xavier输入法+卸载cuda10.2、cudnn8+快速安装cuda10.0、cudnn7+torch、torchvision
之前安装了高版本的cuda和cudnn，目前需要安装低版本的，操作如下：
1、配置输入法
1.sudo apt-get install fcitx-googlepinyin -y
2.settings----language support----keyboard input method system----fcitx
3.计算机重启
4.点击右上角键盘----ConfigureFcitx----此时应只有一个Keyboard-English----点击+----去掉only show current language前面的对号----输入google----选中Google Pinyin----ok----通过crtl+space中英文切换
2、卸载cuda10.2、cudnn8
1.卸载cuda，两种方法
（1）sudo apt-get remove cuda* 或 sudo apt --purge remove cuda*
（2）dpkg -l | grep -i cudav查看目前cuda安装包
sudo apt-get remove [cuda-visual-tools-10-2] (括号中为对对应的包，输入时括号去掉)
2.卸载cudnn
（1）sudo dpkg -l | grep cudnn
（2）sudo dpkg -r libcudnn8-samples
sudo dpkg -r libcudnn8-dev
sudo dpkg -r libcudnn8
Ubuntu——卸载cuda、cudnn的方法_William.csj的博客-CSDN博客_ubuntu卸载cuda
3.安装cuda10.0、cudnn7
1.sudo apt-get install cuda-toolkit-10-0
2.sudo apt-get install libcudnn7=7.6.3.28-1+cuda10.0
Jetson系列Ubuntu--CUDA、cudnn快速安装（Jetson Xavier NX上实践可行）_a111222zw的博客-CSDN博客_jetson安装cudnn
4.torch、torchvision安装
1.torch安装包下载，pip3 install torch-1.2.0a0+8554416-cp36-cp36m-linux_aarch64.whl安装包链接：PyTorch for Jetson - version 1.11 now available - Jetson Nano - NVIDIA Developer ForumsBelow are pre-built PyTorch pip wheel installers for Python on Jetson Nano, Jetson TX1/TX2, Jetson Xavier NX/AGX, and Jetson AGX Orin with JetPack 4.2 and newer. Download one of the PyTorch binaries from below for your…https://forums.developer.nvidia.com/t/pytorch-for-jetson-version-1-11-now-available/72048 2.torchvision安装
（1）git clone --branch v0.4.2 https://github.com/pytorch/vision torchvision
（2）cd vision
（3）export BUILD_VERSION=0.4.2
（4）sudo python3 setup.py install
5.opencv4.1.1卸载，源码安装支持cuda的opencv3.4.5
卸载原有opencv4.1.1
https://blog.csdn.net/qq_45948812/article/details/120376115
源码安装支持cuda的opencv3.4.5
https://blog.csdn.net/xiejiejun/article/details/120688635
出错Unable to locate package libjasper-dev
https://blog.csdn.net/ysten618/article/details/120082709