Bugfix for dreambooth flux2 img2img2

Author: leisuzz

src/diffusers/loaders/lora_pipeline.py

@@ -1208,7 +1208,7 @@ def load_lora_into_text_encoder(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.StableDiffusionXLLoraLoaderMixin.save_lora_weights with unet->transformer
     def save_lora_weights(
-        cls,
+       cls,
         save_directory: Union[str, os.PathLike],
         transformer_lora_layers: Dict[str, Union[torch.nn.Module, torch.Tensor]] = None,
         text_encoder_lora_layers: Dict[str, Union[torch.nn.Module, torch.Tensor]] = None,
@@ -1256,7 +1256,7 @@ def save_lora_weights(
 
     # Copied from diffusers.loaders.lora_pipeline.StableDiffusionXLLoraLoaderMixin.fuse_lora with unet->transformer
     def fuse_lora(
-        self,
+       self,
         components: List[str] = ["transformer", "text_encoder", "text_encoder_2"],
         lora_scale: float = 1.0,
         safe_fusing: bool = False,
@@ -1276,7 +1276,7 @@ def fuse_lora(
 
     # Copied from diffusers.loaders.lora_pipeline.StableDiffusionXLLoraLoaderMixin.unfuse_lora with unet->transformer
     def unfuse_lora(self, components: List[str] = ["transformer", "text_encoder", "text_encoder_2"], **kwargs):
-        r"""
+       r"""
         See [`~loaders.StableDiffusionLoraLoaderMixin.unfuse_lora`] for more details.
         """
         super().unfuse_lora(components=components, **kwargs)
@@ -1390,7 +1390,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->AuraFlowTransformer2DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,
@@ -1893,19 +1893,6 @@ def load_lora_into_text_encoder(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.StableDiffusionLoraLoaderMixin.save_lora_weights with unet->transformer
     def save_lora_weights(
-        cls,
-        save_directory: Union[str, os.PathLike],
-        transformer_lora_layers: Dict[str, Union[torch.nn.Module, torch.Tensor]] = None,
-        text_encoder_lora_layers: Dict[str, torch.nn.Module] = None,
-        is_main_process: bool = True,
-        weight_name: str = None,
-        save_function: Callable = None,
-        safe_serialization: bool = True,
-        transformer_lora_adapter_metadata=None,
-        text_encoder_lora_adapter_metadata=None,
-    ):
-        r"""
-        Save the LoRA parameters corresponding to the UNet and text encoder.
 
         Arguments:
             save_directory (`str` or `os.PathLike`):
@@ -2291,7 +2278,7 @@ class AmusedLoraLoaderMixin(StableDiffusionLoraLoaderMixin):
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.FluxLoraLoaderMixin.load_lora_into_transformer with FluxTransformer2DModel->UVit2DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         network_alphas,
         transformer,
@@ -2542,7 +2529,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->CogVideoXTransformer3DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,
@@ -2739,7 +2726,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->MochiTransformer3DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,
@@ -2942,7 +2929,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->LTXVideoTransformer3DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,
@@ -3142,7 +3129,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->SanaTransformer2DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,
@@ -3345,7 +3332,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->HunyuanVideoTransformer3DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,
@@ -3549,7 +3536,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->Lumina2Transformer2DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,
@@ -4108,7 +4095,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->WanTransformer3DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,
@@ -4385,7 +4372,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->SkyReelsV2Transformer3DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,
@@ -4585,7 +4572,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->CogView4Transformer2DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,
@@ -4788,7 +4775,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->HiDreamImageTransformer2DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,
@@ -4994,7 +4981,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->QwenImageTransformer2DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,
@@ -5200,7 +5187,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->ZImageTransformer2DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,
@@ -5403,7 +5390,7 @@ def load_lora_weights(
     @classmethod
     # Copied from diffusers.loaders.lora_pipeline.SD3LoraLoaderMixin.load_lora_into_transformer with SD3Transformer2DModel->CogView4Transformer2DModel
     def load_lora_into_transformer(
-        cls,
+       cls,
         state_dict,
         transformer,
         adapter_name=None,

src/diffusers/models/controlnets/multicontrolnet_union.py

@@ -85,16 +85,6 @@ def forward(
 
     # Copied from diffusers.models.controlnets.multicontrolnet.MultiControlNetModel.save_pretrained with ControlNet->ControlNetUnion
     def save_pretrained(
-        self,
-        save_directory: Union[str, os.PathLike],
-        is_main_process: bool = True,
-        save_function: Callable = None,
-        safe_serialization: bool = True,
-        variant: Optional[str] = None,
-    ):
-        """
-        Save a model and its configuration file to a directory, so that it can be re-loaded using the
-        `[`~models.controlnets.multicontrolnet.MultiControlNetUnionModel.from_pretrained`]` class method.
 
         Arguments:
             save_directory (`str` or `os.PathLike`):
@@ -125,8 +115,6 @@ def save_pretrained(
     @classmethod
     # Copied from diffusers.models.controlnets.multicontrolnet.MultiControlNetModel.from_pretrained with ControlNet->ControlNetUnion
     def from_pretrained(cls, pretrained_model_path: Optional[Union[str, os.PathLike]], **kwargs):
-        r"""
-        Instantiate a pretrained MultiControlNetUnion model from multiple pre-trained controlnet models.
 
         The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated). To train
         the model, you should first set it back in training mode with `model.train()`.

src/diffusers/models/transformers/auraflow_transformer_2d.py

@@ -367,9 +367,6 @@ def __init__(
 
     # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.fuse_qkv_projections with FusedAttnProcessor2_0->FusedAuraFlowAttnProcessor2_0
     def fuse_qkv_projections(self):
-        """
-        Enables fused QKV projections. For self-attention modules, all projection matrices (i.e., query, key, value)
-        are fused. For cross-attention modules, key and value projection matrices are fused.
 
         > [!WARNING] > This API is 🧪 experimental.
         """

src/diffusers/models/transformers/cogvideox_transformer_3d.py

@@ -333,9 +333,6 @@ def __init__(
 
     # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.fuse_qkv_projections with FusedAttnProcessor2_0->FusedCogVideoXAttnProcessor2_0
     def fuse_qkv_projections(self):
-        """
-        Enables fused QKV projections. For self-attention modules, all projection matrices (i.e., query, key, value)
-        are fused. For cross-attention modules, key and value projection matrices are fused.
 
         > [!WARNING] > This API is 🧪 experimental.
         """

src/diffusers/models/transformers/hunyuan_transformer_2d.py

@@ -320,9 +320,6 @@ def __init__(
 
     # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.fuse_qkv_projections with FusedAttnProcessor2_0->FusedHunyuanAttnProcessor2_0
     def fuse_qkv_projections(self):
-        """
-        Enables fused QKV projections. For self-attention modules, all projection matrices (i.e., query, key, value)
-        are fused. For cross-attention modules, key and value projection matrices are fused.
 
         > [!WARNING] > This API is 🧪 experimental.
         """

src/diffusers/models/transformers/stable_audio_transformer.py

@@ -276,7 +276,7 @@ def __init__(
 
     # Copied from diffusers.models.transformers.hunyuan_transformer_2d.HunyuanDiT2DModel.set_default_attn_processor with Hunyuan->StableAudio
     def set_default_attn_processor(self):
-        """
+       """
         Disables custom attention processors and sets the default attention implementation.
         """
         self.set_attn_processor(StableAudioAttnProcessor2_0())

src/diffusers/models/transformers/transformer_sd3.py

@@ -215,9 +215,6 @@ def fn_recursive_feed_forward(module: torch.nn.Module, chunk_size: int, dim: int
 
     # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.fuse_qkv_projections with FusedAttnProcessor2_0->FusedJointAttnProcessor2_0
     def fuse_qkv_projections(self):
-        """
-        Enables fused QKV projections. For self-attention modules, all projection matrices (i.e., query, key, value)
-        are fused. For cross-attention modules, key and value projection matrices are fused.
 
         > [!WARNING] > This API is 🧪 experimental.
         """

src/diffusers/modular_pipelines/flux/before_denoise.py

@@ -273,7 +273,7 @@ def intermediate_outputs(self) -> List[OutputParam]:
     @staticmethod
     # Copied from diffusers.pipelines.stable_diffusion_3.pipeline_stable_diffusion_3_img2img.StableDiffusion3Img2ImgPipeline.get_timesteps with self.scheduler->scheduler
     def get_timesteps(scheduler, num_inference_steps, strength, device):
-        # get the original timestep using init_timestep
+       # get the original timestep using init_timestep
         init_timestep = min(num_inference_steps * strength, num_inference_steps)
 
         t_start = int(max(num_inference_steps - init_timestep, 0))

src/diffusers/modular_pipelines/stable_diffusion_xl/before_denoise.py

@@ -443,7 +443,7 @@ def intermediate_outputs(self) -> List[str]:
     @staticmethod
     # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_img2img.StableDiffusionXLImg2ImgPipeline.get_timesteps with self->components
     def get_timesteps(components, num_inference_steps, strength, device, denoising_start=None):
-        # get the original timestep using init_timestep
+       # get the original timestep using init_timestep
         if denoising_start is None:
             init_timestep = min(int(num_inference_steps * strength), num_inference_steps)
             t_start = max(num_inference_steps - init_timestep, 0)
@@ -1005,7 +1005,7 @@ def check_inputs(components, block_state):
     @staticmethod
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents with self->comp
     def prepare_latents(comp, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
-        shape = (
+       shape = (
             batch_size,
             num_channels_latents,
             int(height) // comp.vae_scale_factor,
@@ -1135,7 +1135,7 @@ def intermediate_outputs(self) -> List[OutputParam]:
     @staticmethod
     # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl_img2img.StableDiffusionXLImg2ImgPipeline._get_add_time_ids with self->components
     def _get_add_time_ids(
-        components,
+       components,
         original_size,
         crops_coords_top_left,
         target_size,
@@ -1346,7 +1346,7 @@ def intermediate_outputs(self) -> List[OutputParam]:
     @staticmethod
     # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline._get_add_time_ids with self->components
     def _get_add_time_ids(
-        components, original_size, crops_coords_top_left, target_size, dtype, text_encoder_projection_dim=None
+       components, original_size, crops_coords_top_left, target_size, dtype, text_encoder_projection_dim=None
     ):
         add_time_ids = list(original_size + crops_coords_top_left + target_size)
 

src/diffusers/modular_pipelines/stable_diffusion_xl/decoders.py

@@ -76,7 +76,7 @@ def intermediate_outputs(self) -> List[str]:
     @staticmethod
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_upscale.StableDiffusionUpscalePipeline.upcast_vae with self->components
     def upcast_vae(components):
-        deprecate(
+       deprecate(
             "upcast_vae",
             "1.0.0",
             "`upcast_vae` is deprecated. Please use `pipe.vae.to(torch.float32)`. For more details, please refer to: https://github.com/huggingface/diffusers/pull/12619#issue-3606633695.",