Currently there's no possibility to validate a splittable algorithm locally.

There's nothing in the experiment execution part and in the algorithm execution the number of slots is locked to one and not even used.

The splitting part is currently done at the beat/beat.web> level and then the scheduler takes the generated tasks and sends them to the appropriate worker machines.

That part shall be ported to beat/beat.cmdline> so that the behaviour can be mimicked locally through maybe multiprocessing.