sappersapper

I tried above example with dtype=pd.SparseDtype(bool) and get similar error ``` df = pd.DataFrame([[1, 0, 1], [0, 1, 0]], dtype=pd.SparseDtype(bool)) df.iloc[0] ``` NotImplementedError: Default 'empty' implementation is invalid for dtype='Sparse[bool,...

Is it this according to the paper? ``` min_loss.backward(retain_graph=True) max_loss.backward() ```

Does the `self.sigma` need to be constrained positive? How to constrain it?