Hello Tom,
for not high end Linux user, software setting is something really hard. I spent hours to make my first Jetson Nano setup ! It was really hard for me.
We find (too) many informations on the internet and it is hard to know what works and what does not work.
First, i recommend to avoid Youtube lessons and try to find serious internet websites to catch useful informations.
It is important to go step by step and take time to find the good informations.
From what i can remember and considering some quick search on the net :
With JetPack 4.6.1, you will get :
CUDA 10.2
OpenCV 4.1.1
From what i can remember, Python version is 3.6
So, you already have Python3, CUDA and OpenCV. That’s a good start.
No need to install newer version of OpenCV for now.
First :
sudo apt-get update
sudo apt-get upgrade
Then :
sudo apt install python3-pip
python3 -m pip install --upgrade pip
sudo apt-get install python3-dev libpython3-dev
sudo apt install python3-numpy
python3 -m pip install --upgrade Pillow
Concerning PyCuda :
sudo pip3 install pycuda --user
or
sudo pip3 install --global-option=build_ext --global-option=“-I/usr/local/cuda/include” --global-option=“-L/usr/local/cuda/lib64” pycuda
For the last command line, you will have to check the CUDA directory :
It could be :
/usr/local/cuda
so the command line will be : sudo pip3 install --global-option=build_ext --global-option=“-I/usr/local/cuda/include” --global-option=“-L/usr/local/cuda/lib64” pycuda
or
/usr/local/cuda-10.2 (or /usr/local/cuda-10.0 if CUDA 10.0 is installed on your computer for example)
so the command line will be : sudo pip3 install --global-option=build_ext --global-option=“-I/usr/local/cuda-10.2/include” --global-option=“-L/usr/local/cuda-10.2/lib64” pycuda
If you have a problem with nvcc : no such file ; this means you have a problem with PATH.
You can use this command to verify your PATH :
echo $PATH
It will display your PATH. If CUDA paths are not displayed, you will have to add them.
You will have to modify your PATH variable parmanently. There is several methods to do that. I have to chack which method i used for AGX Orin. I will be back as soon as possible.
You can try this :
export PATH=/usr/local/cuda/bin:${PATH}
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:${LD_LIBRARY_PATH}
Or :
export PATH=/usr/local/cuda-10.2/bin:${PATH}
export LD_LIBRARY_PATH=/usr/local/cuda-10.2/lib64:${LD_LIBRARY_PATH}
Just tell me the issues you still have. We will find solutions for sure.
Alain