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
models
提交
f3ebf87c
M
models
项目概览
PaddlePaddle
/
models
大约 1 年 前同步成功
通知
222
Star
6828
Fork
2962
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
602
列表
看板
标记
里程碑
合并请求
255
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
M
models
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
602
Issue
602
列表
看板
标记
里程碑
合并请求
255
合并请求
255
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
f3ebf87c
编写于
5月 08, 2017
作者:
D
dong zhihong
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
ltr case done.
上级
f5724efa
变更
2
隐藏空白更改
内联
并排
Showing
2 changed file
with
203 addition
and
0 deletion
+203
-0
ltr/lambdaRank.py
ltr/lambdaRank.py
+99
-0
ltr/ranknet.py
ltr/ranknet.py
+104
-0
未找到文件。
ltr/lambdaRank.py
0 → 100644
浏览文件 @
f3ebf87c
import
os
,
sys
import
gzip
import
paddle.v2
as
paddle
import
numpy
as
np
import
functools
#lambdaRank is listwise learning to rank algorithm
def
lambdaRank
(
feature_dim
):
label
=
paddle
.
layer
.
data
(
"label"
,
paddle
.
data_type
.
integer_value_sequence
(
1
))
data
=
paddle
.
layer
.
data
(
"data"
,
paddle
.
data_type
.
dense_vector
(
feature_dim
))
# two hidden layers
hd1
=
paddle
.
layer
.
fc
(
name
=
"/hidden_1"
,
input
=
data
,
size
=
256
,
act
=
paddle
.
activation
.
Tanh
(),
param_attr
=
paddle
.
attr
.
Param
(
initial_std
=
0.01
,
name
=
"hidden_w1"
))
hd2
=
paddle
.
layer
.
fc
(
name
=
"/hidden_2"
,
input
=
hd1
,
size
=
256
,
act
=
paddle
.
activation
.
Tanh
(),
param_attr
=
paddle
.
attr
.
Param
(
initial_std
=
0.01
,
name
=
"hidden_w2"
))
output
=
paddle
.
layer
.
fc
(
name
=
"/output"
,
input
=
hd2
,
size
=
1
,
act
=
paddle
.
activation
.
Linear
(),
param_attr
=
paddle
.
attr
.
Param
(
initial_std
=
0.01
,
name
=
"output"
))
cost
=
paddle
.
layer
.
lambda_cost
(
input
=
output
,
score
=
label
,
NDCG_num
=
10
)
return
cost
,
output
def
train_lambdaRank
(
num_passes
):
fill_default_train
=
functools
.
partial
(
paddle
.
dataset
.
mq2007
.
train
,
format
=
"listwise"
)
fill_default_test
=
functools
.
partial
(
paddle
.
dataset
.
mq2007
.
test
,
format
=
"listwise"
)
train_reader
=
paddle
.
batch
(
paddle
.
reader
.
shuffle
(
fill_default_train
,
buf_size
=
1000
),
batch_size
=
1000
)
test_reader
=
paddle
.
batch
(
paddle
.
reader
.
shuffle
(
fill_default_test
,
buf_size
=
1000
),
batch_size
=
1000
)
# mq2007 feature_dim = 46, dense format
# fc hidden_dim = 128
feature_dim
=
46
cost
,
output
=
lambdaRank
(
feature_dim
)
parameters
=
paddle
.
parameters
.
create
(
cost
)
trainer
=
paddle
.
trainer
.
SGD
(
cost
=
cost
,
parameters
=
parameters
,
update_equation
=
paddle
.
optimizer
.
Adam
(
learning_rate
=
1e-4
)
)
# Define end batch and end pass event handler
def
event_handler
(
event
):
if
isinstance
(
event
,
paddle
.
event
.
EndIteration
):
if
event
.
batch_id
%
100
==
0
:
print
"Pass %d Batch %d Cost %.9f"
%
(
event
.
pass_id
,
event
.
batch_id
,
event
.
cost
)
else
:
sys
.
stdout
.
write
(
"."
)
sys
.
stdout
.
flush
()
if
isinstance
(
event
,
paddle
.
event
.
EndPass
):
result
=
trainer
.
test
(
reader
=
test_reader
,
feeding
=
feeding
)
print
"
\n
Test with Pass %d, %s"
%
(
event
.
pass_id
,
result
.
metrics
)
with
gzip
.
open
(
"lambdaRank_params_%d.tar.gz"
%
(
event
.
pass_id
),
"w"
)
as
f
:
parameters
.
to_tar
(
f
)
feeding
=
{
"label"
:
0
,
"data"
:
1
}
trainer
.
train
(
reader
=
train_reader
,
event_handler
=
event_handler
,
feeding
=
feeding
,
num_passes
=
num_passes
)
def
lambdaRank_infer
(
pass_id
):
print
"Begin to Infer..."
feature_dim
=
46
output
=
lambdaRnak
(
feature_dim
)
parameters
=
paddle
.
parameters
.
Parameters
.
from_tar
(
gzip
.
open
(
"lambdaRank_params_%d.tar.gz"
%
(
pass_id
-
1
)))
infer_data
=
[]
infer_data_num
=
1000
for
label
,
left
,
right
in
paddle
.
dataset
.
mq2007
.
test
():
infer_data
.
append
(
left
)
if
len
(
infer_data
)
==
infer_data_num
:
break
predicitons
=
paddle
.
infer
(
output_layer
=
output
,
parameters
=
parameters
,
input
=
infer_data
)
for
i
,
score
in
enumerate
(
predicitons
):
print
score
if
__name__
==
'__main__'
:
paddle
.
init
(
use_gpu
=
False
,
trainer_count
=
4
)
train_lambdaRank
(
2
)
lambdaRank_infer
(
pass_id
=
2
)
ltr/ranknet.py
0 → 100644
浏览文件 @
f3ebf87c
import
os
,
sys
import
gzip
import
paddle.v2
as
paddle
import
numpy
as
np
# ranknet is the classic pairwise learning to rank algorithm
# http://icml.cc/2015/wp-content/uploads/2015/06/icml_ranking.pdf
def
half_ranknet
(
name_prefix
,
input_dim
):
data
=
paddle
.
layer
.
data
(
name_prefix
+
"/data"
,
paddle
.
data_type
.
dense_vector
(
input_dim
))
# two hidden layers
hd1
=
paddle
.
layer
.
fc
(
name
=
name_prefix
+
"/hidden_1"
,
input
=
data
,
size
=
32
,
act
=
paddle
.
activation
.
Tanh
(),
param_attr
=
paddle
.
attr
.
Param
(
initial_std
=
0.01
,
name
=
"hidden_w1"
))
hd2
=
paddle
.
layer
.
fc
(
name
=
name_prefix
+
"/hidden_2"
,
input
=
hd1
,
size
=
16
,
act
=
paddle
.
activation
.
Tanh
(),
param_attr
=
paddle
.
attr
.
Param
(
initial_std
=
0.01
,
name
=
"hidden_w2"
))
output
=
paddle
.
layer
.
fc
(
name
=
name_prefix
+
"/output"
,
input
=
hd2
,
size
=
1
,
act
=
paddle
.
activation
.
Linear
(),
param_attr
=
paddle
.
attr
.
Param
(
initial_std
=
0.01
,
name
=
"output"
))
return
output
def
ranknet
(
input_dim
):
label
=
paddle
.
layer
.
data
(
"label"
,
paddle
.
data_type
.
integer_value
(
1
))
output_left
=
half_ranknet
(
"left"
,
input_dim
)
output_right
=
half_ranknet
(
"right"
,
input_dim
)
cost
=
paddle
.
layer
.
rank_cost
(
name
=
"cost"
,
left
=
output_left
,
right
=
output_right
,
label
=
label
)
return
cost
def
train_ranknet
(
num_passes
):
train_reader
=
paddle
.
batch
(
paddle
.
reader
.
shuffle
(
paddle
.
dataset
.
mq2007
.
train
,
buf_size
=
1000
),
batch_size
=
1000
)
test_reader
=
paddle
.
batch
(
paddle
.
reader
.
shuffle
(
paddle
.
dataset
.
mq2007
.
test
,
buf_size
=
1000
),
batch_size
=
1000
)
# mq2007 feature_dim = 46, dense format
# fc hidden_dim = 128
feature_dim
=
46
cost
=
ranknet
(
feature_dim
)
parameters
=
paddle
.
parameters
.
create
(
cost
)
trainer
=
paddle
.
trainer
.
SGD
(
cost
=
cost
,
parameters
=
parameters
,
update_equation
=
paddle
.
optimizer
.
Adam
(
learning_rate
=
2e-4
)
)
# Define end batch and end pass event handler
def
event_handler
(
event
):
if
isinstance
(
event
,
paddle
.
event
.
EndIteration
):
if
event
.
batch_id
%
100
==
0
:
print
"Pass %d Batch %d Cost %.9f"
%
(
event
.
pass_id
,
event
.
batch_id
,
event
.
cost
)
else
:
sys
.
stdout
.
write
(
"."
)
sys
.
stdout
.
flush
()
if
isinstance
(
event
,
paddle
.
event
.
EndPass
):
result
=
trainer
.
test
(
reader
=
test_reader
,
feeding
=
feeding
)
print
"
\n
Test with Pass %d, %s"
%
(
event
.
pass_id
,
result
.
metrics
)
with
gzip
.
open
(
"ranknet_params_%d.tar.gz"
%
(
event
.
pass_id
),
"w"
)
as
f
:
parameters
.
to_tar
(
f
)
feeding
=
{
"label"
:
0
,
"left/data"
:
1
,
"right/data"
:
2
}
trainer
.
train
(
reader
=
train_reader
,
event_handler
=
event_handler
,
feeding
=
feeding
,
num_passes
=
num_passes
)
def
ranknet_infer
(
pass_id
):
print
"Begin to Infer..."
feature_dim
=
46
output
=
half_ranknet
(
"left"
,
feature_dim
)
parameters
=
paddle
.
parameters
.
Parameters
.
from_tar
(
gzip
.
open
(
"ranknet_params_%d.tar.gz"
%
(
pass_id
-
1
)))
infer_data
=
[]
infer_data_num
=
1000
for
label
,
left
,
right
in
paddle
.
dataset
.
mq2007
.
test
():
infer_data
.
append
(
left
)
if
len
(
infer_data
)
==
infer_data_num
:
break
predicitons
=
paddle
.
infer
(
output_layer
=
output
,
parameters
=
parameters
,
input
=
infer_data
)
for
i
,
score
in
enumerate
(
predicitons
):
print
score
if
__name__
==
'__main__'
:
paddle
.
init
(
use_gpu
=
False
,
trainer_count
=
4
)
train_ranknet
(
2
)
ranknet_infer
(
pass_id
=
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