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
ae0ea541
P
Paddle
项目概览
PaddlePaddle
/
Paddle
大约 1 年 前同步成功
通知
2298
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
ae0ea541
编写于
1月 24, 2018
作者:
C
chengduoZH
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
fix unit test
上级
ca017719
变更
2
隐藏空白更改
内联
并排
Showing
2 changed file
with
215 addition
and
57 deletion
+215
-57
paddle/operators/layer_norm_op.cc
paddle/operators/layer_norm_op.cc
+7
-4
python/paddle/v2/fluid/tests/test_layer_norm_op.py
python/paddle/v2/fluid/tests/test_layer_norm_op.py
+208
-53
未找到文件。
paddle/operators/layer_norm_op.cc
浏览文件 @
ae0ea541
...
@@ -233,13 +233,13 @@ class LayerNormGradKernel<platform::CPUDeviceContext, T>
...
@@ -233,13 +233,13 @@ class LayerNormGradKernel<platform::CPUDeviceContext, T>
if
(
d_x
)
{
if
(
d_x
)
{
d_x
->
mutable_data
<
T
>
(
ctx
.
GetPlace
());
d_x
->
mutable_data
<
T
>
(
ctx
.
GetPlace
());
auto
d_x_map
=
EigenMatrixMapRowMajor
<
T
>
(
d_x
->
data
<
T
>
(),
left
,
right
);
auto
d_x_map
=
EigenMatrixMapRowMajor
<
T
>
(
d_x
->
data
<
T
>
(),
left
,
right
);
auto
triple_product
=
[](
T
ele
)
{
return
ele
*
ele
*
ele
;
};
auto
triple_product
=
[](
T
ele
)
{
return
ele
*
ele
;
};
auto
neg_inv_std
=
[](
T
ele
)
{
return
T
(
-
1.0
)
*
std
::
sqrt
(
1
/
ele
);
};
auto
neg_inv_std
=
[](
T
ele
)
{
return
-
std
::
sqrt
(
1
/
ele
);
};
auto
inv_std_scale_func
=
[
scale_data
](
T
ele
)
{
auto
inv_std_scale_func
=
[
scale_data
](
T
ele
)
{
return
std
::
sqrt
(
1
/
ele
)
*
scale_data
;
return
std
::
sqrt
(
1
/
ele
)
*
scale_data
;
};
};
auto
neg_inv_std_scale_func
=
[
scale_data
](
T
ele
)
{
auto
neg_inv_std_scale_func
=
[
scale_data
](
T
ele
)
{
return
T
(
-
1.0
)
*
std
::
sqrt
(
1
/
ele
)
*
scale_data
;
return
-
std
::
sqrt
(
1
/
ele
)
*
scale_data
;
};
};
// dy_dx
// dy_dx
auto
dx_end
=
var_map
.
unaryExpr
(
inv_std_scale_func
)
auto
dx_end
=
var_map
.
unaryExpr
(
inv_std_scale_func
)
...
@@ -260,10 +260,13 @@ class LayerNormGradKernel<platform::CPUDeviceContext, T>
...
@@ -260,10 +260,13 @@ class LayerNormGradKernel<platform::CPUDeviceContext, T>
auto
dvar_end
=
var_map
.
unaryExpr
(
neg_inv_std
)
auto
dvar_end
=
var_map
.
unaryExpr
(
neg_inv_std
)
.
unaryExpr
(
triple_product
)
.
unaryExpr
(
triple_product
)
.
cwiseProduct
(
dvar_end_0
);
.
cwiseProduct
(
dvar_end_0
);
auto
dx_var
=
(
1.0
f
/
right
)
*
auto
dx_var
=
(
T
(
1.0
)
/
right
)
*
(
x_map
-
mean_map
.
replicate
(
1
,
right
))
(
x_map
-
mean_map
.
replicate
(
1
,
right
))
.
cwiseProduct
(
dvar_end
.
replicate
(
1
,
right
));
.
cwiseProduct
(
dvar_end
.
replicate
(
1
,
right
));
// d_x = (1. / N) * scale * inv_var * (N * d_y - np.sum(d_y, axis=0)
// - (X - mean) * inv_var * inv_var * np.sum(d_y * (X - mean), axis=0))
d_x_map
=
dx_end
+
dx_mean
+
dx_var
;
d_x_map
=
dx_end
+
dx_mean
+
dx_var
;
}
}
}
}
...
...
python/paddle/v2/fluid/tests/test_layer_norm_op.py
浏览文件 @
ae0ea541
...
@@ -15,66 +15,221 @@
...
@@ -15,66 +15,221 @@
import
unittest
import
unittest
import
numpy
as
np
import
numpy
as
np
from
operator
import
mul
from
op_test
import
OpTest
from
op_test
import
OpTest
import
paddle.v2.fluid.core
as
core
from
paddle.v2.fluid.op
import
Operator
from
paddle.v2.fluid.framework
import
grad_var_name
def
layer_norm_naive
(
x
,
scale
,
beta
,
epsilon
):
def
get_backward_op
(
scope
,
op
,
no_grad_set
):
n
,
c
,
h
,
w
=
x
.
shape
backward_op
=
core
.
Operator
.
backward
(
op
,
no_grad_set
)
mean
=
np
.
mean
(
x
,
axis
=
(
1
,
2
,
3
))
for
input
in
backward_op
.
input_vars
():
var
=
np
.
var
(
x
,
axis
=
(
1
,
2
,
3
))
+
epsilon
var
=
scope
.
var
(
input
)
output
=
scale
*
np
.
divide
((
x
-
mean
.
reshape
([
n
,
1
,
1
,
1
])),
var
.
get_tensor
()
(
np
.
sqrt
(
var
)).
reshape
([
n
,
1
,
1
,
1
]))
+
beta
for
output
in
backward_op
.
output_vars
():
var
=
scope
.
var
(
output
)
var
.
get_tensor
()
return
backward_op
def
_reference_layer_norm_naive
(
x
,
scale
,
beta
,
epsilon
):
old_shape
=
x
.
shape
N
=
x
.
shape
[
0
]
D
=
reduce
(
mul
,
old_shape
,
1
)
/
N
x
.
shape
=
[
N
,
D
]
mean
=
np
.
mean
(
x
,
axis
=
1
)
var
=
np
.
var
(
x
,
axis
=
1
)
+
epsilon
output
=
scale
*
np
.
divide
((
x
-
mean
.
reshape
([
N
,
1
])),
(
np
.
sqrt
(
var
)).
reshape
([
N
,
1
]))
+
beta
output
.
shape
=
old_shape
return
output
,
mean
,
var
return
output
,
mean
,
var
def
_reference_layer_norm_grad
(
x
,
grad_y
,
scale
,
mean
,
var
,
epsilon
):
x_shape
=
x
.
shape
N
=
x_shape
[
0
]
D
=
reduce
(
mul
,
x_shape
,
1
)
/
N
grad_y
.
shape
=
[
N
,
D
]
x
.
shape
=
[
N
,
D
]
grad_offset
=
np
.
sum
(
grad_y
)
mean
.
shape
=
[
N
,
1
]
var
.
shape
=
[
N
,
1
]
grad_scale
=
np
.
sum
(((
x
-
mean
)
*
np
.
sqrt
(
1
/
var
))
*
grad_y
)
dx_end
=
np
.
sqrt
(
1.0
/
var
)
*
grad_y
d_mean_0
=
np
.
sum
(
-
np
.
sqrt
(
1.0
/
var
)
*
grad_y
,
axis
=
1
).
reshape
([
N
,
1
])
d_mean_1
=
np
.
sum
(
-
1.0
/
var
*
(
x
-
mean
)
*
grad_y
,
axis
=
1
).
reshape
(
[
N
,
1
])
*
(
-
1.0
/
D
*
np
.
sqrt
(
1.0
/
var
)
*
np
.
sum
(
x
-
mean
,
axis
=
1
).
reshape
([
N
,
1
])).
reshape
([
N
,
1
])
d_mean
=
1.0
/
D
*
(
d_mean_0
+
d_mean_1
)
d_std
=
np
.
sum
(
-
1.0
/
var
*
(
x
-
mean
)
*
grad_y
,
axis
=
1
).
reshape
([
N
,
1
])
*
(
1.0
/
D
*
np
.
sqrt
(
1.0
/
var
).
reshape
([
N
,
1
])
*
(
x
-
mean
))
grad_x
=
scale
*
(
dx_end
+
d_mean
+
d_std
)
grad_y
.
shape
=
x_shape
x
.
shape
=
x_shape
return
grad_x
,
grad_scale
,
grad_offset
def
create_or_get_tensor
(
scope
,
var_name
,
var
,
place
):
tensor
=
scope
.
var
(
var_name
).
get_tensor
()
if
var
is
not
None
:
assert
isinstance
(
var
,
np
.
ndarray
)
tensor
.
set_lod
([[]])
tensor
.
set_dims
(
var
.
shape
)
tensor
.
set
(
var
,
place
)
return
tensor
def
set_output_grad
(
scope
,
outputs
,
place
,
feed_dict
=
None
):
def
__set_tensor__
(
name
,
data
=
None
):
out_tensor
=
scope
.
find_var
(
name
).
get_tensor
()
grad_tensor
=
scope
.
var
(
grad_var_name
(
name
)).
get_tensor
()
out_dtype
=
out_tensor
.
dtype
()
if
data
is
None
:
if
out_dtype
==
core
.
DataType
.
FP64
:
data
=
np
.
ones
(
out_tensor
.
shape
(),
dtype
=
np
.
float64
)
elif
out_dtype
==
core
.
DataType
.
FP32
:
data
=
np
.
ones
(
out_tensor
.
shape
(),
dtype
=
np
.
float32
)
else
:
raise
ValueError
(
"Not supported data type "
+
str
(
out_dtype
))
grad_tensor
.
set
(
data
,
place
)
for
output
in
outputs
:
data
=
None
if
output
in
feed_dict
:
data
=
feed_dict
[
output
]
__set_tensor__
(
output
,
data
)
class
TestLayerNormdOp
(
OpTest
):
class
TestLayerNormdOp
(
OpTest
):
def
setUp
(
self
):
def
__assert_close
(
self
,
tensor
,
np_array
,
msg
,
atol
=
1e-4
):
self
.
init_test_case
()
self
.
assertTrue
(
np
.
allclose
(
input
=
np
.
random
.
random
(
self
.
input_size
).
astype
(
"float32"
)
np
.
array
(
tensor
).
reshape
(
np_array
.
shape
),
np_array
,
atol
=
atol
),
self
.
inputs
=
{
msg
)
'X'
:
input
,
'Scale'
:
np
.
array
([
self
.
scale
]).
astype
(
"float32"
),
def
__assert_grad_close
(
self
,
'Bias'
:
np
.
array
([
self
.
bias
]).
astype
(
"float32"
)
tensor
,
}
np_array
,
output
,
mean
,
var
=
layer_norm_naive
(
input
,
self
.
scale
,
self
.
bias
,
name
,
self
.
epsilon
)
place
,
self
.
outputs
=
{
'Y'
:
output
,
'Mean'
:
mean
,
'Variance'
:
var
}
max_relative_error
=
0.02
):
a
=
np
.
array
(
tensor
).
reshape
(
np_array
.
shape
)
def
test_check_output
(
self
):
b
=
np_array
self
.
check_output
()
abs_a
=
np
.
abs
(
a
)
abs_a
[
abs_a
<
1e-5
]
=
1
# def test_check_grad(self):
# self.check_grad(
diff_mat
=
np
.
abs
(
a
-
b
)
/
abs_a
# ['Scale', 'Bias', 'X'], ['Y', 'Mean', 'Variance'],
max_diff
=
np
.
max
(
diff_mat
)
# max_relative_error=0.02)
def
err_msg
():
def
test_check_grad_no_x
(
self
):
offset
=
np
.
argmax
(
diff_mat
>
max_relative_error
)
self
.
check_grad
(
return
(
"%s Variable %s max gradient diff %f over limit %f, "
[
'Scale'
,
'Bias'
],
[
'Y'
,
'Mean'
,
'Variance'
],
"the first error element is %d, %f, %f"
)
%
(
max_relative_error
=
0.02
,
"Gradient Check On %s"
%
str
(
place
),
name
,
max_diff
,
no_grad_set
=
set
([
'X'
]))
max_relative_error
,
offset
,
a
.
flatten
()[
offset
],
b
.
flatten
()[
offset
])
# def test_check_grad_no_scale(self):
# self.check_grad(
self
.
assertLessEqual
(
max_diff
,
max_relative_error
,
err_msg
())
# ['Bias','X'],
# 'Y',
def
test_forward_backward
(
self
):
# max_relative_error=0.02,
def
test_with_place
(
place
,
shape
):
# no_grad_set=set(['Scale']))
# attr
#
epsilon
=
0.00001
# def test_check_grad_no_bias(self):
x_shape
=
shape
# self.check_grad(
scale_shape
=
[
1
]
# ['Scale','X'],
# 'Y',
x_val
=
np
.
random
.
random_sample
(
x_shape
).
astype
(
np
.
float32
)
# max_relative_error=0.02,
scale_val
=
np
.
random
.
random_sample
(
scale_shape
).
astype
(
np
.
float32
)
# no_grad_set=set(['Bias']))
bias_val
=
np
.
random
.
random_sample
(
scale_shape
).
astype
(
np
.
float32
)
def
init_test_case
(
self
):
# run forward
self
.
op_type
=
"layer_norm"
y_out
,
saved_mean
,
var_ref
=
_reference_layer_norm_naive
(
self
.
input_size
=
[
2
,
3
,
4
,
5
]
x_val
,
scale_val
,
bias_val
,
epsilon
)
self
.
scale
=
0.21
self
.
bias
=
0.1
# for gradient test
self
.
epsilon
=
0.00001
# y_grad = np.ones(x_shape).astype(np.float32) * 0.00277778
y_grad
=
np
.
random
.
random_sample
(
x_shape
).
astype
(
np
.
float32
)
x_grad_ref
,
scale_grad_ref
,
bias_grad_ref
=
_reference_layer_norm_grad
(
x_val
,
y_grad
,
scale_val
,
saved_mean
,
var_ref
,
epsilon
)
scope
=
core
.
Scope
()
# create input
x_tensor
=
create_or_get_tensor
(
scope
,
"X"
,
x_val
,
place
)
scale_tensor
=
create_or_get_tensor
(
scope
,
"Scale"
,
scale_val
,
place
)
bias_tensor
=
create_or_get_tensor
(
scope
,
"Bias"
,
bias_val
,
place
)
# create output
y_tensor
=
create_or_get_tensor
(
scope
,
"Y"
,
None
,
place
)
mean_tensor
=
create_or_get_tensor
(
scope
,
"Mean"
,
None
,
place
)
variance_tensor
=
create_or_get_tensor
(
scope
,
"Variance"
,
None
,
place
)
layer_norm_op
=
Operator
(
"layer_norm"
,
# inputs
X
=
"X"
,
Scale
=
"Scale"
,
Bias
=
"Bias"
,
# outputs
Y
=
"Y"
,
Mean
=
"Mean"
,
Variance
=
"Variance"
,
# attrs
epsilon
=
epsilon
)
layer_norm_op
.
run
(
scope
,
place
)
# check forward result
if
isinstance
(
place
,
core
.
CUDAPlace
):
atol
=
5e-2
else
:
atol
=
1e-4
self
.
__assert_close
(
y_tensor
,
y_out
,
"Y"
,
atol
)
self
.
__assert_close
(
mean_tensor
,
saved_mean
,
"Mean"
,
atol
)
self
.
__assert_close
(
variance_tensor
,
var_ref
,
"Variance"
,
atol
)
# run backward
layer_norm_op_grad
=
get_backward_op
(
scope
,
layer_norm_op
,
set
())
set_output_grad
(
scope
,
[
"Y"
,
"Mean"
,
"Variance"
],
place
,
feed_dict
=
{
"Y"
:
y_grad
})
layer_norm_op_grad
.
run
(
scope
,
place
)
x_grad_tensor
=
create_or_get_tensor
(
scope
,
grad_var_name
(
"X"
),
None
,
place
)
scale_grad_tensor
=
create_or_get_tensor
(
scope
,
grad_var_name
(
"Scale"
),
None
,
place
)
bias_grad_tensor
=
create_or_get_tensor
(
scope
,
grad_var_name
(
"Bias"
),
None
,
place
)
# check gradient output
self
.
__assert_grad_close
(
x_grad_tensor
,
x_grad_ref
,
"x_grad"
,
place
)
self
.
__assert_grad_close
(
scale_grad_tensor
,
scale_grad_ref
,
"scale_grad"
,
place
)
self
.
__assert_grad_close
(
bias_grad_tensor
,
bias_grad_ref
,
"bias_grad"
,
place
)
places
=
[
core
.
CPUPlace
()]
if
core
.
is_compile_gpu
()
and
core
.
op_support_gpu
(
"layer_norm"
):
places
.
append
(
core
.
CUDAPlace
(
0
))
for
place
in
places
:
test_with_place
(
place
,
[
2
,
3
,
4
,
5
])
test_with_place
(
place
,
[
2
,
3
])
if
__name__
==
'__main__'
:
if
__name__
==
'__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