PyTorch 2.0 adds extensive compilation support to the framework; we should test this with our TorchTensor implementations and make sure that

1. it works
2. we don't somehow negate the performance gains