Introduce DeprecatedPipelineMixin to simplify pipeline deprecation process (#11596)

* update

* update

* update

* update

* update

* update

* update

Author: DN6

src/diffusers/pipelines/amused/pipeline_amused.py

@@ -21,7 +21,7 @@
 from ...models import UVit2DModel, VQModel
 from ...schedulers import AmusedScheduler
 from ...utils import is_torch_xla_available, replace_example_docstring
-from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
+from ..pipeline_utils import DeprecatedPipelineMixin, DiffusionPipeline, ImagePipelineOutput
 
 
 if is_torch_xla_available():
@@ -47,7 +47,8 @@
 """
 
 
-class AmusedPipeline(DiffusionPipeline):
+class AmusedPipeline(DeprecatedPipelineMixin, DiffusionPipeline):
+    _last_supported_version = "0.33.1"
     image_processor: VaeImageProcessor
     vqvae: VQModel
     tokenizer: CLIPTokenizer

src/diffusers/pipelines/amused/pipeline_amused_img2img.py

@@ -21,7 +21,7 @@
 from ...models import UVit2DModel, VQModel
 from ...schedulers import AmusedScheduler
 from ...utils import is_torch_xla_available, replace_example_docstring
-from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
+from ..pipeline_utils import DeprecatedPipelineMixin, DiffusionPipeline, ImagePipelineOutput
 
 
 if is_torch_xla_available():
@@ -57,7 +57,8 @@
 """
 
 
-class AmusedImg2ImgPipeline(DiffusionPipeline):
+class AmusedImg2ImgPipeline(DeprecatedPipelineMixin, DiffusionPipeline):
+    _last_supported_version = "0.33.1"
     image_processor: VaeImageProcessor
     vqvae: VQModel
     tokenizer: CLIPTokenizer

src/diffusers/pipelines/amused/pipeline_amused_inpaint.py

@@ -22,7 +22,7 @@
 from ...models import UVit2DModel, VQModel
 from ...schedulers import AmusedScheduler
 from ...utils import is_torch_xla_available, replace_example_docstring
-from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
+from ..pipeline_utils import DeprecatedPipelineMixin, DiffusionPipeline, ImagePipelineOutput
 
 
 if is_torch_xla_available():
@@ -65,7 +65,8 @@
 """
 
 
-class AmusedInpaintPipeline(DiffusionPipeline):
+class AmusedInpaintPipeline(DeprecatedPipelineMixin, DiffusionPipeline):
+    _last_supported_version = "0.33.1"
     image_processor: VaeImageProcessor
     vqvae: VQModel
     tokenizer: CLIPTokenizer

src/diffusers/pipelines/audioldm/pipeline_audioldm.py

@@ -24,7 +24,7 @@
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import AudioPipelineOutput, DiffusionPipeline, StableDiffusionMixin
+from ..pipeline_utils import AudioPipelineOutput, DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin
 
 
 if is_torch_xla_available():
@@ -57,7 +57,7 @@
 """
 
 
-class AudioLDMPipeline(DiffusionPipeline, StableDiffusionMixin):
+class AudioLDMPipeline(DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin):
     r"""
     Pipeline for text-to-audio generation using AudioLDM.
 
@@ -81,6 +81,7 @@ class AudioLDMPipeline(DiffusionPipeline, StableDiffusionMixin):
             Vocoder of class `SpeechT5HifiGan`.
     """
 
+    _last_supported_version = "0.33.1"
     model_cpu_offload_seq = "text_encoder->unet->vae"
 
     def __init__(

src/diffusers/pipelines/blip_diffusion/pipeline_blip_diffusion.py

@@ -25,7 +25,7 @@
     replace_example_docstring,
 )
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
+from ..pipeline_utils import DeprecatedPipelineMixin, DiffusionPipeline, ImagePipelineOutput
 from .blip_image_processing import BlipImageProcessor
 from .modeling_blip2 import Blip2QFormerModel
 from .modeling_ctx_clip import ContextCLIPTextModel
@@ -81,7 +81,7 @@
 """
 
 
-class BlipDiffusionPipeline(DiffusionPipeline):
+class BlipDiffusionPipeline(DeprecatedPipelineMixin, DiffusionPipeline):
     """
     Pipeline for Zero-Shot Subject Driven Generation using Blip Diffusion.
 
@@ -107,6 +107,7 @@ class BlipDiffusionPipeline(DiffusionPipeline):
             Position of the context token in the text encoder.
     """
 
+    _last_supported_version = "0.33.1"
     model_cpu_offload_seq = "qformer->text_encoder->unet->vae"
 
     def __init__(

src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs.py

@@ -37,7 +37,7 @@
     unscale_lora_layers,
 )
 from ...utils.torch_utils import is_compiled_module, is_torch_version, randn_tensor
-from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
+from ..pipeline_utils import DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin
 from ..stable_diffusion.pipeline_output import StableDiffusionPipelineOutput
 from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 
@@ -98,6 +98,7 @@
 
 
 class StableDiffusionControlNetXSPipeline(
+    DeprecatedPipelineMixin,
     DiffusionPipeline,
     StableDiffusionMixin,
     TextualInversionLoaderMixin,
@@ -138,6 +139,7 @@ class StableDiffusionControlNetXSPipeline(
             A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
     """
 
+    _last_supported_version = "0.33.1"
     model_cpu_offload_seq = "text_encoder->unet->vae"
     _optional_components = ["safety_checker", "feature_extractor"]
     _exclude_from_cpu_offload = ["safety_checker"]

src/diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs_sd_xl.py

@@ -46,7 +46,7 @@
     unscale_lora_layers,
 )
 from ...utils.torch_utils import is_compiled_module, is_torch_version, randn_tensor
-from ..pipeline_utils import DiffusionPipeline
+from ..pipeline_utils import DeprecatedPipelineMixin, DiffusionPipeline
 from ..stable_diffusion_xl.pipeline_output import StableDiffusionXLPipelineOutput
 
 
@@ -114,6 +114,7 @@
 
 
 class StableDiffusionXLControlNetXSPipeline(
+    DeprecatedPipelineMixin,
     DiffusionPipeline,
     TextualInversionLoaderMixin,
     StableDiffusionXLLoraLoaderMixin,
@@ -158,6 +159,7 @@ class StableDiffusionXLControlNetXSPipeline(
             watermarker is used.
     """
 
+    _last_supported_version = "0.33.1"
     model_cpu_offload_seq = "text_encoder->text_encoder_2->unet->vae"
     _optional_components = [
         "tokenizer",

src/diffusers/pipelines/dance_diffusion/pipeline_dance_diffusion.py

@@ -21,7 +21,7 @@
 from ...schedulers import SchedulerMixin
 from ...utils import is_torch_xla_available, logging
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import AudioPipelineOutput, DiffusionPipeline
+from ..pipeline_utils import AudioPipelineOutput, DeprecatedPipelineMixin, DiffusionPipeline
 
 
 if is_torch_xla_available():
@@ -34,7 +34,7 @@
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-class DanceDiffusionPipeline(DiffusionPipeline):
+class DanceDiffusionPipeline(DeprecatedPipelineMixin, DiffusionPipeline):
     r"""
     Pipeline for audio generation.
 
@@ -49,6 +49,7 @@ class DanceDiffusionPipeline(DiffusionPipeline):
             [`IPNDMScheduler`].
     """
 
+    _last_supported_version = "0.33.1"
     model_cpu_offload_seq = "unet"
 
     def __init__(self, unet: UNet1DModel, scheduler: SchedulerMixin):

src/diffusers/pipelines/i2vgen_xl/pipeline_i2vgen_xl.py

@@ -33,7 +33,7 @@
 )
 from ...utils.torch_utils import randn_tensor
 from ...video_processor import VideoProcessor
-from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
+from ..pipeline_utils import DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin
 
 
 if is_torch_xla_available():
@@ -97,9 +97,11 @@ class I2VGenXLPipelineOutput(BaseOutput):
 
 
 class I2VGenXLPipeline(
+    DeprecatedPipelineMixin,
     DiffusionPipeline,
     StableDiffusionMixin,
 ):
+    _last_supported_version = "0.33.1"
     r"""
     Pipeline for image-to-video generation as proposed in [I2VGenXL](https://i2vgen-xl.github.io/).
 

src/diffusers/pipelines/musicldm/pipeline_musicldm.py

@@ -36,7 +36,7 @@
     replace_example_docstring,
 )
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import AudioPipelineOutput, DiffusionPipeline, StableDiffusionMixin
+from ..pipeline_utils import AudioPipelineOutput, DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin
 
 
 if is_librosa_available():
@@ -76,7 +76,8 @@
 """
 
 
-class MusicLDMPipeline(DiffusionPipeline, StableDiffusionMixin):
+class MusicLDMPipeline(DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin):
+    _last_supported_version = "0.33.1"
     r"""
     Pipeline for text-to-audio generation using MusicLDM.
 

src/diffusers/pipelines/paint_by_example/pipeline_paint_by_example.py

@@ -25,7 +25,7 @@
 from ...schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
 from ...utils import deprecate, is_torch_xla_available, logging
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
+from ..pipeline_utils import DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin
 from ..stable_diffusion import StableDiffusionPipelineOutput
 from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 from .image_encoder import PaintByExampleImageEncoder
@@ -155,7 +155,8 @@ def prepare_mask_and_masked_image(image, mask):
     return mask, masked_image
 
 
-class PaintByExamplePipeline(DiffusionPipeline, StableDiffusionMixin):
+class PaintByExamplePipeline(DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin):
+    _last_supported_version = "0.33.1"
     r"""
     <Tip warning={true}>
 

src/diffusers/pipelines/pia/pipeline_pia.py

@@ -46,7 +46,7 @@
 from ...utils.torch_utils import randn_tensor
 from ...video_processor import VideoProcessor
 from ..free_init_utils import FreeInitMixin
-from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
+from ..pipeline_utils import DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin
 
 
 if is_torch_xla_available():
@@ -132,6 +132,7 @@ class PIAPipelineOutput(BaseOutput):
 
 
 class PIAPipeline(
+    DeprecatedPipelineMixin,
     DiffusionPipeline,
     StableDiffusionMixin,
     TextualInversionLoaderMixin,
@@ -140,6 +141,7 @@ class PIAPipeline(
     FromSingleFileMixin,
     FreeInitMixin,
 ):
+    _last_supported_version = "0.33.1"
     r"""
     Pipeline for text-to-video generation.
 

src/diffusers/pipelines/pipeline_utils.py

@@ -139,6 +139,43 @@ class AudioPipelineOutput(BaseOutput):
     audios: np.ndarray
 
 
+class DeprecatedPipelineMixin:
+    """
+    A mixin that can be used to mark a pipeline as deprecated.
+
+    Pipelines inheriting from this mixin will raise a warning when instantiated, indicating that they are deprecated
+    and won't receive updates past the specified version. Tests will be skipped for pipelines that inherit from this
+    mixin.
+
+    Example usage:
+    ```python
+    class MyDeprecatedPipeline(DeprecatedPipelineMixin, DiffusionPipeline):
+        _last_supported_version = "0.20.0"
+
+        def __init__(self, *args, **kwargs):
+            super().__init__(*args, **kwargs)
+    ```
+    """
+
+    # Override this in the inheriting class to specify the last version that will support this pipeline
+    _last_supported_version = None
+
+    def __init__(self, *args, **kwargs):
+        # Get the class name for the warning message
+        class_name = self.__class__.__name__
+
+        # Get the last supported version or use the current version if not specified
+        version_info = getattr(self.__class__, "_last_supported_version", __version__)
+
+        # Raise a warning that this pipeline is deprecated
+        logger.warning(
+            f"The {class_name} has been deprecated and will not receive bug fixes or feature updates after Diffusers version {version_info}. "
+        )
+
+        # Call the parent class's __init__ method
+        super().__init__(*args, **kwargs)
+
+
 class DiffusionPipeline(ConfigMixin, PushToHubMixin):
     r"""
     Base class for all pipelines.

src/diffusers/pipelines/semantic_stable_diffusion/pipeline_semantic_stable_diffusion.py

@@ -11,7 +11,7 @@
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import deprecate, is_torch_xla_available, logging
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
+from ..pipeline_utils import DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin
 from .pipeline_output import SemanticStableDiffusionPipelineOutput
 
 
@@ -25,7 +25,8 @@
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
-class SemanticStableDiffusionPipeline(DiffusionPipeline, StableDiffusionMixin):
+class SemanticStableDiffusionPipeline(DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin):
+    _last_supported_version = "0.33.1"
     r"""
     Pipeline for text-to-image generation using Stable Diffusion with latent editing.
 

src/diffusers/pipelines/stable_diffusion_attend_and_excite/pipeline_stable_diffusion_attend_and_excite.py

@@ -37,7 +37,7 @@
     unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
+from ..pipeline_utils import DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin
 from ..stable_diffusion import StableDiffusionPipelineOutput
 from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 
@@ -179,7 +179,9 @@ def __call__(self, attn: Attention, hidden_states, encoder_hidden_states=None, a
         return hidden_states
 
 
-class StableDiffusionAttendAndExcitePipeline(DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin):
+class StableDiffusionAttendAndExcitePipeline(
+    DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin
+):
     r"""
     Pipeline for text-to-image generation using Stable Diffusion and Attend-and-Excite.
 
@@ -209,6 +211,8 @@ class StableDiffusionAttendAndExcitePipeline(DiffusionPipeline, StableDiffusionM
             A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
     """
 
+    _last_supported_version = "0.33.1"
+
     model_cpu_offload_seq = "text_encoder->unet->vae"
     _optional_components = ["safety_checker", "feature_extractor"]
     _exclude_from_cpu_offload = ["safety_checker"]

src/diffusers/pipelines/stable_diffusion_diffedit/pipeline_stable_diffusion_diffedit.py

@@ -40,7 +40,7 @@
     unscale_lora_layers,
 )
 from ...utils.torch_utils import randn_tensor
-from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
+from ..pipeline_utils import DeprecatedPipelineMixin, DiffusionPipeline, StableDiffusionMixin
 from ..stable_diffusion import StableDiffusionPipelineOutput
 from ..stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 
@@ -242,7 +242,11 @@ def preprocess_mask(mask, batch_size: int = 1):
 
 
 class StableDiffusionDiffEditPipeline(
-    DiffusionPipeline, StableDiffusionMixin, TextualInversionLoaderMixin, StableDiffusionLoraLoaderMixin
+    DeprecatedPipelineMixin,
+    DiffusionPipeline,
+    StableDiffusionMixin,
+    TextualInversionLoaderMixin,
+    StableDiffusionLoraLoaderMixin,
 ):
     r"""
     <Tip warning={true}>
@@ -282,6 +286,8 @@ class StableDiffusionDiffEditPipeline(
             A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.
     """
 
+    _last_supported_version = "0.33.1"
+
     model_cpu_offload_seq = "text_encoder->unet->vae"
     _optional_components = ["safety_checker", "feature_extractor", "inverse_scheduler"]
     _exclude_from_cpu_offload = ["safety_checker"]