forked from to-aoki/my-pytorch-bert
/
run_classifier.py
186 lines (176 loc) · 7.89 KB
/
run_classifier.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
# Author Toshihiko Aoki
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""BertClassifier."""
from mptb import BertClassifier
def classification(
config_path='config/bert_base.json',
train_dataset_path='tests/sample_text_class.txt',
eval_dataset_path='tests/sample_text_class.txt',
pretrain_path='pretrain/bert.pt',
tf_pretrain_path=None,
model_path=None,
vocab_path='tests/sample_text.vocab',
sp_model_path='tests/sample_text.model',
save_dir='classifier/',
log_dir=None,
batch_size=2,
max_pos=128,
lr=5e-5,
warmup_proportion=0.1, # warmup_steps = len(dataset) / batch_size * epoch * warmup_proportion
epochs=5,
per_save_epochs=1,
mode='train',
label_num=-1,
balance_weight=False,
balance_sample=False,
under_sampling=False,
under_sampling_cycle=False,
tokenizer_name='google',
read_head=False,
fp16=False
):
if under_sampling_cycle:
under_sampling = True
if mode == 'train':
estimator = BertClassifier(
config_path=config_path,
max_pos=max_pos,
vocab_path=vocab_path,
sp_model_path=sp_model_path,
pretrain_path=pretrain_path,
tf_pretrain_path=tf_pretrain_path,
dataset_path=train_dataset_path,
header_skip=not read_head,
label_num=label_num,
tokenizer_name=tokenizer_name,
under_sampling=under_sampling,
fp16=fp16
)
estimator.train(
traing_model_path=model_path,
batch_size=batch_size,
epochs=epochs,
lr=lr, warmup_proportion=warmup_proportion,
balance_weight=balance_weight,
balance_sample=balance_sample,
under_sampling_cycle=under_sampling_cycle,
save_dir=save_dir,
per_save_epochs=per_save_epochs
)
if eval_dataset_path is None:
return
eval_data_set = estimator.get_dataset(
dataset_path=eval_dataset_path, header_skip=not read_head)
score = estimator.evaluate(dataset=eval_data_set, batch_size=batch_size, log_dir=log_dir)
print(score)
else:
estimator = BertClassifier(
config_path=config_path,
max_pos=max_pos,
vocab_path=vocab_path,
sp_model_path=sp_model_path,
model_path=model_path,
dataset_path=eval_dataset_path,
header_skip=not read_head,
label_num=label_num,
tokenizer_name=tokenizer_name,
under_sampling=under_sampling,
fp16=fp16
)
score = estimator.evaluate(batch_size=batch_size, log_dir=log_dir)
print(score)
if __name__ == '__main__':
import argparse
parser = argparse.ArgumentParser(description='BERT classification.', usage='%(prog)s [options]')
parser.add_argument('--config_path', help='JSON file path for defines networks.', nargs='?',
type=str, default='config/bert_base.json')
parser.add_argument('--train_dataset_path', help='Training Dataset file (TSV file) path for classification.',
nargs='?', type=str, default=None)
parser.add_argument('--eval_dataset_path', help='Evaluate Dataset file (TSV file) path for classification.',
nargs='?', type=str, default=None)
parser.add_argument('--pretrain_path', help='Pre-training PyTorch model path.', nargs='?',
type=str, default=None)
parser.add_argument('--tf_pretrain_path', help='Pre-training TensorFlow(Google) model path.', nargs='?',
type=str, default=None)
parser.add_argument('--model_path', help='Classifier PyTorch model path.', nargs='?',
type=str, default=None)
parser.add_argument('--vocab_path', help='Vocabulary file path for BERT to pre-training.', nargs='?', required=True,
type=str)
parser.add_argument('--sp_model_path', help='Trained SentencePiece model path.', nargs='?',
type=str, default=None)
parser.add_argument('--save_dir', help='Classification model saving directory path.', nargs='?',
type=str, default='classifier/')
parser.add_argument('--log_dir', help='Logging file path.', nargs='?',
type=str, default=None)
parser.add_argument('--batch_size', help='Batch size', nargs='?',
type=int, default=4)
parser.add_argument('--max_pos', help='The maximum sequence length for BERT (slow as big).', nargs='?',
type=int, default=512)
parser.add_argument('--lr', help='Learning rate', nargs='?',
type=float, default=2e-5)
parser.add_argument('--warmup_steps', help='Warm-up steps proportion.', nargs='?',
type=float, default=0.1)
parser.add_argument('--epochs', help='Epochs', nargs='?',
type=int, default=10)
parser.add_argument('--per_save_epochs', help=
'Saving training model timing is the number divided by the epoch number', nargs='?',
type=int, default=1)
parser.add_argument('--mode', help='train or eval', nargs='?',
type=str, default='train')
parser.add_argument('--label_num', help='labels number', nargs='?',
type=int, default=-1)
parser.add_argument('--balance_weight', action='store_true',
help='Use automatically adjust weights')
parser.add_argument('--balance_sample', action='store_true',
help='Use automatically adjust samples(random)')
parser.add_argument('--under_sampling', action='store_true',
help='Use automatically adjust under samples')
parser.add_argument('--under_sampling_cycle', action='store_true',
help='Use automatically adjust under samples cycle peer')
parser.add_argument('--tokenizer', nargs='?', type=str, default='google',
help=
'Select from the following name groups tokenizer that uses only vocabulary files.(mecab, juman)'
)
parser.add_argument('--read_head', action='store_true',
help='Use not include header TSV file')
parser.add_argument('--fp16', action='store_true',
help='Use nVidia fp16 (require apex module)')
args = parser.parse_args()
classification(
config_path=args.config_path,
train_dataset_path=args.train_dataset_path,
eval_dataset_path=args.eval_dataset_path,
pretrain_path= args.pretrain_path,
tf_pretrain_path=args.tf_pretrain_path,
model_path=args.model_path,
vocab_path=args.vocab_path,
sp_model_path=args.sp_model_path,
save_dir=args.save_dir,
log_dir=args.log_dir,
batch_size=args.batch_size,
max_pos=args.max_pos,
lr=args.lr,
warmup_proportion=args.warmup_steps,
epochs=args.epochs,
per_save_epochs=args.per_save_epochs,
mode=args.mode,
label_num=args.label_num,
balance_weight=args.balance_weight,
balance_sample=args.balance_sample,
under_sampling=args.under_sampling,
under_sampling_cycle=args.under_sampling_cycle,
tokenizer_name=args.tokenizer,
read_head=args.read_head,
fp16=args.fp16
)