def split_in_balancing_learners(
learners: List[adaptive.BaseLearner],
fnames: List[str],
n_parts: int,
strategy: str = "npoints",
) -> Tuple[List[adaptive.BaseLearner], List[str]]:
r"""Split a list of learners and fnames into `adaptive.BalancingLearner`\s.
Parameters
----------
learners : list
List of learners.
fnames : list
List of filenames.
n_parts : int
Total number of `~adaptive.BalancingLearner`\s.
strategy : str
Learning strategy of the `~adaptive.BalancingLearner`.
Returns
-------
new_learners, new_fnames
"""
new_learners = []
new_fnames = []
for x in split(zip(learners, fnames), n_parts):
learners_part, fnames_part = zip(*x)
learner = adaptive.BalancingLearner(learners_part, strategy=strategy)
new_learners.append(learner)
new_fnames.append(fnames_part)
return new_learners, new_fnames
def runners_in_executor(learners,
client,
nrunners,
goal=None,
interval=3600,
save_kwargs=default_save_kwargs,
client_kwargs=default_client_kwargs):
if goal is None and interval == 0:
raise Exception('Turn on periodic saving if there is no goal.')
futs = []
split_leaners = split(learners, nrunners)
split_targets = split(range(nrunners, len(client)), nrunners)
for i, (_learners, targets) in enumerate(zip(split_leaners,
split_targets)):
learner = adaptive.BalancingLearner(_learners)
fut = client[i].executor.submit(run_learner_in_ipyparallel_client,
learner, goal, interval, save_kwargs,
client_kwargs, targets)
futs.append(fut)
return futs
