diffusers 0.23.0__py3-none-any.whl → 0.24.0__py3-none-any.whl

diffusers/models/vae_flax.py

@@ -214,6 +214,7 @@ class FlaxAttentionBlock(nn.Module):
             Parameters `dtype`
 
     """
+
     channels: int
     num_head_channels: int = None
     num_groups: int = 32
@@ -291,6 +292,7 @@ class FlaxDownEncoderBlock2D(nn.Module):
         dtype (:obj:`jnp.dtype`, *optional*, defaults to jnp.float32):
             Parameters `dtype`
     """
+
     in_channels: int
     out_channels: int
     dropout: float = 0.0
@@ -347,6 +349,7 @@ class FlaxUpDecoderBlock2D(nn.Module):
         dtype (:obj:`jnp.dtype`, *optional*, defaults to jnp.float32):
             Parameters `dtype`
     """
+
     in_channels: int
     out_channels: int
     dropout: float = 0.0
@@ -401,6 +404,7 @@ class FlaxUNetMidBlock2D(nn.Module):
         dtype (:obj:`jnp.dtype`, *optional*, defaults to jnp.float32):
             Parameters `dtype`
     """
+
     in_channels: int
     dropout: float = 0.0
     num_layers: int = 1
@@ -488,6 +492,7 @@ class FlaxEncoder(nn.Module):
         dtype (:obj:`jnp.dtype`, *optional*, defaults to jnp.float32):
             Parameters `dtype`
     """
+
     in_channels: int = 3
     out_channels: int = 3
     down_block_types: Tuple[str] = ("DownEncoderBlock2D",)
@@ -600,6 +605,7 @@ class FlaxDecoder(nn.Module):
         dtype (:obj:`jnp.dtype`, *optional*, defaults to jnp.float32):
             parameters `dtype`
     """
+
     in_channels: int = 3
     out_channels: int = 3
     up_block_types: Tuple[str] = ("UpDecoderBlock2D",)
@@ -767,6 +773,7 @@ class FlaxAutoencoderKL(nn.Module, FlaxModelMixin, ConfigMixin):
         dtype (`jnp.dtype`, *optional*, defaults to `jnp.float32`):
             The `dtype` of the parameters.
     """
+
     in_channels: int = 3
     out_channels: int = 3
     down_block_types: Tuple[str] = ("DownEncoderBlock2D",)

diffusers/models/vq_model.py

@@ -148,7 +148,9 @@ class VQModel(ModelMixin, ConfigMixin):
 
         return DecoderOutput(sample=dec)
 
-    def forward(self, sample: torch.FloatTensor, return_dict: bool = True) -> Union[DecoderOutput, torch.FloatTensor]:
+    def forward(
+        self, sample: torch.FloatTensor, return_dict: bool = True
+    ) -> Union[DecoderOutput, Tuple[torch.FloatTensor, ...]]:
         r"""
         The [`VQModel`] forward method.
 

diffusers/optimization.py

@@ -37,7 +37,7 @@ class SchedulerType(Enum):
     PIECEWISE_CONSTANT = "piecewise_constant"
 
 
-def get_constant_schedule(optimizer: Optimizer, last_epoch: int = -1):
+def get_constant_schedule(optimizer: Optimizer, last_epoch: int = -1) -> LambdaLR:
     """
     Create a schedule with a constant learning rate, using the learning rate set in optimizer.
 
@@ -53,7 +53,7 @@ def get_constant_schedule(optimizer: Optimizer, last_epoch: int = -1):
     return LambdaLR(optimizer, lambda _: 1, last_epoch=last_epoch)
 
 
-def get_constant_schedule_with_warmup(optimizer: Optimizer, num_warmup_steps: int, last_epoch: int = -1):
+def get_constant_schedule_with_warmup(optimizer: Optimizer, num_warmup_steps: int, last_epoch: int = -1) -> LambdaLR:
     """
     Create a schedule with a constant learning rate preceded by a warmup period during which the learning rate
     increases linearly between 0 and the initial lr set in the optimizer.
@@ -78,7 +78,7 @@ def get_constant_schedule_with_warmup(optimizer: Optimizer, num_warmup_steps: in
     return LambdaLR(optimizer, lr_lambda, last_epoch=last_epoch)
 
 
-def get_piecewise_constant_schedule(optimizer: Optimizer, step_rules: str, last_epoch: int = -1):
+def get_piecewise_constant_schedule(optimizer: Optimizer, step_rules: str, last_epoch: int = -1) -> LambdaLR:
     """
     Create a schedule with a constant learning rate, using the learning rate set in optimizer.
 
@@ -120,7 +120,9 @@ def get_piecewise_constant_schedule(optimizer: Optimizer, step_rules: str, last_
     return LambdaLR(optimizer, rules_func, last_epoch=last_epoch)
 
 
-def get_linear_schedule_with_warmup(optimizer, num_warmup_steps, num_training_steps, last_epoch=-1):
+def get_linear_schedule_with_warmup(
+    optimizer: Optimizer, num_warmup_steps: int, num_training_steps: int, last_epoch: int = -1
+) -> LambdaLR:
     """
     Create a schedule with a learning rate that decreases linearly from the initial lr set in the optimizer to 0, after
     a warmup period during which it increases linearly from 0 to the initial lr set in the optimizer.
@@ -151,7 +153,7 @@ def get_linear_schedule_with_warmup(optimizer, num_warmup_steps, num_training_st
 
 def get_cosine_schedule_with_warmup(
     optimizer: Optimizer, num_warmup_steps: int, num_training_steps: int, num_cycles: float = 0.5, last_epoch: int = -1
-):
+) -> LambdaLR:
     """
     Create a schedule with a learning rate that decreases following the values of the cosine function between the
     initial lr set in the optimizer to 0, after a warmup period during which it increases linearly between 0 and the
@@ -185,7 +187,7 @@ def get_cosine_schedule_with_warmup(
 
 def get_cosine_with_hard_restarts_schedule_with_warmup(
     optimizer: Optimizer, num_warmup_steps: int, num_training_steps: int, num_cycles: int = 1, last_epoch: int = -1
-):
+) -> LambdaLR:
     """
     Create a schedule with a learning rate that decreases following the values of the cosine function between the
     initial lr set in the optimizer to 0, with several hard restarts, after a warmup period during which it increases
@@ -219,8 +221,13 @@ def get_cosine_with_hard_restarts_schedule_with_warmup(
 
 
 def get_polynomial_decay_schedule_with_warmup(
-    optimizer, num_warmup_steps, num_training_steps, lr_end=1e-7, power=1.0, last_epoch=-1
-):
+    optimizer: Optimizer,
+    num_warmup_steps: int,
+    num_training_steps: int,
+    lr_end: float = 1e-7,
+    power: float = 1.0,
+    last_epoch: int = -1,
+) -> LambdaLR:
     """
     Create a schedule with a learning rate that decreases as a polynomial decay from the initial lr set in the
     optimizer to end lr defined by *lr_end*, after a warmup period during which it increases linearly from 0 to the
@@ -288,7 +295,7 @@ def get_scheduler(
     num_cycles: int = 1,
     power: float = 1.0,
     last_epoch: int = -1,
-):
+) -> LambdaLR:
     """
     Unified API to get any scheduler from its name.
 

diffusers/pipelines/__init__.py

@@ -17,7 +17,12 @@ from ..utils import (
 
 # These modules contain pipelines from multiple libraries/frameworks
 _dummy_objects = {}
-_import_structure = {"stable_diffusion": [], "stable_diffusion_xl": [], "latent_diffusion": [], "controlnet": []}
+_import_structure = {
+    "controlnet": [],
+    "latent_diffusion": [],
+    "stable_diffusion": [],
+    "stable_diffusion_xl": [],
+}
 
 try:
     if not is_torch_available():
@@ -39,7 +44,11 @@ else:
     _import_structure["dit"] = ["DiTPipeline"]
     _import_structure["latent_diffusion"].extend(["LDMSuperResolutionPipeline"])
     _import_structure["latent_diffusion_uncond"] = ["LDMPipeline"]
-    _import_structure["pipeline_utils"] = ["AudioPipelineOutput", "DiffusionPipeline", "ImagePipelineOutput"]
+    _import_structure["pipeline_utils"] = [
+        "AudioPipelineOutput",
+        "DiffusionPipeline",
+        "ImagePipelineOutput",
+    ]
     _import_structure["pndm"] = ["PNDMPipeline"]
     _import_structure["repaint"] = ["RePaintPipeline"]
     _import_structure["score_sde_ve"] = ["ScoreSdeVePipeline"]
@@ -61,7 +70,10 @@ except OptionalDependencyNotAvailable:
 
     _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_objects))
 else:
-    _import_structure["alt_diffusion"] = ["AltDiffusionImg2ImgPipeline", "AltDiffusionPipeline"]
+    _import_structure["alt_diffusion"] = [
+        "AltDiffusionImg2ImgPipeline",
+        "AltDiffusionPipeline",
+    ]
     _import_structure["animatediff"] = ["AnimateDiffPipeline"]
     _import_structure["audioldm"] = ["AudioLDMPipeline"]
     _import_structure["audioldm2"] = [
@@ -110,6 +122,10 @@ else:
         "KandinskyV22PriorEmb2EmbPipeline",
         "KandinskyV22PriorPipeline",
     ]
+    _import_structure["kandinsky3"] = [
+        "Kandinsky3Img2ImgPipeline",
+        "Kandinsky3Pipeline",
+    ]
     _import_structure["latent_consistency_models"] = [
         "LatentConsistencyModelImg2ImgPipeline",
         "LatentConsistencyModelPipeline",
@@ -149,6 +165,7 @@ else:
         ]
     )
     _import_structure["stable_diffusion_safe"] = ["StableDiffusionPipelineSafe"]
+    _import_structure["stable_video_diffusion"] = ["StableVideoDiffusionPipeline"]
     _import_structure["stable_diffusion_xl"].extend(
         [
             "StableDiffusionXLImg2ImgPipeline",
@@ -157,10 +174,14 @@ else:
             "StableDiffusionXLPipeline",
         ]
     )
-    _import_structure["t2i_adapter"] = ["StableDiffusionAdapterPipeline", "StableDiffusionXLAdapterPipeline"]
+    _import_structure["t2i_adapter"] = [
+        "StableDiffusionAdapterPipeline",
+        "StableDiffusionXLAdapterPipeline",
+    ]
     _import_structure["text_to_video_synthesis"] = [
         "TextToVideoSDPipeline",
         "TextToVideoZeroPipeline",
+        "TextToVideoZeroSDXLPipeline",
         "VideoToVideoSDPipeline",
     ]
     _import_structure["unclip"] = ["UnCLIPImageVariationPipeline", "UnCLIPPipeline"]
@@ -214,7 +235,9 @@ try:
     if not (is_torch_available() and is_transformers_available() and is_k_diffusion_available()):
         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:
@@ -257,7 +280,10 @@ except OptionalDependencyNotAvailable:
 
     _dummy_objects.update(get_objects_from_module(dummy_transformers_and_torch_and_note_seq_objects))
 else:
-    _import_structure["spectrogram_diffusion"] = ["MidiProcessor", "SpectrogramDiffusionPipeline"]
+    _import_structure["spectrogram_diffusion"] = [
+        "MidiProcessor",
+        "SpectrogramDiffusionPipeline",
+    ]
 
 if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     try:
@@ -267,7 +293,11 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
         from ..utils.dummy_pt_objects import *  # noqa F403
 
     else:
-        from .auto_pipeline import AutoPipelineForImage2Image, AutoPipelineForInpainting, AutoPipelineForText2Image
+        from .auto_pipeline import (
+            AutoPipelineForImage2Image,
+            AutoPipelineForInpainting,
+            AutoPipelineForText2Image,
+        )
         from .consistency_models import ConsistencyModelPipeline
         from .dance_diffusion import DanceDiffusionPipeline
         from .ddim import DDIMPipeline
@@ -275,7 +305,11 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
         from .dit import DiTPipeline
         from .latent_diffusion import LDMSuperResolutionPipeline
         from .latent_diffusion_uncond import LDMPipeline
-        from .pipeline_utils import AudioPipelineOutput, DiffusionPipeline, ImagePipelineOutput
+        from .pipeline_utils import (
+            AudioPipelineOutput,
+            DiffusionPipeline,
+            ImagePipelineOutput,
+        )
         from .pndm import PNDMPipeline
         from .repaint import RePaintPipeline
         from .score_sde_ve import ScoreSdeVePipeline
@@ -298,7 +332,11 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
         from .alt_diffusion import AltDiffusionImg2ImgPipeline, AltDiffusionPipeline
         from .animatediff import AnimateDiffPipeline
         from .audioldm import AudioLDMPipeline
-        from .audioldm2 import AudioLDM2Pipeline, AudioLDM2ProjectionModel, AudioLDM2UNet2DConditionModel
+        from .audioldm2 import (
+            AudioLDM2Pipeline,
+            AudioLDM2ProjectionModel,
+            AudioLDM2UNet2DConditionModel,
+        )
         from .blip_diffusion import BlipDiffusionPipeline
         from .controlnet import (
             BlipDiffusionControlNetPipeline,
@@ -338,7 +376,14 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             KandinskyV22PriorEmb2EmbPipeline,
             KandinskyV22PriorPipeline,
         )
-        from .latent_consistency_models import LatentConsistencyModelImg2ImgPipeline, LatentConsistencyModelPipeline
+        from .kandinsky3 import (
+            Kandinsky3Img2ImgPipeline,
+            Kandinsky3Pipeline,
+        )
+        from .latent_consistency_models import (
+            LatentConsistencyModelImg2ImgPipeline,
+            LatentConsistencyModelPipeline,
+        )
         from .latent_diffusion import LDMTextToImagePipeline
         from .musicldm import MusicLDMPipeline
         from .paint_by_example import PaintByExamplePipeline
@@ -377,10 +422,15 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             StableDiffusionXLInstructPix2PixPipeline,
             StableDiffusionXLPipeline,
         )
-        from .t2i_adapter import StableDiffusionAdapterPipeline, StableDiffusionXLAdapterPipeline
+        from .stable_video_diffusion import StableVideoDiffusionPipeline
+        from .t2i_adapter import (
+            StableDiffusionAdapterPipeline,
+            StableDiffusionXLAdapterPipeline,
+        )
         from .text_to_video_synthesis import (
             TextToVideoSDPipeline,
             TextToVideoZeroPipeline,
+            TextToVideoZeroSDXLPipeline,
             VideoToVideoSDPipeline,
         )
         from .unclip import UnCLIPImageVariationPipeline, UnCLIPPipeline
@@ -466,7 +516,10 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             from ..utils.dummy_transformers_and_torch_and_note_seq_objects import *  # noqa F403
 
         else:
-            from .spectrogram_diffusion import MidiProcessor, SpectrogramDiffusionPipeline
+            from .spectrogram_diffusion import (
+                MidiProcessor,
+                SpectrogramDiffusionPipeline,
+            )
 
 else:
     import sys

diffusers/pipelines/alt_diffusion/pipeline_alt_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, XLMRobertaTokenizer
+from transformers import CLIPImageProcessor, CLIPVisionModelWithProjection, XLMRobertaTokenizer
 
 from ...configuration_utils import FrozenDict
-from ...image_processor import VaeImageProcessor
-from ...loaders import 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
@@ -73,8 +73,55 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     return noise_cfg
 
 
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+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
+
+
 # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline with Stable->Alt, CLIPTextModel->RobertaSeriesModelWithTransformation, CLIPTokenizer->XLMRobertaTokenizer, AltDiffusionSafetyChecker->StableDiffusionSafetyChecker
-class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin):
+class AltDiffusionPipeline(
+    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, IPAdapterMixin, FromSingleFileMixin
+):
     r"""
     Pipeline for text-to-image generation using Alt Diffusion.
 
@@ -86,6 +133,7 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
         - [`~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`]):
@@ -108,7 +156,7 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
     """
 
     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"]
 
@@ -121,6 +169,7 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
         scheduler: KarrasDiffusionSchedulers,
         safety_checker: StableDiffusionSafetyChecker,
         feature_extractor: CLIPImageProcessor,
+        image_encoder: CLIPVisionModelWithProjection = None,
         requires_safety_checker: bool = True,
     ):
         super().__init__()
@@ -197,6 +246,7 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
             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)
@@ -243,10 +293,7 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
         lora_scale: Optional[float] = None,
         **kwargs,
     ):
-        deprecation_message = (
-            "`_encode_prompt()` is deprecated and it will be removed in a future version. Use `encode_prompt()`"
-            " instead. Also, be aware that the output format changed from a concatenated tensor to a tuple."
-        )
+        deprecation_message = "`_encode_prompt()` is deprecated and it will be removed in a future version. Use `encode_prompt()` instead. Also, be aware that the output format changed from a concatenated tensor to a tuple."
         deprecate("_encode_prompt()", "1.0.0", deprecation_message, standard_warn=False)
 
         prompt_embeds_tuple = self.encode_prompt(
@@ -447,6 +494,19 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
 
         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
@@ -462,10 +522,7 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
         return image, has_nsfw_concept
 
     def decode_latents(self, latents):
-        deprecation_message = (
-            "The decode_latents method is deprecated and will be removed in 1.0.0. Please use"
-            " VaeImageProcessor.postprocess(...) instead"
-        )
+        deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
 
         latents = 1 / self.vae.config.scaling_factor * latents
@@ -515,8 +572,7 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
             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"
-                f" {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}"
+                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:
@@ -651,6 +707,7 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
         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,
@@ -659,6 +716,7 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
         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,
@@ -681,6 +739,10 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
             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`.
@@ -705,6 +767,7 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
             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`):
@@ -728,7 +791,7 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
             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:
 
@@ -747,15 +810,13 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
             deprecate(
                 "callback",
                 "1.0.0",
-                "Passing `callback` as an input argument to `__call__` is deprecated, consider using"
-                " `callback_on_step_end`",
+                "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`",
+                "Passing `callback_steps` as an input argument to `__call__` is deprecated, consider using `callback_on_step_end`",
             )
 
         # 0. Default height and width to unet
@@ -806,15 +867,20 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
             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
@@ -832,7 +898,10 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
         # 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)
@@ -856,6 +925,7 @@ class AltDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
                     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]