Revert "[LoRA] introduce LoraBaseMixin to promote reusability." (#8976)

Revert "[LoRA] introduce LoraBaseMixin to promote reusability. (#8774)"

This reverts commit 527430d0.

examples/community/lpw_stable_diffusion_xl.py

@@ -22,12 +22,7 @@ from transformers import (
 
 from diffusers import DiffusionPipeline, StableDiffusionXLPipeline
 from diffusers.image_processor import PipelineImageInput, VaeImageProcessor
-from diffusers.loaders import (
-    FromSingleFileMixin,
-    IPAdapterMixin,
-    StableDiffusionLoraLoaderMixin,
-    TextualInversionLoaderMixin,
-)
+from diffusers.loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models import AutoencoderKL, ImageProjection, UNet2DConditionModel
 from diffusers.models.attention_processor import AttnProcessor2_0, XFormersAttnProcessor
 from diffusers.pipelines.pipeline_utils import StableDiffusionMixin
@@ -549,7 +544,7 @@ class SDXLLongPromptWeightingPipeline(
     StableDiffusionMixin,
     FromSingleFileMixin,
     IPAdapterMixin,
-    StableDiffusionLoraLoaderMixin,
+    LoraLoaderMixin,
     TextualInversionLoaderMixin,
 ):
     r"""
@@ -561,8 +556,8 @@ class SDXLLongPromptWeightingPipeline(
     The pipeline also inherits the following loading methods:
         - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
         - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
-        - [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
-        - [`~loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
         - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
 
     Args:
@@ -743,7 +738,7 @@ class SDXLLongPromptWeightingPipeline(
 
         # set lora scale so that monkey patched LoRA
         # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
+        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
             self._lora_scale = lora_scale
 
         if prompt is not None and isinstance(prompt, str):

examples/community/pipeline_animatediff_controlnet.py

@@ -22,7 +22,7 @@ from PIL import Image
 from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
 
 from diffusers.image_processor import PipelineImageInput, VaeImageProcessor
-from diffusers.loaders import IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from diffusers.loaders import IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models import AutoencoderKL, ControlNetModel, ImageProjection, UNet2DConditionModel, UNetMotionModel
 from diffusers.models.lora import adjust_lora_scale_text_encoder
 from diffusers.models.unets.unet_motion_model import MotionAdapter
@@ -114,11 +114,7 @@ def tensor2vid(video: torch.Tensor, processor, output_type="np"):
 
 
 class AnimateDiffControlNetPipeline(
-    DiffusionPipeline,
-    StableDiffusionMixin,
-    TextualInversionLoaderMixin,
-    IPAdapterMixin,
-    StableDiffusionLoraLoaderMixin,
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, IPAdapterMixin, LoraLoaderMixin
 ):
     r"""
     Pipeline for text-to-video generation.
@@ -128,8 +124,8 @@ class AnimateDiffControlNetPipeline(
 
     The pipeline also inherits the following loading methods:
         - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
-        - [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
-        - [`~loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
         - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
 
     Args:
@@ -238,7 +234,7 @@ class AnimateDiffControlNetPipeline(
         """
         # set lora scale so that monkey patched LoRA
         # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
+        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
             self._lora_scale = lora_scale
 
             # dynamically adjust the LoRA scale
@@ -370,7 +366,7 @@ class AnimateDiffControlNetPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
-        if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
+        if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
             # Retrieve the original scale by scaling back the LoRA layers
             unscale_lora_layers(self.text_encoder, lora_scale)
 

examples/community/pipeline_animatediff_img2video.py

@@ -27,7 +27,7 @@ import torch
 from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
 
 from diffusers.image_processor import PipelineImageInput, VaeImageProcessor
-from diffusers.loaders import IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from diffusers.loaders import IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models import AutoencoderKL, ImageProjection, UNet2DConditionModel, UNetMotionModel
 from diffusers.models.lora import adjust_lora_scale_text_encoder
 from diffusers.models.unet_motion_model import MotionAdapter
@@ -240,11 +240,7 @@ def retrieve_timesteps(
 
 
 class AnimateDiffImgToVideoPipeline(
-    DiffusionPipeline,
-    StableDiffusionMixin,
-    TextualInversionLoaderMixin,
-    IPAdapterMixin,
-    StableDiffusionLoraLoaderMixin,
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, IPAdapterMixin, LoraLoaderMixin
 ):
     r"""
     Pipeline for image-to-video generation.
@@ -254,8 +250,8 @@ class AnimateDiffImgToVideoPipeline(
 
     The pipeline also inherits the following loading methods:
         - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
-        - [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
-        - [`~loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
         - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
 
     Args:
@@ -355,7 +351,7 @@ class AnimateDiffImgToVideoPipeline(
         """
         # set lora scale so that monkey patched LoRA
         # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
+        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
             self._lora_scale = lora_scale
 
             # dynamically adjust the LoRA scale
@@ -487,7 +483,7 @@ class AnimateDiffImgToVideoPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
-        if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
+        if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
             # Retrieve the original scale by scaling back the LoRA layers
             unscale_lora_layers(self.text_encoder, lora_scale)
 

examples/community/pipeline_demofusion_sdxl.py

@@ -12,7 +12,7 @@ from transformers import CLIPTextModel, CLIPTextModelWithProjection, CLIPTokeniz
 from diffusers.image_processor import VaeImageProcessor
 from diffusers.loaders import (
     FromSingleFileMixin,
-    StableDiffusionLoraLoaderMixin,
+    LoraLoaderMixin,
     TextualInversionLoaderMixin,
 )
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
@@ -89,11 +89,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
 
 
 class DemoFusionSDXLPipeline(
-    DiffusionPipeline,
-    StableDiffusionMixin,
-    FromSingleFileMixin,
-    StableDiffusionLoraLoaderMixin,
-    TextualInversionLoaderMixin,
+    DiffusionPipeline, StableDiffusionMixin, FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 ):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion XL.
@@ -235,7 +231,7 @@ class DemoFusionSDXLPipeline(
 
         # set lora scale so that monkey patched LoRA
         # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
+        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
             self._lora_scale = lora_scale
 
             # dynamically adjust the LoRA scale

examples/community/pipeline_fabric.py

@@ -21,7 +21,7 @@ from transformers import CLIPTextModel, CLIPTokenizer
 from diffusers import AutoencoderKL, UNet2DConditionModel
 from diffusers.configuration_utils import FrozenDict
 from diffusers.image_processor import VaeImageProcessor
-from diffusers.loaders import StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from diffusers.loaders import LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models.attention import BasicTransformerBlock
 from diffusers.models.attention_processor import LoRAAttnProcessor
 from diffusers.pipelines.pipeline_utils import DiffusionPipeline
@@ -222,7 +222,7 @@ class FabricPipeline(DiffusionPipeline):
         """
         # set lora scale so that monkey patched LoRA
         # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
+        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
             self._lora_scale = lora_scale
 
         if prompt is not None and isinstance(prompt, str):

examples/community/pipeline_prompt2prompt.py

@@ -35,7 +35,7 @@ from diffusers.image_processor import VaeImageProcessor
 from diffusers.loaders import (
     FromSingleFileMixin,
     IPAdapterMixin,
-    StableDiffusionLoraLoaderMixin,
+    LoraLoaderMixin,
     TextualInversionLoaderMixin,
 )
 from diffusers.models.attention import Attention
@@ -75,7 +75,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
 class Prompt2PromptPipeline(
     DiffusionPipeline,
     TextualInversionLoaderMixin,
-    StableDiffusionLoraLoaderMixin,
+    LoraLoaderMixin,
     IPAdapterMixin,
     FromSingleFileMixin,
 ):
@@ -87,8 +87,8 @@ class Prompt2PromptPipeline(
 
     The pipeline also inherits the following loading methods:
         - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
-        - [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
-        - [`~loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
         - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
         - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
 
@@ -286,7 +286,7 @@ class Prompt2PromptPipeline(
         """
         # set lora scale so that monkey patched LoRA
         # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
+        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
             self._lora_scale = lora_scale
 
             # dynamically adjust the LoRA scale
@@ -420,7 +420,7 @@ class Prompt2PromptPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
-        if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
+        if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
             # Retrieve the original scale by scaling back the LoRA layers
             unscale_lora_layers(self.text_encoder, lora_scale)
 

examples/community/pipeline_stable_diffusion_boxdiff.py

@@ -27,12 +27,7 @@ from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPV
 
 from diffusers.configuration_utils import FrozenDict
 from diffusers.image_processor import PipelineImageInput, VaeImageProcessor
-from diffusers.loaders import (
-    FromSingleFileMixin,
-    IPAdapterMixin,
-    StableDiffusionLoraLoaderMixin,
-    TextualInversionLoaderMixin,
-)
+from diffusers.loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models import AutoencoderKL, ImageProjection, UNet2DConditionModel
 from diffusers.models.attention_processor import Attention, FusedAttnProcessor2_0
 from diffusers.models.lora import adjust_lora_scale_text_encoder
@@ -363,7 +358,7 @@ def retrieve_timesteps(
 
 
 class StableDiffusionBoxDiffPipeline(
-    DiffusionPipeline, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin, IPAdapterMixin, FromSingleFileMixin
+    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, IPAdapterMixin, FromSingleFileMixin
 ):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion with BoxDiff.
@@ -373,8 +368,8 @@ class StableDiffusionBoxDiffPipeline(
 
     The pipeline also inherits the following loading methods:
         - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
-        - [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
-        - [`~loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
         - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
         - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
 
@@ -599,7 +594,7 @@ class StableDiffusionBoxDiffPipeline(
         """
         # set lora scale so that monkey patched LoRA
         # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
+        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
             self._lora_scale = lora_scale
 
             # dynamically adjust the LoRA scale
@@ -731,7 +726,7 @@ class StableDiffusionBoxDiffPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
-        if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
+        if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
             # Retrieve the original scale by scaling back the LoRA layers
             unscale_lora_layers(self.text_encoder, lora_scale)
 

examples/community/pipeline_stable_diffusion_pag.py

@@ -11,12 +11,7 @@ from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPV
 
 from diffusers.configuration_utils import FrozenDict
 from diffusers.image_processor import PipelineImageInput, VaeImageProcessor
-from diffusers.loaders import (
-    FromSingleFileMixin,
-    IPAdapterMixin,
-    StableDiffusionLoraLoaderMixin,
-    TextualInversionLoaderMixin,
-)
+from diffusers.loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models import AutoencoderKL, ImageProjection, UNet2DConditionModel
 from diffusers.models.attention_processor import Attention, AttnProcessor2_0, FusedAttnProcessor2_0
 from diffusers.models.lora import adjust_lora_scale_text_encoder
@@ -333,7 +328,7 @@ def retrieve_timesteps(
 
 
 class StableDiffusionPAGPipeline(
-    DiffusionPipeline, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin, IPAdapterMixin, FromSingleFileMixin
+    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, IPAdapterMixin, FromSingleFileMixin
 ):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion.
@@ -341,8 +336,8 @@ class StableDiffusionPAGPipeline(
     implemented for all pipelines (downloading, saving, running on a particular device, etc.).
     The pipeline also inherits the following loading methods:
         - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
-        - [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
-        - [`~loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
         - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
         - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
     Args:
@@ -565,7 +560,7 @@ class StableDiffusionPAGPipeline(
         """
         # set lora scale so that monkey patched LoRA
         # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
+        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
             self._lora_scale = lora_scale
 
             # dynamically adjust the LoRA scale
@@ -697,7 +692,7 @@ class StableDiffusionPAGPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
-        if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
+        if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
             # Retrieve the original scale by scaling back the LoRA layers
             unscale_lora_layers(self.text_encoder, lora_scale)
 

examples/community/pipeline_stable_diffusion_upscale_ldm3d.py

@@ -22,7 +22,7 @@ from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
 
 from diffusers import DiffusionPipeline
 from diffusers.image_processor import PipelineDepthInput, PipelineImageInput, VaeImageProcessorLDM3D
-from diffusers.loaders import FromSingleFileMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from diffusers.loaders import FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
 from diffusers.models.lora import adjust_lora_scale_text_encoder
 from diffusers.pipelines.stable_diffusion import StableDiffusionSafetyChecker
@@ -69,7 +69,7 @@ EXAMPLE_DOC_STRING = """
 
 
 class StableDiffusionUpscaleLDM3DPipeline(
-    DiffusionPipeline, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin, FromSingleFileMixin
+    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin
 ):
     r"""
     Pipeline for text-to-image and 3D generation using LDM3D.
@@ -79,8 +79,8 @@ class StableDiffusionUpscaleLDM3DPipeline(
 
     The pipeline also inherits the following loading methods:
         - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
-        - [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
-        - [`~loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
         - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
 
     Args:
@@ -233,7 +233,7 @@ class StableDiffusionUpscaleLDM3DPipeline(
         """
         # set lora scale so that monkey patched LoRA
         # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
+        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
             self._lora_scale = lora_scale
 
             # dynamically adjust the LoRA scale
@@ -365,7 +365,7 @@ class StableDiffusionUpscaleLDM3DPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
-        if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
+        if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
             # Retrieve the original scale by scaling back the LoRA layers
             unscale_lora_layers(self.text_encoder, lora_scale)
 

examples/community/pipeline_stable_diffusion_xl_controlnet_adapter_inpaint.py

@@ -33,7 +33,7 @@ from diffusers import DiffusionPipeline
 from diffusers.image_processor import PipelineImageInput, VaeImageProcessor
 from diffusers.loaders import (
     FromSingleFileMixin,
-    StableDiffusionLoraLoaderMixin,
+    LoraLoaderMixin,
     StableDiffusionXLLoraLoaderMixin,
     TextualInversionLoaderMixin,
 )
@@ -300,7 +300,7 @@ def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):
 
 
 class StableDiffusionXLControlNetAdapterInpaintPipeline(
-    DiffusionPipeline, StableDiffusionMixin, FromSingleFileMixin, StableDiffusionLoraLoaderMixin
+    DiffusionPipeline, StableDiffusionMixin, FromSingleFileMixin, LoraLoaderMixin
 ):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion augmented with T2I-Adapter

examples/community/pipeline_stable_diffusion_xl_differential_img2img.py

@@ -178,11 +178,11 @@ class StableDiffusionXLDifferentialImg2ImgPipeline(
 
     In addition the pipeline inherits the following loading methods:
         - *Textual-Inversion*: [`loaders.TextualInversionLoaderMixin.load_textual_inversion`]
-        - *LoRA*: [`loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`]
+        - *LoRA*: [`loaders.LoraLoaderMixin.load_lora_weights`]
         - *Ckpt*: [`loaders.FromSingleFileMixin.from_single_file`]
 
     as well as the following saving methods:
-        - *LoRA*: [`loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`]
+        - *LoRA*: [`loaders.LoraLoaderMixin.save_lora_weights`]
 
     Args:
         vae ([`AutoencoderKL`]):

examples/community/sde_drag.py

@@ -11,7 +11,7 @@ from tqdm.auto import tqdm
 from transformers import CLIPTextModel, CLIPTokenizer
 
 from diffusers import AutoencoderKL, DiffusionPipeline, DPMSolverMultistepScheduler, UNet2DConditionModel
-from diffusers.loaders import AttnProcsLayers, StableDiffusionLoraLoaderMixin
+from diffusers.loaders import AttnProcsLayers, LoraLoaderMixin
 from diffusers.models.attention_processor import (
     AttnAddedKVProcessor,
     AttnAddedKVProcessor2_0,
@@ -321,7 +321,7 @@ class SdeDragPipeline(DiffusionPipeline):
             optimizer.zero_grad()
 
         with tempfile.TemporaryDirectory() as save_lora_dir:
-            StableDiffusionLoraLoaderMixin.save_lora_weights(
+            LoraLoaderMixin.save_lora_weights(
                 save_directory=save_lora_dir,
                 unet_lora_layers=unet_lora_layers,
                 text_encoder_lora_layers=None,

examples/community/stable_diffusion_ipex.py

@@ -21,7 +21,7 @@ from packaging import version
 from transformers import CLIPFeatureExtractor, CLIPTextModel, CLIPTokenizer
 
 from diffusers.configuration_utils import FrozenDict
-from diffusers.loaders import StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from diffusers.loaders import LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
 from diffusers.pipelines.pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
@@ -61,7 +61,7 @@ EXAMPLE_DOC_STRING = """
 
 
 class StableDiffusionIPEXPipeline(
-    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, LoraLoaderMixin
 ):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion on IPEX.

examples/community/stable_diffusion_reference.py

@@ -11,12 +11,7 @@ from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
 from diffusers import AutoencoderKL, DiffusionPipeline, UNet2DConditionModel
 from diffusers.configuration_utils import FrozenDict, deprecate
 from diffusers.image_processor import VaeImageProcessor
-from diffusers.loaders import (
-    FromSingleFileMixin,
-    IPAdapterMixin,
-    StableDiffusionLoraLoaderMixin,
-    TextualInversionLoaderMixin,
-)
+from diffusers.loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models.attention import BasicTransformerBlock
 from diffusers.models.lora import adjust_lora_scale_text_encoder
 from diffusers.models.unets.unet_2d_blocks import CrossAttnDownBlock2D, CrossAttnUpBlock2D, DownBlock2D, UpBlock2D
@@ -81,7 +76,7 @@ def torch_dfs(model: torch.nn.Module):
 
 
 class StableDiffusionReferencePipeline(
-    DiffusionPipeline, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin, IPAdapterMixin, FromSingleFileMixin
+    DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, IPAdapterMixin, FromSingleFileMixin
 ):
     r"""
     Pipeline for Stable Diffusion Reference.
@@ -91,8 +86,8 @@ class StableDiffusionReferencePipeline(
 
     The pipeline also inherits the following loading methods:
     - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
-    - [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
-    - [`~loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+    - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+    - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
     - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
     - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters
 
@@ -448,7 +443,7 @@ class StableDiffusionReferencePipeline(
         """
         # set lora scale so that monkey patched LoRA
         # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
+        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
             self._lora_scale = lora_scale
 
             # dynamically adjust the LoRA scale
@@ -580,7 +575,7 @@ class StableDiffusionReferencePipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
-        if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
+        if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
             # Retrieve the original scale by scaling back the LoRA layers
             unscale_lora_layers(self.text_encoder, lora_scale)
 

examples/community/stable_diffusion_repaint.py

@@ -23,7 +23,7 @@ from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
 
 from diffusers import AutoencoderKL, DiffusionPipeline, UNet2DConditionModel
 from diffusers.configuration_utils import FrozenDict, deprecate
-from diffusers.loaders import StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from diffusers.loaders import LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.pipelines.pipeline_utils import StableDiffusionMixin
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
 from diffusers.pipelines.stable_diffusion.safety_checker import (
@@ -140,7 +140,7 @@ def prepare_mask_and_masked_image(image, mask):
 
 
 class StableDiffusionRepaintPipeline(
-    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, LoraLoaderMixin
 ):
     r"""
     Pipeline for text-guided image inpainting using Stable Diffusion. *This is an experimental feature*.
@@ -148,9 +148,9 @@ class StableDiffusionRepaintPipeline(
     library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.)
     In addition the pipeline inherits the following loading methods:
         - *Textual-Inversion*: [`loaders.TextualInversionLoaderMixin.load_textual_inversion`]
-        - *LoRA*: [`loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`]
+        - *LoRA*: [`loaders.LoraLoaderMixin.load_lora_weights`]
     as well as the following saving methods:
-        - *LoRA*: [`loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`]
+        - *LoRA*: [`loaders.LoraLoaderMixin.save_lora_weights`]
     Args:
         vae ([`AutoencoderKL`]):
             Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.

examples/dreambooth/README.md

@@ -425,8 +425,8 @@ pipe.load_lora_weights(lora_model_id)
 image = pipe("A picture of a sks dog in a bucket", num_inference_steps=25).images[0]
 ```
 
-Note that the use of [`StableDiffusionLoraLoaderMixin.load_lora_weights`](https://huggingface.co/docs/diffusers/main/en/api/loaders#diffusers.loaders.StableDiffusionLoraLoaderMixin.load_lora_weights) is preferred to [`UNet2DConditionLoadersMixin.load_attn_procs`](https://huggingface.co/docs/diffusers/main/en/api/loaders#diffusers.loaders.UNet2DConditionLoadersMixin.load_attn_procs) for loading LoRA parameters. This is because
-`StableDiffusionLoraLoaderMixin.load_lora_weights` can handle the following situations:
+Note that the use of [`LoraLoaderMixin.load_lora_weights`](https://huggingface.co/docs/diffusers/main/en/api/loaders#diffusers.loaders.LoraLoaderMixin.load_lora_weights) is preferred to [`UNet2DConditionLoadersMixin.load_attn_procs`](https://huggingface.co/docs/diffusers/main/en/api/loaders#diffusers.loaders.UNet2DConditionLoadersMixin.load_attn_procs) for loading LoRA parameters. This is because
+`LoraLoaderMixin.load_lora_weights` can handle the following situations:
 
 * LoRA parameters that don't have separate identifiers for the UNet and the text encoder (such as [`"patrickvonplaten/lora_dreambooth_dog_example"`](https://huggingface.co/patrickvonplaten/lora_dreambooth_dog_example)). So, you can just do:
 

examples/dreambooth/train_dreambooth_lora.py

@@ -52,7 +52,7 @@ from diffusers import (
     StableDiffusionPipeline,
     UNet2DConditionModel,
 )
-from diffusers.loaders import StableDiffusionLoraLoaderMixin
+from diffusers.loaders import LoraLoaderMixin
 from diffusers.optimization import get_scheduler
 from diffusers.training_utils import _set_state_dict_into_text_encoder, cast_training_params
 from diffusers.utils import (
@@ -956,7 +956,7 @@ def main(args):
                 # make sure to pop weight so that corresponding model is not saved again
                 weights.pop()
 
-            StableDiffusionLoraLoaderMixin.save_lora_weights(
+            LoraLoaderMixin.save_lora_weights(
                 output_dir,
                 unet_lora_layers=unet_lora_layers_to_save,
                 text_encoder_lora_layers=text_encoder_lora_layers_to_save,
@@ -976,7 +976,7 @@ def main(args):
             else:
                 raise ValueError(f"unexpected save model: {model.__class__}")
 
-        lora_state_dict, network_alphas = StableDiffusionLoraLoaderMixin.lora_state_dict(input_dir)
+        lora_state_dict, network_alphas = LoraLoaderMixin.lora_state_dict(input_dir)
 
         unet_state_dict = {f'{k.replace("unet.", "")}': v for k, v in lora_state_dict.items() if k.startswith("unet.")}
         unet_state_dict = convert_unet_state_dict_to_peft(unet_state_dict)
@@ -1376,7 +1376,7 @@ def main(args):
         else:
             text_encoder_state_dict = None
 
-        StableDiffusionLoraLoaderMixin.save_lora_weights(
+        LoraLoaderMixin.save_lora_weights(
             save_directory=args.output_dir,
             unet_lora_layers=unet_lora_state_dict,
             text_encoder_lora_layers=text_encoder_state_dict,

examples/dreambooth/train_dreambooth_lora_sdxl.py

@@ -58,7 +58,7 @@ from diffusers import (
     StableDiffusionXLPipeline,
     UNet2DConditionModel,
 )
-from diffusers.loaders import StableDiffusionLoraLoaderMixin
+from diffusers.loaders import LoraLoaderMixin
 from diffusers.optimization import get_scheduler
 from diffusers.training_utils import _set_state_dict_into_text_encoder, cast_training_params, compute_snr
 from diffusers.utils import (
@@ -1260,7 +1260,7 @@ def main(args):
             else:
                 raise ValueError(f"unexpected save model: {model.__class__}")
 
-        lora_state_dict, network_alphas = StableDiffusionLoraLoaderMixin.lora_state_dict(input_dir)
+        lora_state_dict, network_alphas = LoraLoaderMixin.lora_state_dict(input_dir)
 
         unet_state_dict = {f'{k.replace("unet.", "")}': v for k, v in lora_state_dict.items() if k.startswith("unet.")}
         unet_state_dict = convert_unet_state_dict_to_peft(unet_state_dict)

examples/research_projects/diffusion_dpo/train_diffusion_dpo.py

@@ -49,7 +49,7 @@ from diffusers import (
     DPMSolverMultistepScheduler,
     UNet2DConditionModel,
 )
-from diffusers.loaders import StableDiffusionLoraLoaderMixin
+from diffusers.loaders import LoraLoaderMixin
 from diffusers.optimization import get_scheduler
 from diffusers.utils import check_min_version, convert_state_dict_to_diffusers
 from diffusers.utils.import_utils import is_xformers_available
@@ -604,7 +604,7 @@ def main(args):
                 # make sure to pop weight so that corresponding model is not saved again
                 weights.pop()
 
-            StableDiffusionLoraLoaderMixin.save_lora_weights(
+            LoraLoaderMixin.save_lora_weights(
                 output_dir,
                 unet_lora_layers=unet_lora_layers_to_save,
                 text_encoder_lora_layers=None,
@@ -621,8 +621,8 @@ def main(args):
             else:
                 raise ValueError(f"unexpected save model: {model.__class__}")
 
-        lora_state_dict, network_alphas = StableDiffusionLoraLoaderMixin.lora_state_dict(input_dir)
-        StableDiffusionLoraLoaderMixin.load_lora_into_unet(lora_state_dict, network_alphas=network_alphas, unet=unet_)
+        lora_state_dict, network_alphas = LoraLoaderMixin.lora_state_dict(input_dir)
+        LoraLoaderMixin.load_lora_into_unet(lora_state_dict, network_alphas=network_alphas, unet=unet_)
 
     accelerator.register_save_state_pre_hook(save_model_hook)
     accelerator.register_load_state_pre_hook(load_model_hook)
@@ -951,7 +951,7 @@ def main(args):
         unet = unet.to(torch.float32)
         unet_lora_state_dict = convert_state_dict_to_diffusers(get_peft_model_state_dict(unet))
 
-        StableDiffusionLoraLoaderMixin.save_lora_weights(
+        LoraLoaderMixin.save_lora_weights(
             save_directory=args.output_dir, unet_lora_layers=unet_lora_state_dict, text_encoder_lora_layers=None
         )
 

examples/research_projects/promptdiffusion/pipeline_prompt_diffusion.py

@@ -28,7 +28,7 @@ import torch.nn.functional as F
 from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
 
 from diffusers.image_processor import PipelineImageInput, VaeImageProcessor
-from diffusers.loaders import FromSingleFileMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from diffusers.loaders import FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from diffusers.models import AutoencoderKL, ControlNetModel, UNet2DConditionModel
 from diffusers.models.lora import adjust_lora_scale_text_encoder
 from diffusers.pipelines.controlnet.multicontrolnet import MultiControlNetModel
@@ -142,9 +142,7 @@ def retrieve_timesteps(
     return timesteps, num_inference_steps
 
 
-class PromptDiffusionPipeline(
-    DiffusionPipeline, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin, FromSingleFileMixin
-):
+class PromptDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion with ControlNet guidance.
 
@@ -155,8 +153,8 @@ class PromptDiffusionPipeline(
 
     The pipeline also inherits the following loading methods:
         - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
-        - [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
-        - [`~loaders.StableDiffusionLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
+        - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights
+        - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights
         - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
 
     Args:
@@ -350,7 +348,7 @@ class PromptDiffusionPipeline(
         """
         # set lora scale so that monkey patched LoRA
         # function of text encoder can correctly access it
-        if lora_scale is not None and isinstance(self, StableDiffusionLoraLoaderMixin):
+        if lora_scale is not None and isinstance(self, LoraLoaderMixin):
             self._lora_scale = lora_scale
 
             # dynamically adjust the LoRA scale
@@ -482,7 +480,7 @@ class PromptDiffusionPipeline(
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
-        if isinstance(self, StableDiffusionLoraLoaderMixin) and USE_PEFT_BACKEND:
+        if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:
             # Retrieve the original scale by scaling back the LoRA layers
             unscale_lora_layers(self.text_encoder, lora_scale)