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
sven357
Paddle
提交
1e92baf7
P
Paddle
项目概览
sven357
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
1
Star
1
Fork
0
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1
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
1
Issue
1
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
1e92baf7
编写于
8月 18, 2018
作者:
N
nhzlx
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
fix comments
上级
f55e8901
变更
3
隐藏空白更改
内联
并排
Showing
3 changed file
with
13 addition
and
12 deletion
+13
-12
paddle/fluid/inference/tensorrt/convert/conv2d_op.cc
paddle/fluid/inference/tensorrt/convert/conv2d_op.cc
+6
-4
paddle/fluid/inference/tensorrt/convert/elementwise_op.cc
paddle/fluid/inference/tensorrt/convert/elementwise_op.cc
+5
-5
paddle/fluid/inference/tensorrt/convert/fc_op.cc
paddle/fluid/inference/tensorrt/convert/fc_op.cc
+2
-3
未找到文件。
paddle/fluid/inference/tensorrt/convert/conv2d_op.cc
浏览文件 @
1e92baf7
...
...
@@ -37,11 +37,11 @@ class Conv2dOpConverter : public OpConverter {
auto
*
Y_t
=
Y_v
->
GetMutable
<
framework
::
LoDTensor
>
();
platform
::
CPUPlace
cpu_place
;
framework
::
LoDTensor
*
weight_tensor
=
new
framework
::
LoDTensor
();
std
::
unique_ptr
<
framework
::
LoDTensor
>
weight_tensor
(
new
framework
::
LoDTensor
());
weight_tensor
->
Resize
(
Y_t
->
dims
());
TensorCopySync
((
*
Y_t
),
cpu_place
,
weight_tensor
);
engine_
->
weight_map
[
op_desc
.
Input
(
"Filter"
).
front
()]
=
std
::
move
(
std
::
unique_ptr
<
framework
::
Tensor
>
(
weight_tensor
));
TensorCopySync
((
*
Y_t
),
cpu_place
,
weight_tensor
.
get
());
auto
*
weight_data
=
weight_tensor
->
mutable_data
<
float
>
(
platform
::
CPUPlace
());
...
...
@@ -78,6 +78,8 @@ class Conv2dOpConverter : public OpConverter {
layer
->
setNbGroups
(
groups
);
auto
output_name
=
op_desc
.
Output
(
"Output"
).
front
();
engine_
->
weight_map
[
op_desc
.
Input
(
"Filter"
).
front
()]
=
std
::
move
(
weight_tensor
);
engine_
->
SetITensor
(
output_name
,
layer
->
getOutput
(
0
));
if
(
test_mode
)
{
engine_
->
DeclareOutput
(
output_name
);
...
...
paddle/fluid/inference/tensorrt/convert/elementwise_op.cc
浏览文件 @
1e92baf7
...
...
@@ -41,12 +41,10 @@ class ElementwiseWeightOpConverter : public OpConverter {
auto
*
Y_t
=
Y_v
->
GetMutable
<
framework
::
LoDTensor
>
();
platform
::
CPUPlace
cpu_place
;
framework
::
LoDTensor
*
weight_tensor
=
new
framework
::
LoDTensor
();
std
::
unique_ptr
<
framework
::
LoDTensor
>
weight_tensor
(
new
framework
::
LoDTensor
());
weight_tensor
->
Resize
(
Y_t
->
dims
());
TensorCopySync
((
*
Y_t
),
cpu_place
,
weight_tensor
);
engine_
->
weight_map
[
op_desc
.
Input
(
"Y"
).
front
()]
=
std
::
move
(
std
::
unique_ptr
<
framework
::
Tensor
>
(
weight_tensor
));
TensorCopySync
((
*
Y_t
),
cpu_place
,
weight_tensor
.
get
());
auto
*
weight_data
=
weight_tensor
->
mutable_data
<
float
>
(
platform
::
CPUPlace
());
auto
scale_mode
=
nvinfer1
::
ScaleMode
::
kELEMENTWISE
;
...
...
@@ -90,6 +88,8 @@ class ElementwiseWeightOpConverter : public OpConverter {
engine_
,
Scale
,
*
const_cast
<
nvinfer1
::
ITensor
*>
(
X
),
scale_mode
,
shift_weights
.
get
(),
scale_weights
.
get
(),
power_weights
.
get
());
auto
output_name
=
op_desc
.
Output
(
"Out"
)[
0
];
engine_
->
weight_map
[
op_desc
.
Input
(
"Y"
).
front
()]
=
std
::
move
(
weight_tensor
);
engine_
->
SetITensor
(
output_name
,
layer
->
getOutput
(
0
));
if
(
test_mode
)
{
// the test framework can not determine which is the
// output, so place the declaration inside.
...
...
paddle/fluid/inference/tensorrt/convert/fc_op.cc
浏览文件 @
1e92baf7
...
...
@@ -78,10 +78,8 @@ class FcOpConverter : public OpConverter {
PADDLE_ENFORCE_EQ
(
weight_tensor
.
dims
().
size
(),
2UL
);
// a matrix
size_t
n_output
=
weight_tensor
.
dims
()[
1
];
framework
::
LoDTensor
*
tmp
=
new
framework
::
LoDTensor
(
);
std
::
unique_ptr
<
framework
::
Tensor
>
tmp
(
new
framework
::
LoDTensor
()
);
tmp
->
Resize
(
weight_tensor
.
dims
());
engine_
->
weight_map
[
op_desc
.
Input
(
"Y"
).
front
()]
=
std
::
move
(
std
::
unique_ptr
<
framework
::
Tensor
>
(
tmp
));
memcpy
(
tmp
->
mutable_data
<
float
>
(
platform
::
CPUPlace
()),
weight_data
,
Y_t
->
dims
()[
0
]
*
Y_t
->
dims
()[
1
]
*
sizeof
(
float
));
...
...
@@ -110,6 +108,7 @@ class FcOpConverter : public OpConverter {
auto
output_name
=
op_desc
.
Output
(
"Out"
).
front
();
engine_
->
SetITensor
(
output_name
,
layer
->
getOutput
(
0
));
engine_
->
weight_map
[
op_desc
.
Input
(
"Y"
).
front
()]
=
std
::
move
(
tmp
);
if
(
test_mode
)
{
engine_
->
DeclareOutput
(
output_name
);
}
...
...
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