Help installing PyTorch on Jetson Orin Nano

Robotics & Edge Computing Jetson & Embedded Systems Jetson Orin Nano
1

Hello,

I’m very frustrated ‘cause after updating Jetson Orin Nano on latest Jetpack can’t instal proper working versin of PyTorch. I even think that PyTorch is instaled but Pyton cant use (find) it.

In directory usr/local/cuda-12.6/targets/aarch64-linux/lib I have libcusparseLt.so.12 and and symbolic links

libcusparse.so.12 (libc6,AArch64) => /usr/local/cuda/targets/aarch64-linux/lib/libcusparse.so.12
libcusparse.so (libc6,AArch64) => /usr/local/cuda/targets/aarch64-linux/lib/libcusparse.so

But simple code:

python3 -c “import torch; print(‘Torch:’, torch.version); print(‘CUDA available:’, torch.cuda.is_available()); print(‘GPU:’, torch.cuda.get_device_name(0) if torch.cuda.is_available() else ‘None’); print(‘cuDNN:’, torch.backends.cudnn.version())”

says: GPU: None

Where to find and how to install proper version of PyTorch.
Thanks.
Simon

4

@simon.pogorelcnik download a wheel from jp6/cu129 index . There are two versions availablecp310 and cp312

This is what I’m using:

Torch: 2.8.0
CUDA available: True
GPU: Orin
cuDNN: 90300

Don’t install torch from a Ubuntu repo.

5

Thank you very much. I will try this week.

One more question: Is it better to install also proper version of torchvision & torchaudio.
I would like to start playing with simple project with YOLO8 models and object recognition and classification.
Do I need any more compatible tools to use this kind of model?
Thank you for your kind answer.
Regards.

8

Hi,

Please refer to the following download center for torchvision and torchaudio packages:

Thank you.

9

Hi,

I’m tottaly stuck… I did:

  1. Remove OLD CPU version:

sudo pip uninstall torch torchvision torchaudio -y
pip uninstall torch torchvision torchaudio -y
sudo rm -rf /usr/local/lib/python3.10/dist-packages/torch*
sudo rm -rf ~/.local/lib/python3.10/site-packages/torch*
sudo rm -rf ~/.cache/torch

  1. Check: python3 -m pip show torch
    Torch not found

  2. Clean Install from:
    pip install torch torchvision torchaudio --extra-index-url jp6/cu126 index

  3. Check with:
    python3 -c “import torch; print(‘Torch:’, torch.version); print(‘CUDA available:’, torch.cuda.is_available()); print(‘GPU:’, torch.cuda.get_device_name(0) if torch.cuda.is_available() else ‘None’); print(‘cuDNN:’, torch.backends.cudnn.version())”

Result:
Torch: 2.9.0+cpu
Path: /usr/local/lib/python3.10/dist-packages/torch/init.py
CUDA available: False
GPU: None
cuDNN: None

Please help. Tnx

10

Hi,

It is recommended to perform a full flash and install the jp6/cu126 index package.

Thanks

11

Yes. I did.. and it’s running. Thank you.

12

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.