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
OPTHREE
Paddle
提交
e1290c4f
P
Paddle
项目概览
OPTHREE
/
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
e1290c4f
编写于
3月 28, 2018
作者:
W
wanghaoshuang
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Make Average Model support for 'moving mean' and 'moving variance' of batch_normal op
上级
123cf165
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
22 addition
and
6 deletion
+22
-6
python/paddle/fluid/optimizer.py
python/paddle/fluid/optimizer.py
+22
-6
未找到文件。
python/paddle/fluid/optimizer.py
浏览文件 @
e1290c4f
...
...
@@ -850,23 +850,39 @@ class ModelAverage(Optimizer):
self
.
min_average_window
=
min_average_window
self
.
max_average_window
=
max_average_window
self
.
params_grads
=
params_grads
# append 'moving mean' and 'moving variance' to self.params_grads
pattern
=
re
.
compile
(
r
"batch_norm_\d+\.w_[1,2]"
)
for
param
in
framework
.
default_main_program
().
global_block
(
).
all_parameters
():
if
pattern
.
match
(
param
.
name
)
is
not
None
:
self
.
params_grads
.
append
((
param
,
None
))
# create a tmp gradient variable to backup parameter value
# for parameter whose grad is None
for
i
,
param_grad
in
enumerate
(
self
.
params_grads
):
param
,
grad
=
param_grad
if
grad
is
None
:
grad
=
param
.
block
.
create_var
(
name
=
unique_name
.
generate
(
"."
.
join
([
param
.
name
,
'tmp'
])),
dtype
=
param
.
dtype
,
persistable
=
False
,
stop_gradient
=
stop_gradient
)
self
.
params_grads
[
i
]
=
(
param
,
grad
)
for
param
,
grad
in
self
.
params_grads
:
if
grad
is
not
None
:
self
.
_append_average_accumulate_op
(
param
)
self
.
_append_average_accumulate_op
(
param
)
self
.
apply_program
=
Program
()
block
=
self
.
apply_program
.
global_block
()
with
program_guard
(
main_program
=
self
.
apply_program
):
for
param_grad
in
self
.
params_grads
:
if
param_grad
[
1
]
is
not
None
:
self
.
_add_average_apply_op
(
block
,
param_grad
)
self
.
_add_average_apply_op
(
block
,
param_grad
)
self
.
restore_program
=
Program
()
block
=
self
.
restore_program
.
global_block
()
with
program_guard
(
main_program
=
self
.
restore_program
):
for
param_grad
in
self
.
params_grads
:
if
param_grad
[
1
]
is
not
None
:
self
.
_add_average_restore_op
(
block
,
param_grad
)
self
.
_add_average_restore_op
(
block
,
param_grad
)
def
_add_average_apply_op
(
self
,
block
,
param_grad
):
param
=
block
.
clone_variable
(
param_grad
[
0
])
...
...
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