brunelautonomousracing@brunelautonomousracing-desktop:~/yolov7-gpu$ python3 detect.py --weights runs/train/yolov7-cones/weights/best.pt --source 1
Traceback (most recent call last):
File “detect.py”, line 6, in
import torch
File “/home/brunelautonomousracing/.local/lib/python3.6/site-packages/torch/init.py”, line 195, in
_load_global_deps()
File “/home/brunelautonomousracing/.local/lib/python3.6/site-packages/torch/init.py”, line 148, 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
how to solve it?
Hi @naikrohanp97, which version of JetPack-L4T are you running (you can check this with cat /etc/nv_tegra_release
) and which PyTorch wheel did you install? You would need to install a PyTorch wheel from here which is compatible with your version of JetPack.
Also, can you check that you have CUDA Toolkit installed?
ls /usr/local/cuda/lib64/
libcublasLt.so libcufft_static.a libcurand_static.a libnppial.so.10 libnppif.so.10 libnppisu.so.10 libnvgraph.so.10.2.300
libcublasLt.so.10 libcufft_static_nocallback.a libcusolver.so libnppial.so.10.2.1.300 libnppif.so.10.2.1.300 libnppisu.so.10.2.1.300 libnvgraph_static.a
libcublasLt.so.10.2.3.300 libcufftw.so libcusolver.so.10 libnppial_static.a libnppif_static.a libnppisu_static.a libnvperf_host.so
libcublasLt_static.a libcufftw.so.10 libcusolver.so.10.3.0.300 libnppicc.so libnppig.so libnppitc.so libnvperf_target.so
libcublas.so libcufftw.so.10.1.2.300 libcusolver_static.a libnppicc.so.10 libnppig.so.10 libnppitc.so.10 libnvrtc-builtins.so
libcublas.so.10 libcufftw_static.a libcusparse.so libnppicc.so.10.2.1.300 libnppig.so.10.2.1.300 libnppitc.so.10.2.1.300 libnvrtc-builtins.so.10.2
libcublas.so.10.2.3.300 libcuinj64.so libcusparse.so.10 libnppicc_static.a libnppig_static.a libnppitc_static.a libnvrtc-builtins.so.10.2.300
libcublas_static.a libcuinj64.so.10.2 libcusparse.so.10.3.1.300 libnppicom.so libnppim.so libnpps.so libnvrtc.so
libcudadevrt.a libcuinj64.so.10.2.300 libcusparse_static.a libnppicom.so.10 libnppim.so.10 libnpps.so.10 libnvrtc.so.10.2
libcudart.so libculibos.a liblapack_static.a libnppicom.so.10.2.1.300 libnppim.so.10.2.1.300 libnpps.so.10.2.1.300 libnvrtc.so.10.2.300
libcudart.so.10.2 libcupti.so libmetis_static.a libnppicom_static.a libnppim_static.a libnpps_static.a libnvToolsExt.so
libcudart.so.10.2.300 libcupti.so.10.2 libnppc.so libnppidei.so libnppist.so libnvblas.so libnvToolsExt.so.1
libcudart_static.a libcupti.so.10.2.175 libnppc.so.10 libnppidei.so.10 libnppist.so.10 libnvblas.so.10 libnvToolsExt.so.1.0.0
libcufft.so libcurand.so libnppc.so.10.2.1.300 libnppidei.so.10.2.1.300 libnppist.so.10.2.1.300 libnvblas.so.10.2.3.300 stubs
libcufft.so.10 libcurand.so.10 libnppc_static.a libnppidei_static.a libnppist_static.a libnvgraph.so
libcufft.so.10.1.2.300 libcurand.so.10.1.2.300 libnppial.so libnppif.so libnppisu.so libnvgraph.so.10
brunelautonomousracing@brunelautonomousracing-desktop:~$ cat /etc/nv_tegra_release
R32 (release), REVISION: 6.1, GCID: 27863751, BOARD: t210ref, EABI: aarch64, DATE: Mon Jul 26 19:20:30 UTC 2021
Hi dusty,
Package: nvidia-jetpack
Version: 4.6-b199
Priority: standard
Section: metapackages
Maintainer: NVIDIA Corporation
Installed-Size: 199 kB
this is the version of jetpack installed
hi dusty, I installed same Pytorch wheel compatible with the version of jetpack but when i check CUDA installation through this line:
ls /usr/local/cuda/lib64/
it gives:
brunelautonomousracing@brunelautonomousracing-desktop:~$ ls /usr/local/cuda/lib64/
ls: cannot access ‘/usr/local/cuda/lib64/’: No such file or directory
l
OK, yea it looks like you need to install the CUDA Toolkit (and probably cuDNN) - can you try running this?
sudo apt-get install cuda-toolkit-10-2 libcudnn8-dev
After flash we tried to jetson nano and this is the result
Typically the Ubuntu GUI would appear at this point and allow you to proceed with the initial system setup (like creating your username/password, setting the keyboard locale, ect). If you’re having problems with that, you can also do this initial config headlessly without using a display, by using the micro-USB port instead - see here for more info: https://developer.nvidia.com/embedded/learn/get-started-jetson-nano-devkit#setup-headless
If you encounter further issues with this, I would recommend opening a new topic about it - thanks, and good luck!
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.