diffusers 0.27.0__py3-none-any.whl → 0.32.2__py3-none-any.whl

diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_img2img.py

@@ -25,6 +25,7 @@ from transformers import (
     CLIPVisionModelWithProjection,
 )
 
+from ...callbacks import MultiPipelineCallbacks, PipelineCallback
 from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...loaders import (
     FromSingleFileMixin,
@@ -35,8 +36,6 @@ from ...loaders import (
 from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel
 from ...models.attention_processor import (
     AttnProcessor2_0,
-    LoRAAttnProcessor2_0,
-    LoRAXFormersAttnProcessor,
     XFormersAttnProcessor,
 )
 from ...models.lora import adjust_lora_scale_text_encoder
@@ -91,9 +90,21 @@ EXAMPLE_DOC_STRING = """
 
 # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
-    """
-    Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and
-    Sample Steps are Flawed](https://arxiv.org/pdf/2305.08891.pdf). See Section 3.4
+    r"""
+    Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
+    Section 3.4 from [Common Diffusion Noise Schedules and Sample Steps are
+    Flawed](https://arxiv.org/pdf/2305.08891.pdf).
+
+    Args:
+        noise_cfg (`torch.Tensor`):
+            The predicted noise tensor for the guided diffusion process.
+        noise_pred_text (`torch.Tensor`):
+            The predicted noise tensor for the text-guided diffusion process.
+        guidance_rescale (`float`, *optional*, defaults to 0.0):
+            A rescale factor applied to the noise predictions.
+
+    Returns:
+        noise_cfg (`torch.Tensor`): The rescaled noise prediction tensor.
     """
     std_text = noise_pred_text.std(dim=list(range(1, noise_pred_text.ndim)), keepdim=True)
     std_cfg = noise_cfg.std(dim=list(range(1, noise_cfg.ndim)), keepdim=True)
@@ -124,9 +135,10 @@ def retrieve_timesteps(
     num_inference_steps: Optional[int] = None,
     device: Optional[Union[str, torch.device]] = None,
     timesteps: Optional[List[int]] = None,
+    sigmas: Optional[List[float]] = None,
     **kwargs,
 ):
-    """
+    r"""
     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`.
 
@@ -134,19 +146,23 @@ def retrieve_timesteps(
         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`.
+            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`.
+            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
+            `num_inference_steps` and `sigmas` must be `None`.
+        sigmas (`List[float]`, *optional*):
+            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
+            `num_inference_steps` and `timesteps` 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 and sigmas is not None:
+        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
     if timesteps is not None:
         accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
         if not accepts_timesteps:
@@ -157,6 +173,16 @@ def retrieve_timesteps(
         scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
         timesteps = scheduler.timesteps
         num_inference_steps = len(timesteps)
+    elif sigmas is not None:
+        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accept_sigmas:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" sigmas schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
     else:
         scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
         timesteps = scheduler.timesteps
@@ -231,11 +257,8 @@ class StableDiffusionXLImg2ImgPipeline(
     _callback_tensor_inputs = [
         "latents",
         "prompt_embeds",
-        "negative_prompt_embeds",
         "add_text_embeds",
         "add_time_ids",
-        "negative_pooled_prompt_embeds",
-        "add_neg_time_ids",
     ]
 
     def __init__(
@@ -288,10 +311,10 @@ class StableDiffusionXLImg2ImgPipeline(
         do_classifier_free_guidance: bool = True,
         negative_prompt: Optional[str] = None,
         negative_prompt_2: Optional[str] = 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,
+        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,
         lora_scale: Optional[float] = None,
         clip_skip: Optional[int] = None,
     ):
@@ -317,17 +340,17 @@ class StableDiffusionXLImg2ImgPipeline(
             negative_prompt_2 (`str` or `List[str]`, *optional*):
                 The prompt or prompts not to guide the image generation to be sent to `tokenizer_2` and
                 `text_encoder_2`. If not defined, `negative_prompt` is used in both text-encoders
-            prompt_embeds (`torch.FloatTensor`, *optional*):
+            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.FloatTensor`, *optional*):
+            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.FloatTensor`, *optional*):
+            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.FloatTensor`, *optional*):
+            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.
@@ -626,14 +649,16 @@ class StableDiffusionXLImg2ImgPipeline(
         if denoising_start is None:
             init_timestep = min(int(num_inference_steps * strength), num_inference_steps)
             t_start = max(num_inference_steps - init_timestep, 0)
-        else:
-            t_start = 0
 
-        timesteps = self.scheduler.timesteps[t_start * self.scheduler.order :]
+            timesteps = self.scheduler.timesteps[t_start * self.scheduler.order :]
+            if hasattr(self.scheduler, "set_begin_index"):
+                self.scheduler.set_begin_index(t_start * self.scheduler.order)
+
+            return timesteps, num_inference_steps - t_start
 
-        # Strength is irrelevant if we directly request a timestep to start at;
-        # that is, strength is determined by the denoising_start instead.
-        if denoising_start is not None:
+        else:
+            # Strength is irrelevant if we directly request a timestep to start at;
+            # that is, strength is determined by the denoising_start instead.
             discrete_timestep_cutoff = int(
                 round(
                     self.scheduler.config.num_train_timesteps
@@ -641,22 +666,23 @@ class StableDiffusionXLImg2ImgPipeline(
                 )
             )
 
-            num_inference_steps = (timesteps < discrete_timestep_cutoff).sum().item()
+            num_inference_steps = (self.scheduler.timesteps < discrete_timestep_cutoff).sum().item()
             if self.scheduler.order == 2 and num_inference_steps % 2 == 0:
                 # if the scheduler is a 2nd order scheduler we might have to do +1
                 # because `num_inference_steps` might be even given that every timestep
                 # (except the highest one) is duplicated. If `num_inference_steps` is even it would
                 # mean that we cut the timesteps in the middle of the denoising step
-                # (between 1st and 2nd devirative) which leads to incorrect results. By adding 1
+                # (between 1st and 2nd derivative) which leads to incorrect results. By adding 1
                 # we ensure that the denoising process always ends after the 2nd derivate step of the scheduler
                 num_inference_steps = num_inference_steps + 1
 
             # because t_n+1 >= t_n, we slice the timesteps starting from the end
-            timesteps = timesteps[-num_inference_steps:]
+            t_start = len(self.scheduler.timesteps) - num_inference_steps
+            timesteps = self.scheduler.timesteps[t_start:]
+            if hasattr(self.scheduler, "set_begin_index"):
+                self.scheduler.set_begin_index(t_start)
             return timesteps, num_inference_steps
 
-        return timesteps, num_inference_steps - t_start
-
     def prepare_latents(
         self, image, timestep, batch_size, num_images_per_prompt, dtype, device, generator=None, add_noise=True
     ):
@@ -665,6 +691,12 @@ class StableDiffusionXLImg2ImgPipeline(
                 f"`image` has to be of type `torch.Tensor`, `PIL.Image.Image` or list but is {type(image)}"
             )
 
+        latents_mean = latents_std = None
+        if hasattr(self.vae.config, "latents_mean") and self.vae.config.latents_mean is not None:
+            latents_mean = torch.tensor(self.vae.config.latents_mean).view(1, 4, 1, 1)
+        if hasattr(self.vae.config, "latents_std") and self.vae.config.latents_std is not None:
+            latents_std = torch.tensor(self.vae.config.latents_std).view(1, 4, 1, 1)
+
         # Offload text encoder if `enable_model_cpu_offload` was enabled
         if hasattr(self, "final_offload_hook") and self.final_offload_hook is not None:
             self.text_encoder_2.to("cpu")
@@ -690,6 +722,13 @@ class StableDiffusionXLImg2ImgPipeline(
                 )
 
             elif isinstance(generator, list):
+                if image.shape[0] < batch_size and batch_size % image.shape[0] == 0:
+                    image = torch.cat([image] * (batch_size // image.shape[0]), dim=0)
+                elif image.shape[0] < batch_size and batch_size % image.shape[0] != 0:
+                    raise ValueError(
+                        f"Cannot duplicate `image` of batch size {image.shape[0]} to effective batch_size {batch_size} "
+                    )
+
                 init_latents = [
                     retrieve_latents(self.vae.encode(image[i : i + 1]), generator=generator[i])
                     for i in range(batch_size)
@@ -702,7 +741,12 @@ class StableDiffusionXLImg2ImgPipeline(
                 self.vae.to(dtype)
 
             init_latents = init_latents.to(dtype)
-            init_latents = self.vae.config.scaling_factor * init_latents
+            if latents_mean is not None and latents_std is not None:
+                latents_mean = latents_mean.to(device=device, dtype=dtype)
+                latents_std = latents_std.to(device=device, dtype=dtype)
+                init_latents = (init_latents - latents_mean) * self.vae.config.scaling_factor / latents_std
+            else:
+                init_latents = self.vae.config.scaling_factor * init_latents
 
         if batch_size > init_latents.shape[0] and batch_size % init_latents.shape[0] == 0:
             # expand init_latents for batch_size
@@ -754,6 +798,9 @@ class StableDiffusionXLImg2ImgPipeline(
     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]
@@ -763,7 +810,6 @@ class StableDiffusionXLImg2ImgPipeline(
                     f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters."
                 )
 
-            image_embeds = []
             for single_ip_adapter_image, image_proj_layer in zip(
                 ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers
             ):
@@ -771,36 +817,28 @@ class StableDiffusionXLImg2ImgPipeline(
                 single_image_embeds, single_negative_image_embeds = self.encode_image(
                     single_ip_adapter_image, device, 1, output_hidden_state
                 )
-                single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)
-                single_negative_image_embeds = torch.stack(
-                    [single_negative_image_embeds] * num_images_per_prompt, dim=0
-                )
 
+                image_embeds.append(single_image_embeds[None, :])
                 if do_classifier_free_guidance:
-                    single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
-                    single_image_embeds = single_image_embeds.to(device)
-
-                image_embeds.append(single_image_embeds)
+                    negative_image_embeds.append(single_negative_image_embeds[None, :])
         else:
-            repeat_dims = [1]
-            image_embeds = []
             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)
-                    single_image_embeds = single_image_embeds.repeat(
-                        num_images_per_prompt, *(repeat_dims * len(single_image_embeds.shape[1:]))
-                    )
-                    single_negative_image_embeds = single_negative_image_embeds.repeat(
-                        num_images_per_prompt, *(repeat_dims * len(single_negative_image_embeds.shape[1:]))
-                    )
-                    single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])
-                else:
-                    single_image_embeds = single_image_embeds.repeat(
-                        num_images_per_prompt, *(repeat_dims * len(single_image_embeds.shape[1:]))
-                    )
+                    negative_image_embeds.append(single_negative_image_embeds)
                 image_embeds.append(single_image_embeds)
 
-        return image_embeds
+        ip_adapter_image_embeds = []
+        for i, single_image_embeds in enumerate(image_embeds):
+            single_image_embeds = torch.cat([single_image_embeds] * num_images_per_prompt, dim=0)
+            if do_classifier_free_guidance:
+                single_negative_image_embeds = torch.cat([negative_image_embeds[i]] * num_images_per_prompt, dim=0)
+                single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds], dim=0)
+
+            single_image_embeds = single_image_embeds.to(device=device)
+            ip_adapter_image_embeds.append(single_image_embeds)
+
+        return ip_adapter_image_embeds
 
     def _get_add_time_ids(
         self,
@@ -862,8 +900,6 @@ class StableDiffusionXLImg2ImgPipeline(
             (
                 AttnProcessor2_0,
                 XFormersAttnProcessor,
-                LoRAXFormersAttnProcessor,
-                LoRAAttnProcessor2_0,
             ),
         )
         # if xformers or torch_2_0 is used attention block does not need
@@ -874,20 +910,22 @@ class StableDiffusionXLImg2ImgPipeline(
             self.vae.decoder.mid_block.to(dtype)
 
     # Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding
-    def get_guidance_scale_embedding(self, w, embedding_dim=512, dtype=torch.float32):
+    def get_guidance_scale_embedding(
+        self, w: torch.Tensor, embedding_dim: int = 512, dtype: torch.dtype = torch.float32
+    ) -> torch.Tensor:
         """
         See https://github.com/google-research/vdm/blob/dc27b98a554f65cdc654b800da5aa1846545d41b/model_vdm.py#L298
 
         Args:
-            timesteps (`torch.Tensor`):
-                generate embedding vectors at these timesteps
+            w (`torch.Tensor`):
+                Generate embedding vectors with a specified guidance scale to subsequently enrich timestep embeddings.
             embedding_dim (`int`, *optional*, defaults to 512):
-                dimension of the embeddings to generate
-            dtype:
-                data type of the generated embeddings
+                Dimension of the embeddings to generate.
+            dtype (`torch.dtype`, *optional*, defaults to `torch.float32`):
+                Data type of the generated embeddings.
 
         Returns:
-            `torch.FloatTensor`: Embedding vectors with shape `(len(timesteps), embedding_dim)`
+            `torch.Tensor`: Embedding vectors with shape `(len(w), embedding_dim)`.
         """
         assert len(w.shape) == 1
         w = w * 1000.0
@@ -951,6 +989,7 @@ class StableDiffusionXLImg2ImgPipeline(
         strength: float = 0.3,
         num_inference_steps: int = 50,
         timesteps: List[int] = None,
+        sigmas: List[float] = None,
         denoising_start: Optional[float] = None,
         denoising_end: Optional[float] = None,
         guidance_scale: float = 5.0,
@@ -959,13 +998,13 @@ class StableDiffusionXLImg2ImgPipeline(
         num_images_per_prompt: Optional[int] = 1,
         eta: float = 0.0,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        latents: Optional[torch.FloatTensor] = 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,
+        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.FloatTensor]] = 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,
@@ -979,7 +1018,9 @@ class StableDiffusionXLImg2ImgPipeline(
         aesthetic_score: float = 6.0,
         negative_aesthetic_score: float = 2.5,
         clip_skip: Optional[int] = None,
-        callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = 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,
     ):
@@ -993,7 +1034,7 @@ class StableDiffusionXLImg2ImgPipeline(
             prompt_2 (`str` or `List[str]`, *optional*):
                 The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
                 used in both text-encoders
-            image (`torch.FloatTensor` or `PIL.Image.Image` or `np.ndarray` or `List[torch.FloatTensor]` or `List[PIL.Image.Image]` or `List[np.ndarray]`):
+            image (`torch.Tensor` or `PIL.Image.Image` or `np.ndarray` or `List[torch.Tensor]` or `List[PIL.Image.Image]` or `List[np.ndarray]`):
                 The image(s) to modify with the pipeline.
             strength (`float`, *optional*, defaults to 0.3):
                 Conceptually, indicates how much to transform the reference `image`. Must be between 0 and 1. `image`
@@ -1009,6 +1050,10 @@ class StableDiffusionXLImg2ImgPipeline(
                 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.
+            sigmas (`List[float]`, *optional*):
+                Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in
+                their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed
+                will be used.
             denoising_start (`float`, *optional*):
                 When specified, indicates the fraction (between 0.0 and 1.0) of the total denoising process to be
                 bypassed before it is initiated. Consequently, the initial part of the denoising process is skipped and
@@ -1045,30 +1090,30 @@ class StableDiffusionXLImg2ImgPipeline(
             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.FloatTensor`, *optional*):
+            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 ge generated by sampling using the supplied random `generator`.
-            prompt_embeds (`torch.FloatTensor`, *optional*):
+            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.FloatTensor`, *optional*):
+            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.FloatTensor`, *optional*):
+            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.FloatTensor`, *optional*):
+            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.FloatTensor]`, *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.
+            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`.
@@ -1124,11 +1169,11 @@ class StableDiffusionXLImg2ImgPipeline(
             clip_skip (`int`, *optional*):
                 Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
                 the output of the pre-final layer will be used for computing the prompt embeddings.
-            callback_on_step_end (`Callable`, *optional*):
-                A function that calls at the end of each denoising steps during the inference. The function is called
-                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 (`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
@@ -1158,6 +1203,9 @@ class StableDiffusionXLImg2ImgPipeline(
                 "Passing `callback_steps` as an input argument to `__call__` is deprecated, consider use `callback_on_step_end`",
             )
 
+        if isinstance(callback_on_step_end, (PipelineCallback, MultiPipelineCallbacks)):
+            callback_on_step_end_tensor_inputs = callback_on_step_end.tensor_inputs
+
         # 1. Check inputs. Raise error if not correct
         self.check_inputs(
             prompt,
@@ -1224,7 +1272,9 @@ class StableDiffusionXLImg2ImgPipeline(
         def denoising_value_valid(dnv):
             return isinstance(dnv, float) and 0 < dnv < 1
 
-        timesteps, num_inference_steps = retrieve_timesteps(self.scheduler, num_inference_steps, device, timesteps)
+        timesteps, num_inference_steps = retrieve_timesteps(
+            self.scheduler, num_inference_steps, device, timesteps, sigmas
+        )
         timesteps, num_inference_steps = self.get_timesteps(
             num_inference_steps,
             strength,
@@ -1234,17 +1284,19 @@ class StableDiffusionXLImg2ImgPipeline(
         latent_timestep = timesteps[:1].repeat(batch_size * num_images_per_prompt)
 
         add_noise = True if self.denoising_start is None else False
+
         # 6. Prepare latent variables
-        latents = self.prepare_latents(
-            image,
-            latent_timestep,
-            batch_size,
-            num_images_per_prompt,
-            prompt_embeds.dtype,
-            device,
-            generator,
-            add_noise,
-        )
+        if latents is None:
+            latents = self.prepare_latents(
+                image,
+                latent_timestep,
+                batch_size,
+                num_images_per_prompt,
+                prompt_embeds.dtype,
+                device,
+                generator,
+                add_noise,
+            )
         # 7. Prepare extra step kwargs.
         extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)
 
@@ -1368,7 +1420,12 @@ class StableDiffusionXLImg2ImgPipeline(
                     noise_pred = rescale_noise_cfg(noise_pred, noise_pred_text, guidance_rescale=self.guidance_rescale)
 
                 # compute the previous noisy sample x_t -> x_t-1
+                latents_dtype = latents.dtype
                 latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
+                if latents.dtype != latents_dtype:
+                    if torch.backends.mps.is_available():
+                        # some platforms (eg. apple mps) misbehave due to a pytorch bug: https://github.com/pytorch/pytorch/pull/99272
+                        latents = latents.to(latents_dtype)
 
                 if callback_on_step_end is not None:
                     callback_kwargs = {}
@@ -1378,13 +1435,8 @@ class StableDiffusionXLImg2ImgPipeline(
 
                     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)
-                    add_neg_time_ids = callback_outputs.pop("add_neg_time_ids", add_neg_time_ids)
 
                 # call the callback, if provided
                 if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
@@ -1403,6 +1455,10 @@ class StableDiffusionXLImg2ImgPipeline(
             if needs_upcasting:
                 self.upcast_vae()
                 latents = latents.to(next(iter(self.vae.post_quant_conv.parameters())).dtype)
+            elif latents.dtype != self.vae.dtype:
+                if torch.backends.mps.is_available():
+                    # some platforms (eg. apple mps) misbehave due to a pytorch bug: https://github.com/pytorch/pytorch/pull/99272
+                    self.vae = self.vae.to(latents.dtype)
 
             # unscale/denormalize the latents
             # denormalize with the mean and std if available and not None