forked from mindspore-Ecosystem/mindspore
add micro examples & delete useless tests
This commit is contained in:
lz
parent
a29c84af3f
commit
208ede7c83
|
|
@ -66,3 +66,4 @@
|
|||
"mindspore/mindspore/lite/examples/runtime_gpu_extend/src/cl" "readability/fn_size"
|
||||
"mindspore/mindspore/lite/examples/quick_start_c/main.c" "readability/casting"
|
||||
"mindspore/mindspore/lite/examples/quick_start_c/main.c" "runtime/threadsafe_fn"
|
||||
"mindspore/mindspore/lite/examples/quick_start_micro" "readability/casting"
|
||||
|
|
|
|||
|
|
@ -0,0 +1,268 @@
|
|||
# X86编译部署
|
||||
|
||||
`Linux` `IoT` `C++` `全流程` `模型编译` `模型代码生成` `模型部署` `推理应用` `初级` `中级` `高级`
|
||||
|
||||
<!-- TOC -->
|
||||
|
||||
- [X86编译部署](#X86编译部署)
|
||||
- [概述](#概述)
|
||||
- [模型编译体验](#模型编译体验)
|
||||
- [详细步骤](#详细步骤)
|
||||
- [生成代码](#生成代码)
|
||||
- [部署应用](#部署应用)
|
||||
- [编译依赖](#编译依赖)
|
||||
- [构建与运行](#构建与运行)
|
||||
- [编写推理代码示例](#编写推理代码示例)
|
||||
- [更多详情](#更多详情)
|
||||
- [Android平台编译部署](#android平台编译部署)
|
||||
- [Arm Cortex-M平台编译部署](#armcortex-m平台编译部署)
|
||||
|
||||
<!-- /TOC -->
|
||||
|
||||
## 概述
|
||||
|
||||
本教程以MNIST分类模型推理代码为例，帮助用户了解codegen生成代码、编译构建、部署等流程。
|
||||
|
||||
## 模型编译体验
|
||||
|
||||
用户可以使用脚本一键式编译生成MNIST分类模型的推理代码并执行推理，得到单次推理输出。下载[MindSpore源码](https://gitee.com/mindspore/mindspore)，进入[`mindspore/mindspore/lite/micro/examples/mnist_x86`](https://gitee.com/mindspore/mindspore/tree/master/mindspore/lite/micro/example/mnist_x86)目录，执行脚本`mnist.sh`自动生成模型推理代码并编译工程目录，即可得到单次推理输出。
|
||||
|
||||
```bash
|
||||
bash mnist.sh
|
||||
```
|
||||
|
||||
推理结果如下：
|
||||
|
||||
```text
|
||||
======run benchmark======
|
||||
input 0: mnist_input.bin
|
||||
|
||||
outputs:
|
||||
name: Softmax-7, DataType: 43, Size: 40, Shape: [1 10], Data:
|
||||
0.000000, 0.000000, 0.000000, 1.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000,
|
||||
======run success=======
|
||||
```
|
||||
|
||||
也可以按照**详细步骤**从生成代码开始逐步完成使用codegen编译一个MNIST分类模型的全流程。
|
||||
|
||||
## 详细步骤
|
||||
|
||||
在编译此工程之前需要预先获取Ubuntu-x64 CPU平台的[Release包](https://www.mindspore.cn/lite/docs/zh-CN/master/use/downloads.html)，解压后得到`mindspore-lite-{version}-linux-x64`，将其拷贝到当前目录下。
|
||||
|
||||
> `{version}`为版本号字符串，如`1.2.0`。
|
||||
|
||||
以本教程为例，预置x86平台的Release包目录如下：
|
||||
|
||||
```text
|
||||
mindspore-lite-{version}-linux-x64
|
||||
└── tools
|
||||
└── codegen # 代码生成工具
|
||||
├── codegen # 可执行程序
|
||||
├── include # 推理框架头文件
|
||||
│ ├── nnacl # nnacl 算子头文件
|
||||
│ └── wrapper
|
||||
├── lib
|
||||
│ └── libwrapper.a # MindSpore Lite CodeGen生成代码依赖的部分算子静态库
|
||||
└── third_party
|
||||
├── include
|
||||
│ └── CMSIS # ARM CMSIS NN 算子头文件
|
||||
└── lib
|
||||
└── libcmsis_nn.a # ARM CMSIS NN 算子静态库
|
||||
```
|
||||
|
||||
### 生成代码
|
||||
|
||||
下载[MNIST分类网络](https://download.mindspore.cn/model_zoo/official/lite/mnist_lite/mnist.ms)。使用Release包中的codegen编译MNIST分类模型，生成对应的x86平台推理代码。生成代码的具体命令如下：
|
||||
|
||||
```bash
|
||||
./codegen --codePath=. --modelPath=mnist.ms --target=x86
|
||||
```
|
||||
|
||||
codegen在当前目录下将生成mnist目录，其中包含了可编译构建的mnist分类模型的代码。
|
||||
> 关于codegen的更多使用命令说明，可参见[codegen使用说明](https://www.mindspore.cn/lite/docs/zh-CN/master/use/micro.html#id4)。
|
||||
|
||||
### 部署应用
|
||||
|
||||
接下来介绍如何构建MindSpore Lite CodeGen生成的模型推理代码工程，并在x86平台完成部署。上文中codegen生成的代码与`mindspore/mindspore/lite/micro/example/mnist_x86`相同，本章节编译、构建步骤将对该目录展开，用户也可参照相同操作，编译上文codegen生成mnist目录代码。
|
||||
|
||||
#### 编译依赖
|
||||
|
||||
- [CMake](https://cmake.org/download/) >= 3.18.3
|
||||
- [GCC](https://gcc.gnu.org/releases.html) >= 7.3.0
|
||||
|
||||
#### 构建与运行
|
||||
|
||||
1. **生成代码工程说明**
|
||||
|
||||
进入`mindspore/mindspore/lite/micro/example/mnist_x86`目录中。
|
||||
|
||||
生成代码工程目录说明：
|
||||
|
||||
当前目录下预置了MNIST分类网络生成的代码。
|
||||
|
||||
```text
|
||||
mnist_x86/ # 生成代码的根目录
|
||||
├── benchmark # 生成代码的benchmark目录
|
||||
└── src # 模型推理代码目录
|
||||
```
|
||||
|
||||
2. **代码编译**
|
||||
|
||||
组织模型生成的推理代码以及算子静态库，编译生成模型推理静态库并编译生成benchmark可执行文件,
|
||||
|
||||
进入代码工程目录下，新建并进入build目录：
|
||||
|
||||
```bash
|
||||
mkdir build && cd build
|
||||
```
|
||||
|
||||
开始编译：
|
||||
|
||||
```bash
|
||||
cmake -DPKG_PATH={path to}/mindspore-lite-{version}-linux-x64 ..
|
||||
make
|
||||
```
|
||||
|
||||
> `{path to}`和`{version}`需要用户根据实际情况填写。
|
||||
|
||||
代码工程编译成功结果：
|
||||
|
||||
```text
|
||||
Scanning dependencies of target net
|
||||
[ 12%] Building C object src/CMakeFiles/net.dir/net.c.o
|
||||
[ 25%] Building CXX object src/CMakeFiles/net.dir/session.cc.o
|
||||
[ 37%] Building CXX object src/CMakeFiles/net.dir/tensor.cc.o
|
||||
[ 50%] Building C object src/CMakeFiles/net.dir/weight.c.o
|
||||
[ 62%] Linking CXX static library libnet.a
|
||||
unzip raw static library libnet.a
|
||||
raw static library libnet.a size:
|
||||
-rw-r--r-- 1 user user 58K Mar 22 10:09 libnet.a
|
||||
generate specified static library libnet.a
|
||||
new static library libnet.a size:
|
||||
-rw-r--r-- 1 user user 162K Mar 22 10:09 libnet.a
|
||||
[ 62%] Built target net
|
||||
Scanning dependencies of target benchmark
|
||||
[ 75%] Building CXX object CMakeFiles/benchmark.dir/benchmark/benchmark.cc.o
|
||||
[ 87%] Building C object CMakeFiles/benchmark.dir/benchmark/load_input.c.o
|
||||
[100%] Linking CXX executable benchmark
|
||||
[100%] Built target benchmark
|
||||
```
|
||||
|
||||
此时在`mnist_x86/build/src/`目录下生成了`libnet.a`，推理执行库，在`mnist_x86/build`目录下生成了`benchmark`可执行文件。
|
||||
|
||||
3. **代码部署**
|
||||
|
||||
本示例部署于x86平台。由代码工程编译成功以后的产物为`benchmark`可执行文件，将其拷贝到用户的目标Linux服务器中即可执行。
|
||||
|
||||
在目标Linux服务上执行编译成功的二进制文件：
|
||||
|
||||
```bash
|
||||
./benchmark mnist_input.bin net.bin
|
||||
```
|
||||
|
||||
> mnist_input.bin在`example/mnist_x86`目录下，`net.bin`为模型参数文件，在`example/mnist_x86/src`目录下。
|
||||
|
||||
生成结果如下：
|
||||
|
||||
```text
|
||||
start run benchmark
|
||||
input 0: mnist_input.bin
|
||||
output size: 1
|
||||
uint8:
|
||||
Name: Softmax-7, DataType: 43, Size: 40, Shape: 1 10, Data:
|
||||
0.000000, 0.000000, 0.000000, 1.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000,
|
||||
run benchmark success
|
||||
```
|
||||
|
||||
#### 编写推理代码示例
|
||||
|
||||
本教程中的`benchmark`内部实现主要用于指导用户如何编写以及调用codegen编译的模型推理代码接口。以下为接口调用的详细介绍，详情代码可以参见[examples/mnist_x86](https://gitee.com/mindspore/mindspore/tree/master/mindspore/lite/micro/example/mnist_x86)下的示例代码示例：
|
||||
|
||||
1. **构建推理的上下文以及会话**
|
||||
|
||||
本教程生成的代码为非并行代码，无需上下文context，可直接设为空。
|
||||
|
||||
```cpp
|
||||
size_t model_size = 0;
|
||||
Context *context = nullptr;
|
||||
session::LiteSession *session = mindspore::session::LiteSession::CreateSession(model_buffer, model_size, context);
|
||||
if (session == nullptr) {
|
||||
std::cerr << "create lite session failed" << std::endl;
|
||||
return RET_ERROR;
|
||||
}
|
||||
```
|
||||
|
||||
2. **输入数据准备**
|
||||
|
||||
用户所需要准备的输入数据内存空间，若输入是持久化文件，可通过读文件方式获取。若输入数据已经存在内存中，则此处无需读取，可直接传入数据指针。
|
||||
|
||||
```cpp
|
||||
std::vector<MSTensor *> inputs = session->GetInputs();
|
||||
MSTensor *input = inputs.at(0);
|
||||
if (input == nullptr) {
|
||||
return RET_ERROR;
|
||||
}
|
||||
// Assume we have got input data in memory.
|
||||
memcpy(input->MutableData(), input_buffer, input->Size());
|
||||
```
|
||||
|
||||
3. **执行推理**
|
||||
|
||||
```cpp
|
||||
session->RunGraph();
|
||||
```
|
||||
|
||||
4. **推理结束获取输出**
|
||||
|
||||
```cpp
|
||||
Vector<String> outputs_name = session->GetOutputTensorNames();
|
||||
for (const auto &name : outputs_name) {
|
||||
auto output = session->GetOutputByTensorName(name);
|
||||
// deal with output
|
||||
......
|
||||
}
|
||||
```
|
||||
|
||||
5. **释放内存session**
|
||||
|
||||
```cpp
|
||||
delete session;
|
||||
```
|
||||
|
||||
6. **推理代码整体调用流程**
|
||||
|
||||
```cpp
|
||||
// Assume we have got model_buffer data in memory.
|
||||
size_t model_size = 0;
|
||||
Context *context = nullptr;
|
||||
session::LiteSession *session = mindspore::session::LiteSession::CreateSession(model_buffer, model_size, context);
|
||||
if (session == nullptr) {
|
||||
std::cerr << "create lite session failed" << std::endl;
|
||||
return RET_ERROR;
|
||||
}
|
||||
|
||||
std::vector<MSTensor *> inputs = session->GetInputs();
|
||||
MSTensor *input = inputs.at(0);
|
||||
if (input == nullptr) {
|
||||
return RET_ERROR;
|
||||
}
|
||||
// Assume we have got input data in memory.
|
||||
memcpy(input->MutableData(), input_buffer, input->Size());
|
||||
|
||||
session->RunGraph();
|
||||
|
||||
Vector<String> outputs_name = session->GetOutputTensorNames();
|
||||
for (const auto &name : outputs_name) {
|
||||
auto output = session->GetOutputByTensorName(name);
|
||||
// deal with output
|
||||
......
|
||||
}
|
||||
|
||||
delete session;
|
||||
```
|
||||
|
||||
## 更多详情
|
||||
|
||||
### [Android平台编译部署](https://gitee.com/mindspore/mindspore/blob/master/mindspore/lite/micro/example/mobilenetv2/README.md#)
|
||||
|
||||
### [Arm Cortex-M平台编译部署](https://www.mindspore.cn/lite/docs/zh-CN/master/use/micro.html)
|
||||
|
|
@ -0,0 +1,27 @@
|
|||
[common_quant_param]
|
||||
# Supports WEIGHT_QUANT or FULL_QUANT
|
||||
#quant_type=WEIGHT_QUANT
|
||||
# Weight quantization support the number of bits [0,16], Set to 0 is mixed bit quantization, otherwise it is fixed bit quantization
|
||||
# Full quantization support the number of bits [1,8]
|
||||
#bit_num=8
|
||||
# Layers with size of weights exceeds threshold `min_quant_weight_size` will be quantized.
|
||||
#min_quant_weight_size=0
|
||||
# Layers with channel size of weights exceeds threshold `min_quant_weight_channel` will be quantized.
|
||||
#min_quant_weight_channel=16
|
||||
|
||||
|
||||
[micro_param]
|
||||
# enable code-generation for MCU HW
|
||||
enable_micro=true
|
||||
|
||||
# specify HW target, support x86,ARM32M, AMR32A, ARM64 only.
|
||||
target=x86
|
||||
|
||||
# code generation for Inference or Train
|
||||
codegen_mode=Inference
|
||||
|
||||
# enable parallel inference or not
|
||||
support_parallel=false
|
||||
|
||||
# enable debug
|
||||
debug_mode=false
|
||||
|
|
@ -0,0 +1,101 @@
|
|||
#!/bin/bash
|
||||
# Copyright 2021 Huawei Technologies Co., Ltd
|
||||
#
|
||||
# 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.
|
||||
# ============================================================================
|
||||
set -e
|
||||
|
||||
GEN=OFF
|
||||
TARBALL=""
|
||||
while getopts 'r:g:' OPT
|
||||
do
|
||||
case "${OPT}" in
|
||||
g)
|
||||
GEN=$OPTARG
|
||||
;;
|
||||
r)
|
||||
TARBALL=$OPTARG
|
||||
;;
|
||||
?)
|
||||
echo "Usage: add -g on , -r specific release.tar.gz"
|
||||
esac
|
||||
done
|
||||
|
||||
BASEPATH="$( cd "$( dirname "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )"
|
||||
ROOT_DIR=${BASEPATH%%/mindspore/lite/examples/quick_start_micro/mnist_x86}
|
||||
DEMO_DIR=${ROOT_DIR}/mindspore/lite/examples/quick_start_micro/mnist_x86
|
||||
MODEL_DIR=${ROOT_DIR}/mindspore/lite/examples/quick_start_micro/models
|
||||
PKG_DIR=${ROOT_DIR}/mindspore/lite/examples/quick_start_micro/pkgs
|
||||
COFIG_FILE=${DEMO_DIR}/micro.cfg
|
||||
echo "root dir is: ${ROOT_DIR}"
|
||||
echo "current dir is: ${BASEPATH}"
|
||||
echo "demo dir is: ${DEMO_DIR}"
|
||||
echo "model dir is: ${MODEL_DIR}"
|
||||
|
||||
MODEL_NAME=mnist
|
||||
INPUT_BIN=${MODEL_DIR}/${MODEL_NAME}/mnist.tflite.ms.bin
|
||||
VALICATION_DATA=${MODEL_DIR}/${MODEL_NAME}/mnist.tflite.ms.out
|
||||
MODEL=${MODEL_DIR}/${MODEL_NAME}/mnist.tflite
|
||||
MODEL_FILE=${MODEL_NAME}.tar.gz
|
||||
|
||||
get_version() {
|
||||
local VERSION_HEADER=${ROOT_DIR}/mindspore/lite/include/version.h
|
||||
local VERSION_MAJOR=$(grep "const int ms_version_major =" ${VERSION_HEADER} | tr -dc "[0-9]")
|
||||
local VERSION_MINOR=$(grep "const int ms_version_minor =" ${VERSION_HEADER} | tr -dc "[0-9]")
|
||||
local VERSION_REVISION=$(grep "const int ms_version_revision =" ${VERSION_HEADER} | tr -dc "[0-9]")
|
||||
VERSION_STR=${VERSION_MAJOR}.${VERSION_MINOR}.${VERSION_REVISION}
|
||||
}
|
||||
|
||||
download_mnist() {
|
||||
rm -rf ${MODEL_DIR:?}/${MODEL_NAME}
|
||||
mkdir -p ${MODEL_DIR}/${MODEL_NAME}
|
||||
tar xzvf ${MODEL_DIR}/${MODEL_FILE} -C ${MODEL_DIR}/${MODEL_NAME} || exit 1
|
||||
}
|
||||
|
||||
gen_mnist() {
|
||||
tar xzvf ${PKG_DIR}/${MINDSPORE_FILE} -C ${PKG_DIR} || exit 1
|
||||
export LD_LIBRARY_PATH=${PKG_DIR}/${MINDSPORE_FILE_NAME}/tools/converter/lib:${LD_LIBRARY_PATH}
|
||||
${PKG_DIR}/${MINDSPORE_FILE_NAME}/tools/converter/converter/converter_lite --fmk=TFLITE --modelFile=${MODEL} --outputFile=${DEMO_DIR} --configFile=${COFIG_FILE}
|
||||
}
|
||||
|
||||
mkdir -p ${BASEPATH}/build
|
||||
|
||||
get_version
|
||||
MINDSPORE_FILE_NAME="mindspore-lite-${VERSION_STR}-linux-x64"
|
||||
MINDSPORE_FILE="${MINDSPORE_FILE_NAME}.tar.gz"
|
||||
PKG_PATH=${PKG_DIR}/${MINDSPORE_FILE_NAME}
|
||||
|
||||
echo "tar ball is: ${TARBALL}"
|
||||
if [ -n "$TARBALL" ]; then
|
||||
echo "cp file"
|
||||
rm -rf ${PKG_DIR}
|
||||
mkdir -p ${PKG_DIR}
|
||||
cp ${TARBALL} ${PKG_DIR}
|
||||
fi
|
||||
|
||||
# 1. code-generation
|
||||
if [[ "${GEN}" == "ON" ]] || [[ "${GEN}" == "on" ]]; then
|
||||
echo "downloading mnist.ms!"
|
||||
download_mnist
|
||||
echo "generating mnist"
|
||||
gen_mnist
|
||||
fi
|
||||
|
||||
# 2. build benchmark
|
||||
mkdir -p ${DEMO_DIR}/build && cd ${DEMO_DIR}/build || exit 1
|
||||
cmake -DPKG_PATH=${PKG_PATH} ..
|
||||
make
|
||||
|
||||
# 3. run benchmark
|
||||
echo "net file: ${DEMO_DIR}/src/mnist.bin"
|
||||
./benchmark ${INPUT_BIN} ../src/net.bin 1 ${VALICATION_DATA}
|
||||
|
|
@ -1,24 +0,0 @@
|
|||
add_definitions(-DUSE_GLOG)
|
||||
string(REPLACE "/test" "" MICRO_DIR ${CMAKE_CURRENT_SOURCE_DIR})
|
||||
string(REPLACE " -fvisibility=hidden " " -fvisibility=default " CMAKE_C_FLAGS "${CMAKE_C_FLAGS}")
|
||||
string(REPLACE " -fvisibility=hidden " " -fvisibility=default " CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
|
||||
|
||||
set(3RD_DIR ${TOP_DIR}/third_party)
|
||||
set(LITE_DIR ${TOP_DIR}/mindspore/lite)
|
||||
set(BUILD_LITE "on")
|
||||
|
||||
include(${TOP_DIR}/cmake/external_libs/gtest.cmake)
|
||||
include(${MICRO_DIR}/cmake/file_list.cmake)
|
||||
include(${MICRO_DIR}/cmake/package_wrapper.cmake)
|
||||
|
||||
include_directories(${NNACL_DIR}/../)
|
||||
include_directories(${TOP_DIR})
|
||||
include_directories(${TOP_DIR}/mindspore/core/)
|
||||
include_directories(${LITE_DIR})
|
||||
include_directories(${MICRO_DIR})
|
||||
include_directories(${3RD_DIR})
|
||||
|
||||
add_executable(micro_test code_gen_test.cc ${FILE_SET})
|
||||
add_dependencies(micro_test fbs_src)
|
||||
add_dependencies(micro_test fbs_inner_src)
|
||||
target_link_libraries(micro_test dl mindspore::gtest ${SECUREC_LIBRARY} mindspore::glog)
|
||||
|
|
@ -1,33 +0,0 @@
|
|||
/**
|
||||
* Copyright 2019 Huawei Technologies Co., Ltd
|
||||
*
|
||||
* 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 "gtest/gtest.h"
|
||||
#include "micro/coder/coder.h"
|
||||
|
||||
namespace mindspore::lite::micro::test {
|
||||
TEST(GenerateCodeTest, mnist_x86) {
|
||||
const char *argv[] = {"./codegen", "--modelPath=../example/mnist/mnist.ms", "--moduleName=mnist", "--codePath=.",
|
||||
"--isWeightFile"};
|
||||
STATUS status = RunCoder(5, argv);
|
||||
ASSERT_EQ(status, RET_OK);
|
||||
}
|
||||
} // namespace mindspore::lite::micro::test
|
||||
|
||||
GTEST_API_ int main(int argc, char **argv) {
|
||||
testing::InitGoogleTest(&argc, argv);
|
||||
int ret = RUN_ALL_TESTS();
|
||||
return ret;
|
||||
}
|
||||
Loading…
Reference in New Issue