1、刷机完成后开始环境配置
- 基本环境配置:
- sudo apt-get update
- sudo apt-get install nvidia-jetpack
- sudo apt-get install python3-pip
2、设置SSD启动
- 首先讲SSD格式化(Disk工具)
- df -h (检查是否存在该硬盘分区)
- 使用 rootOnNVME工具 ( 参考链接Jetson Xavier NX – 从 SSD 运行_jetson nx ssd_AI边缘计算服务器_图为科技的博客-CSDN博客)
- sudo reboot
3、CUDA环境配置
- sudo gedit ~/.bashrc
- export CUDA_HOME=/usr/local/cuda
- export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
- export PATH=/usr/local/cuda/bin:$PATH
- source ~/.bashrc
- nvcc -V #查看版本
4、pip安装及换源
- sudo apt-get install python-pip
- pip install --upgrade pip
- sudo apt-get install python3-pip
- pip3 install --upgrade pip
- # pip 换源
- mkdir ~/.pip
- vim ~/.pip/pip.conf
[global]
index-url = https://pypi.tuna.tsinghua.edu.cn/simple
[install]
trusted-host = https://pypi.tuna.tsinghua.edu.cn
5、jtop安装
- sudo -H pip install jetson-stats
- sudo jtop
6、系统自带opencv不支持cuda加速,卸载重装
- sudo apt purge libopencv*
- sudo apt autoremove
- sudo apt update
7、系统的Cmake版本较低,进行高版本安装,不需要卸载,会自己覆盖。
- wget http://www.cmake.org/files/v3.13/cmake-3.13.0.tar.gz
- tar xpvf cmake-3.13.0.tar.gz cmake-3.13.0/
- cd cmake-3.13.0
- ./configure
- make
- sudo make install
- echo 'export PATH=/home/nvidia/cmake-3.13.0/bin/:$PATH' >> ~/. bashrc
- source ~/.bashrc
- 参考链接:【8-12】树莓派ubuntu升级Cmake_树梅派 ubuntu cmake 升级-CSDN博客
8、Opencv安装,主要在编译过程中出现了两种error
- 安装参考:Jetson带CUDA编译的opencv4.5安装教程与踩坑指南,cmake配置很重要!_jetson带cuda编译的opencv4.5安装教程与踩坑指南,cmake配置很重要!-CSDN博客
- error解决参考链接:【精选】Ubuntu系统opencv编译错误:Makefile:162:recipe for target ‘all‘ failed make:*** [all] Error 2_makefile:162: recipe for target 'all' failed-CSDN博客
- 下载对应版本的opencv-4.5.0.zip and opencv_contrilb-4.5.0.zip
- unzip opencv-4.5.0.zip and opencv_contrilb-4.5.0.zip
cd opencv-4.5.0
mkdir build && cd build - 编译(建议先把容易出现的error看一下,没有的文件提前下载到对应文件夹,不然有error还要重新编译,太浪费时间。)
cmake -D CMAKE_BUILD_TYPE=RELEASE\
-D CMAKE_INSTALL_PREFIX=/usr/local\
-D OPENCV_EXTRA_MODULES_PATH=../opencv_contrib/modules\
-D CUDA_ARCH_BIN='7.2'\ #查一下自己的板子对应的,安装参考链接有解释
-D WITH_CUDA=1\
-D BUILD_opencv_python3=ON -DBUILD_opencv_python2=ON\ #3.6
-D WITH_V4L=ON\
-D WITH_QT=ON\
-D WITH_OPENGL=ON\
-D CUDA_FAST_MATH=1\
-D WITH_CUBLAS=1\
-D OPENCV_GENERATE_PKGCONFIG=1\
-D WITH_GTK_2_X=ON\
-D WITH_GSTREAMER=ON .. - sudo make -j8
- 编译过程在出现的错误
# 在编译过程中出现的错误及解决办法
# error-1: fatal error: features2d/test/test_detectors_regression.impl.hpp: 没有那个文件或目录
#include "features2d/test/test_detectors_regression.impl.hpp"
解决办法:【精选】Ubuntu系统opencv编译错误:Makefile:162:recipe for target ‘all‘ failed make:*** [all] Error 2_makefile:162: recipe for target 'all' failed-CSDN博客# error-2: fatal error: features2d/test/test_detectors_regression.impl.hpp: 没有那个文件或目录
#include "features2d/test/test_detectors_regression.impl.hpp"
解决办法:opencv_config安装问题_fatal error: features2d/test/test_detectors_regres-CSDN博客 - 编译完成后
sudo make install
sudo make install /home/nvidia/opencv/opencv-4.5.0/samples/cpp/example_cmake
sudo cmake .
sudo make - 测试以下
sudo jtop
看看INFO界面是否之存在opencv及支持cuda - over
9、安装Qt5
-
sudo apt-get install qt5-default qtcreator -y
sudo apt-get install libqt5serialport5-dev libudev-dev
sudo apt-get install libqt5multimedia5-plugins
sudo apt-get install libcanberra-gtk-modulesudo apt-get install qt5-default
sudo apt-get install qtbase5-dev-tools qtcreator qt5-doc qt5-doc-html qtbase5-examples
sudo apt-get install qt5-assistant
sudo apt-get install qttools5-dev qttools5-dev-tools
sudo apt-get install qtmultimedia5-dev
sudo apt-get install qtquickcontrols2-5-dev
10、conda 安装
- # install conda conda 安装
- miniforge的GitHub地址:https://github.com/conda-forge/miniforge
- 清华源下载https://mirror.tuna.tsinghua.edu.cn/github-release/conda-forge/miniforge/LatestRelease/
- 例:Miniforge-pypy3-23.3.0-0-Linux-aarch64.sh 下载
- sh Miniforge-pypy3-23.3.0-0-Linux-aarch64.sh # install miniforge
- conda create -n torch_1.10 python=3.6.9
- conda activate torch_1.10
- conda remove -n torch_1.10 --all (这是删除环境,别看错了)
11、torch 安装
- # 依赖环境配置
- sudo apt-get install python3-pip libopenblas-base libopenmpi-dev
- pip install Cython
- #开始安装
- pip install numpy torch-1.10.0-cp36-cp36m-linux_aarch64.whl
- 安装过程中出现的问题
error:Illegal instruction (core dumped)
参考链接解办法:解决英伟达Jetson平台使用Python时的出现“Illegal instruction(cpre dumped)”错误_python illegal instruction-CSDN博客 - 具体步骤
sudo gedit ~/.bashrc
将“export OPENBLAS_CORETYPE=ARMV8”加入到“~/.bashrc”中
source ~/.bashrc
12、torchvision安装
- # 依赖环境配置
- sudo apt-get install libjpeg-dev zlib1g-dev
- 解压下载的对应安装包
- unzip vision-0.11.1.zip
- cd vision-0.11.1
- sudo gedit ~/.bashrc #修改bashrc文件
- export BUILD_VERSION=0.11.1 #把下边这一行添加到最后,#版本号是0.11.1就写0.11.1,是0.10.0就写0.10.0
- source ~/.bashrc # 刷新bashrc文件
- python3 setup.py install --user #开始编译
13、测试 torch and torchvision
- import torch
- import torchvision
- print(torch.__version__) # 查看torch版本
- print(torchvision.__version__) # 查看torchvision版本
- print(torch.cuda.is_available()) # 查看torch下cuda是否可用