How does PyTorch autograd deal with mutation? In particular, what happens when a mutation occurs on a view, which aliases with some other tensor?
How does PyTorch autograd deal with mutation? In particular, what happens when a mutation occurs on a view, which aliases with some other tensor?