File size: 58,261 Bytes
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
86df0b7
aeb03e3
 
 
 
 
86df0b7
 
 
 
 
 
 
 
bd16ff6
aeb03e3
 
bd16ff6
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bd16ff6
aeb03e3
 
 
 
 
bd16ff6
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bd16ff6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
aeb03e3
bd16ff6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
aeb03e3
 
bd16ff6
 
 
 
 
aeb03e3
bd16ff6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
aeb03e3
 
bd16ff6
 
 
 
 
 
 
 
 
 
 
 
 
 
aeb03e3
 
bd16ff6
 
 
 
 
 
 
aeb03e3
bd16ff6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bd16ff6
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bd16ff6
 
 
 
 
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bd16ff6
 
 
 
 
 
 
 
 
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bd16ff6
 
aeb03e3
 
 
 
 
 
bd16ff6
 
 
 
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bd16ff6
 
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bd16ff6
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bd16ff6
 
 
 
 
 
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
44011aa
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bd16ff6
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
bd16ff6
aeb03e3
 
 
 
 
911e84c
aeb03e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
""" PyTorch ChatGLM model. """
import json
import math
import copy
import warnings
import re
import sys

import torch
import torch.utils.checkpoint
import torch.nn.functional as F
from torch import nn
from torch.nn import CrossEntropyLoss, LayerNorm, MSELoss, BCEWithLogitsLoss
from torch.nn.utils import skip_init
from typing import Optional, Tuple, Union, List, Callable, Dict, Any
from copy import deepcopy

from transformers.modeling_outputs import (
    BaseModelOutputWithPast,
    CausalLMOutputWithPast,
    SequenceClassifierOutputWithPast,
)
from transformers.modeling_utils import PreTrainedModel
from transformers.utils import logging, is_torch_npu_available
from transformers.generation.logits_process import LogitsProcessor
from transformers.generation.utils import LogitsProcessorList, StoppingCriteriaList, GenerationConfig, ModelOutput

from .configuration_chatglm import ChatGLMConfig

try:
    from transformers.utils import is_flash_attn_greater_or_equal_2_10, is_flash_attn_2_available
    if is_flash_attn_2_available():
        from flash_attn import flash_attn_func, flash_attn_varlen_func
        from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input  # noqa
except:
    pass


# flags required to enable jit fusion kernels

if sys.platform != 'darwin' and not is_torch_npu_available():
    torch._C._jit_set_profiling_mode(False)
    torch._C._jit_set_profiling_executor(False)
    torch._C._jit_override_can_fuse_on_cpu(True)
    torch._C._jit_override_can_fuse_on_gpu(True)

logger = logging.get_logger(__name__)

_CHECKPOINT_FOR_DOC = "THUDM/ChatGLM"
_CONFIG_FOR_DOC = "ChatGLMConfig"


def default_init(cls, *args, **kwargs):
    return cls(*args, **kwargs)


class InvalidScoreLogitsProcessor(LogitsProcessor):
    def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor) -> torch.FloatTensor:
        if torch.isnan(scores).any() or torch.isinf(scores).any():
            scores.zero_()
            scores[..., 198] = 5e4
        return scores


def split_tensor_along_last_dim(
        tensor: torch.Tensor,
        num_partitions: int,
        contiguous_split_chunks: bool = False,
) -> List[torch.Tensor]:
    """Split a tensor along its last dimension.

    Arguments:
        tensor: input tensor.
        num_partitions: number of partitions to split the tensor
        contiguous_split_chunks: If True, make each chunk contiguous
                                 in memory.

    Returns:
        A list of Tensors
    """
    # Get the size and dimension.
    last_dim = tensor.dim() - 1
    last_dim_size = tensor.size()[last_dim] // num_partitions
    # Split.
    tensor_list = torch.split(tensor, last_dim_size, dim=last_dim)
    # Note: torch.split does not create contiguous tensors by default.
    if contiguous_split_chunks:
        return tuple(chunk.contiguous() for chunk in tensor_list)

    return tensor_list


class RotaryEmbedding(nn.Module):
    def __init__(self, dim, rope_ratio=1, original_impl=False, device=None, dtype=None):
        super().__init__()
        inv_freq = 1.0 / (10000 ** (torch.arange(0, dim, 2, device=device).to(dtype=dtype) / dim))
        self.register_buffer("inv_freq", inv_freq)
        self.dim = dim
        self.original_impl = original_impl
        self.rope_ratio = rope_ratio

    def forward_impl(
            self, seq_len: int, n_elem: int, dtype: torch.dtype, device: torch.device, base: int = 10000
    ):
        """Enhanced Transformer with Rotary Position Embedding.

        Derived from: https://github.com/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/
        transformers/rope/__init__.py. MIT License:
        https://github.com/labmlai/annotated_deep_learning_paper_implementations/blob/master/license.
        """
        # $\Theta = {\theta_i = 10000^{\frac{2(i-1)}{d}}, i \in [1, 2, ..., \frac{d}{2}]}$
        base = base * self.rope_ratio
        theta = 1.0 / (base ** (torch.arange(0, n_elem, 2, dtype=torch.float, device=device) / n_elem))

        # Create position indexes `[0, 1, ..., seq_len - 1]`
        seq_idx = torch.arange(seq_len, dtype=torch.float, device=device)

        # Calculate the product of position index and $\theta_i$
        idx_theta = torch.outer(seq_idx, theta).float()

        cache = torch.stack([torch.cos(idx_theta), torch.sin(idx_theta)], dim=-1)

        # this is to mimic the behaviour of complex32, else we will get different results
        if dtype in (torch.float16, torch.bfloat16, torch.int8):
            cache = cache.bfloat16() if dtype == torch.bfloat16 else cache.half()
        return cache

    def forward(self, max_seq_len, offset=0):
        return self.forward_impl(
            max_seq_len, self.dim, dtype=self.inv_freq.dtype, device=self.inv_freq.device
        )


@torch.jit.script
def apply_rotary_pos_emb(x: torch.Tensor, rope_cache: torch.Tensor) -> torch.Tensor:
    # x: [b, np, sq, hn]
    b, np, sq, hn = x.size(0), x.size(1), x.size(2), x.size(3)
    rot_dim = rope_cache.shape[-2] * 2
    x, x_pass = x[..., :rot_dim], x[..., rot_dim:]
    # truncate to support variable sizes
    rope_cache = rope_cache[:, :sq]
    xshaped = x.reshape(b, np, sq, rot_dim // 2, 2)
    rope_cache = rope_cache.view(-1, 1, sq, xshaped.size(3), 2)
    x_out2 = torch.stack(
        [
            xshaped[..., 0] * rope_cache[..., 0] - xshaped[..., 1] * rope_cache[..., 1],
            xshaped[..., 1] * rope_cache[..., 0] + xshaped[..., 0] * rope_cache[..., 1],
        ],
        -1,
    )
    x_out2 = x_out2.flatten(3)
    return torch.cat((x_out2, x_pass), dim=-1)


class RMSNorm(torch.nn.Module):
    def __init__(self, normalized_shape, eps=1e-5, device=None, dtype=None, **kwargs):
        super().__init__()
        self.weight = torch.nn.Parameter(torch.empty(normalized_shape, device=device, dtype=dtype))
        self.eps = eps

    def forward(self, hidden_states: torch.Tensor):
        input_dtype = hidden_states.dtype
        variance = hidden_states.to(torch.float32).pow(2).mean(-1, keepdim=True)
        hidden_states = hidden_states * torch.rsqrt(variance + self.eps)

        return (self.weight * hidden_states).to(input_dtype)


class CoreAttention(torch.nn.Module):
    def __init__(self, config: ChatGLMConfig, layer_number):
        super(CoreAttention, self).__init__()
        self.config = config
        self.apply_query_key_layer_scaling = config.apply_query_key_layer_scaling
        self.attention_softmax_in_fp32 = config.attention_softmax_in_fp32
        if self.apply_query_key_layer_scaling:
            self.attention_softmax_in_fp32 = True
        self.layer_number = max(1, layer_number)
        self.is_causal = True

        projection_size = config.kv_channels * config.num_attention_heads

        # Per attention head and per partition values.
        self.hidden_size_per_partition = projection_size
        self.hidden_size_per_attention_head = projection_size // config.num_attention_heads
        self.num_attention_heads_per_partition = config.num_attention_heads

        coeff = None
        self.norm_factor = math.sqrt(self.hidden_size_per_attention_head)
        if self.apply_query_key_layer_scaling:
            coeff = self.layer_number
            self.norm_factor *= coeff
        self.coeff = coeff

        self.attention_dropout = torch.nn.Dropout(config.attention_dropout)

    def forward(self, query_layer, key_layer, value_layer, attention_mask):
        # [b, np, sq, sk]
        output_size = (query_layer.size(0), query_layer.size(1), query_layer.size(2), key_layer.size(2))

        # [b, np, sq, hn] -> [b * np, sq, hn]
        query_layer = query_layer.view(output_size[0] * output_size[1], output_size[2], -1)
        # [b, np, sk, hn] -> [b * np, sk, hn]
        key_layer = key_layer.view(output_size[0] * output_size[1], output_size[3], -1)

        # preallocting input tensor: [b * np, sq, sk]
        matmul_input_buffer = torch.empty(
            output_size[0] * output_size[1], output_size[2], output_size[3], dtype=query_layer.dtype,
            device=query_layer.device
        )

        # Raw attention scores. [b * np, sq, sk]
        matmul_result = torch.baddbmm(
            matmul_input_buffer,
            query_layer,  # [b * np, sq, hn]
            key_layer.transpose(1, 2),  # [b * np, hn, sk]
            beta=0.0,
            alpha=(1.0 / self.norm_factor),
        )

        # change view to [b, np, sq, sk]
        attention_scores = matmul_result.view(*output_size)

        # ===========================
        # Attention probs and dropout
        # ===========================

        # attention scores and attention mask [b, np, sq, sk]
        if self.attention_softmax_in_fp32:
            attention_scores = attention_scores.float()
        if self.coeff is not None:
            attention_scores = attention_scores * self.coeff
        if attention_mask is None and attention_scores.shape[2] == attention_scores.shape[3]:
            attention_mask = torch.ones(output_size[0], 1, output_size[2], output_size[3],
                                        device=attention_scores.device, dtype=torch.bool)
            attention_mask.tril_()
            attention_mask = ~attention_mask
        if attention_mask is not None:
            attention_scores = attention_scores.masked_fill(attention_mask, float("-inf"))
        attention_probs = F.softmax(attention_scores, dim=-1)
        attention_probs = attention_probs.type_as(value_layer)

        # 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.attention_dropout(attention_probs)

        # query layer shape: [b * np, sq, hn]
        # value layer shape: [b, np, sk, hn]
        # attention shape: [b, np, sq, sk]
        # context layer shape: [b, np, sq, hn]
        output_size = (value_layer.size(0), value_layer.size(1), query_layer.size(1), value_layer.size(3))
        # change view [b * np, sk, hn]
        value_layer = value_layer.view(output_size[0] * output_size[1], value_layer.size(2), -1)
        # change view [b * np, sq, sk]
        attention_probs = attention_probs.view(output_size[0] * output_size[1], output_size[2], -1)
        # matmul: [b * np, sq, hn]
        context_layer = torch.bmm(attention_probs, value_layer)
        # change view [b, np, sq, hn]
        context_layer = context_layer.view(*output_size)
        # [b, np, sq, hn] --> [b, sq, np, hn]
        context_layer = context_layer.transpose(1, 2).contiguous()
        # [b, sq, np, hn] --> [b, sq, hp]
        new_context_layer_shape = context_layer.size()[:-2] + (self.hidden_size_per_partition,)
        context_layer = context_layer.reshape(*new_context_layer_shape)

        return context_layer


class SdpaAttention(CoreAttention):
    def forward(self, query_layer, key_layer, value_layer, attention_mask):
        if attention_mask is None and query_layer.shape[2] == key_layer.shape[2]:
            context_layer = torch.nn.functional.scaled_dot_product_attention(query_layer, key_layer, value_layer,
                                                                             is_causal=True,
                                                                             dropout_p=self.config.attention_dropout if self.training else 0.0)
        else:
            if attention_mask is not None:
                attention_mask = ~attention_mask
            context_layer = torch.nn.functional.scaled_dot_product_attention(query_layer, key_layer, value_layer,
                                                                             attention_mask,
                                                                             dropout_p=self.config.attention_dropout if self.training else 0.0)
        context_layer = context_layer.transpose(1, 2).contiguous()
        new_context_layer_shape = context_layer.size()[:-2] + (self.hidden_size_per_partition,)
        context_layer = context_layer.reshape(*new_context_layer_shape)
        return context_layer


def _get_unpad_data(attention_mask):
    seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
    indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
    max_seqlen_in_batch = seqlens_in_batch.max().item()
    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
    return (
        indices,
        cu_seqlens,
        max_seqlen_in_batch,
    )


# Copied from transformers.models.llama.modeling_llama.LlamaFlashAttention2
class FlashAttention2(CoreAttention):
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self._flash_attn_uses_top_left_mask = not is_flash_attn_greater_or_equal_2_10()

    def forward(self, query_states, key_states, value_states, attention_mask):
        query_states = query_states.transpose(1, 2)
        key_states = key_states.transpose(1, 2)
        value_states = value_states.transpose(1, 2)
        batch_size, query_length = query_states.shape[:2]
        if not self._flash_attn_uses_top_left_mask:
            causal = self.is_causal
        else:
            # TODO: Remove the `query_length != 1` check once Flash Attention for RoCm is bumped to 2.1. For details, please see the comment in LlamaFlashAttention2 __init__.
            causal = self.is_causal and query_length != 1
        dropout = self.config.attention_dropout if self.training else 0.0
        # Contains at least one padding token in the sequence
        if attention_mask is not None:
            query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input(
                query_states, key_states, value_states, attention_mask, query_length
            )

            cu_seqlens_q, cu_seqlens_k = cu_seq_lens
            max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens

            attn_output_unpad = flash_attn_varlen_func(
                query_states,
                key_states,
                value_states,
                cu_seqlens_q=cu_seqlens_q,
                cu_seqlens_k=cu_seqlens_k,
                max_seqlen_q=max_seqlen_in_batch_q,
                max_seqlen_k=max_seqlen_in_batch_k,
                dropout_p=dropout,
                softmax_scale=None,
                causal=causal,
            )

            attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length)
        else:
            attn_output = flash_attn_func(
                query_states, key_states, value_states, dropout, softmax_scale=None, causal=causal
            )
        attn_output = attn_output.reshape(batch_size, query_length, self.hidden_size_per_partition).contiguous()
        return attn_output

    def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length):
        indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask)
        batch_size, kv_seq_len, num_key_value_heads, head_dim = key_layer.shape

        key_layer = index_first_axis(
            key_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k
        )
        value_layer = index_first_axis(
            value_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k
        )
        if query_length == kv_seq_len:
            query_layer = index_first_axis(
                query_layer.reshape(batch_size * kv_seq_len, self.num_attention_heads_per_partition, head_dim), indices_k
            )
            cu_seqlens_q = cu_seqlens_k
            max_seqlen_in_batch_q = max_seqlen_in_batch_k
            indices_q = indices_k
        elif query_length == 1:
            max_seqlen_in_batch_q = 1
            cu_seqlens_q = torch.arange(
                batch_size + 1, dtype=torch.int32, device=query_layer.device
            )  # There is a memcpy here, that is very bad.
            indices_q = cu_seqlens_q[:-1]
            query_layer = query_layer.squeeze(1)
        else:
            # The -q_len: slice assumes left padding.
            attention_mask = attention_mask[:, -query_length:]
            query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask)

        return (
            query_layer,
            key_layer,
            value_layer,
            indices_q,
            (cu_seqlens_q, cu_seqlens_k),
            (max_seqlen_in_batch_q, max_seqlen_in_batch_k),
        )


CORE_ATTENTION_CLASSES = {
    "eager": CoreAttention,
    "sdpa": SdpaAttention,
    "flash_attention_2": FlashAttention2
}


class SelfAttention(torch.nn.Module):
    """Parallel self-attention layer abstract class.

    Self-attention layer takes input with size [s, b, h]
    and returns output of the same size.
    """

    def __init__(self, config: ChatGLMConfig, layer_number, device=None):
        super(SelfAttention, self).__init__()
        self.layer_number = max(1, layer_number)

        self.projection_size = config.kv_channels * config.num_attention_heads

        # Per attention head and per partition values.
        self.hidden_size_per_attention_head = self.projection_size // config.num_attention_heads
        self.num_attention_heads_per_partition = config.num_attention_heads

        self.multi_query_attention = config.multi_query_attention
        self.qkv_hidden_size = 3 * self.projection_size
        if self.multi_query_attention:
            self.num_multi_query_groups_per_partition = config.multi_query_group_num
            self.qkv_hidden_size = (
                    self.projection_size + 2 * self.hidden_size_per_attention_head * config.multi_query_group_num
            )
        self.query_key_value = nn.Linear(config.hidden_size, self.qkv_hidden_size,
                                         bias=config.add_bias_linear or config.add_qkv_bias,
                                         device=device, **_config_to_kwargs(config)
                                         )

        self.core_attention = CORE_ATTENTION_CLASSES[config._attn_implementation](config, self.layer_number)

        # Output.
        self.dense = nn.Linear(self.projection_size, config.hidden_size, bias=config.add_bias_linear,
                               device=device, **_config_to_kwargs(config)
                               )

    def _allocate_memory(self, inference_max_sequence_len, batch_size, device=None, dtype=None):
        if self.multi_query_attention:
            num_attention_heads = self.num_multi_query_groups_per_partition
        else:
            num_attention_heads = self.num_attention_heads_per_partition
        return torch.empty(
            inference_max_sequence_len,
            batch_size,
            num_attention_heads,
            self.hidden_size_per_attention_head,
            dtype=dtype,
            device=device,
        )

    def forward(
            self, hidden_states, attention_mask, rotary_pos_emb, kv_cache=None, use_cache=True
    ):
        # hidden_states: [b, sq, h]

        # =================================================
        # Pre-allocate memory for key-values for inference.
        # =================================================
        # =====================
        # Query, Key, and Value
        # =====================

        # Attention heads [b, sq, h] --> [b, sq, (np * 3 * hn)]
        mixed_x_layer = self.query_key_value(hidden_states)

        if self.multi_query_attention:
            (query_layer, key_layer, value_layer) = mixed_x_layer.split(
                [
                    self.num_attention_heads_per_partition * self.hidden_size_per_attention_head,
                    self.num_multi_query_groups_per_partition * self.hidden_size_per_attention_head,
                    self.num_multi_query_groups_per_partition * self.hidden_size_per_attention_head,
                ],
                dim=-1,
            )
            query_layer = query_layer.view(
                query_layer.size()[:-1] + (self.num_attention_heads_per_partition, self.hidden_size_per_attention_head)
            )
            key_layer = key_layer.view(
                key_layer.size()[:-1] + (self.num_multi_query_groups_per_partition, self.hidden_size_per_attention_head)
            )
            value_layer = value_layer.view(
                value_layer.size()[:-1]
                + (self.num_multi_query_groups_per_partition, self.hidden_size_per_attention_head)
            )
        else:
            new_tensor_shape = mixed_x_layer.size()[:-1] + \
                               (self.num_attention_heads_per_partition,
                                3 * self.hidden_size_per_attention_head)
            mixed_x_layer = mixed_x_layer.view(*new_tensor_shape)

            # [b, sq, np, 3 * hn] --> 3 [b, sq, np, hn]
            (query_layer, key_layer, value_layer) = split_tensor_along_last_dim(mixed_x_layer, 3)

        # [b, sq, np, hn] -> [b, np, sq, hn]
        query_layer, key_layer, value_layer = [k.transpose(1, 2) for k in [query_layer, key_layer, value_layer]]

        # apply relative positional encoding (rotary embedding)
        if rotary_pos_emb is not None:
            query_layer = apply_rotary_pos_emb(query_layer, rotary_pos_emb)
            key_layer = apply_rotary_pos_emb(key_layer, rotary_pos_emb)

        # adjust key and value for inference
        if kv_cache is not None:
            cache_k, cache_v = kv_cache
            key_layer = torch.cat((cache_k, key_layer), dim=2)
            value_layer = torch.cat((cache_v, value_layer), dim=2)
        if use_cache:
            if kv_cache is None:
                kv_cache = torch.cat((key_layer.unsqueeze(0).unsqueeze(0), value_layer.unsqueeze(0).unsqueeze(0)),
                                     dim=1)
            else:
                kv_cache = (key_layer, value_layer)
        else:
            kv_cache = None

        if self.multi_query_attention:
            key_layer = key_layer.unsqueeze(2)
            key_layer = key_layer.expand(
                -1, -1, self.num_attention_heads_per_partition // self.num_multi_query_groups_per_partition, -1, -1
            )
            key_layer = key_layer.contiguous().view(
                key_layer.size()[:1] + (self.num_attention_heads_per_partition,) + key_layer.size()[3:]
            )
            value_layer = value_layer.unsqueeze(2)
            value_layer = value_layer.expand(
                -1, -1, self.num_attention_heads_per_partition // self.num_multi_query_groups_per_partition, -1, -1
            )
            value_layer = value_layer.contiguous().view(
                value_layer.size()[:1] + (self.num_attention_heads_per_partition,) + value_layer.size()[3:]
            )

        # ==================================
        # core attention computation
        # ==================================

        context_layer = self.core_attention(query_layer, key_layer, value_layer, attention_mask)

        # =================
        # Output. [sq, b, h]
        # =================

        output = self.dense(context_layer)

        return output, kv_cache


def _config_to_kwargs(args):
    common_kwargs = {
        "dtype": args.torch_dtype,
    }
    return common_kwargs


class MLP(torch.nn.Module):
    """MLP.

    MLP will take the input with h hidden state, project it to 4*h
    hidden dimension, perform nonlinear transformation, and project the
    state back into h hidden dimension.
    """

    def __init__(self, config: ChatGLMConfig, device=None):
        super(MLP, self).__init__()

        self.add_bias = config.add_bias_linear

        # Project to 4h. If using swiglu double the output width, see https://arxiv.org/pdf/2002.05202.pdf
        self.dense_h_to_4h = nn.Linear(
            config.hidden_size,
            config.ffn_hidden_size * 2,
            bias=self.add_bias,
            device=device,
            **_config_to_kwargs(config)
        )

        def swiglu(x):
            x = torch.chunk(x, 2, dim=-1)
            return F.silu(x[0]) * x[1]

        self.activation_func = swiglu

        # Project back to h.
        self.dense_4h_to_h = nn.Linear(
            config.ffn_hidden_size,
            config.hidden_size,
            bias=self.add_bias,
            device=device,
            **_config_to_kwargs(config)
        )

    def forward(self, hidden_states):
        # [s, b, 4hp]
        intermediate_parallel = self.dense_h_to_4h(hidden_states)
        intermediate_parallel = self.activation_func(intermediate_parallel)
        # [s, b, h]
        output = self.dense_4h_to_h(intermediate_parallel)
        return output


class GLMBlock(torch.nn.Module):
    """A single transformer layer.

    Transformer layer takes input with size [s, b, h] and returns an
    output of the same size.
    """

    def __init__(self, config: ChatGLMConfig, layer_number, device=None):
        super(GLMBlock, self).__init__()
        self.layer_number = layer_number

        self.apply_residual_connection_post_layernorm = config.apply_residual_connection_post_layernorm

        self.fp32_residual_connection = config.fp32_residual_connection

        LayerNormFunc = RMSNorm if config.rmsnorm else LayerNorm
        # Layernorm on the input data.
        self.input_layernorm = LayerNormFunc(config.hidden_size, eps=config.layernorm_epsilon, device=device,
                                             dtype=config.torch_dtype)

        # Self attention.
        self.self_attention = SelfAttention(config, layer_number, device=device)
        self.hidden_dropout = config.hidden_dropout

        # Layernorm on the attention output
        self.post_attention_layernorm = LayerNormFunc(config.hidden_size, eps=config.layernorm_epsilon, device=device,
                                                      dtype=config.torch_dtype)

        # MLP
        self.mlp = MLP(config, device=device)

    def forward(
            self, hidden_states, attention_mask, rotary_pos_emb, kv_cache=None, use_cache=True,
    ):
        # hidden_states: [s, b, h]

        # Layer norm at the beginning of the transformer layer.
        layernorm_output = self.input_layernorm(hidden_states)
        # Self attention.
        attention_output, kv_cache = self.self_attention(
            layernorm_output,
            attention_mask,
            rotary_pos_emb,
            kv_cache=kv_cache,
            use_cache=use_cache
        )

        # Residual connection.
        if self.apply_residual_connection_post_layernorm:
            residual = layernorm_output
        else:
            residual = hidden_states

        layernorm_input = torch.nn.functional.dropout(attention_output, p=self.hidden_dropout, training=self.training)
        layernorm_input = residual + layernorm_input

        # Layer norm post the self attention.
        layernorm_output = self.post_attention_layernorm(layernorm_input)

        # MLP.
        mlp_output = self.mlp(layernorm_output)

        # Second residual connection.
        if self.apply_residual_connection_post_layernorm:
            residual = layernorm_output
        else:
            residual = layernorm_input

        output = torch.nn.functional.dropout(mlp_output, p=self.hidden_dropout, training=self.training)
        output = residual + output

        return output, kv_cache


class GLMTransformer(torch.nn.Module):
    """Transformer class."""

    def __init__(self, config: ChatGLMConfig, device=None):
        super(GLMTransformer, self).__init__()

        self.fp32_residual_connection = config.fp32_residual_connection
        self.post_layer_norm = config.post_layer_norm

        # Number of layers.
        self.num_layers = config.num_layers

        # Transformer layers.
        def build_layer(layer_number):
            return GLMBlock(config, layer_number, device=device)

        self.layers = torch.nn.ModuleList([build_layer(i + 1) for i in range(self.num_layers)])

        if self.post_layer_norm:
            LayerNormFunc = RMSNorm if config.rmsnorm else LayerNorm
            # Final layer norm before output.
            self.final_layernorm = LayerNormFunc(config.hidden_size, eps=config.layernorm_epsilon, device=device,
                                                 dtype=config.torch_dtype)

        self.gradient_checkpointing = False

    def _get_layer(self, layer_number):
        return self.layers[layer_number]

    def forward(
            self, hidden_states, attention_mask, rotary_pos_emb, kv_caches=None,
            use_cache: Optional[bool] = True,
            output_hidden_states: Optional[bool] = False,
    ):
        if not kv_caches:
            kv_caches = [None for _ in range(self.num_layers)]
        presents = () if use_cache 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

        all_self_attentions = None
        all_hidden_states = () if output_hidden_states else None
        for index in range(self.num_layers):
            if output_hidden_states:
                all_hidden_states = all_hidden_states + (hidden_states,)

            layer = self._get_layer(index)
            if self.gradient_checkpointing and self.training:
                layer_ret = torch.utils.checkpoint.checkpoint(
                    layer,
                    hidden_states,
                    attention_mask,
                    rotary_pos_emb,
                    kv_caches[index],
                    use_cache,
                    use_reentrant=False
                )
            else:
                layer_ret = layer(
                    hidden_states,
                    attention_mask,
                    rotary_pos_emb,
                    kv_cache=kv_caches[index],
                    use_cache=use_cache
                )
            hidden_states, kv_cache = layer_ret
            if use_cache:
                # token by token decoding, use tuple format
                if kv_caches[0] is not None:
                    presents = presents + (kv_cache,)
                # prefilling in decoding, use tensor format to save cuda memory
                else:
                    if len(presents) == 0:
                        presents = kv_cache
                    else:
                        presents = torch.cat((presents, kv_cache.to(presents.device)), dim=0)

        if output_hidden_states:
            all_hidden_states = all_hidden_states + (hidden_states,)

        # Final layer norm.
        if self.post_layer_norm:
            hidden_states = self.final_layernorm(hidden_states)

        return hidden_states, presents, all_hidden_states, all_self_attentions


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

    is_parallelizable = False
    supports_gradient_checkpointing = True
    config_class = ChatGLMConfig
    base_model_prefix = "transformer"
    _no_split_modules = ["GLMBlock"]
    _supports_flash_attn_2 = True
    _supports_sdpa = True

    def _init_weights(self, module: nn.Module):
        """Initialize the weights."""
        return

    def get_masks(self, input_ids, past_key_values, padding_mask=None):
        if self.config._attn_implementation == "flash_attention_2":
            if padding_mask is not None and not padding_mask.all():
                return padding_mask
            return None
        batch_size, seq_length = input_ids.shape
        full_attention_mask = torch.ones(batch_size, seq_length, seq_length, device=input_ids.device)
        full_attention_mask.tril_()
        past_length = 0
        if past_key_values:
            past_length = past_key_values[0][0].shape[2]
        if past_length:
            full_attention_mask = torch.cat((torch.ones(batch_size, seq_length, past_length,
                                                        device=input_ids.device), full_attention_mask), dim=-1)
        if padding_mask is not None:
            full_attention_mask = full_attention_mask * padding_mask.unsqueeze(1)
        if not past_length and padding_mask is not None:
            full_attention_mask -= padding_mask.unsqueeze(-1) - 1
        full_attention_mask = (full_attention_mask < 0.5).bool()
        full_attention_mask.unsqueeze_(1)
        return full_attention_mask

    def get_position_ids(self, input_ids, device):
        batch_size, seq_length = input_ids.shape
        position_ids = torch.arange(seq_length, dtype=torch.long, device=device).unsqueeze(0).repeat(batch_size, 1)
        return position_ids


class Embedding(torch.nn.Module):
    """Language model embeddings."""

    def __init__(self, config: ChatGLMConfig, device=None):
        super(Embedding, self).__init__()

        self.hidden_size = config.hidden_size
        # Word embeddings (parallel).
        self.word_embeddings = nn.Embedding(
            config.padded_vocab_size,
            self.hidden_size,
            dtype=config.torch_dtype,
            device=device
        )
        self.fp32_residual_connection = config.fp32_residual_connection

    def forward(self, input_ids):
        # Embeddings.
        words_embeddings = self.word_embeddings(input_ids)
        embeddings = words_embeddings
        # If the input flag for fp32 residual connection is set, convert for float.
        if self.fp32_residual_connection:
            embeddings = embeddings.float()
        return embeddings


class ChatGLMModel(ChatGLMPreTrainedModel):
    def __init__(self, config: ChatGLMConfig, device=None, empty_init=True):
        super().__init__(config)
        if empty_init:
            init_method = skip_init
        else:
            init_method = default_init
        init_kwargs = {}
        if device is not None:
            init_kwargs["device"] = device
        self.embedding = init_method(Embedding, config, **init_kwargs)
        self.num_layers = config.num_layers
        self.multi_query_group_num = config.multi_query_group_num
        self.kv_channels = config.kv_channels

        # Rotary positional embeddings
        self.seq_length = config.seq_length
        rotary_dim = (
            config.hidden_size // config.num_attention_heads if config.kv_channels is None else config.kv_channels
        )

        self.rotary_pos_emb = RotaryEmbedding(rotary_dim // 2, rope_ratio=config.rope_ratio,
                                              original_impl=config.original_rope,
                                              device=device, dtype=config.torch_dtype)
        self.encoder = init_method(GLMTransformer, config, **init_kwargs)
        self.output_layer = init_method(nn.Linear, config.hidden_size, config.padded_vocab_size, bias=False,
                                        dtype=config.torch_dtype, **init_kwargs)

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

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

    def forward(
            self,
            input_ids,
            position_ids: Optional[torch.Tensor] = None,
            attention_mask: Optional[torch.BoolTensor] = None,
            full_attention_mask: Optional[torch.BoolTensor] = None,
            past_key_values: Optional[Tuple[Tuple[torch.Tensor, torch.Tensor], ...]] = None,
            inputs_embeds: Optional[torch.Tensor] = None,
            use_cache: Optional[bool] = None,
            output_attentions: Optional[bool] = None,
            output_hidden_states: Optional[bool] = None,
            return_dict: Optional[bool] = None,
    ):
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        use_cache = use_cache if use_cache is not None else self.config.use_cache
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        batch_size, seq_length = input_ids.shape

        if inputs_embeds is None:
            inputs_embeds = self.embedding(input_ids)

        if full_attention_mask is None:
            if (attention_mask is not None and not attention_mask.all()) or (past_key_values and seq_length != 1):
                full_attention_mask = self.get_masks(input_ids, past_key_values, padding_mask=attention_mask)

        # Rotary positional embeddings
        rotary_pos_emb = self.rotary_pos_emb(self.seq_length)
        if position_ids is not None:
            rotary_pos_emb = rotary_pos_emb[position_ids]
        else:
            rotary_pos_emb = rotary_pos_emb[None, :seq_length]

        # Run encoder.
        hidden_states, presents, all_hidden_states, all_self_attentions = self.encoder(
            inputs_embeds, full_attention_mask, rotary_pos_emb=rotary_pos_emb,
            kv_caches=past_key_values, use_cache=use_cache, output_hidden_states=output_hidden_states
        )
        if presents is not None and type(presents) is torch.Tensor:
            presents = presents.split(1, dim=0)
            presents = list(presents)
            presents = [list(x.squeeze(0).split(1, dim=0)) for x in presents]
            presents = [tuple([x.squeeze(0) for x in y]) for y in presents]
            presents = tuple(presents)

        if not return_dict:
            return tuple(v for v in [hidden_states, presents, all_hidden_states, all_self_attentions] if v is not None)

        return BaseModelOutputWithPast(
            last_hidden_state=hidden_states,
            past_key_values=presents,
            hidden_states=all_hidden_states,
            attentions=all_self_attentions,
        )


class ChatGLMForConditionalGeneration(ChatGLMPreTrainedModel):
    def __init__(self, config: ChatGLMConfig, empty_init=True, device=None):
        super().__init__(config)

        self.max_sequence_length = config.max_length
        self.transformer = ChatGLMModel(config, empty_init=empty_init, device=device)
        self.config = config

    def _update_model_kwargs_for_generation(
            self,
            outputs: ModelOutput,
            model_kwargs: Dict[str, Any],
            is_encoder_decoder: bool = False,
            standardize_cache_format: bool = False,
    ) -> Dict[str, Any]:
        # update past_key_values
        model_kwargs["past_key_values"] = self._extract_past_from_model_output(
            outputs, standardize_cache_format=standardize_cache_format
        )

        # update attention mask
        if "attention_mask" in model_kwargs:
            attention_mask = model_kwargs["attention_mask"]
            model_kwargs["attention_mask"] = torch.cat(
                [attention_mask, attention_mask.new_ones((attention_mask.shape[0], 1))], dim=-1
            )

        # update position ids
        if "position_ids" in model_kwargs:
            position_ids = model_kwargs["position_ids"]
            new_position_id = position_ids[..., -1:].clone()
            new_position_id += 1
            model_kwargs["position_ids"] = torch.cat(
                [position_ids, new_position_id], dim=-1
            )

        model_kwargs["is_first_forward"] = False
        return model_kwargs

    def prepare_inputs_for_generation(
            self,
            input_ids: torch.LongTensor,
            past_key_values: Optional[torch.Tensor] = None,
            attention_mask: Optional[torch.Tensor] = None,
            position_ids: Optional[torch.Tensor] = None,
            use_cache: Optional[bool] = None,
            is_first_forward: bool = True,
            **kwargs
    ) -> dict:
        # only last token for input_ids if past is not None
        if position_ids is None:
            position_ids = self.get_position_ids(input_ids, device=input_ids.device)
        if not is_first_forward:
            if past_key_values is not None:
                position_ids = position_ids[..., -1:]
                input_ids = input_ids[:, -1:]
        return {
            "input_ids": input_ids,
            "past_key_values": past_key_values,
            "position_ids": position_ids,
            "attention_mask": attention_mask,
            "return_last_logit": True,
            "use_cache": use_cache
        }

    def forward(
            self,
            input_ids: Optional[torch.Tensor] = None,
            position_ids: Optional[torch.Tensor] = None,
            attention_mask: Optional[torch.Tensor] = None,
            past_key_values: Optional[Tuple[torch.FloatTensor]] = None,
            inputs_embeds: Optional[torch.Tensor] = None,
            labels: Optional[torch.Tensor] = None,
            use_cache: Optional[bool] = None,
            output_attentions: Optional[bool] = None,
            output_hidden_states: Optional[bool] = None,
            return_dict: Optional[bool] = None,
            return_last_logit: Optional[bool] = False,
    ):
        use_cache = use_cache if use_cache is not None else self.config.use_cache
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        transformer_outputs = self.transformer(
            input_ids=input_ids,
            position_ids=position_ids,
            attention_mask=attention_mask,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        hidden_states = transformer_outputs[0]
        if return_last_logit:
            hidden_states = hidden_states[:, -1:]
        lm_logits = self.transformer.output_layer(hidden_states)

        loss = None
        if labels is not None:
            lm_logits = lm_logits.to(torch.float32)

            # Shift so that tokens < n predict n
            shift_logits = lm_logits[..., :-1, :].contiguous()
            shift_labels = labels[..., 1:].contiguous()
            # Flatten the tokens
            loss_fct = CrossEntropyLoss(ignore_index=-100)
            loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), shift_labels.view(-1))

            lm_logits = lm_logits.to(hidden_states.dtype)
            loss = loss.to(hidden_states.dtype)

        if not return_dict:
            output = (lm_logits,) + transformer_outputs[1:]
            return ((loss,) + output) if loss is not None else output

        return CausalLMOutputWithPast(
            loss=loss,
            logits=lm_logits,
            past_key_values=transformer_outputs.past_key_values,
            hidden_states=transformer_outputs.hidden_states,
            attentions=transformer_outputs.attentions,
        )

    @staticmethod
    def _reorder_cache(
            past: Tuple[Tuple[torch.Tensor, torch.Tensor], ...], beam_idx: torch.LongTensor
    ) -> Tuple[Tuple[torch.Tensor, torch.Tensor], ...]:
        """
        This function is used to re-order the `past_key_values` cache if [`~PreTrainedModel.beam_search`] or
        [`~PreTrainedModel.beam_sample`] is called. This is required to match `past_key_values` with the correct
        beam_idx at every generation step.

        Output shares the same memory storage as `past`.
        """
        return tuple(
            (
                layer_past[0].index_select(0, beam_idx.to(layer_past[0].device)),
                layer_past[1].index_select(0, beam_idx.to(layer_past[1].device)),
            )
            for layer_past in past
        )

    def process_response(self, output, history):
        content = ""
        history = deepcopy(history)
        for response in output.split("<|assistant|>"):
            if "\n" in response:
                metadata, content = response.split("\n", maxsplit=1)
            else:
                metadata, content = "", response
            if not metadata.strip():
                content = content.strip()
                history.append({"role": "assistant", "metadata": metadata, "content": content})
                content = content.replace("[[训练时间]]", "2023年")
            else:
                history.append({"role": "assistant", "metadata": metadata, "content": content})
                if history[0]["role"] == "system" and "tools" in history[0]:
                    parameters = json.loads(content)
                    content = {"name": metadata.strip(), "parameters": parameters}
                else:
                    content = {"name": metadata.strip(), "content": content}
        return content, history

    @torch.inference_mode()
    def chat(self, tokenizer, query: str, history: List[Dict] = None, role: str = "user",
             max_length: int = 8192, num_beams=1, do_sample=True, top_p=0.8, temperature=0.8, logits_processor=None,
             **kwargs):
        if history is None:
            history = []
        if logits_processor is None:
            logits_processor = LogitsProcessorList()
        logits_processor.append(InvalidScoreLogitsProcessor())
        gen_kwargs = {"max_length": max_length, "num_beams": num_beams, "do_sample": do_sample, "top_p": top_p,
                      "temperature": temperature, "logits_processor": logits_processor, **kwargs}
        history.append({"role": role, "content": query})
        inputs = tokenizer.apply_chat_template(history, add_generation_prompt=True, tokenize=True,
                                               return_tensors="pt", return_dict=True)
        inputs = inputs.to(self.device)
        eos_token_id = [tokenizer.eos_token_id, tokenizer.convert_tokens_to_ids("<|user|>"),
                        tokenizer.convert_tokens_to_ids("<|observation|>")]
        outputs = self.generate(**inputs, **gen_kwargs, eos_token_id=eos_token_id)
        outputs = outputs.tolist()[0][len(inputs["input_ids"][0]):-1]
        response = tokenizer.decode(outputs)
        response, history = self.process_response(response, history)
        return response, history

    @torch.inference_mode()
    def stream_chat(self, tokenizer, query: str, history: List[Dict] = None, role: str = "user",
                    past_key_values=None, max_length: int = 8192, do_sample=True, top_p=0.8, temperature=0.8,
                    logits_processor=None, return_past_key_values=False, **kwargs):
        if history is None:
            history = []
        if logits_processor is None:
            logits_processor = LogitsProcessorList()
        logits_processor.append(InvalidScoreLogitsProcessor())
        eos_token_id = [tokenizer.eos_token_id, tokenizer.convert_tokens_to_ids("<|user|>"),
                        tokenizer.convert_tokens_to_ids("<|observation|>")]
        gen_kwargs = {"max_length": max_length, "do_sample": do_sample, "top_p": top_p,
                      "temperature": temperature, "logits_processor": logits_processor, **kwargs}
        if past_key_values is None:
            inputs = tokenizer.apply_chat_template(history + [{"role": role, "content": query}],
                                                   add_generation_prompt=True, tokenize=True, return_tensors="pt",
                                                   return_dict=True)
        else:
            inputs = tokenizer.apply_chat_template([{"role": role, "content": query}], add_special_tokens=False,
                                                   add_generation_prompt=True, tokenize=True, return_tensors="pt",
                                                   return_dict=True)
        inputs = inputs.to(self.device)
        if past_key_values is not None:
            past_length = past_key_values[0][0].shape[2]
            inputs.position_ids += past_length
            attention_mask = inputs.attention_mask
            attention_mask = torch.cat((attention_mask.new_ones(1, past_length), attention_mask), dim=1)
            inputs['attention_mask'] = attention_mask
        history.append({"role": role, "content": query})
        for outputs in self.stream_generate(**inputs, past_key_values=past_key_values,
                                            eos_token_id=eos_token_id, return_past_key_values=return_past_key_values,
                                            **gen_kwargs):
            if return_past_key_values:
                outputs, past_key_values = outputs
            outputs = outputs.tolist()[0][len(inputs["input_ids"][0]):-1]
            response = tokenizer.decode(outputs)
            if response and response[-1] != "�":
                response, new_history = self.process_response(response, history)
                if return_past_key_values:
                    yield response, new_history, past_key_values
                else:
                    yield response, new_history

    @torch.inference_mode()
    def stream_generate(
            self,
            input_ids,
            generation_config: Optional[GenerationConfig] = None,
            logits_processor: Optional[LogitsProcessorList] = None,
            stopping_criteria: Optional[StoppingCriteriaList] = None,
            prefix_allowed_tokens_fn: Optional[Callable[[int, torch.Tensor], List[int]]] = None,
            return_past_key_values=False,
            **kwargs,
    ):
        batch_size, input_ids_seq_length = input_ids.shape[0], input_ids.shape[-1]

        if generation_config is None:
            generation_config = self.generation_config
        generation_config = copy.deepcopy(generation_config)
        model_kwargs = generation_config.update(**kwargs)
        model_kwargs["use_cache"] = generation_config.use_cache
        bos_token_id, eos_token_id = generation_config.bos_token_id, generation_config.eos_token_id

        if isinstance(eos_token_id, int):
            eos_token_id = [eos_token_id]
        eos_token_id_tensor = torch.tensor(eos_token_id).to(input_ids.device) if eos_token_id is not None else None

        has_default_max_length = kwargs.get("max_length") is None and generation_config.max_length is not None
        if has_default_max_length and generation_config.max_new_tokens is None:
            warnings.warn(
                f"Using `max_length`'s default ({generation_config.max_length}) to control the generation length. "
                "This behaviour is deprecated and will be removed from the config in v5 of Transformers -- we"
                " recommend using `max_new_tokens` to control the maximum length of the generation.",
                UserWarning,
            )
        elif generation_config.max_new_tokens is not None:
            generation_config.max_length = generation_config.max_new_tokens + input_ids_seq_length
            if not has_default_max_length:
                logger.warn(
                    f"Both `max_new_tokens` (={generation_config.max_new_tokens}) and `max_length`(="
                    f"{generation_config.max_length}) seem to have been set. `max_new_tokens` will take precedence. "
                    "Please refer to the documentation for more information. "
                    "(https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)",
                    UserWarning,
                )

        if input_ids_seq_length >= generation_config.max_length:
            input_ids_string = "decoder_input_ids" if self.config.is_encoder_decoder else "input_ids"
            logger.warning(
                f"Input length of {input_ids_string} is {input_ids_seq_length}, but `max_length` is set to"
                f" {generation_config.max_length}. This can lead to unexpected behavior. You should consider"
                " increasing `max_new_tokens`."
            )

        # 2. Set generation parameters if not already defined
        logits_processor = logits_processor if logits_processor is not None else LogitsProcessorList()
        stopping_criteria = stopping_criteria if stopping_criteria is not None else StoppingCriteriaList()

        logits_processor = self._get_logits_processor(
            generation_config=generation_config,
            input_ids_seq_length=input_ids_seq_length,
            encoder_input_ids=input_ids,
            prefix_allowed_tokens_fn=prefix_allowed_tokens_fn,
            logits_processor=logits_processor,
        )

        stopping_criteria = self._get_stopping_criteria(
            generation_config=generation_config, stopping_criteria=stopping_criteria
        )
        logits_warper = self._get_logits_warper(generation_config)

        unfinished_sequences = input_ids.new(input_ids.shape[0]).fill_(1)
        scores = None
        while True:
            model_inputs = self.prepare_inputs_for_generation(input_ids, **model_kwargs)
            # forward pass to get next token
            outputs = self(
                **model_inputs,
                return_dict=True,
                output_attentions=False,
                output_hidden_states=False,
            )

            next_token_logits = outputs.logits[:, -1, :]

            # pre-process distribution
            next_token_scores = logits_processor(input_ids, next_token_logits)
            next_token_scores = logits_warper(input_ids, next_token_scores)

            # sample
            probs = nn.functional.softmax(next_token_scores, dim=-1)
            if generation_config.do_sample:
                next_tokens = torch.multinomial(probs, num_samples=1).squeeze(1)
            else:
                next_tokens = torch.argmax(probs, dim=-1)
            # update generated ids, model inputs, and length for next step
            input_ids = torch.cat([input_ids, next_tokens[:, None]], dim=-1)
            model_kwargs = self._update_model_kwargs_for_generation(
                outputs, model_kwargs, is_encoder_decoder=self.config.is_encoder_decoder
            )
            unfinished_sequences = unfinished_sequences.mul(
                next_tokens.tile(eos_token_id_tensor.shape[0], 1).ne(eos_token_id_tensor.unsqueeze(1)).prod(dim=0)
            )
            if return_past_key_values:
                yield input_ids, outputs.past_key_values
            else:
                yield input_ids
            # stop when each sentence is finished, or if we exceed the maximum length
            if unfinished_sequences.max() == 0 or stopping_criteria(input_ids, scores):
                break


class ChatGLMForSequenceClassification(ChatGLMPreTrainedModel):
    def __init__(self, config: ChatGLMConfig, empty_init=True, device=None):
        super().__init__(config)

        self.num_labels = config.num_labels
        self.transformer = ChatGLMModel(config, empty_init=empty_init, device=device)

        self.classifier_head = nn.Linear(config.hidden_size, config.num_labels, bias=True, dtype=config.torch_dtype)
        if config.classifier_dropout is not None:
            self.dropout = nn.Dropout(config.classifier_dropout)
        else:
            self.dropout = None
        self.config = config

    def forward(
            self,
            input_ids: Optional[torch.LongTensor] = None,
            position_ids: Optional[torch.LongTensor] = None,
            attention_mask: Optional[torch.Tensor] = None,
            full_attention_mask: Optional[torch.Tensor] = None,
            past_key_values: Optional[Tuple[Tuple[torch.Tensor, torch.Tensor], ...]] = None,
            inputs_embeds: Optional[torch.LongTensor] = None,
            labels: Optional[torch.LongTensor] = 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, ...], SequenceClassifierOutputWithPast]:
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        transformer_outputs = self.transformer(
            input_ids=input_ids,
            position_ids=position_ids,
            attention_mask=attention_mask,
            full_attention_mask=full_attention_mask,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        hidden_states = transformer_outputs[0]
        pooled_hidden_states = hidden_states[:, -1]
        if self.dropout is not None:
            pooled_hidden_states = self.dropout(pooled_hidden_states)
        logits = self.classifier_head(pooled_hidden_states)

        loss = None
        if labels is not None:
            if self.config.problem_type is None:
                if self.num_labels == 1:
                    self.config.problem_type = "regression"
                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
                    self.config.problem_type = "single_label_classification"
                else:
                    self.config.problem_type = "multi_label_classification"

            if self.config.problem_type == "regression":
                loss_fct = MSELoss()
                if self.num_labels == 1:
                    loss = loss_fct(logits.squeeze().float(), labels.squeeze())
                else:
                    loss = loss_fct(logits.float(), labels)
            elif self.config.problem_type == "single_label_classification":
                loss_fct = CrossEntropyLoss()
                loss = loss_fct(logits.view(-1, self.num_labels).float(), labels.view(-1))
            elif self.config.problem_type == "multi_label_classification":
                loss_fct = BCEWithLogitsLoss()
                loss = loss_fct(logits.float(), labels.view(-1, self.num_labels))

        if not return_dict:
            output = (logits,) + transformer_outputs[1:]
            return ((loss,) + output) if loss is not None else output

        return SequenceClassifierOutputWithPast(
            loss=loss,
            logits=logits,
            past_key_values=transformer_outputs.past_key_values,
            hidden_states=transformer_outputs.hidden_states,
            attentions=transformer_outputs.attentions,
        )