_nms.so undefined symbol __cudapopcallconfiguration

I am trying to use this repo. But i am getting this error
File "/home/maq/PycharmProjects/pytorch-retinanet/visualize.py", line 106, in <module> main(coco_path, path_to_model) File "/home/maq/PycharmProjects/pytorch-retinanet/visualize.py", line 53, in main retinanet = torch.load(path_to_model) File "/home/maq/PycharmProjects/pytorch4python2/pytorch-retinanet/local/lib/python2.7/site-packages/torch/serialization.py", line 358, in load return _load(f, map_location, pickle_module) File "/home/maq/PycharmProjects/pytorch4python2/pytorch-retinanet/local/lib/python2.7/site-packages/torch/serialization.py", line 542, in _load result = unpickler.load() File "/home/maq/PycharmProjects/pytorch-retinanet/model.py", line 9, in <module> from lib.nms.pth_nms import pth_nms File "/home/maq/PycharmProjects/pytorch-retinanet/lib/nms/pth_nms.py", line 2, in <module> from ._ext import nms File "/home/maq/PycharmProjects/pytorch-retinanet/lib/nms/_ext/nms/__init__.py", line 3, in <module> from ._nms import lib as _lib, ffi as _ffi ImportError: /home/maq/PycharmProjects/pytorch-retinanet/lib/nms/_ext/nms/_nms.so: undefined symbol: __cudaPopCallConfiguration

I tried to troubleshoot it using this link. All the paths added. In FAQ point 7 of link

Result in python environment in which repo is being compiled:
$ python -c “import torch; print(torch.version.cuda)”

$nvcc --version
nvcc: NVIDIA ® Cuda compiler driver
Copyright © 2005-2017 NVIDIA Corporation
Built on Fri_Sep__1_21:08:03_CDT_2017
Cuda compilation tools, release 9.0, V9.0.176

$ python -c “import torch; print(torch.version)”

$python -c “import torch; print(torch.cuda.is_available())”

I am using ubuntu 16.04.

PS: Outside environment in terminal. Result of all the commands are same except
$ python -c “import torch; print(torch.version)”
which result 1.0.1.post2. But i am running repo in environment i dont think this should matter.

1 Like