examples/community/pipeline_controlnet_xl_kolors.py

# Copyright 2025 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.


import inspect
from typing import Any, Callable, Dict, List, Optional, Tuple, Union

import numpy as np
import PIL.Image
import torch
import torch.nn.functional as F
from transformers import (
    CLIPImageProcessor,
    CLIPVisionModelWithProjection,
)

from diffusers.callbacks import MultiPipelineCallbacks, PipelineCallback
from diffusers.image_processor import PipelineImageInput, VaeImageProcessor
from diffusers.loaders import (
    FromSingleFileMixin,
    IPAdapterMixin,
    StableDiffusionXLLoraLoaderMixin,
    TextualInversionLoaderMixin,
)
from diffusers.models import (
    AutoencoderKL,
    ControlNetModel,
    ImageProjection,
    MultiControlNetModel,
    UNet2DConditionModel,
)
from diffusers.pipelines.kolors import ChatGLMModel, ChatGLMTokenizer
from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
from diffusers.pipelines.stable_diffusion_xl.pipeline_output import StableDiffusionXLPipelineOutput
from diffusers.schedulers import KarrasDiffusionSchedulers
from diffusers.utils import (
    deprecate,
    is_invisible_watermark_available,
    logging,
    replace_example_docstring,
)
from diffusers.utils.torch_utils import is_compiled_module, randn_tensor


if is_invisible_watermark_available():
    from diffusers.pipelines.stable_diffusion_xl.watermark import StableDiffusionXLWatermarker

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


EXAMPLE_DOC_STRING = """
    Examples:
        ```py
        >>> import torch
        >>> from diffusers import KolorsControlNetPipeline, ControlNetModel
        >>> from diffusers.utils import load_image
        >>> import numpy as np
        >>> import cv2
        >>> from PIL import Image

        >>> prompt = "aerial view, a futuristic research complex in a bright foggy jungle, hard lighting"
        >>> negative_prompt = "low quality, bad quality, sketches"

        >>> # download an image
        >>> image = load_image(
        ...     "https://hf.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd_controlnet/hf-logo.png"
        ... )

        >>> # initialize the models and pipeline
        >>> controlnet_conditioning_scale = 0.5  # recommended for good generalization
        >>> controlnet = ControlNetModel.from_pretrained(
        ...     "Kwai-Kolors/Kolors-ControlNet-Canny",
        ...     use_safetensors=True,
        ...     torch_dtype=torch.float16
        ... )

        >>> pipe = KolorsControlNetPipeline.from_pretrained(
        ...     "Kwai-Kolors/Kolors-diffusers",
        ...     controlnet=controlnet,
        ...     variant="fp16",
        ...     use_safetensors=True,
        ...     torch_dtype=torch.float16
        ... )
        >>> pipe.enable_model_cpu_offload()

        >>> # get canny image
        >>> image = np.array(image)
        >>> image = cv2.Canny(image, 100, 200)
        >>> image = image[:, :, None]
        >>> image = np.concatenate([image, image, image], axis=2)
        >>> canny_image = Image.fromarray(image)

        >>> # generate image
        >>> image = pipe(
        ...     prompt, controlnet_conditioning_scale=controlnet_conditioning_scale, image=canny_image
        ... ).images[0]
        ```
"""


# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
def retrieve_latents(
    encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
):
    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
        return encoder_output.latent_dist.sample(generator)
    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
        return encoder_output.latent_dist.mode()
    elif hasattr(encoder_output, "latents"):
        return encoder_output.latents
    else:
        raise AttributeError("Could not access latents of provided encoder_output")


class KolorsControlNetPipeline(
    DiffusionPipeline,
    StableDiffusionMixin,
    StableDiffusionXLLoraLoaderMixin,
    FromSingleFileMixin,
    IPAdapterMixin,
):
    r"""
    Pipeline for image-to-image generation using Kolors with ControlNet guidance.

    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.)

    The pipeline also inherits the following loading methods:
        - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.safetensors` files
        - [`~loaders.StableDiffusionXLLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
        - [`~loaders.StableDiffusionXLLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
        - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters

    Args:
        vae ([`AutoencoderKL`]):
            Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.
        text_encoder ([`ChatGLMModel`]):
            Frozen text-encoder. Kolors uses [ChatGLM3-6B](https://huggingface.co/THUDM/chatglm3-6b).
        tokenizer (`ChatGLMTokenizer`):
            Tokenizer of class
            [ChatGLMTokenizer](https://huggingface.co/THUDM/chatglm3-6b/blob/main/tokenization_chatglm.py).
        unet ([`UNet2DConditionModel`]): Conditional U-Net architecture to denoise the encoded image latents.
        controlnet ([`ControlNetModel`] or `List[ControlNetModel]`):
            Provides additional conditioning to the unet during the denoising process. If you set multiple ControlNets
            as a list, the outputs from each ControlNet are added together to create one combined additional
            conditioning.
        scheduler ([`SchedulerMixin`]):
            A scheduler to be used in combination with `unet` to denoise the encoded image latents. Can be one of
            [`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
        requires_aesthetics_score (`bool`, *optional*, defaults to `"False"`):
            Whether the `unet` requires an `aesthetic_score` condition to be passed during inference.
        force_zeros_for_empty_prompt (`bool`, *optional*, defaults to `"True"`):
            Whether the negative prompt embeddings shall be forced to always be set to 0. Also see the config of
            `Kwai-Kolors/Kolors-diffusers`.
        feature_extractor ([`~transformers.CLIPImageProcessor`]):
            A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
    """

    model_cpu_offload_seq = "text_encoder->image_encoder->unet->vae"

    _optional_components = [
        "tokenizer",
        "text_encoder",
        "feature_extractor",
        "image_encoder",
    ]
    _callback_tensor_inputs = [
        "latents",
        "prompt_embeds",
        "negative_prompt_embeds",
        "add_text_embeds",
        "add_time_ids",
        "negative_pooled_prompt_embeds",
        "negative_add_time_ids",
        "image",
    ]

    def __init__(
        self,
        vae: AutoencoderKL,
        text_encoder: ChatGLMModel,
        tokenizer: ChatGLMTokenizer,
        unet: UNet2DConditionModel,
        controlnet: Union[ControlNetModel, List[ControlNetModel], Tuple[ControlNetModel], MultiControlNetModel],
        scheduler: KarrasDiffusionSchedulers,
        requires_aesthetics_score: bool = False,
        force_zeros_for_empty_prompt: bool = True,
        feature_extractor: CLIPImageProcessor = None,
        image_encoder: CLIPVisionModelWithProjection = None,
        add_watermarker: Optional[bool] = None,
    ):
        super().__init__()

        if isinstance(controlnet, (list, tuple)):
            controlnet = MultiControlNetModel(controlnet)

        self.register_modules(
            vae=vae,
            text_encoder=text_encoder,
            tokenizer=tokenizer,
            unet=unet,
            controlnet=controlnet,
            scheduler=scheduler,
            feature_extractor=feature_extractor,
            image_encoder=image_encoder,
        )
        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
        self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True)
        self.control_image_processor = VaeImageProcessor(
            vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True, do_normalize=False
        )

        if add_watermarker:
            self.watermark = StableDiffusionXLWatermarker()
        else:
            self.watermark = None

        self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
        self.register_to_config(requires_aesthetics_score=requires_aesthetics_score)

    def encode_prompt(
        self,
        prompt,
        device: Optional[torch.device] = None,
        num_images_per_prompt: int = 1,
        do_classifier_free_guidance: bool = True,
        negative_prompt=None,
        prompt_embeds: Optional[torch.FloatTensor] = None,
        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
        pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
        negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
        lora_scale: Optional[float] = None,
    ):
        r"""
        Encodes the prompt into text encoder hidden states.

        Args:
             prompt (`str` or `List[str]`, *optional*):
                prompt to be encoded
            device: (`torch.device`):
                torch device
            num_images_per_prompt (`int`):
                number of images that should be generated per prompt
            do_classifier_free_guidance (`bool`):
                whether to use classifier free guidance or not
            negative_prompt (`str` or `List[str]`, *optional*):
                The prompt or prompts not to guide the image generation. If not defined, one has to pass
                `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
                less than `1`).
            prompt_embeds (`torch.FloatTensor`, *optional*):
                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
                provided, text embeddings will be generated from `prompt` input argument.
            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
                argument.
            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
                If not provided, pooled text embeddings will be generated from `prompt` input argument.
            negative_pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
                Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
                input argument.
            lora_scale (`float`, *optional*):
                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
        """
        device = device or self._execution_device

        # set lora scale so that monkey patched LoRA
        # function of text encoder can correctly access it
        if lora_scale is not None and isinstance(self, StableDiffusionXLLoraLoaderMixin):
            self._lora_scale = lora_scale

        if prompt is not None and isinstance(prompt, str):
            batch_size = 1
        elif prompt is not None and isinstance(prompt, list):
            batch_size = len(prompt)
        else:
            batch_size = prompt_embeds.shape[0]

        # Define tokenizers and text encoders
        tokenizers = [self.tokenizer]
        text_encoders = [self.text_encoder]

        if prompt_embeds is None:
            # textual inversion: procecss multi-vector tokens if necessary
            prompt_embeds_list = []
            for tokenizer, text_encoder in zip(tokenizers, text_encoders):
                if isinstance(self, TextualInversionLoaderMixin):
                    prompt = self.maybe_convert_prompt(prompt, tokenizer)

                text_inputs = tokenizer(
                    prompt,
                    padding="max_length",
                    max_length=256,
                    truncation=True,
                    return_tensors="pt",
                ).to(self._execution_device)
                output = text_encoder(
                    input_ids=text_inputs["input_ids"],
                    attention_mask=text_inputs["attention_mask"],
                    position_ids=text_inputs["position_ids"],
                    output_hidden_states=True,
                )
                prompt_embeds = output.hidden_states[-2].permute(1, 0, 2).clone()
                pooled_prompt_embeds = output.hidden_states[-1][-1, :, :].clone()  # [batch_size, 4096]
                bs_embed, seq_len, _ = prompt_embeds.shape
                prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
                prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)

                prompt_embeds_list.append(prompt_embeds)

            prompt_embeds = prompt_embeds_list[0]

        # get unconditional embeddings for classifier free guidance
        zero_out_negative_prompt = negative_prompt is None and self.config.force_zeros_for_empty_prompt
        if do_classifier_free_guidance and negative_prompt_embeds is None and zero_out_negative_prompt:
            negative_prompt_embeds = torch.zeros_like(prompt_embeds)
            negative_pooled_prompt_embeds = torch.zeros_like(pooled_prompt_embeds)
        elif do_classifier_free_guidance and negative_prompt_embeds is None:
            uncond_tokens: List[str]
            if negative_prompt is None:
                uncond_tokens = [""] * batch_size
            elif prompt is not None and 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

            negative_prompt_embeds_list = []
            for tokenizer, text_encoder in zip(tokenizers, text_encoders):
                # textual inversion: procecss multi-vector tokens if necessary
                if isinstance(self, TextualInversionLoaderMixin):
                    uncond_tokens = self.maybe_convert_prompt(uncond_tokens, tokenizer)

                max_length = prompt_embeds.shape[1]
                uncond_input = tokenizer(
                    uncond_tokens,
                    padding="max_length",
                    max_length=max_length,
                    truncation=True,
                    return_tensors="pt",
                ).to(self._execution_device)
                output = text_encoder(
                    input_ids=uncond_input["input_ids"],
                    attention_mask=uncond_input["attention_mask"],
                    position_ids=uncond_input["position_ids"],
                    output_hidden_states=True,
                )
                negative_prompt_embeds = output.hidden_states[-2].permute(1, 0, 2).clone()
                negative_pooled_prompt_embeds = output.hidden_states[-1][-1, :, :].clone()  # [batch_size, 4096]

                if do_classifier_free_guidance:
                    # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
                    seq_len = negative_prompt_embeds.shape[1]

                    negative_prompt_embeds = negative_prompt_embeds.to(dtype=text_encoder.dtype, device=device)

                    negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
                    negative_prompt_embeds = negative_prompt_embeds.view(
                        batch_size * num_images_per_prompt, seq_len, -1
                    )

                negative_prompt_embeds_list.append(negative_prompt_embeds)

            negative_prompt_embeds = negative_prompt_embeds_list[0]

        bs_embed = pooled_prompt_embeds.shape[0]
        pooled_prompt_embeds = pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
            bs_embed * num_images_per_prompt, -1
        )
        if do_classifier_free_guidance:
            negative_pooled_prompt_embeds = negative_pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
                bs_embed * num_images_per_prompt, -1
            )

        return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds

    def prepare_ip_adapter_image_embeds(
        self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt, do_classifier_free_guidance
    ):
        image_embeds = []
        if do_classifier_free_guidance:
            negative_image_embeds = []
        if ip_adapter_image_embeds is None:
            if not isinstance(ip_adapter_image, list):
                ip_adapter_image = [ip_adapter_image]

            if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):
                raise ValueError(
                    f"`ip_adapter_image` must have same length as the number of IP Adapters. "
                    f"Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
                )

            for single_ip_adapter_image, image_proj_layer in zip(
                ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
            ):
                output_hidden_state = not isinstance(image_proj_layer, ImageProjection)
                single_image_embeds, single_negative_image_embeds = self.encode_image(
                    single_ip_adapter_image, device, 1, output_hidden_state
                )

                image_embeds.append(single_image_embeds[None, :])
                if do_classifier_free_guidance:
                    negative_image_embeds.append(single_negative_image_embeds[None, :])
        else:
            for single_image_embeds in ip_adapter_image_embeds:
                if do_classifier_free_guidance:
                    single_negative_image_embeds, single_image_embeds = single_image_embeds.chunk(2)

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_image
    def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):
        dtype = next(self.image_encoder.parameters()).dtype

        if not isinstance(image, torch.Tensor):
            image = self.feature_extractor(image, return_tensors="pt").pixel_values

        image = image.to(device=device, dtype=dtype)
        if output_hidden_states:
            image_enc_hidden_states = self.image_encoder(image, output_hidden_states=True).hidden_states[-2]
            image_enc_hidden_states = image_enc_hidden_states.repeat_interleave(num_images_per_prompt, dim=0)
            uncond_image_enc_hidden_states = self.image_encoder(
                torch.zeros_like(image), output_hidden_states=True
            ).hidden_states[-2]
            uncond_image_enc_hidden_states = uncond_image_enc_hidden_states.repeat_interleave(
                num_images_per_prompt, dim=0
            )
            return image_enc_hidden_states, uncond_image_enc_hidden_states
        else:
            image_embeds = self.image_encoder(image).image_embeds
            image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)
            uncond_image_embeds = torch.zeros_like(image_embeds)

            return image_embeds, uncond_image_embeds

    # 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://huggingface.co/papers/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

    def check_inputs(
        self,
        prompt,
        image,
        num_inference_steps,
        callback_steps,
        negative_prompt=None,
        prompt_embeds=None,
        negative_prompt_embeds=None,
        pooled_prompt_embeds=None,
        negative_pooled_prompt_embeds=None,
        ip_adapter_image=None,
        ip_adapter_image_embeds=None,
        controlnet_conditioning_scale=1.0,
        control_guidance_start=0.0,
        control_guidance_end=1.0,
        callback_on_step_end_tensor_inputs=None,
    ):
        if num_inference_steps is None:
            raise ValueError("`num_inference_steps` cannot be None.")
        elif not isinstance(num_inference_steps, int) or num_inference_steps <= 0:
            raise ValueError(
                f"`num_inference_steps` has to be a positive integer but is {num_inference_steps} of type"
                f" {type(num_inference_steps)}."
            )

        if 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 callback_on_step_end_tensor_inputs is not None and not all(
            k in self._callback_tensor_inputs for k in callback_on_step_end_tensor_inputs
        ):
            raise ValueError(
                f"`callback_on_step_end_tensor_inputs` has to be in {self._callback_tensor_inputs}, but found {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}"
            )

        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}."
                )

        if prompt_embeds is not None and pooled_prompt_embeds is None:
            raise ValueError(
                "If `prompt_embeds` are provided, `pooled_prompt_embeds` also have to be passed. Make sure to generate `pooled_prompt_embeds` from the same text encoder that was used to generate `prompt_embeds`."
            )

        if negative_prompt_embeds is not None and negative_pooled_prompt_embeds is None:
            raise ValueError(
                "If `negative_prompt_embeds` are provided, `negative_pooled_prompt_embeds` also have to be passed. Make sure to generate `negative_pooled_prompt_embeds` from the same text encoder that was used to generate `negative_prompt_embeds`."
            )

        # `prompt` needs more sophisticated handling when there are multiple
        # conditionings.
        if isinstance(self.controlnet, MultiControlNetModel):
            if isinstance(prompt, list):
                logger.warning(
                    f"You have {len(self.controlnet.nets)} ControlNets and you have passed {len(prompt)}"
                    " prompts. The conditionings will be fixed across the prompts."
                )

        is_compiled = hasattr(F, "scaled_dot_product_attention") and isinstance(
            self.controlnet, torch._dynamo.eval_frame.OptimizedModule
        )

        # Check `controlnet_conditioning_scale`
        if (
            isinstance(self.controlnet, ControlNetModel)
            or is_compiled
            and isinstance(self.controlnet._orig_mod, ControlNetModel)
        ):
            if not isinstance(controlnet_conditioning_scale, float):
                raise TypeError("For single controlnet: `controlnet_conditioning_scale` must be type `float`.")
        elif (
            isinstance(self.controlnet, MultiControlNetModel)
            or is_compiled
            and isinstance(self.controlnet._orig_mod, MultiControlNetModel)
        ):
            if isinstance(controlnet_conditioning_scale, list):
                if any(isinstance(i, list) for i in controlnet_conditioning_scale):
                    raise ValueError("A single batch of multiple conditionings are supported at the moment.")
            elif isinstance(controlnet_conditioning_scale, list) and len(controlnet_conditioning_scale) != len(
                self.controlnet.nets
            ):
                raise ValueError(
                    "For multiple controlnets: When `controlnet_conditioning_scale` is specified as `list`, it must have"
                    " the same length as the number of controlnets"
                )
        else:
            assert False

        if not isinstance(control_guidance_start, (tuple, list)):
            control_guidance_start = [control_guidance_start]

        if not isinstance(control_guidance_end, (tuple, list)):
            control_guidance_end = [control_guidance_end]

        if len(control_guidance_start) != len(control_guidance_end):
            raise ValueError(
                f"`control_guidance_start` has {len(control_guidance_start)} elements, but `control_guidance_end` has {len(control_guidance_end)} elements. Make sure to provide the same number of elements to each list."
            )

        if isinstance(self.controlnet, MultiControlNetModel):
            if len(control_guidance_start) != len(self.controlnet.nets):
                raise ValueError(
                    f"`control_guidance_start`: {control_guidance_start} has {len(control_guidance_start)} elements but there are {len(self.controlnet.nets)} controlnets available. Make sure to provide {len(self.controlnet.nets)}."
                )

        for start, end in zip(control_guidance_start, control_guidance_end):
            if start >= end:
                raise ValueError(
                    f"control guidance start: {start} cannot be larger or equal to control guidance end: {end}."
                )
            if start < 0.0:
                raise ValueError(f"control guidance start: {start} can't be smaller than 0.")
            if end > 1.0:
                raise ValueError(f"control guidance end: {end} can't be larger than 1.0.")

        if ip_adapter_image is not None and ip_adapter_image_embeds is not None:
            raise ValueError(
                "Provide either `ip_adapter_image` or `ip_adapter_image_embeds`. Cannot leave both `ip_adapter_image` and `ip_adapter_image_embeds` defined."
            )

        if ip_adapter_image_embeds is not None:
            if not isinstance(ip_adapter_image_embeds, list):
                raise ValueError(
                    f"`ip_adapter_image_embeds` has to be of type `list` but is {type(ip_adapter_image_embeds)}"
                )
            elif ip_adapter_image_embeds[0].ndim not in [3, 4]:
                raise ValueError(
                    f"`ip_adapter_image_embeds` has to be a list of 3D or 4D tensors but is {ip_adapter_image_embeds[0].ndim}D"
                )

    # Copied from diffusers.pipelines.controlnet.pipeline_controlnet_sd_xl.StableDiffusionXLControlNetPipeline.check_image
    def check_image(self, image, prompt, prompt_embeds):
        image_is_pil = isinstance(image, PIL.Image.Image)
        image_is_tensor = isinstance(image, torch.Tensor)
        image_is_np = isinstance(image, np.ndarray)
        image_is_pil_list = isinstance(image, list) and isinstance(image[0], PIL.Image.Image)
        image_is_tensor_list = isinstance(image, list) and isinstance(image[0], torch.Tensor)
        image_is_np_list = isinstance(image, list) and isinstance(image[0], np.ndarray)

        if (
            not image_is_pil
            and not image_is_tensor
            and not image_is_np
            and not image_is_pil_list
            and not image_is_tensor_list
            and not image_is_np_list
        ):
            raise TypeError(
                f"image must be passed and be one of PIL image, numpy array, torch tensor, list of PIL images, list of numpy arrays or list of torch tensors, but is {type(image)}"
            )

        if image_is_pil:
            image_batch_size = 1
        else:
            image_batch_size = len(image)

        if prompt is not None and isinstance(prompt, str):
            prompt_batch_size = 1
        elif prompt is not None and isinstance(prompt, list):
            prompt_batch_size = len(prompt)
        elif prompt_embeds is not None:
            prompt_batch_size = prompt_embeds.shape[0]

        if image_batch_size != 1 and image_batch_size != prompt_batch_size:
            raise ValueError(
                f"If image batch size is not 1, image batch size must be same as prompt batch size. image batch size: {image_batch_size}, prompt batch size: {prompt_batch_size}"
            )

    # Copied from diffusers.pipelines.controlnet.pipeline_controlnet_sd_xl.StableDiffusionXLControlNetPipeline.prepare_image
    def prepare_control_image(
        self,
        image,
        width,
        height,
        batch_size,
        num_images_per_prompt,
        device,
        dtype,
        do_classifier_free_guidance=False,
        guess_mode=False,
    ):
        image = self.control_image_processor.preprocess(image, height=height, width=width).to(dtype=torch.float32)
        image_batch_size = image.shape[0]

        if image_batch_size == 1:
            repeat_by = batch_size
        else:
            # image batch size is the same as prompt batch size
            repeat_by = num_images_per_prompt

        image = image.repeat_interleave(repeat_by, dim=0)

        image = image.to(device=device, dtype=dtype)

        if do_classifier_free_guidance and not guess_mode:
            image = torch.cat([image] * 2)

        return image

    # 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

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
    def prepare_latents_t2i(
        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

    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline._get_add_time_ids
    def _get_add_time_ids(
        self, original_size, crops_coords_top_left, target_size, dtype, text_encoder_projection_dim=None
    ):
        add_time_ids = list(original_size + crops_coords_top_left + target_size)

        passed_add_embed_dim = (
            self.unet.config.addition_time_embed_dim * len(add_time_ids) + text_encoder_projection_dim
        )
        expected_add_embed_dim = self.unet.add_embedding.linear_1.in_features

        if expected_add_embed_dim != passed_add_embed_dim:
            raise ValueError(
                f"Model expects an added time embedding vector of length {expected_add_embed_dim}, but a vector of {passed_add_embed_dim} was created. The model has an incorrect config. Please check `unet.config.time_embedding_type` and `text_encoder_2.config.projection_dim`."
            )

        add_time_ids = torch.tensor([add_time_ids], dtype=dtype)
        return add_time_ids

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_upscale.StableDiffusionUpscalePipeline.upcast_vae
    def upcast_vae(self):
        deprecate("upcast_vae", "1.0.0", "`upcast_vae` is deprecated. Please use `pipe.vae.to(torch.float32)`")
        self.vae.to(dtype=torch.float32)

    @property
    def guidance_scale(self):
        return self._guidance_scale

    # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
    # of the Imagen paper: https://huggingface.co/papers/2205.11487 . `guidance_scale = 1`
    # corresponds to doing no classifier free guidance.
    @property
    def do_classifier_free_guidance(self):
        return self._guidance_scale > 1

    @property
    def cross_attention_kwargs(self):
        return self._cross_attention_kwargs

    @property
    def num_timesteps(self):
        return self._num_timesteps

    @torch.no_grad()
    @replace_example_docstring(EXAMPLE_DOC_STRING)
    def __call__(
        self,
        prompt: Union[str, List[str]] = None,
        image: PipelineImageInput = None,
        height: Optional[int] = None,
        width: Optional[int] = None,
        num_inference_steps: int = 50,
        guidance_scale: float = 5.0,
        negative_prompt: Optional[Union[str, List[str]]] = None,
        num_images_per_prompt: Optional[int] = 1,
        eta: float = 0.0,
        guess_mode: bool = False,
        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
        latents: Optional[torch.Tensor] = None,
        prompt_embeds: Optional[torch.Tensor] = None,
        negative_prompt_embeds: Optional[torch.Tensor] = None,
        pooled_prompt_embeds: Optional[torch.Tensor] = None,
        negative_pooled_prompt_embeds: Optional[torch.Tensor] = None,
        ip_adapter_image: Optional[PipelineImageInput] = None,
        ip_adapter_image_embeds: Optional[List[torch.Tensor]] = None,
        output_type: Optional[str] = "pil",
        return_dict: bool = True,
        cross_attention_kwargs: Optional[Dict[str, Any]] = None,
        controlnet_conditioning_scale: Union[float, List[float]] = 0.8,
        control_guidance_start: Union[float, List[float]] = 0.0,
        control_guidance_end: Union[float, List[float]] = 1.0,
        original_size: Tuple[int, int] = None,
        crops_coords_top_left: Tuple[int, int] = (0, 0),
        target_size: Tuple[int, int] = None,
        negative_original_size: Optional[Tuple[int, int]] = None,
        negative_crops_coords_top_left: Tuple[int, int] = (0, 0),
        negative_target_size: Optional[Tuple[int, int]] = None,
        callback_on_step_end: Optional[
            Union[Callable[[int, int, Dict], None], PipelineCallback, MultiPipelineCallbacks]
        ] = None,
        callback_on_step_end_tensor_inputs: List[str] = ["latents"],
        **kwargs,
    ):
        r"""
        Function invoked when calling the pipeline for generation.

        Args:
            prompt (`str` or `List[str]`, *optional*):
                The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
                instead.
            image (`torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.Tensor]`, `List[PIL.Image.Image]`, `List[np.ndarray]`,:
                    `List[List[torch.Tensor]]`, `List[List[np.ndarray]]` or `List[List[PIL.Image.Image]]`):
                The ControlNet input condition. ControlNet uses this input condition to generate guidance to Unet. If
                the type is specified as `torch.Tensor`, it is passed to ControlNet as is. `PIL.Image.Image` can also
                be accepted as an image. The dimensions of the output image defaults to `image`'s dimensions. If height
                and/or width are passed, `image` is resized according to them. If multiple ControlNets are specified in
                init, images must be passed as a list such that each element of the list can be correctly batched for
                input to a single controlnet.
            height (`int`, *optional*, defaults to the size of image):
                The height in pixels of the generated image. Anything below 512 pixels won't work well for
                [stabilityai/stable-diffusion-xl-base-1.0](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0)
                and checkpoints that are not specifically fine-tuned on low resolutions.
            width (`int`, *optional*, defaults to the size of image):
                The width in pixels of the generated image. Anything below 512 pixels won't work well for
                [stabilityai/stable-diffusion-xl-base-1.0](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0)
                and checkpoints that are not specifically fine-tuned on low resolutions.
            strength (`float`, *optional*, defaults to 0.8):
                Indicates extent to transform the reference `image`. Must be between 0 and 1. `image` is used as a
                starting point and more noise is added the higher the `strength`. The number of denoising steps depends
                on the amount of noise initially added. When `strength` is 1, added noise is maximum and the denoising
                process runs for the full number of iterations specified in `num_inference_steps`. A value of 1
                essentially ignores `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://huggingface.co/papers/2207.12598).
                `guidance_scale` is defined as `w` of equation 2. of [Imagen
                Paper](https://huggingface.co/papers/2205.11487). 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. If not defined, one has to pass
                `negative_prompt_embeds` instead. 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://huggingface.co/papers/2010.02502. Only applies to
                [`schedulers.DDIMScheduler`], will be ignored for others.
            generator (`torch.Generator` or `List[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.Tensor`, *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 be generated by sampling using the supplied random `generator`.
            prompt_embeds (`torch.Tensor`, *optional*):
                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
                provided, text embeddings will be generated from `prompt` input argument.
            negative_prompt_embeds (`torch.Tensor`, *optional*):
                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
                argument.
            pooled_prompt_embeds (`torch.Tensor`, *optional*):
                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
                If not provided, pooled text embeddings will be generated from `prompt` input argument.
            negative_pooled_prompt_embeds (`torch.Tensor`, *optional*):
                Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
                input argument.
            ip_adapter_image: (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
            ip_adapter_image_embeds (`List[torch.Tensor]`, *optional*):
                Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of
                IP-adapters. Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. It should
                contain the negative image embedding if `do_classifier_free_guidance` is set to `True`. If not
                provided, embeddings are computed from the `ip_adapter_image` input argument.
            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.
            cross_attention_kwargs (`dict`, *optional*):
                A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
                `self.processor` in
                [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
            controlnet_conditioning_scale (`float` or `List[float]`, *optional*, defaults to 1.0):
                The outputs of the controlnet are multiplied by `controlnet_conditioning_scale` before they are added
                to the residual in the original unet. If multiple ControlNets are specified in init, you can set the
                corresponding scale as a list.
            control_guidance_start (`float` or `List[float]`, *optional*, defaults to 0.0):
                The percentage of total steps at which the controlnet starts applying.
            control_guidance_end (`float` or `List[float]`, *optional*, defaults to 1.0):
                The percentage of total steps at which the controlnet stops applying.
            original_size (`Tuple[int]`, *optional*, defaults to (1024, 1024)):
                If `original_size` is not the same as `target_size` the image will appear to be down- or upsampled.
                `original_size` defaults to `(height, width)` if not specified. Part of SDXL's micro-conditioning as
                explained in section 2.2 of
                [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952).
            crops_coords_top_left (`Tuple[int]`, *optional*, defaults to (0, 0)):
                `crops_coords_top_left` can be used to generate an image that appears to be "cropped" from the position
                `crops_coords_top_left` downwards. Favorable, well-centered images are usually achieved by setting
                `crops_coords_top_left` to (0, 0). Part of SDXL's micro-conditioning as explained in section 2.2 of
                [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952).
            target_size (`Tuple[int]`, *optional*, defaults to (1024, 1024)):
                For most cases, `target_size` should be set to the desired height and width of the generated image. If
                not specified it will default to `(height, width)`. Part of SDXL's micro-conditioning as explained in
                section 2.2 of [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952).
            negative_original_size (`Tuple[int]`, *optional*, defaults to (1024, 1024)):
                To negatively condition the generation process based on a specific image resolution. Part of SDXL's
                micro-conditioning as explained in section 2.2 of
                [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952). For more
                information, refer to this issue thread: https://github.com/huggingface/diffusers/issues/4208.
            negative_crops_coords_top_left (`Tuple[int]`, *optional*, defaults to (0, 0)):
                To negatively condition the generation process based on a specific crop coordinates. Part of SDXL's
                micro-conditioning as explained in section 2.2 of
                [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952). For more
                information, refer to this issue thread: https://github.com/huggingface/diffusers/issues/4208.
            negative_target_size (`Tuple[int]`, *optional*, defaults to (1024, 1024)):
                To negatively condition the generation process based on a target image resolution. It should be as same
                as the `target_size` for most cases. Part of SDXL's micro-conditioning as explained in section 2.2 of
                [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952). For more
                information, refer to this issue thread: https://github.com/huggingface/diffusers/issues/4208.
            callback_on_step_end (`Callable`, `PipelineCallback`, `MultiPipelineCallbacks`, *optional*):
                A function or a subclass of `PipelineCallback` or `MultiPipelineCallbacks` that is called at the end of
                each denoising step during the inference. with the following arguments: `callback_on_step_end(self:
                DiffusionPipeline, step: int, timestep: int, callback_kwargs: Dict)`. `callback_kwargs` will include a
                list of all tensors as specified by `callback_on_step_end_tensor_inputs`.
            callback_on_step_end_tensor_inputs (`List`, *optional*):
                The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list
                will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the
                `._callback_tensor_inputs` attribute of your pipeline class.

        Examples:

        Returns:
            [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`:
            [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] if `return_dict` is True, otherwise a `tuple`
            containing the output images.
        """

        callback = kwargs.pop("callback", None)
        callback_steps = kwargs.pop("callback_steps", None)

        if callback is not None:
            deprecate(
                "callback",
                "1.0.0",
                "Passing `callback` as an input argument to `__call__` is deprecated, consider using `callback_on_step_end`",
            )
        if callback_steps is not None:
            deprecate(
                "callback_steps",
                "1.0.0",
                "Passing `callback_steps` as an input argument to `__call__` is deprecated, consider using `callback_on_step_end`",
            )

        if isinstance(callback_on_step_end, (PipelineCallback, MultiPipelineCallbacks)):
            callback_on_step_end_tensor_inputs = callback_on_step_end.tensor_inputs

        controlnet = self.controlnet._orig_mod if is_compiled_module(self.controlnet) else self.controlnet

        # align format for control guidance
        if not isinstance(control_guidance_start, list) and isinstance(control_guidance_end, list):
            control_guidance_start = len(control_guidance_end) * [control_guidance_start]
        elif not isinstance(control_guidance_end, list) and isinstance(control_guidance_start, list):
            control_guidance_end = len(control_guidance_start) * [control_guidance_end]
        elif not isinstance(control_guidance_start, list) and not isinstance(control_guidance_end, list):
            mult = len(controlnet.nets) if isinstance(controlnet, MultiControlNetModel) else 1
            control_guidance_start, control_guidance_end = (
                mult * [control_guidance_start],
                mult * [control_guidance_end],
            )

        # from IPython import embed; embed()
        # 1. Check inputs. Raise error if not correct
        self.check_inputs(
            prompt,
            image,
            num_inference_steps,
            callback_steps,
            negative_prompt,
            prompt_embeds,
            negative_prompt_embeds,
            pooled_prompt_embeds,
            negative_pooled_prompt_embeds,
            ip_adapter_image,
            ip_adapter_image_embeds,
            controlnet_conditioning_scale,
            control_guidance_start,
            control_guidance_end,
            callback_on_step_end_tensor_inputs,
        )

        self._guidance_scale = guidance_scale
        self._cross_attention_kwargs = cross_attention_kwargs

        # 2. Define call parameters
        if prompt is not None and isinstance(prompt, str):
            batch_size = 1
        elif prompt is not None and isinstance(prompt, list):
            batch_size = len(prompt)
        else:
            batch_size = prompt_embeds.shape[0]

        device = self._execution_device

        if isinstance(controlnet, MultiControlNetModel) and isinstance(controlnet_conditioning_scale, float):
            controlnet_conditioning_scale = [controlnet_conditioning_scale] * len(controlnet.nets)

        # 3.1. Encode input prompt
        text_encoder_lora_scale = (
            self.cross_attention_kwargs.get("scale", None) if self.cross_attention_kwargs is not None else None
        )
        (
            prompt_embeds,
            negative_prompt_embeds,
            pooled_prompt_embeds,
            negative_pooled_prompt_embeds,
        ) = self.encode_prompt(
            prompt,
            device,
            num_images_per_prompt,
            self.do_classifier_free_guidance,
            negative_prompt,
            prompt_embeds=prompt_embeds,
            negative_prompt_embeds=negative_prompt_embeds,
            pooled_prompt_embeds=pooled_prompt_embeds,
            negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
            lora_scale=text_encoder_lora_scale,
        )

        # 3.2 Encode ip_adapter_image
        if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
            image_embeds = self.prepare_ip_adapter_image_embeds(
                ip_adapter_image,
                ip_adapter_image_embeds,
                device,
                batch_size * num_images_per_prompt,
                self.do_classifier_free_guidance,
            )

        if isinstance(controlnet, ControlNetModel):
            image = self.prepare_control_image(
                image=image,
                width=width,
                height=height,
                batch_size=batch_size * num_images_per_prompt,
                num_images_per_prompt=num_images_per_prompt,
                device=device,
                dtype=controlnet.dtype,
                do_classifier_free_guidance=self.do_classifier_free_guidance,
                guess_mode=guess_mode,
            )
            height, width = image.shape[-2:]
        elif isinstance(controlnet, MultiControlNetModel):
            control_images = []

            for control_image_ in image:
                control_image_ = self.prepare_control_image(
                    image=control_image_,
                    width=width,
                    height=height,
                    batch_size=batch_size * num_images_per_prompt,
                    num_images_per_prompt=num_images_per_prompt,
                    device=device,
                    dtype=controlnet.dtype,
                    do_classifier_free_guidance=self.do_classifier_free_guidance,
                    guess_mode=guess_mode,
                )

                control_images.append(control_image_)

            image = control_images
            height, width = image[0].shape[-2:]
        else:
            assert False

        # 4. Prepare timesteps
        self.scheduler.set_timesteps(num_inference_steps, device=device)

        timesteps = self.scheduler.timesteps

        # 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,
            prompt_embeds.dtype,
            device,
            generator,
            latents,
        )

        # 6.5 Optionally get Guidance Scale Embedding
        timestep_cond = None
        if self.unet.config.time_cond_proj_dim is not None:
            guidance_scale_tensor = torch.tensor(self.guidance_scale - 1).repeat(batch_size * num_images_per_prompt)
            timestep_cond = self.get_guidance_scale_embedding(
                guidance_scale_tensor, embedding_dim=self.unet.config.time_cond_proj_dim
            ).to(device=device, dtype=latents.dtype)

        # 7. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline
        extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)

        # 7.1 Create tensor stating which controlnets to keep
        controlnet_keep = []
        for i in range(len(timesteps)):
            keeps = [
                1.0 - float(i / len(timesteps) < s or (i + 1) / len(timesteps) > e)
                for s, e in zip(control_guidance_start, control_guidance_end)
            ]
            controlnet_keep.append(keeps[0] if isinstance(controlnet, ControlNetModel) else keeps)

        # 7.2 Prepare added time ids & embeddings
        if isinstance(image, list):
            original_size = original_size or image[0].shape[-2:]
        else:
            original_size = original_size or image.shape[-2:]
        target_size = target_size or (height, width)

        # 7. Prepare added time ids & embeddings
        text_encoder_projection_dim = int(pooled_prompt_embeds.shape[-1])

        add_text_embeds = pooled_prompt_embeds
        add_time_ids = self._get_add_time_ids(
            original_size,
            crops_coords_top_left,
            target_size,
            dtype=prompt_embeds.dtype,
            text_encoder_projection_dim=text_encoder_projection_dim,
        )

        if negative_original_size is not None and negative_target_size is not None:
            negative_add_time_ids = self._get_add_time_ids(
                negative_original_size,
                negative_crops_coords_top_left,
                negative_target_size,
                dtype=prompt_embeds.dtype,
                text_encoder_projection_dim=text_encoder_projection_dim,
            )
        else:
            negative_add_time_ids = add_time_ids

        if self.do_classifier_free_guidance:
            prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds], dim=0)
            add_text_embeds = torch.cat([negative_pooled_prompt_embeds, add_text_embeds], dim=0)
            add_time_ids = torch.cat([add_time_ids, add_time_ids], dim=0)

        prompt_embeds = prompt_embeds.to(device)
        add_text_embeds = add_text_embeds.to(device)
        add_time_ids = add_time_ids.to(device).repeat(batch_size * num_images_per_prompt, 1)

        # patch diffusers controlnet instance forward, undo
        # after denoising loop

        patched_cn_models = []
        if isinstance(self.controlnet, MultiControlNetModel):
            cn_models_to_patch = self.controlnet.nets
        else:
            cn_models_to_patch = [self.controlnet]

        for cn_model in cn_models_to_patch:
            cn_og_forward = cn_model.forward

            def _cn_patch_forward(*args, **kwargs):
                encoder_hidden_states = kwargs["encoder_hidden_states"]
                if cn_model.encoder_hid_proj is not None and cn_model.config.encoder_hid_dim_type == "text_proj":
                    # Ensure encoder_hidden_states is on the same device as the projection layer
                    encoder_hidden_states = encoder_hidden_states.to(cn_model.encoder_hid_proj.weight.device)
                    encoder_hidden_states = cn_model.encoder_hid_proj(encoder_hidden_states)
                kwargs.pop("encoder_hidden_states")
                return cn_og_forward(*args, encoder_hidden_states=encoder_hidden_states, **kwargs)

            cn_model.forward = _cn_patch_forward
            patched_cn_models.append((cn_model, cn_og_forward))

        # 8. Denoising loop
        num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
        try:
            with self.progress_bar(total=num_inference_steps) as progress_bar:
                for i, t in enumerate(timesteps):
                    # expand the latents if we are doing classifier free guidance
                    latent_model_input = torch.cat([latents] * 2) if self.do_classifier_free_guidance else latents
                    latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)

                    added_cond_kwargs = {"text_embeds": add_text_embeds, "time_ids": add_time_ids}

                    # controlnet(s) inference
                    if guess_mode and self.do_classifier_free_guidance:
                        # Infer ControlNet only for the conditional batch.
                        control_model_input = latents
                        control_model_input = self.scheduler.scale_model_input(control_model_input, t)
                        controlnet_prompt_embeds = prompt_embeds.chunk(2)[1]
                        controlnet_added_cond_kwargs = {
                            "text_embeds": add_text_embeds.chunk(2)[1],
                            "time_ids": add_time_ids.chunk(2)[1],
                        }
                    else:
                        control_model_input = latent_model_input
                        controlnet_prompt_embeds = prompt_embeds
                        controlnet_added_cond_kwargs = added_cond_kwargs

                    if isinstance(controlnet_keep[i], list):
                        cond_scale = [c * s for c, s in zip(controlnet_conditioning_scale, controlnet_keep[i])]
                    else:
                        controlnet_cond_scale = controlnet_conditioning_scale
                        if isinstance(controlnet_cond_scale, list):
                            controlnet_cond_scale = controlnet_cond_scale[0]
                        cond_scale = controlnet_cond_scale * controlnet_keep[i]

                    down_block_res_samples, mid_block_res_sample = self.controlnet(
                        control_model_input,
                        t,
                        encoder_hidden_states=controlnet_prompt_embeds,
                        controlnet_cond=image,
                        conditioning_scale=cond_scale,
                        guess_mode=guess_mode,
                        added_cond_kwargs=controlnet_added_cond_kwargs,
                        return_dict=False,
                    )

                    if guess_mode and self.do_classifier_free_guidance:
                        # Inferred ControlNet only for the conditional batch.
                        # To apply the output of ControlNet to both the unconditional and conditional batches,
                        # add 0 to the unconditional batch to keep it unchanged.
                        down_block_res_samples = [torch.cat([torch.zeros_like(d), d]) for d in down_block_res_samples]
                        mid_block_res_sample = torch.cat(
                            [torch.zeros_like(mid_block_res_sample), mid_block_res_sample]
                        )

                    if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
                        added_cond_kwargs["image_embeds"] = image_embeds

                    # predict the noise residual
                    noise_pred = self.unet(
                        latent_model_input,
                        t,
                        encoder_hidden_states=prompt_embeds,
                        timestep_cond=timestep_cond,
                        cross_attention_kwargs=self.cross_attention_kwargs,
                        down_block_additional_residuals=down_block_res_samples,
                        mid_block_additional_residual=mid_block_res_sample,
                        added_cond_kwargs=added_cond_kwargs,
                        return_dict=False,
                    )[0]

                    # perform guidance
                    if self.do_classifier_free_guidance:
                        noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
                        noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)

                    # compute the previous noisy sample x_t -> x_t-1
                    latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]

                    if callback_on_step_end is not None:
                        callback_kwargs = {}
                        for k in callback_on_step_end_tensor_inputs:
                            callback_kwargs[k] = locals()[k]
                        callback_outputs = callback_on_step_end(self, i, t, callback_kwargs)

                        latents = callback_outputs.pop("latents", latents)
                        prompt_embeds = callback_outputs.pop("prompt_embeds", prompt_embeds)
                        negative_prompt_embeds = callback_outputs.pop("negative_prompt_embeds", negative_prompt_embeds)
                        add_text_embeds = callback_outputs.pop("add_text_embeds", add_text_embeds)
                        negative_pooled_prompt_embeds = callback_outputs.pop(
                            "negative_pooled_prompt_embeds", negative_pooled_prompt_embeds
                        )
                        add_time_ids = callback_outputs.pop("add_time_ids", add_time_ids)
                        negative_add_time_ids = callback_outputs.pop("negative_add_time_ids", negative_add_time_ids)
                        image = callback_outputs.pop("image", image)

                    # call the callback, if provided
                    if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
                        progress_bar.update()
                        if callback is not None and i % callback_steps == 0:
                            step_idx = i // getattr(self.scheduler, "order", 1)
                            callback(step_idx, t, latents)

        finally:
            for cn_and_og in patched_cn_models:
                cn_and_og[0].forward = cn_and_og[1]

        # If we do sequential model offloading, let's offload unet and controlnet
        # manually for max memory savings
        if hasattr(self, "final_offload_hook") and self.final_offload_hook is not None:
            self.unet.to("cpu")
            self.controlnet.to("cpu")
            torch.cuda.empty_cache()
            torch.cuda.ipc_collect()

        if not output_type == "latent":
            # make sure the VAE is in float32 mode, as it overflows in float16
            needs_upcasting = self.vae.dtype == torch.float16 and self.vae.config.force_upcast

            if needs_upcasting:
                self.upcast_vae()
                latents = latents.to(next(iter(self.vae.post_quant_conv.parameters())).dtype)

            latents = latents / self.vae.config.scaling_factor
            image = self.vae.decode(latents, return_dict=False)[0]

            # cast back to fp16 if needed
            if needs_upcasting:
                self.vae.to(dtype=torch.float16)
        else:
            image = latents
            return StableDiffusionXLPipelineOutput(images=image)

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

        # Offload all models
        self.maybe_free_model_hooks()

        if not return_dict:
            return (image,)

        return StableDiffusionXLPipelineOutput(images=image)