gets the same output multiple time.

Ideally, we should replicate lots of things that are in
TensorType.values_eq_approx. See #664.

fixed bug reported by Olivier, by having a typo in the name of the variable when raising an exception ..

throws an exception or silently returns a bunch of 0s)

dictionary and from time to time tried to sync the two. As Josh pointed out,
this was confusing, so I changed the code to use only self. I left the usage
of the info dictionary to construct the op and in __eq__ and __hash__
because at those points I don't care about the individual terms in the
dictionary and it would make the code considerably bigger to spell out all
keys in the info dictionary.

equal and the other way around.

__eq__ ( namely if two things compare equal they need to have the same hash,
but the reverse is not mandatory )

to a variable when it did not know how to compute those gradients.