Cuda_assert: unknown error

[h8c2]

Loading model from: "/home/DISCOVER_summer2022/houc/mega-nerf/Datasets/building-pixsfm/building-pixsfm-8.pt" Model loaded cuda_assert: unknown error /home/DISCOVER_summer2022/houc/mega-nerf-viewer/src/renderer/cuda_renderer.cpp 444 Segmentation fault

[h8c2]

line 442-449 in cuda_render.cpp
cuda_assert(cudaGraphicsGLRegisterImage(&cgr[index * 2], rb[index], GL_RENDERBUFFER, cudaGraphicsRegisterFlagsSurfaceLoadStore | cudaGraphicsRegisterFlagsWriteDiscard), __FILE__, __LINE__); cuda_assert(cudaGraphicsGLRegisterImage(&cgr[index * 2 + 1], depth_rb[index], GL_RENDERBUFFER, cudaGraphicsRegisterFlagsSurfaceLoadStore | cudaGraphicsRegisterFlagsWriteDiscard), __FILE__, __LINE__);

[alan355]

Have you solved this question?I meet the same problem with you

[pengzhao-life]

I got the same error.

[h8c2]

Have you solved this question?I meet the same problem with you

I failed to run this code on a remote server while I succeeded in running it on the local computer. The problem may be caused by the graphic display setting of the computer. The remote server is likely to use Integrated graphics instead of independent GPU by default.

[alan355]

Have you solved this question?I meet the same problem with you

I failed to run this code on a remote server while I succeeded in running it on the local computer. The problem may be caused by the graphic display setting of the computer. The remote server is likely to use Integrated graphics instead of independent GPU by default.

Thank you a lot. I have got your idea.

[chl2]

I got the same error. How to save as obj or video format?

[abovzv]

Have you solved this question?I meet the same problem with you

I failed to run this code on a remote server while I succeeded in running it on the local computer. The problem may be caused by the graphic display setting of the computer. The remote server is likely to use Integrated graphics instead of independent GPU by default.

Maybe I got the same problem with yours, and this GPU is 3090, which is not Integrated though it's still on the remote server.
How can I work it out? Thanks~

[hvkwak]

I had the same problem and asked a question at nigels-com/glew#375 to see if there's a solution. Anyone found a solution?

[annkri]

I had the same issue, and fixed it by adding __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia before the command; the second one absolutely necessary for it to start.
(Inspired by this thread: https://forums.developer.nvidia.com/t/running-sample-gives-code-999-cudaerrorunknown-how-can-this-be-solved/128947)

__NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia ./mega-nerf-viewer ~/data/rubble/output/rubble_octree.npz --model_path ~/data/rubble/rubble-pixsfm-8.pt