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
urhero
Paddle
提交
c469334c
P
Paddle
项目概览
urhero
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
1
Star
0
Fork
0
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
0
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
0
Issue
0
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
c469334c
编写于
11月 27, 2018
作者:
J
JiabinYang
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
polish python code and comment, test=develop
上级
87648f8e
变更
4
隐藏空白更改
内联
并排
Showing
4 changed file
with
88 addition
and
73 deletion
+88
-73
paddle/fluid/operators/hierarchical_sigmoid_op.h
paddle/fluid/operators/hierarchical_sigmoid_op.h
+8
-8
python/paddle/fluid/layers/nn.py
python/paddle/fluid/layers/nn.py
+34
-24
python/paddle/fluid/tests/unittests/test_hsigmoid_op.py
python/paddle/fluid/tests/unittests/test_hsigmoid_op.py
+39
-36
python/paddle/fluid/tests/unittests/test_layers.py
python/paddle/fluid/tests/unittests/test_layers.py
+7
-5
未找到文件。
paddle/fluid/operators/hierarchical_sigmoid_op.h
浏览文件 @
c469334c
...
...
@@ -47,11 +47,11 @@ template <typename DeviceContext, typename T>
class
HierarchicalSigmoidOpKernel
:
public
framework
::
OpKernel
<
T
>
{
public:
void
Compute
(
const
framework
::
ExecutionContext
&
ctx
)
const
override
{
auto
in
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"X"
));
auto
w
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"W"
));
auto
&
in
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"X"
));
auto
&
w
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"W"
));
auto
*
path
=
ctx
.
Input
<
framework
::
LoDTensor
>
(
"PTable"
);
auto
*
code
=
ctx
.
Input
<
framework
::
LoDTensor
>
(
"PathCode"
);
auto
label
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"Label"
));
auto
&
label
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"Label"
));
auto
*
bias
=
ctx
.
Input
<
framework
::
LoDTensor
>
(
"Bias"
);
auto
*
out
=
ctx
.
Output
<
framework
::
LoDTensor
>
(
"Out"
);
auto
*
pre_out
=
ctx
.
Output
<
framework
::
LoDTensor
>
(
"PreOut"
);
...
...
@@ -114,8 +114,8 @@ template <typename DeviceContext, typename T>
class
HierarchicalSigmoidGradOpKernel
:
public
framework
::
OpKernel
<
T
>
{
public:
void
Compute
(
const
framework
::
ExecutionContext
&
ctx
)
const
override
{
auto
in
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"X"
));
auto
w
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"W"
));
auto
&
in
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"X"
));
auto
&
w
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"W"
));
auto
*
path
=
ctx
.
Input
<
framework
::
LoDTensor
>
(
"PTable"
);
auto
*
code
=
ctx
.
Input
<
framework
::
LoDTensor
>
(
"PathCode"
);
auto
*
bias
=
ctx
.
Input
<
framework
::
LoDTensor
>
(
"Bias"
);
...
...
@@ -124,9 +124,9 @@ class HierarchicalSigmoidGradOpKernel : public framework::OpKernel<T> {
bool
is_sparse
=
ctx
.
Attr
<
bool
>
(
"is_sparse"
);
auto
&
dev_ctx
=
ctx
.
template
device_context
<
DeviceContext
>();
math
::
SetConstant
<
DeviceContext
,
T
>
zero
;
auto
label
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"Label"
));
auto
pre_out
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"PreOut"
));
auto
out_grad
=
detail
::
Ref
(
auto
&
label
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"Label"
));
auto
&
pre_out
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
"PreOut"
));
auto
&
out_grad
=
detail
::
Ref
(
ctx
.
Input
<
framework
::
LoDTensor
>
(
framework
::
GradVarName
(
"Out"
)));
framework
::
LoDTensor
pre_out_grad
;
...
...
python/paddle/fluid/layers/nn.py
浏览文件 @
c469334c
...
...
@@ -4589,23 +4589,33 @@ def hsigmoid(input,
bias_attr
=
None
,
name
=
None
,
non_leaf_num
=
None
,
ptable
=
None
,
pcode
=
None
,
is_c
ostu
m
=
False
,
p
ath_
table
=
None
,
p
ath_
code
=
None
,
is_c
usto
m
=
False
,
is_sparse
=
False
):
"""
The hierarchical sigmoid operator is used to accelerate the training
process of language model. This operator organizes the classes into a
complete binary tree, each leaf node represents a class(a word) and each
complete binary tree, or you can use is_custom to pass your own tree to
implement hierarchical. Each leaf node represents a class(a word) and each
internal node acts as a binary classifier. For each word there's a unique
path from root to it's leaf node, hsigmoid calculate the cost for each
internal node on the path, and sum them to get a total cost. hsigmoid can
achive a acceleration from :math:`O(N)` to :math:`O(logN)`, where :math:`N`
represents the size of word dict.
Refer to `Hierarchical Probabilistic Neural Network Language Model
Using default tree you can
Refer to `Hierarchical Probabilistic Neural Network Language Model
<http://www.iro.umontreal.ca/~lisa/pointeurs/hierarchical-nnlm-aistats05.pdf>`_
And if you want to use the costumed tree by set 'is_custom' as true you may need to do following things first:
1. using your word dict to build a binary tree, each leaf node should be an word of your word dict
2. build a dict to store word_id -> word's leaf to root path, we call it path_table.
3. build a dict to store word_id -> code of word's leaf to root path, we call it path_code. Code
means label of each binary classification, using 1 indicate true, 0 indicate false.
4. now, each word should has its path and code along the path, you can pass a batch of path and code
related to the same batch of inputs.
Args:
input (Variable): The input tensor variable with shape
:math:`[N
\\
times D]`, where :math:`N` is the size of mini-batch,
...
...
@@ -4613,13 +4623,6 @@ def hsigmoid(input,
label (Variable): The tensor variable contains labels of training data.
It's a tensor with shape is :math:`[N
\\
times 1]`.
num_classes: (int), The number of classes, must not be less than 2. with default tree this has to be set
non_leaf_num: this defines the number of non-leaf nodes in costumed tree
ptable: (Variable|None) this variable can store each batch of samples' path to root,
it should be in leaf -> root order
ptable should have the same shape with pcode, and for each sample i ptable[i] indicates a np.array like
structure and each element in this array is indexes in parent nodes' Weight Matrix.
pcode: (Variable|None) this variable can store each batch of samples' code,
each code consist with every code of parent nodes. it should be in leaf -> root order
param_attr (ParamAttr|None): The parameter attribute for learnable parameters/weights
of hsigmoid. If it is set to None or one attribute of ParamAttr, hsigmoid
will create ParamAttr as param_attr. If the Initializer of the param_attr
...
...
@@ -4631,8 +4634,15 @@ def hsigmoid(input,
is not set, the bias is initialized zero. Default: None.
name (str|None): A name for this layer(optional). If set None, the layer
will be named automatically. Default: None.
is_costum: (bool|False)using user defined binary tree instead of default complete binary tree, if costum is
set you need to set ptable/pcode/non_leaf_num, otherwise num_classes should be set
non_leaf_num: this defines the number of non-leaf nodes in costumed tree
path_table: (Variable|None) this variable can store each batch of samples' path to root,
it should be in leaf -> root order
path_table should have the same shape with path_code, and for each sample i path_table[i] indicates a np.array like
structure and each element in this array is indexes in parent nodes' Weight Matrix.
path_code: (Variable|None) this variable can store each batch of samples' code,
each code consist with every code of parent nodes. it should be in leaf -> root order
is_custom: (bool|False)using user defined binary tree instead of default complete binary tree, if costum is
set you need to set path_table/path_code/non_leaf_num, otherwise num_classes should be set
is_sparse: (bool|False)using sparse update instead of dense update, if set, the gradient
of W and input will be sparse.
...
...
@@ -4653,22 +4663,22 @@ def hsigmoid(input,
out
=
helper
.
create_variable_for_type_inference
(
dtype
)
pre_out
=
helper
.
create_variable_for_type_inference
(
dtype
)
dim
=
input
.
shape
[
1
]
if
((
num_classes
is
None
)
or
(
num_classes
<
2
))
and
(
not
is_c
ostu
m
):
if
((
num_classes
is
None
)
or
(
num_classes
<
2
))
and
(
not
is_c
usto
m
):
raise
ValueError
(
"num_classes must not be less than 2 with default tree"
)
if
(
is_c
ostum
)
and
(
p
code
is
None
):
raise
ValueError
(
"pcode should not be None with costum tree"
)
elif
(
is_c
ostum
)
and
(
p
table
is
None
):
raise
ValueError
(
"ptable should not be None with costum tree"
)
elif
(
is_c
ostu
m
)
and
(
non_leaf_num
is
None
):
if
(
is_c
ustom
)
and
(
path_
code
is
None
):
raise
ValueError
(
"p
ath_
code should not be None with costum tree"
)
elif
(
is_c
ustom
)
and
(
path_
table
is
None
):
raise
ValueError
(
"p
ath_
table should not be None with costum tree"
)
elif
(
is_c
usto
m
)
and
(
non_leaf_num
is
None
):
raise
ValueError
(
"non_leaf_num should not be None with costum tree"
)
else
:
pass
weights
=
None
if
not
is_c
ostu
m
:
if
not
is_c
usto
m
:
weights
=
helper
.
create_parameter
(
attr
=
helper
.
param_attr
,
shape
=
[
num_classes
-
1
,
dim
],
...
...
@@ -4683,12 +4693,12 @@ def hsigmoid(input,
inputs
=
{
"X"
:
input
,
"W"
:
weights
,
"PTable"
:
ptable
,
"PathCode"
:
pcode
,
"PTable"
:
p
ath_
table
,
"PathCode"
:
p
ath_
code
,
"Label"
:
label
}
if
helper
.
bias_attr
:
if
not
is_c
ostu
m
:
if
not
is_c
usto
m
:
bias
=
helper
.
create_parameter
(
attr
=
helper
.
bias_attr
,
shape
=
[
num_classes
-
1
,
1
],
...
...
python/paddle/fluid/tests/unittests/test_hsigmoid_op.py
浏览文件 @
c469334c
...
...
@@ -43,9 +43,9 @@ class CodeTable(object):
class
CodeTableWithCustomTree
(
object
):
def
__init__
(
self
,
p
table
,
p
code
,
index
):
self
.
ptable_
=
ptable
self
.
pcode_
=
pcode
def
__init__
(
self
,
p
ath_table
,
path_
code
,
index
):
self
.
ptable_
=
p
ath_
table
self
.
pcode_
=
p
ath_
code
self
.
index_
=
index
def
cal_index
(
self
,
bit
):
...
...
@@ -102,9 +102,10 @@ def hsigmoid(x, w, label, bias, num_classes):
return
pre_output
,
out
def
hsigmoidWithCustomTree
(
x
,
w
,
ptable
,
pcode
,
label
,
bias
,
num_classes
):
def
hsigmoidWithCustomTree
(
x
,
w
,
path_table
,
path_code
,
label
,
bias
,
num_classes
):
batch_size
=
x
.
shape
[
0
]
code_length
=
len
(
ptable
[
0
])
code_length
=
len
(
p
ath_
table
[
0
])
code_table
=
[
0
for
_
in
range
(
code_length
)]
# init pre_out with shape [N, code_length]
pre_output
=
np
.
zeros
((
batch_size
,
code_length
))
...
...
@@ -112,13 +113,13 @@ def hsigmoidWithCustomTree(x, w, ptable, pcode, label, bias, num_classes):
out
=
np
.
zeros
((
batch_size
,
1
)).
astype
(
"float32"
)
if
isinstance
(
bias
,
np
.
ndarray
):
for
i
in
range
(
batch_size
):
code_table
=
CodeTableWithCustomTree
(
p
table
,
p
code
,
i
)
code_table
=
CodeTableWithCustomTree
(
p
ath_table
,
path_
code
,
i
)
length
=
code_table
.
get_length
()
for
j
in
range
(
length
):
idx
=
code_table
.
cal_index
(
j
)
pre_output
[
i
][
j
]
+=
bias
[
idx
][
0
]
for
i
in
range
(
batch_size
):
code_table
=
CodeTableWithCustomTree
(
p
table
,
p
code
,
i
)
code_table
=
CodeTableWithCustomTree
(
p
ath_table
,
path_
code
,
i
)
length
=
code_table
.
get_length
()
for
j
in
range
(
length
):
idx
=
code_table
.
cal_index
(
j
)
...
...
@@ -127,7 +128,7 @@ def hsigmoidWithCustomTree(x, w, ptable, pcode, label, bias, num_classes):
pre_output
=
np
.
clip
(
pre_output
,
-
40.0
,
40.0
)
# out(i, 0) = \sum_j bit(i, j) * preout(i, j)
for
i
in
range
(
batch_size
):
code_table
=
CodeTableWithCustomTree
(
p
table
,
p
code
,
i
)
code_table
=
CodeTableWithCustomTree
(
p
ath_table
,
path_
code
,
i
)
length
=
code_table
.
get_length
()
sum
=
0.0
for
j
in
range
(
length
):
...
...
@@ -173,24 +174,24 @@ class TestHSigmoidOpSparse(OpTest):
x
=
np
.
random
.
random
((
batch_size
,
feature_size
)).
astype
(
"float32"
)
w
=
np
.
random
.
random
((
num_classes
-
1
,
feature_size
)).
astype
(
"float32"
)
label
=
np
.
array
([
0
,
1
,
4
,
5
])
ptable
=
np
.
array
(
p
ath_
table
=
np
.
array
(
[(
0
,
2
,
-
1
,
-
1
,
-
1
),
(
0
,
1
,
3
,
-
1
,
-
1
),
(
0
,
1
,
4
,
-
1
,
-
1
),
(
0
,
2
,
-
1
,
-
1
,
-
1
)])
#np.array to store 1,2,5,6s' non-leaf path(root -> leaf)
pcode
=
np
.
array
([(
0
,
0
,
-
1
,
-
1
,
-
1
),
(
1
,
1
,
1
,
-
1
,
-
1
),
(
p
ath_
code
=
np
.
array
([(
0
,
0
,
-
1
,
-
1
,
-
1
),
(
1
,
1
,
1
,
-
1
,
-
1
),
(
1
,
0
,
0
,
-
1
,
-
1
),
(
0
,
1
,
-
1
,
-
1
,
-
1
)])
#np.array to store
bias
=
np
.
random
.
random
((
num_classes
-
1
,
1
)).
astype
(
"float32"
)
self
.
attrs
=
{
'num_classes'
:
num_classes
,
'is_sparse'
:
True
}
self
.
inputs
=
{
'X'
:
x
,
'W'
:
w
,
'PTable'
:
ptable
,
'PathCode'
:
pcode
,
'PTable'
:
p
ath_
table
,
'PathCode'
:
p
ath_
code
,
'Label'
:
label
,
'Bias'
:
bias
}
pre_output
,
out
=
hsigmoidWithCustomTree
(
x
,
w
,
p
table
,
pcode
,
label
,
bias
,
num_classes
)
pre_output
,
out
=
hsigmoidWithCustomTree
(
x
,
w
,
p
ath_table
,
path_code
,
label
,
bias
,
num_classes
)
self
.
outputs
=
{
'PreOut'
:
pre_output
,
'Out'
:
out
}
def
test_check_output
(
self
):
...
...
@@ -200,11 +201,13 @@ class TestHSigmoidOpSparse(OpTest):
class
TestHSigmoidOpWithSparseGrad
(
unittest
.
TestCase
):
def
hs_net_conf
(
self
,
is_sparse
):
input_word
=
fluid
.
layers
.
data
(
name
=
"x"
,
shape
=
[
1
],
dtype
=
'int64'
)
ptable
=
fluid
.
layers
.
data
(
name
=
'ptable'
,
shape
=
[
3
],
dtype
=
'int64'
)
pcode
=
fluid
.
layers
.
data
(
name
=
'pcode'
,
shape
=
[
3
],
dtype
=
'int64'
)
path_table
=
fluid
.
layers
.
data
(
name
=
'path_table'
,
shape
=
[
3
],
dtype
=
'int64'
)
path_code
=
fluid
.
layers
.
data
(
name
=
'path_code'
,
shape
=
[
3
],
dtype
=
'int64'
)
label
=
fluid
.
layers
.
data
(
name
=
'label'
,
shape
=
[
1
],
dtype
=
'int64'
)
data_list
=
[
input_word
,
p
table
,
p
code
,
label
]
data_list
=
[
input_word
,
p
ath_table
,
path_
code
,
label
]
emb
=
fluid
.
layers
.
embedding
(
input
=
input_word
,
...
...
@@ -218,9 +221,9 @@ class TestHSigmoidOpWithSparseGrad(unittest.TestCase):
label
=
label
,
bias_attr
=
True
,
non_leaf_num
=
3
,
p
table
=
p
table
,
p
code
=
p
code
,
is_c
ostu
m
=
True
,
p
ath_table
=
path_
table
,
p
ath_code
=
path_
code
,
is_c
usto
m
=
True
,
is_sparse
=
is_sparse
)
avg_cost
=
fluid
.
layers
.
reduce_mean
(
cost
)
...
...
@@ -232,8 +235,8 @@ class TestHSigmoidOpWithSparseGrad(unittest.TestCase):
start_up
=
fluid
.
default_startup_program
()
start_up
.
random_seed
=
1
# Fix random seed
x
=
np
.
arange
(
6
).
reshape
(
6
)
ptable
=
np
.
array
([(
1
,
2
,
-
1
),
(
1
,
2
,
-
1
)])
pcode
=
np
.
array
([(
1
,
0
,
-
1
),
(
0
,
0
,
-
1
)])
p
ath_
table
=
np
.
array
([(
1
,
2
,
-
1
),
(
1
,
2
,
-
1
)])
p
ath_
code
=
np
.
array
([(
1
,
0
,
-
1
),
(
0
,
0
,
-
1
)])
label
=
np
.
array
([
1
,
4
])
loss
,
data_list
=
self
.
hs_net_conf
(
is_sparse
)
...
...
@@ -248,8 +251,8 @@ class TestHSigmoidOpWithSparseGrad(unittest.TestCase):
exe
.
run
(
start_up
)
result
=
list
()
for
i
in
range
(
10
):
data
=
[([[
x
[
i
%
2
]]],
[
list
(
ptable
[
i
%
2
])],
[
list
(
pcode
[
i
%
2
])],
[
label
[
i
%
2
]])]
data
=
[([[
x
[
i
%
2
]]],
[
list
(
p
ath_
table
[
i
%
2
])],
[
list
(
p
ath_
code
[
i
%
2
])],
[
label
[
i
%
2
]])]
loss_val
=
exe
.
run
(
main_program
,
feed
=
feeder
.
feed
(
data
),
...
...
@@ -273,24 +276,24 @@ class TestHSigmoidOpWithCostumTree(OpTest):
w
=
np
.
random
.
random
(
(
num_classes
-
1
,
feature_size
)).
astype
(
"float32"
)
*
2
label
=
np
.
array
([
0
,
1
,
4
,
5
])
ptable
=
np
.
array
(
p
ath_
table
=
np
.
array
(
[(
0
,
2
,
-
1
,
-
1
,
-
1
),
(
0
,
1
,
3
,
-
1
,
-
1
),
(
0
,
1
,
4
,
-
1
,
-
1
),
(
0
,
2
,
-
1
,
-
1
,
-
1
)])
#np.array to store 1,2,5,6s' non-leaf path(root -> leaf)
pcode
=
np
.
array
([(
0
,
0
,
-
1
,
-
1
,
-
1
),
(
1
,
1
,
1
,
-
1
,
-
1
),
(
p
ath_
code
=
np
.
array
([(
0
,
0
,
-
1
,
-
1
,
-
1
),
(
1
,
1
,
1
,
-
1
,
-
1
),
(
1
,
0
,
0
,
-
1
,
-
1
),
(
0
,
1
,
-
1
,
-
1
,
-
1
)])
#np.array to store
bias
=
np
.
random
.
random
((
num_classes
-
1
,
1
)).
astype
(
"float32"
)
self
.
attrs
=
{
'num_classes'
:
num_classes
,
'is_sparse'
:
False
}
self
.
inputs
=
{
'X'
:
x
,
'W'
:
w
,
'PTable'
:
ptable
,
'PathCode'
:
pcode
,
'PTable'
:
p
ath_
table
,
'PathCode'
:
p
ath_
code
,
'Label'
:
label
,
'Bias'
:
bias
}
pre_output
,
out
=
hsigmoidWithCustomTree
(
x
,
w
,
p
table
,
pcode
,
label
,
bias
,
num_classes
)
pre_output
,
out
=
hsigmoidWithCustomTree
(
x
,
w
,
p
ath_table
,
path_code
,
label
,
bias
,
num_classes
)
self
.
outputs
=
{
'PreOut'
:
pre_output
,
'Out'
:
out
}
def
test_check_output
(
self
):
...
...
@@ -310,26 +313,26 @@ class TestHSigmoidOpWithCostumTreeWithoutBias(OpTest):
w
=
np
.
random
.
random
(
(
num_classes
-
1
,
feature_size
)).
astype
(
"float32"
)
*
2
label
=
np
.
array
([
0
,
1
,
4
,
5
])
ptable
=
np
.
array
(
p
ath_
table
=
np
.
array
(
[(
0
,
2
,
-
1
,
-
1
,
-
1
),
(
0
,
1
,
3
,
-
1
,
-
1
),
(
0
,
1
,
4
,
-
1
,
-
1
),
(
0
,
2
,
-
1
,
-
1
,
-
1
)])
#np.array to store 1,2,5,6s' non-leaf path(root -> leaf)
pcode
=
np
.
array
([(
0
,
0
,
-
1
,
-
1
,
-
1
),
(
1
,
1
,
1
,
-
1
,
-
1
),
(
p
ath_
code
=
np
.
array
([(
0
,
0
,
-
1
,
-
1
,
-
1
),
(
1
,
1
,
1
,
-
1
,
-
1
),
(
1
,
0
,
0
,
-
1
,
-
1
),
(
0
,
1
,
-
1
,
-
1
,
-
1
)])
#np.array to store
# bias = np.random.random((num_classes - 1, 1)).astype("float32")
self
.
attrs
=
{
'num_classes'
:
num_classes
,
'is_sparse'
:
False
}
self
.
inputs
=
{
'X'
:
x
,
'W'
:
w
,
'PTable'
:
ptable
,
'PathCode'
:
pcode
,
'PTable'
:
p
ath_
table
,
'PathCode'
:
p
ath_
code
,
'Label'
:
label
,
}
pre_output
,
out
=
hsigmoidWithCustomTree
(
x
=
x
,
w
=
w
,
p
table
=
p
table
,
p
code
=
p
code
,
p
ath_table
=
path_
table
,
p
ath_code
=
path_
code
,
label
=
label
,
bias
=
None
,
num_classes
=
num_classes
)
...
...
python/paddle/fluid/tests/unittests/test_layers.py
浏览文件 @
c469334c
...
...
@@ -190,16 +190,18 @@ class TestBook(unittest.TestCase):
with
program_guard
(
program2
):
x2
=
layers
.
data
(
name
=
'x2'
,
shape
=
[
4
,
8
],
dtype
=
'float32'
)
y2
=
layers
.
data
(
name
=
'y2'
,
shape
=
[
4
],
dtype
=
'int64'
)
ptable
=
layers
.
data
(
name
=
'ptable'
,
shape
=
[
4
,
6
],
dtype
=
'int64'
)
pcode
=
layers
.
data
(
name
=
'pcode'
,
shape
=
[
4
,
6
],
dtype
=
'int64'
)
path_table
=
layers
.
data
(
name
=
'path_table'
,
shape
=
[
4
,
6
],
dtype
=
'int64'
)
path_code
=
layers
.
data
(
name
=
'path_code'
,
shape
=
[
4
,
6
],
dtype
=
'int64'
)
self
.
assertIsNotNone
(
layers
.
hsigmoid
(
input
=
x2
,
label
=
y2
,
non_leaf_num
=
6
,
p
table
=
p
table
,
p
code
=
p
code
,
is_c
ostu
m
=
True
))
p
ath_table
=
path_
table
,
p
ath_code
=
path_
code
,
is_c
usto
m
=
True
))
print
(
str
(
program2
))
def
test_sequence_expand
(
self
):
...
...
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