tensor_backward.py

class Tensor:
    ...
    def backward(self):
        # topological order all of the parents in the graph
        topo = []
        visited = set()
        def build_topo(v):
            if v not in visited:
                visited.add(v)
                for parent in v._prev:
                    build_topo(parent)
                topo.append(v)
    
        build_topo(self)
        # Initialize the gradient of the last node as 1
        self.grad = 1
        # Go one variable at a time and apply the chain rule to get its gradient
        for v in reversed(topo):
            v._backward()