Jetson环境安装(一):Ubuntu18.04安装pytorch、opencv、onnx、tensorflow、setuptools、pycuda....

1.安装pytorch

• 对于安装pytorch你要结合你的cuda的版本，以及你的python版本，由于我是jetson nano的平台，所以安装的时候要注意有aarch64的后缀，pytorch就是蛮难下下来的，需要你翻墙才能下载，下面我会给出相应的pytorch的腾讯微云分享版本

这是全部的下载链接，直接找对应下载:
https://elinux.org/Jetson_Zoo#PyTorch_.28Caffe2.29

首先给出下载pytorch的代码，如果运行不起就直接https://nvidia.box.com/shared/static/j2dn48btaxosqp0zremqqm8pjelriyvs.whl

wget https://nvidia.box.com/shared/static/j2dn48btaxosqp0zremqqm8pjelriyvs.whl -O torch-1.6.0-cp36-cp36m-linux_aarch64.whl
 
sudo pip3 install torch-1.6.0-cp36-cp36m-linux_aarch64.whl

• ImportError: libopenblas.so.0: cannot open shared object file: No such file or directory
• solve: sudo apt-get install libopenblas-dev
• OSError: libmpi_cxx.so.20: cannot open shared object file: No such file or directory
• solve:
  

安装完了之后就可以通过下面的这个验证

import torch
print(torch.__version__)
print('CUDA available: ' + str(torch.cuda.is_available()))
a = torch.cuda.FloatTensor(2).zero_()
print('Tensor a = ' + str(a))
b = torch.randn(2).cuda()
print('Tensor b = ' + str(b))
c = a + b
print('Tensor c = ' + str(c))

2.安装torchvision

# 安装一些必要的依赖包
sudo apt-get install libjpeg-dev zlib1g-dev
 
# 下载torchvision
git clone --branch v0.7.0 https://gitee.com/zero-one-game/vision torchvision   
 
# 安装torchvision
cd torchvision
sudo python3 setup.py install
# OSError: libmpi_cxx.so.20: cannot open shared object file: No such file or directory  ->>>>sudo apt-get install libopenmpi2
cd ../
pip3 install 'pillow<7'

• error: command ‘aarch64-linux-gnu-gcc‘ failed with exit status 1
• way1: sudo apt-get install python3-dev python3-pip libxml2-dev libxslt1-dev zlib1g-dev libffi-dev libssl-dev
• way2: if has ffmpeg 改为if False即可

用以下代码测试是否成功

import torchvision
 
print(torchvision.__version__) 

在Jetson Xavier NX上安装torchvision编译报错：fatal error: libavcodec/avcodec.h: No such file or directory
look at the authou csdn

3.安装opencv-python（自带有）

一条命令直接搞定

sudo apt-get install python3-opencv

通过终端验证直接通过

Python 3.6.9 (default, Oct  8 2020, 12:12:24) 
[GCC 8.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import cv2
>>> 

若要删除用sudo apt-get remove python3-opencv

4.安装onnx/onnxruntime

• 先安装onnx的环境依赖
  sudo apt-get install protobuf-compiler libprotoc-dev
• 再安装onnx为1.9.0的版本
  pip install onnx1.9.0
  pip3 install onnxruntime1.9.0 -i https://pypi.tuna.tsinghua.edu.cn/simple/ --trusted-host pypi.tuna.tsinghua.edu.cn

  Created wheel for onnx: filename=onnx-1.4.1-cp36-cp36m-linux_aarch64.whl size=3646291 sha256=3632a794d71e5c744920b2abf7626a4fd3ca39b7ea47a85e7cb9f3a7c325dbf7
  Stored in directory: /home/lqs/.cache/pip/wheels/45/d1/9f/9508c4176d79c0373629eff6cf2645e2dcbfb291bbe9dc2f8a
Successfully built onnx
Installing collected packages: onnx
Successfully installed onnx-1.4.1

5.安装tensorflow-gpu

if your platform is jetson xavier nx,I think you should see the offcial word,the website is :https://forums.developer.nvidia.com/t/official-tensorflow-for-jetson-agx-xavier/65523

jetson nano look at here

• 我这里安装的是tensorflow-gpu1.15版本的，下面是下载的地址
  https://developer.download.nvidia.com/compute/redist/jp/v42
  下好了之后在终端输入
  先安装依赖再安装tensorflow：

1.sudo apt install python3-pip python3-dev
2.python3 -m pip install --upgrade pip -i https://pypi.tuna.tsinghua.edu.cn/simple
#使用下述指令可以更改python镜像源为清华pypi镜像源
3.pip3 config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
4.#安装机器学习常用包 
sudo apt install python3-scipy -y
sudo apt install python3-pandas -y
sudo apt install python3-sklearn -y
sudo apt install python3-seaborn -y
sudo pip install cython
上述指令会自动安装numpy和matplotlib等依赖包
5.
sudo apt-get update
sudo apt install libhdf5-serial-dev hdf5-tools libhdf5-dev zlib1g-dev zip libjpeg8-dev liblapack-dev libblas-dev gfortran
6.#安装Python依赖包
sudo pip3 install -U numpy==1.16.1 future==0.18.2 mock==3.0.5 h5py==2.10.0 keras_preprocessing==1.1.1 keras_applications==1.0.8 gast==0.2.2 futures protobuf pybind11 astor
7.#download tensorflow https://developer.download.nvidia.com/compute/redist/jp/v45 
pip3 install tensorflow-1.15.5+nv21.4-cp36-cp36m-linux_aarch64.whl --user
8.test
import tensorflow as tf
# 输出提示：
# 2020-10-11 15:25:36.253267: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.2

a = tf.constant(1.)
b = tf.constant(2.)
print(a+b)
# 输出结果：
# tf.Tensor(3.0, shape=(), dtype=float32)
 
print('GPU:', tf.test.is_gpu_available())
# 输出最后一句为：
# GPU: True

#this is NV
pip3 install tensorflow-1.15.5+nv21.3-cp36-cp36m-linux_aarch64.whl 
# jetson xavier nx  jetpack 4.5.2 cuda 10.2 

• 验证
  

6.安装setuptools

pip install --no-cache-dir setuptools==20.7.0 -i https://pypi.doubanio.com/simple/
#下面是升级 
pip install --ignore-installed  setuptools

7.安装Pycuda

下载地址
https://files.pythonhosted.org/packages/5e/3f/5658c38579b41866ba21ee1b5020b8225cec86fe717e4b1c5c972de0a33c/pycuda-2019.1.2.tar.gz

tar zxvf pycuda-2019.1.2.tar.gz    
cd pycuda-2019.1.2/  
python3 configure.py --cuda-root=/usr/local/cuda-10.2
sudo python3 setup.py install

• 若报错FileNotFoundError: [Errno 2] No such file or directory: ‘nvcc’
• 打开Pycuda的compiler.py文件中的compile_plain()
  中，大约在第 73 行的位置中加入下面段代码！
  

8. 安装Cmake

wget http://www.cmake.org/files/v3.13/cmake-3.13.0.tar.gz
tar xpvf cmake-3.13.0.tar.gz cmake-3.13.0/  
cd cmake-3.13.0/
./bootstrap --system-curl	
make -j4 
echo 'export PATH=/home/z/Downloads/cmake-3.13.0/bin/:$PATH' >> ~/.bashrc
source ~/.bashrc

9. 安装matplotlib

sudo apt-get install python3-matplotlib

10. 安装pyqt5

sudo apt install pyqt5*
sudo apt-get install python3-pyqt5