"learn()" takes the "state" parameter twice

The "learn()" function in "graph_tool.inference.blockmodel_em" takes the state parameter both implicitly and explicitly. This makes it impossible to run "em_infer()" with "learn_first = True".

Example:

g = collection.data["polbooks"]
state = EMBlockState(g, B=3)
delta, niter = em_infer(state, learn_first = True)

Output:

TypeError                                 Traceback (most recent call last)
<ipython-input-393-b787977162fc> in <module>()
      1 g = collection.data["polbooks"]
      2 state = EMBlockState(g, B=3)
----> 3 delta, niter = em_infer(state, learn_first = True)

/usr/lib/python3.6/site-packages/graph_tool/inference/blockmodel_em.py in em_infer(state, max_iter, max_e_iter, epsilon, learn_first, verbose)
    282 
    283     if learn_first:
--> 284         state.learn(state, epsilon)
    285 
    286     niter = 0

TypeError: learn() takes from 1 to 2 positional arguments but 3 were given