File size: 77,010 Bytes
f8f5cdf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
# Copyright 2020 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import random
import warnings
from collections.abc import Mapping
from dataclasses import dataclass
from random import randint
from typing import Any, Callable, Dict, List, NewType, Optional, Tuple, Union

import numpy as np

from ..models.bert import BertTokenizer, BertTokenizerFast
from ..tokenization_utils_base import PreTrainedTokenizerBase
from ..utils import PaddingStrategy


InputDataClass = NewType("InputDataClass", Any)

"""
A DataCollator is a function that takes a list of samples from a Dataset and collate them into a batch, as a dictionary
of PyTorch/TensorFlow tensors or NumPy arrays.
"""
DataCollator = NewType("DataCollator", Callable[[List[InputDataClass]], Dict[str, Any]])


class DataCollatorMixin:
    def __call__(self, features, return_tensors=None):
        if return_tensors is None:
            return_tensors = self.return_tensors
        if return_tensors == "tf":
            return self.tf_call(features)
        elif return_tensors == "pt":
            return self.torch_call(features)
        elif return_tensors == "np":
            return self.numpy_call(features)
        else:
            raise ValueError(f"Framework '{return_tensors}' not recognized!")


def default_data_collator(features: List[InputDataClass], return_tensors="pt") -> Dict[str, Any]:
    """
    Very simple data collator that simply collates batches of dict-like objects and performs special handling for
    potential keys named:

        - `label`: handles a single value (int or float) per object
        - `label_ids`: handles a list of values per object

    Does not do any additional preprocessing: property names of the input object will be used as corresponding inputs
    to the model. See glue and ner for example of how it's useful.
    """

    # In this function we'll make the assumption that all `features` in the batch
    # have the same attributes.
    # So we will look at the first element as a proxy for what attributes exist
    # on the whole batch.

    if return_tensors == "pt":
        return torch_default_data_collator(features)
    elif return_tensors == "tf":
        return tf_default_data_collator(features)
    elif return_tensors == "np":
        return numpy_default_data_collator(features)


@dataclass
class DefaultDataCollator(DataCollatorMixin):
    """
    Very simple data collator that simply collates batches of dict-like objects and performs special handling for
    potential keys named:

        - `label`: handles a single value (int or float) per object
        - `label_ids`: handles a list of values per object

    Does not do any additional preprocessing: property names of the input object will be used as corresponding inputs
    to the model. See glue and ner for example of how it's useful.

    This is an object (like other data collators) rather than a pure function like default_data_collator. This can be
    helpful if you need to set a return_tensors value at initialization.

    Args:
        return_tensors (`str`, *optional*, defaults to `"pt"`):
            The type of Tensor to return. Allowable values are "np", "pt" and "tf".
    """

    return_tensors: str = "pt"

    def __call__(self, features: List[Dict[str, Any]], return_tensors=None) -> Dict[str, Any]:
        if return_tensors is None:
            return_tensors = self.return_tensors
        return default_data_collator(features, return_tensors)


def torch_default_data_collator(features: List[InputDataClass]) -> Dict[str, Any]:
    import torch

    if not isinstance(features[0], Mapping):
        features = [vars(f) for f in features]
    first = features[0]
    batch = {}

    # Special handling for labels.
    # Ensure that tensor is created with the correct type
    # (it should be automatically the case, but let's make sure of it.)
    if "label" in first and first["label"] is not None:
        label = first["label"].item() if isinstance(first["label"], torch.Tensor) else first["label"]
        dtype = torch.long if isinstance(label, int) else torch.float
        batch["labels"] = torch.tensor([f["label"] for f in features], dtype=dtype)
    elif "label_ids" in first and first["label_ids"] is not None:
        if isinstance(first["label_ids"], torch.Tensor):
            batch["labels"] = torch.stack([f["label_ids"] for f in features])
        else:
            dtype = torch.long if type(first["label_ids"][0]) is int else torch.float
            batch["labels"] = torch.tensor([f["label_ids"] for f in features], dtype=dtype)

    # Handling of all other possible keys.
    # Again, we will use the first element to figure out which key/values are not None for this model.
    for k, v in first.items():
        if k not in ("label", "label_ids") and v is not None and not isinstance(v, str):
            if isinstance(v, torch.Tensor):
                batch[k] = torch.stack([f[k] for f in features])
            elif isinstance(v, np.ndarray):
                batch[k] = torch.tensor(np.stack([f[k] for f in features]))
            else:
                batch[k] = torch.tensor([f[k] for f in features])

    return batch


def tf_default_data_collator(features: List[InputDataClass]) -> Dict[str, Any]:
    import tensorflow as tf

    if not isinstance(features[0], Mapping):
        features = [vars(f) for f in features]
    first = features[0]
    batch = {}

    # Special handling for labels.
    # Ensure that tensor is created with the correct type
    # (it should be automatically the case, but let's make sure of it.)
    if "label" in first and first["label"] is not None:
        label_col_name = "label"
    elif "label_ids" in first and first["label_ids"] is not None:
        label_col_name = "label_ids"
    elif "labels" in first and first["labels"] is not None:
        label_col_name = "labels"
    else:
        label_col_name = None
    if label_col_name is not None:
        if isinstance(first[label_col_name], tf.Tensor):
            dtype = tf.int64 if first[label_col_name].dtype.is_integer else tf.float32
        elif isinstance(first[label_col_name], np.ndarray) or isinstance(first[label_col_name], np.generic):
            dtype = tf.int64 if np.issubdtype(first[label_col_name].dtype, np.integer) else tf.float32
        elif isinstance(first[label_col_name], (tuple, list)):
            dtype = tf.int64 if isinstance(first[label_col_name][0], int) else tf.float32
        else:
            dtype = tf.int64 if isinstance(first[label_col_name], int) else tf.float32
        batch["labels"] = tf.convert_to_tensor([f[label_col_name] for f in features], dtype=dtype)
    # Handling of all other possible keys.
    # Again, we will use the first element to figure out which key/values are not None for this model.
    for k, v in first.items():
        if k not in ("label", "label_ids", "labels") and v is not None and not isinstance(v, str):
            if isinstance(v, (tf.Tensor, np.ndarray)):
                batch[k] = tf.stack([f[k] for f in features])
            else:
                batch[k] = tf.convert_to_tensor([f[k] for f in features])

    return batch


def numpy_default_data_collator(features: List[InputDataClass]) -> Dict[str, Any]:
    if not isinstance(features[0], Mapping):
        features = [vars(f) for f in features]
    first = features[0]
    batch = {}

    # Special handling for labels.
    # Ensure that tensor is created with the correct type
    # (it should be automatically the case, but let's make sure of it.)
    if "label" in first and first["label"] is not None:
        label = first["label"].item() if isinstance(first["label"], np.ndarray) else first["label"]
        dtype = np.int64 if isinstance(label, int) else np.float32
        batch["labels"] = np.array([f["label"] for f in features], dtype=dtype)
    elif "label_ids" in first and first["label_ids"] is not None:
        if isinstance(first["label_ids"], np.ndarray):
            batch["labels"] = np.stack([f["label_ids"] for f in features])
        else:
            dtype = np.int64 if type(first["label_ids"][0]) is int else np.float32
            batch["labels"] = np.array([f["label_ids"] for f in features], dtype=dtype)

    # Handling of all other possible keys.
    # Again, we will use the first element to figure out which key/values are not None for this model.
    for k, v in first.items():
        if k not in ("label", "label_ids") and v is not None and not isinstance(v, str):
            if isinstance(v, np.ndarray):
                batch[k] = np.stack([f[k] for f in features])
            else:
                batch[k] = np.array([f[k] for f in features])

    return batch


@dataclass
class DataCollatorWithPadding:
    """
    Data collator that will dynamically pad the inputs received.

    Args:
        tokenizer ([`PreTrainedTokenizer`] or [`PreTrainedTokenizerFast`]):
            The tokenizer used for encoding the data.
        padding (`bool`, `str` or [`~utils.PaddingStrategy`], *optional*, defaults to `True`):
            Select a strategy to pad the returned sequences (according to the model's padding side and padding index)
            among:

            - `True` or `'longest'` (default): Pad to the longest sequence in the batch (or no padding if only a single
              sequence is provided).
            - `'max_length'`: Pad to a maximum length specified with the argument `max_length` or to the maximum
              acceptable input length for the model if that argument is not provided.
            - `False` or `'do_not_pad'`: No padding (i.e., can output a batch with sequences of different lengths).
        max_length (`int`, *optional*):
            Maximum length of the returned list and optionally padding length (see above).
        pad_to_multiple_of (`int`, *optional*):
            If set will pad the sequence to a multiple of the provided value.

            This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >=
            7.5 (Volta).
        return_tensors (`str`, *optional*, defaults to `"pt"`):
            The type of Tensor to return. Allowable values are "np", "pt" and "tf".
    """

    tokenizer: PreTrainedTokenizerBase
    padding: Union[bool, str, PaddingStrategy] = True
    max_length: Optional[int] = None
    pad_to_multiple_of: Optional[int] = None
    return_tensors: str = "pt"

    def __call__(self, features: List[Dict[str, Any]]) -> Dict[str, Any]:
        batch = self.tokenizer.pad(
            features,
            padding=self.padding,
            max_length=self.max_length,
            pad_to_multiple_of=self.pad_to_multiple_of,
            return_tensors=self.return_tensors,
        )
        if "label" in batch:
            batch["labels"] = batch["label"]
            del batch["label"]
        if "label_ids" in batch:
            batch["labels"] = batch["label_ids"]
            del batch["label_ids"]
        return batch


@dataclass
class DataCollatorForTokenClassification(DataCollatorMixin):
    """
    Data collator that will dynamically pad the inputs received, as well as the labels.

    Args:
        tokenizer ([`PreTrainedTokenizer`] or [`PreTrainedTokenizerFast`]):
            The tokenizer used for encoding the data.
        padding (`bool`, `str` or [`~utils.PaddingStrategy`], *optional*, defaults to `True`):
            Select a strategy to pad the returned sequences (according to the model's padding side and padding index)
            among:

            - `True` or `'longest'` (default): Pad to the longest sequence in the batch (or no padding if only a single
              sequence is provided).
            - `'max_length'`: Pad to a maximum length specified with the argument `max_length` or to the maximum
              acceptable input length for the model if that argument is not provided.
            - `False` or `'do_not_pad'`: No padding (i.e., can output a batch with sequences of different lengths).
        max_length (`int`, *optional*):
            Maximum length of the returned list and optionally padding length (see above).
        pad_to_multiple_of (`int`, *optional*):
            If set will pad the sequence to a multiple of the provided value.

            This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >=
            7.5 (Volta).
        label_pad_token_id (`int`, *optional*, defaults to -100):
            The id to use when padding the labels (-100 will be automatically ignore by PyTorch loss functions).
        return_tensors (`str`, *optional*, defaults to `"pt"`):
            The type of Tensor to return. Allowable values are "np", "pt" and "tf".
    """

    tokenizer: PreTrainedTokenizerBase
    padding: Union[bool, str, PaddingStrategy] = True
    max_length: Optional[int] = None
    pad_to_multiple_of: Optional[int] = None
    label_pad_token_id: int = -100
    return_tensors: str = "pt"

    def torch_call(self, features):
        import torch

        label_name = "label" if "label" in features[0].keys() else "labels"
        labels = [feature[label_name] for feature in features] if label_name in features[0].keys() else None

        no_labels_features = [{k: v for k, v in feature.items() if k != label_name} for feature in features]

        batch = self.tokenizer.pad(
            no_labels_features,
            padding=self.padding,
            max_length=self.max_length,
            pad_to_multiple_of=self.pad_to_multiple_of,
            return_tensors="pt",
        )

        if labels is None:
            return batch

        sequence_length = batch["input_ids"].shape[1]
        padding_side = self.tokenizer.padding_side

        def to_list(tensor_or_iterable):
            if isinstance(tensor_or_iterable, torch.Tensor):
                return tensor_or_iterable.tolist()
            return list(tensor_or_iterable)

        if padding_side == "right":
            batch[label_name] = [
                to_list(label) + [self.label_pad_token_id] * (sequence_length - len(label)) for label in labels
            ]
        else:
            batch[label_name] = [
                [self.label_pad_token_id] * (sequence_length - len(label)) + to_list(label) for label in labels
            ]

        batch[label_name] = torch.tensor(batch[label_name], dtype=torch.int64)
        return batch

    def tf_call(self, features):
        import tensorflow as tf

        label_name = "label" if "label" in features[0].keys() else "labels"
        labels = [feature[label_name] for feature in features] if label_name in features[0].keys() else None
        batch = self.tokenizer.pad(
            features,
            padding=self.padding,
            max_length=self.max_length,
            pad_to_multiple_of=self.pad_to_multiple_of,
            # Conversion to tensors will fail if we have labels as they are not of the same length yet.
            return_tensors="tf" if labels is None else None,
        )

        if labels is None:
            return batch

        sequence_length = tf.convert_to_tensor(batch["input_ids"]).shape[1]
        padding_side = self.tokenizer.padding_side
        if padding_side == "right":
            batch["labels"] = [
                list(label) + [self.label_pad_token_id] * (sequence_length - len(label)) for label in labels
            ]
        else:
            batch["labels"] = [
                [self.label_pad_token_id] * (sequence_length - len(label)) + list(label) for label in labels
            ]

        batch = {k: tf.convert_to_tensor(v, dtype=tf.int64) for k, v in batch.items()}
        return batch

    def numpy_call(self, features):
        label_name = "label" if "label" in features[0].keys() else "labels"
        labels = [feature[label_name] for feature in features] if label_name in features[0].keys() else None
        batch = self.tokenizer.pad(
            features,
            padding=self.padding,
            max_length=self.max_length,
            pad_to_multiple_of=self.pad_to_multiple_of,
            # Conversion to tensors will fail if we have labels as they are not of the same length yet.
            return_tensors="np" if labels is None else None,
        )

        if labels is None:
            return batch

        sequence_length = np.array(batch["input_ids"]).shape[1]
        padding_side = self.tokenizer.padding_side
        if padding_side == "right":
            batch["labels"] = [
                list(label) + [self.label_pad_token_id] * (sequence_length - len(label)) for label in labels
            ]
        else:
            batch["labels"] = [
                [self.label_pad_token_id] * (sequence_length - len(label)) + list(label) for label in labels
            ]

        batch = {k: np.array(v, dtype=np.int64) for k, v in batch.items()}
        return batch


def _torch_collate_batch(examples, tokenizer, pad_to_multiple_of: Optional[int] = None):
    """Collate `examples` into a batch, using the information in `tokenizer` for padding if necessary."""
    import torch

    # Tensorize if necessary.
    if isinstance(examples[0], (list, tuple, np.ndarray)):
        examples = [torch.tensor(e, dtype=torch.long) for e in examples]

    length_of_first = examples[0].size(0)

    # Check if padding is necessary.

    are_tensors_same_length = all(x.size(0) == length_of_first for x in examples)
    if are_tensors_same_length and (pad_to_multiple_of is None or length_of_first % pad_to_multiple_of == 0):
        return torch.stack(examples, dim=0)

    # If yes, check if we have a `pad_token`.
    if tokenizer._pad_token is None:
        raise ValueError(
            "You are attempting to pad samples but the tokenizer you are using"
            f" ({tokenizer.__class__.__name__}) does not have a pad token."
        )

    # Creating the full tensor and filling it with our data.
    max_length = max(x.size(0) for x in examples)
    if pad_to_multiple_of is not None and (max_length % pad_to_multiple_of != 0):
        max_length = ((max_length // pad_to_multiple_of) + 1) * pad_to_multiple_of
    result = examples[0].new_full([len(examples), max_length], tokenizer.pad_token_id)
    for i, example in enumerate(examples):
        if tokenizer.padding_side == "right":
            result[i, : example.shape[0]] = example
        else:
            result[i, -example.shape[0] :] = example
    return result


def _tf_collate_batch(examples, tokenizer, pad_to_multiple_of: Optional[int] = None):
    import tensorflow as tf

    """Collate `examples` into a batch, using the information in `tokenizer` for padding if necessary."""
    # Tensorize if necessary.
    if isinstance(examples[0], (list, tuple)):
        examples = [tf.convert_to_tensor(e, dtype=tf.int64) for e in examples]

    # Check if padding is necessary.
    length_of_first = len(examples[0])
    are_tensors_same_length = all(len(x) == length_of_first for x in examples)
    if are_tensors_same_length and (pad_to_multiple_of is None or length_of_first % pad_to_multiple_of == 0):
        return tf.stack(examples, axis=0)

    # If yes, check if we have a `pad_token`.
    if tokenizer._pad_token is None:
        raise ValueError(
            "You are attempting to pad samples but the tokenizer you are using"
            f" ({tokenizer.__class__.__name__}) does not have a pad token."
        )

    # Creating the full tensor and filling it with our data.
    max_length = max(len(x) for x in examples)
    if pad_to_multiple_of is not None and (max_length % pad_to_multiple_of != 0):
        max_length = ((max_length // pad_to_multiple_of) + 1) * pad_to_multiple_of
    # result = examples[0].new_full([len(examples), max_length], tokenizer.pad_token_id)
    result = []
    rank = tf.rank(examples[0])
    paddings = np.zeros((rank, 2), dtype=np.int32)
    for example in examples:
        if tokenizer.padding_side == "right":
            paddings[0, 1] = max_length - len(example)
        else:
            paddings[0, 0] = max_length - len(example)
        result.append(tf.pad(example, paddings, constant_values=tokenizer.pad_token_id))
    return tf.stack(result, axis=0)


def _numpy_collate_batch(examples, tokenizer, pad_to_multiple_of: Optional[int] = None):
    """Collate `examples` into a batch, using the information in `tokenizer` for padding if necessary."""
    # Tensorize if necessary.
    if isinstance(examples[0], (list, tuple)):
        examples = [np.array(e, dtype=np.int64) for e in examples]

    # Check if padding is necessary.
    length_of_first = len(examples[0])
    are_tensors_same_length = all(len(x) == length_of_first for x in examples)
    if are_tensors_same_length and (pad_to_multiple_of is None or length_of_first % pad_to_multiple_of == 0):
        return np.stack(examples, axis=0)

    # If yes, check if we have a `pad_token`.
    if tokenizer._pad_token is None:
        raise ValueError(
            "You are attempting to pad samples but the tokenizer you are using"
            f" ({tokenizer.__class__.__name__}) does not have a pad token."
        )

    # Creating the full tensor and filling it with our data.
    max_length = max(len(x) for x in examples)
    if pad_to_multiple_of is not None and (max_length % pad_to_multiple_of != 0):
        max_length = ((max_length // pad_to_multiple_of) + 1) * pad_to_multiple_of
    result = np.full(shape=(len(examples), max_length), fill_value=tokenizer.pad_token_id, dtype=examples[0].dtype)
    for i, example in enumerate(examples):
        if tokenizer.padding_side == "right":
            result[i, : example.shape[0]] = example
        else:
            result[i, -example.shape[0] :] = example
    return result


def tolist(x):
    if isinstance(x, list):
        return x
    elif hasattr(x, "numpy"):  # Checks for TF tensors without needing the import
        x = x.numpy()
    return x.tolist()


@dataclass
class DataCollatorForSeq2Seq:
    """
    Data collator that will dynamically pad the inputs received, as well as the labels.

    Args:
        tokenizer ([`PreTrainedTokenizer`] or [`PreTrainedTokenizerFast`]):
            The tokenizer used for encoding the data.
        model ([`PreTrainedModel`], *optional*):
            The model that is being trained. If set and has the *prepare_decoder_input_ids_from_labels*, use it to
            prepare the *decoder_input_ids*

            This is useful when using *label_smoothing* to avoid calculating loss twice.
        padding (`bool`, `str` or [`~utils.PaddingStrategy`], *optional*, defaults to `True`):
            Select a strategy to pad the returned sequences (according to the model's padding side and padding index)
            among:

            - `True` or `'longest'` (default): Pad to the longest sequence in the batch (or no padding if only a single
              sequence is provided).
            - `'max_length'`: Pad to a maximum length specified with the argument `max_length` or to the maximum
              acceptable input length for the model if that argument is not provided.
            - `False` or `'do_not_pad'`: No padding (i.e., can output a batch with sequences of different lengths).
        max_length (`int`, *optional*):
            Maximum length of the returned list and optionally padding length (see above).
        pad_to_multiple_of (`int`, *optional*):
            If set will pad the sequence to a multiple of the provided value.

            This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >=
            7.5 (Volta).
        label_pad_token_id (`int`, *optional*, defaults to -100):
            The id to use when padding the labels (-100 will be automatically ignored by PyTorch loss functions).
        return_tensors (`str`, *optional*, defaults to `"pt"`):
            The type of Tensor to return. Allowable values are "np", "pt" and "tf".
    """

    tokenizer: PreTrainedTokenizerBase
    model: Optional[Any] = None
    padding: Union[bool, str, PaddingStrategy] = True
    max_length: Optional[int] = None
    pad_to_multiple_of: Optional[int] = None
    label_pad_token_id: int = -100
    return_tensors: str = "pt"

    def __call__(self, features, return_tensors=None):
        if return_tensors is None:
            return_tensors = self.return_tensors
        labels = [feature["labels"] for feature in features] if "labels" in features[0].keys() else None
        # We have to pad the labels before calling `tokenizer.pad` as this method won't pad them and needs them of the
        # same length to return tensors.
        if labels is not None:
            max_label_length = max(len(l) for l in labels)
            if self.pad_to_multiple_of is not None:
                max_label_length = (
                    (max_label_length + self.pad_to_multiple_of - 1)
                    // self.pad_to_multiple_of
                    * self.pad_to_multiple_of
                )

            padding_side = self.tokenizer.padding_side
            for feature in features:
                remainder = [self.label_pad_token_id] * (max_label_length - len(feature["labels"]))
                if isinstance(feature["labels"], list):
                    feature["labels"] = (
                        feature["labels"] + remainder if padding_side == "right" else remainder + feature["labels"]
                    )
                elif padding_side == "right":
                    feature["labels"] = np.concatenate([feature["labels"], remainder]).astype(np.int64)
                else:
                    feature["labels"] = np.concatenate([remainder, feature["labels"]]).astype(np.int64)

        features = self.tokenizer.pad(
            features,
            padding=self.padding,
            max_length=self.max_length,
            pad_to_multiple_of=self.pad_to_multiple_of,
            return_tensors=return_tensors,
        )

        # prepare decoder_input_ids
        if (
            labels is not None
            and self.model is not None
            and hasattr(self.model, "prepare_decoder_input_ids_from_labels")
        ):
            decoder_input_ids = self.model.prepare_decoder_input_ids_from_labels(labels=features["labels"])
            features["decoder_input_ids"] = decoder_input_ids

        return features


@dataclass
class DataCollatorForLanguageModeling(DataCollatorMixin):
    """
    Data collator used for language modeling. Inputs are dynamically padded to the maximum length of a batch if they
    are not all of the same length.

    Args:
        tokenizer ([`PreTrainedTokenizer`] or [`PreTrainedTokenizerFast`]):
            The tokenizer used for encoding the data.
        mlm (`bool`, *optional*, defaults to `True`):
            Whether or not to use masked language modeling. If set to `False`, the labels are the same as the inputs
            with the padding tokens ignored (by setting them to -100). Otherwise, the labels are -100 for non-masked
            tokens and the value to predict for the masked token.
        mlm_probability (`float`, *optional*, defaults to 0.15):
            The probability with which to (randomly) mask tokens in the input, when `mlm` is set to `True`.
        pad_to_multiple_of (`int`, *optional*):
            If set will pad the sequence to a multiple of the provided value.
        return_tensors (`str`):
            The type of Tensor to return. Allowable values are "np", "pt" and "tf".

    <Tip>

    For best performance, this data collator should be used with a dataset having items that are dictionaries or
    BatchEncoding, with the `"special_tokens_mask"` key, as returned by a [`PreTrainedTokenizer`] or a
    [`PreTrainedTokenizerFast`] with the argument `return_special_tokens_mask=True`.

    </Tip>"""

    tokenizer: PreTrainedTokenizerBase
    mlm: bool = True
    mlm_probability: float = 0.15
    pad_to_multiple_of: Optional[int] = None
    tf_experimental_compile: bool = False
    return_tensors: str = "pt"

    def __post_init__(self):
        if self.mlm and self.tokenizer.mask_token is None:
            raise ValueError(
                "This tokenizer does not have a mask token which is necessary for masked language modeling. "
                "You should pass `mlm=False` to train on causal language modeling instead."
            )
        if self.tf_experimental_compile:
            import tensorflow as tf

            self.tf_mask_tokens = tf.function(self.tf_mask_tokens, jit_compile=True)

    @staticmethod
    def tf_bernoulli(shape, probability):
        import tensorflow as tf

        prob_matrix = tf.fill(shape, probability)
        return tf.cast(prob_matrix - tf.random.uniform(shape, 0, 1) >= 0, tf.bool)

    def tf_mask_tokens(
        self, inputs: Any, vocab_size, mask_token_id, special_tokens_mask: Optional[Any] = None
    ) -> Tuple[Any, Any]:
        """
        Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original.
        """
        import tensorflow as tf

        mask_token_id = tf.cast(mask_token_id, inputs.dtype)

        input_shape = tf.shape(inputs)
        # 1 for a special token, 0 for a normal token in the special tokens mask
        # We sample a few tokens in each sequence for MLM training (with probability `self.mlm_probability`)
        masked_indices = self.tf_bernoulli(input_shape, self.mlm_probability) & ~special_tokens_mask
        # Replace unmasked indices with -100 in the labels since we only compute loss on masked tokens
        labels = tf.where(masked_indices, inputs, -100)

        # 80% of the time, we replace masked input tokens with tokenizer.mask_token ([MASK])
        indices_replaced = self.tf_bernoulli(input_shape, 0.8) & masked_indices

        inputs = tf.where(indices_replaced, mask_token_id, inputs)

        # 10% of the time, we replace masked input tokens with random word
        indices_random = self.tf_bernoulli(input_shape, 0.1) & masked_indices & ~indices_replaced
        random_words = tf.random.uniform(input_shape, maxval=vocab_size, dtype=inputs.dtype)

        inputs = tf.where(indices_random, random_words, inputs)

        # The rest of the time (10% of the time) we keep the masked input tokens unchanged
        return inputs, labels

    def tf_call(self, examples: List[Union[List[int], Any, Dict[str, Any]]]) -> Dict[str, Any]:
        import tensorflow as tf

        # Handle dict or lists with proper padding and conversion to tensor.
        if isinstance(examples[0], Mapping):
            batch = self.tokenizer.pad(examples, return_tensors="tf", pad_to_multiple_of=self.pad_to_multiple_of)
        else:
            batch = {
                "input_ids": _tf_collate_batch(examples, self.tokenizer, pad_to_multiple_of=self.pad_to_multiple_of)
            }

        # If special token mask has been preprocessed, pop it from the dict.
        special_tokens_mask = batch.pop("special_tokens_mask", None)
        if self.mlm:
            if special_tokens_mask is None:
                special_tokens_mask = [
                    self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True)
                    for val in batch["input_ids"].numpy().tolist()
                ]
                # Cannot directly create as bool
                special_tokens_mask = tf.cast(tf.convert_to_tensor(special_tokens_mask, dtype=tf.int64), tf.bool)
            else:
                special_tokens_mask = tf.cast(special_tokens_mask, tf.bool)
            batch["input_ids"], batch["labels"] = self.tf_mask_tokens(
                tf.cast(batch["input_ids"], tf.int64),
                special_tokens_mask=special_tokens_mask,
                mask_token_id=self.tokenizer.mask_token_id,
                vocab_size=len(self.tokenizer),
            )
        else:
            labels = batch["input_ids"]
            if self.tokenizer.pad_token_id is not None:
                # Replace self.tokenizer.pad_token_id with -100
                labels = tf.where(labels == self.tokenizer.pad_token_id, -100, labels)
            else:
                labels = tf.identity(labels)  # Makes a copy, just in case
            batch["labels"] = labels
        return batch

    def torch_call(self, examples: List[Union[List[int], Any, Dict[str, Any]]]) -> Dict[str, Any]:
        # Handle dict or lists with proper padding and conversion to tensor.
        if isinstance(examples[0], Mapping):
            batch = self.tokenizer.pad(examples, return_tensors="pt", pad_to_multiple_of=self.pad_to_multiple_of)
        else:
            batch = {
                "input_ids": _torch_collate_batch(examples, self.tokenizer, pad_to_multiple_of=self.pad_to_multiple_of)
            }

        # If special token mask has been preprocessed, pop it from the dict.
        special_tokens_mask = batch.pop("special_tokens_mask", None)
        if self.mlm:
            batch["input_ids"], batch["labels"] = self.torch_mask_tokens(
                batch["input_ids"], special_tokens_mask=special_tokens_mask
            )
        else:
            labels = batch["input_ids"].clone()
            if self.tokenizer.pad_token_id is not None:
                labels[labels == self.tokenizer.pad_token_id] = -100
            batch["labels"] = labels
        return batch

    def torch_mask_tokens(self, inputs: Any, special_tokens_mask: Optional[Any] = None) -> Tuple[Any, Any]:
        """
        Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original.
        """
        import torch

        labels = inputs.clone()
        # We sample a few tokens in each sequence for MLM training (with probability `self.mlm_probability`)
        probability_matrix = torch.full(labels.shape, self.mlm_probability)
        if special_tokens_mask is None:
            special_tokens_mask = [
                self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True) for val in labels.tolist()
            ]
            special_tokens_mask = torch.tensor(special_tokens_mask, dtype=torch.bool)
        else:
            special_tokens_mask = special_tokens_mask.bool()

        probability_matrix.masked_fill_(special_tokens_mask, value=0.0)
        masked_indices = torch.bernoulli(probability_matrix).bool()
        labels[~masked_indices] = -100  # We only compute loss on masked tokens

        # 80% of the time, we replace masked input tokens with tokenizer.mask_token ([MASK])
        indices_replaced = torch.bernoulli(torch.full(labels.shape, 0.8)).bool() & masked_indices
        inputs[indices_replaced] = self.tokenizer.convert_tokens_to_ids(self.tokenizer.mask_token)

        # 10% of the time, we replace masked input tokens with random word
        indices_random = torch.bernoulli(torch.full(labels.shape, 0.5)).bool() & masked_indices & ~indices_replaced
        random_words = torch.randint(len(self.tokenizer), labels.shape, dtype=torch.long)
        inputs[indices_random] = random_words[indices_random]

        # The rest of the time (10% of the time) we keep the masked input tokens unchanged
        return inputs, labels

    def numpy_call(self, examples: List[Union[List[int], Any, Dict[str, Any]]]) -> Dict[str, Any]:
        # Handle dict or lists with proper padding and conversion to tensor.
        if isinstance(examples[0], Mapping):
            batch = self.tokenizer.pad(examples, return_tensors="np", pad_to_multiple_of=self.pad_to_multiple_of)
        else:
            batch = {
                "input_ids": _numpy_collate_batch(examples, self.tokenizer, pad_to_multiple_of=self.pad_to_multiple_of)
            }

        # If special token mask has been preprocessed, pop it from the dict.
        special_tokens_mask = batch.pop("special_tokens_mask", None)
        if self.mlm:
            batch["input_ids"], batch["labels"] = self.numpy_mask_tokens(
                batch["input_ids"], special_tokens_mask=special_tokens_mask
            )
        else:
            labels = np.copy(batch["input_ids"])
            if self.tokenizer.pad_token_id is not None:
                labels[labels == self.tokenizer.pad_token_id] = -100
            batch["labels"] = labels
        return batch

    def numpy_mask_tokens(self, inputs: Any, special_tokens_mask: Optional[Any] = None) -> Tuple[Any, Any]:
        """
        Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original.
        """
        labels = np.copy(inputs)
        # We sample a few tokens in each sequence for MLM training (with probability `self.mlm_probability`)
        probability_matrix = np.full(labels.shape, self.mlm_probability)
        if special_tokens_mask is None:
            special_tokens_mask = [
                self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True) for val in labels.tolist()
            ]
            special_tokens_mask = np.array(special_tokens_mask, dtype=bool)
        else:
            special_tokens_mask = special_tokens_mask.astype(bool)

        probability_matrix[special_tokens_mask] = 0
        # Numpy doesn't have bernoulli, so we use a binomial with 1 trial
        masked_indices = np.random.binomial(1, probability_matrix, size=probability_matrix.shape).astype(bool)
        labels[~masked_indices] = -100  # We only compute loss on masked tokens

        # 80% of the time, we replace masked input tokens with tokenizer.mask_token ([MASK])
        indices_replaced = np.random.binomial(1, 0.8, size=labels.shape).astype(bool) & masked_indices
        inputs[indices_replaced] = self.tokenizer.mask_token_id

        # 10% of the time, we replace masked input tokens with random word
        # indices_random = torch.bernoulli(torch.full(labels.shape, 0.5)).bool() & masked_indices & ~indices_replaced
        indices_random = (
            np.random.binomial(1, 0.5, size=labels.shape).astype(bool) & masked_indices & ~indices_replaced
        )
        random_words = np.random.randint(
            low=0, high=len(self.tokenizer), size=np.count_nonzero(indices_random), dtype=np.int64
        )
        inputs[indices_random] = random_words

        # The rest of the time (10% of the time) we keep the masked input tokens unchanged
        return inputs, labels


@dataclass
class DataCollatorForWholeWordMask(DataCollatorForLanguageModeling):
    """
    Data collator used for language modeling that masks entire words.

    - collates batches of tensors, honoring their tokenizer's pad_token
    - preprocesses batches for masked language modeling

    <Tip>

    This collator relies on details of the implementation of subword tokenization by [`BertTokenizer`], specifically
    that subword tokens are prefixed with *##*. For tokenizers that do not adhere to this scheme, this collator will
    produce an output that is roughly equivalent to [`.DataCollatorForLanguageModeling`].

    </Tip>"""

    def torch_call(self, examples: List[Union[List[int], Any, Dict[str, Any]]]) -> Dict[str, Any]:
        if isinstance(examples[0], Mapping):
            input_ids = [e["input_ids"] for e in examples]
        else:
            input_ids = examples
            examples = [{"input_ids": e} for e in examples]

        batch_input = _torch_collate_batch(input_ids, self.tokenizer, pad_to_multiple_of=self.pad_to_multiple_of)

        mask_labels = []
        for e in examples:
            ref_tokens = []
            for id in tolist(e["input_ids"]):
                token = self.tokenizer._convert_id_to_token(id)
                ref_tokens.append(token)

            # For Chinese tokens, we need extra inf to mark sub-word, e.g [喜,欢]-> [喜,##欢]
            if "chinese_ref" in e:
                ref_pos = tolist(e["chinese_ref"])
                len_seq = len(e["input_ids"])
                for i in range(len_seq):
                    if i in ref_pos:
                        ref_tokens[i] = "##" + ref_tokens[i]
            mask_labels.append(self._whole_word_mask(ref_tokens))
        batch_mask = _torch_collate_batch(mask_labels, self.tokenizer, pad_to_multiple_of=self.pad_to_multiple_of)
        inputs, labels = self.torch_mask_tokens(batch_input, batch_mask)
        return {"input_ids": inputs, "labels": labels}

    def tf_call(self, examples: List[Union[List[int], Any, Dict[str, Any]]]) -> Dict[str, Any]:
        import tensorflow as tf

        if isinstance(examples[0], Mapping):
            input_ids = [e["input_ids"] for e in examples]
        else:
            input_ids = examples
            examples = [{"input_ids": e} for e in examples]

        batch_input = _tf_collate_batch(input_ids, self.tokenizer, pad_to_multiple_of=self.pad_to_multiple_of)

        mask_labels = []
        for e in examples:
            ref_tokens = []
            for id in tolist(e["input_ids"]):
                token = self.tokenizer._convert_id_to_token(id)
                ref_tokens.append(token)

            # For Chinese tokens, we need extra inf to mark sub-word, e.g [喜,欢]-> [喜,##欢]
            if "chinese_ref" in e:
                ref_pos = tolist(e["chinese_ref"])
                len_seq = len(e["input_ids"])
                for i in range(len_seq):
                    if i in ref_pos:
                        ref_tokens[i] = "##" + ref_tokens[i]
            mask_labels.append(self._whole_word_mask(ref_tokens))
        batch_mask = _tf_collate_batch(mask_labels, self.tokenizer, pad_to_multiple_of=self.pad_to_multiple_of)
        inputs, labels = self.tf_mask_tokens(tf.cast(batch_input, tf.int64), batch_mask)
        return {"input_ids": inputs, "labels": labels}

    def numpy_call(self, examples: List[Union[List[int], Any, Dict[str, Any]]]) -> Dict[str, Any]:
        if isinstance(examples[0], Mapping):
            input_ids = [e["input_ids"] for e in examples]
        else:
            input_ids = examples
            examples = [{"input_ids": e} for e in examples]

        batch_input = _numpy_collate_batch(input_ids, self.tokenizer, pad_to_multiple_of=self.pad_to_multiple_of)

        mask_labels = []
        for e in examples:
            ref_tokens = []
            for id in tolist(e["input_ids"]):
                token = self.tokenizer._convert_id_to_token(id)
                ref_tokens.append(token)

            # For Chinese tokens, we need extra inf to mark sub-word, e.g [喜,欢]-> [喜,##欢]
            if "chinese_ref" in e:
                ref_pos = tolist(e["chinese_ref"])
                len_seq = len(e["input_ids"])
                for i in range(len_seq):
                    if i in ref_pos:
                        ref_tokens[i] = "##" + ref_tokens[i]
            mask_labels.append(self._whole_word_mask(ref_tokens))
        batch_mask = _numpy_collate_batch(mask_labels, self.tokenizer, pad_to_multiple_of=self.pad_to_multiple_of)
        inputs, labels = self.numpy_mask_tokens(batch_input, batch_mask)
        return {"input_ids": inputs, "labels": labels}

    def _whole_word_mask(self, input_tokens: List[str], max_predictions=512):
        """
        Get 0/1 labels for masked tokens with whole word mask proxy
        """
        if not isinstance(self.tokenizer, (BertTokenizer, BertTokenizerFast)):
            warnings.warn(
                "DataCollatorForWholeWordMask is only suitable for BertTokenizer-like tokenizers. "
                "Please refer to the documentation for more information."
            )

        cand_indexes = []
        for i, token in enumerate(input_tokens):
            if token == "[CLS]" or token == "[SEP]":
                continue

            if len(cand_indexes) >= 1 and token.startswith("##"):
                cand_indexes[-1].append(i)
            else:
                cand_indexes.append([i])

        random.shuffle(cand_indexes)
        num_to_predict = min(max_predictions, max(1, int(round(len(input_tokens) * self.mlm_probability))))
        masked_lms = []
        covered_indexes = set()
        for index_set in cand_indexes:
            if len(masked_lms) >= num_to_predict:
                break
            # If adding a whole-word mask would exceed the maximum number of
            # predictions, then just skip this candidate.
            if len(masked_lms) + len(index_set) > num_to_predict:
                continue
            is_any_index_covered = False
            for index in index_set:
                if index in covered_indexes:
                    is_any_index_covered = True
                    break
            if is_any_index_covered:
                continue
            for index in index_set:
                covered_indexes.add(index)
                masked_lms.append(index)

        if len(covered_indexes) != len(masked_lms):
            raise ValueError("Length of covered_indexes is not equal to length of masked_lms.")
        mask_labels = [1 if i in covered_indexes else 0 for i in range(len(input_tokens))]
        return mask_labels

    def torch_mask_tokens(self, inputs: Any, mask_labels: Any) -> Tuple[Any, Any]:
        """
        Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original. Set
        'mask_labels' means we use whole word mask (wwm), we directly mask idxs according to it's ref.
        """
        import torch

        if self.tokenizer.mask_token is None:
            raise ValueError(
                "This tokenizer does not have a mask token which is necessary for masked language modeling. Remove the"
                " --mlm flag if you want to use this tokenizer."
            )
        labels = inputs.clone()
        # We sample a few tokens in each sequence for masked-LM training (with probability args.mlm_probability defaults to 0.15 in Bert/RoBERTa)

        probability_matrix = mask_labels

        special_tokens_mask = [
            self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True) for val in labels.tolist()
        ]
        probability_matrix.masked_fill_(torch.tensor(special_tokens_mask, dtype=torch.bool), value=0.0)
        if self.tokenizer._pad_token is not None:
            padding_mask = labels.eq(self.tokenizer.pad_token_id)
            probability_matrix.masked_fill_(padding_mask, value=0.0)

        masked_indices = probability_matrix.bool()
        labels[~masked_indices] = -100  # We only compute loss on masked tokens

        # 80% of the time, we replace masked input tokens with tokenizer.mask_token ([MASK])
        indices_replaced = torch.bernoulli(torch.full(labels.shape, 0.8)).bool() & masked_indices
        inputs[indices_replaced] = self.tokenizer.convert_tokens_to_ids(self.tokenizer.mask_token)

        # 10% of the time, we replace masked input tokens with random word
        indices_random = torch.bernoulli(torch.full(labels.shape, 0.5)).bool() & masked_indices & ~indices_replaced
        random_words = torch.randint(len(self.tokenizer), labels.shape, dtype=torch.long)
        inputs[indices_random] = random_words[indices_random]

        # The rest of the time (10% of the time) we keep the masked input tokens unchanged
        return inputs, labels

    def tf_mask_tokens(self, inputs: Any, mask_labels: Any) -> Tuple[Any, Any]:
        """
        Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original. Set
        'mask_labels' means we use whole word mask (wwm), we directly mask idxs according to it's ref.
        """
        import tensorflow as tf

        input_shape = tf.shape(inputs)
        if self.tokenizer.mask_token is None:
            raise ValueError(
                "This tokenizer does not have a mask token which is necessary for masked language modeling. Remove the"
                " --mlm flag if you want to use this tokenizer."
            )
        labels = tf.identity(inputs)
        # We sample a few tokens in each sequence for masked-LM training (with probability args.mlm_probability defaults to 0.15 in Bert/RoBERTa)

        masked_indices = tf.cast(mask_labels, tf.bool)

        special_tokens_mask = [
            self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True) for val in labels
        ]
        masked_indices = masked_indices & ~tf.cast(special_tokens_mask, dtype=tf.bool)
        if self.tokenizer._pad_token is not None:
            padding_mask = inputs == self.tokenizer.pad_token_id
            masked_indices = masked_indices & ~padding_mask

        # Replace unmasked indices with -100 in the labels since we only compute loss on masked tokens
        labels = tf.where(masked_indices, inputs, -100)

        # 80% of the time, we replace masked input tokens with tokenizer.mask_token ([MASK])
        indices_replaced = self.tf_bernoulli(input_shape, 0.8) & masked_indices

        inputs = tf.where(indices_replaced, self.tokenizer.mask_token_id, inputs)

        # 10% of the time, we replace masked input tokens with random word
        indices_random = self.tf_bernoulli(input_shape, 0.5) & masked_indices & ~indices_replaced
        random_words = tf.random.uniform(input_shape, maxval=len(self.tokenizer), dtype=tf.int64)
        inputs = tf.where(indices_random, random_words, inputs)

        # The rest of the time (10% of the time) we keep the masked input tokens unchanged
        return inputs, labels

    def numpy_mask_tokens(self, inputs: Any, mask_labels: Any) -> Tuple[Any, Any]:
        """
        Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original. Set
        'mask_labels' means we use whole word mask (wwm), we directly mask idxs according to it's ref.
        """
        if self.tokenizer.mask_token is None:
            raise ValueError(
                "This tokenizer does not have a mask token which is necessary for masked language modeling. Remove the"
                " --mlm flag if you want to use this tokenizer."
            )
        labels = np.copy(inputs)
        # We sample a few tokens in each sequence for masked-LM training (with probability args.mlm_probability defaults to 0.15 in Bert/RoBERTa)

        masked_indices = mask_labels.astype(bool)

        special_tokens_mask = [
            self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True) for val in labels.tolist()
        ]
        masked_indices[np.array(special_tokens_mask, dtype=bool)] = 0
        if self.tokenizer._pad_token is not None:
            padding_mask = labels == self.tokenizer.pad_token_id
            masked_indices[padding_mask] = 0

        labels[~masked_indices] = -100  # We only compute loss on masked tokens

        # 80% of the time, we replace masked input tokens with tokenizer.mask_token ([MASK])
        indices_replaced = np.random.binomial(1, 0.8, size=labels.shape).astype(bool) & masked_indices
        inputs[indices_replaced] = self.tokenizer.convert_tokens_to_ids(self.tokenizer.mask_token)

        # 10% of the time, we replace masked input tokens with random word
        # indices_random = torch.bernoulli(torch.full(labels.shape, 0.5)).bool() & masked_indices & ~indices_replaced
        indices_random = (
            np.random.binomial(1, 0.5, size=labels.shape).astype(bool) & masked_indices & ~indices_replaced
        )
        random_words = np.random.randint(low=0, high=len(self.tokenizer), size=labels.shape, dtype=np.int64)
        inputs[indices_random] = random_words[indices_random]

        # The rest of the time (10% of the time) we keep the masked input tokens unchanged
        return inputs, labels


@dataclass
class DataCollatorForSOP(DataCollatorForLanguageModeling):
    """
    Data collator used for sentence order prediction task.

    - collates batches of tensors, honoring their tokenizer's pad_token
    - preprocesses batches for both masked language modeling and sentence order prediction
    """

    def __init__(self, *args, **kwargs):
        warnings.warn(
            "DataCollatorForSOP is deprecated and will be removed in a future version, you can now use "
            "DataCollatorForLanguageModeling instead.",
            FutureWarning,
        )

    def __call__(self, examples: List[Dict[str, Any]]) -> Dict[str, Any]:
        import torch
        from torch.nn.utils.rnn import pad_sequence

        input_ids = [example["input_ids"] for example in examples]
        input_ids = _torch_collate_batch(input_ids, self.tokenizer)
        input_ids, labels, attention_mask = self.mask_tokens(input_ids)

        token_type_ids = [example["token_type_ids"] for example in examples]
        # size of segment_ids varied because randomness, padding zero to the end as the original implementation
        token_type_ids = pad_sequence(token_type_ids, batch_first=True, padding_value=self.tokenizer.pad_token_id)

        sop_label_list = [example["sentence_order_label"] for example in examples]
        sentence_order_label = torch.stack(sop_label_list)

        return {
            "input_ids": input_ids,
            "labels": labels,
            "attention_mask": attention_mask,
            "token_type_ids": token_type_ids,
            "sentence_order_label": sentence_order_label,
        }

    def mask_tokens(self, inputs: Any) -> Tuple[Any, Any, Any]:
        """
        Prepare masked tokens inputs/labels/attention_mask for masked language modeling: 80% MASK, 10% random, 10%
        original. N-gram not applied yet.
        """
        import torch

        if self.tokenizer.mask_token is None:
            raise ValueError(
                "This tokenizer does not have a mask token which is necessary for masked language modeling. Remove the"
                " --mlm flag if you want to use this tokenizer."
            )

        labels = inputs.clone()
        # We sample a few tokens in each sequence for masked-LM training (with probability args.mlm_probability defaults to 0.15 in Bert/RoBERTa)
        probability_matrix = torch.full(labels.shape, self.mlm_probability)
        special_tokens_mask = [
            self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True) for val in labels.tolist()
        ]
        probability_matrix.masked_fill_(torch.tensor(special_tokens_mask, dtype=torch.bool), value=0.0)
        if self.tokenizer._pad_token is not None:
            padding_mask = labels.eq(self.tokenizer.pad_token_id)
            probability_matrix.masked_fill_(padding_mask, value=0.0)
        masked_indices = torch.bernoulli(probability_matrix).bool()
        # probability be `1` (masked), however in albert model attention mask `0` means masked, revert the value
        attention_mask = (~masked_indices).float()
        if self.tokenizer._pad_token is not None:
            attention_padding_mask = labels.eq(self.tokenizer.pad_token_id)
            attention_mask.masked_fill_(attention_padding_mask, value=1.0)
        labels[~masked_indices] = -100  # We only compute loss on masked tokens, -100 is default for CE compute

        # 80% of the time, we replace masked input tokens with tokenizer.mask_token ([MASK])
        indices_replaced = torch.bernoulli(torch.full(labels.shape, 0.8)).bool() & masked_indices
        inputs[indices_replaced] = self.tokenizer.convert_tokens_to_ids(self.tokenizer.mask_token)

        # 10% of the time, we replace masked input tokens with random word
        indices_random = torch.bernoulli(torch.full(labels.shape, 0.5)).bool() & masked_indices & ~indices_replaced
        random_words = torch.randint(len(self.tokenizer), labels.shape, dtype=torch.long)
        inputs[indices_random] = random_words[indices_random]

        # The rest of the time (10% of the time) we keep the masked input tokens unchanged
        return inputs, labels, attention_mask


@dataclass
class DataCollatorForPermutationLanguageModeling(DataCollatorMixin):
    """
    Data collator used for permutation language modeling.

    - collates batches of tensors, honoring their tokenizer's pad_token
    - preprocesses batches for permutation language modeling with procedures specific to XLNet
    """

    tokenizer: PreTrainedTokenizerBase
    plm_probability: float = 1 / 6
    max_span_length: int = 5  # maximum length of a span of masked tokens
    return_tensors: str = "pt"

    def torch_call(self, examples: List[Union[List[int], Any, Dict[str, Any]]]) -> Dict[str, Any]:
        if isinstance(examples[0], Mapping):
            examples = [e["input_ids"] for e in examples]
        batch = _torch_collate_batch(examples, self.tokenizer)
        inputs, perm_mask, target_mapping, labels = self.torch_mask_tokens(batch)
        return {"input_ids": inputs, "perm_mask": perm_mask, "target_mapping": target_mapping, "labels": labels}

    def tf_call(self, examples: List[Union[List[int], Any, Dict[str, Any]]]) -> Dict[str, Any]:
        if isinstance(examples[0], Mapping):
            examples = [e["input_ids"] for e in examples]
        batch = _tf_collate_batch(examples, self.tokenizer)
        inputs, perm_mask, target_mapping, labels = self.tf_mask_tokens(batch)
        return {"input_ids": inputs, "perm_mask": perm_mask, "target_mapping": target_mapping, "labels": labels}

    def numpy_call(self, examples: List[Union[List[int], Any, Dict[str, Any]]]) -> Dict[str, Any]:
        if isinstance(examples[0], Mapping):
            examples = [e["input_ids"] for e in examples]
        batch = _numpy_collate_batch(examples, self.tokenizer)
        inputs, perm_mask, target_mapping, labels = self.numpy_mask_tokens(batch)
        return {"input_ids": inputs, "perm_mask": perm_mask, "target_mapping": target_mapping, "labels": labels}

    def torch_mask_tokens(self, inputs: Any) -> Tuple[Any, Any, Any, Any]:
        """
        The masked tokens to be predicted for a particular sequence are determined by the following algorithm:

            0. Start from the beginning of the sequence by setting `cur_len = 0` (number of tokens processed so far).
            1. Sample a `span_length` from the interval `[1, max_span_length]` (length of span of tokens to be masked)
            2. Reserve a context of length `context_length = span_length / plm_probability` to surround span to be
               masked
            3. Sample a starting point `start_index` from the interval `[cur_len, cur_len + context_length -
               span_length]` and mask tokens `start_index:start_index + span_length`
            4. Set `cur_len = cur_len + context_length`. If `cur_len < max_len` (i.e. there are tokens remaining in the
               sequence to be processed), repeat from Step 1.
        """
        import torch

        if self.tokenizer.mask_token is None:
            raise ValueError(
                "This tokenizer does not have a mask token which is necessary for permutation language modeling."
                " Please add a mask token if you want to use this tokenizer."
            )

        if inputs.size(1) % 2 != 0:
            raise ValueError(
                "This collator requires that sequence lengths be even to create a leakage-free perm_mask. Please see"
                " relevant comments in source code for details."
            )

        labels = inputs.clone()
        # Creating the mask and target_mapping tensors
        masked_indices = torch.full(labels.shape, 0, dtype=torch.bool)
        target_mapping = torch.zeros((labels.size(0), labels.size(1), labels.size(1)), dtype=torch.float32)

        for i in range(labels.size(0)):
            # Start from the beginning of the sequence by setting `cur_len = 0` (number of tokens processed so far).
            cur_len = 0
            max_len = labels.size(1)

            while cur_len < max_len:
                # Sample a `span_length` from the interval `[1, max_span_length]` (length of span of tokens to be masked)
                span_length = torch.randint(1, self.max_span_length + 1, (1,)).item()
                # Reserve a context of length `context_length = span_length / plm_probability` to surround the span to be masked
                context_length = int(span_length / self.plm_probability)
                # Sample a starting point `start_index` from the interval `[cur_len, cur_len + context_length - span_length]` and mask tokens `start_index:start_index + span_length`
                start_index = cur_len + torch.randint(context_length - span_length + 1, (1,)).item()
                masked_indices[i, start_index : start_index + span_length] = 1
                # Set `cur_len = cur_len + context_length`
                cur_len += context_length

            # Since we're replacing non-masked tokens with -100 in the labels tensor instead of skipping them altogether,
            # the i-th predict corresponds to the i-th token.
            target_mapping[i] = torch.eye(labels.size(1))

        special_tokens_mask = torch.tensor(
            [self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True) for val in labels.tolist()],
            dtype=torch.bool,
        )
        masked_indices.masked_fill_(special_tokens_mask, value=0.0)
        if self.tokenizer._pad_token is not None:
            padding_mask = labels.eq(self.tokenizer.pad_token_id)
            masked_indices.masked_fill_(padding_mask, value=0.0)

        # Mask indicating non-functional tokens, where functional tokens are [SEP], [CLS], padding, etc.
        non_func_mask = ~(padding_mask | special_tokens_mask)

        inputs[masked_indices] = self.tokenizer.mask_token_id
        labels[~masked_indices] = -100  # We only compute loss on masked tokens

        perm_mask = torch.zeros((labels.size(0), labels.size(1), labels.size(1)), dtype=torch.float32)

        for i in range(labels.size(0)):
            # Generate permutation indices i.e. sample a random factorisation order for the sequence. This will
            # determine which tokens a given token can attend to (encoded in `perm_mask`).
            # Note: Length of token sequence being permuted has to be less than or equal to reused sequence length
            # (see documentation for `mems`), otherwise information may leak through due to reuse. In this implementation,
            # we assume that reused length is half of sequence length and permutation length is equal to reused length.
            # This requires that the sequence length be even.

            # Create a linear factorisation order
            perm_index = torch.arange(labels.size(1))
            # Split this into two halves, assuming that half the sequence is reused each time
            perm_index = perm_index.reshape((-1, labels.size(1) // 2)).transpose(0, 1)
            # Permute the two halves such that they do not cross over
            perm_index = perm_index[torch.randperm(labels.size(1) // 2)]
            # Flatten this out into the desired permuted factorisation order
            perm_index = torch.flatten(perm_index.transpose(0, 1))
            # Set the permutation indices of non-masked (non-functional) tokens to the
            # smallest index (-1) so that:
            # (1) They can be seen by all other positions
            # (2) They cannot see masked positions, so there won't be information leak
            perm_index.masked_fill_(~masked_indices[i] & non_func_mask[i], -1)
            # The logic for whether the i-th token can attend on the j-th token based on the factorisation order:
            # 0 (can attend): If perm_index[i] > perm_index[j] or j is neither masked nor a functional token
            # 1 (cannot attend): If perm_index[i] <= perm_index[j] and j is either masked or a functional token
            perm_mask[i] = (
                perm_index.reshape((labels.size(1), 1)) <= perm_index.reshape((1, labels.size(1)))
            ) & masked_indices[i]

        return inputs.long(), perm_mask, target_mapping, labels.long()

    def tf_mask_tokens(self, inputs: Any) -> Tuple[Any, Any, Any, Any]:
        """
        The masked tokens to be predicted for a particular sequence are determined by the following algorithm:

            0. Start from the beginning of the sequence by setting `cur_len = 0` (number of tokens processed so far).
            1. Sample a `span_length` from the interval `[1, max_span_length]` (length of span of tokens to be masked)
            2. Reserve a context of length `context_length = span_length / plm_probability` to surround span to be
               masked
            3. Sample a starting point `start_index` from the interval `[cur_len, cur_len + context_length -
               span_length]` and mask tokens `start_index:start_index + span_length`
            4. Set `cur_len = cur_len + context_length`. If `cur_len < max_len` (i.e. there are tokens remaining in the
               sequence to be processed), repeat from Step 1.
        """
        import tensorflow as tf

        if self.tokenizer.mask_token is None:
            raise ValueError(
                "This tokenizer does not have a mask token which is necessary for permutation language modeling."
                " Please add a mask token if you want to use this tokenizer."
            )

        if tf.shape(inputs)[1] % 2 != 0:
            raise ValueError(
                "This collator requires that sequence lengths be even to create a leakage-free perm_mask. Please see"
                " relevant comments in source code for details."
            )

        labels = tf.identity(inputs)
        # Creating the mask and target_mapping tensors
        masked_indices = np.full(labels.shape.as_list(), 0, dtype=bool)
        labels_shape = tf.shape(labels)
        target_mapping = np.zeros((labels_shape[0], labels_shape[1], labels_shape[1]), dtype=np.float32)

        for i in range(len(labels)):
            # Start from the beginning of the sequence by setting `cur_len = 0` (number of tokens processed so far).
            cur_len = 0
            max_len = tf.shape(labels)[1]

            while cur_len < max_len:
                # Sample a `span_length` from the interval `[1, max_span_length]` (length of span of tokens to be masked)
                span_length = randint(1, self.max_span_length + 1)
                # Reserve a context of length `context_length = span_length / plm_probability` to surround the span to be masked
                context_length = int(span_length / self.plm_probability)
                # Sample a starting point `start_index` from the interval `[cur_len, cur_len + context_length - span_length]` and mask tokens `start_index:start_index + span_length`
                start_index = cur_len + randint(0, context_length - span_length + 1)
                masked_indices[i, start_index : start_index + span_length] = 1
                # Set `cur_len = cur_len + context_length`
                cur_len += context_length

            # Since we're replacing non-masked tokens with -100 in the labels tensor instead of skipping them altogether,
            # the i-th predict corresponds to the i-th token.
            target_mapping[i] = np.eye(labels_shape[1])
        masked_indices = tf.cast(tf.convert_to_tensor(masked_indices), dtype=tf.bool)
        target_mapping = tf.convert_to_tensor(target_mapping)
        special_tokens_mask = tf.convert_to_tensor(
            [
                self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True)
                for val in labels.numpy().tolist()
            ],
        )
        special_tokens_mask = tf.cast(special_tokens_mask, dtype=tf.bool)
        masked_indices = masked_indices & ~special_tokens_mask
        if self.tokenizer._pad_token is not None:
            padding_mask = labels == self.tokenizer.pad_token_id
            masked_indices = masked_indices & ~padding_mask

        # Mask indicating non-functional tokens, where functional tokens are [SEP], [CLS], padding, etc.
        non_func_mask = ~(padding_mask | special_tokens_mask)

        inputs = tf.where(masked_indices, self.tokenizer.mask_token_id, inputs)
        labels = tf.where(masked_indices, labels, -100)  # We only compute loss on masked tokens

        perm_mask = []

        for i in range(len(labels)):
            # Generate permutation indices i.e. sample a random factorisation order for the sequence. This will
            # determine which tokens a given token can attend to (encoded in `perm_mask`).
            # Note: Length of token sequence being permuted has to be less than or equal to reused sequence length
            # (see documentation for `mems`), otherwise information may leak through due to reuse. In this implementation,
            # we assume that reused length is half of sequence length and permutation length is equal to reused length.
            # This requires that the sequence length be even.

            # Create a linear factorisation order
            # tf.range is the equivalent of torch.arange
            perm_index = tf.range(labels_shape[1])
            # Split this into two halves, assuming that half the sequence is reused each time
            perm_index = tf.transpose(tf.reshape(perm_index, (-1, labels_shape[1] // 2)))
            # Permute the two halves such that they do not cross over
            perm_index = tf.random.shuffle(perm_index)  # Shuffles along the first dimension
            # Flatten this out into the desired permuted factorisation order
            perm_index = tf.reshape(tf.transpose(perm_index), (-1,))
            # Set the permutation indices of non-masked (non-functional) tokens to the
            # smallest index (-1) so that:
            # (1) They can be seen by all other positions
            # (2) They cannot see masked positions, so there won't be information leak
            perm_index = tf.where(~masked_indices[i] & non_func_mask[i], -1, perm_index)
            # The logic for whether the i-th token can attend on the j-th token based on the factorisation order:
            # 0 (can attend): If perm_index[i] > perm_index[j] or j is neither masked nor a functional token
            # 1 (cannot attend): If perm_index[i] <= perm_index[j] and j is either masked or a functional token
            perm_mask.append(
                (tf.reshape(perm_index, (labels_shape[1], 1)) <= tf.reshape(perm_index, (1, labels_shape[1])))
                & masked_indices[i]
            )
        perm_mask = tf.stack(perm_mask, axis=0)

        return tf.cast(inputs, tf.int64), tf.cast(perm_mask, tf.float32), target_mapping, tf.cast(labels, tf.int64)

    def numpy_mask_tokens(self, inputs: Any) -> Tuple[Any, Any, Any, Any]:
        """
        The masked tokens to be predicted for a particular sequence are determined by the following algorithm:

            0. Start from the beginning of the sequence by setting `cur_len = 0` (number of tokens processed so far).
            1. Sample a `span_length` from the interval `[1, max_span_length]` (length of span of tokens to be masked)
            2. Reserve a context of length `context_length = span_length / plm_probability` to surround span to be
               masked
            3. Sample a starting point `start_index` from the interval `[cur_len, cur_len + context_length -
               span_length]` and mask tokens `start_index:start_index + span_length`
            4. Set `cur_len = cur_len + context_length`. If `cur_len < max_len` (i.e. there are tokens remaining in the
               sequence to be processed), repeat from Step 1.
        """
        if self.tokenizer.mask_token is None:
            raise ValueError(
                "This tokenizer does not have a mask token which is necessary for permutation language modeling."
                " Please add a mask token if you want to use this tokenizer."
            )

        if inputs.shape[1] % 2 != 0:
            raise ValueError(
                "This collator requires that sequence lengths be even to create a leakage-free perm_mask. Please see"
                " relevant comments in source code for details."
            )

        labels = np.copy(inputs)
        # Creating the mask and target_mapping tensors
        masked_indices = np.full(labels.shape, 0, dtype=bool)
        target_mapping = np.zeros((labels.shape[0], labels.shape[1], labels.shape[1]), dtype=np.float32)

        for i in range(labels.shape[0]):
            # Start from the beginning of the sequence by setting `cur_len = 0` (number of tokens processed so far).
            cur_len = 0
            max_len = labels.shape[1]

            while cur_len < max_len:
                # Sample a `span_length` from the interval `[1, max_span_length]` (length of span of tokens to be masked)
                span_length = randint(1, self.max_span_length + 1)
                # Reserve a context of length `context_length = span_length / plm_probability` to surround the span to be masked
                context_length = int(span_length / self.plm_probability)
                # Sample a starting point `start_index` from the interval `[cur_len, cur_len + context_length - span_length]` and mask tokens `start_index:start_index + span_length`
                start_index = cur_len + randint(0, context_length - span_length + 1)
                masked_indices[i, start_index : start_index + span_length] = 1
                # Set `cur_len = cur_len + context_length`
                cur_len += context_length

            # Since we're replacing non-masked tokens with -100 in the labels tensor instead of skipping them altogether,
            # the i-th predict corresponds to the i-th token.
            target_mapping[i] = np.eye(labels.shape[1])

        special_tokens_mask = np.array(
            [self.tokenizer.get_special_tokens_mask(val, already_has_special_tokens=True) for val in labels.tolist()],
            dtype=bool,
        )
        masked_indices[special_tokens_mask] = 0
        if self.tokenizer._pad_token is not None:
            padding_mask = labels == self.tokenizer.pad_token_id
            masked_indices[padding_mask] = 0.0

        # Mask indicating non-functional tokens, where functional tokens are [SEP], [CLS], padding, etc.
        non_func_mask = ~(padding_mask | special_tokens_mask)

        inputs[masked_indices] = self.tokenizer.mask_token_id
        labels[~masked_indices] = -100  # We only compute loss on masked tokens

        perm_mask = np.zeros((labels.shape[0], labels.shape[1], labels.shape[1]), dtype=np.float32)

        for i in range(labels.shape[0]):
            # Generate permutation indices i.e. sample a random factorisation order for the sequence. This will
            # determine which tokens a given token can attend to (encoded in `perm_mask`).
            # Note: Length of token sequence being permuted has to be less than or equal to reused sequence length
            # (see documentation for `mems`), otherwise information may leak through due to reuse. In this implementation,
            # we assume that reused length is half of sequence length and permutation length is equal to reused length.
            # This requires that the sequence length be even.

            # Create a linear factorisation order
            perm_index = np.arange(labels.shape[1])
            # Split this into two halves, assuming that half the sequence is reused each time
            perm_index = perm_index.reshape((-1, labels.shape[1] // 2)).T
            # Permute the two halves such that they do not cross over
            np.random.shuffle(perm_index)
            # Flatten this out into the desired permuted factorisation order
            perm_index = perm_index.T.flatten()
            # Set the permutation indices of non-masked (non-functional) tokens to the
            # smallest index (-1) so that:
            # (1) They can be seen by all other positions
            # (2) They cannot see masked positions, so there won't be information leak
            perm_index[~masked_indices[i] & non_func_mask[i]] = -1
            # The logic for whether the i-th token can attend on the j-th token based on the factorisation order:
            # 0 (can attend): If perm_index[i] > perm_index[j] or j is neither masked nor a functional token
            # 1 (cannot attend): If perm_index[i] <= perm_index[j] and j is either masked or a functional token
            perm_mask[i] = (
                perm_index.reshape((labels.shape[1], 1)) <= perm_index.reshape((1, labels.shape[1]))
            ) & masked_indices[i]

        return inputs.astype(np.int64), perm_mask, target_mapping, labels.astype(np.int64)