Add support for multiple outputs in structured kernels, port fractional_max_pool2d (#55581)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/55581
Multiple outputs now OK, as long as their all Tensor. Ported
fractional_max_pool2d to make sure the shindig all works.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Test Plan: Imported from OSS
Reviewed By: SplitInfinity
Differential Revision: D27641267
Pulled By: ezyang
fbshipit-source-id: f88bfcd2b11e9ae90b023c9310c033d12637a53e