# Parse arguments
    args = parse_arguments()
    # Load config file
    load_config_file(args.f_config)
    # Adjust paths
    change_paths(args)

    from instances import load_instance, load_tiers, load_seeds
    from policy_search_functions import policy_search, capacity_policy_search
    from objective_functions import multi_tier_objective, multi_tier_objective_ACS
    from policies import MultiTierPolicy as MTP
    from policies import MultiTierPolicy_ACS as MTP_ACS

    # Parse city and get corresponding instance
    instance = load_instance(args.city,
                             setup_file_name=args.f,
                             transmission_file_name=args.tr,
                             hospitalization_file_name=args.hos)
    train_seeds, test_seeds = load_seeds(args.city, args.seed)
    tiers = load_tiers(args.city, tier_file_name=args.t)

    # TODO Read command line args for n_proc for better integration with crunch
    n_proc = args.n_proc

    # TODO: pull out n_replicas_train and n_replicas_test to a config file
    n_replicas_train = args.train_reps
    n_replicas_test = args.test_reps

    # Create the pool (Note: pool needs to be created only once to run on a cluster)
    mp_pool = mp.Pool(n_proc) if n_proc > 1 else None

    # check if the "do-nothing" / 'Stage 1 option is in the tiers. If not, add it
from InterventionsMIP import project_path, instances_path
import multiprocessing as mp
from threshold_policy import threshold_policy_search
from interventions import Intervension
from epi_params import EpiSetup, ParamDistribution
from utils import parse_arguments
from reporting.plotting import plot_stoch_simulations

from instances import load_instance

if __name__ == '__main__':
    # Parse arguments
    args = parse_arguments()

    # Parse city and get corresponding instance
    instance = load_instance(args.city, setup_file_name=args.f)

    # TODO Read command line args for n_proc for better integration with crunch
    n_proc = args.n_proc
    # TODO: pull out n_replicas_train and n_replicas_test to a config file
    n_replicas_train = args.train_reps
    n_replicas_test = args.test_reps
    # Create the pool (Note: pool needs to be created only once to run on a cluster)
    mp_pool = mp.Pool(n_proc) if n_proc > 1 else None
    for sc in [0]:
        for co in [0.95]:
            for base_line_train in [0.4]:
                for base_line_test in [0.4]:
                    for const in [
                            'test'
                    ]:  #[10 * i for i in range(0, 21)] + [215, 1000]: