File size: 41,039 Bytes
162c70e
 
 
 
 
 
 
 
 
 
 
 
 
be7e901
 
162c70e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
66aec19
162c70e
 
 
 
 
 
 
3fcb5ce
 
162c70e
3fcb5ce
 
162c70e
3fcb5ce
 
 
162c70e
3fcb5ce
 
162c70e
3fcb5ce
162c70e
3fcb5ce
 
 
 
 
 
 
162c70e
3fcb5ce
 
 
 
162c70e
3fcb5ce
162c70e
3fcb5ce
 
162c70e
3fcb5ce
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

import inspect
import warnings
from itertools import repeat
from typing import Callable, List, Optional, Union

import torch
from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer

from diffusers.image_processor import VaeImageProcessor
from diffusers.models import AutoencoderKL, UNet2DConditionModel
from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
from diffusers.schedulers import KarrasDiffusionSchedulers
from diffusers.utils import logging
from diffusers.utils.torch_utils import randn_tensor
from diffusers.pipelines.pipeline_utils import DiffusionPipeline
from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
# from . import SemanticStableDiffusionPipelineOutput


logger = logging.get_logger(__name__)  # pylint: disable=invalid-name


class SemanticStableDiffusionPipeline(DiffusionPipeline):
    r"""
    Pipeline for text-to-image generation with latent editing.

    This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods the
    library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.)

    This model builds on the implementation of ['StableDiffusionPipeline']

    Args:
        vae ([`AutoencoderKL`]):
            Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.
        text_encoder ([`CLIPTextModel`]):
            Frozen text-encoder. Stable Diffusion uses the text portion of
            [CLIP](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModel), specifically
            the [clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14) variant.
        tokenizer (`CLIPTokenizer`):
            Tokenizer of class
            [CLIPTokenizer](https://huggingface.co/docs/transformers/v4.21.0/en/model_doc/clip#transformers.CLIPTokenizer).
        unet ([`UNet2DConditionModel`]): Conditional U-Net architecture to denoise the encoded image latents.
        scheduler ([`SchedulerMixin`]):
            A scheduler to be used in combination with `unet` to denoise the encoded image latens. Can be one of
            [`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
        safety_checker ([`Q16SafetyChecker`]):
            Classification module that estimates whether generated images could be considered offensive or harmful.
            Please, refer to the [model card](https://huggingface.co/CompVis/stable-diffusion-v1-4) for details.
        feature_extractor ([`CLIPImageProcessor`]):
            Model that extracts features from generated images to be used as inputs for the `safety_checker`.
    """

    _optional_components = ["safety_checker", "feature_extractor"]

    def __init__(
        self,
        vae: AutoencoderKL,
        text_encoder: CLIPTextModel,
        tokenizer: CLIPTokenizer,
        unet: UNet2DConditionModel,
        scheduler: KarrasDiffusionSchedulers,
        safety_checker: StableDiffusionSafetyChecker,
        feature_extractor: CLIPImageProcessor,
        requires_safety_checker: bool = True,
    ):
        super().__init__()

        if safety_checker is None and requires_safety_checker:
            logger.warning(
                f"You have disabled the safety checker for {self.__class__} by passing `safety_checker=None`. Ensure"
                " that you abide to the conditions of the Stable Diffusion license and do not expose unfiltered"
                " results in services or applications open to the public. Both the diffusers team and Hugging Face"
                " strongly recommend to keep the safety filter enabled in all public facing circumstances, disabling"
                " it only for use-cases that involve analyzing network behavior or auditing its results. For more"
                " information, please have a look at https://github.com/huggingface/diffusers/pull/254 ."
            )

        if safety_checker is not None and feature_extractor is None:
            raise ValueError(
                "Make sure to define a feature extractor when loading {self.__class__} if you want to use the safety"
                " checker. If you do not want to use the safety checker, you can pass `'safety_checker=None'` instead."
            )

        self.register_modules(
            vae=vae,
            text_encoder=text_encoder,
            tokenizer=tokenizer,
            unet=unet,
            scheduler=scheduler,
            safety_checker=safety_checker,
            feature_extractor=feature_extractor,
        )
        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
        self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
        self.register_to_config(requires_safety_checker=requires_safety_checker)

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
    def run_safety_checker(self, image, device, dtype):
        if self.safety_checker is None:
            has_nsfw_concept = None
        else:
            if torch.is_tensor(image):
                feature_extractor_input = self.image_processor.postprocess(image, output_type="pil")
            else:
                feature_extractor_input = self.image_processor.numpy_to_pil(image)
            safety_checker_input = self.feature_extractor(feature_extractor_input, return_tensors="pt").to(device)
            image, has_nsfw_concept = self.safety_checker(
                images=image, clip_input=safety_checker_input.pixel_values.to(dtype)
            )
        return image, has_nsfw_concept

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
    def decode_latents(self, latents):
        warnings.warn(
            "The decode_latents method is deprecated and will be removed in a future version. Please"
            " use VaeImageProcessor instead",
            FutureWarning,
        )
        latents = 1 / self.vae.config.scaling_factor * latents
        image = self.vae.decode(latents, return_dict=False)[0]
        image = (image / 2 + 0.5).clamp(0, 1)
        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
        return image

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
    def prepare_extra_step_kwargs(self, generator, eta):
        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
        # eta corresponds to η in DDIM paper: https://arxiv.org/abs/2010.02502
        # and should be between [0, 1]

        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
        extra_step_kwargs = {}
        if accepts_eta:
            extra_step_kwargs["eta"] = eta

        # check if the scheduler accepts generator
        accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys())
        if accepts_generator:
            extra_step_kwargs["generator"] = generator
        return extra_step_kwargs

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.check_inputs
    def check_inputs(
        self,
        prompt,
        height,
        width,
        callback_steps,
        negative_prompt=None,
        prompt_embeds=None,
        negative_prompt_embeds=None,
    ):
        if height % 8 != 0 or width % 8 != 0:
            raise ValueError(f"`height` and `width` have to be divisible by 8 but are {height} and {width}.")

        if (callback_steps is None) or (
            callback_steps is not None and (not isinstance(callback_steps, int) or callback_steps <= 0)
        ):
            raise ValueError(
                f"`callback_steps` has to be a positive integer but is {callback_steps} of type"
                f" {type(callback_steps)}."
            )

        if prompt is not None and prompt_embeds is not None:
            raise ValueError(
                f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
                " only forward one of the two."
            )
        elif prompt is None and prompt_embeds is None:
            raise ValueError(
                "Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined."
            )
        elif prompt is not None and (not isinstance(prompt, str) and not isinstance(prompt, list)):
            raise ValueError(f"`prompt` has to be of type `str` or `list` but is {type(prompt)}")

        if negative_prompt is not None and negative_prompt_embeds is not None:
            raise ValueError(
                f"Cannot forward both `negative_prompt`: {negative_prompt} and `negative_prompt_embeds`:"
                f" {negative_prompt_embeds}. Please make sure to only forward one of the two."
            )

        if prompt_embeds is not None and negative_prompt_embeds is not None:
            if prompt_embeds.shape != negative_prompt_embeds.shape:
                raise ValueError(
                    "`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly, but"
                    f" got: `prompt_embeds` {prompt_embeds.shape} != `negative_prompt_embeds`"
                    f" {negative_prompt_embeds.shape}."
                )

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
    def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
        shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
        if isinstance(generator, list) and len(generator) != batch_size:
            raise ValueError(
                f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
                f" size of {batch_size}. Make sure the batch size matches the length of the generators."
            )

        if latents is None:
            latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
        else:
            latents = latents.to(device)

        # scale the initial noise by the standard deviation required by the scheduler
        latents = latents * self.scheduler.init_noise_sigma
        return latents

    @torch.no_grad()
    def __call__(
        self,
        prompt: Union[str, List[str]],
        height: Optional[int] = None,
        width: Optional[int] = None,
        num_inference_steps: int = 50,
        guidance_scale: float = 7.5,
        negative_prompt: Optional[Union[str, List[str]]] = None,
        num_images_per_prompt: int = 1,
        eta: float = 0.0,
        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
        latents: Optional[torch.FloatTensor] = None,
        output_type: Optional[str] = "pil",
        return_dict: bool = True,
        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
        callback_steps: int = 1,
        editing_prompt: Optional[Union[str, List[str]]] = None,
        editing_prompt_embeddings: Optional[torch.Tensor] = None,
        reverse_editing_direction: Optional[Union[bool, List[bool]]] = False,
        edit_guidance_scale: Optional[Union[float, List[float]]] = 5,
        edit_warmup_steps: Optional[Union[int, List[int]]] = 10,
        edit_cooldown_steps: Optional[Union[int, List[int]]] = None,
        edit_threshold: Optional[Union[float, List[float]]] = 0.9,
        edit_momentum_scale: Optional[float] = 0.1,
        edit_mom_beta: Optional[float] = 0.4,
        edit_weights: Optional[List[float]] = None,
        sem_guidance: Optional[List[torch.Tensor]] = None,

        # DDPM additions
        use_ddpm: bool = False,
        wts: Optional[List[torch.Tensor]] = None,
        zs: Optional[List[torch.Tensor]] = None
    ):
        r"""
        Function invoked when calling the pipeline for generation.

        Args:
            prompt (`str` or `List[str]`):
                The prompt or prompts to guide the image generation.
            height (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor):
                The height in pixels of the generated image.
            width (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor):
                The width in pixels of the generated image.
            num_inference_steps (`int`, *optional*, defaults to 50):
                The number of denoising steps. More denoising steps usually lead to a higher quality image at the
                expense of slower inference.
            guidance_scale (`float`, *optional*, defaults to 7.5):
                Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
                `guidance_scale` is defined as `w` of equation 2. of [Imagen
                Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
                1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
                usually at the expense of lower image quality.
            negative_prompt (`str` or `List[str]`, *optional*):
                The prompt or prompts not to guide the image generation. Ignored when not using guidance (i.e., ignored
                if `guidance_scale` is less than `1`).
            num_images_per_prompt (`int`, *optional*, defaults to 1):
                The number of images to generate per prompt.
            eta (`float`, *optional*, defaults to 0.0):
                Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to
                [`schedulers.DDIMScheduler`], will be ignored for others.
            generator (`torch.Generator`, *optional*):
                One or a list of [torch generator(s)](https://pytorch.org/docs/stable/generated/torch.Generator.html)
                to make generation deterministic.
            latents (`torch.FloatTensor`, *optional*):
                Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
                generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                tensor will ge generated by sampling using the supplied random `generator`.
            output_type (`str`, *optional*, defaults to `"pil"`):
                The output format of the generate image. Choose between
                [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
            return_dict (`bool`, *optional*, defaults to `True`):
                Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
                plain tuple.
            callback (`Callable`, *optional*):
                A function that will be called every `callback_steps` steps during inference. The function will be
                called with the following arguments: `callback(step: int, timestep: int, latents: torch.FloatTensor)`.
            callback_steps (`int`, *optional*, defaults to 1):
                The frequency at which the `callback` function will be called. If not specified, the callback will be
                called at every step.
            editing_prompt (`str` or `List[str]`, *optional*):
                The prompt or prompts to use for Semantic guidance. Semantic guidance is disabled by setting
                `editing_prompt = None`. Guidance direction of prompt should be specified via
                `reverse_editing_direction`.
            editing_prompt_embeddings (`torch.Tensor>`, *optional*):
                Pre-computed embeddings to use for semantic guidance. Guidance direction of embedding should be
                specified via `reverse_editing_direction`.
            reverse_editing_direction (`bool` or `List[bool]`, *optional*, defaults to `False`):
                Whether the corresponding prompt in `editing_prompt` should be increased or decreased.
            edit_guidance_scale (`float` or `List[float]`, *optional*, defaults to 5):
                Guidance scale for semantic guidance. If provided as list values should correspond to `editing_prompt`.
                `edit_guidance_scale` is defined as `s_e` of equation 6 of [SEGA
                Paper](https://arxiv.org/pdf/2301.12247.pdf).
            edit_warmup_steps (`float` or `List[float]`, *optional*, defaults to 10):
                Number of diffusion steps (for each prompt) for which semantic guidance will not be applied. Momentum
                will still be calculated for those steps and applied once all warmup periods are over.
                `edit_warmup_steps` is defined as `delta` (δ) of [SEGA Paper](https://arxiv.org/pdf/2301.12247.pdf).
            edit_cooldown_steps (`float` or `List[float]`, *optional*, defaults to `None`):
                Number of diffusion steps (for each prompt) after which semantic guidance will no longer be applied.
            edit_threshold (`float` or `List[float]`, *optional*, defaults to 0.9):
                Threshold of semantic guidance.
            edit_momentum_scale (`float`, *optional*, defaults to 0.1):
                Scale of the momentum to be added to the semantic guidance at each diffusion step. If set to 0.0
                momentum will be disabled. Momentum is already built up during warmup, i.e. for diffusion steps smaller
                than `sld_warmup_steps`. Momentum will only be added to latent guidance once all warmup periods are
                finished. `edit_momentum_scale` is defined as `s_m` of equation 7 of [SEGA
                Paper](https://arxiv.org/pdf/2301.12247.pdf).
            edit_mom_beta (`float`, *optional*, defaults to 0.4):
                Defines how semantic guidance momentum builds up. `edit_mom_beta` indicates how much of the previous
                momentum will be kept. Momentum is already built up during warmup, i.e. for diffusion steps smaller
                than `edit_warmup_steps`. `edit_mom_beta` is defined as `beta_m` (β) of equation 8 of [SEGA
                Paper](https://arxiv.org/pdf/2301.12247.pdf).
            edit_weights (`List[float]`, *optional*, defaults to `None`):
                Indicates how much each individual concept should influence the overall guidance. If no weights are
                provided all concepts are applied equally. `edit_mom_beta` is defined as `g_i` of equation 9 of [SEGA
                Paper](https://arxiv.org/pdf/2301.12247.pdf).
            sem_guidance (`List[torch.Tensor]`, *optional*):
                List of pre-generated guidance vectors to be applied at generation. Length of the list has to
                correspond to `num_inference_steps`.

        Returns:
            [`~pipelines.semantic_stable_diffusion.SemanticStableDiffusionPipelineOutput`] or `tuple`:
            [`~pipelines.semantic_stable_diffusion.SemanticStableDiffusionPipelineOutput`] if `return_dict` is True,
            otherwise a `tuple. When returning a tuple, the first element is a list with the generated images, and the
            second element is a list of `bool`s denoting whether the corresponding generated image likely represents
            "not-safe-for-work" (nsfw) content, according to the `safety_checker`.
        """
        # 0. Default height and width to unet
        height = height or self.unet.config.sample_size * self.vae_scale_factor
        width = width or self.unet.config.sample_size * self.vae_scale_factor

        # 1. Check inputs. Raise error if not correct
        self.check_inputs(prompt, height, width, callback_steps)

        # 2. Define call parameters
        batch_size = 1 if isinstance(prompt, str) else len(prompt)

        if editing_prompt:
            enable_edit_guidance = True
            if isinstance(editing_prompt, str):
                editing_prompt = [editing_prompt]
            enabled_editing_prompts = len(editing_prompt)
        elif editing_prompt_embeddings is not None:
            enable_edit_guidance = True
            enabled_editing_prompts = editing_prompt_embeddings.shape[0]
        else:
            enabled_editing_prompts = 0
            enable_edit_guidance = False

        # get prompt text embeddings
        text_inputs = self.tokenizer(
            prompt,
            padding="max_length",
            max_length=self.tokenizer.model_max_length,
            return_tensors="pt",
        )
        text_input_ids = text_inputs.input_ids

        if text_input_ids.shape[-1] > self.tokenizer.model_max_length:
            removed_text = self.tokenizer.batch_decode(text_input_ids[:, self.tokenizer.model_max_length :])
            logger.warning(
                "The following part of your input was truncated because CLIP can only handle sequences up to"
                f" {self.tokenizer.model_max_length} tokens: {removed_text}"
            )
            text_input_ids = text_input_ids[:, : self.tokenizer.model_max_length]
        text_embeddings = self.text_encoder(text_input_ids.to(self.device))[0]

        # duplicate text embeddings for each generation per prompt, using mps friendly method
        bs_embed, seq_len, _ = text_embeddings.shape
        text_embeddings = text_embeddings.repeat(1, num_images_per_prompt, 1)
        text_embeddings = text_embeddings.view(bs_embed * num_images_per_prompt, seq_len, -1)

        if enable_edit_guidance:
            # get safety text embeddings
            if editing_prompt_embeddings is None:
                edit_concepts_input = self.tokenizer(
                    [x for item in editing_prompt for x in repeat(item, batch_size)],
                    padding="max_length",
                    max_length=self.tokenizer.model_max_length,
                    return_tensors="pt",
                )

                edit_concepts_input_ids = edit_concepts_input.input_ids

                if edit_concepts_input_ids.shape[-1] > self.tokenizer.model_max_length:
                    removed_text = self.tokenizer.batch_decode(
                        edit_concepts_input_ids[:, self.tokenizer.model_max_length :]
                    )
                    logger.warning(
                        "The following part of your input was truncated because CLIP can only handle sequences up to"
                        f" {self.tokenizer.model_max_length} tokens: {removed_text}"
                    )
                    edit_concepts_input_ids = edit_concepts_input_ids[:, : self.tokenizer.model_max_length]
                edit_concepts = self.text_encoder(edit_concepts_input_ids.to(self.device))[0]
            else:
                edit_concepts = editing_prompt_embeddings.to(self.device).repeat(batch_size, 1, 1)

            # duplicate text embeddings for each generation per prompt, using mps friendly method
            bs_embed_edit, seq_len_edit, _ = edit_concepts.shape
            edit_concepts = edit_concepts.repeat(1, num_images_per_prompt, 1)
            edit_concepts = edit_concepts.view(bs_embed_edit * num_images_per_prompt, seq_len_edit, -1)

        # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
        # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
        # corresponds to doing no classifier free guidance.
        do_classifier_free_guidance = guidance_scale > 1.0
        # get unconditional embeddings for classifier free guidance

        if do_classifier_free_guidance:
            uncond_tokens: List[str]
            if negative_prompt is None:
                uncond_tokens = [""]
            elif type(prompt) is not type(negative_prompt):
                raise TypeError(
                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
                    f" {type(prompt)}."
                )
            elif isinstance(negative_prompt, str):
                uncond_tokens = [negative_prompt]
            elif batch_size != len(negative_prompt):
                raise ValueError(
                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
                    " the batch size of `prompt`."
                )
            else:
                uncond_tokens = negative_prompt

            max_length = text_input_ids.shape[-1]
            uncond_input = self.tokenizer(
                uncond_tokens,
                padding="max_length",
                max_length=max_length,
                truncation=True,
                return_tensors="pt",
            )
            uncond_embeddings = self.text_encoder(uncond_input.input_ids.to(self.device))[0]

            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
            seq_len = uncond_embeddings.shape[1]
            uncond_embeddings = uncond_embeddings.repeat(batch_size, num_images_per_prompt, 1)
            uncond_embeddings = uncond_embeddings.view(batch_size * num_images_per_prompt, seq_len, -1)

            # For classifier free guidance, we need to do two forward passes.
            # Here we concatenate the unconditional and text embeddings into a single batch
            # to avoid doing two forward passes
            if enable_edit_guidance:
                text_embeddings = torch.cat([uncond_embeddings, text_embeddings, edit_concepts])
            else:
                text_embeddings = torch.cat([uncond_embeddings, text_embeddings])
        # get the initial random noise unless the user supplied it

        # 4. Prepare timesteps
        self.scheduler.set_timesteps(num_inference_steps, device=self.device)
        timesteps = self.scheduler.timesteps
        if use_ddpm:
          t_to_idx = {int(v):k for k,v in enumerate(timesteps[-zs.shape[0]:])}
          timesteps = timesteps[-zs.shape[0]:] 

        # 5. Prepare latent variables
        num_channels_latents = self.unet.config.in_channels
        latents = self.prepare_latents(
            batch_size * num_images_per_prompt,
            num_channels_latents,
            height,
            width,
            text_embeddings.dtype,
            self.device,
            generator,
            latents,
        )

        # 6. Prepare extra step kwargs.
        extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)

        # Initialize edit_momentum to None
        edit_momentum = None

        self.uncond_estimates = None
        self.text_estimates = None
        self.edit_estimates = None
        self.sem_guidance = None

        for i, t in enumerate(self.progress_bar(timesteps)):
            # expand the latents if we are doing classifier free guidance
            latent_model_input = (
                torch.cat([latents] * (2 + enabled_editing_prompts)) if do_classifier_free_guidance else latents
            )
            latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)

            # predict the noise residual
            noise_pred = self.unet(latent_model_input, t, encoder_hidden_states=text_embeddings).sample

            # perform guidance
            if do_classifier_free_guidance:
                noise_pred_out = noise_pred.chunk(2 + enabled_editing_prompts)  # [b,4, 64, 64]
                noise_pred_uncond, noise_pred_text = noise_pred_out[0], noise_pred_out[1]
                noise_pred_edit_concepts = noise_pred_out[2:]

                # default text guidance
                noise_guidance = guidance_scale * (noise_pred_text - noise_pred_uncond)
                # noise_guidance = (noise_pred_text - noise_pred_edit_concepts[0])

                if self.uncond_estimates is None:
                    self.uncond_estimates = torch.zeros((num_inference_steps + 1, *noise_pred_uncond.shape))
                self.uncond_estimates[i] = noise_pred_uncond.detach().cpu()

                if self.text_estimates is None:
                    self.text_estimates = torch.zeros((num_inference_steps + 1, *noise_pred_text.shape))
                self.text_estimates[i] = noise_pred_text.detach().cpu()

                if self.edit_estimates is None and enable_edit_guidance:
                    self.edit_estimates = torch.zeros(
                        (num_inference_steps + 1, len(noise_pred_edit_concepts), *noise_pred_edit_concepts[0].shape)
                    )

                if self.sem_guidance is None:
                    self.sem_guidance = torch.zeros((num_inference_steps + 1, *noise_pred_text.shape))

                if edit_momentum is None:
                    edit_momentum = torch.zeros_like(noise_guidance)

                if enable_edit_guidance:
                    concept_weights = torch.zeros(
                        (len(noise_pred_edit_concepts), noise_guidance.shape[0]),
                        device=self.device,
                        dtype=noise_guidance.dtype,
                    )
                    noise_guidance_edit = torch.zeros(
                        (len(noise_pred_edit_concepts), *noise_guidance.shape),
                        device=self.device,
                        dtype=noise_guidance.dtype,
                    )
                    # noise_guidance_edit = torch.zeros_like(noise_guidance)
                    warmup_inds = []
                    for c, noise_pred_edit_concept in enumerate(noise_pred_edit_concepts):
                        self.edit_estimates[i, c] = noise_pred_edit_concept
                        if isinstance(edit_guidance_scale, list):
                            edit_guidance_scale_c = edit_guidance_scale[c]
                        else:
                            edit_guidance_scale_c = edit_guidance_scale

                        if isinstance(edit_threshold, list):
                            edit_threshold_c = edit_threshold[c]
                        else:
                            edit_threshold_c = edit_threshold
                        if isinstance(reverse_editing_direction, list):
                            reverse_editing_direction_c = reverse_editing_direction[c]
                        else:
                            reverse_editing_direction_c = reverse_editing_direction
                        if edit_weights:
                            edit_weight_c = edit_weights[c]
                        else:
                            edit_weight_c = 1.0
                        if isinstance(edit_warmup_steps, list):
                            edit_warmup_steps_c = edit_warmup_steps[c]
                        else:
                            edit_warmup_steps_c = edit_warmup_steps

                        if isinstance(edit_cooldown_steps, list):
                            edit_cooldown_steps_c = edit_cooldown_steps[c]
                        elif edit_cooldown_steps is None:
                            edit_cooldown_steps_c = i + 1
                        else:
                            edit_cooldown_steps_c = edit_cooldown_steps
                        if i >= edit_warmup_steps_c:
                            warmup_inds.append(c)
                        if i >= edit_cooldown_steps_c:
                            noise_guidance_edit[c, :, :, :, :] = torch.zeros_like(noise_pred_edit_concept)
                            continue

                        noise_guidance_edit_tmp = noise_pred_edit_concept - noise_pred_uncond
                        # tmp_weights = (noise_pred_text - noise_pred_edit_concept).sum(dim=(1, 2, 3))
                        tmp_weights = (noise_guidance - noise_pred_edit_concept).sum(dim=(1, 2, 3))

                        tmp_weights = torch.full_like(tmp_weights, edit_weight_c)  # * (1 / enabled_editing_prompts)
                        if reverse_editing_direction_c:
                            noise_guidance_edit_tmp = noise_guidance_edit_tmp * -1
                        concept_weights[c, :] = tmp_weights

                        noise_guidance_edit_tmp = noise_guidance_edit_tmp * edit_guidance_scale_c

                        # torch.quantile function expects float32
                        if noise_guidance_edit_tmp.dtype == torch.float32:
                            tmp = torch.quantile(
                                torch.abs(noise_guidance_edit_tmp).flatten(start_dim=2),
                                edit_threshold_c,
                                dim=2,
                                keepdim=False,
                            )
                        else:
                            tmp = torch.quantile(
                                torch.abs(noise_guidance_edit_tmp).flatten(start_dim=2).to(torch.float32),
                                edit_threshold_c,
                                dim=2,
                                keepdim=False,
                            ).to(noise_guidance_edit_tmp.dtype)

                        noise_guidance_edit_tmp = torch.where(
                            torch.abs(noise_guidance_edit_tmp) >= tmp[:, :, None, None],
                            noise_guidance_edit_tmp,
                            torch.zeros_like(noise_guidance_edit_tmp),
                        )
                        noise_guidance_edit[c, :, :, :, :] = noise_guidance_edit_tmp

                        # noise_guidance_edit = noise_guidance_edit + noise_guidance_edit_tmp

                    warmup_inds = torch.tensor(warmup_inds).to(self.device)
                    if len(noise_pred_edit_concepts) > warmup_inds.shape[0] > 0:
                        concept_weights = concept_weights.to("cpu")  # Offload to cpu
                        noise_guidance_edit = noise_guidance_edit.to("cpu")

                        concept_weights_tmp = torch.index_select(concept_weights.to(self.device), 0, warmup_inds)
                        concept_weights_tmp = torch.where(
                            concept_weights_tmp < 0, torch.zeros_like(concept_weights_tmp), concept_weights_tmp
                        )
                        concept_weights_tmp = concept_weights_tmp / concept_weights_tmp.sum(dim=0)
                        # concept_weights_tmp = torch.nan_to_num(concept_weights_tmp)

                        noise_guidance_edit_tmp = torch.index_select(
                            noise_guidance_edit.to(self.device), 0, warmup_inds
                        )
                        noise_guidance_edit_tmp = torch.einsum(
                            "cb,cbijk->bijk", concept_weights_tmp, noise_guidance_edit_tmp
                        )
                        noise_guidance_edit_tmp = noise_guidance_edit_tmp
                        noise_guidance = noise_guidance + noise_guidance_edit_tmp

                        self.sem_guidance[i] = noise_guidance_edit_tmp.detach().cpu()

                        del noise_guidance_edit_tmp
                        del concept_weights_tmp
                        concept_weights = concept_weights.to(self.device)
                        noise_guidance_edit = noise_guidance_edit.to(self.device)

                    concept_weights = torch.where(
                        concept_weights < 0, torch.zeros_like(concept_weights), concept_weights
                    )

                    concept_weights = torch.nan_to_num(concept_weights)

                    noise_guidance_edit = torch.einsum("cb,cbijk->bijk", concept_weights, noise_guidance_edit)

                    noise_guidance_edit = noise_guidance_edit + edit_momentum_scale * edit_momentum

                    edit_momentum = edit_mom_beta * edit_momentum + (1 - edit_mom_beta) * noise_guidance_edit

                    if warmup_inds.shape[0] == len(noise_pred_edit_concepts):
                        noise_guidance = noise_guidance + noise_guidance_edit
                        self.sem_guidance[i] = noise_guidance_edit.detach().cpu()

                if sem_guidance is not None:
                    edit_guidance = sem_guidance[i].to(self.device)
                    noise_guidance = noise_guidance + edit_guidance

                noise_pred = noise_pred_uncond + noise_guidance
            ## ddpm ###########################################################
            if use_ddpm:
              
              idx = t_to_idx[int(t)]
              z = zs[idx] if not zs is None else None

              # 1. get previous step value (=t-1)
              prev_timestep = t - self.scheduler.config.num_train_timesteps // self.scheduler.num_inference_steps
              # 2. compute alphas, betas
              alpha_prod_t = self.scheduler.alphas_cumprod[t]
              alpha_prod_t_prev = self.scheduler.alphas_cumprod[prev_timestep] if prev_timestep >= 0 else self.scheduler.final_alpha_cumprod
              beta_prod_t = 1 - alpha_prod_t
              
              # 3. compute predicted original sample from predicted noise also called
              # "predicted x_0" of formula (12) from https://arxiv.org/pdf/2010.02502.pdf
              pred_original_sample = (latents - beta_prod_t ** (0.5) * noise_pred) / alpha_prod_t ** (0.5)
              
              
              # 5. compute variance: "sigma_t(η)" -> see formula (16)
              # σ_t = sqrt((1 − α_t−1)/(1 − α_t)) * sqrt(1 − α_t/α_t−1)    
              # variance = self.scheduler._get_variance(timestep, prev_timestep)
              # variance = get_variance(model, t) #, prev_timestep)
              beta_prod_t_prev = 1 - alpha_prod_t_prev
              variance = (beta_prod_t_prev / beta_prod_t) * (1 - alpha_prod_t / alpha_prod_t_prev)
              
            
              
              std_dev_t = eta * variance ** (0.5)
              # Take care of asymetric reverse process (asyrp)
              noise_pred_direction = noise_pred
              
              # 6. compute "direction pointing to x_t" of formula (12) from https://arxiv.org/pdf/2010.02502.pdf
              # pred_sample_direction = (1 - alpha_prod_t_prev - std_dev_t**2) ** (0.5) * model_output_direction
              pred_sample_direction = (1 - alpha_prod_t_prev - eta * variance) ** (0.5) * noise_pred_direction
              
              # 7. compute x_t without "random noise" of formula (12) from https://arxiv.org/pdf/2010.02502.pdf
              prev_sample = alpha_prod_t_prev ** (0.5) * pred_original_sample + pred_sample_direction
              # 8. Add noice if eta > 0
              if eta > 0:
                  if z is None:
                      z = torch.randn(noise_pred.shape, device=self.device)
                  sigma_z =  eta * variance ** (0.5) * z
                  latents = prev_sample + sigma_z

            ## ddpm ##########################################################
                # compute the previous noisy sample x_t -> x_t-1
            else: #if not use_ddpm:
              latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).prev_sample

            # call the callback, if provided
            if callback is not None and i % callback_steps == 0:
                callback(i, t, latents)

        
        #  # 8. Post-processing
        # image = self.decode_latents(latents)

        # # 9. Run safety checker
        # image, has_nsfw_concept = self.run_safety_checker(image, self.device, text_embeddings.dtype)

        # # 10. Convert to PIL
        # if output_type == "pil":
        #     image = self.numpy_to_pil(image)

        # if not return_dict:
        #     return (image, has_nsfw_concept)

        # return StableDiffusionPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept)
        
        # 8. Post-processing
        if not output_type == "latent":
            image = self.vae.decode(latents / self.vae.config.scaling_factor, return_dict=False)[0]
            image, has_nsfw_concept = self.run_safety_checker(image, self.device, text_embeddings.dtype)
        else:
            image = latents
            has_nsfw_concept = None

        if has_nsfw_concept is None:
            do_denormalize = [True] * image.shape[0]
        else:
            do_denormalize = [not has_nsfw for has_nsfw in has_nsfw_concept]

        image = self.image_processor.postprocess(image, output_type=output_type, do_denormalize=do_denormalize)

        if not return_dict:
            return (image, has_nsfw_concept)

        return StableDiffusionPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept)