CBraMod/finetune_main.py at main · wjq-learning/CBraMod · 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
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
import argparse
import random

import numpy as np
import torch

from datasets import faced_dataset, seedv_dataset, physio_dataset, shu_dataset, isruc_dataset, chb_dataset, \
    speech_dataset, mumtaz_dataset, seedvig_dataset, stress_dataset, tuev_dataset, tuab_dataset, bciciv2a_dataset
from finetune_trainer import Trainer
from models import model_for_faced, model_for_seedv, model_for_physio, model_for_shu, model_for_isruc, model_for_chb, \
    model_for_speech, model_for_mumtaz, model_for_seedvig, model_for_stress, model_for_tuev, model_for_tuab, \
    model_for_bciciv2a


def main():
    parser = argparse.ArgumentParser(description='Big model downstream')
    parser.add_argument('--seed', type=int, default=3407, help='random seed (default: 0)')
    parser.add_argument('--cuda', type=int, default=1, help='cuda number (default: 1)')
    parser.add_argument('--epochs', type=int, default=50, help='number of epochs (default: 5)')
    parser.add_argument('--batch_size', type=int, default=64, help='batch size for training (default: 32)')
    parser.add_argument('--lr', type=float, default=1e-4, help='learning rate (default: 1e-3)')
    parser.add_argument('--weight_decay', type=float, default=5e-2, help='weight decay (default: 1e-2)')
    parser.add_argument('--optimizer', type=str, default='AdamW', help='optimizer (AdamW, SGD)')
    parser.add_argument('--clip_value', type=float, default=1, help='clip_value')
    parser.add_argument('--dropout', type=float, default=0.1, help='dropout')
    parser.add_argument('--classifier', type=str, default='all_patch_reps',
                        help='[all_patch_reps, all_patch_reps_twolayer, '
                             'all_patch_reps_onelayer, avgpooling_patch_reps]')
    # all_patch_reps: use all patch features with a three-layer classifier;
    # all_patch_reps_twolayer: use all patch features with a two-layer classifier;
    # all_patch_reps_onelayer: use all patch features with a one-layer classifier;
    # avgpooling_patch_reps: use average pooling for patch features;

    """############ Downstream dataset settings ############"""
    parser.add_argument('--downstream_dataset', type=str, default='MentalArithmetic',
                        help='[FACED, SEED-V, PhysioNet-MI, SHU-MI, ISRUC, CHB-MIT, BCIC2020-3, Mumtaz2016, '
                             'SEED-VIG, MentalArithmetic, TUEV, TUAB, BCIC-IV-2a]')
    parser.add_argument('--datasets_dir', type=str,
                        default='/data/datasets/BigDownstream/mental-arithmetic/processed',
                        help='datasets_dir')
    parser.add_argument('--num_of_classes', type=int, default=2, help='number of classes')
    parser.add_argument('--model_dir', type=str, default='/data/wjq/models_weights/Big/BigFaced', help='model_dir')
    """############ Downstream dataset settings ############"""

    parser.add_argument('--num_workers', type=int, default=16, help='num_workers')
    parser.add_argument('--label_smoothing', type=float, default=0.1, help='label_smoothing')
    parser.add_argument('--multi_lr', type=bool, default=True,
                        help='multi_lr')  # set different learning rates for different modules
    parser.add_argument('--frozen', type=bool,
                        default=False, help='frozen')
    parser.add_argument('--use_pretrained_weights', type=bool,
                        default=True, help='use_pretrained_weights')
    parser.add_argument('--foundation_dir', type=str,
                        default='pretrained_weights/pretrained_weights.pth',
                        help='foundation_dir')

    params = parser.parse_args()
    print(params)

    setup_seed(params.seed)
    torch.cuda.set_device(params.cuda)
    print('The downstream dataset is {}'.format(params.downstream_dataset))
    if params.downstream_dataset == 'FACED':
        load_dataset = faced_dataset.LoadDataset(params)
        data_loader = load_dataset.get_data_loader()
        model = model_for_faced.Model(params)
        t = Trainer(params, data_loader, model)
        t.train_for_multiclass()
    elif params.downstream_dataset == 'SEED-V':
        load_dataset = seedv_dataset.LoadDataset(params)
        data_loader = load_dataset.get_data_loader()
        model = model_for_seedv.Model(params)
        t = Trainer(params, data_loader, model)
        t.train_for_multiclass()
    elif params.downstream_dataset == 'PhysioNet-MI':
        load_dataset = physio_dataset.LoadDataset(params)
        data_loader = load_dataset.get_data_loader()
        model = model_for_physio.Model(params)
        t = Trainer(params, data_loader, model)
        t.train_for_multiclass()
    elif params.downstream_dataset == 'SHU-MI':
        load_dataset = shu_dataset.LoadDataset(params)
        data_loader = load_dataset.get_data_loader()
        model = model_for_shu.Model(params)
        t = Trainer(params, data_loader, model)
        t.train_for_binaryclass()
    elif params.downstream_dataset == 'ISRUC':
        load_dataset = isruc_dataset.LoadDataset(params)
        data_loader = load_dataset.get_data_loader()
        model = model_for_isruc.Model(params)
        t = Trainer(params, data_loader, model)
        t.train_for_multiclass()
    elif params.downstream_dataset == 'CHB-MIT':
        load_dataset = chb_dataset.LoadDataset(params)
        data_loader = load_dataset.get_data_loader()
        model = model_for_chb.Model(params)
        t = Trainer(params, data_loader, model)
        t.train_for_binaryclass()
    elif params.downstream_dataset == 'BCIC2020-3':
        load_dataset = speech_dataset.LoadDataset(params)
        data_loader = load_dataset.get_data_loader()
        model = model_for_speech.Model(params)
        t = Trainer(params, data_loader, model)
        t.train_for_multiclass()
    elif params.downstream_dataset == 'Mumtaz2016':
        load_dataset = mumtaz_dataset.LoadDataset(params)
        data_loader = load_dataset.get_data_loader()
        model = model_for_mumtaz.Model(params)
        t = Trainer(params, data_loader, model)
        t.train_for_binaryclass()
    elif params.downstream_dataset == 'SEED-VIG':
        load_dataset = seedvig_dataset.LoadDataset(params)
        data_loader = load_dataset.get_data_loader()
        model = model_for_seedvig.Model(params)
        t = Trainer(params, data_loader, model)
        t.train_for_regression()
    elif params.downstream_dataset == 'MentalArithmetic':
        load_dataset = stress_dataset.LoadDataset(params)
        data_loader = load_dataset.get_data_loader()
        model = model_for_stress.Model(params)
        t = Trainer(params, data_loader, model)
        t.train_for_binaryclass()
    elif params.downstream_dataset == 'TUEV':
        load_dataset = tuev_dataset.LoadDataset(params)
        data_loader = load_dataset.get_data_loader()
        model = model_for_tuev.Model(params)
        t = Trainer(params, data_loader, model)
        t.train_for_multiclass()
    elif params.downstream_dataset == 'TUAB':
        load_dataset = tuab_dataset.LoadDataset(params)
        data_loader = load_dataset.get_data_loader()
        model = model_for_tuab.Model(params)
        t = Trainer(params, data_loader, model)
        t.train_for_binaryclass()
    elif params.downstream_dataset == 'BCIC-IV-2a':
        load_dataset = bciciv2a_dataset.LoadDataset(params)
        data_loader = load_dataset.get_data_loader()
        model = model_for_bciciv2a.Model(params)
        t = Trainer(params, data_loader, model)
        t.train_for_multiclass()
    print('Done!!!!!')


def setup_seed(seed):
    torch.manual_seed(seed)
    torch.cuda.manual_seed_all(seed)
    np.random.seed(seed)
    random.seed(seed)
    torch.backends.cudnn.deterministic = True


if __name__ == '__main__':
    main()