Issue installing Pytorch on Jetson Orin Nano with Jetpack 6.0

Robotics & Edge Computing Jetson & Embedded Systems Jetson Orin NX
, , ,
1

Cuda Arch Bin: 8.7
L4T: 36.3.0
Jetpack : 6.0
Machine: aarch64
Distribution: Ubuntu 22.04 Jammy Jellyfish
Python: 3.10.12

Followed the instructions from below link

As per my system requirements, I’ve tried installing * torch 2.3 - torch-2.3.0-cp310-cp310-linux_aarch64.whl

my base environment has 3.12 version, so I’ve created a new virtual environment in anaconda with python 3.10.12 and tried installing the above pytorch package there

However, after installation, I noticed that PyTorch was showing version 1.8 instead of 2.3.

And then I’m tried installing torchvision, but it threw an error as “OSError: libmpi_cxx.so.20: cannot open shared object file: No such file or directory” when I was running “python3 setup.py install --user”

I’ve tried two versions of torchvision v0.18.0 and v0.9.0, for both of those version, I was getting the same error.

Any help would be much appreciated!

3

Hi,

Do you use the package listed in the below topic?

If yes, there is a TorchVision prebuilt so you don’t need to build it from the source.

For OSError: libmpi_cxx.so.20: cannot open shared object file: No such file or directory.
Please run the below command to install the dependency and try it again.

$ sudo apt-get install python3-pip libopenblas-base libopenmpi-dev libomp-dev

Thanks.

4

Okay, this time I’ve created a new virtual environment with python 3.10.15 version and executed below instructions
wget https://nvidia.box.com/shared/static/p57jwntv436lfrd78inwl7iml6p13fzh.whl -O torch-2.3.0-cp310-cp310-linux_aarch64.whl
sudo apt-get install python3-pip libopenblas-base libopenmpi-dev libomp-dev
pip3 install ‘Cython<3’
pip3 install numpy torch-2.3.0-cp310-cp310-linux_aarch64.whl

Again it installed torch-1.8.0 version instead of 2.3
Output:

pip3 install numpy torch-2.3.0-cp310-cp310-linux_aarch64.whl
Processing ./torch-2.3.0-cp310-cp310-linux_aarch64.whl
Collecting numpy
  Using cached numpy-2.1.3-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl.metadata (63 kB)
Collecting typing-extensions (from torch==2.3.0)
  Using cached typing_extensions-4.12.2-py3-none-any.whl.metadata (3.0 kB)
Using cached numpy-2.1.3-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (13.9 MB)
Using cached typing_extensions-4.12.2-py3-none-any.whl (37 kB)
Installing collected packages: typing-extensions, numpy, torch
Successfully installed numpy-2.1.3 torch-1.8.0 typing-extensions-4.12.2

after that when I was trying to import torch
I’m again getting same error

self._handle = _dlopen(self._name, mode)
OSError: libmpi_cxx.so.20: cannot open shared object file: No such file or directory
5

Any suggestions please?

6

Hi,

Are you able to find the libmpi_cxx.so.20 file in your virtual environment?
If yes, could you try to add the corresponding path to the LD_LIBRARY_PATH to see if it helps?

Thanks.

7

I was resolve libmpi_cxx.so.20 error, by downgrading my openmpi version to 2.0.4
but now I’m getting “OSError: libcufft.so.10: cannot open shared object file: No such file or directory” error

I think all this is happening because of pytorch version 1.8 is getting installed while I was trying to install pytorch 2.3 as I mentioned earlier

pip3 install numpy torch-2.3.0-cp310-cp310-linux_aarch64.whl --no-cache --force-reinstall
Processing ./torch-2.3.0-cp310-cp310-linux_aarch64.whl
Collecting numpy
  Downloading numpy-2.1.3-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl.metadata (63 kB)
Collecting typing-extensions (from torch==2.3.0)
  Downloading typing_extensions-4.12.2-py3-none-any.whl.metadata (3.0 kB)
Downloading numpy-2.1.3-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (13.9 MB)
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 13.9/13.9 MB 12.5 MB/s eta 0:00:00
Downloading typing_extensions-4.12.2-py3-none-any.whl (37 kB)
Installing collected packages: typing-extensions, numpy, torch
  Attempting uninstall: typing-extensions
    Found existing installation: typing_extensions 4.12.2
    Uninstalling typing_extensions-4.12.2:
      Successfully uninstalled typing_extensions-4.12.2
  Attempting uninstall: numpy
    Found existing installation: numpy 2.1.3
    Uninstalling numpy-2.1.3:
      Successfully uninstalled numpy-2.1.3
  Attempting uninstall: torch
    Found existing installation: torch 1.8.0
    Uninstalling torch-1.8.0:
      Successfully uninstalled torch-1.8.0
Successfully installed numpy-2.1.3 torch-1.8.0 typing-extensions-4.12.2

I even tried uninstalling the miniconda and installing it back, but still it was downloading 1.8 version of torch

Below are the dependencies that I’ve found in libtorch_global_deps.so file

linux-vdso.so.1 (0x0000ffff9bf3c000)
	libmpi_cxx.so.20 => /opt/openmpi/lib/libmpi_cxx.so.20 (0x0000ffff9bec0000)
	libmpi.so.20 => /opt/openmpi/lib/libmpi.so.20 (0x0000ffff9bdc0000)
	libcurand.so.10 => /usr/local/cuda/lib64/libcurand.so.10 (0x0000ffff94eb0000)
	libcudnn.so.8 => /lib/aarch64-linux-gnu/libcudnn.so.8 (0x0000ffff94e50000)
	libcufft.so.10 => not found
	libcublas.so.10 => not found
	libcudart.so.10.2 => not found
	libnvToolsExt.so.1 => /usr/local/cuda/lib64/libnvToolsExt.so.1 (0x0000ffff94e20000)
	libgomp.so.1 => /lib/aarch64-linux-gnu/libgomp.so.1 (0x0000ffff94dc0000)
	libpthread.so.0 => /lib/aarch64-linux-gnu/libpthread.so.0 (0x0000ffff94da0000)
	libc.so.6 => /lib/aarch64-linux-gnu/libc.so.6 (0x0000ffff94bf0000)
	libopen-pal.so.20 => /opt/openmpi/lib/libopen-pal.so.20 (0x0000ffff94af0000)
	libstdc++.so.6 => /lib/aarch64-linux-gnu/libstdc++.so.6 (0x0000ffff948c0000)
	/lib/ld-linux-aarch64.so.1 (0x0000ffff9bf03000)
	libgcc_s.so.1 => /lib/aarch64-linux-gnu/libgcc_s.so.1 (0x0000ffff94890000)
	libopen-rte.so.20 => /opt/openmpi/lib/libopen-rte.so.20 (0x0000ffff947f0000)
	libm.so.6 => /lib/aarch64-linux-gnu/libm.so.6 (0x0000ffff94750000)
	librt.so.1 => /lib/aarch64-linux-gnu/librt.so.1 (0x0000ffff94730000)
	libdl.so.2 => /lib/aarch64-linux-gnu/libdl.so.2 (0x0000ffff94710000)
	libpciaccess.so.0 => /lib/aarch64-linux-gnu/libpciaccess.so.0 (0x0000ffff946f0000)

Which seems to be pytorch of1.8

8

Hi,

Do you use Python 3.10 or a custom Python version?
Could you also try to upgrade pip3 to the latest to see if it works?

$ sudo -H pip3 install --upgrade pip

Thanks.

9

I was able to resolve the issue by directly downloading the file and installing it, not through wget command

wget https://nvidia.box.com/shared/static/p57jwntv436lfrd78inwl7iml6p13fzh.whl -O torch-2.3.0-cp310-cp310-linux_aarch64.whl

when I did wget, for some reason it was downloading pytorch 1.8 version (I renamed the extension from whl to zip and verified it)

Thanks

10

Hi,

Thanks for the update.
Good to know it works now.

Maybe wget use the cache file instead of re-downloading.

12

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