spconv
spconv copied to clipboard
nvlink fatal : Could not open input file '/usr/lib/x86_64-linux-gnu/libpthread.a'
Following the instructions, when running ‘python setup.py bdist_wheel’ , I got caught in:
[ 95%] Linking CUDA device code CMakeFiles/spconv.dir/cmake_device_link.o
nvlink fatal : Could not open input file '/usr/lib/x86_64-linux-gnu/libpthread.a'
make[2]: *** [src/spconv/CMakeFiles/spconv.dir/build.make:293: src/spconv/CMakeFiles/spconv.dir/cmake_device_link.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:180: src/spconv/CMakeFiles/spconv.dir/all] Error 2
make: *** [Makefile:136: all] Error 2
Traceback (most recent call last):
File "setup.py", line 108, in
I have no idea how to solve the problem and looking forward to hearing from you soon. Thanks! BTW, my system configuration: Operating system: ubuntu 20.04 Compiler: gcc 9.4.0 CUDA version (if applicable): 11.3 CUDNN version (if applicable):8.4.1.50 Python version: 3.7.12
I just solved the issue, by removing Line 7 - 10
of src/spconv/CMakeLists.txt
.
These four lines of code should be removed somehow,
find_package(OpenMP)
if(OpenMP_CXX_FOUND)
target_link_libraries(spconv PUBLIC OpenMP::OpenMP_CXX)
endif()
hi, have u solved it ? I have the same error when runningpython setup.py bdist_wheel
, could u give me some advice
I just solved the issue, by removing
Line 7 - 10
ofsrc/spconv/CMakeLists.txt
.These four lines of code should be removed somehow,
find_package(OpenMP) if(OpenMP_CXX_FOUND) target_link_libraries(spconv PUBLIC OpenMP::OpenMP_CXX) endif()
Hi, I successfully installed spconv1.2.1 as your advice. However, when training the Error comes as follows: "ImportError: cannot import name 'spconv_utils' from partially initialized module 'spconv' (most likely due to a circular import) (/home/durbin/projs/spconv/spconv/init.py)" So maybe it is not a good solution.
How about this?
- source ~/.bashrc
- conda activate <your_env>
- python setup.py bdist_wheel
I have solved the issue. This may caused by the low version of cuda. You can use a higer version of cuda when compiling by adding a line in the line 57
in spconv/setup.py
of v1.2.1
. For example, if you want to use the cuda-11.8
, please add:
'-DCMAKE_CUDA_COMPILER=/usr/local/cuda-11.8/bin/nvcc'