Tabular-Temporal-Modulation/train_model_deep.py at main · LAMDA-Tabular/Tabular-Temporal-Modulation · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
from tqdm import tqdm
from model.utils import (
    get_deep_args,
    show_results,
    tune_hyper_parameters,
    get_method,
    set_seeds
)
from model.lib.data import (
    get_dataset,
    get_dataset_tabred
)


if __name__ == '__main__':
    loss_list, results_list, time_list = [], [], []
    args, default_para, opt_space = get_deep_args()
    if args.enable_timestamp:
        train_val_data, test_data, info = get_dataset_tabred(args)
    else:
        train_val_data, test_data, info = get_dataset(args.dataset, args.dataset_path)
    if args.tune:
        args, best_epoch = tune_hyper_parameters(args, opt_space, train_val_data, info, validate_option=args.validate_option)

    ## Training Stage over different random seeds
    for seed in tqdm(range(args.seed_num)):
        args.seed = seed    # update seed
        set_seeds(args.seed)
        method = get_method(args.model_type)(args, info['task_type'] == 'regression')
        if args.validate_option.startswith('holdout'):
            time_cost = method.fit(train_val_data, info)
        elif args.validate_option in ['temporal_cross_validation']:
            N_trainval, C_trainval, M_trainval, y_trainval, split_idx = train_val_data
            train_val_data_all = ({"train": N_trainval}, {"train": C_trainval}, {"train": M_trainval}, {"train": y_trainval})
            print(f'{best_epoch=}')
            time_cost = method.fit(train_val_data_all, info, best_epoch=best_epoch)
        else:
            raise ValueError(f'Invalid validate_option: {args.validate_option}')
        vl, vres, metric_name, predict_logits = method.predict(test_data, info, model_name=args.evaluate_option)

        loss_list.append(vl)
        results_list.append(vres)
        time_list.append(time_cost)

    show_results(args, info, metric_name, loss_list, results_list, time_list)


# python train_model_deep.py --dataset $DATASET_NAME \
#                            --enable_timestamp \
#                            --validate_option $VAL_OPTION \
#                            --model_type $MODEL_NAME \
#                            --cat_policy $CAT_POLICY \
#                            --temporal_policy $TEMPORAL_POLICY \
#                            --gpu 0 --max_epoch 200 --seed_num 15 \
#                            --tune --retune --n_trials 100