OSError: libcurand.so.10: cannot open shared object file: No such file or directory

[bth]

Hi

I’m currently testing out a new camera from Baumer (VAX32.C.I.NVN) wich has the jetson nano SOM.

They have provided a OS with L4T installed, when i check the version it says:

cat /etc/nv_tegra_release
# R32 (release), REVISION: 4.4, GCID: 23942405, BOARD: t210ref, EABI: aarch64, DATE: Fri Oct 16 19:44:43 UTC 2020

I would like to try the classification_interactive from the DLI (Getting Started with AI on Jetson Nano). I followed the toturial and as i understand my docker file sould be (to match the L4T version 4.4):

JetPack Release 	Container Version Tag 	Language
4.4 	            v2.0.0-r32.4.3 	        en-US

However when i try to import:

import torchvision.transforms as transforms

i get the following error:

OSError: libcurand.so.10: cannot open shared object file: No such file or directory

My docker command is:

sudo docker run --runtime nvidia -it --privileged --rm --network host --volume ~/nvdli-data:/nvdli-nano/data --volume /tmp/argus_socket:/tmp/argus_socket nvcr.io/nvidia/dli/dli-nano-ai:v2.0.0-r32.4.3

Ive added --privileged to allow access to speak with the Baumer camera module.

Can anyone help resolve this issue?
Thanks in advance!

[bth]

Full error readout:

---------------------------------------------------------------------------
OSError                                   Traceback (most recent call last)
<ipython-input-1-413b1040f905> in <module>
----> 1 import torchvision.transforms as transforms
      2 from dataset import ImageClassificationDataset
      3 
      4 TASK = 'thumbs'
      5 # TASK = 'emotions'

/usr/local/lib/python3.6/dist-packages/torchvision-0.7.0a0+6631b74-py3.6-linux-aarch64.egg/torchvision/__init__.py in <module>
      3 from .extension import _HAS_OPS
      4 
----> 5 from torchvision import models
      6 from torchvision import datasets
      7 from torchvision import ops

/usr/local/lib/python3.6/dist-packages/torchvision-0.7.0a0+6631b74-py3.6-linux-aarch64.egg/torchvision/models/__init__.py in <module>
----> 1 from .alexnet import *
      2 from .resnet import *
      3 from .vgg import *
      4 from .squeezenet import *
      5 from .inception import *

/usr/local/lib/python3.6/dist-packages/torchvision-0.7.0a0+6631b74-py3.6-linux-aarch64.egg/torchvision/models/alexnet.py in <module>
----> 1 import torch
      2 import torch.nn as nn
      3 from .utils import load_state_dict_from_url
      4 
      5 

/usr/local/lib/python3.6/dist-packages/torch/__init__.py in <module>
    186     # See Note [Global dependencies]
    187     if USE_GLOBAL_DEPS:
--> 188         _load_global_deps()
    189     from torch._C import *
    190 

/usr/local/lib/python3.6/dist-packages/torch/__init__.py in _load_global_deps()
    139     lib_path = os.path.join(os.path.dirname(here), 'lib', lib_name)
    140 
--> 141     ctypes.CDLL(lib_path, mode=ctypes.RTLD_GLOBAL)
    142 
    143 

/usr/lib/python3.6/ctypes/__init__.py in __init__(self, name, mode, handle, use_errno, use_last_error)
    346 
    347         if handle is None:
--> 348             self._handle = _dlopen(self._name, mode)
    349         else:
    350             self._handle = handle

OSError: libcurand.so.10: cannot open shared object file: No such file or directory

[bth]

From inside the container:

root@varo-desktop:~# ll /usr/local/cuda/lib64/
total 1444
drwxr-xr-x 1 root root   4096 Jun 27  2020 ./
drwxr-xr-x 1 root root   4096 Jun 27  2020 ../
-rw-r--r-- 1 root root 677036 Jun 27  2020 libcudadevrt.a
-rw-r--r-- 1 root root 786194 Jun 27  2020 libcudart_static.a
drwxr-xr-x 2 root root   4096 Jun 27  2020 stubs/

Outside the container:

varo@varo-desktop:~$ ll /usr/local/cuda/lib64/
ls: cannot access '/usr/local/cuda/lib64/': No such file or directory

[bth]

As per my last reply i realized that i needed cuda installed properly, and following the steps from:

https://forums.developer.nvidia.com/t/oserror-libcurand-so-10-cannot-open-shared-object-file-no-such-file-or-directory-when-running-from-l4t-image/198578/5?u=bth

I ran following:

sudo apt-get update && sudo apt-get install cuda-toolkit-10-2
export LD_LIBRARY_PATH=/usr/local/cuda-10.2/lib64:$LD_LIBRARY_PATH
sudo apt install nvidia-container-csv-cuda
sudo apt install nvidia-container-csv-cudnn

That resolved my issues.

If anyone else are trying out the Baumer camera, i have written a camera file similar to usb_camera.py that works with jetcam, so that the training in the DLI course can be run with the built in camera.

[kayccc]

Gald to know issue fixed, thanks for the update.