-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrun.py
188 lines (172 loc) · 8.88 KB
/
run.py
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
175
176
177
178
179
180
181
182
183
184
185
186
187
188
import torch
import random
import os
import utils
import config
import logging
import numpy as np
from data_process import Processor
from data_loader import NERDataset
from model import BertNER
from train import train, evaluate
from torch.utils.data import DataLoader
from transformers.optimization import AdamW,get_linear_schedule_with_warmup
import warnings
warnings.filterwarnings('ignore')
def seed_everything(seed=42):
"""
Set the seed for the entire development environment.
:param seed:
:return:
"""
random.seed(seed)
os.environ['PYTHONHASHSEED'] = str(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
# some cudnn methods can be random even after fixing the seed
# unless you tell it to be deterministic
torch.backends.cudnn.deterministic = True
def test():
data = np.load(config.test_dir, allow_pickle=True)
word_test = data["words"]
label_test = data["labels"]
test_dataset = NERDataset(word_test, label_test, config)
logging.info("--------Dataset Build!--------")
# build data_loader
test_loader = DataLoader(test_dataset, batch_size=config.batch_size,
shuffle=False, collate_fn=test_dataset.collate_fn)
label_embedding=torch.load(config.label_embedding_dir).to(config.device)
#label_embedding_bt=label_embedding.repeat_interleave(repeats=config.batch_size, dim=0)
logging.info("--------Get Data-loader!--------")
# Prepare model
if config.model_dir is not None:
model = BertNER.from_pretrained(config.model_dir)
model.to(config.device)
logging.info("--------Load model from {}--------".format(config.model_dir))
else:
logging.info("--------No model to test !--------")
return
val_metrics = evaluate(test_loader, model, label_embedding,mode='test')
val_precision = val_metrics['precision']
val_recall = val_metrics['recall']
val_f1 = val_metrics['f1']
logging.info("test loss: {}, precision: {}, recall: {}, f1 score: {}".format(val_metrics['loss'],
val_precision, val_recall, val_f1))
val_metr_labels = val_metrics['metr_labels']
for label in config.labels:
val_metr_label=val_metr_labels[label]
logging.info("metrics of {}: precision: {}, recall: {}, f1 score: {}".format(label,
val_metr_label[0],val_metr_label[1],val_metr_label[2]))
def load_data():
train_data = np.load(config.train_dir, allow_pickle=True)
dev_data = np.load(config.dev_dir, allow_pickle=True) #dev
word_train = train_data["words"]
label_train = train_data["labels"]
word_dev = dev_data["words"]
label_dev = dev_data["labels"]
return word_train, word_dev, label_train, label_dev
def run():
"""train the model"""
# set the logger
utils.set_logger(config.log_dir)
logging.info("device: {}".format(config.device))
# process the data, separating the text and labels
processor = Processor(config)
processor.process()
logging.info("--------Process Done!--------")
# load train set and dev set
word_train, word_dev, label_train, label_dev = load_data()
# build dataset
train_dataset = NERDataset(word_train, label_train, config)
dev_dataset = NERDataset(word_dev, label_dev, config)
logging.info("--------Dataset Build!--------")
# get dataset size
train_size = len(train_dataset)
# build data_loader
train_loader = DataLoader(train_dataset, batch_size=config.batch_size,
shuffle=True, collate_fn=train_dataset.collate_fn)
dev_loader = DataLoader(dev_dataset, batch_size=config.batch_size,
shuffle=True, collate_fn=dev_dataset.collate_fn)
label_embedding=torch.load(config.label_embedding_dir).to(config.device)
#label_embedding_bt=label_embedding.repeat_interleave(repeats=config.batch_size, dim=0)
logging.info("--------Get Dataloader!--------")
# Prepare model
device = config.device
model = BertNER.from_pretrained(config.bert_model, num_labels=len(config.label2id))
model.to(device)
# Prepare optimizer
if config.full_fine_tuning:
# model.named_parameters(): [bert, classifier, crf]
bert_optimizer = list(model.bert.named_parameters())
attention_optimizer= list(model.multiheadAttn.named_parameters())
lstm_optimizer = list(model.bilstm.named_parameters())
start_optimizer = list(model.start.named_parameters())
end_optimizer = list(model.end.named_parameters())
classifier_optimizer = list(model.classifier.named_parameters())
no_decay = ['bias', 'LayerNorm.bias', 'LayerNorm.weight']
optimizer_grouped_parameters = [
#bert_optimizer
{'params': [p for n, p in bert_optimizer if not any(nd in n for nd in no_decay)],
'weight_decay': config.weight_decay},
{'params': [p for n, p in bert_optimizer if any(nd in n for nd in no_decay)],
'weight_decay': 0.0},
#lstm_optimizer
{'params': [p for n, p in lstm_optimizer if not any(nd in n for nd in no_decay)],
'lr': config.learning_rate * 5, 'weight_decay': config.weight_decay},
{'params': [p for n, p in lstm_optimizer if any(nd in n for nd in no_decay)],
'lr': config.learning_rate * 5, 'weight_decay': 0.0},
# #attention_optimizer
# {'params': [p for n, p in attention_optimizer if not any(nd in n for nd in no_decay)],
# 'lr': config.learning_rate * 5, 'weight_decay': config.weight_decay},
# {'params': [p for n, p in attention_optimizer if any(nd in n for nd in no_decay)],
# 'lr': config.learning_rate * 5, 'weight_decay': 0.0},
#start_optimizer
{'params': [p for n, p in start_optimizer if not any(nd in n for nd in no_decay)],
'lr': config.learning_rate * 5, 'weight_decay': config.weight_decay},
{'params': [p for n, p in start_optimizer if any(nd in n for nd in no_decay)],
'lr': config.learning_rate * 5, 'weight_decay': 0.0},
#end_optimizer
{'params': [p for n, p in end_optimizer if not any(nd in n for nd in no_decay)],
'lr': config.learning_rate * 5, 'weight_decay': config.weight_decay},
{'params': [p for n, p in end_optimizer if any(nd in n for nd in no_decay)],
'lr': config.learning_rate * 5, 'weight_decay': 0.0},
#classifier_optimizer
{'params': [p for n, p in classifier_optimizer if not any(nd in n for nd in no_decay)],
'lr': config.learning_rate * 5, 'weight_decay': config.weight_decay},
{'params': [p for n, p in classifier_optimizer if any(nd in n for nd in no_decay)],
'lr': config.learning_rate * 5, 'weight_decay': 0.0},
#crf_optimizer
{'params': model.crf.parameters(), 'lr': config.learning_rate * 100} #由5改为100
]
# #lstm_optimizer
# {'params': [p for n, p in lstm_optimizer if not any(nd in n for nd in no_decay)],
# 'lr': config.learning_rate * 5, 'weight_decay': config.weight_decay},
# {'params': [p for n, p in lstm_optimizer if any(nd in n for nd in no_decay)],
# 'lr': config.learning_rate * 5, 'weight_decay': 0.0},
# #lstm_optimizer
# {'params': model.bilstm.parameters(), 'lr': config.learning_rate * 5},
# #attention_optimizer
# {'params': [p for n, p in attention_optimizer if not any(nd in n for nd in no_decay)],
# 'lr': config.learning_rate * 5, 'weight_decay': config.weight_decay},
# {'params': [p for n, p in attention_optimizer if any(nd in n for nd in no_decay)],
# 'lr': config.learning_rate * 5, 'weight_decay': 0.0},
# #attention_optimizer
# {'params': model.multiheadAttn.parameters(), 'lr': config.learning_rate},
# only fine-tune the head classifier
else:
param_optimizer = list(model.classifier.named_parameters())
optimizer_grouped_parameters = [{'params': [p for n, p in param_optimizer]}]
optimizer = AdamW(optimizer_grouped_parameters, lr=config.learning_rate, correct_bias=False)
train_steps_per_epoch = train_size // config.batch_size
scheduler = get_linear_schedule_with_warmup(optimizer,
num_warmup_steps=(config.epoch_num // 10) * train_steps_per_epoch,
num_training_steps=config.epoch_num * train_steps_per_epoch)
# Train the model
logging.info("--------Start Training!--------")
train(train_loader, dev_loader, model, optimizer, scheduler, config.model_dir,label_embedding)
if __name__ == '__main__':
seed_everything(config.seed)
run()
test()