Unable to use cuda cores for yolo. tried everything found on internet for orin nano dev kit

We are encountering difficulties installing the necessary libraries for CUDA 11.4 on our system, which currently runs JetPack 5.1.2. Our goal is to optimize the performance of YOLOv8 by leveraging CUDA cores, but we have been unable to utilize them effectively. As a result, the model’s performance is not as fast as expected. We need assistance in resolving this issue and ensuring that CUDA is properly configured to accelerate YOLOv8.
we think its something related to pytorch but we followed many videos and when we type this command we get none python3 -c “import torch; print(torch.version.cuda)”

Hi,

How do you install pytorch?
Could you provide your commands or referred link for us to review.

To install pytorch for jetson, you can refer this topic

Thanks

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu113

Hi,

The command you mentioned is for x86 arch gpu, which not support jetson device.

Please follow above mentioned link to install pytorch.

Thanks

What if we are running Cuda 12.6?

Hi,

Seems you are using JP6.2, you could follow the example provided to install the pytorch version.

Thanks