Https://forums.developer.nvidia.com/t/installing-torch-and-torchvision-on-orin-nano/335317?u=joel.alejandro11366

Robotics & Edge Computing Jetson Systems Jetson Orin Nano
1

I’ve got a Jetson Orin Nano Developer Kit 8GB with of JetPack 6.2 and I’m trying to install the Python packages torch and torchvision. Every time I try to install torchvision I get the following error:
if torch._C._dispatch_has_kernel_for_dispatch_key(self.qualname, “Meta”):
RuntimeError: operator torchvision::nms does not exist

Similarly as this same forum:

However, in this forum two links were shared that should solve the problem. But, these two links are broken, they do not link to anything. I’ve tried for the past two weeks, every day, on using these links but they do not install anything. How can I install torchvision succesfully?

1 Like
5

Hi,

You can find our PyTorch package for JetPack 6.2 below:

Thanks.

6

Hi, I followed the link and ran:
pip install https://pypi.jetson-ai-lab.io/jp6/cu126/+f/590/92ab729aee2b8/torch-2.8.0-cp310-cp310-linux_aarch64.whl#sha256=59092ab729aee2b8937d80cc1b35d1128275bd02a7e1bc911e7efa375bd97226

However, I get the following error:

>>> import torch Traceback (most recent call last):   File "<stdin>", line 1, in <module>   File "/home/jackal/miniconda3/envs/test/lib/python3.10/site-packages/torch/__init__.py", line 416, in <module>     from torch._C import *  # noqa: F403 ImportError: libcudss.so.0: cannot open shared object file: No such file or directory >>> import torchvision Traceback (most recent call last):   File "<stdin>", line 1, in <module>   File "/home/jackal/miniconda3/envs/test/lib/python3.10/site-packages/torchvision/__init__.py", line 5, in <module>     import torch   File "/home/jackal/miniconda3/envs/test/lib/python3.10/site-packages/torch/__init__.py", line 416, in <module>     from torch._C import *  # noqa: F403 ImportError: libcudss.so.0: cannot open shared object file: No such file or directory

I’m not sure where can I install the package: libcudss.so.0? I initially thought it was part of cuda-toolkit and try running the following command but it didn’t solved the problem

sudo apt install nvidia-jetpack

Next, I installed the library I thought was causing the error to my cuda environment using this:

conda install conda-forge::libcudss-dev
export LD_LIBRARY_PATH=~/miniconda3/envs/test/lib:$LD_LIBRARY_PATH

this solved the problem of import torch, but now when I try to run the model, I get error with the library, as if it has a dependency issue. Am I installing the correct libcudss library? For torch 2.8?

Error that I’m getting:

 File "/home/jackal/miniconda3/envs/test/lib/python3.10/site-packages/torch/nn/modules/linear.py", line 125, in forward
    return F.linear(input, self.weight, self.bias)
RuntimeError: CUDA error: CUBLAS_STATUS_ALLOC_FAILED when calling cublasCreate(handle)
1 Like
7

Hi,

cuDSS can be found in the link below:

Or you can follow the steps shared in the link below to install PyTorch on the JetPack 6.2:

Thanks.

9

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