NextFace
NextFace copied to clipboard
abdallahdib
I can't start the GPU 3060
It may be caused by redner: https://github.com/BachiLi/redner/issues/169. Unfortunately, the issue has not been fixed.
I've hit this same issue using a 3080.
RuntimeError: Function "_rtpModelUpdate" caught exception: Encountered a CUDA error: radix_sort_temp_size -> cub::DeviceRadixSort::SortPairs returned (8): invalid device function
I think redner needs to be compiled for cuda 11. meantime try to get pytorch with cuda 10 support. plz execute the commands in INSTALL file which will create a compatible environment for u. Hope this help.
Yeah, looks like 3000-series GPUs are incompatible with Cuda 10, so unless Redner gets Cuda 11 support, it's not going to be possible to run NextFace on 3000-series GPUs.
there is no reason for not being able to install pytorch 1.3 with cudatoolkit 10.1 on ur gpu 3000. running this command should work on ur machine :
conda install pytorch==1.3.1 torchvision==0.4.2 cudatoolkit=10.1 -c pytorch
u can always run NextFace on Cpu (change device to cpu in optimConfig.ini)
Nope, that doesn't work, I get the error I mentioned before when running the reconstruction. Same issue brought up here: https://github.com/BachiLi/redner/issues/169. Will try with the CPU though.
The first Cuda version to recognize GTX 30xx cards was 11.1 https://docs.nvidia.com/cuda/archive/11.1.0/cuda-toolkit-release-notes/index.html#cuda-general-new-features
I'm having the same issue. I'm working with RTX 3080 Laptop GPU and pytorch built with CUDA 11.3
I've hit this same issue using a 3080.
RuntimeError: Function "_rtpModelUpdate" caught exception: Encountered a CUDA error: radix_sort_temp_size -> cub::DeviceRadixSort::SortPairs returned (8): invalid device function
I think redner needs to be compiled for cuda 11. meantime try to get pytorch with cuda 10 support. plz execute the commands in INSTALL file which will create a compatible environment for u. Hope this help.
I have compile redner from source for cuda 11. But the issue didn't resolve. My GPU is 3080 Laptop.
@arstropica
The first Cuda version to recognize GTX 30xx cards was 11.1 https://docs.nvidia.com/cuda/archive/11.1.0/cuda-toolkit-release-notes/index.html#cuda-general-new-features
Could you run the code under cuda 11.1 for RTX 30 series ?
also having this same issue with an RTX 3080Ti :( cpu looks like it's running fine.
not only redner but also pytorch from version 1.7 is compatibile with CUDA 11.x so as long as repo won't be updated to those, it won't run on ampere arch.
@arstropica
The first Cuda version to recognize GTX 30xx cards was 11.1 https://docs.nvidia.com/cuda/archive/11.1.0/cuda-toolkit-release-notes/index.html#cuda-general-new-features
Could you run the code under cuda 11.1 for RTX 30 series ?
No. Redner uses Optix 6.5 which will never be compatible with RTX30xx. https://forums.developer.nvidia.com/t/optix-6-5-prime-samples-fail-with-rtx-3080/177078
hm hardware requirements ought to be listed if RTX30X0 is not supported
what is the hardware you are running on?!
hm hardware requirements ought to be listed if RTX30X0 is not supported
what is the hardware you are running on?!
I ended up changing my Optix config to run on cpu and also updated my torch version to 1.12.1. Shame because I have an RTX3090, but no other way to go.
hm hardware requirements ought to be listed if RTX30X0 is not supported what is the hardware you are running on?!
I ended up changing my Optix config to run on cpu and also updated my torch version to 1.12.1. Shame because I have an RTX3090, but no other way to go.
Hello. can you tell me how i can run optix on cpu?
Hello. can you tell me how i can run optix on cpu?
Try my forked branch @: https://github.com/arstropica/NextFace/tree/RTX30XX
Note the changes to the INSTALL doc.