Hello, I Brought a new jetson Nano Board for real time object detection project But It is not compatible with the latest modules for example Torch and torchvision. Since it has cuda 10.2 Most of the latest versions are not compliable with this. Because of this I am unable to run the model on GPU. Can Anyone help me
Hi,
Jetson Nano uses Jetpack 4 releases(with Ubuntu 18.04 and Kernel 4.9) and the latest version of some software stacks may not work properly. You would need to stay on previous releases.
If you must use latest versions, would suggest use Orin Nano or Orin NX which uses Jetpack 6(with Ubuntu 22.04 and Kernel 5.15).
even tough i am using the older torch versions compatible with jetpack 4.6 and python 3.6 also it is not running on GPU
Hi,
Please share more information about your use-case so that we can suggest next. For example, if you would like to run deep learning inference, we suggest use Deepstream SDK.
Yes I am Running a deep learning model (Yolo Version).Installed the compatible torch versions and we got cuda as true. But during running it says no compatibility b/w torch and to Torch vision. We followed the guild lines of the Pytorch for Jetpack 4.6 version
Hi,
Running Yolo on the whl package should use GPU. You may run sudo tegrastats to check the GPU loading.
yes, it shows GPU is loading with percentage
Hi,
So GPU is up and running. If it does not meet target performance, it probably is due to hardware capability. Jetson Nano may not be able to offer the target performance.