Ubuntu18.04 配置SNPE并将ONNX模型转为DLC

文章目录0.前提条件1.ONNX下载安装2.SNPE下载3.安装SNPE相关依赖4.设置环境变量5.将ONNX模型转为DLC0.前提条件已安装好Anaconda和Python3.101.ONNX下载安装ONNX官方链接https://github.com/onnx/onnx#installation根据官方指导使用Conda进行安装condainstall-cconda-forge onnx2.SNPE下载下载地址为https://www.qualcomm.com/developer/software/neural-processing-sdk-for-ai点击Get Software 直接下载然后复制到ubuntu系统中并解压3.安装SNPE相关依赖进到刚才下载解压的snpe文件夹bin目录下cdv2.22.6.240515/qairt/2.22.6.240515/bin安装linux依赖sourcecheck-linux-dependency.sh成功后会提示All Dependency Packages FoundDone!!安装python依赖python check-python-dependency注这里有些库可能在国内无法下载修改check-python-dependency文件使用清华源即可vimcheck-python-dependency将如下代码subprocess.check_call([sys.executable,-m,pip,install,toBeInstalledPackagetoBeInstalledPackages[toBeInstalledPackage],],增加一行改为subprocess.check_call([sys.executable,-m,pip,install,toBeInstalledPackagetoBeInstalledPackages[toBeInstalledPackage],-i https://pypi.tuna.tsinghua.edu.cn/simple,],所有python依赖都安装完成后提示Summary:Package Recommended Installed~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~ ~~~~~~~~~~~absl-py 2.1.0 2.1.0attrs 23.2.0 23.1.0dash 2.12.1 2.17.1decorator 4.4.2 5.1.1invoke 1.7.3 2.2.0joblib 1.4.0 1.2.0jsonschema 4.19.0 4.19.2lxml 5.2.1 4.9.3mako 1.1.0 1.3.5matplotlib 3.3.4 3.8.0mock 3.0.5 5.1.0numpy 1.26.4 1.26.4opencv-python 4.5.4.58 4.9.0.80optuna 3.3.0 3.6.1packaging 24.0 23.1pandas 2.0.1 2.1.4paramiko 3.4.0 3.4.1pathlib2 2.3.6 2.3.7.post1pillow 10.2.0 10.2.0plotly 5.20.0 5.9.0protobuf 3.19.6 3.20.3psutil 5.6.4 5.9.0pytest 8.1.1 7.4.0pyyaml 5.3 6.0.1scikit-optimize 0.9.0 0.10.2scipy 1.10.1 1.11.4six 1.16.0 1.16.0tabulate 0.9.0 0.9.0typing-extensions 4.10.0 4.9.0xlsxwriter 1.2.2 3.2.04.设置环境变量执行指令sourceenvsetup.sh5.将ONNX模型转为DLC执行指令snpe-onnx-to-dlc-ixxx.onnx其中xxx.onnx为onnx文件所在位置成功完成后会提示2024-08-13 09:39:46,382 - 235 - INFO - INFO_INITIALIZATION_SUCCESS:注这里可能会遇到问题/home/gy/v2.22.6.240515/qairt/2.22.6.240515/bin/x86_64-linux-clang/snpe-onnx-to-dlc: Permission denied进入x86_64-linux-clang路径下查看发现snpe-onnx-to-dlc缺少执行权限-rw-r–r-- 1 root root 2931 May 16 01:29 snpe-onnx-to-dlc将所有snpe指令都加上执行权限就好啦cdx86_64-linux-clangchmodax snpe*