diffusers 0.31.0__py3-none-any.whl → 0.32.0__py3-none-any.whl

diffusers/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.31.0"
+__version__ = "0.32.0"
 
 from typing import TYPE_CHECKING
 
@@ -31,7 +31,7 @@ _import_structure = {
     "loaders": ["FromOriginalModelMixin"],
     "models": [],
     "pipelines": [],
-    "quantizers.quantization_config": ["BitsAndBytesConfig"],
+    "quantizers.quantization_config": ["BitsAndBytesConfig", "GGUFQuantizationConfig", "TorchAoConfig"],
     "schedulers": [],
     "utils": [
         "OptionalDependencyNotAvailable",
@@ -77,10 +77,16 @@ except OptionalDependencyNotAvailable:
 else:
     _import_structure["models"].extend(
         [
+            "AllegroTransformer3DModel",
             "AsymmetricAutoencoderKL",
             "AuraFlowTransformer2DModel",
+            "AutoencoderDC",
             "AutoencoderKL",
+            "AutoencoderKLAllegro",
             "AutoencoderKLCogVideoX",
+            "AutoencoderKLHunyuanVideo",
+            "AutoencoderKLLTXVideo",
+            "AutoencoderKLMochi",
             "AutoencoderKLTemporalDecoder",
             "AutoencoderOobleck",
             "AutoencoderTiny",
@@ -88,6 +94,7 @@ else:
             "CogView3PlusTransformer2DModel",
             "ConsistencyDecoderVAE",
             "ControlNetModel",
+            "ControlNetUnionModel",
             "ControlNetXSAdapter",
             "DiTTransformer2DModel",
             "FluxControlNetModel",
@@ -96,15 +103,20 @@ else:
             "HunyuanDiT2DControlNetModel",
             "HunyuanDiT2DModel",
             "HunyuanDiT2DMultiControlNetModel",
+            "HunyuanVideoTransformer3DModel",
             "I2VGenXLUNet",
             "Kandinsky3UNet",
             "LatteTransformer3DModel",
+            "LTXVideoTransformer3DModel",
             "LuminaNextDiT2DModel",
+            "MochiTransformer3DModel",
             "ModelMixin",
             "MotionAdapter",
             "MultiAdapter",
+            "MultiControlNetModel",
             "PixArtTransformer2DModel",
             "PriorTransformer",
+            "SanaTransformer2DModel",
             "SD3ControlNetModel",
             "SD3MultiControlNetModel",
             "SD3Transformer2DModel",
@@ -237,6 +249,7 @@ except OptionalDependencyNotAvailable:
 else:
     _import_structure["pipelines"].extend(
         [
+            "AllegroPipeline",
             "AltDiffusionImg2ImgPipeline",
             "AltDiffusionPipeline",
             "AmusedImg2ImgPipeline",
@@ -263,15 +276,21 @@ else:
             "CogVideoXVideoToVideoPipeline",
             "CogView3PlusPipeline",
             "CycleDiffusionPipeline",
+            "FluxControlImg2ImgPipeline",
+            "FluxControlInpaintPipeline",
             "FluxControlNetImg2ImgPipeline",
             "FluxControlNetInpaintPipeline",
             "FluxControlNetPipeline",
+            "FluxControlPipeline",
+            "FluxFillPipeline",
             "FluxImg2ImgPipeline",
             "FluxInpaintPipeline",
             "FluxPipeline",
+            "FluxPriorReduxPipeline",
             "HunyuanDiTControlNetPipeline",
             "HunyuanDiTPAGPipeline",
             "HunyuanDiTPipeline",
+            "HunyuanVideoPipeline",
             "I2VGenXLPipeline",
             "IFImg2ImgPipeline",
             "IFImg2ImgSuperResolutionPipeline",
@@ -305,15 +324,21 @@ else:
             "LDMTextToImagePipeline",
             "LEditsPPPipelineStableDiffusion",
             "LEditsPPPipelineStableDiffusionXL",
+            "LTXImageToVideoPipeline",
+            "LTXPipeline",
             "LuminaText2ImgPipeline",
             "MarigoldDepthPipeline",
             "MarigoldNormalsPipeline",
+            "MochiPipeline",
             "MusicLDMPipeline",
             "PaintByExamplePipeline",
             "PIAPipeline",
             "PixArtAlphaPipeline",
             "PixArtSigmaPAGPipeline",
             "PixArtSigmaPipeline",
+            "ReduxImageEncoder",
+            "SanaPAGPipeline",
+            "SanaPipeline",
             "SemanticStableDiffusionPipeline",
             "ShapEImg2ImgPipeline",
             "ShapEPipeline",
@@ -326,6 +351,8 @@ else:
             "StableDiffusion3ControlNetPipeline",
             "StableDiffusion3Img2ImgPipeline",
             "StableDiffusion3InpaintPipeline",
+            "StableDiffusion3PAGImg2ImgPipeline",
+            "StableDiffusion3PAGImg2ImgPipeline",
             "StableDiffusion3PAGPipeline",
             "StableDiffusion3Pipeline",
             "StableDiffusionAdapterPipeline",
@@ -349,6 +376,7 @@ else:
             "StableDiffusionLDM3DPipeline",
             "StableDiffusionModelEditingPipeline",
             "StableDiffusionPAGImg2ImgPipeline",
+            "StableDiffusionPAGInpaintPipeline",
             "StableDiffusionPAGPipeline",
             "StableDiffusionPanoramaPipeline",
             "StableDiffusionParadigmsPipeline",
@@ -363,6 +391,9 @@ else:
             "StableDiffusionXLControlNetPAGImg2ImgPipeline",
             "StableDiffusionXLControlNetPAGPipeline",
             "StableDiffusionXLControlNetPipeline",
+            "StableDiffusionXLControlNetUnionImg2ImgPipeline",
+            "StableDiffusionXLControlNetUnionInpaintPipeline",
+            "StableDiffusionXLControlNetUnionPipeline",
             "StableDiffusionXLControlNetXSPipeline",
             "StableDiffusionXLImg2ImgPipeline",
             "StableDiffusionXLInpaintPipeline",
@@ -481,7 +512,7 @@ except OptionalDependencyNotAvailable:
 
 
 else:
-    _import_structure["models.controlnet_flax"] = ["FlaxControlNetModel"]
+    _import_structure["models.controlnets.controlnet_flax"] = ["FlaxControlNetModel"]
     _import_structure["models.modeling_flax_utils"] = ["FlaxModelMixin"]
     _import_structure["models.unets.unet_2d_condition_flax"] = ["FlaxUNet2DConditionModel"]
     _import_structure["models.vae_flax"] = ["FlaxAutoencoderKL"]
@@ -539,7 +570,7 @@ else:
 
 if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     from .configuration_utils import ConfigMixin
-    from .quantizers.quantization_config import BitsAndBytesConfig
+    from .quantizers.quantization_config import BitsAndBytesConfig, GGUFQuantizationConfig, TorchAoConfig
 
     try:
         if not is_onnx_available():
@@ -556,10 +587,16 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
         from .utils.dummy_pt_objects import *  # noqa F403
     else:
         from .models import (
+            AllegroTransformer3DModel,
             AsymmetricAutoencoderKL,
             AuraFlowTransformer2DModel,
+            AutoencoderDC,
             AutoencoderKL,
+            AutoencoderKLAllegro,
             AutoencoderKLCogVideoX,
+            AutoencoderKLHunyuanVideo,
+            AutoencoderKLLTXVideo,
+            AutoencoderKLMochi,
             AutoencoderKLTemporalDecoder,
             AutoencoderOobleck,
             AutoencoderTiny,
@@ -567,6 +604,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             CogView3PlusTransformer2DModel,
             ConsistencyDecoderVAE,
             ControlNetModel,
+            ControlNetUnionModel,
             ControlNetXSAdapter,
             DiTTransformer2DModel,
             FluxControlNetModel,
@@ -575,15 +613,20 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             HunyuanDiT2DControlNetModel,
             HunyuanDiT2DModel,
             HunyuanDiT2DMultiControlNetModel,
+            HunyuanVideoTransformer3DModel,
             I2VGenXLUNet,
             Kandinsky3UNet,
             LatteTransformer3DModel,
+            LTXVideoTransformer3DModel,
             LuminaNextDiT2DModel,
+            MochiTransformer3DModel,
             ModelMixin,
             MotionAdapter,
             MultiAdapter,
+            MultiControlNetModel,
             PixArtTransformer2DModel,
             PriorTransformer,
+            SanaTransformer2DModel,
             SD3ControlNetModel,
             SD3MultiControlNetModel,
             SD3Transformer2DModel,
@@ -697,6 +740,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
         from .utils.dummy_torch_and_transformers_objects import *  # noqa F403
     else:
         from .pipelines import (
+            AllegroPipeline,
             AltDiffusionImg2ImgPipeline,
             AltDiffusionPipeline,
             AmusedImg2ImgPipeline,
@@ -721,15 +765,21 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             CogVideoXVideoToVideoPipeline,
             CogView3PlusPipeline,
             CycleDiffusionPipeline,
+            FluxControlImg2ImgPipeline,
+            FluxControlInpaintPipeline,
             FluxControlNetImg2ImgPipeline,
             FluxControlNetInpaintPipeline,
             FluxControlNetPipeline,
+            FluxControlPipeline,
+            FluxFillPipeline,
             FluxImg2ImgPipeline,
             FluxInpaintPipeline,
             FluxPipeline,
+            FluxPriorReduxPipeline,
             HunyuanDiTControlNetPipeline,
             HunyuanDiTPAGPipeline,
             HunyuanDiTPipeline,
+            HunyuanVideoPipeline,
             I2VGenXLPipeline,
             IFImg2ImgPipeline,
             IFImg2ImgSuperResolutionPipeline,
@@ -763,15 +813,21 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             LDMTextToImagePipeline,
             LEditsPPPipelineStableDiffusion,
             LEditsPPPipelineStableDiffusionXL,
+            LTXImageToVideoPipeline,
+            LTXPipeline,
             LuminaText2ImgPipeline,
             MarigoldDepthPipeline,
             MarigoldNormalsPipeline,
+            MochiPipeline,
             MusicLDMPipeline,
             PaintByExamplePipeline,
             PIAPipeline,
             PixArtAlphaPipeline,
             PixArtSigmaPAGPipeline,
             PixArtSigmaPipeline,
+            ReduxImageEncoder,
+            SanaPAGPipeline,
+            SanaPipeline,
             SemanticStableDiffusionPipeline,
             ShapEImg2ImgPipeline,
             ShapEPipeline,
@@ -783,6 +839,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             StableDiffusion3ControlNetPipeline,
             StableDiffusion3Img2ImgPipeline,
             StableDiffusion3InpaintPipeline,
+            StableDiffusion3PAGImg2ImgPipeline,
             StableDiffusion3PAGPipeline,
             StableDiffusion3Pipeline,
             StableDiffusionAdapterPipeline,
@@ -806,6 +863,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             StableDiffusionLDM3DPipeline,
             StableDiffusionModelEditingPipeline,
             StableDiffusionPAGImg2ImgPipeline,
+            StableDiffusionPAGInpaintPipeline,
             StableDiffusionPAGPipeline,
             StableDiffusionPanoramaPipeline,
             StableDiffusionParadigmsPipeline,
@@ -820,6 +878,9 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             StableDiffusionXLControlNetPAGImg2ImgPipeline,
             StableDiffusionXLControlNetPAGPipeline,
             StableDiffusionXLControlNetPipeline,
+            StableDiffusionXLControlNetUnionImg2ImgPipeline,
+            StableDiffusionXLControlNetUnionInpaintPipeline,
+            StableDiffusionXLControlNetUnionPipeline,
             StableDiffusionXLControlNetXSPipeline,
             StableDiffusionXLImg2ImgPipeline,
             StableDiffusionXLInpaintPipeline,
@@ -902,7 +963,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     except OptionalDependencyNotAvailable:
         from .utils.dummy_flax_objects import *  # noqa F403
     else:
-        from .models.controlnet_flax import FlaxControlNetModel
+        from .models.controlnets.controlnet_flax import FlaxControlNetModel
         from .models.modeling_flax_utils import FlaxModelMixin
         from .models.unets.unet_2d_condition_flax import FlaxUNet2DConditionModel
         from .models.vae_flax import FlaxAutoencoderKL

diffusers/callbacks.py

@@ -97,13 +97,17 @@ class SDCFGCutoffCallback(PipelineCallback):
 
 class SDXLCFGCutoffCallback(PipelineCallback):
     """
-    Callback function for Stable Diffusion XL Pipelines. After certain number of steps (set by `cutoff_step_ratio` or
-    `cutoff_step_index`), this callback will disable the CFG.
+    Callback function for the base Stable Diffusion XL Pipelines. After certain number of steps (set by
+    `cutoff_step_ratio` or `cutoff_step_index`), this callback will disable the CFG.
 
     Note: This callback mutates the pipeline by changing the `_guidance_scale` attribute to 0.0 after the cutoff step.
     """
 
-    tensor_inputs = ["prompt_embeds", "add_text_embeds", "add_time_ids"]
+    tensor_inputs = [
+        "prompt_embeds",
+        "add_text_embeds",
+        "add_time_ids",
+    ]
 
     def callback_fn(self, pipeline, step_index, timestep, callback_kwargs) -> Dict[str, Any]:
         cutoff_step_ratio = self.config.cutoff_step_ratio
@@ -129,6 +133,55 @@ class SDXLCFGCutoffCallback(PipelineCallback):
             callback_kwargs[self.tensor_inputs[0]] = prompt_embeds
             callback_kwargs[self.tensor_inputs[1]] = add_text_embeds
             callback_kwargs[self.tensor_inputs[2]] = add_time_ids
+
+        return callback_kwargs
+
+
+class SDXLControlnetCFGCutoffCallback(PipelineCallback):
+    """
+    Callback function for the Controlnet Stable Diffusion XL Pipelines. After certain number of steps (set by
+    `cutoff_step_ratio` or `cutoff_step_index`), this callback will disable the CFG.
+
+    Note: This callback mutates the pipeline by changing the `_guidance_scale` attribute to 0.0 after the cutoff step.
+    """
+
+    tensor_inputs = [
+        "prompt_embeds",
+        "add_text_embeds",
+        "add_time_ids",
+        "image",
+    ]
+
+    def callback_fn(self, pipeline, step_index, timestep, callback_kwargs) -> Dict[str, Any]:
+        cutoff_step_ratio = self.config.cutoff_step_ratio
+        cutoff_step_index = self.config.cutoff_step_index
+
+        # Use cutoff_step_index if it's not None, otherwise use cutoff_step_ratio
+        cutoff_step = (
+            cutoff_step_index if cutoff_step_index is not None else int(pipeline.num_timesteps * cutoff_step_ratio)
+        )
+
+        if step_index == cutoff_step:
+            prompt_embeds = callback_kwargs[self.tensor_inputs[0]]
+            prompt_embeds = prompt_embeds[-1:]  # "-1" denotes the embeddings for conditional text tokens.
+
+            add_text_embeds = callback_kwargs[self.tensor_inputs[1]]
+            add_text_embeds = add_text_embeds[-1:]  # "-1" denotes the embeddings for conditional pooled text tokens
+
+            add_time_ids = callback_kwargs[self.tensor_inputs[2]]
+            add_time_ids = add_time_ids[-1:]  # "-1" denotes the embeddings for conditional added time vector
+
+            # For Controlnet
+            image = callback_kwargs[self.tensor_inputs[3]]
+            image = image[-1:]
+
+            pipeline._guidance_scale = 0.0
+
+            callback_kwargs[self.tensor_inputs[0]] = prompt_embeds
+            callback_kwargs[self.tensor_inputs[1]] = add_text_embeds
+            callback_kwargs[self.tensor_inputs[2]] = add_time_ids
+            callback_kwargs[self.tensor_inputs[3]] = image
+
         return callback_kwargs
 
 

diffusers/configuration_utils.py

@@ -170,7 +170,7 @@ class ConfigMixin:
 
         if push_to_hub:
             commit_message = kwargs.pop("commit_message", None)
-            private = kwargs.pop("private", False)
+            private = kwargs.pop("private", None)
             create_pr = kwargs.pop("create_pr", False)
             token = kwargs.pop("token", None)
             repo_id = kwargs.pop("repo_id", save_directory.split(os.path.sep)[-1])

diffusers/dependency_versions_table.py

@@ -38,7 +38,7 @@ deps = {
     "regex": "regex!=2019.12.17",
     "requests": "requests",
     "tensorboard": "tensorboard",
-    "torch": "torch>=1.4,<2.5.0",
+    "torch": "torch>=1.4",
     "torchvision": "torchvision",
     "transformers": "transformers>=4.41.2",
     "urllib3": "urllib3<=2.0.0",

diffusers/image_processor.py

@@ -236,7 +236,7 @@ class VaeImageProcessor(ConfigMixin):
             `np.ndarray` or `torch.Tensor`:
                 The denormalized image array.
         """
-        return (images / 2 + 0.5).clamp(0, 1)
+        return (images * 0.5 + 0.5).clamp(0, 1)
 
     @staticmethod
     def convert_to_rgb(image: PIL.Image.Image) -> PIL.Image.Image:
@@ -537,6 +537,26 @@ class VaeImageProcessor(ConfigMixin):
 
         return image
 
+    def _denormalize_conditionally(
+        self, images: torch.Tensor, do_denormalize: Optional[List[bool]] = None
+    ) -> torch.Tensor:
+        r"""
+        Denormalize a batch of images based on a condition list.
+
+        Args:
+            images (`torch.Tensor`):
+                The input image tensor.
+            do_denormalize (`Optional[List[bool]`, *optional*, defaults to `None`):
+                A list of booleans indicating whether to denormalize each image in the batch. If `None`, will use the
+                value of `do_normalize` in the `VaeImageProcessor` config.
+        """
+        if do_denormalize is None:
+            return self.denormalize(images) if self.config.do_normalize else images
+
+        return torch.stack(
+            [self.denormalize(images[i]) if do_denormalize[i] else images[i] for i in range(images.shape[0])]
+        )
+
     def get_default_height_width(
         self,
         image: Union[PIL.Image.Image, np.ndarray, torch.Tensor],
@@ -752,12 +772,7 @@ class VaeImageProcessor(ConfigMixin):
         if output_type == "latent":
             return image
 
-        if do_denormalize is None:
-            do_denormalize = [self.config.do_normalize] * image.shape[0]
-
-        image = torch.stack(
-            [self.denormalize(image[i]) if do_denormalize[i] else image[i] for i in range(image.shape[0])]
-        )
+        image = self._denormalize_conditionally(image, do_denormalize)
 
         if output_type == "pt":
             return image
@@ -795,13 +810,11 @@ class VaeImageProcessor(ConfigMixin):
                 The final image with the overlay applied.
         """
 
-        width, height = image.width, image.height
-
-        init_image = self.resize(init_image, width=width, height=height)
-        mask = self.resize(mask, width=width, height=height)
+        width, height = init_image.width, init_image.height
 
         init_image_masked = PIL.Image.new("RGBa", (width, height))
         init_image_masked.paste(init_image.convert("RGBA").convert("RGBa"), mask=ImageOps.invert(mask.convert("L")))
+
         init_image_masked = init_image_masked.convert("RGBA")
 
         if crop_coords is not None:
@@ -968,12 +981,7 @@ class VaeImageProcessorLDM3D(VaeImageProcessor):
             deprecate("Unsupported output_type", "1.0.0", deprecation_message, standard_warn=False)
             output_type = "np"
 
-        if do_denormalize is None:
-            do_denormalize = [self.config.do_normalize] * image.shape[0]
-
-        image = torch.stack(
-            [self.denormalize(image[i]) if do_denormalize[i] else image[i] for i in range(image.shape[0])]
-        )
+        image = self._denormalize_conditionally(image, do_denormalize)
 
         image = self.pt_to_numpy(image)
 

diffusers/loaders/__init__.py

@@ -55,7 +55,8 @@ _import_structure = {}
 
 if is_torch_available():
     _import_structure["single_file_model"] = ["FromOriginalModelMixin"]
-
+    _import_structure["transformer_flux"] = ["FluxTransformer2DLoadersMixin"]
+    _import_structure["transformer_sd3"] = ["SD3Transformer2DLoadersMixin"]
     _import_structure["unet"] = ["UNet2DConditionLoadersMixin"]
     _import_structure["utils"] = ["AttnProcsLayers"]
     if is_transformers_available():
@@ -65,12 +66,20 @@ if is_torch_available():
             "StableDiffusionLoraLoaderMixin",
             "SD3LoraLoaderMixin",
             "StableDiffusionXLLoraLoaderMixin",
+            "LTXVideoLoraLoaderMixin",
             "LoraLoaderMixin",
             "FluxLoraLoaderMixin",
             "CogVideoXLoraLoaderMixin",
+            "Mochi1LoraLoaderMixin",
+            "HunyuanVideoLoraLoaderMixin",
+            "SanaLoraLoaderMixin",
         ]
         _import_structure["textual_inversion"] = ["TextualInversionLoaderMixin"]
-        _import_structure["ip_adapter"] = ["IPAdapterMixin"]
+        _import_structure["ip_adapter"] = [
+            "IPAdapterMixin",
+            "FluxIPAdapterMixin",
+            "SD3IPAdapterMixin",
+        ]
 
 _import_structure["peft"] = ["PeftAdapterMixin"]
 
@@ -78,16 +87,26 @@ _import_structure["peft"] = ["PeftAdapterMixin"]
 if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     if is_torch_available():
         from .single_file_model import FromOriginalModelMixin
+        from .transformer_flux import FluxTransformer2DLoadersMixin
+        from .transformer_sd3 import SD3Transformer2DLoadersMixin
         from .unet import UNet2DConditionLoadersMixin
         from .utils import AttnProcsLayers
 
         if is_transformers_available():
-            from .ip_adapter import IPAdapterMixin
+            from .ip_adapter import (
+                FluxIPAdapterMixin,
+                IPAdapterMixin,
+                SD3IPAdapterMixin,
+            )
             from .lora_pipeline import (
                 AmusedLoraLoaderMixin,
                 CogVideoXLoraLoaderMixin,
                 FluxLoraLoaderMixin,
+                HunyuanVideoLoraLoaderMixin,
                 LoraLoaderMixin,
+                LTXVideoLoraLoaderMixin,
+                Mochi1LoraLoaderMixin,
+                SanaLoraLoaderMixin,
                 SD3LoraLoaderMixin,
                 StableDiffusionLoraLoaderMixin,
                 StableDiffusionXLLoraLoaderMixin,