resolve conflicts

This reverts commit 65efbcead5.

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

@@ -120,12 +120,6 @@ The `guidance_scale` parameter in the pipeline is there to support future guidan
   - all
   - __call__
 
-## QwenImageEditInpaintPipeline
-
-[[autodoc]] QwenImageEditInpaintPipeline
-  - all
-  - __call__
-
 ## QwenImaggeControlNetPipeline
   - all
   - __call__

docs/source/en/modular_diffusers/components_manager.md

@@ -51,10 +51,10 @@ t2i_pipeline = t2i_blocks.init_pipeline(modular_repo_id, components_manager=comp
 </hfoption>
 </hfoptions>
 
-Components are only loaded and registered when using [`~ModularPipeline.load_components`] or [`~ModularPipeline.load_components`]. The example below uses [`~ModularPipeline.load_components`] to create a second pipeline that reuses all the components from the first one, and assigns it to a different collection
+Components are only loaded and registered when using [`~ModularPipeline.load_components`] or [`~ModularPipeline.load_default_components`]. The example below uses [`~ModularPipeline.load_default_components`] to create a second pipeline that reuses all the components from the first one, and assigns it to a different collection
 
 ```py
-pipe.load_components()
+pipe.load_default_components()
 pipe2 = ModularPipeline.from_pretrained("YiYiXu/modular-demo-auto", components_manager=comp, collection="test2")
 ```
 
@@ -187,4 +187,4 @@ comp.enable_auto_cpu_offload(device="cuda")
 
 All models begin on the CPU and [`ComponentsManager`] moves them to the appropriate device right before they're needed, and moves other models back to the CPU when GPU memory is low.
 
-You can set your own rules for which models to offload first.
+You can set your own rules for which models to offload first.

docs/source/en/modular_diffusers/guiders.md

@@ -75,13 +75,13 @@ Guiders that are already saved on the Hub with a `modular_model_index.json` file
 }
 ```
 
-The guider is only created after calling [`~ModularPipeline.load_components`] based on the loading specification in `modular_model_index.json`.
+The guider is only created after calling [`~ModularPipeline.load_default_components`] based on the loading specification in `modular_model_index.json`.
 
 ```py
 t2i_pipeline = t2i_blocks.init_pipeline("YiYiXu/modular-doc-guider")
 # not created during init
 assert t2i_pipeline.guider is None
-t2i_pipeline.load_components()
+t2i_pipeline.load_default_components()
 # loaded as PAG guider
 t2i_pipeline.guider
 ```
@@ -172,4 +172,4 @@ t2i_pipeline.push_to_hub("YiYiXu/modular-doc-guider")
 ```
 
 </hfoption>
-</hfoptions>
+</hfoptions>

docs/source/en/modular_diffusers/modular_pipeline.md

@@ -29,7 +29,7 @@ blocks = SequentialPipelineBlocks.from_blocks_dict(TEXT2IMAGE_BLOCKS)
 modular_repo_id = "YiYiXu/modular-loader-t2i-0704"
 pipeline = blocks.init_pipeline(modular_repo_id)
 
-pipeline.load_components(torch_dtype=torch.float16)
+pipeline.load_default_components(torch_dtype=torch.float16)
 pipeline.to("cuda")
 
 image = pipeline(prompt="Astronaut in a jungle, cold color palette, muted colors, detailed, 8k", output="images")[0]
@@ -49,7 +49,7 @@ blocks = SequentialPipelineBlocks.from_blocks_dict(IMAGE2IMAGE_BLOCKS)
 modular_repo_id = "YiYiXu/modular-loader-t2i-0704"
 pipeline = blocks.init_pipeline(modular_repo_id)
 
-pipeline.load_components(torch_dtype=torch.float16)
+pipeline.load_default_components(torch_dtype=torch.float16)
 pipeline.to("cuda")
 
 url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/sdxl-text2img.png"
@@ -73,7 +73,7 @@ blocks = SequentialPipelineBlocks.from_blocks_dict(INPAINT_BLOCKS)
 modular_repo_id = "YiYiXu/modular-loader-t2i-0704"
 pipeline = blocks.init_pipeline(modular_repo_id)
 
-pipeline.load_components(torch_dtype=torch.float16)
+pipeline.load_default_components(torch_dtype=torch.float16)
 pipeline.to("cuda")
 
 img_url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/sdxl-text2img.png"
@@ -176,15 +176,15 @@ diffdiff_pipeline = ModularPipeline.from_pretrained(modular_repo_id, trust_remot
 
 ## Loading components
 
-A [`ModularPipeline`] doesn't automatically instantiate with components. It only loads the configuration and component specifications. You can load all components with [`~ModularPipeline.load_components`] or only load specific components with [`~ModularPipeline.load_components`].
+A [`ModularPipeline`] doesn't automatically instantiate with components. It only loads the configuration and component specifications. You can load all components with [`~ModularPipeline.load_default_components`] or only load specific components with [`~ModularPipeline.load_components`].
 
 <hfoptions id="load">
-<hfoption id="load_components">
+<hfoption id="load_default_components">
 
 ```py
 import torch
 
-t2i_pipeline.load_components(torch_dtype=torch.float16)
+t2i_pipeline.load_default_components(torch_dtype=torch.float16)
 t2i_pipeline.to("cuda")
 ```
 
@@ -355,4 +355,4 @@ The [config.json](https://huggingface.co/YiYiXu/modular-diffdiff-0704/blob/main/
     "ModularPipelineBlocks": "block.DiffDiffBlocks"
   }
 }
-```
+```

docs/source/en/modular_diffusers/quickstart.md

@@ -173,9 +173,9 @@ print(dd_blocks)
 
 ## ModularPipeline
 
-Convert the [`SequentialPipelineBlocks`] into a [`ModularPipeline`] with the [`ModularPipeline.init_pipeline`] method. This initializes the expected components to load from a `modular_model_index.json` file. Explicitly load the components by calling [`ModularPipeline.load_components`].
+Convert the [`SequentialPipelineBlocks`] into a [`ModularPipeline`] with the [`ModularPipeline.init_pipeline`] method. This initializes the expected components to load from a `modular_model_index.json` file. Explicitly load the components by calling [`ModularPipeline.load_default_components`].
 
-It is a good idea to initialize the [`ComponentManager`] with the pipeline to help manage the different components. Once you call [`~ModularPipeline.load_components`], the components are registered to the [`ComponentManager`] and can be shared between workflows. The example below uses the `collection` argument to assign the components a `"diffdiff"` label for better organization.
+It is a good idea to initialize the [`ComponentManager`] with the pipeline to help manage the different components. Once you call [`~ModularPipeline.load_default_components`], the components are registered to the [`ComponentManager`] and can be shared between workflows. The example below uses the `collection` argument to assign the components a `"diffdiff"` label for better organization.
 
 ```py
 from diffusers.modular_pipelines import ComponentsManager
@@ -209,11 +209,11 @@ Use the [`sub_blocks.insert`] method to insert it into the [`ModularPipeline`].
 dd_blocks.sub_blocks.insert("ip_adapter", ip_adapter_block, 0)
 ```
 
-Call [`~ModularPipeline.init_pipeline`] to initialize a [`ModularPipeline`] and use [`~ModularPipeline.load_components`] to load the model components. Load and set the IP-Adapter to run the pipeline.
+Call [`~ModularPipeline.init_pipeline`] to initialize a [`ModularPipeline`] and use [`~ModularPipeline.load_default_components`] to load the model components. Load and set the IP-Adapter to run the pipeline.
 
 ```py
 dd_pipeline = dd_blocks.init_pipeline("YiYiXu/modular-demo-auto", collection="diffdiff")
-dd_pipeline.load_components(torch_dtype=torch.float16)
+dd_pipeline.load_default_components(torch_dtype=torch.float16)
 dd_pipeline.loader.load_ip_adapter("h94/IP-Adapter", subfolder="sdxl_models", weight_name="ip-adapter_sdxl.bin")
 dd_pipeline.loader.set_ip_adapter_scale(0.6)
 dd_pipeline = dd_pipeline.to(device)
@@ -260,14 +260,14 @@ class SDXLDiffDiffControlNetDenoiseStep(StableDiffusionXLDenoiseLoopWrapper):
 controlnet_denoise_block = SDXLDiffDiffControlNetDenoiseStep()
 ```
 
-Insert the `controlnet_input` block and replace the `denoise` block with the new `controlnet_denoise_block`. Initialize a [`ModularPipeline`] and [`~ModularPipeline.load_components`] into it.
+Insert the `controlnet_input` block and replace the `denoise` block with the new `controlnet_denoise_block`. Initialize a [`ModularPipeline`] and [`~ModularPipeline.load_default_components`] into it.
 
 ```py
 dd_blocks.sub_blocks.insert("controlnet_input", control_input_block, 7)
 dd_blocks.sub_blocks["denoise"] = controlnet_denoise_block
 
 dd_pipeline = dd_blocks.init_pipeline("YiYiXu/modular-demo-auto", collection="diffdiff")
-dd_pipeline.load_components(torch_dtype=torch.float16)
+dd_pipeline.load_default_components(torch_dtype=torch.float16)
 dd_pipeline = dd_pipeline.to(device)
 
 control_image = load_image("https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/diffdiff_tomato_canny.jpeg")
@@ -320,7 +320,7 @@ Call [`SequentialPipelineBlocks.from_blocks_dict`] to create a [`SequentialPipel
 ```py
 dd_auto_blocks = SequentialPipelineBlocks.from_blocks_dict(DIFFDIFF_AUTO_BLOCKS)
 dd_pipeline = dd_auto_blocks.init_pipeline("YiYiXu/modular-demo-auto", collection="diffdiff")
-dd_pipeline.load_components(torch_dtype=torch.float16)
+dd_pipeline.load_default_components(torch_dtype=torch.float16)
 ```
 
 ## Share
@@ -340,5 +340,5 @@ from diffusers.modular_pipelines import ModularPipeline, ComponentsManager
 components = ComponentsManager()
 
 diffdiff_pipeline = ModularPipeline.from_pretrained("YiYiXu/modular-diffdiff-0704", trust_remote_code=True, components_manager=components, collection="diffdiff")
-diffdiff_pipeline.load_components(torch_dtype=torch.float16)
-```
+diffdiff_pipeline.load_default_components(torch_dtype=torch.float16)
+```

docs/source/en/training/distributed_inference.md

@@ -223,7 +223,7 @@ from diffusers.image_processor import VaeImageProcessor
 import torch 
 
 vae = AutoencoderKL.from_pretrained(ckpt_id, subfolder="vae", torch_dtype=torch.bfloat16).to("cuda")
-vae_scale_factor = 2 ** (len(vae.config.block_out_channels) - 1)
+vae_scale_factor = 2 ** (len(vae.config.block_out_channels))
 image_processor = VaeImageProcessor(vae_scale_factor=vae_scale_factor)
 
 with torch.no_grad():

docs/source/zh/modular_diffusers/components_manager.md

@@ -48,10 +48,10 @@ t2i_pipeline = t2i_blocks.init_pipeline(modular_repo_id, components_manager=comp
 </hfoption>
 </hfoptions>
 
-组件仅在调用 [`~ModularPipeline.load_components`] 或 [`~ModularPipeline.load_components`] 时加载和注册。以下示例使用 [`~ModularPipeline.load_components`] 创建第二个管道，重用第一个管道的所有组件，并将其分配到不同的集合。
+组件仅在调用 [`~ModularPipeline.load_components`] 或 [`~ModularPipeline.load_default_components`] 时加载和注册。以下示例使用 [`~ModularPipeline.load_default_components`] 创建第二个管道，重用第一个管道的所有组件，并将其分配到不同的集合。
 
 ```py
-pipe.load_components()
+pipe.load_default_components()
 pipe2 = ModularPipeline.from_pretrained("YiYiXu/modular-demo-auto", components_manager=comp, collection="test2")
 ```
 
@@ -185,4 +185,4 @@ comp.enable_auto_cpu_offload(device="cuda")
 
 所有模型开始时都在 CPU 上，[`ComponentsManager`] 在需要它们之前将它们移动到适当的设备，并在 GPU 内存不足时将其他模型移回 CPU。
 
-您可以设置自己的规则来决定哪些模型要卸载。
+您可以设置自己的规则来决定哪些模型要卸载。

docs/source/zh/modular_diffusers/guiders.md

@@ -73,13 +73,13 @@ ComponentSpec(name='guider', type_hint=<class 'diffusers.guiders.perturbed_atten
 }
 ```
 
-引导器只有在调用 [`~ModularPipeline.load_components`] 之后才会创建，基于 `modular_model_index.json` 中的加载规范。
+引导器只有在调用 [`~ModularPipeline.load_default_components`] 之后才会创建，基于 `modular_model_index.json` 中的加载规范。
 
 ```py
 t2i_pipeline = t2i_blocks.init_pipeline("YiYiXu/modular-doc-guider")
 # 在初始化时未创建
 assert t2i_pipeline.guider is None
-t2i_pipeline.load_components()
+t2i_pipeline.load_default_components()
 # 加载为 PAG 引导器
 t2i_pipeline.guider
 ```
@@ -170,4 +170,4 @@ t2i_pipeline.push_to_hub("YiYiXu/modular-doc-guider")
 ```
 
 </hfoption>
-</hfoptions>
+</hfoptions>

docs/source/zh/modular_diffusers/modular_pipeline.md

@@ -28,7 +28,7 @@ blocks = SequentialPipelineBlocks.from_blocks_dict(TEXT2IMAGE_BLOCKS)
 modular_repo_id = "YiYiXu/modular-loader-t2i-0704"
 pipeline = blocks.init_pipeline(modular_repo_id)
 
-pipeline.load_components(torch_dtype=torch.float16)
+pipeline.load_default_components(torch_dtype=torch.float16)
 pipeline.to("cuda")
 
 image = pipeline(prompt="Astronaut in a jungle, cold color palette, muted colors, detailed, 8k", output="images")[0]
@@ -48,7 +48,7 @@ blocks = SequentialPipelineBlocks.from_blocks_dict(IMAGE2IMAGE_BLOCKS)
 modular_repo_id = "YiYiXu/modular-loader-t2i-0704"
 pipeline = blocks.init_pipeline(modular_repo_id)
 
-pipeline.load_components(torch_dtype=torch.float16)
+pipeline.load_default_components(torch_dtype=torch.float16)
 pipeline.to("cuda")
 
 url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/sdxl-text2img.png"
@@ -72,7 +72,7 @@ blocks = SequentialPipelineBlocks.from_blocks_dict(INPAINT_BLOCKS)
 modular_repo_id = "YiYiXu/modular-loader-t2i-0704"
 pipeline = blocks.init_pipeline(modular_repo_id)
 
-pipeline.load_components(torch_dtype=torch.float16)
+pipeline.load_default_components(torch_dtype=torch.float16)
 pipeline.to("cuda")
 
 img_url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/sdxl-text2img.png"
@@ -176,15 +176,15 @@ diffdiff_pipeline = ModularPipeline.from_pretrained(modular_repo_id, trust_remot
 
 ## 加载组件
 
-一个[`ModularPipeline`]不会自动实例化组件。它只加载配置和组件规范。您可以使用[`~ModularPipeline.load_components`]加载所有组件，或仅使用[`~ModularPipeline.load_components`]加载特定组件。
+一个[`ModularPipeline`]不会自动实例化组件。它只加载配置和组件规范。您可以使用[`~ModularPipeline.load_default_components`]加载所有组件，或仅使用[`~ModularPipeline.load_components`]加载特定组件。
 
 <hfoptions id="load">
-<hfoption id="load_components">
+<hfoption id="load_default_components">
 
 ```py
 import torch
 
-t2i_pipeline.load_components(torch_dtype=torch.float16)
+t2i_pipeline.load_default_components(torch_dtype=torch.float16)
 t2i_pipeline.to("cuda")
 ```
 

docs/source/zh/modular_diffusers/quickstart.md

@@ -175,7 +175,7 @@ print(dd_blocks)
 将 [`SequentialPipelineBlocks`] 转换为 [`ModularPipeline`]，使用 [`ModularPipeline.init_pipeline`] 方法。这会初始化从 `modular_model_index.json` 文件加载的预期组件。通过调用 [`ModularPipeline.load_defau
 lt_components`]。
 
-初始化[`ComponentManager`]时传入pipeline是一个好主意，以帮助管理不同的组件。一旦调用[`~ModularPipeline.load_components`]，组件就会被注册到[`ComponentManager`]中，并且可以在工作流之间共享。下面的例子使用`collection`参数为组件分配了一个`"diffdiff"`标签，以便更好地组织。
+初始化[`ComponentManager`]时传入pipeline是一个好主意，以帮助管理不同的组件。一旦调用[`~ModularPipeline.load_default_components`]，组件就会被注册到[`ComponentManager`]中，并且可以在工作流之间共享。下面的例子使用`collection`参数为组件分配了一个`"diffdiff"`标签，以便更好地组织。
 
 ```py
 from diffusers.modular_pipelines import ComponentsManager
@@ -209,11 +209,11 @@ ip_adapter_block = StableDiffusionXLAutoIPAdapterStep()
 dd_blocks.sub_blocks.insert("ip_adapter", ip_adapter_block, 0)
 ```
 
-调用[`~ModularPipeline.init_pipeline`]来初始化一个[`ModularPipeline`]，并使用[`~ModularPipeline.load_components`]加载模型组件。加载并设置IP-Adapter以运行pipeline。
+调用[`~ModularPipeline.init_pipeline`]来初始化一个[`ModularPipeline`]，并使用[`~ModularPipeline.load_default_components`]加载模型组件。加载并设置IP-Adapter以运行pipeline。
 
 ```py
 dd_pipeline = dd_blocks.init_pipeline("YiYiXu/modular-demo-auto", collection="diffdiff")
-dd_pipeline.load_components(torch_dtype=torch.float16)
+dd_pipeline.load_default_components(torch_dtype=torch.float16)
 dd_pipeline.loader.load_ip_adapter("h94/IP-Adapter", subfolder="sdxl_models", weight_name="ip-adapter_sdxl.bin")
 dd_pipeline.loader.set_ip_adapter_scale(0.6)
 dd_pipeline = dd_pipeline.to(device)
@@ -261,14 +261,14 @@ class SDXLDiffDiffControlNetDenoiseStep(StableDiffusionXLDenoiseLoopWrapper):
 controlnet_denoise_block = SDXLDiffDiffControlNetDenoiseStep()
 ```
 
-插入 `controlnet_input` 块并用新的 `controlnet_denoise_block` 替换 `denoise` 块。初始化一个 [`ModularPipeline`] 并将 [`~ModularPipeline.load_components`] 加载到其中。
+插入 `controlnet_input` 块并用新的 `controlnet_denoise_block` 替换 `denoise` 块。初始化一个 [`ModularPipeline`] 并将 [`~ModularPipeline.load_default_components`] 加载到其中。
 
 ```py
 dd_blocks.sub_blocks.insert("controlnet_input", control_input_block, 7)
 dd_blocks.sub_blocks["denoise"] = controlnet_denoise_block
 
 dd_pipeline = dd_blocks.init_pipeline("YiYiXu/modular-demo-auto", collection="diffdiff")
-dd_pipeline.load_components(torch_dtype=torch.float16)
+dd_pipeline.load_default_components(torch_dtype=torch.float16)
 dd_pipeline = dd_pipeline.to(device)
 
 control_image = load_image("https://huggingface.co/datasets/YiYiXu/testing-images/resolve/main/diffdiff_tomato_canny.jpeg")
@@ -322,7 +322,7 @@ DIFFDIFF_AUTO_BLOCKS.insert("controlnet_input",StableDiffusionXLControlNetAutoIn
 ```py
 dd_auto_blocks = SequentialPipelineBlocks.from_blocks_dict(DIFFDIFF_AUTO_BLOCKS)
 dd_pipeline = dd_auto_blocks.init_pipeline("YiYiXu/modular-demo-auto", collection="diffdiff")
-dd_pipeline.load_components(torch_dtype=torch.float16)
+dd_pipeline.load_default_components(torch_dtype=torch.float16)
 ```
 
 ## 分享
@@ -342,5 +342,5 @@ from diffusers.modular_pipelines import ModularPipeline, ComponentsManager
 components = ComponentsManager()
 
 diffdiff_pipeline = ModularPipeline.from_pretrained("YiYiXu/modular-diffdiff-0704", trust_remote_code=True, components_manager=components, collection="diffdiff")
-diffdiff_pipeline.load_components(torch_dtype=torch.float16)
+diffdiff_pipeline.load_default_components(torch_dtype=torch.float16)
 ```

docs/source/zh/training/distributed_inference.md

@@ -223,7 +223,7 @@ from diffusers.image_processor import VaeImageProcessor
 import torch 
 
 vae = AutoencoderKL.from_pretrained(ckpt_id, subfolder="vae", torch_dtype=torch.bfloat16).to("cuda")
-vae_scale_factor = 2 ** (len(vae.config.block_out_channels) - 1)
+vae_scale_factor = 2 ** (len(vae.config.block_out_channels))
 image_processor = VaeImageProcessor(vae_scale_factor=vae_scale_factor)
 
 with torch.no_grad():

examples/dreambooth/train_dreambooth_lora_flux_kontext.py

@@ -1270,7 +1270,6 @@ def main(args):
                 subfolder="transformer",
                 revision=args.revision,
                 variant=args.variant,
-                torch_dtype=torch_dtype,
             )
             pipeline = FluxKontextPipeline.from_pretrained(
                 args.pretrained_model_name_or_path,
@@ -1293,8 +1292,7 @@ def main(args):
             for example in tqdm(
                 sample_dataloader, desc="Generating class images", disable=not accelerator.is_local_main_process
             ):
-                with torch.autocast(device_type=accelerator.device.type, dtype=torch_dtype):
-                    images = pipeline(prompt=example["prompt"]).images
+                images = pipeline(example["prompt"]).images
 
                 for i, image in enumerate(images):
                     hash_image = insecure_hashlib.sha1(image.tobytes()).hexdigest()
@@ -1901,10 +1899,6 @@ def main(args):
                             device=accelerator.device,
                             prompt=args.instance_prompt,
                         )
-                    else:
-                        prompt_embeds, pooled_prompt_embeds, text_ids = compute_text_embeddings(
-                            prompts, text_encoders, tokenizers
-                        )
 
                 # Convert images to latent space
                 if args.cache_latents:

src/diffusers/__init__.py

@@ -494,7 +494,6 @@ else:
             "PixArtSigmaPAGPipeline",
             "PixArtSigmaPipeline",
             "QwenImageControlNetPipeline",
-            "QwenImageEditInpaintPipeline",
             "QwenImageEditPipeline",
             "QwenImageImg2ImgPipeline",
             "QwenImageInpaintPipeline",
@@ -1135,7 +1134,6 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             PixArtSigmaPAGPipeline,
             PixArtSigmaPipeline,
             QwenImageControlNetPipeline,
-            QwenImageEditInpaintPipeline,
             QwenImageEditPipeline,
             QwenImageImg2ImgPipeline,
             QwenImageInpaintPipeline,

src/diffusers/guiders/auto_guidance.py

@@ -82,15 +82,15 @@ class AutoGuidance(BaseGuidance):
         self.guidance_rescale = guidance_rescale
         self.use_original_formulation = use_original_formulation
 
-        is_layer_or_config_provided = auto_guidance_layers is not None or auto_guidance_config is not None
-        is_layer_and_config_provided = auto_guidance_layers is not None and auto_guidance_config is not None
-        if not is_layer_or_config_provided:
+        if auto_guidance_layers is None and auto_guidance_config is None:
             raise ValueError(
-                "Either `auto_guidance_layers` or `auto_guidance_config` must be provided to enable AutoGuidance."
+                "Either `auto_guidance_layers` or `auto_guidance_config` must be provided to enable Skip Layer Guidance."
             )
-        if is_layer_and_config_provided:
+        if auto_guidance_layers is not None and auto_guidance_config is not None:
             raise ValueError("Only one of `auto_guidance_layers` or `auto_guidance_config` can be provided.")
-        if auto_guidance_config is None and dropout is None:
+        if (dropout is None and auto_guidance_layers is not None) or (
+            dropout is not None and auto_guidance_layers is None
+        ):
             raise ValueError("`dropout` must be provided if `auto_guidance_layers` is provided.")
 
         if auto_guidance_layers is not None:

src/diffusers/loaders/lora_conversion_utils.py

@@ -2129,10 +2129,6 @@ def _convert_non_diffusers_ltxv_lora_to_diffusers(state_dict, non_diffusers_pref
 
 
 def _convert_non_diffusers_qwen_lora_to_diffusers(state_dict):
-    has_diffusion_model = any(k.startswith("diffusion_model.") for k in state_dict)
-    if has_diffusion_model:
-        state_dict = {k.removeprefix("diffusion_model."): v for k, v in state_dict.items()}
-
     has_lora_unet = any(k.startswith("lora_unet_") for k in state_dict)
     if has_lora_unet:
         state_dict = {k.removeprefix("lora_unet_"): v for k, v in state_dict.items()}
@@ -2205,44 +2201,29 @@ def _convert_non_diffusers_qwen_lora_to_diffusers(state_dict):
     all_keys = list(state_dict.keys())
     down_key = ".lora_down.weight"
     up_key = ".lora_up.weight"
-    a_key = ".lora_A.weight"
-    b_key = ".lora_B.weight"
 
-    has_non_diffusers_lora_id = any(down_key in k or up_key in k for k in all_keys)
-    has_diffusers_lora_id = any(a_key in k or b_key in k for k in all_keys)
+    def get_alpha_scales(down_weight, alpha_key):
+        rank = down_weight.shape[0]
+        alpha = state_dict.pop(alpha_key).item()
+        scale = alpha / rank  # LoRA is scaled by 'alpha / rank' in forward pass, so we need to scale it back here
+        scale_down = scale
+        scale_up = 1.0
+        while scale_down * 2 < scale_up:
+            scale_down *= 2
+            scale_up /= 2
+        return scale_down, scale_up
 
-    if has_non_diffusers_lora_id:
+    for k in all_keys:
+        if k.endswith(down_key):
+            diffusers_down_key = k.replace(down_key, ".lora_A.weight")
+            diffusers_up_key = k.replace(down_key, up_key).replace(up_key, ".lora_B.weight")
+            alpha_key = k.replace(down_key, ".alpha")
 
-        def get_alpha_scales(down_weight, alpha_key):
-            rank = down_weight.shape[0]
-            alpha = state_dict.pop(alpha_key).item()
-            scale = alpha / rank  # LoRA is scaled by 'alpha / rank' in forward pass, so we need to scale it back here
-            scale_down = scale
-            scale_up = 1.0
-            while scale_down * 2 < scale_up:
-                scale_down *= 2
-                scale_up /= 2
-            return scale_down, scale_up
-
-        for k in all_keys:
-            if k.endswith(down_key):
-                diffusers_down_key = k.replace(down_key, ".lora_A.weight")
-                diffusers_up_key = k.replace(down_key, up_key).replace(up_key, ".lora_B.weight")
-                alpha_key = k.replace(down_key, ".alpha")
-
-                down_weight = state_dict.pop(k)
-                up_weight = state_dict.pop(k.replace(down_key, up_key))
-                scale_down, scale_up = get_alpha_scales(down_weight, alpha_key)
-                converted_state_dict[diffusers_down_key] = down_weight * scale_down
-                converted_state_dict[diffusers_up_key] = up_weight * scale_up
-
-    # Already in diffusers format (lora_A/lora_B), just pop
-    elif has_diffusers_lora_id:
-        for k in all_keys:
-            if a_key in k or b_key in k:
-                converted_state_dict[k] = state_dict.pop(k)
-            elif ".alpha" in k:
-                state_dict.pop(k)
+            down_weight = state_dict.pop(k)
+            up_weight = state_dict.pop(k.replace(down_key, up_key))
+            scale_down, scale_up = get_alpha_scales(down_weight, alpha_key)
+            converted_state_dict[diffusers_down_key] = down_weight * scale_down
+            converted_state_dict[diffusers_up_key] = up_weight * scale_up
 
     if len(state_dict) > 0:
         raise ValueError(f"`state_dict` should be empty at this point but has {state_dict.keys()=}")

src/diffusers/loaders/lora_pipeline.py

@@ -6684,8 +6684,7 @@ class QwenImageLoraLoaderMixin(LoraBaseMixin):
 
         has_alphas_in_sd = any(k.endswith(".alpha") for k in state_dict)
         has_lora_unet = any(k.startswith("lora_unet_") for k in state_dict)
-        has_diffusion_model = any(k.startswith("diffusion_model.") for k in state_dict)
-        if has_alphas_in_sd or has_lora_unet or has_diffusion_model:
+        if has_alphas_in_sd or has_lora_unet:
             state_dict = _convert_non_diffusers_qwen_lora_to_diffusers(state_dict)
 
         out = (state_dict, metadata) if return_lora_metadata else state_dict

src/diffusers/models/attention_dispatch.py

@@ -955,13 +955,12 @@ def _native_npu_attention(
     dropout_p: float = 0.0,
     scale: Optional[float] = None,
 ) -> torch.Tensor:
-    query, key, value = (x.transpose(1, 2).contiguous() for x in (query, key, value))
-    out = npu_fusion_attention(
+    return npu_fusion_attention(
         query,
         key,
         value,
-        query.size(1),  # num_heads
-        input_layout="BNSD",
+        query.size(2),  # num_heads
+        input_layout="BSND",
         pse=None,
         scale=1.0 / math.sqrt(query.shape[-1]) if scale is None else scale,
         pre_tockens=65536,
@@ -970,8 +969,6 @@ def _native_npu_attention(
         sync=False,
         inner_precise=0,
     )[0]
-    out = out.transpose(1, 2).contiguous()
-    return out
 
 
 # Reference: https://github.com/pytorch/xla/blob/06c5533de6588f6b90aa1655d9850bcf733b90b4/torch_xla/experimental/custom_kernel.py#L853

src/diffusers/modular_pipelines/modular_pipeline.py

@@ -299,7 +299,7 @@ class ModularPipelineBlocks(ConfigMixin, PushToHubMixin):
     def from_pretrained(
         cls,
         pretrained_model_name_or_path: str,
-        trust_remote_code: bool = False,
+        trust_remote_code: Optional[bool] = None,
         **kwargs,
     ):
         hub_kwargs_names = [
@@ -1418,7 +1418,7 @@ class LoopSequentialPipelineBlocks(ModularPipelineBlocks):
 # YiYi TODO:
 # 1. look into the serialization of modular_model_index.json, make sure the items are properly ordered like model_index.json (currently a mess)
 # 2. do we need ConfigSpec? the are basically just key/val kwargs
-# 3. imnprove docstring and potentially add validator for methods where we accpet kwargs to be passed to from_pretrained/save_pretrained/load_components()
+# 3. imnprove docstring and potentially add validator for methods where we accpet kwargs to be passed to from_pretrained/save_pretrained/load_default_components(), load_components()
 class ModularPipeline(ConfigMixin, PushToHubMixin):
     """
     Base class for all Modular pipelines.
@@ -1488,7 +1488,7 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
             - Components with default_creation_method="from_config" are created immediately, its specs are not included
               in config dict and will not be saved in `modular_model_index.json`
             - Components with default_creation_method="from_pretrained" are set to None and can be loaded later with
-              `load_components()` (with or without specific component names)
+              `load_default_components()`/`load_components()`
             - The pipeline's config dict is populated with component specs (only for from_pretrained components) and
               config values, which will be saved as `modular_model_index.json` during `save_pretrained`
             - The pipeline's config dict is also used to store the pipeline blocks's class name, which will be saved as
@@ -1603,6 +1603,20 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
             params[input_param.name] = input_param.default
         return params
 
+    def load_default_components(self, **kwargs):
+        """
+        Load from_pretrained components using the loading specs in the config dict.
+
+        Args:
+            **kwargs: Additional arguments passed to `from_pretrained` method, e.g. torch_dtype, cache_dir, etc.
+        """
+        names = [
+            name
+            for name in self._component_specs.keys()
+            if self._component_specs[name].default_creation_method == "from_pretrained"
+        ]
+        self.load_components(names=names, **kwargs)
+
     @classmethod
     @validate_hf_hub_args
     def from_pretrained(
@@ -1756,8 +1770,8 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
            - non from_pretrained components are created during __init__ and registered as the object itself
         - Components are updated with the `update_components()` method: e.g. loader.update_components(unet=unet) or
           loader.update_components(guider=guider_spec)
-        - (from_pretrained) Components are loaded with the `load_components()` method: e.g.
-          loader.load_components(names=["unet"]) or loader.load_components() to load all default components
+        - (from_pretrained) Components are loaded with the `load_default_components()` method: e.g.
+          loader.load_default_components(names=["unet"])
 
         Args:
             **kwargs: Keyword arguments where keys are component names and values are component objects.
@@ -2083,14 +2097,13 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
         self.register_to_config(**config_to_register)
 
     # YiYi TODO: support map for additional from_pretrained kwargs
-    def load_components(self, names: Optional[Union[List[str], str]] = None, **kwargs):
+    # YiYi/Dhruv TODO: consolidate load_components and load_default_components?
+    def load_components(self, names: Union[List[str], str], **kwargs):
         """
         Load selected components from specs.
 
         Args:
-            names: List of component names to load. If None, will load all components with
-                   default_creation_method == "from_pretrained". If provided as a list or string, will load only the
-                   specified components.
+            names: List of component names to load; by default will not load any components
             **kwargs: additional kwargs to be passed to `from_pretrained()`.Can be:
              - a single value to be applied to all components to be loaded, e.g. torch_dtype=torch.bfloat16
              - a dict, e.g. torch_dtype={"unet": torch.bfloat16, "default": torch.float32}
@@ -2098,13 +2111,7 @@ class ModularPipeline(ConfigMixin, PushToHubMixin):
                `variant`, `revision`, etc.
         """
 
-        if names is None:
-            names = [
-                name
-                for name in self._component_specs.keys()
-                if self._component_specs[name].default_creation_method == "from_pretrained"
-            ]
-        elif isinstance(names, str):
+        if isinstance(names, str):
             names = [names]
         elif not isinstance(names, list):
             raise ValueError(f"Invalid type for names: {type(names)}")

src/diffusers/pipelines/__init__.py

@@ -393,7 +393,6 @@ else:
         "QwenImageImg2ImgPipeline",
         "QwenImageInpaintPipeline",
         "QwenImageEditPipeline",
-        "QwenImageEditInpaintPipeline",
         "QwenImageControlNetPipeline",
     ]
 try:
@@ -715,7 +714,6 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
         from .pixart_alpha import PixArtAlphaPipeline, PixArtSigmaPipeline
         from .qwenimage import (
             QwenImageControlNetPipeline,
-            QwenImageEditInpaintPipeline,
             QwenImageEditPipeline,
             QwenImageImg2ImgPipeline,
             QwenImageInpaintPipeline,

src/diffusers/pipelines/qwenimage/__init__.py

@@ -26,7 +26,6 @@ else:
     _import_structure["pipeline_qwenimage"] = ["QwenImagePipeline"]
     _import_structure["pipeline_qwenimage_controlnet"] = ["QwenImageControlNetPipeline"]
     _import_structure["pipeline_qwenimage_edit"] = ["QwenImageEditPipeline"]
-    _import_structure["pipeline_qwenimage_edit_inpaint"] = ["QwenImageEditInpaintPipeline"]
     _import_structure["pipeline_qwenimage_img2img"] = ["QwenImageImg2ImgPipeline"]
     _import_structure["pipeline_qwenimage_inpaint"] = ["QwenImageInpaintPipeline"]
 
@@ -40,7 +39,6 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
         from .pipeline_qwenimage import QwenImagePipeline
         from .pipeline_qwenimage_controlnet import QwenImageControlNetPipeline
         from .pipeline_qwenimage_edit import QwenImageEditPipeline
-        from .pipeline_qwenimage_edit_inpaint import QwenImageEditInpaintPipeline
         from .pipeline_qwenimage_img2img import QwenImageImg2ImgPipeline
         from .pipeline_qwenimage_inpaint import QwenImageInpaintPipeline
 else:

src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit.py

@@ -551,12 +551,6 @@ class QwenImageEditPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Function invoked when calling the pipeline for generation.
 
         Args:
-            image (`torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.Tensor]`, `List[PIL.Image.Image]`, or `List[np.ndarray]`):
-                `Image`, numpy array or tensor representing an image batch to be used as the starting point. For both
-                numpy array and pytorch tensor, the expected value range is between `[0, 1]` If it's a tensor or a list
-                or tensors, the expected shape should be `(B, C, H, W)` or `(C, H, W)`. If it is a numpy array or a
-                list of arrays, the expected shape should be `(B, H, W, C)` or `(H, W, C)` It can also accept image
-                latents as `image`, but if passing latents directly it is not encoded again.
             prompt (`str` or `List[str]`, *optional*):
                 The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
                 instead.

src/diffusers/utils/constants.py

@@ -45,7 +45,6 @@ DIFFUSERS_ATTN_BACKEND = os.getenv("DIFFUSERS_ATTN_BACKEND", "native")
 DIFFUSERS_ATTN_CHECKS = os.getenv("DIFFUSERS_ATTN_CHECKS", "0") in ENV_VARS_TRUE_VALUES
 DEFAULT_HF_PARALLEL_LOADING_WORKERS = 8
 HF_ENABLE_PARALLEL_LOADING = os.environ.get("HF_ENABLE_PARALLEL_LOADING", "").upper() in ENV_VARS_TRUE_VALUES
-DIFFUSERS_DISABLE_REMOTE_CODE = os.getenv("DIFFUSERS_DISABLE_REMOTE_CODE", "false").lower() in ENV_VARS_TRUE_VALUES
 
 # Below should be `True` if the current version of `peft` and `transformers` are compatible with
 # PEFT backend. Will automatically fall back to PEFT backend if the correct versions of the libraries are

src/diffusers/utils/dummy_torch_and_transformers_objects.py

@@ -1772,21 +1772,6 @@ class QwenImageControlNetPipeline(metaclass=DummyObject):
         requires_backends(cls, ["torch", "transformers"])
 
 
-class QwenImageEditInpaintPipeline(metaclass=DummyObject):
-    _backends = ["torch", "transformers"]
-
-    def __init__(self, *args, **kwargs):
-        requires_backends(self, ["torch", "transformers"])
-
-    @classmethod
-    def from_config(cls, *args, **kwargs):
-        requires_backends(cls, ["torch", "transformers"])
-
-    @classmethod
-    def from_pretrained(cls, *args, **kwargs):
-        requires_backends(cls, ["torch", "transformers"])
-
-
 class QwenImageEditPipeline(metaclass=DummyObject):
     _backends = ["torch", "transformers"]
 

src/diffusers/utils/dynamic_modules_utils.py

@@ -20,6 +20,7 @@ import json
 import os
 import re
 import shutil
+import signal
 import sys
 import threading
 from pathlib import Path
@@ -33,7 +34,6 @@ from packaging import version
 
 from .. import __version__
 from . import DIFFUSERS_DYNAMIC_MODULE_NAME, HF_MODULES_CACHE, logging
-from .constants import DIFFUSERS_DISABLE_REMOTE_CODE
 
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
@@ -159,25 +159,52 @@ def check_imports(filename):
     return get_relative_imports(filename)
 
 
+def _raise_timeout_error(signum, frame):
+    raise ValueError(
+        "Loading this model requires you to execute custom code contained in the model repository on your local "
+        "machine. Please set the option `trust_remote_code=True` to permit loading of this model."
+    )
+
+
 def resolve_trust_remote_code(trust_remote_code, model_name, has_remote_code):
-    trust_remote_code = trust_remote_code and not DIFFUSERS_DISABLE_REMOTE_CODE
-    if DIFFUSERS_DISABLE_REMOTE_CODE:
-        logger.warning(
-            "Downloading remote code is disabled globally via the DIFFUSERS_DISABLE_REMOTE_CODE environment variable. Ignoring `trust_remote_code`."
-        )
+    if trust_remote_code is None:
+        if has_remote_code and TIME_OUT_REMOTE_CODE > 0:
+            prev_sig_handler = None
+            try:
+                prev_sig_handler = signal.signal(signal.SIGALRM, _raise_timeout_error)
+                signal.alarm(TIME_OUT_REMOTE_CODE)
+                while trust_remote_code is None:
+                    answer = input(
+                        f"The repository for {model_name} contains custom code which must be executed to correctly "
+                        f"load the model. You can inspect the repository content at https://hf.co/{model_name}.\n"
+                        f"You can avoid this prompt in future by passing the argument `trust_remote_code=True`.\n\n"
+                        f"Do you wish to run the custom code? [y/N] "
+                    )
+                    if answer.lower() in ["yes", "y", "1"]:
+                        trust_remote_code = True
+                    elif answer.lower() in ["no", "n", "0", ""]:
+                        trust_remote_code = False
+                signal.alarm(0)
+            except Exception:
+                # OS which does not support signal.SIGALRM
+                raise ValueError(
+                    f"The repository for {model_name} contains custom code which must be executed to correctly "
+                    f"load the model. You can inspect the repository content at https://hf.co/{model_name}.\n"
+                    f"Please pass the argument `trust_remote_code=True` to allow custom code to be run."
+                )
+            finally:
+                if prev_sig_handler is not None:
+                    signal.signal(signal.SIGALRM, prev_sig_handler)
+                    signal.alarm(0)
+        elif has_remote_code:
+            # For the CI which puts the timeout at 0
+            _raise_timeout_error(None, None)
 
     if has_remote_code and not trust_remote_code:
-        error_msg = f"The repository for {model_name} contains custom code. "
-        error_msg += (
-            "Downloading remote code is disabled globally via the DIFFUSERS_DISABLE_REMOTE_CODE environment variable."
-            if DIFFUSERS_DISABLE_REMOTE_CODE
-            else "Pass `trust_remote_code=True` to allow loading remote code modules."
-        )
-        raise ValueError(error_msg)
-
-    elif has_remote_code and trust_remote_code:
-        logger.warning(
-            f"`trust_remote_code` is enabled. Downloading code from {model_name}. Please ensure you trust the contents of this repository"
+        raise ValueError(
+            f"Loading {model_name} requires you to execute the configuration file in that"
+            " repo on your local machine. Make sure you have read the code there to avoid malicious use, then"
+            " set the option `trust_remote_code=True` to remove this error."
         )
 
     return trust_remote_code

tests/modular_pipelines/stable_diffusion_xl/test_modular_pipeline_stable_diffusion_xl.py

@@ -67,7 +67,7 @@ class SDXLModularTests:
 
     def get_pipeline(self, components_manager=None, torch_dtype=torch.float32):
         pipeline = self.pipeline_blocks_class().init_pipeline(self.repo, components_manager=components_manager)
-        pipeline.load_components(torch_dtype=torch_dtype)
+        pipeline.load_default_components(torch_dtype=torch_dtype)
         return pipeline
 
     def get_dummy_inputs(self, device, seed=0):
@@ -158,7 +158,7 @@ class SDXLModularIPAdapterTests:
         blocks = self.pipeline_blocks_class()
         _ = blocks.sub_blocks.pop("ip_adapter")
         pipe = blocks.init_pipeline(self.repo)
-        pipe.load_components(torch_dtype=torch.float32)
+        pipe.load_default_components(torch_dtype=torch.float32)
         pipe = pipe.to(torch_device)
         pipe.set_progress_bar_config(disable=None)
         cross_attention_dim = pipe.unet.config.get("cross_attention_dim")

tests/modular_pipelines/test_modular_pipelines_common.py

@@ -344,7 +344,7 @@ class ModularPipelineTesterMixin:
         with tempfile.TemporaryDirectory() as tmpdirname:
             base_pipe.save_pretrained(tmpdirname)
             pipe = ModularPipeline.from_pretrained(tmpdirname).to(torch_device)
-            pipe.load_components(torch_dtype=torch.float32)
+            pipe.load_default_components(torch_dtype=torch.float32)
             pipe.to(torch_device)
 
         pipes.append(pipe)

tests/quantization/quanto/test_quanto.py

@@ -13,7 +13,7 @@ from ...testing_utils import (
     nightly,
     numpy_cosine_similarity_distance,
     require_accelerate,
-    require_accelerator,
+    require_big_accelerator,
     require_torch_cuda_compatibility,
     torch_device,
 )
@@ -31,7 +31,7 @@ enable_full_determinism()
 
 
 @nightly
-@require_accelerator
+@require_big_accelerator
 @require_accelerate
 class QuantoBaseTesterMixin:
     model_id = None