No module named 'lib_kernel.lib_nms.nms.gpu_nms

[huitang]

I add $cpn_root/lib to $PYTHONPATH I tried to google, and it seems that I need to run make in $cpn_root/lib I did Still not solved yet

[tianjiahao]

@huitang I have the same problem as you ,have you solved it?

[Zwei1996]

You need to run the compile.sh in lib/lib_kernel/lib_nms

[huitang]

I did run that still does not work but I solved it by doing this: in compile.sh, remove the last two lines which moves the .so files

On Tue, Oct 16, 2018 at 1:45 AM Zwei1996 notifications@github.com wrote:

You need to run the compile.sh in lib/lib_kernel/lib_nms

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/chenyilun95/tf-cpn/issues/56#issuecomment-430153671, or mute the thread https://github.com/notifications/unsubscribe-auth/ABz1ZGQaOzUUMmGA4c8hExKXLc1Tnua9ks5ulZy_gaJpZM4XEgsC .

[mathpopo]

move old so ,re-build

[Deicide-LC]

Sorry to interrupt you, I have run the compile.sh in lib/lib_kernel/lib_nms . After that, cpython-36m-x86_64-linux-gnu.so were made, but when I run mptest.py, there is an import error: lib/lib_kernel/lib_nms/gpu_nms.cpython-36m-x86_64-linux-gnu.so: undefined symbol: __cudaPopCallConfiguration, I don't know how to fix it and I wish you could help me

[weixinxu666]

You need to run the compile.sh in lib/lib_kernel/lib_nms bash compile.sh

[TianshengSun]

First, you need delete all xxx.so in ROOT/lib/lib_kernel/lib_nms & ROOT/lib/nms & ROOT/lib/utils. Second, open the Makefile which in ROOT/lib and change command from Python3.5 to python3(or your python version), then run make in you conda environment. Third, run compile.sh in ROOT/lib/lib_kernel/lib_nms.