Hello,
I’m trying to use CUDA on my Jetson Orin NX, but after quite a bit of research, I still see that the output of torch.cuda.is_available() is FALSE, and I am unable to run samples successfully. For context, I have Ubuntu 20.04, Python 3.8, JetPack 5.1.2, and PyTorch 2.1.0a (I ensured that this is the compatible version). I attempted to install CUDA 11.4 using the following command: sudo sh cuda_11.4.0_470.42.01_linux_sbsa.run. I installed the Toolkit and the samples but did not install the drivers because I read that JetPack already includes them. Below are all the steps I’ve taken to verify the installation and their outputs:
export PATH=/usr/local/cuda-11.4/bin${PATH:+:${PATH}},export LD_LIBRARY_PATH=/usr/local/cuda-11.4/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}},source ~/.bashrc.nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Wed_Jun__2_19:15:41_PDT_2021
Cuda compilation tools, release 11.4, V11.4.48
Build cuda_11.4.r11.4/compiler.30033411_0
sudo tegrastats
10-10-2024 09:30:24 RAM 1194/15523MB (lfb 892x4MB) SWAP 0/7762MB (cached 0MB) CPU [1%@1984,0%@1984,0%@729,0%@729,0%@729,0%@729,0%@729,0%@729] EMC_FREQ 0%@2133 GR3D_FREQ 0%@[305,0] VIC_FREQ 729 APE 174 CV0@48.312C CPU@50.656C SOC2@49.031C SOC0@50.25C CV1@48.968C GPU@48.625C tj@51.875C SOC1@51.875C CV2@48.781C VDD_IN 4820mW/4820mW VDD_CPU_GPU_CV 607mW/607mW VDD_SOC 1498mW/1498mW
lsmod | grep nvidia
nvidia_modeset 1093632 3
nvidia 1327104 7 nvidia_modeset
- Checked that
/usr/local/cuda-11.4exists andls -l /usr/local/cudareturnslrwxrwxrwx 1 root root 21 oct 7 16:46 /usr/local/cuda -> /usr/local/cuda-11.4/ /usr/lib/aarch64-linux-gnu/: Is a directorydpkg -l | grep libcudnn
ii libcudnn8 8.6.0.166-1+cuda11.4 arm64 cuDNN runtime libraries
ii libcudnn8-dev 8.6.0.166-1+cuda11.4 arm64 cuDNN development libraries and headers
ii libcudnn8-samples 8.6.0.166-1+cuda11.4 arm64 cuDNN samples
ls: cannot access '/usr/lib/aarch64-linux-gnu/libnvinfer*': No such file or directoryls /usr/local/cuda/lib64 | grep libcudart
libcudart.so
libcudart.so.11.0
libcudart.so.11.4.43
libcudart_static.a
ls /usr/local/cuda/lib64 | grep libcuda
libcudadevrt.a
libcudart.so
libcudart.so.11.0
libcudart.so.11.4.43
libcudart_static.a
11.glxinfo | grep "OpenGL version" returns: OpenGL version string: 3.1 Mesa 21.2.6
12. cat /proc/driver/nvidia/version
NVRM version: NVIDIA UNIX Open Kernel Module for aarch64 35.4.1 Release Build (buildbrain@mobile-u64-6422-d7000) Tue Aug 1 12:45:41 PDT 2023
GCC version: gcc version 9.3.0 (Buildroot 2020.08)
sudo ./deviceQuery
./deviceQuery Starting...
CUDA Device Query (Runtime API) version (CUDART static linking)
cudaGetDeviceCount returned 100
-> no CUDA-capable device is detected
Result = FAIL
cat /var/log/cuda-installer.log
[INFO]: Driver not installed.
[INFO]: Checking compiler version...
[INFO]: gcc location: /usr/bin/gcc
[INFO]: gcc version: gcc version 9.4.0 (Ubuntu 9.4.0-1ubuntu1~20.04.2)
[INFO]: Initializing menu
[INFO]: Setup complete
[INFO]: Components to install:
[INFO]: CUDA Samples 11.4
[WARNING]: Missing recommended library: libGLU.so
[WARNING]: Missing recommended library: libXmu.so
[INFO]: CUDA Documentation 11.4
[INFO]: Uninstall entry: dir /usr/local/cuda-11.4/
[INFO]: Uninstall entry: dir /usr/local/cuda-11.4/tools
[INFO]: Uninstall entry: dir /usr/local/cuda-11.4/bin
[INFO]: Uninstall entry: file /usr/local/cuda-11.4/DOCS eba0599f5ca8b9bda647f20a e3dd6345
[INFO]: md5 matches, removing file
[INFO]: Uninstall entry: file /usr/local/cuda-11.4/EULA.txt 1203d26f82bb4d2d485f e3be00efc3ae
[INFO]: md5 matches, removing file
[INFO]: Uninstall entry: file /usr/local/cuda-11.4/README 8514c9e2df920c17a65c54 e30843d822
[INFO]: md5 matches, removing file
[INFO]: Uninstall entry: file /usr/local/cuda-11.4/bin/cuda-uninstaller c582e0ad d9ed4a47af8e4d2e2e698067
[INFO]: md5 matches, removing file
[INFO]: Uninstall entry: file /usr/local/cuda-11.4/tools/CUDA_Occupancy_Calculat or.xls 056066ac68c81aa2d3da4793112748cd
[INFO]: md5 matches, removing file
[INFO]: Removing empty directory: /usr/local/cuda-11.4/tools
[INFO]: Successfully created directory: /usr/local/cuda-11.4/tools
[INFO]: Installed: /usr/local/cuda-11.4/DOCS
[INFO]: Installed: /usr/local/cuda-11.4/EULA.txt
[INFO]: Installed: /usr/local/cuda-11.4/README
[INFO]: Installed: /usr/local/cuda-11.4/bin/cuda-uninstaller
[INFO]: Installed: /usr/local/cuda-11.4/tools/CUDA_Occupancy_Calculator.xls
[WARNING]: Cannot find manpages to install.
cat /etc/X11/xorg.conf
# Copyright (c) 2011-2013 NVIDIA CORPORATION. All Rights Reserved.
#
# This is the minimal configuration necessary to use the Tegra driver.
# Please refer to the xorg.conf man page for more configuration
# options provided by the X server, including display-related options
# provided by RandR 1.2 and higher.
# Disable extensions not useful on Tegra.
Section "Module"
Disable "dri"
SubSection "extmod"
Option "omit xfree86-dga"
EndSubSection
EndSection
Section "Device"
Identifier "Tegra0"
Driver "nvidia"
# Allow X server to be started even if no display devices are connected.
Option "AllowEmptyInitialConfiguration" "true"
EndSection
sudo dmesg | grep nvidia
[ 0.002849] DTS File Name: /dvs/git/dirty/git-master_linux/kernel/kernel-5.10 /arch/arm64/boot/dts/../../../../../../hardware/nvidia/platform/t23x/p3768/kerne l-dts/tegra234-p3767-0000-p3509-a02.dts
[ 14.934640] nvidia: loading out-of-tree module taints kernel.
[ 19.318537] nvidia-modeset: Loading NVIDIA UNIX Open Kernel Mode Setting Driv er for aarch64 35.4.1 Release Build (buildbrain@mobile-u64-6422-d7000) Tue A ug 1 12:45:42 PDT 2023
gcc --version
gcc (Ubuntu 9.4.0-1ubuntu1~20.04.2) 9.4.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
lspci | grep -i nvidia
0004:00:00.0 PCI bridge: NVIDIA Corporation Device 229c (rev a1)
0008:00:00.0 PCI bridge: NVIDIA Corporation Device 229c (rev a1)
I have no clue what the issue might be, so any input would be helpful. Thank you in advance!