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
847c48a8
P
Paddle
项目概览
PaddlePaddle
/
Paddle
大约 1 年 前同步成功
通知
2298
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
未验证
提交
847c48a8
编写于
5月 16, 2023
作者:
M
Mahmoud Ashraf
提交者:
GitHub
5月 16, 2023
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
fix simple typos (#53783)
* correct 1th to 1st * correct 1th to 1st * fix typo * fix typos
上级
926b8861
变更
4
隐藏空白更改
内联
并排
Showing
4 changed file
with
14 addition
and
14 deletion
+14
-14
paddle/fluid/framework/ir/seq_concat_fc_fuse_pass.cc
paddle/fluid/framework/ir/seq_concat_fc_fuse_pass.cc
+2
-2
python/paddle/distributed/utils/moe_utils.py
python/paddle/distributed/utils/moe_utils.py
+9
-9
python/paddle/fluid/tests/unittests/test_sparse_utils_op.py
python/paddle/fluid/tests/unittests/test_sparse_utils_op.py
+2
-2
tools/CrossStackProfiler/DCGMFileReader.py
tools/CrossStackProfiler/DCGMFileReader.py
+1
-1
未找到文件。
paddle/fluid/framework/ir/seq_concat_fc_fuse_pass.cc
浏览文件 @
847c48a8
...
...
@@ -32,8 +32,8 @@ PDNode* BuildSeqExpandConcatPattern(PDPattern* pattern) {
// The following variables will be treat as inputs:
// concat mid input, 0th input for fused op
// sequence_expand input, 1
th
input for fused op
// sequence_expand input, 2
th
input for fused op
// sequence_expand input, 1
st
input for fused op
// sequence_expand input, 2
nd
input for fused op
// The following variables will be treat as outputs:
// concat output
...
...
python/paddle/distributed/utils/moe_utils.py
浏览文件 @
847c48a8
...
...
@@ -35,19 +35,19 @@ def global_scatter(
local_count[0] represents taking out 2 batches from x and sending 2 batches to the 0th expert of the 0th card;
local_count[1] represents taking out 0 batches from x and sending 0 batches to the 1
th
expert of the 0th card;
local_count[1] represents taking out 0 batches from x and sending 0 batches to the 1
st
expert of the 0th card;
local_count[2] represents taking out 2 batches from x and sending 2 batches to the 0th expert of the 1
th
card;
local_count[2] represents taking out 2 batches from x and sending 2 batches to the 0th expert of the 1
st
card;
local_count[3] represents taking out 0 batches from x and sending 0 batches to the 1
th expert of the 1th
card;
local_count[3] represents taking out 0 batches from x and sending 0 batches to the 1
st expert of the 1st
card;
Therefore, the global_count[0] of the 0th card is equal to 2, which means that 2 batches of data are received from the 0th card to the 0th expert;
the global_count[1] of the 0th card is equal to 0, which means that 0 batches of data are received from the 0th card to the 1
th
expert;
the global_count[1] of the 0th card is equal to 0, which means that 0 batches of data are received from the 0th card to the 1
st
expert;
the global_count[0] of the 1
th
card is equal to 2, which means that 2 batches of data are received from the 0th card to the 0th expert;
the global_count[0] of the 1
st
card is equal to 2, which means that 2 batches of data are received from the 0th card to the 0th expert;
the global_count[1] of the 1
th card is equal to 0, which means that 0 batches of data are received from the 0th card to the 1th
expert.
the global_count[1] of the 1
st card is equal to 0, which means that 0 batches of data are received from the 0th card to the 1st
expert.
.. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/global_scatter_gather.png
:width: 800
...
...
@@ -160,11 +160,11 @@ def global_gather(
The global_count[0] of the 0th card represents sending 2 data to the 0th expert of the 0th card;
The global_count[1] of the 0th card represents sending 0 data to the 1
th
expert of the 0th card;
The global_count[1] of the 0th card represents sending 0 data to the 1
st
expert of the 0th card;
The global_count[0] of the 1
th
card represents sending 2 data to the 0th expert of the 0th card;
The global_count[0] of the 1
st
card represents sending 2 data to the 0th expert of the 0th card;
The global_count[1] of the 1
th card represents sending 0 data to the 1th
expert of the 0th card.
The global_count[1] of the 1
st card represents sending 0 data to the 1st
expert of the 0th card.
.. image:: https://githubraw.cdn.bcebos.com/PaddlePaddle/docs/develop/docs/api/paddle/distributed/img/global_scatter_gather.png
:width: 800
...
...
python/paddle/fluid/tests/unittests/test_sparse_utils_op.py
浏览文件 @
847c48a8
...
...
@@ -342,12 +342,12 @@ class TestSparseConvert(unittest.TestCase):
verify
(
dense_x
)
dense_x
=
paddle
.
randn
(
shape
)
# set the 1
th
batch to zero
# set the 1
st
batch to zero
dense_x
[
1
]
=
0
verify
(
dense_x
)
dense_x
=
paddle
.
randn
(
shape
)
# set the 2
th
batch to zero
# set the 2
nd
batch to zero
dense_x
[
2
]
=
0
verify
(
dense_x
)
...
...
tools/CrossStackProfiler/DCGMFileReader.py
浏览文件 @
847c48a8
...
...
@@ -136,7 +136,7 @@ class dcgmFileReader(FileReader):
)
and
not
line
.
strip
().
startswith
(
"# Entity"
):
continue
# skip non-needed headers (only the header in 1
th
line was needed)
# skip non-needed headers (only the header in 1
st
line was needed)
if
line
.
strip
().
startswith
(
"# Entity"
):
line
=
line
.
strip
()[
2
:]
...
...
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