Jetson Orin Nano系统烧录教程
1 系统烧录
1.1 方式1：使用SD卡烧录
1.1.1 相关下载
推荐在win系统下进行烧录，方便SD卡的格式化和镜像软件的烧录。
（1）SD卡格式化软件下载：https://www.sdcard.org/downloads/formatter/sd-memory-card-formatter-for-windows-download/
（2）镜像烧录软件下载：https://etcher.balena.io/#download-etcher
（3）jetpack5.1.2 镜像下载：https://developer.nvidia.com/embedded/jetpack-sdk-512
1.1.2 烧录步骤
先将SD卡进行格式化，然后将jetpack5.1.2镜像使用烧录软件实现系统烧录。
1.2 方式2：使用SDK Manager烧录
1.2.1 下载SDK Manager
官网链接：https://developer.nvidia.com/sdk-manager
- 短接FC_REC和GND（第二三引脚），接上TypeC数据线，然后接通电源
学习文档：https://docs.nvidia.com/jetson/archives/r34.1/DeveloperGuide/text/HR/JetsonDeveloperKitSetup.html
1.2.2 烧录步骤
1、准备一台主机、待烧录的Jetson Orin Nano开发板、SD卡、数据线
2、在主机Ubuntu下载SDK Manager
3、通过数据线将主机和Orin 来连接，短接Nano(如果没有短接帽，使用钥匙也可以实现短接)
4、按照Manager流程烧录
2 Orin Nano开机缓慢问题
【！！！文中采用 关闭开机启动项的方式 解决开机缓慢问题，非必要别动！！不然 系统 会崩！
一般情况下，采用 方式1 使用SD卡烧录系统的开机时间在一分钟以内，不需要解决开机缓慢问题】
2.1 查看开机启动项
systemd-analyze blame
会按照时间长短顺序显示服务启动项：
9.871s alsa-restore.service
7.216s networkd-dispatcher.service
6.057s nv.service
5.725s ModemManager.service
3.450s dev-mmcblk1p1.device
2.769s docker.service
2.096s apt-daily-upgrade.service
2.062s accounts-daemon.service
1.959s nv-l4t-usb-device-mode.service
1.701s udisks2.service
1.613s avahi-daemon.service
1.578s NetworkManager.service
1.500s snapd.service
1.455s nv-l4t-bootloader-config.service
1.381s polkit.service
1.279s apt-daily.service
1.248s containerd.service
1.183s apport.service
1.181s switcheroo-control.service
1.110s wpa_supplicant.service
1.099s ua-timer.service
1.093s systemd-logind.service
998ms resolvconf-pull-resolved.service
913ms nvpower.service
845ms rsyslog.service
710ms user@124.service
680ms systemd-resolved.service
577ms binfmt-support.service
539ms systemd-modules-load.service
528ms nvpmodel.service
515ms dev-hugepages.mount
510ms dev-mqueue.mount
502ms run-rpc_pipefs.mount
501ms kerneloops.service
495ms systemd-udev-trigger.service
494ms sys-kernel-debug.mount
490ms sys-kernel-tracing.mount
467ms keyboard-setup.service
457ms kmod-static-nodes.service
442ms modprobe@chromeos_pstore.service
433ms nvphs.service
426ms modprobe@efi_pstore.service
422ms modprobe@pstore_blk.service
419ms modprobe@ramoops.service
418ms modprobe@pstore_zone.service
415ms nvfb-udev.service
413ms e2scrub_reap.service
392ms systemd-journald.service
336ms ssh.service
325ms nvfb-early.service
324ms nv_nvsciipc_init.service
321ms systemd-remount-fs.service
306ms bluetooth.service
287ms gdm.service
272ms pppd-dns.service
253ms systemd-udevd.service
243ms systemd-timesyncd.service
209ms systemd-update-utmp.service
199ms systemd-random-seed.service
197ms systemd-tmpfiles-clean.service
193ms snapd.seeded.service
168ms user@1000.service
164ms systemd-tmpfiles-setup-dev.service
162ms systemd-sysusers.service
159ms openvpn.service
156ms console-setup.service
154ms motd-news.service
153ms proc-sys-fs-binfmt_misc.mount
135ms plymouth-read-write.service
131ms systemd-tmpfiles-setup.service
126ms colord.service
117ms upower.service
115ms nfs-config.service
103ms systemd-sysctl.service
89ms systemd-journal-flush.service
85ms docker.socket
85ms rpcbind.service
75ms systemd-user-sessions.service
74ms ubuntu-fan.service
68ms setvtrgb.service
54ms nvfb.service
51ms snapd.socket
47ms sys-kernel-config.mount
34ms rtkit-daemon.service
30ms user-runtime-dir@124.service
19ms user-runtime-dir@1000.service
13ms plymouth-quit-wait.service
12ms systemd-update-utmp-runlevel.service
10ms systemd-rfkill.service
6ms sys-fs-fuse-connections.mount
2.2 禁用服务
//禁用服务
sudo systemctl disable gdm.service 注意此服务是桌面服务，要确定自己是否需用
sudo systemctl disable NetworkManager-wait-online.service
sudo systemctl disable alsa-restore.service
sudo systemctl disable docker.service
sudo systemctl disable cron-daily.service
sudo systemctl disable bluetooth.service
//然后重启
reboot
2.3 启用服务
//启用服务
sudo systemctl enable cron-daily.service
//然后重启
reboot
2.4 对Orin Nano的禁用服务
//sudo systemctl disable NetworkManager.service
sudo systemctl disable alsa-restore.service
sudo systemctl disable docker.service
sudo systemctl disable apt-daily-upgrade.service
sudo systemctl disable apt-daily.service
sudo systemctl disable bluetooth.service
3 环境配置
Jetpack 5.1.2
Ubuntu20.04
3.1 配置CUDA
# open环境变量：
vim ~/.bashrc #或者采用 source gedit ~/.bashrc
#在最后写入并保存：
export CUDA_HOME=/usr/local/cuda-11.4
export PATH=$PATH:$CUDA_HOME/bin
export LD_LIBRARY_PATH=/usr/local/cuda-11.4/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
#使其生效：
source ~/.bashrc
#查看是否生效：
cat /proc/driver/nvidia/version
nvcc -V
# ~/.bahrc文件中的内容还可以填：
export CUDA_HOME=/usr/local/cuda
export PATH=${CUDA_HOME}/bin:${PATH}
export LD_LIBRARY_PATH=${CUDA_HOME}/lib64:$LD_LIBRARY_PATH
将对应的头文件、库文件放到cuda目录
cuDNN的头文件在：/usr/include，库文件位于：/usr/lib/aarch64-linux-gnu
将头文件与库文件复制到cuda目录下：
cd /usr/include && sudo cp cudnn.h /usr/local/cuda/include
cd /usr/lib/aarch64-linux-gnu && sudo cp libcudnn* /usr/local/cuda/lib64
修改文件权限，修改复制完的头文件与库文件的权限，所有用户都可读，可写，可执行：
sudo chmod 777 /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn*
重新链接：
cd /usr/local/cuda/lib64
sudo ln -sf libcudnn.so.8.4.0 libcudnn.so.8
sudo ln -sf libcudnn_ops_train.so.8.4.0 libcudnn_ops_train.so.8
sudo ln -sf libcudnn_ops_infer.so.8.4.0 libcudnn_ops_infer.so.8
sudo ln -sf libcudnn_adv_infer.so.8.4.0 libcudnn_adv_infer.so.8
sudo ln -sf libcudnn_cnn_infer.so.8.4.0 libcudnn_cnn_infer.so.8
sudo ln -sf libcudnn_cnn_train.so.8.4.0 libcudnn_cnn_train.so.8
sudo ln -sf libcudnn_adv_train.so.8.4.0 libcudnn_adv_train.so.8
sudo ldconfig
测试cuDNN:
sudo cp -r /usr/src/cudnn_samples_v8/ ~/
cd ~/cudnn_samples_v8/mnistCUDNN
sudo chmod 777 ~/cudnn_samples_v8
sudo make clean && sudo make
./mnistCUDNN
3.2 依赖环境配置
# update system and install depends
sudo apt-get update
# 安装JetPack组件包，其中包括了Cuda、CuDNN和TensorRT
sudo apt install nvidia-jetpack
sudo apt-get install libhdf5-serial-dev hdf5-tools zlib1g-dev zip libjpeg8-dev libopenblas-dev python3-pip
sudo apt-get install libfreeimage3 libfreeimage-dev
#分别执行以下命令，即可查看自己的jetson nano 预搭载的CUDA版本
sudo apt-get install python3-pip
sudo pip3 install jetson-stats
sudo jtop
3.3 安装conda
清华镜像：
https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/
可下载最新版本的：
| Anaconda3-2023.09-0-Linux-aarch64.sh | 838.8 MiB | 2023-09-29 23:47 |
|---|
conda create -n detect python=3.8
3.4 安装torch 和 torchvision
torch: '2.1.0a0+41361538.nv23.06'
torchvision: '0.15.2a0+fa99a53'
The method of torchvision install:
$ sudo apt-get install libjpeg-dev zlib1g-dev libpython3-dev libavcodec-dev libavformat-dev libswscale-dev
$ git clone --branch v0.15.2 https://ghproxy.com/https://github.com/pytorch/vision torchvision # see below for version of torchvision to download
$ cd torchvision
$ export BUILD_VERSION=0.15.2 # where 0.x.0 is the torchvision version
$ python3 setup.py install --user #long time for it
$ cd ../ # attempting to load torchvision from build dir will result in import error
$ pip install 'pillow<7' # always needed for Python 2.7, not needed torchvision v0.5.0+ with Python 3.6
3.5 查看系统是否存在tensorrt
dpkg -l | grep TensorRT
sudo vim ~/.bashrc
export PATH=/usr/src/tensorrt/bin:$PATH
source ~/.bashrc
注：
1、在使用SD卡镜像烧录时，会自带TensorRT，不需要用户后期再安装，如果import tensorrt找不到tensorrt，执行下面的命名试试：
# 安装JetPack组件包，其中包括了Cuda、CuDNN和TensorRT
sudo apt install nvidia-jetpack
2、资源有版本为8.2.3.0的tensorrt安装包whl。
扫一扫
1366
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
