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

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

This reverts commit 527430d0.

examples/research_projects/scheduled_huber_loss_training/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 (
@@ -999,7 +999,7 @@ 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,
@@ -1019,7 +1019,7 @@ 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)
@@ -1451,7 +1451,7 @@ 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/research_projects/scheduled_huber_loss_training/dreambooth/train_dreambooth_lora_sdxl.py

@@ -59,7 +59,7 @@ 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 (
@@ -1334,7 +1334,7 @@ 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/scheduled_huber_loss_training/text_to_image/train_text_to_image_lora_sdxl.py

@@ -49,7 +49,7 @@ 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 (
@@ -749,7 +749,7 @@ def main(args):
             else:
                 raise ValueError(f"unexpected save model: {model.__class__}")
 
-        lora_state_dict, _ = StableDiffusionLoraLoaderMixin.lora_state_dict(input_dir)
+        lora_state_dict, _ = 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)
         incompatible_keys = set_peft_model_state_dict(unet_, unet_state_dict, adapter_name="default")

examples/text_to_image/train_text_to_image_lora_sdxl.py

@@ -50,7 +50,7 @@ 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 (
@@ -766,7 +766,7 @@ def main(args):
             else:
                 raise ValueError(f"unexpected save model: {model.__class__}")
 
-        lora_state_dict, _ = StableDiffusionLoraLoaderMixin.lora_state_dict(input_dir)
+        lora_state_dict, _ = 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)
         incompatible_keys = set_peft_model_state_dict(unet_, unet_state_dict, adapter_name="default")

src/diffusers/loaders/__init__.py

@@ -55,18 +55,11 @@ _import_structure = {}
 
 if is_torch_available():
     _import_structure["single_file_model"] = ["FromOriginalModelMixin"]
-    _import_structure["transformer_sd3"] = ["SD3TransformerLoRALoadersMixin"]
-
     _import_structure["unet"] = ["UNet2DConditionLoadersMixin"]
     _import_structure["utils"] = ["AttnProcsLayers"]
     if is_transformers_available():
         _import_structure["single_file"] = ["FromSingleFileMixin"]
-        _import_structure["lora_pipeline"] = [
-            "AmusedLoraLoaderMixin",
-            "StableDiffusionLoraLoaderMixin",
-            "SD3LoraLoaderMixin",
-            "StableDiffusionXLLoraLoaderMixin",
-        ]
+        _import_structure["lora"] = ["LoraLoaderMixin", "StableDiffusionXLLoraLoaderMixin", "SD3LoraLoaderMixin"]
         _import_structure["textual_inversion"] = ["TextualInversionLoaderMixin"]
         _import_structure["ip_adapter"] = ["IPAdapterMixin"]
 
@@ -81,12 +74,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
 
         if is_transformers_available():
             from .ip_adapter import IPAdapterMixin
-            from .lora_pipeline import (
-                AmusedLoraLoaderMixin,
-                SD3LoraLoaderMixin,
-                StableDiffusionLoraLoaderMixin,
-                StableDiffusionXLLoraLoaderMixin,
-            )
+            from .lora import LoraLoaderMixin, SD3LoraLoaderMixin, StableDiffusionXLLoraLoaderMixin
             from .single_file import FromSingleFileMixin
             from .textual_inversion import TextualInversionLoaderMixin
 

src/diffusers/loaders/lora_pipeline.py → src/diffusers/loaders/lora.py

@@ -11,32 +11,49 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+import copy
+import inspect
 import os
+from pathlib import Path
 from typing import Callable, Dict, List, Optional, Union
 
+import safetensors
 import torch
+from huggingface_hub import model_info
+from huggingface_hub.constants import HF_HUB_OFFLINE
 from huggingface_hub.utils import validate_hf_hub_args
+from torch import nn
 
+from ..models.modeling_utils import load_state_dict
 from ..utils import (
     USE_PEFT_BACKEND,
+    _get_model_file,
     convert_state_dict_to_diffusers,
     convert_state_dict_to_peft,
     convert_unet_state_dict_to_peft,
-    deprecate,
+    delete_adapter_layers,
     get_adapter_name,
     get_peft_kwargs,
+    is_accelerate_available,
     is_peft_version,
     is_transformers_available,
     logging,
+    recurse_remove_peft_layers,
     scale_lora_layers,
+    set_adapter_layers,
+    set_weights_and_activate_adapters,
 )
-from .lora_base import LoraBaseMixin
 from .lora_conversion_utils import _convert_non_diffusers_lora_to_diffusers, _maybe_map_sgm_blocks_to_diffusers
 
 
 if is_transformers_available():
+    from transformers import PreTrainedModel
+
     from ..models.lora import text_encoder_attn_modules, text_encoder_mlp_modules
 
+if is_accelerate_available():
+    from accelerate.hooks import AlignDevicesHook, CpuOffload, remove_hook_from_module
+
 logger = logging.get_logger(__name__)
 
 TEXT_ENCODER_NAME = "text_encoder"
@@ -46,16 +63,19 @@ TRANSFORMER_NAME = "transformer"
 LORA_WEIGHT_NAME = "pytorch_lora_weights.bin"
 LORA_WEIGHT_NAME_SAFE = "pytorch_lora_weights.safetensors"
 
+LORA_DEPRECATION_MESSAGE = "You are using an old version of LoRA backend. This will be deprecated in the next releases in favor of PEFT make sure to install the latest PEFT and transformers packages in the future."
 
-class StableDiffusionLoraLoaderMixin(LoraBaseMixin):
+
+class LoraLoaderMixin:
     r"""
-    Load LoRA layers into Stable Diffusion [`UNet2DConditionModel`] and
+    Load LoRA layers into [`UNet2DConditionModel`] and
     [`CLIPTextModel`](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModel).
     """
 
-    _lora_loadable_modules = ["unet", "text_encoder"]
-    unet_name = UNET_NAME
     text_encoder_name = TEXT_ENCODER_NAME
+    unet_name = UNET_NAME
+    transformer_name = TRANSFORMER_NAME
+    num_fused_loras = 0
 
     def load_lora_weights(
         self, pretrained_model_name_or_path_or_dict: Union[str, Dict[str, torch.Tensor]], adapter_name=None, **kwargs
@@ -66,20 +86,19 @@ class StableDiffusionLoraLoaderMixin(LoraBaseMixin):
 
         All kwargs are forwarded to `self.lora_state_dict`.
 
-        See [`~loaders.StableDiffusionLoraLoaderMixin.lora_state_dict`] for more details on how the state dict is
-        loaded.
+        See [`~loaders.LoraLoaderMixin.lora_state_dict`] for more details on how the state dict is loaded.
 
-        See [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_into_unet`] for more details on how the state dict is
-        loaded into `self.unet`.
+        See [`~loaders.LoraLoaderMixin.load_lora_into_unet`] for more details on how the state dict is loaded into
+        `self.unet`.
 
-        See [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_into_text_encoder`] for more details on how the state
-        dict is loaded into `self.text_encoder`.
+        See [`~loaders.LoraLoaderMixin.load_lora_into_text_encoder`] for more details on how the state dict is loaded
+        into `self.text_encoder`.
 
         Parameters:
             pretrained_model_name_or_path_or_dict (`str` or `os.PathLike` or `dict`):
-                See [`~loaders.StableDiffusionLoraLoaderMixin.lora_state_dict`].
+                See [`~loaders.LoraLoaderMixin.lora_state_dict`].
             kwargs (`dict`, *optional*):
-                See [`~loaders.StableDiffusionLoraLoaderMixin.lora_state_dict`].
+                See [`~loaders.LoraLoaderMixin.lora_state_dict`].
             adapter_name (`str`, *optional*):
                 Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
                 `default_{i}` where i is the total number of adapters being loaded.
@@ -192,20 +211,62 @@ class StableDiffusionLoraLoaderMixin(LoraBaseMixin):
             "framework": "pytorch",
         }
 
-        state_dict = cls._fetch_state_dict(
-            pretrained_model_name_or_path_or_dict=pretrained_model_name_or_path_or_dict,
-            weight_name=weight_name,
-            use_safetensors=use_safetensors,
+        model_file = None
+        if not isinstance(pretrained_model_name_or_path_or_dict, dict):
+            # Let's first try to load .safetensors weights
+            if (use_safetensors and weight_name is None) or (
+                weight_name is not None and weight_name.endswith(".safetensors")
+            ):
+                try:
+                    # Here we're relaxing the loading check to enable more Inference API
+                    # friendliness where sometimes, it's not at all possible to automatically
+                    # determine `weight_name`.
+                    if weight_name is None:
+                        weight_name = cls._best_guess_weight_name(
+                            pretrained_model_name_or_path_or_dict,
+                            file_extension=".safetensors",
+                            local_files_only=local_files_only,
+                        )
+                    model_file = _get_model_file(
+                        pretrained_model_name_or_path_or_dict,
+                        weights_name=weight_name or LORA_WEIGHT_NAME_SAFE,
+                        cache_dir=cache_dir,
+                        force_download=force_download,
+                        proxies=proxies,
                         local_files_only=local_files_only,
+                        token=token,
+                        revision=revision,
+                        subfolder=subfolder,
+                        user_agent=user_agent,
+                    )
+                    state_dict = safetensors.torch.load_file(model_file, device="cpu")
+                except (IOError, safetensors.SafetensorError) as e:
+                    if not allow_pickle:
+                        raise e
+                    # try loading non-safetensors weights
+                    model_file = None
+                    pass
+
+            if model_file is None:
+                if weight_name is None:
+                    weight_name = cls._best_guess_weight_name(
+                        pretrained_model_name_or_path_or_dict, file_extension=".bin", local_files_only=local_files_only
+                    )
+                model_file = _get_model_file(
+                    pretrained_model_name_or_path_or_dict,
+                    weights_name=weight_name or LORA_WEIGHT_NAME,
                     cache_dir=cache_dir,
                     force_download=force_download,
                     proxies=proxies,
+                    local_files_only=local_files_only,
                     token=token,
                     revision=revision,
                     subfolder=subfolder,
                     user_agent=user_agent,
-            allow_pickle=allow_pickle,
                 )
+                state_dict = load_state_dict(model_file)
+        else:
+            state_dict = pretrained_model_name_or_path_or_dict
 
         network_alphas = None
         # TODO: replace it with a method from `state_dict_utils`
@@ -226,6 +287,82 @@ class StableDiffusionLoraLoaderMixin(LoraBaseMixin):
 
         return state_dict, network_alphas
 
+    @classmethod
+    def _best_guess_weight_name(
+        cls, pretrained_model_name_or_path_or_dict, file_extension=".safetensors", local_files_only=False
+    ):
+        if local_files_only or HF_HUB_OFFLINE:
+            raise ValueError("When using the offline mode, you must specify a `weight_name`.")
+
+        targeted_files = []
+
+        if os.path.isfile(pretrained_model_name_or_path_or_dict):
+            return
+        elif os.path.isdir(pretrained_model_name_or_path_or_dict):
+            targeted_files = [
+                f for f in os.listdir(pretrained_model_name_or_path_or_dict) if f.endswith(file_extension)
+            ]
+        else:
+            files_in_repo = model_info(pretrained_model_name_or_path_or_dict).siblings
+            targeted_files = [f.rfilename for f in files_in_repo if f.rfilename.endswith(file_extension)]
+        if len(targeted_files) == 0:
+            return
+
+        # "scheduler" does not correspond to a LoRA checkpoint.
+        # "optimizer" does not correspond to a LoRA checkpoint
+        # only top-level checkpoints are considered and not the other ones, hence "checkpoint".
+        unallowed_substrings = {"scheduler", "optimizer", "checkpoint"}
+        targeted_files = list(
+            filter(lambda x: all(substring not in x for substring in unallowed_substrings), targeted_files)
+        )
+
+        if any(f.endswith(LORA_WEIGHT_NAME) for f in targeted_files):
+            targeted_files = list(filter(lambda x: x.endswith(LORA_WEIGHT_NAME), targeted_files))
+        elif any(f.endswith(LORA_WEIGHT_NAME_SAFE) for f in targeted_files):
+            targeted_files = list(filter(lambda x: x.endswith(LORA_WEIGHT_NAME_SAFE), targeted_files))
+
+        if len(targeted_files) > 1:
+            raise ValueError(
+                f"Provided path contains more than one weights file in the {file_extension} format. Either specify `weight_name` in `load_lora_weights` or make sure there's only one  `.safetensors` or `.bin` file in  {pretrained_model_name_or_path_or_dict}."
+            )
+        weight_name = targeted_files[0]
+        return weight_name
+
+    @classmethod
+    def _optionally_disable_offloading(cls, _pipeline):
+        """
+        Optionally removes offloading in case the pipeline has been already sequentially offloaded to CPU.
+
+        Args:
+            _pipeline (`DiffusionPipeline`):
+                The pipeline to disable offloading for.
+
+        Returns:
+            tuple:
+                A tuple indicating if `is_model_cpu_offload` or `is_sequential_cpu_offload` is True.
+        """
+        is_model_cpu_offload = False
+        is_sequential_cpu_offload = False
+
+        if _pipeline is not None and _pipeline.hf_device_map is None:
+            for _, component in _pipeline.components.items():
+                if isinstance(component, nn.Module) and hasattr(component, "_hf_hook"):
+                    if not is_model_cpu_offload:
+                        is_model_cpu_offload = isinstance(component._hf_hook, CpuOffload)
+                    if not is_sequential_cpu_offload:
+                        is_sequential_cpu_offload = (
+                            isinstance(component._hf_hook, AlignDevicesHook)
+                            or hasattr(component._hf_hook, "hooks")
+                            and isinstance(component._hf_hook.hooks[0], AlignDevicesHook)
+                        )
+
+                    logger.info(
+                        "Accelerate hooks detected. Since you have called `load_lora_weights()`, the previous hooks will be first removed. Then the LoRA parameters will be loaded and the hooks will be applied again."
+                    )
+                    remove_hook_from_module(component, recurse=is_sequential_cpu_offload)
+
+        return (is_model_cpu_offload, is_sequential_cpu_offload)
+
     @classmethod
     def load_lora_into_unet(cls, state_dict, network_alphas, unet, adapter_name=None, _pipeline=None):
         """
@@ -379,12 +516,115 @@ class StableDiffusionLoraLoaderMixin(LoraBaseMixin):
                     _pipeline.enable_sequential_cpu_offload()
                 # Unsafe code />
 
+    @classmethod
+    def load_lora_into_transformer(cls, state_dict, network_alphas, transformer, adapter_name=None, _pipeline=None):
+        """
+        This will load the LoRA layers specified in `state_dict` into `transformer`.
+
+        Parameters:
+            state_dict (`dict`):
+                A standard state dict containing the lora layer parameters. The keys can either be indexed directly
+                into the unet or prefixed with an additional `unet` which can be used to distinguish between text
+                encoder lora layers.
+            network_alphas (`Dict[str, float]`):
+                See `LoRALinearLayer` for more details.
+            unet (`UNet2DConditionModel`):
+                The UNet model to load the LoRA layers into.
+            adapter_name (`str`, *optional*):
+                Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
+                `default_{i}` where i is the total number of adapters being loaded.
+        """
+        from peft import LoraConfig, inject_adapter_in_model, set_peft_model_state_dict
+
+        keys = list(state_dict.keys())
+
+        transformer_keys = [k for k in keys if k.startswith(cls.transformer_name)]
+        state_dict = {
+            k.replace(f"{cls.transformer_name}.", ""): v for k, v in state_dict.items() if k in transformer_keys
+        }
+
+        if network_alphas is not None:
+            alpha_keys = [k for k in network_alphas.keys() if k.startswith(cls.transformer_name)]
+            network_alphas = {
+                k.replace(f"{cls.transformer_name}.", ""): v for k, v in network_alphas.items() if k in alpha_keys
+            }
+
+        if len(state_dict.keys()) > 0:
+            if adapter_name in getattr(transformer, "peft_config", {}):
+                raise ValueError(
+                    f"Adapter name {adapter_name} already in use in the transformer - please select a new adapter name."
+                )
+
+            rank = {}
+            for key, val in state_dict.items():
+                if "lora_B" in key:
+                    rank[key] = val.shape[1]
+
+            lora_config_kwargs = get_peft_kwargs(rank, network_alphas, state_dict)
+            if "use_dora" in lora_config_kwargs:
+                if lora_config_kwargs["use_dora"] and is_peft_version("<", "0.9.0"):
+                    raise ValueError(
+                        "You need `peft` 0.9.0 at least to use DoRA-enabled LoRAs. Please upgrade your installation of `peft`."
+                    )
+                else:
+                    lora_config_kwargs.pop("use_dora")
+            lora_config = LoraConfig(**lora_config_kwargs)
+
+            # adapter_name
+            if adapter_name is None:
+                adapter_name = get_adapter_name(transformer)
+
+            # In case the pipeline has been already offloaded to CPU - temporarily remove the hooks
+            # otherwise loading LoRA weights will lead to an error
+            is_model_cpu_offload, is_sequential_cpu_offload = cls._optionally_disable_offloading(_pipeline)
+
+            inject_adapter_in_model(lora_config, transformer, adapter_name=adapter_name)
+            incompatible_keys = set_peft_model_state_dict(transformer, state_dict, adapter_name)
+
+            if incompatible_keys is not None:
+                # check only for unexpected keys
+                unexpected_keys = getattr(incompatible_keys, "unexpected_keys", None)
+                if unexpected_keys:
+                    logger.warning(
+                        f"Loading adapter weights from state_dict led to unexpected keys not found in the model: "
+                        f" {unexpected_keys}. "
+                    )
+
+            # Offload back.
+            if is_model_cpu_offload:
+                _pipeline.enable_model_cpu_offload()
+            elif is_sequential_cpu_offload:
+                _pipeline.enable_sequential_cpu_offload()
+            # Unsafe code />
+
+    @property
+    def lora_scale(self) -> float:
+        # property function that returns the lora scale which can be set at run time by the pipeline.
+        # if _lora_scale has not been set, return 1
+        return self._lora_scale if hasattr(self, "_lora_scale") else 1.0
+
+    def _remove_text_encoder_monkey_patch(self):
+        remove_method = recurse_remove_peft_layers
+        if hasattr(self, "text_encoder"):
+            remove_method(self.text_encoder)
+            # In case text encoder have no Lora attached
+            if getattr(self.text_encoder, "peft_config", None) is not None:
+                del self.text_encoder.peft_config
+                self.text_encoder._hf_peft_config_loaded = None
+
+        if hasattr(self, "text_encoder_2"):
+            remove_method(self.text_encoder_2)
+            if getattr(self.text_encoder_2, "peft_config", None) is not None:
+                del self.text_encoder_2.peft_config
+                self.text_encoder_2._hf_peft_config_loaded = None
+
     @classmethod
     def save_lora_weights(
         cls,
         save_directory: Union[str, os.PathLike],
         unet_lora_layers: Dict[str, Union[torch.nn.Module, torch.Tensor]] = None,
         text_encoder_lora_layers: Dict[str, torch.nn.Module] = None,
+        transformer_lora_layers: Dict[str, torch.nn.Module] = None,
         is_main_process: bool = True,
         weight_name: str = None,
         save_function: Callable = None,
@@ -414,14 +654,24 @@ class StableDiffusionLoraLoaderMixin(LoraBaseMixin):
         """
         state_dict = {}
 
-        if not (unet_lora_layers or text_encoder_lora_layers):
-            raise ValueError("You must pass at least one of `unet_lora_layers` and `text_encoder_lora_layers`.")
+        def pack_weights(layers, prefix):
+            layers_weights = layers.state_dict() if isinstance(layers, torch.nn.Module) else layers
+            layers_state_dict = {f"{prefix}.{module_name}": param for module_name, param in layers_weights.items()}
+            return layers_state_dict
+
+        if not (unet_lora_layers or text_encoder_lora_layers or transformer_lora_layers):
+            raise ValueError(
+                "You must pass at least one of `unet_lora_layers`, `text_encoder_lora_layers`, or `transformer_lora_layers`."
+            )
 
         if unet_lora_layers:
-            state_dict.update(cls.pack_weights(unet_lora_layers, cls.unet_name))
+            state_dict.update(pack_weights(unet_lora_layers, cls.unet_name))
 
         if text_encoder_lora_layers:
-            state_dict.update(cls.pack_weights(text_encoder_lora_layers, cls.text_encoder_name))
+            state_dict.update(pack_weights(text_encoder_lora_layers, cls.text_encoder_name))
+
+        if transformer_lora_layers:
+            state_dict.update(pack_weights(transformer_lora_layers, "transformer"))
 
         # Save the model
         cls.write_lora_layers(
@@ -433,13 +683,68 @@ class StableDiffusionLoraLoaderMixin(LoraBaseMixin):
             safe_serialization=safe_serialization,
         )
 
+    @staticmethod
+    def write_lora_layers(
+        state_dict: Dict[str, torch.Tensor],
+        save_directory: str,
+        is_main_process: bool,
+        weight_name: str,
+        save_function: Callable,
+        safe_serialization: bool,
+    ):
+        if os.path.isfile(save_directory):
+            logger.error(f"Provided path ({save_directory}) should be a directory, not a file")
+            return
+
+        if save_function is None:
+            if safe_serialization:
+
+                def save_function(weights, filename):
+                    return safetensors.torch.save_file(weights, filename, metadata={"format": "pt"})
+
+            else:
+                save_function = torch.save
+
+        os.makedirs(save_directory, exist_ok=True)
+
+        if weight_name is None:
+            if safe_serialization:
+                weight_name = LORA_WEIGHT_NAME_SAFE
+            else:
+                weight_name = LORA_WEIGHT_NAME
+
+        save_path = Path(save_directory, weight_name).as_posix()
+        save_function(state_dict, save_path)
+        logger.info(f"Model weights saved in {save_path}")
+
+    def unload_lora_weights(self):
+        """
+        Unloads the LoRA parameters.
+
+        Examples:
+
+        ```python
+        >>> # Assuming `pipeline` is already loaded with the LoRA parameters.
+        >>> pipeline.unload_lora_weights()
+        >>> ...
+        ```
+        """
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for this method.")
+
+        unet = getattr(self, self.unet_name) if not hasattr(self, "unet") else self.unet
+        unet.unload_lora()
+
+        # Safe to call the following regardless of LoRA.
+        self._remove_text_encoder_monkey_patch()
+
     def fuse_lora(
         self,
-        components: List[str] = ["unet", "text_encoder"],
+        fuse_unet: bool = True,
+        fuse_text_encoder: bool = True,
         lora_scale: float = 1.0,
         safe_fusing: bool = False,
         adapter_names: Optional[List[str]] = None,
-        **kwargs,
     ):
         r"""
         Fuses the LoRA parameters into the original parameters of the corresponding blocks.
@@ -451,7 +756,10 @@ class StableDiffusionLoraLoaderMixin(LoraBaseMixin):
         </Tip>
 
         Args:
-            components: (`List[str]`): List of LoRA-injectable components to fuse the LoRAs into.
+            fuse_unet (`bool`, defaults to `True`): Whether to fuse the UNet LoRA parameters.
+            fuse_text_encoder (`bool`, defaults to `True`):
+                Whether to fuse the text encoder LoRA parameters. If the text encoder wasn't monkey-patched with the
+                LoRA parameters then it won't have any effect.
             lora_scale (`float`, defaults to 1.0):
                 Controls how much to influence the outputs with the LoRA parameters.
             safe_fusing (`bool`, defaults to `False`):
@@ -472,14 +780,50 @@ class StableDiffusionLoraLoaderMixin(LoraBaseMixin):
         pipeline.fuse_lora(lora_scale=0.7)
         ```
         """
-        super().fuse_lora(
-            components=components, lora_scale=lora_scale, safe_fusing=safe_fusing, adapter_names=adapter_names
+        from peft.tuners.tuners_utils import BaseTunerLayer
+
+        if fuse_unet or fuse_text_encoder:
+            self.num_fused_loras += 1
+            if self.num_fused_loras > 1:
+                logger.warning(
+                    "The current API is supported for operating with a single LoRA file. You are trying to load and fuse more than one LoRA which is not well-supported.",
+                )
+
+        if fuse_unet:
+            unet = getattr(self, self.unet_name) if not hasattr(self, "unet") else self.unet
+            unet.fuse_lora(lora_scale, safe_fusing=safe_fusing, adapter_names=adapter_names)
+
+        def fuse_text_encoder_lora(text_encoder, lora_scale=1.0, safe_fusing=False, adapter_names=None):
+            merge_kwargs = {"safe_merge": safe_fusing}
+
+            for module in text_encoder.modules():
+                if isinstance(module, BaseTunerLayer):
+                    if lora_scale != 1.0:
+                        module.scale_layer(lora_scale)
+
+                    # For BC with previous PEFT versions, we need to check the signature
+                    # of the `merge` method to see if it supports the `adapter_names` argument.
+                    supported_merge_kwargs = list(inspect.signature(module.merge).parameters)
+                    if "adapter_names" in supported_merge_kwargs:
+                        merge_kwargs["adapter_names"] = adapter_names
+                    elif "adapter_names" not in supported_merge_kwargs and adapter_names is not None:
+                        raise ValueError(
+                            "The `adapter_names` argument is not supported with your PEFT version. "
+                            "Please upgrade to the latest version of PEFT. `pip install -U peft`"
                         )
 
-    def unfuse_lora(self, components: List[str] = ["unet", "text_encoder"], **kwargs):
+                    module.merge(**merge_kwargs)
+
+        if fuse_text_encoder:
+            if hasattr(self, "text_encoder"):
+                fuse_text_encoder_lora(self.text_encoder, lora_scale, safe_fusing, adapter_names=adapter_names)
+            if hasattr(self, "text_encoder_2"):
+                fuse_text_encoder_lora(self.text_encoder_2, lora_scale, safe_fusing, adapter_names=adapter_names)
+
+    def unfuse_lora(self, unfuse_unet: bool = True, unfuse_text_encoder: bool = True):
         r"""
         Reverses the effect of
-        [`pipe.fuse_lora()`](https://huggingface.co/docs/diffusers/main/en/api/loaders#diffusers.loaders.LoraBaseMixin.fuse_lora).
+        [`pipe.fuse_lora()`](https://huggingface.co/docs/diffusers/main/en/api/loaders#diffusers.loaders.LoraLoaderMixin.fuse_lora).
 
         <Tip warning={true}>
 
@@ -488,871 +832,434 @@ class StableDiffusionLoraLoaderMixin(LoraBaseMixin):
         </Tip>
 
         Args:
-            components (`List[str]`): List of LoRA-injectable components to unfuse LoRA from.
             unfuse_unet (`bool`, defaults to `True`): Whether to unfuse the UNet LoRA parameters.
             unfuse_text_encoder (`bool`, defaults to `True`):
                 Whether to unfuse the text encoder LoRA parameters. If the text encoder wasn't monkey-patched with the
                 LoRA parameters then it won't have any effect.
         """
-        super().unfuse_lora(components=components)
+        from peft.tuners.tuners_utils import BaseTunerLayer
 
+        unet = getattr(self, self.unet_name) if not hasattr(self, "unet") else self.unet
+        if unfuse_unet:
+            for module in unet.modules():
+                if isinstance(module, BaseTunerLayer):
+                    module.unmerge()
 
-class StableDiffusionXLLoraLoaderMixin(LoraBaseMixin):
-    r"""
-    Load LoRA layers into Stable Diffusion XL [`UNet2DConditionModel`],
-    [`CLIPTextModel`](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModel), and
-    [`CLIPTextModelWithProjection`](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModelWithProjection).
-    """
+        def unfuse_text_encoder_lora(text_encoder):
+            for module in text_encoder.modules():
+                if isinstance(module, BaseTunerLayer):
+                    module.unmerge()
 
-    _lora_loadable_modules = ["unet", "text_encoder", "text_encoder_2"]
-    unet_name = UNET_NAME
-    text_encoder_name = TEXT_ENCODER_NAME
+        if unfuse_text_encoder:
+            if hasattr(self, "text_encoder"):
+                unfuse_text_encoder_lora(self.text_encoder)
+            if hasattr(self, "text_encoder_2"):
+                unfuse_text_encoder_lora(self.text_encoder_2)
 
-    def load_lora_weights(
+        self.num_fused_loras -= 1
+
+    def set_adapters_for_text_encoder(
         self,
-        pretrained_model_name_or_path_or_dict: Union[str, Dict[str, torch.Tensor]],
-        adapter_name: Optional[str] = None,
-        **kwargs,
+        adapter_names: Union[List[str], str],
+        text_encoder: Optional["PreTrainedModel"] = None,  # noqa: F821
+        text_encoder_weights: Optional[Union[float, List[float], List[None]]] = None,
     ):
         """
-        Load LoRA weights specified in `pretrained_model_name_or_path_or_dict` into `self.unet` and
-        `self.text_encoder`.
+        Sets the adapter layers for the text encoder.
 
-        All kwargs are forwarded to `self.lora_state_dict`.
+        Args:
+            adapter_names (`List[str]` or `str`):
+                The names of the adapters to use.
+            text_encoder (`torch.nn.Module`, *optional*):
+                The text encoder module to set the adapter layers for. If `None`, it will try to get the `text_encoder`
+                attribute.
+            text_encoder_weights (`List[float]`, *optional*):
+                The weights to use for the text encoder. If `None`, the weights are set to `1.0` for all the adapters.
+        """
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for this method.")
+
+        def process_weights(adapter_names, weights):
+            # Expand weights into a list, one entry per adapter
+            # e.g. for 2 adapters:  7 -> [7,7] ; [3, None] -> [3, None]
+            if not isinstance(weights, list):
+                weights = [weights] * len(adapter_names)
 
-        See [`~loaders.StableDiffusionLoraLoaderMixin.lora_state_dict`] for more details on how the state dict is
-        loaded.
+            if len(adapter_names) != len(weights):
+                raise ValueError(
+                    f"Length of adapter names {len(adapter_names)} is not equal to the length of the weights {len(weights)}"
+                )
 
-        See [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_into_unet`] for more details on how the state dict is
-        loaded into `self.unet`.
+            # Set None values to default of 1.0
+            # e.g. [7,7] -> [7,7] ; [3, None] -> [3,1]
+            weights = [w if w is not None else 1.0 for w in weights]
 
-        See [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_into_text_encoder`] for more details on how the state
-        dict is loaded into `self.text_encoder`.
+            return weights
 
-        Parameters:
-            pretrained_model_name_or_path_or_dict (`str` or `os.PathLike` or `dict`):
-                See [`~loaders.StableDiffusionLoraLoaderMixin.lora_state_dict`].
-            adapter_name (`str`, *optional*):
-                Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
-                `default_{i}` where i is the total number of adapters being loaded.
-            kwargs (`dict`, *optional*):
-                See [`~loaders.StableDiffusionLoraLoaderMixin.lora_state_dict`].
+        adapter_names = [adapter_names] if isinstance(adapter_names, str) else adapter_names
+        text_encoder_weights = process_weights(adapter_names, text_encoder_weights)
+        text_encoder = text_encoder or getattr(self, "text_encoder", None)
+        if text_encoder is None:
+            raise ValueError(
+                "The pipeline does not have a default `pipe.text_encoder` class. Please make sure to pass a `text_encoder` instead."
+            )
+        set_weights_and_activate_adapters(text_encoder, adapter_names, text_encoder_weights)
+
+    def disable_lora_for_text_encoder(self, text_encoder: Optional["PreTrainedModel"] = None):
+        """
+        Disables the LoRA layers for the text encoder.
+
+        Args:
+            text_encoder (`torch.nn.Module`, *optional*):
+                The text encoder module to disable the LoRA layers for. If `None`, it will try to get the
+                `text_encoder` attribute.
         """
         if not USE_PEFT_BACKEND:
             raise ValueError("PEFT backend is required for this method.")
 
-        # We could have accessed the unet config from `lora_state_dict()` too. We pass
-        # it here explicitly to be able to tell that it's coming from an SDXL
-        # pipeline.
+        text_encoder = text_encoder or getattr(self, "text_encoder", None)
+        if text_encoder is None:
+            raise ValueError("Text Encoder not found.")
+        set_adapter_layers(text_encoder, enabled=False)
 
-        # if a dict is passed, copy it instead of modifying it inplace
-        if isinstance(pretrained_model_name_or_path_or_dict, dict):
-            pretrained_model_name_or_path_or_dict = pretrained_model_name_or_path_or_dict.copy()
+    def enable_lora_for_text_encoder(self, text_encoder: Optional["PreTrainedModel"] = None):
+        """
+        Enables the LoRA layers for the text encoder.
 
-        # First, ensure that the checkpoint is a compatible one and can be successfully loaded.
-        state_dict, network_alphas = self.lora_state_dict(
-            pretrained_model_name_or_path_or_dict,
-            unet_config=self.unet.config,
-            **kwargs,
-        )
-        is_correct_format = all("lora" in key or "dora_scale" in key for key in state_dict.keys())
-        if not is_correct_format:
-            raise ValueError("Invalid LoRA checkpoint.")
+        Args:
+            text_encoder (`torch.nn.Module`, *optional*):
+                The text encoder module to enable the LoRA layers for. If `None`, it will try to get the `text_encoder`
+                attribute.
+        """
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for this method.")
+        text_encoder = text_encoder or getattr(self, "text_encoder", None)
+        if text_encoder is None:
+            raise ValueError("Text Encoder not found.")
+        set_adapter_layers(self.text_encoder, enabled=True)
 
-        self.load_lora_into_unet(
-            state_dict, network_alphas=network_alphas, unet=self.unet, adapter_name=adapter_name, _pipeline=self
+    def set_adapters(
+        self,
+        adapter_names: Union[List[str], str],
+        adapter_weights: Optional[Union[float, Dict, List[float], List[Dict]]] = None,
+    ):
+        adapter_names = [adapter_names] if isinstance(adapter_names, str) else adapter_names
+
+        adapter_weights = copy.deepcopy(adapter_weights)
+
+        # Expand weights into a list, one entry per adapter
+        if not isinstance(adapter_weights, list):
+            adapter_weights = [adapter_weights] * len(adapter_names)
+
+        if len(adapter_names) != len(adapter_weights):
+            raise ValueError(
+                f"Length of adapter names {len(adapter_names)} is not equal to the length of the weights {len(adapter_weights)}"
             )
-        text_encoder_state_dict = {k: v for k, v in state_dict.items() if "text_encoder." in k}
-        if len(text_encoder_state_dict) > 0:
-            self.load_lora_into_text_encoder(
-                text_encoder_state_dict,
-                network_alphas=network_alphas,
-                text_encoder=self.text_encoder,
-                prefix="text_encoder",
-                lora_scale=self.lora_scale,
-                adapter_name=adapter_name,
-                _pipeline=self,
+
+        # Decompose weights into weights for unet, text_encoder and text_encoder_2
+        unet_lora_weights, text_encoder_lora_weights, text_encoder_2_lora_weights = [], [], []
+
+        list_adapters = self.get_list_adapters()  # eg {"unet": ["adapter1", "adapter2"], "text_encoder": ["adapter2"]}
+        all_adapters = {
+            adapter for adapters in list_adapters.values() for adapter in adapters
+        }  # eg ["adapter1", "adapter2"]
+        invert_list_adapters = {
+            adapter: [part for part, adapters in list_adapters.items() if adapter in adapters]
+            for adapter in all_adapters
+        }  # eg {"adapter1": ["unet"], "adapter2": ["unet", "text_encoder"]}
+
+        for adapter_name, weights in zip(adapter_names, adapter_weights):
+            if isinstance(weights, dict):
+                unet_lora_weight = weights.pop("unet", None)
+                text_encoder_lora_weight = weights.pop("text_encoder", None)
+                text_encoder_2_lora_weight = weights.pop("text_encoder_2", None)
+
+                if len(weights) > 0:
+                    raise ValueError(
+                        f"Got invalid key '{weights.keys()}' in lora weight dict for adapter {adapter_name}."
                     )
 
-        text_encoder_2_state_dict = {k: v for k, v in state_dict.items() if "text_encoder_2." in k}
-        if len(text_encoder_2_state_dict) > 0:
-            self.load_lora_into_text_encoder(
-                text_encoder_2_state_dict,
-                network_alphas=network_alphas,
-                text_encoder=self.text_encoder_2,
-                prefix="text_encoder_2",
-                lora_scale=self.lora_scale,
-                adapter_name=adapter_name,
-                _pipeline=self,
+                if text_encoder_2_lora_weight is not None and not hasattr(self, "text_encoder_2"):
+                    logger.warning(
+                        "Lora weight dict contains text_encoder_2 weights but will be ignored because pipeline does not have text_encoder_2."
                     )
 
-    @classmethod
-    @validate_hf_hub_args
-    # Copied from diffusers.loaders.lora_pipeline.StableDiffusionLoraLoaderMixin.lora_state_dict
-    def lora_state_dict(
-        cls,
-        pretrained_model_name_or_path_or_dict: Union[str, Dict[str, torch.Tensor]],
-        **kwargs,
+                # warn if adapter doesn't have parts specified by adapter_weights
+                for part_weight, part_name in zip(
+                    [unet_lora_weight, text_encoder_lora_weight, text_encoder_2_lora_weight],
+                    ["unet", "text_encoder", "text_encoder_2"],
                 ):
-        r"""
-        Return state dict for lora weights and the network alphas.
-
-        <Tip warning={true}>
-
-        We support loading A1111 formatted LoRA checkpoints in a limited capacity.
-
-        This function is experimental and might change in the future.
-
-        </Tip>
-
-        Parameters:
-            pretrained_model_name_or_path_or_dict (`str` or `os.PathLike` or `dict`):
-                Can be either:
-
-                    - A string, the *model id* (for example `google/ddpm-celebahq-256`) of a pretrained model hosted on
-                      the Hub.
-                    - A path to a *directory* (for example `./my_model_directory`) containing the model weights saved
-                      with [`ModelMixin.save_pretrained`].
-                    - A [torch state
-                      dict](https://pytorch.org/tutorials/beginner/saving_loading_models.html#what-is-a-state-dict).
-
-            cache_dir (`Union[str, os.PathLike]`, *optional*):
-                Path to a directory where a downloaded pretrained model configuration is cached if the standard cache
-                is not used.
-            force_download (`bool`, *optional*, defaults to `False`):
-                Whether or not to force the (re-)download of the model weights and configuration files, overriding the
-                cached versions if they exist.
-
-            proxies (`Dict[str, str]`, *optional*):
-                A dictionary of proxy servers to use by protocol or endpoint, for example, `{'http': 'foo.bar:3128',
-                'http://hostname': 'foo.bar:4012'}`. The proxies are used on each request.
-            local_files_only (`bool`, *optional*, defaults to `False`):
-                Whether to only load local model weights and configuration files or not. If set to `True`, the model
-                won't be downloaded from the Hub.
-            token (`str` or *bool*, *optional*):
-                The token to use as HTTP bearer authorization for remote files. If `True`, the token generated from
-                `diffusers-cli login` (stored in `~/.huggingface`) is used.
-            revision (`str`, *optional*, defaults to `"main"`):
-                The specific model version to use. It can be a branch name, a tag name, a commit id, or any identifier
-                allowed by Git.
-            subfolder (`str`, *optional*, defaults to `""`):
-                The subfolder location of a model file within a larger model repository on the Hub or locally.
-            weight_name (`str`, *optional*, defaults to None):
-                Name of the serialized state dict file.
-        """
-        # Load the main state dict first which has the LoRA layers for either of
-        # UNet and text encoder or both.
-        cache_dir = kwargs.pop("cache_dir", None)
-        force_download = kwargs.pop("force_download", False)
-        proxies = kwargs.pop("proxies", None)
-        local_files_only = kwargs.pop("local_files_only", None)
-        token = kwargs.pop("token", None)
-        revision = kwargs.pop("revision", None)
-        subfolder = kwargs.pop("subfolder", None)
-        weight_name = kwargs.pop("weight_name", None)
-        unet_config = kwargs.pop("unet_config", None)
-        use_safetensors = kwargs.pop("use_safetensors", None)
-
-        allow_pickle = False
-        if use_safetensors is None:
-            use_safetensors = True
-            allow_pickle = True
-
-        user_agent = {
-            "file_type": "attn_procs_weights",
-            "framework": "pytorch",
-        }
-
-        state_dict = cls._fetch_state_dict(
-            pretrained_model_name_or_path_or_dict=pretrained_model_name_or_path_or_dict,
-            weight_name=weight_name,
-            use_safetensors=use_safetensors,
-            local_files_only=local_files_only,
-            cache_dir=cache_dir,
-            force_download=force_download,
-            proxies=proxies,
-            token=token,
-            revision=revision,
-            subfolder=subfolder,
-            user_agent=user_agent,
-            allow_pickle=allow_pickle,
-        )
-
-        network_alphas = None
-        # TODO: replace it with a method from `state_dict_utils`
-        if all(
-            (
-                k.startswith("lora_te_")
-                or k.startswith("lora_unet_")
-                or k.startswith("lora_te1_")
-                or k.startswith("lora_te2_")
-            )
-            for k in state_dict.keys()
-        ):
-            # Map SDXL blocks correctly.
-            if unet_config is not None:
-                # use unet config to remap block numbers
-                state_dict = _maybe_map_sgm_blocks_to_diffusers(state_dict, unet_config)
-            state_dict, network_alphas = _convert_non_diffusers_lora_to_diffusers(state_dict)
-
-        return state_dict, network_alphas
-
-    @classmethod
-    # Copied from diffusers.loaders.lora_pipeline.StableDiffusionLoraLoaderMixin.load_lora_into_unet
-    def load_lora_into_unet(cls, state_dict, network_alphas, unet, adapter_name=None, _pipeline=None):
-        """
-        This will load the LoRA layers specified in `state_dict` into `unet`.
-
-        Parameters:
-            state_dict (`dict`):
-                A standard state dict containing the lora layer parameters. The keys can either be indexed directly
-                into the unet or prefixed with an additional `unet` which can be used to distinguish between text
-                encoder lora layers.
-            network_alphas (`Dict[str, float]`):
-                The value of the network alpha used for stable learning and preventing underflow. This value has the
-                same meaning as the `--network_alpha` option in the kohya-ss trainer script. Refer to [this
-                link](https://github.com/darkstorm2150/sd-scripts/blob/main/docs/train_network_README-en.md#execute-learning).
-            unet (`UNet2DConditionModel`):
-                The UNet model to load the LoRA layers into.
-            adapter_name (`str`, *optional*):
-                Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
-                `default_{i}` where i is the total number of adapters being loaded.
-        """
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for this method.")
-
-        # If the serialization format is new (introduced in https://github.com/huggingface/diffusers/pull/2918),
-        # then the `state_dict` keys should have `cls.unet_name` and/or `cls.text_encoder_name` as
-        # their prefixes.
-        keys = list(state_dict.keys())
-        only_text_encoder = all(key.startswith(cls.text_encoder_name) for key in keys)
-        if not only_text_encoder:
-            # Load the layers corresponding to UNet.
-            logger.info(f"Loading {cls.unet_name}.")
-            unet.load_attn_procs(
-                state_dict, network_alphas=network_alphas, adapter_name=adapter_name, _pipeline=_pipeline
-            )
-
-    @classmethod
-    # Copied from diffusers.loaders.lora_pipeline.StableDiffusionLoraLoaderMixin.load_lora_into_text_encoder
-    def load_lora_into_text_encoder(
-        cls,
-        state_dict,
-        network_alphas,
-        text_encoder,
-        prefix=None,
-        lora_scale=1.0,
-        adapter_name=None,
-        _pipeline=None,
-    ):
-        """
-        This will load the LoRA layers specified in `state_dict` into `text_encoder`
-
-        Parameters:
-            state_dict (`dict`):
-                A standard state dict containing the lora layer parameters. The key should be prefixed with an
-                additional `text_encoder` to distinguish between unet lora layers.
-            network_alphas (`Dict[str, float]`):
-                See `LoRALinearLayer` for more details.
-            text_encoder (`CLIPTextModel`):
-                The text encoder model to load the LoRA layers into.
-            prefix (`str`):
-                Expected prefix of the `text_encoder` in the `state_dict`.
-            lora_scale (`float`):
-                How much to scale the output of the lora linear layer before it is added with the output of the regular
-                lora layer.
-            adapter_name (`str`, *optional*):
-                Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
-                `default_{i}` where i is the total number of adapters being loaded.
-        """
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for this method.")
-
-        from peft import LoraConfig
-
-        # If the serialization format is new (introduced in https://github.com/huggingface/diffusers/pull/2918),
-        # then the `state_dict` keys should have `self.unet_name` and/or `self.text_encoder_name` as
-        # their prefixes.
-        keys = list(state_dict.keys())
-        prefix = cls.text_encoder_name if prefix is None else prefix
-
-        # Safe prefix to check with.
-        if any(cls.text_encoder_name in key for key in keys):
-            # Load the layers corresponding to text encoder and make necessary adjustments.
-            text_encoder_keys = [k for k in keys if k.startswith(prefix) and k.split(".")[0] == prefix]
-            text_encoder_lora_state_dict = {
-                k.replace(f"{prefix}.", ""): v for k, v in state_dict.items() if k in text_encoder_keys
-            }
-
-            if len(text_encoder_lora_state_dict) > 0:
-                logger.info(f"Loading {prefix}.")
-                rank = {}
-                text_encoder_lora_state_dict = convert_state_dict_to_diffusers(text_encoder_lora_state_dict)
-
-                # convert state dict
-                text_encoder_lora_state_dict = convert_state_dict_to_peft(text_encoder_lora_state_dict)
-
-                for name, _ in text_encoder_attn_modules(text_encoder):
-                    for module in ("out_proj", "q_proj", "k_proj", "v_proj"):
-                        rank_key = f"{name}.{module}.lora_B.weight"
-                        if rank_key not in text_encoder_lora_state_dict:
-                            continue
-                        rank[rank_key] = text_encoder_lora_state_dict[rank_key].shape[1]
-
-                for name, _ in text_encoder_mlp_modules(text_encoder):
-                    for module in ("fc1", "fc2"):
-                        rank_key = f"{name}.{module}.lora_B.weight"
-                        if rank_key not in text_encoder_lora_state_dict:
-                            continue
-                        rank[rank_key] = text_encoder_lora_state_dict[rank_key].shape[1]
-
-                if network_alphas is not None:
-                    alpha_keys = [
-                        k for k in network_alphas.keys() if k.startswith(prefix) and k.split(".")[0] == prefix
-                    ]
-                    network_alphas = {
-                        k.replace(f"{prefix}.", ""): v for k, v in network_alphas.items() if k in alpha_keys
-                    }
-
-                lora_config_kwargs = get_peft_kwargs(rank, network_alphas, text_encoder_lora_state_dict, is_unet=False)
-                if "use_dora" in lora_config_kwargs:
-                    if lora_config_kwargs["use_dora"]:
-                        if is_peft_version("<", "0.9.0"):
-                            raise ValueError(
-                                "You need `peft` 0.9.0 at least to use DoRA-enabled LoRAs. Please upgrade your installation of `peft`."
-                            )
-                    else:
-                        if is_peft_version("<", "0.9.0"):
-                            lora_config_kwargs.pop("use_dora")
-                lora_config = LoraConfig(**lora_config_kwargs)
-
-                # adapter_name
-                if adapter_name is None:
-                    adapter_name = get_adapter_name(text_encoder)
-
-                is_model_cpu_offload, is_sequential_cpu_offload = cls._optionally_disable_offloading(_pipeline)
-
-                # inject LoRA layers and load the state dict
-                # in transformers we automatically check whether the adapter name is already in use or not
-                text_encoder.load_adapter(
-                    adapter_name=adapter_name,
-                    adapter_state_dict=text_encoder_lora_state_dict,
-                    peft_config=lora_config,
-                )
-
-                # scale LoRA layers with `lora_scale`
-                scale_lora_layers(text_encoder, weight=lora_scale)
-
-                text_encoder.to(device=text_encoder.device, dtype=text_encoder.dtype)
-
-                # Offload back.
-                if is_model_cpu_offload:
-                    _pipeline.enable_model_cpu_offload()
-                elif is_sequential_cpu_offload:
-                    _pipeline.enable_sequential_cpu_offload()
-                # Unsafe code />
-
-    @classmethod
-    def save_lora_weights(
-        cls,
-        save_directory: Union[str, os.PathLike],
-        unet_lora_layers: Dict[str, Union[torch.nn.Module, torch.Tensor]] = None,
-        text_encoder_lora_layers: Dict[str, Union[torch.nn.Module, torch.Tensor]] = None,
-        text_encoder_2_lora_layers: Dict[str, Union[torch.nn.Module, torch.Tensor]] = None,
-        is_main_process: bool = True,
-        weight_name: str = None,
-        save_function: Callable = None,
-        safe_serialization: bool = True,
-    ):
-        r"""
-        Save the LoRA parameters corresponding to the UNet and text encoder.
-
-        Arguments:
-            save_directory (`str` or `os.PathLike`):
-                Directory to save LoRA parameters to. Will be created if it doesn't exist.
-            unet_lora_layers (`Dict[str, torch.nn.Module]` or `Dict[str, torch.Tensor]`):
-                State dict of the LoRA layers corresponding to the `unet`.
-            text_encoder_lora_layers (`Dict[str, torch.nn.Module]` or `Dict[str, torch.Tensor]`):
-                State dict of the LoRA layers corresponding to the `text_encoder`. Must explicitly pass the text
-                encoder LoRA state dict because it comes from 🤗 Transformers.
-            text_encoder_2_lora_layers (`Dict[str, torch.nn.Module]` or `Dict[str, torch.Tensor]`):
-                State dict of the LoRA layers corresponding to the `text_encoder_2`. Must explicitly pass the text
-                encoder LoRA state dict because it comes from 🤗 Transformers.
-            is_main_process (`bool`, *optional*, defaults to `True`):
-                Whether the process calling this is the main process or not. Useful during distributed training and you
-                need to call this function on all processes. In this case, set `is_main_process=True` only on the main
-                process to avoid race conditions.
-            save_function (`Callable`):
-                The function to use to save the state dictionary. Useful during distributed training when you need to
-                replace `torch.save` with another method. Can be configured with the environment variable
-                `DIFFUSERS_SAVE_MODE`.
-            safe_serialization (`bool`, *optional*, defaults to `True`):
-                Whether to save the model using `safetensors` or the traditional PyTorch way with `pickle`.
-        """
-        state_dict = {}
-
-        if not (unet_lora_layers or text_encoder_lora_layers or text_encoder_2_lora_layers):
-            raise ValueError(
-                "You must pass at least one of `unet_lora_layers`, `text_encoder_lora_layers` or `text_encoder_2_lora_layers`."
-            )
-
-        if unet_lora_layers:
-            state_dict.update(cls.pack_weights(unet_lora_layers, "unet"))
-
-        if text_encoder_lora_layers:
-            state_dict.update(cls.pack_weights(text_encoder_lora_layers, "text_encoder"))
-
-        if text_encoder_2_lora_layers:
-            state_dict.update(cls.pack_weights(text_encoder_2_lora_layers, "text_encoder_2"))
-
-        cls.write_lora_layers(
-            state_dict=state_dict,
-            save_directory=save_directory,
-            is_main_process=is_main_process,
-            weight_name=weight_name,
-            save_function=save_function,
-            safe_serialization=safe_serialization,
-        )
-
-    def fuse_lora(
-        self,
-        components: List[str] = ["unet", "text_encoder", "text_encoder_2"],
-        lora_scale: float = 1.0,
-        safe_fusing: bool = False,
-        adapter_names: Optional[List[str]] = None,
-        **kwargs,
-    ):
-        r"""
-        Fuses the LoRA parameters into the original parameters of the corresponding blocks.
-
-        <Tip warning={true}>
-
-        This is an experimental API.
-
-        </Tip>
-
-        Args:
-            components: (`List[str]`): List of LoRA-injectable components to fuse the LoRAs into.
-            lora_scale (`float`, defaults to 1.0):
-                Controls how much to influence the outputs with the LoRA parameters.
-            safe_fusing (`bool`, defaults to `False`):
-                Whether to check fused weights for NaN values before fusing and if values are NaN not fusing them.
-            adapter_names (`List[str]`, *optional*):
-                Adapter names to be used for fusing. If nothing is passed, all active adapters will be fused.
-
-        Example:
-
-        ```py
-        from diffusers import DiffusionPipeline
-        import torch
-
-        pipeline = DiffusionPipeline.from_pretrained(
-            "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
-        ).to("cuda")
-        pipeline.load_lora_weights("nerijs/pixel-art-xl", weight_name="pixel-art-xl.safetensors", adapter_name="pixel")
-        pipeline.fuse_lora(lora_scale=0.7)
-        ```
-        """
-        super().fuse_lora(
-            components=components, lora_scale=lora_scale, safe_fusing=safe_fusing, adapter_names=adapter_names
-        )
-
-    def unfuse_lora(self, components: List[str] = ["unet", "text_encoder", "text_encoder_2"], **kwargs):
-        r"""
-        Reverses the effect of
-        [`pipe.fuse_lora()`](https://huggingface.co/docs/diffusers/main/en/api/loaders#diffusers.loaders.LoraBaseMixin.fuse_lora).
-
-        <Tip warning={true}>
-
-        This is an experimental API.
-
-        </Tip>
-
-        Args:
-            components (`List[str]`): List of LoRA-injectable components to unfuse LoRA from.
-            unfuse_unet (`bool`, defaults to `True`): Whether to unfuse the UNet LoRA parameters.
-            unfuse_text_encoder (`bool`, defaults to `True`):
-                Whether to unfuse the text encoder LoRA parameters. If the text encoder wasn't monkey-patched with the
-                LoRA parameters then it won't have any effect.
-        """
-        super().unfuse_lora(components=components)
-
-
-class SD3LoraLoaderMixin(LoraBaseMixin):
-    r"""
-    Load LoRA layers into [`SD3Transformer2DModel`],
-    [`CLIPTextModel`](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModel), and
-    [`CLIPTextModelWithProjection`](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModelWithProjection).
-
-    Specific to [`StableDiffusion3Pipeline`].
-    """
-
-    _lora_loadable_modules = ["transformer", "text_encoder", "text_encoder_2"]
-    transformer_name = TRANSFORMER_NAME
-    text_encoder_name = TEXT_ENCODER_NAME
-
-    @classmethod
-    @validate_hf_hub_args
-    def lora_state_dict(
-        cls,
-        pretrained_model_name_or_path_or_dict: Union[str, Dict[str, torch.Tensor]],
-        **kwargs,
-    ):
-        r"""
-        Return state dict for lora weights and the network alphas.
-
-        <Tip warning={true}>
-
-        We support loading A1111 formatted LoRA checkpoints in a limited capacity.
-
-        This function is experimental and might change in the future.
+                    if part_weight is not None and part_name not in invert_list_adapters[adapter_name]:
+                        logger.warning(
+                            f"Lora weight dict for adapter '{adapter_name}' contains {part_name}, but this will be ignored because {adapter_name} does not contain weights for {part_name}. Valid parts for {adapter_name} are: {invert_list_adapters[adapter_name]}."
+                        )
 
-        </Tip>
+            else:
+                unet_lora_weight = weights
+                text_encoder_lora_weight = weights
+                text_encoder_2_lora_weight = weights
 
-        Parameters:
-            pretrained_model_name_or_path_or_dict (`str` or `os.PathLike` or `dict`):
-                Can be either:
+            unet_lora_weights.append(unet_lora_weight)
+            text_encoder_lora_weights.append(text_encoder_lora_weight)
+            text_encoder_2_lora_weights.append(text_encoder_2_lora_weight)
 
-                    - A string, the *model id* (for example `google/ddpm-celebahq-256`) of a pretrained model hosted on
-                      the Hub.
-                    - A path to a *directory* (for example `./my_model_directory`) containing the model weights saved
-                      with [`ModelMixin.save_pretrained`].
-                    - A [torch state
-                      dict](https://pytorch.org/tutorials/beginner/saving_loading_models.html#what-is-a-state-dict).
+        unet = getattr(self, self.unet_name) if not hasattr(self, "unet") else self.unet
+        # Handle the UNET
+        unet.set_adapters(adapter_names, unet_lora_weights)
 
-            cache_dir (`Union[str, os.PathLike]`, *optional*):
-                Path to a directory where a downloaded pretrained model configuration is cached if the standard cache
-                is not used.
-            force_download (`bool`, *optional*, defaults to `False`):
-                Whether or not to force the (re-)download of the model weights and configuration files, overriding the
-                cached versions if they exist.
+        # Handle the Text Encoder
+        if hasattr(self, "text_encoder"):
+            self.set_adapters_for_text_encoder(adapter_names, self.text_encoder, text_encoder_lora_weights)
+        if hasattr(self, "text_encoder_2"):
+            self.set_adapters_for_text_encoder(adapter_names, self.text_encoder_2, text_encoder_2_lora_weights)
 
-            proxies (`Dict[str, str]`, *optional*):
-                A dictionary of proxy servers to use by protocol or endpoint, for example, `{'http': 'foo.bar:3128',
-                'http://hostname': 'foo.bar:4012'}`. The proxies are used on each request.
-            local_files_only (`bool`, *optional*, defaults to `False`):
-                Whether to only load local model weights and configuration files or not. If set to `True`, the model
-                won't be downloaded from the Hub.
-            token (`str` or *bool*, *optional*):
-                The token to use as HTTP bearer authorization for remote files. If `True`, the token generated from
-                `diffusers-cli login` (stored in `~/.huggingface`) is used.
-            revision (`str`, *optional*, defaults to `"main"`):
-                The specific model version to use. It can be a branch name, a tag name, a commit id, or any identifier
-                allowed by Git.
-            subfolder (`str`, *optional*, defaults to `""`):
-                The subfolder location of a model file within a larger model repository on the Hub or locally.
+    def disable_lora(self):
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for this method.")
 
-        """
-        # Load the main state dict first which has the LoRA layers for either of
-        # transformer and text encoder or both.
-        cache_dir = kwargs.pop("cache_dir", None)
-        force_download = kwargs.pop("force_download", False)
-        proxies = kwargs.pop("proxies", None)
-        local_files_only = kwargs.pop("local_files_only", None)
-        token = kwargs.pop("token", None)
-        revision = kwargs.pop("revision", None)
-        subfolder = kwargs.pop("subfolder", None)
-        weight_name = kwargs.pop("weight_name", None)
-        use_safetensors = kwargs.pop("use_safetensors", None)
+        # Disable unet adapters
+        unet = getattr(self, self.unet_name) if not hasattr(self, "unet") else self.unet
+        unet.disable_lora()
 
-        allow_pickle = False
-        if use_safetensors is None:
-            use_safetensors = True
-            allow_pickle = True
+        # Disable text encoder adapters
+        if hasattr(self, "text_encoder"):
+            self.disable_lora_for_text_encoder(self.text_encoder)
+        if hasattr(self, "text_encoder_2"):
+            self.disable_lora_for_text_encoder(self.text_encoder_2)
 
-        user_agent = {
-            "file_type": "attn_procs_weights",
-            "framework": "pytorch",
-        }
+    def enable_lora(self):
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for this method.")
 
-        state_dict = cls._fetch_state_dict(
-            pretrained_model_name_or_path_or_dict=pretrained_model_name_or_path_or_dict,
-            weight_name=weight_name,
-            use_safetensors=use_safetensors,
-            local_files_only=local_files_only,
-            cache_dir=cache_dir,
-            force_download=force_download,
-            proxies=proxies,
-            token=token,
-            revision=revision,
-            subfolder=subfolder,
-            user_agent=user_agent,
-            allow_pickle=allow_pickle,
-        )
+        # Enable unet adapters
+        unet = getattr(self, self.unet_name) if not hasattr(self, "unet") else self.unet
+        unet.enable_lora()
 
-        return state_dict
+        # Enable text encoder adapters
+        if hasattr(self, "text_encoder"):
+            self.enable_lora_for_text_encoder(self.text_encoder)
+        if hasattr(self, "text_encoder_2"):
+            self.enable_lora_for_text_encoder(self.text_encoder_2)
 
-    def load_lora_weights(
-        self, pretrained_model_name_or_path_or_dict: Union[str, Dict[str, torch.Tensor]], adapter_name=None, **kwargs
-    ):
+    def delete_adapters(self, adapter_names: Union[List[str], str]):
         """
-        Load LoRA weights specified in `pretrained_model_name_or_path_or_dict` into `self.unet` and
-        `self.text_encoder`.
-
-        All kwargs are forwarded to `self.lora_state_dict`.
-
-        See [`~loaders.StableDiffusionLoraLoaderMixin.lora_state_dict`] for more details on how the state dict is
-        loaded.
-
-        See [`~loaders.StableDiffusionLoraLoaderMixin.load_lora_into_transformer`] for more details on how the state
-        dict is loaded into `self.transformer`.
-
-        Parameters:
-            pretrained_model_name_or_path_or_dict (`str` or `os.PathLike` or `dict`):
-                See [`~loaders.StableDiffusionLoraLoaderMixin.lora_state_dict`].
-            kwargs (`dict`, *optional*):
-                See [`~loaders.StableDiffusionLoraLoaderMixin.lora_state_dict`].
-            adapter_name (`str`, *optional*):
-                Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
-                `default_{i}` where i is the total number of adapters being loaded.
+        Args:
+        Deletes the LoRA layers of `adapter_name` for the unet and text-encoder(s).
+            adapter_names (`Union[List[str], str]`):
+                The names of the adapter to delete. Can be a single string or a list of strings
         """
         if not USE_PEFT_BACKEND:
             raise ValueError("PEFT backend is required for this method.")
 
-        # if a dict is passed, copy it instead of modifying it inplace
-        if isinstance(pretrained_model_name_or_path_or_dict, dict):
-            pretrained_model_name_or_path_or_dict = pretrained_model_name_or_path_or_dict.copy()
-
-        # First, ensure that the checkpoint is a compatible one and can be successfully loaded.
-        state_dict = self.lora_state_dict(pretrained_model_name_or_path_or_dict, **kwargs)
-
-        is_correct_format = all("lora" in key or "dora_scale" in key for key in state_dict.keys())
-        if not is_correct_format:
-            raise ValueError("Invalid LoRA checkpoint.")
-
-        self.load_lora_into_transformer(
-            state_dict,
-            transformer=getattr(self, self.transformer_name) if not hasattr(self, "transformer") else self.transformer,
-            adapter_name=adapter_name,
-            _pipeline=self,
-        )
-
-        text_encoder_state_dict = {k: v for k, v in state_dict.items() if "text_encoder." in k}
-        if len(text_encoder_state_dict) > 0:
-            self.load_lora_into_text_encoder(
-                text_encoder_state_dict,
-                network_alphas=None,
-                text_encoder=self.text_encoder,
-                prefix="text_encoder",
-                lora_scale=self.lora_scale,
-                adapter_name=adapter_name,
-                _pipeline=self,
-            )
+        if isinstance(adapter_names, str):
+            adapter_names = [adapter_names]
 
-        text_encoder_2_state_dict = {k: v for k, v in state_dict.items() if "text_encoder_2." in k}
-        if len(text_encoder_2_state_dict) > 0:
-            self.load_lora_into_text_encoder(
-                text_encoder_2_state_dict,
-                network_alphas=None,
-                text_encoder=self.text_encoder_2,
-                prefix="text_encoder_2",
-                lora_scale=self.lora_scale,
-                adapter_name=adapter_name,
-                _pipeline=self,
-            )
+        # Delete unet adapters
+        unet = getattr(self, self.unet_name) if not hasattr(self, "unet") else self.unet
+        unet.delete_adapters(adapter_names)
 
-    @classmethod
-    def load_lora_into_transformer(cls, state_dict, transformer, adapter_name=None, _pipeline=None):
-        """
-        This will load the LoRA layers specified in `state_dict` into `transformer`.
+        for adapter_name in adapter_names:
+            # Delete text encoder adapters
+            if hasattr(self, "text_encoder"):
+                delete_adapter_layers(self.text_encoder, adapter_name)
+            if hasattr(self, "text_encoder_2"):
+                delete_adapter_layers(self.text_encoder_2, adapter_name)
 
-        Parameters:
-            state_dict (`dict`):
-                A standard state dict containing the lora layer parameters. The keys can either be indexed directly
-                into the unet or prefixed with an additional `unet` which can be used to distinguish between text
-                encoder lora layers.
-            transformer (`SD3Transformer2DModel`):
-                The Transformer model to load the LoRA layers into.
-            adapter_name (`str`, *optional*):
-                Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
-                `default_{i}` where i is the total number of adapters being loaded.
+    def get_active_adapters(self) -> List[str]:
         """
-        from peft import LoraConfig, inject_adapter_in_model, set_peft_model_state_dict
+        Gets the list of the current active adapters.
 
-        keys = list(state_dict.keys())
-
-        transformer_keys = [k for k in keys if k.startswith(cls.transformer_name)]
-        state_dict = {
-            k.replace(f"{cls.transformer_name}.", ""): v for k, v in state_dict.items() if k in transformer_keys
-        }
+        Example:
 
-        if len(state_dict.keys()) > 0:
-            # check with first key if is not in peft format
-            first_key = next(iter(state_dict.keys()))
-            if "lora_A" not in first_key:
-                state_dict = convert_unet_state_dict_to_peft(state_dict)
+        ```python
+        from diffusers import DiffusionPipeline
 
-            if adapter_name in getattr(transformer, "peft_config", {}):
+        pipeline = DiffusionPipeline.from_pretrained(
+            "stabilityai/stable-diffusion-xl-base-1.0",
+        ).to("cuda")
+        pipeline.load_lora_weights("CiroN2022/toy-face", weight_name="toy_face_sdxl.safetensors", adapter_name="toy")
+        pipeline.get_active_adapters()
+        ```
+        """
+        if not USE_PEFT_BACKEND:
             raise ValueError(
-                    f"Adapter name {adapter_name} already in use in the transformer - please select a new adapter name."
+                "PEFT backend is required for this method. Please install the latest version of PEFT `pip install -U peft`"
             )
 
-            rank = {}
-            for key, val in state_dict.items():
-                if "lora_B" in key:
-                    rank[key] = val.shape[1]
+        from peft.tuners.tuners_utils import BaseTunerLayer
 
-            lora_config_kwargs = get_peft_kwargs(rank, network_alpha_dict=None, peft_state_dict=state_dict)
-            if "use_dora" in lora_config_kwargs:
-                if lora_config_kwargs["use_dora"] and is_peft_version("<", "0.9.0"):
+        active_adapters = []
+        unet = getattr(self, self.unet_name) if not hasattr(self, "unet") else self.unet
+        for module in unet.modules():
+            if isinstance(module, BaseTunerLayer):
+                active_adapters = module.active_adapters
+                break
+
+        return active_adapters
+
+    def get_list_adapters(self) -> Dict[str, List[str]]:
+        """
+        Gets the current list of all available adapters in the pipeline.
+        """
+        if not USE_PEFT_BACKEND:
             raise ValueError(
-                        "You need `peft` 0.9.0 at least to use DoRA-enabled LoRAs. Please upgrade your installation of `peft`."
+                "PEFT backend is required for this method. Please install the latest version of PEFT `pip install -U peft`"
             )
-                else:
-                    lora_config_kwargs.pop("use_dora")
-            lora_config = LoraConfig(**lora_config_kwargs)
 
-            # adapter_name
-            if adapter_name is None:
-                adapter_name = get_adapter_name(transformer)
+        set_adapters = {}
 
-            # In case the pipeline has been already offloaded to CPU - temporarily remove the hooks
-            # otherwise loading LoRA weights will lead to an error
-            is_model_cpu_offload, is_sequential_cpu_offload = cls._optionally_disable_offloading(_pipeline)
+        if hasattr(self, "text_encoder") and hasattr(self.text_encoder, "peft_config"):
+            set_adapters["text_encoder"] = list(self.text_encoder.peft_config.keys())
 
-            inject_adapter_in_model(lora_config, transformer, adapter_name=adapter_name)
-            incompatible_keys = set_peft_model_state_dict(transformer, state_dict, adapter_name)
+        if hasattr(self, "text_encoder_2") and hasattr(self.text_encoder_2, "peft_config"):
+            set_adapters["text_encoder_2"] = list(self.text_encoder_2.peft_config.keys())
 
-            if incompatible_keys is not None:
-                # check only for unexpected keys
-                unexpected_keys = getattr(incompatible_keys, "unexpected_keys", None)
-                if unexpected_keys:
-                    logger.warning(
-                        f"Loading adapter weights from state_dict led to unexpected keys not found in the model: "
-                        f" {unexpected_keys}. "
-                    )
+        unet = getattr(self, self.unet_name) if not hasattr(self, "unet") else self.unet
+        if hasattr(self, self.unet_name) and hasattr(unet, "peft_config"):
+            set_adapters[self.unet_name] = list(self.unet.peft_config.keys())
 
-            # Offload back.
-            if is_model_cpu_offload:
-                _pipeline.enable_model_cpu_offload()
-            elif is_sequential_cpu_offload:
-                _pipeline.enable_sequential_cpu_offload()
-            # Unsafe code />
+        return set_adapters
 
-    @classmethod
-    # Copied from diffusers.loaders.lora_pipeline.StableDiffusionLoraLoaderMixin.load_lora_into_text_encoder
-    def load_lora_into_text_encoder(
-        cls,
-        state_dict,
-        network_alphas,
-        text_encoder,
-        prefix=None,
-        lora_scale=1.0,
-        adapter_name=None,
-        _pipeline=None,
-    ):
+    def set_lora_device(self, adapter_names: List[str], device: Union[torch.device, str, int]) -> None:
         """
-        This will load the LoRA layers specified in `state_dict` into `text_encoder`
+        Moves the LoRAs listed in `adapter_names` to a target device. Useful for offloading the LoRA to the CPU in case
+        you want to load multiple adapters and free some GPU memory.
 
-        Parameters:
-            state_dict (`dict`):
-                A standard state dict containing the lora layer parameters. The key should be prefixed with an
-                additional `text_encoder` to distinguish between unet lora layers.
-            network_alphas (`Dict[str, float]`):
-                See `LoRALinearLayer` for more details.
-            text_encoder (`CLIPTextModel`):
-                The text encoder model to load the LoRA layers into.
-            prefix (`str`):
-                Expected prefix of the `text_encoder` in the `state_dict`.
-            lora_scale (`float`):
-                How much to scale the output of the lora linear layer before it is added with the output of the regular
-                lora layer.
-            adapter_name (`str`, *optional*):
-                Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
-                `default_{i}` where i is the total number of adapters being loaded.
+        Args:
+            adapter_names (`List[str]`):
+                List of adapters to send device to.
+            device (`Union[torch.device, str, int]`):
+                Device to send the adapters to. Can be either a torch device, a str or an integer.
         """
         if not USE_PEFT_BACKEND:
             raise ValueError("PEFT backend is required for this method.")
 
-        from peft import LoraConfig
+        from peft.tuners.tuners_utils import BaseTunerLayer
+
+        # Handle the UNET
+        unet = getattr(self, self.unet_name) if not hasattr(self, "unet") else self.unet
+        for unet_module in unet.modules():
+            if isinstance(unet_module, BaseTunerLayer):
+                for adapter_name in adapter_names:
+                    unet_module.lora_A[adapter_name].to(device)
+                    unet_module.lora_B[adapter_name].to(device)
+                    # this is a param, not a module, so device placement is not in-place -> re-assign
+                    if hasattr(unet_module, "lora_magnitude_vector") and unet_module.lora_magnitude_vector is not None:
+                        unet_module.lora_magnitude_vector[adapter_name] = unet_module.lora_magnitude_vector[
+                            adapter_name
+                        ].to(device)
+
+        # Handle the text encoder
+        modules_to_process = []
+        if hasattr(self, "text_encoder"):
+            modules_to_process.append(self.text_encoder)
+
+        if hasattr(self, "text_encoder_2"):
+            modules_to_process.append(self.text_encoder_2)
+
+        for text_encoder in modules_to_process:
+            # loop over submodules
+            for text_encoder_module in text_encoder.modules():
+                if isinstance(text_encoder_module, BaseTunerLayer):
+                    for adapter_name in adapter_names:
+                        text_encoder_module.lora_A[adapter_name].to(device)
+                        text_encoder_module.lora_B[adapter_name].to(device)
+                        # this is a param, not a module, so device placement is not in-place -> re-assign
+                        if (
+                            hasattr(text_encoder_module, "lora_magnitude_vector")
+                            and text_encoder_module.lora_magnitude_vector is not None
+                        ):
+                            text_encoder_module.lora_magnitude_vector[
+                                adapter_name
+                            ] = text_encoder_module.lora_magnitude_vector[adapter_name].to(device)
 
-        # If the serialization format is new (introduced in https://github.com/huggingface/diffusers/pull/2918),
-        # then the `state_dict` keys should have `self.unet_name` and/or `self.text_encoder_name` as
-        # their prefixes.
-        keys = list(state_dict.keys())
-        prefix = cls.text_encoder_name if prefix is None else prefix
 
-        # Safe prefix to check with.
-        if any(cls.text_encoder_name in key for key in keys):
-            # Load the layers corresponding to text encoder and make necessary adjustments.
-            text_encoder_keys = [k for k in keys if k.startswith(prefix) and k.split(".")[0] == prefix]
-            text_encoder_lora_state_dict = {
-                k.replace(f"{prefix}.", ""): v for k, v in state_dict.items() if k in text_encoder_keys
-            }
+class StableDiffusionXLLoraLoaderMixin(LoraLoaderMixin):
+    """This class overrides `LoraLoaderMixin` with LoRA loading/saving code that's specific to SDXL"""
 
-            if len(text_encoder_lora_state_dict) > 0:
-                logger.info(f"Loading {prefix}.")
-                rank = {}
-                text_encoder_lora_state_dict = convert_state_dict_to_diffusers(text_encoder_lora_state_dict)
+    # Override to properly handle the loading and unloading of the additional text encoder.
+    def load_lora_weights(
+        self,
+        pretrained_model_name_or_path_or_dict: Union[str, Dict[str, torch.Tensor]],
+        adapter_name: Optional[str] = None,
+        **kwargs,
+    ):
+        """
+        Load LoRA weights specified in `pretrained_model_name_or_path_or_dict` into `self.unet` and
+        `self.text_encoder`.
+
+        All kwargs are forwarded to `self.lora_state_dict`.
 
-                # convert state dict
-                text_encoder_lora_state_dict = convert_state_dict_to_peft(text_encoder_lora_state_dict)
+        See [`~loaders.LoraLoaderMixin.lora_state_dict`] for more details on how the state dict is loaded.
 
-                for name, _ in text_encoder_attn_modules(text_encoder):
-                    for module in ("out_proj", "q_proj", "k_proj", "v_proj"):
-                        rank_key = f"{name}.{module}.lora_B.weight"
-                        if rank_key not in text_encoder_lora_state_dict:
-                            continue
-                        rank[rank_key] = text_encoder_lora_state_dict[rank_key].shape[1]
+        See [`~loaders.LoraLoaderMixin.load_lora_into_unet`] for more details on how the state dict is loaded into
+        `self.unet`.
 
-                for name, _ in text_encoder_mlp_modules(text_encoder):
-                    for module in ("fc1", "fc2"):
-                        rank_key = f"{name}.{module}.lora_B.weight"
-                        if rank_key not in text_encoder_lora_state_dict:
-                            continue
-                        rank[rank_key] = text_encoder_lora_state_dict[rank_key].shape[1]
+        See [`~loaders.LoraLoaderMixin.load_lora_into_text_encoder`] for more details on how the state dict is loaded
+        into `self.text_encoder`.
 
-                if network_alphas is not None:
-                    alpha_keys = [
-                        k for k in network_alphas.keys() if k.startswith(prefix) and k.split(".")[0] == prefix
-                    ]
-                    network_alphas = {
-                        k.replace(f"{prefix}.", ""): v for k, v in network_alphas.items() if k in alpha_keys
-                    }
+        Parameters:
+            pretrained_model_name_or_path_or_dict (`str` or `os.PathLike` or `dict`):
+                See [`~loaders.LoraLoaderMixin.lora_state_dict`].
+            adapter_name (`str`, *optional*):
+                Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
+                `default_{i}` where i is the total number of adapters being loaded.
+            kwargs (`dict`, *optional*):
+                See [`~loaders.LoraLoaderMixin.lora_state_dict`].
+        """
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for this method.")
 
-                lora_config_kwargs = get_peft_kwargs(rank, network_alphas, text_encoder_lora_state_dict, is_unet=False)
-                if "use_dora" in lora_config_kwargs:
-                    if lora_config_kwargs["use_dora"]:
-                        if is_peft_version("<", "0.9.0"):
-                            raise ValueError(
-                                "You need `peft` 0.9.0 at least to use DoRA-enabled LoRAs. Please upgrade your installation of `peft`."
-                            )
-                    else:
-                        if is_peft_version("<", "0.9.0"):
-                            lora_config_kwargs.pop("use_dora")
-                lora_config = LoraConfig(**lora_config_kwargs)
+        # We could have accessed the unet config from `lora_state_dict()` too. We pass
+        # it here explicitly to be able to tell that it's coming from an SDXL
+        # pipeline.
 
-                # adapter_name
-                if adapter_name is None:
-                    adapter_name = get_adapter_name(text_encoder)
+        # if a dict is passed, copy it instead of modifying it inplace
+        if isinstance(pretrained_model_name_or_path_or_dict, dict):
+            pretrained_model_name_or_path_or_dict = pretrained_model_name_or_path_or_dict.copy()
 
-                is_model_cpu_offload, is_sequential_cpu_offload = cls._optionally_disable_offloading(_pipeline)
+        # First, ensure that the checkpoint is a compatible one and can be successfully loaded.
+        state_dict, network_alphas = self.lora_state_dict(
+            pretrained_model_name_or_path_or_dict,
+            unet_config=self.unet.config,
+            **kwargs,
+        )
+        is_correct_format = all("lora" in key or "dora_scale" in key for key in state_dict.keys())
+        if not is_correct_format:
+            raise ValueError("Invalid LoRA checkpoint.")
 
-                # inject LoRA layers and load the state dict
-                # in transformers we automatically check whether the adapter name is already in use or not
-                text_encoder.load_adapter(
+        self.load_lora_into_unet(
+            state_dict, network_alphas=network_alphas, unet=self.unet, adapter_name=adapter_name, _pipeline=self
+        )
+        text_encoder_state_dict = {k: v for k, v in state_dict.items() if "text_encoder." in k}
+        if len(text_encoder_state_dict) > 0:
+            self.load_lora_into_text_encoder(
+                text_encoder_state_dict,
+                network_alphas=network_alphas,
+                text_encoder=self.text_encoder,
+                prefix="text_encoder",
+                lora_scale=self.lora_scale,
                 adapter_name=adapter_name,
-                    adapter_state_dict=text_encoder_lora_state_dict,
-                    peft_config=lora_config,
+                _pipeline=self,
             )
 
-                # scale LoRA layers with `lora_scale`
-                scale_lora_layers(text_encoder, weight=lora_scale)
-
-                text_encoder.to(device=text_encoder.device, dtype=text_encoder.dtype)
-
-                # Offload back.
-                if is_model_cpu_offload:
-                    _pipeline.enable_model_cpu_offload()
-                elif is_sequential_cpu_offload:
-                    _pipeline.enable_sequential_cpu_offload()
-                # Unsafe code />
+        text_encoder_2_state_dict = {k: v for k, v in state_dict.items() if "text_encoder_2." in k}
+        if len(text_encoder_2_state_dict) > 0:
+            self.load_lora_into_text_encoder(
+                text_encoder_2_state_dict,
+                network_alphas=network_alphas,
+                text_encoder=self.text_encoder_2,
+                prefix="text_encoder_2",
+                lora_scale=self.lora_scale,
+                adapter_name=adapter_name,
+                _pipeline=self,
+            )
 
     @classmethod
     def save_lora_weights(
         cls,
         save_directory: Union[str, os.PathLike],
-        transformer_lora_layers: Dict[str, torch.nn.Module] = None,
+        unet_lora_layers: Dict[str, Union[torch.nn.Module, torch.Tensor]] = None,
         text_encoder_lora_layers: Dict[str, Union[torch.nn.Module, torch.Tensor]] = None,
         text_encoder_2_lora_layers: Dict[str, Union[torch.nn.Module, torch.Tensor]] = None,
         is_main_process: bool = True,
@@ -1366,8 +1273,8 @@ class SD3LoraLoaderMixin(LoraBaseMixin):
         Arguments:
             save_directory (`str` or `os.PathLike`):
                 Directory to save LoRA parameters to. Will be created if it doesn't exist.
-            transformer_lora_layers (`Dict[str, torch.nn.Module]` or `Dict[str, torch.Tensor]`):
-                State dict of the LoRA layers corresponding to the `transformer`.
+            unet_lora_layers (`Dict[str, torch.nn.Module]` or `Dict[str, torch.Tensor]`):
+                State dict of the LoRA layers corresponding to the `unet`.
             text_encoder_lora_layers (`Dict[str, torch.nn.Module]` or `Dict[str, torch.Tensor]`):
                 State dict of the LoRA layers corresponding to the `text_encoder`. Must explicitly pass the text
                 encoder LoRA state dict because it comes from 🤗 Transformers.
@@ -1387,21 +1294,25 @@ class SD3LoraLoaderMixin(LoraBaseMixin):
         """
         state_dict = {}
 
-        if not (transformer_lora_layers or text_encoder_lora_layers or text_encoder_2_lora_layers):
+        def pack_weights(layers, prefix):
+            layers_weights = layers.state_dict() if isinstance(layers, torch.nn.Module) else layers
+            layers_state_dict = {f"{prefix}.{module_name}": param for module_name, param in layers_weights.items()}
+            return layers_state_dict
+
+        if not (unet_lora_layers or text_encoder_lora_layers or text_encoder_2_lora_layers):
             raise ValueError(
-                "You must pass at least one of `transformer_lora_layers`, `text_encoder_lora_layers`, `text_encoder_2_lora_layers`."
+                "You must pass at least one of `unet_lora_layers`, `text_encoder_lora_layers` or `text_encoder_2_lora_layers`."
             )
 
-        if transformer_lora_layers:
-            state_dict.update(cls.pack_weights(transformer_lora_layers, cls.transformer_name))
+        if unet_lora_layers:
+            state_dict.update(pack_weights(unet_lora_layers, "unet"))
 
         if text_encoder_lora_layers:
-            state_dict.update(cls.pack_weights(text_encoder_lora_layers, "text_encoder"))
+            state_dict.update(pack_weights(text_encoder_lora_layers, "text_encoder"))
 
         if text_encoder_2_lora_layers:
-            state_dict.update(cls.pack_weights(text_encoder_2_lora_layers, "text_encoder_2"))
+            state_dict.update(pack_weights(text_encoder_2_lora_layers, "text_encoder_2"))
 
-        # Save the model
         cls.write_lora_layers(
             state_dict=state_dict,
             save_directory=save_directory,
@@ -1411,79 +1322,193 @@ class SD3LoraLoaderMixin(LoraBaseMixin):
             safe_serialization=safe_serialization,
         )
 
-    def fuse_lora(
-        self,
-        components: List[str] = ["transformer", "text_encoder", "text_encoder_2"],
-        lora_scale: float = 1.0,
-        safe_fusing: bool = False,
-        adapter_names: Optional[List[str]] = None,
-        **kwargs,
-    ):
-        r"""
-        Fuses the LoRA parameters into the original parameters of the corresponding blocks.
+    def _remove_text_encoder_monkey_patch(self):
+        recurse_remove_peft_layers(self.text_encoder)
+        # TODO: @younesbelkada handle this in transformers side
+        if getattr(self.text_encoder, "peft_config", None) is not None:
+            del self.text_encoder.peft_config
+            self.text_encoder._hf_peft_config_loaded = None
 
-        <Tip warning={true}>
+        recurse_remove_peft_layers(self.text_encoder_2)
+        if getattr(self.text_encoder_2, "peft_config", None) is not None:
+            del self.text_encoder_2.peft_config
+            self.text_encoder_2._hf_peft_config_loaded = None
 
-        This is an experimental API.
 
-        </Tip>
+class SD3LoraLoaderMixin:
+    r"""
+    Load LoRA layers into [`SD3Transformer2DModel`].
+    """
 
-        Args:
-            components: (`List[str]`): List of LoRA-injectable components to fuse the LoRAs into.
-            lora_scale (`float`, defaults to 1.0):
-                Controls how much to influence the outputs with the LoRA parameters.
-            safe_fusing (`bool`, defaults to `False`):
-                Whether to check fused weights for NaN values before fusing and if values are NaN not fusing them.
-            adapter_names (`List[str]`, *optional*):
-                Adapter names to be used for fusing. If nothing is passed, all active adapters will be fused.
+    transformer_name = TRANSFORMER_NAME
+    num_fused_loras = 0
 
-        Example:
+    def load_lora_weights(
+        self, pretrained_model_name_or_path_or_dict: Union[str, Dict[str, torch.Tensor]], adapter_name=None, **kwargs
+    ):
+        """
+        Load LoRA weights specified in `pretrained_model_name_or_path_or_dict` into `self.unet` and
+        `self.text_encoder`.
 
-        ```py
-        from diffusers import DiffusionPipeline
-        import torch
+        All kwargs are forwarded to `self.lora_state_dict`.
 
-        pipeline = DiffusionPipeline.from_pretrained(
-            "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
-        ).to("cuda")
-        pipeline.load_lora_weights("nerijs/pixel-art-xl", weight_name="pixel-art-xl.safetensors", adapter_name="pixel")
-        pipeline.fuse_lora(lora_scale=0.7)
-        ```
+        See [`~loaders.LoraLoaderMixin.lora_state_dict`] for more details on how the state dict is loaded.
+
+        See [`~loaders.LoraLoaderMixin.load_lora_into_transformer`] for more details on how the state dict is loaded
+        into `self.transformer`.
+
+        Parameters:
+            pretrained_model_name_or_path_or_dict (`str` or `os.PathLike` or `dict`):
+                See [`~loaders.LoraLoaderMixin.lora_state_dict`].
+            kwargs (`dict`, *optional*):
+                See [`~loaders.LoraLoaderMixin.lora_state_dict`].
+            adapter_name (`str`, *optional*):
+                Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
+                `default_{i}` where i is the total number of adapters being loaded.
         """
-        super().fuse_lora(
-            components=components, lora_scale=lora_scale, safe_fusing=safe_fusing, adapter_names=adapter_names
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for this method.")
+
+        # if a dict is passed, copy it instead of modifying it inplace
+        if isinstance(pretrained_model_name_or_path_or_dict, dict):
+            pretrained_model_name_or_path_or_dict = pretrained_model_name_or_path_or_dict.copy()
+
+        # First, ensure that the checkpoint is a compatible one and can be successfully loaded.
+        state_dict = self.lora_state_dict(pretrained_model_name_or_path_or_dict, **kwargs)
+
+        is_correct_format = all("lora" in key or "dora_scale" in key for key in state_dict.keys())
+        if not is_correct_format:
+            raise ValueError("Invalid LoRA checkpoint.")
+
+        self.load_lora_into_transformer(
+            state_dict,
+            transformer=getattr(self, self.transformer_name) if not hasattr(self, "transformer") else self.transformer,
+            adapter_name=adapter_name,
+            _pipeline=self,
         )
 
-    def unfuse_lora(self, components: List[str] = ["transformer", "text_encoder", "text_encoder_2"], **kwargs):
+    @classmethod
+    @validate_hf_hub_args
+    def lora_state_dict(
+        cls,
+        pretrained_model_name_or_path_or_dict: Union[str, Dict[str, torch.Tensor]],
+        **kwargs,
+    ):
         r"""
-        Reverses the effect of
-        [`pipe.fuse_lora()`](https://huggingface.co/docs/diffusers/main/en/api/loaders#diffusers.loaders.LoraBaseMixin.fuse_lora).
+        Return state dict for lora weights and the network alphas.
 
         <Tip warning={true}>
 
-        This is an experimental API.
+        We support loading A1111 formatted LoRA checkpoints in a limited capacity.
+
+        This function is experimental and might change in the future.
 
         </Tip>
 
-        Args:
-            components (`List[str]`): List of LoRA-injectable components to unfuse LoRA from.
-            unfuse_unet (`bool`, defaults to `True`): Whether to unfuse the UNet LoRA parameters.
-            unfuse_text_encoder (`bool`, defaults to `True`):
-                Whether to unfuse the text encoder LoRA parameters. If the text encoder wasn't monkey-patched with the
-                LoRA parameters then it won't have any effect.
+        Parameters:
+            pretrained_model_name_or_path_or_dict (`str` or `os.PathLike` or `dict`):
+                Can be either:
+
+                    - A string, the *model id* (for example `google/ddpm-celebahq-256`) of a pretrained model hosted on
+                      the Hub.
+                    - A path to a *directory* (for example `./my_model_directory`) containing the model weights saved
+                      with [`ModelMixin.save_pretrained`].
+                    - A [torch state
+                      dict](https://pytorch.org/tutorials/beginner/saving_loading_models.html#what-is-a-state-dict).
+
+            cache_dir (`Union[str, os.PathLike]`, *optional*):
+                Path to a directory where a downloaded pretrained model configuration is cached if the standard cache
+                is not used.
+            force_download (`bool`, *optional*, defaults to `False`):
+                Whether or not to force the (re-)download of the model weights and configuration files, overriding the
+                cached versions if they exist.
+
+            proxies (`Dict[str, str]`, *optional*):
+                A dictionary of proxy servers to use by protocol or endpoint, for example, `{'http': 'foo.bar:3128',
+                'http://hostname': 'foo.bar:4012'}`. The proxies are used on each request.
+            local_files_only (`bool`, *optional*, defaults to `False`):
+                Whether to only load local model weights and configuration files or not. If set to `True`, the model
+                won't be downloaded from the Hub.
+            token (`str` or *bool*, *optional*):
+                The token to use as HTTP bearer authorization for remote files. If `True`, the token generated from
+                `diffusers-cli login` (stored in `~/.huggingface`) is used.
+            revision (`str`, *optional*, defaults to `"main"`):
+                The specific model version to use. It can be a branch name, a tag name, a commit id, or any identifier
+                allowed by Git.
+            subfolder (`str`, *optional*, defaults to `""`):
+                The subfolder location of a model file within a larger model repository on the Hub or locally.
+
         """
-        super().unfuse_lora(components=components)
+        # Load the main state dict first which has the LoRA layers for either of
+        # UNet and text encoder or both.
+        cache_dir = kwargs.pop("cache_dir", None)
+        force_download = kwargs.pop("force_download", False)
+        proxies = kwargs.pop("proxies", None)
+        local_files_only = kwargs.pop("local_files_only", None)
+        token = kwargs.pop("token", None)
+        revision = kwargs.pop("revision", None)
+        subfolder = kwargs.pop("subfolder", None)
+        weight_name = kwargs.pop("weight_name", None)
+        use_safetensors = kwargs.pop("use_safetensors", None)
+
+        allow_pickle = False
+        if use_safetensors is None:
+            use_safetensors = True
+            allow_pickle = True
+
+        user_agent = {
+            "file_type": "attn_procs_weights",
+            "framework": "pytorch",
+        }
 
+        model_file = None
+        if not isinstance(pretrained_model_name_or_path_or_dict, dict):
+            # Let's first try to load .safetensors weights
+            if (use_safetensors and weight_name is None) or (
+                weight_name is not None and weight_name.endswith(".safetensors")
+            ):
+                try:
+                    model_file = _get_model_file(
+                        pretrained_model_name_or_path_or_dict,
+                        weights_name=weight_name or LORA_WEIGHT_NAME_SAFE,
+                        cache_dir=cache_dir,
+                        force_download=force_download,
+                        proxies=proxies,
+                        local_files_only=local_files_only,
+                        token=token,
+                        revision=revision,
+                        subfolder=subfolder,
+                        user_agent=user_agent,
+                    )
+                    state_dict = safetensors.torch.load_file(model_file, device="cpu")
+                except (IOError, safetensors.SafetensorError) as e:
+                    if not allow_pickle:
+                        raise e
+                    # try loading non-safetensors weights
+                    model_file = None
+                    pass
+
+            if model_file is None:
+                model_file = _get_model_file(
+                    pretrained_model_name_or_path_or_dict,
+                    weights_name=weight_name or LORA_WEIGHT_NAME,
+                    cache_dir=cache_dir,
+                    force_download=force_download,
+                    proxies=proxies,
+                    local_files_only=local_files_only,
+                    token=token,
+                    revision=revision,
+                    subfolder=subfolder,
+                    user_agent=user_agent,
+                )
+                state_dict = load_state_dict(model_file)
+        else:
+            state_dict = pretrained_model_name_or_path_or_dict
 
-# The reason why we subclass from `StableDiffusionLoraLoaderMixin` here is because Amused initially
-# relied on `StableDiffusionLoraLoaderMixin` for its LoRA support.
-class AmusedLoraLoaderMixin(StableDiffusionLoraLoaderMixin):
-    _lora_loadable_modules = ["transformer", "text_encoder"]
-    transformer_name = TRANSFORMER_NAME
-    text_encoder_name = TEXT_ENCODER_NAME
+        return state_dict
 
     @classmethod
-    def load_lora_into_transformer(cls, state_dict, network_alphas, transformer, adapter_name=None, _pipeline=None):
+    def load_lora_into_transformer(cls, state_dict, transformer, adapter_name=None, _pipeline=None):
         """
         This will load the LoRA layers specified in `state_dict` into `transformer`.
 
@@ -1492,17 +1517,12 @@ class AmusedLoraLoaderMixin(StableDiffusionLoraLoaderMixin):
                 A standard state dict containing the lora layer parameters. The keys can either be indexed directly
                 into the unet or prefixed with an additional `unet` which can be used to distinguish between text
                 encoder lora layers.
-            network_alphas (`Dict[str, float]`):
-                See `LoRALinearLayer` for more details.
-            unet (`UNet2DConditionModel`):
-                The UNet model to load the LoRA layers into.
+            transformer (`SD3Transformer2DModel`):
+                The Transformer model to load the LoRA layers into.
             adapter_name (`str`, *optional*):
                 Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
                 `default_{i}` where i is the total number of adapters being loaded.
         """
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for this method.")
-
         from peft import LoraConfig, inject_adapter_in_model, set_peft_model_state_dict
 
         keys = list(state_dict.keys())
@@ -1512,13 +1532,12 @@ class AmusedLoraLoaderMixin(StableDiffusionLoraLoaderMixin):
             k.replace(f"{cls.transformer_name}.", ""): v for k, v in state_dict.items() if k in transformer_keys
         }
 
-        if network_alphas is not None:
-            alpha_keys = [k for k in network_alphas.keys() if k.startswith(cls.transformer_name)]
-            network_alphas = {
-                k.replace(f"{cls.transformer_name}.", ""): v for k, v in network_alphas.items() if k in alpha_keys
-            }
-
         if len(state_dict.keys()) > 0:
+            # check with first key if is not in peft format
+            first_key = next(iter(state_dict.keys()))
+            if "lora_A" not in first_key:
+                state_dict = convert_unet_state_dict_to_peft(state_dict)
+
             if adapter_name in getattr(transformer, "peft_config", {}):
                 raise ValueError(
                     f"Adapter name {adapter_name} already in use in the transformer - please select a new adapter name."
@@ -1529,7 +1548,7 @@ class AmusedLoraLoaderMixin(StableDiffusionLoraLoaderMixin):
                 if "lora_B" in key:
                     rank[key] = val.shape[1]
 
-            lora_config_kwargs = get_peft_kwargs(rank, network_alphas, state_dict)
+            lora_config_kwargs = get_peft_kwargs(rank, network_alpha_dict=None, peft_state_dict=state_dict)
             if "use_dora" in lora_config_kwargs:
                 if lora_config_kwargs["use_dora"] and is_peft_version("<", "0.9.0"):
                     raise ValueError(
@@ -1566,131 +1585,13 @@ class AmusedLoraLoaderMixin(StableDiffusionLoraLoaderMixin):
                 _pipeline.enable_sequential_cpu_offload()
             # Unsafe code />
 
-    @classmethod
-    # Copied from diffusers.loaders.lora_pipeline.StableDiffusionLoraLoaderMixin.load_lora_into_text_encoder
-    def load_lora_into_text_encoder(
-        cls,
-        state_dict,
-        network_alphas,
-        text_encoder,
-        prefix=None,
-        lora_scale=1.0,
-        adapter_name=None,
-        _pipeline=None,
-    ):
-        """
-        This will load the LoRA layers specified in `state_dict` into `text_encoder`
-
-        Parameters:
-            state_dict (`dict`):
-                A standard state dict containing the lora layer parameters. The key should be prefixed with an
-                additional `text_encoder` to distinguish between unet lora layers.
-            network_alphas (`Dict[str, float]`):
-                See `LoRALinearLayer` for more details.
-            text_encoder (`CLIPTextModel`):
-                The text encoder model to load the LoRA layers into.
-            prefix (`str`):
-                Expected prefix of the `text_encoder` in the `state_dict`.
-            lora_scale (`float`):
-                How much to scale the output of the lora linear layer before it is added with the output of the regular
-                lora layer.
-            adapter_name (`str`, *optional*):
-                Adapter name to be used for referencing the loaded adapter model. If not specified, it will use
-                `default_{i}` where i is the total number of adapters being loaded.
-        """
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for this method.")
-
-        from peft import LoraConfig
-
-        # If the serialization format is new (introduced in https://github.com/huggingface/diffusers/pull/2918),
-        # then the `state_dict` keys should have `self.unet_name` and/or `self.text_encoder_name` as
-        # their prefixes.
-        keys = list(state_dict.keys())
-        prefix = cls.text_encoder_name if prefix is None else prefix
-
-        # Safe prefix to check with.
-        if any(cls.text_encoder_name in key for key in keys):
-            # Load the layers corresponding to text encoder and make necessary adjustments.
-            text_encoder_keys = [k for k in keys if k.startswith(prefix) and k.split(".")[0] == prefix]
-            text_encoder_lora_state_dict = {
-                k.replace(f"{prefix}.", ""): v for k, v in state_dict.items() if k in text_encoder_keys
-            }
-
-            if len(text_encoder_lora_state_dict) > 0:
-                logger.info(f"Loading {prefix}.")
-                rank = {}
-                text_encoder_lora_state_dict = convert_state_dict_to_diffusers(text_encoder_lora_state_dict)
-
-                # convert state dict
-                text_encoder_lora_state_dict = convert_state_dict_to_peft(text_encoder_lora_state_dict)
-
-                for name, _ in text_encoder_attn_modules(text_encoder):
-                    for module in ("out_proj", "q_proj", "k_proj", "v_proj"):
-                        rank_key = f"{name}.{module}.lora_B.weight"
-                        if rank_key not in text_encoder_lora_state_dict:
-                            continue
-                        rank[rank_key] = text_encoder_lora_state_dict[rank_key].shape[1]
-
-                for name, _ in text_encoder_mlp_modules(text_encoder):
-                    for module in ("fc1", "fc2"):
-                        rank_key = f"{name}.{module}.lora_B.weight"
-                        if rank_key not in text_encoder_lora_state_dict:
-                            continue
-                        rank[rank_key] = text_encoder_lora_state_dict[rank_key].shape[1]
-
-                if network_alphas is not None:
-                    alpha_keys = [
-                        k for k in network_alphas.keys() if k.startswith(prefix) and k.split(".")[0] == prefix
-                    ]
-                    network_alphas = {
-                        k.replace(f"{prefix}.", ""): v for k, v in network_alphas.items() if k in alpha_keys
-                    }
-
-                lora_config_kwargs = get_peft_kwargs(rank, network_alphas, text_encoder_lora_state_dict, is_unet=False)
-                if "use_dora" in lora_config_kwargs:
-                    if lora_config_kwargs["use_dora"]:
-                        if is_peft_version("<", "0.9.0"):
-                            raise ValueError(
-                                "You need `peft` 0.9.0 at least to use DoRA-enabled LoRAs. Please upgrade your installation of `peft`."
-                            )
-                    else:
-                        if is_peft_version("<", "0.9.0"):
-                            lora_config_kwargs.pop("use_dora")
-                lora_config = LoraConfig(**lora_config_kwargs)
-
-                # adapter_name
-                if adapter_name is None:
-                    adapter_name = get_adapter_name(text_encoder)
-
-                is_model_cpu_offload, is_sequential_cpu_offload = cls._optionally_disable_offloading(_pipeline)
-
-                # inject LoRA layers and load the state dict
-                # in transformers we automatically check whether the adapter name is already in use or not
-                text_encoder.load_adapter(
-                    adapter_name=adapter_name,
-                    adapter_state_dict=text_encoder_lora_state_dict,
-                    peft_config=lora_config,
-                )
-
-                # scale LoRA layers with `lora_scale`
-                scale_lora_layers(text_encoder, weight=lora_scale)
-
-                text_encoder.to(device=text_encoder.device, dtype=text_encoder.dtype)
-
-                # Offload back.
-                if is_model_cpu_offload:
-                    _pipeline.enable_model_cpu_offload()
-                elif is_sequential_cpu_offload:
-                    _pipeline.enable_sequential_cpu_offload()
-                # Unsafe code />
-
     @classmethod
     def save_lora_weights(
         cls,
         save_directory: Union[str, os.PathLike],
-        text_encoder_lora_layers: Dict[str, torch.nn.Module] = None,
         transformer_lora_layers: Dict[str, torch.nn.Module] = None,
+        text_encoder_lora_layers: Dict[str, Union[torch.nn.Module, torch.Tensor]] = None,
+        text_encoder_2_lora_layers: Dict[str, Union[torch.nn.Module, torch.Tensor]] = None,
         is_main_process: bool = True,
         weight_name: str = None,
         save_function: Callable = None,
@@ -1702,11 +1603,8 @@ class AmusedLoraLoaderMixin(StableDiffusionLoraLoaderMixin):
         Arguments:
             save_directory (`str` or `os.PathLike`):
                 Directory to save LoRA parameters to. Will be created if it doesn't exist.
-            unet_lora_layers (`Dict[str, torch.nn.Module]` or `Dict[str, torch.Tensor]`):
-                State dict of the LoRA layers corresponding to the `unet`.
-            text_encoder_lora_layers (`Dict[str, torch.nn.Module]` or `Dict[str, torch.Tensor]`):
-                State dict of the LoRA layers corresponding to the `text_encoder`. Must explicitly pass the text
-                encoder LoRA state dict because it comes from 🤗 Transformers.
+            transformer_lora_layers (`Dict[str, torch.nn.Module]` or `Dict[str, torch.Tensor]`):
+                State dict of the LoRA layers corresponding to the `transformer`.
             is_main_process (`bool`, *optional*, defaults to `True`):
                 Whether the process calling this is the main process or not. Useful during distributed training and you
                 need to call this function on all processes. In this case, set `is_main_process=True` only on the main
@@ -1720,14 +1618,24 @@ class AmusedLoraLoaderMixin(StableDiffusionLoraLoaderMixin):
         """
         state_dict = {}
 
-        if not (transformer_lora_layers or text_encoder_lora_layers):
-            raise ValueError("You must pass at least one of `transformer_lora_layers` or `text_encoder_lora_layers`.")
+        def pack_weights(layers, prefix):
+            layers_weights = layers.state_dict() if isinstance(layers, torch.nn.Module) else layers
+            layers_state_dict = {f"{prefix}.{module_name}": param for module_name, param in layers_weights.items()}
+            return layers_state_dict
+
+        if not (transformer_lora_layers or text_encoder_lora_layers or text_encoder_2_lora_layers):
+            raise ValueError(
+                "You must pass at least one of `transformer_lora_layers`, `text_encoder_lora_layers`, `text_encoder_2_lora_layers`."
+            )
 
         if transformer_lora_layers:
-            state_dict.update(cls.pack_weights(transformer_lora_layers, cls.transformer_name))
+            state_dict.update(pack_weights(transformer_lora_layers, cls.transformer_name))
 
         if text_encoder_lora_layers:
-            state_dict.update(cls.pack_weights(text_encoder_lora_layers, cls.text_encoder_name))
+            state_dict.update(pack_weights(text_encoder_lora_layers, "text_encoder"))
+
+        if text_encoder_2_lora_layers:
+            state_dict.update(pack_weights(text_encoder_2_lora_layers, "text_encoder_2"))
 
         # Save the model
         cls.write_lora_layers(
@@ -1739,9 +1647,164 @@ class AmusedLoraLoaderMixin(StableDiffusionLoraLoaderMixin):
             safe_serialization=safe_serialization,
         )
 
+    @staticmethod
+    def write_lora_layers(
+        state_dict: Dict[str, torch.Tensor],
+        save_directory: str,
+        is_main_process: bool,
+        weight_name: str,
+        save_function: Callable,
+        safe_serialization: bool,
+    ):
+        if os.path.isfile(save_directory):
+            logger.error(f"Provided path ({save_directory}) should be a directory, not a file")
+            return
+
+        if save_function is None:
+            if safe_serialization:
+
+                def save_function(weights, filename):
+                    return safetensors.torch.save_file(weights, filename, metadata={"format": "pt"})
+
+            else:
+                save_function = torch.save
+
+        os.makedirs(save_directory, exist_ok=True)
+
+        if weight_name is None:
+            if safe_serialization:
+                weight_name = LORA_WEIGHT_NAME_SAFE
+            else:
+                weight_name = LORA_WEIGHT_NAME
+
+        save_path = Path(save_directory, weight_name).as_posix()
+        save_function(state_dict, save_path)
+        logger.info(f"Model weights saved in {save_path}")
+
+    def unload_lora_weights(self):
+        """
+        Unloads the LoRA parameters.
+
+        Examples:
+
+        ```python
+        >>> # Assuming `pipeline` is already loaded with the LoRA parameters.
+        >>> pipeline.unload_lora_weights()
+        >>> ...
+        ```
+        """
+        transformer = getattr(self, self.transformer_name) if not hasattr(self, "transformer") else self.transformer
+        recurse_remove_peft_layers(transformer)
+        if hasattr(transformer, "peft_config"):
+            del transformer.peft_config
+
+    @classmethod
+    # Copied from diffusers.loaders.lora.LoraLoaderMixin._optionally_disable_offloading
+    def _optionally_disable_offloading(cls, _pipeline):
+        """
+        Optionally removes offloading in case the pipeline has been already sequentially offloaded to CPU.
+
+        Args:
+            _pipeline (`DiffusionPipeline`):
+                The pipeline to disable offloading for.
+
+        Returns:
+            tuple:
+                A tuple indicating if `is_model_cpu_offload` or `is_sequential_cpu_offload` is True.
+        """
+        is_model_cpu_offload = False
+        is_sequential_cpu_offload = False
+
+        if _pipeline is not None and _pipeline.hf_device_map is None:
+            for _, component in _pipeline.components.items():
+                if isinstance(component, nn.Module) and hasattr(component, "_hf_hook"):
+                    if not is_model_cpu_offload:
+                        is_model_cpu_offload = isinstance(component._hf_hook, CpuOffload)
+                    if not is_sequential_cpu_offload:
+                        is_sequential_cpu_offload = (
+                            isinstance(component._hf_hook, AlignDevicesHook)
+                            or hasattr(component._hf_hook, "hooks")
+                            and isinstance(component._hf_hook.hooks[0], AlignDevicesHook)
+                        )
+
+                    logger.info(
+                        "Accelerate hooks detected. Since you have called `load_lora_weights()`, the previous hooks will be first removed. Then the LoRA parameters will be loaded and the hooks will be applied again."
+                    )
+                    remove_hook_from_module(component, recurse=is_sequential_cpu_offload)
+
+        return (is_model_cpu_offload, is_sequential_cpu_offload)
+
+    def fuse_lora(
+        self,
+        fuse_transformer: bool = True,
+        lora_scale: float = 1.0,
+        safe_fusing: bool = False,
+        adapter_names: Optional[List[str]] = None,
+    ):
+        r"""
+        Fuses the LoRA parameters into the original parameters of the corresponding blocks.
+
+        <Tip warning={true}>
+
+        This is an experimental API.
+
+        </Tip>
+
+        Args:
+            fuse_transformer (`bool`, defaults to `True`): Whether to fuse the transformer LoRA parameters.
+            lora_scale (`float`, defaults to 1.0):
+                Controls how much to influence the outputs with the LoRA parameters.
+            safe_fusing (`bool`, defaults to `False`):
+                Whether to check fused weights for NaN values before fusing and if values are NaN not fusing them.
+            adapter_names (`List[str]`, *optional*):
+                Adapter names to be used for fusing. If nothing is passed, all active adapters will be fused.
+
+        Example:
+
+        ```py
+        from diffusers import DiffusionPipeline
+        import torch
+
+        pipeline = DiffusionPipeline.from_pretrained(
+            "stabilityai/stable-diffusion-3-medium-diffusers", torch_dtype=torch.float16
+        ).to("cuda")
+        pipeline.load_lora_weights(
+            "nerijs/pixel-art-medium-128-v0.1",
+            weight_name="pixel-art-medium-128-v0.1.safetensors",
+            adapter_name="pixel",
+        )
+        pipeline.fuse_lora(lora_scale=0.7)
+        ```
+        """
+        if fuse_transformer:
+            self.num_fused_loras += 1
+
+        if fuse_transformer:
+            transformer = (
+                getattr(self, self.transformer_name) if not hasattr(self, "transformer") else self.transformer
+            )
+            transformer.fuse_lora(lora_scale, safe_fusing=safe_fusing, adapter_names=adapter_names)
+
+    def unfuse_lora(self, unfuse_transformer: bool = True):
+        r"""
+        Reverses the effect of
+        [`pipe.fuse_lora()`](https://huggingface.co/docs/diffusers/main/en/api/loaders#diffusers.loaders.LoraLoaderMixin.fuse_lora).
+
+        <Tip warning={true}>
+
+        This is an experimental API.
+
+        </Tip>
+
+        Args:
+            unfuse_transformer (`bool`, defaults to `True`): Whether to unfuse the transformer LoRA parameters.
+        """
+        from peft.tuners.tuners_utils import BaseTunerLayer
+
+        transformer = getattr(self, self.transformer_name) if not hasattr(self, "transformer") else self.transformer
+        if unfuse_transformer:
+            for module in transformer.modules():
+                if isinstance(module, BaseTunerLayer):
+                    module.unmerge()
 
-class LoraLoaderMixin(StableDiffusionLoraLoaderMixin):
-    def __init__(self, *args, **kwargs):
-        deprecation_message = "LoraLoaderMixin is deprecated and this will be removed in a future version. Please use `StableDiffusionLoraLoaderMixin`, instead."
-        deprecate("LoraLoaderMixin", "1.0.0", deprecation_message)
-        super().__init__(*args, **kwargs)
+        self.num_fused_loras -= 1

src/diffusers/loaders/lora_base.py

-# 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.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import copy
-import inspect
-import os
-from pathlib import Path
-from typing import Callable, Dict, List, Optional, Union
-
-import safetensors
-import torch
-import torch.nn as nn
-from huggingface_hub import model_info
-from huggingface_hub.constants import HF_HUB_OFFLINE
-
-from ..models.modeling_utils import ModelMixin, load_state_dict
-from ..utils import (
-    USE_PEFT_BACKEND,
-    _get_model_file,
-    delete_adapter_layers,
-    deprecate,
-    is_accelerate_available,
-    is_peft_available,
-    is_transformers_available,
-    logging,
-    recurse_remove_peft_layers,
-    set_adapter_layers,
-    set_weights_and_activate_adapters,
-)
-
-
-if is_transformers_available():
-    from transformers import PreTrainedModel
-
-if is_peft_available():
-    from peft.tuners.tuners_utils import BaseTunerLayer
-
-if is_accelerate_available():
-    from accelerate.hooks import AlignDevicesHook, CpuOffload, remove_hook_from_module
-
-logger = logging.get_logger(__name__)
-
-
-def fuse_text_encoder_lora(text_encoder, lora_scale=1.0, safe_fusing=False, adapter_names=None):
-    """
-    Fuses LoRAs for the text encoder.
-
-    Args:
-        text_encoder (`torch.nn.Module`):
-            The text encoder module to set the adapter layers for. If `None`, it will try to get the `text_encoder`
-            attribute.
-        lora_scale (`float`, defaults to 1.0):
-            Controls how much to influence the outputs with the LoRA parameters.
-        safe_fusing (`bool`, defaults to `False`):
-            Whether to check fused weights for NaN values before fusing and if values are NaN not fusing them.
-        adapter_names (`List[str]` or `str`):
-            The names of the adapters to use.
-    """
-    merge_kwargs = {"safe_merge": safe_fusing}
-
-    for module in text_encoder.modules():
-        if isinstance(module, BaseTunerLayer):
-            if lora_scale != 1.0:
-                module.scale_layer(lora_scale)
-
-            # For BC with previous PEFT versions, we need to check the signature
-            # of the `merge` method to see if it supports the `adapter_names` argument.
-            supported_merge_kwargs = list(inspect.signature(module.merge).parameters)
-            if "adapter_names" in supported_merge_kwargs:
-                merge_kwargs["adapter_names"] = adapter_names
-            elif "adapter_names" not in supported_merge_kwargs and adapter_names is not None:
-                raise ValueError(
-                    "The `adapter_names` argument is not supported with your PEFT version. "
-                    "Please upgrade to the latest version of PEFT. `pip install -U peft`"
-                )
-
-            module.merge(**merge_kwargs)
-
-
-def unfuse_text_encoder_lora(text_encoder):
-    """
-    Unfuses LoRAs for the text encoder.
-
-    Args:
-        text_encoder (`torch.nn.Module`):
-            The text encoder module to set the adapter layers for. If `None`, it will try to get the `text_encoder`
-            attribute.
-    """
-    for module in text_encoder.modules():
-        if isinstance(module, BaseTunerLayer):
-            module.unmerge()
-
-
-def set_adapters_for_text_encoder(
-    adapter_names: Union[List[str], str],
-    text_encoder: Optional["PreTrainedModel"] = None,  # noqa: F821
-    text_encoder_weights: Optional[Union[float, List[float], List[None]]] = None,
-):
-    """
-    Sets the adapter layers for the text encoder.
-
-    Args:
-        adapter_names (`List[str]` or `str`):
-            The names of the adapters to use.
-        text_encoder (`torch.nn.Module`, *optional*):
-            The text encoder module to set the adapter layers for. If `None`, it will try to get the `text_encoder`
-            attribute.
-        text_encoder_weights (`List[float]`, *optional*):
-            The weights to use for the text encoder. If `None`, the weights are set to `1.0` for all the adapters.
-    """
-    if text_encoder is None:
-        raise ValueError(
-            "The pipeline does not have a default `pipe.text_encoder` class. Please make sure to pass a `text_encoder` instead."
-        )
-
-    def process_weights(adapter_names, weights):
-        # Expand weights into a list, one entry per adapter
-        # e.g. for 2 adapters:  7 -> [7,7] ; [3, None] -> [3, None]
-        if not isinstance(weights, list):
-            weights = [weights] * len(adapter_names)
-
-        if len(adapter_names) != len(weights):
-            raise ValueError(
-                f"Length of adapter names {len(adapter_names)} is not equal to the length of the weights {len(weights)}"
-            )
-
-        # Set None values to default of 1.0
-        # e.g. [7,7] -> [7,7] ; [3, None] -> [3,1]
-        weights = [w if w is not None else 1.0 for w in weights]
-
-        return weights
-
-    adapter_names = [adapter_names] if isinstance(adapter_names, str) else adapter_names
-    text_encoder_weights = process_weights(adapter_names, text_encoder_weights)
-    set_weights_and_activate_adapters(text_encoder, adapter_names, text_encoder_weights)
-
-
-def disable_lora_for_text_encoder(text_encoder: Optional["PreTrainedModel"] = None):
-    """
-    Disables the LoRA layers for the text encoder.
-
-    Args:
-        text_encoder (`torch.nn.Module`, *optional*):
-            The text encoder module to disable the LoRA layers for. If `None`, it will try to get the `text_encoder`
-            attribute.
-    """
-    if text_encoder is None:
-        raise ValueError("Text Encoder not found.")
-    set_adapter_layers(text_encoder, enabled=False)
-
-
-def enable_lora_for_text_encoder(text_encoder: Optional["PreTrainedModel"] = None):
-    """
-    Enables the LoRA layers for the text encoder.
-
-    Args:
-        text_encoder (`torch.nn.Module`, *optional*):
-            The text encoder module to enable the LoRA layers for. If `None`, it will try to get the `text_encoder`
-            attribute.
-    """
-    if text_encoder is None:
-        raise ValueError("Text Encoder not found.")
-    set_adapter_layers(text_encoder, enabled=True)
-
-
-def _remove_text_encoder_monkey_patch(text_encoder):
-    recurse_remove_peft_layers(text_encoder)
-    if getattr(text_encoder, "peft_config", None) is not None:
-        del text_encoder.peft_config
-        text_encoder._hf_peft_config_loaded = None
-
-
-class LoraBaseMixin:
-    """Utility class for handling LoRAs."""
-
-    _lora_loadable_modules = []
-    num_fused_loras = 0
-
-    def load_lora_weights(self, **kwargs):
-        raise NotImplementedError("`load_lora_weights()` is not implemented.")
-
-    @classmethod
-    def save_lora_weights(cls, **kwargs):
-        raise NotImplementedError("`save_lora_weights()` not implemented.")
-
-    @classmethod
-    def lora_state_dict(cls, **kwargs):
-        raise NotImplementedError("`lora_state_dict()` is not implemented.")
-
-    @classmethod
-    def _optionally_disable_offloading(cls, _pipeline):
-        """
-        Optionally removes offloading in case the pipeline has been already sequentially offloaded to CPU.
-
-        Args:
-            _pipeline (`DiffusionPipeline`):
-                The pipeline to disable offloading for.
-
-        Returns:
-            tuple:
-                A tuple indicating if `is_model_cpu_offload` or `is_sequential_cpu_offload` is True.
-        """
-        is_model_cpu_offload = False
-        is_sequential_cpu_offload = False
-
-        if _pipeline is not None and _pipeline.hf_device_map is None:
-            for _, component in _pipeline.components.items():
-                if isinstance(component, nn.Module) and hasattr(component, "_hf_hook"):
-                    if not is_model_cpu_offload:
-                        is_model_cpu_offload = isinstance(component._hf_hook, CpuOffload)
-                    if not is_sequential_cpu_offload:
-                        is_sequential_cpu_offload = (
-                            isinstance(component._hf_hook, AlignDevicesHook)
-                            or hasattr(component._hf_hook, "hooks")
-                            and isinstance(component._hf_hook.hooks[0], AlignDevicesHook)
-                        )
-
-                    logger.info(
-                        "Accelerate hooks detected. Since you have called `load_lora_weights()`, the previous hooks will be first removed. Then the LoRA parameters will be loaded and the hooks will be applied again."
-                    )
-                    remove_hook_from_module(component, recurse=is_sequential_cpu_offload)
-
-        return (is_model_cpu_offload, is_sequential_cpu_offload)
-
-    @classmethod
-    def _fetch_state_dict(
-        cls,
-        pretrained_model_name_or_path_or_dict,
-        weight_name,
-        use_safetensors,
-        local_files_only,
-        cache_dir,
-        force_download,
-        proxies,
-        token,
-        revision,
-        subfolder,
-        user_agent,
-        allow_pickle,
-    ):
-        from .lora_pipeline import LORA_WEIGHT_NAME, LORA_WEIGHT_NAME_SAFE
-
-        model_file = None
-        if not isinstance(pretrained_model_name_or_path_or_dict, dict):
-            # Let's first try to load .safetensors weights
-            if (use_safetensors and weight_name is None) or (
-                weight_name is not None and weight_name.endswith(".safetensors")
-            ):
-                try:
-                    # Here we're relaxing the loading check to enable more Inference API
-                    # friendliness where sometimes, it's not at all possible to automatically
-                    # determine `weight_name`.
-                    if weight_name is None:
-                        weight_name = cls._best_guess_weight_name(
-                            pretrained_model_name_or_path_or_dict,
-                            file_extension=".safetensors",
-                            local_files_only=local_files_only,
-                        )
-                    model_file = _get_model_file(
-                        pretrained_model_name_or_path_or_dict,
-                        weights_name=weight_name or LORA_WEIGHT_NAME_SAFE,
-                        cache_dir=cache_dir,
-                        force_download=force_download,
-                        proxies=proxies,
-                        local_files_only=local_files_only,
-                        token=token,
-                        revision=revision,
-                        subfolder=subfolder,
-                        user_agent=user_agent,
-                    )
-                    state_dict = safetensors.torch.load_file(model_file, device="cpu")
-                except (IOError, safetensors.SafetensorError) as e:
-                    if not allow_pickle:
-                        raise e
-                    # try loading non-safetensors weights
-                    model_file = None
-                    pass
-
-            if model_file is None:
-                if weight_name is None:
-                    weight_name = cls._best_guess_weight_name(
-                        pretrained_model_name_or_path_or_dict, file_extension=".bin", local_files_only=local_files_only
-                    )
-                model_file = _get_model_file(
-                    pretrained_model_name_or_path_or_dict,
-                    weights_name=weight_name or LORA_WEIGHT_NAME,
-                    cache_dir=cache_dir,
-                    force_download=force_download,
-                    proxies=proxies,
-                    local_files_only=local_files_only,
-                    token=token,
-                    revision=revision,
-                    subfolder=subfolder,
-                    user_agent=user_agent,
-                )
-                state_dict = load_state_dict(model_file)
-        else:
-            state_dict = pretrained_model_name_or_path_or_dict
-
-        return state_dict
-
-    @classmethod
-    def _best_guess_weight_name(
-        cls, pretrained_model_name_or_path_or_dict, file_extension=".safetensors", local_files_only=False
-    ):
-        from .lora_pipeline import LORA_WEIGHT_NAME, LORA_WEIGHT_NAME_SAFE
-
-        if local_files_only or HF_HUB_OFFLINE:
-            raise ValueError("When using the offline mode, you must specify a `weight_name`.")
-
-        targeted_files = []
-
-        if os.path.isfile(pretrained_model_name_or_path_or_dict):
-            return
-        elif os.path.isdir(pretrained_model_name_or_path_or_dict):
-            targeted_files = [
-                f for f in os.listdir(pretrained_model_name_or_path_or_dict) if f.endswith(file_extension)
-            ]
-        else:
-            files_in_repo = model_info(pretrained_model_name_or_path_or_dict).siblings
-            targeted_files = [f.rfilename for f in files_in_repo if f.rfilename.endswith(file_extension)]
-        if len(targeted_files) == 0:
-            return
-
-        # "scheduler" does not correspond to a LoRA checkpoint.
-        # "optimizer" does not correspond to a LoRA checkpoint
-        # only top-level checkpoints are considered and not the other ones, hence "checkpoint".
-        unallowed_substrings = {"scheduler", "optimizer", "checkpoint"}
-        targeted_files = list(
-            filter(lambda x: all(substring not in x for substring in unallowed_substrings), targeted_files)
-        )
-
-        if any(f.endswith(LORA_WEIGHT_NAME) for f in targeted_files):
-            targeted_files = list(filter(lambda x: x.endswith(LORA_WEIGHT_NAME), targeted_files))
-        elif any(f.endswith(LORA_WEIGHT_NAME_SAFE) for f in targeted_files):
-            targeted_files = list(filter(lambda x: x.endswith(LORA_WEIGHT_NAME_SAFE), targeted_files))
-
-        if len(targeted_files) > 1:
-            raise ValueError(
-                f"Provided path contains more than one weights file in the {file_extension} format. Either specify `weight_name` in `load_lora_weights` or make sure there's only one  `.safetensors` or `.bin` file in  {pretrained_model_name_or_path_or_dict}."
-            )
-        weight_name = targeted_files[0]
-        return weight_name
-
-    def unload_lora_weights(self):
-        """
-        Unloads the LoRA parameters.
-
-        Examples:
-
-        ```python
-        >>> # Assuming `pipeline` is already loaded with the LoRA parameters.
-        >>> pipeline.unload_lora_weights()
-        >>> ...
-        ```
-        """
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for this method.")
-
-        for component in self._lora_loadable_modules:
-            model = getattr(self, component, None)
-            if model is not None:
-                if issubclass(model.__class__, ModelMixin):
-                    model.unload_lora()
-                elif issubclass(model.__class__, PreTrainedModel):
-                    _remove_text_encoder_monkey_patch(model)
-
-    def fuse_lora(
-        self,
-        components: List[str] = [],
-        lora_scale: float = 1.0,
-        safe_fusing: bool = False,
-        adapter_names: Optional[List[str]] = None,
-        **kwargs,
-    ):
-        r"""
-        Fuses the LoRA parameters into the original parameters of the corresponding blocks.
-
-        <Tip warning={true}>
-
-        This is an experimental API.
-
-        </Tip>
-
-        Args:
-            components: (`List[str]`): List of LoRA-injectable components to fuse the LoRAs into.
-            lora_scale (`float`, defaults to 1.0):
-                Controls how much to influence the outputs with the LoRA parameters.
-            safe_fusing (`bool`, defaults to `False`):
-                Whether to check fused weights for NaN values before fusing and if values are NaN not fusing them.
-            adapter_names (`List[str]`, *optional*):
-                Adapter names to be used for fusing. If nothing is passed, all active adapters will be fused.
-
-        Example:
-
-        ```py
-        from diffusers import DiffusionPipeline
-        import torch
-
-        pipeline = DiffusionPipeline.from_pretrained(
-            "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
-        ).to("cuda")
-        pipeline.load_lora_weights("nerijs/pixel-art-xl", weight_name="pixel-art-xl.safetensors", adapter_name="pixel")
-        pipeline.fuse_lora(lora_scale=0.7)
-        ```
-        """
-        if "fuse_unet" in kwargs:
-            depr_message = "Passing `fuse_unet` to `fuse_lora()` is deprecated and will be ignored. Please use the `components` argument and provide a list of the components whose LoRAs are to be fused. `fuse_unet` will be removed in a future version."
-            deprecate(
-                "fuse_unet",
-                "1.0.0",
-                depr_message,
-            )
-        if "fuse_transformer" in kwargs:
-            depr_message = "Passing `fuse_transformer` to `fuse_lora()` is deprecated and will be ignored. Please use the `components` argument and provide a list of the components whose LoRAs are to be fused. `fuse_transformer` will be removed in a future version."
-            deprecate(
-                "fuse_transformer",
-                "1.0.0",
-                depr_message,
-            )
-        if "fuse_text_encoder" in kwargs:
-            depr_message = "Passing `fuse_text_encoder` to `fuse_lora()` is deprecated and will be ignored. Please use the `components` argument and provide a list of the components whose LoRAs are to be fused. `fuse_text_encoder` will be removed in a future version."
-            deprecate(
-                "fuse_text_encoder",
-                "1.0.0",
-                depr_message,
-            )
-
-        if len(components) == 0:
-            raise ValueError("`components` cannot be an empty list.")
-
-        for fuse_component in components:
-            if fuse_component not in self._lora_loadable_modules:
-                raise ValueError(f"{fuse_component} is not found in {self._lora_loadable_modules=}.")
-
-            model = getattr(self, fuse_component, None)
-            if model is not None:
-                # check if diffusers model
-                if issubclass(model.__class__, ModelMixin):
-                    model.fuse_lora(lora_scale, safe_fusing=safe_fusing, adapter_names=adapter_names)
-                # handle transformers models.
-                if issubclass(model.__class__, PreTrainedModel):
-                    fuse_text_encoder_lora(
-                        model, lora_scale=lora_scale, safe_fusing=safe_fusing, adapter_names=adapter_names
-                    )
-
-        self.num_fused_loras += 1
-
-    def unfuse_lora(self, components: List[str] = [], **kwargs):
-        r"""
-        Reverses the effect of
-        [`pipe.fuse_lora()`](https://huggingface.co/docs/diffusers/main/en/api/loaders#diffusers.loaders.LoraBaseMixin.fuse_lora).
-
-        <Tip warning={true}>
-
-        This is an experimental API.
-
-        </Tip>
-
-        Args:
-            components (`List[str]`): List of LoRA-injectable components to unfuse LoRA from.
-            unfuse_unet (`bool`, defaults to `True`): Whether to unfuse the UNet LoRA parameters.
-            unfuse_text_encoder (`bool`, defaults to `True`):
-                Whether to unfuse the text encoder LoRA parameters. If the text encoder wasn't monkey-patched with the
-                LoRA parameters then it won't have any effect.
-        """
-        if "unfuse_unet" in kwargs:
-            depr_message = "Passing `unfuse_unet` to `unfuse_lora()` is deprecated and will be ignored. Please use the `components` argument. `unfuse_unet` will be removed in a future version."
-            deprecate(
-                "unfuse_unet",
-                "1.0.0",
-                depr_message,
-            )
-        if "unfuse_transformer" in kwargs:
-            depr_message = "Passing `unfuse_transformer` to `unfuse_lora()` is deprecated and will be ignored. Please use the `components` argument. `unfuse_transformer` will be removed in a future version."
-            deprecate(
-                "unfuse_transformer",
-                "1.0.0",
-                depr_message,
-            )
-        if "unfuse_text_encoder" in kwargs:
-            depr_message = "Passing `unfuse_text_encoder` to `unfuse_lora()` is deprecated and will be ignored. Please use the `components` argument. `unfuse_text_encoder` will be removed in a future version."
-            deprecate(
-                "unfuse_text_encoder",
-                "1.0.0",
-                depr_message,
-            )
-
-        if len(components) == 0:
-            raise ValueError("`components` cannot be an empty list.")
-
-        for fuse_component in components:
-            if fuse_component not in self._lora_loadable_modules:
-                raise ValueError(f"{fuse_component} is not found in {self._lora_loadable_modules=}.")
-
-            model = getattr(self, fuse_component, None)
-            if model is not None:
-                if issubclass(model.__class__, (ModelMixin, PreTrainedModel)):
-                    for module in model.modules():
-                        if isinstance(module, BaseTunerLayer):
-                            module.unmerge()
-
-        self.num_fused_loras -= 1
-
-    def set_adapters(
-        self,
-        adapter_names: Union[List[str], str],
-        adapter_weights: Optional[Union[float, Dict, List[float], List[Dict]]] = None,
-    ):
-        adapter_names = [adapter_names] if isinstance(adapter_names, str) else adapter_names
-
-        adapter_weights = copy.deepcopy(adapter_weights)
-
-        # Expand weights into a list, one entry per adapter
-        if not isinstance(adapter_weights, list):
-            adapter_weights = [adapter_weights] * len(adapter_names)
-
-        if len(adapter_names) != len(adapter_weights):
-            raise ValueError(
-                f"Length of adapter names {len(adapter_names)} is not equal to the length of the weights {len(adapter_weights)}"
-            )
-
-        list_adapters = self.get_list_adapters()  # eg {"unet": ["adapter1", "adapter2"], "text_encoder": ["adapter2"]}
-        all_adapters = {
-            adapter for adapters in list_adapters.values() for adapter in adapters
-        }  # eg ["adapter1", "adapter2"]
-        invert_list_adapters = {
-            adapter: [part for part, adapters in list_adapters.items() if adapter in adapters]
-            for adapter in all_adapters
-        }  # eg {"adapter1": ["unet"], "adapter2": ["unet", "text_encoder"]}
-
-        # Decompose weights into weights for denoiser and text encoders.
-        _component_adapter_weights = {}
-        for component in self._lora_loadable_modules:
-            model = getattr(self, component)
-
-            for adapter_name, weights in zip(adapter_names, adapter_weights):
-                if isinstance(weights, dict):
-                    component_adapter_weights = weights.pop(component, None)
-
-                    if component_adapter_weights is not None and not hasattr(self, component):
-                        logger.warning(
-                            f"Lora weight dict contains {component} weights but will be ignored because pipeline does not have {component}."
-                        )
-
-                    if component_adapter_weights is not None and component not in invert_list_adapters[adapter_name]:
-                        logger.warning(
-                            (
-                                f"Lora weight dict for adapter '{adapter_name}' contains {component},"
-                                f"but this will be ignored because {adapter_name} does not contain weights for {component}."
-                                f"Valid parts for {adapter_name} are: {invert_list_adapters[adapter_name]}."
-                            )
-                        )
-
-                else:
-                    component_adapter_weights = weights
-
-                _component_adapter_weights.setdefault(component, [])
-                _component_adapter_weights[component].append(component_adapter_weights)
-
-            if issubclass(model.__class__, ModelMixin):
-                model.set_adapters(adapter_names, _component_adapter_weights[component])
-            elif issubclass(model.__class__, PreTrainedModel):
-                set_adapters_for_text_encoder(adapter_names, model, _component_adapter_weights[component])
-
-    def disable_lora(self):
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for this method.")
-
-        for component in self._lora_loadable_modules:
-            model = getattr(self, component, None)
-            if model is not None:
-                if issubclass(model.__class__, ModelMixin):
-                    model.disable_lora()
-                elif issubclass(model.__class__, PreTrainedModel):
-                    disable_lora_for_text_encoder(model)
-
-    def enable_lora(self):
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for this method.")
-
-        for component in self._lora_loadable_modules:
-            model = getattr(self, component, None)
-            if model is not None:
-                if issubclass(model.__class__, ModelMixin):
-                    model.enable_lora()
-                elif issubclass(model.__class__, PreTrainedModel):
-                    enable_lora_for_text_encoder(model)
-
-    def delete_adapters(self, adapter_names: Union[List[str], str]):
-        """
-        Args:
-        Deletes the LoRA layers of `adapter_name` for the unet and text-encoder(s).
-            adapter_names (`Union[List[str], str]`):
-                The names of the adapter to delete. Can be a single string or a list of strings
-        """
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for this method.")
-
-        if isinstance(adapter_names, str):
-            adapter_names = [adapter_names]
-
-        for component in self._lora_loadable_modules:
-            model = getattr(self, component, None)
-            if model is not None:
-                if issubclass(model.__class__, ModelMixin):
-                    model.delete_adapters(adapter_names)
-                elif issubclass(model.__class__, PreTrainedModel):
-                    for adapter_name in adapter_names:
-                        delete_adapter_layers(model, adapter_name)
-
-    def get_active_adapters(self) -> List[str]:
-        """
-        Gets the list of the current active adapters.
-
-        Example:
-
-        ```python
-        from diffusers import DiffusionPipeline
-
-        pipeline = DiffusionPipeline.from_pretrained(
-            "stabilityai/stable-diffusion-xl-base-1.0",
-        ).to("cuda")
-        pipeline.load_lora_weights("CiroN2022/toy-face", weight_name="toy_face_sdxl.safetensors", adapter_name="toy")
-        pipeline.get_active_adapters()
-        ```
-        """
-        if not USE_PEFT_BACKEND:
-            raise ValueError(
-                "PEFT backend is required for this method. Please install the latest version of PEFT `pip install -U peft`"
-            )
-
-        active_adapters = []
-
-        for component in self._lora_loadable_modules:
-            model = getattr(self, component, None)
-            if model is not None and issubclass(model.__class__, ModelMixin):
-                for module in model.modules():
-                    if isinstance(module, BaseTunerLayer):
-                        active_adapters = module.active_adapters
-                        break
-
-        return active_adapters
-
-    def get_list_adapters(self) -> Dict[str, List[str]]:
-        """
-        Gets the current list of all available adapters in the pipeline.
-        """
-        if not USE_PEFT_BACKEND:
-            raise ValueError(
-                "PEFT backend is required for this method. Please install the latest version of PEFT `pip install -U peft`"
-            )
-
-        set_adapters = {}
-
-        for component in self._lora_loadable_modules:
-            model = getattr(self, component, None)
-            if (
-                model is not None
-                and issubclass(model.__class__, (ModelMixin, PreTrainedModel))
-                and hasattr(model, "peft_config")
-            ):
-                set_adapters[component] = list(model.peft_config.keys())
-
-        return set_adapters
-
-    def set_lora_device(self, adapter_names: List[str], device: Union[torch.device, str, int]) -> None:
-        """
-        Moves the LoRAs listed in `adapter_names` to a target device. Useful for offloading the LoRA to the CPU in case
-        you want to load multiple adapters and free some GPU memory.
-
-        Args:
-            adapter_names (`List[str]`):
-                List of adapters to send device to.
-            device (`Union[torch.device, str, int]`):
-                Device to send the adapters to. Can be either a torch device, a str or an integer.
-        """
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for this method.")
-
-        for component in self._lora_loadable_modules:
-            model = getattr(self, component, None)
-            if model is not None:
-                for module in model.modules():
-                    if isinstance(module, BaseTunerLayer):
-                        for adapter_name in adapter_names:
-                            module.lora_A[adapter_name].to(device)
-                            module.lora_B[adapter_name].to(device)
-                            # this is a param, not a module, so device placement is not in-place -> re-assign
-                            if hasattr(module, "lora_magnitude_vector") and module.lora_magnitude_vector is not None:
-                                module.lora_magnitude_vector[adapter_name] = module.lora_magnitude_vector[
-                                    adapter_name
-                                ].to(device)
-
-    @staticmethod
-    def pack_weights(layers, prefix):
-        layers_weights = layers.state_dict() if isinstance(layers, torch.nn.Module) else layers
-        layers_state_dict = {f"{prefix}.{module_name}": param for module_name, param in layers_weights.items()}
-        return layers_state_dict
-
-    @staticmethod
-    def write_lora_layers(
-        state_dict: Dict[str, torch.Tensor],
-        save_directory: str,
-        is_main_process: bool,
-        weight_name: str,
-        save_function: Callable,
-        safe_serialization: bool,
-    ):
-        from .lora_pipeline import LORA_WEIGHT_NAME, LORA_WEIGHT_NAME_SAFE
-
-        if os.path.isfile(save_directory):
-            logger.error(f"Provided path ({save_directory}) should be a directory, not a file")
-            return
-
-        if save_function is None:
-            if safe_serialization:
-
-                def save_function(weights, filename):
-                    return safetensors.torch.save_file(weights, filename, metadata={"format": "pt"})
-
-            else:
-                save_function = torch.save
-
-        os.makedirs(save_directory, exist_ok=True)
-
-        if weight_name is None:
-            if safe_serialization:
-                weight_name = LORA_WEIGHT_NAME_SAFE
-            else:
-                weight_name = LORA_WEIGHT_NAME
-
-        save_path = Path(save_directory, weight_name).as_posix()
-        save_function(state_dict, save_path)
-        logger.info(f"Model weights saved in {save_path}")
-
-    @property
-    def lora_scale(self) -> float:
-        # property function that returns the lora scale which can be set at run time by the pipeline.
-        # if _lora_scale has not been set, return 1
-        return self._lora_scale if hasattr(self, "_lora_scale") else 1.0

src/diffusers/loaders/peft.py

@@ -12,32 +12,15 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-import inspect
-from functools import partial
-from typing import Dict, List, Optional, Union
+from typing import List, Union
 
-from ..utils import (
-    MIN_PEFT_VERSION,
-    USE_PEFT_BACKEND,
-    check_peft_version,
-    delete_adapter_layers,
-    is_peft_available,
-    set_adapter_layers,
-    set_weights_and_activate_adapters,
-)
-from .unet_loader_utils import _maybe_expand_lora_scales
-
-
-_SET_ADAPTER_SCALE_FN_MAPPING = {
-    "UNet2DConditionModel": _maybe_expand_lora_scales,
-    "SD3Transformer2DModel": lambda model_cls, weights: weights,
-}
+from ..utils import MIN_PEFT_VERSION, check_peft_version, is_peft_available
 
 
 class PeftAdapterMixin:
     """
     A class containing all functions for loading and using adapters weights that are supported in PEFT library. For
-    more details about adapters and injecting them in a base model, check out the PEFT
+    more details about adapters and injecting them in a transformer-based model, check out the PEFT
     [documentation](https://huggingface.co/docs/peft/index).
 
     Install the latest version of PEFT, and use this mixin to:
@@ -50,62 +33,6 @@ class PeftAdapterMixin:
 
     _hf_peft_config_loaded = False
 
-    def set_adapters(
-        self,
-        adapter_names: Union[List[str], str],
-        weights: Optional[Union[float, Dict, List[float], List[Dict], List[None]]] = None,
-    ):
-        """
-        Set the currently active adapters for use in the UNet.
-
-        Args:
-            adapter_names (`List[str]` or `str`):
-                The names of the adapters to use.
-            adapter_weights (`Union[List[float], float]`, *optional*):
-                The adapter(s) weights to use with the UNet. If `None`, the weights are set to `1.0` for all the
-                adapters.
-
-        Example:
-
-        ```py
-        from diffusers import AutoPipelineForText2Image
-        import torch
-
-        pipeline = AutoPipelineForText2Image.from_pretrained(
-            "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
-        ).to("cuda")
-        pipeline.load_lora_weights(
-            "jbilcke-hf/sdxl-cinematic-1", weight_name="pytorch_lora_weights.safetensors", adapter_name="cinematic"
-        )
-        pipeline.load_lora_weights("nerijs/pixel-art-xl", weight_name="pixel-art-xl.safetensors", adapter_name="pixel")
-        pipeline.set_adapters(["cinematic", "pixel"], adapter_weights=[0.5, 0.5])
-        ```
-        """
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for `set_adapters()`.")
-
-        adapter_names = [adapter_names] if isinstance(adapter_names, str) else adapter_names
-
-        # Expand weights into a list, one entry per adapter
-        # examples for e.g. 2 adapters:  [{...}, 7] -> [7,7] ; None -> [None, None]
-        if not isinstance(weights, list):
-            weights = [weights] * len(adapter_names)
-
-        if len(adapter_names) != len(weights):
-            raise ValueError(
-                f"Length of adapter names {len(adapter_names)} is not equal to the length of their weights {len(weights)}."
-            )
-
-        # Set None values to default of 1.0
-        # e.g. [{...}, 7] -> [{...}, 7] ; [None, None] -> [1.0, 1.0]
-        weights = [w if w is not None else 1.0 for w in weights]
-
-        # e.g. [{...}, 7] -> [{expanded dict...}, 7]
-        scale_expansion_fn = _SET_ADAPTER_SCALE_FN_MAPPING[self.__class__.__name__]
-        weights = scale_expansion_fn(self, weights)
-
-        set_weights_and_activate_adapters(self, adapter_names, weights)
-
     def add_adapter(self, adapter_config, adapter_name: str = "default") -> None:
         r"""
         Adds a new adapter to the current model for training. If no adapter name is passed, a default name is assigned
@@ -139,7 +66,7 @@ class PeftAdapterMixin:
             )
 
         # Unlike transformers, here we don't need to retrieve the name_or_path of the unet as the loading logic is
-        # handled by the `load_lora_layers` or `StableDiffusionLoraLoaderMixin`. Therefore we set it to `None` here.
+        # handled by the `load_lora_layers` or `LoraLoaderMixin`. Therefore we set it to `None` here.
         adapter_config.base_model_name_or_path = None
         inject_adapter_in_model(adapter_config, self, adapter_name)
         self.set_adapter(adapter_name)
@@ -258,136 +185,3 @@ class PeftAdapterMixin:
         for _, module in self.named_modules():
             if isinstance(module, BaseTunerLayer):
                 return module.active_adapter
-
-    def fuse_lora(self, lora_scale=1.0, safe_fusing=False, adapter_names=None):
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for `fuse_lora()`.")
-
-        self.lora_scale = lora_scale
-        self._safe_fusing = safe_fusing
-        self.apply(partial(self._fuse_lora_apply, adapter_names=adapter_names))
-
-    def _fuse_lora_apply(self, module, adapter_names=None):
-        from peft.tuners.tuners_utils import BaseTunerLayer
-
-        merge_kwargs = {"safe_merge": self._safe_fusing}
-
-        if isinstance(module, BaseTunerLayer):
-            if self.lora_scale != 1.0:
-                module.scale_layer(self.lora_scale)
-
-            # For BC with prevous PEFT versions, we need to check the signature
-            # of the `merge` method to see if it supports the `adapter_names` argument.
-            supported_merge_kwargs = list(inspect.signature(module.merge).parameters)
-            if "adapter_names" in supported_merge_kwargs:
-                merge_kwargs["adapter_names"] = adapter_names
-            elif "adapter_names" not in supported_merge_kwargs and adapter_names is not None:
-                raise ValueError(
-                    "The `adapter_names` argument is not supported with your PEFT version. Please upgrade"
-                    " to the latest version of PEFT. `pip install -U peft`"
-                )
-
-            module.merge(**merge_kwargs)
-
-    def unfuse_lora(self):
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for `unfuse_lora()`.")
-        self.apply(self._unfuse_lora_apply)
-
-    def _unfuse_lora_apply(self, module):
-        from peft.tuners.tuners_utils import BaseTunerLayer
-
-        if isinstance(module, BaseTunerLayer):
-            module.unmerge()
-
-    def unload_lora(self):
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for `unload_lora()`.")
-
-        from ..utils import recurse_remove_peft_layers
-
-        recurse_remove_peft_layers(self)
-        if hasattr(self, "peft_config"):
-            del self.peft_config
-
-    def disable_lora(self):
-        """
-        Disables the active LoRA layers of the underlying model.
-
-        Example:
-
-        ```py
-        from diffusers import AutoPipelineForText2Image
-        import torch
-
-        pipeline = AutoPipelineForText2Image.from_pretrained(
-            "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
-        ).to("cuda")
-        pipeline.load_lora_weights(
-            "jbilcke-hf/sdxl-cinematic-1", weight_name="pytorch_lora_weights.safetensors", adapter_name="cinematic"
-        )
-        pipeline.disable_lora()
-        ```
-        """
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for this method.")
-        set_adapter_layers(self, enabled=False)
-
-    def enable_lora(self):
-        """
-        Enables the active LoRA layers of the underlying model.
-
-        Example:
-
-        ```py
-        from diffusers import AutoPipelineForText2Image
-        import torch
-
-        pipeline = AutoPipelineForText2Image.from_pretrained(
-            "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
-        ).to("cuda")
-        pipeline.load_lora_weights(
-            "jbilcke-hf/sdxl-cinematic-1", weight_name="pytorch_lora_weights.safetensors", adapter_name="cinematic"
-        )
-        pipeline.enable_lora()
-        ```
-        """
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for this method.")
-        set_adapter_layers(self, enabled=True)
-
-    def delete_adapters(self, adapter_names: Union[List[str], str]):
-        """
-        Delete an adapter's LoRA layers from the underlying model.
-
-        Args:
-            adapter_names (`Union[List[str], str]`):
-                The names (single string or list of strings) of the adapter to delete.
-
-        Example:
-
-        ```py
-        from diffusers import AutoPipelineForText2Image
-        import torch
-
-        pipeline = AutoPipelineForText2Image.from_pretrained(
-            "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
-        ).to("cuda")
-        pipeline.load_lora_weights(
-            "jbilcke-hf/sdxl-cinematic-1", weight_name="pytorch_lora_weights.safetensors", adapter_names="cinematic"
-        )
-        pipeline.delete_adapters("cinematic")
-        ```
-        """
-        if not USE_PEFT_BACKEND:
-            raise ValueError("PEFT backend is required for this method.")
-
-        if isinstance(adapter_names, str):
-            adapter_names = [adapter_names]
-
-        for adapter_name in adapter_names:
-            delete_adapter_layers(self, adapter_name)
-
-            # Pop also the corresponding adapter from the config
-            if hasattr(self, "peft_config"):
-                self.peft_config.pop(adapter_name, None)

src/diffusers/loaders/unet.py

@@ -11,11 +11,13 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+import inspect
 import os
 from collections import defaultdict
 from contextlib import nullcontext
+from functools import partial
 from pathlib import Path
-from typing import Callable, Dict, Union
+from typing import Callable, Dict, List, Optional, Union
 
 import safetensors
 import torch
@@ -36,14 +38,18 @@ from ..utils import (
     USE_PEFT_BACKEND,
     _get_model_file,
     convert_unet_state_dict_to_peft,
+    delete_adapter_layers,
     get_adapter_name,
     get_peft_kwargs,
     is_accelerate_available,
     is_peft_version,
     is_torch_version,
     logging,
+    set_adapter_layers,
+    set_weights_and_activate_adapters,
 )
-from .lora_pipeline import LORA_WEIGHT_NAME, LORA_WEIGHT_NAME_SAFE, TEXT_ENCODER_NAME, UNET_NAME
+from .lora import LORA_WEIGHT_NAME, LORA_WEIGHT_NAME_SAFE, TEXT_ENCODER_NAME, UNET_NAME
+from .unet_loader_utils import _maybe_expand_lora_scales
 from .utils import AttnProcsLayers
 
 
@@ -351,7 +357,7 @@ class UNet2DConditionLoadersMixin:
         return is_model_cpu_offload, is_sequential_cpu_offload
 
     @classmethod
-    # Copied from diffusers.loaders.lora_base.LoraBaseMixin._optionally_disable_offloading
+    # Copied from diffusers.loaders.lora.LoraLoaderMixin._optionally_disable_offloading
     def _optionally_disable_offloading(cls, _pipeline):
         """
         Optionally removes offloading in case the pipeline has been already sequentially offloaded to CPU.
@@ -513,6 +519,194 @@ class UNet2DConditionLoadersMixin:
 
         return state_dict
 
+    def fuse_lora(self, lora_scale=1.0, safe_fusing=False, adapter_names=None):
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for `fuse_lora()`.")
+
+        self.lora_scale = lora_scale
+        self._safe_fusing = safe_fusing
+        self.apply(partial(self._fuse_lora_apply, adapter_names=adapter_names))
+
+    def _fuse_lora_apply(self, module, adapter_names=None):
+        from peft.tuners.tuners_utils import BaseTunerLayer
+
+        merge_kwargs = {"safe_merge": self._safe_fusing}
+
+        if isinstance(module, BaseTunerLayer):
+            if self.lora_scale != 1.0:
+                module.scale_layer(self.lora_scale)
+
+            # For BC with prevous PEFT versions, we need to check the signature
+            # of the `merge` method to see if it supports the `adapter_names` argument.
+            supported_merge_kwargs = list(inspect.signature(module.merge).parameters)
+            if "adapter_names" in supported_merge_kwargs:
+                merge_kwargs["adapter_names"] = adapter_names
+            elif "adapter_names" not in supported_merge_kwargs and adapter_names is not None:
+                raise ValueError(
+                    "The `adapter_names` argument is not supported with your PEFT version. Please upgrade"
+                    " to the latest version of PEFT. `pip install -U peft`"
+                )
+
+            module.merge(**merge_kwargs)
+
+    def unfuse_lora(self):
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for `unfuse_lora()`.")
+        self.apply(self._unfuse_lora_apply)
+
+    def _unfuse_lora_apply(self, module):
+        from peft.tuners.tuners_utils import BaseTunerLayer
+
+        if isinstance(module, BaseTunerLayer):
+            module.unmerge()
+
+    def unload_lora(self):
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for `unload_lora()`.")
+
+        from ..utils import recurse_remove_peft_layers
+
+        recurse_remove_peft_layers(self)
+        if hasattr(self, "peft_config"):
+            del self.peft_config
+
+    def set_adapters(
+        self,
+        adapter_names: Union[List[str], str],
+        weights: Optional[Union[float, Dict, List[float], List[Dict], List[None]]] = None,
+    ):
+        """
+        Set the currently active adapters for use in the UNet.
+
+        Args:
+            adapter_names (`List[str]` or `str`):
+                The names of the adapters to use.
+            adapter_weights (`Union[List[float], float]`, *optional*):
+                The adapter(s) weights to use with the UNet. If `None`, the weights are set to `1.0` for all the
+                adapters.
+
+        Example:
+
+        ```py
+        from diffusers import AutoPipelineForText2Image
+        import torch
+
+        pipeline = AutoPipelineForText2Image.from_pretrained(
+            "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
+        ).to("cuda")
+        pipeline.load_lora_weights(
+            "jbilcke-hf/sdxl-cinematic-1", weight_name="pytorch_lora_weights.safetensors", adapter_name="cinematic"
+        )
+        pipeline.load_lora_weights("nerijs/pixel-art-xl", weight_name="pixel-art-xl.safetensors", adapter_name="pixel")
+        pipeline.set_adapters(["cinematic", "pixel"], adapter_weights=[0.5, 0.5])
+        ```
+        """
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for `set_adapters()`.")
+
+        adapter_names = [adapter_names] if isinstance(adapter_names, str) else adapter_names
+
+        # Expand weights into a list, one entry per adapter
+        # examples for e.g. 2 adapters:  [{...}, 7] -> [7,7] ; None -> [None, None]
+        if not isinstance(weights, list):
+            weights = [weights] * len(adapter_names)
+
+        if len(adapter_names) != len(weights):
+            raise ValueError(
+                f"Length of adapter names {len(adapter_names)} is not equal to the length of their weights {len(weights)}."
+            )
+
+        # Set None values to default of 1.0
+        # e.g. [{...}, 7] -> [{...}, 7] ; [None, None] -> [1.0, 1.0]
+        weights = [w if w is not None else 1.0 for w in weights]
+
+        # e.g. [{...}, 7] -> [{expanded dict...}, 7]
+        weights = _maybe_expand_lora_scales(self, weights)
+
+        set_weights_and_activate_adapters(self, adapter_names, weights)
+
+    def disable_lora(self):
+        """
+        Disable the UNet's active LoRA layers.
+
+        Example:
+
+        ```py
+        from diffusers import AutoPipelineForText2Image
+        import torch
+
+        pipeline = AutoPipelineForText2Image.from_pretrained(
+            "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
+        ).to("cuda")
+        pipeline.load_lora_weights(
+            "jbilcke-hf/sdxl-cinematic-1", weight_name="pytorch_lora_weights.safetensors", adapter_name="cinematic"
+        )
+        pipeline.disable_lora()
+        ```
+        """
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for this method.")
+        set_adapter_layers(self, enabled=False)
+
+    def enable_lora(self):
+        """
+        Enable the UNet's active LoRA layers.
+
+        Example:
+
+        ```py
+        from diffusers import AutoPipelineForText2Image
+        import torch
+
+        pipeline = AutoPipelineForText2Image.from_pretrained(
+            "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
+        ).to("cuda")
+        pipeline.load_lora_weights(
+            "jbilcke-hf/sdxl-cinematic-1", weight_name="pytorch_lora_weights.safetensors", adapter_name="cinematic"
+        )
+        pipeline.enable_lora()
+        ```
+        """
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for this method.")
+        set_adapter_layers(self, enabled=True)
+
+    def delete_adapters(self, adapter_names: Union[List[str], str]):
+        """
+        Delete an adapter's LoRA layers from the UNet.
+
+        Args:
+            adapter_names (`Union[List[str], str]`):
+                The names (single string or list of strings) of the adapter to delete.
+
+        Example:
+
+        ```py
+        from diffusers import AutoPipelineForText2Image
+        import torch
+
+        pipeline = AutoPipelineForText2Image.from_pretrained(
+            "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
+        ).to("cuda")
+        pipeline.load_lora_weights(
+            "jbilcke-hf/sdxl-cinematic-1", weight_name="pytorch_lora_weights.safetensors", adapter_names="cinematic"
+        )
+        pipeline.delete_adapters("cinematic")
+        ```
+        """
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for this method.")
+
+        if isinstance(adapter_names, str):
+            adapter_names = [adapter_names]
+
+        for adapter_name in adapter_names:
+            delete_adapter_layers(self, adapter_name)
+
+            # Pop also the corresponding adapter from the config
+            if hasattr(self, "peft_config"):
+                self.peft_config.pop(adapter_name, None)
+
     def _convert_ip_adapter_image_proj_to_diffusers(self, state_dict, low_cpu_mem_usage=False):
         if low_cpu_mem_usage:
             if is_accelerate_available():

src/diffusers/models/transformers/transformer_sd3.py

@@ -13,6 +13,8 @@
 # limitations under the License.
 
 
+import inspect
+from functools import partial
 from typing import Any, Dict, List, Optional, Union
 
 import torch
@@ -253,6 +255,47 @@ class SD3Transformer2DModel(ModelMixin, ConfigMixin, PeftAdapterMixin, FromOrigi
         if hasattr(module, "gradient_checkpointing"):
             module.gradient_checkpointing = value
 
+    def fuse_lora(self, lora_scale=1.0, safe_fusing=False, adapter_names=None):
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for `fuse_lora()`.")
+
+        self.lora_scale = lora_scale
+        self._safe_fusing = safe_fusing
+        self.apply(partial(self._fuse_lora_apply, adapter_names=adapter_names))
+
+    def _fuse_lora_apply(self, module, adapter_names=None):
+        from peft.tuners.tuners_utils import BaseTunerLayer
+
+        merge_kwargs = {"safe_merge": self._safe_fusing}
+
+        if isinstance(module, BaseTunerLayer):
+            if self.lora_scale != 1.0:
+                module.scale_layer(self.lora_scale)
+
+            # For BC with prevous PEFT versions, we need to check the signature
+            # of the `merge` method to see if it supports the `adapter_names` argument.
+            supported_merge_kwargs = list(inspect.signature(module.merge).parameters)
+            if "adapter_names" in supported_merge_kwargs:
+                merge_kwargs["adapter_names"] = adapter_names
+            elif "adapter_names" not in supported_merge_kwargs and adapter_names is not None:
+                raise ValueError(
+                    "The `adapter_names` argument is not supported with your PEFT version. Please upgrade"
+                    " to the latest version of PEFT. `pip install -U peft`"
+                )
+
+            module.merge(**merge_kwargs)
+
+    def unfuse_lora(self):
+        if not USE_PEFT_BACKEND:
+            raise ValueError("PEFT backend is required for `unfuse_lora()`.")
+        self.apply(self._unfuse_lora_apply)
+
+    def _unfuse_lora_apply(self, module):
+        from peft.tuners.tuners_utils import BaseTunerLayer
+
+        if isinstance(module, BaseTunerLayer):
+            module.unmerge()
+
     def forward(
         self,
         hidden_states: torch.FloatTensor,

src/diffusers/pipelines/animatediff/pipeline_animatediff.py

@@ -19,7 +19,7 @@ import torch
 from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
 
 from ...image_processor import PipelineImageInput
-from ...loaders import IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from ...loaders import IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel, UNetMotionModel
 from ...models.lora import adjust_lora_scale_text_encoder
 from ...models.unets.unet_motion_model import MotionAdapter
@@ -70,7 +70,7 @@ class AnimateDiffPipeline(
     StableDiffusionMixin,
     TextualInversionLoaderMixin,
     IPAdapterMixin,
-    StableDiffusionLoraLoaderMixin,
+    LoraLoaderMixin,
     FreeInitMixin,
 ):
     r"""
@@ -81,8 +81,8 @@ class AnimateDiffPipeline(
 
     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:
@@ -184,7 +184,7 @@ class AnimateDiffPipeline(
         """
         # 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
@@ -317,7 +317,7 @@ class AnimateDiffPipeline(
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
         if self.text_encoder is not None:
-            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)
 

src/diffusers/pipelines/animatediff/pipeline_animatediff_video2video.py

@@ -19,7 +19,7 @@ import torch
 from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
 
 from ...image_processor import PipelineImageInput
-from ...loaders import IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from ...loaders import IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel, UNetMotionModel
 from ...models.lora import adjust_lora_scale_text_encoder
 from ...models.unets.unet_motion_model import MotionAdapter
@@ -174,7 +174,7 @@ class AnimateDiffVideoToVideoPipeline(
     StableDiffusionMixin,
     TextualInversionLoaderMixin,
     IPAdapterMixin,
-    StableDiffusionLoraLoaderMixin,
+    LoraLoaderMixin,
     FreeInitMixin,
 ):
     r"""
@@ -185,8 +185,8 @@ class AnimateDiffVideoToVideoPipeline(
 
     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:
@@ -288,7 +288,7 @@ class AnimateDiffVideoToVideoPipeline(
         """
         # 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
@@ -421,7 +421,7 @@ class AnimateDiffVideoToVideoPipeline(
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
         if self.text_encoder is not None:
-            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)
 

src/diffusers/pipelines/controlnet/pipeline_controlnet.py

@@ -24,7 +24,7 @@ from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPV
 
 from ...callbacks import MultiPipelineCallbacks, PipelineCallback
 from ...image_processor import PipelineImageInput, VaeImageProcessor
-from ...loaders import FromSingleFileMixin, IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from ...loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, ControlNetModel, ImageProjection, UNet2DConditionModel
 from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
@@ -156,7 +156,7 @@ class StableDiffusionControlNetPipeline(
     DiffusionPipeline,
     StableDiffusionMixin,
     TextualInversionLoaderMixin,
-    StableDiffusionLoraLoaderMixin,
+    LoraLoaderMixin,
     IPAdapterMixin,
     FromSingleFileMixin,
 ):
@@ -168,8 +168,8 @@ class StableDiffusionControlNetPipeline(
 
     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
 
@@ -331,7 +331,7 @@ class StableDiffusionControlNetPipeline(
         """
         # 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
@@ -464,7 +464,7 @@ class StableDiffusionControlNetPipeline(
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
         if self.text_encoder is not None:
-            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)
 

src/diffusers/pipelines/controlnet/pipeline_controlnet_img2img.py

@@ -23,7 +23,7 @@ from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPV
 
 from ...callbacks import MultiPipelineCallbacks, PipelineCallback
 from ...image_processor import PipelineImageInput, VaeImageProcessor
-from ...loaders import FromSingleFileMixin, IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from ...loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, ControlNetModel, ImageProjection, UNet2DConditionModel
 from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
@@ -134,7 +134,7 @@ class StableDiffusionControlNetImg2ImgPipeline(
     DiffusionPipeline,
     StableDiffusionMixin,
     TextualInversionLoaderMixin,
-    StableDiffusionLoraLoaderMixin,
+    LoraLoaderMixin,
     IPAdapterMixin,
     FromSingleFileMixin,
 ):
@@ -146,8 +146,8 @@ class StableDiffusionControlNetImg2ImgPipeline(
 
     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
 
@@ -309,7 +309,7 @@ class StableDiffusionControlNetImg2ImgPipeline(
         """
         # 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
@@ -442,7 +442,7 @@ class StableDiffusionControlNetImg2ImgPipeline(
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
         if self.text_encoder is not None:
-            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)
 

src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint.py

@@ -25,7 +25,7 @@ from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPV
 
 from ...callbacks import MultiPipelineCallbacks, PipelineCallback
 from ...image_processor import PipelineImageInput, VaeImageProcessor
-from ...loaders import FromSingleFileMixin, IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from ...loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, ControlNetModel, ImageProjection, UNet2DConditionModel
 from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
@@ -122,7 +122,7 @@ class StableDiffusionControlNetInpaintPipeline(
     DiffusionPipeline,
     StableDiffusionMixin,
     TextualInversionLoaderMixin,
-    StableDiffusionLoraLoaderMixin,
+    LoraLoaderMixin,
     IPAdapterMixin,
     FromSingleFileMixin,
 ):
@@ -134,8 +134,8 @@ class StableDiffusionControlNetInpaintPipeline(
 
     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
 
@@ -311,7 +311,7 @@ class StableDiffusionControlNetInpaintPipeline(
         """
         # 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
@@ -444,7 +444,7 @@ class StableDiffusionControlNetInpaintPipeline(
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
         if self.text_encoder is not None:
-            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)
 

src/diffusers/pipelines/controlnet_sd3/pipeline_stable_diffusion_3_controlnet.py

@@ -30,12 +30,9 @@ from ...models.controlnet_sd3 import SD3ControlNetModel, SD3MultiControlNetModel
 from ...models.transformers import SD3Transformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
 from ...utils import (
-    USE_PEFT_BACKEND,
     is_torch_xla_available,
     logging,
     replace_example_docstring,
-    scale_lora_layers,
-    unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
@@ -349,7 +346,6 @@ class StableDiffusion3ControlNetPipeline(DiffusionPipeline, SD3LoraLoaderMixin,
         negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
         clip_skip: Optional[int] = None,
         max_sequence_length: int = 256,
-        lora_scale: Optional[float] = None,
     ):
         r"""
 
@@ -395,22 +391,9 @@ class StableDiffusion3ControlNetPipeline(DiffusionPipeline, SD3LoraLoaderMixin,
             clip_skip (`int`, *optional*):
                 Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
                 the output of the pre-final layer will be used for computing the prompt embeddings.
-            lora_scale (`float`, *optional*):
-                A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
         """
         device = device or self._execution_device
 
-        # 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, SD3LoraLoaderMixin):
-            self._lora_scale = lora_scale
-
-            # dynamically adjust the LoRA scale
-            if self.text_encoder is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder, lora_scale)
-            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
-                scale_lora_layers(self.text_encoder_2, lora_scale)
-
         prompt = [prompt] if isinstance(prompt, str) else prompt
         if prompt is not None:
             batch_size = len(prompt)
@@ -513,16 +496,6 @@ class StableDiffusion3ControlNetPipeline(DiffusionPipeline, SD3LoraLoaderMixin,
                 [negative_pooled_prompt_embed, negative_pooled_prompt_2_embed], dim=-1
             )
 
-        if self.text_encoder is not None:
-            if isinstance(self, SD3LoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder, lora_scale)
-
-        if self.text_encoder_2 is not None:
-            if isinstance(self, SD3LoraLoaderMixin) and USE_PEFT_BACKEND:
-                # Retrieve the original scale by scaling back the LoRA layers
-                unscale_lora_layers(self.text_encoder_2, lora_scale)
-
         return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
 
     def check_inputs(

src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs.py

@@ -23,7 +23,7 @@ from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer
 
 from ...callbacks import MultiPipelineCallbacks, PipelineCallback
 from ...image_processor import PipelineImageInput, VaeImageProcessor
-from ...loaders import FromSingleFileMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from ...loaders import FromSingleFileMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, ControlNetXSAdapter, UNet2DConditionModel, UNetControlNetXSModel
 from ...models.lora import adjust_lora_scale_text_encoder
 from ...schedulers import KarrasDiffusionSchedulers
@@ -90,11 +90,7 @@ EXAMPLE_DOC_STRING = """
 
 
 class StableDiffusionControlNetXSPipeline(
-    DiffusionPipeline,
-    StableDiffusionMixin,
-    TextualInversionLoaderMixin,
-    StableDiffusionLoraLoaderMixin,
-    FromSingleFileMixin,
+    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, LoraLoaderMixin, FromSingleFileMixin
 ):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion with ControlNet-XS guidance.
@@ -104,8 +100,8 @@ class StableDiffusionControlNetXSPipeline(
 
     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:
@@ -262,7 +258,7 @@ class StableDiffusionControlNetXSPipeline(
         """
         # 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
@@ -395,7 +391,7 @@ class StableDiffusionControlNetXSPipeline(
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
 
         if self.text_encoder is not None:
-            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)
 

src/diffusers/pipelines/deepfloyd_if/pipeline_if.py

@@ -7,7 +7,7 @@ from typing import Any, Callable, Dict, List, Optional, Union
 import torch
 from transformers import CLIPImageProcessor, T5EncoderModel, T5Tokenizer
 
-from ...loaders import StableDiffusionLoraLoaderMixin
+from ...loaders import LoraLoaderMixin
 from ...models import UNet2DConditionModel
 from ...schedulers import DDPMScheduler
 from ...utils import (
@@ -84,7 +84,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-class IFPipeline(DiffusionPipeline, StableDiffusionLoraLoaderMixin):
+class IFPipeline(DiffusionPipeline, LoraLoaderMixin):
     tokenizer: T5Tokenizer
     text_encoder: T5EncoderModel
 

src/diffusers/pipelines/deepfloyd_if/pipeline_if_img2img.py

@@ -9,7 +9,7 @@ import PIL.Image
 import torch
 from transformers import CLIPImageProcessor, T5EncoderModel, T5Tokenizer
 
-from ...loaders import StableDiffusionLoraLoaderMixin
+from ...loaders import LoraLoaderMixin
 from ...models import UNet2DConditionModel
 from ...schedulers import DDPMScheduler
 from ...utils import (
@@ -108,7 +108,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-class IFImg2ImgPipeline(DiffusionPipeline, StableDiffusionLoraLoaderMixin):
+class IFImg2ImgPipeline(DiffusionPipeline, LoraLoaderMixin):
     tokenizer: T5Tokenizer
     text_encoder: T5EncoderModel
 

src/diffusers/pipelines/deepfloyd_if/pipeline_if_img2img_superresolution.py

@@ -10,7 +10,7 @@ import torch
 import torch.nn.functional as F
 from transformers import CLIPImageProcessor, T5EncoderModel, T5Tokenizer
 
-from ...loaders import StableDiffusionLoraLoaderMixin
+from ...loaders import LoraLoaderMixin
 from ...models import UNet2DConditionModel
 from ...schedulers import DDPMScheduler
 from ...utils import (
@@ -111,7 +111,7 @@ EXAMPLE_DOC_STRING = """
 """
 
 
-class IFImg2ImgSuperResolutionPipeline(DiffusionPipeline, StableDiffusionLoraLoaderMixin):
+class IFImg2ImgSuperResolutionPipeline(DiffusionPipeline, LoraLoaderMixin):
     tokenizer: T5Tokenizer
     text_encoder: T5EncoderModel