diffusers 0.23.1__py3-none-any.whl → 0.25.0__py3-none-any.whl

diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_upscale.py

@@ -76,6 +76,12 @@ class StableDiffusionUpscalePipeline(
     This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
     implemented for all pipelines (downloading, saving, running on a particular device, etc.).
 
+    The pipeline also inherits the following loading methods:
+        - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
+
     Args:
         vae ([`AutoencoderKL`]):
             Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations.
@@ -92,6 +98,7 @@ class StableDiffusionUpscalePipeline(
             A scheduler to be used in combination with `unet` to denoise the encoded image latents. Can be one of
             [`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
     """
+
     model_cpu_offload_seq = "text_encoder->unet->vae"
     _optional_components = ["watermarker", "safety_checker", "feature_extractor"]
     _exclude_from_cpu_offload = ["safety_checker"]

diffusers/pipelines/stable_diffusion/pipeline_stable_unclip.py

@@ -65,6 +65,11 @@ class StableUnCLIPPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
     This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
     implemented for all pipelines (downloading, saving, running on a particular device, etc.).
 
+    The pipeline also inherits the following loading methods:
+        - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+
     Args:
         prior_tokenizer ([`CLIPTokenizer`]):
             A [`CLIPTokenizer`].

diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py

@@ -76,6 +76,11 @@ class StableUnCLIPImg2ImgPipeline(DiffusionPipeline, TextualInversionLoaderMixin
     This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
     implemented for all pipelines (downloading, saving, running on a particular device, etc.).
 
+    The pipeline also inherits the following loading methods:
+        - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+
     Args:
         feature_extractor ([`CLIPImageProcessor`]):
             Feature extractor for image pre-processing before being encoded.

diffusers/pipelines/stable_diffusion_attend_and_excite/__init__.py

@@ -0,0 +1,48 @@
+from typing import TYPE_CHECKING
+
+from ...utils import (
+    DIFFUSERS_SLOW_IMPORT,
+    OptionalDependencyNotAvailable,
+    _LazyModule,
+    get_objects_from_module,
+    is_torch_available,
+    is_transformers_available,
+)
+
+
+_dummy_objects = {}
+_import_structure = {}
+
+
+try:
+    if not (is_transformers_available() and is_torch_available()):
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    from ...utils import dummy_torch_and_transformers_objects  # noqa F403
+
+    _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_objects))
+else:
+    _import_structure["pipeline_stable_diffusion_attend_and_excite"] = ["StableDiffusionAttendAndExcitePipeline"]
+
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
+    try:
+        if not (is_transformers_available() and is_torch_available()):
+            raise OptionalDependencyNotAvailable()
+
+    except OptionalDependencyNotAvailable:
+        from ...utils.dummy_torch_and_transformers_objects import *
+    else:
+        from .pipeline_stable_diffusion_attend_and_excite import StableDiffusionAttendAndExcitePipeline
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(
+        __name__,
+        globals()["__file__"],
+        _import_structure,
+        module_spec=__spec__,
+    )
+
+    for name, value in _dummy_objects.items():
+        setattr(sys.modules[__name__], name, value)

diffusers/pipelines/{stable_diffusion → stable_diffusion_attend_and_excite}/pipeline_stable_diffusion_attend_and_excite.py

@@ -37,8 +37,8 @@ from ...utils import (
 )
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
-from . import StableDiffusionPipelineOutput
-from .safety_checker import StableDiffusionSafetyChecker
+from ..stable_diffusion import StableDiffusionPipelineOutput
+from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 
 
 logger = logging.get_logger(__name__)
@@ -177,6 +177,9 @@ class StableDiffusionAttendAndExcitePipeline(DiffusionPipeline, TextualInversion
     This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods
     implemented for all pipelines (downloading, saving, running on a particular device, etc.).
 
+    The pipeline also inherits the following loading methods:
+        - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
+
     Args:
         vae ([`AutoencoderKL`]):
             Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations.
@@ -196,6 +199,7 @@ class StableDiffusionAttendAndExcitePipeline(DiffusionPipeline, TextualInversion
         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->unet->vae"
     _optional_components = ["safety_checker", "feature_extractor"]
     _exclude_from_cpu_offload = ["safety_checker"]

diffusers/pipelines/stable_diffusion_diffedit/__init__.py

@@ -0,0 +1,48 @@
+from typing import TYPE_CHECKING
+
+from ...utils import (
+    DIFFUSERS_SLOW_IMPORT,
+    OptionalDependencyNotAvailable,
+    _LazyModule,
+    get_objects_from_module,
+    is_torch_available,
+    is_transformers_available,
+)
+
+
+_dummy_objects = {}
+_import_structure = {}
+
+
+try:
+    if not (is_transformers_available() and is_torch_available()):
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    from ...utils import dummy_torch_and_transformers_objects  # noqa F403
+
+    _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_objects))
+else:
+    _import_structure["pipeline_stable_diffusion_diffedit"] = ["StableDiffusionDiffEditPipeline"]
+
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
+    try:
+        if not (is_transformers_available() and is_torch_available()):
+            raise OptionalDependencyNotAvailable()
+
+    except OptionalDependencyNotAvailable:
+        from ...utils.dummy_torch_and_transformers_objects import *
+    else:
+        from .pipeline_stable_diffusion_diffedit import StableDiffusionDiffEditPipeline
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(
+        __name__,
+        globals()["__file__"],
+        _import_structure,
+        module_spec=__spec__,
+    )
+
+    for name, value in _dummy_objects.items():
+        setattr(sys.modules[__name__], name, value)

diffusers/pipelines/{stable_diffusion → stable_diffusion_diffedit}/pipeline_stable_diffusion_diffedit.py

@@ -40,8 +40,8 @@ from ...utils import (
 )
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
-from . import StableDiffusionPipelineOutput
-from .safety_checker import StableDiffusionSafetyChecker
+from ..stable_diffusion import StableDiffusionPipelineOutput
+from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
@@ -273,6 +273,7 @@ class StableDiffusionDiffEditPipeline(DiffusionPipeline, TextualInversionLoaderM
         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->unet->vae"
     _optional_components = ["safety_checker", "feature_extractor", "inverse_scheduler"]
     _exclude_from_cpu_offload = ["safety_checker"]
@@ -787,7 +788,6 @@ class StableDiffusionDiffEditPipeline(DiffusionPipeline, TextualInversionLoaderM
         latents = latents * self.scheduler.init_noise_sigma
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_pix2pix_zero.StableDiffusionPix2PixZeroPipeline.prepare_image_latents
     def prepare_image_latents(self, image, batch_size, dtype, device, generator=None):
         if not isinstance(image, (torch.Tensor, PIL.Image.Image, list)):
             raise ValueError(

diffusers/pipelines/stable_diffusion_gligen/__init__.py

@@ -0,0 +1,50 @@
+from typing import TYPE_CHECKING
+
+from ...utils import (
+    DIFFUSERS_SLOW_IMPORT,
+    OptionalDependencyNotAvailable,
+    _LazyModule,
+    get_objects_from_module,
+    is_torch_available,
+    is_transformers_available,
+)
+
+
+_dummy_objects = {}
+_import_structure = {}
+
+
+try:
+    if not (is_transformers_available() and is_torch_available()):
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    from ...utils import dummy_torch_and_transformers_objects  # noqa F403
+
+    _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_objects))
+else:
+    _import_structure["pipeline_stable_diffusion_gligen"] = ["StableDiffusionGLIGENPipeline"]
+    _import_structure["pipeline_stable_diffusion_gligen_text_image"] = ["StableDiffusionGLIGENTextImagePipeline"]
+
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
+    try:
+        if not (is_transformers_available() and is_torch_available()):
+            raise OptionalDependencyNotAvailable()
+
+    except OptionalDependencyNotAvailable:
+        from ...utils.dummy_torch_and_transformers_objects import *
+    else:
+        from .pipeline_stable_diffusion_gligen import StableDiffusionGLIGENPipeline
+        from .pipeline_stable_diffusion_gligen_text_image import StableDiffusionGLIGENTextImagePipeline
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(
+        __name__,
+        globals()["__file__"],
+        _import_structure,
+        module_spec=__spec__,
+    )
+
+    for name, value in _dummy_objects.items():
+        setattr(sys.modules[__name__], name, value)

diffusers/pipelines/{stable_diffusion → stable_diffusion_gligen}/pipeline_stable_diffusion_gligen.py

@@ -36,8 +36,8 @@ from ...utils import (
 )
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
-from . import StableDiffusionPipelineOutput
-from .safety_checker import StableDiffusionSafetyChecker
+from ..stable_diffusion import StableDiffusionPipelineOutput
+from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
@@ -125,6 +125,7 @@ class StableDiffusionGLIGENPipeline(DiffusionPipeline):
         feature_extractor ([`~transformers.CLIPImageProcessor`]):
             A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
     """
+
     _optional_components = ["safety_checker", "feature_extractor"]
     model_cpu_offload_seq = "text_encoder->unet->vae"
     _exclude_from_cpu_offload = ["safety_checker"]

diffusers/pipelines/{stable_diffusion → stable_diffusion_gligen}/pipeline_stable_diffusion_gligen_text_image.py

@@ -35,9 +35,9 @@ from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import USE_PEFT_BACKEND, logging, replace_example_docstring, scale_lora_layers, unscale_lora_layers
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
-from . import StableDiffusionPipelineOutput
-from .clip_image_project_model import CLIPImageProjection
-from .safety_checker import StableDiffusionSafetyChecker
+from ..stable_diffusion import StableDiffusionPipelineOutput
+from ..stable_diffusion.clip_image_project_model import CLIPImageProjection
+from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
@@ -177,6 +177,7 @@ class StableDiffusionGLIGENTextImagePipeline(DiffusionPipeline):
         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->unet->vae"
     _optional_components = ["safety_checker", "feature_extractor"]
     _exclude_from_cpu_offload = ["safety_checker"]

diffusers/pipelines/stable_diffusion_k_diffusion/__init__.py

@@ -0,0 +1,60 @@
+from typing import TYPE_CHECKING
+
+from ...utils import (
+    DIFFUSERS_SLOW_IMPORT,
+    OptionalDependencyNotAvailable,
+    _LazyModule,
+    get_objects_from_module,
+    is_k_diffusion_available,
+    is_k_diffusion_version,
+    is_torch_available,
+    is_transformers_available,
+)
+
+
+_dummy_objects = {}
+_import_structure = {}
+
+
+try:
+    if not (
+        is_transformers_available()
+        and is_torch_available()
+        and is_k_diffusion_available()
+        and is_k_diffusion_version(">=", "0.0.12")
+    ):
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    from ...utils import dummy_torch_and_transformers_and_k_diffusion_objects  # noqa F403
+
+    _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_and_k_diffusion_objects))
+else:
+    _import_structure["pipeline_stable_diffusion_k_diffusion"] = ["StableDiffusionKDiffusionPipeline"]
+
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
+    try:
+        if not (
+            is_transformers_available()
+            and is_torch_available()
+            and is_k_diffusion_available()
+            and is_k_diffusion_version(">=", "0.0.12")
+        ):
+            raise OptionalDependencyNotAvailable()
+
+    except OptionalDependencyNotAvailable:
+        from ...utils.dummy_torch_and_transformers_and_k_diffusion_objects import *
+    else:
+        from .pipeline_stable_diffusion_k_diffusion import StableDiffusionKDiffusionPipeline
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(
+        __name__,
+        globals()["__file__"],
+        _import_structure,
+        module_spec=__spec__,
+    )
+
+    for name, value in _dummy_objects.items():
+        setattr(sys.modules[__name__], name, value)

diffusers/pipelines/{stable_diffusion → stable_diffusion_k_diffusion}/pipeline_stable_diffusion_k_diffusion.py

@@ -27,7 +27,7 @@ from ...schedulers import LMSDiscreteScheduler
 from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
-from . import StableDiffusionPipelineOutput
+from ..stable_diffusion import StableDiffusionPipelineOutput
 
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
@@ -54,6 +54,11 @@ class StableDiffusionKDiffusionPipeline(DiffusionPipeline, TextualInversionLoade
     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.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+
     <Tip warning={true}>
 
         This is an experimental pipeline and is likely to change in the future.
@@ -80,6 +85,7 @@ class StableDiffusionKDiffusionPipeline(DiffusionPipeline, TextualInversionLoade
         feature_extractor ([`CLIPImageProcessor`]):
             Model that extracts features from generated images to be used as inputs for the `safety_checker`.
     """
+
     model_cpu_offload_seq = "text_encoder->unet->vae"
     _optional_components = ["safety_checker", "feature_extractor"]
     _exclude_from_cpu_offload = ["safety_checker"]

diffusers/pipelines/stable_diffusion_ldm3d/__init__.py

@@ -0,0 +1,48 @@
+from typing import TYPE_CHECKING
+
+from ...utils import (
+    DIFFUSERS_SLOW_IMPORT,
+    OptionalDependencyNotAvailable,
+    _LazyModule,
+    get_objects_from_module,
+    is_torch_available,
+    is_transformers_available,
+)
+
+
+_dummy_objects = {}
+_import_structure = {}
+
+
+try:
+    if not (is_transformers_available() and is_torch_available()):
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    from ...utils import dummy_torch_and_transformers_objects  # noqa F403
+
+    _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_objects))
+else:
+    _import_structure["pipeline_stable_diffusion_ldm3d"] = ["StableDiffusionLDM3DPipeline"]
+
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
+    try:
+        if not (is_transformers_available() and is_torch_available()):
+            raise OptionalDependencyNotAvailable()
+
+    except OptionalDependencyNotAvailable:
+        from ...utils.dummy_torch_and_transformers_objects import *
+    else:
+        from .pipeline_stable_diffusion_ldm3d import StableDiffusionLDM3DPipeline
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(
+        __name__,
+        globals()["__file__"],
+        _import_structure,
+        module_spec=__spec__,
+    )
+
+    for name, value in _dummy_objects.items():
+        setattr(sys.modules[__name__], name, value)

diffusers/pipelines/{stable_diffusion → stable_diffusion_ldm3d}/pipeline_stable_diffusion_ldm3d.py

@@ -19,11 +19,11 @@ from typing import Any, Callable, Dict, List, Optional, Union
 import numpy as np
 import PIL.Image
 import torch
-from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
+from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
 
-from ...image_processor import VaeImageProcessorLDM3D
-from ...loaders import FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
-from ...models import AutoencoderKL, UNet2DConditionModel
+from ...image_processor import PipelineImageInput, VaeImageProcessorLDM3D
+from ...loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
+from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel
 from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import (
@@ -37,7 +37,7 @@ from ...utils import (
 )
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
-from .safety_checker import StableDiffusionSafetyChecker
+from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
@@ -82,7 +82,7 @@ class LDM3DPipelineOutput(BaseOutput):
 
 
 class StableDiffusionLDM3DPipeline(
-    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin
+    DiffusionPipeline, TextualInversionLoaderMixin, IPAdapterMixin, LoraLoaderMixin, FromSingleFileMixin
 ):
     r"""
     Pipeline for text-to-image and 3D generation using LDM3D.
@@ -95,6 +95,7 @@ class StableDiffusionLDM3DPipeline(
         - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
         - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
         - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
+        - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
 
     Args:
         vae ([`AutoencoderKL`]):
@@ -115,8 +116,9 @@ class StableDiffusionLDM3DPipeline(
         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->unet->vae"
-    _optional_components = ["safety_checker", "feature_extractor"]
+    _optional_components = ["safety_checker", "feature_extractor", "image_encoder"]
     _exclude_from_cpu_offload = ["safety_checker"]
 
     def __init__(
@@ -128,6 +130,7 @@ class StableDiffusionLDM3DPipeline(
         scheduler: KarrasDiffusionSchedulers,
         safety_checker: StableDiffusionSafetyChecker,
         feature_extractor: CLIPImageProcessor,
+        image_encoder: Optional[CLIPVisionModelWithProjection],
         requires_safety_checker: bool = True,
     ):
         super().__init__()
@@ -156,6 +159,7 @@ class StableDiffusionLDM3DPipeline(
             scheduler=scheduler,
             safety_checker=safety_checker,
             feature_extractor=feature_extractor,
+            image_encoder=image_encoder,
         )
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
         self.image_processor = VaeImageProcessorLDM3D(vae_scale_factor=self.vae_scale_factor)
@@ -409,6 +413,31 @@ class StableDiffusionLDM3DPipeline(
 
         return prompt_embeds, negative_prompt_embeds
 
+    # 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
+
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -528,6 +557,7 @@ class StableDiffusionLDM3DPipeline(
         latents: Optional[torch.FloatTensor] = None,
         prompt_embeds: Optional[torch.FloatTensor] = None,
         negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        ip_adapter_image: Optional[PipelineImageInput] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
@@ -572,6 +602,8 @@ class StableDiffusionLDM3DPipeline(
             negative_prompt_embeds (`torch.FloatTensor`, *optional*):
                 Pre-generated negative text embeddings. Can be used to easily tweak text inputs (prompt weighting). If
                 not provided, `negative_prompt_embeds` are generated from the `negative_prompt` input argument.
+            ip_adapter_image: (`PipelineImageInput`, *optional*):
+                Optional image input to work with IP Adapters.
             output_type (`str`, *optional*, defaults to `"pil"`):
                 The output format of the generated image. Choose between `PIL.Image` or `np.array`.
             return_dict (`bool`, *optional*, defaults to `True`):
@@ -621,6 +653,14 @@ class StableDiffusionLDM3DPipeline(
         # corresponds to doing no classifier free guidance.
         do_classifier_free_guidance = guidance_scale > 1.0
 
+        if ip_adapter_image is not None:
+            output_hidden_state = False if isinstance(self.unet.encoder_hid_proj, ImageProjection) else True
+            image_embeds, negative_image_embeds = self.encode_image(
+                ip_adapter_image, device, num_images_per_prompt, output_hidden_state
+            )
+            if do_classifier_free_guidance:
+                image_embeds = torch.cat([negative_image_embeds, image_embeds])
+
         # 3. Encode input prompt
         prompt_embeds, negative_prompt_embeds = self.encode_prompt(
             prompt,
@@ -658,6 +698,9 @@ class StableDiffusionLDM3DPipeline(
         # 6. 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)
 
+        # 6.1 Add image embeds for IP-Adapter
+        added_cond_kwargs = {"image_embeds": image_embeds} if ip_adapter_image is not None else None
+
         # 7. Denoising loop
         num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
         with self.progress_bar(total=num_inference_steps) as progress_bar:
@@ -672,6 +715,7 @@ class StableDiffusionLDM3DPipeline(
                     t,
                     encoder_hidden_states=prompt_embeds,
                     cross_attention_kwargs=cross_attention_kwargs,
+                    added_cond_kwargs=added_cond_kwargs,
                     return_dict=False,
                 )[0]
 

diffusers/pipelines/stable_diffusion_panorama/__init__.py

@@ -0,0 +1,48 @@
+from typing import TYPE_CHECKING
+
+from ...utils import (
+    DIFFUSERS_SLOW_IMPORT,
+    OptionalDependencyNotAvailable,
+    _LazyModule,
+    get_objects_from_module,
+    is_torch_available,
+    is_transformers_available,
+)
+
+
+_dummy_objects = {}
+_import_structure = {}
+
+
+try:
+    if not (is_transformers_available() and is_torch_available()):
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    from ...utils import dummy_torch_and_transformers_objects  # noqa F403
+
+    _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_objects))
+else:
+    _import_structure["pipeline_stable_diffusion_panorama"] = ["StableDiffusionPanoramaPipeline"]
+
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
+    try:
+        if not (is_transformers_available() and is_torch_available()):
+            raise OptionalDependencyNotAvailable()
+
+    except OptionalDependencyNotAvailable:
+        from ...utils.dummy_torch_and_transformers_objects import *
+    else:
+        from .pipeline_stable_diffusion_panorama import StableDiffusionPanoramaPipeline
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(
+        __name__,
+        globals()["__file__"],
+        _import_structure,
+        module_spec=__spec__,
+    )
+
+    for name, value in _dummy_objects.items():
+        setattr(sys.modules[__name__], name, value)