Closed Kou-99 closed 3 years ago
Hi @Kou-99, thank you for your interest in our work! The only thing that seems strange to me is that, your NVCC version (9.0) is not the same as your CUDA Runtime (10.1). As there are many GPU custom ops, compiling them using different version of NVCC may cause a problem. I'll suggest you to install cuda-10.1
and use its nvcc to compile mmdet3d.
Thank you for your information! I solved this problem by changing my Pytorch version from 1.8.1(cuda10.1) to 1.7.0(cuda9.2) and use Cuda-9.2's nvcc to compile. It seems important to choose Pytorch Cuda version according to nvcc version rather than Cuda runtime version.
@Kou-99 Great to hear! I am glad it works :)
Thanks for your error report and we appreciate it a lot.
Checklist
Describe the bug I installed mmdetection and mmtracking according to your doc and everything went on well. However, when I tried to install mmdetection3d, it failed to compile. I noticed #273 also has a compile issue and you suggested to use a lower version of gcc, but his traceback is quite different from mine so I think that might not be the solution.
Reproduction
Environment
python mmdet3d/utils/collect_env.py
to collect necessary environment infomation and paste it here. sys.platform: linux Python: 3.7.10 (default, Feb 26 2021, 18:47:35) [GCC 7.3.0] CUDA available: True GPU 0,1: GeForce GTX 1080 CUDA_HOME: /usr/local/cuda NVCC: Cuda compilation tools, release 9.0, V9.0.176 GCC: gcc (Ubuntu 5.4.0-6ubuntu1~16.04.11) 5.4.0 20160609 PyTorch: 1.8.1 PyTorch compiling details: PyTorch built with:TorchVision: 0.9.1 OpenCV: 4.5.1 MMCV: 1.3.1 MMCV Compiler: GCC 7.3 MMCV CUDA Compiler: 10.1 MMDetection: 2.11.0+8a432d2
$PATH
,$LD_LIBRARY_PATH
,$PYTHONPATH
, etc.)Error traceback If applicable, paste the error trackback here.
Bug fix If you have already identified the reason, you can provide the information here. If you are willing to create a PR to fix it, please also leave a comment here and that would be much appreciated!