diffusers 0.27.1__py3-none-any.whl → 0.32.2__py3-none-any.whl

diffusers/pipelines/stable_diffusion/pipeline_stable_unclip.py

@@ -20,7 +20,7 @@ from transformers import CLIPTextModel, CLIPTextModelWithProjection, CLIPTokeniz
 from transformers.models.clip.modeling_clip import CLIPTextModelOutput
 
 from ...image_processor import VaeImageProcessor
-from ...loaders import LoraLoaderMixin, TextualInversionLoaderMixin
+from ...loaders import StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, PriorTransformer, UNet2DConditionModel
 from ...models.embeddings import get_timestep_embedding
 from ...models.lora import adjust_lora_scale_text_encoder
@@ -58,7 +58,9 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, LoraLoaderMixin):
+class StableUnCLIPPipeline(
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin
+):
     """
     Pipeline for text-to-image generation using stable unCLIP.
 
@@ -67,8 +69,8 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
 
     The pipeline also inherits the following loading methods:
         - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
-        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
-        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
 
     Args:
         prior_tokenizer ([`CLIPTokenizer`]):
@@ -76,7 +78,7 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
         prior_text_encoder ([`CLIPTextModelWithProjection`]):
             Frozen [`CLIPTextModelWithProjection`] text-encoder.
         prior ([`PriorTransformer`]):
-            The canonincal unCLIP prior to approximate the image embedding from the text embedding.
+            The canonical unCLIP prior to approximate the image embedding from the text embedding.
         prior_scheduler ([`KarrasDiffusionSchedulers`]):
             Scheduler used in the prior denoising process.
         image_normalizer ([`StableUnCLIPImageNormalizer`]):
@@ -257,8 +259,8 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
         num_images_per_prompt,
         do_classifier_free_guidance,
         negative_prompt=None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         lora_scale: Optional[float] = None,
         **kwargs,
     ):
@@ -290,8 +292,8 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
         num_images_per_prompt,
         do_classifier_free_guidance,
         negative_prompt=None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         lora_scale: Optional[float] = None,
         clip_skip: Optional[int] = None,
     ):
@@ -311,10 +313,10 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
                 The prompt or prompts not to guide the image generation. If not defined, one has to pass
                 `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
                 less than `1`).
-            prompt_embeds (`torch.FloatTensor`, *optional*):
+            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.
@@ -326,7 +328,7 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
         """
         # set lora scale so that monkey patched LoRA
         # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
+        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
             self._lora_scale = lora_scale
 
             # dynamically adjust the LoRA scale
@@ -458,9 +460,10 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
-        if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
-            # Retrieve the original scale by scaling back the LoRA layers
-            unscale_lora_layers(self.text_encoder, lora_scale)
+        if self.text_encoder is not None:
+            if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
+                # Retrieve the original scale by scaling back the LoRA layers
+                unscale_lora_layers(self.text_encoder, lora_scale)
 
         return prompt_embeds, negative_prompt_embeds
 
@@ -588,7 +591,7 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
         self,
         image_embeds: torch.Tensor,
         noise_level: int,
-        noise: Optional[torch.FloatTensor] = None,
+        noise: Optional[torch.Tensor] = None,
         generator: Optional[torch.Generator] = None,
     ):
         """
@@ -644,19 +647,19 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
         num_images_per_prompt: Optional[int] = 1,
         eta: float = 0.0,
         generator: Optional[torch.Generator] = None,
-        latents: Optional[torch.FloatTensor] = None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        latents: Optional[torch.Tensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
-        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
         noise_level: int = 0,
         # prior args
         prior_num_inference_steps: int = 25,
         prior_guidance_scale: float = 4.0,
-        prior_latents: Optional[torch.FloatTensor] = None,
+        prior_latents: Optional[torch.Tensor] = None,
         clip_skip: Optional[int] = None,
     ):
         """
@@ -686,14 +689,14 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
             generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
                 A [`torch.Generator`](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 is 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 (prompt weighting). If not
                 provided, text embeddings are generated from the `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 (prompt weighting). If
                 not provided, `negative_prompt_embeds` are generated from the `negative_prompt` input argument.
             output_type (`str`, *optional*, defaults to `"pil"`):
@@ -702,7 +705,7 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
                 Whether or not to return a [`~pipelines.ImagePipelineOutput`] instead of a plain tuple.
             callback (`Callable`, *optional*):
                 A function that calls every `callback_steps` steps during inference. The function is called with the
-                following arguments: `callback(step: int, timestep: int, latents: torch.FloatTensor)`.
+                following arguments: `callback(step: int, timestep: int, latents: torch.Tensor)`.
             callback_steps (`int`, *optional*, defaults to 1):
                 The frequency at which the `callback` function is called. If not specified, the callback is called at
                 every step.
@@ -718,7 +721,7 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
             prior_guidance_scale (`float`, *optional*, defaults to 4.0):
                 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`.
-            prior_latents (`torch.FloatTensor`, *optional*):
+            prior_latents (`torch.Tensor`, *optional*):
                 Pre-generated noisy latents sampled from a Gaussian distribution, to be used as inputs for image
                 embedding generation in the prior denoising process. Can be used to tweak the same generation with
                 different prompts. If not provided, a latents tensor is generated by sampling using the supplied random
@@ -876,7 +879,12 @@ class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInver
 
         # 11. Prepare latent variables
         num_channels_latents = self.unet.config.in_channels
-        shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
+        shape = (
+            batch_size,
+            num_channels_latents,
+            int(height) // self.vae_scale_factor,
+            int(width) // self.vae_scale_factor,
+        )
         latents = self.prepare_latents(
             shape=shape,
             dtype=prompt_embeds.dtype,

diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py

@@ -20,7 +20,7 @@ import torch
 from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
 
 from ...image_processor import VaeImageProcessor
-from ...loaders import LoraLoaderMixin, TextualInversionLoaderMixin
+from ...loaders import StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, UNet2DConditionModel
 from ...models.embeddings import get_timestep_embedding
 from ...models.lora import adjust_lora_scale_text_encoder
@@ -51,8 +51,8 @@ EXAMPLE_DOC_STRING = """
         >>> from diffusers import StableUnCLIPImg2ImgPipeline
 
         >>> pipe = StableUnCLIPImg2ImgPipeline.from_pretrained(
-        ...     "fusing/stable-unclip-2-1-l-img2img", torch_dtype=torch.float16
-        ... )  # TODO update model path
+        ...     "stabilityai/stable-diffusion-2-1-unclip-small", torch_dtype=torch.float16
+        ... )
         >>> pipe = pipe.to("cuda")
 
         >>> url = "https://raw.githubusercontent.com/CompVis/stable-diffusion/main/assets/stable-samples/img2img/sketch-mountains-input.jpg"
@@ -63,14 +63,14 @@ EXAMPLE_DOC_STRING = """
 
         >>> prompt = "A fantasy landscape, trending on artstation"
 
-        >>> images = pipe(prompt, init_image).images
+        >>> images = pipe(init_image, prompt).images
         >>> images[0].save("fantasy_landscape.png")
         ```
 """
 
 
 class StableUnCLIPImg2ImgPipeline(
-    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, LoraLoaderMixin
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin
 ):
     """
     Pipeline for text-guided image-to-image generation using stable unCLIP.
@@ -80,8 +80,8 @@ class StableUnCLIPImg2ImgPipeline(
 
     The pipeline also inherits the following loading methods:
         - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
-        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
-        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
 
     Args:
         feature_extractor ([`CLIPImageProcessor`]):
@@ -166,8 +166,8 @@ class StableUnCLIPImg2ImgPipeline(
         num_images_per_prompt,
         do_classifier_free_guidance,
         negative_prompt=None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         lora_scale: Optional[float] = None,
         **kwargs,
     ):
@@ -254,8 +254,8 @@ class StableUnCLIPImg2ImgPipeline(
         num_images_per_prompt,
         do_classifier_free_guidance,
         negative_prompt=None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         lora_scale: Optional[float] = None,
         clip_skip: Optional[int] = None,
     ):
@@ -275,10 +275,10 @@ class StableUnCLIPImg2ImgPipeline(
                 The prompt or prompts not to guide the image generation. If not defined, one has to pass
                 `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
                 less than `1`).
-            prompt_embeds (`torch.FloatTensor`, *optional*):
+            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.
@@ -290,7 +290,7 @@ class StableUnCLIPImg2ImgPipeline(
         """
         # set lora scale so that monkey patched LoRA
         # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
+        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
             self._lora_scale = lora_scale
 
             # dynamically adjust the LoRA scale
@@ -422,9 +422,10 @@ class StableUnCLIPImg2ImgPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
-        if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
-            # Retrieve the original scale by scaling back the LoRA layers
-            unscale_lora_layers(self.text_encoder, lora_scale)
+        if self.text_encoder is not None:
+            if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
+                # Retrieve the original scale by scaling back the LoRA layers
+                unscale_lora_layers(self.text_encoder, lora_scale)
 
         return prompt_embeds, negative_prompt_embeds
 
@@ -537,13 +538,18 @@ class StableUnCLIPImg2ImgPipeline(
                 and not isinstance(image, list)
             ):
                 raise ValueError(
-                    "`image` has to be of type `torch.FloatTensor` or `PIL.Image.Image` or `List[PIL.Image.Image]` but is"
+                    "`image` has to be of type `torch.Tensor` or `PIL.Image.Image` or `List[PIL.Image.Image]` but is"
                     f" {type(image)}"
                 )
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
     def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
-        shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
+        shape = (
+            batch_size,
+            num_channels_latents,
+            int(height) // self.vae_scale_factor,
+            int(width) // self.vae_scale_factor,
+        )
         if isinstance(generator, list) and len(generator) != batch_size:
             raise ValueError(
                 f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
@@ -564,7 +570,7 @@ class StableUnCLIPImg2ImgPipeline(
         self,
         image_embeds: torch.Tensor,
         noise_level: int,
-        noise: Optional[torch.FloatTensor] = None,
+        noise: Optional[torch.Tensor] = None,
         generator: Optional[torch.Generator] = None,
     ):
         """
@@ -610,7 +616,7 @@ class StableUnCLIPImg2ImgPipeline(
     @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
         self,
-        image: Union[torch.FloatTensor, PIL.Image.Image] = None,
+        image: Union[torch.Tensor, PIL.Image.Image] = None,
         prompt: Union[str, List[str]] = None,
         height: Optional[int] = None,
         width: Optional[int] = None,
@@ -620,16 +626,16 @@ class StableUnCLIPImg2ImgPipeline(
         num_images_per_prompt: Optional[int] = 1,
         eta: float = 0.0,
         generator: Optional[torch.Generator] = None,
-        latents: Optional[torch.FloatTensor] = None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        latents: Optional[torch.Tensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
-        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: int = 1,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
         noise_level: int = 0,
-        image_embeds: Optional[torch.FloatTensor] = None,
+        image_embeds: Optional[torch.Tensor] = None,
         clip_skip: Optional[int] = None,
     ):
         r"""
@@ -639,7 +645,7 @@ class StableUnCLIPImg2ImgPipeline(
             prompt (`str` or `List[str]`, *optional*):
                 The prompt or prompts to guide the image generation. If not defined, either `prompt_embeds` will be
                 used or prompt is initialized to `""`.
-            image (`torch.FloatTensor` or `PIL.Image.Image`):
+            image (`torch.Tensor` or `PIL.Image.Image`):
                 `Image` or tensor representing an image batch. The image is encoded to its CLIP embedding which the
                 `unet` is conditioned on. The image is _not_ encoded by the `vae` and then used as the latents in the
                 denoising process like it is in the standard Stable Diffusion text-guided image variation process.
@@ -664,14 +670,14 @@ class StableUnCLIPImg2ImgPipeline(
             generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
                 A [`torch.Generator`](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 is 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 (prompt weighting). If not
                 provided, text embeddings are generated from the `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 (prompt weighting). If
                 not provided, `negative_prompt_embeds` are generated from the `negative_prompt` input argument.
             output_type (`str`, *optional*, defaults to `"pil"`):
@@ -680,7 +686,7 @@ class StableUnCLIPImg2ImgPipeline(
                 Whether or not to return a [`~pipelines.ImagePipelineOutput`] instead of a plain tuple.
             callback (`Callable`, *optional*):
                 A function that calls every `callback_steps` steps during inference. The function is called with the
-                following arguments: `callback(step: int, timestep: int, latents: torch.FloatTensor)`.
+                following arguments: `callback(step: int, timestep: int, latents: torch.Tensor)`.
             callback_steps (`int`, *optional*, defaults to 1):
                 The frequency at which the `callback` function is called. If not specified, the callback is called at
                 every step.
@@ -690,7 +696,7 @@ class StableUnCLIPImg2ImgPipeline(
             noise_level (`int`, *optional*, defaults to `0`):
                 The amount of noise to add to the image embeddings. A higher `noise_level` increases the variance in
                 the final un-noised images. See [`StableUnCLIPPipeline.noise_image_embeddings`] for more details.
-            image_embeds (`torch.FloatTensor`, *optional*):
+            image_embeds (`torch.Tensor`, *optional*):
                 Pre-generated CLIP embeddings to condition the `unet` on. These latents are not used in the denoising
                 process. If you want to provide pre-generated latents, pass them to `__call__` as `latents`.
             clip_skip (`int`, *optional*):
@@ -781,16 +787,17 @@ class StableUnCLIPImg2ImgPipeline(
 
         # 6. Prepare latent variables
         num_channels_latents = self.unet.config.in_channels
-        latents = self.prepare_latents(
-            batch_size=batch_size,
-            num_channels_latents=num_channels_latents,
-            height=height,
-            width=width,
-            dtype=prompt_embeds.dtype,
-            device=device,
-            generator=generator,
-            latents=latents,
-        )
+        if latents is None:
+            latents = self.prepare_latents(
+                batch_size=batch_size,
+                num_channels_latents=num_channels_latents,
+                height=height,
+                width=width,
+                dtype=prompt_embeds.dtype,
+                device=device,
+                generator=generator,
+                latents=latents,
+            )
 
         # 7. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline
         extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)

diffusers/pipelines/stable_diffusion/safety_checker.py

@@ -31,6 +31,7 @@ def cosine_distance(image_embeds, text_embeds):
 
 class StableDiffusionSafetyChecker(PreTrainedModel):
     config_class = CLIPConfig
+    main_input_name = "clip_input"
 
     _no_split_modules = ["CLIPEncoderLayer"]
 
@@ -99,7 +100,7 @@ class StableDiffusionSafetyChecker(PreTrainedModel):
         return images, has_nsfw_concepts
 
     @torch.no_grad()
-    def forward_onnx(self, clip_input: torch.FloatTensor, images: torch.FloatTensor):
+    def forward_onnx(self, clip_input: torch.Tensor, images: torch.Tensor):
         pooled_output = self.vision_model(clip_input)[1]  # pooled_output
         image_embeds = self.visual_projection(pooled_output)
 

diffusers/pipelines/stable_diffusion_3/__init__.py

@@ -0,0 +1,54 @@
+from typing import TYPE_CHECKING
+
+from ...utils import (
+    DIFFUSERS_SLOW_IMPORT,
+    OptionalDependencyNotAvailable,
+    _LazyModule,
+    get_objects_from_module,
+    is_flax_available,
+    is_torch_available,
+    is_transformers_available,
+)
+
+
+_dummy_objects = {}
+_additional_imports = {}
+_import_structure = {"pipeline_output": ["StableDiffusion3PipelineOutput"]}
+
+try:
+    if not (is_transformers_available() and is_torch_available()):
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    from ...utils import dummy_torch_and_transformers_objects  # noqa F403
+
+    _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_objects))
+else:
+    _import_structure["pipeline_stable_diffusion_3"] = ["StableDiffusion3Pipeline"]
+    _import_structure["pipeline_stable_diffusion_3_img2img"] = ["StableDiffusion3Img2ImgPipeline"]
+    _import_structure["pipeline_stable_diffusion_3_inpaint"] = ["StableDiffusion3InpaintPipeline"]
+
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
+    try:
+        if not (is_transformers_available() and is_torch_available()):
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        from ...utils.dummy_torch_and_transformers_objects import *  # noqa F403
+    else:
+        from .pipeline_stable_diffusion_3 import StableDiffusion3Pipeline
+        from .pipeline_stable_diffusion_3_img2img import StableDiffusion3Img2ImgPipeline
+        from .pipeline_stable_diffusion_3_inpaint import StableDiffusion3InpaintPipeline
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(
+        __name__,
+        globals()["__file__"],
+        _import_structure,
+        module_spec=__spec__,
+    )
+
+    for name, value in _dummy_objects.items():
+        setattr(sys.modules[__name__], name, value)
+    for name, value in _additional_imports.items():
+        setattr(sys.modules[__name__], name, value)

diffusers/pipelines/stable_diffusion_3/pipeline_output.py

@@ -0,0 +1,21 @@
+from dataclasses import dataclass
+from typing import List, Union
+
+import numpy as np
+import PIL.Image
+
+from ...utils import BaseOutput
+
+
+@dataclass
+class StableDiffusion3PipelineOutput(BaseOutput):
+    """
+    Output class for Stable Diffusion pipelines.
+
+    Args:
+        images (`List[PIL.Image.Image]` or `np.ndarray`)
+            List of denoised PIL images of length `batch_size` or numpy array of shape `(batch_size, height, width,
+            num_channels)`. PIL images or numpy array present the denoised images of the diffusion pipeline.
+    """
+
+    images: Union[List[PIL.Image.Image], np.ndarray]