diffusers 0.33.1__py3-none-any.whl → 0.34.0__py3-none-any.whl

diffusers/hooks/hooks.py

@@ -1,4 +1,4 @@
-# Copyright 2024 The HuggingFace Team. All rights reserved.
+# Copyright 2025 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -45,7 +45,7 @@ class ModelHook:
 
     def deinitalize_hook(self, module: torch.nn.Module) -> torch.nn.Module:
         r"""
-        Hook that is executed when a model is deinitalized.
+        Hook that is executed when a model is deinitialized.
 
         Args:
             module (`torch.nn.Module`):

diffusers/hooks/layerwise_casting.py

@@ -1,4 +1,4 @@
-# Copyright 2024 The HuggingFace Team. All rights reserved.
+# Copyright 2025 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -62,7 +62,7 @@ class LayerwiseCastingHook(ModelHook):
 
     def deinitalize_hook(self, module: torch.nn.Module):
         raise NotImplementedError(
-            "LayerwiseCastingHook does not support deinitalization. A model once enabled with layerwise casting will "
+            "LayerwiseCastingHook does not support deinitialization. A model once enabled with layerwise casting will "
             "have casted its weights to a lower precision dtype for storage. Casting this back to the original dtype "
             "will lead to precision loss, which might have an impact on the model's generation quality. The model should "
             "be re-initialized and loaded in the original dtype."
@@ -90,7 +90,7 @@ class PeftInputAutocastDisableHook(ModelHook):
           that the inputs are casted to the computation dtype correctly always. However, there are two goals we are
           hoping to achieve:
             1. Making forward implementations independent of device/dtype casting operations as much as possible.
-            2. Peforming inference without losing information from casting to different precisions. With the current
+            2. Performing inference without losing information from casting to different precisions. With the current
                PEFT implementation (as linked in the reference above), and assuming running layerwise casting inference
                with storage_dtype=torch.float8_e4m3fn and compute_dtype=torch.bfloat16, inputs are cast to
                torch.float8_e4m3fn in the lora layer. We will then upcast back to torch.bfloat16 when we continue the

diffusers/hooks/pyramid_attention_broadcast.py

@@ -1,4 +1,4 @@
-# Copyright 2024 The HuggingFace Team. All rights reserved.
+# Copyright 2025 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.

diffusers/image_processor.py

@@ -1,4 +1,4 @@
-# Copyright 2024 The HuggingFace Team. All rights reserved.
+# Copyright 2025 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -116,6 +116,7 @@ class VaeImageProcessor(ConfigMixin):
         vae_scale_factor: int = 8,
         vae_latent_channels: int = 4,
         resample: str = "lanczos",
+        reducing_gap: int = None,
         do_normalize: bool = True,
         do_binarize: bool = False,
         do_convert_rgb: bool = False,
@@ -498,7 +499,11 @@ class VaeImageProcessor(ConfigMixin):
             raise ValueError(f"Only PIL image input is supported for resize_mode {resize_mode}")
         if isinstance(image, PIL.Image.Image):
             if resize_mode == "default":
-                image = image.resize((width, height), resample=PIL_INTERPOLATION[self.config.resample])
+                image = image.resize(
+                    (width, height),
+                    resample=PIL_INTERPOLATION[self.config.resample],
+                    reducing_gap=self.config.reducing_gap,
+                )
             elif resize_mode == "fill":
                 image = self._resize_and_fill(image, width, height)
             elif resize_mode == "crop":

diffusers/loaders/__init__.py

@@ -65,6 +65,7 @@ if is_torch_available():
             "AmusedLoraLoaderMixin",
             "StableDiffusionLoraLoaderMixin",
             "SD3LoraLoaderMixin",
+            "AuraFlowLoraLoaderMixin",
             "StableDiffusionXLLoraLoaderMixin",
             "LTXVideoLoraLoaderMixin",
             "LoraLoaderMixin",
@@ -76,6 +77,7 @@ if is_torch_available():
             "SanaLoraLoaderMixin",
             "Lumina2LoraLoaderMixin",
             "WanLoraLoaderMixin",
+            "HiDreamImageLoraLoaderMixin",
         ]
         _import_structure["textual_inversion"] = ["TextualInversionLoaderMixin"]
         _import_structure["ip_adapter"] = [
@@ -103,9 +105,11 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             )
             from .lora_pipeline import (
                 AmusedLoraLoaderMixin,
+                AuraFlowLoraLoaderMixin,
                 CogVideoXLoraLoaderMixin,
                 CogView4LoraLoaderMixin,
                 FluxLoraLoaderMixin,
+                HiDreamImageLoraLoaderMixin,
                 HunyuanVideoLoraLoaderMixin,
                 LoraLoaderMixin,
                 LTXVideoLoraLoaderMixin,

diffusers/loaders/ip_adapter.py

@@ -1,4 +1,4 @@
-# Copyright 2024 The HuggingFace Team. All rights reserved.
+# Copyright 2025 The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -159,10 +159,7 @@ class IPAdapterMixin:
                 " `low_cpu_mem_usage=False`."
             )
 
-        user_agent = {
-            "file_type": "attn_procs_weights",
-            "framework": "pytorch",
-        }
+        user_agent = {"file_type": "attn_procs_weights", "framework": "pytorch"}
         state_dicts = []
         for pretrained_model_name_or_path_or_dict, weight_name, subfolder in zip(
             pretrained_model_name_or_path_or_dict, weight_name, subfolder
@@ -465,10 +462,7 @@ class FluxIPAdapterMixin:
                 " `low_cpu_mem_usage=False`."
             )
 
-        user_agent = {
-            "file_type": "attn_procs_weights",
-            "framework": "pytorch",
-        }
+        user_agent = {"file_type": "attn_procs_weights", "framework": "pytorch"}
         state_dicts = []
         for pretrained_model_name_or_path_or_dict, weight_name, subfolder in zip(
             pretrained_model_name_or_path_or_dict, weight_name, subfolder
@@ -526,7 +520,7 @@ class FluxIPAdapterMixin:
                                 low_cpu_mem_usage=low_cpu_mem_usage,
                                 cache_dir=cache_dir,
                                 local_files_only=local_files_only,
-                                dtype=image_encoder_dtype,
+                                torch_dtype=image_encoder_dtype,
                             )
                             .to(self.device)
                             .eval()
@@ -750,10 +744,7 @@ class SD3IPAdapterMixin:
                 " `low_cpu_mem_usage=False`."
             )
 
-        user_agent = {
-            "file_type": "attn_procs_weights",
-            "framework": "pytorch",
-        }
+        user_agent = {"file_type": "attn_procs_weights", "framework": "pytorch"}
 
         if not isinstance(pretrained_model_name_or_path_or_dict, dict):
             model_file = _get_model_file(