simple_NER/train.py at main · TaiToTo/simple_NER · 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
168
169
import os
import pandas as pd
import glob
import numpy as np
import torch
from torch.utils.data import DataLoader
import torch.nn as nn
from tqdm import tqdm

# from transformers import AutoTokenizer, BertTokenizer
import transformers


from prepare_invoice_ner_dataset import label_idx_dict
from prepare_invoice_ner_dataset import split_tokenize_label_dataset
from prepare_invoice_ner_dataset import split_tokenize_label_file
from prepare_invoice_ner_dataset import form_input


def train_fn(data_loader, model, optimizer):
    '''
    Functiont to train the model
    '''
    print("Training phase")
    train_loss = 0
    for index, dataset in enumerate(data_loader):
        batch_input_ids = dataset['ids'].to(config['device'], dtype=torch.long)
        batch_att_mask = dataset['att_mask'].to(config['device'], dtype=torch.long)
        batch_tok_type_id = dataset['tok_type_id'].to(config['device'], dtype=torch.long)
        batch_target = dataset['target'].to(config['device'], dtype=torch.long)

        output = model(batch_input_ids,
                       token_type_ids=None,
                       attention_mask=batch_att_mask,
                       labels=batch_target)

        step_loss = output[0]
        prediction = output[1]

        if ((index + 1) % 10 == 0):
            print("Step {}, train loss {}".format(index + 1, step_loss))

        # print(prediction.shape)

        step_loss.sum().backward()
        optimizer.step()
        train_loss += step_loss
        optimizer.zero_grad()

    return train_loss.sum()


def eval_fn(data_loader, model):
    '''
    Functiont to evaluate the model on each epoch.
    We can also use Jaccard metric to see the performance on each epoch.
    '''

    model.eval()

    eval_loss = 0
    predictions = np.array([], dtype=np.int64).reshape(0, config['MAX_LEN'])
    true_labels = np.array([], dtype=np.int64).reshape(0, config['MAX_LEN'])
    print("\nEvaluation phase")
    with torch.no_grad():
        for index, dataset in enumerate(data_loader):
            batch_input_ids = dataset['ids'].to(config['device'], dtype=torch.long)
            batch_att_mask = dataset['att_mask'].to(config['device'], dtype=torch.long)
            batch_tok_type_id = dataset['tok_type_id'].to(config['device'], dtype=torch.long)
            batch_target = dataset['target'].to(config['device'], dtype=torch.long)

            output = model(batch_input_ids,
                           token_type_ids=None,
                           attention_mask=batch_att_mask,
                           labels=batch_target)

            step_loss = output[0]
            eval_prediction = output[1]

            if ((index + 1) % 10 == 0):
                print("Step {}, eval loss {}".format(index + 1, step_loss))

            eval_loss += step_loss

            eval_prediction = np.argmax(eval_prediction.detach().to('cpu').numpy(), axis=2)
            actual = batch_target.to('cpu').numpy()

            predictions = np.concatenate((predictions, eval_prediction), axis=0)
            true_labels = np.concatenate((true_labels, actual), axis=0)

    return eval_loss.sum(), predictions, true_labels


def train_engine(epoch, train_data, valid_data):
    model = transformers.BertForTokenClassification.from_pretrained('bert-base-uncased', num_labels=len(label_idx_dict))
    model = nn.DataParallel(model)
    model = model.to(config['device'])

    params = model.parameters()
    optimizer = torch.optim.Adam(params, lr=3e-5)

    best_eval_loss = 1000000
    for i in range(epoch):
        train_loss = train_fn(data_loader=train_data,
                              model=model,
                              optimizer=optimizer)
        eval_loss, eval_predictions, true_labels = eval_fn(data_loader=valid_data,
                                                           model=model)

        if eval_loss < best_eval_loss:
            best_eval_loss = eval_loss

            print("\nSaving the model")
            torch.save(model.state_dict(), config['model_name'])

    return model, eval_predictions, true_labels

if __name__ == "__main__":
    train_split_path = 'data/split/train.txt'
    val_split_path = 'data/split/val.txt'
    test_split_path = 'data/split/test.txt'

    with open(train_split_path, "r", encoding="utf-8") as f:
        train_file_path_list = f.read().splitlines()

    model_checkpoint = "bert-base-cased"
    train_split_path = 'data/split/train.txt'
    val_split_path = 'data/split/val.txt'
    test_split_path = 'data/split/test.txt'

    with open(train_split_path, "r", encoding="utf-8") as f:
        train_file_path_list = f.read().splitlines()

    from transformers import AutoTokenizer, BertTokenizer

    model_checkpoint = "bert-base-cased"
    tokenizer = transformers.AutoTokenizer.from_pretrained(model_checkpoint, do_lower_case=True)

    config = {'MAX_LEN': 128,
              'tokenizer': tokenizer,
              'batch_size': 32,
              'Epoch': 3,
              'device': 'cuda' if torch.cuda.is_available() else 'cpu',
              'model_name': 'model1_bert_base_uncased_3_epochs.bin'
              }

    final_train_split_and_tokenized_file_list, \
        final_train_split_and_tokenized_labels, \
        final_train_split_word_id_list, \
        final_train_split_token_ids_list = split_tokenize_label_dataset(train_split_path, tokenizer)

    final_val_split_and_tokenized_file_list, \
        final_val_split_and_tokenized_labels, \
        final_val_split_word_id_list, \
        final_val_split_token_ids_list = split_tokenize_label_dataset(val_split_path, tokenizer)

    train_prod_input = form_input(final_train_split_and_tokenized_file_list, final_train_split_and_tokenized_labels,
                                  final_train_split_token_ids_list, final_train_split_word_id_list, config,
                                  data_type='train')

    val_prod_input = form_input(final_val_split_and_tokenized_file_list, final_val_split_and_tokenized_labels,
                                final_val_split_token_ids_list, final_val_split_word_id_list, config, data_type='train')

    train_prod_input_data_loader = DataLoader(train_prod_input, batch_size=config['batch_size'], shuffle=True)
    val_prod_input_data_loader = DataLoader(val_prod_input, batch_size=config['batch_size'], shuffle=True)

    model, val_predictions, val_true_labels = train_engine(epoch=config['Epoch'],
                                                           train_data=train_prod_input_data_loader,
                                                           valid_data=val_prod_input_data_loader)