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
TonyTonyFun
Paddle
提交
88e27a07
P
Paddle
项目概览
TonyTonyFun
/
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 >>
未验证
提交
88e27a07
编写于
7月 01, 2022
作者:
光明和真理
提交者:
GitHub
7月 01, 2022
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
[MLU] add mlu kernel for fill_constant_batch_size_like (#43820)
上级
3a59ede9
变更
2
隐藏空白更改
内联
并排
Showing
2 changed file
with
318 addition
and
0 deletion
+318
-0
paddle/fluid/operators/fill_constant_batch_size_like_op_mlu.cc
...e/fluid/operators/fill_constant_batch_size_like_op_mlu.cc
+99
-0
python/paddle/fluid/tests/unittests/mlu/test_fill_constant_batch_size_like_op_mlu.py
...nittests/mlu/test_fill_constant_batch_size_like_op_mlu.py
+219
-0
未找到文件。
paddle/fluid/operators/fill_constant_batch_size_like_op_mlu.cc
0 → 100644
浏览文件 @
88e27a07
/* Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/operators/fill_constant_op.h"
#include "paddle/fluid/operators/mlu/mlu_baseop.h"
#include "paddle/fluid/operators/utils.h"
namespace
paddle
{
namespace
operators
{
template
<
typename
T
>
class
FillConstantBatchSizeLikeOpMLUKernel
:
public
framework
::
OpKernel
<
T
>
{
public:
void
Compute
(
const
framework
::
ExecutionContext
&
ctx
)
const
override
{
auto
data_type
=
static_cast
<
framework
::
proto
::
VarType
::
Type
>
(
ctx
.
Attr
<
int
>
(
"dtype"
));
auto
float_value
=
ctx
.
Attr
<
float
>
(
"value"
);
auto
str_value
=
ctx
.
Attr
<
std
::
string
>
(
"str_value"
);
auto
force_cpu
=
ctx
.
Attr
<
bool
>
(
"force_cpu"
);
auto
*
out
=
ctx
.
Output
<
Tensor
>
(
"Out"
);
auto
*
in
=
ctx
.
Input
<
framework
::
LoDTensor
>
(
"Input"
);
if
(
in
->
lod
().
size
()
&&
ctx
.
Attr
<
int
>
(
"input_dim_idx"
)
==
0
)
{
// set the correct batch size for the LoDTensor.
auto
odims
=
out
->
dims
();
int
output_dim_idx
=
ctx
.
Attr
<
int
>
(
"output_dim_idx"
);
odims
[
output_dim_idx
]
=
static_cast
<
int
>
(
in
->
lod
().
back
().
size
())
-
1
;
out
->
mutable_data
<
T
>
(
odims
,
ctx
.
GetPlace
());
}
T
value
;
if
(
str_value
.
empty
())
{
value
=
static_cast
<
T
>
(
float_value
);
}
else
{
// handle NaN/Inf first, which cannot be read from stream.
if
(
str_value
==
"inf"
)
{
value
=
static_cast
<
T
>
(
std
::
numeric_limits
<
double
>::
infinity
());
}
else
if
(
str_value
==
"-inf"
)
{
value
=
static_cast
<
T
>
(
-
std
::
numeric_limits
<
double
>::
infinity
());
}
else
if
(
str_value
==
"nan"
)
{
value
=
static_cast
<
T
>
(
std
::
numeric_limits
<
double
>::
quiet_NaN
());
}
else
{
std
::
stringstream
convert_stream
(
str_value
);
if
(
std
::
is_same
<
int64_t
,
T
>::
value
)
{
int64_t
tmp_value
;
convert_stream
>>
tmp_value
;
value
=
static_cast
<
T
>
(
tmp_value
);
}
else
{
double
tmp_value
;
convert_stream
>>
tmp_value
;
value
=
static_cast
<
T
>
(
tmp_value
);
}
}
}
platform
::
DeviceContextPool
&
pool
=
platform
::
DeviceContextPool
::
Instance
();
bool
cpu_place
=
force_cpu
||
ctx
.
GetPlace
()
==
platform
::
CPUPlace
();
if
(
cpu_place
)
{
auto
&
dev_ctx
=
*
pool
.
Get
(
platform
::
CPUPlace
());
phi
::
funcs
::
SetConstant
<
platform
::
CPUDeviceContext
,
T
>
functor
;
out
->
mutable_data
(
platform
::
CPUPlace
(),
framework
::
TransToPhiDataType
(
data_type
));
functor
(
reinterpret_cast
<
const
platform
::
CPUDeviceContext
&>
(
dev_ctx
),
out
,
static_cast
<
T
>
(
value
));
}
else
{
out
->
mutable_data
(
ctx
.
GetPlace
(),
framework
::
TransToPhiDataType
(
data_type
));
const
T
*
value_data
=
&
value
;
cnnlPointerMode_t
pointer_mode
=
CNNL_POINTER_MODE_HOST
;
MLUCnnlTensorDesc
output_desc
(
*
out
);
MLUCnnl
::
Fill
(
ctx
,
pointer_mode
,
value_data
,
output_desc
.
get
(),
GetBasePtr
(
out
));
}
}
};
}
// namespace operators
}
// namespace paddle
namespace
ops
=
paddle
::
operators
;
namespace
plat
=
paddle
::
platform
;
REGISTER_OP_MLU_KERNEL
(
fill_constant_batch_size_like
,
ops
::
FillConstantBatchSizeLikeOpMLUKernel
<
int
>
,
ops
::
FillConstantBatchSizeLikeOpMLUKernel
<
float
>
,
ops
::
FillConstantBatchSizeLikeOpMLUKernel
<
plat
::
float16
>
);
python/paddle/fluid/tests/unittests/mlu/test_fill_constant_batch_size_like_op_mlu.py
0 → 100644
浏览文件 @
88e27a07
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from
__future__
import
print_function
import
sys
sys
.
path
.
append
(
".."
)
import
paddle
import
paddle.fluid.core
as
core
from
paddle.static
import
program_guard
,
Program
import
paddle.compat
as
cpt
import
unittest
import
numpy
as
np
from
op_test
import
OpTest
from
paddle.fluid.framework
import
convert_np_dtype_to_dtype_
paddle
.
enable_static
()
def
fill_constant_batch_size_like
(
input
,
shape
,
value
,
data_type
,
input_dim_idx
=
0
,
output_dim_idx
=
0
,
force_cpu
=
False
):
return
paddle
.
fluid
.
layers
.
fill_constant_batch_size_like
(
input
,
shape
,
data_type
,
value
,
input_dim_idx
,
output_dim_idx
,
force_cpu
)
class
TestFillConstantBatchSizeLike
(
OpTest
):
def
setUp
(
self
):
self
.
place
=
paddle
.
device
.
MLUPlace
(
0
)
self
.
__class__
.
use_mlu
=
True
self
.
op_type
=
"fill_constant_batch_size_like"
self
.
init_shape
()
self
.
init_value
()
self
.
init_dtype
()
self
.
init_force_cpu
()
self
.
init_dim_idx
()
self
.
inputs
=
{
'Input'
:
np
.
random
.
random
(
self
.
input_shape
).
astype
(
"float32"
)
}
self
.
attrs
=
{
'shape'
:
self
.
shape
,
'value'
:
self
.
value
,
'str_value'
:
self
.
str_value
,
'dtype'
:
self
.
dtype
,
'force_cpu'
:
self
.
force_cpu
,
'input_dim_idx'
:
self
.
input_dim_idx
,
'output_dim_idx'
:
self
.
output_dim_idx
}
self
.
outputs
=
{
'Out'
:
np
.
full
(
self
.
output_shape
,
self
.
output_value
,
self
.
output_dtype
)
}
def
init_shape
(
self
):
self
.
input_shape
=
[
4
,
5
]
self
.
shape
=
[
123
,
92
]
self
.
output_shape
=
(
4
,
92
)
def
init_value
(
self
):
self
.
value
=
3.8
self
.
str_value
=
''
self
.
output_value
=
3.8
def
init_dtype
(
self
):
self
.
dtype
=
core
.
VarDesc
.
VarType
.
FP32
self
.
output_dtype
=
np
.
float32
def
init_force_cpu
(
self
):
self
.
force_cpu
=
False
def
init_dim_idx
(
self
):
self
.
input_dim_idx
=
0
self
.
output_dim_idx
=
0
def
test_check_output
(
self
):
self
.
check_output_with_place
(
self
.
place
)
class
TestFillConstantBatchSizeLike2
(
TestFillConstantBatchSizeLike
):
def
init_shape
(
self
):
# test shape
self
.
input_shape
=
[
4
,
5
,
6
,
7
]
self
.
shape
=
[
10
,
123
,
92
]
self
.
output_shape
=
(
4
,
123
,
92
)
class
TestFillConstantBatchSizeLike3
(
TestFillConstantBatchSizeLike
):
def
init_value
(
self
):
# use 'str_value' rather than 'value'
self
.
value
=
3.8
self
.
str_value
=
'4.5'
self
.
output_value
=
4.5
class
TestFillConstantBatchSizeLike4
(
TestFillConstantBatchSizeLike
):
def
init_value
(
self
):
# str_value = 'inf'
self
.
value
=
3.8
self
.
str_value
=
'inf'
self
.
output_value
=
float
(
'inf'
)
class
TestFillConstantBatchSizeLike5
(
TestFillConstantBatchSizeLike
):
def
init_value
(
self
):
# str_value = '-inf'
self
.
value
=
3.8
self
.
str_value
=
'-inf'
self
.
output_value
=
-
float
(
'inf'
)
class
TestFillConstantBatchSizeLike6
(
TestFillConstantBatchSizeLike
):
def
init_dtype
(
self
):
self
.
dtype
=
core
.
VarDesc
.
VarType
.
FP16
self
.
output_dtype
=
np
.
float16
def
test_check_output
(
self
):
self
.
check_output_with_place
(
self
.
place
,
atol
=
1e-2
)
class
TestFillConstantBatchSizeLike7
(
TestFillConstantBatchSizeLike
):
def
init_dtype
(
self
):
self
.
dtype
=
core
.
VarDesc
.
VarType
.
INT32
self
.
output_dtype
=
np
.
int32
class
TestFillConstantBatchSizeLike8
(
TestFillConstantBatchSizeLike
):
def
init_force_cpu
(
self
):
self
.
force_cpu
=
True
class
TestFillConstantBatchSizeLike9
(
TestFillConstantBatchSizeLike
):
def
init_shape
(
self
):
self
.
input_shape
=
[
4
,
5
]
self
.
shape
=
[
123
,
92
]
self
.
output_shape
=
(
123
,
4
)
def
init_dim_idx
(
self
):
self
.
input_dim_idx
=
0
self
.
output_dim_idx
=
1
class
TestFillConstantBatchSizeLikeLodTensor
(
TestFillConstantBatchSizeLike
):
# test LodTensor
def
setUp
(
self
):
self
.
place
=
paddle
.
device
.
MLUPlace
(
0
)
self
.
__class__
.
use_mlu
=
True
self
.
op_type
=
"fill_constant_batch_size_like"
self
.
init_shape
()
self
.
init_value
()
self
.
init_dtype
()
self
.
init_force_cpu
()
self
.
init_dim_idx
()
lod
=
[[
3
,
2
,
5
]]
self
.
inputs
=
{
'Input'
:
(
np
.
random
.
random
(
self
.
input_shape
).
astype
(
"float32"
),
lod
)
}
self
.
attrs
=
{
'shape'
:
self
.
shape
,
'value'
:
self
.
value
,
'str_value'
:
self
.
str_value
,
'dtype'
:
self
.
dtype
,
'force_cpu'
:
self
.
force_cpu
,
'input_dim_idx'
:
self
.
input_dim_idx
,
'output_dim_idx'
:
self
.
output_dim_idx
}
self
.
outputs
=
{
'Out'
:
np
.
full
(
self
.
output_shape
,
self
.
output_value
,
self
.
output_dtype
)
}
def
init_shape
(
self
):
self
.
input_shape
=
[
10
,
20
]
self
.
shape
=
[
123
,
92
]
self
.
output_shape
=
(
3
,
92
)
class
TestFillConstantBatchSizeLikeLodTensor2
(
TestFillConstantBatchSizeLikeLodTensor
):
# test LodTensor with 'input_dim_idx' != 0
def
init_shape
(
self
):
self
.
input_shape
=
[
10
,
20
]
self
.
shape
=
[
123
,
92
]
self
.
output_shape
=
(
20
,
92
)
def
init_dim_idx
(
self
):
self
.
input_dim_idx
=
1
self
.
output_dim_idx
=
0
if
__name__
==
"__main__"
:
unittest
.
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