How to use the deephyper.benchmarks.keras_cmdline function in deephyper

def run(param_dict):
    default_params = defaults()
    for key in default_params:
        if key not in param_dict:
            param_dict[key] = default_params[key]
    optimizer = keras_cmdline.return_optimizer(param_dict)
    print(param_dict)

    BATCH_SIZE = param_dict['batch_size']
    HIDDEN_SIZE = param_dict['hidden_size']
    NUNITS = param_dict['nunits']
    DROPOUT = param_dict['dropout']

    fpath = os.path.dirname(os.path.abspath(__file__))

    tag = 'ml-climate-hm-01'
    inp_df = pd.read_csv(fpath+'/data/1980-2005_2d_002_new.txt', sep="  ", header=None, engine='python')
    out_df1 = pd.read_csv(fpath+'/data/1980-2005_3d_vy_002.txt', sep=r"\s*", header=None, engine='python')
    out_df2 = pd.read_csv(fpath+'/data/1980-2005_3d_ux_002.txt', sep=r"\s*", header=None, engine='python')
    out_df3 = pd.read_csv(fpath+'/data/1980-2005_3d_wz_002.txt', sep=r"\s*", header=None, engine='python')
    out_df4 = pd.read_csv(fpath+'/data/1980-2005_3d_tk_002.txt', sep=r"\s*", header=None, engine='python')
    out_df5 = pd.read_csv(fpath+'/data/1980-2005_3d_qv_002.txt', sep=r"\s*", header=None, engine='python')

if model_path:
        timer.start('model save')
        model = Model(a, b)
        model.save(model_path)
        util.save_meta_data(param_dict, model_mda_path)
        timer.end()
        print(f"saved model to {model_path} and MDA to {model_mda_path}")
    return mse

def augment_parser(parser):
    parser.add_argument('--penalty', type=float, default=0.0)
    return parser


if __name__ == "__main__":
    parser = keras_cmdline.create_parser()
    parser = augment_parser(parser)
    cmdline_args = parser.parse_args()
    param_dict = vars(cmdline_args)
    run(param_dict)

parser.add_argument('--rnn_type', action='store',
                        dest='rnn_type',
                        nargs='?', const=1, type=str, default='LSTM',
                        choices=['LSTM', 'GRU', 'SimpleRNN'],
                        help='type of RNN')

    parser.add_argument('--nhidden', action='store', dest='nhidden',
                        nargs='?', const=2, type=int, default='128',)

    parser.add_argument('--nlayers', action='store', dest='nlayers',
                        nargs='?', const=2, type=int, default='1',)
    return parser


if __name__ == "__main__":
    parser = keras_cmdline.create_parser()
    parser = augment_parser(parser)
    cmdline_args = parser.parse_args()
    param_dict = vars(cmdline_args)
    print(param_dict)
    run(param_dict)

help='Filter 2 units')

    parser.add_argument('--p_size', action='store', dest='p_size',
                        nargs='?', const=2, type=int, default='2',
                        help='pool size')

    parser.add_argument('--nunits', action='store', dest='nunits',
                        nargs='?', const=2, type=int, default='128',
                        help='number of units in FC layer')
    parser.add_argument('--dropout2', type=float, default=0.5, 
                        help='dropout after FC layer')

    return parser

if __name__ == "__main__":
    parser = keras_cmdline.create_parser()
    parser = augment_parser(parser)
    cmdline_args = parser.parse_args()
    param_dict = vars(cmdline_args)
    run(param_dict)

def run(param_dict):
    timer.start('preprocessing')
    param_dict = keras_cmdline.fill_missing_defaults(augment_parser, param_dict)
    pprint(param_dict)
    optimizer = keras_cmdline.return_optimizer(param_dict)

    x_train, y_train, x_val, y_val, chars = generate_data()

    if param_dict['rnn_type'] == 'GRU':
        RNN = layers.GRU
    elif param_dict['rnn_type'] == 'SimpleRNN':
        RNN = layers.SimpleRNN
    else:
        RNN = layers.LSTM

    HIDDEN_SIZE = param_dict['nhidden']
    BATCH_SIZE = param_dict['batch_size']
    NLAYERS = param_dict['nlayers']
    DROPOUT = param_dict['dropout']