pytorch

eded97ac - as_strided: Fix default storage_offset for reference implementation (#89513)

as_strided: Fix default storage_offset for reference implementation (#89513) This fixes the default storage_offset to take it from the input. This was previously untested, so I've also added a new OpInfo which includes samples with non-zero storage_offsets on the input tensor. Pull Request resolved: https://github.com/pytorch/pytorch/pull/89513 Approved by: https://github.com/ezyang, https://github.com/ngimel