up

* up

* style + copies

* fix

---------

Co-authored-by: yiyi@huggingface.co <yiyi@ip-26-0-160-103.ec2.internal>

src/diffusers/__init__.py

@@ -294,10 +294,17 @@ else:
     )
     _import_structure["modular_pipelines"].extend(
         [
+            "AutoPipelineBlocks",
             "ComponentsManager",
             "ComponentSpec",
+            "ConditionalPipelineBlocks",
+            "ConfigSpec",
+            "InputParam",
+            "LoopSequentialPipelineBlocks",
             "ModularPipeline",
             "ModularPipelineBlocks",
+            "OutputParam",
+            "SequentialPipelineBlocks",
         ]
     )
     _import_structure["optimization"] = [
@@ -1063,7 +1070,19 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             ZImageTransformer2DModel,
             attention_backend,
         )
-        from .modular_pipelines import ComponentsManager, ComponentSpec, ModularPipeline, ModularPipelineBlocks
+        from .modular_pipelines import (
+            AutoPipelineBlocks,
+            ComponentsManager,
+            ComponentSpec,
+            ConditionalPipelineBlocks,
+            ConfigSpec,
+            InputParam,
+            LoopSequentialPipelineBlocks,
+            ModularPipeline,
+            ModularPipelineBlocks,
+            OutputParam,
+            SequentialPipelineBlocks,
+        )
         from .optimization import (
             get_constant_schedule,
             get_constant_schedule_with_warmup,

src/diffusers/modular_pipelines/__init__.py

@@ -33,6 +33,7 @@ else:
         "ModularPipeline",
         "AutoPipelineBlocks",
         "SequentialPipelineBlocks",
+        "ConditionalPipelineBlocks",
         "LoopSequentialPipelineBlocks",
         "PipelineState",
         "BlockState",
@@ -105,6 +106,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
         from .modular_pipeline import (
             AutoPipelineBlocks,
             BlockState,
+            ConditionalPipelineBlocks,
             LoopSequentialPipelineBlocks,
             ModularPipeline,
             ModularPipelineBlocks,

src/diffusers/pipelines/pag/pipeline_pag_hunyuandit.py

@@ -17,7 +17,7 @@ from typing import Callable
 
 import numpy as np
 import torch
-from transformers import BertModel, BertTokenizer, CLIPImageProcessor, MT5Tokenizer, T5EncoderModel
+from transformers import BertModel, BertTokenizer, CLIPImageProcessor, T5EncoderModel, T5Tokenizer
 
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
 
@@ -208,7 +208,7 @@ class HunyuanDiTPAGPipeline(DiffusionPipeline, PAGMixin):
         feature_extractor: CLIPImageProcessor | None = None,
         requires_safety_checker: bool = True,
         text_encoder_2: T5EncoderModel | None = None,
-        tokenizer_2: MT5Tokenizer | None = None,
+        tokenizer_2: T5Tokenizer | None = None,
         pag_applied_layers: str | list[str] = "blocks.1",  # "blocks.16.attn1", "blocks.16", "16", 16
     ):
         super().__init__()

src/diffusers/utils/dummy_pt_objects.py

@@ -1905,6 +1905,21 @@ def attention_backend(*args, **kwargs):
     requires_backends(attention_backend, ["torch"])
 
 
+class AutoPipelineBlocks(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+    @classmethod
+    def from_config(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+
 class ComponentsManager(metaclass=DummyObject):
     _backends = ["torch"]
 
@@ -1935,6 +1950,66 @@ class ComponentSpec(metaclass=DummyObject):
         requires_backends(cls, ["torch"])
 
 
+class ConditionalPipelineBlocks(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+    @classmethod
+    def from_config(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+
+class ConfigSpec(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+    @classmethod
+    def from_config(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+
+class InputParam(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+    @classmethod
+    def from_config(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+
+class LoopSequentialPipelineBlocks(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+    @classmethod
+    def from_config(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+
 class ModularPipeline(metaclass=DummyObject):
     _backends = ["torch"]
 
@@ -1965,6 +2040,36 @@ class ModularPipelineBlocks(metaclass=DummyObject):
         requires_backends(cls, ["torch"])
 
 
+class OutputParam(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+    @classmethod
+    def from_config(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+
+class SequentialPipelineBlocks(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+    @classmethod
+    def from_config(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+
 def get_constant_schedule(*args, **kwargs):
     requires_backends(get_constant_schedule, ["torch"])