search

facebookresearch / DensePose

A real-time approach for mapping all human pixels of 2D RGB images to a 3D surface-based model of the body
http://densepose.org
Other
6.98k stars 1.3k forks source link

OSError: "undefined symbol" when running "test_zero_even_op.py" #143

Open wswdx opened 6 years ago

wswdx commented 6 years ago

Traceback of the error:

Traceback (most recent call last): File "detectron/tests/test_zero_even_op.py", line 117, in c2_utils.import_custom_ops() File "/data/weidx/densepose/detectron/utils/c2.py", line 40, in import_custom_ops dyndep.InitOpsLibrary(custom_ops_lib) File "/home/weidx/anaconda3/envs/caffe2/lib/python2.7/site-packages/caffe2/python/dyndep.py", line 35, in InitOpsLibrary _init_impl(name) File "/home/weidx/anaconda3/envs/caffe2/lib/python2.7/site-packages/caffe2/python/dyndep.py", line 48, in _init_impl ctypes.CDLL(path) File "/home/weidx/anaconda3/envs/caffe2/lib/python2.7/ctypes/init.py", line 365, in init self._handle = _dlopen(self._name, mode) OSError: /data/weidx/densepose/build/libcaffe2_detectron_custom_ops_gpu.so: undefined symbol: _ZN6google8protobuf8internal9ArenaImpl28AllocateAlignedAndAddCleanupEmPFvPvE

I've tried to reinstall protobuf from source and also tried the solution of #119 , both of them don't help. And my protobuf's version is 3.6.1 now, while my caffe2 is installed from Pytorch source code. Please help me out of this! Thanks in advance!

ZhiyLiu commented 6 years ago

Same problem.

ZhiyLiu commented 5 years ago

@wswdx Did you solve your problem?

ckyleda commented 5 years ago

I am also encountering this issue, is clearly a fatal flaw within the way this project is structured.

wswdx commented 5 years ago

@wswdx Did you solve your problem?

Sorry that I haven't notice your comment for so long. I just ignore it because I found testing pose estimation didn't require making ops on my own. If you only wants to use the well trained model to detect pose, there is no need to pass this "test_zero_even_op.py".

wswdx commented 5 years ago

I am also encountering this issue, is clearly a fatal flaw within the way this project is structured.

Yes, maybe. I think you can follow the writers' cuda setting(cuda 8.0 + cudnn 6.0.21) and rebuild, maybe it will work.

ckyleda commented 5 years ago

This doesn't look like a CUDA/cudnn issue - it's definitely seems like a protobuf issue, which is no real surprise considering the amount of problems protobuf's always cause!