Unsupported gpu architecture 'compute_86'

[DIVE128]

Thanks for your great job. Our environment is linux with gpu RTX3090 and the fixed lowest version of pytorch is 1.7 with cuda11.0. When we run mask.sh , it prompts Unsupported gpu architecture compute 86 . Could you release the code that suits a higher version of pytorch? Thanks a lot.

[XinyiYing]

We can use D3Dnet in linux with gpu RTX3090 and the version of pytorch is 1.7.1 with cuda11.0.221 and cudnn8.0.5_0. The problem may be caused by the problem of "The current CUDA hashrate is higher than the CUDA hashrate supported by the current PyTorch version"
The driver version of NVIDIA-SMI is 470.57.02.
Please refer to https://blog.csdn.net/cwm19950318/article/details/111287797 for more details. We hope this may help.