前提系统烧录完成1、安装jetpack组件安装jetpack过程中会安装cuda、cudnn、TensorRT等cuDNN 安装路径 /usr/lib/aarch64-linux-gnuCUDA安装路径 /usr/local/cudasudo apt update sudo apt install nvidia-jetpack2、安装jtopsudo apt-get install python3-pip sudo pip3 install -U pip sudo pip3 install -U jetson-stats终端输入“jtop”会看到设备的信息3、配置CUDA路径sudo vim ~/.bashrc export CUDA_HOME/usr/local/cuda export PATH$CUDA_HOME/bin:$PATH export LD_LIBRARY_PATH$CUDA_HOME/lib64:$LD_LIBRARY_PATH source ~/.bashrc执行完命令后终端输入nvcc -V查看cuda信息nvcc -V4、配置cudnncd /usr/include sudo cp cudnn* /usr/local/cuda/include cd /usr/lib/aarch64-linux-gnu sudo cp libcudnn* /usr/local/cuda/lib64 sudo chmod 777 /usr/local/cuda/include/cudnn.h sudo chmod 777 /usr/local/cuda/lib64/libcudnn* cd /usr/local/cuda/lib64 sudo ln -sf libcudnn.so.8.6.0 libcudnn.so.8 sudo ln -sf libcudnn_ops_train.so.8.6.0 libcudnn_ops_train.so.8 sudo ln -sf libcudnn_ops_infer.so.8.6.0 libcudnn_ops_infer.so.8 sudo ln -sf libcudnn_adv_train.so.8.6.0 libcudnn_adv_train.so.8 sudo ln -sf libcudnn_adv_infer.so.8.6.0 libcudnn_adv_infer.so.8 sudo ln -sf libcudnn_cnn_train.so.8.6.0 libcudnn_cnn_train.so.8 sudo ln -sf libcudnn_cnn_infer.so.8.6.0 libcudnn_cnn_infer.so.8 sudo ldconfig测试cudnn是否配置成功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如果在sudo make clean sudo make 遇到如下报错则需要安装相应的库文件test.c:1:10: fatal error: FreeImage.h: No such file or directory 1 | #include FreeImage.h | ^~~~~~~~~~~~~ compilation terminated. WARNING - FreeImage is not set up correctly. Please ensure FreeImage is set up correctly. 运行以下命令后继续编译就会编译成功。sudo apt-get install libfreeimage3 libfreeimage-dev sudo make clean sudo make ./mnistCUDNN输出Test passed则cudnn配置完成5、pytorch安装sudo apt-cache show nvidia-jetpack #查看jetpack版本找对应的torch版本根据jetpack版本下载对应的torch版本pytorch下载的官网地址安装所需要的系统软件包sudo apt-get -y update sudo apt-get -y install autoconf bc build-essential g-8 gcc-8 clang-8 lld-8 gettext-base gfortran-8 iputils-ping libbz2-dev libc-dev libcgal-dev libffi-dev libfreetype6-dev libhdf5-dev libjpeg-dev liblzma-dev libncurses5-dev libncursesw5-dev libpng-dev libreadline-dev libssl-dev libsqlite3-dev libxml2-dev libxslt-dev locales moreutils openssl python-openssl rsync scons python3-pip libopenblas-dev系统安装包安装完成后进入到pytorch安装包目录执行命令pip install torch-1.12.0a02c916ef.nv22.3-cp38-cp38-linux_aarch64.whl6、torchvision安装安装所需要的软件包sudo apt-get update sudo apt-get upgrade sudo apt-get install libjpeg-dev zlib1g-dev libpython3-dev libavcodec-dev libavformat-dev libswscale-dev下载torchvision安装文件网络异常无法克隆的话进入torchvision文件地址git clone --branch v0.13.0 https://github.com/pytorch/vision torchvisiontorchvision配置cd torchvision export BUILD_VERSION0.13.0 python3 setup.py install --user出现下面标志表示安装结束在import torchvision 时报下面错重启设备就可以解决了。 import torchvision /home/nvidia/torchvision/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension: warn(fFailed to load image Python extension: {e}) /home/nvidia/torchvision/torchvision/__init__.py:28: UserWarning: You are importing torchvision within its own root folder (/home/nvidia/torchvision). This is not expected to work and may give errors. Please exit the torchvision project source and relaunch your python interpreter. warnings.warn(message.format(os.getcwd()))验证torch是否可以使用GPUtorch.cuda.is_available()(py38) nvidianvidia-desktop:~$ python Python 3.8.20 (default, Oct 3 2024, 15:18:56) [GCC 11.2.0] :: Anaconda, Inc. on linux Type help, copyright, credits or license for more information. import torch torch.cuda.is_available() True7、conda虚拟环景中配置TensorRT安装jetpack时已经安装了tensorRT只需要将tensorrt与本地进行软连接就可以在虚拟环境中使用了sudo ln -s /usr/lib/python3.8/dist-packages/tensorrt* /home/nvidia/.local/lib/python3.8/site-packages/或者也可以将TensorRT文件复制到虚拟环境的包文件下