-
Notifications
You must be signed in to change notification settings - Fork 0
/
train_cla.py
156 lines (138 loc) · 5.9 KB
/
train_cla.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
import os
import copy
import torch
import torch.nn.functional as F
import math
import json
import logging
import argparse
import datetime
import random
import numpy as np
from transformers import BertTokenizer, BertModel, AdamW
import utils
import models
import trainers
def main(args, logger):
tok_name_tmp = args['tok_name'].replace('//', '_').replace('-', '_')
agnews_train = utils.load_data(
'data/agnews.train.%s.json' % (tok_name_tmp), 'agnews')
amazon_train = utils.load_data(
'data/amazon.train.%s.json' % (tok_name_tmp), 'amazon')
dbpedia_train = utils.load_data(
'data/dbpedia.train.%s.json' % (tok_name_tmp), 'dbpedia')
yahoo_train = utils.load_data(
'data/yahoo.train.%s.json' % (tok_name_tmp), 'yahoo')
yelp_train = utils.load_data(
'data/yelp.train.%s.json' % (tok_name_tmp), 'yelp')
agnews_test = utils.load_data(
'data/agnews.test.%s.json' % (tok_name_tmp), 'agnews')
amazon_test = utils.load_data(
'data/amazon.test.%s.json' % (tok_name_tmp), 'amazon')
dbpedia_test = utils.load_data(
'data/dbpedia.test.%s.json' % (tok_name_tmp), 'dbpedia')
yahoo_test = utils.load_data(
'data/yahoo.test.%s.json' % (tok_name_tmp), 'yahoo')
yelp_test = utils.load_data(
'data/yelp.test.%s.json' % (tok_name_tmp), 'yelp')
train_data = [
agnews_train,
amazon_train,
dbpedia_train,
yahoo_train,
yelp_train
]
test_data = [
agnews_test,
amazon_test,
dbpedia_test,
yahoo_test,
yelp_test
]
args['logger'].info('Load %d items from training dataset, %d items from test dataset.' \
% (sum([len(i) for i in train_data]), sum([len(i) for i in test_data])))
trainer_select = {
'multiclass' : trainers.multiclass_train,
'sequential' : trainers.sequential_train,
'replay' : trainers.original_replay_train,
}
if args['model'] == 'multiclass':
model = models.MultiClassModel(args)
layer_optimizer = None
elif args['model'] == 'noptr':
model = models.NoPretrainedModel(args)
layer_optimizer = None
'''
elif args['model'] == 'layer':
model = models.LayerSymModel(args)
layer_optimizer = []
for i in range(12):
l_param = \
[p for p in model.layer_pooler[i].parameters() if p.requires_grad] + \
[p for p in model.layer_classifier[i].parameters() if p.requires_grad]
l_optim = torch.optim.Adadelta(l_param, lr=1e-5)
layer_optimizer.append(l_optim)
elif args['model'] == 'layersum':
model = models.LayerSymModel(args)
layer_optimizer = None'''
model = model.to(args['device'])
parameters = filter(lambda p: p.requires_grad, model.parameters())
optimizer = AdamW(parameters, lr=args['learning_rate'])
trainer = trainer_select[args['trainer']]
trainer(args, model, optimizer, train_data, test_data, layer_optimizer)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--plm_name', default='roberta-base', type=str)
parser.add_argument('--tok_name', default='', type=str)
parser.add_argument('--pad_token', default=1, type=int)
parser.add_argument('--plm_type', default='roberta', type=str)
parser.add_argument('--hidden_size', default=768, type=int)
parser.add_argument('--device', default='cuda')
parser.add_argument('--seed', default=1023, type=int)
parser.add_argument('--padding_len', default=128, type=int)
parser.add_argument('--batch_size', default=32, type=int)
parser.add_argument('--dropout', default=0.1, type=float)
parser.add_argument('--learning_rate', default=1.5e-5, type=float)
parser.add_argument('--eval_step', default=500000, type=int)
parser.add_argument('--load_path', default='', type=str)
parser.add_argument('--trainer', default='', type=str)
parser.add_argument('--model', default='multiclass', type=str)
parser.add_argument('--epoch', default=2, type=int)
parser.add_argument('--order', default=0, type=int)
parser.add_argument('--memory_save', default=1150, type=int)
#parser.add_argument('--reg_rate', default=1e-6, type=float)
#parser.add_argument('--start_layer', default=12, type=int)
#parser.add_argument('--end_layer', default=12, type=int)
parser.add_argument('--rep_itv', default=10000, type=int)
parser.add_argument('--rep_num', default=100, type=int)
#parser.add_argument('--cla_num', default=50, type=int)
#parser.add_argument('--prox_rate', default=1e-4, type=float)
#parser.add_argument('--cla_lr', default=5e-5, type=float)
#parser.add_argument('--rep_lr', default=3e-5, type=float)
args = parser.parse_args()
if args.device != 'cuda':
args.device = 'cpu'
args = args.__dict__
if args['tok_name'] == '':
args['tok_name'] = args['plm_name']
now_time = datetime.datetime.now().strftime('%Y-%m-%d-%H-%M-%S')
logger = logging.getLogger()
fh = logging.FileHandler('./logs/TrainModel-%s.log' % now_time, encoding='utf-8')
args['now_time'] = now_time
sh = logging.StreamHandler()
formatter = logging.Formatter('%(asctime)s %(filename)s[line:%(lineno)d] %(levelname)s %(message)s')
fh.setFormatter(formatter)
sh.setFormatter(formatter)
logger.addHandler(fh)
logger.addHandler(sh)
logger.setLevel(10)
logger.info('Configuration:\n' + json.dumps(args, indent=2))
args['logger'] = logger
def setup_seed(seed):
torch.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
np.random.seed(seed)
random.seed(seed)
torch.backends.cudnn.deterministic = True
setup_seed(args['seed'])
main(args, logger)