diffusers 0.26.3__py3-none-any.whl → 0.27.0__py3-none-any.whl

diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_instruct_pix2pix.py

@@ -1,4 +1,4 @@
-# Copyright 2023 The InstructPix2Pix Authors and The HuggingFace Team. All rights reserved.
+# Copyright 2024 The InstructPix2Pix Authors and The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -26,7 +26,7 @@ from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import PIL_INTERPOLATION, deprecate, logging
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import DiffusionPipeline
+from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from . import StableDiffusionPipelineOutput
 from .safety_checker import StableDiffusionSafetyChecker
 
@@ -73,7 +73,7 @@ def retrieve_latents(
 
 
 class StableDiffusionInstructPix2PixPipeline(
-    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, IPAdapterMixin
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, LoraLoaderMixin, IPAdapterMixin
 ):
     r"""
     Pipeline for pixel-level image editing by following text instructions (based on Stable Diffusion).
@@ -323,8 +323,6 @@ class StableDiffusionInstructPix2PixPipeline(
             batch_size = prompt_embeds.shape[0]
 
         device = self._execution_device
-        # check if scheduler is in sigmas space
-        scheduler_is_in_sigma_space = hasattr(self.scheduler, "sigmas")
 
         # 2. Encode input prompt
         prompt_embeds = self._encode_prompt(
@@ -411,15 +409,6 @@ class StableDiffusionInstructPix2PixPipeline(
                     return_dict=False,
                 )[0]
 
-                # Hack:
-                # For karras style schedulers the model does classifer free guidance using the
-                # predicted_original_sample instead of the noise_pred. So we need to compute the
-                # predicted_original_sample here if we are using a karras style scheduler.
-                if scheduler_is_in_sigma_space:
-                    step_index = (self.scheduler.timesteps == t).nonzero()[0].item()
-                    sigma = self.scheduler.sigmas[step_index]
-                    noise_pred = latent_model_input - sigma * noise_pred
-
                 # perform guidance
                 if self.do_classifier_free_guidance:
                     noise_pred_text, noise_pred_image, noise_pred_uncond = noise_pred.chunk(3)
@@ -429,15 +418,6 @@ class StableDiffusionInstructPix2PixPipeline(
                         + self.image_guidance_scale * (noise_pred_image - noise_pred_uncond)
                     )
 
-                # Hack:
-                # For karras style schedulers the model does classifer free guidance using the
-                # predicted_original_sample instead of the noise_pred. But the scheduler.step function
-                # expects the noise_pred and computes the predicted_original_sample internally. So we
-                # need to overwrite the noise_pred here such that the value of the computed
-                # predicted_original_sample is correct.
-                if scheduler_is_in_sigma_space:
-                    noise_pred = (noise_pred - latents) / (-sigma)
-
                 # compute the previous noisy sample x_t -> x_t-1
                 latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
 
@@ -523,7 +503,7 @@ class StableDiffusionInstructPix2PixPipeline(
             batch_size = prompt_embeds.shape[0]
 
         if prompt_embeds is None:
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
 
@@ -553,13 +533,15 @@ class StableDiffusionInstructPix2PixPipeline(
             else:
                 attention_mask = None
 
-            prompt_embeds = self.text_encoder(
-                text_input_ids.to(device),
-                attention_mask=attention_mask,
-            )
+            prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)
             prompt_embeds = prompt_embeds[0]
 
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
+        if self.text_encoder is not None:
+            prompt_embeds_dtype = self.text_encoder.dtype
+        else:
+            prompt_embeds_dtype = self.unet.dtype
+
+        prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
 
         bs_embed, seq_len, _ = prompt_embeds.shape
         # duplicate text embeddings for each generation per prompt, using mps friendly method
@@ -587,7 +569,7 @@ class StableDiffusionInstructPix2PixPipeline(
             else:
                 uncond_tokens = negative_prompt
 
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
 
@@ -615,7 +597,7 @@ class StableDiffusionInstructPix2PixPipeline(
             # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
             seq_len = negative_prompt_embeds.shape[1]
 
-            negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
+            negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
 
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
@@ -805,34 +787,6 @@ class StableDiffusionInstructPix2PixPipeline(
 
         return image_latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_freeu
-    def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
-        r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
-
-        The suffixes after the scaling factors represent the stages where they are being applied.
-
-        Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
-        that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
-
-        Args:
-            s1 (`float`):
-                Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            s2 (`float`):
-                Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
-            b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
-        """
-        if not hasattr(self, "unet"):
-            raise ValueError("The pipeline must have `unet` for using FreeU.")
-        self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_freeu
-    def disable_freeu(self):
-        """Disables the FreeU mechanism if enabled."""
-        self.unet.disable_freeu()
-
     @property
     def guidance_scale(self):
         return self._guidance_scale

diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_latent_upscale.py

@@ -1,4 +1,4 @@
-# Copyright 2023 The HuggingFace Team. All rights reserved.
+# Copyright 2024 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -27,7 +27,7 @@ from ...models import AutoencoderKL, UNet2DConditionModel
 from ...schedulers import EulerDiscreteScheduler
 from ...utils import deprecate, logging
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
+from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput, StableDiffusionMixin
 
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
@@ -60,7 +60,7 @@ def preprocess(image):
     return image
 
 
-class StableDiffusionLatentUpscalePipeline(DiffusionPipeline, FromSingleFileMixin):
+class StableDiffusionLatentUpscalePipeline(DiffusionPipeline, StableDiffusionMixin, FromSingleFileMixin):
     r"""
     Pipeline for upscaling Stable Diffusion output image resolution by a factor of 2.
 
@@ -258,34 +258,6 @@ class StableDiffusionLatentUpscalePipeline(DiffusionPipeline, FromSingleFileMixi
         latents = latents * self.scheduler.init_noise_sigma
         return latents
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_freeu
-    def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
-        r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
-
-        The suffixes after the scaling factors represent the stages where they are being applied.
-
-        Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
-        that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
-
-        Args:
-            s1 (`float`):
-                Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            s2 (`float`):
-                Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
-            b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
-        """
-        if not hasattr(self, "unet"):
-            raise ValueError("The pipeline must have `unet` for using FreeU.")
-        self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_freeu
-    def disable_freeu(self):
-        """Disables the FreeU mechanism if enabled."""
-        self.unet.disable_freeu()
-
     @torch.no_grad()
     def __call__(
         self,

diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_upscale.py

@@ -1,4 +1,4 @@
-# Copyright 2023 The HuggingFace Team. All rights reserved.
+# Copyright 2024 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -34,7 +34,7 @@ from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import DDPMScheduler, KarrasDiffusionSchedulers
 from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import DiffusionPipeline
+from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from . import StableDiffusionPipelineOutput
 
 
@@ -68,7 +68,7 @@ def preprocess(image):
 
 
 class StableDiffusionUpscalePipeline(
-    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin
 ):
     r"""
     Pipeline for text-guided image super-resolution using Stable Diffusion 2.
@@ -262,7 +262,7 @@ class StableDiffusionUpscalePipeline(
             batch_size = prompt_embeds.shape[0]
 
         if prompt_embeds is None:
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
 
@@ -344,7 +344,7 @@ class StableDiffusionUpscalePipeline(
             else:
                 uncond_tokens = negative_prompt
 
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
 
@@ -530,34 +530,6 @@ class StableDiffusionUpscalePipeline(
             self.vae.decoder.conv_in.to(dtype)
             self.vae.decoder.mid_block.to(dtype)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_freeu
-    def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
-        r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
-
-        The suffixes after the scaling factors represent the stages where they are being applied.
-
-        Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
-        that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
-
-        Args:
-            s1 (`float`):
-                Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            s2 (`float`):
-                Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
-                mitigate "oversmoothing effect" in the enhanced denoising process.
-            b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
-            b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
-        """
-        if not hasattr(self, "unet"):
-            raise ValueError("The pipeline must have `unet` for using FreeU.")
-        self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_freeu
-    def disable_freeu(self):
-        """Disables the FreeU mechanism if enabled."""
-        self.unet.disable_freeu()
-
     @torch.no_grad()
     def __call__(
         self,

diffusers/pipelines/stable_diffusion/pipeline_stable_unclip.py

@@ -1,4 +1,4 @@
-# Copyright 2023 The HuggingFace Team. All rights reserved.
+# Copyright 2024 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -34,7 +34,7 @@ from ...utils import (
     unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
+from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput, StableDiffusionMixin
 from .stable_unclip_image_normalizer import StableUnCLIPImageNormalizer
 
 
@@ -58,7 +58,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-class StableUnCLIPPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin):
+class StableUnCLIPPipeline(DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, LoraLoaderMixin):
     """
     Pipeline for text-to-image generation using stable unCLIP.
 
@@ -155,22 +155,6 @@ class StableUnCLIPPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
     # Copied from diffusers.pipelines.unclip.pipeline_unclip.UnCLIPPipeline._encode_prompt with _encode_prompt->_encode_prior_prompt, tokenizer->prior_tokenizer, text_encoder->prior_text_encoder
     def _encode_prior_prompt(
         self,
@@ -359,7 +343,7 @@ class StableUnCLIPPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
             batch_size = prompt_embeds.shape[0]
 
         if prompt_embeds is None:
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
 
@@ -441,7 +425,7 @@ class StableUnCLIPPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraL
             else:
                 uncond_tokens = negative_prompt
 
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
 

diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py

@@ -1,4 +1,4 @@
-# Copyright 2023 The HuggingFace Team. All rights reserved.
+# Copyright 2024 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -34,7 +34,7 @@ from ...utils import (
     unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
+from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput, StableDiffusionMixin
 from .stable_unclip_image_normalizer import StableUnCLIPImageNormalizer
 
 
@@ -69,7 +69,9 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-class StableUnCLIPImg2ImgPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin):
+class StableUnCLIPImg2ImgPipeline(
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, LoraLoaderMixin
+):
     """
     Pipeline for text-guided image-to-image generation using stable unCLIP.
 
@@ -156,22 +158,6 @@ class StableUnCLIPImg2ImgPipeline(DiffusionPipeline, TextualInversionLoaderMixin
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
 
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
     def _encode_prompt(
         self,
@@ -321,7 +307,7 @@ class StableUnCLIPImg2ImgPipeline(DiffusionPipeline, TextualInversionLoaderMixin
             batch_size = prompt_embeds.shape[0]
 
         if prompt_embeds is None:
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
 
@@ -403,7 +389,7 @@ class StableUnCLIPImg2ImgPipeline(DiffusionPipeline, TextualInversionLoaderMixin
             else:
                 uncond_tokens = negative_prompt
 
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
 

diffusers/pipelines/stable_diffusion/safety_checker.py

@@ -1,4 +1,4 @@
-# Copyright 2023 The HuggingFace Team. All rights reserved.
+# Copyright 2024 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.

diffusers/pipelines/stable_diffusion/safety_checker_flax.py

@@ -1,4 +1,4 @@
-# Copyright 2023 The HuggingFace Team. All rights reserved.
+# Copyright 2024 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.

diffusers/pipelines/stable_diffusion/stable_unclip_image_normalizer.py

@@ -1,4 +1,4 @@
-# Copyright 2023 The HuggingFace Team. All rights reserved.
+# Copyright 2024 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.

diffusers/pipelines/stable_diffusion_attend_and_excite/pipeline_stable_diffusion_attend_and_excite.py

@@ -1,4 +1,4 @@
-# Copyright 2023 The HuggingFace Team. All rights reserved.
+# Copyright 2024 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -36,7 +36,7 @@ from ...utils import (
     unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import DiffusionPipeline
+from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from ..stable_diffusion import StableDiffusionPipelineOutput
 from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 
@@ -170,7 +170,7 @@ class AttendExciteAttnProcessor:
         return hidden_states
 
 
-class StableDiffusionAttendAndExcitePipeline(DiffusionPipeline, TextualInversionLoaderMixin):
+class StableDiffusionAttendAndExcitePipeline(DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion and Attend-and-Excite.
 
@@ -246,22 +246,6 @@ class StableDiffusionAttendAndExcitePipeline(DiffusionPipeline, TextualInversion
         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.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
     def _encode_prompt(
         self,
@@ -356,7 +340,7 @@ class StableDiffusionAttendAndExcitePipeline(DiffusionPipeline, TextualInversion
             batch_size = prompt_embeds.shape[0]
 
         if prompt_embeds is None:
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
 
@@ -438,7 +422,7 @@ class StableDiffusionAttendAndExcitePipeline(DiffusionPipeline, TextualInversion
             else:
                 uncond_tokens = negative_prompt
 
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
 

diffusers/pipelines/stable_diffusion_diffedit/pipeline_stable_diffusion_diffedit.py

@@ -1,4 +1,4 @@
-# Copyright 2023 DiffEdit Authors and Pix2Pix Zero Authors and The HuggingFace Team. All rights reserved.
+# Copyright 2024 DiffEdit Authors and Pix2Pix Zero Authors and The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -39,7 +39,7 @@ from ...utils import (
     unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import DiffusionPipeline
+from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from ..stable_diffusion import StableDiffusionPipelineOutput
 from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 
@@ -235,7 +235,9 @@ def preprocess_mask(mask, batch_size: int = 1):
     return mask
 
 
-class StableDiffusionDiffEditPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin):
+class StableDiffusionDiffEditPipeline(
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, LoraLoaderMixin
+):
     r"""
     <Tip warning={true}>
 
@@ -371,39 +373,6 @@ class StableDiffusionDiffEditPipeline(DiffusionPipeline, TextualInversionLoaderM
         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.enable_vae_slicing
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_tiling
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_tiling
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
     def _encode_prompt(
         self,
@@ -498,7 +467,7 @@ class StableDiffusionDiffEditPipeline(DiffusionPipeline, TextualInversionLoaderM
             batch_size = prompt_embeds.shape[0]
 
         if prompt_embeds is None:
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
 
@@ -580,7 +549,7 @@ class StableDiffusionDiffEditPipeline(DiffusionPipeline, TextualInversionLoaderM
             else:
                 uncond_tokens = negative_prompt
 
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)
 
@@ -754,6 +723,8 @@ class StableDiffusionDiffEditPipeline(DiffusionPipeline, TextualInversionLoaderM
 
         t_start = max(num_inference_steps - init_timestep, 0)
         timesteps = self.scheduler.timesteps[t_start * self.scheduler.order :]
+        if hasattr(self.scheduler, "set_begin_index"):
+            self.scheduler.set_begin_index(t_start * self.scheduler.order)
 
         return timesteps, num_inference_steps - t_start
 

diffusers/pipelines/stable_diffusion_gligen/pipeline_stable_diffusion_gligen.py

@@ -1,4 +1,4 @@
-# Copyright 2023 The GLIGEN Authors and HuggingFace Team. All rights reserved.
+# Copyright 2024 The GLIGEN Authors and HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -35,7 +35,7 @@ from ...utils import (
     unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import DiffusionPipeline
+from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from ..stable_diffusion import StableDiffusionPipelineOutput
 from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 
@@ -99,7 +99,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-class StableDiffusionGLIGENPipeline(DiffusionPipeline):
+class StableDiffusionGLIGENPipeline(DiffusionPipeline, StableDiffusionMixin):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion with Grounded-Language-to-Image Generation (GLIGEN).
 
@@ -172,35 +172,6 @@ class StableDiffusionGLIGENPipeline(DiffusionPipeline):
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 
-    def enable_vae_slicing(self):
-        r"""
-        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
-        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
-        """
-        self.vae.enable_slicing()
-
-    def disable_vae_slicing(self):
-        r"""
-        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_slicing()
-
-    def enable_vae_tiling(self):
-        r"""
-        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
-        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
-        processing larger images.
-        """
-        self.vae.enable_tiling()
-
-    def disable_vae_tiling(self):
-        r"""
-        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
-        computing decoding in one step.
-        """
-        self.vae.disable_tiling()
-
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._encode_prompt
     def _encode_prompt(
         self,
@@ -295,7 +266,7 @@ class StableDiffusionGLIGENPipeline(DiffusionPipeline):
             batch_size = prompt_embeds.shape[0]
 
         if prompt_embeds is None:
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 prompt = self.maybe_convert_prompt(prompt, self.tokenizer)
 
@@ -377,7 +348,7 @@ class StableDiffusionGLIGENPipeline(DiffusionPipeline):
             else:
                 uncond_tokens = negative_prompt
 
-            # textual inversion: procecss multi-vector tokens if necessary
+            # textual inversion: process multi-vector tokens if necessary
             if isinstance(self, TextualInversionLoaderMixin):
                 uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)