Merge branch 'main' into wan22-lightx2v

Author: linoytsaban

.github/workflows/nightly_tests.yml

@@ -333,7 +333,7 @@ jobs:
             additional_deps: ["peft"]
           - backend: "gguf"
             test_location: "gguf"
-            additional_deps: ["peft"]
+            additional_deps: ["peft", "kernels"]
           - backend: "torchao"
             test_location: "torchao"
             additional_deps: []

docs/source/en/quantization/gguf.md

@@ -53,6 +53,16 @@ image = pipe(prompt, generator=torch.manual_seed(0)).images[0]
 image.save("flux-gguf.png")
 ```
 
+## Using Optimized CUDA Kernels with GGUF
+
+Optimized CUDA kernels can accelerate GGUF quantized model inference by approximately 10%. This functionality requires a compatible GPU with `torch.cuda.get_device_capability` greater than 7 and the kernels library:
+
+```shell
+pip install -U kernels
+```
+
+Once installed, set `DIFFUSERS_GGUF_CUDA_KERNELS=true`  to use optimized kernels when available. Note that CUDA kernels may introduce minor numerical differences compared to the original GGUF implementation, potentially causing subtle visual variations in generated images. To disable CUDA kernel usage, set the environment variable `DIFFUSERS_GGUF_CUDA_KERNELS=false`.
+
 ## Supported Quantization Types
 
 - BF16

src/diffusers/__init__.py

@@ -139,6 +139,7 @@
             "AutoGuidance",
             "ClassifierFreeGuidance",
             "ClassifierFreeZeroStarGuidance",
+            "FrequencyDecoupledGuidance",
             "PerturbedAttentionGuidance",
             "SkipLayerGuidance",
             "SmoothedEnergyGuidance",
@@ -804,6 +805,7 @@
             AutoGuidance,
             ClassifierFreeGuidance,
             ClassifierFreeZeroStarGuidance,
+            FrequencyDecoupledGuidance,
             PerturbedAttentionGuidance,
             SkipLayerGuidance,
             SmoothedEnergyGuidance,

src/diffusers/guiders/__init__.py

@@ -22,6 +22,7 @@
     from .auto_guidance import AutoGuidance
     from .classifier_free_guidance import ClassifierFreeGuidance
     from .classifier_free_zero_star_guidance import ClassifierFreeZeroStarGuidance
+    from .frequency_decoupled_guidance import FrequencyDecoupledGuidance
     from .perturbed_attention_guidance import PerturbedAttentionGuidance
     from .skip_layer_guidance import SkipLayerGuidance
     from .smoothed_energy_guidance import SmoothedEnergyGuidance
@@ -32,6 +33,7 @@
         AutoGuidance,
         ClassifierFreeGuidance,
         ClassifierFreeZeroStarGuidance,
+        FrequencyDecoupledGuidance,
         PerturbedAttentionGuidance,
         SkipLayerGuidance,
         SmoothedEnergyGuidance,

src/diffusers/guiders/frequency_decoupled_guidance.py

@@ -133,6 +133,7 @@ def _register_attention_processors_metadata():
             skip_processor_output_fn=_skip_proc_output_fn_Attention_WanAttnProcessor2_0,
         ),
     )
+
     # FluxAttnProcessor
     AttentionProcessorRegistry.register(
         model_class=FluxAttnProcessor,

src/diffusers/hooks/_helpers.py

@@ -0,0 +1,43 @@
+# 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.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import torch
+
+from ._common import _ALL_TRANSFORMER_BLOCK_IDENTIFIERS, _ATTENTION_CLASSES, _FEEDFORWARD_CLASSES
+
+
+def _get_identifiable_transformer_blocks_in_module(module: torch.nn.Module):
+    module_list_with_transformer_blocks = []
+    for name, submodule in module.named_modules():
+        name_endswith_identifier = any(name.endswith(identifier) for identifier in _ALL_TRANSFORMER_BLOCK_IDENTIFIERS)
+        is_modulelist = isinstance(submodule, torch.nn.ModuleList)
+        if name_endswith_identifier and is_modulelist:
+            module_list_with_transformer_blocks.append((name, submodule))
+    return module_list_with_transformer_blocks
+
+
+def _get_identifiable_attention_layers_in_module(module: torch.nn.Module):
+    attention_layers = []
+    for name, submodule in module.named_modules():
+        if isinstance(submodule, _ATTENTION_CLASSES):
+            attention_layers.append((name, submodule))
+    return attention_layers
+
+
+def _get_identifiable_feedforward_layers_in_module(module: torch.nn.Module):
+    feedforward_layers = []
+    for name, submodule in module.named_modules():
+        if isinstance(submodule, _FEEDFORWARD_CLASSES):
+            feedforward_layers.append((name, submodule))
+    return feedforward_layers

src/diffusers/hooks/group_offloading.py

@@ -310,7 +310,7 @@ def _get_clip_prompt_embeds(
     def encode_prompt(
         self,
         prompt: Union[str, List[str]],
-        prompt_2: Union[str, List[str]],
+        prompt_2: Optional[Union[str, List[str]]] = None,
         device: Optional[torch.device] = None,
         num_images_per_prompt: int = 1,
         prompt_embeds: Optional[torch.FloatTensor] = None,

src/diffusers/hooks/utils.py

@@ -324,7 +324,7 @@ def _get_clip_prompt_embeds(
     def encode_prompt(
         self,
         prompt: Union[str, List[str]],
-        prompt_2: Union[str, List[str]],
+        prompt_2: Optional[Union[str, List[str]]] = None,
         device: Optional[torch.device] = None,
         num_images_per_prompt: int = 1,
         prompt_embeds: Optional[torch.FloatTensor] = None,