up

This reverts commit c6fc91031e.

examples/community/adaptive_mask_inpainting.py

@@ -489,7 +489,6 @@ class AdaptiveMaskInpaintPipeline(
         # We'll offload the last model manually.
         self.final_offload_hook = hook
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -651,7 +650,7 @@ class AdaptiveMaskInpaintPipeline(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -666,7 +665,7 @@ class AdaptiveMaskInpaintPipeline(
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/fresco_v2v.py

@@ -1380,7 +1380,7 @@ class FrescoV2VPipeline(StableDiffusionControlNetImg2ImgPipeline):
         flow_model.eval()
         self.flow_model = flow_model
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -1413,7 +1413,7 @@ class FrescoV2VPipeline(StableDiffusionControlNetImg2ImgPipeline):
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -1672,7 +1672,7 @@ class FrescoV2VPipeline(StableDiffusionControlNetImg2ImgPipeline):
 
         return image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -1687,7 +1687,7 @@ class FrescoV2VPipeline(StableDiffusionControlNetImg2ImgPipeline):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     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"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -1699,7 +1699,7 @@ class FrescoV2VPipeline(StableDiffusionControlNetImg2ImgPipeline):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/latent_consistency_interpolate.py

@@ -277,7 +277,7 @@ class LatentConsistencyModelWalkPipeline(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -459,7 +459,7 @@ class LatentConsistencyModelWalkPipeline(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -525,7 +525,7 @@ class LatentConsistencyModelWalkPipeline(
         assert emb.shape == (w.shape[0], embedding_dim)
         return emb
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/llm_grounded_diffusion.py

@@ -1195,7 +1195,7 @@ class LLMGroundedDiffusionPipeline(
     # Below are methods copied from StableDiffusionPipeline
     # The design choice of not inheriting from StableDiffusionPipeline is discussed here: https://github.com/huggingface/diffusers/pull/5993#issuecomment-1834258517
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -1228,7 +1228,7 @@ class LLMGroundedDiffusionPipeline(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -1426,7 +1426,7 @@ class LLMGroundedDiffusionPipeline(
         uncond_image_embeds = torch.zeros_like(image_embeds)
         return image_embeds, uncond_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -1441,7 +1441,7 @@ class LLMGroundedDiffusionPipeline(
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     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"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -1453,7 +1453,7 @@ class LLMGroundedDiffusionPipeline(
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -1534,17 +1534,17 @@ class LLMGroundedDiffusionPipeline(
         return emb
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.guidance_scale
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.guidance_scale
     def guidance_scale(self):
         return self._guidance_scale
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.guidance_rescale
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.guidance_rescale
     def guidance_rescale(self):
         return self._guidance_rescale
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.clip_skip
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.clip_skip
     def clip_skip(self):
         return self._clip_skip
 
@@ -1552,16 +1552,16 @@ class LLMGroundedDiffusionPipeline(
     # of the Imagen paper: https://huggingface.co/papers/2205.11487 . `guidance_scale = 1`
     # corresponds to doing no classifier free guidance.
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.do_classifier_free_guidance
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.do_classifier_free_guidance
     def do_classifier_free_guidance(self):
         return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.cross_attention_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.cross_attention_kwargs
     def cross_attention_kwargs(self):
         return self._cross_attention_kwargs
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.num_timesteps
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.num_timesteps
     def num_timesteps(self):
         return self._num_timesteps

examples/community/lpw_stable_diffusion_xl.py

@@ -503,7 +503,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -518,7 +518,7 @@ 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_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -532,7 +532,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -721,7 +721,7 @@ class SDXLLongPromptWeightingPipeline(
         # We'll offload the last model manually.
         self.final_offload_hook = hook
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -941,7 +941,7 @@ class SDXLLongPromptWeightingPipeline(
 
             return image_embeds, uncond_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/matryoshka.py

@@ -121,7 +121,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -136,7 +136,7 @@ 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
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

examples/community/mixture_tiling_sdxl.py

@@ -196,7 +196,7 @@ def _get_crops_coords_list(num_rows, num_cols, output_width):
     return result_list
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -223,7 +223,7 @@ 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
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -627,7 +627,7 @@ class StableDiffusionXLTilingPipeline(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/mod_controlnet_tile_sr_sdxl.py

@@ -244,7 +244,7 @@ def _tile2latent_indices(
     return latent_row_init, latent_row_end, latent_col_init, latent_col_end
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -394,7 +394,7 @@ class StableDiffusionXLControlNetTileSRPipeline(
         COSINE = "Cosine"
         GAUSSIAN = "Gaussian"
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -633,7 +633,7 @@ class StableDiffusionXLControlNetTileSRPipeline(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_animatediff_controlnet.py

@@ -194,7 +194,7 @@ class AnimateDiffControlNetPipeline(
             vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True, do_normalize=False
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -460,7 +460,7 @@ class AnimateDiffControlNetPipeline(
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_animatediff_img2video.py

@@ -180,7 +180,7 @@ def tensor2vid(video: torch.Tensor, processor, output_type="np"):
     return outputs
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -194,7 +194,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -311,7 +311,7 @@ class AnimateDiffImgToVideoPipeline(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -577,7 +577,7 @@ class AnimateDiffImgToVideoPipeline(
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_animatediff_ipex.py

@@ -165,7 +165,7 @@ class AnimateDiffPipelineIpex(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.video_processor = VideoProcessor(do_resize=False, vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -438,7 +438,7 @@ class AnimateDiffPipelineIpex(
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_controlnet_xl_kolors.py

@@ -113,7 +113,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -458,7 +458,7 @@ class KolorsControlNetPipeline(
 
             return image_embeds, uncond_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_controlnet_xl_kolors_img2img.py

@@ -133,7 +133,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -501,7 +501,7 @@ class KolorsControlNetImg2ImgPipeline(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for others.

examples/community/pipeline_controlnet_xl_kolors_inpaint.py

@@ -120,7 +120,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -134,7 +134,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -552,7 +552,7 @@ class KolorsControlNetInpaintPipeline(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_demofusion_sdxl.py

@@ -73,7 +73,7 @@ def gaussian_filter(latents, kernel_size=3, sigma=1.0):
     return blurred_latents
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -379,7 +379,7 @@ class DemoFusionSDXLPipeline(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_fabric.py

@@ -186,7 +186,7 @@ class FabricPipeline(DiffusionPipeline):
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,

examples/community/pipeline_faithdiff_stable_diffusion_xl.py

@@ -1078,7 +1078,7 @@ class LocalAttention:
         return out
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -1101,7 +1101,7 @@ 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_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -1125,7 +1125,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -1505,7 +1505,7 @@ class FaithDiffStableDiffusionXLPipeline(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_flux_differential_img2img.py

@@ -81,7 +81,7 @@ EXAMPLE_DOC_STRING = """
         """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -95,7 +95,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -109,7 +109,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -502,7 +502,7 @@ class FluxDifferentialImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._prepare_latent_image_ids
     def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         latent_image_ids = torch.zeros(height // 2, width // 2, 3)
         latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height // 2)[:, None]
@@ -517,7 +517,7 @@ class FluxDifferentialImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
         return latent_image_ids.to(device=device, dtype=dtype)
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._pack_latents
     def _pack_latents(latents, batch_size, num_channels_latents, height, width):
         latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
         latents = latents.permute(0, 2, 4, 1, 3, 5)

examples/community/pipeline_flux_kontext_multiple_images.py

@@ -126,7 +126,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -186,7 +186,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -567,7 +567,7 @@ class FluxKontextPipeline(
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._prepare_latent_image_ids
     def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         latent_image_ids = torch.zeros(height, width, 3)
         latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
@@ -582,7 +582,7 @@ class FluxKontextPipeline(
         return latent_image_ids.to(device=device, dtype=dtype)
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._pack_latents
     def _pack_latents(latents, batch_size, num_channels_latents, height, width):
         latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
         latents = latents.permute(0, 2, 4, 1, 3, 5)

examples/community/pipeline_flux_rf_inversion.py

@@ -89,7 +89,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -103,7 +103,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

examples/community/pipeline_flux_semantic_guidance.py

@@ -86,7 +86,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -100,7 +100,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -640,7 +640,7 @@ class FluxSemanticGuidancePipeline(
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._prepare_latent_image_ids
     def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         latent_image_ids = torch.zeros(height, width, 3)
         latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
@@ -655,7 +655,7 @@ class FluxSemanticGuidancePipeline(
         return latent_image_ids.to(device=device, dtype=dtype)
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._pack_latents
     def _pack_latents(latents, batch_size, num_channels_latents, height, width):
         latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
         latents = latents.permute(0, 2, 4, 1, 3, 5)

examples/community/pipeline_flux_with_cfg.py

@@ -65,7 +65,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -79,7 +79,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

examples/community/pipeline_hunyuandit_differential_img2img.py

@@ -146,7 +146,7 @@ def get_resize_crop_region_for_grid(src, tgt_size):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -161,7 +161,7 @@ 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_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor,
     generator: Optional[torch.Generator] = None,
@@ -177,7 +177,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -512,7 +512,7 @@ class HunyuanDiTDifferentialImg2ImgPipeline(DiffusionPipeline):
             negative_prompt_attention_mask,
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -527,7 +527,7 @@ class HunyuanDiTDifferentialImg2ImgPipeline(DiffusionPipeline):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_kolors_differential_img2img.py

@@ -66,7 +66,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -80,7 +80,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -458,7 +458,7 @@ class KolorsDifferentialImg2ImgPipeline(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -709,7 +709,7 @@ class KolorsDifferentialImg2ImgPipeline(
         add_time_ids = torch.tensor([add_time_ids], dtype=dtype)
         return add_time_ids
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.upcast_vae
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.upcast_vae
     def upcast_vae(self):
         dtype = self.vae.dtype
         self.vae.to(dtype=torch.float32)

examples/community/pipeline_kolors_inpainting.py

@@ -94,7 +94,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -243,7 +243,7 @@ def prepare_mask_and_masked_image(image, mask, height, width, return_image: bool
     return mask, masked_image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -257,7 +257,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -669,7 +669,7 @@ class KolorsInpaintPipeline(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_prompt2prompt.py

@@ -57,7 +57,7 @@ from diffusers.utils.torch_utils import randn_tensor
 logger = logging.get_logger(__name__)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -213,7 +213,7 @@ class Prompt2PromptPipeline(
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -246,7 +246,7 @@ class Prompt2PromptPipeline(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -430,7 +430,7 @@ class Prompt2PromptPipeline(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -445,7 +445,7 @@ class Prompt2PromptPipeline(
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_sdxl_style_aligned.py

@@ -311,7 +311,7 @@ class SharedAttentionProcessor(AttnProcessor2_0):
         return hidden_states
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -326,7 +326,7 @@ 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
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -371,7 +371,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -769,7 +769,7 @@ class StyleAlignedSDXLPipeline(
 
             return image_embeds, uncond_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_stable_diffusion_3_differential_img2img.py

@@ -66,7 +66,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -80,7 +80,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

examples/community/pipeline_stable_diffusion_3_instruct_pix2pix.py

@@ -86,7 +86,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -100,7 +100,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -114,7 +114,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

examples/community/pipeline_stable_diffusion_upscale_ldm3d.py

@@ -386,7 +386,7 @@ class StableDiffusionUpscaleLDM3DPipeline(
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_stable_diffusion_xl_attentive_eraser.py

@@ -297,7 +297,7 @@ class AAS_XL(AttentionBase):
         return out
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -439,7 +439,7 @@ def prepare_mask_and_masked_image(image, mask, height, width, return_image: bool
     return mask, masked_image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -453,7 +453,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -696,7 +696,7 @@ class StableDiffusionXL_AE_Pipeline(
 
         return image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -931,7 +931,7 @@ class StableDiffusionXL_AE_Pipeline(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_stable_diffusion_xl_controlnet_adapter.py

@@ -137,7 +137,7 @@ def _preprocess_adapter_image(image, height, width):
     return image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -237,7 +237,7 @@ class StableDiffusionXLControlNetAdapterPipeline(
             else 128
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -475,7 +475,7 @@ class StableDiffusionXLControlNetAdapterPipeline(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_stable_diffusion_xl_controlnet_adapter_inpaint.py

@@ -283,7 +283,7 @@ def prepare_mask_and_masked_image(image, mask, height, width, return_image: bool
     return mask, masked_image
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -384,7 +384,7 @@ class StableDiffusionXLControlNetAdapterInpaintPipeline(
             else 128
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -622,7 +622,7 @@ class StableDiffusionXLControlNetAdapterInpaintPipeline(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_stable_diffusion_xl_differential_img2img.py

@@ -88,7 +88,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -103,7 +103,7 @@ 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_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -117,7 +117,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -268,7 +268,7 @@ class StableDiffusionXLDifferentialImg2ImgPipeline(
         else:
             self.watermark = None
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -506,7 +506,7 @@ class StableDiffusionXLDifferentialImg2ImgPipeline(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_stable_diffusion_xl_ipex.py

@@ -98,7 +98,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -113,7 +113,7 @@ 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
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -520,7 +520,7 @@ class StableDiffusionXLPipelineIpex(
         uncond_image_embeds = torch.zeros_like(image_embeds)
         return image_embeds, uncond_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_stg_cogvideox.py

@@ -138,7 +138,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -409,7 +409,7 @@ class CogVideoXSTGPipeline(DiffusionPipeline, CogVideoXLoraLoaderMixin):
         frames = self.vae.decode(latents).sample
         return frames
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/pipeline_stg_hunyuan_video.py

@@ -142,7 +142,7 @@ def forward_without_stg(
     return hidden_states, encoder_hidden_states
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

examples/community/pipeline_stg_ltx.py

@@ -119,7 +119,7 @@ def forward_with_stg(
     return hidden_states
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -133,7 +133,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

examples/community/pipeline_stg_ltx_image2video.py

@@ -124,7 +124,7 @@ def forward_with_stg(
     return hidden_states
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -138,7 +138,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -198,7 +198,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):

examples/community/pipeline_stg_mochi.py

@@ -137,7 +137,7 @@ def linear_quadratic_schedule(num_steps, threshold_noise, linear_steps=None):
     return sigma_schedule
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

examples/community/pipline_flux_fill_controlnet_Inpaint.py

@@ -83,7 +83,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -97,7 +97,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -124,7 +124,7 @@ def retrieve_latents_fill(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -534,7 +534,7 @@ class FluxControlNetFillInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin,
             raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._prepare_latent_image_ids
     def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         latent_image_ids = torch.zeros(height, width, 3)
         latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
@@ -549,7 +549,7 @@ class FluxControlNetFillInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin,
         return latent_image_ids.to(device=device, dtype=dtype)
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._pack_latents
     def _pack_latents(latents, batch_size, num_channels_latents, height, width):
         latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
         latents = latents.permute(0, 2, 4, 1, 3, 5)

examples/community/run_onnx_controlnet.py

@@ -234,7 +234,7 @@ class OnnxStableDiffusionControlNetImg2ImgPipeline(DiffusionPipeline):
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         warnings.warn(
             "The decode_latents method is deprecated and will be removed in a future version. Please"
@@ -248,7 +248,7 @@ class OnnxStableDiffusionControlNetImg2ImgPipeline(DiffusionPipeline):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/run_tensorrt_controlnet.py

@@ -338,7 +338,7 @@ class TensorRTStableDiffusionControlNetImg2ImgPipeline(DiffusionPipeline):
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         warnings.warn(
             "The decode_latents method is deprecated and will be removed in a future version. Please"
@@ -352,7 +352,7 @@ class TensorRTStableDiffusionControlNetImg2ImgPipeline(DiffusionPipeline):
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/community/stable_diffusion_reference.py

@@ -358,7 +358,7 @@ class StableDiffusionReferencePipeline(
                 "Provide either `ip_adapter_image` or `ip_adapter_image_embeds`. Cannot leave both `ip_adapter_image` and `ip_adapter_image_embeds` defined."
             )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt: Union[str, List[str]],
@@ -408,7 +408,7 @@ class StableDiffusionReferencePipeline(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: Optional[str],
@@ -639,7 +639,7 @@ class StableDiffusionReferencePipeline(
         latents = latents * self.scheduler.init_noise_sigma
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(
         self, generator: Union[torch.Generator, List[torch.Generator]], eta: float
     ) -> Dict[str, Any]:
@@ -789,7 +789,7 @@ class StableDiffusionReferencePipeline(
         ref_image_latents = ref_image_latents.to(device=device, dtype=dtype)
         return ref_image_latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(
         self, image: Union[torch.Tensor, PIL.Image.Image], device: torch.device, dtype: torch.dtype
     ) -> Tuple[Union[torch.Tensor, PIL.Image.Image], Optional[bool]]:

examples/community/stable_diffusion_repaint.py

@@ -281,7 +281,7 @@ class StableDiffusionRepaintPipeline(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -427,7 +427,7 @@ class StableDiffusionRepaintPipeline(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is not None:
             safety_checker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(device)
@@ -438,7 +438,7 @@ class StableDiffusionRepaintPipeline(
             has_nsfw_concept = None
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -456,7 +456,7 @@ class StableDiffusionRepaintPipeline(
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     def decode_latents(self, latents):
         latents = 1 / self.vae.config.scaling_factor * latents
         image = self.vae.decode(latents).sample

examples/community/stable_diffusion_tensorrt_img2img.py

@@ -832,7 +832,7 @@ class TensorRTStableDiffusionImg2ImgPipeline(DiffusionPipeline):
         if "vae_encoder" in self.stages:
             self.models["vae_encoder"] = make_VAEEncoder(self.vae, **models_args)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(
         self, image: Union[torch.Tensor, PIL.Image.Image], device: torch.device, dtype: torch.dtype
     ) -> Tuple[Union[torch.Tensor, PIL.Image.Image], Optional[bool]]:

examples/community/stable_diffusion_tensorrt_inpaint.py

@@ -915,7 +915,7 @@ class TensorRTStableDiffusionInpaintPipeline(DiffusionPipeline):
 
         return outputs
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(
         self, image: Union[torch.Tensor, PIL.Image.Image], device: torch.device, dtype: torch.dtype
     ) -> Tuple[Union[torch.Tensor, PIL.Image.Image], Optional[bool]]:

examples/community/stable_diffusion_tensorrt_txt2img.py

@@ -788,7 +788,7 @@ class TensorRTStableDiffusionPipeline(DiffusionPipeline):
         latents = latents * self.scheduler.init_noise_sigma
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(
         self, image: Union[torch.Tensor, PIL.Image.Image], device: torch.device, dtype: torch.dtype
     ) -> Tuple[Union[torch.Tensor, PIL.Image.Image], Optional[bool]]:

examples/community/stable_diffusion_xl_controlnet_reference.py

@@ -87,7 +87,7 @@ def torch_dfs(model: torch.nn.Module):
     return result
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

examples/community/stable_diffusion_xl_reference.py

@@ -61,7 +61,7 @@ def torch_dfs(model: torch.nn.Module):
     return result
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.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
@@ -76,7 +76,7 @@ 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
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

examples/research_projects/anytext/anytext.py

@@ -1016,7 +1016,7 @@ class TextEmbeddingModule(ModelMixin, ConfigMixin):
         return new_string[:-nSpace]
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -1124,7 +1124,7 @@ class AuxiliaryLatentModule(ModelMixin, ConfigMixin):
         return new_string[:-nSpace]
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -1323,7 +1323,7 @@ class AnyTextPipeline(
                 return True
         return False
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -1356,7 +1356,7 @@ class AnyTextPipeline(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -1610,7 +1610,7 @@ class AnyTextPipeline(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -1625,7 +1625,7 @@ class AnyTextPipeline(
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     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"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -1637,7 +1637,7 @@ class AnyTextPipeline(
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/research_projects/pixart/pipeline_pixart_alpha_controlnet.py

@@ -185,7 +185,7 @@ def get_closest_hw(width, height, image_size):
     return width, height
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -457,7 +457,7 @@ class PixArtAlphaControlnetPipeline(DiffusionPipeline):
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

examples/research_projects/promptdiffusion/pipeline_prompt_diffusion.py

@@ -97,7 +97,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -285,7 +285,7 @@ class PromptDiffusionPipeline(
         )
         self.vae.disable_tiling()
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -318,7 +318,7 @@ class PromptDiffusionPipeline(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -525,7 +525,7 @@ class PromptDiffusionPipeline(
 
             return image_embeds, uncond_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -540,7 +540,7 @@ class PromptDiffusionPipeline(
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     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"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -552,7 +552,7 @@ class PromptDiffusionPipeline(
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/modular_pipelines/flux/before_denoise.py

@@ -19,6 +19,7 @@ import numpy as np
 import torch
 
 from ...pipelines import FluxPipeline
+from ...pipelines.flux.pipeline_flux_utils import calculate_shift
 from ...schedulers import FlowMatchEulerDiscreteScheduler
 from ...utils import logging
 from ...utils.torch_utils import randn_tensor
@@ -30,7 +31,7 @@ from .modular_pipeline import FluxModularPipeline
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -90,21 +91,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.15,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):

src/diffusers/modular_pipelines/flux/encoders.py

@@ -53,7 +53,7 @@ def prompt_clean(text):
     return text
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):

src/diffusers/modular_pipelines/qwenimage/before_denoise.py

@@ -26,7 +26,7 @@ from ..modular_pipeline_utils import ComponentSpec, InputParam, OutputParam
 from .modular_pipeline import QwenImageModularPipeline, QwenImagePachifier
 
 
-# Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.calculate_shift
+# Copied from diffusers.pipelines.qwenimage.pipeline_qwen_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -40,7 +40,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

src/diffusers/modular_pipelines/qwenimage/encoders.py

@@ -183,7 +183,7 @@ def get_qwen_prompt_embeds_edit_plus(
     return prompt_embeds, encoder_attention_mask
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):

src/diffusers/modular_pipelines/stable_diffusion_xl/before_denoise.py

@@ -43,7 +43,7 @@ logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 # configuration of guider is.
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -103,7 +103,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):

src/diffusers/modular_pipelines/stable_diffusion_xl/decoders.py

@@ -81,10 +81,7 @@ class StableDiffusionXLDecodeStep(ModularPipelineBlocks):
         components.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             components.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory

src/diffusers/modular_pipelines/stable_diffusion_xl/encoders.py

@@ -43,7 +43,7 @@ from .modular_pipeline import StableDiffusionXLModularPipeline
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):

src/diffusers/modular_pipelines/wan/before_denoise.py

@@ -126,7 +126,7 @@ def calculate_dimension_from_latents(
     return num_frames, height, width
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

src/diffusers/modular_pipelines/wan/encoders.py

@@ -102,7 +102,7 @@ def encode_image(
     return image_embeds.hidden_states[-2]
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):

src/diffusers/pipelines/allegro/pipeline_allegro.py

@@ -81,7 +81,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -347,7 +347,7 @@ class AllegroPipeline(DiffusionPipeline):
 
         return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/animatediff/pipeline_animatediff.py

@@ -152,7 +152,7 @@ class AnimateDiffPipeline(
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.video_processor = VideoProcessor(do_resize=False, vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -424,7 +424,7 @@ class AnimateDiffPipeline(
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/animatediff/pipeline_animatediff_controlnet.py

@@ -195,7 +195,7 @@ class AnimateDiffControlNetPipeline(
             vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True, do_normalize=False
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -468,7 +468,7 @@ class AnimateDiffControlNetPipeline(
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/animatediff/pipeline_animatediff_sdxl.py

@@ -120,7 +120,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -147,7 +147,7 @@ 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
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -325,7 +325,7 @@ class AnimateDiffSDXLPipeline(
             else 128
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt with num_images_per_prompt->num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt with num_images_per_prompt->num_videos_per_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -648,7 +648,7 @@ class AnimateDiffSDXLPipeline(
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/animatediff/pipeline_animatediff_sparsectrl.py

@@ -115,7 +115,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -204,7 +204,7 @@ class AnimateDiffSparseControlNetPipeline(
             vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True, do_normalize=False
         )
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
         self,
         prompt,
@@ -471,7 +471,7 @@ class AnimateDiffSparseControlNetPipeline(
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/animatediff/pipeline_animatediff_video2video.py

@@ -104,7 +104,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -118,7 +118,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -535,7 +535,7 @@ class AnimateDiffVideoToVideoPipeline(
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/animatediff/pipeline_animatediff_video2video_controlnet.py

@@ -122,7 +122,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -136,7 +136,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -567,7 +567,7 @@ class AnimateDiffVideoToVideoControlNetPipeline(
         video = video.float()
         return video
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/audioldm/pipeline_audioldm.py

@@ -258,7 +258,7 @@ class AudioLDMPipeline(DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusi
         waveform = waveform.cpu().float()
         return waveform
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/audioldm2/pipeline_audioldm2.py

@@ -719,7 +719,7 @@ class AudioLDM2Pipeline(DiffusionPipeline):
         audio = torch.index_select(audio, 0, indices.reshape(-1).cpu())
         return audio
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/aura_flow/pipeline_aura_flow.py

@@ -60,7 +60,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -404,17 +404,13 @@ class AuraFlowPipeline(DiffusionPipeline, AuraFlowLoraLoaderMixin):
 
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.upcast_vae
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.upcast_vae
     def upcast_vae(self):
         dtype = self.vae.dtype
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-                FusedAttnProcessor2_0,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor, FusedAttnProcessor2_0),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory

src/diffusers/pipelines/bria_fibo/pipeline_bria_fibo.py

@@ -367,7 +367,7 @@ class BriaFiboPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
         return latents
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._prepare_latent_image_ids
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._prepare_latent_image_ids
     def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
         latent_image_ids = torch.zeros(height, width, 3)
         latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
@@ -400,7 +400,7 @@ class BriaFiboPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
         return latents
 
     @staticmethod
-    # Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline._pack_latents
+    # Copied from diffusers.pipelines.flux.pipeline_flux_utils.FluxMixin._pack_latents
     def _pack_latents(latents, batch_size, num_channels_latents, height, width):
         latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
         latents = latents.permute(0, 2, 4, 1, 3, 5)

src/diffusers/pipelines/chroma/pipeline_chroma.py

@@ -74,7 +74,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -88,7 +88,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

src/diffusers/pipelines/chroma/pipeline_chroma_img2img.py

@@ -72,7 +72,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.flux.pipeline_flux.calculate_shift
+# Copied from diffusers.pipelines.flux.pipeline_flux_utils.calculate_shift
 def calculate_shift(
     image_seq_len,
     base_seq_len: int = 256,
@@ -86,7 +86,7 @@ def calculate_shift(
     return mu
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -100,7 +100,7 @@ def retrieve_latents(
         raise AttributeError("Could not access latents of provided encoder_output")
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

src/diffusers/pipelines/chronoedit/pipeline_chronoedit.py

@@ -111,7 +111,7 @@ def prompt_clean(text):
     return text
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):

src/diffusers/pipelines/cogvideo/pipeline_cogvideox.py

@@ -84,7 +84,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -355,7 +355,7 @@ class CogVideoXPipeline(DiffusionPipeline, CogVideoXLoraLoaderMixin):
         frames = self.vae.decode(latents).sample
         return frames
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/cogvideo/pipeline_cogvideox_fun_control.py

@@ -91,7 +91,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -394,7 +394,7 @@ class CogVideoXFunControlPipeline(DiffusionPipeline, CogVideoXLoraLoaderMixin):
         frames = self.vae.decode(latents).sample
         return frames
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/cogvideo/pipeline_cogvideox_image2video.py

@@ -83,7 +83,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -143,7 +143,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -434,7 +434,7 @@ class CogVideoXImageToVideoPipeline(DiffusionPipeline, CogVideoXLoraLoaderMixin)
 
         return timesteps, num_inference_steps - t_start
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/cogvideo/pipeline_cogvideox_video2video.py

@@ -92,7 +92,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -152,7 +152,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -414,7 +414,7 @@ class CogVideoXVideoToVideoPipeline(DiffusionPipeline, CogVideoXLoraLoaderMixin)
 
         return timesteps, num_inference_steps - t_start
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/cogview3/pipeline_cogview3plus.py

@@ -55,7 +55,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -315,7 +315,7 @@ class CogView3PlusPipeline(DiffusionPipeline):
         latents = latents * self.scheduler.init_noise_sigma
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/consisid/pipeline_consisid.py

@@ -173,7 +173,7 @@ def get_resize_crop_region_for_grid(src, tgt_width, tgt_height):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -233,7 +233,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -536,7 +536,7 @@ class ConsisIDPipeline(DiffusionPipeline, CogVideoXLoraLoaderMixin):
 
         return timesteps, num_inference_steps - t_start
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/controlnet/pipeline_controlnet.py

@@ -99,7 +99,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -261,7 +261,7 @@ class StableDiffusionControlNetPipeline(
         )
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -294,7 +294,7 @@ class StableDiffusionControlNetPipeline(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -548,7 +548,7 @@ class StableDiffusionControlNetPipeline(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -563,7 +563,7 @@ class StableDiffusionControlNetPipeline(
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     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"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -575,7 +575,7 @@ class StableDiffusionControlNetPipeline(
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/controlnet/pipeline_controlnet_img2img.py

@@ -99,7 +99,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -239,7 +239,7 @@ class StableDiffusionControlNetImg2ImgPipeline(
         )
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -272,7 +272,7 @@ class StableDiffusionControlNetImg2ImgPipeline(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -526,7 +526,7 @@ class StableDiffusionControlNetImg2ImgPipeline(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -541,7 +541,7 @@ class StableDiffusionControlNetImg2ImgPipeline(
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     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"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -553,7 +553,7 @@ class StableDiffusionControlNetImg2ImgPipeline(
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint.py

@@ -111,7 +111,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -245,7 +245,7 @@ class StableDiffusionControlNetInpaintPipeline(
         )
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -278,7 +278,7 @@ class StableDiffusionControlNetInpaintPipeline(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -532,7 +532,7 @@ class StableDiffusionControlNetInpaintPipeline(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -547,7 +547,7 @@ class StableDiffusionControlNetInpaintPipeline(
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     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"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -559,7 +559,7 @@ class StableDiffusionControlNetInpaintPipeline(
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint_sd_xl.py

@@ -73,7 +73,7 @@ else:
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -144,7 +144,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -291,7 +291,7 @@ class StableDiffusionXLControlNetInpaintPipeline(
         else:
             self.watermark = None
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -601,7 +601,7 @@ class StableDiffusionXLControlNetInpaintPipeline(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -1136,10 +1136,7 @@ class StableDiffusionXLControlNetInpaintPipeline(
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory

src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl.py

@@ -121,7 +121,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -299,7 +299,7 @@ class StableDiffusionXLControlNetPipeline(
 
         self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -609,7 +609,7 @@ class StableDiffusionXLControlNetPipeline(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -933,10 +933,7 @@ class StableDiffusionXLControlNetPipeline(
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory

src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl_img2img.py

@@ -150,7 +150,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -293,7 +293,7 @@ class StableDiffusionXLControlNetImg2ImgPipeline(
         self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
         self.register_to_config(requires_aesthetics_score=requires_aesthetics_score)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -603,7 +603,7 @@ class StableDiffusionXLControlNetImg2ImgPipeline(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -1048,10 +1048,7 @@ class StableDiffusionXLControlNetImg2ImgPipeline(
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory

src/diffusers/pipelines/controlnet/pipeline_controlnet_union_inpaint_sd_xl.py

@@ -78,7 +78,7 @@ else:
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -134,7 +134,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -280,7 +280,7 @@ class StableDiffusionXLControlNetUnionInpaintPipeline(
         else:
             self.watermark = None
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -590,7 +590,7 @@ class StableDiffusionXLControlNetUnionInpaintPipeline(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -1115,10 +1115,7 @@ class StableDiffusionXLControlNetUnionInpaintPipeline(
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory

src/diffusers/pipelines/controlnet/pipeline_controlnet_union_sd_xl.py

@@ -115,7 +115,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -289,7 +289,7 @@ class StableDiffusionXLControlNetUnionPipeline(
 
         self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -599,7 +599,7 @@ class StableDiffusionXLControlNetUnionPipeline(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -908,10 +908,7 @@ class StableDiffusionXLControlNetUnionPipeline(
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory

src/diffusers/pipelines/controlnet/pipeline_controlnet_union_sd_xl_img2img.py

@@ -175,7 +175,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):
@@ -309,7 +309,7 @@ class StableDiffusionXLControlNetUnionImg2ImgPipeline(
         self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
         self.register_to_config(requires_aesthetics_score=requires_aesthetics_score)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -619,7 +619,7 @@ class StableDiffusionXLControlNetUnionImg2ImgPipeline(
 
         return ip_adapter_image_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -1036,10 +1036,7 @@ class StableDiffusionXLControlNetUnionImg2ImgPipeline(
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory

src/diffusers/pipelines/controlnet_hunyuandit/pipeline_hunyuandit_controlnet.py

@@ -135,7 +135,7 @@ def get_resize_crop_region_for_grid(src, tgt_size):
     return (crop_top, crop_left), (crop_top + resize_height, crop_left + resize_width)
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.rescale_noise_cfg
 def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
     r"""
     Rescales `noise_cfg` tensor based on `guidance_rescale` to improve image quality and fix overexposure. Based on
@@ -440,7 +440,7 @@ class HunyuanDiTControlNetPipeline(DiffusionPipeline):
 
         return prompt_embeds, negative_prompt_embeds, prompt_attention_mask, negative_prompt_attention_mask
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -455,7 +455,7 @@ class HunyuanDiTControlNetPipeline(DiffusionPipeline):
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.

src/diffusers/pipelines/controlnet_sd3/pipeline_stable_diffusion_3_controlnet.py

@@ -80,7 +80,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

src/diffusers/pipelines/controlnet_sd3/pipeline_stable_diffusion_3_controlnet_inpainting.py

@@ -101,7 +101,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs.py

@@ -195,7 +195,7 @@ class StableDiffusionControlNetXSPipeline(
         )
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin._encode_prompt
     def _encode_prompt(
         self,
         prompt,
@@ -228,7 +228,7 @@ class StableDiffusionControlNetXSPipeline(
 
         return prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.encode_prompt
     def encode_prompt(
         self,
         prompt,
@@ -411,7 +411,7 @@ class StableDiffusionControlNetXSPipeline(
 
         return prompt_embeds, negative_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.run_safety_checker
     def run_safety_checker(self, image, device, dtype):
         if self.safety_checker is None:
             has_nsfw_concept = None
@@ -426,7 +426,7 @@ class StableDiffusionControlNetXSPipeline(
             )
         return image, has_nsfw_concept
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.decode_latents
     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"
         deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
@@ -438,7 +438,7 @@ class StableDiffusionControlNetXSPipeline(
         image = image.cpu().permute(0, 2, 3, 1).float().numpy()
         return image
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -616,27 +616,27 @@ class StableDiffusionControlNetXSPipeline(
         return latents
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.guidance_scale
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.guidance_scale
     def guidance_scale(self):
         return self._guidance_scale
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.clip_skip
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.clip_skip
     def clip_skip(self):
         return self._clip_skip
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.do_classifier_free_guidance
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.do_classifier_free_guidance
     def do_classifier_free_guidance(self):
         return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.cross_attention_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.cross_attention_kwargs
     def cross_attention_kwargs(self):
         return self._cross_attention_kwargs
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.num_timesteps
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.num_timesteps
     def num_timesteps(self):
         return self._num_timesteps
 

src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs_sd_xl.py

@@ -222,7 +222,7 @@ class StableDiffusionXLControlNetXSPipeline(
 
         self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
 
-    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
+    # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_utils.StableDiffusionXLMixin.encode_prompt
     def encode_prompt(
         self,
         prompt: str,
@@ -461,7 +461,7 @@ class StableDiffusionXLControlNetXSPipeline(
 
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
         # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
         # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
@@ -689,10 +689,7 @@ class StableDiffusionXLControlNetXSPipeline(
         self.vae.to(dtype=torch.float32)
         use_torch_2_0_or_xformers = isinstance(
             self.vae.decoder.mid_block.attentions[0].processor,
-            (
-                AttnProcessor2_0,
-                XFormersAttnProcessor,
-            ),
+            (AttnProcessor2_0, XFormersAttnProcessor),
         )
         # if xformers or torch_2_0 is used attention block does not need
         # to be in float32 which can save lots of memory
@@ -702,27 +699,27 @@ class StableDiffusionXLControlNetXSPipeline(
             self.vae.decoder.mid_block.to(dtype)
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.guidance_scale
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.guidance_scale
     def guidance_scale(self):
         return self._guidance_scale
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.clip_skip
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.clip_skip
     def clip_skip(self):
         return self._clip_skip
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.do_classifier_free_guidance
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.do_classifier_free_guidance
     def do_classifier_free_guidance(self):
         return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.cross_attention_kwargs
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.cross_attention_kwargs
     def cross_attention_kwargs(self):
         return self._cross_attention_kwargs
 
     @property
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.num_timesteps
+    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.SDMixin.num_timesteps
     def num_timesteps(self):
         return self._num_timesteps
 

src/diffusers/pipelines/cosmos/pipeline_cosmos2_text2image.py

@@ -72,7 +72,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

src/diffusers/pipelines/cosmos/pipeline_cosmos2_video2world.py

@@ -77,7 +77,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -137,7 +137,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):

src/diffusers/pipelines/cosmos/pipeline_cosmos_text2world.py

@@ -69,7 +69,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,

src/diffusers/pipelines/cosmos/pipeline_cosmos_video2world.py

@@ -98,7 +98,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_timesteps
 def retrieve_timesteps(
     scheduler,
     num_inference_steps: Optional[int] = None,
@@ -158,7 +158,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_utils.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
 ):