Deprecate slicing and tiling methods from `DiffusionPipeline` (#12271)

* deprecate slicing from flux pipeline.

* propagate.

* tiling

* up

* up

src/diffusers/pipelines/ltx/pipeline_ltx_latent_upsample.py

@@ -18,7 +18,7 @@ import torch
 
 from ...image_processor import PipelineImageInput
 from ...models import AutoencoderKLLTXVideo
-from ...utils import get_logger
+from ...utils import deprecate, get_logger
 from ...utils.torch_utils import randn_tensor
 from ...video_processor import VideoProcessor
 from ..pipeline_utils import DiffusionPipeline
@@ -148,6 +148,12 @@ class LTXLatentUpsamplePipeline(DiffusionPipeline):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -155,6 +161,12 @@ class LTXLatentUpsamplePipeline(DiffusionPipeline):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -163,6 +175,12 @@ class LTXLatentUpsamplePipeline(DiffusionPipeline):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -170,6 +188,12 @@ class LTXLatentUpsamplePipeline(DiffusionPipeline):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     def check_inputs(self, video, height, width, latents):

src/diffusers/pipelines/lumina2/pipeline_lumina2.py

@@ -433,6 +433,12 @@ class Lumina2Pipeline(DiffusionPipeline, Lumina2LoraLoaderMixin):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -440,6 +446,12 @@ class Lumina2Pipeline(DiffusionPipeline, Lumina2LoraLoaderMixin):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -448,6 +460,12 @@ class Lumina2Pipeline(DiffusionPipeline, Lumina2LoraLoaderMixin):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -455,6 +473,12 @@ class Lumina2Pipeline(DiffusionPipeline, Lumina2LoraLoaderMixin):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):

src/diffusers/pipelines/mochi/pipeline_mochi.py

@@ -23,11 +23,7 @@ from ...callbacks import MultiPipelineCallbacks, PipelineCallback
 from ...loaders import Mochi1LoraLoaderMixin
 from ...models import AutoencoderKLMochi, MochiTransformer3DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import (
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-)
+from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ...video_processor import VideoProcessor
 from ..pipeline_utils import DiffusionPipeline
@@ -396,6 +392,12 @@ class MochiPipeline(DiffusionPipeline, Mochi1LoraLoaderMixin):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -403,6 +405,12 @@ class MochiPipeline(DiffusionPipeline, Mochi1LoraLoaderMixin):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -411,6 +419,12 @@ class MochiPipeline(DiffusionPipeline, Mochi1LoraLoaderMixin):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -418,6 +432,12 @@ class MochiPipeline(DiffusionPipeline, Mochi1LoraLoaderMixin):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     def prepare_latents(

src/diffusers/pipelines/omnigen/pipeline_omnigen.py

@@ -23,7 +23,7 @@ from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...models.autoencoders import AutoencoderKL
 from ...models.transformers import OmniGenTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import is_torch_xla_available, is_torchvision_available, logging, replace_example_docstring
+from ...utils import deprecate, is_torch_xla_available, is_torchvision_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
 
@@ -235,6 +235,12 @@ class OmniGenPipeline(
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -242,6 +248,12 @@ class OmniGenPipeline(
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -250,6 +262,12 @@ class OmniGenPipeline(
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -257,6 +275,12 @@ class OmniGenPipeline(
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     # Copied from diffusers.pipelines.stable_diffusion_3.pipeline_stable_diffusion_3.StableDiffusion3Pipeline.prepare_latents

src/diffusers/pipelines/pag/pipeline_pag_hunyuandit.py

@@ -28,11 +28,7 @@ from ...models.attention_processor import PAGCFGHunyuanAttnProcessor2_0, PAGHuny
 from ...models.embeddings import get_2d_rotary_pos_embed
 from ...pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 from ...schedulers import DDPMScheduler
-from ...utils import (
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-)
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pag_utils import PAGMixin

src/diffusers/pipelines/pag/pipeline_pag_sana.py

@@ -29,6 +29,7 @@ from ...models.attention_processor import PAGCFGSanaLinearAttnProcessor2_0, PAGI
 from ...schedulers import FlowMatchEulerDiscreteScheduler
 from ...utils import (
     BACKENDS_MAPPING,
+    deprecate,
     is_bs4_available,
     is_ftfy_available,
     is_torch_xla_available,
@@ -190,6 +191,12 @@ class SanaPAGPipeline(DiffusionPipeline, PAGMixin):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -197,6 +204,12 @@ class SanaPAGPipeline(DiffusionPipeline, PAGMixin):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -205,6 +218,12 @@ class SanaPAGPipeline(DiffusionPipeline, PAGMixin):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -212,6 +231,12 @@ class SanaPAGPipeline(DiffusionPipeline, PAGMixin):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     def encode_prompt(

src/diffusers/pipelines/pipeline_utils.py

@@ -57,6 +57,7 @@ from ..utils import (
     PushToHubMixin,
     _get_detailed_type,
     _is_valid_type,
+    deprecate,
     is_accelerate_available,
     is_accelerate_version,
     is_hpu_available,
@@ -2201,6 +2202,12 @@ class StableDiffusionMixin:
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -2208,6 +2215,12 @@ class StableDiffusionMixin:
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -2216,6 +2229,12 @@ class StableDiffusionMixin:
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -2223,6 +2242,12 @@ class StableDiffusionMixin:
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):

src/diffusers/pipelines/qwenimage/pipeline_qwenimage.py

@@ -23,7 +23,7 @@ from ...image_processor import VaeImageProcessor
 from ...loaders import QwenImageLoraLoaderMixin
 from ...models import AutoencoderKLQwenImage, QwenImageTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import is_torch_xla_available, logging, replace_example_docstring
+from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
@@ -348,6 +348,12 @@ class QwenImagePipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -355,6 +361,12 @@ class QwenImagePipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -363,6 +375,12 @@ class QwenImagePipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -370,6 +388,12 @@ class QwenImagePipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     def prepare_latents(

src/diffusers/pipelines/qwenimage/pipeline_qwenimage_controlnet.py

@@ -24,7 +24,7 @@ from ...loaders import QwenImageLoraLoaderMixin
 from ...models import AutoencoderKLQwenImage, QwenImageTransformer2DModel
 from ...models.controlnets.controlnet_qwenimage import QwenImageControlNetModel, QwenImageMultiControlNetModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import is_torch_xla_available, logging, replace_example_docstring
+from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
@@ -412,6 +412,12 @@ class QwenImageControlNetPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -419,6 +425,12 @@ class QwenImageControlNetPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -427,6 +439,12 @@ class QwenImageControlNetPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -434,6 +452,12 @@ class QwenImageControlNetPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage.QwenImagePipeline.prepare_latents

src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit.py

@@ -24,7 +24,7 @@ from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...loaders import QwenImageLoraLoaderMixin
 from ...models import AutoencoderKLQwenImage, QwenImageTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import is_torch_xla_available, logging, replace_example_docstring
+from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
@@ -421,6 +421,12 @@ class QwenImageEditPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -428,6 +434,12 @@ class QwenImageEditPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -436,6 +448,12 @@ class QwenImageEditPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -443,6 +461,12 @@ class QwenImageEditPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     def prepare_latents(

src/diffusers/pipelines/qwenimage/pipeline_qwenimage_edit_inpaint.py

@@ -25,7 +25,7 @@ from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...loaders import QwenImageLoraLoaderMixin
 from ...models import AutoencoderKLQwenImage, QwenImageTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import is_torch_xla_available, logging, replace_example_docstring
+from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
@@ -466,6 +466,12 @@ class QwenImageEditInpaintPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -473,6 +479,12 @@ class QwenImageEditInpaintPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -481,6 +493,12 @@ class QwenImageEditInpaintPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -488,6 +506,12 @@ class QwenImageEditInpaintPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     # Copied from diffusers.pipelines.qwenimage.pipeline_qwenimage_inpaint.QwenImageInpaintPipeline.prepare_latents

src/diffusers/pipelines/qwenimage/pipeline_qwenimage_img2img.py

@@ -9,7 +9,7 @@ from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...loaders import QwenImageLoraLoaderMixin
 from ...models import AutoencoderKLQwenImage, QwenImageTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import is_torch_xla_available, logging, replace_example_docstring
+from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
@@ -397,6 +397,12 @@ class QwenImageImg2ImgPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -404,6 +410,12 @@ class QwenImageImg2ImgPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -412,6 +424,12 @@ class QwenImageImg2ImgPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -419,6 +437,12 @@ class QwenImageImg2ImgPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     def prepare_latents(

src/diffusers/pipelines/qwenimage/pipeline_qwenimage_inpaint.py

@@ -10,7 +10,7 @@ from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...loaders import QwenImageLoraLoaderMixin
 from ...models import AutoencoderKLQwenImage, QwenImageTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import is_torch_xla_available, logging, replace_example_docstring
+from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline
 from .pipeline_output import QwenImagePipelineOutput
@@ -424,6 +424,12 @@ class QwenImageInpaintPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -431,6 +437,12 @@ class QwenImageInpaintPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -439,6 +451,12 @@ class QwenImageInpaintPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -446,6 +464,12 @@ class QwenImageInpaintPipeline(DiffusionPipeline, QwenImageLoraLoaderMixin):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     def prepare_latents(

src/diffusers/pipelines/sana/pipeline_sana.py

@@ -30,6 +30,7 @@ from ...schedulers import DPMSolverMultistepScheduler
 from ...utils import (
     BACKENDS_MAPPING,
     USE_PEFT_BACKEND,
+    deprecate,
     is_bs4_available,
     is_ftfy_available,
     is_torch_xla_available,
@@ -224,6 +225,12 @@ class SanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -231,6 +238,12 @@ class SanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -239,6 +252,12 @@ class SanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -246,6 +265,12 @@ class SanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     def _get_gemma_prompt_embeds(

src/diffusers/pipelines/sana/pipeline_sana_controlnet.py

@@ -30,6 +30,7 @@ from ...schedulers import DPMSolverMultistepScheduler
 from ...utils import (
     BACKENDS_MAPPING,
     USE_PEFT_BACKEND,
+    deprecate,
     is_bs4_available,
     is_ftfy_available,
     is_torch_xla_available,
@@ -237,6 +238,12 @@ class SanaControlNetPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -244,6 +251,12 @@ class SanaControlNetPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -252,6 +265,12 @@ class SanaControlNetPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -259,6 +278,12 @@ class SanaControlNetPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     # Copied from diffusers.pipelines.sana.pipeline_sana.SanaPipeline._get_gemma_prompt_embeds

src/diffusers/pipelines/sana/pipeline_sana_sprint.py

@@ -30,6 +30,7 @@ from ...schedulers import DPMSolverMultistepScheduler
 from ...utils import (
     BACKENDS_MAPPING,
     USE_PEFT_BACKEND,
+    deprecate,
     is_bs4_available,
     is_ftfy_available,
     is_torch_xla_available,
@@ -175,6 +176,12 @@ class SanaSprintPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     def disable_vae_slicing(self):
@@ -182,6 +189,12 @@ class SanaSprintPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def enable_vae_tiling(self):
@@ -190,6 +203,12 @@ class SanaSprintPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -197,6 +216,12 @@ class SanaSprintPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     # Copied from diffusers.pipelines.sana.pipeline_sana.SanaPipeline._get_gemma_prompt_embeds

src/diffusers/pipelines/sana/pipeline_sana_sprint_img2img.py

@@ -31,6 +31,7 @@ from ...schedulers import DPMSolverMultistepScheduler
 from ...utils import (
     BACKENDS_MAPPING,
     USE_PEFT_BACKEND,
+    deprecate,
     is_bs4_available,
     is_ftfy_available,
     is_torch_xla_available,
@@ -183,6 +184,12 @@ class SanaSprintImg2ImgPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     # Copied from diffusers.pipelines.sana.pipeline_sana.SanaPipeline.disable_vae_slicing
@@ -191,6 +198,12 @@ class SanaSprintImg2ImgPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     # Copied from diffusers.pipelines.sana.pipeline_sana.SanaPipeline.enable_vae_tiling
@@ -200,6 +213,12 @@ class SanaSprintImg2ImgPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     def disable_vae_tiling(self):
@@ -207,6 +226,12 @@ class SanaSprintImg2ImgPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     # Copied from diffusers.pipelines.sana.pipeline_sana.SanaPipeline._get_gemma_prompt_embeds

src/diffusers/pipelines/stable_audio/pipeline_stable_audio.py

@@ -25,11 +25,7 @@ from transformers import (
 from ...models import AutoencoderOobleck, StableAudioDiTModel
 from ...models.embeddings import get_1d_rotary_pos_embed
 from ...schedulers import EDMDPMSolverMultistepScheduler
-from ...utils import (
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-)
+from ...utils import deprecate, is_torch_xla_available, logging, replace_example_docstring
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import AudioPipelineOutput, DiffusionPipeline
 from .modeling_stable_audio import StableAudioProjectionModel
@@ -134,6 +130,12 @@ class StableAudioPipeline(DiffusionPipeline):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     # Copied from diffusers.pipelines.pipeline_utils.StableDiffusionMixin.disable_vae_slicing
@@ -142,6 +144,12 @@ class StableAudioPipeline(DiffusionPipeline):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     def encode_prompt(

src/diffusers/pipelines/unidiffuser/pipeline_unidiffuser.py

@@ -232,6 +232,12 @@ class UniDiffuserPipeline(DeprecatedPipelineMixin, DiffusionPipeline):
         Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
         compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
         """
+        depr_message = f"Calling `enable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_slicing()`."
+        deprecate(
+            "enable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_slicing()
 
     # Copied from diffusers.pipelines.pipeline_utils.StableDiffusionMixin.disable_vae_slicing
@@ -240,6 +246,12 @@ class UniDiffuserPipeline(DeprecatedPipelineMixin, DiffusionPipeline):
         Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_slicing()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_slicing()`."
+        deprecate(
+            "disable_vae_slicing",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_slicing()
 
     # Copied from diffusers.pipelines.pipeline_utils.StableDiffusionMixin.enable_vae_tiling
@@ -249,6 +261,12 @@ class UniDiffuserPipeline(DeprecatedPipelineMixin, DiffusionPipeline):
         compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
         processing larger images.
         """
+        depr_message = f"Calling `enable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.enable_tiling()`."
+        deprecate(
+            "enable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.enable_tiling()
 
     # Copied from diffusers.pipelines.pipeline_utils.StableDiffusionMixin.disable_vae_tiling
@@ -257,6 +275,12 @@ class UniDiffuserPipeline(DeprecatedPipelineMixin, DiffusionPipeline):
         Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
         computing decoding in one step.
         """
+        depr_message = f"Calling `disable_vae_tiling()` on a `{self.__class__.__name__}` is deprecated and this method will be removed in a future version. Please use `pipe.vae.disable_tiling()`."
+        deprecate(
+            "disable_vae_tiling",
+            "0.40.0",
+            depr_message,
+        )
         self.vae.disable_tiling()
 
     # Functions to manually set the mode

src/diffusers/pipelines/visualcloze/pipeline_visualcloze_combined.py

@@ -22,11 +22,7 @@ from ...loaders import FluxLoraLoaderMixin, FromSingleFileMixin, TextualInversio
 from ...models.autoencoders import AutoencoderKL
 from ...models.transformers import FluxTransformer2DModel
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import (
-    is_torch_xla_available,
-    logging,
-    replace_example_docstring,
-)
+from ...utils import is_torch_xla_available, logging, replace_example_docstring
 from ..flux.pipeline_flux_fill import FluxFillPipeline as VisualClozeUpsamplingPipeline
 from ..flux.pipeline_output import FluxPipelineOutput
 from ..pipeline_utils import DiffusionPipeline