CASE/main.py at master · jfzhouyoo/CASE · 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
154
155
156
157
158
159
160
161
162
163
164
165
166
167
from tqdm import tqdm
from copy import deepcopy
from tensorboardX import SummaryWriter
from torch.nn.init import xavier_uniform_

from src.utils.config import config
from src.utils.common import set_seed
from src.models.CASE.model import CASE
from src.utils.data.loader import prepare_data_seq
from src.models.common import evaluate, count_parameters, make_infinite

def make_model(vocab, emo_num, strategy_num):
    is_eval = config.test
    if config.model == "case":
        model = CASE(
            vocab,
            emotion_num=emo_num,
            strategy_num=strategy_num,
            is_eval=is_eval,
            model_file_path=config.model_file_path if is_eval else None,
        )
    model.to(config.device)

    # Intialization
    for n, p in model.named_parameters():
        if p.dim() > 1 and (n != "embedding.lut.weight" and config.pretrain_emb):
            xavier_uniform_(p)

    print("# PARAMETERS", count_parameters(model))

    return model

def pretrain(model, train_set):
    pretrain_epoch = config.pretrain_epoch
    check_iter = 200
    try:
        model.train()
        writer = SummaryWriter(log_dir=config.save_path)
        weights_best = deepcopy(model.state_dict())
        data_iter = make_infinite(train_set)
        steps = len(train_set)
        for epoch in range(pretrain_epoch):
            for n_iter in tqdm(range(steps)):
                bow_loss = model.train_one_batch(next(data_iter), n_iter)
                writer.add_scalars("bow_loss", {"loss_train": bow_loss}, n_iter)
                if config.noam:
                    writer.add_scalars(
                        "lr", {"learning_rata": model.optimizer._rate}, n_iter
                    )
            weights_best = deepcopy(model.state_dict())
    except KeyboardInterrupt:
        print("-" * 89)
        print("Exiting from training early")
        model.save_model(0, 0)
        weights_best = deepcopy(model.state_dict())

    return weights_best


def train(model, train_set, dev_set):
    check_iter = 2000
    iters = 13000 if config.dataset=="ED" else 6000
    # check_iter = 1
    try:
        model.train()
        best_ppl = 1000
        patient = 0
        writer = SummaryWriter(log_dir=config.save_path)
        weights_best = deepcopy(model.state_dict())
        data_iter = make_infinite(train_set)
        for n_iter in tqdm(range(1000000)):
            bow_loss, kl_loss, mim_loss, ctx_loss, ppl, str_loss, str_acc, emo_loss, emo_acc = model.train_one_batch(next(data_iter), n_iter)
            writer.add_scalars("bow_loss", {"loss_train": bow_loss}, n_iter)
            writer.add_scalars("kl_loss", {"loss_train": kl_loss}, n_iter)
            writer.add_scalars("mim_loss", {"loss_train": mim_loss}, n_iter)
            writer.add_scalars("ctx_loss", {"loss_train": ctx_loss}, n_iter)
            writer.add_scalars("ppl", {"ppl_train": ppl}, n_iter)
            if config.dataset == "ESConv":
                writer.add_scalars("str_loss", {"loss_train": str_loss}, n_iter)
                writer.add_scalars("str_acc", {"str_acc_train": str_acc}, n_iter)
            else:
                writer.add_scalars("emo_loss", {"loss_train": emo_loss}, n_iter)
                writer.add_scalars("emo_acc", {"emo_acc_train": emo_acc}, n_iter)
            if config.noam:
                writer.add_scalars(
                    "lr", {"learning_rata": model.optimizer._rate}, n_iter
                )

            if (n_iter + 1) % check_iter == 0:
                model.eval()
                model.epoch = n_iter
                bow_loss_val, kl_loss_val, mim_loss_val, ctx_loss_val, ppl_val, str_loss_val, str_acc_val, emo_loss_val, emo_acc_val, _ = evaluate(
                    model, dev_set, ty="valid", max_dec_step=50
                )
                writer.add_scalars("bow_loss", {"bow_loss_valid": bow_loss_val}, n_iter)
                writer.add_scalars("kl_loss", {"kl_loss_valid": kl_loss_val}, n_iter)
                writer.add_scalars("mim_loss", {"mim_loss_valid": mim_loss_val}, n_iter)
                writer.add_scalars("ctx_loss", {"ctx_loss_valid": ctx_loss_val}, n_iter)
                writer.add_scalars("ppl", {"ppl_valid": ppl_val}, n_iter)
                if config.dataset == "ESConv":
                    writer.add_scalars("str_loss", {"str_loss_valid": str_loss_val}, n_iter)
                    writer.add_scalars("str_acc", {"str_acc_valid": str_acc_val}, n_iter)
                else:
                    writer.add_scalars("emo_loss", {"emo_loss_valid": emo_loss_val}, n_iter)
                    writer.add_scalars("emo_acc", {"emo_acc_valid": emo_acc_val}, n_iter)
                model.train()
                if n_iter < iters:
                    continue
                if ppl_val <= best_ppl:
                    best_ppl = ppl_val
                    patient = 0
                    model.save_model(best_ppl, n_iter)
                    weights_best = deepcopy(model.state_dict())
                else:
                    patient += 1
                if patient > 2:
                    break

    except KeyboardInterrupt:
        print("-" * 89)
        print("Exiting from training early")
        model.save_model(best_ppl, n_iter)
        weights_best = deepcopy(model.state_dict())

    return weights_best

def test(model, test_set):
    model.eval()
    model.is_eval = True
    bow_loss_test, kl_loss_test, mim_loss_test, ctx_loss_test, ppl_test, str_loss_test, str_acc_test, emo_loss_test, emo_acc_test, results = evaluate(
        model, test_set, ty="test", max_dec_step=50
    )
    file_summary = config.save_path + "/results.txt"
    with open(file_summary, "w") as f:
        f.write("EVAL\tBOW_Loss\tKL_Loss\tMIM_Loss\tCTX_Loss\tPPL\tSTR_loss\tSTR_acc\tEMO_loss\tEMO_acc\n")
        f.write(
            "{}\t{:.4f}\t{:.4f}\t{:.4f}\t{:.4f}\t{:.4f}\t{:.4f}\t{:.4f}\t{:.4f}\n".format(
                bow_loss_test, kl_loss_test, mim_loss_test, ctx_loss_test, ppl_test, str_loss_test, str_acc_test, emo_loss_test, emo_acc_test
            )
        )
        for r in results:
            f.write(r)

def main():
    set_seed()  # for reproducibility

    train_set, dev_set, test_set, vocab, emo_num, strategy_num = prepare_data_seq(
        batch_size=config.batch_size
    )

    model = make_model(vocab, emo_num, strategy_num)

    if config.test:
        test(model, test_set)
    else:
        if config.pretrain:
            weights_best = pretrain(model, train_set)
            model.load_state_dict({name: weights_best[name] for name in weights_best})
            config.pretrain = False
        weights_best = train(model, train_set, dev_set)
        model.epoch = 1
        model.load_state_dict({name: weights_best[name] for name in weights_best})
        test(model, test_set)

if __name__ == '__main__':
    # prepare_data_seq(batch_size=config.batch_size)
    main()