Closed Taekbum closed 2 years ago
Hi @Taekbum, Thanks for your interests!
Actually I've ever met a very similar problem before when I developed this project. May I ask you for more details to confirm:
I think I met this problem when I was using CUDA 11.1 and PyTorch 1.8.1. I can run the Python script but met exactly the same error when embedding the Python script into C++. I Googled it for a long time but did not find any solutions. In the end I solved this simply by re-installing PyTorch 1.8.0. I've got the feeling that similar things happen with PyTorch 1.8.2
I assume your nvidia driver does not support CUDA 11.3 and you don't want to upgrade it at the moment, and that's why you didn't run the buiding script? So I would suggest you try reinstalling PyTorch 1.8.0 first and see if that solves that error. Feel free to contact me if you have any further questions.
Nice! Feel free to reopen this if you meet any new problems!
Hi, thank you for this nice project. I successfully build dsp_slam with cuda 11.1, pytorch 1.8.2 lts referring to build_cuda113.sh BUT "munmap_chunk(): invalid pointer" error occurred when I tried to run both dsp_slam and dsp_slam_mono. Do you have any ideas about this?