File size: 50,197 Bytes
8173ae1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# This code is modified from the Huggingface repository: https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth_lora.py, and
import argparse
import hashlib
import itertools
import json
import logging
import math
import os
import warnings
from pathlib import Path

import numpy as np
import torch
import torch.nn.functional as F
import torch.utils.checkpoint
import transformers
from accelerate import Accelerator
from accelerate.logging import get_logger
from accelerate.utils import ProjectConfiguration, set_seed
from huggingface_hub import HfApi, create_repo
from model_pipeline import (
    CustomDiffusionAttnProcessor,
    CustomDiffusionPipeline,
    set_use_memory_efficient_attention_xformers,
)
from packaging import version
from tqdm.auto import tqdm
from transformers import AutoTokenizer, PretrainedConfig
from utils import (
    CustomDiffusionDataset,
    PromptDataset,
    collate_fn,
    filter,
    getanchorprompts,
)

import diffusers
from diffusers import (
    AutoencoderKL,
    DDPMScheduler,
    DiffusionPipeline,
    DPMSolverMultistepScheduler,
    UNet2DConditionModel,
)
from diffusers.models.cross_attention import CrossAttention
from diffusers.optimization import get_scheduler
from diffusers.utils import check_min_version, is_wandb_available
from diffusers.utils.import_utils import is_xformers_available

# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
check_min_version("0.14.0")

logger = get_logger(__name__)


def create_custom_diffusion(unet, parameter_group):
    for name, params in unet.named_parameters():
        if parameter_group == "cross-attn":
            if 'attn2.to_k' in name or 'attn2.to_v' in name:
                params.requires_grad = True
            else:
                params.requires_grad = False
        elif parameter_group == 'full-weight':
            params.requires_grad = True
        elif parameter_group == 'embedding':
            params.requires_grad = False
        else:
            raise ValueError(
                "parameter_group argument only cross-attn, full-weight, embedding"
            )

    # change attn class
    def change_attn(unet):
        for layer in unet.children():
            if type(layer) == CrossAttention:
                bound_method = set_use_memory_efficient_attention_xformers.__get__(
                    layer, layer.__class__)
                setattr(
                    layer, 'set_use_memory_efficient_attention_xformers', bound_method)
            else:
                change_attn(layer)

    change_attn(unet)
    unet.set_attn_processor(CustomDiffusionAttnProcessor())
    return unet


def save_model_card(repo_id: str, images=None, base_model=str, prompt=str, repo_folder=None):
    img_str = ""
    for i, image in enumerate(images):
        image.save(os.path.join(repo_folder, f"image_{i}.png"))
        img_str += f"./image_{i}.png\n"

    yaml = f"""
        ---
        license: creativeml-openrail-m
        base_model: {base_model}
        instance_prompt: {prompt}
        tags:
        - stable-diffusion
        - stable-diffusion-diffusers
        - text-to-image
        - diffusers
        - custom diffusion
        inference: true
        ---
            """
    model_card = f"""
        # Custom Diffusion - {repo_id}

        These are Custom Diffusion adaption weights for {base_model}. The weights were trained on {prompt} using [Custom Diffusion](https://www.cs.cmu.edu/~custom-diffusion). You can find some example images in the following. \n
        {img_str[0]}
        """
    with open(os.path.join(repo_folder, "README.md"), "w") as f:
        f.write(yaml + model_card)


def import_model_class_from_model_name_or_path(pretrained_model_name_or_path: str, revision: str):
    text_encoder_config = PretrainedConfig.from_pretrained(
        pretrained_model_name_or_path,
        subfolder="text_encoder",
        revision=revision,
    )
    model_class = text_encoder_config.architectures[0]

    if model_class == "CLIPTextModel":
        from transformers import CLIPTextModel

        return CLIPTextModel
    elif model_class == "RobertaSeriesModelWithTransformation":
        from diffusers.pipelines.alt_diffusion.modeling_roberta_series import (
            RobertaSeriesModelWithTransformation,
        )

        return RobertaSeriesModelWithTransformation
    else:
        raise ValueError(f"{model_class} is not supported.")


def freeze_params(params):
    for param in params:
        param.requires_grad = False


def parse_args(input_args=None):
    parser = argparse.ArgumentParser(
        description="Simple example of a training script.")
    parser.add_argument(
        "--pretrained_model_name_or_path",
        type=str,
        default=None,
        required=True,
        help="Path to pretrained model or model identifier from huggingface.co/models.",
    )
    parser.add_argument(
        "--revision",
        type=str,
        default=None,
        required=False,
        help="Revision of pretrained model identifier from huggingface.co/models.",
    )
    parser.add_argument(
        "--tokenizer_name",
        type=str,
        default=None,
        help="Pretrained tokenizer name or path if not the same as model_name",
    )
    parser.add_argument(
        "--concept_type",
        type=str,
        required=True,
        choices=['style', 'object', 'memorization'],
        help='the type of removed concepts'
    )
    parser.add_argument(
        "--caption_target",
        type=str,
        required=True,
        help="target style to remove, used when kldiv loss",
    )
    parser.add_argument(
        "--instance_data_dir",
        type=str,
        default=None,
        help="A folder containing the training data of instance images.",
    )
    parser.add_argument(
        "--class_data_dir",
        type=str,
        default=None,
        help="A folder containing the training data of class images.",
    )
    parser.add_argument(
        "--instance_prompt",
        type=str,
        help="The prompt with identifier specifying the instance",
    )
    parser.add_argument(
        "--class_prompt",
        type=str,
        default=None,
        help="The prompt to specify images in the same class as provided instance images.",
    )
    parser.add_argument(
        "--mem_impath",
        type=str,
        default="",
        help='the path to saved memorized image. Required when concept_type is memorization'
    )
    parser.add_argument(
        "--validation_prompt",
        type=str,
        default=None,
        help="A prompt that is used during validation to verify that the model is learning.",
    )
    parser.add_argument(
        "--num_validation_images",
        type=int,
        default=2,
        help="Number of images that should be generated during validation with `validation_prompt`.",
    )
    parser.add_argument(
        "--validation_steps",
        type=int,
        default=500,
        help=(
            "Run dreambooth validation every X epochs. Dreambooth validation consists of running the prompt"
            " `args.validation_prompt` multiple times: `args.num_validation_images`."
        ),
    )
    parser.add_argument(
        "--with_prior_preservation",
        default=False,
        action="store_true",
        help="Flag to add prior preservation loss.",
    )
    parser.add_argument("--prior_loss_weight", type=float,
                        default=1.0, help="The weight of prior preservation loss.")
    parser.add_argument(
        "--train_size",
        type=int,
        default=1000,
        help='the number of generated images used for ablating the concept'
    )
    parser.add_argument(
        "--output_dir",
        type=str,
        default="custom-diffusion-model",
        help="The output directory where the model predictions and checkpoints will be written.",
    )
    parser.add_argument(
        "--num_class_images",
        type=int,
        default=1000,
        help=(
            "Minimal anchor class images. If there are not enough images already present in"
            " class_data_dir, additional images will be sampled with class_prompt."
        ),
    )
    parser.add_argument(
        "--num_class_prompts",
        type=int,
        default=200,
        help=(
            "Minimal prompts used to generate anchor class images"
        ),
    )
    parser.add_argument("--seed", type=int, default=42,
                        help="A seed for reproducible training.")
    parser.add_argument(
        "--resolution",
        type=int,
        default=512,
        help=(
            "The resolution for input images, all the images in the train/validation dataset will be resized to this"
            " resolution"
        ),
    )
    parser.add_argument(
        "--center_crop",
        default=False,
        action="store_true",
        help=(
            "Whether to center crop the input images to the resolution. If not set, the images will be randomly"
            " cropped. The images will be resized to the resolution first before cropping."
        ),
    )
    parser.add_argument(
        "--train_batch_size", type=int, default=4, help="Batch size (per device) for the training dataloader."
    )
    parser.add_argument(
        "--sample_batch_size", type=int, default=4, help="Batch size (per device) for sampling images."
    )
    parser.add_argument("--num_train_epochs", type=int, default=1)
    parser.add_argument(
        "--max_train_steps",
        type=int,
        default=None,
        help="Total number of training steps to perform.  If provided, overrides num_train_epochs.",
    )
    parser.add_argument(
        "--checkpointing_steps",
        type=int,
        default=250,
        help=(
            "Save a checkpoint of the training state every X updates. These checkpoints can be used both as final"
            " checkpoints in case they are better than the last checkpoint, and are also suitable for resuming"
            " training using `--resume_from_checkpoint`."
        ),
    )
    parser.add_argument(
        "--checkpoints_total_limit",
        type=int,
        default=None,
        help=(
            "Max number of checkpoints to store. Passed as `total_limit` to the `Accelerator` `ProjectConfiguration`."
            " See Accelerator::save_state https://huggingface.co/docs/accelerate/package_reference/accelerator#accelerate.Accelerator.save_state"
            " for more docs"
        ),
    )
    parser.add_argument(
        "--resume_from_checkpoint",
        type=str,
        default=None,
        help=(
            "Whether training should be resumed from a previous checkpoint. Use a path saved by"
            ' `--checkpointing_steps`, or `"latest"` to automatically select the last available checkpoint.'
        ),
    )
    parser.add_argument(
        "--gradient_accumulation_steps",
        type=int,
        default=1,
        help="Number of updates steps to accumulate before performing a backward/update pass.",
    )
    parser.add_argument(
        "--gradient_checkpointing",
        action="store_true",
        help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
    )
    parser.add_argument(
        "--learning_rate",
        type=float,
        default=1e-5,
        help="Initial learning rate (after the potential warmup period) to use.",
    )
    parser.add_argument(
        "--scale_lr",
        action="store_true",
        default=False,
        help="Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.",
    )
    parser.add_argument(
        "--dataloader_num_workers",
        type=int,
        default=2,
        help=(
            "Number of subprocesses to use for data loading. 0 means that the data will be loaded in the main process."
        ),
    )
    parser.add_argument(
        "--parameter_group",
        type=str,
        default='cross-attn',
        choices=['full-weight', 'cross-attn', 'embedding'],
        help='parameter groups to finetune. Default: full-weight for memorization and cross-attn for others'
    )
    parser.add_argument(
        "--loss_type_reverse",
        type=str,
        default='model-based',
        help="loss type for reverse fine-tuning",
    )
    parser.add_argument(
        "--lr_scheduler",
        type=str,
        default="constant",
        help=(
            'The scheduler type to use. Choose between ["linear", "cosine", "cosine_with_restarts", "polynomial",'
            ' "constant", "constant_with_warmup"]'
        ),
    )
    parser.add_argument(
        "--lr_warmup_steps", type=int, default=500, help="Number of steps for the warmup in the lr scheduler."
    )
    parser.add_argument(
        "--use_8bit_adam", action="store_true", help="Whether or not to use 8-bit Adam from bitsandbytes."
    )
    parser.add_argument("--adam_beta1", type=float, default=0.9,
                        help="The beta1 parameter for the Adam optimizer.")
    parser.add_argument("--adam_beta2", type=float, default=0.999,
                        help="The beta2 parameter for the Adam optimizer.")
    parser.add_argument("--adam_weight_decay", type=float,
                        default=1e-2, help="Weight decay to use.")
    parser.add_argument("--adam_epsilon", type=float, default=1e-08,
                        help="Epsilon value for the Adam optimizer")
    parser.add_argument("--max_grad_norm", default=1.0,
                        type=float, help="Max gradient norm.")
    parser.add_argument("--push_to_hub", action="store_true",
                        help="Whether or not to push the model to the Hub.")
    parser.add_argument("--hub_token", type=str, default=None,
                        help="The token to use to push to the Model Hub.")
    parser.add_argument(
        "--hub_model_id",
        type=str,
        default=None,
        help="The name of the repository to keep in sync with the local `output_dir`.",
    )
    parser.add_argument(
        "--logging_dir",
        type=str,
        default="logs",
        help=(
            "[TensorBoard](https://www.tensorflow.org/tensorboard) log directory. Will default to"
            " *output_dir/runs/**CURRENT_DATETIME_HOSTNAME***."
        ),
    )
    parser.add_argument(
        "--allow_tf32",
        action="store_true",
        help=(
            "Whether or not to allow TF32 on Ampere GPUs. Can be used to speed up training. For more information, see"
            " https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices"
        ),
    )
    parser.add_argument(
        "--report_to",
        type=str,
        default="tensorboard",
        help=(
            'The integration to report the results and logs to. Supported platforms are `"tensorboard"`'
            ' (default), `"wandb"` and `"comet_ml"`. Use `"all"` to report to all integrations.'
        ),
    )
    parser.add_argument(
        "--mixed_precision",
        type=str,
        default=None,
        choices=["no", "fp16", "bf16"],
        help=(
            "Whether to use mixed precision. Choose between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
            " 1.10.and an Nvidia Ampere GPU.  Default to the value of accelerate config of the current system or the"
            " flag passed with the `accelerate.launch` command. Use this argument to override the accelerate config."
        ),
    )
    parser.add_argument(
        "--prior_generation_precision",
        type=str,
        default=None,
        choices=["no", "fp32", "fp16", "bf16"],
        help=(
            "Choose prior generation precision between fp32, fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
            " 1.10.and an Nvidia Ampere GPU.  Default to  fp16 if a GPU is available else fp32."
        ),
    )
    parser.add_argument(
        "--concepts_list",
        type=str,
        default=None,
        help="Path to json containing multiple concepts, will overwrite parameters like instance_prompt, class_prompt, etc.",
    )
    parser.add_argument(
        "--openai_key",
        type=str,
        default="",
        help=(
            "OPENAI API key. required for ablating objects and memorized images."
        ),
    )
    parser.add_argument("--local_rank", type=int, default=-1,
                        help="For distributed training: local_rank")
    parser.add_argument(
        "--enable_xformers_memory_efficient_attention", action="store_true", help="Whether or not to use xformers."
    )
    parser.add_argument("--hflip", action="store_true",
                        help="Apply horizontal flip data augmentation.")
    parser.add_argument("--noaug", action="store_true",
                        help="Dont apply augmentation during data augmentation when this flag is enabled.")

    if input_args is not None:
        args = parser.parse_args(input_args)
    else:
        args = parser.parse_args()

    env_local_rank = int(os.environ.get("LOCAL_RANK", -1))
    if env_local_rank != -1 and env_local_rank != args.local_rank:
        args.local_rank = env_local_rank

    if args.with_prior_preservation:
        if args.concepts_list is None:
            if args.class_data_dir is None:
                raise ValueError(
                    "You must specify a data directory for class images.")
            if args.class_prompt is None:
                raise ValueError("You must specify prompt for class images.")
    else:
        # logger is not available yet
        if args.class_data_dir is not None:
            warnings.warn(
                "You need not use --class_data_dir without --with_prior_preservation.")
        if args.class_prompt is not None:
            warnings.warn(
                "You need not use --class_prompt without --with_prior_preservation.")

    return args


def main(args):
    logging_dir = Path(args.output_dir, args.logging_dir)

    accelerator_project_config = ProjectConfiguration(
        total_limit=args.checkpoints_total_limit)

    accelerator = Accelerator(
        gradient_accumulation_steps=args.gradient_accumulation_steps,
        mixed_precision=args.mixed_precision,
        log_with=args.report_to,
        project_dir=logging_dir,
        project_config=accelerator_project_config,
    )

    if args.report_to == "wandb":
        if not is_wandb_available():
            raise ImportError(
                "Make sure to install wandb if you want to use it for logging during training.")
        import wandb

    # Currently, it's not possible to do gradient accumulation when training two models with accelerate.accumulate
    # This will be enabled soon in accelerate. For now, we don't allow gradient accumulation when training two models.
    # TODO (patil-suraj): Remove this check when gradient accumulation with two models is enabled in accelerate.
    # Make one log on every process with the configuration for debugging.
    logging.basicConfig(
        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
        datefmt="%m/%d/%Y %H:%M:%S",
        level=logging.INFO,
    )
    logger.info(accelerator.state, main_process_only=False)
    if accelerator.is_local_main_process:
        transformers.utils.logging.set_verbosity_warning()
        diffusers.utils.logging.set_verbosity_info()
    else:
        transformers.utils.logging.set_verbosity_error()
        diffusers.utils.logging.set_verbosity_error()

    # We need to initialize the trackers we use, and also store our configuration.
    # The trackers initializes automatically on the main process.
    if accelerator.is_main_process:
        print(vars(args))
        accelerator.init_trackers("custom-diffusion", config=vars(args))

    # If passed along, set the training seed now.
    if args.seed is not None:
        set_seed(args.seed)
    if args.concepts_list is None:
        args.concepts_list = [
            {
                "instance_prompt": args.instance_prompt,
                "class_prompt": args.class_prompt,
                "instance_data_dir": args.instance_data_dir,
                "class_data_dir": args.class_data_dir,
                "caption_target": args.caption_target,
            }
        ]
    else:
        with open(args.concepts_list, "r") as f:
            args.concepts_list = json.load(f)

    # Generate class images if prior preservation is enabled.
    for i, concept in enumerate(args.concepts_list):
        # directly path to ablation images and its corresponding prompts is provided.
        if (concept['instance_prompt'] is not None and concept['instance_data_dir'] is not None):
            break

        class_images_dir = Path(concept['class_data_dir'])
        if not class_images_dir.exists():
            class_images_dir.mkdir(parents=True, exist_ok=True)
        os.makedirs(f'{class_images_dir}/images', exist_ok=True)

        # we need to generate training images
        if len(list(Path(os.path.join(class_images_dir, 'images')).iterdir())) < args.num_class_images:
            torch_dtype = torch.float16 if accelerator.device.type == "cuda" else torch.float32
            if args.prior_generation_precision == "fp32":
                torch_dtype = torch.float32
            elif args.prior_generation_precision == "fp16":
                torch_dtype = torch.float16
            elif args.prior_generation_precision == "bf16":
                torch_dtype = torch.bfloat16
            pipeline = DiffusionPipeline.from_pretrained(
                args.pretrained_model_name_or_path,
                torch_dtype=torch_dtype,
                safety_checker=None,
                revision=args.revision,
            )
            pipeline.scheduler = DPMSolverMultistepScheduler.from_config(
                                    pipeline.scheduler.config)

            pipeline.set_progress_bar_config(disable=True)
            pipeline.to(accelerator.device)

            # need to create prompts using class_prompt.
            if not os.path.isfile(concept['class_prompt']):
                # style based prompts are retrieved from laion dataset
                if args.concept_type == 'style':
                    with open(os.path.join(class_images_dir, 'painting.txt')) as f:
                        class_prompt_collection = [
                                x.strip() for x in f.readlines()]

                # LLM based prompt collection.
                else:
                    class_prompt = concept['class_prompt']
                    # in case of object query chatGPT to generate captions containing the anchor category
                    if args.concept_type == 'object':
                        class_prompt_collection, _ = getanchorprompts(
                            pipeline, accelerator, class_prompt, args.concept_type, class_images_dir, args.openai_key, args.num_class_prompts)
                        with open(class_images_dir / 'caption_anchor.txt', 'w') as f:
                            for prompt in class_prompt_collection:
                                f.write(prompt + '\n')
                    # in case of memorization query chatGPT to generate different captions that can be paraphrase of the origianl caption
                    elif args.concept_type == 'memorization':
                        class_prompt_collection, caption_target = getanchorprompts(
                            pipeline, accelerator, class_prompt, args.concept_type, class_images_dir, args.openai_key, args.num_class_prompts, mem_impath=args.mem_impath)
                        concept['caption_target'] += f';*+{caption_target}'
                        with open(class_images_dir / 'caption_target.txt', 'w') as f:
                            f.write(concept['caption_target'])
                        print(class_prompt_collection,
                              concept['caption_target'])
            # class_prompt is filepath to prompts.
            else:
                with open(concept['class_prompt']) as f:
                    class_prompt_collection = [
                        x.strip() for x in f.readlines()]

            num_new_images = args.num_class_images
            logger.info(
                f"Number of class images to sample: {num_new_images}.")

            sample_dataset = PromptDataset(
                class_prompt_collection, num_new_images)
            sample_dataloader = torch.utils.data.DataLoader(
                sample_dataset, batch_size=args.sample_batch_size)

            sample_dataloader = accelerator.prepare(sample_dataloader)

            if os.path.exists(f'{class_images_dir}/caption.txt'):
                os.remove(f'{class_images_dir}/caption.txt')
            if os.path.exists(f'{class_images_dir}/images.txt'):
                os.remove(f'{class_images_dir}/images.txt')

            for example in tqdm(
                sample_dataloader, desc="Generating class images", disable=not accelerator.is_local_main_process
            ):
                accelerator.wait_for_everyone()
                with open(f'{class_images_dir}/caption.txt', 'a') as f1, open(f'{class_images_dir}/images.txt', 'a') as f2:
                    images = pipeline(example["prompt"], num_inference_steps=25, guidance_scale=6., eta=1.).images

                    for i, image in enumerate(images):
                        hash_image = hashlib.sha1(
                            image.tobytes()).hexdigest()
                        image_filename = class_images_dir / \
                            f"images/{example['index'][i]}-{hash_image}.jpg"
                        image.save(image_filename)
                        f2.write(str(image_filename)+'\n')
                    f1.write('\n'.join(example["prompt"]) + '\n')
                    accelerator.wait_for_everyone()

            del pipeline

        if args.concept_type == 'memorization':
            filter(class_images_dir, args.mem_impath,
                    outpath=str(class_images_dir / 'filtered'))
            if os.path.exists(class_images_dir / 'caption_target.txt'):
                with open(class_images_dir / 'caption_target.txt', 'r') as f:
                    concept['caption_target'] = f.readlines()[0].strip()
            class_images_dir = class_images_dir / 'filtered'

        concept['class_prompt'] = os.path.join(
            class_images_dir, 'caption.txt')
        concept['class_data_dir'] = os.path.join(
            class_images_dir, 'images.txt')
        concept['instance_prompt'] = os.path.join(
            class_images_dir, 'caption.txt')
        concept['instance_data_dir'] = os.path.join(
            class_images_dir, 'images.txt')

        if torch.cuda.is_available():
            torch.cuda.empty_cache()

    # Handle the repository creation
    if accelerator.is_main_process:
        if args.output_dir is not None:
            os.makedirs(args.output_dir, exist_ok=True)

        if args.push_to_hub:
            print(args.hub_model_id or Path(args.output_dir).name)
            repo_id = create_repo(
                repo_id=args.hub_model_id or Path(args.output_dir).name, exist_ok=True, token=args.hub_token
            )
            print(repo_id)
            repo_id = args.hub_model_id

    # Load the tokenizer
    if args.tokenizer_name:
        tokenizer = AutoTokenizer.from_pretrained(
            args.tokenizer_name,
            revision=args.revision,
            use_fast=False,
        )
    elif args.pretrained_model_name_or_path:
        tokenizer = AutoTokenizer.from_pretrained(
            args.pretrained_model_name_or_path,
            subfolder="tokenizer",
            revision=args.revision,
            use_fast=False,
        )

    # import correct text encoder class
    text_encoder_cls = import_model_class_from_model_name_or_path(
        args.pretrained_model_name_or_path, args.revision)

    # Load scheduler and models
    noise_scheduler = DDPMScheduler.from_pretrained(
        args.pretrained_model_name_or_path, subfolder="scheduler")
    text_encoder = text_encoder_cls.from_pretrained(
        args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision
    )
    vae = AutoencoderKL.from_pretrained(
        args.pretrained_model_name_or_path, subfolder="vae", revision=args.revision)
    unet = UNet2DConditionModel.from_pretrained(
        args.pretrained_model_name_or_path, subfolder="unet", revision=args.revision
    )

    vae.requires_grad_(False)
    if args.parameter_group != 'embedding':
        text_encoder.requires_grad_(False)
    unet = create_custom_diffusion(unet, args.parameter_group)

    # For mixed precision training we cast the text_encoder and vae weights to half-precision
    # as these models are only used for inference, keeping weights in full precision is not required.
    weight_dtype = torch.float32
    if accelerator.mixed_precision == "fp16":
        weight_dtype = torch.float16
    elif accelerator.mixed_precision == "bf16":
        weight_dtype = torch.bfloat16

    # Move unet, vae and text_encoder to device and cast to weight_dtype
    if accelerator.mixed_precision != "fp16":
        unet.to(accelerator.device, dtype=weight_dtype)
        text_encoder.to(accelerator.device, dtype=weight_dtype)
    vae.to(accelerator.device, dtype=weight_dtype)

    if args.enable_xformers_memory_efficient_attention:
        if is_xformers_available():
            import xformers
            xformers_version = version.parse(xformers.__version__)
            if xformers_version == version.parse("0.0.16"):
                logger.warn(
                    "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                )
            unet.enable_xformers_memory_efficient_attention()
        else:
            raise ValueError(
                "xformers is not available. Make sure it is installed correctly")

    if args.gradient_checkpointing:
        unet.enable_gradient_checkpointing()
        if args.parameter_group == 'embedding':
            text_encoder.gradient_checkpointing_enable()
    # Enable TF32 for faster training on Ampere GPUs,
    # cf https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices
    if args.allow_tf32:
        torch.backends.cuda.matmul.allow_tf32 = True

    if args.scale_lr:
        args.learning_rate = (
            args.learning_rate * args.gradient_accumulation_steps *
            args.train_batch_size * accelerator.num_processes
        )
        if args.with_prior_preservation:
            args.learning_rate = args.learning_rate * 2.

    # Use 8-bit Adam for lower memory usage or to fine-tune the model in 16GB GPUs
    if args.use_8bit_adam:
        try:
            import bitsandbytes as bnb
        except ImportError:
            raise ImportError(
                "To use 8-bit Adam, please install the bitsandbytes library: `pip install bitsandbytes`."
            )

        optimizer_class = bnb.optim.AdamW8bit
    else:
        optimizer_class = torch.optim.AdamW

    # Adding a modifier token which is optimized ####
    # Code taken from https://github.com/huggingface/diffusers/blob/main/examples/textual_inversion/textual_inversion.py
    modifier_token_id = []
    if args.parameter_group == 'embedding':
        assert args.concept_type != 'memorization', "embedding finetuning is not supported for memorization"

        for concept in args.concept_list:
            # Convert the caption_target to ids
            token_ids = tokenizer.encode(
                [concept['caption_target']], add_special_tokens=False)
            print(token_ids)
        # Check if initializer_token is a single token or a sequence of tokens
        modifier_token_id += token_ids

        # Freeze all parameters except for the token embeddings in text encoder
        params_to_freeze = itertools.chain(
            text_encoder.text_model.encoder.parameters(),
            text_encoder.text_model.final_layer_norm.parameters(),
            text_encoder.text_model.embeddings.position_embedding.parameters(),
        )
        freeze_params(params_to_freeze)
        params_to_optimize = itertools.chain(
            text_encoder.get_input_embeddings().parameters())
    else:
        if args.parameter_group == 'cross-attn':
            params_to_optimize = itertools.chain([x[1] for x in unet.named_parameters() if (
                'attn2.to_k' in x[0] or 'attn2.to_v' in x[0])])
        if args.parameter_group == 'full-weight':
            params_to_optimize = itertools.chain(unet.parameters())

    # Optimizer creation
    optimizer = optimizer_class(
        params_to_optimize,
        lr=args.learning_rate,
        betas=(args.adam_beta1, args.adam_beta2),
        weight_decay=args.adam_weight_decay,
        eps=args.adam_epsilon,
    )

    # Dataset and DataLoaders creation:
    train_dataset = CustomDiffusionDataset(
        concepts_list=args.concepts_list,
        concept_type=args.concept_type,
        tokenizer=tokenizer,
        with_prior_preservation=args.with_prior_preservation,
        size=args.resolution,
        center_crop=args.center_crop,
        num_class_images=args.num_class_images,
        hflip=args.hflip, aug=not args.noaug,
    )

    train_dataloader = torch.utils.data.DataLoader(
        train_dataset,
        batch_size=args.train_batch_size,
        shuffle=True,
        collate_fn=lambda examples: collate_fn(
            examples, args.with_prior_preservation),
        num_workers=args.dataloader_num_workers,
    )

    # Scheduler and math around the number of training steps.
    overrode_max_train_steps = False
    num_update_steps_per_epoch = math.ceil(
        len(train_dataloader) / args.gradient_accumulation_steps)
    if args.max_train_steps is None:
        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
        overrode_max_train_steps = True

    lr_scheduler = get_scheduler(
        args.lr_scheduler,
        optimizer=optimizer,
        num_warmup_steps=args.lr_warmup_steps * args.gradient_accumulation_steps,
        num_training_steps=args.max_train_steps * args.gradient_accumulation_steps,
    )

    # Prepare everything with our `accelerator`.
    if args.parameter_group == 'embedding':
        text_encoder, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
            text_encoder, optimizer, train_dataloader, lr_scheduler
        )
    else:
        unet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
            unet, optimizer, train_dataloader, lr_scheduler
        )

    # We need to recalculate our total training steps as the size of the training dataloader may have changed.
    num_update_steps_per_epoch = math.ceil(
        len(train_dataloader) / args.gradient_accumulation_steps)
    if overrode_max_train_steps:
        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
    # Afterwards we recalculate our number of training epochs
    args.num_train_epochs = math.ceil(
        args.max_train_steps / num_update_steps_per_epoch)

    # Train!
    total_batch_size = args.train_batch_size * \
        accelerator.num_processes * args.gradient_accumulation_steps

    logger.info("***** Running training *****")
    logger.info(f"  Num examples = {len(train_dataset)}")
    logger.info(f"  Num batches each epoch = {len(train_dataloader)}")
    logger.info(f"  Num Epochs = {args.num_train_epochs}")
    logger.info(
        f"  Instantaneous batch size per device = {args.train_batch_size}")
    logger.info(
        f"  Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
    logger.info(
        f"  Gradient Accumulation steps = {args.gradient_accumulation_steps}")
    logger.info(f"  Total optimization steps = {args.max_train_steps}")
    global_step = 0
    first_epoch = 0

    # Potentially load in the weights and states from a previous save
    if args.resume_from_checkpoint:
        if args.resume_from_checkpoint != "latest":
            path = os.path.basename(args.resume_from_checkpoint)
        else:
            # Get the mos recent checkpoint
            dirs = os.listdir(args.output_dir)
            dirs = [d for d in dirs if d.startswith("checkpoint")]
            dirs = sorted(dirs, key=lambda x: int(x.split("-")[1]))
            path = dirs[-1] if len(dirs) > 0 else None

        if path is None:
            accelerator.print(
                f"Checkpoint '{args.resume_from_checkpoint}' does not exist. Starting a new training run."
            )
            args.resume_from_checkpoint = None
        else:
            accelerator.print(f"Resuming from checkpoint {path}")
            accelerator.load_state(os.path.join(args.output_dir, path))
            global_step = int(path.split("-")[1])

            resume_global_step = global_step * args.gradient_accumulation_steps
            first_epoch = global_step // num_update_steps_per_epoch
            resume_step = resume_global_step % (
                num_update_steps_per_epoch * args.gradient_accumulation_steps)

    # Only show the progress bar once on each machine.
    progress_bar = tqdm(range(global_step, args.max_train_steps),
                        disable=not accelerator.is_local_main_process)
    progress_bar.set_description("Steps")

    for epoch in range(first_epoch, args.num_train_epochs):
        if args.parameter_group == 'embedding':
            text_encoder.train()
        else:
            unet.train()
        for step, batch in enumerate(train_dataloader):
            # Skip steps until we reach the resumed step
            if args.resume_from_checkpoint and epoch == first_epoch and step < resume_step:
                if step % args.gradient_accumulation_steps == 0:
                    progress_bar.update(1)
                continue

            with accelerator.accumulate(unet) if args.parameter_group != 'embedding' else accelerator.accumulate(text_encoder):
                # Convert images to latent space
                latents = vae.encode(batch["pixel_values"].to(
                    dtype=weight_dtype)).latent_dist.sample()
                latents = latents * vae.config.scaling_factor

                # Sample noise that we'll add to the latents
                noise = torch.randn_like(latents)
                bsz = latents.shape[0]
                # Sample a random timestep for each image
                timesteps = torch.randint(
                    0, noise_scheduler.config.num_train_timesteps, (bsz,), device=latents.device)
                timesteps = timesteps.long()

                # Add noise to the latents according to the noise magnitude at each timestep
                # (this is the forward diffusion process)
                noisy_latents = noise_scheduler.add_noise(
                    latents, noise, timesteps)

                # Get the text embedding for conditioning
                encoder_hidden_states = text_encoder(batch["input_ids"])[0]
                encoder_anchor_hidden_states = text_encoder(
                    batch["input_anchor_ids"])[0]

                # Predict the noise residual
                model_pred = unet(noisy_latents, timesteps,
                                  encoder_hidden_states).sample
                with torch.no_grad():
                    model_pred_anchor = unet(noisy_latents[:encoder_anchor_hidden_states.size(
                        0)], timesteps[:encoder_anchor_hidden_states.size(0)], encoder_anchor_hidden_states).sample

                # Get the target for loss depending on the prediction type
                if args.loss_type_reverse == 'model-based':
                    if args.with_prior_preservation:
                        target_prior = torch.chunk(noise, 2, dim=0)[1]
                    target = model_pred_anchor
                else:
                    if noise_scheduler.config.prediction_type == "epsilon":
                        target = noise
                    elif noise_scheduler.config.prediction_type == "v_prediction":
                        target = noise_scheduler.get_velocity(
                            latents, noise, timesteps)
                    else:
                        raise ValueError(
                            f"Unknown prediction type {noise_scheduler.config.prediction_type}")
                    if args.with_prior_preservation:
                        target, target_prior = torch.chunk(target, 2, dim=0)

                if args.with_prior_preservation:
                    # Chunk the noise and model_pred into two parts and compute the loss on each part separately.
                    model_pred, model_pred_prior = torch.chunk(
                        model_pred, 2, dim=0)
                    mask = torch.chunk(batch["mask"], 2, dim=0)[0]
                    # Compute instance loss
                    loss = F.mse_loss(model_pred.float(),
                                      target.float(), reduction="none")
                    loss = (
                        (loss * mask).sum([1, 2, 3]) / mask.sum([1, 2, 3])).mean()

                    # Compute prior loss
                    prior_loss = F.mse_loss(
                        model_pred_prior.float(), target_prior.float(), reduction="mean")

                    # Add the prior loss to the instance loss.
                    loss = loss + args.prior_loss_weight * prior_loss
                else:
                    mask = batch["mask"]
                    loss = F.mse_loss(model_pred.float(),
                                      target.float(), reduction="none")
                    loss = (
                        (loss * mask).sum([1, 2, 3]) / mask.sum([1, 2, 3])).mean()

                accelerator.backward(loss)
                # Zero out the gradients for all token embeddings except the newly added
                # embeddings for the concept, as we only want to optimize the concept embeddings
                if args.parameter_group == 'embedding':
                    if accelerator.num_processes > 1:
                        grads_text_encoder = text_encoder.module.get_input_embeddings().weight.grad
                    else:
                        grads_text_encoder = text_encoder.get_input_embeddings().weight.grad
                    # Get the index for tokens that we want to zero the grads for
                    index_grads_to_zero = torch.arange(
                        len(tokenizer)) != modifier_token_id[0]
                    for i in range(len(modifier_token_id[1:])):
                        index_grads_to_zero = index_grads_to_zero & (
                            torch.arange(len(tokenizer)) != modifier_token_id[i])
                    grads_text_encoder.data[index_grads_to_zero,
                                            :] = grads_text_encoder.data[index_grads_to_zero, :].fill_(0)

                if accelerator.sync_gradients:
                    params_to_clip = (
                        itertools.chain(text_encoder.parameters())
                        if args.parameter_group == 'embedding'
                        else itertools.chain([x[1] for x in unet.named_parameters() if ('attn2' in x[0])])
                        if args.parameter_group == 'cross-attn'
                        else itertools.chain(unet.parameters())
                    )
                    accelerator.clip_grad_norm_(
                        params_to_clip, args.max_grad_norm)
                optimizer.step()
                lr_scheduler.step()
                optimizer.zero_grad()

            # Checks if the accelerator has performed an optimization step behind the scenes
            if accelerator.sync_gradients:
                progress_bar.update(1)
                global_step += 1

                if global_step % args.checkpointing_steps == 0:
                    if accelerator.is_main_process:
                        pipeline = CustomDiffusionPipeline.from_pretrained(
                            args.pretrained_model_name_or_path,
                            unet=accelerator.unwrap_model(unet),
                            text_encoder=accelerator.unwrap_model(
                                text_encoder),
                            tokenizer=tokenizer,
                            revision=args.revision,
                            modifier_token_id=modifier_token_id,
                        )
                        save_path = os.path.join(
                            args.output_dir, f"delta-{global_step}")
                        pipeline.save_pretrained(
                            save_path, parameter_group=args.parameter_group)
                        logger.info(f"Saved state to {save_path}")

            logs = {"loss": loss.detach().item(
            ), "lr": lr_scheduler.get_last_lr()[0]}
            progress_bar.set_postfix(**logs)
            accelerator.log(logs, step=global_step)

            if global_step >= args.max_train_steps:
                break

        if accelerator.is_main_process:
            if args.validation_prompt is not None and global_step % args.validation_steps == 0:
                logger.info(
                    f"Running validation... \n Generating {args.num_validation_images} images with prompt:"
                    f" {args.validation_prompt}."
                )
                # create pipeline
                pipeline = CustomDiffusionPipeline.from_pretrained(
                    args.pretrained_model_name_or_path,
                    unet=accelerator.unwrap_model(unet),
                    text_encoder=accelerator.unwrap_model(text_encoder),
                    tokenizer=tokenizer,
                    revision=args.revision,
                    modifier_token_id=modifier_token_id,
                )
                pipeline.scheduler = DPMSolverMultistepScheduler.from_config(
                    pipeline.scheduler.config)
                pipeline = pipeline.to(accelerator.device)
                pipeline.set_progress_bar_config(disable=True)

                # run inference
                generator = torch.Generator(
                    device=accelerator.device).manual_seed(args.seed)
                images = [
                    pipeline(args.validation_prompt, num_inference_steps=25,
                             generator=generator, eta=1.).images[0]
                    for _ in range(args.num_validation_images)
                ]

                for tracker in accelerator.trackers:
                    if tracker.name == "tensorboard":
                        np_images = np.stack([np.asarray(img)
                                             for img in images])
                        tracker.writer.add_images(
                            "validation", np_images, epoch, dataformats="NHWC")
                    if tracker.name == "wandb":
                        tracker.log(
                            {
                                "validation": [
                                    wandb.Image(
                                        image, caption=f"{i}: {args.validation_prompt}")
                                    for i, image in enumerate(images)
                                ]
                            }
                        )

                del pipeline
                torch.cuda.empty_cache()

    accelerator.wait_for_everyone()
    if accelerator.is_main_process:
        unet = unet.to(torch.float32)
        pipeline = CustomDiffusionPipeline.from_pretrained(
            args.pretrained_model_name_or_path,
            unet=accelerator.unwrap_model(unet),
            text_encoder=accelerator.unwrap_model(text_encoder),
            tokenizer=tokenizer,
            revision=args.revision,
            modifier_token_id=modifier_token_id,
        )
        save_path = os.path.join(args.output_dir, "delta.bin")
        pipeline.save_pretrained(
            save_path, parameter_group=args.parameter_group)

        # run inference
        if args.validation_prompt and args.num_validation_images > 0:
            pipeline.scheduler = DPMSolverMultistepScheduler.from_config(
                pipeline.scheduler.config)
            pipeline = pipeline.to(accelerator.device)
            pipeline.set_progress_bar_config(disable=True)

            # run inference
            generator = torch.Generator(
                device=accelerator.device).manual_seed(args.seed)
            images = [
                pipeline(args.validation_prompt, num_inference_steps=25,
                         generator=generator, eta=1.).images[0]
                for _ in range(args.num_validation_images)
            ]

            for tracker in accelerator.trackers:
                if tracker.name == "tensorboard":
                    np_images = np.stack([np.asarray(img) for img in images])
                    tracker.writer.add_images(
                        "test", np_images, epoch, dataformats="NHWC")
                if tracker.name == "wandb":
                    tracker.log(
                        {
                            "test": [
                                wandb.Image(
                                    image, caption=f"{i}: {args.validation_prompt}")
                                for i, image in enumerate(images)
                            ]
                        }
                    )

        if args.push_to_hub:
            save_model_card(
                repo_id,
                images=images,
                base_model=args.pretrained_model_name_or_path,
                prompt=args.instance_prompt,
                repo_folder=args.output_dir,
            )
            api = HfApi(token=args.hub_token)
            api.upload_folder(
                repo_id=repo_id,
                folder_path=args.output_dir,
                path_in_repo='.',
                repo_type='model'
            )

    accelerator.end_training()


if __name__ == "__main__":
    args = parse_args()
    main(args)