The model.cuda() do not work

My computer configures GTX1080ti, PyTorch =1.0. The model can work. After changing the GPU to GTX3070ti, I run “torch.cuda.is_available()”, and it return True. But, the running code stops to “model = model.cuda()” for a long time, and not possible to keep going.

I have updated the GPU driver to 497.29, CUDA to 10.0, and reinstalled PyTorch to 1.4.0. But the situation has no changing.

Your Ampere GPU needs CUDA11, so install the latest binaries with the 11.3 or 11.5 CUDA runtime.