torch-mesh-isect
torch-mesh-isect copied to clipboard
What's your CUDA version?
Hi @vchoutas , there is a prolem, hope you can help me. When I run setup.py, CUDA INC PATH seems lacks, so I mannualy point it to '/usr/lcoal/cuda/samples/common/inc'. it worked. However, a cuda error happened when i try examples/detect_and_plot_collisions.py. The error is: Error inside sort: radix_sort: failed on 2nd step: invalid argument Cuda failure src/bvh_cuda_op.cu:877: 'invalid argument'
I meet the same problem .
this seems to come from a thrust::sort function of cuda... i am trying to look into it because this is blocking me :( help
I have tested the package with cuda 10, and its compatible cudnn 7.5. I believe your LD_LIBRARY_PATH environment variable should include the path to these libraries. here is mine for an example: ROOT_PATH/nvidia/cudnn-7.5-cu10.0/lib64:ROOT_PATH/nvidia/cuda-10.0/extras/CUPTI/lib64:ROOT_PATH/nvidia/cuda-10.0/lib64:ROOT_PATH/nvidia/cuda-10.0 ofcourse ROOT_PATH should be updated to your situation...
I have tested this with Pytorch 0.4.1 / CUDA 9.2 and Pytorch 1.1 / CUDA 10. I will make sure to update the README. @mehameha998 Could you please share the mesh that you use and gives you the problem?
@vchoutas I use Pytorch 1.0.1.post2 and cuda 10.0.130. I test the mesh of smpl
Can I install without CUDA? is there a way to disable the option to use CUDA?
Please take a look at the answers here. Other users having the same problem were able to resolve it by following the instructions posted in one of the answers.