Added support for RTX 50 Series

[kireal]

Requires CUDA Toolkit 12.9 version.

CUDA Toolkit 12.9 adds compiler target support for SM architecture 10.3 (sm_103, sm_103f, sm_103a) and 12.1 (sm_121), enabling development for the latest GPU architectures with specific optimizations for each variant.

Compiled successfully with RTX 5090 under: nvcr.io/nvidia/pytorch:25.04-py3

[peterbjorgensen]

You should also add sm_120 to cuda 12.8

/usr/local/cuda-12.8/bin/nvcc --list-gpu-code
sm_50
sm_52
sm_53
sm_60
sm_61
sm_62
sm_70
sm_72
sm_75
sm_80
sm_86
sm_87
sm_89
sm_90
sm_100
sm_101
sm_120

[loscrossos]

yep.. what @peterbjorgensen says.
i submitted a PR using all kernels for 12.8 and 12.9

#60