Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

DCNv2 undefined symbol #117

Open
caleballen opened this issue Nov 10, 2021 · 3 comments
Open

DCNv2 undefined symbol #117

caleballen opened this issue Nov 10, 2021 · 3 comments

Comments

@caleballen
Copy link

caleballen commented Nov 10, 2021

Hi there,
I can build the extension successfully but upon testing with either test file I get the following error. This would suggest the build was not successful.

python testcuda.py                                                                                                                                                                                                    
Traceback (most recent call last):
  File "testcuda.py", line 11, in <module>
    from dcn_v2 import dcn_v2_conv, DCNv2, DCN
  File "/home/david/CenterTrack/src/lib/model/networks/DCNv2/dcn_v2.py", line 13, in <module>
    import _ext as _backend
ImportError: /home/david/CenterTrack/src/lib/model/networks/DCNv2/_ext.cpython-36m-x86_64-linux-gnu.so: undefined symbol: _ZNSt15__exception_ptr13exception_ptr10_M_releaseEv

I am running on Arch Linux x64 in a conda environment with pytorch==1.4.0 torchvision==0.5.0 and cudatoolkit=10.1.
Build logs

Any pointers on what could be causing this? I've tried many different version combinations from here.

@lwCVer
Copy link

lwCVer commented Nov 19, 2021

Hi there, I can build the extension successfully but upon testing with either test file I get the following error. This would suggest the build was not successful.

python testcuda.py                                                                                                                                                                                                    
Traceback (most recent call last):
  File "testcuda.py", line 11, in <module>
    from dcn_v2 import dcn_v2_conv, DCNv2, DCN
  File "/home/david/CenterTrack/src/lib/model/networks/DCNv2/dcn_v2.py", line 13, in <module>
    import _ext as _backend
ImportError: /home/david/CenterTrack/src/lib/model/networks/DCNv2/_ext.cpython-36m-x86_64-linux-gnu.so: undefined symbol: _ZNSt15__exception_ptr13exception_ptr10_M_releaseEv

I am running on Arch Linux x64 in a conda environment with pytorch==1.4.0 torchvision==0.5.0 and cudatoolkit=10.1. Build logs

Any pointers on what could be causing this? I've tried many different version combinations from here.

I also met the same problem. Have you solved this problem now

@caleballen
Copy link
Author

caleballen commented Nov 19, 2021

I have solved the problem. As I was building DCN to run this, I reinstalled my system as per their environment using Ubuntu 18.04 LTS, Anaconda Python 3.6, CUDA 10.0, and PyTorch v1.0. In this environment DCN built correctly.

As for running it on Arch, I believe the undefined symbol could be due to Arch's modern GCC version, installing an older version may work but in my case it was faster to reload the system with Ubuntu.

@lwCVer
Copy link

lwCVer commented Nov 20, 2021

I have solved the problem. As I was building DCN to run this, I reinstalled my system as per their environment using Ubuntu 18.04 LTS, Anaconda Python 3.6, CUDA 10.0, and PyTorch v1.0. In this environment DCN built correctly.

As for running it on Arch, I believe the undefined symbol could be due to Arch's modern GCC version, installing an older version may work but in my case it was faster to reload the system with Ubuntu.

thank you

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants