpoint
=
args
[
"currentendpoint"
]
nranks
=
2
paddle
.
distributed
.
init_parallel_env
()
nranks
=
2
if
args
[
'backend'
]
==
'nccl'
:
device_id
=
int
(
os
.
getenv
(
"FLAGS_selected_gpus"
,
"0"
))
place
=
fluid
.
CUDAPlace
(
...
...
@@ -78,33 +77,37 @@ class TestCollectiveGlobalGatherAPI(TestCollectiveAPIRunnerBase):
n_expert
=
2
world_size
=
2
tot_expert
=
n_expert
*
world_size
paddle
.
disable_static
()
# Call paddle.distributed.alltoall() under legacy dygraph
_enable_legacy_dygraph
()
tmp_main_prog
=
fluid
.
Program
()
with
fluid
.
program_guard
(
tmp_main_prog
,
fluid
.
Program
()):
local_expert_count
=
paddle
.
static
.
data
(
name
=
"local_expert_count"
,
shape
=
[
tot_expert
],
dtype
=
"int64"
)
global_expert_count
=
[]
paddle
.
distributed
.
alltoall
(
paddle
.
split
(
local_expert_count
,
2
,
axis
=
0
),
global_expert_count
)
global_expert_count
=
paddle
.
concat
(
global_expert_count
,
axis
=
0
)
exe
=
fluid
.
Executor
(
place
)
exe
.
run
(
startup_prog
)
np
.
random
.
seed
(
os
.
getpid
())
local_expert_count
=
np
.
random
.
randint
(
1
,
4
,
size
=
tot_expert
).
astype
(
"int64"
)
local_expert_count
=
paddle
.
to_tensor
(
local_expert_count
)
global_expert_count
=
[]
paddle
.
distributed
.
alltoall
(
paddle
.
split
(
local_expert_count
,
2
,
axis
=
0
),
global_expert_count
(
global_expert_count
,)
=
exe
.
run
(
tmp_main_prog
,
feed
=
{
"local_expert_count"
:
local_expert_count
},
fetch_list
=
[
global_expert_count
.
name
],
)
global_expert_count
=
paddle
.
concat
(
global_expert_count
,
axis
=
0
)
global_expert_count
=
global_expert_count
.
numpy
()
local_expert_count
=
local_expert_count
.
numpy
()
fwd_expert_count
=
sum
(
global_expert_count
)
np
.
random
.
seed
(
os
.
getpid
())
local_input_buf
=
np
.
random
.
rand
(
fwd_expert_count
,
in_feat
).
astype
(
"float32"
)
paddle
.
enable_static
()
if
args
[
'static_mode'
]:
result
=
self
.
get_model
(
train_prog
,
startup_prog
,
rank
)
exe
=
fluid
.
Executor
(
place
)
exe
.
run
(
startup_prog
)
fetch_list
=
[]
for
elem
in
result
:
fetch_list
.
append
(
elem
.
name
)
...
...
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