pytorch-faster-rcnn
pytorch-faster-rcnn copied to clipboard
ImportError: ../_roi_pooling.so: undefined symbol: __cudaPopCallConfiguration
Hi, Thank you for your great work. I have some error with roi_pooling layer (also nms and roi_align).
Traceback (most recent call last): File "./tools/test_net.py", line 18, in <module> from nets.vgg16 import vgg16 File "/home/sarairon/PycharmProjects/pytorch-faster-rcnn/tools/../lib/nets/vgg16.py", line 10, in <module> from nets.network import Network File "/home/sarairon/PycharmProjects/pytorch-faster-rcnn/tools/../lib/nets/network.py", line 27, in <module> from layer_utils.roi_pooling.roi_pool import RoIPoolFunction File "/home/sarairon/PycharmProjects/pytorch-faster-rcnn/tools/../lib/layer_utils/roi_pooling/roi_pool.py", line 3, in <module> from ._ext import roi_pooling File "/home/sarairon/PycharmProjects/pytorch-faster-rcnn/tools/../lib/layer_utils/roi_pooling/_ext/roi_pooling/__init__.py", line 3, in <module> from ._roi_pooling import lib as _lib, ffi as _ffi ImportError: /home/sarairon/PycharmProjects/pytorch-faster-rcnn/tools/../lib/layer_utils/roi_pooling/_ext/roi_pooling/_roi_pooling.so: undefined symbol: __cudaPopCallConfiguration
Above error is appeared, so I cannot run the code. Have you ever seen this error before? I used cuda 9.0 and pytorch 0.4, cudnn 7.1 for running.
hi @hesedjds have u solved this problem?
Have you followed the second step as per given in readme file?
Compile modules(nms, roi_pooling(from longcw/yolo2-pytorch), roi_align(from longcw/RoIAlign.pytorch)):
cd pytorch-faster-rcnn/lib bash make.sh cd ../
If anybody wants to compile it with pytorch 0.4.1 and cuda 9.1 then replace the existing C file at 'lib/layers_utils/roi_align/src/crop_and_resize.c' by file given at https://github.com/lxtGH/RoIAlign.pytorch/blob/6e6aaaec6f4947e0462b1dc2f29cd90f195adcb9/roi_align/src/crop_and_resize.c and then compile. Hope it helps!