Semi-Supervised_Models/main.py at master · le-big-mac/Semi-Supervised_Models · 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
import sys
import pickle
import argparse
from utils.datautils import *
from Models import *
import torch.nn.functional as F
import csv
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

parser = argparse.ArgumentParser(description='Choose mode to run model')
parser.add_argument('mode', type=str, choices=['train', 'classify'], help='Mode to run in')
parser.add_argument('data_filepath', type=str, help='File to load data from')
parser.add_argument('output_folder', type=str, help='Folder to save outputs to')
parser.add_argument('--classification_file', type=str, default='outputs.csv', help='File to save classification '
                                                                                   'results to')
args = parser.parse_args()

mode = args.mode
output_folder = args.output_folder

if not os.path.exists(output_folder):
    print('{} does not exist - making directories'.format(output_folder))
    os.makedirs(output_folder)

state_path = '{}/state'.format(output_folder)

if mode == 'train':
    if not os.path.exists(state_path):
        os.mkdir(state_path)
    else:
        print('WARNING: Possibly overriding previous data in {}'.format(output_folder))
        press = input('Press Enter to continue, or \'e\' followed by Enter to exit:')

        if press == 'e':
            sys.exit()

    print('==Loading Data==')

    (labelled_data, labels), unlabelled_data, label_map, col_means = load_train_data_from_file(args.data_filepath)

    train_val_fold = list(stratified_k_fold(labelled_data, labels, 2))

    print("==M2 optimisation==")

    m2, m2_normalizer, m2_accuracies = m2_tool_loop(train_val_fold, labelled_data, labels, unlabelled_data, output_folder,
                                                    device)

    print("==Ladder optimisation==")

    ladder, ladder_normalizer, ladder_accuracies = ladder_tool_loop(train_val_fold, labelled_data, labels, unlabelled_data,
                                                                    output_folder, device)

    print('==Saving State==')

    torch.save(m2, '{}/m2.pt'.format(state_path))
    pickle.dump(m2_normalizer, open('{}/m2_normalizer.p'.format(state_path), 'wb'))

    torch.save(ladder, '{}/ladder.pt'.format(state_path))
    pickle.dump(ladder_normalizer, open('{}/ladder_normalizer.p'.format(state_path), 'wb'))

    pickle.dump(col_means, open('{}/imputation_means.p'.format(state_path), 'wb'))
    pickle.dump(label_map, open('{}/label_map.p'.format(state_path), 'wb'))

    print('M2 2-fold accuracies: {}'.format(m2_accuracies))
    print('Ladder 2-fold accuracies: {}'.format(ladder_accuracies))

    m2_accuracies.insert(0, 'M2')
    ladder_accuracies.insert(0, "Ladder")

    f = open("{}/accs.csv".format(output_folder), 'w')
    writer = csv.writer(f)
    writer.writerow(m2_accuracies)
    writer.writerow(ladder_accuracies)
    f.close()

if mode == 'classify':
    class_file = args.classification_file

    if class_file == 'outputs.csv':
        print('WARNING: Using default output file outputs.csv. This may override previous data')
        press = input('Press Enter to continue, or \'e\' followed by Enter to exit:')

        if press == 'e':
            sys.exit()

    if not (os.path.exists('{}/m2.pt'.format(state_path)) or os.path.exists('{}/ladder.pt'.format(state_path))):
        sys.exit('Models have not been trained')

    print('==Loading Data==')

    col_means = pickle.load(open('{}/imputation_means.p'.format(state_path), 'rb'))
    int_string_map = pickle.load(open('{}/label_map.p'.format(state_path), 'rb'))
    sample_names, data = load_data_to_classify_from_file(args.data_filepath, col_means)

    if device.type == 'cpu':
        m2 = torch.load('{}/m2.pt'.format(state_path), map_location='cpu')
        ladder = torch.load('{}/ladder.pt'.format(state_path), map_location='cpu')
    else:
        m2 = torch.load('{}/m2.pt'.format(state_path))
        ladder = torch.load('{}/ladder.pt'.format(state_path))

    print('==Classifying==')

    m2_normalizer = pickle.load(open('{}/m2_normalizer.p'.format(state_path), 'rb'))
    ladder_normalizer = pickle.load(open('{}/ladder_normalizer.p'.format(state_path), 'rb'))

    m2_data = torch.tensor(m2_normalizer.transform(data)).float()
    m2_results = m2.classify(data)

    ladder_data = torch.tensor(ladder_normalizer.transform(data)).float()
    ladder_results = ladder.classify(data)

    predictions = (F.softmax(m2_results, dim=1) + F.softmax(ladder_results, dim=1))/2

    _, predictions = torch.max(predictions.data, 1)

    labels = [[sample_names[i], int_string_map[l]] for i, l in enumerate(predictions.cpu().numpy())]

    print('==Writing Output==')

    file = open('{}/{}'.format(output_folder, class_file), 'w')
    writer = csv.writer(file)

    for row in labels:
        writer.writerow(row)

    file.close()