-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathdataset.py
More file actions
427 lines (359 loc) · 23.8 KB
/
dataset.py
File metadata and controls
427 lines (359 loc) · 23.8 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
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
from datasets import load_dataset, concatenate_datasets, ClassLabel, Dataset
from transformers import set_seed
import logging
import utils
"""
Divide the model training set like a MHAT algorithm out of the same public
"""
class DatasetPartition:
def __init__(self, args):
self.args = args
self.private_datasets = []
self.validation_datasets = []
self.test_datasets = []
self.public_datasets = []
self.merged_val_datasets = []
self.merged_test_datasets = []
# No partitioning of the public dataset from the training dataset for the centralized or centralized_mixed algorithm
self.centralized_train_datasets = []
# mixed training and validation dataset for the centralized_mixed algorithm
self.mix_train_datasets = Dataset.from_dict({})
self.mix_validation_datasets = Dataset.from_dict({})
self.mix_test_datasets = Dataset.from_dict({})
# ALL label statistics are in three categories (0:neural 1:positive 2:negative)
# All datasets modified to two columns（label and text）for easy splicing public_dataset
if self.args.algorithm not in ['centralized_mixed']:
for dataset in args.datasets:
raw_datasets = load_dataset(args.data_dir + '/{}'.format(dataset))
raw_datasets = raw_datasets.remove_columns(['idx'])
if dataset == 'automotive':
# convert label value in datasets
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
automotive_train_dataset = raw_datasets['train']
automotive_val_dataset = raw_datasets['validation']
automotive_test_dataset = raw_datasets['test']
# split public dataset
automotive_public_private_dataset = automotive_train_dataset.train_test_split(test_size=args.public_ratio, seed=args.seed, stratify_by_column='label')
automotive_private_dataset = automotive_public_private_dataset['train']
automotive_public_dataset = automotive_public_private_dataset['test']
del automotive_public_private_dataset
self.public_datasets.append(automotive_public_dataset)
self.private_datasets.append(automotive_private_dataset)
self.validation_datasets.append(automotive_val_dataset)
self.test_datasets.append(automotive_test_dataset)
elif dataset == 'baby':
# convert label value in datasets
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
baby_train_dataset = raw_datasets['train']
baby_val_dataset = raw_datasets['validation']
baby_test_dataset = raw_datasets['test']
# split public dataset
baby_public_private_dataset = baby_train_dataset.train_test_split(test_size=args.public_ratio, seed=args.seed, stratify_by_column='label')
baby_private_dataset = baby_public_private_dataset['train']
baby_public_dataset = baby_public_private_dataset['test']
del baby_public_private_dataset
self.public_datasets.append(baby_public_dataset)
self.private_datasets.append(baby_private_dataset)
self.validation_datasets.append(baby_val_dataset)
self.test_datasets.append(baby_test_dataset)
elif dataset == 'clothing':
# convert label value in datasets
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
clothing_train_dataset = raw_datasets['train']
clothing_val_dataset = raw_datasets['validation']
clothing_test_dataset = raw_datasets['test']
# split public dataset
clothing_public_private_dataset = clothing_train_dataset.train_test_split(test_size=args.public_ratio, seed=args.seed, stratify_by_column='label')
clothing_private_dataset = clothing_public_private_dataset['train']
clothing_public_dataset = clothing_public_private_dataset['test']
del clothing_public_private_dataset
self.public_datasets.append(clothing_public_dataset)
self.private_datasets.append(clothing_private_dataset)
self.validation_datasets.append(clothing_val_dataset)
self.test_datasets.append(clothing_test_dataset)
elif dataset == 'health':
# convert label value in datasets
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
health_train_dataset = raw_datasets['train']
health_val_dataset = raw_datasets['validation']
health_test_dataset = raw_datasets['test']
# split public dataset
health_public_private_dataset = health_train_dataset.train_test_split(test_size=args.public_ratio, seed=args.seed, stratify_by_column='label')
health_private_dataset = health_public_private_dataset['train']
health_public_dataset = health_public_private_dataset['test']
del health_public_private_dataset
self.public_datasets.append(health_public_dataset)
self.private_datasets.append(health_private_dataset)
self.validation_datasets.append(health_val_dataset)
self.test_datasets.append(health_test_dataset)
elif dataset == 'sport':
# convert label value in datasets
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
sport_train_dataset = raw_datasets['train']
sport_val_dataset = raw_datasets['validation']
sport_test_dataset = raw_datasets['test']
# split public dataset
sport_public_private_dataset = sport_train_dataset.train_test_split(test_size=args.public_ratio, seed=args.seed, stratify_by_column='label')
sport_private_dataset = sport_public_private_dataset['train']
sport_public_dataset = sport_public_private_dataset['test']
del sport_public_private_dataset
self.public_datasets.append(sport_public_dataset)
self.private_datasets.append(sport_private_dataset)
self.validation_datasets.append(sport_val_dataset)
self.test_datasets.append(sport_test_dataset)
elif dataset == 'beauty':
# convert label value in datasets
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
train_dataset = raw_datasets['train']
val_dataset = raw_datasets['validation']
test_dataset = raw_datasets['test']
# split public dataset
public_private_dataset = train_dataset.train_test_split(test_size=args.public_ratio, seed=args.seed, stratify_by_column='label')
private_dataset = public_private_dataset['train']
pub_dataset = public_private_dataset['test']
del public_private_dataset
self.public_datasets.append(pub_dataset)
self.private_datasets.append(private_dataset)
self.validation_datasets.append(val_dataset)
self.test_datasets.append(test_dataset)
elif dataset == 'patio':
# convert label value in datasets
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
train_dataset = raw_datasets['train']
val_dataset = raw_datasets['validation']
test_dataset = raw_datasets['test']
# split public dataset
public_private_dataset = train_dataset.train_test_split(test_size=args.public_ratio, seed=args.seed, stratify_by_column='label')
private_dataset = public_private_dataset['train']
pub_dataset = public_private_dataset['test']
del public_private_dataset
self.public_datasets.append(pub_dataset)
self.private_datasets.append(private_dataset)
self.validation_datasets.append(val_dataset)
self.test_datasets.append(test_dataset)
elif dataset == 'pet':
# convert label value in datasets
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
train_dataset = raw_datasets['train']
val_dataset = raw_datasets['validation']
test_dataset = raw_datasets['test']
# split public dataset
public_private_dataset = train_dataset.train_test_split(test_size=args.public_ratio, seed=args.seed, stratify_by_column='label')
private_dataset = public_private_dataset['train']
pub_dataset = public_private_dataset['test']
del public_private_dataset
self.public_datasets.append(pub_dataset)
self.private_datasets.append(private_dataset)
self.validation_datasets.append(val_dataset)
self.test_datasets.append(test_dataset)
elif dataset == 'shoes':
# convert label value in datasets
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
train_dataset = raw_datasets['train']
val_dataset = raw_datasets['validation']
test_dataset = raw_datasets['test']
# split public dataset
public_private_dataset = train_dataset.train_test_split(test_size=args.public_ratio, seed=args.seed, stratify_by_column='label')
private_dataset = public_private_dataset['train']
pub_dataset = public_private_dataset['test']
del public_private_dataset
self.public_datasets.append(pub_dataset)
self.private_datasets.append(private_dataset)
self.validation_datasets.append(val_dataset)
self.test_datasets.append(test_dataset)
elif dataset == 'software':
# convert label value in datasets
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
train_dataset = raw_datasets['train']
val_dataset = raw_datasets['validation']
test_dataset = raw_datasets['test']
# split public dataset
public_private_dataset = train_dataset.train_test_split(test_size=args.public_ratio, seed=args.seed, stratify_by_column='label')
private_dataset = public_private_dataset['train']
pub_dataset = public_private_dataset['test']
del public_private_dataset
self.public_datasets.append(pub_dataset)
self.private_datasets.append(private_dataset)
self.validation_datasets.append(val_dataset)
self.test_datasets.append(test_dataset)
# merge all public、validation and test data
# self.public_datasets = concatenate_datasets(self.public_datasets).shuffle(seed=self.args.seed)
# self.public_datasets.to_csv("/gemini/code/public_datasets.csv", index=False)
self.merged_val_datasets = concatenate_datasets(self.validation_datasets).shuffle(seed=self.args.seed)
self.merged_id_test_datasets = concatenate_datasets(self.test_datasets).shuffle(seed=self.args.seed)
self.merged_id_test_datasets.to_csv("/gemini/code/merged_id_test_datasets.csv", index=False)
# Use text set from the all domains
all_domain_test = load_dataset('csv', data_files=args.data_dir + '/all_domain_test.csv', split="train")
all_domain_test = all_domain_test.remove_columns(['idx'])
self.merged_ood_test_datasets = all_domain_test.class_encode_column('label')
logging.info(f'The merged_test_datasets after full-domain reconstruction: {self.merged_ood_test_datasets}')
# Use public data across all domains
all_domain_pub_data = load_dataset('csv', data_files=args.data_dir + '/all_domain_pub_data.csv', split="train")
self.public_datasets = all_domain_pub_data.class_encode_column('label')
# Unconstrained LLM-based public data
# self.public_datasets = load_dataset('csv', data_files='/gemini/code/ood_pub_data/gpt_all_domain_pub_data.csv', split="train")
# Vocabulary-constrained LLM-based public data
# self.public_datasets = load_dataset('csv', data_files='/gemini/code/ood_pub_data/gpt_all_domain_pub_data_vocab.csv', split="train")
logging.info(f'The public_datasets after full-domain reconstruction: {self.public_datasets}')
# centralized and centralized_mixed dataset
else:
for dataset in args.datasets:
raw_datasets = load_dataset(args.data_dir + '/{}'.format(dataset))
raw_datasets = raw_datasets.remove_columns(['idx'])
if dataset == 'automotive':
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
automotive_train_dataset = raw_datasets['train']
automotive_val_dataset = raw_datasets['validation']
automotive_test_dataset = raw_datasets['test']
# split public dataset
automotive_public_private_dataset = automotive_train_dataset.train_test_split(test_size=args.public_ratio,
seed=args.seed,
stratify_by_column='label')
automotive_private_dataset = automotive_public_private_dataset['train']
del automotive_public_private_dataset
self.centralized_train_datasets.append(automotive_private_dataset)
self.validation_datasets.append(automotive_val_dataset)
self.test_datasets.append(automotive_test_dataset)
if dataset == 'baby':
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
baby_train_dataset = raw_datasets['train']
baby_val_dataset = raw_datasets['validation']
baby_test_dataset = raw_datasets['test']
# split public dataset
baby_public_private_dataset = baby_train_dataset.train_test_split(
test_size=args.public_ratio,
seed=args.seed,
stratify_by_column='label')
baby_private_dataset = baby_public_private_dataset['train']
del baby_public_private_dataset
self.centralized_train_datasets.append(baby_private_dataset)
self.validation_datasets.append(baby_val_dataset)
self.test_datasets.append(baby_test_dataset)
if dataset == 'clothing':
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
clothing_train_dataset = raw_datasets['train']
clothing_val_dataset = raw_datasets['validation']
clothing_test_dataset = raw_datasets['test']
# split public dataset
clothing_public_private_dataset = clothing_train_dataset.train_test_split(test_size=args.public_ratio,
seed=args.seed,
stratify_by_column='label')
tsa_private_dataset = clothing_public_private_dataset['train']
del clothing_public_private_dataset
self.centralized_train_datasets.append(tsa_private_dataset)
self.validation_datasets.append(clothing_val_dataset)
self.test_datasets.append(clothing_test_dataset)
if dataset == 'health':
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
health_train_dataset = raw_datasets['train']
health_val_dataset = raw_datasets['validation']
health_test_dataset = raw_datasets['test']
# split public dataset
health_public_private_dataset = health_train_dataset.train_test_split(
test_size=args.public_ratio,
seed=args.seed,
stratify_by_column='label')
health_private_dataset = health_public_private_dataset['train']
del health_public_private_dataset
self.centralized_train_datasets.append(health_private_dataset)
self.validation_datasets.append(health_val_dataset)
self.test_datasets.append(health_test_dataset)
if dataset == 'sport':
# raw_datasets = raw_datasets.map(utils.replace_labels)
# label type: Value ---> ClassLabel
raw_datasets = raw_datasets.class_encode_column('label')
sport_train_dataset = raw_datasets['train']
sport_val_dataset = raw_datasets['validation']
sport_test_dataset = raw_datasets['test']
# split public dataset
sport_public_private_dataset = sport_train_dataset.train_test_split(
test_size=args.public_ratio,
seed=args.seed,
stratify_by_column='label')
sport_private_dataset = sport_public_private_dataset['train']
del sport_public_private_dataset
self.centralized_train_datasets.append(sport_private_dataset)
self.validation_datasets.append(sport_val_dataset)
self.test_datasets.append(sport_test_dataset)
self.mix_train_datasets = concatenate_datasets(self.centralized_train_datasets).shuffle(seed=self.args.seed)
self.mix_validation_datasets = concatenate_datasets(self.validation_datasets).shuffle(seed=self.args.seed)
# self.mix_test_datasets = concatenate_datasets(self.test_datasets).shuffle(seed=self.args.seed)
# Use test set from all domains
all_domain_test = load_dataset('csv', data_files=args.data_dir + '/all_domain_test.csv')['train']
all_domain_test = all_domain_test.remove_columns(['idx'])
self.mix_test_datasets = all_domain_test.class_encode_column('label')
logging.info(f'The mix_test_datasets after full-domain reconstruction: {self.mix_test_datasets}')
if __name__ == '__main__':
class args:
seed = 42
# datasets = ['automotive', 'baby', 'clothing', 'health', 'sport', 'beauty', 'patio', 'pet', 'shoes', 'software']
datasets = ['automotive', 'baby', 'clothing', 'health', 'sport']
data_dir = '/gemini/data-1/all_domain_test'
K = 5
public_ratio = 0.2
algorithm = 'fd_pub_data'
set_seed(args.seed)
sa = DatasetPartition(args)
train_datasets = sa.private_datasets
public_dataset = sa.public_datasets
validation_datasets = sa.validation_datasets
test_datasets = sa.test_datasets
merged_val_dataset = sa.merged_val_datasets
merged_in_test_dataset = sa.merged_id_test_datasets
merged_ood_test_dataset = sa.merged_ood_test_datasets
centralized_train_datasets = sa.centralized_train_datasets
mix_train_datasets = sa.mix_train_datasets
mix_validation_datasets = sa.mix_validation_datasets
mix_test_datasets = sa.mix_test_datasets
if centralized_train_datasets:
print('length of centralized_train_datasets: {}'.format(
[len(centralized_train_datasets[k]) for k in range(args.K)]))
else:
if public_dataset:
print('length of public_dataset: {}'.format(len(public_dataset)))
if train_datasets:
print('length of train_datasets: {}'.format([len(train_datasets[k]) for k in range(args.K)]))
if merged_val_dataset:
print('length of merged_val_dataset: {}'.format(len(merged_val_dataset)))
if merged_in_test_dataset:
print('length of merged_test_dataset: {}'.format(len(merged_in_test_dataset)))
if merged_ood_test_dataset:
print('length of merged_test_dataset: {}'.format(len(merged_ood_test_dataset)))
if validation_datasets:
print('length of val_datasets: {}'.format([len(validation_datasets[k]) for k in range(args.K)]))
if test_datasets:
print('length of test_datasets: {}'.format([len(test_datasets[k]) for k in range(args.K)]))
if mix_train_datasets:
print('length of mix_train_datasets: {}'.format(len(mix_train_datasets)))
if mix_validation_datasets:
print('length of mix_validation_datasets: {}'.format(len(mix_validation_datasets)))
if mix_test_datasets:
print('length of mix_test_datasets: {}'.format(len(mix_test_datasets)))