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
马赛克医院
Paddle
提交
54afcb7e
P
Paddle
项目概览
马赛克医院
/
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
体验新版 GitCode,发现更多精彩内容 >>
提交
54afcb7e
编写于
1月 09, 2019
作者:
T
tensor-tang
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
add compare zerocopy test with native result
test=develop
上级
13706013
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
36 addition
and
17 deletion
+36
-17
paddle/fluid/inference/tests/api/analyzer_seq_pool1_tester.cc
...le/fluid/inference/tests/api/analyzer_seq_pool1_tester.cc
+36
-17
未找到文件。
paddle/fluid/inference/tests/api/analyzer_seq_pool1_tester.cc
浏览文件 @
54afcb7e
...
...
@@ -168,15 +168,13 @@ TEST(Analyzer_seq_pool1, compare) {
reinterpret_cast
<
const
PaddlePredictor
::
Config
*>
(
&
cfg
),
input_slots_all
);
}
// Check the fuse status
TEST
(
Analyzer_seq_pool1
,
fuse_statis
)
{
void
analysis_fuse_statis
(
bool
use_zerocopy
)
{
AnalysisConfig
cfg
;
SetConfig
(
&
cfg
);
cfg
.
SwitchUseFeedFetchOps
(
!
use_zerocopy
);
int
num_ops
;
auto
predictor
=
CreatePaddlePredictor
<
AnalysisConfig
>
(
cfg
);
auto
fuse_statis
=
GetFuseStatis
(
static_cast
<
AnalysisPredictor
*>
(
predictor
.
get
()),
&
num_ops
);
auto
fuse_statis
=
GetFuseStatis
(
predictor
.
get
(),
&
num_ops
);
ASSERT_TRUE
(
fuse_statis
.
count
(
"fc_fuse"
));
ASSERT_EQ
(
fuse_statis
.
at
(
"fc_fuse"
),
10
);
ASSERT_TRUE
(
fuse_statis
.
count
(
"seqpool_concat_fuse"
));
...
...
@@ -185,6 +183,9 @@ TEST(Analyzer_seq_pool1, fuse_statis) {
EXPECT_EQ
(
num_ops
,
195
);
}
// Check the fuse status
TEST
(
Analyzer_seq_pool1
,
fuse_statis
)
{
analysis_fuse_statis
(
false
);
}
void
PrepareZeroCopyInputs
(
const
std
::
unique_ptr
<
PaddlePredictor
>
&
predictor
,
std
::
vector
<
std
::
unique_ptr
<
ZeroCopyTensor
>>
*
inputs
)
{
...
...
@@ -202,7 +203,8 @@ void PrepareZeroCopyInputs(
}
}
std
::
unique_ptr
<
ZeroCopyTensor
>
zerocopy_profile
(
int
repeat_times
)
{
// return the output values
std
::
vector
<
float
>
zerocopy_profile
(
int
repeat_times
)
{
AnalysisConfig
config
;
SetConfig
(
&
config
);
config
.
SwitchUseFeedFetchOps
(
false
);
...
...
@@ -225,23 +227,40 @@ std::unique_ptr<ZeroCopyTensor> zerocopy_profile(int repeat_times) {
}
PrintTime
(
FLAGS_batch_size
,
repeat_times
,
1
,
0
,
timer
.
toc
()
/
repeat_times
,
1
);
return
output_tensor
;
VLOG
(
3
)
<<
"ZeroCopy output: "
<<
DescribeZeroCopyTensor
(
*
output_tensor
);
PaddlePlace
place
;
int
output_size
{
0
};
auto
*
pdata
=
output_tensor
->
data
<
float
>
(
&
place
,
&
output_size
);
std
::
vector
<
float
>
res
(
output_size
);
for
(
int
i
=
0
;
i
<
output_size
;
++
i
)
{
res
[
i
]
=
pdata
[
i
];
}
return
res
;
}
TEST
(
Analyzer_seq_pool1
,
zerocopy_profile
)
{
zerocopy_profile
(
FLAGS_repeat
);
}
TEST
(
Analyzer_seq_pool1
,
zerocopy_fuse_statis
)
{
TEST
(
Analyzer_seq_pool1
,
zerocopy_fuse_statis
)
{
analysis_fuse_statis
(
true
);
}
TEST
(
Analyzer_seq_pool1
,
zerocopy_compare_native
)
{
AnalysisConfig
config
;
SetConfig
(
&
config
);
config
.
SwitchUseFeedFetchOps
(
false
);
auto
predictor
=
CreatePaddlePredictor
<
AnalysisConfig
>
(
config
);
int
num_ops
;
auto
fuse_statis
=
GetFuseStatis
(
predictor
.
get
(),
&
num_ops
);
ASSERT_TRUE
(
fuse_statis
.
count
(
"fc_fuse"
));
ASSERT_EQ
(
fuse_statis
.
at
(
"fc_fuse"
),
10
);
ASSERT_TRUE
(
fuse_statis
.
count
(
"seqpool_concat_fuse"
));
EXPECT_EQ
(
fuse_statis
.
at
(
"seqpool_concat_fuse"
),
2
);
ASSERT_EQ
(
num_ops
,
195
);
config
.
SwitchUseFeedFetchOps
(
true
);
auto
predictor
=
CreatePaddlePredictor
<
NativeConfig
>
(
config
.
ToNativeConfig
());
std
::
vector
<
PaddleTensor
>
native_outputs
;
std
::
vector
<
std
::
vector
<
PaddleTensor
>>
input_slots_all
;
SetInput
(
&
input_slots_all
);
ASSERT_TRUE
(
predictor
->
Run
(
input_slots_all
[
0
],
&
native_outputs
));
EXPECT_EQ
(
native_outputs
.
size
(),
1UL
);
auto
zerocopy_output
=
zerocopy_profile
(
1
);
EXPECT_EQ
(
zerocopy_output
.
size
()
*
sizeof
(
float
),
native_outputs
.
front
().
data
.
length
());
auto
*
native_data
=
static_cast
<
float
*>
(
native_outputs
.
front
().
data
.
data
());
for
(
size_t
i
=
0
;
i
<
zerocopy_output
.
size
();
++
i
)
{
EXPECT_NEAR
(
zerocopy_output
[
i
],
native_data
[
i
],
1e-3
);
}
}
}
// namespace analysis
...
...
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