一个demo，mindspore lite 部署在树莓派4B ubuntu22.04中，为后续操作开个门！

环境

• 开发环境：wsl-ubuntu22.04分发版
• 部署环境：树莓派4B，操作系统为ubuntu22.04
• mindspore lite版本：mindspore-lite-2.4.1-linux-aarch64.tar.gz
• demo模型：mnist手写数字识别

步骤

1. 准备交叉编译环境

• 安装交叉编译器

sudo apt install gcc-aarch64-linux-gnu
# 验证是否安装成功
(base) joke@ShineZhang:~/mindspore_deploy_demo_aarch64$ aarch64-linux-gnu-gcc --version
aarch64-linux-gnu-gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Copyright (C) 2021 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

• 设置交叉编译环境

# file: toolchain.cmake
set(CMAKE_SYSTEM_NAME Linux)
set(CMAKE_SYSTEM_PROCESSOR aarch64)
set(CMAKE_C_COMPILER aarch64-linux-gnu-gcc)

2. 下载demo并交叉编译

• 下载

# 有用的话给个star呗
git clone https://gitee.com/Shine_Zhang/mindspore_aarch64_runtime_demo.git

• 文件结构说明：

.
├── CMakeLists.txt 
├── README.md
├── build.sh # 执行编译
├── main.c
├── model
│   └── mnist.ms # mnist模型文件
└── toolchain.cmake # 编译工具链

• 模型文件转换：
  MindSpore Lite端侧模型转换

执行交叉编译：

chomd +x build.sh
./build.sh

3. 部署环境配置

• 准备环境
  编译后目录如下：

.
├── CMakeLists.txt
├── README.md
├── bin
│   └── mindspore_mnist_demo # es
├── build
│   ├── CMakeCache.txt
│   ├── CMakeFiles
│   ├── Makefile
│   ├── bin
│   ├── cmake_install.cmake
│   ├── mindspore-lite-2.4.1-linux-aarch64
│   ├── mindspore-lite-2.4.1-linux-aarch64.tar.gz
│   ├── mindspore_mnist_demo
│   └── mindspore_quick_start_c
├── build.sh
├── include
│   ├── api
│   ├── c_api
│   ├── dataset
│   ├── ir
│   ├── kernel_interface.h
│   ├── mindapi
│   ├── registry
│   ├── schema
│   └── third_party
├── lib
│   ├── libmindspore-lite.so # es
│   ├── libmindspore_glog.so -> /home/mindspore_deploy_demo_aarch64/lib/libmindspore_glog.so.0
│   └── libmindspore_glog.so.0 # es
├── main.c
├── model
│   └── mnist.ms # es
└── toolchain.cmake

将目录中标记es的文件拷贝至要部署的树莓派4B中

• 配置运行环境

将libmindspore-lite.so和libmindspore_glog.so.0拷贝至/usr/local/lib

# 编辑~/.bashrc 增加环境变量
vim ~/.bashrc

# ~/.bashrc 末尾增加以下内容后保存：
export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH

# 重新加载当前用户的 ~/.bashrc 配置文件
source ~/.bashrc

# 检验动态库是否加载成功
joke in ~ λ ldd mindspore_mnist_demo
        linux-vdso.so.1 (0x0000ffffb580b000)
        libmindspore-lite.so => /usr/local/lib/libmindspore-lite.so (0x0000ffffb5040000)
        libc.so.6 => /lib/aarch64-linux-gnu/libc.so.6 (0x0000ffffb4e90000)
        /lib/ld-linux-aarch64.so.1 (0x0000ffffb57d2000)
        libmindspore_glog.so.0 => /usr/local/lib/libmindspore_glog.so.0 (0x0000ffffb4e30000)
        libdl.so.2 => /lib/aarch64-linux-gnu/libdl.so.2 (0x0000ffffb4e10000)
        libstdc++.so.6 => /lib/aarch64-linux-gnu/libstdc++.so.6 (0x0000ffffb4be0000)
        libm.so.6 => /lib/aarch64-linux-gnu/libm.so.6 (0x0000ffffb4b40000)
        libgcc_s.so.1 => /lib/aarch64-linux-gnu/libgcc_s.so.1 (0x0000ffffb4b10000)
        libpthread.so.0 => /lib/aarch64-linux-gnu/libpthread.so.0 (0x0000ffffb4af0000)

4. 执行demo

joke in ~ λ ./mindspore_mnist_demo mnist.ms           
----------------------------------------------------------
Model Inputs (1):
  Input 0:
    Name: serving_default_keras_tensor:0
    Shape: [1, 28, 28]
    Data Type: 43
    Data Size: 3136 bytes

Model Outputs (1):
  Output 0:
    Name: StatefulPartitionedCall_1:0
    Shape: [-1]
    Data Type: 43
    Data Size: 0 bytes

Estimated Model Size: 3136 bytes
----------------------------------------------------------
inputs: Tensor 1:
Tensor shape: [1, 28, 28]
outputs: Tensor 1:
Tensor shape: [1, 10]
outputs[1]: -5.782
outputs[2]: 0.069
outputs[3]: 1.802
outputs[4]: -12.207
outputs[5]: -7.331
outputs[6]: -18.879
outputs[7]: 6.847
outputs[8]: -5.083
outputs[9]: -4.575
Predicted digit: 7 # 预测结果为数字 7