File size: 88,459 Bytes
06ba6ce
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
# coding=utf-8
# Copyright 2023 The Intel Labs Team Authors, The Microsoft Research Team Authors and HuggingFace Inc. 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.
"""PyTorch BridgeTower Model"""

import math
from collections import OrderedDict
from dataclasses import dataclass
from typing import List, Optional, Tuple, Union

import torch
import torch.utils.checkpoint
from torch import nn
from torch.nn import CrossEntropyLoss

from ...activations import ACT2FN, QuickGELUActivation
from ...modeling_outputs import (
    BaseModelOutputWithPastAndCrossAttentions,
    BaseModelOutputWithPoolingAndCrossAttentions,
    MaskedLMOutput,
    ModelOutput,
    SequenceClassifierOutput,
)
from ...modeling_utils import PreTrainedModel, apply_chunking_to_forward
from ...pytorch_utils import find_pruneable_heads_and_indices, prune_linear_layer
from ...utils import add_start_docstrings, add_start_docstrings_to_model_forward, logging, replace_return_docstrings
from .configuration_bridgetower import BridgeTowerConfig, BridgeTowerTextConfig, BridgeTowerVisionConfig


logger = logging.get_logger(__name__)

_CONFIG_FOR_DOC = "BridgeTowerConfig"
_CHECKPOINT_FOR_DOC = "BridgeTower/bridgetower-base"
_TOKENIZER_FOR_DOC = "RobertaTokenizer"

BRIDGETOWER_PRETRAINED_MODEL_ARCHIVE_LIST = [
    "BridgeTower/bridgetower-base",
    "BridgeTower/bridgetower-base-itm-mlm"
    # See all bridgetower models at https://huggingface.co/BridgeTower
]


BRIDGETOWER_START_DOCSTRING = r"""
    This model is a PyTorch `torch.nn.Module <https://pytorch.org/docs/stable/nn.html#torch.nn.Module>`_ subclass. Use
    it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and
    behavior.

    Parameters:
        config ([`BridgeTowerConfig`]): Model configuration class with all the parameters of the model.
            Initializing with a config file does not load the weights associated with the model, only the
            configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights.
"""

BRIDGETOWER_INPUTS_DOCSTRING = r"""
    Args:
        input_ids (`torch.LongTensor` of shape `({0})`):
            Indices of input sequence tokens in the vocabulary. Indices can be obtained using [`AutoTokenizer`]. See
            [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. [What are input
            IDs?](../glossary#input-ids)

        attention_mask (`torch.FloatTensor` of shape `({0})`, *optional*):
            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
            - 1 for tokens that are **not masked**,
            - 0 for tokens that are **masked**.
            [What are attention masks?](../glossary#attention-mask)

        token_type_ids (`torch.LongTensor` of shape `({0})`, *optional*):
            Segment token indices to indicate first and second portions of the inputs. Indices are selected in `[0,
            1]`:
            - 0 corresponds to a *sentence A* token,
            - 1 corresponds to a *sentence B* token.
            [What are token type IDs?](../glossary#token-type-ids)

        pixel_values (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`):
            Pixel values. Pixel values can be obtained using [`BridgeTowerImageProcessor`]. See
            [`BridgeTowerImageProcessor.__call__`] for details.

        pixel_mask (`torch.LongTensor` of shape `(batch_size, height, width)`, *optional*):
            Mask to avoid performing attention on padding pixel values. Mask values selected in `[0, 1]`:

            - 1 for pixels that are real (i.e. **not masked**),
            - 0 for pixels that are padding (i.e. **masked**).
            `What are attention masks? <../glossary.html#attention-mask>`__

        head_mask (`torch.FloatTensor` of shape `(num_heads,)` or `(num_layers, num_heads)`, *optional*):
            Mask to nullify selected heads of the self-attention modules. Mask values selected in `[0, 1]`:
            - 1 indicates the head is **not masked**,
            - 0 indicates the head is **masked**.

        inputs_embeds (`torch.FloatTensor` of shape `({0}, hidden_size)`, *optional*):
            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
            model's internal embedding lookup matrix.

        image_embeds (`torch.FloatTensor` of shape `(batch_size, num_patches, hidden_size)`, *optional*):
            Optionally, instead of passing `pixel_values`, you can choose to directly pass an embedded representation.
            This is useful if you want more control over how to convert `pixel_values` into patch embeddings.

        image_token_type_idx (`int`, *optional*):
            - The token type ids for images.

        output_attentions (`bool`, *optional*):
            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
            tensors for more detail.

        output_hidden_states (`bool`, *optional*):
            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
            more detail.
        return_dict (`bool`, *optional*):
            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
"""


@dataclass
class BridgeTowerModelOutput(ModelOutput):
    """
    Output type of [`BridgeTowerModel`].

    Args:
        text_features (`torch.FloatTensor` of shape `(batch_size, text_sequence_length, hidden_size)`):
            Sequence of hidden-states at the text output of the last layer of the model.
        image_features (`torch.FloatTensor` of shape `(batch_size, image_sequence_length, hidden_size)`):
            Sequence of hidden-states at the image output of the last layer of the model.
        pooler_output (`torch.FloatTensor` of shape `(batch_size, hidden_size x 2)`):
            Concatenation of last layer hidden-state of the first token of the text and image sequence (classification
            token), respectively, after further processing through layers used for auxiliary pretraining tasks.
        hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
            Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, +
            one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`. Hidden-states of
            the model at the output of each layer plus the optional initial embedding outputs.
        attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
            Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
            sequence_length)`.

            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
            heads.
    """

    text_features: torch.FloatTensor = None
    image_features: torch.FloatTensor = None
    pooler_output: torch.FloatTensor = None
    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
    attentions: Optional[Tuple[torch.FloatTensor]] = None


@dataclass
class BridgeTowerContrastiveOutput(ModelOutput):
    """
    Output type of ['BridgeTowerForContrastiveLearning']

    Args:
        loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `return_loss` is `True`:
            Image-text contrastive loss.
        logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`):
            Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
        text_embeds (`torch.FloatTensor)`, *optional*, returned when model is initialized with `with_projection=True`):
            The text embeddings obtained by applying the projection layer to the pooler_output.
        image_embeds (`torch.FloatTensor)`, *optional*, returned when model is initialized with `with_projection=True`):
            The image embeddings obtained by applying the projection layer to the pooler_output.
        cross_embeds  (`torch.FloatTensor)`, *optional*, returned when model is initialized with `with_projection=True`):
            The text-image cross-modal embeddings obtained by applying the projection layer to the pooler_output.
        hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
            Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, +
            one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`. Hidden-states of
            the model at the output of each layer plus the optional initial embedding outputs.
        attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
            Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
            sequence_length)`.
    """

    loss: Optional[torch.FloatTensor] = None
    logits: torch.FloatTensor = None
    text_embeds: Optional[Tuple[torch.FloatTensor]] = None
    image_embeds: Optional[Tuple[torch.FloatTensor]] = None
    cross_embeds: Optional[Tuple[torch.FloatTensor]] = None
    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
    attentions: Optional[Tuple[torch.FloatTensor]] = None


class BridgeTowerResidualAttention(nn.Module):
    def __init__(self, config):
        super().__init__()

        self.attn = nn.MultiheadAttention(config.hidden_size, config.hidden_size // 64)
        self.ln_1 = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
        self.mlp = nn.ModuleDict(
            OrderedDict(
                [
                    ("c_fc", nn.Linear(config.hidden_size, config.hidden_size * 4)),
                    ("gelu", QuickGELUActivation()),
                    ("c_proj", nn.Linear(config.hidden_size * 4, config.hidden_size)),
                ]
            )
        )
        self.ln_2 = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
        self.attn_mask = None

    def attention(self, hidden_state: torch.Tensor, attention_mask: torch.Tensor):
        if attention_mask is not None:
            attention_mask = attention_mask.to(dtype=torch.bool, device=hidden_state.device)
        self.attn_mask = (
            self.attn_mask.to(dtype=hidden_state.dtype, device=hidden_state.device)
            if self.attn_mask is not None
            else None
        )
        return self.attn(
            hidden_state,
            hidden_state,
            hidden_state,
            need_weights=False,
            attn_mask=self.attn_mask,
            key_padding_mask=attention_mask,
        )[0]

    def forward(self, hidden_state: torch.Tensor, attention_mask: torch.Tensor = None):
        residual_state = hidden_state + self.attention(self.ln_1(hidden_state), attention_mask)
        hidden_state = self.ln_2(residual_state)
        for _, layer in self.mlp.items():
            hidden_state = layer(hidden_state)
        hidden_state = residual_state + hidden_state
        return hidden_state


class BridgeTowerTransformer(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.hidden_size = config.hidden_size
        self.num_hidden_layers = config.num_hidden_layers
        if config.remove_last_layer:
            self.resblocks = nn.ModuleList(
                [BridgeTowerResidualAttention(config) for _ in range(self.num_hidden_layers - 1)]
            )
        else:
            self.resblocks = nn.ModuleList(
                [BridgeTowerResidualAttention(config) for _ in range(self.num_hidden_layers)]
            )
        self.stop_gradient = config.stop_gradient

    def forward(self, hidden_state: torch.Tensor, attention_mask: Optional[torch.Tensor] = None):
        hidden_states = []
        for block in self.resblocks:
            hidden_state = block(hidden_state, attention_mask)
            if self.stop_gradient:
                hidden_states.append(hidden_state.detach())
            else:
                hidden_states.append(hidden_state)
        return hidden_states


# Copied from transformers.models.clip.modeling_clip.CLIPVisionEmbeddings with CLIP->BridgeTower
class BridgeTowerVisionEmbeddings(nn.Module):
    def __init__(self, config: BridgeTowerVisionConfig):
        super().__init__()
        self.config = config
        self.embed_dim = config.hidden_size
        self.image_size = config.image_size
        self.patch_size = config.patch_size

        self.class_embedding = nn.Parameter(torch.randn(self.embed_dim))

        self.patch_embedding = nn.Conv2d(
            in_channels=config.num_channels,
            out_channels=self.embed_dim,
            kernel_size=self.patch_size,
            stride=self.patch_size,
            bias=False,
        )

        self.num_patches = (self.image_size // self.patch_size) ** 2
        self.num_positions = self.num_patches + 1
        self.position_embedding = nn.Embedding(self.num_positions, self.embed_dim)
        self.register_buffer("position_ids", torch.arange(self.num_positions).expand((1, -1)), persistent=False)

    def forward(self, pixel_values: torch.FloatTensor) -> torch.Tensor:
        batch_size = pixel_values.shape[0]
        target_dtype = self.patch_embedding.weight.dtype
        patch_embeds = self.patch_embedding(pixel_values.to(dtype=target_dtype))  # shape = [*, width, grid, grid]
        patch_embeds = patch_embeds.flatten(2).transpose(1, 2)

        class_embeds = self.class_embedding.expand(batch_size, 1, -1)
        embeddings = torch.cat([class_embeds, patch_embeds], dim=1)
        embeddings = embeddings + self.position_embedding(self.position_ids)
        return embeddings


class BridgeTowerVisionTransformer(nn.Module):
    def __init__(self, config):
        super().__init__()

        self.embeddings = BridgeTowerVisionEmbeddings(config)
        self.ln_pre = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
        self.transformer = BridgeTowerTransformer(config)
        self.ln_post = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
        self.share_layernorm = config.share_layernorm
        if not config.share_layernorm:
            self.ln_separate = nn.ModuleList(
                [nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) for _ in range(config.num_hidden_layers)]
            )

    def forward(self, pixel_values: torch.Tensor, attention_mask):
        hidden_states = self.embeddings(pixel_values)
        hidden_states = self.ln_pre(hidden_states)
        # NLD -> LND
        hidden_states = hidden_states.permute(1, 0, 2)

        hidden_states = self.transformer(hidden_states, attention_mask)
        # shape = [num_hidden_layers, hidden_size, *, grid ** 2]
        hidden_states = torch.stack(hidden_states, dim=0)
        # shape = [num_hidden_layers, *, hidden_size, grid ** 2]
        hidden_states = hidden_states.permute(0, 2, 1, 3)
        if self.share_layernorm:
            hidden_states = self.ln_post(hidden_states)
        else:
            hidden_states_stack = []
            for hidden_states, ln in zip(hidden_states, self.ln_separate):
                hidden_states = ln(hidden_states)
                hidden_states_stack.append(hidden_states)
            # shape = [num_hidden_layers, *, hidden_size, grid ** 2]
            hidden_states = torch.stack(hidden_states_stack, dim=0)
        return hidden_states

    def forward_pre(self, pixel_values: torch.Tensor):
        hidden_states = self.embeddings(pixel_values)
        hidden_states = self.ln_pre(hidden_states)
        # NLD -> LND
        hidden_states = hidden_states.permute(1, 0, 2)
        return hidden_states

    def forward_post(self, hidden_state: torch.Tensor):
        visual_output_post = hidden_state.permute(1, 0, 2)
        visual_output_post = self.ln_post(visual_output_post)
        return visual_output_post


class BridgeTowerLinkTower(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.link_tower_type = config.link_tower_type
        self.hidden_size = config.hidden_size
        if config.link_tower_type in ["add", "scaled_add", "interpolate"]:
            if config.link_tower_type == "scaled_add":
                self.scaled_factor = nn.Parameter(torch.tensor(1.0))
            elif config.link_tower_type == "interpolate":
                self.beta = nn.Parameter(torch.tensor(0.5))
            self.LayerNorm = nn.LayerNorm(self.hidden_size, eps=config.layer_norm_eps)
        else:
            raise NotImplementedError(f"link_tower_type {config.link_tower_type} is not implemented")

    def forward(self, hidden_states, cross_modal_hidden_states, attention_mask):
        if self.link_tower_type == "add":
            return self.LayerNorm(hidden_states + cross_modal_hidden_states)
        elif self.link_tower_type == "scaled_add":
            return self.LayerNorm(hidden_states * self.scaled_factor + cross_modal_hidden_states)
        elif self.link_tower_type == "interpolate":
            return self.LayerNorm(hidden_states * (1 - self.beta) + cross_modal_hidden_states * self.beta)
        else:
            raise NotImplementedError(f"link_tower_type {self.link_tower_type} is not implemented")


# Copied from transformers.models.bert.modeling_bert.BertSelfOutput with Bert->BridgeTower
class BridgeTowerSelfOutput(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)

    def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor) -> torch.Tensor:
        hidden_states = self.dense(hidden_states)
        hidden_states = self.dropout(hidden_states)
        hidden_states = self.LayerNorm(hidden_states + input_tensor)
        return hidden_states


# Copied from transformers.models.bert.modeling_bert.BertIntermediate with Bert->BridgeTower
class BridgeTowerIntermediate(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.intermediate_size)
        if isinstance(config.hidden_act, str):
            self.intermediate_act_fn = ACT2FN[config.hidden_act]
        else:
            self.intermediate_act_fn = config.hidden_act

    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
        hidden_states = self.dense(hidden_states)
        hidden_states = self.intermediate_act_fn(hidden_states)
        return hidden_states


# Copied from transformers.models.bert.modeling_bert.BertOutput with Bert->BridgeTower
class BridgeTowerOutput(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.intermediate_size, config.hidden_size)
        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)

    def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor) -> torch.Tensor:
        hidden_states = self.dense(hidden_states)
        hidden_states = self.dropout(hidden_states)
        hidden_states = self.LayerNorm(hidden_states + input_tensor)
        return hidden_states


# Copied from transformers.models.bert.modeling_bert.BertPooler with Bert->BridgeTower
class BridgeTowerPooler(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        self.activation = nn.Tanh()

    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
        # We "pool" the model by simply taking the hidden state corresponding
        # to the first token.
        first_token_tensor = hidden_states[:, 0]
        pooled_output = self.dense(first_token_tensor)
        pooled_output = self.activation(pooled_output)
        return pooled_output


# Copied from transformers.models.roberta.modeling_roberta.RobertaSelfAttention with Roberta->BridgeTower
class BridgeTowerSelfAttention(nn.Module):
    def __init__(self, config, position_embedding_type=None):
        super().__init__()
        if config.hidden_size % config.num_attention_heads != 0 and not hasattr(config, "embedding_size"):
            raise ValueError(
                f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention "
                f"heads ({config.num_attention_heads})"
            )

        self.num_attention_heads = config.num_attention_heads
        self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
        self.all_head_size = self.num_attention_heads * self.attention_head_size

        self.query = nn.Linear(config.hidden_size, self.all_head_size)
        self.key = nn.Linear(config.hidden_size, self.all_head_size)
        self.value = nn.Linear(config.hidden_size, self.all_head_size)

        self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
        self.position_embedding_type = position_embedding_type or getattr(
            config, "position_embedding_type", "absolute"
        )
        if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
            self.max_position_embeddings = config.max_position_embeddings
            self.distance_embedding = nn.Embedding(2 * config.max_position_embeddings - 1, self.attention_head_size)

        self.is_decoder = config.is_decoder

    def transpose_for_scores(self, x: torch.Tensor) -> torch.Tensor:
        new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.attention_head_size)
        x = x.view(new_x_shape)
        return x.permute(0, 2, 1, 3)

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        encoder_hidden_states: Optional[torch.FloatTensor] = None,
        encoder_attention_mask: Optional[torch.FloatTensor] = None,
        past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
        output_attentions: Optional[bool] = False,
    ) -> Tuple[torch.Tensor]:
        mixed_query_layer = self.query(hidden_states)

        # If this is instantiated as a cross-attention module, the keys
        # and values come from an encoder; the attention mask needs to be
        # such that the encoder's padding tokens are not attended to.
        is_cross_attention = encoder_hidden_states is not None

        if is_cross_attention and past_key_value is not None:
            # reuse k,v, cross_attentions
            key_layer = past_key_value[0]
            value_layer = past_key_value[1]
            attention_mask = encoder_attention_mask
        elif is_cross_attention:
            key_layer = self.transpose_for_scores(self.key(encoder_hidden_states))
            value_layer = self.transpose_for_scores(self.value(encoder_hidden_states))
            attention_mask = encoder_attention_mask
        elif past_key_value is not None:
            key_layer = self.transpose_for_scores(self.key(hidden_states))
            value_layer = self.transpose_for_scores(self.value(hidden_states))
            key_layer = torch.cat([past_key_value[0], key_layer], dim=2)
            value_layer = torch.cat([past_key_value[1], value_layer], dim=2)
        else:
            key_layer = self.transpose_for_scores(self.key(hidden_states))
            value_layer = self.transpose_for_scores(self.value(hidden_states))

        query_layer = self.transpose_for_scores(mixed_query_layer)

        use_cache = past_key_value is not None
        if self.is_decoder:
            # if cross_attention save Tuple(torch.Tensor, torch.Tensor) of all cross attention key/value_states.
            # Further calls to cross_attention layer can then reuse all cross-attention
            # key/value_states (first "if" case)
            # if uni-directional self-attention (decoder) save Tuple(torch.Tensor, torch.Tensor) of
            # all previous decoder key/value_states. Further calls to uni-directional self-attention
            # can concat previous decoder key/value_states to current projected key/value_states (third "elif" case)
            # if encoder bi-directional self-attention `past_key_value` is always `None`
            past_key_value = (key_layer, value_layer)

        # Take the dot product between "query" and "key" to get the raw attention scores.
        attention_scores = torch.matmul(query_layer, key_layer.transpose(-1, -2))

        if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
            query_length, key_length = query_layer.shape[2], key_layer.shape[2]
            if use_cache:
                position_ids_l = torch.tensor(key_length - 1, dtype=torch.long, device=hidden_states.device).view(
                    -1, 1
                )
            else:
                position_ids_l = torch.arange(query_length, dtype=torch.long, device=hidden_states.device).view(-1, 1)
            position_ids_r = torch.arange(key_length, dtype=torch.long, device=hidden_states.device).view(1, -1)
            distance = position_ids_l - position_ids_r

            positional_embedding = self.distance_embedding(distance + self.max_position_embeddings - 1)
            positional_embedding = positional_embedding.to(dtype=query_layer.dtype)  # fp16 compatibility

            if self.position_embedding_type == "relative_key":
                relative_position_scores = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
                attention_scores = attention_scores + relative_position_scores
            elif self.position_embedding_type == "relative_key_query":
                relative_position_scores_query = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
                relative_position_scores_key = torch.einsum("bhrd,lrd->bhlr", key_layer, positional_embedding)
                attention_scores = attention_scores + relative_position_scores_query + relative_position_scores_key

        attention_scores = attention_scores / math.sqrt(self.attention_head_size)
        if attention_mask is not None:
            # Apply the attention mask is (precomputed for all layers in BridgeTowerModel forward() function)
            attention_scores = attention_scores + attention_mask

        # Normalize the attention scores to probabilities.
        attention_probs = nn.functional.softmax(attention_scores, dim=-1)

        # This is actually dropping out entire tokens to attend to, which might
        # seem a bit unusual, but is taken from the original Transformer paper.
        attention_probs = self.dropout(attention_probs)

        # Mask heads if we want to
        if head_mask is not None:
            attention_probs = attention_probs * head_mask

        context_layer = torch.matmul(attention_probs, value_layer)

        context_layer = context_layer.permute(0, 2, 1, 3).contiguous()
        new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
        context_layer = context_layer.view(new_context_layer_shape)

        outputs = (context_layer, attention_probs) if output_attentions else (context_layer,)

        if self.is_decoder:
            outputs = outputs + (past_key_value,)
        return outputs


# Copied from transformers.models.bert.modeling_bert.BertAttention with Bert->BridgeTower
class BridgeTowerAttention(nn.Module):
    def __init__(self, config, position_embedding_type=None):
        super().__init__()
        self.self = BridgeTowerSelfAttention(config, position_embedding_type=position_embedding_type)
        self.output = BridgeTowerSelfOutput(config)
        self.pruned_heads = set()

    def prune_heads(self, heads):
        if len(heads) == 0:
            return
        heads, index = find_pruneable_heads_and_indices(
            heads, self.self.num_attention_heads, self.self.attention_head_size, self.pruned_heads
        )

        # Prune linear layers
        self.self.query = prune_linear_layer(self.self.query, index)
        self.self.key = prune_linear_layer(self.self.key, index)
        self.self.value = prune_linear_layer(self.self.value, index)
        self.output.dense = prune_linear_layer(self.output.dense, index, dim=1)

        # Update hyper params and store pruned heads
        self.self.num_attention_heads = self.self.num_attention_heads - len(heads)
        self.self.all_head_size = self.self.attention_head_size * self.self.num_attention_heads
        self.pruned_heads = self.pruned_heads.union(heads)

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        encoder_hidden_states: Optional[torch.FloatTensor] = None,
        encoder_attention_mask: Optional[torch.FloatTensor] = None,
        past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
        output_attentions: Optional[bool] = False,
    ) -> Tuple[torch.Tensor]:
        self_outputs = self.self(
            hidden_states,
            attention_mask,
            head_mask,
            encoder_hidden_states,
            encoder_attention_mask,
            past_key_value,
            output_attentions,
        )
        attention_output = self.output(self_outputs[0], hidden_states)
        outputs = (attention_output,) + self_outputs[1:]  # add attentions if we output them
        return outputs


class BridgeTowerBertCrossLayer(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.chunk_size_feed_forward = config.chunk_size_feed_forward
        self.seq_len_dim = 1
        self.attention = BridgeTowerAttention(config)
        self.is_decoder = config.is_decoder
        self.add_cross_attention = config.add_cross_attention
        self.crossattention = BridgeTowerAttention(config)
        self.intermediate = BridgeTowerIntermediate(config)
        self.output = BridgeTowerOutput(config)

    def forward(
        self,
        hidden_states,
        encoder_hidden_states,
        attention_mask=None,
        head_mask=None,
        encoder_attention_mask=None,
        past_key_value=None,
        output_attentions=False,
    ):
        # decoder uni-directional self-attention cached key/values tuple is at positions 1,2
        self_attention_outputs = self.attention(
            hidden_states,
            attention_mask=attention_mask,
            head_mask=None,
            output_attentions=output_attentions,
            past_key_value=None,
        )
        attention_output = self_attention_outputs[0]

        # if decoder, the last output is tuple of self-attn cache
        # add self attentions if we output attention weights
        outputs = self_attention_outputs[1:]

        cross_attention_outputs = self.crossattention(
            attention_output,
            attention_mask=attention_mask,
            head_mask=head_mask,
            encoder_hidden_states=encoder_hidden_states,
            encoder_attention_mask=encoder_attention_mask,
            past_key_value=past_key_value,
            output_attentions=output_attentions,
        )
        attention_output = cross_attention_outputs[0]
        # add cross attentions if we output attention weights
        outputs = outputs + cross_attention_outputs[1:-1]

        layer_output = apply_chunking_to_forward(
            self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output
        )
        outputs = (layer_output,) + outputs

        return outputs

    def feed_forward_chunk(self, attention_output):
        intermediate_output = self.intermediate(attention_output)
        layer_output = self.output(intermediate_output, attention_output)
        return layer_output


class BridgeTowerTextLayer(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.chunk_size_feed_forward = config.chunk_size_feed_forward
        self.seq_len_dim = 1
        self.attention = BridgeTowerAttention(config)
        self.is_decoder = config.is_decoder
        self.add_cross_attention = config.add_cross_attention
        if self.add_cross_attention:
            if not self.is_decoder:
                raise ValueError(f"{self} should be used as a decoder model if cross attention is added")
            self.crossattention = BridgeTowerAttention(config, position_embedding_type="absolute")
        self.intermediate = BridgeTowerIntermediate(config)
        self.output = BridgeTowerOutput(config)

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        encoder_hidden_states: Optional[torch.FloatTensor] = None,
        encoder_attention_mask: Optional[torch.FloatTensor] = None,
        past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
        output_attentions: Optional[bool] = False,
    ) -> Tuple[torch.Tensor]:
        # decoder uni-directional self-attention cached key/values tuple is at positions 1,2
        self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None
        self_attention_outputs = self.attention(
            hidden_states,
            attention_mask,
            head_mask,
            output_attentions=output_attentions,
            past_key_value=self_attn_past_key_value,
        )
        attention_output = self_attention_outputs[0]

        # if decoder, the last output is tuple of self-attn cache
        if self.is_decoder:
            outputs = self_attention_outputs[1:-1]
            present_key_value = self_attention_outputs[-1]
        else:
            outputs = self_attention_outputs[1:]  # add self attentions if we output attention weights

        cross_attn_present_key_value = None
        if self.is_decoder and encoder_hidden_states is not None:
            if not hasattr(self, "crossattention"):
                raise ValueError(
                    f"If `encoder_hidden_states` are passed, {self} has to be instantiated with cross-attention layers"
                    " by setting `config.add_cross_attention=True`"
                )

            # cross_attn cached key/values tuple is at positions 3,4 of past_key_value tuple
            cross_attn_past_key_value = past_key_value[-2:] if past_key_value is not None else None
            cross_attention_outputs = self.crossattention(
                attention_output,
                attention_mask,
                head_mask,
                encoder_hidden_states,
                encoder_attention_mask,
                cross_attn_past_key_value,
                output_attentions,
            )
            attention_output = cross_attention_outputs[0]
            outputs = outputs + cross_attention_outputs[1:-1]  # add cross attentions if we output attention weights

            # add cross-attn cache to positions 3,4 of present_key_value tuple
            cross_attn_present_key_value = cross_attention_outputs[-1]
            present_key_value = present_key_value + cross_attn_present_key_value

        layer_output = apply_chunking_to_forward(
            self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output
        )
        outputs = (layer_output,) + outputs

        # if decoder, return the attn key/values as the last output
        if self.is_decoder:
            outputs = outputs + (present_key_value,)

        return outputs

    def feed_forward_chunk(self, attention_output):
        intermediate_output = self.intermediate(attention_output)
        layer_output = self.output(intermediate_output, attention_output)
        return layer_output


# Copied from transformers.models.roberta.modeling_roberta.RobertaEncoder with Roberta->BridgeTowerText
class BridgeTowerTextEncoder(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.config = config
        self.layer = nn.ModuleList([BridgeTowerTextLayer(config) for _ in range(config.num_hidden_layers)])
        self.gradient_checkpointing = False

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        encoder_hidden_states: Optional[torch.FloatTensor] = None,
        encoder_attention_mask: Optional[torch.FloatTensor] = None,
        past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = False,
        output_hidden_states: Optional[bool] = False,
        return_dict: Optional[bool] = True,
    ) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPastAndCrossAttentions]:
        all_hidden_states = () if output_hidden_states else None
        all_self_attentions = () if output_attentions else None
        all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None

        if self.gradient_checkpointing and self.training:
            if use_cache:
                logger.warning_once(
                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
                )
                use_cache = False

        next_decoder_cache = () if use_cache else None
        for i, layer_module in enumerate(self.layer):
            if output_hidden_states:
                all_hidden_states = all_hidden_states + (hidden_states,)

            layer_head_mask = head_mask[i] if head_mask is not None else None
            past_key_value = past_key_values[i] if past_key_values is not None else None

            if self.gradient_checkpointing and self.training:

                def create_custom_forward(module):
                    def custom_forward(*inputs):
                        return module(*inputs, past_key_value, output_attentions)

                    return custom_forward

                layer_outputs = torch.utils.checkpoint.checkpoint(
                    create_custom_forward(layer_module),
                    hidden_states,
                    attention_mask,
                    layer_head_mask,
                    encoder_hidden_states,
                    encoder_attention_mask,
                )
            else:
                layer_outputs = layer_module(
                    hidden_states,
                    attention_mask,
                    layer_head_mask,
                    encoder_hidden_states,
                    encoder_attention_mask,
                    past_key_value,
                    output_attentions,
                )

            hidden_states = layer_outputs[0]
            if use_cache:
                next_decoder_cache += (layer_outputs[-1],)
            if output_attentions:
                all_self_attentions = all_self_attentions + (layer_outputs[1],)
                if self.config.add_cross_attention:
                    all_cross_attentions = all_cross_attentions + (layer_outputs[2],)

        if output_hidden_states:
            all_hidden_states = all_hidden_states + (hidden_states,)

        if not return_dict:
            return tuple(
                v
                for v in [
                    hidden_states,
                    next_decoder_cache,
                    all_hidden_states,
                    all_self_attentions,
                    all_cross_attentions,
                ]
                if v is not None
            )
        return BaseModelOutputWithPastAndCrossAttentions(
            last_hidden_state=hidden_states,
            past_key_values=next_decoder_cache,
            hidden_states=all_hidden_states,
            attentions=all_self_attentions,
            cross_attentions=all_cross_attentions,
        )


# Copied from transformers.models.roberta.modeling_roberta.RobertaEmbeddings with Roberta->BridgeTowerText
class BridgeTowerTextEmbeddings(nn.Module):
    """
    Same as BertEmbeddings with a tiny tweak for positional embeddings indexing.
    """

    # Copied from transformers.models.bert.modeling_bert.BertEmbeddings.__init__
    def __init__(self, config):
        super().__init__()
        self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id)
        self.position_embeddings = nn.Embedding(config.max_position_embeddings, config.hidden_size)
        self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size)

        # self.LayerNorm is not snake-cased to stick with TensorFlow model variable name and be able to load
        # any TensorFlow checkpoint file
        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)
        # position_ids (1, len position emb) is contiguous in memory and exported when serialized
        self.position_embedding_type = getattr(config, "position_embedding_type", "absolute")
        self.register_buffer(
            "position_ids", torch.arange(config.max_position_embeddings).expand((1, -1)), persistent=False
        )
        self.register_buffer(
            "token_type_ids", torch.zeros(self.position_ids.size(), dtype=torch.long), persistent=False
        )

        # End copy
        self.padding_idx = config.pad_token_id
        self.position_embeddings = nn.Embedding(
            config.max_position_embeddings, config.hidden_size, padding_idx=self.padding_idx
        )

    def forward(
        self, input_ids=None, token_type_ids=None, position_ids=None, inputs_embeds=None, past_key_values_length=0
    ):
        if position_ids is None:
            if input_ids is not None:
                # Create the position ids from the input token ids. Any padded tokens remain padded.
                position_ids = create_position_ids_from_input_ids(input_ids, self.padding_idx, past_key_values_length)
            else:
                position_ids = self.create_position_ids_from_inputs_embeds(inputs_embeds)

        if input_ids is not None:
            input_shape = input_ids.size()
        else:
            input_shape = inputs_embeds.size()[:-1]

        seq_length = input_shape[1]

        # Setting the token_type_ids to the registered buffer in constructor where it is all zeros, which usually occurs
        # when its auto-generated, registered buffer helps users when tracing the model without passing token_type_ids, solves
        # issue #5664
        if token_type_ids is None:
            if hasattr(self, "token_type_ids"):
                buffered_token_type_ids = self.token_type_ids[:, :seq_length]
                buffered_token_type_ids_expanded = buffered_token_type_ids.expand(input_shape[0], seq_length)
                token_type_ids = buffered_token_type_ids_expanded
            else:
                token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=self.position_ids.device)

        if inputs_embeds is None:
            inputs_embeds = self.word_embeddings(input_ids)
        token_type_embeddings = self.token_type_embeddings(token_type_ids)

        embeddings = inputs_embeds + token_type_embeddings
        if self.position_embedding_type == "absolute":
            position_embeddings = self.position_embeddings(position_ids)
            embeddings += position_embeddings
        embeddings = self.LayerNorm(embeddings)
        embeddings = self.dropout(embeddings)
        return embeddings

    def create_position_ids_from_inputs_embeds(self, inputs_embeds):
        """
        We are provided embeddings directly. We cannot infer which are padded so just generate sequential position ids.

        Args:
            inputs_embeds: torch.Tensor

        Returns: torch.Tensor
        """
        input_shape = inputs_embeds.size()[:-1]
        sequence_length = input_shape[1]

        position_ids = torch.arange(
            self.padding_idx + 1, sequence_length + self.padding_idx + 1, dtype=torch.long, device=inputs_embeds.device
        )
        return position_ids.unsqueeze(0).expand(input_shape)


# Copied from transformers.models.roberta.modeling_roberta.create_position_ids_from_input_ids
def create_position_ids_from_input_ids(input_ids, padding_idx, past_key_values_length=0):
    """
    Replace non-padding symbols with their position numbers. Position numbers begin at padding_idx+1. Padding symbols
    are ignored. This is modified from fairseq's `utils.make_positions`.

    Args:
        x: torch.Tensor x:

    Returns: torch.Tensor
    """
    # The series of casts and type-conversions here are carefully balanced to both work with ONNX export and XLA.
    mask = input_ids.ne(padding_idx).int()
    incremental_indices = (torch.cumsum(mask, dim=1).type_as(mask) + past_key_values_length) * mask
    return incremental_indices.long() + padding_idx


class BridgeTowerPreTrainedModel(PreTrainedModel):
    """
    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
    models.
    """

    config_class = BridgeTowerConfig
    base_model_prefix = "bridgetower"
    supports_gradient_checkpointing = False
    _no_split_modules = ["BridgeTowerSelfAttention", "BridgeTowerResidualAttention"]
    _skip_keys_device_placement = "past_key_values"

    def _init_weights(self, module):
        if isinstance(module, BridgeTowerVisionModel):
            proj_std = (module.visual.transformer.hidden_size**-0.5) * (
                (2 * module.visual.transformer.num_hidden_layers) ** -0.5
            )
            attn_std = module.visual.transformer.hidden_size**-0.5
            fc_std = (2 * module.visual.transformer.hidden_size) ** -0.5
            for block in module.visual.transformer.resblocks:
                nn.init.normal_(block.attn.in_proj_weight, std=attn_std * self.config.initializer_factor)
                nn.init.normal_(block.attn.out_proj.weight, std=proj_std * self.config.initializer_factor)
                nn.init.normal_(block.mlp.c_fc.weight, std=fc_std * self.config.initializer_factor)
                nn.init.normal_(block.mlp.c_proj.weight, std=proj_std * self.config.initializer_factor)

            nn.init.normal_(module.visual.embeddings.class_embedding, std=attn_std * self.config.initializer_factor)
            nn.init.normal_(
                module.visual.embeddings.position_embedding.weight, std=attn_std * self.config.initializer_factor
            )
        elif isinstance(module, (nn.Linear, nn.Conv2d, nn.Embedding)):
            module.weight.data.normal_(mean=0.0, std=0.05 * self.config.initializer_factor)
        elif isinstance(module, nn.LayerNorm):
            module.bias.data.zero_()
            module.weight.data.fill_(1.0)

        if isinstance(module, nn.Linear) and module.bias is not None:
            module.bias.data.zero_()


class BridgeTowerVisionModel(BridgeTowerPreTrainedModel):
    config_class = BridgeTowerVisionConfig

    def __init__(self, config):
        super().__init__(config)
        self.visual = BridgeTowerVisionTransformer(config)

    @property
    def dtype(self):
        return self.visual.embeddings.patch_embedding.weight.dtype

    def forward(self, image, image_mask=None):
        return self.visual(image.type(self.dtype), image_mask)


class BridgeTowerTextModel(BridgeTowerPreTrainedModel):
    """

    The model can behave as an encoder (with only self-attention) as well as a decoder, in which case a layer of
    cross-attention is added between the self-attention layers, following the architecture described in *Attention is
    all you need*_ by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz
    Kaiser and Illia Polosukhin.

    To behave as an decoder the model needs to be initialized with the `is_decoder` argument of the configuration set
    to `True`. To be used in a Seq2Seq model, the model needs to initialized with both `is_decoder` argument and
    `add_cross_attention` set to `True`; an `encoder_hidden_states` is then expected as an input to the forward pass.

    .. _*Attention is all you need*: https://arxiv.org/abs/1706.03762

    """

    config_class = BridgeTowerTextConfig

    def __init__(self, config, add_pooling_layer=True):
        super().__init__(config)
        self.config = config

        self.embeddings = BridgeTowerTextEmbeddings(config)
        self.encoder = BridgeTowerTextEncoder(config)

        self.pooler = BridgeTowerPooler(config) if add_pooling_layer else None

        # Initialize weights and apply final processing
        self.post_init()

    def get_input_embeddings(self):
        return self.embeddings.word_embeddings

    def set_input_embeddings(self, value):
        self.embeddings.word_embeddings = value

    def _prune_heads(self, heads_to_prune):
        """
        Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base
        class PreTrainedModel
        """
        for layer, heads in heads_to_prune.items():
            self.encoder.layer[layer].attention.prune_heads(heads)

    # Copied from transformers.models.roberta.modeling_roberta.RobertaModel.forward
    def forward(
        self,
        input_ids: Optional[torch.Tensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        token_type_ids: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.Tensor] = None,
        head_mask: Optional[torch.Tensor] = None,
        inputs_embeds: Optional[torch.Tensor] = None,
        encoder_hidden_states: Optional[torch.Tensor] = None,
        encoder_attention_mask: Optional[torch.Tensor] = None,
        past_key_values: Optional[List[torch.FloatTensor]] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPoolingAndCrossAttentions]:
        r"""
        encoder_hidden_states  (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
            Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if
            the model is configured as a decoder.
        encoder_attention_mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in
            the cross-attention if the model is configured as a decoder. Mask values selected in `[0, 1]`:

            - 1 for tokens that are **not masked**,
            - 0 for tokens that are **masked**.
        past_key_values (`tuple(tuple(torch.FloatTensor))` of length `config.n_layers` with each tuple having 4 tensors of shape `(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`):
            Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding.

            If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
            don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
            `decoder_input_ids` of shape `(batch_size, sequence_length)`.
        use_cache (`bool`, *optional*):
            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
            `past_key_values`).
        """
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        if self.config.is_decoder:
            use_cache = use_cache if use_cache is not None else self.config.use_cache
        else:
            use_cache = False

        if input_ids is not None and inputs_embeds is not None:
            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
        elif input_ids is not None:
            self.warn_if_padding_and_no_attention_mask(input_ids, attention_mask)
            input_shape = input_ids.size()
        elif inputs_embeds is not None:
            input_shape = inputs_embeds.size()[:-1]
        else:
            raise ValueError("You have to specify either input_ids or inputs_embeds")

        batch_size, seq_length = input_shape
        device = input_ids.device if input_ids is not None else inputs_embeds.device

        # past_key_values_length
        past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0

        if attention_mask is None:
            attention_mask = torch.ones(((batch_size, seq_length + past_key_values_length)), device=device)

        if token_type_ids is None:
            if hasattr(self.embeddings, "token_type_ids"):
                buffered_token_type_ids = self.embeddings.token_type_ids[:, :seq_length]
                buffered_token_type_ids_expanded = buffered_token_type_ids.expand(batch_size, seq_length)
                token_type_ids = buffered_token_type_ids_expanded
            else:
                token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device)

        # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
        # ourselves in which case we just need to make it broadcastable to all heads.
        extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(attention_mask, input_shape)

        # If a 2D or 3D attention mask is provided for the cross-attention
        # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
        if self.config.is_decoder and encoder_hidden_states is not None:
            encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
            encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
            if encoder_attention_mask is None:
                encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)
            encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask)
        else:
            encoder_extended_attention_mask = None

        # Prepare head mask if needed
        # 1.0 in head_mask indicate we keep the head
        # attention_probs has shape bsz x n_heads x N x N
        # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
        # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
        head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)

        embedding_output = self.embeddings(
            input_ids=input_ids,
            position_ids=position_ids,
            token_type_ids=token_type_ids,
            inputs_embeds=inputs_embeds,
            past_key_values_length=past_key_values_length,
        )
        encoder_outputs = self.encoder(
            embedding_output,
            attention_mask=extended_attention_mask,
            head_mask=head_mask,
            encoder_hidden_states=encoder_hidden_states,
            encoder_attention_mask=encoder_extended_attention_mask,
            past_key_values=past_key_values,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        sequence_output = encoder_outputs[0]
        pooled_output = self.pooler(sequence_output) if self.pooler is not None else None

        if not return_dict:
            return (sequence_output, pooled_output) + encoder_outputs[1:]

        return BaseModelOutputWithPoolingAndCrossAttentions(
            last_hidden_state=sequence_output,
            pooler_output=pooled_output,
            past_key_values=encoder_outputs.past_key_values,
            hidden_states=encoder_outputs.hidden_states,
            attentions=encoder_outputs.attentions,
            cross_attentions=encoder_outputs.cross_attentions,
        )


@add_start_docstrings(
    "The bare BridgeTower Model transformer outputting BridgeTowerModelOutput object without any specific head on"
    " top.",
    BRIDGETOWER_START_DOCSTRING,
)
class BridgeTowerModel(BridgeTowerPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.config = config
        vision_config = config.vision_config
        text_config = config.text_config

        if config.share_cross_modal_transformer_layers:
            self.cross_modal_text_transform = nn.Linear(text_config.hidden_size, config.hidden_size)
            self.cross_modal_image_transform = nn.Linear(vision_config.hidden_size, config.hidden_size)
        else:
            self.cross_modal_text_transform = nn.ModuleList(
                [nn.Linear(text_config.hidden_size, config.hidden_size) for _ in range(config.num_hidden_layers)]
            )
            self.cross_modal_image_transform = nn.ModuleList(
                [nn.Linear(vision_config.hidden_size, config.hidden_size) for _ in range(config.num_hidden_layers)]
            )

        self.token_type_embeddings = nn.Embedding(2, config.hidden_size)

        self.vision_model = BridgeTowerVisionModel(vision_config)

        self.text_model = BridgeTowerTextModel(text_config)

        if not vision_config.share_layernorm and config.init_layernorm_from_vision_encoder:
            for ln in self.vision_model.visual.cross_modal_ln_separate:
                ln.weight.data = self.vision_model.visual.ln_post.weight.data
                ln.bias.data = self.vision_model.visual.ln_post.bias.data

        self.cross_modal_image_layers = nn.ModuleList(
            [BridgeTowerBertCrossLayer(text_config) for _ in range(config.num_hidden_layers)]
        )
        self.cross_modal_text_layers = nn.ModuleList(
            [BridgeTowerBertCrossLayer(text_config) for _ in range(config.num_hidden_layers)]
        )

        # Class token => Linear => Tanh
        self.cross_modal_image_pooler = BridgeTowerPooler(config)
        self.cross_modal_text_pooler = BridgeTowerPooler(config)

        # Initialize BridgeTower Components
        self.cross_modal_text_layernorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
        self.cross_modal_image_layernorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)

        if config.share_link_tower_layers:
            self.cross_modal_text_link_tower = BridgeTowerLinkTower(config)
            self.cross_modal_image_link_tower = BridgeTowerLinkTower(config)
        else:
            self.cross_modal_text_link_tower = nn.ModuleList(
                [BridgeTowerLinkTower(config) for _ in range(config.num_hidden_layers - 1)]
            )
            self.cross_modal_image_link_tower = nn.ModuleList(
                [BridgeTowerLinkTower(config) for _ in range(config.num_hidden_layers - 1)]
            )

        self.post_init()

    def get_input_embeddings(self):
        return self.text_model.get_input_embeddings()

    def set_input_embeddings(self, value):
        self.text_model.set_input_embeddings(value)

    @add_start_docstrings_to_model_forward(BRIDGETOWER_INPUTS_DOCSTRING)
    @replace_return_docstrings(output_type=BridgeTowerModelOutput, config_class=_CONFIG_FOR_DOC)
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        pixel_values: Optional[torch.FloatTensor] = None,
        pixel_mask: Optional[torch.LongTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        image_embeds: Optional[torch.FloatTensor] = None,
        image_token_type_idx: Optional[int] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        labels: Optional[torch.LongTensor] = None,
    ) -> Union[Tuple[torch.Tensor], BridgeTowerModelOutput]:
        r"""
        output_hidden_states (`bool`, *optional*):
            If set to `True`, hidden states are returned as a list containing the hidden states of text, image, and
            cross-modal components respectively. i.e. `(hidden_states_text, hidden_states_image,
            hidden_states_cross_modal)` where each element is a list of the hidden states of the corresponding
            modality. `hidden_states_txt/img` are a list of tensors corresponding to unimodal hidden states and
            `hidden_states_cross_modal` is a list of tuples containing `cross_modal_text_hidden_states` and
            `cross_modal_image_hidden_states` of each brdige layer.
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels are currently not supported.
        Returns:

        Examples:

        ```python
        >>> from transformers import BridgeTowerProcessor, BridgeTowerModel
        >>> from PIL import Image
        >>> import requests

        >>> # prepare image and text
        >>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
        >>> image = Image.open(requests.get(url, stream=True).raw)
        >>> text = "hello world"
        >>> processor = BridgeTowerProcessor.from_pretrained("BridgeTower/bridgetower-base")
        >>> model = BridgeTowerModel.from_pretrained("BridgeTower/bridgetower-base")

        >>> inputs = processor(image, text, return_tensors="pt")
        >>> outputs = model(**inputs)
        >>> outputs.keys()
        odict_keys(['text_features', 'image_features', 'pooler_output'])
        ```"""
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        all_hidden_states_text = () if output_hidden_states else None
        all_hidden_states_image = () if output_hidden_states else None
        all_hidden_states_cross = () if output_hidden_states else None
        all_hidden_states = () if output_hidden_states else None
        all_self_attentions = () if output_attentions else None

        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
        image_token_type_idx = image_token_type_idx if image_token_type_idx else 1
        input_shape = input_ids.size()
        text_embeds = self.text_model.embeddings(input_ids=input_ids)

        if output_hidden_states:
            all_hidden_states_text += (text_embeds,)

        if attention_mask is None:
            attention_mask = torch.ones(input_shape, dtype=torch.long, device=input_ids.device)
        extend_text_masks = self.text_model.get_extended_attention_mask(attention_mask, input_shape).to(
            input_ids.device
        )

        # The split_index determines how many layers of the uni-modal encoder are applied before the cross-modal encoder
        split_index = len(self.text_model.encoder.layer) - self.config.num_hidden_layers + 1

        # Run the first 'split_index' layers of the textual encoder
        for layer in self.text_model.encoder.layer[:split_index]:
            text_embeds = layer(text_embeds, extend_text_masks)[0]

            if output_hidden_states:
                all_hidden_states_text += (text_embeds,)

        if image_embeds is None:
            image_embeds = self.vision_model.visual.forward_pre(pixel_values.type(self.vision_model.dtype))
        else:
            # Permute as BridgeTowerResidualAttention has batch_first=True
            image_embeds = image_embeds.permute(1, 0, 2)

        if output_hidden_states:
            all_hidden_states_image += (image_embeds,)

        # Run the first 'split_index' layers of the visual encoder
        for block in self.vision_model.visual.transformer.resblocks[:split_index]:
            image_embeds = block(image_embeds)
            if output_hidden_states:
                all_hidden_states_image += (image_embeds,)

        image_embeds_with_ln = self.vision_model.visual.forward_post(image_embeds.type(self.vision_model.dtype))

        # first layer is a special case because we don't have the output from the cross-encoder yet
        cross_modal_text = self.cross_modal_text_transform(text_embeds)

        text_token_type_embeddings = self.token_type_embeddings(
            torch.zeros(1, dtype=torch.long, device=input_ids.device)
        ).expand_as(cross_modal_text)

        cross_modal_text = self.cross_modal_text_layernorm(cross_modal_text + text_token_type_embeddings)

        image_embeds_with_ln = self.cross_modal_image_transform(image_embeds_with_ln)
        image_token_type_embeddings = self.token_type_embeddings(
            torch.full((1,), image_token_type_idx, dtype=torch.long, device=input_ids.device)
        ).expand_as(image_embeds_with_ln)

        image_embeds_with_ln = image_embeds_with_ln + image_token_type_embeddings
        cross_modal_image = self.cross_modal_image_layernorm(image_embeds_with_ln)

        pixel_mask = torch.ones(
            (cross_modal_image.size(0), cross_modal_image.size(1)),
            dtype=torch.long,
            device=input_ids.device,
        )
        extend_image_masks = self.text_model.get_extended_attention_mask(pixel_mask, pixel_mask.size()).to(
            input_ids.device
        )

        layer_outputs_text = self.cross_modal_text_layers[0](
            cross_modal_text,
            cross_modal_image,
            attention_mask=extend_text_masks,
            encoder_attention_mask=extend_image_masks,
            output_attentions=output_attentions,
        )
        cross_text_features = layer_outputs_text[0]

        layer_outputs_image = self.cross_modal_image_layers[0](
            cross_modal_image,
            cross_modal_text,
            attention_mask=extend_image_masks,
            encoder_attention_mask=extend_text_masks,
            output_attentions=output_attentions,
        )
        cross_image_features = layer_outputs_image[0]

        if output_hidden_states:
            all_hidden_states_cross += ((cross_text_features, cross_image_features),)

        if output_attentions:
            all_self_attentions += ((layer_outputs_text[1], layer_outputs_image[1]),)

        link_layer_index = 0

        #  Each of the top 6 layers of the visual and textual encoders ([split_index:]) is connected to each layer of
        #  the cross-modal encoder via bridge layers, which brings bottom-up alignment and fusion to the cross-modal encoder.
        for i in range(split_index, len(self.text_model.encoder.layer)):
            text_embeds = self.text_model.encoder.layer[i](text_embeds, extend_text_masks)[0]
            image_embeds = self.vision_model.visual.transformer.resblocks[i](image_embeds).type(
                self.vision_model.dtype
            )
            image_embeds_with_ln = (
                self.cross_modal_image_transform(self.vision_model.visual.forward_post(image_embeds))
                + image_token_type_embeddings
            )

            text_link_tower = self.cross_modal_text_link_tower[link_layer_index]
            image_link_tower = self.cross_modal_image_link_tower[link_layer_index]

            # Bridge layers for textual and visual encoders
            cross_text_features_ = text_link_tower(
                self.cross_modal_text_transform(text_embeds) + text_token_type_embeddings,
                cross_text_features,
                extend_text_masks,
            )
            cross_image_features_ = image_link_tower(image_embeds_with_ln, cross_image_features, extend_image_masks)

            # Cross-modal encoder via bridge layers of textual and visual encoders
            layer_outputs_text = self.cross_modal_text_layers[link_layer_index + 1](
                cross_text_features_,
                cross_image_features_,
                attention_mask=extend_text_masks,
                encoder_attention_mask=extend_image_masks,
                output_attentions=output_attentions,
            )
            cross_text_features = layer_outputs_text[0]

            layer_outputs_image = self.cross_modal_image_layers[link_layer_index + 1](
                cross_image_features_,
                cross_text_features_,
                attention_mask=extend_image_masks,
                encoder_attention_mask=extend_text_masks,
                output_attentions=output_attentions,
            )
            cross_image_features = layer_outputs_image[0]

            link_layer_index += 1

            if output_hidden_states:
                all_hidden_states_text += (text_embeds,)
                all_hidden_states_image += (image_embeds,)
                all_hidden_states_cross += ((cross_text_features, cross_image_features),)

            if output_attentions:
                all_self_attentions += ((layer_outputs_text[1], layer_outputs_image[1]),)

        #  Concatenate the cls token of the text and image features to get the final represtation
        text_features, image_features = cross_text_features, cross_image_features
        cls_features = self.get_cls_features(text_features, image_features)

        if output_hidden_states:
            all_hidden_states = (all_hidden_states_text, all_hidden_states_image, all_hidden_states_cross)

        if not return_dict:
            return tuple(
                v
                for v in [text_features, image_features, cls_features, all_hidden_states, all_self_attentions]
                if v is not None
            )

        return BridgeTowerModelOutput(
            text_features=text_features,
            image_features=image_features,
            pooler_output=cls_features,
            hidden_states=all_hidden_states,
            attentions=all_self_attentions,
        )

    def get_cls_features(self, text_features, image_features):
        cls_features_text = self.cross_modal_text_pooler(text_features)
        cls_features_image = self.cross_modal_image_pooler(image_features)
        return torch.cat([cls_features_text, cls_features_image], dim=-1)


# Copied from transformers.models.vilt.modeling_vilt.ViltPredictionHeadTransform with Vilt->BridgeTower
class BridgeTowerPredictionHeadTransform(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        if isinstance(config.hidden_act, str):
            self.transform_act_fn = ACT2FN[config.hidden_act]
        else:
            self.transform_act_fn = config.hidden_act
        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)

    def forward(self, hidden_states):
        hidden_states = self.dense(hidden_states)
        hidden_states = self.transform_act_fn(hidden_states)
        hidden_states = self.LayerNorm(hidden_states)
        return hidden_states


class BridgeTowerMLMHead(nn.Module):
    def __init__(self, config, weight=None):
        super().__init__()
        self.config = config
        self.transform = BridgeTowerPredictionHeadTransform(config)
        self.decoder = nn.Linear(config.hidden_size, config.text_config.vocab_size, bias=False)
        self.bias = nn.Parameter(torch.zeros(config.text_config.vocab_size))
        if weight is not None:
            self.decoder.weight = weight

    def forward(self, x):
        mlm_score = self.transform(x)
        mlm_score = self.decoder(mlm_score) + self.bias
        return mlm_score


class BridgeTowerITMHead(nn.Module):
    def __init__(self, hidden_size):
        super().__init__()
        self.fc = nn.Linear(hidden_size, 2)

    def forward(self, x):
        itm_score = self.fc(x)
        return itm_score


@add_start_docstrings(
    """
    BridgeTower Model with a language modeling head on top as done during pretraining.
    """,
    BRIDGETOWER_START_DOCSTRING,
)
class BridgeTowerForMaskedLM(BridgeTowerPreTrainedModel):
    _tied_weights_keys = ["mlm_score.decoder.weight"]

    def __init__(self, config):
        super().__init__(config)

        self.bridgetower = BridgeTowerModel(config)
        self.mlm_score = BridgeTowerMLMHead(config)

        # Initialize weights and apply final processing
        self.post_init()

    def get_output_embeddings(self):
        return self.mlm_score.decoder

    def set_output_embeddings(self, new_embeddings):
        self.mlm_score.decoder = new_embeddings

    @add_start_docstrings_to_model_forward(BRIDGETOWER_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @replace_return_docstrings(output_type=MaskedLMOutput, config_class=_CONFIG_FOR_DOC)
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        pixel_values: Optional[torch.FloatTensor] = None,
        pixel_mask: Optional[torch.LongTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        image_embeds: Optional[torch.FloatTensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        labels: Optional[torch.LongTensor] = None,
    ) -> Union[MaskedLMOutput, Tuple[torch.FloatTensor]]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ...,
            config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the
            loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
        Returns:

        Examples:

        ```python
        >>> from transformers import BridgeTowerProcessor, BridgeTowerForMaskedLM
        >>> from PIL import Image
        >>> import requests

        >>> url = "http://images.cocodataset.org/val2017/000000360943.jpg"
        >>> image = Image.open(requests.get(url, stream=True).raw).convert("RGB")
        >>> text = "a <mask> looking out of the window"

        >>> processor = BridgeTowerProcessor.from_pretrained("BridgeTower/bridgetower-base-itm-mlm")
        >>> model = BridgeTowerForMaskedLM.from_pretrained("BridgeTower/bridgetower-base-itm-mlm")

        >>> # prepare inputs
        >>> encoding = processor(image, text, return_tensors="pt")

        >>> # forward pass
        >>> outputs = model(**encoding)

        >>> results = processor.decode(outputs.logits.argmax(dim=-1).squeeze(0).tolist())

        >>> print(results)
        .a cat looking out of the window.
        ```"""
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
        outputs = self.bridgetower(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            pixel_values=pixel_values,
            pixel_mask=pixel_mask,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            image_embeds=image_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        mlm_logits = self.mlm_score(outputs.text_features if return_dict else outputs[0])
        masked_lm_loss = None
        if labels is not None:
            loss_fct = CrossEntropyLoss()  # -100 index = padding token

            labels = labels.to(mlm_logits.device)
            masked_lm_loss = loss_fct(mlm_logits.view(-1, self.config.text_config.vocab_size), labels.view(-1))

        if not return_dict:
            output = tuple(mlm_logits)
            return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output

        return MaskedLMOutput(
            loss=masked_lm_loss,
            logits=mlm_logits,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )


@add_start_docstrings(
    """
    BridgeTower Model transformer with a classifier head on top (a linear layer on top of the final hidden state of the
    [CLS] token) for image-to-text matching.
    """,
    BRIDGETOWER_START_DOCSTRING,
)
class BridgeTowerForImageAndTextRetrieval(BridgeTowerPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)

        self.bridgetower = BridgeTowerModel(config)

        self.itm_score = BridgeTowerITMHead(config.hidden_size * 2)

        # Initialize weights and apply final processing
        self.post_init()

    @add_start_docstrings_to_model_forward(BRIDGETOWER_INPUTS_DOCSTRING)
    @replace_return_docstrings(output_type=SequenceClassifierOutput, config_class=_CONFIG_FOR_DOC)
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        pixel_values: Optional[torch.FloatTensor] = None,
        pixel_mask: Optional[torch.LongTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        image_embeds: Optional[torch.FloatTensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        labels: Optional[torch.LongTensor] = None,
    ) -> Union[SequenceClassifierOutput, Tuple[torch.FloatTensor]]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size, 1)`, *optional*):
            Labels for computing the image-text matching loss. 0 means the pairs don't match and 1 means they match.
            The pairs with 0 will be skipped for calculation.
        Returns:

        Examples:

        ```python
        >>> from transformers import BridgeTowerProcessor, BridgeTowerForImageAndTextRetrieval
        >>> import requests
        >>> from PIL import Image

        >>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
        >>> image = Image.open(requests.get(url, stream=True).raw)
        >>> texts = ["An image of two cats chilling on a couch", "A football player scoring a goal"]

        >>> processor = BridgeTowerProcessor.from_pretrained("BridgeTower/bridgetower-base-itm-mlm")
        >>> model = BridgeTowerForImageAndTextRetrieval.from_pretrained("BridgeTower/bridgetower-base-itm-mlm")

        >>> # forward pass
        >>> scores = dict()
        >>> for text in texts:
        ...     # prepare inputs
        ...     encoding = processor(image, text, return_tensors="pt")
        ...     outputs = model(**encoding)
        ...     scores[text] = outputs.logits[0, 1].item()
        ```"""
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.bridgetower(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            pixel_values=pixel_values,
            pixel_mask=pixel_mask,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            image_embeds=image_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        pooler_output = outputs.pooler_output if return_dict else outputs[2]

        logits = self.itm_score(pooler_output)

        itm_loss = None
        if labels is not None:
            loss_fct = CrossEntropyLoss()

            labels = labels.to(logits.device)
            itm_loss = loss_fct(logits, labels)

        if not return_dict:
            output = tuple(logits)
            return ((itm_loss,) + output) if itm_loss is not None else output

        return SequenceClassifierOutput(
            loss=itm_loss,
            logits=logits,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )


class BridgeTowerContrastiveHead(nn.Module):
    def __init__(self, hidden_size, embed_size):
        super().__init__()
        self.fc = nn.Linear(hidden_size, embed_size)

    def forward(self, x):
        x = self.fc(x)
        return x


@add_start_docstrings(
    """
    BridgeTower Model with a image-text contrastive head on top computing image-text contrastive loss.
    """,
    BRIDGETOWER_START_DOCSTRING,
)
class BridgeTowerForContrastiveLearning(BridgeTowerPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)

        self.bridgetower = BridgeTowerModel(config)

        self.itc_text_head = BridgeTowerContrastiveHead(config.hidden_size, config.contrastive_hidden_size)
        self.itc_image_head = BridgeTowerContrastiveHead(config.hidden_size, config.contrastive_hidden_size)
        self.itc_cross_modal_head = BridgeTowerContrastiveHead(config.hidden_size * 2, config.contrastive_hidden_size)

        self.logit_scale = nn.Parameter(torch.tensor(self.config.logit_scale_init_value))
        # Initialize weights and apply final processing
        self.post_init()

    @add_start_docstrings_to_model_forward(BRIDGETOWER_INPUTS_DOCSTRING)
    @replace_return_docstrings(output_type=BridgeTowerContrastiveOutput, config_class=_CONFIG_FOR_DOC)
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        pixel_values: Optional[torch.FloatTensor] = None,
        pixel_mask: Optional[torch.LongTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        image_embeds: Optional[torch.FloatTensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = True,
        return_dict: Optional[bool] = None,
        return_loss: Optional[bool] = None,
    ) -> Union[BridgeTowerContrastiveOutput, Tuple[torch.FloatTensor]]:
        r"""
        return_loss (`bool`, *optional*):
            Whether or not to return the contrastive loss.
        Returns:

        Examples:

        ```python
        >>> from transformers import BridgeTowerProcessor, BridgeTowerForContrastiveLearning
        >>> import requests
        >>> from PIL import Image
        >>> import torch

        >>> image_urls = [
        ...     "https://farm4.staticflickr.com/3395/3428278415_81c3e27f15_z.jpg",
        ...     "http://images.cocodataset.org/val2017/000000039769.jpg",
        ... ]
        >>> texts = ["two dogs in a car", "two cats sleeping on a couch"]
        >>> images = [Image.open(requests.get(url, stream=True).raw) for url in image_urls]

        >>> processor = BridgeTowerProcessor.from_pretrained("BridgeTower/bridgetower-large-itm-mlm-itc")
        >>> model = BridgeTowerForContrastiveLearning.from_pretrained("BridgeTower/bridgetower-large-itm-mlm-itc")

        >>> inputs = processor(images, texts, padding=True, return_tensors="pt")
        >>> loss = model(**inputs, return_loss=True).loss

        >>> inputs = processor(images, texts[::-1], padding=True, return_tensors="pt")
        >>> loss_swapped = model(**inputs, return_loss=True).loss

        >>> print("Loss", round(loss.item(), 4))
        Loss 0.0019

        >>> print("Loss with swapped images", round(loss_swapped.item(), 4))
        Loss with swapped images 2.126
        ```"""
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.bridgetower(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            pixel_values=pixel_values,
            pixel_mask=pixel_mask,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            image_embeds=image_embeds,
            output_attentions=output_attentions,
            output_hidden_states=True,
            return_dict=return_dict,
        )

        pooler_output = outputs.pooler_output if return_dict else outputs[2]
        hidden_states_txt, hidden_states_img, hidden_states_cross_modal = (
            outputs.hidden_states if return_dict else outputs[3]
        )

        text_embeds = hidden_states_txt[-1]
        image_embeds = hidden_states_img[-1]

        image_embeds_with_ln = self.bridgetower.vision_model.visual.forward_post(image_embeds)
        image_token_type_embeddings = self.bridgetower.token_type_embeddings(
            torch.full((1,), 1, dtype=torch.long, device=self.bridgetower.token_type_embeddings.weight.device)
        ).expand_as(image_embeds_with_ln)

        image_embeds = self.bridgetower.cross_modal_image_transform(image_embeds_with_ln) + image_token_type_embeddings

        # normalized features
        text_embeds = nn.functional.normalize(self.itc_text_head(text_embeds[:, 0, :]), dim=-1, p=2)
        image_embeds = nn.functional.normalize(self.itc_image_head(image_embeds[:, 0, :]), dim=-1, p=2).to(
            device=text_embeds.device
        )
        cross_embeds = nn.functional.normalize(self.itc_cross_modal_head(pooler_output), dim=-1, p=2).to(
            device=text_embeds.device
        )

        logits = torch.stack([text_embeds, image_embeds, cross_embeds], dim=-2)

        logit_scale = self.logit_scale.exp().to(device=text_embeds.device)
        logits_text_to_image = torch.matmul(text_embeds, image_embeds.t()) * logit_scale
        logits_text_to_cross = torch.matmul(text_embeds, cross_embeds.t()) * logit_scale
        logits_image_to_cross = torch.matmul(image_embeds, cross_embeds.t()) * logit_scale

        itc_loss = None

        if return_loss:
            labels = torch.arange(len(logits), device=logits.device)
            text_to_image_loss = nn.functional.cross_entropy(logits_text_to_image, labels)
            text_to_cross_loss = nn.functional.cross_entropy(logits_text_to_cross, labels)
            image_to_cross_loss = nn.functional.cross_entropy(logits_image_to_cross, labels)
            itc_loss = (text_to_image_loss + text_to_cross_loss + image_to_cross_loss) / 3.0

        if not return_dict:
            output = (logits, text_embeds, image_embeds, cross_embeds) + outputs[3:]
            return ((itc_loss,) + output) if itc_loss is not None else output

        return BridgeTowerContrastiveOutput(
            loss=itc_loss,
            logits=logits,
            text_embeds=text_embeds,
            image_embeds=image_embeds,
            cross_embeds=cross_embeds,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )