Hi
I have some problems with cuda on jetson orin agx developer kit so in proccess of trying to solve problems i deleted all cuda related files and all nvidia drivers. Right after i installed cuda but looks like i forgot to install drivers and after first reboot orin booted up but there was no background wallpaper, wifi and everything looked strange so i rebooted once more and then i end up with very weird look like i have 3 screens instead of one and I bearly see what is on screen. So mu qustion is- is there a way to install just drivers from sdk manager or I need to flash system?
Easiest will be reflash, otherwise you’ll have to figure out all you deleted and force a reinstall associated packages
But reflash will delete all files? I know what i deleted, I deleted all related to nvidia like this
sudo apt-get --purge remove “nvidia”
sudo apt-get autoremove
sudo apt-get autoclean
and after that i din’t installed new drivers.
I would like to keep files if it is possible and just install drivers, but at the end I will reflash if no other solution. Just in case if I do reflash it, which jetpack should I install?
To install Jetpack on latest linux for tegra 36.3
Use the command
$sudo apt-get install nvidia-jetpack
It will. install complete CUDa and other library frame work perfectly except.open CV which you need to download and rebuild using on shell script (search in forum or Google you will find it )
Ah in case you used the package manager the jetpack install as below should work
for this i need first to make jetson to boot and show image on screen correctly.
If I want to install jetpack from sdk manager i have 3 options-jetpack 6, jetpack 5.2 and jetpack 5.1, which one to choose?Latest?
1 Like
I flashed whole system and now i can boot in ubuntu. I installed jetpack 5.1.3 because I must use ubuntu 20.04. Jetson now works ok but now i need help in proper configuration.
When i run nvcc --version i have this
Build cuda_11.4.r11.4
I also installed cuda 11.4 toolkit from nvidia website tutorial but when i run
torch.cuda.is_available() i have
False
My torch version is 2.0.1 and torchvision is 0.15.2
If I uninstall torch and install version 2.1.0 I then have cuda available but then I have problem with torchvision.
Can somebody explain proper procedure to correct all problems?
Installing individual components is not suggestable as there are lot of dependencies between each other and you end up getting messed up
You can execute
sudo apt install nvidia-jetpack
Which installs all library framework like cuda, cudann, tensorrt etc
For Open CV with cuda you need to use a shell.script and execute (search for that in this forum you will get it )
Install Jtop tool to verify the installation of CUDa and entire jetpack tools.as a full package…
Hi,
Do you use our prebuilt? Please note that we only have v2.1.0 PyTorch for JetPack 5.1.x users.
https://developer.download.nvidia.com/compute/redist/jp/v512/pytorch/
Then, please build TorchVision as below to ensure compatibility.
- PyTorch v2.1 - torchvision v0.16.1
$ sudo apt-get install libjpeg-dev zlib1g-dev libpython3-dev libopenblas-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
Thanks.
I did install jetpack with sudo apt install nvidia-jetpack
and i did all this that @AastaLLL suggested but torch.cuda.is_available() still gives False
There is no update from you for a period, assuming this is not an issue any more.
Hence we are closing this topic. If need further support, please open a new one.
Thanks
Hi,
Could you share the output of the following commands with us?
$ sudo apt show nvidia-jetpack
$ python3
import torch
torch.__version__
Thanks.