Unable to compile NMS and ROI-Align on RTX 3080

[bundle-adjuster]

Hello team, I was trying to reproduce this on my system with a RTX 3080. But it is failing to compile with arch = sm_86 which is the compute architecture for Ampere based GPUs. Given that is there a way I can get this to work?

[bundle-adjuster]

nvcc -c -o nms_kernel.cu.o nms_kernel.cu -x cu -Xcompiler -fPIC -arch=sm_86

Fails due to CUDA 9.0 not supporting it. And the pipeline doesn't build on Higher CUDA versions.

[SJingjia]

nvcc -c -o nms_kernel.cu.o nms_kernel.cu -x cu -Xcompiler -fPIC -arch=sm_86

Fails due to CUDA 9.0 not supporting it. And the pipeline doesn't build on Higher CUDA versions.

using arch=sm_61 with RTX 3080Ti can work

[bundle-adjuster]

nvcc -c -o nms_kernel.cu.o nms_kernel.cu -x cu -Xcompiler -fPIC -arch=sm_86
Fails due to CUDA 9.0 not supporting it. And the pipeline doesn't build on Higher CUDA versions.

using arch=sm_61 with RTX 3080Ti can work

It will work, but it will be extremely slow in data loading. Is there no work around for the same?