-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathconfig.py
218 lines (177 loc) · 8.36 KB
/
config.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
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
import numpy as np
import warnings
from pathlib import Path
from torch.utils.data import TensorDataset, DataLoader, RandomSampler, SequentialSampler
from keras.preprocessing.sequence import pad_sequences
from transformers import BertTokenizer, BertForSequenceClassification, BertConfig, \
BertModel # convert pretrain_bert to transformers to add special tokens
class Option(object):
def __init__(self):
super(Option, self).__init__()
self.train_loss_type = "Rank List Loss"
self.testset_loss_type = 'none'
# self.data_type = 'trex'
# self.train_data_file = './data/datasets/Trex/train-spo.json'
# self.val_data_file = './data/datasets/Trex/dev-spo.json'
# self.test_data_file = './data/datasets/Trex/test-spo.json'
# self.wordvec_file = "./data/wordvec/word_vec.json"
# self.rel2id_file = './data/datasets/Trex/rel2id.json'
self.gpu = "0"
self.dataset = "nyt"
self.data_type = 'nyt_fb'
self.train_data_file = '/home/wyt/lrz/data/datasets/nyt_fb/nyt_ori/nyt_all_train.json'
self.val_data_file = '/home/wyt/lrz/data/datasets/nyt_fb/nyt_ori/nyt_all_dev.json'
self.test_data_file = '/home/wyt/lrz/data/datasets/nyt_fb/nyt_ori/nyt_all_test.json'
self.wordvec_file = "/home/wyt/lrz/data/wordvec/word_vec.json"
self.rel2id_file = '/home/wyt/lrz/data/datasets/nyt_fb/nyt_ori/rel2id_cnn.json'
# self.data_type = 'fewrel'
# self.train_data_file = './data/datasets/fewrel_ori/fewrel80_train.json'
# self.val_data_file = './data/datasets/fewrel_ori/fewrel80_test_train.json'
# self.test_data_file = './data/datasets/fewrel_ori/fewrel80_test_test.json'
# self.wordvec_file = "./data/wordvec/word_vec.json"
# self.rel2id_file = './data/support_files/rel2id.json'
self.similarity_file = None # './data/support_files/trainset_similarity.pkl'
self.same_level_pair_file = None # './data/support_files/same_level_pair.json'
self.gt_hierarchy_file = None # './data/support_files/all_structure.json'
self.load_model_name = None # 'MORE_best.pt' , the file_name (under checkpoints)
self.save_model_name = 'MORE'
self.save_dir = str(Path().absolute())
self.lr = [0.0003]
self.lr_chose = 0
self.batch_size = [60, 60, 60, 60]
self.class_num_ratio = [0.5, 0.5, 0.5, 0.5]
# self.batch_size = [240, 240, 240, 240]
# self.class_num_ratio = [0.1, 0.1, 0.1, 0.1]
self.batch_num = [2000, 2000, 2000, 2000]
self.epoch_num = 4
self.squared = False
self.alpha = None
self.lr_decay_num = 0 # real decay successfully num
self.lr_decay_record = [] # global step when it go to decay
self.batch_shuffle = False
self.seed = 42
self.early_stop = None # None，if no early stop
self.whether_visualize = True
self.record_test = True
self.embedding_dim = 64
self.max_len = 120
self.pos_emb_dim = 5
self.drop_out = 0.2
self.same_level_part = 200
self.mask_same_level_epoch = 100
self.random_init = False
self.print_losses = False
self.select_cluster = 'K-means'
self.eval_num = 5
self.K_num = 16
self.band = 0.7784975910442384
self.margin = 0.2
self.margin_un = 0.4
self.alpha_rank = 1.0
self.alpha_rank_un = 1.2
self.temp_neg = 10
self.temp_pos = 0
self.landa = 0.5
self.VAT = 0.05
self.p_mult = 0.02
self.power_iterations = 1
self.warm_up = 0
self.lambda_V = 1
self.inclass_augment = False
self.uninfor_landa = 0.0
self.uninfor_drop = 0.0
self.BERT = False
def parse(self, kwargs):
for k, v in kwargs.items():
if not hasattr(self, k):
warnings.warn("Warning: opt has no attribute %s" % k)
setattr(self, k.v)
print("user config:")
for k, v in self.__class__.__dict__.items():
if not k.startswith("__"):
print(k, getattr(self, k))
class Bert_option(Option):
def __init__(self, gpu):
super(Bert_option, self).__init__()
import os
os.environ["CUDA_VISIBLE_DEVICES"] = gpu
import torch
import torch.nn as nn
if torch.cuda.is_available():
torch.cuda.set_device(int(gpu))
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
self.BERT = True
print(">>>Note that you chose to use BERT for extractor!")
self.tokenizer = BertTokenizer.from_pretrained('bert-base-uncased',
do_lower_case=True,
never_split=(
"[UNK]", "[SEP]", "[PAD]", "[CLS]", "[MASK]", "<e1>",
"</e1>",
"<e2>", "</e2>"))
# using pretrained bert model
self.model = BertForSequenceClassification.from_pretrained("bert-base-uncased")
self.Bert_model = nn.Sequential(*list(self.model.children())[:-2])
# add token to vocabulary =======================================================
special_tokens_dict = {'additional_special_tokens': ['<e1>', '</e1>', '<e2>', '</e2>']}
num_added_toks = self.tokenizer.add_special_tokens(special_tokens_dict)
assert num_added_toks == 4
self.model.resize_token_embeddings(len(self.tokenizer))
e1_id = self.tokenizer.convert_tokens_to_ids('<e1>')
e2_id = self.tokenizer.convert_tokens_to_ids('<e2>')
print("e1:", e1_id, " e2:", e2_id)
assert e1_id != e2_id != 1
def count_parameters(model):
return sum(p.numel() for p in model.parameters() if p.requires_grad)
count = count_parameters(self.Bert_model)
print("model_num_parameters:", count)
self.bert_batch_size = 16
self.word_vec_mat = np.random.randn(114042, 50)
# have some special training methods
self.lr = [1e-5]
self.lr_linear = 1e-4
self.lr_chose = 0
# self.batch_size = [60,60,60,60]
# self.class_num_ratio = [0.5, 0.5, 0.5,0.5]
self.batch_size = [40, 40, 40, 40]
self.class_num_ratio = [0.5, 0.5, 0.5, 0.5]
self.batch_num = [1000, 1000, 1000, 1000]
def batch_dataset(self, inputs, indexs, labels, batch_size):
data = TensorDataset(inputs, indexs, labels)
# sampler = RandomSampler(data)
sampler = SequentialSampler(data)
train_dataloader = DataLoader(data, sampler=sampler, batch_size=batch_size)
return train_dataloader
def pre_processing(self, sentence, sentence_label):
tokenized_texts = [self.tokenizer.tokenize(sent) for sent in sentence] # tokenization
# reset the sentence index =========================================================
sen_index = []
for sen in tokenized_texts:
e1_begin = sen.index('<e1>')
e1_end = sen.index('</e1>')
e2_begin = sen.index('<e2>')
e2_end = sen.index('</e2>')
sen_index.append([e1_begin, e2_begin])
try:
assert len(sen_index) == len(tokenized_texts)
sentence_index = sen_index
except:
raise Exception("error!")
MAX_LEN = 128
# Use the BERT tokenizer to convert the tokens to their index numbers in the BERT vocabulary
input_ids = [self.tokenizer.convert_tokens_to_ids(x) for x in tokenized_texts]
# Pad our input tokens
input_ids = pad_sequences(input_ids, maxlen=MAX_LEN, dtype="long", truncating="post", padding="post")
attention_masks = []
# Create a mask of 1s for each token followed by 0s for padding
for seq in input_ids:
seq_mask = [float(i > 0) for i in seq]
attention_masks.append(seq_mask)
# Convert all of our data into torch tensors, the required datatype for our model
import torch
inputs = torch.tensor(input_ids)
index = torch.tensor(sentence_index)
labels = torch.tensor(sentence_label)
masks = torch.tensor(attention_masks)
return inputs, index, labels
if __name__ == '__main__':
Bert_option()