Installing Torch Vision of Jetson Nano

[rahul.goel]

Hello Everyone,

I am trying to install torch an torch vision module in Jetson Nano to run our PyTorch(resent50 ) ML model .We have followed the exact steps mentioned from forum link mentioned below but every time we a hitting a same roadblock which shown below,

Please if any one can help us to solve this problem ,will be really obliged.

Thanks in Advance
Rahul Goel

[AastaLLL]

Hi,

Please check if you have CUDA toolkit installed first.
It should be located at /usr/local/cuda-[version]/.

If yes, the installed package version might be incorrect.
Please noted that you will need to use the same CUDA version as the package built with.

Based on your log, the pyTorch package needs CUDA v10.0 library, which is included in the JetPack 4.3.
If newer JetPack is used, please install the corresponding package listed in this topic:

PyTorch for Jetson

Below are pre-built PyTorch pip wheel installers for Jetson Nano, TX1/TX2, Xavier, and Orin with JetPack 4.2 and newer. Download one of the PyTorch binaries from below for your version of JetPack, and see the installation instructions to run on your Jetson. These pip wheels are built for ARM aarch64 architecture, so run these commands on your Jetson (not on a host PC). You can also use the pre-built l4t-pytorch and l4t-ml container images and Dockerfiles. PyTorch pip wheels JetPack 5

PyTo…

Thanks.

[rahul.goel]

Hey ,

Thanks for the Reply ,

We already have Cuda preinstalled in Jetson nano-10.2.89 as we used Jetpack 4.4.Still got the problem mentioned in above post.

But still we reflashed and installed Jetpack 4.3 with Cuda -10.0 installed it is also giving us the error which is attached in image. Moreover,thera is one thing which i would like to understand is if Cuda is preinstalled then command nvcc -V should work ,but here it return us with error “Command not found” and even “Nvidia -smi” command is not working. Can you help us know why so?

Also the error we got using Jet Pack 4.3 is attached below.

[AastaLLL]

Hi,

Please add the CUDA path into .bashrc to see if helps first.

$ echo 'export LD_LIBRARY_PATH=/usr/local/cuda-10.2/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc
$ echo 'export PATH=/usr/local/cuda-10.2/bin:$PATH' >> ~/.bashrc
$ source ~/.bashrc

Thanks.

[rahul.goel]

Hey AastaLLL,

as you suggested i tried doing the same thing which you have suggested ,screenshot is attached below.but still it is failing ,Moreover the the command for nvcc -V still fails.screen shot for which is attached below.

Thanks

![

[AastaLLL]

Hi,

First, please make sure that you have CUDA 10.2 installed.
It should be located at the /usr/local/.

If the toolkit is well-installed, please remember to run the $ source ~/.bashrc to update the new configure.

Thanks.

[rahul.goel]

Hey,

I have double check that cuda 10.2 is already installed there.Moreover the source ~/.bashrc command i already ran… screenshot is attached below.
Still i got yesterday’s problem.

Moreover,Apart from this i am trying to run a CSI camera IMX219-170,which is also showing the below error.

Error generated. /dvs/git/dirty/git-master_linux/multimedia/nvgstreamer/gst-nvarguscamera/gstnvarguscamerasrc.cpp, execute:557 No cameras available

I have make sure that camera is properly connected.

Thanks

[AastaLLL]

Hi,

Sorry for the late update.

The example is for CUDA 10.2.
For JetPack 4.3, please update it into cuda-10.0.

Thanks.