Repository setup issues. Ubuntu 22.04, CUDA-12.0

[PelkiuBebras]

Information about system

• Ubuntu 22.04
• gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
• g++ (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0

Steps made:

1. Remove current nvidia drivers and CUDA (source) sudo apt-get purge nvidia* sudo apt-get autoremove sudo apt-get autoclean sudo rm -rf /usr/local/cuda*

2. Upgrade and update sudo apt update && sudo apt upgrade

3. Installed Ubuntu drivers (source) sudo ubuntu-drivers autoinstall

4. Installed nvidia-drivers (version according to version of example notebook) sudo apt install nvidia-driver-525

5. Rebooted sudo reboot

6. Checked nvidia drivers installation nvidia-smi

+-----------------------------------------------------------------------------+ | NVIDIA-SMI 525.125.06 Driver Version: 525.125.06 CUDA Version: 12.0 | |-------------------------------+----------------------+----------------------+

7. Installed CUDA 12.0 version (again according to notebook CUDA version) (additional install, because /usr/local/ contained no cuda folder) (source) wget https://developer.download.nvidia.com/compute/cuda/12.0.0/local_installers/cuda_12.0.0_525.60.13_linux.run sudo sh cuda_12.0.0_525.60.13_linux.run In install wizard, deselected driver installation

8. Checked system paths echo $PATH contains: /usr/local/cuda/bin echo $LD_LIBRARY_PATH contains: /usr/local/cuda/lib64

9. Set $CUDA_HOME path (according to GroundingDINO repository) echo 'export CUDA_HOME=/usr/local/cuda' >> ~/.bashrc source ~/.bashrc echo $CUDA_HOME path is set (echo $CUDA_HOME command output): /usr/local/cuda

Further steps according to repository installation steps. Made some alteration according to other issues of repository

10. Pull GroundingDINO repository git clone https://github.com/IDEA-Research/GroundingDINO.git

11. Open via PyCharm and create interpreter (python3.9) and virtual env using UI

12. Change working directory cd GroundingDINO/

13. Upgrade pip (22.3.1 -> 23.2.1) pip install --upgrade pip

14. Install requirements pip install -r requirements.txt

15. Package build pip install -e .

From here getting error:

RuntimeError: The detected CUDA version (12.0) mismatches the version that was used to compile PyTorch (11.7). Please make sure to use the same CUDA versions.

Also, tryied similar setup with CUDA-11.7 (nvidia drivers 515). No CUDA mismatch errors, but got different error

Questions

• Can I run project with CUDA-12.0, as in notebool example?
• Perhaps, there are device setup mistakes in provided steps?

Thanks in advance!

[rentainhe]

I think maybe your GCC version is too high for Grounding-DINO, you can try GCC 9.4

[sssssshf]

I think maybe your GCC version is too high for Grounding-DINO, you can try GCC 9.4

how to install the 9.4 GCC? pip?or sudo apt-get install.whta is the shell code

[sssssshf]

Information about system

• Ubuntu 22.04
• gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
• g++ (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0

Steps made:

1. Remove current nvidia drivers and CUDA (source) sudo apt-get purge nvidia* sudo apt-get autoremove sudo apt-get autoclean sudo rm -rf /usr/local/cuda*
2. Upgrade and update sudo apt update && sudo apt upgrade
3. Installed Ubuntu drivers (source) sudo ubuntu-drivers autoinstall
4. Installed nvidia-drivers (version according to version of example notebook) sudo apt install nvidia-driver-525
5. Rebooted sudo reboot
6. Checked nvidia drivers installation nvidia-smi

+-----------------------------------------------------------------------------+ | NVIDIA-SMI 525.125.06 Driver Version: 525.125.06 CUDA Version: 12.0 | |-------------------------------+----------------------+----------------------+

7. Installed CUDA 12.0 version (again according to notebook CUDA version) (additional install, because /usr/local/ contained no cuda folder) (source) wget https://developer.download.nvidia.com/compute/cuda/12.0.0/local_installers/cuda_12.0.0_525.60.13_linux.run sudo sh cuda_12.0.0_525.60.13_linux.run In install wizard, deselected driver installation
8. Checked system paths echo $PATH contains: /usr/local/cuda/bin echo $LD_LIBRARY_PATH contains: /usr/local/cuda/lib64
9. Set $CUDA_HOME path (according to GroundingDINO repository) echo 'export CUDA_HOME=/usr/local/cuda' >> ~/.bashrc source ~/.bashrc echo $CUDA_HOME path is set (echo $CUDA_HOME command output): /usr/local/cuda

Further steps according to repository installation steps. Made some alteration according to other issues of repository

10. Pull GroundingDINO repository git clone https://github.com/IDEA-Research/GroundingDINO.git
11. Open via PyCharm and create interpreter (python3.9) and virtual env using UI
12. Change working directory cd GroundingDINO/
13. Upgrade pip (22.3.1 -> 23.2.1) pip install --upgrade pip
14. Install requirements pip install -r requirements.txt
15. Package build pip install -e .

From here getting error:

RuntimeError: The detected CUDA version (12.0) mismatches the version that was used to compile PyTorch (11.7). Please make sure to use the same CUDA versions.

Also, tryied similar setup with CUDA-11.7 (nvidia drivers 515). No CUDA mismatch errors, but got different error

Questions

• Can I run project with CUDA-12.0, as in notebool example?
• Perhaps, there are device setup mistakes in provided steps?

Thanks in advance!

do you selvo this problem????

[ygtxr1997]

Downgrading to gcc-9, g++-9, and cuda-11.3 works for me.

[flyonok]

@ygtxr1997 mine is ubuntu 22.04 and gcc , g++ is 9, cuda is cuda-11.3,but not woked. (inpaint) root@vm-14a-28gr8:~/Grounded-Segment-Anything-main# g++ --version g++ (Ubuntu 9.5.0-1ubuntu1~22.04) 9.5.0 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.

(inpaint) root@vm-14a-28gr8:~/Grounded-Segment-Anything-main# dpkg -l cuda-11-3 Desired=Unknown/Install/Remove/Purge/Hold | Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend |/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad) ||/ Name Version Architecture Description +++-==============-============-============-================================= ii cuda-11-3 11.3.1-1 amd64 CUDA 11.3 meta-package (inpaint) root@vm-14a-28gr8:~/Grounded-Segment-Anything-main# gcc --version gcc (Ubuntu 9.5.0-1ubuntu1~22.04) 9.5.0 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.

and error message: The detected CUDA version (11.3) mismatches the version that was used to compile PyTorch (12.1). Please make sure to use the same CUDA versions.

[xiwen1]

Is there any workaround except changing cuda-12.0 to cuda-12.1?