Minimal reverse-mode autograd engine

With PyTorch available, write the core of an automatic differentiation engine yourself. Build a small Tensor wrapper whose operations record how each result was computed, and a backward() method that walks that graph in reverse and fills in the gradient of every input.

Support a handful of basic operations, make sure gradients add up correctly when a value is used more than once, and check your gradients against PyTorch's own autograd.