Merge branch 'main' into qwen-pipeline-mixin

Author: sayakpaul

docs/source/en/api/pipelines/qwenimage.md

@@ -26,6 +26,7 @@ Qwen-Image comes in the following variants:
 |:----------:|:--------:|
 | Qwen-Image | [`Qwen/Qwen-Image`](https://huggingface.co/Qwen/Qwen-Image) |
 | Qwen-Image-Edit | [`Qwen/Qwen-Image-Edit`](https://huggingface.co/Qwen/Qwen-Image-Edit) |
+| Qwen-Image-Edit Plus | [Qwen/Qwen-Image-Edit-2509](https://huggingface.co/Qwen/Qwen-Image-Edit-2509) |
 
 <Tip>
 
@@ -96,6 +97,29 @@ The `guidance_scale` parameter in the pipeline is there to support future guidan
 
 </Tip>
 
+## Multi-image reference with QwenImageEditPlusPipeline
+
+With [`QwenImageEditPlusPipeline`], one can provide multiple images as input reference.
+
+```
+import torch
+from PIL import Image
+from diffusers import QwenImageEditPlusPipeline
+from diffusers.utils import load_image
+
+pipe = QwenImageEditPlusPipeline.from_pretrained(
+    "Qwen/Qwen-Image-Edit-2509", torch_dtype=torch.bfloat16
+).to("cuda")
+
+image_1 = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/grumpy.jpg")
+image_2 = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/peng.png")
+image = pipe(
+    image=[image_1, image_2], 
+    prompt="put the penguin and the cat at a game show called "Qwen Edit Plus Games"", 
+    num_inference_steps=50
+).images[0]
+```
+
 ## QwenImagePipeline
 
 [[autodoc]] QwenImagePipeline
@@ -126,7 +150,15 @@ The `guidance_scale` parameter in the pipeline is there to support future guidan
   - all
   - __call__
 
-## QwenImaggeControlNetPipeline
+## QwenImageControlNetPipeline
+
+[[autodoc]] QwenImageControlNetPipeline
+  - all
+  - __call__
+
+## QwenImageEditPlusPipeline
+
+[[autodoc]] QwenImageEditPlusPipeline
   - all
   - __call__
 

src/diffusers/loaders/lora_base.py

@@ -1064,6 +1064,41 @@ def save_function(weights, filename):
         save_function(state_dict, save_path)
         logger.info(f"Model weights saved in {save_path}")
 
+    @classmethod
+    def _save_lora_weights(
+        cls,
+        save_directory: Union[str, os.PathLike],
+        lora_layers: Dict[str, Dict[str, Union[torch.nn.Module, torch.Tensor]]],
+        lora_metadata: Dict[str, Optional[dict]],
+        is_main_process: bool = True,
+        weight_name: str = None,
+        save_function: Callable = None,
+        safe_serialization: bool = True,
+    ):
+        """
+        Helper method to pack and save LoRA weights and metadata. This method centralizes the saving logic for all
+        pipeline types.
+        """
+        state_dict = {}
+        final_lora_adapter_metadata = {}
+
+        for prefix, layers in lora_layers.items():
+            state_dict.update(cls.pack_weights(layers, prefix))
+
+        for prefix, metadata in lora_metadata.items():
+            if metadata:
+                final_lora_adapter_metadata.update(_pack_dict_with_prefix(metadata, prefix))
+
+        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,
+            lora_adapter_metadata=final_lora_adapter_metadata if final_lora_adapter_metadata else None,
+        )
+
     @classmethod
     def _optionally_disable_offloading(cls, _pipeline):
         return _func_optionally_disable_offloading(_pipeline=_pipeline)