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
Paddle
提交
4a94f8a4
P
Paddle
项目概览
PaddlePaddle
/
Paddle
1 年多 前同步成功
通知
2302
Star
20931
Fork
5422
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1423
列表
看板
标记
里程碑
合并请求
543
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
1,423
Issue
1,423
列表
看板
标记
里程碑
合并请求
543
合并请求
543
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
4a94f8a4
编写于
3月 02, 2017
作者:
D
dangqingqing
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
refine api training
上级
bca1fce6
变更
4
显示空白变更内容
内联
并排
Showing
4 changed file
with
147 addition
and
203 deletion
+147
-203
demo/semantic_role_labeling/api_train_v2.py
demo/semantic_role_labeling/api_train_v2.py
+143
-78
demo/semantic_role_labeling/model_v2.py
demo/semantic_role_labeling/model_v2.py
+0
-121
python/paddle/v2/dataset/__init__.py
python/paddle/v2/dataset/__init__.py
+1
-1
python/paddle/v2/dataset/conll05.py
python/paddle/v2/dataset/conll05.py
+3
-3
未找到文件。
demo/semantic_role_labeling/api_train_v2.py
浏览文件 @
4a94f8a4
import
sys
import
math
import
numpy
as
np
import
numpy
as
np
import
paddle.v2
as
paddle
import
paddle.v2
as
paddle
from
model_v2
import
db_lstm
import
paddle.v2.dataset.conll05
as
conll05
UNK_IDX
=
0
UNK_IDX
=
0
word_dict_file
=
'./data/wordDict.txt'
label_dict_file
=
'./data/targetDict.txt'
predicate_file
=
'./data/verbDict.txt'
word_dict
=
dict
()
def
db_lstm
():
label_dict
=
dict
()
word_dict
,
verb_dict
,
label_dict
=
conll05
.
get_dict
()
predicate_dict
=
dict
()
word_dict_len
=
len
(
word_dict
)
label_dict_len
=
len
(
label_dict
)
with
open
(
word_dict_file
,
'r'
)
as
f_word
,
\
pred_len
=
len
(
verb_dict
)
open
(
label_dict_file
,
'r'
)
as
f_label
,
\
print
'word_dict_len,'
,
word_dict_len
open
(
predicate_file
,
'r'
)
as
f_pre
:
print
'label_dict_len,'
,
label_dict_len
for
i
,
line
in
enumerate
(
f_word
):
print
'pred_len,'
,
pred_len
w
=
line
.
strip
()
word_dict
[
w
]
=
i
mark_dict_len
=
2
word_dim
=
32
for
i
,
line
in
enumerate
(
f_label
):
mark_dim
=
5
w
=
line
.
strip
()
hidden_dim
=
512
label_dict
[
w
]
=
i
depth
=
8
for
i
,
line
in
enumerate
(
f_pre
):
#8 features
w
=
line
.
strip
()
def
d_type
(
size
):
predicate_dict
[
w
]
=
i
return
paddle
.
data_type
.
integer_value_sequence
(
size
)
word_dict_len
=
len
(
word_dict
)
word
=
paddle
.
layer
.
data
(
name
=
'word_data'
,
type
=
d_type
(
word_dict_len
))
label_dict_len
=
len
(
label_dict
)
predicate
=
paddle
.
layer
.
data
(
name
=
'verb_data'
,
type
=
d_type
(
pred_len
))
pred_len
=
len
(
predicate_dict
)
ctx_n2
=
paddle
.
layer
.
data
(
name
=
'ctx_n2_data'
,
type
=
d_type
(
word_dict_len
))
ctx_n1
=
paddle
.
layer
.
data
(
name
=
'ctx_n1_data'
,
type
=
d_type
(
word_dict_len
))
def
train_reader
(
file_name
=
"data/feature"
):
ctx_0
=
paddle
.
layer
.
data
(
name
=
'ctx_0_data'
,
type
=
d_type
(
word_dict_len
))
def
reader
():
ctx_p1
=
paddle
.
layer
.
data
(
name
=
'ctx_p1_data'
,
type
=
d_type
(
word_dict_len
))
with
open
(
file_name
,
'r'
)
as
fdata
:
ctx_p2
=
paddle
.
layer
.
data
(
name
=
'ctx_p2_data'
,
type
=
d_type
(
word_dict_len
))
for
line
in
fdata
:
mark
=
paddle
.
layer
.
data
(
name
=
'mark_data'
,
type
=
d_type
(
mark_dict_len
))
sentence
,
predicate
,
ctx_n2
,
ctx_n1
,
ctx_0
,
ctx_p1
,
ctx_p2
,
mark
,
label
=
\
line
.
strip
().
split
(
'
\t
'
)
target
=
paddle
.
layer
.
data
(
name
=
'target'
,
type
=
d_type
(
label_dict_len
))
words
=
sentence
.
split
()
default_std
=
1
/
math
.
sqrt
(
hidden_dim
)
/
3.0
sen_len
=
len
(
words
)
word_slot
=
[
word_dict
.
get
(
w
,
UNK_IDX
)
for
w
in
words
]
emb_para
=
paddle
.
attr
.
Param
(
name
=
'emb'
,
initial_std
=
0.
,
learning_rate
=
0.
)
std_0
=
paddle
.
attr
.
Param
(
initial_std
=
0.
)
predicate_slot
=
[
predicate_dict
.
get
(
predicate
)]
*
sen_len
std_default
=
paddle
.
attr
.
Param
(
initial_std
=
default_std
)
ctx_n2_slot
=
[
word_dict
.
get
(
ctx_n2
,
UNK_IDX
)]
*
sen_len
ctx_n1_slot
=
[
word_dict
.
get
(
ctx_n1
,
UNK_IDX
)]
*
sen_len
predicate_embedding
=
paddle
.
layer
.
embedding
(
ctx_0_slot
=
[
word_dict
.
get
(
ctx_0
,
UNK_IDX
)]
*
sen_len
size
=
word_dim
,
ctx_p1_slot
=
[
word_dict
.
get
(
ctx_p1
,
UNK_IDX
)]
*
sen_len
input
=
predicate
,
ctx_p2_slot
=
[
word_dict
.
get
(
ctx_p2
,
UNK_IDX
)]
*
sen_len
param_attr
=
paddle
.
attr
.
Param
(
name
=
'vemb'
,
initial_std
=
default_std
))
marks
=
mark
.
split
()
mark_embedding
=
paddle
.
layer
.
embedding
(
mark_slot
=
[
int
(
w
)
for
w
in
marks
]
size
=
mark_dim
,
input
=
mark
,
param_attr
=
std_0
)
label_list
=
label
.
split
()
word_input
=
[
word
,
ctx_n2
,
ctx_n1
,
ctx_0
,
ctx_p1
,
ctx_p2
]
label_slot
=
[
label_dict
.
get
(
w
)
for
w
in
label_list
]
emb_layers
=
[
yield
word_slot
,
ctx_n2_slot
,
ctx_n1_slot
,
\
paddle
.
layer
.
embedding
(
ctx_0_slot
,
ctx_p1_slot
,
ctx_p2_slot
,
predicate_slot
,
mark_slot
,
label_slot
size
=
word_dim
,
input
=
x
,
param_attr
=
emb_para
)
for
x
in
word_input
]
return
reader
emb_layers
.
append
(
predicate_embedding
)
emb_layers
.
append
(
mark_embedding
)
hidden_0
=
paddle
.
layer
.
mixed
(
size
=
hidden_dim
,
bias_attr
=
std_default
,
input
=
[
paddle
.
layer
.
full_matrix_projection
(
input
=
emb
,
param_attr
=
std_default
)
for
emb
in
emb_layers
])
mix_hidden_lr
=
1e-3
lstm_para_attr
=
paddle
.
attr
.
Param
(
initial_std
=
0.0
,
learning_rate
=
1.0
)
hidden_para_attr
=
paddle
.
attr
.
Param
(
initial_std
=
default_std
,
learning_rate
=
mix_hidden_lr
)
lstm_0
=
paddle
.
layer
.
lstmemory
(
input
=
hidden_0
,
act
=
paddle
.
activation
.
Relu
(),
gate_act
=
paddle
.
activation
.
Sigmoid
(),
state_act
=
paddle
.
activation
.
Sigmoid
(),
bias_attr
=
std_0
,
param_attr
=
lstm_para_attr
)
#stack L-LSTM and R-LSTM with direct edges
input_tmp
=
[
hidden_0
,
lstm_0
]
for
i
in
range
(
1
,
depth
):
mix_hidden
=
paddle
.
layer
.
mixed
(
size
=
hidden_dim
,
bias_attr
=
std_default
,
input
=
[
paddle
.
layer
.
full_matrix_projection
(
input
=
input_tmp
[
0
],
param_attr
=
hidden_para_attr
),
paddle
.
layer
.
full_matrix_projection
(
input
=
input_tmp
[
1
],
param_attr
=
lstm_para_attr
)
])
lstm
=
paddle
.
layer
.
lstmemory
(
input
=
mix_hidden
,
act
=
paddle
.
activation
.
Relu
(),
gate_act
=
paddle
.
activation
.
Sigmoid
(),
state_act
=
paddle
.
activation
.
Sigmoid
(),
reverse
=
((
i
%
2
)
==
1
),
bias_attr
=
std_0
,
param_attr
=
lstm_para_attr
)
input_tmp
=
[
mix_hidden
,
lstm
]
feature_out
=
paddle
.
layer
.
mixed
(
size
=
label_dict_len
,
bias_attr
=
std_default
,
input
=
[
paddle
.
layer
.
full_matrix_projection
(
input
=
input_tmp
[
0
],
param_attr
=
hidden_para_attr
),
paddle
.
layer
.
full_matrix_projection
(
input
=
input_tmp
[
1
],
param_attr
=
lstm_para_attr
)
],
)
crf_cost
=
paddle
.
layer
.
crf
(
size
=
label_dict_len
,
input
=
feature_out
,
label
=
target
,
param_attr
=
paddle
.
attr
.
Param
(
name
=
'crfw'
,
initial_std
=
default_std
,
learning_rate
=
mix_hidden_lr
))
crf_dec
=
paddle
.
layer
.
crf_decoding
(
name
=
'crf_dec_l'
,
size
=
label_dict_len
,
input
=
feature_out
,
label
=
target
,
param_attr
=
paddle
.
attr
.
Param
(
name
=
'crfw'
))
return
crf_cost
,
crf_dec
def
load_parameter
(
file_name
,
h
,
w
):
def
load_parameter
(
file_name
,
h
,
w
):
with
open
(
file_name
,
'rb'
)
as
f
:
with
open
(
file_name
,
'rb'
)
as
f
:
f
.
read
(
16
)
# skip header
for float type
.
f
.
read
(
16
)
# skip header.
return
np
.
fromfile
(
f
,
dtype
=
np
.
float32
).
reshape
(
h
,
w
)
return
np
.
fromfile
(
f
,
dtype
=
np
.
float32
).
reshape
(
h
,
w
)
...
@@ -71,44 +144,36 @@ def main():
...
@@ -71,44 +144,36 @@ def main():
paddle
.
init
(
use_gpu
=
False
,
trainer_count
=
1
)
paddle
.
init
(
use_gpu
=
False
,
trainer_count
=
1
)
# define network topology
# define network topology
crf_cost
,
crf_dec
=
db_lstm
(
word_dict_len
,
label_dict_len
,
pred_len
)
crf_cost
,
crf_dec
=
db_lstm
()
# create parameters
parameters
=
paddle
.
parameters
.
create
([
crf_cost
,
crf_dec
])
parameters
=
paddle
.
parameters
.
create
([
crf_cost
,
crf_dec
])
optimizer
=
paddle
.
optimizer
.
Momentum
(
momentum
=
0.01
,
learning_rate
=
2e-2
)
# create optimizer
optimizer
=
paddle
.
optimizer
.
Momentum
(
momentum
=
0
,
learning_rate
=
2e-2
,
regularization
=
paddle
.
optimizer
.
L2Regularization
(
rate
=
8e-4
),
model_average
=
paddle
.
optimizer
.
ModelAverage
(
average_window
=
0.5
,
max_average_window
=
10000
),
)
def
event_handler
(
event
):
def
event_handler
(
event
):
if
isinstance
(
event
,
paddle
.
event
.
EndIteration
):
if
isinstance
(
event
,
paddle
.
event
.
EndIteration
):
if
event
.
batch_id
%
100
==
0
:
if
event
.
batch_id
%
100
==
0
:
print
"Pass %d, Batch %d, Cost %f, %s"
%
(
print
"Pass %d, Batch %d, Cost %f, %s"
%
(
event
.
pass_id
,
event
.
batch_id
,
event
.
cost
,
event
.
metrics
)
event
.
pass_id
,
event
.
batch_id
,
event
.
cost
,
event
.
metrics
)
else
:
pass
trainer
=
paddle
.
trainer
.
SGD
(
cost
=
crf_cost
,
trainer
=
paddle
.
trainer
.
SGD
(
cost
=
crf_cost
,
parameters
=
parameters
,
parameters
=
parameters
,
update_equation
=
optimizer
)
update_equation
=
optimizer
)
parameters
.
set
(
'emb'
,
load_parameter
(
conll05
.
get_embedding
(),
44068
,
32
))
parameters
.
set
(
'emb'
,
load_parameter
(
"data/emb"
,
44068
,
32
))
reader_dict
=
{
'word_data'
:
0
,
'ctx_n2_data'
:
1
,
'ctx_n1_data'
:
2
,
'ctx_0_data'
:
3
,
'ctx_p1_data'
:
4
,
'ctx_p2_data'
:
5
,
'verb_data'
:
6
,
'mark_data'
:
7
,
'target'
:
8
,
}
trn_reader
=
paddle
.
reader
.
batched
(
trn_reader
=
paddle
.
reader
.
batched
(
paddle
.
reader
.
shuffle
(
paddle
.
reader
.
shuffle
(
train_reader
(),
buf_size
=
8192
),
batch_size
=
10
)
conll05
.
test
,
buf_size
=
8192
),
batch_size
=
10
)
trainer
.
train
(
trainer
.
train
(
reader
=
trn_reader
,
reader
=
trn_reader
,
event_handler
=
event_handler
,
num_passes
=
10000
)
event_handler
=
event_handler
,
num_passes
=
10000
,
reader_dict
=
reader_dict
)
if
__name__
==
'__main__'
:
if
__name__
==
'__main__'
:
...
...
demo/semantic_role_labeling/model_v2.py
已删除
100644 → 0
浏览文件 @
bca1fce6
import
math
import
paddle.v2
as
paddle
def
db_lstm
(
word_dict_len
,
label_dict_len
,
pred_len
):
mark_dict_len
=
2
word_dim
=
32
mark_dim
=
5
hidden_dim
=
512
depth
=
8
#8 features
def
d_type
(
size
):
return
paddle
.
data_type
.
integer_value_sequence
(
size
)
word
=
paddle
.
layer
.
data
(
name
=
'word_data'
,
type
=
d_type
(
word_dict_len
))
predicate
=
paddle
.
layer
.
data
(
name
=
'verb_data'
,
type
=
d_type
(
pred_len
))
ctx_n2
=
paddle
.
layer
.
data
(
name
=
'ctx_n2_data'
,
type
=
d_type
(
word_dict_len
))
ctx_n1
=
paddle
.
layer
.
data
(
name
=
'ctx_n1_data'
,
type
=
d_type
(
word_dict_len
))
ctx_0
=
paddle
.
layer
.
data
(
name
=
'ctx_0_data'
,
type
=
d_type
(
word_dict_len
))
ctx_p1
=
paddle
.
layer
.
data
(
name
=
'ctx_p1_data'
,
type
=
d_type
(
word_dict_len
))
ctx_p2
=
paddle
.
layer
.
data
(
name
=
'ctx_p2_data'
,
type
=
d_type
(
word_dict_len
))
mark
=
paddle
.
layer
.
data
(
name
=
'mark_data'
,
type
=
d_type
(
mark_dict_len
))
target
=
paddle
.
layer
.
data
(
name
=
'target'
,
type
=
d_type
(
label_dict_len
))
default_std
=
1
/
math
.
sqrt
(
hidden_dim
)
/
3.0
emb_para
=
paddle
.
attr
.
Param
(
name
=
'emb'
,
initial_std
=
0.
,
learning_rate
=
0.
)
std_0
=
paddle
.
attr
.
Param
(
initial_std
=
0.
)
std_default
=
paddle
.
attr
.
Param
(
initial_std
=
default_std
)
predicate_embedding
=
paddle
.
layer
.
embedding
(
size
=
word_dim
,
input
=
predicate
,
param_attr
=
paddle
.
attr
.
Param
(
name
=
'vemb'
,
initial_std
=
default_std
))
mark_embedding
=
paddle
.
layer
.
embedding
(
size
=
mark_dim
,
input
=
mark
,
param_attr
=
std_0
)
word_input
=
[
word
,
ctx_n2
,
ctx_n1
,
ctx_0
,
ctx_p1
,
ctx_p2
]
emb_layers
=
[
paddle
.
layer
.
embedding
(
size
=
word_dim
,
input
=
x
,
param_attr
=
emb_para
)
for
x
in
word_input
]
emb_layers
.
append
(
predicate_embedding
)
emb_layers
.
append
(
mark_embedding
)
hidden_0
=
paddle
.
layer
.
mixed
(
size
=
hidden_dim
,
bias_attr
=
std_default
,
input
=
[
paddle
.
layer
.
full_matrix_projection
(
input
=
emb
,
param_attr
=
std_default
)
for
emb
in
emb_layers
])
mix_hidden_lr
=
1e-3
lstm_para_attr
=
paddle
.
attr
.
Param
(
initial_std
=
0.0
,
learning_rate
=
1.0
)
hidden_para_attr
=
paddle
.
attr
.
Param
(
initial_std
=
default_std
,
learning_rate
=
mix_hidden_lr
)
lstm_0
=
paddle
.
layer
.
lstmemory
(
input
=
hidden_0
,
act
=
paddle
.
activation
.
Relu
(),
gate_act
=
paddle
.
activation
.
Sigmoid
(),
state_act
=
paddle
.
activation
.
Sigmoid
(),
bias_attr
=
std_0
,
param_attr
=
lstm_para_attr
)
#stack L-LSTM and R-LSTM with direct edges
input_tmp
=
[
hidden_0
,
lstm_0
]
for
i
in
range
(
1
,
depth
):
mix_hidden
=
paddle
.
layer
.
mixed
(
size
=
hidden_dim
,
bias_attr
=
std_default
,
input
=
[
paddle
.
layer
.
full_matrix_projection
(
input
=
input_tmp
[
0
],
param_attr
=
hidden_para_attr
),
paddle
.
layer
.
full_matrix_projection
(
input
=
input_tmp
[
1
],
param_attr
=
lstm_para_attr
)
])
lstm
=
paddle
.
layer
.
lstmemory
(
input
=
mix_hidden
,
act
=
paddle
.
activation
.
Relu
(),
gate_act
=
paddle
.
activation
.
Sigmoid
(),
state_act
=
paddle
.
activation
.
Sigmoid
(),
reverse
=
((
i
%
2
)
==
1
),
bias_attr
=
std_0
,
param_attr
=
lstm_para_attr
)
input_tmp
=
[
mix_hidden
,
lstm
]
feature_out
=
paddle
.
layer
.
mixed
(
size
=
label_dict_len
,
bias_attr
=
std_default
,
input
=
[
paddle
.
layer
.
full_matrix_projection
(
input
=
input_tmp
[
0
],
param_attr
=
hidden_para_attr
),
paddle
.
layer
.
full_matrix_projection
(
input
=
input_tmp
[
1
],
param_attr
=
lstm_para_attr
)
],
)
crf_cost
=
paddle
.
layer
.
crf
(
size
=
label_dict_len
,
input
=
feature_out
,
label
=
target
,
param_attr
=
paddle
.
attr
.
Param
(
name
=
'crfw'
,
initial_std
=
default_std
,
learning_rate
=
mix_hidden_lr
))
crf_dec
=
paddle
.
layer
.
crf_decoding
(
name
=
'crf_dec_l'
,
size
=
label_dict_len
,
input
=
feature_out
,
label
=
target
,
param_attr
=
paddle
.
attr
.
Param
(
name
=
'crfw'
))
return
crf_cost
,
crf_dec
python/paddle/v2/dataset/__init__.py
浏览文件 @
4a94f8a4
...
@@ -14,4 +14,4 @@
...
@@ -14,4 +14,4 @@
import
mnist
import
mnist
__all__
=
[
'mnist'
]
__all__
=
[
'mnist'
,
'cifar'
,
'imdb'
,
'conll05'
,
'imikolov'
,
'movielens'
]
python/paddle/v2/dataset/conll05.py
浏览文件 @
4a94f8a4
...
@@ -160,7 +160,6 @@ def reader_creator(corpus_reader,
...
@@ -160,7 +160,6 @@ def reader_creator(corpus_reader,
ctx_p2
=
'eos'
ctx_p2
=
'eos'
word_idx
=
[
word_dict
.
get
(
w
,
UNK_IDX
)
for
w
in
sentence
]
word_idx
=
[
word_dict
.
get
(
w
,
UNK_IDX
)
for
w
in
sentence
]
pred_idx
=
[
predicate_dict
.
get
(
predicate
)]
*
sen_len
ctx_n2_idx
=
[
word_dict
.
get
(
ctx_n2
,
UNK_IDX
)]
*
sen_len
ctx_n2_idx
=
[
word_dict
.
get
(
ctx_n2
,
UNK_IDX
)]
*
sen_len
ctx_n1_idx
=
[
word_dict
.
get
(
ctx_n1
,
UNK_IDX
)]
*
sen_len
ctx_n1_idx
=
[
word_dict
.
get
(
ctx_n1
,
UNK_IDX
)]
*
sen_len
...
@@ -168,10 +167,11 @@ def reader_creator(corpus_reader,
...
@@ -168,10 +167,11 @@ def reader_creator(corpus_reader,
ctx_p1_idx
=
[
word_dict
.
get
(
ctx_p1
,
UNK_IDX
)]
*
sen_len
ctx_p1_idx
=
[
word_dict
.
get
(
ctx_p1
,
UNK_IDX
)]
*
sen_len
ctx_p2_idx
=
[
word_dict
.
get
(
ctx_p2
,
UNK_IDX
)]
*
sen_len
ctx_p2_idx
=
[
word_dict
.
get
(
ctx_p2
,
UNK_IDX
)]
*
sen_len
pred_idx
=
[
predicate_dict
.
get
(
predicate
)]
*
sen_len
label_idx
=
[
label_dict
.
get
(
w
)
for
w
in
labels
]
label_idx
=
[
label_dict
.
get
(
w
)
for
w
in
labels
]
yield
word_idx
,
pred_idx
,
ctx_n2_idx
,
ctx_n1_idx
,
\
yield
word_idx
,
ctx_n2_idx
,
ctx_n1_idx
,
\
ctx_0_idx
,
ctx_p1_idx
,
ctx_p2_idx
,
mark
,
label_idx
ctx_0_idx
,
ctx_p1_idx
,
ctx_p2_idx
,
pred_idx
,
mark
,
label_idx
return
reader
()
return
reader
()
...
...
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