NVIDIA Jetson Xavier NX入门-镜像为jetpack4（3）——pytorch和torchvision安装

NVIDIA Jetson Xavier NX入门-镜像为jetpack4（3）——pytorch和torchvision安装

镜像为jetpack5系列：
NVIDIA Jetson Xavier NX入门-镜像为jetpack5（1）——镜像烧写
NVIDIA Jetson Xavier NX入门-镜像为jetpack5（2）——系统设置；环境搭建；CUDA；Cudnn配置；Ubuntu&pip&conda换源；miniforge安装
NVIDIA Jetson Xavier NX入门-镜像为jetpack5（3）——pytorch和torchvision安装
镜像为jetpack4系列：
NVIDIA Jetson Xavier NX入门-镜像为jetpack4（1）——镜像烧写
NVIDIA Jetson Xavier NX入门-镜像为jetpack4（2）——系统设置；环境搭建
NVIDIA Jetson Xavier NX入门-镜像为jetpack4（3）——pytorch和torchvision安装

jetpack5安装pytorch

参考：
https://www.bilibili.com/read/cv25497125/?spm_id_from=333.999.collection.opus.click
https://docs.nvidia.com/deeplearning/frameworks/install-pytorch-jetson-platform/#overview__section_xavier_nx

jetpack4请参考下面，jetpack请参考上方两个连接安装torch。

NVIDIA下载pytorch官方网址是：https://forums.developer.nvidia.com/t/pytorch-for-jetson-version-1-9-0-now-available/72048

下载好后，在下载文件目录下输入指令：

pip3 install torch-1.8.0-cp36-cp36m-linux_aarch64.whl

安装完毕后可以在终端输入以下命令检验PyTorch是否正确安装：

python -c 'import torch; print(torch.cuda.is_available())'

正常来说，会返回True。

出现报错1：博主本人在安装后出现错误(NVIDIA Jetson Xavier NX安装pytorch时import torch时报OSError: libmpi_cxx.so.20: cannot open shared object file: No such file or directory)：

import torch
Traceback (most recent call last):
File “”, line 1, in
File “/home/*/.local/lib/python3.6/site-packages/torch/init.py”, line 188, in
_load_global_deps()
File “/home/*/.local/lib/python3.6/site-packages/torch/init.py”, line 141, in _load_global_deps
ctypes.CDLL(lib_path, mode=ctypes.RTLD_GLOBAL)
File “/usr/lib/python3.6/ctypes/init.py”, line 348, in init
self._handle = _dlopen(self._name, mode)
OSError: libmpi_cxx.so.20: cannot open shared object file: No such file or directory

有库没有安装，输入指令：

sudo apt-get install libopenblas-base libopenmpi-dev 

出现报错2：Illegal instruction (core dumped)如果出现Illegal instruction (core dumped)的错误，这是由于numpy 1.19.5和OpenBLAS冲突引起的，修改其中一项即可。选择以下两种做法之一：
（1）降低numpy版本：

pip3 install -U numpy==1.19.4

（2）设置OpenBLAS

vim ~/.bashrc

加入：

export OPENBLAS_CORETYPE=ARMV8

激活.bashrc：

source ~/.bashrc

2. 安装torchvision

Pytorch除了torch包之外还需要安装对应的torchvision才能运行，两者的版本对应关系如下图：

输入指令：

sudo apt-get install libjpeg-dev zlib1g-dev libpython3-dev libavcodec-dev libavformat-dev libswscale-dev
git clone --branch <version> https://github.com/pytorch/vision torchvision   # see below for version of torchvision to download
cd torchvision
export BUILD_VERSION=0.x.0  # where 0.x.0 is the torchvision version  
python3 setup.py install --user
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

到此NVIDIA Jetson Xavier NX环境的基本配置已经完成，当出现无法定位软件包（Unable to locate package）的情况，首先使用sudo apt-get update指令。