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
2d392828
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
体验新版 GitCode,发现更多精彩内容 >>
提交
2d392828
编写于
4月 27, 2018
作者:
Y
Yao Cheng
提交者:
chengyao
4月 27, 2018
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
add cloud
上级
8b873b29
变更
2
隐藏空白更改
内联
并排
Showing
2 changed file
with
1003 addition
and
0 deletion
+1003
-0
fluid/text_classification/clouds/scdb_parallel_executor.py
fluid/text_classification/clouds/scdb_parallel_executor.py
+509
-0
fluid/text_classification/clouds/scdb_single_card.py
fluid/text_classification/clouds/scdb_single_card.py
+494
-0
未找到文件。
fluid/text_classification/clouds/scdb_parallel_executor.py
0 → 100644
浏览文件 @
2d392828
import
unittest
import
contextlib
import
paddle.fluid
as
fluid
import
paddle.v2
as
paddle
import
numpy
as
np
import
sys
import
time
import
os
import
json
import
random
def
to_lodtensor
(
data
,
place
):
"""
convert to LODtensor
"""
seq_lens
=
[
len
(
seq
)
for
seq
in
data
]
cur_len
=
0
lod
=
[
cur_len
]
for
l
in
seq_lens
:
cur_len
+=
l
lod
.
append
(
cur_len
)
flattened_data
=
np
.
concatenate
(
data
,
axis
=
0
).
astype
(
"int64"
)
flattened_data
=
flattened_data
.
reshape
([
len
(
flattened_data
),
1
])
res
=
fluid
.
LoDTensor
()
res
.
set
(
flattened_data
,
place
)
res
.
set_lod
([
lod
])
return
res
def
load_vocab
(
filename
):
"""
load imdb vocabulary
"""
vocab
=
{}
with
open
(
filename
)
as
f
:
wid
=
0
for
line
in
f
:
vocab
[
line
.
strip
()]
=
wid
wid
+=
1
vocab
[
"<unk>"
]
=
len
(
vocab
)
return
vocab
def
data2tensor
(
data
,
place
):
"""
data2tensor
"""
input_seq
=
to_lodtensor
(
map
(
lambda
x
:
x
[
0
],
data
),
place
)
y_data
=
np
.
array
(
map
(
lambda
x
:
x
[
1
],
data
)).
astype
(
"int64"
)
y_data
=
y_data
.
reshape
([
-
1
,
1
])
return
{
"words"
:
input_seq
,
"label"
:
y_data
}
def
data2pred
(
data
,
place
):
"""
data2tensor
"""
input_seq
=
to_lodtensor
(
map
(
lambda
x
:
x
[
0
],
data
),
place
)
y_data
=
np
.
array
(
map
(
lambda
x
:
x
[
1
],
data
)).
astype
(
"int64"
)
y_data
=
y_data
.
reshape
([
-
1
,
1
])
return
{
"words"
:
input_seq
}
def
load_dict
(
vocab
):
"""
Load dict from vocab
"""
word_dict
=
dict
()
with
open
(
vocab
,
"r"
)
as
fin
:
for
line
in
fin
:
cols
=
line
.
strip
(
"
\r\n
"
).
decode
(
"gb18030"
).
split
(
"
\t
"
)
word_dict
[
cols
[
0
]]
=
int
(
cols
[
1
])
return
word_dict
def
save_dict
(
word_dict
,
vocab
):
"""
Save dict into file
"""
with
open
(
vocab
,
"w"
)
as
fout
:
for
k
,
v
in
word_dict
.
iteritems
():
outstr
=
(
"%s
\t
%s
\n
"
%
(
k
,
v
)).
encode
(
"gb18030"
)
fout
.
write
(
outstr
)
def
build_dict
(
fname
):
"""
build word dict using trainset
"""
word_dict
=
dict
()
with
open
(
fname
,
"r"
)
as
fin
:
for
line
in
fin
:
try
:
words
=
line
.
strip
(
"
\r\n
"
).
decode
(
"gb18030"
).
split
(
"
\t
"
)[
1
].
split
(
" "
)
except
:
sys
.
stderr
.
write
(
"[warning] build_dict: decode error
\n
"
)
continue
for
w
in
words
:
if
w
not
in
word_dict
:
word_dict
[
w
]
=
len
(
word_dict
)
return
word_dict
def
scdb_word_dict
(
vocab
=
"scdb_data/train_set/train.vocab"
):
"""
get word_dict
"""
if
not
os
.
path
.
exists
(
vocab
):
w_dict
=
build_dict
(
train_file
)
save_dict
(
w_dict
,
vocab
)
else
:
w_dict
=
load_dict
(
vocab
)
w_dict
[
"<unk>"
]
=
len
(
w_dict
)
return
w_dict
def
data_reader
(
fname
,
word_dict
,
is_dir
=
False
):
"""
Convert word sequence into slot
"""
unk_id
=
len
(
word_dict
)
all_data
=
[]
filelist
=
[]
if
is_dir
:
filelist
=
[
fname
+
os
.
sep
+
f
for
f
in
os
.
listdir
(
fname
)]
else
:
filelist
=
[
fname
]
for
each_name
in
filelist
:
with
open
(
each_name
,
"r"
)
as
fin
:
for
line
in
fin
:
try
:
cols
=
line
.
strip
(
"
\r\n
"
).
decode
(
"gb18030"
).
split
(
"
\t
"
)
except
:
sys
.
stderr
.
write
(
"warning: ignore decode error
\n
"
)
continue
label
=
int
(
cols
[
0
])
wids
=
[
word_dict
[
x
]
if
x
in
word_dict
else
unk_id
for
x
in
cols
[
1
].
split
(
" "
)]
all_data
.
append
((
wids
,
label
))
random
.
shuffle
(
all_data
)
def
reader
():
for
doc
,
label
in
all_data
:
yield
doc
,
label
return
reader
def
scdb_train_data
(
train_dir
=
"scdb_data/train_set/corpus.train.seg"
,
w_dict
=
None
):
"""
create train data
"""
return
data_reader
(
train_dir
,
w_dict
,
True
)
def
scdb_test_data
(
test_file
,
w_dict
):
"""
test_set=["car", "lbs", "spot", "weibo",
"baby", "toutiao", "3c", "movie", "haogan"]
"""
return
data_reader
(
test_file
,
w_dict
)
def
bow_net
(
data
,
label
,
dict_dim
,
emb_dim
=
128
,
hid_dim
=
128
,
hid_dim2
=
96
,
class_dim
=
2
):
"""
bow net
"""
emb
=
fluid
.
layers
.
embedding
(
input
=
data
,
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
)
acc
=
fluid
.
layers
.
accuracy
(
input
=
prediction
,
label
=
label
)
return
avg_cost
,
acc
,
prediction
def
cnn_net
(
data
,
label
,
dict_dim
,
emb_dim
=
128
,
hid_dim
=
128
,
hid_dim2
=
96
,
class_dim
=
2
,
win_size
=
3
):
"""
conv net
"""
emb
=
fluid
.
layers
.
embedding
(
input
=
data
,
size
=
[
dict_dim
,
emb_dim
])
conv_3
=
fluid
.
nets
.
sequence_conv_pool
(
input
=
emb
,
num_filters
=
hid_dim
,
filter_size
=
win_size
,
act
=
"tanh"
,
pool_type
=
"max"
)
fc_1
=
fluid
.
layers
.
fc
(
input
=
[
conv_3
],
size
=
hid_dim2
)
prediction
=
fluid
.
layers
.
fc
(
input
=
[
fc_1
],
size
=
class_dim
,
act
=
"softmax"
)
cost
=
fluid
.
layers
.
cross_entropy
(
input
=
prediction
,
label
=
label
)
avg_cost
=
fluid
.
layers
.
mean
(
x
=
cost
)
acc
=
fluid
.
layers
.
accuracy
(
input
=
prediction
,
label
=
label
)
return
avg_cost
,
acc
,
prediction
def
lstm_net
(
data
,
label
,
dict_dim
,
emb_dim
=
128
,
hid_dim
=
128
,
hid_dim2
=
96
,
class_dim
=
2
,
emb_lr
=
30.0
):
"""
lstm net
"""
emb
=
fluid
.
layers
.
embedding
(
input
=
data
,
size
=
[
dict_dim
,
emb_dim
],
param_attr
=
fluid
.
ParamAttr
(
learning_rate
=
emb_lr
))
fc0
=
fluid
.
layers
.
fc
(
input
=
emb
,
size
=
hid_dim
*
4
,
act
=
'tanh'
)
lstm_h
,
c
=
fluid
.
layers
.
dynamic_lstm
(
input
=
fc0
,
size
=
hid_dim
*
4
,
is_reverse
=
False
)
lstm_max
=
fluid
.
layers
.
sequence_pool
(
input
=
lstm_h
,
pool_type
=
'max'
)
lstm_max_tanh
=
fluid
.
layers
.
tanh
(
lstm_max
)
fc1
=
fluid
.
layers
.
fc
(
input
=
lstm_max_tanh
,
size
=
hid_dim2
,
act
=
'tanh'
)
prediction
=
fluid
.
layers
.
fc
(
input
=
fc1
,
size
=
class_dim
,
act
=
'softmax'
)
cost
=
fluid
.
layers
.
cross_entropy
(
input
=
prediction
,
label
=
label
)
avg_cost
=
fluid
.
layers
.
mean
(
x
=
cost
)
acc
=
fluid
.
layers
.
accuracy
(
input
=
prediction
,
label
=
label
)
return
avg_cost
,
acc
,
prediction
def
bilstm_net
(
data
,
label
,
dict_dim
,
emb_dim
=
128
,
hid_dim
=
128
,
hid_dim2
=
96
,
class_dim
=
2
,
emb_lr
=
30.0
):
"""
lstm net
"""
emb
=
fluid
.
layers
.
embedding
(
input
=
data
,
size
=
[
dict_dim
,
emb_dim
],
param_attr
=
fluid
.
ParamAttr
(
learning_rate
=
emb_lr
))
fc0
=
fluid
.
layers
.
fc
(
input
=
emb
,
size
=
hid_dim
*
4
,
act
=
'tanh'
)
rfc0
=
fluid
.
layers
.
fc
(
input
=
emb
,
size
=
hid_dim
*
4
,
act
=
'tanh'
)
lstm_h
,
c
=
fluid
.
layers
.
dynamic_lstm
(
input
=
fc0
,
size
=
hid_dim
*
4
,
is_reverse
=
False
)
rlstm_h
,
c
=
fluid
.
layers
.
dynamic_lstm
(
input
=
rfc0
,
size
=
hid_dim
*
4
,
is_reverse
=
True
)
lstm_last
=
fluid
.
layers
.
sequence_last_step
(
input
=
lstm_h
)
rlstm_last
=
fluid
.
layers
.
sequence_last_step
(
input
=
rlstm_h
)
lstm_last_tanh
=
fluid
.
layers
.
tanh
(
lstm_last
)
rlstm_last_tanh
=
fluid
.
layers
.
tanh
(
rlstm_last
)
lstm_concat
=
fluid
.
layers
.
concat
(
input
=
[
lstm_last
,
rlstm_last
],
axis
=
1
)
fc1
=
fluid
.
layers
.
fc
(
input
=
lstm_concat
,
size
=
hid_dim2
,
act
=
'tanh'
)
prediction
=
fluid
.
layers
.
fc
(
input
=
fc1
,
size
=
class_dim
,
act
=
'softmax'
)
cost
=
fluid
.
layers
.
cross_entropy
(
input
=
prediction
,
label
=
label
)
avg_cost
=
fluid
.
layers
.
mean
(
x
=
cost
)
acc
=
fluid
.
layers
.
accuracy
(
input
=
prediction
,
label
=
label
)
return
avg_cost
,
acc
,
prediction
def
gru_net
(
data
,
label
,
dict_dim
,
emb_dim
=
128
,
hid_dim
=
128
,
hid_dim2
=
96
,
class_dim
=
2
,
emb_lr
=
30.0
):
"""
gru net
"""
emb
=
fluid
.
layers
.
embedding
(
input
=
data
,
size
=
[
dict_dim
,
emb_dim
],
param_attr
=
fluid
.
ParamAttr
(
learning_rate
=
emb_lr
))
fc0
=
fluid
.
layers
.
fc
(
input
=
emb
,
size
=
hid_dim
*
3
)
gru_h
=
fluid
.
layers
.
dynamic_gru
(
input
=
fc0
,
size
=
hid_dim
,
is_reverse
=
False
)
gru_max
=
fluid
.
layers
.
sequence_pool
(
input
=
gru_h
,
pool_type
=
'max'
)
gru_max_tanh
=
fluid
.
layers
.
tanh
(
gru_max
)
fc1
=
fluid
.
layers
.
fc
(
input
=
gru_max_tanh
,
size
=
hid_dim2
,
act
=
'tanh'
)
prediction
=
fluid
.
layers
.
fc
(
input
=
fc1
,
size
=
class_dim
,
act
=
'softmax'
)
cost
=
fluid
.
layers
.
cross_entropy
(
input
=
prediction
,
label
=
label
)
avg_cost
=
fluid
.
layers
.
mean
(
x
=
cost
)
acc
=
fluid
.
layers
.
accuracy
(
input
=
prediction
,
label
=
label
)
return
avg_cost
,
acc
,
prediction
def
infer
(
test_reader
,
use_cuda
,
model_path
=
None
):
"""
inference function
"""
if
model_path
is
None
:
print
(
str
(
model_path
)
+
" cannot be found"
)
return
place
=
fluid
.
CPUPlace
()
exe
=
fluid
.
Executor
(
place
)
inference_scope
=
fluid
.
core
.
Scope
()
with
fluid
.
scope_guard
(
inference_scope
):
[
inference_program
,
feed_target_names
,
fetch_targets
]
=
fluid
.
io
.
load_inference_model
(
model_path
,
exe
)
class2_list
,
class3_list
=
[],
[]
for
each_test_reader
in
test_reader
:
class2_acc
,
class3_acc
=
0.0
,
0.0
total_count
,
neu_count
=
0
,
0
for
data
in
each_test_reader
():
pred
=
exe
.
run
(
inference_program
,
feed
=
data2pred
(
data
,
place
),
fetch_list
=
fetch_targets
,
return_numpy
=
True
)
for
i
,
val
in
enumerate
(
data
):
pos_score
=
pred
[
0
][
i
,
1
]
true_label
=
val
[
1
]
if
true_label
==
2.0
and
pos_score
>
0.5
:
class2_acc
+=
1
if
true_label
==
0.0
and
pos_score
<
0.5
:
class2_acc
+=
1
if
true_label
==
2.0
and
pos_score
>
0.55
:
class3_acc
+=
1
if
true_label
==
1.0
and
pos_score
>
0.45
and
pos_score
<=
0.55
:
class3_acc
+=
1
if
true_label
==
0.0
and
pos_score
<=
0.45
:
class3_acc
+=
1
if
true_label
==
1.0
:
neu_count
+=
1
total_count
+=
len
(
data
)
class2_acc
=
class2_acc
/
(
total_count
-
neu_count
)
class3_acc
=
class3_acc
/
total_count
class2_list
.
append
(
class2_acc
)
class3_list
.
append
(
class3_acc
)
class2_acc
=
sum
(
class2_list
)
/
len
(
class2_list
)
class3_acc
=
sum
(
class3_list
)
/
len
(
class3_list
)
print
(
"[test info] model_path: %s, class2_acc: %f, class3_acc: %f"
%
(
model_path
,
class2_acc
,
class3_acc
))
def
start_train
(
train_reader
,
test_reader
,
word_dict
,
network
,
use_cuda
,
parallel
,
save_dirname
,
lr
=
0.2
,
batch_size
=
128
,
pass_num
=
30
):
"""
train network
"""
data
=
fluid
.
layers
.
data
(
name
=
"words"
,
shape
=
[
1
],
dtype
=
"int64"
,
lod_level
=
1
)
label
=
fluid
.
layers
.
data
(
name
=
"label"
,
shape
=
[
1
],
dtype
=
"int64"
)
cost
,
acc
,
pred
=
network
(
data
,
label
,
len
(
word_dict
)
+
1
)
sgd_optimizer
=
fluid
.
optimizer
.
Adagrad
(
learning_rate
=
lr
)
sgd_optimizer
.
minimize
(
cost
)
place
=
fluid
.
CPUPlace
()
feeder
=
fluid
.
DataFeeder
(
feed_list
=
[
data
,
label
],
place
=
place
)
start_exe
=
fluid
.
Executor
(
place
)
start_exe
.
run
(
fluid
.
default_startup_program
())
exe
=
fluid
.
ParallelExecutor
(
use_cuda
,
loss_name
=
cost
.
name
)
for
pass_id
in
xrange
(
pass_num
):
total_acc
,
total_cost
,
total_count
,
avg_cost
,
avg_acc
=
0.0
,
0.0
,
0.0
,
0.0
,
0.0
for
data
in
train_reader
():
cost_val
,
acc_val
=
exe
.
run
(
feed
=
feeder
.
feed
(
data
),
fetch_list
=
[
cost
.
name
,
acc
.
name
])
cost_val_list
,
acc_val_list
=
np
.
array
(
cost_val
),
np
.
array
(
acc_val
)
total_cost
+=
cost_val_list
.
sum
()
*
len
(
data
)
total_acc
+=
acc_val_list
.
sum
()
*
len
(
data
)
total_count
+=
len
(
data
)
avg_cost
=
total_cost
/
total_count
avg_acc
=
total_acc
/
total_count
print
(
"[train info]: pass_id: %d, avg_acc: %f, avg_cost: %f"
%
(
pass_id
,
avg_acc
,
avg_cost
))
gpu_place
=
fluid
.
CUDAPlace
(
0
)
save_exe
=
fluid
.
Executor
(
gpu_place
)
epoch_model
=
save_dirname
+
"/"
+
"epoch"
+
str
(
pass_id
)
fluid
.
io
.
save_inference_model
(
epoch_model
,
[
"words"
],
pred
,
save_exe
)
infer
(
test_reader
,
False
,
epoch_model
)
def
train_net
(
vocab
=
"./thirdparty/train.vocab"
,
train_dir
=
"./train"
,
test_list
=
[
"car"
,
"spot"
,
"weibo"
,
"lbs"
]):
"""
w_dict = scdb_word_dict(vocab=vocab)
test_files = [ "./thirdparty" + os.sep + f for f in test_list]
train_reader = paddle.batch(
scdb_train_data(train_dir, w_dict),
batch_size = 256)
test_reader = [paddle.batch(scdb_test_data(test_file, w_dict), batch_size = 50)
\
for test_file in test_files]
"""
w_dict
=
paddle
.
dataset
.
imdb
.
word_dict
()
print
(
"dict ready"
)
train_reader
=
paddle
.
batch
(
paddle
.
reader
.
shuffle
(
paddle
.
dataset
.
imdb
.
train
(
w_dict
),
buf_size
=
50000
),
batch_size
=
128
)
test_reader
=
paddle
.
batch
(
paddle
.
reader
.
shuffle
(
paddle
.
dataset
.
imdb
.
test
(
w_dict
),
buf_size
=
50000
),
batch_size
=
128
)
test_reader
=
[
test_reader
]
start_train
(
train_reader
,
test_reader
,
w_dict
,
bilstm_net
,
use_cuda
=
True
,
parallel
=
False
,
save_dirname
=
"scdb_bilstm_model"
,
lr
=
0.05
,
pass_num
=
10
,
batch_size
=
256
)
if
__name__
==
"__main__"
:
train_net
()
fluid/text_classification/clouds/scdb_single_card.py
0 → 100644
浏览文件 @
2d392828
import
unittest
import
contextlib
import
paddle.fluid
as
fluid
import
paddle.v2
as
paddle
import
numpy
as
np
import
sys
import
time
import
os
import
json
import
random
def
to_lodtensor
(
data
,
place
):
"""
convert to LODtensor
"""
seq_lens
=
[
len
(
seq
)
for
seq
in
data
]
cur_len
=
0
lod
=
[
cur_len
]
for
l
in
seq_lens
:
cur_len
+=
l
lod
.
append
(
cur_len
)
flattened_data
=
np
.
concatenate
(
data
,
axis
=
0
).
astype
(
"int64"
)
flattened_data
=
flattened_data
.
reshape
([
len
(
flattened_data
),
1
])
res
=
fluid
.
LoDTensor
()
res
.
set
(
flattened_data
,
place
)
res
.
set_lod
([
lod
])
return
res
def
load_vocab
(
filename
):
"""
load imdb vocabulary
"""
vocab
=
{}
with
open
(
filename
)
as
f
:
wid
=
0
for
line
in
f
:
vocab
[
line
.
strip
()]
=
wid
wid
+=
1
vocab
[
"<unk>"
]
=
len
(
vocab
)
return
vocab
def
data2tensor
(
data
,
place
):
"""
data2tensor
"""
input_seq
=
to_lodtensor
(
map
(
lambda
x
:
x
[
0
],
data
),
place
)
y_data
=
np
.
array
(
map
(
lambda
x
:
x
[
1
],
data
)).
astype
(
"int64"
)
y_data
=
y_data
.
reshape
([
-
1
,
1
])
return
{
"words"
:
input_seq
,
"label"
:
y_data
}
def
data2pred
(
data
,
place
):
"""
data2tensor
"""
input_seq
=
to_lodtensor
(
map
(
lambda
x
:
x
[
0
],
data
),
place
)
y_data
=
np
.
array
(
map
(
lambda
x
:
x
[
1
],
data
)).
astype
(
"int64"
)
y_data
=
y_data
.
reshape
([
-
1
,
1
])
return
{
"words"
:
input_seq
}
def
load_dict
(
vocab
):
"""
Load dict from vocab
"""
word_dict
=
dict
()
with
open
(
vocab
,
"r"
)
as
fin
:
for
line
in
fin
:
cols
=
line
.
strip
(
"
\r\n
"
).
decode
(
"gb18030"
).
split
(
"
\t
"
)
word_dict
[
cols
[
0
]]
=
int
(
cols
[
1
])
return
word_dict
def
save_dict
(
word_dict
,
vocab
):
"""
Save dict into file
"""
with
open
(
vocab
,
"w"
)
as
fout
:
for
k
,
v
in
word_dict
.
iteritems
():
outstr
=
(
"%s
\t
%s
\n
"
%
(
k
,
v
)).
encode
(
"gb18030"
)
fout
.
write
(
outstr
)
def
build_dict
(
fname
):
"""
build word dict using trainset
"""
word_dict
=
dict
()
with
open
(
fname
,
"r"
)
as
fin
:
for
line
in
fin
:
try
:
words
=
line
.
strip
(
"
\r\n
"
).
decode
(
"gb18030"
).
split
(
"
\t
"
)[
1
].
split
(
" "
)
except
:
sys
.
stderr
.
write
(
"[warning] build_dict: decode error
\n
"
)
continue
for
w
in
words
:
if
w
not
in
word_dict
:
word_dict
[
w
]
=
len
(
word_dict
)
return
word_dict
def
scdb_word_dict
(
vocab
=
"scdb_data/train_set/train.vocab"
):
"""
get word_dict
"""
if
not
os
.
path
.
exists
(
vocab
):
w_dict
=
build_dict
(
train_file
)
save_dict
(
w_dict
,
vocab
)
else
:
w_dict
=
load_dict
(
vocab
)
w_dict
[
"<unk>"
]
=
len
(
w_dict
)
return
w_dict
def
data_reader
(
fname
,
word_dict
,
is_dir
=
False
):
"""
Convert word sequence into slot
"""
unk_id
=
len
(
word_dict
)
all_data
=
[]
filelist
=
[]
if
is_dir
:
filelist
=
[
fname
+
os
.
sep
+
f
for
f
in
os
.
listdir
(
fname
)]
else
:
filelist
=
[
fname
]
for
each_name
in
filelist
:
with
open
(
each_name
,
"r"
)
as
fin
:
for
line
in
fin
:
try
:
cols
=
line
.
strip
(
"
\r\n
"
).
decode
(
"gb18030"
).
split
(
"
\t
"
)
except
:
sys
.
stderr
.
write
(
"warning: ignore decode error
\n
"
)
continue
label
=
int
(
cols
[
0
])
wids
=
[
word_dict
[
x
]
if
x
in
word_dict
else
unk_id
for
x
in
cols
[
1
].
split
(
" "
)]
all_data
.
append
((
wids
,
label
))
random
.
shuffle
(
all_data
)
def
reader
():
for
doc
,
label
in
all_data
:
yield
doc
,
label
return
reader
def
scdb_train_data
(
train_dir
=
"scdb_data/train_set/corpus.train.seg"
,
w_dict
=
None
):
"""
create train data
"""
return
data_reader
(
train_dir
,
w_dict
,
True
)
def
scdb_test_data
(
test_file
,
w_dict
):
"""
test_set=["car", "lbs", "spot", "weibo",
"baby", "toutiao", "3c", "movie", "haogan"]
"""
return
data_reader
(
test_file
,
w_dict
)
def
bow_net
(
data
,
label
,
dict_dim
,
emb_dim
=
128
,
hid_dim
=
128
,
hid_dim2
=
96
,
class_dim
=
2
):
"""
bow net
"""
emb
=
fluid
.
layers
.
embedding
(
input
=
data
,
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
)
acc
=
fluid
.
layers
.
accuracy
(
input
=
prediction
,
label
=
label
)
return
avg_cost
,
acc
,
prediction
def
cnn_net
(
data
,
label
,
dict_dim
,
emb_dim
=
128
,
hid_dim
=
128
,
hid_dim2
=
96
,
class_dim
=
2
,
win_size
=
3
):
"""
conv net
"""
emb
=
fluid
.
layers
.
embedding
(
input
=
data
,
size
=
[
dict_dim
,
emb_dim
])
conv_3
=
fluid
.
nets
.
sequence_conv_pool
(
input
=
emb
,
num_filters
=
hid_dim
,
filter_size
=
win_size
,
act
=
"tanh"
,
pool_type
=
"max"
)
fc_1
=
fluid
.
layers
.
fc
(
input
=
[
conv_3
],
size
=
hid_dim2
)
prediction
=
fluid
.
layers
.
fc
(
input
=
[
fc_1
],
size
=
class_dim
,
act
=
"softmax"
)
cost
=
fluid
.
layers
.
cross_entropy
(
input
=
prediction
,
label
=
label
)
avg_cost
=
fluid
.
layers
.
mean
(
x
=
cost
)
acc
=
fluid
.
layers
.
accuracy
(
input
=
prediction
,
label
=
label
)
return
avg_cost
,
acc
,
prediction
def
lstm_net
(
data
,
label
,
dict_dim
,
emb_dim
=
128
,
hid_dim
=
128
,
hid_dim2
=
96
,
class_dim
=
2
,
emb_lr
=
30.0
):
"""
lstm net
"""
emb
=
fluid
.
layers
.
embedding
(
input
=
data
,
size
=
[
dict_dim
,
emb_dim
],
param_attr
=
fluid
.
ParamAttr
(
learning_rate
=
emb_lr
))
fc0
=
fluid
.
layers
.
fc
(
input
=
emb
,
size
=
hid_dim
*
4
,
act
=
'tanh'
)
lstm_h
,
c
=
fluid
.
layers
.
dynamic_lstm
(
input
=
fc0
,
size
=
hid_dim
*
4
,
is_reverse
=
False
)
lstm_max
=
fluid
.
layers
.
sequence_pool
(
input
=
lstm_h
,
pool_type
=
'max'
)
lstm_max_tanh
=
fluid
.
layers
.
tanh
(
lstm_max
)
fc1
=
fluid
.
layers
.
fc
(
input
=
lstm_max_tanh
,
size
=
hid_dim2
,
act
=
'tanh'
)
prediction
=
fluid
.
layers
.
fc
(
input
=
fc1
,
size
=
class_dim
,
act
=
'softmax'
)
cost
=
fluid
.
layers
.
cross_entropy
(
input
=
prediction
,
label
=
label
)
avg_cost
=
fluid
.
layers
.
mean
(
x
=
cost
)
acc
=
fluid
.
layers
.
accuracy
(
input
=
prediction
,
label
=
label
)
return
avg_cost
,
acc
,
prediction
def
bilstm_net
(
data
,
label
,
dict_dim
,
emb_dim
=
128
,
hid_dim
=
128
,
hid_dim2
=
96
,
class_dim
=
2
,
emb_lr
=
30.0
):
"""
lstm net
"""
emb
=
fluid
.
layers
.
embedding
(
input
=
data
,
size
=
[
dict_dim
,
emb_dim
],
param_attr
=
fluid
.
ParamAttr
(
learning_rate
=
emb_lr
))
fc0
=
fluid
.
layers
.
fc
(
input
=
emb
,
size
=
hid_dim
*
4
,
act
=
'tanh'
)
rfc0
=
fluid
.
layers
.
fc
(
input
=
emb
,
size
=
hid_dim
*
4
,
act
=
'tanh'
)
lstm_h
,
c
=
fluid
.
layers
.
dynamic_lstm
(
input
=
fc0
,
size
=
hid_dim
*
4
,
is_reverse
=
False
)
rlstm_h
,
c
=
fluid
.
layers
.
dynamic_lstm
(
input
=
rfc0
,
size
=
hid_dim
*
4
,
is_reverse
=
True
)
lstm_last
=
fluid
.
layers
.
sequence_last_step
(
input
=
lstm_h
)
rlstm_last
=
fluid
.
layers
.
sequence_last_step
(
input
=
rlstm_h
)
lstm_last_tanh
=
fluid
.
layers
.
tanh
(
lstm_last
)
rlstm_last_tanh
=
fluid
.
layers
.
tanh
(
rlstm_last
)
lstm_concat
=
fluid
.
layers
.
concat
(
input
=
[
lstm_last
,
rlstm_last
],
axis
=
1
)
fc1
=
fluid
.
layers
.
fc
(
input
=
lstm_concat
,
size
=
hid_dim2
,
act
=
'tanh'
)
prediction
=
fluid
.
layers
.
fc
(
input
=
fc1
,
size
=
class_dim
,
act
=
'softmax'
)
cost
=
fluid
.
layers
.
cross_entropy
(
input
=
prediction
,
label
=
label
)
avg_cost
=
fluid
.
layers
.
mean
(
x
=
cost
)
acc
=
fluid
.
layers
.
accuracy
(
input
=
prediction
,
label
=
label
)
return
avg_cost
,
acc
,
prediction
def
gru_net
(
data
,
label
,
dict_dim
,
emb_dim
=
128
,
hid_dim
=
128
,
hid_dim2
=
96
,
class_dim
=
2
,
emb_lr
=
30.0
):
"""
gru net
"""
emb
=
fluid
.
layers
.
embedding
(
input
=
data
,
size
=
[
dict_dim
,
emb_dim
],
param_attr
=
fluid
.
ParamAttr
(
learning_rate
=
emb_lr
))
fc0
=
fluid
.
layers
.
fc
(
input
=
emb
,
size
=
hid_dim
*
3
)
gru_h
=
fluid
.
layers
.
dynamic_gru
(
input
=
fc0
,
size
=
hid_dim
,
is_reverse
=
False
)
gru_max
=
fluid
.
layers
.
sequence_pool
(
input
=
gru_h
,
pool_type
=
'max'
)
gru_max_tanh
=
fluid
.
layers
.
tanh
(
gru_max
)
fc1
=
fluid
.
layers
.
fc
(
input
=
gru_max_tanh
,
size
=
hid_dim2
,
act
=
'tanh'
)
prediction
=
fluid
.
layers
.
fc
(
input
=
fc1
,
size
=
class_dim
,
act
=
'softmax'
)
cost
=
fluid
.
layers
.
cross_entropy
(
input
=
prediction
,
label
=
label
)
avg_cost
=
fluid
.
layers
.
mean
(
x
=
cost
)
acc
=
fluid
.
layers
.
accuracy
(
input
=
prediction
,
label
=
label
)
return
avg_cost
,
acc
,
prediction
def
infer
(
test_reader
,
use_cuda
,
model_path
=
None
):
"""
inference function
"""
if
model_path
is
None
:
print
(
str
(
model_path
)
+
" cannot be found"
)
return
place
=
fluid
.
CUDAPlace
(
0
)
if
use_cuda
else
fluid
.
CPUPlace
()
exe
=
fluid
.
Executor
(
place
)
inference_scope
=
fluid
.
core
.
Scope
()
with
fluid
.
scope_guard
(
inference_scope
):
[
inference_program
,
feed_target_names
,
fetch_targets
]
=
fluid
.
io
.
load_inference_model
(
model_path
,
exe
)
class2_list
,
class3_list
=
[],
[]
for
each_test_reader
in
test_reader
:
class2_acc
,
class3_acc
=
0.0
,
0.0
total_count
,
neu_count
=
0
,
0
for
data
in
each_test_reader
():
pred
=
exe
.
run
(
inference_program
,
feed
=
data2pred
(
data
,
place
),
fetch_list
=
fetch_targets
,
return_numpy
=
True
)
for
i
,
val
in
enumerate
(
data
):
pos_score
=
pred
[
0
][
i
,
1
]
true_label
=
val
[
1
]
if
true_label
==
2.0
and
pos_score
>
0.5
:
class2_acc
+=
1
if
true_label
==
0.0
and
pos_score
<
0.5
:
class2_acc
+=
1
if
true_label
==
2.0
and
pos_score
>
0.55
:
class3_acc
+=
1
if
true_label
==
1.0
and
pos_score
>
0.45
and
pos_score
<=
0.55
:
class3_acc
+=
1
if
true_label
==
0.0
and
pos_score
<=
0.45
:
class3_acc
+=
1
if
true_label
==
1.0
:
neu_count
+=
1
total_count
+=
len
(
data
)
class2_acc
=
class2_acc
/
(
total_count
-
neu_count
)
class3_acc
=
class3_acc
/
total_count
class2_list
.
append
(
class2_acc
)
class3_list
.
append
(
class3_acc
)
class2_acc
=
sum
(
class2_list
)
/
len
(
class2_list
)
class3_acc
=
sum
(
class3_list
)
/
len
(
class3_list
)
print
(
"[test info] model_path: %s, class2_acc: %f, class3_acc: %f"
%
(
model_path
,
class2_acc
,
class3_acc
))
def
start_train
(
train_reader
,
test_reader
,
word_dict
,
network
,
use_cuda
,
parallel
,
save_dirname
,
lr
=
0.2
,
batch_size
=
128
,
pass_num
=
30
):
"""
train network
"""
data
=
fluid
.
layers
.
data
(
name
=
"words"
,
shape
=
[
1
],
dtype
=
"int64"
,
lod_level
=
1
)
label
=
fluid
.
layers
.
data
(
name
=
"label"
,
shape
=
[
1
],
dtype
=
"int64"
)
cost
,
acc
,
pred
=
network
(
data
,
label
,
len
(
word_dict
)
+
1
)
sgd_optimizer
=
fluid
.
optimizer
.
Adagrad
(
learning_rate
=
lr
)
sgd_optimizer
.
minimize
(
cost
)
place
=
fluid
.
CUDAPlace
(
0
)
if
use_cuda
else
fluid
.
CPUPlace
()
exe
=
fluid
.
Executor
(
place
)
feeder
=
fluid
.
DataFeeder
(
feed_list
=
[
data
,
label
],
place
=
place
)
exe
.
run
(
fluid
.
default_startup_program
())
for
pass_id
in
xrange
(
pass_num
):
data_size
,
data_count
,
total_acc
,
total_cost
=
0
,
0
,
0.0
,
0.0
for
data
in
train_reader
():
avg_cost_np
,
avg_acc_np
=
exe
.
run
(
fluid
.
default_main_program
(),
feed
=
feeder
.
feed
(
data
),
fetch_list
=
[
cost
,
acc
])
data_size
=
len
(
data
)
total_acc
+=
data_size
*
avg_acc_np
total_cost
+=
data_size
*
avg_cost_np
data_count
+=
data_size
avg_cost
=
total_cost
/
data_count
avg_acc
=
total_acc
/
data_count
print
(
"[train info]: pass_id: %d, avg_acc: %f, avg_cost: %f"
%
(
pass_id
,
avg_acc
,
avg_cost
))
epoch_model
=
save_dirname
+
"/"
+
"epoch"
+
str
(
pass_id
)
fluid
.
io
.
save_inference_model
(
epoch_model
,
[
"words"
],
pred
,
exe
)
infer
(
test_reader
,
False
,
epoch_model
)
def
train_net
(
vocab
=
"./thirdparty/train.vocab"
,
train_dir
=
"./train"
,
test_list
=
[
"car"
,
"spot"
,
"weibo"
,
"lbs"
]):
w_dict
=
scdb_word_dict
(
vocab
=
vocab
)
test_files
=
[
"./thirdparty"
+
os
.
sep
+
f
for
f
in
test_list
]
train_reader
=
paddle
.
batch
(
scdb_train_data
(
train_dir
,
w_dict
),
batch_size
=
256
)
test_reader
=
[
paddle
.
batch
(
scdb_test_data
(
test_file
,
w_dict
),
batch_size
=
50
)
\
for
test_file
in
test_files
]
start_train
(
train_reader
,
test_reader
,
w_dict
,
bow_net
,
use_cuda
=
False
,
parallel
=
False
,
save_dirname
=
"scdb_bow_model"
,
lr
=
0.002
,
pass_num
=
10
,
batch_size
=
256
)
if
__name__
==
"__main__"
:
train_net
()
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