Enable transposed tensor copy for complex types (#45487)
Summary:
This enables a special copy operator for transposed tensors with more than 360 elements:
https://github.com/pytorch/pytorch/blob/417e3f85e52cdd7439c16c692f7a154d90729dfc/aten/src/ATen/native/Copy.cpp#L19
Steps to repro: python -c "import torch; print(torch.svd(torch.randn(61, 61, dtype=torch.complex64)))"
Fixes https://github.com/pytorch/pytorch/issues/45269
Pull Request resolved: https://github.com/pytorch/pytorch/pull/45487
Reviewed By: anjali411
Differential Revision: D23984441
Pulled By: malfet
fbshipit-source-id: 10ce1d5f4425fb6de78e96adffd119e545b6624f
