[Imports] Fix many import bugs and make sure that doc builder CI test works correctly (#5176)

* [Doc builder] Ensure slow import for doc builder

* Apply suggestions from code review

* env for doc builder

* fix more

* [Diffusers] Set import to slow as env variable

* fix docs

* fix docs

* Apply suggestions from code review

* Apply suggestions from code review

* fix docs

* fix docs

docs/source/en/api/logging.md

@@ -67,30 +67,30 @@ By default, `tqdm` progress bars are displayed during model download. [`logging.
 
 ## Base setters
 
-[[autodoc]] logging.set_verbosity_error
+[[autodoc]] utils.logging.set_verbosity_error
 
-[[autodoc]] logging.set_verbosity_warning
+[[autodoc]] utils.logging.set_verbosity_warning
 
-[[autodoc]] logging.set_verbosity_info
+[[autodoc]] utils.logging.set_verbosity_info
 
-[[autodoc]] logging.set_verbosity_debug
+[[autodoc]] utils.logging.set_verbosity_debug
 
 ## Other functions
 
-[[autodoc]] logging.get_verbosity
+[[autodoc]] utils.logging.get_verbosity
 
-[[autodoc]] logging.set_verbosity
+[[autodoc]] utils.logging.set_verbosity
 
-[[autodoc]] logging.get_logger
+[[autodoc]] utils.logging.get_logger
 
-[[autodoc]] logging.enable_default_handler
+[[autodoc]] utils.logging.enable_default_handler
 
-[[autodoc]] logging.disable_default_handler
+[[autodoc]] utils.logging.disable_default_handler
 
-[[autodoc]] logging.enable_explicit_format
+[[autodoc]] utils.logging.enable_explicit_format
 
-[[autodoc]] logging.reset_format
+[[autodoc]] utils.logging.reset_format
 
-[[autodoc]] logging.enable_progress_bar
+[[autodoc]] utils.logging.enable_progress_bar
 
-[[autodoc]] logging.disable_progress_bar
+[[autodoc]] utils.logging.disable_progress_bar

docs/source/en/api/pipelines/pix2pix.md

@@ -34,13 +34,7 @@ Make sure to check out the Schedulers [guide](/using-diffusers/schedulers) to le
 	- load_lora_weights
 	- save_lora_weights
 
-## StableDiffusionPipelineOutput
-[[autodoc]] pipelines.stable_diffusion.StableDiffusionPipelineOutput
-
 ## StableDiffusionXLInstructPix2PixPipeline
 [[autodoc]] StableDiffusionXLInstructPix2PixPipeline
 	- __call__
 	- all
-
-## StableDiffusionXLPipelineOutput
-[[autodoc]] pipelines.stable_diffusion_xl.StableDiffusionXLPipelineOutput
\ No newline at end of file

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

@@ -31,5 +31,5 @@ Make sure to check out the Schedulers [guide](/using-diffusers/schedulers) to le
 	- __call__
 
 ## StableDiffusionSafePipelineOutput
-[[autodoc]] pipelines.semantic_stable_diffusion.SemanticStableDiffusionPipelineOutput
-	- all
\ No newline at end of file
+[[autodoc]] pipelines.semantic_stable_diffusion.pipeline_output.SemanticStableDiffusionPipelineOutput
+	- all

src/diffusers/__init__.py

@@ -3,6 +3,7 @@ __version__ = "0.22.0.dev0"
 from typing import TYPE_CHECKING
 
 from .utils import (
+    DIFFUSERS_SLOW_IMPORT,
     OptionalDependencyNotAvailable,
     _LazyModule,
     is_flax_available,
@@ -414,7 +415,7 @@ except OptionalDependencyNotAvailable:
 else:
     _import_structure["pipelines"].extend(["MidiProcessor"])
 
-if TYPE_CHECKING:
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     from .configuration_utils import ConfigMixin
 
     try:

src/diffusers/models/__init__.py

@@ -14,7 +14,7 @@
 
 from typing import TYPE_CHECKING
 
-from ..utils import _LazyModule, is_flax_available, is_torch_available
+from ..utils import DIFFUSERS_SLOW_IMPORT, _LazyModule, is_flax_available, is_torch_available
 
 
 _import_structure = {}
@@ -43,7 +43,7 @@ if is_flax_available():
     _import_structure["vae_flax"] = ["FlaxAutoencoderKL"]
 
 
-if TYPE_CHECKING:
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     if is_torch_available():
         from .adapter import MultiAdapter, T2IAdapter
         from .autoencoder_asym_kl import AsymmetricAutoencoderKL

src/diffusers/pipelines/__init__.py

 from typing import TYPE_CHECKING
 
 from ..utils import (
+    DIFFUSERS_SLOW_IMPORT,
     OptionalDependencyNotAvailable,
     _LazyModule,
     get_objects_from_module,
@@ -252,7 +253,7 @@ except OptionalDependencyNotAvailable:
 else:
     _import_structure["spectrogram_diffusion"] = ["MidiProcessor", "SpectrogramDiffusionPipeline"]
 
-if TYPE_CHECKING:
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     try:
         if not is_torch_available():
             raise OptionalDependencyNotAvailable()

src/diffusers/pipelines/alt_diffusion/__init__.py

 from typing import TYPE_CHECKING
 
 from ...utils import (
+    DIFFUSERS_SLOW_IMPORT,
     OptionalDependencyNotAvailable,
     _LazyModule,
     get_objects_from_module,
@@ -26,7 +27,7 @@ else:
 
     _import_structure["pipeline_output"] = ["AltDiffusionPipelineOutput"]
 
-if TYPE_CHECKING:
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     try:
         if not (is_transformers_available() and is_torch_available()):
             raise OptionalDependencyNotAvailable()

src/diffusers/pipelines/alt_diffusion/pipeline_alt_diffusion.py

@@ -29,7 +29,8 @@ from ...utils import deprecate, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
-from . import AltDiffusionPipelineOutput, RobertaSeriesModelWithTransformation
+from .modeling_roberta_series import RobertaSeriesModelWithTransformation
+from .pipeline_output import AltDiffusionPipelineOutput
 
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name

src/diffusers/pipelines/alt_diffusion/pipeline_alt_diffusion_img2img.py

@@ -31,7 +31,8 @@ from ...utils import PIL_INTERPOLATION, deprecate, logging, replace_example_docs
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
-from . import AltDiffusionPipelineOutput, RobertaSeriesModelWithTransformation
+from .modeling_roberta_series import RobertaSeriesModelWithTransformation
+from .pipeline_output import AltDiffusionPipelineOutput
 
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name

src/diffusers/pipelines/audio_diffusion/__init__.py

 from typing import TYPE_CHECKING
 
-from ...utils import _LazyModule
+from ...utils import DIFFUSERS_SLOW_IMPORT, _LazyModule
 
 
 _import_structure = {
@@ -8,7 +8,7 @@ _import_structure = {
     "pipeline_audio_diffusion": ["AudioDiffusionPipeline"],
 }
 
-if TYPE_CHECKING:
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     from .mel import Mel
     from .pipeline_audio_diffusion import AudioDiffusionPipeline
 

src/diffusers/pipelines/audioldm/__init__.py

 from typing import TYPE_CHECKING
 
 from ...utils import (
+    DIFFUSERS_SLOW_IMPORT,
     OptionalDependencyNotAvailable,
     _LazyModule,
     is_torch_available,
@@ -25,7 +26,7 @@ else:
     _import_structure["pipeline_audioldm"] = ["AudioLDMPipeline"]
 
 
-if TYPE_CHECKING:
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     try:
         if not (is_transformers_available() and is_torch_available() and is_transformers_version(">=", "4.27.0")):
             raise OptionalDependencyNotAvailable()

src/diffusers/pipelines/audioldm2/__init__.py

 from typing import TYPE_CHECKING
 
 from ...utils import (
+    DIFFUSERS_SLOW_IMPORT,
     OptionalDependencyNotAvailable,
     _LazyModule,
     get_objects_from_module,
@@ -25,7 +26,7 @@ else:
     _import_structure["pipeline_audioldm2"] = ["AudioLDM2Pipeline"]
 
 
-if TYPE_CHECKING:
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     try:
         if not (is_transformers_available() and is_torch_available() and is_transformers_version(">=", "4.27.0")):
             raise OptionalDependencyNotAvailable()

src/diffusers/pipelines/consistency_models/__init__.py

 from typing import TYPE_CHECKING
 
 from ...utils import (
+    DIFFUSERS_SLOW_IMPORT,
     _LazyModule,
 )
 
 
 _import_structure = {"pipeline_consistency_models": ["ConsistencyModelPipeline"]}
 
-if TYPE_CHECKING:
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     from .pipeline_consistency_models import ConsistencyModelPipeline
 
 else:

src/diffusers/pipelines/controlnet/__init__.py

 from typing import TYPE_CHECKING
 
 from ...utils import (
+    DIFFUSERS_SLOW_IMPORT,
     OptionalDependencyNotAvailable,
     _LazyModule,
     get_objects_from_module,
@@ -40,7 +41,7 @@ else:
     _import_structure["pipeline_flax_controlnet"] = ["FlaxStableDiffusionControlNetPipeline"]
 
 
-if TYPE_CHECKING:
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     try:
         if not (is_transformers_available() and is_torch_available()):
             raise OptionalDependencyNotAvailable()

src/diffusers/pipelines/controlnet/pipeline_controlnet.py

@@ -34,7 +34,7 @@ from ...utils import (
 )
 from ...utils.torch_utils import is_compiled_module, randn_tensor
 from ..pipeline_utils import DiffusionPipeline
-from ..stable_diffusion import StableDiffusionPipelineOutput
+from ..stable_diffusion.pipeline_output import StableDiffusionPipelineOutput
 from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 from .multicontrolnet import MultiControlNetModel
 

src/diffusers/pipelines/controlnet/pipeline_controlnet_inpaint_sd_xl.py

@@ -21,8 +21,6 @@ import torch
 import torch.nn.functional as F
 from transformers import CLIPTextModel, CLIPTextModelWithProjection, CLIPTokenizer
 
-from diffusers.pipelines.stable_diffusion_xl import StableDiffusionXLPipelineOutput
-
 from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...loaders import FromSingleFileMixin, StableDiffusionXLLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, ControlNetModel, UNet2DConditionModel
@@ -41,6 +39,7 @@ from ...utils import (
 )
 from ...utils.torch_utils import is_compiled_module, randn_tensor
 from ..pipeline_utils import DiffusionPipeline
+from ..stable_diffusion_xl.pipeline_output import StableDiffusionXLPipelineOutput
 from .multicontrolnet import MultiControlNetModel
 
 

src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl.py

@@ -41,7 +41,7 @@ from ...utils import (
 )
 from ...utils.torch_utils import is_compiled_module, randn_tensor
 from ..pipeline_utils import DiffusionPipeline
-from ..stable_diffusion_xl import StableDiffusionXLPipelineOutput
+from ..stable_diffusion_xl.pipeline_output import StableDiffusionXLPipelineOutput
 
 
 if is_invisible_watermark_available():

src/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl_img2img.py

@@ -41,7 +41,7 @@ from ...utils import (
 )
 from ...utils.torch_utils import is_compiled_module, randn_tensor
 from ..pipeline_utils import DiffusionPipeline
-from ..stable_diffusion_xl import StableDiffusionXLPipelineOutput
+from ..stable_diffusion_xl.pipeline_output import StableDiffusionXLPipelineOutput
 
 
 if is_invisible_watermark_available():

src/diffusers/pipelines/dance_diffusion/__init__.py

 from typing import TYPE_CHECKING
 
-from ...utils import _LazyModule
+from ...utils import DIFFUSERS_SLOW_IMPORT, _LazyModule
 
 
 _import_structure = {"pipeline_dance_diffusion": ["DanceDiffusionPipeline"]}
 
-if TYPE_CHECKING:
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     from .pipeline_dance_diffusion import DanceDiffusionPipeline
 else:
     import sys

src/diffusers/pipelines/ddim/__init__.py

 from typing import TYPE_CHECKING
 
-from ...utils import _LazyModule
+from ...utils import DIFFUSERS_SLOW_IMPORT, _LazyModule
 
 
 _import_structure = {"pipeline_ddim": ["DDIMPipeline"]}
 
-if TYPE_CHECKING:
+if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
     from .pipeline_ddim import DDIMPipeline
 else:
     import sys