NO WAY TO USE GPU - TORCH NOT INSTALLING - CUDA 12.6 - JP 6.1 ORIN NANO

want to use GPU for YOLO8, but no way to install TORCH properly. Maybe I do something wrong? YOLO is to use with camera and detect/follow objects. Any idea?

import torch
print(torch.version)
2.0.1
print(torch.version.cuda)
None
print(torch.cuda.is_available())
False

om@ubuntu:~/Downloads$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2024 NVIDIA Corporation
Built on Wed_Aug_14_10:14:07_PDT_2024
Cuda compilation tools, release 12.6, V12.6.68
Build cuda_12.6.r12.6/compiler.34714021_0

Torch jp6/cu126 index

What commands are you using to install pytorch?
Are you installing into venv?

Hello,
I use
pip3 install torch-2.6.0rc1-cp310-cp310-linux_aarch64.whl
pip3 install torchaudio-2.5.0-cp310-cp310-linux_aarch64.whl
pip3 install torchvision-0.20.0-cp310-cp310-linux_aarch64.whl
All goes well without errors.
When I check with
import torch
print(torch.cuda.is_available()) # Should return True
print(torch.backends.cudnn.version()) # Should return a version number
I get

print(torch.version.cuda)
None
print(torch.cuda.is_available())
False

i do not use venv

I have used venv and could install torch 2.3.0. Now I can use GPU. Thanks a lot for your help!!!

I will try to replicate your build and see what happens.
Make sure your path env are set to point to the cuda compiler.

If this works your cuda path is correct.

nvcc --version

Yes, it works now perfectly. Thanks again for your support and follow up :-)

1 Like

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