I’m working with a Jetson Nano with 16GB of emmc storage, and I’m facing an issue in my Docker container based on l4t-pytorch:r32.6.1-pth1.9-py3
When trying to import torch inside of the container, the following error occurs:
>>> import torch
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/usr/local/lib/python3.6/dist-packages/torch/__init__.py", line 196, in <module>
_load_global_deps()
File "/usr/local/lib/python3.6/dist-packages/torch/__init__.py", line 149, in _load_global_deps
ctypes.CDLL(lib_path, mode=ctypes.RTLD_GLOBAL)
File "/usr/lib/python3.6/ctypes/__init__.py", line 348, in __init__
self._handle = _dlopen(self._name, mode)
OSError: libcurand.so.10: cannot open shared object file: No such file or directory
My Jetson Nano is running Ubuntu 18.04. I also noticed that when I run jtop, it says “Jetpack UNKNOWN [L4T 32.5.2]” at the top
I’ve tried to install the nvidia-cuda and nvidia-jetpack metapackages, but they simply won’t fit in the Jetson’s 16GB disk
What can I do to allow Pytorch to work inside the container? Is there something that needs to be installed inside the container that might be missing?
Update: I switched to another Jetson Nano, and after pulling nvcr.io/nvidia/l4t-pytorch:r32.5.0-pth1.7-py3 and trying to run import torch inside Python 3 I got the exact same error.
A colleague suggested I’d add some more debug info:
$ uname -a
Linux <HOSTNAME> 4.9.201-tegra #1 SMP PREEMPT Fri Jan 15 14:41:02 PST 2021 aarch64 aarch64 aarch64 GNU/Linux
$ cat /etc/nv_tegra_release
# R32 (release), REVISION: 5.0, GCID: 25531747, BOARD: t210ref, EABI: aarch64, DATE: Fri Jan 15 22:55:35 UTC 2021