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
csdn_franckjun
Paddle
提交
b1fc2386
P
Paddle
项目概览
csdn_franckjun
/
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
1
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
1
Issue
1
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
b1fc2386
编写于
8月 23, 2018
作者:
G
guochaorong
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Revert "Disable in_place in batch_norm API. (#12736)"
This reverts commit
f5d5d7b2
.
上级
7570e5ef
变更
4
隐藏空白更改
内联
并排
Showing
4 changed file
with
5 addition
and
13 deletion
+5
-13
paddle/fluid/operators/batch_norm_op.cc
paddle/fluid/operators/batch_norm_op.cc
+1
-1
python/paddle/fluid/layers/nn.py
python/paddle/fluid/layers/nn.py
+2
-7
python/paddle/fluid/nets.py
python/paddle/fluid/nets.py
+1
-1
python/paddle/fluid/tests/book/test_image_classification.py
python/paddle/fluid/tests/book/test_image_classification.py
+1
-4
未找到文件。
paddle/fluid/operators/batch_norm_op.cc
浏览文件 @
b1fc2386
...
...
@@ -135,7 +135,7 @@ class BatchNormOpMaker : public framework::OpProtoAndCheckerMaker {
AddInput
(
"Variance"
,
"The global variance (for training) "
"or estimated Variance (for testing)"
);
AddOutput
(
"Y"
,
"result after normalization"
);
AddOutput
(
"Y"
,
"result after normalization"
)
.
Reuse
(
"X"
)
;
AddOutput
(
"MeanOut"
,
"Share memory with Mean. "
"Store the global mean when training"
)
...
...
python/paddle/fluid/layers/nn.py
浏览文件 @
b1fc2386
...
...
@@ -27,7 +27,6 @@ from . import utils
import
random
from
..
import
unique_name
from
functools
import
reduce
import
warnings
__all__
=
[
'fc'
,
...
...
@@ -2048,7 +2047,7 @@ def batch_norm(input,
param_attr(ParamAttr): The parameter attribute for Parameter `scale`.
bias_attr(ParamAttr): The parameter attribute for Parameter `bias`.
data_layout(string, default NCHW): NCHW|NHWC
in_place(bool, Default False):
This argument is deprecated since 0.15.0
.
in_place(bool, Default False):
Make the input and output of batch norm reuse memory
.
use_mkldnn(bool, Default false): ${use_mkldnn_comment}
name(string, Default None): A name for this layer(optional). If set None, the layer
will be named automatically.
...
...
@@ -2070,10 +2069,6 @@ def batch_norm(input,
helper
=
LayerHelper
(
'batch_norm'
,
**
locals
())
dtype
=
helper
.
input_dtype
()
if
in_place
:
raise
warnings
.
warn
(
"The argument in_place is deprecated since 0.15.0, "
"please do not set it True."
)
input_shape
=
input
.
shape
if
data_layout
==
'NCHW'
:
channel_num
=
input_shape
[
1
]
...
...
@@ -2123,7 +2118,7 @@ def batch_norm(input,
saved_mean
=
helper
.
create_tmp_variable
(
dtype
=
dtype
,
stop_gradient
=
True
)
saved_variance
=
helper
.
create_tmp_variable
(
dtype
=
dtype
,
stop_gradient
=
True
)
batch_norm_out
=
helper
.
create_tmp_variable
(
dtype
)
batch_norm_out
=
input
if
in_place
else
helper
.
create_tmp_variable
(
dtype
)
helper
.
append_op
(
type
=
"batch_norm"
,
...
...
python/paddle/fluid/nets.py
浏览文件 @
b1fc2386
...
...
@@ -229,7 +229,7 @@ def img_conv_group(input,
use_mkldnn
=
use_mkldnn
)
if
conv_with_batchnorm
[
i
]:
tmp
=
layers
.
batch_norm
(
input
=
tmp
,
act
=
conv_act
)
tmp
=
layers
.
batch_norm
(
input
=
tmp
,
act
=
conv_act
,
in_place
=
True
)
drop_rate
=
conv_batchnorm_drop_rate
[
i
]
if
abs
(
drop_rate
)
>
1e-5
:
tmp
=
layers
.
dropout
(
x
=
tmp
,
dropout_prob
=
drop_rate
)
...
...
python/paddle/fluid/tests/book/test_image_classification.py
浏览文件 @
b1fc2386
...
...
@@ -256,10 +256,7 @@ def main(net_type, use_cuda, is_local=True):
save_dirname
=
"image_classification_"
+
net_type
+
".inference.model"
train
(
net_type
,
use_cuda
,
save_dirname
,
is_local
)
# There is bug in fluid.InferenceTranspiler for VGG.
if
net_type
==
"resnet"
:
infer
(
use_cuda
,
save_dirname
)
infer
(
use_cuda
,
save_dirname
)
class
TestImageClassification
(
unittest
.
TestCase
):
...
...
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