Sentence_Matching/train_siamese_net.py at master · doudoucao/Sentence_Matching · 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
from data_helper import DatasetReader
import torch
import torch.nn as nn
from torch.utils.data import DataLoader
from tensorboardX import SummaryWriter
import argparse
from models.siamese_network import Siamese_Net

class Instructor:
    def __init__(self, opt):
        self.opt = opt

        dataset = DatasetReader(embed_dim=opt.embed_dim, max_seq_len=opt.max_seq_len)
        self.train_data_loader = DataLoader(dataset=dataset.train_data, batch_size=opt.batch_size, shuffle=False)
        self.test_data_loader = DataLoader(dataset=dataset.test_data, batch_size=len(dataset.test_data), shuffle=False)
        self.val_data_loader = DataLoader(dataset=dataset.val_data, batch_size=opt.batch_size, shuffle=False)

        self.model = Siamese_Net(opt, dataset.embedding_matrix).to(self.opt.device)
        self._init_and_print_parameters()

    def _init_and_print_parameters(self):
        n_trainable_params, n_nontrainable_params = 0, 0
        for p in self.model.parameters():
            n_params = torch.prod(torch.tensor(p.shape))
            if p.requires_grad:
                n_trainable_params += n_params
                if len(p.shape) > 1:
                    self.opt.initializer(p)
                else:
                    n_nontrainable_params += n_params

        print('n_trainable_params: {0}, n_nontrainable_params: {1}'.format(n_trainable_params, n_nontrainable_params))

    def _train(self, criterion, optimizer):
        # writer = SummaryWriter(log_dir=self.opt.logdir)
        max_val_acc = 0
        max_val_epoch = 0
        global_step = 0
        for epoch in range(self.opt.num_epoch):
            print('>' * 50)
            print('epoch:', epoch)
            n_correct, n_total = 0, 0
            for i_batch, sample_batched in enumerate(self.train_data_loader):
                global_step += 1

                self.model.train()
                optimizer.zero_grad()
                # print(sample_batched['p'].size())
                inputs = [sample_batched['p'].to(self.opt.device), sample_batched['h'].to(self.opt.device)]
                # print(sample_batched['p'].size())
                # print(sample_batched['h'])
                outputs = self.model(inputs)
                # print(outputs.size())
                label = sample_batched['label'].to(self.opt.device)

                loss = criterion(outputs, label)
                loss.backward()
                optimizer.step()

                if global_step % self.opt.log_step == 0:
                    # pred = torch.argmax(outputs, dim=1).item()
                    # acc = accuracy_score(label, outputs)
                    # recall = recall_score(label, outputs)
                    # f1 = f1_score(label, outputs)
                    n_correct += (torch.argmax(outputs, -1) == label).sum().item()
                    n_total += len(outputs)
                    train_acc = n_correct / n_total

                    val_acc = self._evaluate_acc()
                    if val_acc > max_val_acc:
                        max_val_acc = val_acc
                        max_val_epoch = epoch

                    print('loss: {:.4f}, train_acc:{:.4f}, val_acc:{:.4f}'.format(loss.item(), train_acc, val_acc))
        return max_val_acc, max_val_epoch

    def _evaluate_acc(self):
        self.model.eval()
        n_val_correct, n_val_total = 0, 0
        with torch.no_grad():
            for t_batch, t_sample_batched in enumerate(self.val_data_loader):
                t_inputs = [t_sample_batched['p'].to(self.opt.device), t_sample_batched['h'].to(self.opt.device)]
                t_label = t_sample_batched['label'].to(self.opt.device)
                t_outputs = self.model(t_inputs)
                # print(t_outputs.size())

                n_val_correct += (torch.argmax(t_outputs, -1) == t_label).sum().item()
                n_val_total += len(t_outputs)
        val_acc = n_val_correct / n_val_total
        return val_acc

    def _test(self):
        self.model.eval()
        for t_batch, t_sample_batched in enumerate(self.test_data_loader):
            t_inputs = [t_sample_batched['p'].to(self.opt.device), t_sample_batched['h'].to(self.opt.device)]
            t_outputs = self.model(t_inputs)

    def run(self):
        criterion = nn.CrossEntropyLoss()
        _params = filter(lambda p: p.requires_grad, self.model.parameters())
        optimizer = self.opt.optimizer(_params, lr=self.opt.learning_rate)

        max_val_acc, max_val_epoch = self._train(criterion, optimizer)
        print("max_val_acc: {0}".format(max_val_acc))
        print('max_val_epoch: {0}'.format(max_val_epoch))
        return max_val_acc

if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument('--optimizer', default='adam', type=str)
    parser.add_argument('--initializer', default='xavier_uniform_', type=str)
    parser.add_argument('--learning_rate', default=0.01, type=float)
    parser.add_argument('--dropout', default=0.2, type=float)
    parser.add_argument('--num_epoch', default=40, type=int)
    parser.add_argument('--batch_size', default=300, type=int)
    parser.add_argument('--log_step', default=5, type=int)
    parser.add_argument('--embed_dim', default=300, type=int)
    parser.add_argument('--hidden_size', default=100, type=int)
    parser.add_argument('--max_seq_len', default=10, type=int)
    parser.add_argument('--num_perspective', default=10, type=int)
    parser.add_argument('--class_size', default=2, type=int)
    parser.add_argument('--device', default=0, type=int)
    parser.add_argument('--use_char_emb', default=False, type=bool)
    parser.add_argument('--char_hidden_size', default=50, type=int)
    opt = parser.parse_args()

    optimizers = {
        'adadelta': torch.optim.Adadelta,
        'adam': torch.optim.Adam,
        'sgd': torch.optim.SGD
    }
    initializers = {
        'xavier_uniform_', torch.nn.init.xavier_uniform_,
        'xavier_normal_', torch.nn.init.xavier_normal_,
        'orthogonal_', torch.nn.init.orthogonal_,
    }
    opt.initializer = torch.nn.init.xavier_uniform_
    opt.optimizer = optimizers[opt.optimizer]
    opt.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') \
        if opt.device is None else torch.device(opt.device)

    ins = Instructor(opt)
    ins.run()