diffusers 0.23.1__py3-none-any.whl → 0.24.0__py3-none-any.whl

diffusers/pipelines/stable_diffusion/__init__.py

@@ -55,7 +55,9 @@ try:
     if not (is_transformers_available() and is_torch_available() and is_transformers_version(">=", "4.25.0")):
         raise OptionalDependencyNotAvailable()
 except OptionalDependencyNotAvailable:
-    from ...utils.dummy_torch_and_transformers_objects import StableDiffusionImageVariationPipeline
+    from ...utils.dummy_torch_and_transformers_objects import (
+        StableDiffusionImageVariationPipeline,
+    )
 
     _dummy_objects.update({"StableDiffusionImageVariationPipeline": StableDiffusionImageVariationPipeline})
 else:
@@ -90,7 +92,9 @@ try:
     ):
         raise OptionalDependencyNotAvailable()
 except OptionalDependencyNotAvailable:
-    from ...utils import dummy_torch_and_transformers_and_k_diffusion_objects  # noqa F403
+    from ...utils import (
+        dummy_torch_and_transformers_and_k_diffusion_objects,
+    )
 
     _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_and_k_diffusion_objects))
 else:
@@ -137,18 +141,32 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             StableDiffusionPipelineOutput,
             StableDiffusionSafetyChecker,
         )
-        from .pipeline_stable_diffusion_attend_and_excite import StableDiffusionAttendAndExcitePipeline
+        from .pipeline_stable_diffusion_attend_and_excite import (
+            StableDiffusionAttendAndExcitePipeline,
+        )
         from .pipeline_stable_diffusion_gligen import StableDiffusionGLIGENPipeline
-        from .pipeline_stable_diffusion_gligen_text_image import StableDiffusionGLIGENTextImagePipeline
+        from .pipeline_stable_diffusion_gligen_text_image import (
+            StableDiffusionGLIGENTextImagePipeline,
+        )
         from .pipeline_stable_diffusion_img2img import StableDiffusionImg2ImgPipeline
         from .pipeline_stable_diffusion_inpaint import StableDiffusionInpaintPipeline
-        from .pipeline_stable_diffusion_inpaint_legacy import StableDiffusionInpaintPipelineLegacy
-        from .pipeline_stable_diffusion_instruct_pix2pix import StableDiffusionInstructPix2PixPipeline
-        from .pipeline_stable_diffusion_latent_upscale import StableDiffusionLatentUpscalePipeline
+        from .pipeline_stable_diffusion_inpaint_legacy import (
+            StableDiffusionInpaintPipelineLegacy,
+        )
+        from .pipeline_stable_diffusion_instruct_pix2pix import (
+            StableDiffusionInstructPix2PixPipeline,
+        )
+        from .pipeline_stable_diffusion_latent_upscale import (
+            StableDiffusionLatentUpscalePipeline,
+        )
         from .pipeline_stable_diffusion_ldm3d import StableDiffusionLDM3DPipeline
-        from .pipeline_stable_diffusion_model_editing import StableDiffusionModelEditingPipeline
+        from .pipeline_stable_diffusion_model_editing import (
+            StableDiffusionModelEditingPipeline,
+        )
         from .pipeline_stable_diffusion_panorama import StableDiffusionPanoramaPipeline
-        from .pipeline_stable_diffusion_paradigms import StableDiffusionParadigmsPipeline
+        from .pipeline_stable_diffusion_paradigms import (
+            StableDiffusionParadigmsPipeline,
+        )
         from .pipeline_stable_diffusion_sag import StableDiffusionSAGPipeline
         from .pipeline_stable_diffusion_upscale import StableDiffusionUpscalePipeline
         from .pipeline_stable_unclip import StableUnCLIPPipeline
@@ -160,9 +178,13 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
         if not (is_transformers_available() and is_torch_available() and is_transformers_version(">=", "4.25.0")):
             raise OptionalDependencyNotAvailable()
     except OptionalDependencyNotAvailable:
-        from ...utils.dummy_torch_and_transformers_objects import StableDiffusionImageVariationPipeline
+        from ...utils.dummy_torch_and_transformers_objects import (
+            StableDiffusionImageVariationPipeline,
+        )
     else:
-        from .pipeline_stable_diffusion_image_variation import StableDiffusionImageVariationPipeline
+        from .pipeline_stable_diffusion_image_variation import (
+            StableDiffusionImageVariationPipeline,
+        )
 
     try:
         if not (is_transformers_available() and is_torch_available() and is_transformers_version(">=", "4.26.0")):
@@ -174,9 +196,13 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             StableDiffusionPix2PixZeroPipeline,
         )
     else:
-        from .pipeline_stable_diffusion_depth2img import StableDiffusionDepth2ImgPipeline
+        from .pipeline_stable_diffusion_depth2img import (
+            StableDiffusionDepth2ImgPipeline,
+        )
         from .pipeline_stable_diffusion_diffedit import StableDiffusionDiffEditPipeline
-        from .pipeline_stable_diffusion_pix2pix_zero import StableDiffusionPix2PixZeroPipeline
+        from .pipeline_stable_diffusion_pix2pix_zero import (
+            StableDiffusionPix2PixZeroPipeline,
+        )
 
     try:
         if not (
@@ -189,7 +215,9 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     except OptionalDependencyNotAvailable:
         from ...utils.dummy_torch_and_transformers_and_k_diffusion_objects import *
     else:
-        from .pipeline_stable_diffusion_k_diffusion import StableDiffusionKDiffusionPipeline
+        from .pipeline_stable_diffusion_k_diffusion import (
+            StableDiffusionKDiffusionPipeline,
+        )
 
     try:
         if not (is_transformers_available() and is_onnx_available()):
@@ -197,11 +225,22 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     except OptionalDependencyNotAvailable:
         from ...utils.dummy_onnx_objects import *
     else:
-        from .pipeline_onnx_stable_diffusion import OnnxStableDiffusionPipeline, StableDiffusionOnnxPipeline
-        from .pipeline_onnx_stable_diffusion_img2img import OnnxStableDiffusionImg2ImgPipeline
-        from .pipeline_onnx_stable_diffusion_inpaint import OnnxStableDiffusionInpaintPipeline
-        from .pipeline_onnx_stable_diffusion_inpaint_legacy import OnnxStableDiffusionInpaintPipelineLegacy
-        from .pipeline_onnx_stable_diffusion_upscale import OnnxStableDiffusionUpscalePipeline
+        from .pipeline_onnx_stable_diffusion import (
+            OnnxStableDiffusionPipeline,
+            StableDiffusionOnnxPipeline,
+        )
+        from .pipeline_onnx_stable_diffusion_img2img import (
+            OnnxStableDiffusionImg2ImgPipeline,
+        )
+        from .pipeline_onnx_stable_diffusion_inpaint import (
+            OnnxStableDiffusionInpaintPipeline,
+        )
+        from .pipeline_onnx_stable_diffusion_inpaint_legacy import (
+            OnnxStableDiffusionInpaintPipelineLegacy,
+        )
+        from .pipeline_onnx_stable_diffusion_upscale import (
+            OnnxStableDiffusionUpscalePipeline,
+        )
 
     try:
         if not (is_transformers_available() and is_flax_available()):
@@ -210,8 +249,12 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
         from ...utils.dummy_flax_objects import *
     else:
         from .pipeline_flax_stable_diffusion import FlaxStableDiffusionPipeline
-        from .pipeline_flax_stable_diffusion_img2img import FlaxStableDiffusionImg2ImgPipeline
-        from .pipeline_flax_stable_diffusion_inpaint import FlaxStableDiffusionInpaintPipeline
+        from .pipeline_flax_stable_diffusion_img2img import (
+            FlaxStableDiffusionImg2ImgPipeline,
+        )
+        from .pipeline_flax_stable_diffusion_inpaint import (
+            FlaxStableDiffusionInpaintPipeline,
+        )
         from .pipeline_output import FlaxStableDiffusionPipelineOutput
         from .safety_checker_flax import FlaxStableDiffusionSafetyChecker
 

diffusers/pipelines/stable_diffusion/convert_from_ckpt.py

@@ -1232,13 +1232,11 @@ def download_from_original_stable_diffusion_ckpt(
         StableDiffusionPipeline,
         StableDiffusionUpscalePipeline,
         StableDiffusionXLImg2ImgPipeline,
+        StableDiffusionXLPipeline,
         StableUnCLIPImg2ImgPipeline,
         StableUnCLIPPipeline,
     )
 
-    if pipeline_class is None:
-        pipeline_class = StableDiffusionPipeline if not controlnet else StableDiffusionControlNetPipeline
-
     if prediction_type == "v-prediction":
         prediction_type = "v_prediction"
 
@@ -1333,6 +1331,13 @@ def download_from_original_stable_diffusion_ckpt(
         if image_size is None:
             image_size = 1024
 
+    if pipeline_class is None:
+        # Check if we have a SDXL or SD model and initialize default pipeline
+        if model_type not in ["SDXL", "SDXL-Refiner"]:
+            pipeline_class = StableDiffusionPipeline if not controlnet else StableDiffusionControlNetPipeline
+        else:
+            pipeline_class = StableDiffusionXLPipeline if model_type == "SDXL" else StableDiffusionXLImg2ImgPipeline
+
     if num_in_channels is None and pipeline_class == StableDiffusionInpaintPipeline:
         num_in_channels = 9
     if num_in_channels is None and pipeline_class == StableDiffusionUpscalePipeline:

diffusers/pipelines/stable_diffusion/pipeline_cycle_diffusion.py

@@ -61,6 +61,20 @@ def preprocess(image):
     return image
 
 
+# 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")
+
+
 def posterior_sample(scheduler, latents, timestep, clean_latents, generator, eta):
     # 1. get previous step value (=t-1)
     prev_timestep = timestep - scheduler.config.num_train_timesteps // scheduler.num_inference_steps
@@ -148,6 +162,7 @@ class CycleDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lor
         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"]
 
@@ -566,11 +581,12 @@ class CycleDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lor
 
             if isinstance(generator, list):
                 init_latents = [
-                    self.vae.encode(image[i : i + 1]).latent_dist.sample(generator[i]) for i in range(batch_size)
+                    retrieve_latents(self.vae.encode(image[i : i + 1]), generator=generator[i])
+                    for i in range(image.shape[0])
                 ]
                 init_latents = torch.cat(init_latents, dim=0)
             else:
-                init_latents = self.vae.encode(image).latent_dist.sample(generator)
+                init_latents = retrieve_latents(self.vae.encode(image), generator=generator)
 
             init_latents = self.vae.config.scaling_factor * init_latents
 

diffusers/pipelines/stable_diffusion/pipeline_flax_stable_diffusion.py

@@ -410,13 +410,13 @@ class FlaxStableDiffusionPipeline(FlaxDiffusionPipeline):
 
             images_uint8_casted = np.asarray(images_uint8_casted).reshape(num_devices * batch_size, height, width, 3)
             images_uint8_casted, has_nsfw_concept = self._run_safety_checker(images_uint8_casted, safety_params, jit)
-            images = np.asarray(images)
+            images = np.asarray(images).copy()
 
             # block images
             if any(has_nsfw_concept):
                 for i, is_nsfw in enumerate(has_nsfw_concept):
                     if is_nsfw:
-                        images[i] = np.asarray(images_uint8_casted[i])
+                        images[i, 0] = np.asarray(images_uint8_casted[i])
 
             images = images.reshape(num_devices, batch_size, height, width, 3)
         else:

diffusers/pipelines/stable_diffusion/pipeline_onnx_stable_diffusion_img2img.py

@@ -33,10 +33,7 @@ logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.preprocess with 8->64
 def preprocess(image):
-    deprecation_message = (
-        "The preprocess method is deprecated and will be removed in diffusers 1.0.0. Please use"
-        " VaeImageProcessor.preprocess(...) instead"
-    )
+    deprecation_message = "The preprocess method is deprecated and will be removed in diffusers 1.0.0. Please use VaeImageProcessor.preprocess(...) instead"
     deprecate("preprocess", "1.0.0", deprecation_message, standard_warn=False)
     if isinstance(image, torch.Tensor):
         return image
@@ -85,6 +82,7 @@ class OnnxStableDiffusionImg2ImgPipeline(DiffusionPipeline):
         feature_extractor ([`CLIPImageProcessor`]):
             Model that extracts features from generated images to be used as inputs for the `safety_checker`.
     """
+
     vae_encoder: OnnxRuntimeModel
     vae_decoder: OnnxRuntimeModel
     text_encoder: OnnxRuntimeModel

diffusers/pipelines/stable_diffusion/pipeline_onnx_stable_diffusion_inpaint.py

@@ -80,6 +80,7 @@ class OnnxStableDiffusionInpaintPipeline(DiffusionPipeline):
         feature_extractor ([`CLIPImageProcessor`]):
             Model that extracts features from generated images to be used as inputs for the `safety_checker`.
     """
+
     vae_encoder: OnnxRuntimeModel
     vae_decoder: OnnxRuntimeModel
     text_encoder: OnnxRuntimeModel

diffusers/pipelines/stable_diffusion/pipeline_onnx_stable_diffusion_inpaint_legacy.py

@@ -66,6 +66,7 @@ class OnnxStableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
         feature_extractor ([`CLIPImageProcessor`]):
             Model that extracts features from generated images to be used as inputs for the `safety_checker`.
     """
+
     _optional_components = ["safety_checker", "feature_extractor"]
     _is_onnx = True
 

diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py

@@ -17,11 +17,11 @@ from typing import Any, Callable, Dict, List, Optional, Union
 
 import torch
 from packaging import version
-from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
+from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
 
 from ...configuration_utils import FrozenDict
-from ...image_processor import VaeImageProcessor
-from ...loaders import FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
+from ...image_processor import PipelineImageInput, VaeImageProcessor
+from ...loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, UNet2DConditionModel
 from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
@@ -70,7 +70,53 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
-class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin):
+def retrieve_timesteps(
+    scheduler,
+    num_inference_steps: Optional[int] = None,
+    device: Optional[Union[str, torch.device]] = None,
+    timesteps: Optional[List[int]] = None,
+    **kwargs,
+):
+    """
+    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
+    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.
+
+    Args:
+        scheduler (`SchedulerMixin`):
+            The scheduler to get timesteps from.
+        num_inference_steps (`int`):
+            The number of diffusion steps used when generating samples with a pre-trained model. If used,
+            `timesteps` must be `None`.
+        device (`str` or `torch.device`, *optional*):
+            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
+        timesteps (`List[int]`, *optional*):
+                Custom timesteps used to support arbitrary spacing between timesteps. If `None`, then the default
+                timestep spacing strategy of the scheduler is used. If `timesteps` is passed, `num_inference_steps`
+                must be `None`.
+
+    Returns:
+        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
+        second element is the number of inference steps.
+    """
+    if timesteps is not None:
+        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accepts_timesteps:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" timestep schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    else:
+        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+    return timesteps, num_inference_steps
+
+
+class StableDiffusionPipeline(
+    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, IPAdapterMixin, FromSingleFileMixin
+):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion.
 
@@ -82,6 +128,7 @@ class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
         - [`~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`]):
@@ -102,8 +149,9 @@ class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
         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"]
     _callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"]
 
@@ -116,6 +164,7 @@ class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
         scheduler: KarrasDiffusionSchedulers,
         safety_checker: StableDiffusionSafetyChecker,
         feature_extractor: CLIPImageProcessor,
+        image_encoder: CLIPVisionModelWithProjection = None,
         requires_safety_checker: bool = True,
     ):
         super().__init__()
@@ -192,6 +241,7 @@ class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
             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 = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
@@ -439,6 +489,19 @@ class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
 
         return prompt_embeds, negative_prompt_embeds
 
+    def encode_image(self, image, device, num_images_per_prompt):
+        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)
+        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
@@ -640,6 +703,7 @@ class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
         height: Optional[int] = None,
         width: Optional[int] = None,
         num_inference_steps: int = 50,
+        timesteps: List[int] = None,
         guidance_scale: float = 7.5,
         negative_prompt: Optional[Union[str, List[str]]] = None,
         num_images_per_prompt: Optional[int] = 1,
@@ -648,6 +712,7 @@ class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
         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,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
@@ -670,6 +735,10 @@ class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
             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.
+            timesteps (`List[int]`, *optional*):
+                Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument
+                in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is
+                passed will be used. Must be in descending order.
             guidance_scale (`float`, *optional*, defaults to 7.5):
                 A higher guidance scale value encourages the model to generate images closely linked to the text
                 `prompt` at the expense of lower image quality. Guidance scale is enabled when `guidance_scale > 1`.
@@ -694,6 +763,7 @@ class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
             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`):
@@ -717,7 +787,7 @@ class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
             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 pipeine class.
+                `._callback_tensor_inputs` attribute of your pipeline class.
 
         Examples:
 
@@ -793,15 +863,20 @@ class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
             lora_scale=lora_scale,
             clip_skip=self.clip_skip,
         )
+
         # For classifier free guidance, we need to do two forward passes.
         # Here we concatenate the unconditional and text embeddings into a single batch
         # to avoid doing two forward passes
         if self.do_classifier_free_guidance:
             prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds])
 
+        if ip_adapter_image is not None:
+            image_embeds, negative_image_embeds = self.encode_image(ip_adapter_image, device, num_images_per_prompt)
+            if self.do_classifier_free_guidance:
+                image_embeds = torch.cat([negative_image_embeds, image_embeds])
+
         # 4. Prepare timesteps
-        self.scheduler.set_timesteps(num_inference_steps, device=device)
-        timesteps = self.scheduler.timesteps
+        timesteps, num_inference_steps = retrieve_timesteps(self.scheduler, num_inference_steps, device, timesteps)
 
         # 5. Prepare latent variables
         num_channels_latents = self.unet.config.in_channels
@@ -819,7 +894,10 @@ class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
         # 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)
 
-        # 6.5 Optionally get Guidance Scale Embedding
+        # 6.1 Add image embeds for IP-Adapter
+        added_cond_kwargs = {"image_embeds": image_embeds} if ip_adapter_image is not None else None
+
+        # 6.2 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)
@@ -843,6 +921,7 @@ class StableDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Lo
                     encoder_hidden_states=prompt_embeds,
                     timestep_cond=timestep_cond,
                     cross_attention_kwargs=self.cross_attention_kwargs,
+                    added_cond_kwargs=added_cond_kwargs,
                     return_dict=False,
                 )[0]
 

diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_attend_and_excite.py

@@ -196,6 +196,7 @@ 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/pipeline_stable_diffusion_depth2img.py

@@ -37,9 +37,13 @@ logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
 # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
-def retrieve_latents(encoder_output, generator):
-    if hasattr(encoder_output, "latent_dist"):
+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:
@@ -95,6 +99,7 @@ class StableDiffusionDepth2ImgPipeline(DiffusionPipeline, TextualInversionLoader
             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"
     _callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds", "depth_mask"]
 
@@ -674,7 +679,7 @@ class StableDiffusionDepth2ImgPipeline(DiffusionPipeline, TextualInversionLoader
             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 pipeine class.
+                `._callback_tensor_inputs` attribute of your pipeline class.
         Examples:
 
         ```py

diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_diffedit.py

@@ -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"]

diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_gligen.py

@@ -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/pipeline_stable_diffusion_gligen_text_image.py

@@ -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/pipeline_stable_diffusion_image_variation.py

@@ -62,6 +62,7 @@ class StableDiffusionImageVariationPipeline(DiffusionPipeline):
         feature_extractor ([`~transformers.CLIPImageProcessor`]):
             A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
     """
+
     # TODO: feature_extractor is required to encode images (if they are in PIL format),
     # we should give a descriptive message if the pipeline doesn't have one.
     _optional_components = ["safety_checker"]