def main(): parser = argparse.ArgumentParser(description='Placement') parser.add_argument("-c", "--config", help="path to config file", default="") parser.add_argument("-i", "--input", help="path to data file", default="") parser.add_argument("-d", "--debug", help="debug mode", action="store_const", dest="loglevel", const=logging.DEBUG, default=logging.WARNING) parser.add_argument("-p", "--pickle", help="create pickle files", action="store_true", dest="create_pickle") parser.add_argument("-v", "--verbose", help="verbose mode", action="store_const", dest="loglevel", const=logging.INFO) args = parser.parse_args(sys.argv[1:]) logging.basicConfig(format='%(asctime)s main: %(message)s', level=args.loglevel) config = pandas.read_csv(args.config, sep=',') X = pandas.read_csv(args.input, sep=',', header=None).values.astype(np.float64) """ discretize the space into cells """ geo_utils = helpers.GeoUtilities(config['lat_min'].iloc[0], config['lat_max'].iloc[0], config['lng_min'].iloc[0], config['lng_max'].iloc[0], cell_length_meters) geo_utils.set_grids() """ segregate data based on time bins """ train_time_utils = helpers.TimeUtilities(time_bin_width_secs) train_time_utils.set_bounds(X) logging.info("Loaded %d data points", len(X)) sim = Sim(X, len(geo_utils.lng_grids), train_time_utils, geo_utils, action_dim, time_bins_per_hour) """ starting time-step for training and testing time bins (tb) """ test_tb_starts = [time_bins_per_hour] train_tb_starts = [ time_bins_per_day * 4 + time_bins_per_hour, time_bins_per_day * 5 + time_bins_per_hour, time_bins_per_day * 6 + time_bins_per_hour ] train_bins = range(time_bins_per_day * 4 + time_bins_per_hour, time_bins_per_day * 7, time_bins_per_day) if args.create_pickle: helpers.load_create_pickle(sim, train_time_utils, geo_utils, X, train_tb_starts, test_tb_starts) with open(r"rrs.pickle", "rb") as input_file: sim.rrs = cPickle.load(input_file) sim.n_reqs = cPickle.load(input_file) sim.pre_sim_pickups = cPickle.load(input_file) with tf.Session() as sess: worker = Worker('worker', sim, 10, train_bins, test_tb_starts, sim.num_cells, sim.n_actions) sess.run(tf.global_variables_initializer()) worker.train(sess) """ model = A2C(sim, 10, train_windows, test_window, sim.num_cells, sim.n_actions, hidden_units) model.train() """ '''