-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathdata_loader.py
More file actions
274 lines (242 loc) · 8.89 KB
/
data_loader.py
File metadata and controls
274 lines (242 loc) · 8.89 KB
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
# data_loader.py
# Functions for loading and preprocessing datasets
from datasets import load_dataset
from transformers import BertTokenizer
from torch.utils.data import random_split
def load_glue_dataset(config):
"""
Load datasets from the GLUE benchmark.
Args:
config: Configuration object
Returns:
train_dataset, eval_dataset, test_dataset
"""
# Load dataset using the Hugging Face datasets library
if config.task_name == "mnli":
datasets = load_dataset("glue", "mnli")
# For MNLI, there are two validation sets: matched and mismatched
train_dataset = datasets["train"]
eval_dataset = datasets["validation_matched"]
test_dataset = datasets["validation_mismatched"]
else:
datasets = load_dataset("glue", config.task_name)
train_dataset = datasets["train"]
if "validation" in datasets:
eval_dataset = datasets["validation"]
test_dataset = datasets["test"] if "test" in datasets else datasets["validation"]
else:
# Split the training set if no validation set is available
train_size = int(config.train_test_split_ratio * len(train_dataset))
eval_size = len(train_dataset) - train_size
train_dataset, eval_dataset = random_split(
train_dataset,
[train_size, eval_size]
)
test_dataset = eval_dataset
# Load tokenizer
tokenizer = BertTokenizer.from_pretrained(config.bert_model)
# Define preprocessing function
def tokenize_function(examples):
if config.task_name == "qqp":
# For sentence pair tasks
return tokenizer(
examples["question1"],
examples["question2"],
padding="max_length",
truncation=True,
max_length=config.max_seq_length
)
elif config.task_name == "mnli":
# For MNLI sentence pair task
return tokenizer(
examples["premise"],
examples["hypothesis"],
padding="max_length",
truncation=True,
max_length=config.max_seq_length
)
else:
# For single sentence tasks
text_key = "sentence" if config.task_name == "cola" else "sentence"
return tokenizer(
examples[text_key],
padding="max_length",
truncation=True,
max_length=config.max_seq_length
)
# Apply tokenization
train_dataset = train_dataset.map(
tokenize_function,
batched=True,
desc="Tokenizing training dataset"
)
eval_dataset = eval_dataset.map(
tokenize_function,
batched=True,
desc="Tokenizing evaluation dataset"
)
test_dataset = test_dataset.map(
tokenize_function,
batched=True,
desc="Tokenizing test dataset"
)
# Set format for PyTorch
train_dataset.set_format(
type="torch",
columns=["input_ids", "attention_mask", "token_type_ids", "label"]
)
eval_dataset.set_format(
type="torch",
columns=["input_ids", "attention_mask", "token_type_ids", "label"]
)
test_dataset.set_format(
type="torch",
columns=["input_ids", "attention_mask", "token_type_ids", "label"]
)
return train_dataset, eval_dataset, test_dataset
def load_conll_dataset(config):
"""
Load the CoNLL-2003 NER dataset.
Args:
config: Configuration object
Returns:
train_dataset, eval_dataset, test_dataset
"""
# Load dataset using the Hugging Face datasets library
datasets = load_dataset("conll2003")
train_dataset = datasets["train"]
eval_dataset = datasets["validation"]
test_dataset = datasets["test"]
# Load tokenizer
tokenizer = BertTokenizer.from_pretrained(config.bert_model)
# Define preprocessing function
def tokenize_and_align_labels(examples):
tokenized_inputs = tokenizer(
examples["tokens"],
truncation=True,
is_split_into_words=True,
padding="max_length",
max_length=config.max_seq_length
)
labels = []
for i, label in enumerate(examples["ner_tags"]):
word_ids = tokenized_inputs.word_ids(batch_index=i)
previous_word_idx = None
label_ids = []
for word_idx in word_ids:
# Special tokens have a word id that is None
if word_idx is None:
label_ids.append(-100)
# We set the label for the first token of each word
elif word_idx != previous_word_idx:
label_ids.append(label[word_idx])
# For the other tokens in a word, we set the label to -100
else:
label_ids.append(-100)
previous_word_idx = word_idx
labels.append(label_ids)
tokenized_inputs["labels"] = labels
return tokenized_inputs
# Apply tokenization
train_dataset = train_dataset.map(
tokenize_and_align_labels,
batched=True,
desc="Tokenizing and aligning training dataset"
)
eval_dataset = eval_dataset.map(
tokenize_and_align_labels,
batched=True,
desc="Tokenizing and aligning evaluation dataset"
)
test_dataset = test_dataset.map(
tokenize_and_align_labels,
batched=True,
desc="Tokenizing and aligning test dataset"
)
# Set format for PyTorch
train_dataset.set_format(
type="torch",
columns=["input_ids", "attention_mask", "token_type_ids", "labels"]
)
eval_dataset.set_format(
type="torch",
columns=["input_ids", "attention_mask", "token_type_ids", "labels"]
)
test_dataset.set_format(
type="torch",
columns=["input_ids", "attention_mask", "token_type_ids", "labels"]
)
return train_dataset, eval_dataset, test_dataset
def create_public_dataset(config, size=10000):
"""
Create a public dataset for distillation.
Args:
config: Configuration object
size: Size of the public dataset
Returns:
Public dataset
"""
# For real implementation, this could be a separate unlabeled dataset
# For simplicity, we use a subset of the train dataset without labels
if config.dataset_name == "glue":
dataset = load_dataset("glue", config.task_name)
public_dataset = dataset["train"].select(range(min(size, len(dataset["train"]))))
elif config.dataset_name == "conll2003":
dataset = load_dataset("conll2003")
public_dataset = dataset["train"].select(range(min(size, len(dataset["train"]))))
else:
raise ValueError(f"Unsupported dataset: {config.dataset_name}")
# Load tokenizer
tokenizer = BertTokenizer.from_pretrained(config.bert_model)
# Tokenize the public dataset
if config.dataset_name == "glue":
if config.task_name == "qqp":
# For sentence pair tasks
def tokenize_function(examples):
return tokenizer(
examples["question1"],
examples["question2"],
padding="max_length",
truncation=True,
max_length=config.max_seq_length
)
elif config.task_name == "mnli":
def tokenize_function(examples):
return tokenizer(
examples["premise"],
examples["hypothesis"],
padding="max_length",
truncation=True,
max_length=config.max_seq_length
)
else:
# For single sentence tasks
text_key = "sentence" if config.task_name == "cola" else "sentence"
def tokenize_function(examples):
return tokenizer(
examples[text_key],
padding="max_length",
truncation=True,
max_length=config.max_seq_length
)
else: # CoNLL
def tokenize_function(examples):
return tokenizer(
examples["tokens"],
truncation=True,
is_split_into_words=True,
padding="max_length",
max_length=config.max_seq_length
)
# Apply tokenization
public_dataset = public_dataset.map(
tokenize_function,
batched=True,
desc="Tokenizing public dataset"
)
# Set format for PyTorch (input features only, no labels)
public_dataset.set_format(
type="torch",
columns=["input_ids", "attention_mask", "token_type_ids"]
)
return public_dataset