PyTorch doesn't see Cuda device

After reading the following article PyTorch for Jetson - version 1.10 now available and installing the package there for torch 1.10.0, I run torch.cuda.is_available() and get false. Is there a setup instruction I’m missing to get it to recognize the GPU? My Jetpack version is 4.4

Hi,

It’s possible to upgrade your device into JetPack 4.6.1 first?
We test PyTorch 1.10.0 with JetPack4.6.1, and it works good.

>>> import torch
>>> torch.cuda.is_available()
True

Thanks.

Unfortunately, I can’t upgrade my jetpack version, but I can use an older version of torch if there is one that’s supported.
torch >=1.7.0

Hi,

Based on the instructions, PyTorch 1.10 should work on JetPack4.4.
We are checking this internally.
Will share more information with you later.

Hi,

We test PyTorch 1.10 on a Xavier set up with JetPack4.4.
It also works normally.

$ python3
Python 3.6.9 (default, Dec  8 2021, 21:08:43) 
[GCC 8.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> torch.cuda.is_available()
True

Thanks.

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