Batching rule for torch.mv (#43780)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/43780
The general strategy is:
- unsqueeze the physical inputs enough
- pass the unsqueezed physical inputs to at::matmul
- squeeze any extra dimensions
Test Plan: - `pytest test/test_vmap.py -v`
Reviewed By: ezyang
Differential Revision: D23400842
Pulled By: zou3519
fbshipit-source-id: c550eeb935747c08e3b083609ed307a4374b9096