search

NVlabs / nvdiffrecmc

Official code for the NeurIPS 2022 paper "Shape, Light, and Material Decomposition from Images using Monte Carlo Rendering and Denoising".
Other
363 stars 28 forks source link

break when run the code #5

Closed semchan closed 1 year ago

semchan commented 1 year ago

Hi, thanks for your work.

When I run the training code, it break here "self.optix_ctx = ou.OptiXContext()". it just pop out "Cuda path /usr/local/cuda" . could you please help me? thanks a lot.

JHnvidia commented 1 year ago

Hello

If that intialization code fails it likely means that our code cannot find OptiX. Nowdays, OptiX is distributed with the NVIDIA drivers and requires a recent version (see https://developer.nvidia.com/designworks/optix/download). We recommend using a recent card with RTX-support, preferably 30XX-series.

We primarily develop on Windows and rely pre-bundled OS/Docker images on our DGX servers, so unfortunately I cannot help if you have a custom installation issue. On Linux we have noted issues with graphics driver version (which controls OptiX version) and installed CUDA version. Using a more recent version of CUDA than Driver/OptiX usually does not work.

If you would like to debug the error a bit more, ou.OptiXContext() calls in to the following C-code. You could add some extra printf(""); statements to see what fails. https://github.com/NVlabs/nvdiffrecmc/blob/main/render/optixutils/c_src/optix_wrapper.cpp#L305

semchan commented 1 year ago

Hello

If that intialization code fails it likely means that our code cannot find OptiX. Nowdays, OptiX is distributed with the NVIDIA drivers and requires a recent version (see https://developer.nvidia.com/designworks/optix/download). We recommend using a recent card with RTX-support, preferably 30XX-series.

We primarily develop on Windows and rely pre-bundled OS/Docker images on our DGX servers, so unfortunately I cannot help if you have a custom installation issue. On Linux we have noted issues with graphics driver version (which controls OptiX version) and installed CUDA version. Using a more recent version of CUDA than Driver/OptiX usually does not work.

If you would like to debug the error a bit more, ou.OptiXContext() calls in to the following C-code. You could add some extra printf(""); statements to see what fails. https://github.com/NVlabs/nvdiffrecmc/blob/main/render/optixutils/c_src/optix_wrapper.cpp#L305

Noted with thanks.