search

Owen-Liuyuxuan / visualDet3D

Official Repo for Ground-aware Monocular 3D Object Detection for Autonomous Driving / YOLOStereo3D: A Step Back to 2D for Efficient Stereo 3D Detection
https://owen-liuyuxuan.github.io/papers_reading_sharing.github.io/3dDetection/GroundAwareConvultion/
Apache License 2.0
362 stars 77 forks source link

【Problem】iou3d_cuda: undefined symbol #9

Closed RH-Lin closed 3 years ago

RH-Lin commented 3 years ago

Hi,thanks for you nice work! I've run sh ./make.shand the program have created a iou3d_cuda.cpython-37m-x86_64-linux-gnu.so document. But when I run sh ./launchers/det_precompute.sh config/Yolo3D_example.py train ,it turns out a traceback saying */iou3d/iou3d_cuda.cpython-37m-x86_64-linux-gnu.so: undefined symbol: _Z17nmsNormalLauncherPKfPyif. Could you please tell me how to solve this problem?Thank you!

Owen-Liuyuxuan commented 3 years ago

Could you provide the Ubuntu/Cuda/Python/PyTorch version of your environment so that I could reproduce?

Personally, I have tried on Ubuntu18.04/[Cuda 10.2 11.2]/Python[3.6]/[Pytorch 1.5 1.6 1.8], and Ubuntu20.04/[Cuda 11.1]/Python[3.8]/[Pytorch 1.8]

inSight-mk1 commented 3 years ago

Same issue on CentOS 7, CUDA 10.0, PyTorch 1.8.1, Python 3.6, GCC 5.2.0.

Owen-Liuyuxuan commented 3 years ago

Same issue on CentOS 7, CUDA 10.0, PyTorch 1.8.1, Python 3.6, GCC 5.2.0.

I have not tried out CentOS. I will try to have a follow up

Owen-Liuyuxuan commented 3 years ago

Same issue on CentOS 7, CUDA 10.0, PyTorch 1.8.1, Python 3.6, GCC 5.2.0.

I may need a dockerfile to reproduce this one.. Pytorch does not have a prebuilt 1.8.1 version with CUDA==10.0

inSight-mk1 commented 3 years ago

Same issue on CentOS 7, CUDA 10.0, PyTorch 1.8.1, Python 3.6, GCC 5.2.0.

I have not tried out CentOS. I will try to have a follow up

THX. I will try PyTorch 1.5 or lower on Monday. I suppose this is not a system issue (CentOS or Ubuntu).

inSight-mk1 commented 3 years ago

Same issue on CentOS 7, CUDA 10.0, PyTorch 1.8.1, Python 3.6, GCC 5.2.0.

I may need a dockerfile to reproduce this one.. Pytorch does not have a prebuilt 1.8.1 version with CUDA==10.0

Thank you for your reply. It works with CUDA 10.1 and PyTorch 1.5.0. The reason that this issue occurs may relate to wrong corresponding relations between PyTorch and CUDA version.