Merge branch 'main' into layerwise-upcasting

Author: sayakpaul

docs/source/en/api/pipelines/controlnet_sd3.md

@@ -22,7 +22,7 @@ The abstract from the paper is:
 
 *We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models. ControlNet locks the production-ready large diffusion models, and reuses their deep and robust encoding layers pretrained with billions of images as a strong backbone to learn a diverse set of conditional controls. The neural architecture is connected with "zero convolutions" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning. We test various conditioning controls, eg, edges, depth, segmentation, human pose, etc, with Stable Diffusion, using single or multiple conditions, with or without prompts. We show that the training of ControlNets is robust with small (<50k) and large (>1m) datasets. Extensive results show that ControlNet may facilitate wider applications to control image diffusion models.*
 
-This controlnet code is mainly implemented by [The InstantX Team](https://huggingface.co/InstantX). The inpainting-related code was developed by [The Alimama Creative Team](https://huggingface.co/alimama-creative). You can find pre-trained checkpoints for SD3-ControlNet in the table below: 
+This controlnet code is mainly implemented by [The InstantX Team](https://huggingface.co/InstantX). The inpainting-related code was developed by [The Alimama Creative Team](https://huggingface.co/alimama-creative). You can find pre-trained checkpoints for SD3-ControlNet in the table below:
 
 
 | ControlNet type | Developer | Link |

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

@@ -14,7 +14,7 @@ specific language governing permissions and limitations under the License.
 
 ![](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/kolors/kolors_header_collage.png)
 
-Kolors is a large-scale text-to-image generation model based on latent diffusion, developed by [the Kuaishou Kolors team](kwai-kolors@kuaishou.com). Trained on billions of text-image pairs, Kolors exhibits significant advantages over both open-source and closed-source models in visual quality, complex semantic accuracy, and text rendering for both Chinese and English characters. Furthermore, Kolors supports both Chinese and English inputs, demonstrating strong performance in understanding and generating Chinese-specific content. For more details, please refer to this [technical report](https://github.com/Kwai-Kolors/Kolors/blob/master/imgs/Kolors_paper.pdf).
+Kolors is a large-scale text-to-image generation model based on latent diffusion, developed by [the Kuaishou Kolors team](https://github.com/Kwai-Kolors/Kolors). Trained on billions of text-image pairs, Kolors exhibits significant advantages over both open-source and closed-source models in visual quality, complex semantic accuracy, and text rendering for both Chinese and English characters. Furthermore, Kolors supports both Chinese and English inputs, demonstrating strong performance in understanding and generating Chinese-specific content. For more details, please refer to this [technical report](https://github.com/Kwai-Kolors/Kolors/blob/master/imgs/Kolors_paper.pdf).
 
 The abstract from the technical report is:
 
@@ -74,7 +74,7 @@ image_encoder = CLIPVisionModelWithProjection.from_pretrained(
 
 pipe = KolorsPipeline.from_pretrained(
     "Kwai-Kolors/Kolors-diffusers", image_encoder=image_encoder, torch_dtype=torch.float16, variant="fp16"
-).to("cuda")
+)
 pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config, use_karras_sigmas=True)
 
 pipe.load_ip_adapter(

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

@@ -20,7 +20,7 @@ The abstract from the paper is:
 
 *Recent studies have demonstrated that diffusion models are capable of generating high-quality samples, but their quality heavily depends on sampling guidance techniques, such as classifier guidance (CG) and classifier-free guidance (CFG). These techniques are often not applicable in unconditional generation or in various downstream tasks such as image restoration. In this paper, we propose a novel sampling guidance, called Perturbed-Attention Guidance (PAG), which improves diffusion sample quality across both unconditional and conditional settings, achieving this without requiring additional training or the integration of external modules. PAG is designed to progressively enhance the structure of samples throughout the denoising process. It involves generating intermediate samples with degraded structure by substituting selected self-attention maps in diffusion U-Net with an identity matrix, by considering the self-attention mechanisms' ability to capture structural information, and guiding the denoising process away from these degraded samples. In both ADM and Stable Diffusion, PAG surprisingly improves sample quality in conditional and even unconditional scenarios. Moreover, PAG significantly improves the baseline performance in various downstream tasks where existing guidances such as CG or CFG cannot be fully utilized, including ControlNet with empty prompts and image restoration such as inpainting and deblurring.*
 
-PAG can be used by specifying the `pag_applied_layers` as a parameter when instantiating a PAG pipeline. It can be a single string or a list of strings. Each string can be a unique layer identifier or a regular expression to identify one or more layers. 
+PAG can be used by specifying the `pag_applied_layers` as a parameter when instantiating a PAG pipeline. It can be a single string or a list of strings. Each string can be a unique layer identifier or a regular expression to identify one or more layers.
 
 - Full identifier as a normal string: `down_blocks.2.attentions.0.transformer_blocks.0.attn1.processor`
 - Full identifier as a RegEx: `down_blocks.2.(attentions|motion_modules).0.transformer_blocks.0.attn1.processor`
@@ -46,7 +46,7 @@ Since RegEx is supported as a way for matching layer identifiers, it is crucial
 ## KolorsPAGPipeline
 [[autodoc]] KolorsPAGPipeline
   - all
-  - __call__  
+  - __call__
 
 ## StableDiffusionPAGPipeline
 [[autodoc]] StableDiffusionPAGPipeline

examples/dreambooth/README_flux.md

@@ -3,17 +3,17 @@
 [DreamBooth](https://arxiv.org/abs/2208.12242) is a method to personalize text2image models like stable diffusion given just a few (3~5) images of a subject.
 
 The `train_dreambooth_flux.py` script shows how to implement the training procedure and adapt it for [FLUX.1 [dev]](https://blackforestlabs.ai/announcing-black-forest-labs/). We also provide a LoRA implementation in the `train_dreambooth_lora_flux.py` script.
-> [!NOTE] 
+> [!NOTE]
 > **Memory consumption**
-> 
-> Flux can be quite expensive to run on consumer hardware devices and as a result finetuning it comes with high memory requirements - 
+>
+> Flux can be quite expensive to run on consumer hardware devices and as a result finetuning it comes with high memory requirements -
 > a LoRA with a rank of 16 (w/ all components trained) can exceed 40GB of VRAM for training.
-> For more tips & guidance on training on a resource-constrained device please visit [`@bghira`'s guide](https://github.com/bghira/SimpleTuner/blob/main/documentation/quickstart/FLUX.md) 
+> For more tips & guidance on training on a resource-constrained device please visit [`@bghira`'s guide](https://github.com/bghira/SimpleTuner/blob/main/documentation/quickstart/FLUX.md)
 
 
 > [!NOTE]
 > **Gated model**
-> 
+>
 > As the model is gated, before using it with diffusers you first need to go to the [FLUX.1 [dev] Hugging Face page](https://huggingface.co/black-forest-labs/FLUX.1-dev), fill in the form and accept the gate. Once you are in, you need to log in so that your system knows you’ve accepted the gate. Use the command below to log in:
 
 ```bash
@@ -163,7 +163,7 @@ To do so, just specify `--train_text_encoder` while launching training. Please k
 
 > [!NOTE]
 > FLUX.1 has 2 text encoders (CLIP L/14 and T5-v1.1-XXL).
-By enabling `--train_text_encoder`, fine-tuning of the **CLIP encoder** is performed. 
+By enabling `--train_text_encoder`, fine-tuning of the **CLIP encoder** is performed.
 > At the moment, T5 fine-tuning is not supported and weights remain frozen when text encoder training is enabled.
 
 To perform DreamBooth LoRA with text-encoder training, run:

examples/dreambooth/train_dreambooth_lora_sd3.py

@@ -1454,7 +1454,7 @@ def compute_text_embeddings(prompt, text_encoders, tokenizers):
             )
 
     # Clear the memory here
-    if not args.train_text_encoder and train_dataset.custom_instance_prompts:
+    if not args.train_text_encoder and not train_dataset.custom_instance_prompts:
         del tokenizers, text_encoders
         # Explicitly delete the objects as well, otherwise only the lists are deleted and the original references remain, preventing garbage collection
         del text_encoder_one, text_encoder_two, text_encoder_three

src/diffusers/loaders/ip_adapter.py

@@ -222,7 +222,11 @@ def load_ip_adapter(
 
             # create feature extractor if it has not been registered to the pipeline yet
             if hasattr(self, "feature_extractor") and getattr(self, "feature_extractor", None) is None:
-                clip_image_size = self.image_encoder.config.image_size
+                # FaceID IP adapters don't need the image encoder so it's not present, in this case we default to 224
+                default_clip_size = 224
+                clip_image_size = (
+                    self.image_encoder.config.image_size if self.image_encoder is not None else default_clip_size
+                )
                 feature_extractor = CLIPImageProcessor(size=clip_image_size, crop_size=clip_image_size)
                 self.register_modules(feature_extractor=feature_extractor)
 

src/diffusers/models/transformers/auraflow_transformer_2d.py

@@ -274,6 +274,7 @@ class AuraFlowTransformer2DModel(ModelMixin, ConfigMixin):
         pos_embed_max_size (`int`, defaults to 4096): Maximum positions to embed from the image latents.
     """
 
+    _no_split_modules = ["AuraFlowJointTransformerBlock", "AuraFlowSingleTransformerBlock", "AuraFlowPatchEmbed"]
     _supports_gradient_checkpointing = True
     _always_upcast_modules = ["AuraFlowPatchEmbed"]
 

src/diffusers/pipelines/auto_pipeline.py

@@ -49,6 +49,7 @@
 )
 from .kandinsky3 import Kandinsky3Img2ImgPipeline, Kandinsky3Pipeline
 from .latent_consistency_models import LatentConsistencyModelImg2ImgPipeline, LatentConsistencyModelPipeline
+from .lumina import LuminaText2ImgPipeline
 from .pag import (
     HunyuanDiTPAGPipeline,
     PixArtSigmaPAGPipeline,
@@ -106,6 +107,7 @@
         ("pixart-sigma-pag", PixArtSigmaPAGPipeline),
         ("auraflow", AuraFlowPipeline),
         ("flux", FluxPipeline),
+        ("lumina", LuminaText2ImgPipeline),
     ]
 )
 

src/diffusers/pipelines/latte/pipeline_latte.py

@@ -56,7 +56,7 @@
         >>> from diffusers.utils import export_to_gif
 
         >>> # You can replace the checkpoint id with "maxin-cn/Latte-1" too.
-        >>> pipe = LattePipeline.from_pretrained("maxin-cn/Latte-1", torch_dtype=torch.float16).to("cuda")
+        >>> pipe = LattePipeline.from_pretrained("maxin-cn/Latte-1", torch_dtype=torch.float16)
         >>> # Enable memory optimizations.
         >>> pipe.enable_model_cpu_offload()
 

src/diffusers/pipelines/lumina/pipeline_lumina.py

@@ -54,7 +54,7 @@
 
         >>> pipe = LuminaText2ImgPipeline.from_pretrained(
         ...     "Alpha-VLLM/Lumina-Next-SFT-diffusers", torch_dtype=torch.bfloat16
-        ... ).cuda()
+        ... )
         >>> # Enable memory optimizations.
         >>> pipe.enable_model_cpu_offload()