BalancingLearner of SequenceLearners does not finish calculation

[andrenmelo]

The following code fails to finish the calculation

import adaptive
adaptive.notebook_extension()

def dummy(x):
    return x

sequence = range(20)
learners = [adaptive.SequenceLearner(dummy, sequence)]
bl = adaptive.BalancingLearner(learners)
runner = adaptive.Runner(bl, lambda bl: all([l.done() for l in bl.learners]), 
                         log=True)
runner.live_info()

The error is happening in runner._get_futures(). I think something goes wrong when the runner asks for more points then there are points left in sequence.

[andrenmelo]

The log is just
[('ask', 48)]

[akhmerov]

As a temporary workaround you can use a single sequence learner (there's nothing to balance really).

[andrenmelo]

Oh that was just a minimal example. In reality I have a bunch of sequence learners.

[akhmerov]

I mean that also if you have a bunch of sequence learners, you can directly combine and shuffle the sequences. Since all the values are predefined, balancing learner does not bring any advantage as far as I can tell.

[basnijholt]

It should work with the fix in #214.

I will test a bit more.