remove more.

src/diffusers/models/controlnets/controlnet_flux.py

@@ -21,12 +21,9 @@ import torch.nn as nn
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import PeftAdapterMixin
 from ...utils import (
-    USE_PEFT_BACKEND,
     BaseOutput,
     apply_lora_scale,
     logging,
-    scale_lora_layers,
-    unscale_lora_layers,
 )
 from ..attention import AttentionMixin
 from ..controlnets.controlnet import ControlNetConditioningEmbedding, zero_module
@@ -205,20 +202,6 @@ class FluxControlNetModel(ModelMixin, AttentionMixin, ConfigMixin, PeftAdapterMi
             If `return_dict` is True, an [`~models.transformer_2d.Transformer2DModelOutput`] is returned, otherwise a
             `tuple` where the first element is the sample tensor.
         """
-        if joint_attention_kwargs is not None:
-            joint_attention_kwargs = joint_attention_kwargs.copy()
-            lora_scale = joint_attention_kwargs.pop("scale", 1.0)
-        else:
-            lora_scale = 1.0
-
-        if USE_PEFT_BACKEND:
-            # weight the lora layers by setting `lora_scale` for each PEFT layer
-            scale_lora_layers(self, lora_scale)
-        else:
-            if joint_attention_kwargs is not None and joint_attention_kwargs.get("scale", None) is not None:
-                logger.warning(
-                    "Passing `scale` via `joint_attention_kwargs` when not using the PEFT backend is ineffective."
-                )
         hidden_states = self.x_embedder(hidden_states)
 
         if self.input_hint_block is not None:
@@ -331,10 +314,6 @@ class FluxControlNetModel(ModelMixin, AttentionMixin, ConfigMixin, PeftAdapterMi
             None if len(controlnet_single_block_samples) == 0 else controlnet_single_block_samples
         )
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (controlnet_block_samples, controlnet_single_block_samples)
 

src/diffusers/models/controlnets/controlnet_qwenimage.py

@@ -21,13 +21,10 @@ import torch.nn as nn
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import FromOriginalModelMixin, PeftAdapterMixin
 from ...utils import (
-    USE_PEFT_BACKEND,
     BaseOutput,
     apply_lora_scale,
     deprecate,
     logging,
-    scale_lora_layers,
-    unscale_lora_layers,
 )
 from ..attention import AttentionMixin
 from ..cache_utils import CacheMixin
@@ -190,20 +187,6 @@ class QwenImageControlNetModel(
                 standard_warn=False,
             )
 
-        if joint_attention_kwargs is not None:
-            joint_attention_kwargs = joint_attention_kwargs.copy()
-            lora_scale = joint_attention_kwargs.pop("scale", 1.0)
-        else:
-            lora_scale = 1.0
-
-        if USE_PEFT_BACKEND:
-            # weight the lora layers by setting `lora_scale` for each PEFT layer
-            scale_lora_layers(self, lora_scale)
-        else:
-            if joint_attention_kwargs is not None and joint_attention_kwargs.get("scale", None) is not None:
-                logger.warning(
-                    "Passing `scale` via `joint_attention_kwargs` when not using the PEFT backend is ineffective."
-                )
         hidden_states = self.img_in(hidden_states)
 
         # add
@@ -265,10 +248,6 @@ class QwenImageControlNetModel(
         controlnet_block_samples = [sample * conditioning_scale for sample in controlnet_block_samples]
         controlnet_block_samples = None if len(controlnet_block_samples) == 0 else controlnet_block_samples
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return controlnet_block_samples
 

src/diffusers/models/controlnets/controlnet_sd3.py

@@ -21,7 +21,7 @@ import torch.nn as nn
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import FromOriginalModelMixin, PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, apply_lora_scale, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import apply_lora_scale, logging
 from ..attention import AttentionMixin, JointTransformerBlock
 from ..attention_processor import Attention, FusedJointAttnProcessor2_0
 from ..embeddings import CombinedTimestepTextProjEmbeddings, PatchEmbed
@@ -309,21 +309,6 @@ class SD3ControlNetModel(ModelMixin, AttentionMixin, ConfigMixin, PeftAdapterMix
             If `return_dict` is True, an [`~models.transformer_2d.Transformer2DModelOutput`] is returned, otherwise a
             `tuple` where the first element is the sample tensor.
         """
-        if joint_attention_kwargs is not None:
-            joint_attention_kwargs = joint_attention_kwargs.copy()
-            lora_scale = joint_attention_kwargs.pop("scale", 1.0)
-        else:
-            lora_scale = 1.0
-
-        if USE_PEFT_BACKEND:
-            # weight the lora layers by setting `lora_scale` for each PEFT layer
-            scale_lora_layers(self, lora_scale)
-        else:
-            if joint_attention_kwargs is not None and joint_attention_kwargs.get("scale", None) is not None:
-                logger.warning(
-                    "Passing `scale` via `joint_attention_kwargs` when not using the PEFT backend is ineffective."
-                )
-
         if self.pos_embed is not None and hidden_states.ndim != 4:
             raise ValueError("hidden_states must be 4D when pos_embed is used")
 
@@ -383,10 +368,6 @@ class SD3ControlNetModel(ModelMixin, AttentionMixin, ConfigMixin, PeftAdapterMix
         # 6. scaling
         controlnet_block_res_samples = [sample * conditioning_scale for sample in controlnet_block_res_samples]
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (controlnet_block_res_samples,)
 

src/diffusers/models/transformers/transformer_bria_fibo.py

@@ -22,11 +22,8 @@ from ...models.modeling_outputs import Transformer2DModelOutput
 from ...models.modeling_utils import ModelMixin
 from ...models.transformers.transformer_bria import BriaAttnProcessor
 from ...utils import (
-    USE_PEFT_BACKEND,
     apply_lora_scale,
     logging,
-    scale_lora_layers,
-    unscale_lora_layers,
 )
 from ...utils.torch_utils import maybe_allow_in_graph
 from ..attention import AttentionModuleMixin, FeedForward
@@ -547,20 +544,7 @@ class BriaFiboTransformer2DModel(ModelMixin, ConfigMixin, PeftAdapterMixin, From
             If `return_dict` is True, an [`~models.transformer_2d.Transformer2DModelOutput`] is returned, otherwise a
             `tuple` where the first element is the sample tensor.
         """
-        if joint_attention_kwargs is not None:
-            joint_attention_kwargs = joint_attention_kwargs.copy()
-            lora_scale = joint_attention_kwargs.pop("scale", 1.0)
-        else:
-            lora_scale = 1.0
 
-        if USE_PEFT_BACKEND:
-            # weight the lora layers by setting `lora_scale` for each PEFT layer
-            scale_lora_layers(self, lora_scale)
-        else:
-            if joint_attention_kwargs is not None and joint_attention_kwargs.get("scale", None) is not None:
-                logger.warning(
-                    "Passing `scale` via `joint_attention_kwargs` when not using the PEFT backend is ineffective."
-                )
         hidden_states = self.x_embedder(hidden_states)
 
         timestep = timestep.to(hidden_states.dtype)
@@ -647,10 +631,6 @@ class BriaFiboTransformer2DModel(ModelMixin, ConfigMixin, PeftAdapterMixin, From
         hidden_states = self.norm_out(hidden_states, temb)
         output = self.proj_out(hidden_states)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (output,)
 

src/diffusers/models/transformers/transformer_chroma.py

@@ -21,7 +21,7 @@ import torch.nn as nn
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import FluxTransformer2DLoadersMixin, FromOriginalModelMixin, PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import deprecate, logging
 from ...utils.import_utils import is_torch_npu_available
 from ...utils.torch_utils import maybe_allow_in_graph
 from ..attention import AttentionMixin, FeedForward
@@ -511,20 +511,6 @@ class ChromaTransformer2DModel(
             If `return_dict` is True, an [`~models.transformer_2d.Transformer2DModelOutput`] is returned, otherwise a
             `tuple` where the first element is the sample tensor.
         """
-        if joint_attention_kwargs is not None:
-            joint_attention_kwargs = joint_attention_kwargs.copy()
-            lora_scale = joint_attention_kwargs.pop("scale", 1.0)
-        else:
-            lora_scale = 1.0
-
-        if USE_PEFT_BACKEND:
-            # weight the lora layers by setting `lora_scale` for each PEFT layer
-            scale_lora_layers(self, lora_scale)
-        else:
-            if joint_attention_kwargs is not None and joint_attention_kwargs.get("scale", None) is not None:
-                logger.warning(
-                    "Passing `scale` via `joint_attention_kwargs` when not using the PEFT backend is ineffective."
-                )
 
         hidden_states = self.x_embedder(hidden_states)
 
@@ -631,10 +617,6 @@ class ChromaTransformer2DModel(
         hidden_states = self.norm_out(hidden_states, temb)
         output = self.proj_out(hidden_states)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (output,)
 

src/diffusers/models/transformers/transformer_chronoedit.py

@@ -21,7 +21,7 @@ import torch.nn.functional as F
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import FromOriginalModelMixin, PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers
 from ...utils.torch_utils import maybe_allow_in_graph
 from .._modeling_parallel import ContextParallelInput, ContextParallelOutput
 from ..attention import AttentionMixin, AttentionModuleMixin, FeedForward
@@ -729,10 +729,6 @@ class ChronoEditTransformer3DModel(
         hidden_states = hidden_states.permute(0, 7, 1, 4, 2, 5, 3, 6)
         output = hidden_states.flatten(6, 7).flatten(4, 5).flatten(2, 3)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (output,)
 

src/diffusers/models/transformers/transformer_flux2.py

@@ -21,7 +21,7 @@ import torch.nn.functional as F
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import FluxTransformer2DLoadersMixin, FromOriginalModelMixin, PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, apply_lora_scale, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import apply_lora_scale, logging
 from .._modeling_parallel import ContextParallelInput, ContextParallelOutput
 from ..attention import AttentionMixin, AttentionModuleMixin
 from ..attention_dispatch import dispatch_attention_fn
@@ -811,20 +811,6 @@ class Flux2Transformer2DModel(
             `tuple` where the first element is the sample tensor.
         """
         # 0. Handle input arguments
-        if joint_attention_kwargs is not None:
-            joint_attention_kwargs = joint_attention_kwargs.copy()
-            lora_scale = joint_attention_kwargs.pop("scale", 1.0)
-        else:
-            lora_scale = 1.0
-
-        if USE_PEFT_BACKEND:
-            # weight the lora layers by setting `lora_scale` for each PEFT layer
-            scale_lora_layers(self, lora_scale)
-        else:
-            if joint_attention_kwargs is not None and joint_attention_kwargs.get("scale", None) is not None:
-                logger.warning(
-                    "Passing `scale` via `joint_attention_kwargs` when not using the PEFT backend is ineffective."
-                )
 
         num_txt_tokens = encoder_hidden_states.shape[1]
 
@@ -909,10 +895,6 @@ class Flux2Transformer2DModel(
         hidden_states = self.norm_out(hidden_states, temb)
         output = self.proj_out(hidden_states)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (output,)
 

src/diffusers/models/transformers/transformer_hunyuan_video.py

@@ -22,7 +22,7 @@ from diffusers.loaders import FromOriginalModelMixin
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers
 from ..attention import AttentionMixin, FeedForward
 from ..attention_dispatch import dispatch_attention_fn
 from ..attention_processor import Attention
@@ -1104,10 +1104,6 @@ class HunyuanVideoTransformer3DModel(
         hidden_states = hidden_states.permute(0, 4, 1, 5, 2, 6, 3, 7)
         hidden_states = hidden_states.flatten(6, 7).flatten(4, 5).flatten(2, 3)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (hidden_states,)
 

src/diffusers/models/transformers/transformer_hunyuan_video15.py

@@ -22,7 +22,7 @@ from diffusers.loaders import FromOriginalModelMixin
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers
 from ..attention import AttentionMixin, FeedForward
 from ..attention_dispatch import dispatch_attention_fn
 from ..attention_processor import Attention
@@ -783,10 +783,6 @@ class HunyuanVideo15Transformer3DModel(
         hidden_states = hidden_states.permute(0, 4, 1, 5, 2, 6, 3, 7)
         hidden_states = hidden_states.flatten(6, 7).flatten(4, 5).flatten(2, 3)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (hidden_states,)
 

src/diffusers/models/transformers/transformer_hunyuan_video_framepack.py

@@ -20,7 +20,7 @@ import torch.nn.functional as F
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import FromOriginalModelMixin, PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, get_logger, scale_lora_layers, unscale_lora_layers
+from ...utils import USE_PEFT_BACKEND, get_logger, scale_lora_layers
 from ..cache_utils import CacheMixin
 from ..embeddings import get_1d_rotary_pos_embed
 from ..modeling_outputs import Transformer2DModelOutput
@@ -337,10 +337,6 @@ class HunyuanVideoFramepackTransformer3DModel(
         hidden_states = hidden_states.permute(0, 4, 1, 5, 2, 6, 3, 7)
         hidden_states = hidden_states.flatten(6, 7).flatten(4, 5).flatten(2, 3)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (hidden_states,)
         return Transformer2DModelOutput(sample=hidden_states)

src/diffusers/models/transformers/transformer_hunyuanimage.py

@@ -23,7 +23,7 @@ from diffusers.loaders import FromOriginalModelMixin
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers
 from ...utils.torch_utils import maybe_allow_in_graph
 from ..attention import AttentionMixin, FeedForward
 from ..attention_dispatch import dispatch_attention_fn
@@ -900,10 +900,6 @@ class HunyuanImageTransformer2DModel(
         ]
         hidden_states = hidden_states.reshape(*final_dims)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (hidden_states,)
 

src/diffusers/models/transformers/transformer_ltx.py

@@ -22,7 +22,7 @@ import torch.nn as nn
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import FromOriginalModelMixin, PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, deprecate, is_torch_version, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import USE_PEFT_BACKEND, deprecate, is_torch_version, logging, scale_lora_layers
 from ...utils.torch_utils import maybe_allow_in_graph
 from .._modeling_parallel import ContextParallelInput, ContextParallelOutput
 from ..attention import AttentionMixin, AttentionModuleMixin, FeedForward
@@ -568,10 +568,6 @@ class LTXVideoTransformer3DModel(
         hidden_states = hidden_states * (1 + scale) + shift
         output = self.proj_out(hidden_states)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (output,)
         return Transformer2DModelOutput(sample=output)

src/diffusers/models/transformers/transformer_ltx2.py

@@ -28,7 +28,6 @@ from ...utils import (
     is_torch_version,
     logging,
     scale_lora_layers,
-    unscale_lora_layers,
 )
 from .._modeling_parallel import ContextParallelInput, ContextParallelOutput
 from ..attention import AttentionMixin, AttentionModuleMixin, FeedForward
@@ -1341,10 +1340,6 @@ class LTX2VideoTransformer3DModel(
         audio_hidden_states = audio_hidden_states * (1 + audio_scale) + audio_shift
         audio_output = self.audio_proj_out(audio_hidden_states)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (output, audio_output)
         return AudioVisualModelOutput(sample=output, audio_sample=audio_output)

src/diffusers/models/transformers/transformer_qwenimage.py

@@ -24,7 +24,7 @@ import torch.nn.functional as F
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import FromOriginalModelMixin, PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers
 from ...utils.torch_utils import maybe_allow_in_graph
 from .._modeling_parallel import ContextParallelInput, ContextParallelOutput
 from ..attention import AttentionMixin, FeedForward
@@ -981,10 +981,6 @@ class QwenImageTransformer2DModel(
         hidden_states = self.norm_out(hidden_states, temb)
         output = self.proj_out(hidden_states)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (output,)
 

src/diffusers/models/transformers/transformer_sd3.py

@@ -18,7 +18,7 @@ import torch.nn as nn
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import FromOriginalModelMixin, PeftAdapterMixin, SD3Transformer2DLoadersMixin
-from ...utils import USE_PEFT_BACKEND, apply_lora_scale, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import apply_lora_scale, logging
 from ...utils.torch_utils import maybe_allow_in_graph
 from ..attention import AttentionMixin, FeedForward, JointTransformerBlock
 from ..attention_processor import (
@@ -285,20 +285,6 @@ class SD3Transformer2DModel(
             If `return_dict` is True, an [`~models.transformer_2d.Transformer2DModelOutput`] is returned, otherwise a
             `tuple` where the first element is the sample tensor.
         """
-        if joint_attention_kwargs is not None:
-            joint_attention_kwargs = joint_attention_kwargs.copy()
-            lora_scale = joint_attention_kwargs.pop("scale", 1.0)
-        else:
-            lora_scale = 1.0
-
-        if USE_PEFT_BACKEND:
-            # weight the lora layers by setting `lora_scale` for each PEFT layer
-            scale_lora_layers(self, lora_scale)
-        else:
-            if joint_attention_kwargs is not None and joint_attention_kwargs.get("scale", None) is not None:
-                logger.warning(
-                    "Passing `scale` via `joint_attention_kwargs` when not using the PEFT backend is ineffective."
-                )
 
         height, width = hidden_states.shape[-2:]
 
@@ -353,10 +339,6 @@ class SD3Transformer2DModel(
             shape=(hidden_states.shape[0], self.out_channels, height * patch_size, width * patch_size)
         )
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (output,)
 

src/diffusers/models/transformers/transformer_skyreels_v2.py

@@ -21,7 +21,7 @@ import torch.nn.functional as F
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import FromOriginalModelMixin, PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers
 from ...utils.torch_utils import maybe_allow_in_graph
 from ..attention import AttentionMixin, AttentionModuleMixin, FeedForward
 from ..attention_dispatch import dispatch_attention_fn
@@ -771,10 +771,6 @@ class SkyReelsV2Transformer3DModel(
         hidden_states = hidden_states.permute(0, 7, 1, 4, 2, 5, 3, 6)
         output = hidden_states.flatten(6, 7).flatten(4, 5).flatten(2, 3)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (output,)
 

src/diffusers/models/transformers/transformer_wan.py

@@ -21,7 +21,7 @@ import torch.nn.functional as F
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import FromOriginalModelMixin, PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import USE_PEFT_BACKEND, deprecate, logging, scale_lora_layers
 from ...utils.torch_utils import maybe_allow_in_graph
 from .._modeling_parallel import ContextParallelInput, ContextParallelOutput
 from ..attention import AttentionMixin, AttentionModuleMixin, FeedForward
@@ -713,10 +713,6 @@ class WanTransformer3DModel(
         hidden_states = hidden_states.permute(0, 7, 1, 4, 2, 5, 3, 6)
         output = hidden_states.flatten(6, 7).flatten(4, 5).flatten(2, 3)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (output,)
 

src/diffusers/models/transformers/transformer_wan_animate.py

@@ -21,7 +21,7 @@ import torch.nn.functional as F
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import FromOriginalModelMixin, PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers
 from ..attention import AttentionMixin, AttentionModuleMixin, FeedForward
 from ..attention_dispatch import dispatch_attention_fn
 from ..cache_utils import CacheMixin
@@ -1293,10 +1293,6 @@ class WanAnimateTransformer3DModel(
         hidden_states = hidden_states.permute(0, 7, 1, 4, 2, 5, 3, 6)
         output = hidden_states.flatten(6, 7).flatten(4, 5).flatten(2, 3)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (output,)
 

src/diffusers/models/transformers/transformer_wan_vace.py

@@ -20,7 +20,7 @@ import torch.nn as nn
 
 from ...configuration_utils import ConfigMixin, register_to_config
 from ...loaders import FromOriginalModelMixin, PeftAdapterMixin
-from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers
 from ..attention import AttentionMixin, FeedForward
 from ..cache_utils import CacheMixin
 from ..modeling_outputs import Transformer2DModelOutput
@@ -379,10 +379,6 @@ class WanVACETransformer3DModel(
         hidden_states = hidden_states.permute(0, 7, 1, 4, 2, 5, 3, 6)
         output = hidden_states.flatten(6, 7).flatten(4, 5).flatten(2, 3)
 
-        if USE_PEFT_BACKEND:
-            # remove `lora_scale` from each PEFT layer
-            unscale_lora_layers(self, lora_scale)
-
         if not return_dict:
             return (output,)