ModuleNotFoundError: No module named 'frcnn.nms._ext.nms._nms'

[Mayllisa]

ModuleNotFoundError: No module named 'frcnn.nms._ext.nms._nms'

[timmeinhardt]

Did you successfully complete step 3 of our installation instructions?

[timmeinhardt]

If the error got resolved please close this issue. Thank you!

[nathanlem1]

I am facing the same problem. I have successfully installed the step 3 installation but this problem is still there.

[timmeinhardt]

Did you check whether the compilations of step 3 finished successfully? A missing _nms.py file indicated that something went wrong in step 3.

[nathanlem1]

There is no error during step 3 installation except some warnings. It shows this error:

File "experiments/scripts/test_tracktor.py", line 16, in
from tracktor.datasets.factory import Datasets
File "/storage/AnyVision/tracking_wo_bnw/src/tracktor/datasets/factory.py", line 2, in
from .mot_wrapper import MOT17_Wrapper, MOT19CVPR_Wrapper, MOT17LOWFPS_Wrapper
File "/storage/AnyVision/tracking_wo_bnw/src/tracktor/datasets/mot_wrapper.py", line 4, in
from .mot_sequence import MOT17_Sequence, MOT19CVPR_Sequence, MOT17LOWFPS_Sequence
File "/storage/AnyVision/tracking_wo_bnw/src/tracktor/datasets/mot_sequence.py", line 12, in
from frcnn.model import test
File "/storage/AnyVision/tracking_wo_bnw/src/frcnn/frcnn/model/test.py", line 22, in
from .nms_wrapper import nms
File "/storage/AnyVision/tracking_wo_bnw/src/frcnn/frcnn/model/nms_wrapper.py", line 11, in
from ..nms.pth_nms import pth_nms
File "/storage/AnyVision/tracking_wo_bnw/src/frcnn/frcnn/nms/pth_nms.py", line 2, in
from ._ext import nms
File "/storage/AnyVision/tracking_wo_bnw/src/frcnn/frcnn/nms/_ext/nms/init.py", line 3, in
from ._nms import lib as _lib, ffi as _ffi
ImportError: /storage/AnyVision/tracking_wo_bnw/src/frcnn/frcnn/nms/_ext/nms/_nms.so: undefined symbol: __cudaPopCallConfiguration

[timmeinhardt]

Please check the compilation in step 3 and whether is finished successfully. These errors are most likely due to a failed compilation. And one can easily miss that the compilation failed.

[nathanlem1]

There is no error during installation of step except some warnings. I tried using both cuda-9.0 and cuda-10.0.

[timmeinhardt]

@nathanlem1 If you google the last statement undefined symbol: __cudaPopCallConfiguration this seems to be an issue of wrong PyTorch and CUDA versions. Please make sure to use the same versions for compiling and executing the code (PyTorch 0.3.1 for CUDA 9.0).

[nathanlem1]

There problem is on: from ._nms import lib as _lib, ffi as _ffi

It doesn't find lib and ffi in ._nms.

[nathanlem1]

Yes, I used PyTorch 0.3.1 after making sure it doesn't work with the latest PyTorch version.

[timmeinhardt]

The error message does not say that it can not find lib and ffi. It just says that an error occured while importing the two from ._nms. The error comes when the _nms.so library file encounters a symbol that it does not know. And that symbol is __cudaPopCallConfiguration . And apparently this is an issue of a mismatch between PyTorch und CUDA version.