-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrain.py
190 lines (166 loc) · 8.67 KB
/
train.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
import sys
import time
import os
import argparse
import logging
import numpy as np
import random
from sklearn import metrics
from time import strftime, localtime
import torch
from torch.utils.data import DataLoader
from transformers import BertModel
from model import GCNClassifier
from model_bert import GCNBertClassifier
from data_utils import build_tokenizer, build_embedding_matrix, SentenceDataset, Tokenizer4BertGCN, ABSAGCNData
from prepare_vocab import VocabHelp
from trainer import Trainer
t_start = time.time()
logger = logging.getLogger()
logger.setLevel(logging.INFO)
logger.addHandler(logging.StreamHandler(sys.stdout))
model_classes = {'dagcn': GCNClassifier,
'dagcnbert': GCNBertClassifier}
dataset_files = {
'restaurant': {
'train': './dataset/Restaurants_corenlp/train.json',
'test': './dataset/Restaurants_corenlp/test.json',
},
'laptop': {
'train': './dataset/Laptops_corenlp/train.json',
'test': './dataset/Laptops_corenlp/test.json'
},
'twitter': {
'train': './dataset/Tweets_corenlp/train.json',
'test': './dataset/Tweets_corenlp/test.json',
}
}
input_colses = {
'dagcn': ['text', 'aspect', 'pos', 'head', 'deprel', 'post', 'mask', 'length', 'adj_reshape'],
'dagcnbert': ['text_bert_indices', 'bert_segments_ids', 'attention_mask', 'asp_start', 'asp_end', 'adj_matrix',
'src_mask', 'aspect_mask']
}
initializers = {
'xavier_uniform_': torch.nn.init.xavier_uniform_,
'xavier_normal_': torch.nn.init.xavier_normal_,
}
optimizers = {
'adadelta': torch.optim.Adadelta,
'adagrad': torch.optim.Adagrad,
'adam': torch.optim.Adam,
'adamax': torch.optim.Adamax,
'asgd': torch.optim.ASGD,
'rmsprop': torch.optim.RMSprop,
'sgd': torch.optim.SGD,
}
# Hyperparameters
parser = argparse.ArgumentParser()
parser.add_argument('--model_name', default='dagcn', type=str, help=', '.join(model_classes.keys()))
parser.add_argument('--dataset', default='twitter', type=str, help=', '.join(dataset_files.keys()))
parser.add_argument('--optimizer', default='adam', type=str, help=', '.join(optimizers.keys()))
parser.add_argument('--initializer', default='xavier_uniform_', type=str, help=', '.join(initializers.keys()))
parser.add_argument('--learning_rate', default=0.002, type=float)
parser.add_argument('--l2reg', default=1e-4, type=float)
parser.add_argument('--num_epoch', default=1, type=int)
parser.add_argument('--batch_size', default=16, type=int)
parser.add_argument('--log_step', default=5, type=int)
parser.add_argument('--embed_dim', default=300, type=int)
parser.add_argument('--post_dim', type=int, default=30, help='Position embedding dimension.')
parser.add_argument('--pos_dim', type=int, default=30, help='Pos embedding dimension.')
parser.add_argument('--hidden_dim', type=int, default=50, help='GCN mem dim.')
parser.add_argument('--num_layers', type=int, default=1, help='Num of GCN layers.')
parser.add_argument('--polarities_dim', default=3, type=int, help='3')
parser.add_argument('--input_dropout', type=float, default=0.7, help='Input dropout rate.')
parser.add_argument('--gcn_dropout', type=float, default=0.1, help='GCN layer dropout rate.')
parser.add_argument('--lower', default=True, help='Lowercase all words.')
parser.add_argument('--direct', default=False, help='directed graph or undirected graph')
parser.add_argument('--loop', default=True)
parser.add_argument('--bidirect', default=True, help='Do use bi-RNN layer.')
parser.add_argument('--rnn_hidden', type=int, default=50, help='RNN hidden state size.')
parser.add_argument('--rnn_layers', type=int, default=1, help='Number of RNN layers.')
parser.add_argument('--rnn_dropout', type=float, default=0.1, help='RNN dropout rate.')
parser.add_argument('--attention_heads', default=1, type=int, help='number of multi-attention heads')
parser.add_argument('--max_length', default=85, type=int)
parser.add_argument('--multi_hop', default=False, type=bool)
parser.add_argument('--max_hop', default=4, type=int)
parser.add_argument('--device', default=None, type=str, help='cpu, cuda')
parser.add_argument('--seed', default=1000, type=int)
parser.add_argument("--weight_decay", default=0.0, type=float, help="Weight deay if we apply some.")
parser.add_argument('--vocab_dir', type=str, default='./dataset/Tweets_corenlp')
parser.add_argument('--pad_id', default=0, type=int)
parser.add_argument('--parseadj', default=False, action='store_true', help='dependency probability')
parser.add_argument('--parsehead', default=False, action='store_true', help='dependency tree')
parser.add_argument('--cuda', default='0', type=str)
parser.add_argument('--fusion', default=True, type=bool,
help='fuse distance based weighted matrices belonging to different aspects')
parser.add_argument('--alpha', default=0.8, type=float, help='the weight of distance')
parser.add_argument('--beta', default=0.4, type=float, help='the threshold that whether link aspect with words directly')
parser.add_argument('--gama', default=1.2, type=float, help='the weight of kl divergence loss')
parser.add_argument('--distance_matrix_debug', default=False, type=bool, help='debug mode')
# * bert
parser.add_argument('--pretrained_bert_name', default='bert-base-uncased', type=str)
parser.add_argument("--adam_epsilon", default=1e-8, type=float, help="Epsilon for Adam optimizer.")
parser.add_argument('--bert_dim', type=int, default=768)
parser.add_argument('--bert_dropout', type=float, default=0.3, help='BERT dropout rate.')
parser.add_argument('--diff_lr', default=False, action='store_true')
parser.add_argument('--bert_lr', default=2e-5, type=float)
opt = parser.parse_args()
opt.model_class = model_classes[opt.model_name]
opt.dataset_file = dataset_files[opt.dataset]
opt.inputs_cols = input_colses[opt.model_name]
opt.initializer = initializers[opt.initializer]
opt.optimizer = optimizers[opt.optimizer]
opt.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') if opt.device is None else torch.device(opt.device)
if 'bert' in opt.model_name:
tokenizer = Tokenizer4BertGCN(opt.max_length, opt.pretrained_bert_name)
bert = BertModel.from_pretrained(opt.pretrained_bert_name)
model = opt.model_class(bert, opt).to(opt.device)
trainset = ABSAGCNData(opt.dataset_file['train'], tokenizer, opt=opt)
testset = ABSAGCNData(opt.dataset_file['test'], tokenizer, opt=opt)
else:
logger.info('Building tokenizer...')
tokenizer = build_tokenizer(
fnames=[opt.dataset_file['train'], opt.dataset_file['test']],
max_length=opt.max_length,
data_file='{}/{}_tokenizer.dat'.format(opt.vocab_dir, opt.dataset)
)
logger.info('Building embedding matrix...')
embedding_matrix = build_embedding_matrix(
vocab=tokenizer.vocab,
embed_dim=opt.embed_dim,
data_file='{}/{}d_{}_embedding_matrix.dat'.format(opt.vocab_dir, str(opt.embed_dim), opt.dataset)
)
logger.info("Loading vocab...")
token_vocab = VocabHelp.load_vocab(opt.vocab_dir + '/vocab_tok.vocab') # token
post_vocab = VocabHelp.load_vocab(opt.vocab_dir + '/vocab_post.vocab') # position
pos_vocab = VocabHelp.load_vocab(opt.vocab_dir + '/vocab_pos.vocab') # POS
dep_vocab = VocabHelp.load_vocab(opt.vocab_dir + '/vocab_dep.vocab') # deprel
pol_vocab = VocabHelp.load_vocab(opt.vocab_dir + '/vocab_pol.vocab') # polarity
logger.info("token_vocab: {}, post_vocab: {}, pos_vocab: {}, dep_vocab: {}, pol_vocab: {}".format(
len(token_vocab), len(post_vocab), len(pos_vocab), len(dep_vocab), len(pol_vocab)))
opt.post_size = len(post_vocab)
opt.pos_size = len(pos_vocab)
vocab_help = (post_vocab, pos_vocab, dep_vocab, pol_vocab)
model = opt.model_class(opt, embedding_matrix).to(opt.device)
trainset = SentenceDataset(opt.dataset_file['train'], tokenizer, opt=opt, vocab_help=vocab_help)
testset = SentenceDataset(opt.dataset_file['test'], tokenizer, opt=opt, vocab_help=vocab_help)
train_dataloader = DataLoader(dataset=trainset, batch_size=opt.batch_size, shuffle=True)
test_dataloader = DataLoader(dataset=testset, batch_size=opt.batch_size)
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
# set random seed
setup_seed(opt.seed)
log_dir = './log/'+opt.dataset
if not os.path.exists(log_dir):
os.makedirs(log_dir, mode=0o777)
log_file = '{}.log'.format(strftime("%Y-%m-%d_%H:%M:%S", localtime()))
logger.addHandler(logging.FileHandler("%s/%s" % (log_dir, log_file)))
our_trainer = Trainer(opt, model, train_dataloader, test_dataloader, logger)
our_trainer._print_args()
our_trainer.run()
t_end = time.time()
print('running time is ', round(t_end-t_start), 'secs.')