File size: 52,415 Bytes
122057f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# coding=utf-8
# Copyright 2022 The HuggingFace Inc. team.
#
# 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.
""" Generation configuration class and utilities."""

import copy
import json
import os
import warnings
from typing import Any, Dict, Optional, Union

from .. import __version__
from ..configuration_utils import PretrainedConfig
from ..utils import (
    GENERATION_CONFIG_NAME,
    PushToHubMixin,
    cached_file,
    download_url,
    extract_commit_hash,
    is_remote_url,
    logging,
)


logger = logging.get_logger(__name__)
METADATA_FIELDS = ("_from_model_config", "_commit_hash", "_original_object_hash", "transformers_version")


class GenerationConfig(PushToHubMixin):
    # no-format
    r"""
    Class that holds a configuration for a generation task. A `generate` call supports the following generation methods
    for text-decoder, text-to-text, speech-to-text, and vision-to-text models:

        - *greedy decoding* by calling [`~generation.GenerationMixin.greedy_search`] if `num_beams=1` and
            `do_sample=False`
        - *contrastive search* by calling [`~generation.GenerationMixin.contrastive_search`] if `penalty_alpha>0.`
            and `top_k>1`
        - *multinomial sampling* by calling [`~generation.GenerationMixin.sample`] if `num_beams=1` and
            `do_sample=True`
        - *beam-search decoding* by calling [`~generation.GenerationMixin.beam_search`] if `num_beams>1` and
            `do_sample=False`
        - *beam-search multinomial sampling* by calling [`~generation.GenerationMixin.beam_sample`] if
            `num_beams>1` and `do_sample=True`
        - *diverse beam-search decoding* by calling [`~generation.GenerationMixin.group_beam_search`], if
            `num_beams>1` and `num_beam_groups>1`
        - *constrained beam-search decoding* by calling [`~generation.GenerationMixin.constrained_beam_search`], if
            `constraints!=None` or `force_words_ids!=None`
        - *assisted decoding* by calling [`~generation.GenerationMixin.assisted_decoding`], if
            `assistant_model` is passed to `.generate()`

    You do not need to call any of the above methods directly. Pass custom parameter values to '.generate()'. To learn
    more about decoding strategies refer to the [text generation strategies guide](../generation_strategies).

    <Tip>

    A large number of these flags control the logits or the stopping criteria of the generation. Make sure you check
    the [generate-related classes](https://huggingface.co/docs/transformers/internal/generation_utils) for a full
    description of the possible manipulations, as well as examples of their usage.

    </Tip>

    Arg:
        > Parameters that control the length of the output

        max_length (`int`, *optional*, defaults to 20):
            The maximum length the generated tokens can have. Corresponds to the length of the input prompt +
            `max_new_tokens`. Its effect is overridden by `max_new_tokens`, if also set.
        max_new_tokens (`int`, *optional*):
            The maximum numbers of tokens to generate, ignoring the number of tokens in the prompt.
        min_length (`int`, *optional*, defaults to 0):
            The minimum length of the sequence to be generated. Corresponds to the length of the input prompt +
            `min_new_tokens`. Its effect is overridden by `min_new_tokens`, if also set.
        min_new_tokens (`int`, *optional*):
            The minimum numbers of tokens to generate, ignoring the number of tokens in the prompt.
        early_stopping (`bool` or `str`, *optional*, defaults to `False`):
            Controls the stopping condition for beam-based methods, like beam-search. It accepts the following values:
            `True`, where the generation stops as soon as there are `num_beams` complete candidates; `False`, where an
            heuristic is applied and the generation stops when is it very unlikely to find better candidates;
            `"never"`, where the beam search procedure only stops when there cannot be better candidates (canonical
            beam search algorithm).
        max_time(`float`, *optional*):
            The maximum amount of time you allow the computation to run for in seconds. generation will still finish
            the current pass after allocated time has been passed.

        > Parameters that control the generation strategy used

        do_sample (`bool`, *optional*, defaults to `False`):
            Whether or not to use sampling ; use greedy decoding otherwise.
        num_beams (`int`, *optional*, defaults to 1):
            Number of beams for beam search. 1 means no beam search.
        num_beam_groups (`int`, *optional*, defaults to 1):
            Number of groups to divide `num_beams` into in order to ensure diversity among different groups of beams.
            [this paper](https://arxiv.org/pdf/1610.02424.pdf) for more details.
        penalty_alpha (`float`, *optional*):
            The values balance the model confidence and the degeneration penalty in contrastive search decoding.
        use_cache (`bool`, *optional*, defaults to `True`):
            Whether or not the model should use the past last key/values attentions (if applicable to the model) to
            speed up decoding.

        > Parameters for manipulation of the model output logits

        temperature (`float`, *optional*, defaults to 1.0):
            The value used to modulate the next token probabilities.
        top_k (`int`, *optional*, defaults to 50):
            The number of highest probability vocabulary tokens to keep for top-k-filtering.
        top_p (`float`, *optional*, defaults to 1.0):
            If set to float < 1, only the smallest set of most probable tokens with probabilities that add up to
            `top_p` or higher are kept for generation.
        typical_p (`float`, *optional*, defaults to 1.0):
            Local typicality measures how similar the conditional probability of predicting a target token next is to
            the expected conditional probability of predicting a random token next, given the partial text already
            generated. If set to float < 1, the smallest set of the most locally typical tokens with probabilities that
            add up to `typical_p` or higher are kept for generation. See [this
            paper](https://arxiv.org/pdf/2202.00666.pdf) for more details.
        epsilon_cutoff (`float`, *optional*, defaults to 0.0):
            If set to float strictly between 0 and 1, only tokens with a conditional probability greater than
            `epsilon_cutoff` will be sampled. In the paper, suggested values range from 3e-4 to 9e-4, depending on the
            size of the model. See [Truncation Sampling as Language Model
            Desmoothing](https://arxiv.org/abs/2210.15191) for more details.
        eta_cutoff (`float`, *optional*, defaults to 0.0):
            Eta sampling is a hybrid of locally typical sampling and epsilon sampling. If set to float strictly between
            0 and 1, a token is only considered if it is greater than either `eta_cutoff` or `sqrt(eta_cutoff) *
            exp(-entropy(softmax(next_token_logits)))`. The latter term is intuitively the expected next token
            probability, scaled by `sqrt(eta_cutoff)`. In the paper, suggested values range from 3e-4 to 2e-3,
            depending on the size of the model. See [Truncation Sampling as Language Model
            Desmoothing](https://arxiv.org/abs/2210.15191) for more details.
        diversity_penalty (`float`, *optional*, defaults to 0.0):
            This value is subtracted from a beam's score if it generates a token same as any beam from other group at a
            particular time. Note that `diversity_penalty` is only effective if `group beam search` is enabled.
        repetition_penalty (`float`, *optional*, defaults to 1.0):
            The parameter for repetition penalty. 1.0 means no penalty. See [this
            paper](https://arxiv.org/pdf/1909.05858.pdf) for more details.
        encoder_repetition_penalty (`float`, *optional*, defaults to 1.0):
            The paramater for encoder_repetition_penalty. An exponential penalty on sequences that are not in the
            original input. 1.0 means no penalty.
        length_penalty (`float`, *optional*, defaults to 1.0):
            Exponential penalty to the length that is used with beam-based generation. It is applied as an exponent to
            the sequence length, which in turn is used to divide the score of the sequence. Since the score is the log
            likelihood of the sequence (i.e. negative), `length_penalty` > 0.0 promotes longer sequences, while
            `length_penalty` < 0.0 encourages shorter sequences.
        no_repeat_ngram_size (`int`, *optional*, defaults to 0):
            If set to int > 0, all ngrams of that size can only occur once.
        bad_words_ids(`List[List[int]]`, *optional*):
            List of list of token ids that are not allowed to be generated. Check
            [`~generation.NoBadWordsLogitsProcessor`] for further documentation and examples.
        force_words_ids(`List[List[int]]` or `List[List[List[int]]]`, *optional*):
            List of token ids that must be generated. If given a `List[List[int]]`, this is treated as a simple list of
            words that must be included, the opposite to `bad_words_ids`. If given `List[List[List[int]]]`, this
            triggers a [disjunctive constraint](https://github.com/huggingface/transformers/issues/14081), where one
            can allow different forms of each word.
        renormalize_logits (`bool`, *optional*, defaults to `False`):
            Whether to renormalize the logits after applying all the logits processors or warpers (including the custom
            ones). It's highly recommended to set this flag to `True` as the search algorithms suppose the score logits
            are normalized but some logit processors or warpers break the normalization.
        constraints (`List[Constraint]`, *optional*):
            Custom constraints that can be added to the generation to ensure that the output will contain the use of
            certain tokens as defined by `Constraint` objects, in the most sensible way possible.
        forced_bos_token_id (`int`, *optional*, defaults to `model.config.forced_bos_token_id`):
            The id of the token to force as the first generated token after the `decoder_start_token_id`. Useful for
            multilingual models like [mBART](../model_doc/mbart) where the first generated token needs to be the target
            language token.
        forced_eos_token_id (`Union[int, List[int]]`, *optional*, defaults to `model.config.forced_eos_token_id`):
            The id of the token to force as the last generated token when `max_length` is reached. Optionally, use a
            list to set multiple *end-of-sequence* tokens.
        remove_invalid_values (`bool`, *optional*, defaults to `model.config.remove_invalid_values`):
            Whether to remove possible *nan* and *inf* outputs of the model to prevent the generation method to crash.
            Note that using `remove_invalid_values` can slow down generation.
        exponential_decay_length_penalty (`tuple(int, float)`, *optional*):
            This Tuple adds an exponentially increasing length penalty, after a certain amount of tokens have been
            generated. The tuple shall consist of: `(start_index, decay_factor)` where `start_index` indicates where
            penalty starts and `decay_factor` represents the factor of exponential decay
        suppress_tokens  (`List[int]`, *optional*):
            A list of tokens that will be suppressed at generation. The `SupressTokens` logit processor will set their
            log probs to `-inf` so that they are not sampled.
        begin_suppress_tokens  (`List[int]`, *optional*):
            A list of tokens that will be suppressed at the beginning of the generation. The `SupressBeginTokens` logit
            processor will set their log probs to `-inf` so that they are not sampled.
        forced_decoder_ids (`List[List[int]]`, *optional*):
            A list of pairs of integers which indicates a mapping from generation indices to token indices that will be
            forced before sampling. For example, `[[1, 123]]` means the second generated token will always be a token
            of index 123.
        sequence_bias (`Dict[Tuple[int], float]`, *optional*)):
            Dictionary that maps a sequence of tokens to its bias term. Positive biases increase the odds of the
            sequence being selected, while negative biases do the opposite. Check
            [`~generation.SequenceBiasLogitsProcessor`] for further documentation and examples.
        guidance_scale (`float`, *optional*):
            The guidance scale for classifier free guidance (CFG). CFG is enabled by setting `guidance_scale > 1`.
            Higher guidance scale encourages the model to generate samples that are more closely linked to the input
            prompt, usually at the expense of poorer quality.
        low_memory (`bool`, *optional*):
            Switch to sequential topk for contrastive search to reduce peak memory. Used with contrastive search.


        > Parameters that define the output variables of `generate`

        num_return_sequences(`int`, *optional*, defaults to 1):
            The number of independently computed returned sequences for each element in the batch.
        output_attentions (`bool`, *optional*, defaults to `False`):
            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
            tensors for more details.
        output_hidden_states (`bool`, *optional*, defaults to `False`):
            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
            more details.
        output_scores (`bool`, *optional*, defaults to `False`):
            Whether or not to return the prediction scores. See `scores` under returned tensors for more details.
        return_dict_in_generate (`bool`, *optional*, defaults to `False`):
            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.

        > Special tokens that can be used at generation time

        pad_token_id (`int`, *optional*):
            The id of the *padding* token.
        bos_token_id (`int`, *optional*):
            The id of the *beginning-of-sequence* token.
        eos_token_id (`Union[int, List[int]]`, *optional*):
            The id of the *end-of-sequence* token. Optionally, use a list to set multiple *end-of-sequence* tokens.

        > Generation parameters exclusive to encoder-decoder models

        encoder_no_repeat_ngram_size (`int`, *optional*, defaults to 0):
            If set to int > 0, all ngrams of that size that occur in the `encoder_input_ids` cannot occur in the
            `decoder_input_ids`.
        decoder_start_token_id (`int`, *optional*):
            If an encoder-decoder model starts decoding with a different token than *bos*, the id of that token.

        > Generation parameters exclusive to [assistant generation](https://arxiv.org/abs/2211.17192)

        num_assistant_tokens (`int`, *optional*, defaults to 5):
            Defines the number of _speculative tokens_ that shall be generated by the assistant model before being
            checked by the target model at each iteration. Higher values for `num_assistant_tokens` make the generation
            more _speculative_ : If the assistant model is performant larger speed-ups can be reached, if the assistant
            model requires lots of corrections, lower speed-ups are reached.

        num_assistant_tokens_schedule (`str`, *optional*, defaults to `"heuristic"`):
            Defines the schedule at which max assistant tokens shall be changed during inference.
            - `"_heuristic_`: When all _speculative_ tokens are correct, increase `num_assistant_tokens` by 2 else
              reduce by 1
            - `"constant"`: `num_assistant_tokens` stays unchanged during generation

        > Wild card

        generation_kwargs:
            Additional generation kwargs will be forwarded to the `generate` function of the model. Kwargs that are not
            present in `generate`'s signature will be used in the model forward pass.
    """

    def __init__(self, **kwargs):
        # Parameters that control the length of the output
        # if the default `max_length` is updated here, make sure to update the `generate` tests following https://github.com/huggingface/transformers/pull/25030
        self.max_length = kwargs.pop("max_length", 20)
        self.max_new_tokens = kwargs.pop("max_new_tokens", None)
        self.min_length = kwargs.pop("min_length", 0)
        self.min_new_tokens = kwargs.pop("min_new_tokens", None)
        self.early_stopping = kwargs.pop("early_stopping", False)
        self.max_time = kwargs.pop("max_time", None)

        # Parameters that control the generation strategy used
        self.do_sample = kwargs.pop("do_sample", False)
        self.num_beams = kwargs.pop("num_beams", 1)
        self.num_beam_groups = kwargs.pop("num_beam_groups", 1)
        self.penalty_alpha = kwargs.pop("penalty_alpha", None)
        self.use_cache = kwargs.pop("use_cache", True)

        # Parameters for manipulation of the model output logits
        self.temperature = kwargs.pop("temperature", 1.0)
        self.top_k = kwargs.pop("top_k", 50)
        self.top_p = kwargs.pop("top_p", 1.0)
        self.typical_p = kwargs.pop("typical_p", 1.0)
        self.epsilon_cutoff = kwargs.pop("epsilon_cutoff", 0.0)
        self.eta_cutoff = kwargs.pop("eta_cutoff", 0.0)
        self.diversity_penalty = kwargs.pop("diversity_penalty", 0.0)
        self.repetition_penalty = kwargs.pop("repetition_penalty", 1.0)
        self.encoder_repetition_penalty = kwargs.pop("encoder_repetition_penalty", 1.0)
        self.length_penalty = kwargs.pop("length_penalty", 1.0)
        self.no_repeat_ngram_size = kwargs.pop("no_repeat_ngram_size", 0)
        self.bad_words_ids = kwargs.pop("bad_words_ids", None)
        self.force_words_ids = kwargs.pop("force_words_ids", None)
        self.renormalize_logits = kwargs.pop("renormalize_logits", False)
        self.constraints = kwargs.pop("constraints", None)
        self.forced_bos_token_id = kwargs.pop("forced_bos_token_id", None)
        self.forced_eos_token_id = kwargs.pop("forced_eos_token_id", None)
        self.remove_invalid_values = kwargs.pop("remove_invalid_values", False)
        self.exponential_decay_length_penalty = kwargs.pop("exponential_decay_length_penalty", None)
        self.suppress_tokens = kwargs.pop("suppress_tokens", None)
        self.begin_suppress_tokens = kwargs.pop("begin_suppress_tokens", None)
        self.forced_decoder_ids = kwargs.pop("forced_decoder_ids", None)
        self.sequence_bias = kwargs.pop("sequence_bias", None)
        self.guidance_scale = kwargs.pop("guidance_scale", None)
        self.low_memory = kwargs.pop("low_memory", None)

        # Parameters that define the output variables of `generate`
        self.num_return_sequences = kwargs.pop("num_return_sequences", 1)
        self.output_attentions = kwargs.pop("output_attentions", False)
        self.output_hidden_states = kwargs.pop("output_hidden_states", False)
        self.output_scores = kwargs.pop("output_scores", False)
        self.return_dict_in_generate = kwargs.pop("return_dict_in_generate", False)

        # Special tokens that can be used at generation time
        self.pad_token_id = kwargs.pop("pad_token_id", None)
        self.bos_token_id = kwargs.pop("bos_token_id", None)
        self.eos_token_id = kwargs.pop("eos_token_id", None)

        # Generation parameters exclusive to encoder-decoder models
        self.encoder_no_repeat_ngram_size = kwargs.pop("encoder_no_repeat_ngram_size", 0)
        self.decoder_start_token_id = kwargs.pop("decoder_start_token_id", None)

        # Assistant generation
        self.num_assistant_tokens = kwargs.pop("num_assistant_tokens", 5)
        self.num_assistant_tokens_schedule = kwargs.pop("num_assistant_tokens_schedule", "heuristic")

        # Wild card
        self.generation_kwargs = kwargs.pop("generation_kwargs", {})

        # The remaining attributes do not parametrize `.generate()`, but are informative and/or used by the hub
        # interface.
        self._from_model_config = kwargs.pop("_from_model_config", False)
        self._commit_hash = kwargs.pop("_commit_hash", None)
        self.transformers_version = kwargs.pop("transformers_version", __version__)

        # Additional attributes without default values
        if not self._from_model_config:
            # we don't want to copy values from the model config if we're initializing a `GenerationConfig` from a
            # model's default configuration file
            for key, value in kwargs.items():
                try:
                    setattr(self, key, value)
                except AttributeError as err:
                    logger.error(f"Can't set {key} with value {value} for {self}")
                    raise err

        # Validate the values of the attributes
        self.validate(is_init=True)

    def __hash__(self):
        return hash(self.to_json_string(ignore_metadata=True))

    def __eq__(self, other):
        if not isinstance(other, GenerationConfig):
            return False

        self_without_metadata = self.to_json_string(use_diff=False, ignore_metadata=True)
        other_without_metadata = other.to_json_string(use_diff=False, ignore_metadata=True)
        return self_without_metadata == other_without_metadata

    def __repr__(self):
        return f"{self.__class__.__name__} {self.to_json_string(ignore_metadata=True)}"

    def validate(self, is_init=False):
        """
        Validates the values of the attributes of the [`GenerationConfig`] instance. Raises exceptions in the presence
        of parameterization that can be detected as incorrect from the configuration instance alone.

        Note that some parameters are best validated at generate runtime, as they may depend on other inputs and/or the
        model, such as parameters related to the generation length.
        """

        # Validation of individual attributes
        if self.early_stopping not in {True, False, "never"}:
            raise ValueError(f"`early_stopping` must be a boolean or 'never', but is {self.early_stopping}.")

        # Validation of attribute relations:
        fix_location = ""
        if is_init:
            fix_location = (
                " This was detected when initializing the generation config instance, which means the corresponding "
                "file may hold incorrect parameterization and should be fixed."
            )

        # 1. detect sampling-only parameterization when not in sampling mode
        if self.do_sample is False:
            greedy_wrong_parameter_msg = (
                "`do_sample` is set to `False`. However, `{flag_name}` is set to `{flag_value}` -- this flag is only "
                "used in sample-based generation modes. You should set `do_sample=True` or unset `{flag_name}`."
                + fix_location
            )
            if self.temperature != 1.0:
                warnings.warn(
                    greedy_wrong_parameter_msg.format(flag_name="temperature", flag_value=self.temperature),
                    UserWarning,
                )
            if self.top_p != 1.0:
                warnings.warn(
                    greedy_wrong_parameter_msg.format(flag_name="top_p", flag_value=self.top_p),
                    UserWarning,
                )
            if self.typical_p != 1.0:
                warnings.warn(
                    greedy_wrong_parameter_msg.format(flag_name="typical_p", flag_value=self.typical_p),
                    UserWarning,
                )
            if self.top_k != 50 and self.penalty_alpha is None:  # contrastive search uses top_k
                warnings.warn(
                    greedy_wrong_parameter_msg.format(flag_name="top_k", flag_value=self.top_k),
                    UserWarning,
                )
            if self.epsilon_cutoff != 0.0:
                warnings.warn(
                    greedy_wrong_parameter_msg.format(flag_name="epsilon_cutoff", flag_value=self.epsilon_cutoff),
                    UserWarning,
                )
            if self.eta_cutoff != 0.0:
                warnings.warn(
                    greedy_wrong_parameter_msg.format(flag_name="eta_cutoff", flag_value=self.eta_cutoff),
                    UserWarning,
                )

        # 2. detect beam-only parameterization when not in beam mode
        if self.num_beams is None:
            warnings.warn("`num_beams` is set to None - defaulting to 1.", UserWarning)
            self.num_beams = 1

        if self.num_beams == 1:
            single_beam_wrong_parameter_msg = (
                "`num_beams` is set to 1. However, `{flag_name}` is set to `{flag_value}` -- this flag is only used "
                "in beam-based generation modes. You should set `num_beams>1` or unset `{flag_name}`." + fix_location
            )
            if self.early_stopping is not False:
                warnings.warn(
                    single_beam_wrong_parameter_msg.format(flag_name="early_stopping", flag_value=self.early_stopping),
                    UserWarning,
                )
            if self.num_beam_groups != 1:
                warnings.warn(
                    single_beam_wrong_parameter_msg.format(
                        flag_name="num_beam_groups", flag_value=self.num_beam_groups
                    ),
                    UserWarning,
                )
            if self.diversity_penalty != 0.0:
                warnings.warn(
                    single_beam_wrong_parameter_msg.format(
                        flag_name="diversity_penalty", flag_value=self.diversity_penalty
                    ),
                    UserWarning,
                )
            if self.length_penalty != 1.0:
                warnings.warn(
                    single_beam_wrong_parameter_msg.format(flag_name="length_penalty", flag_value=self.length_penalty),
                    UserWarning,
                )
            if self.constraints is not None:
                warnings.warn(
                    single_beam_wrong_parameter_msg.format(flag_name="constraints", flag_value=self.constraints),
                    UserWarning,
                )

        # 3. detect incorrect paramaterization specific to advanced beam modes
        else:
            # constrained beam search
            if self.constraints is not None:
                constrained_wrong_parameter_msg = (
                    "`constraints` is not `None`, triggering constrained beam search. However, `{flag_name}` is set "
                    "to `{flag_value}`, which is incompatible with this generation mode. Set `constraints=None` or "
                    "unset `{flag_name}` to continue." + fix_location
                )
                if self.do_sample is True:
                    raise ValueError(
                        constrained_wrong_parameter_msg.format(flag_name="do_sample", flag_value=self.do_sample)
                    )
                if self.num_beam_groups != 1:
                    raise ValueError(
                        constrained_wrong_parameter_msg.format(
                            flag_name="num_beam_groups", flag_value=self.num_beam_groups
                        )
                    )
            # group beam search
            if self.diversity_penalty != 0.0 or self.num_beam_groups != 1:
                group_error_prefix = (
                    "`diversity_penalty` is not 0.0 or `num_beam_groups` is not 1, triggering group beam search. In "
                    "this generation mode, "
                )
                if self.do_sample is True:
                    raise ValueError(group_error_prefix + "`do_sample` must be set to `False`")
                if self.num_beams % self.num_beam_groups != 0:
                    raise ValueError(group_error_prefix + "`num_beams` should be divisible by `num_beam_groups`")
                if self.diversity_penalty == 0.0:
                    raise ValueError(
                        group_error_prefix
                        + "`diversity_penalty` should be greater than `0.0`, otherwise your groups will be identical."
                    )

        # 4. check `num_return_sequences`
        if self.num_return_sequences != 1:
            if self.num_beams == 1:
                if self.do_sample is False:
                    raise ValueError(
                        "Greedy methods without beam search do not support `num_return_sequences` different than 1 "
                        f"(got {self.num_return_sequences})."
                    )
            elif self.num_return_sequences > self.num_beams:
                raise ValueError(
                    f"`num_return_sequences` ({self.num_return_sequences}) has to be smaller or equal to `num_beams` "
                    f"({self.num_beams})."
                )

        # 5. check common issue: passing `generate` arguments inside the generation config
        generate_arguments = (
            "logits_processor",
            "stopping_criteria",
            "prefix_allowed_tokens_fn",
            "synced_gpus",
            "assistant_model",
            "streamer",
            "negative_prompt_ids",
            "negative_prompt_attention_mask",
        )
        for arg in generate_arguments:
            if hasattr(self, arg):
                raise ValueError(
                    f"Argument `{arg}` is not a valid argument of `GenerationConfig`. It should be passed to "
                    "`generate()` (or a pipeline) directly."
                )

    def save_pretrained(
        self,
        save_directory: Union[str, os.PathLike],
        config_file_name: Optional[Union[str, os.PathLike]] = None,
        push_to_hub: bool = False,
        **kwargs,
    ):
        r"""
        Save a generation configuration object to the directory `save_directory`, so that it can be re-loaded using the
        [`~GenerationConfig.from_pretrained`] class method.

        Args:
            save_directory (`str` or `os.PathLike`):
                Directory where the configuration JSON file will be saved (will be created if it does not exist).
            config_file_name (`str` or `os.PathLike`, *optional*, defaults to `"generation_config.json"`):
                Name of the generation configuration JSON file to be saved in `save_directory`.
            push_to_hub (`bool`, *optional*, defaults to `False`):
                Whether or not to push your model to the Hugging Face model hub after saving it. You can specify the
                repository you want to push to with `repo_id` (will default to the name of `save_directory` in your
                namespace).
            kwargs (`Dict[str, Any]`, *optional*):
                Additional key word arguments passed along to the [`~utils.PushToHubMixin.push_to_hub`] method.
        """

        # At save time, validate the instance -- if any warning/exception is thrown, we refuse to save the instance
        try:
            with warnings.catch_warnings(record=True) as caught_warnings:
                self.validate()
            for w in caught_warnings:
                raise ValueError(w.message)
        except ValueError as exc:
            warnings.warn(
                "The generation config instance is invalid -- `.validate()` throws warnings and/or exceptions. "
                "Fix these issues to save the configuration. This warning will be raised to an exception in v4.34."
                "\n\nThrown during validation:\n" + str(exc),
                UserWarning,
            )
            return

        use_auth_token = kwargs.pop("use_auth_token", None)

        if use_auth_token is not None:
            warnings.warn(
                "The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. Please use `token` instead.",
                FutureWarning,
            )
            if kwargs.get("token", None) is not None:
                raise ValueError(
                    "`token` and `use_auth_token` are both specified. Please set only the argument `token`."
                )
            kwargs["token"] = use_auth_token

        config_file_name = config_file_name if config_file_name is not None else GENERATION_CONFIG_NAME

        if os.path.isfile(save_directory):
            raise AssertionError(f"Provided path ({save_directory}) should be a directory, not a file")

        os.makedirs(save_directory, exist_ok=True)

        if push_to_hub:
            commit_message = kwargs.pop("commit_message", None)
            repo_id = kwargs.pop("repo_id", save_directory.split(os.path.sep)[-1])
            repo_id = self._create_repo(repo_id, **kwargs)
            files_timestamps = self._get_files_timestamps(save_directory)

        output_config_file = os.path.join(save_directory, config_file_name)

        self.to_json_file(output_config_file, use_diff=True)
        logger.info(f"Configuration saved in {output_config_file}")

        if push_to_hub:
            self._upload_modified_files(
                save_directory,
                repo_id,
                files_timestamps,
                commit_message=commit_message,
                token=kwargs.get("token"),
            )

    @classmethod
    def from_pretrained(
        cls,
        pretrained_model_name: Union[str, os.PathLike],
        config_file_name: Optional[Union[str, os.PathLike]] = None,
        cache_dir: Optional[Union[str, os.PathLike]] = None,
        force_download: bool = False,
        local_files_only: bool = False,
        token: Optional[Union[str, bool]] = None,
        revision: str = "main",
        **kwargs,
    ) -> "GenerationConfig":
        r"""
        Instantiate a [`GenerationConfig`] from a generation configuration file.

        Args:
            pretrained_model_name (`str` or `os.PathLike`):
                This can be either:

                - a string, the *model id* of a pretrained model configuration hosted inside a model repo on
                  huggingface.co. Valid model ids can be located at the root-level, like `bert-base-uncased`, or
                  namespaced under a user or organization name, like `dbmdz/bert-base-german-cased`.
                - a path to a *directory* containing a configuration file saved using the
                  [`~GenerationConfig.save_pretrained`] method, e.g., `./my_model_directory/`.
            config_file_name (`str` or `os.PathLike`, *optional*, defaults to `"generation_config.json"`):
                Name of the generation configuration JSON file to be loaded from `pretrained_model_name`.
            cache_dir (`str` or `os.PathLike`, *optional*):
                Path to a directory in which a downloaded pretrained model configuration should be cached if the
                standard cache should not be used.
            force_download (`bool`, *optional*, defaults to `False`):
                Whether or not to force to (re-)download the configuration files and override the cached versions if
                they exist.
            resume_download (`bool`, *optional*, defaults to `False`):
                Whether or not to delete incompletely received file. Attempts to resume the download if such a file
                exists.
            proxies (`Dict[str, str]`, *optional*):
                A dictionary of proxy servers to use by protocol or endpoint, e.g., `{'http': 'foo.bar:3128',
                'http://hostname': 'foo.bar:4012'}.` The proxies are used on each request.
            token (`str` or `bool`, *optional*):
                The token to use as HTTP bearer authorization for remote files. If `True`, or not specified, will use
                the token generated when running `huggingface-cli login` (stored in `~/.huggingface`).
            revision (`str`, *optional*, defaults to `"main"`):
                The specific model version to use. It can be a branch name, a tag name, or a commit id, since we use a
                git-based system for storing models and other artifacts on huggingface.co, so `revision` can be any
                identifier allowed by git.

                <Tip>

                To test a pull request you made on the Hub, you can pass `revision="refs/pr/<pr_number>".

                </Tip>

            return_unused_kwargs (`bool`, *optional*, defaults to `False`):
                If `False`, then this function returns just the final configuration object.

                If `True`, then this functions returns a `Tuple(config, unused_kwargs)` where *unused_kwargs* is a
                dictionary consisting of the key/value pairs whose keys are not configuration attributes: i.e., the
                part of `kwargs` which has not been used to update `config` and is otherwise ignored.
            subfolder (`str`, *optional*, defaults to `""`):
                In case the relevant files are located inside a subfolder of the model repo on huggingface.co, you can
                specify the folder name here.
            kwargs (`Dict[str, Any]`, *optional*):
                The values in kwargs of any keys which are configuration attributes will be used to override the loaded
                values. Behavior concerning key/value pairs whose keys are *not* configuration attributes is controlled
                by the `return_unused_kwargs` keyword parameter.

        Returns:
            [`GenerationConfig`]: The configuration object instantiated from this pretrained model.

        Examples:

        ```python
        >>> from transformers import GenerationConfig

        >>> # Download configuration from huggingface.co and cache.
        >>> generation_config = GenerationConfig.from_pretrained("gpt2")

        >>> # E.g. config was saved using *save_pretrained('./test/saved_model/')*
        >>> generation_config.save_pretrained("./test/saved_model/")
        >>> generation_config = GenerationConfig.from_pretrained("./test/saved_model/")

        >>> # You can also specify configuration names to your generation configuration file
        >>> generation_config.save_pretrained("./test/saved_model/", config_file_name="my_configuration.json")
        >>> generation_config = GenerationConfig.from_pretrained("./test/saved_model/", "my_configuration.json")

        >>> # If you'd like to try a minor variation to an existing configuration, you can also pass generation
        >>> # arguments to `.from_pretrained()`. Be mindful that typos and unused arguments will be ignored
        >>> generation_config, unused_kwargs = GenerationConfig.from_pretrained(
        ...     "gpt2", top_k=1, foo=False, do_sample=True, return_unused_kwargs=True
        ... )
        >>> generation_config.top_k
        1

        >>> unused_kwargs
        {'foo': False}
        ```"""
        config_file_name = config_file_name if config_file_name is not None else GENERATION_CONFIG_NAME

        resume_download = kwargs.pop("resume_download", False)
        proxies = kwargs.pop("proxies", None)
        use_auth_token = kwargs.pop("use_auth_token", None)
        subfolder = kwargs.pop("subfolder", "")
        from_pipeline = kwargs.pop("_from_pipeline", None)
        from_auto_class = kwargs.pop("_from_auto", False)
        commit_hash = kwargs.pop("_commit_hash", None)

        if use_auth_token is not None:
            warnings.warn(
                "The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. Please use `token` instead.",
                FutureWarning,
            )
            if token is not None:
                raise ValueError(
                    "`token` and `use_auth_token` are both specified. Please set only the argument `token`."
                )
            token = use_auth_token

        user_agent = {"file_type": "config", "from_auto_class": from_auto_class}
        if from_pipeline is not None:
            user_agent["using_pipeline"] = from_pipeline

        config_path = os.path.join(pretrained_model_name, config_file_name)
        config_path = str(config_path)

        is_local = os.path.exists(config_path)
        if os.path.isfile(os.path.join(subfolder, config_path)):
            # Special case when config_path is a local file
            resolved_config_file = config_path
            is_local = True
        elif is_remote_url(config_path):
            configuration_file = config_path
            resolved_config_file = download_url(config_path)
        else:
            configuration_file = config_file_name
            try:
                # Load from local folder or from cache or download from model Hub and cache
                resolved_config_file = cached_file(
                    pretrained_model_name,
                    configuration_file,
                    cache_dir=cache_dir,
                    force_download=force_download,
                    proxies=proxies,
                    resume_download=resume_download,
                    local_files_only=local_files_only,
                    token=token,
                    user_agent=user_agent,
                    revision=revision,
                    subfolder=subfolder,
                    _commit_hash=commit_hash,
                )
                commit_hash = extract_commit_hash(resolved_config_file, commit_hash)
            except EnvironmentError:
                # Raise any environment error raise by `cached_file`. It will have a helpful error message adapted to
                # the original exception.
                raise
            except Exception:
                # For any other exception, we throw a generic error.
                raise EnvironmentError(
                    f"Can't load the configuration of '{pretrained_model_name}'. If you were trying to load it"
                    " from 'https://huggingface.co/models', make sure you don't have a local directory with the same"
                    f" name. Otherwise, make sure '{pretrained_model_name}' is the correct path to a directory"
                    f" containing a {configuration_file} file"
                )

        try:
            # Load config dict
            config_dict = cls._dict_from_json_file(resolved_config_file)
            config_dict["_commit_hash"] = commit_hash
        except (json.JSONDecodeError, UnicodeDecodeError):
            raise EnvironmentError(
                f"It looks like the config file at '{resolved_config_file}' is not a valid JSON file."
            )

        if is_local:
            logger.info(f"loading configuration file {resolved_config_file}")
        else:
            logger.info(f"loading configuration file {configuration_file} from cache at {resolved_config_file}")

        if kwargs.get("return_unused_kwargs") is True:
            config, unused_kwargs = cls.from_dict(config_dict, **kwargs)
            config._original_object_hash = hash(config)  # Hash to detect whether the instance was modified
            return config, unused_kwargs
        else:
            config = cls.from_dict(config_dict, **kwargs)
            config._original_object_hash = hash(config)  # Hash to detect whether the instance was modified
            return config

    @classmethod
    def _dict_from_json_file(cls, json_file: Union[str, os.PathLike]):
        with open(json_file, "r", encoding="utf-8") as reader:
            text = reader.read()
        return json.loads(text)

    @classmethod
    def from_dict(cls, config_dict: Dict[str, Any], **kwargs) -> "GenerationConfig":
        """
        Instantiates a [`GenerationConfig`] from a Python dictionary of parameters.

        Args:
            config_dict (`Dict[str, Any]`):
                Dictionary that will be used to instantiate the configuration object.
            kwargs (`Dict[str, Any]`):
                Additional parameters from which to initialize the configuration object.

        Returns:
            [`GenerationConfig`]: The configuration object instantiated from those parameters.
        """
        return_unused_kwargs = kwargs.pop("return_unused_kwargs", False)
        # Those arguments may be passed along for our internal telemetry.
        # We remove them so they don't appear in `return_unused_kwargs`.
        kwargs.pop("_from_auto", None)
        kwargs.pop("_from_pipeline", None)
        # The commit hash might have been updated in the `config_dict`, we don't want the kwargs to erase that update.
        if "_commit_hash" in kwargs and "_commit_hash" in config_dict:
            kwargs["_commit_hash"] = config_dict["_commit_hash"]

        # The line below allows model-specific config to be loaded as well through kwargs, with safety checks.
        # See https://github.com/huggingface/transformers/pull/21269
        config = cls(**{**config_dict, **kwargs})
        unused_kwargs = config.update(**kwargs)

        logger.info(f"Generate config {config}")
        if return_unused_kwargs:
            return config, unused_kwargs
        else:
            return config

    def dict_torch_dtype_to_str(self, d: Dict[str, Any]) -> None:
        """
        Checks whether the passed dictionary and its nested dicts have a *torch_dtype* key and if it's not None,
        converts torch.dtype to a string of just the type. For example, `torch.float32` get converted into *"float32"*
        string, which can then be stored in the json format.
        """
        if d.get("torch_dtype", None) is not None and not isinstance(d["torch_dtype"], str):
            d["torch_dtype"] = str(d["torch_dtype"]).split(".")[1]
        for value in d.values():
            if isinstance(value, dict):
                self.dict_torch_dtype_to_str(value)

    def to_diff_dict(self) -> Dict[str, Any]:
        """
        Removes all attributes from config which correspond to the default config attributes for better readability and
        serializes to a Python dictionary.

        Returns:
            `Dict[str, Any]`: Dictionary of all the attributes that make up this configuration instance,
        """
        config_dict = self.to_dict()

        # get the default config dict
        default_config_dict = GenerationConfig().to_dict()

        serializable_config_dict = {}

        # only serialize values that differ from the default config
        for key, value in config_dict.items():
            if key not in default_config_dict or key == "transformers_version" or value != default_config_dict[key]:
                serializable_config_dict[key] = value

        self.dict_torch_dtype_to_str(serializable_config_dict)
        return serializable_config_dict

    def to_dict(self) -> Dict[str, Any]:
        """
        Serializes this instance to a Python dictionary.

        Returns:
            `Dict[str, Any]`: Dictionary of all the attributes that make up this configuration instance.
        """
        output = copy.deepcopy(self.__dict__)

        # Fields to ignore at serialization time
        if "_commit_hash" in output:
            del output["_commit_hash"]
        if "_original_object_hash" in output:
            del output["_original_object_hash"]

        # Transformers version when serializing this file
        output["transformers_version"] = __version__

        self.dict_torch_dtype_to_str(output)
        return output

    def to_json_string(self, use_diff: bool = True, ignore_metadata: bool = False) -> str:
        """
        Serializes this instance to a JSON string.

        Args:
            use_diff (`bool`, *optional*, defaults to `True`):
                If set to `True`, only the difference between the config instance and the default `GenerationConfig()`
                is serialized to JSON string.
            ignore_metadata (`bool`, *optional*, defaults to `False`):
                Whether to ignore the metadata fields present in the instance

        Returns:
            `str`: String containing all the attributes that make up this configuration instance in JSON format.
        """
        if use_diff is True:
            config_dict = self.to_diff_dict()
        else:
            config_dict = self.to_dict()

        if ignore_metadata:
            for metadata_field in METADATA_FIELDS:
                config_dict.pop(metadata_field, None)

        return json.dumps(config_dict, indent=2, sort_keys=True) + "\n"

    def to_json_file(self, json_file_path: Union[str, os.PathLike], use_diff: bool = True):
        """
        Save this instance to a JSON file.

        Args:
            json_file_path (`str` or `os.PathLike`):
                Path to the JSON file in which this configuration instance's parameters will be saved.
            use_diff (`bool`, *optional*, defaults to `True`):
                If set to `True`, only the difference between the config instance and the default `GenerationConfig()`
                is serialized to JSON file.
        """
        with open(json_file_path, "w", encoding="utf-8") as writer:
            writer.write(self.to_json_string(use_diff=use_diff))

    @classmethod
    def from_model_config(cls, model_config: PretrainedConfig) -> "GenerationConfig":
        """
        Instantiates a [`GenerationConfig`] from a [`PretrainedConfig`]. This function is useful to convert legacy
        [`PretrainedConfig`] objects, which may contain generation parameters, into a stand-alone [`GenerationConfig`].

        Args:
            model_config (`PretrainedConfig`):
                The model config that will be used to instantiate the generation config.

        Returns:
            [`GenerationConfig`]: The configuration object instantiated from those parameters.
        """
        config_dict = model_config.to_dict()
        config_dict.pop("_from_model_config", None)
        config = cls.from_dict(config_dict, return_unused_kwargs=False, _from_model_config=True)

        # Special case: some models have generation attributes set in the decoder. Use them if still unset in the
        # generation config.
        for decoder_name in ("decoder", "generator", "text_config"):
            if decoder_name in config_dict:
                default_generation_config = GenerationConfig()
                decoder_config = config_dict[decoder_name]
                for attr in config.to_dict().keys():
                    if attr in decoder_config and getattr(config, attr) == getattr(default_generation_config, attr):
                        setattr(config, attr, decoder_config[attr])

        config._original_object_hash = hash(config)  # Hash to detect whether the instance was modified
        return config

    def update(self, **kwargs):
        """
        Updates attributes of this class instance with attributes from `kwargs` if they match existing atributtes,
        returning all the unused kwargs.

        Args:
            kwargs (`Dict[str, Any]`):
                Dictionary of attributes to tentatively update this class.

        Returns:
            `Dict[str, Any]`: Dictionary containing all the key-value pairs that were not used to update the instance.
        """
        to_remove = []
        for key, value in kwargs.items():
            if hasattr(self, key):
                setattr(self, key, value)
                to_remove.append(key)

        # remove all the attributes that were updated, without modifying the input dict
        unused_kwargs = {key: value for key, value in kwargs.items() if key not in to_remove}
        return unused_kwargs