SoftRas icon indicating copy to clipboard operation
SoftRas copied to clipboard

undefined symbol: cudaSetupArgument

Open wangleishan opened this issue 4 years ago • 5 comments

After I finished processing dependencies for soft-renderer==1.0.0, I run you code. But I has a error below. I try python=3.6 or python=3.7. it all failed. my cudatoolkit =10.1.243 .

Traceback (most recent call last): File "trainval.py", line 8, in import soft_renderer as sr File "/home/wangls@1/anaconda3/envs/py37/lib/python3.7/site-packages/soft_renderer-1.0.0-py3.7-linux-x86_64.egg/soft_renderer/init.py", line 1, in from . import functional File "/home/wangls@1/anaconda3/envs/py37/lib/python3.7/site-packages/soft_renderer-1.0.0-py3.7-linux-x86_64.egg/soft_renderer/functional/init.py", line 4, in from .load_obj import load_obj File "/home/wangls@1/anaconda3/envs/py37/lib/python3.7/site-packages/soft_renderer-1.0.0-py3.7-linux-x86_64.egg/soft_renderer/functional/load_obj.py", line 7, in import soft_renderer.cuda.load_textures as load_textures_cuda ImportError: /home/wangls@1/anaconda3/envs/py37/lib/python3.7/site-packages/soft_renderer-1.0.0-py3.7-linux-x86_64.egg/soft_renderer/cuda/load_textures.cpython-37m-x86_64-linux-gnu.so: undefined symbol: cudaSetupArgument

wangleishan avatar Oct 27 '21 06:10 wangleishan

Can you try removing "/home/wangls@1/anaconda3/envs/py37/lib/python3.7/site-packages/soft_renderer-1.0.0-py3.7-linux-x86_64.egg" folder and rebuilding it? It seems that the problem is due to version mismatch between already built modules and current cuda version.

ShichenLiu avatar Oct 27 '21 06:10 ShichenLiu

The error is not resolved, when I create a new environment, the error is still the same .

wangleishan avatar Oct 27 '21 06:10 wangleishan

Can you make sure your system cuda version is also 10.1? Because compiling cuda code does not necessarily use the cudatoolkit version.

ShichenLiu avatar Oct 27 '21 06:10 ShichenLiu

yes. when I use 'nvcc -V' it show that my cuda version is 10.1

wangleishan avatar Oct 27 '21 06:10 wangleishan

yes. when I use 'nvcc -V' it show that my cuda version is 10.1

I also meet the problem, I changed my "nvcc -V" to 10.1 but also have the problem of version dismatch. However, I try again a few days later and the problem is solved

cavalier501 avatar Apr 12 '22 05:04 cavalier501