Pytorch CUDA build for compute capability 3.0

I’m pretty new to Pytorch. After some googling, I found that current builds support compute capability >=3.5. I have a laptop with a weak GPU with compute capability 3.0. It supports up to CUDA 11.4 and works fine with Blender. I tried all available PyTorch builds for different CUDA versions but nothing is working. Should I build my own for 3.0? Would it work?