File size: 6,817 Bytes
7900c16
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
"""
This script provides an example to wrap TencentPretrain for classification with cross validation.
"""
import sys
import os
import random
import argparse
import torch.nn as nn
import numpy as np

tencentpretrain_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
sys.path.append(tencentpretrain_dir)

from tencentpretrain.utils.constants import *
from tencentpretrain.utils import *
from tencentpretrain.utils.optimizers import *
from tencentpretrain.utils.config import load_hyperparam
from tencentpretrain.utils.seed import set_seed
from tencentpretrain.model_saver import save_model
from tencentpretrain.opts import *
from finetune.run_classifier import *


def main():
    parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)

    # Path options.
    parser.add_argument("--pretrained_model_path", default=None, type=str,
                        help="Path of the pretrained model.")
    parser.add_argument("--output_model_path", default="models/classifier_model.bin", type=str,
                        help="Path of the output model.")
    parser.add_argument("--train_path", type=str, required=True,
                        help="Path of the trainset.")
    parser.add_argument("--config_path", default="models/bert/base_config.json", type=str,
                        help="Path of the config file.")
    parser.add_argument("--train_features_path", type=str, required=True,
                        help="Path of the train features for stacking.")

    # Model options.
    model_opts(parser)

    # Tokenizer options.
    tokenizer_opts(parser)

    # Optimization options.
    optimization_opts(parser)
    parser.add_argument("--soft_targets", action='store_true',
                        help="Train model with logits.")
    parser.add_argument("--soft_alpha", type=float, default=0.5,
                        help="Weight of the soft targets loss.")

    # Training options.
    training_opts(parser)

    # Cross validation options.
    parser.add_argument("--folds_num", type=int, default=5,
                        help="The number of folds for cross validation.")

    adv_opts(parser)

    args = parser.parse_args()

    # Load the hyperparameters from the config file.
    args = load_hyperparam(args)

    # Get logger.
    args.logger = init_logger(args)

    set_seed(args.seed)

    # Count the number of labels.
    args.labels_num = count_labels_num(args.train_path)

    # Build tokenizer.
    args.tokenizer = str2tokenizer[args.tokenizer](args)

    # Training phase.
    dataset = read_dataset(args, args.train_path)
    instances_num = len(dataset)
    batch_size = args.batch_size
    instances_num_per_fold = instances_num // args.folds_num + 1

    args.train_steps = int(instances_num * args.epochs_num / batch_size) + 1

    train_features = []

    total_loss, result = 0.0, 0.0
    acc, marco_f1 = 0.0, 0.0

    for fold_id in range(args.folds_num):
        # Build classification model.
        model = Classifier(args)

        args.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
        model = model.to(args.device)
        load_or_initialize_parameters(args, model)
        optimizer, scheduler = build_optimizer(args, model)
        if args.fp16:
            try:
                from apex import amp
            except ImportError:
                raise ImportError("Please install apex from https://www.github.com/nvidia/apex to use fp16 training.")
            model, optimizer = amp.initialize(model, optimizer, opt_level=args.fp16_opt_level)
            args.amp = amp
        if torch.cuda.device_count() > 1:
            model = torch.nn.DataParallel(model)
        args.model = model

        if args.use_adv:
            args.adv_method = str2adv[args.adv_type](model)

        trainset = dataset[0 : fold_id * instances_num_per_fold] + dataset[(fold_id + 1) * instances_num_per_fold :]

        devset = dataset[fold_id * instances_num_per_fold : (fold_id + 1) * instances_num_per_fold]

        dev_src = torch.LongTensor([example[0] for example in devset])
        dev_tgt = torch.LongTensor([example[1] for example in devset])
        dev_seg = torch.LongTensor([example[2] for example in devset])
        dev_soft_tgt = None

        for epoch in range(1, args.epochs_num + 1):
            random.shuffle(trainset)

            train_src = torch.LongTensor([example[0] for example in trainset])
            train_tgt = torch.LongTensor([example[1] for example in trainset])
            train_seg = torch.LongTensor([example[2] for example in trainset])

            if args.soft_targets:
                train_soft_tgt = torch.FloatTensor([example[3] for example in trainset])
            else:
                train_soft_tgt = None

            model.train()
            for i, (src_batch, tgt_batch, seg_batch, soft_tgt_batch) in enumerate(batch_loader(batch_size, train_src, train_tgt, train_seg, train_soft_tgt)):
                loss = train_model(args, model, optimizer, scheduler, src_batch, tgt_batch, seg_batch, soft_tgt_batch)
                total_loss += loss.item()
                if (i + 1) % args.report_steps == 0:
                    args.logger.info("Fold id: {}, Epoch id: {}, Training steps: {}, Avg loss: {:.3f}".format(fold_id, epoch, i + 1, total_loss / args.report_steps))
                    total_loss = 0.0

        model.eval()
        for i, (src_batch, tgt_batch, seg_batch, soft_tgt_batch) in enumerate(batch_loader(batch_size, dev_src, dev_tgt, dev_seg, dev_soft_tgt)):
            src_batch = src_batch.to(args.device)
            seg_batch = seg_batch.to(args.device)
            with torch.no_grad():
                _, logits = model(src_batch, None, seg_batch)
            prob = nn.Softmax(dim=1)(logits)
            prob = prob.cpu().numpy().tolist()
            train_features.extend(prob)

        output_model_name = ".".join(args.output_model_path.split(".")[:-1])
        output_model_suffix = args.output_model_path.split(".")[-1]
        save_model(model, output_model_name + "-fold_" + str(fold_id) + "." + output_model_suffix)
        result = evaluate(args, devset)
        acc += result[0] / args.folds_num
        f1 = []
        confusion = result[1]
        eps = 1e-9
        for i in range(confusion.size()[0]):
            p = confusion[i, i].item() / (confusion[i, :].sum().item() + eps)
            r = confusion[i, i].item() / (confusion[:, i].sum().item() + eps)
            f1.append(2 * p * r / (p + r + eps))

        marco_f1 += sum(f1) / len(f1) / args.folds_num

    train_features = np.array(train_features)
    np.save(args.train_features_path, train_features)
    args.logger.info("Acc. : {:.4f}".format(acc))
    args.logger.info("Marco F1 : {:.4f}".format(marco_f1))


if __name__ == "__main__":
    main()