Jetson Xavier安装ROS+CUDA+PyTorch

安装ROS

• 参考
  • ROS官网教程：http://wiki.ros.org/cn/melodic/Installation/Ubuntu
• 安装

  # 1.设置sources.list
  sudo sh -c '. /etc/lsb-release && echo "deb http://mirrors.tuna.tsinghua.edu.cn/ros/ubuntu/ `lsb_release -cs` main" > /etc/apt/sources.list.d/ros-latest.list'
  
  # 2.设置密钥
  sudo apt-key adv --keyserver 'hkp://keyserver.ubuntu.com:80' --recv-key C1CF6E31E6BADE8868B172B4F42ED6FBAB17C654
  
  # 3.更新Debian包索引
  sudo apt update
  
  # 4.安装ROS桌面完整版
  sudo apt install ros-melodic-desktop-full
  
  # 5.初始化rosdep
  sudo apt install rospack-tools
  sudo rosdep init
  rosdep update # 可能报错，但不影响roscore的使用
  
  # 6.将ROS环境变量添加到bash
  echo "source /opt/ros/melodic/setup.bash" >> ~/.bashrc
  source ~/.bashrc
  

安装CUDA

• 检查Jetson Xavier系统信息

  # 检查内核架构
  uname -m
  
  # 检查Linux版本
  cat /etc/*release
  
  # 检查JetPack
  sudo pip3 install jetson-stats
  sudo jtop
  

• 下载安装包

  # 在另一台Linux计算机上，下载环境管理工具NVIDIA SDK Manager
  
  # 访问并登录https://developer.nvidia.com
  # 点击'JetPack'(in Popular SDKs) -> 'FOR ANY JETSON DEVELOPER KIT'(in Installing JetPack) -> 'Download NVIDIA SDK Manager'
  # 可以查看'Install Jetson Software with SDK Manager'获取更多信息
  
  # 安装SDK Manager
  sudo dpkg -i XXX.deb
  sdkmanager
  
  # 下载Xavier需要的安装包
  # STEP 01: Jetson -> Host Machine -> Target Hardware Jetson AGX Xavier modules -> Linux JetPack 4.4 -> CONTINUE
  # STEP 02: I accept ... -> Download now. Install later. -> CONTINUE
  # STEP 03: Download packages
  # STEP 04: Finish
  
  # 拷贝安装包到Xavier上，这些包分别是
  # 'cuda-repo-l4t-10-2-local-10.2.89_1.0-1_arm64.deb'
  # 'libcudnn8_8.0.0.180-1+cuda10.2_arm64.deb'
  # 'libcudnn8-dev_8.0.0.180-1+cuda10.2_arm64.deb'
  

• 安装

  # 在Xavier上执行安装命令
  
  sudo dpkg -i cuda-repo-l4t-10-2-local-10.2.89_1.0-1_arm64.deb
  sudo dpkg -i libcudnn8_8.0.0.180-1+cuda10.2_arm64.deb
  sudo dpkg -i libcudnn8-dev_8.0.0.180-1+cuda10.2_arm64.deb
  
  sudo apt update
  sudo apt install cuda-toolkit-10-2
  

安装PyTorch和TorchVision

• 参考
  • 查看Jetson相关信息：http://www.gpus.cn/gpus_list_page_techno_support_content?id=39
  • 下载PyTorch安装包：https://elinux.org/Jetson_Zoo#PyTorch_.28Caffe2.29
  • PyTorch官网：https://github.com/pytorch/pytorch
  • TorchVision官网：https://github.com/pytorch/vision
• 安装

  # 根据Jetson版本下载安装包，例如JetPack 4.4 -> PyTorch v1.6.0 -> TorchVision v0.7.0
  
  sudo apt-get install libopenblas-base libopenmpi-dev
  sudo apt-get install python3-pip
  pip3 install Cython
  pip3 install numpy torch-1.6.0-cp36-cp36m-linux_aarch64.whl
  
  git clone --branch v0.7.0 https://github.com/pytorch/vision.git torchvision
  cd torchvision
  pip3 install setuptools
  sudo python3 setup.py install
  

• 测试

  python3
  import torch # no error appears
  print(torch.__version__) # get 1.6.0
  print(torch.cuda.is_available()) # get Ture
  import torchvision # no error appears
  

• 卸载（如果有必要的话）

  sudo pip3 uninstall torch
  sudo pip3 uninstall torchvision