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
PaddleDetection
提交
265fd33c
P
PaddleDetection
项目概览
PaddlePaddle
/
PaddleDetection
大约 1 年 前同步成功
通知
695
Star
11112
Fork
2696
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
184
列表
看板
标记
里程碑
合并请求
40
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
PaddleDetection
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
184
Issue
184
列表
看板
标记
里程碑
合并请求
40
合并请求
40
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
提交
265fd33c
编写于
1月 10, 2019
作者:
M
minqiyang
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
remove weight decay ut
test=release/1.2
上级
f04ec6bd
变更
1
显示空白变更内容
内联
并排
Showing
1 changed file
with
0 addition
and
188 deletion
+0
-188
python/paddle/fluid/tests/unittests/test_weight_decay.py
python/paddle/fluid/tests/unittests/test_weight_decay.py
+0
-188
未找到文件。
python/paddle/fluid/tests/unittests/test_weight_decay.py
已删除
100644 → 0
浏览文件 @
f04ec6bd
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from
__future__
import
print_function
import
contextlib
import
unittest
from
functools
import
partial
import
numpy
as
np
import
paddle
import
paddle.fluid.core
as
core
import
paddle.fluid
as
fluid
def
get_places
():
places
=
[]
if
core
.
is_compiled_with_cuda
():
places
.
append
(
core
.
CUDAPlace
(
0
))
return
places
@
contextlib
.
contextmanager
def
prog_scope_guard
(
main_prog
,
startup_prog
):
scope
=
fluid
.
core
.
Scope
()
with
fluid
.
unique_name
.
guard
():
with
fluid
.
scope_guard
(
scope
):
with
fluid
.
program_guard
(
main_prog
,
startup_prog
):
yield
def
bow_net
(
data
,
label
,
dict_dim
,
is_sparse
=
False
,
emb_dim
=
128
,
hid_dim
=
128
,
hid_dim2
=
96
,
class_dim
=
2
):
"""
BOW net
This model is from https://github.com/PaddlePaddle/models:
fluid/PaddleNLP/text_classification/nets.py
"""
emb
=
fluid
.
layers
.
embedding
(
input
=
data
,
is_sparse
=
is_sparse
,
size
=
[
dict_dim
,
emb_dim
])
bow
=
fluid
.
layers
.
sequence_pool
(
input
=
emb
,
pool_type
=
'sum'
)
bow_tanh
=
fluid
.
layers
.
tanh
(
bow
)
fc_1
=
fluid
.
layers
.
fc
(
input
=
bow_tanh
,
size
=
hid_dim
,
act
=
"tanh"
)
fc_2
=
fluid
.
layers
.
fc
(
input
=
fc_1
,
size
=
hid_dim2
,
act
=
"tanh"
)
prediction
=
fluid
.
layers
.
fc
(
input
=
[
fc_2
],
size
=
class_dim
,
act
=
"softmax"
)
cost
=
fluid
.
layers
.
cross_entropy
(
input
=
prediction
,
label
=
label
)
avg_cost
=
fluid
.
layers
.
mean
(
x
=
cost
)
return
avg_cost
class
TestWeightDecay
(
unittest
.
TestCase
):
def
setUp
(
self
):
self
.
word_dict
=
paddle
.
dataset
.
imdb
.
word_dict
()
reader
=
paddle
.
batch
(
paddle
.
dataset
.
imdb
.
train
(
self
.
word_dict
),
batch_size
=
4
)()
self
.
train_data
=
[
next
(
reader
)
for
_
in
range
(
5
)]
self
.
learning_rate
=
.
5
def
run_executor
(
self
,
place
,
feed_list
,
loss
):
exe
=
fluid
.
Executor
(
place
)
feeder
=
fluid
.
DataFeeder
(
feed_list
=
feed_list
,
place
=
place
)
exe
.
run
(
fluid
.
default_startup_program
())
main_prog
=
fluid
.
default_main_program
()
loss_set
=
[]
for
data
in
self
.
train_data
:
out
=
exe
.
run
(
main_prog
,
feed
=
feeder
.
feed
(
data
),
fetch_list
=
[
loss
.
name
])
print
(
"loss %s"
%
(
np
.
average
(
out
)))
loss_set
.
append
(
np
.
average
(
out
))
return
loss_set
def
run_parallel_exe
(
self
,
place
,
feed_list
,
loss
,
use_cuda
=
True
,
use_reduce
=
False
,
use_fast_executor
=
False
,
use_ir_memory_optimize
=
False
):
exe
=
fluid
.
Executor
(
place
)
feeder
=
fluid
.
DataFeeder
(
feed_list
=
feed_list
,
place
=
place
)
exe
.
run
(
fluid
.
default_startup_program
())
exec_strategy
=
fluid
.
ExecutionStrategy
()
if
use_fast_executor
:
exec_strategy
.
use_experimental_executor
=
True
build_strategy
=
fluid
.
BuildStrategy
()
build_strategy
.
reduce_strategy
=
fluid
.
BuildStrategy
.
ReduceStrategy
.
Reduce
\
if
use_reduce
else
fluid
.
BuildStrategy
.
ReduceStrategy
.
AllReduce
build_strategy
.
memory_optimize
=
use_ir_memory_optimize
parallel_exe
=
fluid
.
ParallelExecutor
(
use_cuda
,
loss_name
=
loss
.
name
,
exec_strategy
=
exec_strategy
,
build_strategy
=
build_strategy
)
loss_set
=
[]
for
data
in
self
.
train_data
:
out
=
parallel_exe
.
run
(
feed
=
feeder
.
feed
(
data
),
fetch_list
=
[
loss
.
name
])
print
(
"loss %s"
%
(
np
.
average
(
out
)))
loss_set
.
append
(
np
.
average
(
out
))
return
loss_set
def
check_weight_decay
(
self
,
place
,
model
,
use_parallel_exe
=
False
,
use_reduce
=
False
):
main_prog
=
fluid
.
framework
.
Program
()
startup_prog
=
fluid
.
framework
.
Program
()
startup_prog
.
random_seed
=
1
with
prog_scope_guard
(
main_prog
=
main_prog
,
startup_prog
=
startup_prog
):
data
=
fluid
.
layers
.
data
(
name
=
"words"
,
shape
=
[
1
],
dtype
=
"int64"
,
lod_level
=
1
)
label
=
fluid
.
layers
.
data
(
name
=
"label"
,
shape
=
[
1
],
dtype
=
"int64"
)
avg_cost
=
model
(
data
,
label
,
len
(
self
.
word_dict
))
param_list
=
[(
var
,
var
*
self
.
learning_rate
)
for
var
in
main_prog
.
block
(
0
).
all_parameters
()]
optimizer
=
fluid
.
optimizer
.
Adagrad
(
learning_rate
=
self
.
learning_rate
)
optimizer
.
minimize
(
avg_cost
)
for
params
in
param_list
:
updated_p
=
fluid
.
layers
.
elementwise_sub
(
x
=
params
[
0
],
y
=
params
[
1
])
fluid
.
layers
.
assign
(
input
=
updated_p
,
output
=
params
[
0
])
if
use_parallel_exe
:
loss
=
self
.
run_parallel_exe
(
place
,
[
data
,
label
],
loss
=
avg_cost
,
use_cuda
=
True
,
use_reduce
=
use_reduce
)
else
:
loss
=
self
.
run_executor
(
place
,
[
data
,
label
],
loss
=
avg_cost
)
return
loss
def
test_weight_decay
(
self
):
model
=
partial
(
bow_net
,
is_sparse
=
False
)
for
place
in
get_places
():
loss
=
self
.
check_weight_decay
(
place
,
model
,
use_parallel_exe
=
False
)
loss2
=
self
.
check_weight_decay
(
place
,
model
,
use_parallel_exe
=
True
,
use_reduce
=
False
)
for
i
in
range
(
len
(
loss
)):
assert
np
.
isclose
(
a
=
loss
[
i
],
b
=
loss2
[
i
],
rtol
=
5e-5
)
loss3
=
self
.
check_weight_decay
(
place
,
model
,
use_parallel_exe
=
True
,
use_reduce
=
True
)
for
i
in
range
(
len
(
loss
)):
assert
np
.
isclose
(
a
=
loss
[
i
],
b
=
loss3
[
i
],
rtol
=
5e-5
)
if
__name__
==
'__main__'
:
unittest
.
main
()
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