YOLOv8 on Jetson Nano with Cuda Support

[javad3]

I am trying to run YOLOv8 on a Jetson Nano with Cuda acceleration.
And I have to switch from python3.6 to python3.8.
And this means I can not use my prebuilt torch library that comes with Cuda.
I was wondering has anyone made an effort to build PyTorch on jetson nano with python3.8?

[AastaLLL]

Hi,

You can find the instructions for building the PyTorch library in the below 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…

> Instructions
>> Build from Source

Thanks.

[javad3]

Thanks for the reply.
Actually, I am following the “Build from Source” instructions to build torch 1.10.1 on my Jetson Nano. I have also increased my Swap area to 4GB, and work in a virtualenv with python 3.8 (to be used with YOLOv8).
But now, the error I face is that while “Building CXX object caffe2/CMakeFiles/torch_cpu.dir/__/torch/csrc/autograd/generated/VariableType_1.cpp.o”
it shows a “c++: internal compiler error: Killed (program cc1plus)” and it does not progress anymore.

Any thoughts on this would be really appreciated.
Maybe trying other versions of pytorch?
Can it be a memory problem? Or something else?

[AastaLLL]

Hi,

Yes, killed is usually related to the memory issue.
Could you increase the swap memory size and try it again?

Thanks.