-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdataset.py
289 lines (217 loc) · 10.4 KB
/
dataset.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
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
import os
import numpy as np
import pandas as pd
import torch
from torch.utils.data import Dataset, DataLoader
import pytorch_lightning as pl
# Dataset Builders -----------------------------------------------------------------------------------------------------
class NERDataset(Dataset):
def __init__(self, bio_data, tokenizer, label_vocab, max_seq_len):
self.tokenizer = tokenizer
self.label_vocab = label_vocab
self.max_seq_len = max_seq_len
# save all case ids
# self.case_ids = case_ids
# for debugging -- to smallset
# N = 70
# bio_data = bio_data[:N]
# get padding token label id
from torch.nn import CrossEntropyLoss
pad_token_label_id = CrossEntropyLoss().ignore_index
# transform all data
from utils.ner_collate import convert_examples_to_features_for_ner
self.features = convert_examples_to_features_for_ner(bio_data,
label_vocab,
max_seq_len,
tokenizer,
pad_token_label_id=pad_token_label_id)
def __getitem__(self, i):
feature = self.features[i]
# case_ids = self.case_ids[i]
# Convert to Tensors and build dataset
input_ids = torch.tensor(feature.input_ids, dtype=torch.long)
attention_mask = torch.tensor(feature.attention_mask, dtype=torch.long)
token_type_ids = torch.tensor(feature.token_type_ids, dtype=torch.long)
label_ids = torch.tensor(feature.label_ids, dtype=torch.long)
item = [input_ids, token_type_ids, attention_mask, label_ids]
return item
def __len__(self):
return (len(self.features))
class ClassificationDataset(Dataset):
def __init__(self, df, tokenizer, label_vocab, max_seq_len):
self.tokenizer = tokenizer
self.label_vocab = label_vocab
self.max_seq_len = max_seq_len
# for debugging -- to smallset
# N = 70
# df = df[:N]
# transform all data
from tqdm.auto import tqdm
df_iterator = tqdm(df.iterrows(), desc="Iteration")
self.texts = []
self.labels = []
for row_idx, (index, row) in enumerate(df_iterator):
text = row[1]
label_text = row[2]
text_obj = tokenizer(text, padding='max_length', max_length=self.max_seq_len, truncation=True)
label_id = self.label_vocab[label_text]
self.texts.append(text_obj)
self.labels.append(label_id)
def __getitem__(self, i):
input_ids = np.array(self.texts[i]['input_ids'])
token_type_ids = np.array(self.texts[i]['token_type_ids'])
attention_mask = np.array(self.texts[i]['attention_mask'])
label_ids = np.array(self.labels[i])
item = [input_ids, token_type_ids, attention_mask, label_ids]
return item
def __len__(self):
return (len(self.texts))
class DocumentClassificationDataset(Dataset):
def __init__(self, df, tokenizer, label_vocab, max_seq_len):
self.tokenizer = tokenizer
self.label_vocab = label_vocab
self.max_seq_len = max_seq_len
# for debugging -- to smallset
# N = 70
# df = df[:N]
# transform all data
from tqdm.auto import tqdm
df_iterator = tqdm(df.iterrows(), desc="Iteration")
self.texts = []
self.labels = []
for row_idx, (index, row) in enumerate(df_iterator):
text = row[1]
label_text = row[2]
text_obj = tokenizer(text, padding='max_length', max_length=self.max_seq_len, truncation=True)
label_id = self.label_vocab[label_text]
self.texts.append(text_obj)
self.labels.append(label_id)
def __getitem__(self, i):
input_ids = np.array(self.texts[i]['input_ids'])
token_type_ids = np.array(self.texts[i]['token_type_ids'])
attention_mask = np.array(self.texts[i]['attention_mask'])
label_ids = np.array(self.labels[i])
item = [input_ids, token_type_ids, attention_mask, label_ids]
return item
def __len__(self):
return (len(self.texts))
# ----------------------------------------------------------------------------------------------------------------------
# Data Modules ---------------------------------------------------------------------------------------------------------
class NER_Data_Module(pl.LightningDataModule):
def __init__(self, task, text_reader, max_seq_length, batch_size):
super().__init__()
self.task = task
# prepare tokenizer
from utils.readers import get_tokenizer
self.tokenizer = get_tokenizer(text_reader)
# data preparing params
self.data_dir = os.path.join("./data", self.task, "run")
self.max_seq_length = max_seq_length
self.batch_size = batch_size
# number of labels for determining model's last dimension
self.label_vocab = None
self.num_labels = None
def prepare_data(self):
# vocab
label_vocab = self._load_vocab(os.path.join(self.data_dir, "label.vocab"))
self.label_vocab = label_vocab
self.num_labels = len(label_vocab)
# read data
train_bio_data = self._read_txt(os.path.join(self.data_dir, "train.bio.txt"), sentence_splitter="----")
valid_bio_data = self._read_txt(os.path.join(self.data_dir, "dev.bio.txt"), sentence_splitter="----")
test_bio_data = self._read_txt(os.path.join(self.data_dir, "dev.bio.txt"), sentence_splitter="----")
# building dataset
dataset = task_to_dataset[self.task]
self.train_dataset = dataset(train_bio_data, self.tokenizer, label_vocab, self.max_seq_length)
self.valid_dataset = dataset(valid_bio_data, self.tokenizer, label_vocab, self.max_seq_length)
self.test_dataset = dataset(test_bio_data, self.tokenizer, label_vocab, self.max_seq_length)
def _load_vocab(self, fn):
print("Vocab loading from {}".format(fn))
vocab = {}
with open(fn, 'r', encoding='utf-8') as f:
for line in f:
line = line.rstrip()
symbol, _id = line.split('\t')
vocab[symbol] = int(_id)
return vocab
def _read_txt(self, fn, sentence_splitter=None):
case_ids = []
data = []
with open(fn, 'r', encoding='utf-8') as f:
sentence = ''
labels = []
for line in f:
if ("\t" not in list(line)) and (line.lstrip().rstrip() != sentence_splitter):
case_ids.append(int(line.lstrip().rstrip()))
continue
if line.lstrip().rstrip() == sentence_splitter:
data.append((sentence, labels))
sentence = ''
labels = []
continue
sentence = sentence + line.split('\t')[0]
labels.append(line.split('\t')[1].lstrip().rstrip())
print("[Text] data is loaded from {} -- {}".format(fn, len(data)))
return data
def train_dataloader(self):
return DataLoader(self.train_dataset, batch_size=self.batch_size, shuffle=True)
def val_dataloader(self):
return DataLoader(self.valid_dataset, batch_size=self.batch_size)
def test_dataloader(self):
return DataLoader(self.test_dataset, batch_size=self.batch_size)
class Classification_Data_Module(pl.LightningDataModule):
def __init__(self, task, text_reader, max_seq_length, batch_size):
super().__init__()
self.task = task
# prepare tokenizer
from utils.readers import get_tokenizer
self.tokenizer = get_tokenizer(text_reader)
# data preparing params
self.data_dir = os.path.join("./data", self.task, "run")
self.max_seq_length = max_seq_length
self.batch_size = batch_size
# number of labels for determining model's last dimension
self.label_vocab = None
self.num_labels = None
def prepare_data(self):
# vocab
label_vocab = self._load_vocab(os.path.join(self.data_dir, "label.vocab"))
self.label_vocab = label_vocab
self.num_labels = len(label_vocab)
# read data
train_df = pd.read_csv(os.path.join(self.data_dir, "train.tsv"), sep='\t', header=None)
valid_df = pd.read_csv(os.path.join(self.data_dir, "dev.tsv"), sep='\t', header=None)
test_df = pd.read_csv(os.path.join(self.data_dir, "dev.tsv"), sep='\t', header=None)
# building dataset
dataset = task_to_dataset[self.task]
self.train_dataset = dataset(train_df, self.tokenizer, label_vocab, self.max_seq_length)
self.valid_dataset = dataset(valid_df, self.tokenizer, label_vocab, self.max_seq_length)
self.test_dataset = dataset(test_df, self.tokenizer, label_vocab, self.max_seq_length)
def _load_vocab(self, fn):
print("Vocab loading from {}".format(fn))
vocab = {}
with open(fn, 'r', encoding='utf-8') as f:
for line in f:
line = line.rstrip()
symbol, _id = line.split('\t')
vocab[symbol] = int(_id)
return vocab
def train_dataloader(self):
return DataLoader(self.train_dataset, batch_size=self.batch_size, shuffle=True)
def val_dataloader(self):
return DataLoader(self.valid_dataset, batch_size=self.batch_size)
def test_dataloader(self):
return DataLoader(self.test_dataset, batch_size=self.batch_size)
#-----------------------------------------------------------------------------------------------------------------------
# Map For Converting Task Name to Specific Dataset Builder -------------------------------------------------------------
# If you wanna add new task,
# (1) You should write dataset builder class.
# (2) Next, As follows, You should add task name(key) and dataset builder class name(value).
# Optionally, If you wanna add task of another types such as Sequence Labeling and Question & Answering,
# You should implement another LightningDataModules.
task_to_dataset = {
"csie" : NERDataset,
"csii" : ClassificationDataset,
"doc_classification" : DocumentClassificationDataset
}
#-----------------------------------------------------------------------------------------------------------------------