NVIDIA Jetson AGX Orin计算平台环境搭建

一、环境介绍

        芯片型号：NVIDIA Jetson AGX Orin(32GB ram)

        AI算力：200TOPS

        架构:aarch64

        系统版本：Ubuntu20.04 focal

        内核版本：5.10.104-1-2-0-Tegra

        Python版本：3.8.10

二、任务介绍

        完成新环境搭建，包括CUDA版OpenCV，pytorch，torchvision;cuda,cudnn,tensorrt,jtop等

三、环境搭建

1.jtop安装

         jtop是Jetson系列的系统监控工具，使用以下指令安装：

sudo apt install python3-pip
sudo -H pip3 install -U pip
sudo -H pip install jetson-stats

        使用以下指令开启工具：

jtop

        显示界面如下：

        环境搭建时主要看的是数字7INFO中的内容，其中显示了大部分的版本号信息。 

2.cuda、cudnn、tensorrt安装

        Jetpack是Jetson板上的开发工具包，apt安装Jetpack时会默认安装适配版本的cuda、cudnn、tensorrt。

        1）检查L4T(Linux for Tegra)版本 

cat /etc/nv_tegra_release

       # R35 (release), REVISION: 1.0, GCID: 31346300, BOARD: t186ref, EABI: aarch64, DATE: Thu Aug 25 18:41:45 UTC 2022

        R35 (release), REVISION: 1.0代表L4T是35.1.0版本

        添加源:(不添加源后面安装依赖会报错Unable to locate package nvidia-jetpack)

sudo bash -c 'echo "deb https://repo.download.nvidia.com/jetson/common r35.1 main" >> /etc/apt/sources.list.d/nvidia-l4t-apt-source.list'
sudo bash -c 'echo "deb https://repo.download.nvidia.com/jetson/t234 r35.1 main" >> /etc/apt/sources.list.d/nvidia-l4t-apt-source.list'

        2）开始安装

sudo apt upgrade
sudo apt update
sudo apt dist-upgrade
sudo reboot
sudo apt install nvidia-jetpack

        3）安装验证

        jtop中查看7INFO中是否正确显示版本号，如下图则安装成功：

         

3.OpenCV安装

        1）版本检查

        jtop中查看OpenCV版本，如果显示带有‘-with CUDA:NO’则表示自带的OpenCV不支持CUDA加速，需要卸载重装。

        2）卸载原版OpenCV

        卸载默认版本OpenCV：

sudo apt purge libopencv*
sudo apt autoremove
sudo apt update

        如果有自己后安装的OpenCV，则使用以下指令卸载（假设源码编译安装的OpenCV4.6.0）:

# 进入编译安装目录
cd .../opencv-4.6.0/build
sudo make uninstall
cd .../opencv-4.6.0
sudo rm -r build

        3）安装依赖以及解决报错

# 依次输入以下命令
sudo apt install -y build-essential checkinstall cmake pkg-config yasm git gfortran
sudo apt update
sudo apt install -y libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev
sudo apt install -y libjpeg8-dev libjasper-dev libpng12-dev libtiff5-dev libavcodec-dev libavformat-dev libswscale-dev libdc1394-22-dev libxine2-dev libv4l-dev
sudo apt install -y libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev libgtk2.0-dev libtbb-dev libatlas-base-dev libfaac-dev libmp3lame-dev libtheora-dev libvorbis-dev libxvidcore-dev libopencore-amrnb-dev libopencore-amrwb-dev x264 v4l-utils
sudo apt install python-dev python-numpy libtbb2 libtbb-dev libjpeg-dev libpngdev libtiff-dev libjasper-dev libdc1394-22-dev

# 以防上面保存为pdf以后显示不全
sudo apt install -y build-essential checkinstall cmake pkg-config yasm git gfortran
sudo apt update
sudo apt install -y libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev
sudo apt install -y libjpeg8-dev libjasper-dev libpng12-dev libtiff5-dev libavcodec-dev libavformat-dev libswscale-dev libdc1394-22-dev libxine2-dev libv4l-dev
sudo apt install -y libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev libgtk2.0-dev libtbb-dev libatlas-base-dev libfaac-dev libmp3lame-dev libtheora-dev libvorbis-dev libxvidcore-dev libopencore-amrnb-dev libopencore-amrwb-dev x264 v4l-utils
sudo apt install python-dev python-numpy libtbb2 libtbb-dev libjpeg-dev libpngdev libtiff-dev libjasper-dev libdc1394-22-dev 

         ·如果第一条命令就报错则需要先换源：

        修改文件/etc/apt/source.list,改为下面的源：

# 默认注释了源码仓库，如有需要可自行取消注释
deb [arch=arm64] http://mirrors.ustc.edu.cn/ubuntu-ports/ focal main restricted universe multiverse
# deb-src http://mirrors.ustc.edu.cn/ubuntu-ports/ focal main main restricted universe multiverse
deb [arch=arm64] http://mirrors.ustc.edu.cn/ubuntu-ports/ focal-updates main restricted universe multiverse
# deb-src http://mirrors.ustc.edu.cn/ubuntu-ports/ focal-updates main restricted universe multiverse
deb [arch=arm64] http://mirrors.ustc.edu.cn/ubuntu-ports/ focalbackports main restricted universe multiverse
# deb-src http://mirrors.ustc.edu.cn/ubuntu-ports/ focal-backports main restricted universe multiverse
deb [arch=arm64] http://mirrors.ustc.edu.cn/ubuntu-ports/ focalsecurity main restricted universe multiverse
# deb-src http://mirrors.ustc.edu.cn/ubuntu-ports/ focal-security main restricted universe multiverse

        改完以后sudo apt-get update即可。 

        ·倒数第三条命令如果报错‘无法定位libjasper-dev’：

sudo add-apt-repository "deb http://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ xenial main multiverse restricted universe"
sudo apt update
sudo apt install libjasper1 libjasper-dev

        ·最后一条命令python-dev报错： 

        python-dev改成python3-dev，重新运行命令 。

        4）下载OpenCV以及OpenCV-contrib源码

        OpenCV：Releases · opencv/opencv (github.com)

        OpenCV-contrib：Tags · opencv/opencv_contrib · GitHub

        因为我的Python版本是3.8，所以下载OpenCV和contrib都下的4.6.0版本

 

 

         5）编译安装OpenCV

        ·把OpenCV和OpenCV-contrib解压，把OpenCV-contrib解压后的文件夹改名成opencv_contrib后放到OpenCV的文件夹。

# 进入opencv文件夹
cd opencv-4.6.0
# 建立编译文件夹
mkdir build
cd build

         ·执行cmake构建

cmake -D CMAKE_BUILD_TYPE=RELEASE \
-D CMAKE_INSTALL_PREFIX=/usr/local \
-D OPENCV_EXTRA_MODULES_PATH=../opencv_contrib/modules \
-D CUDA_ARCH_BIN='8.7' \
-D CUDA_ARCH_PTX='8.7' \
-D WITH_CUDA=1 \
-D BUILD_opencv_python3=1 \
-D BUILD_opencv_python2=1 \
-D WITH_FFMPEG=1 \
-D WITH_V4L=ON \
-D WITH_QT=ON \
-D WITH_OPENGL=ON \
-D ENABLE_FAST_MATH=1 \
-D CUDA_FAST_MATH=1 \
-D WITH_CUBLAS=1 \
-D OPENCV_GENERATE_PKGCONFIG=1 \
-D WITH_GTK_2_X=ON \
-D WITH_GSTREAMER=ON \
..

        其中：

                CMAKE_BUILD_TYPE=RELEASE $~~~~~~$ #表示编译发布版本

                CMAKE_INSTALL_PREFIX$~~~~~~$  #指定安装路径

                OPENCV_EXTRA_MODULED_PATH $~~~~~~$ # opencv_contrib路径

                CUDA_ARCH_BIN $~~~~~~$ #指定GPU算力，在jtop的INFO中查看（很重要）

                WITH_CUDA $~~~~~~$ #使用CUDA加速

                WITH_GSTREAMER $~~~~~~$ #使用OpenCV的gstreamer包

        执行完之后再build目录下查找download_with_wget.sh文件，如果文件不存在说明全部下载成功。

        

        ·编译OpenCV

        注意：如果有安装anaconda，建议关闭conda环境，进入主机环境后再编译：

conda deactivate

        因为conda屏蔽了某些主机环境的库，导致编译时链接到了conda的库而冲突报错。

             

# 开始编译,按设备配置设置编译时使用的线程数，越多越快
make -j8

        编译完成后：

# 开始安装
sudo make install 

        6）安装验证

         在jtop中查看OpenCV的版本以及后面是否为‘-with CUDA:YES’

        python验证：

python
import cv2
print(cv2.__version__)

        输出OpenCV版本则表示安装成功

        如果报错'No module named 'cv2''找不到则参考手动编译CUDA版OpenCV4.6.0，在python中报错‘No module named ‘cv2‘’的解决方法-CSDN博客

4.pytorch,torchvision安装

        1）安装依赖

# 配置设备环境
sudo apt-get install libhdf5-serial-dev hdf5-tools libhdf5-dev zlib1g-dev zip libjpeg8-dev liblapack-dev libblas-dev gfortran

# 安装其他库
pip install -U numpy==1.19.4 future==0.18.2 mock==3.0.5 h5py==2.10.0 keras_preprocessing==1.1.1 keras_applications==1.0.8 gast==0.2.2 futures protobuf pybind11

        2）安装pytorch

       python,pytorch,torchvision对应关系表：

torch	torchvision	Python
main / nightly	main / nightly	>=3.8, <=3.11
2.1	0.16	>=3.8, <=3.11
2.0	0.15	>=3.8, <=3.11
1.13	0.14	>=3.7.2, <=3.10

1.12	0.13	>=3.7, <=3.10
1.11	0.12	>=3.7, <=3.10
1.10	0.11	>=3.6, <=3.9
1.9	0.10	>=3.6, <=3.9
1.8	0.9	>=3.6, <=3.9
1.7	0.8	>=3.6, <=3.9
1.6	0.7	>=3.6, <=3.8
1.5	0.6	>=3.5, <=3.8
1.4	0.5	==2.7, >=3.5, <=3.8
1.3	0.4.2 / 0.4.3	==2.7, >=3.5, <=3.7
1.2	0.4.1	==2.7, >=3.5, <=3.7
1.1	0.3	==2.7, >=3.5, <=3.7
<=1.0	0.2	==2.7, >=3.5, <=3.7

           我是python3.8所以安装pytorch1.12.0和torchvision0.13.0     

        下载：https://developer.download.nvidia.com/compute/redist/jp/v50/pytorch/torch-1.12.0a0+2c916ef.nv22.3-cp38-cp38-linux_aarch64.whl

        下载完以后:

# pip安装pytorch
pip install torch-1.12.0a0+2c916ef.nv22.3-cp38-cp38-linux-aarch64.whl

        安装完以后：

# 验证安装
python
import torch
print(torch.__version__)

        3）安装torchvision

        因为torchvision没有支持CUDA版本的whl包，所以需要下载源码编译安装：

        下载：

git clone --branch v0.13.0 https://github.com/pytorch/vision torchvision

        配置环境：

sudo apt-get install libopenblas-base libopenmpi-dev
pip install Cython

        进入文件夹编译安装：

cd torchvision
export BUILD_VERSION=0.13.0    
python setup.py install --user

        安装完以后验证：

python
import torchvision
print(torchvision.__version__)

        

        至此，所需环境基本安装完毕。