Unable to install pytorch 2 in jetson orin with python 3.10

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

Hello,

I used sdkmanager tool to flash the jetson board with jetpack 6. Then i followed Installing PyTorch for Jetson Platform - NVIDIA Docs this tutorial. and installed torch-2.4.0a0+07cecf4168.nv24.05.14710581-cp310-cp310-linux_aarch64.whl file using pip.

i chose above specific file because when i was installing cusparselt, i got following error.

+ mkdir tmp_cusparselt
+ cd tmp_cusparselt
+ [[ 12.1 =~ ^12\.[1-4]$ ]]
+ arch_path=sbsa
++ uname -m
+ export TARGETARCH=aarch64
+ TARGETARCH=aarch64
+ '[' aarch64 = amd64 ']'
+ '[' aarch64 = x86_64 ']'
+ CUSPARSELT_NAME=libcusparse_lt-linux-sbsa-0.5.2.1-archive
+ curl --retry 3 -OLs https://developer.download.nvidia.com/compute/cusparselt/redist/libcusparse_lt/linux-sbsa/libcusparse_lt-linux-sbsa-0.5.2.1-archive.tar.xz
+ tar xf libcusparse_lt-linux-sbsa-0.5.2.1-archive.tar.xz
+ cp -a libcusparse_lt-linux-sbsa-0.5.2.1-archive/include/cusparseLt.h /usr/local/cuda/include/
cp: cannot create regular file '/usr/local/cuda/include/cusparseLt.h': Permission denied

Hence i chose to install pytorch which is older than 24.06 PyTorch. Installing PyTorch for Jetson Platform - NVIDIA Docs

Installation was successful but when i tried to use torch library, it gave runtimeerror.

Python 3.10.12 (main, Feb  4 2025, 14:57:36) [GCC 11.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
Traceback (most recent call last):
  File "/home/mzjet/.local/lib/python3.10/site-packages/torch/__init__.py", line 176, in _load_global_deps
    ctypes.CDLL(lib_path, mode=ctypes.RTLD_GLOBAL)
  File "/usr/lib/python3.10/ctypes/__init__.py", line 374, in __init__
    self._handle = _dlopen(self._name, mode)
OSError: libnvToolsExt.so.1: cannot open shared object file: No such file or directory

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/mzjet/.local/lib/python3.10/site-packages/torch/__init__.py", line 236, in <module>
    _load_global_deps()
  File "/home/mzjet/.local/lib/python3.10/site-packages/torch/__init__.py", line 197, in _load_global_deps
    _preload_cuda_deps(lib_folder, lib_name)
  File "/home/mzjet/.local/lib/python3.10/site-packages/torch/__init__.py", line 162, in _preload_cuda_deps
    raise ValueError(f"{lib_name} not found in the system path {sys.path}")
ValueError: libcublas.so.*[0-9] not found in the system path ['', '/usr/lib/python310.zip', '/usr/lib/python3.10', '/usr/lib/python3.10/lib-dynload', '/home/mzjet/.local/lib/python3.10/site-packages', '/usr/local/lib/python3.10/dist-packages', '/usr/lib/python3/dist-packages', '/usr/lib/python3.10/dist-packages']

I have added the cuda path to the LD_LIBRARY_PATH.

echo $LD_LIBRARY_PATH
/usr/local/cuda/lib64:/usr/lib/llvm-8/lib:/usr/local/cuda-12.2/lib64:

I am not sure what am i doing wrong. If anyone could help me out here, would be helpful.

Thanks

2

Hi,
Here are some suggestions for the common issues:

1. Performance

Please run the below command before benchmarking deep learning use case:

$ sudo nvpmodel -m 0
$ sudo jetson_clocks

2. Installation

Installation guide of deep learning frameworks on Jetson:

3. Tutorial

Startup deep learning tutorial:

4. Report issue

If these suggestions don’t help and you want to report an issue to us, please attach the model, command/step, and the customized app (if any) with us to reproduce locally.

Thanks!

4

To reproduce>

  1. Install Jetpack 6 on jetson Orin
  2. Install pytorch using torch-2.4.0a0+07cecf4168.nv24.05.14710581-cp310-cp310-linux_aarch64.whl
  3. run python3 -c “import torch” command in terminal.
5

you can install cusparse with
sudo apt update
sudo apt install (pick your cuda version)

  • libcusparse-12-4 libcusparse-dev-12-4
  • libcusparse-12-5 libcusparse-dev-12-5
  • libcusparse-12-6 libcusparse-dev-12-6
  • libcusparse-12-8 libcusparse-dev-12-8

then go here pypi.jetson-ai-lab

click jp6/{your cuda version}

and install that torch

6

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