-
Notifications
You must be signed in to change notification settings - Fork 2.4k
/
ner.py
193 lines (162 loc) · 6.54 KB
/
ner.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
189
190
191
192
193
# pylint: disable=invalid-name
import argparse
import math
import os
import nemo
from nemo.utils.lr_policies import SquareAnnealing, CosineAnnealing, \
WarmupAnnealing
import nemo_nlp
from nemo_nlp import NemoBertTokenizer, SentencePieceTokenizer
from nemo_nlp.callbacks.ner import \
eval_iter_callback, eval_epochs_done_callback
# Parsing arguments
parser = argparse.ArgumentParser(description="NER_with_pretrained_BERT")
parser.add_argument("--local_rank", default=None, type=int)
parser.add_argument("--batch_size", default=32, type=int)
parser.add_argument("--num_gpus", default=1, type=int)
parser.add_argument("--num_epochs", default=10, type=int)
parser.add_argument("--lr_warmup_proportion", default=0.1, type=float)
parser.add_argument("--lr", default=3e-4, type=float)
parser.add_argument("--weight_decay", default=0, type=float)
parser.add_argument("--optimizer_kind", default="adam", type=str)
parser.add_argument("--mixed_precision", action="store_true")
parser.add_argument("--lr_policy", default="lr_warmup", type=str)
parser.add_argument("--pretrained_bert_model", default="bert-base-cased",
type=str)
parser.add_argument("--data_dir", default="./conll2003", type=str)
parser.add_argument("--classification_dropout", default=0.1, type=float)
parser.add_argument("--max_seq_length", default=128, type=int)
parser.add_argument("--output_filename", default="output.txt", type=str)
parser.add_argument("--tensorboard_filename", default="ner_tensorboard",
type=str)
parser.add_argument("--bert_checkpoint", default=None, type=str)
parser.add_argument("--bert_config", default=None, type=str)
args = parser.parse_args()
data_file = os.path.join(args.data_dir, "train.txt")
if not os.path.isfile(data_file):
raise FileNotFoundError("CoNLL-2003 dataset not found. Dataset can be "
+ "obtained at https://github.com/kyzhouhzau/BERT"
+ "-NER/tree/master/data and should be put in a "
+ "folder at the same level as ner.py.")
try:
import tensorboardX
tb_writer = tensorboardX.SummaryWriter(args.tensorboard_filename)
except ModuleNotFoundError:
tb_writer = None
print("Tensorboard is not available.")
if args.local_rank is not None:
device = nemo.core.DeviceType.AllGpu
else:
device = nemo.core.DeviceType.GPU
if args.mixed_precision is True:
optimization_level = nemo.core.Optimization.mxprO1
else:
optimization_level = nemo.core.Optimization.mxprO0
# Instantiate Neural Factory with supported backend
neural_factory = nemo.core.NeuralModuleFactory(
backend=nemo.core.Backend.PyTorch,
local_rank=args.local_rank,
optimization_level=optimization_level,
placement=device)
if args.bert_checkpoint is None:
tokenizer = NemoBertTokenizer(args.pretrained_bert_model)
bert_model = nemo_nlp.huggingface.BERT(
pretrained_model_name=args.pretrained_bert_model,
factory=neural_factory)
else:
tokenizer = SentencePieceTokenizer(model_path="tokenizer.model")
tokenizer.add_special_tokens(["[MASK]", "[CLS]", "[SEP]"])
bert_model = nemo_nlp.huggingface.BERT(
config_filename=args.bert_config,
factory=neural_factory)
bert_model.restore_from(args.bert_checkpoint)
vocab_size = 8 * math.ceil(tokenizer.vocab_size / 8)
# Training pipeline
print("Loading training data...")
train_data_layer = nemo_nlp.BertNERDataLayer(
tokenizer=tokenizer,
path_to_data=os.path.join(args.data_dir, "train.txt"),
max_seq_length=args.max_seq_length,
is_training=True,
batch_size=args.batch_size,
shuffle=False,
num_workers=0,
local_rank=args.local_rank,
factory=neural_factory)
# Create training loss
tag_ids = train_data_layer.dataset.tag_ids
ner_loss = nemo_nlp.TokenClassificationLoss(
d_model=bert_model.bert.config.hidden_size,
num_labels=len(tag_ids),
dropout=args.classification_dropout,
factory=neural_factory)
input_ids, input_type_ids, input_mask, labels, _ = train_data_layer()
hidden_states = bert_model(
input_ids=input_ids,
token_type_ids=input_type_ids,
attention_mask=input_mask)
train_loss, train_logits = ner_loss(
hidden_states=hidden_states,
labels=labels,
input_mask=input_mask)
# Evaluation pipeline
print("Loading eval data...")
eval_data_layer = nemo_nlp.BertNERDataLayer(
tokenizer=tokenizer,
path_to_data=os.path.join(args.data_dir, "test.txt"),
max_seq_length=args.max_seq_length,
is_training=True,
batch_size=args.batch_size,
shuffle=False,
num_workers=0,
local_rank=args.local_rank,
factory=neural_factory)
input_ids, input_type_ids, eval_input_mask, eval_labels, eval_seq_ids = \
eval_data_layer()
hidden_states = bert_model(
input_ids=input_ids,
token_type_ids=input_type_ids,
attention_mask=eval_input_mask)
eval_loss, eval_logits = ner_loss(
hidden_states=hidden_states,
labels=eval_labels,
input_mask=eval_input_mask)
# Create trainer and execute training action
callback_train = nemo.core.SimpleLossLoggerCallback(
tensor_list2str=lambda x: "{:.3f}".format(x[0].item()),
tb_writer=tb_writer,
step_freq=100)
# Instantiate an optimizer to perform `train` action
optimizer = neural_factory.get_trainer(
params={
"optimizer_kind": args.optimizer_kind,
"optimization_params": {
"num_epochs": args.num_epochs,
"lr": args.lr,
"weight_decay": args.weight_decay,
"amsgrad": True
}})
train_data_size = len(train_data_layer)
steps_per_epoch = int(train_data_size / (args.batch_size * args.num_gpus))
print("steps_per_epoch =", steps_per_epoch)
callback_eval = nemo.core.EvaluatorCallback(
eval_tensors=[eval_logits, eval_seq_ids],
user_iter_callback=lambda x, y: eval_iter_callback(
x, y, eval_data_layer, tag_ids),
user_epochs_done_callback=lambda x: eval_epochs_done_callback(
x, tag_ids, args.output_filename),
tb_writer=tb_writer,
eval_step=steps_per_epoch)
if args.lr_policy == "lr_warmup":
lr_policy_func = WarmupAnnealing(args.num_epochs * steps_per_epoch,
warmup_ratio=args.lr_warmup_proportion)
elif args.lr_policy == "lr_poly":
lr_policy_func = SquareAnnealing(args.num_epochs * steps_per_epoch)
elif args.lr_policy == "lr_cosine":
lr_policy_func = CosineAnnealing(args.num_epochs * steps_per_epoch)
else:
raise ValueError("Invalid lr_policy, must be lr_warmup or lr_poly")
optimizer.train(
tensors_to_optimize=[train_loss],
callbacks=[callback_train, callback_eval],
lr_policy=lr_policy_func)