"prior": offcd.const_prior,
"p": 0.8,
"k": 20,
"thresh": 0.6,
"min_peak_dist": 10
}
metric = "loss"
parallel_args = {
"cmp_class_args": cmp_class_args,
"preprocess_args": preprocess_args,
"param": param,
"metric": metric
}
sequential_args = parallel_args
single_args = {"dt_start": dt_start, "dt_end": dt_end}
single_args.update(parallel_args)
specific_client_args = single_args
fp_specific_client = partial(change_point_alg.run_specific_client,
mac=mac,
server=server,
model_class=BayesianOffline,
out_path=script_dir)
cp_utils.parse_args(partial(change_point_alg.run_single,
run=run), single_args,
partial(change_point_alg.run_parallel,
run=run), parallel_args,
partial(change_point_alg.run_sequential,
run=run), sequential_args, fp_specific_client,
specific_client_args)
metric=metric,
eps_hours=eps_hours,
min_fraction_of_clients=min_fraction_of_clients)
utils.parallel_exec(f_localize_events, dt_ranges)
def run_single(dt_start, dt_end, metric, eps_hours, min_fraction_of_clients):
localize_events(dt_start, dt_end, metric, eps_hours,
min_fraction_of_clients)
if __name__ == "__main__":
min_fraction_of_clients = 0.75
eps_hours = 12
metric = "throughput_up"
dt_start = datetime.datetime(2016, 5, 1)
dt_end = datetime.datetime(2016, 5, 11)
min_fraction_of_clients -= np.finfo(float).eps
parallel_args = {"eps_hours": eps_hours, "metric": metric,
"min_fraction_of_clients": min_fraction_of_clients}
sequential_args = parallel_args
single_args = {"dt_start": dt_start,
"dt_end": dt_end}
single_args.update(parallel_args)
cp_utils.parse_args(run_single, single_args,
run_parallel, parallel_args,
run_sequential, sequential_args,
None, None)
