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
s920243400
PaddleDetection
提交
b41894d1
P
PaddleDetection
项目概览
s920243400
/
PaddleDetection
与 Fork 源项目一致
Fork自
PaddlePaddle / PaddleDetection
通知
2
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
PaddleDetection
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
提交
b41894d1
编写于
12月 09, 2017
作者:
S
sweetsky0901
浏览文件
操作
浏览文件
下载
差异文件
Merge branch 'develop' of
https://github.com/PaddlePaddle/Paddle
into detection_output
上级
fe177b62
5ba231d8
变更
18
显示空白变更内容
内联
并排
Showing
18 changed file
with
313 addition
and
110 deletion
+313
-110
paddle/capi/Main.cpp
paddle/capi/Main.cpp
+7
-0
paddle/capi/Matrix.cpp
paddle/capi/Matrix.cpp
+1
-1
paddle/capi/error.cpp
paddle/capi/error.cpp
+32
-0
paddle/capi/error.h
paddle/capi/error.h
+7
-0
paddle/capi/examples/model_inference/multi_thread/CMakeLists.txt
...capi/examples/model_inference/multi_thread/CMakeLists.txt
+25
-4
paddle/capi/examples/model_inference/multi_thread/main_gpu.c
paddle/capi/examples/model_inference/multi_thread/main_gpu.c
+113
-0
paddle/capi/main.h
paddle/capi/main.h
+7
-0
paddle/framework/op_desc.cc
paddle/framework/op_desc.cc
+1
-1
paddle/operators/concat_op.cc
paddle/operators/concat_op.cc
+8
-4
paddle/operators/cross_entropy_op.cc
paddle/operators/cross_entropy_op.cc
+1
-0
paddle/platform/device_context.cc
paddle/platform/device_context.cc
+0
-4
paddle/platform/device_context.h
paddle/platform/device_context.h
+0
-5
paddle/scripts/docker/build.sh
paddle/scripts/docker/build.sh
+3
-8
python/CMakeLists.txt
python/CMakeLists.txt
+6
-0
python/paddle/trainer_config_helpers/layers.py
python/paddle/trainer_config_helpers/layers.py
+30
-28
python/paddle/v2/fluid/layers.py
python/paddle/v2/fluid/layers.py
+2
-1
python/paddle/v2/fluid/tests/book/test_machine_translation.py
...on/paddle/v2/fluid/tests/book/test_machine_translation.py
+68
-52
python/setup.py.in
python/setup.py.in
+2
-2
未找到文件。
paddle/capi/Main.cpp
浏览文件 @
b41894d1
...
...
@@ -43,4 +43,11 @@ paddle_error paddle_init(int argc, char** argv) {
isInit
=
true
;
return
kPD_NO_ERROR
;
}
paddle_error
paddle_init_thread
()
{
if
(
FLAGS_use_gpu
)
{
hl_init
(
FLAGS_gpu_id
);
}
return
kPD_NO_ERROR
;
}
}
paddle/capi/Matrix.cpp
浏览文件 @
b41894d1
...
...
@@ -40,7 +40,7 @@ paddle_error paddle_matrix_destroy(paddle_matrix mat) {
paddle_error
paddle_matrix_set_row
(
paddle_matrix
mat
,
uint64_t
rowID
,
paddle_real
*
rowArray
)
{
if
(
mat
==
nullptr
)
return
kPD_NULLPTR
;
if
(
mat
==
nullptr
||
rowArray
==
nullptr
)
return
kPD_NULLPTR
;
auto
ptr
=
cast
(
mat
);
if
(
ptr
->
mat
==
nullptr
)
return
kPD_NULLPTR
;
if
(
rowID
>=
ptr
->
mat
->
getHeight
())
return
kPD_OUT_OF_RANGE
;
...
...
paddle/capi/error.cpp
0 → 100644
浏览文件 @
b41894d1
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
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 "error.h"
const
char
*
paddle_error_string
(
paddle_error
err
)
{
switch
(
err
)
{
case
kPD_NULLPTR
:
return
"nullptr error"
;
case
kPD_OUT_OF_RANGE
:
return
"out of range error"
;
case
kPD_PROTOBUF_ERROR
:
return
"protobuf error"
;
case
kPD_NOT_SUPPORTED
:
return
"not supported error"
;
case
kPD_UNDEFINED_ERROR
:
return
"undefined error"
;
default:
return
""
;
}
}
paddle/capi/error.h
浏览文件 @
b41894d1
...
...
@@ -15,6 +15,8 @@ limitations under the License. */
#ifndef __PADDLE_CAPI_ERROR_H__
#define __PADDLE_CAPI_ERROR_H__
#include "config.h"
/**
* Error Type for Paddle API.
*/
...
...
@@ -27,4 +29,9 @@ typedef enum {
kPD_UNDEFINED_ERROR
=
-
1
,
}
paddle_error
;
/**
* Error string for Paddle API.
*/
PD_API
const
char
*
paddle_error_string
(
paddle_error
err
);
#endif
paddle/capi/examples/model_inference/multi_thread/CMakeLists.txt
浏览文件 @
b41894d1
project
(
multi_thread
)
cmake_minimum_required
(
VERSION 2.8
)
aux_source_directory
(
. SRC_LIST
)
add_executable
(
${
PROJECT_NAME
}
${
SRC_LIST
}
)
find_package
(
Threads
)
if
(
NOT PADDLE_ROOT
)
set
(
PADDLE_ROOT $ENV{PADDLE_ROOT} CACHE PATH
"Paddle Path"
)
endif
()
if
(
PADDLE_ROOT
)
include_directories
(
${
PADDLE_ROOT
}
/include
)
link_directories
(
${
PADDLE_ROOT
}
/lib
)
endif
()
set
(
CPU_SRCS main.c
)
add_executable
(
${
PROJECT_NAME
}
${
CPU_SRCS
}
)
set_property
(
TARGET
${
PROJECT_NAME
}
PROPERTY C_STANDARD 99
)
target_link_libraries
(
${
PROJECT_NAME
}
-lpaddle_capi_shared
target_link_libraries
(
${
PROJECT_NAME
}
-lpaddle_capi_shared
${
CMAKE_THREAD_LIBS_INIT
}
)
find_package
(
CUDA QUIET
)
if
(
CUDA_FOUND
)
set
(
GPU_SRCS main_gpu.c
)
cuda_add_executable
(
${
PROJECT_NAME
}
_gpu
${
GPU_SRCS
}
)
set_property
(
TARGET
${
PROJECT_NAME
}
_gpu PROPERTY C_STANDARD 99
)
target_link_libraries
(
${
PROJECT_NAME
}
_gpu
-lpaddle_capi_shared
${
CMAKE_THREAD_LIBS_INIT
}
)
endif
(
CUDA_FOUND
)
paddle/capi/examples/model_inference/multi_thread/main_gpu.c
0 → 100644
浏览文件 @
b41894d1
#include <paddle/capi.h>
#include <pthread.h>
#include <time.h>
#include "../common/common.h"
#define CONFIG_BIN "./trainer_config.bin"
#define NUM_THREAD 4
#define NUM_ITER 1000
pthread_mutex_t
mutex
;
/*
* @brief It is an simple inference example that runs multi-threads on a GPU.
* Each thread holds it own local gradient_machine but shares the same
* parameters.
* If you want to run on different GPUs, you need to launch
* multi-processes or set trainer_count > 1.
*/
void
*
thread_main
(
void
*
gm_ptr
)
{
// Initialize the thread environment of Paddle.
CHECK
(
paddle_init_thread
());
paddle_gradient_machine
machine
=
(
paddle_gradient_machine
)(
gm_ptr
);
// Create input arguments.
paddle_arguments
in_args
=
paddle_arguments_create_none
();
// Create input matrix.
paddle_matrix
mat
=
paddle_matrix_create
(
/* sample_num */
1
,
/* size */
784
,
/* useGPU */
true
);
// Create output arguments.
paddle_arguments
out_args
=
paddle_arguments_create_none
();
// Create output matrix.
paddle_matrix
prob
=
paddle_matrix_create_none
();
// CPU buffer to cache the input and output.
paddle_real
*
cpu_input
=
(
paddle_real
*
)
malloc
(
784
*
sizeof
(
paddle_real
));
paddle_real
*
cpu_output
=
(
paddle_real
*
)
malloc
(
10
*
sizeof
(
paddle_real
));
for
(
int
iter
=
0
;
iter
<
NUM_ITER
;
++
iter
)
{
// There is only one input layer of this network.
CHECK
(
paddle_arguments_resize
(
in_args
,
1
));
CHECK
(
paddle_arguments_set_value
(
in_args
,
0
,
mat
));
for
(
int
i
=
0
;
i
<
784
;
++
i
)
{
cpu_input
[
i
]
=
rand
()
/
((
float
)
RAND_MAX
);
}
CHECK
(
paddle_matrix_set_value
(
mat
,
cpu_input
));
CHECK
(
paddle_gradient_machine_forward
(
machine
,
in_args
,
out_args
,
/* isTrain */
false
));
CHECK
(
paddle_arguments_get_value
(
out_args
,
0
,
prob
));
CHECK
(
paddle_matrix_get_value
(
prob
,
cpu_output
));
pthread_mutex_lock
(
&
mutex
);
printf
(
"Prob: "
);
for
(
int
i
=
0
;
i
<
10
;
++
i
)
{
printf
(
"%.2f "
,
cpu_output
[
i
]);
}
printf
(
"
\n
"
);
pthread_mutex_unlock
(
&
mutex
);
}
CHECK
(
paddle_matrix_destroy
(
prob
));
CHECK
(
paddle_arguments_destroy
(
out_args
));
CHECK
(
paddle_matrix_destroy
(
mat
));
CHECK
(
paddle_arguments_destroy
(
in_args
));
CHECK
(
paddle_gradient_machine_destroy
(
machine
));
free
(
cpu_input
);
free
(
cpu_output
);
return
NULL
;
}
int
main
()
{
// Initalize Paddle
char
*
argv
[]
=
{
"--use_gpu=True"
};
CHECK
(
paddle_init
(
1
,
(
char
**
)
argv
));
// Reading config binary file. It is generated by `convert_protobin.sh`
long
size
;
void
*
buf
=
read_config
(
CONFIG_BIN
,
&
size
);
// Create a gradient machine for inference.
paddle_gradient_machine
machine
;
CHECK
(
paddle_gradient_machine_create_for_inference
(
&
machine
,
buf
,
(
int
)
size
));
CHECK
(
paddle_gradient_machine_randomize_param
(
machine
));
// Loading parameter. Uncomment the following line and change the directory.
// CHECK(paddle_gradient_machine_load_parameter_from_disk(machine,
// "./some_where_to_params"));
srand
(
time
(
0
));
pthread_mutex_init
(
&
mutex
,
NULL
);
pthread_t
threads
[
NUM_THREAD
];
for
(
int
i
=
0
;
i
<
NUM_THREAD
;
++
i
)
{
paddle_gradient_machine
thread_local_machine
;
CHECK
(
paddle_gradient_machine_create_shared_param
(
machine
,
buf
,
size
,
&
thread_local_machine
));
pthread_create
(
&
threads
[
i
],
NULL
,
thread_main
,
thread_local_machine
);
}
for
(
int
i
=
0
;
i
<
NUM_THREAD
;
++
i
)
{
pthread_join
(
threads
[
i
],
NULL
);
}
pthread_mutex_destroy
(
&
mutex
);
return
0
;
}
paddle/capi/main.h
浏览文件 @
b41894d1
...
...
@@ -26,6 +26,13 @@ extern "C" {
*/
PD_API
paddle_error
paddle_init
(
int
argc
,
char
**
argv
);
/**
* Initialize the thread environment of Paddle.
* @note it is requisite for GPU runs but optional for CPU runs.
* For GPU runs, all threads will run on the same GPU devices.
*/
PD_API
paddle_error
paddle_init_thread
();
#ifdef __cplusplus
}
#endif
...
...
paddle/framework/op_desc.cc
浏览文件 @
b41894d1
...
...
@@ -59,7 +59,7 @@ class CompileTimeInferShapeContext : public InferShapeContext {
auto
*
in_var
=
block_
.
FindVarRecursive
(
Inputs
(
in
)[
i
]);
auto
*
out_var
=
block_
.
FindVarRecursive
(
Outputs
(
out
)[
j
]);
if
(
in_var
->
GetType
()
!=
VarDesc
::
LOD_TENSOR
)
{
VLOG
(
3
)
<<
"input "
<<
in
<<
"is not LodTensor"
;
VLOG
(
3
)
<<
"input "
<<
in
<<
"
is not LodTensor"
;
return
;
}
PADDLE_ENFORCE_EQ
(
in_var
->
GetType
(),
VarDesc
::
LOD_TENSOR
,
...
...
paddle/operators/concat_op.cc
浏览文件 @
b41894d1
...
...
@@ -41,14 +41,18 @@ class ConcatOp : public framework::OperatorWithKernel {
for
(
size_t
j
=
0
;
j
<
in_zero_dims_size
;
j
++
)
{
if
(
j
==
axis
)
{
out_dims
[
axis
]
+=
ins
[
i
][
j
];
continue
;
}
}
else
{
PADDLE_ENFORCE_EQ
(
out_dims
[
j
],
ins
[
i
][
j
],
"Input tensors should have the same "
"elements except the specify axis."
);
}
}
}
if
(
out_dims
[
axis
]
<
0
)
{
out_dims
[
axis
]
=
-
1
;
}
ctx
->
SetOutputDim
(
"Out"
,
out_dims
);
ctx
->
ShareLoD
(
"X"
,
/*->*/
"Out"
);
}
};
...
...
paddle/operators/cross_entropy_op.cc
浏览文件 @
b41894d1
...
...
@@ -95,6 +95,7 @@ class CrossEntropyGradientOp : public framework::OperatorWithKernel {
"Input(Label) should be 1."
);
}
ctx
->
SetOutputDim
(
framework
::
GradVarName
(
"X"
),
x_dims
);
ctx
->
ShareLoD
(
"X"
,
framework
::
GradVarName
(
"X"
));
}
protected:
...
...
paddle/platform/device_context.cc
浏览文件 @
b41894d1
...
...
@@ -122,10 +122,6 @@ Place CUDADeviceContext::GetPlace() const { return place_; }
void
CUDADeviceContext
::
Wait
()
const
{
PADDLE_ENFORCE
(
cudaStreamSynchronize
(
stream_
));
}
void
CUDADeviceContext
::
Finish
()
const
{
Wait
();
PADDLE_ENFORCE
(
cudaGetLastError
());
}
...
...
paddle/platform/device_context.h
浏览文件 @
b41894d1
...
...
@@ -46,8 +46,6 @@ class DeviceContext {
DeviceType
*
GetEigenDevice
()
const
;
virtual
void
Wait
()
const
{}
virtual
void
Finish
()
const
{}
};
class
CPUDeviceContext
:
public
DeviceContext
{
...
...
@@ -79,9 +77,6 @@ class CUDADeviceContext : public DeviceContext {
/*! \brief Wait for all operations completion in the stream. */
void
Wait
()
const
override
;
/*! \brief Check potential errors for the cuda kernel calls. */
void
Finish
()
const
override
;
/*! \brief Return place in the device context. */
Place
GetPlace
()
const
override
;
...
...
paddle/scripts/docker/build.sh
浏览文件 @
b41894d1
...
...
@@ -113,7 +113,10 @@ EOF
-DWITH_SWIG_PY
=
ON
\
-DWITH_STYLE_CHECK
=
OFF
make
-j
`
nproc
`
gen_proto_py
make
-j
`
nproc
`
paddle_python
make
-j
`
nproc
`
paddle_docs paddle_docs_cn
make
-j
`
nproc
`
print_operators_doc
paddle/pybind/print_operators_doc
>
doc/en/html/operators.json
popd
fi
...
...
@@ -185,14 +188,6 @@ EOF
${
DOCKERFILE_GPU_ENV
}
ADD go/cmd/pserver/pserver /usr/bin/
ADD go/cmd/master/master /usr/bin/
EOF
if
[[
${
WITH_DOC
:-
OFF
}
==
'ON'
]]
;
then
cat
>>
/paddle/build/Dockerfile
<<
EOF
ADD paddle/pybind/print_operators_doc /usr/bin/
EOF
fi
cat
>>
/paddle/build/Dockerfile
<<
EOF
# default command shows the paddle version and exit
CMD ["paddle", "version"]
EOF
...
...
python/CMakeLists.txt
浏览文件 @
b41894d1
...
...
@@ -33,6 +33,12 @@ if(WITH_MKLDNN)
list
(
APPEND MKL_DEPENDS mkldnn
)
endif
()
if
(
WITH_GPU
)
SET
(
PACKAGE_NAME
"paddlepaddle-gpu"
)
else
()
SET
(
PACKAGE_NAME
"paddlepaddle"
)
endif
()
configure_file
(
${
CMAKE_CURRENT_SOURCE_DIR
}
/setup.py.in
${
CMAKE_CURRENT_BINARY_DIR
}
/setup.py
)
...
...
python/paddle/trainer_config_helpers/layers.py
浏览文件 @
b41894d1
...
...
@@ -2722,15 +2722,15 @@ def img_pool_layer(input,
.. math::
w = 1 +
int(ceil(input\_width + 2 * padding - pool\_size) / float(stride))
h = 1 +
int(ceil(input\_height + 2 * padding\_y - pool\_size\_y) / float(stride\_y))
w = 1 +
\f
rac{ceil(input\_width + 2 * padding - pool\_size)}{stride}
\\\\
h = 1 +
\f
rac{ceil(input\_height + 2 * padding\_y - pool\_size\_y)}{stride\_y}
- ceil_mode=False:
.. math::
w = 1 +
int(floor(input\_width + 2 * padding - pool\_size) / float(stride))
h = 1 +
int(floor(input\_height + 2 * padding\_y - pool\_size\_y) / float(stride\_y))
w = 1 +
\f
rac{floor(input\_width + 2 * padding - pool\_size)}{stride}
\\\\
h = 1 +
\f
rac{floor(input\_height + 2 * padding\_y - pool\_size\_y)}{stride\_y}
The example usage is:
...
...
@@ -2863,17 +2863,17 @@ def img_pool3d_layer(input,
.. math::
w = 1 +
int(ceil(input\_width + 2 * padding - pool\_size) / float(stride))
h = 1 +
int(ceil(input\_height + 2 * padding\_y - pool\_size\_y) / float(stride\_y))
d = 1 +
int(ceil(input\_depth + 2 * padding\_z - pool\_size\_z) / float(stride\_z))
w = 1 +
\f
rac{ceil(input\_width + 2 * padding - pool\_size)}{stride}
\\\\
h = 1 +
\f
rac{ceil(input\_height + 2 * padding\_y - pool\_size\_y)}{stride\_y}
\\\\
d = 1 +
\f
rac{ceil(input\_depth + 2 * padding\_z - pool\_size\_z)}{stride\_z}
- ceil_mode=False:
.. math::
w = 1 +
int(floor(input\_width + 2 * padding - pool\_size) / float(stride))
h = 1 +
int(floor(input\_height + 2 * padding\_y - pool\_size\_y) / float(stride\_y))
d = 1 +
int(floor(input\_depth + 2 * padding\_z - pool\_size\_z) / float(stride\_z))
w = 1 +
\f
rac{floor(input\_width + 2 * padding - pool\_size)}{stride}
\\\\
h = 1 +
\f
rac{floor(input\_height + 2 * padding\_y - pool\_size\_y)}{stride\_y}
\\\\
d = 1 +
\f
rac{floor(input\_depth + 2 * padding\_z - pool\_size\_z)}{stride\_z}
\\\\
The example usage is:
...
...
@@ -2996,7 +2996,7 @@ def spp_layer(input,
Reference:
`Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition
https://arxiv.org/abs/1406.4729
`_
<https://arxiv.org/abs/1406.4729>
`_
The example usage is:
...
...
@@ -3098,7 +3098,7 @@ def img_cmrnorm_layer(input,
Reference:
`ImageNet Classification with Deep Convolutional Neural Networks
http://www.cs.toronto.edu/~fritz/absps/imagenet.pdf
`_
<http://www.cs.toronto.edu/~fritz/absps/imagenet.pdf>
`_
The example usage is:
...
...
@@ -3166,7 +3166,7 @@ def batch_norm_layer(input,
Reference:
`Batch Normalization: Accelerating Deep Network Training by Reducing
Internal Covariate Shift
http://arxiv.org/abs/1502.03167
`_
<http://arxiv.org/abs/1502.03167>
`_
The example usage is:
...
...
@@ -5424,17 +5424,19 @@ def maxout_layer(input, groups, num_channels=None, name=None, layer_attr=None):
Reference:
`Maxout Networks
http://www.jmlr.org/proceedings/papers/v28/goodfellow13.pdf
`_
<http://www.jmlr.org/proceedings/papers/v28/goodfellow13.pdf>
`_
`Multi-digit Number Recognition from Street View Imagery using Deep Convolutional Neural Networks
https://arxiv.org/pdf/1312.6082v4.pdf`_
<https://arxiv.org/pdf/1312.6082v4.pdf>`_
.. math::
y_{si+j} = \max_k x_{gsi + sk + j}
g = groups
s = input.size / num_channels
0 \le i < num_channels / groups
0 \le j < s
0 \le k < groups
out = \max_k (in[n, k, o_c , s])
\\\\
out_{i * s + j} = \max_k in_{ k * o_{c} * s + i * s + j}
\\\\
s =
\f
rac{input.size}{ num\_channels}
\\\\
o_{c} =
\f
rac{num\_channels}{groups}
\\\\
0 \le i < o_{c}
\\\\
0 \le j < s
\\\\
0 \le k < groups
\\\\
The simple usage is:
...
...
@@ -5493,7 +5495,7 @@ def ctc_layer(input,
Reference:
`Connectionist Temporal Classification: Labelling Unsegmented Sequence Data
with Recurrent Neural Networks
http://machinelearning.wustl.edu/mlpapers/paper_files/icml2006_GravesFGS06.pdf
`_
<http://machinelearning.wustl.edu/mlpapers/paper_files/icml2006_GravesFGS06.pdf>
`_
Note:
Considering the 'blank' label needed by CTC, you need to use (num_classes + 1)
...
...
@@ -5567,7 +5569,7 @@ def warp_ctc_layer(input,
Reference:
`Connectionist Temporal Classification: Labelling Unsegmented Sequence Data
with Recurrent Neural Networks
http://machinelearning.wustl.edu/mlpapers/paper_files/icml2006_GravesFGS06.pdf
`_
<http://machinelearning.wustl.edu/mlpapers/paper_files/icml2006_GravesFGS06.pdf>
`_
Note:
- Let num_classes represents the category number. Considering the 'blank'
...
...
@@ -5788,7 +5790,7 @@ def nce_layer(input,
Reference:
`A fast and simple algorithm for training neural probabilistic language
models.
https://www.cs.toronto.edu/~amnih/papers/ncelm.pdf
`_
models.
<https://www.cs.toronto.edu/~amnih/papers/ncelm.pdf>
`_
The example usage is:
...
...
@@ -5904,7 +5906,7 @@ def rank_cost(left,
Reference:
`Learning to Rank using Gradient Descent
http://research.microsoft.com/en-us/um/people/cburges/papers/ICML_ranking.pdf
`_
<http://research.microsoft.com/en-us/um/people/cburges/papers/ICML_ranking.pdf>
`_
.. math::
...
...
@@ -6440,7 +6442,7 @@ def smooth_l1_cost(input, label, name=None, coeff=1.0, layer_attr=None):
Reference:
`Fast R-CNN
https://arxiv.org/pdf/1504.08083v2.pdf
`_
<https://arxiv.org/pdf/1504.08083v2.pdf>
`_
The example usage is:
...
...
@@ -6647,7 +6649,7 @@ def prelu_layer(input,
Reference:
`Delving Deep into Rectifiers: Surpassing Human-Level Performance on
ImageNet Classification
http://arxiv.org/pdf/1502.01852v1.pdf
`_
ImageNet Classification
<http://arxiv.org/pdf/1502.01852v1.pdf>
`_
.. math::
z_i &
\\
quad if
\\
quad z_i > 0
\\\\
...
...
@@ -6744,7 +6746,7 @@ def gated_unit_layer(input,
Reference:
`Language Modeling with Gated Convolutional Networks
https://arxiv.org/abs/1612.08083
`_
<https://arxiv.org/abs/1612.08083>
`_
.. math::
y=
\\
text{act}(X \cdot W + b)\otimes \sigma(X \cdot V + c)
...
...
python/paddle/v2/fluid/layers.py
浏览文件 @
b41894d1
...
...
@@ -430,7 +430,8 @@ def _create_op_func_(op_type):
dtype
=
each
.
dtype
elif
dtype
!=
each
.
dtype
:
raise
ValueError
(
"operator {0} must input same dtype"
.
format
(
op_type
))
"operator {0} must input same dtype. {1} vs {2}"
.
format
(
op_type
,
dtype
,
each
.
dtype
))
return
dtype
...
...
python/paddle/v2/fluid/tests/book/test_machine_translation.py
浏览文件 @
b41894d1
import
numpy
as
np
import
paddle.v2
as
paddle
import
paddle.v2.
dataset.conll05
as
conll05
import
paddle.v2.
fluid
as
fluid
import
paddle.v2.fluid.core
as
core
import
paddle.v2.fluid.framework
as
framework
import
paddle.v2.fluid.layers
as
layers
from
paddle.v2.fluid.executor
import
Executor
,
g_scope
from
paddle.v2.fluid.optimizer
import
SGDOptimizer
import
paddle.v2.fluid
as
fluid
import
paddle.v2.fluid.layers
as
pd
from
paddle.v2.fluid.executor
import
Executor
dict_size
=
30000
source_dict_dim
=
target_dict_dim
=
dict_size
src_dict
,
trg_dict
=
paddle
.
dataset
.
wmt14
.
get_dict
(
dict_size
)
hidden_dim
=
51
2
word_dim
=
512
hidden_dim
=
3
2
word_dim
=
16
IS_SPARSE
=
True
batch_size
=
5
0
batch_size
=
1
0
max_length
=
50
topk_size
=
50
trg_dic_size
=
10000
src_word_id
=
layers
.
data
(
name
=
"src_word_id"
,
shape
=
[
1
],
dtype
=
'int64'
)
src_embedding
=
layers
.
embedding
(
decoder_size
=
hidden_dim
def
encoder_decoder
():
# encoder
src_word_id
=
layers
.
data
(
name
=
"src_word_id"
,
shape
=
[
1
],
dtype
=
'int64'
,
lod_level
=
1
)
src_embedding
=
layers
.
embedding
(
input
=
src_word_id
,
size
=
[
dict_size
,
word_dim
],
dtype
=
'float32'
,
is_sparse
=
IS_SPARSE
,
param_attr
=
fluid
.
ParamAttr
(
name
=
'vemb'
))
fc1
=
fluid
.
layers
.
fc
(
input
=
src_embedding
,
size
=
hidden_dim
*
4
,
act
=
'tanh'
)
lstm_hidden0
,
lstm_0
=
layers
.
dynamic_lstm
(
input
=
fc1
,
size
=
hidden_dim
*
4
)
encoder_out
=
layers
.
sequence_pool
(
input
=
lstm_hidden0
,
pool_type
=
"last"
)
def
encoder
():
lstm_hidden0
,
lstm_0
=
layers
.
dynamic_lstm
(
input
=
src_embedding
,
size
=
hidden_dim
,
candidate_activation
=
'sigmoid'
,
cell_activation
=
'sigmoid'
)
lstm_hidden1
,
lstm_1
=
layers
.
dynamic_lstm
(
input
=
src_embedding
,
size
=
hidden_dim
,
candidate_activation
=
'sigmoid'
,
cell_activation
=
'sigmoid'
,
is_reverse
=
True
)
bidirect_lstm_out
=
layers
.
concat
([
lstm_hidden0
,
lstm_hidden1
],
axis
=
0
)
return
bidirect_lstm_out
# decoder
trg_language_word
=
layers
.
data
(
name
=
"target_language_word"
,
shape
=
[
1
],
dtype
=
'int64'
,
lod_level
=
1
)
trg_embedding
=
layers
.
embedding
(
input
=
trg_language_word
,
size
=
[
dict_size
,
word_dim
],
dtype
=
'float32'
,
is_sparse
=
IS_SPARSE
,
param_attr
=
fluid
.
ParamAttr
(
name
=
'vemb'
))
rnn
=
fluid
.
layers
.
DynamicRNN
()
with
rnn
.
block
():
current_word
=
rnn
.
step_input
(
trg_embedding
)
mem
=
rnn
.
memory
(
init
=
encoder_out
)
fc1
=
fluid
.
layers
.
fc
(
input
=
[
current_word
,
mem
],
size
=
decoder_size
,
act
=
'tanh'
)
out
=
fluid
.
layers
.
fc
(
input
=
fc1
,
size
=
target_dict_dim
,
act
=
'softmax'
)
rnn
.
update_memory
(
mem
,
fc1
)
rnn
.
output
(
out
)
def
decoder_trainer
(
context
):
'''
decoder with trainer
'''
pass
return
rnn
()
def
to_lodtensor
(
data
,
place
):
...
...
@@ -72,13 +75,18 @@ def to_lodtensor(data, place):
def
main
():
encoder_out
=
encoder
()
# TODO(jacquesqiao) call here
decoder_trainer
(
encoder_out
)
rnn_out
=
encoder_decoder
()
label
=
layers
.
data
(
name
=
"target_language_next_word"
,
shape
=
[
1
],
dtype
=
'int64'
,
lod_level
=
1
)
cost
=
layers
.
cross_entropy
(
input
=
rnn_out
,
label
=
label
)
avg_cost
=
fluid
.
layers
.
mean
(
x
=
cost
)
optimizer
=
fluid
.
optimizer
.
Adagrad
(
learning_rate
=
1e-4
)
optimizer
.
minimize
(
avg_cost
)
train_data
=
paddle
.
batch
(
paddle
.
reader
.
shuffle
(
paddle
.
dataset
.
wmt14
.
train
(
8000
),
buf_size
=
1000
),
paddle
.
dataset
.
wmt14
.
train
(
dict_size
),
buf_size
=
1000
),
batch_size
=
batch_size
)
place
=
core
.
CPUPlace
()
...
...
@@ -88,15 +96,23 @@ def main():
batch_id
=
0
for
pass_id
in
xrange
(
2
):
print
'pass_id'
,
pass_id
for
data
in
train_data
():
print
'batch'
,
batch_id
batch_id
+=
1
if
batch_id
>
10
:
break
word_data
=
to_lodtensor
(
map
(
lambda
x
:
x
[
0
],
data
),
place
)
trg_word
=
to_lodtensor
(
map
(
lambda
x
:
x
[
1
],
data
),
place
)
trg_word_next
=
to_lodtensor
(
map
(
lambda
x
:
x
[
2
],
data
),
place
)
outs
=
exe
.
run
(
framework
.
default_main_program
(),
feed
=
{
'src_word_id'
:
word_data
,
},
fetch_list
=
[
encoder_out
])
feed
=
{
'src_word_id'
:
word_data
,
'target_language_word'
:
trg_word
,
'target_language_next_word'
:
trg_word_next
},
fetch_list
=
[
avg_cost
])
avg_cost_val
=
np
.
array
(
outs
[
0
])
print
(
'pass_id='
+
str
(
pass_id
)
+
' batch='
+
str
(
batch_id
)
+
" avg_cost="
+
str
(
avg_cost_val
))
if
batch_id
>
3
:
exit
(
0
)
batch_id
+=
1
if
__name__
==
'__main__'
:
...
...
python/setup.py.in
浏览文件 @
b41894d1
...
...
@@ -5,7 +5,7 @@ class BinaryDistribution(Distribution):
return True
MAJOR = 0
MINOR = 1
0
MINOR = 1
1
PATCH = 0
RC = 0
ISTAGED = False
...
...
@@ -89,7 +89,7 @@ paddle_rt_libs = ['${WARPCTC_LIBRARIES}']
if '${MKL_SHARED_LIBS}'!= '':
paddle_rt_libs += '${MKL_SHARED_LIBS}'.split(';')
setup(name='
paddlepaddle
',
setup(name='
${PACKAGE_NAME}
',
version='${PADDLE_VERSION}',
description='Parallel Distributed Deep Learning',
install_requires=setup_requires,
...
...
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