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
OPTHREE
Paddle
提交
bcd67bdd
P
Paddle
项目概览
OPTHREE
/
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
bcd67bdd
编写于
7月 24, 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
add assert for GetOutput
上级
7382f986
变更
5
隐藏空白更改
内联
并排
Showing
5 changed file
with
20 addition
and
13 deletion
+20
-13
paddle/fluid/inference/tensorrt/convert/ut_helper.h
paddle/fluid/inference/tensorrt/convert/ut_helper.h
+2
-2
paddle/fluid/inference/tensorrt/engine.cc
paddle/fluid/inference/tensorrt/engine.cc
+8
-4
paddle/fluid/inference/tensorrt/engine.h
paddle/fluid/inference/tensorrt/engine.h
+2
-2
paddle/fluid/inference/tensorrt/test_engine.cc
paddle/fluid/inference/tensorrt/test_engine.cc
+3
-3
paddle/fluid/operators/tensorrt_engine_op.h
paddle/fluid/operators/tensorrt_engine_op.h
+5
-2
未找到文件。
paddle/fluid/inference/tensorrt/convert/ut_helper.h
浏览文件 @
bcd67bdd
...
...
@@ -139,11 +139,11 @@ class TRTConvertValidation {
cudaStreamSynchronize
(
*
engine_
->
stream
());
ASSERT_FALSE
(
op_desc_
->
OutputArgumentNames
().
empty
());
const
size_t
output_space_size
=
200
;
const
size_t
output_space_size
=
200
0
;
for
(
const
auto
&
output
:
op_desc_
->
OutputArgumentNames
())
{
std
::
vector
<
float
>
fluid_out
;
std
::
vector
<
float
>
trt_out
(
output_space_size
);
engine_
->
GetOutputInCPU
(
output
,
&
trt_out
[
0
]);
engine_
->
GetOutputInCPU
(
output
,
&
trt_out
[
0
]
,
output_space_size
);
cudaStreamSynchronize
(
*
engine_
->
stream
());
auto
*
var
=
scope_
.
FindVar
(
output
);
...
...
paddle/fluid/inference/tensorrt/engine.cc
浏览文件 @
bcd67bdd
/* Copyright (c) 2018 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.
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
...
...
@@ -149,7 +149,8 @@ void *TensorRTEngine::GetOutputInGPU(const std::string &name) {
return
buffer
(
name
).
buffer
;
}
void
TensorRTEngine
::
GetOutputInGPU
(
const
std
::
string
&
name
,
void
*
dst
)
{
void
TensorRTEngine
::
GetOutputInGPU
(
const
std
::
string
&
name
,
void
*
dst
,
size_t
max_size
)
{
// determine data size
auto
*
output
=
TensorRTEngine
::
GetITensor
(
name
);
nvinfer1
::
Dims
dims
=
output
->
getDimensions
();
...
...
@@ -161,6 +162,7 @@ void TensorRTEngine::GetOutputInGPU(const std::string &name, void *dst) {
PADDLE_ENFORCE
(
it
!=
buffer_sizes_
.
end
());
PADDLE_ENFORCE_GT
(
it
->
second
,
0
);
PADDLE_ENFORCE_LE
(
dst_size
,
it
->
second
);
PADDLE_ENFORCE_GE
(
max_size
,
dst_size
);
auto
&
buf
=
buffer
(
name
);
PADDLE_ENFORCE_NOT_NULL
(
buf
.
buffer
,
"buffer should be allocated before"
);
PADDLE_ENFORCE_EQ
(
cudaMemcpyAsync
(
dst
,
buf
.
buffer
,
dst_size
,
...
...
@@ -168,7 +170,8 @@ void TensorRTEngine::GetOutputInGPU(const std::string &name, void *dst) {
0
);
}
void
TensorRTEngine
::
GetOutputInCPU
(
const
std
::
string
&
name
,
void
*
dst
)
{
void
TensorRTEngine
::
GetOutputInCPU
(
const
std
::
string
&
name
,
void
*
dst
,
size_t
max_size
)
{
// determine data size
auto
*
output
=
TensorRTEngine
::
GetITensor
(
name
);
...
...
@@ -180,6 +183,7 @@ void TensorRTEngine::GetOutputInCPU(const std::string &name, void *dst) {
PADDLE_ENFORCE
(
it
!=
buffer_sizes_
.
end
());
PADDLE_ENFORCE_GT
(
it
->
second
,
0
);
PADDLE_ENFORCE_LE
(
dst_size
,
it
->
second
);
PADDLE_ENFORCE_GE
(
max_size
,
dst_size
);
auto
&
buf
=
buffer
(
name
);
PADDLE_ENFORCE_NOT_NULL
(
buf
.
buffer
,
"buffer should be allocated before"
);
PADDLE_ENFORCE_EQ
(
0
,
cudaMemcpyAsync
(
dst
,
buf
.
buffer
,
dst_size
,
...
...
paddle/fluid/inference/tensorrt/engine.h
浏览文件 @
bcd67bdd
...
...
@@ -106,10 +106,10 @@ class TensorRTEngine : public EngineBase {
// Return the output's GPU memory address without copy.
void
*
GetOutputInGPU
(
const
std
::
string
&
name
);
// Copy data into dst inside the GPU device.
void
GetOutputInGPU
(
const
std
::
string
&
name
,
void
*
dst
);
void
GetOutputInGPU
(
const
std
::
string
&
name
,
void
*
dst
,
size_t
max_size
);
// LOW EFFICENCY! Get output to CPU, this will trigger a memory copy from GPU
// to CPU.
void
GetOutputInCPU
(
const
std
::
string
&
name
,
void
*
dst
);
void
GetOutputInCPU
(
const
std
::
string
&
name
,
void
*
dst
,
size_t
max_size
);
// Fill an ITensor into map itensor_map_.
void
SetITensor
(
const
std
::
string
&
name
,
nvinfer1
::
ITensor
*
tensor
);
// Get an ITensor called name.
...
...
paddle/fluid/inference/tensorrt/test_engine.cc
浏览文件 @
bcd67bdd
...
...
@@ -71,7 +71,7 @@ TEST_F(TensorRTEngineTest, add_layer) {
LOG
(
INFO
)
<<
"to get output"
;
float
y_cpu
;
engine_
->
GetOutputInCPU
(
"y"
,
&
y_cpu
);
engine_
->
GetOutputInCPU
(
"y"
,
&
y_cpu
,
1
*
sizeof
(
float
)
);
LOG
(
INFO
)
<<
"to checkout output"
;
ASSERT_EQ
(
y_cpu
,
x_v
*
2
+
3
);
...
...
@@ -108,7 +108,7 @@ TEST_F(TensorRTEngineTest, add_layer_multi_dim) {
ASSERT_EQ
(
dims
.
nbDims
,
3
);
ASSERT_EQ
(
dims
.
d
[
0
],
2
);
ASSERT_EQ
(
dims
.
d
[
1
],
1
);
engine_
->
GetOutputInCPU
(
"y"
,
&
y_cpu
[
0
]);
engine_
->
GetOutputInCPU
(
"y"
,
&
y_cpu
[
0
]
,
2
*
sizeof
(
float
)
);
ASSERT_EQ
(
y_cpu
[
0
],
4.5
);
ASSERT_EQ
(
y_cpu
[
1
],
14.5
);
}
...
...
@@ -141,7 +141,7 @@ TEST_F(TensorRTEngineTest, test_conv2d_temp) {
LOG
(
INFO
)
<<
"to get output"
;
float
*
y_cpu
=
new
float
[
18
];
engine_
->
GetOutputInCPU
(
"y"
,
&
y_cpu
[
0
]);
engine_
->
GetOutputInCPU
(
"y"
,
&
y_cpu
[
0
]
,
18
*
sizeof
(
float
)
);
ASSERT_EQ
(
y_cpu
[
0
],
4.0
);
ASSERT_EQ
(
y_cpu
[
1
],
6.0
);
}
...
...
paddle/fluid/operators/tensorrt_engine_op.h
浏览文件 @
bcd67bdd
...
...
@@ -100,8 +100,11 @@ class TensorRTEngineKernel : public framework::OpKernel<T> {
// tensor.
// if (platform::is_cpu_place(fluid_t->place())) {
// TODO(Superjomn) change this float to dtype size.
engine
->
GetOutputInCPU
(
y
,
fluid_t
->
mutable_data
<
float
>
(
platform
::
CPUPlace
()));
auto
size
=
inference
::
analysis
::
AccuDims
(
dims
.
d
,
dims
.
nbDims
)
*
FLAGS_tensorrt_engine_batch_size
;
engine
->
GetOutputInCPU
(
y
,
fluid_t
->
mutable_data
<
float
>
(
platform
::
CPUPlace
()),
size
*
sizeof
(
float
));
//} else {
// engine->GetOutputInGPU(
// y, fluid_t->mutable_data<float>(platform::CUDAPlace()),
...
...
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