/
data.py
226 lines (188 loc) · 8.62 KB
/
data.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
219
220
221
222
223
224
225
226
import configparser
import os
import re
import string
import pickle
import copy
import random
import numpy as np
import torch
from torch.utils.data import DataLoader
from fastNLP import Vocabulary
from dataset import Dataset
from dataloader import TrainDataLoader
from utils import padding, batch_padding
def _parse_list(data_path, list_name):
domain = set()
with open(os.path.join(data_path, list_name), 'r', encoding='utf-8') as f:
for line in f:
domain.add(line.strip('\n'))
return domain
def get_domains(data_path, filtered_name, target_name):
all_domains = _parse_list(data_path, filtered_name)
test_domains = _parse_list(data_path, target_name)
train_domains = all_domains - test_domains
print('train domains', len(train_domains), 'test_domains', len(test_domains))
return sorted(list(train_domains)), sorted(list(test_domains))
def _parse_data(data_path, filename):
neg = {
'filename': filename,
'data': [],
'target': []
}
pos = {
'filename': filename,
'data': [],
'target': []
}
with open(os.path.join(data_path, filename), 'r', encoding='utf-8') as f:
for line in f:
line = line.strip('\n')
if line[-2:] == '-1':
neg['data'].append(line[:-2])
neg['target'].append(0)
else:
pos['data'].append(line[:-1])
pos['target'].append(1)
# check
print(filename, 'neg', len(neg['data']), 'pos', len(pos['data']))
return neg, pos
def _process_data(data_dict):
for i in range(len(data_dict['data'])):
text = data_dict['data'][i]
# ignore string.punctuation
text = re.sub('[%s]' % re.escape(string.punctuation), ' ', text)
# string.whitespace -> space
text = re.sub('[%s]' % re.escape(string.whitespace), ' ', text)
# lower case
text = text.lower()
# split by whitespace
text = text.split()
# replace
data_dict['data'][i] = text
return data_dict
def _get_data(data_path, domains, usage):
# usage in ['train', 'dev', 'test']
data = {}
for domain in domains:
for t in ['t2', 't4', 't5']:
filename = '.'.join([domain, t, usage])
neg, pos = _parse_data(data_path, filename)
neg = _process_data(neg)
pos = _process_data(pos)
data[filename] = {'neg': neg, 'pos': pos}
return data
def get_train_data(data_path, domains):
train_data = _get_data(data_path, domains, 'train')
print('train data', len(train_data))
return train_data
def _combine_data(support_data, data):
# support -> dev, test
for key in data:
key_split = key.split('.')[0:-1] + ['train']
support_key = '.'.join(key_split)
for value in data[key]:
data[key][value]['support_data'] = copy.deepcopy(support_data[support_key][value]['data'])
data[key][value]['support_target'] = copy.deepcopy(support_data[support_key][value]['target'])
return data
def get_test_data(data_path, domains):
# get dev, test data
support_data = _get_data(data_path, domains, 'train')
dev_data = _get_data(data_path, domains, 'dev')
test_data = _get_data(data_path, domains, 'test')
# support -> dev, test
dev_data = _combine_data(support_data, dev_data)
test_data = _combine_data(support_data, test_data)
print('dev data', len(dev_data), 'test data', len(test_data))
return dev_data, test_data
def get_vocabulary(data, min_freq):
# train data -> vocabulary
vocabulary = Vocabulary(min_freq=min_freq, padding='<pad>', unknown='<unk>')
for filename in data:
for value in data[filename]:
for word_list in data[filename][value]['data']:
vocabulary.add_word_lst(word_list)
vocabulary.build_vocab()
print('vocab size', len(vocabulary), 'pad', vocabulary.padding_idx, 'unk', vocabulary.unknown_idx)
return vocabulary
def _idx_text(text_list, vocabulary):
for i in range(len(text_list)):
for j in range(len(text_list[i])):
text_list[i][j] = vocabulary.to_index(text_list[i][j])
return text_list
def idx_all_data(data, vocabulary):
for filename in data:
for value in data[filename]:
for key in data[filename][value]:
if key in ['data', 'support_data']:
data[filename][value][key] = _idx_text(data[filename][value][key], vocabulary)
return data
def get_train_loader(train_data, support, query, pad_idx):
batch_size = support + query
train_loaders = {}
for filename in train_data:
neg_dl = DataLoader(Dataset(train_data[filename]['neg'], pad_idx), batch_size=batch_size, shuffle=True, drop_last=False, **kwargs)
pos_dl = DataLoader(Dataset(train_data[filename]['pos'], pad_idx), batch_size=batch_size, shuffle=True, drop_last=False, **kwargs)
if min(len(neg_dl), len(pos_dl)) > 0:
train_loaders[filename] = {
'neg': neg_dl,
'pos': pos_dl
}
print('train loaders', len(train_loaders))
return TrainDataLoader(train_loaders, support=support, query=query, pad_idx=pad_idx)
def get_test_loader(full_data, support, query, pad_idx):
loader = []
for filename in full_data:
# support
support_data = full_data[filename]['neg']['support_data'][0:support] + full_data[filename]['pos']['support_data'][0:support]
support_data = batch_padding(support_data, pad_idx)
support_target = full_data[filename]['neg']['support_target'][0:support] + full_data[filename]['pos']['support_target'][0:support]
support_target = torch.tensor(support_target)
# query
neg_dl = DataLoader(Dataset(full_data[filename]['neg'], pad_idx), batch_size=query * 2, shuffle=False, drop_last=False, **kwargs)
pos_dl = DataLoader(Dataset(full_data[filename]['pos'], pad_idx), batch_size=query * 2, shuffle=False, drop_last=False, **kwargs)
# combine
for dl in [neg_dl, pos_dl]:
for batch_data, batch_target in dl:
support_data_cp, support_target_cp = copy.deepcopy(support_data), copy.deepcopy(support_target)
support_data_cp, batch_data = padding(support_data_cp, batch_data, pad_idx)
data = torch.cat([support_data_cp, batch_data], dim=0)
target = torch.cat([support_target_cp, batch_target], dim=0)
loader.append((data, target))
print('test loader length', len(loader))
return loader
def main():
train_domains, test_domains = get_domains(data_path, config['data']['filtered_list'], config['data']['target_list'])
train_data = get_train_data(data_path, train_domains)
dev_data, test_data = get_test_data(data_path, test_domains)
# print(dev_data['books.t2.dev']['neg']['support_data'])
# print(dev_data['books.t2.dev']['neg']['support_target'])
vocabulary = get_vocabulary(train_data, min_freq=int(config['data']['min_freq']))
pad_idx = vocabulary.padding_idx
pickle.dump(vocabulary, open(os.path.join(config['data']['path'], config['data']['vocabulary']), 'wb'))
train_data = idx_all_data(train_data, vocabulary)
dev_data = idx_all_data(dev_data, vocabulary)
test_data = idx_all_data(test_data, vocabulary)
# print(dev_data['books.t2.dev']['neg']['support_data'])
# print(dev_data['books.t2.dev']['neg']['support_target'])
support = int(config['model']['support'])
query = int(config['model']['query'])
train_loader = get_train_loader(train_data, support, query, pad_idx)
dev_loader = get_test_loader(dev_data, support, query, pad_idx)
test_loader = get_test_loader(test_data, support, query, pad_idx)
pickle.dump(train_loader, open(os.path.join(config['data']['path'], config['data']['train_loader']), 'wb'))
pickle.dump(dev_loader, open(os.path.join(config['data']['path'], config['data']['dev_loader']), 'wb'))
pickle.dump(test_loader, open(os.path.join(config['data']['path'], config['data']['test_loader']), 'wb'))
if __name__ == "__main__":
# config
config = configparser.ConfigParser()
config.read("config.ini")
# seed
seed = int(config['data']['seed'])
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
kwargs = {'num_workers': 4, 'pin_memory': True} if torch.cuda.is_available() else {}
data_path = config['data']['path']
main()