diffusers 0.32.1__py3-none-any.whl → 0.33.0__py3-none-any.whl

diffusers/pipelines/controlnet_xs/pipeline_controlnet_xs_sd_xl.py

@@ -54,6 +54,16 @@ if is_invisible_watermark_available():
     from ..stable_diffusion_xl.watermark import StableDiffusionXLWatermarker
 
 
+from ...utils import is_torch_xla_available
+
+
+if is_torch_xla_available():
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+else:
+    XLA_AVAILABLE = False
+
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
@@ -196,7 +206,7 @@ class StableDiffusionXLControlNetXSPipeline(
             scheduler=scheduler,
             feature_extractor=feature_extractor,
         )
-        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
+        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True)
         self.control_image_processor = VaeImageProcessor(
             vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True, do_normalize=False
@@ -336,7 +346,9 @@ class StableDiffusionXLControlNetXSPipeline(
                 prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)
 
                 # We are only ALWAYS interested in the pooled output of the final text encoder
-                pooled_prompt_embeds = prompt_embeds[0]
+                if pooled_prompt_embeds is None and prompt_embeds[0].ndim == 2:
+                    pooled_prompt_embeds = prompt_embeds[0]
+
                 if clip_skip is None:
                     prompt_embeds = prompt_embeds.hidden_states[-2]
                 else:
@@ -395,8 +407,10 @@ class StableDiffusionXLControlNetXSPipeline(
                     uncond_input.input_ids.to(device),
                     output_hidden_states=True,
                 )
+
                 # We are only ALWAYS interested in the pooled output of the final text encoder
-                negative_pooled_prompt_embeds = negative_prompt_embeds[0]
+                if negative_pooled_prompt_embeds is None and negative_prompt_embeds[0].ndim == 2:
+                    negative_pooled_prompt_embeds = negative_prompt_embeds[0]
                 negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]
 
                 negative_prompt_embeds_list.append(negative_prompt_embeds)
@@ -1074,6 +1088,9 @@ class StableDiffusionXLControlNetXSPipeline(
                 if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
                     progress_bar.update()
 
+                if XLA_AVAILABLE:
+                    xm.mark_step()
+
         # manually for max memory savings
         if self.vae.dtype == torch.float16 and self.vae.config.force_upcast:
             self.upcast_vae()

diffusers/pipelines/dance_diffusion/pipeline_dance_diffusion.py

@@ -17,11 +17,20 @@ from typing import List, Optional, Tuple, Union
 
 import torch
 
-from ...utils import logging
+from ...models import UNet1DModel
+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
 
 
+if is_torch_xla_available():
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+else:
+    XLA_AVAILABLE = False
+
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
@@ -42,7 +51,7 @@ class DanceDiffusionPipeline(DiffusionPipeline):
 
     model_cpu_offload_seq = "unet"
 
-    def __init__(self, unet, scheduler):
+    def __init__(self, unet: UNet1DModel, scheduler: SchedulerMixin):
         super().__init__()
         self.register_modules(unet=unet, scheduler=scheduler)
 
@@ -146,6 +155,9 @@ class DanceDiffusionPipeline(DiffusionPipeline):
             # 2. compute previous audio sample: x_t -> t_t-1
             audio = self.scheduler.step(model_output, t, audio).prev_sample
 
+            if XLA_AVAILABLE:
+                xm.mark_step()
+
         audio = audio.clamp(-1, 1).float().cpu().numpy()
 
         audio = audio[:, :, :original_sample_size]

diffusers/pipelines/ddim/pipeline_ddim.py

@@ -16,11 +16,21 @@ from typing import List, Optional, Tuple, Union
 
 import torch
 
+from ...models import UNet2DModel
 from ...schedulers import DDIMScheduler
+from ...utils import is_torch_xla_available
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
 
 
+if is_torch_xla_available():
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+else:
+    XLA_AVAILABLE = False
+
+
 class DDIMPipeline(DiffusionPipeline):
     r"""
     Pipeline for image generation.
@@ -38,7 +48,7 @@ class DDIMPipeline(DiffusionPipeline):
 
     model_cpu_offload_seq = "unet"
 
-    def __init__(self, unet, scheduler):
+    def __init__(self, unet: UNet2DModel, scheduler: DDIMScheduler):
         super().__init__()
 
         # make sure scheduler can always be converted to DDIM
@@ -143,6 +153,9 @@ class DDIMPipeline(DiffusionPipeline):
                 model_output, t, image, eta=eta, use_clipped_model_output=use_clipped_model_output, generator=generator
             ).prev_sample
 
+            if XLA_AVAILABLE:
+                xm.mark_step()
+
         image = (image / 2 + 0.5).clamp(0, 1)
         image = image.cpu().permute(0, 2, 3, 1).numpy()
         if output_type == "pil":

diffusers/pipelines/ddpm/pipeline_ddpm.py

@@ -17,10 +17,21 @@ from typing import List, Optional, Tuple, Union
 
 import torch
 
+from ...models import UNet2DModel
+from ...schedulers import DDPMScheduler
+from ...utils import is_torch_xla_available
 from ...utils.torch_utils import randn_tensor
 from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
 
 
+if is_torch_xla_available():
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+else:
+    XLA_AVAILABLE = False
+
+
 class DDPMPipeline(DiffusionPipeline):
     r"""
     Pipeline for image generation.
@@ -38,7 +49,7 @@ class DDPMPipeline(DiffusionPipeline):
 
     model_cpu_offload_seq = "unet"
 
-    def __init__(self, unet, scheduler):
+    def __init__(self, unet: UNet2DModel, scheduler: DDPMScheduler):
         super().__init__()
         self.register_modules(unet=unet, scheduler=scheduler)
 
@@ -116,6 +127,9 @@ class DDPMPipeline(DiffusionPipeline):
             # 2. compute previous image: x_t -> x_t-1
             image = self.scheduler.step(model_output, t, image, generator=generator).prev_sample
 
+            if XLA_AVAILABLE:
+                xm.mark_step()
+
         image = (image / 2 + 0.5).clamp(0, 1)
         image = image.cpu().permute(0, 2, 3, 1).numpy()
         if output_type == "pil":

diffusers/pipelines/deepfloyd_if/pipeline_if.py

@@ -14,6 +14,7 @@ from ...utils import (
     BACKENDS_MAPPING,
     is_bs4_available,
     is_ftfy_available,
+    is_torch_xla_available,
     logging,
     replace_example_docstring,
 )
@@ -24,8 +25,16 @@ from .safety_checker import IFSafetyChecker
 from .watermark import IFWatermarker
 
 
+if is_torch_xla_available():
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+else:
+    XLA_AVAILABLE = False
+
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
+
 if is_bs4_available():
     from bs4 import BeautifulSoup
 
@@ -735,6 +744,9 @@ class IFPipeline(DiffusionPipeline, StableDiffusionLoraLoaderMixin):
                     if callback is not None and i % callback_steps == 0:
                         callback(i, t, intermediate_images)
 
+                if XLA_AVAILABLE:
+                    xm.mark_step()
+
         image = intermediate_images
 
         if output_type == "pil":

diffusers/pipelines/deepfloyd_if/pipeline_if_img2img.py

@@ -17,6 +17,7 @@ from ...utils import (
     PIL_INTERPOLATION,
     is_bs4_available,
     is_ftfy_available,
+    is_torch_xla_available,
     logging,
     replace_example_docstring,
 )
@@ -27,8 +28,16 @@ from .safety_checker import IFSafetyChecker
 from .watermark import IFWatermarker
 
 
+if is_torch_xla_available():
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+else:
+    XLA_AVAILABLE = False
+
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
+
 if is_bs4_available():
     from bs4 import BeautifulSoup
 
@@ -856,6 +865,9 @@ class IFImg2ImgPipeline(DiffusionPipeline, StableDiffusionLoraLoaderMixin):
                     if callback is not None and i % callback_steps == 0:
                         callback(i, t, intermediate_images)
 
+                if XLA_AVAILABLE:
+                    xm.mark_step()
+
         image = intermediate_images
 
         if output_type == "pil":

diffusers/pipelines/deepfloyd_if/pipeline_if_img2img_superresolution.py

@@ -35,6 +35,16 @@ if is_ftfy_available():
     import ftfy
 
 
+from ...utils import is_torch_xla_available
+
+
+if is_torch_xla_available():
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+else:
+    XLA_AVAILABLE = False
+
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
@@ -174,7 +184,7 @@ class IFImg2ImgSuperResolutionPipeline(DiffusionPipeline, StableDiffusionLoraLoa
                 " checker. If you do not want to use the safety checker, you can pass `'safety_checker=None'` instead."
             )
 
-        if unet.config.in_channels != 6:
+        if unet is not None and unet.config.in_channels != 6:
             logger.warning(
                 "It seems like you have loaded a checkpoint that shall not be used for super resolution from {unet.config._name_or_path} as it accepts {unet.config.in_channels} input channels instead of 6. Please make sure to pass a super resolution checkpoint as the `'unet'`: IFSuperResolutionPipeline.from_pretrained(unet=super_resolution_unet, ...)`."
             )
@@ -974,6 +984,9 @@ class IFImg2ImgSuperResolutionPipeline(DiffusionPipeline, StableDiffusionLoraLoa
                     if callback is not None and i % callback_steps == 0:
                         callback(i, t, intermediate_images)
 
+                if XLA_AVAILABLE:
+                    xm.mark_step()
+
         image = intermediate_images
 
         if output_type == "pil":

diffusers/pipelines/deepfloyd_if/pipeline_if_inpainting.py

@@ -17,6 +17,7 @@ from ...utils import (
     PIL_INTERPOLATION,
     is_bs4_available,
     is_ftfy_available,
+    is_torch_xla_available,
     logging,
     replace_example_docstring,
 )
@@ -27,8 +28,16 @@ from .safety_checker import IFSafetyChecker
 from .watermark import IFWatermarker
 
 
+if is_torch_xla_available():
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+else:
+    XLA_AVAILABLE = False
+
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
+
 if is_bs4_available():
     from bs4 import BeautifulSoup
 
@@ -975,6 +984,9 @@ class IFInpaintingPipeline(DiffusionPipeline, StableDiffusionLoraLoaderMixin):
                     if callback is not None and i % callback_steps == 0:
                         callback(i, t, intermediate_images)
 
+                if XLA_AVAILABLE:
+                    xm.mark_step()
+
         image = intermediate_images
 
         if output_type == "pil":

diffusers/pipelines/deepfloyd_if/pipeline_if_inpainting_superresolution.py

@@ -35,6 +35,16 @@ if is_ftfy_available():
     import ftfy
 
 
+from ...utils import is_torch_xla_available
+
+
+if is_torch_xla_available():
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+else:
+    XLA_AVAILABLE = False
+
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
@@ -176,7 +186,7 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, StableDiffusionLora
                 " checker. If you do not want to use the safety checker, you can pass `'safety_checker=None'` instead."
             )
 
-        if unet.config.in_channels != 6:
+        if unet is not None and unet.config.in_channels != 6:
             logger.warning(
                 "It seems like you have loaded a checkpoint that shall not be used for super resolution from {unet.config._name_or_path} as it accepts {unet.config.in_channels} input channels instead of 6. Please make sure to pass a super resolution checkpoint as the `'unet'`: IFSuperResolutionPipeline.from_pretrained(unet=super_resolution_unet, ...)`."
             )
@@ -1085,6 +1095,9 @@ class IFInpaintingSuperResolutionPipeline(DiffusionPipeline, StableDiffusionLora
                     if callback is not None and i % callback_steps == 0:
                         callback(i, t, intermediate_images)
 
+                if XLA_AVAILABLE:
+                    xm.mark_step()
+
         image = intermediate_images
 
         if output_type == "pil":

diffusers/pipelines/deepfloyd_if/pipeline_if_superresolution.py

@@ -34,6 +34,16 @@ if is_ftfy_available():
     import ftfy
 
 
+from ...utils import is_torch_xla_available
+
+
+if is_torch_xla_available():
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+else:
+    XLA_AVAILABLE = False
+
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
 
@@ -132,7 +142,7 @@ class IFSuperResolutionPipeline(DiffusionPipeline, StableDiffusionLoraLoaderMixi
                 " checker. If you do not want to use the safety checker, you can pass `'safety_checker=None'` instead."
             )
 
-        if unet.config.in_channels != 6:
+        if unet is not None and unet.config.in_channels != 6:
             logger.warning(
                 "It seems like you have loaded a checkpoint that shall not be used for super resolution from {unet.config._name_or_path} as it accepts {unet.config.in_channels} input channels instead of 6. Please make sure to pass a super resolution checkpoint as the `'unet'`: IFSuperResolutionPipeline.from_pretrained(unet=super_resolution_unet, ...)`."
             )
@@ -831,6 +841,9 @@ class IFSuperResolutionPipeline(DiffusionPipeline, StableDiffusionLoraLoaderMixi
                     if callback is not None and i % callback_steps == 0:
                         callback(i, t, intermediate_images)
 
+                if XLA_AVAILABLE:
+                    xm.mark_step()
+
         image = intermediate_images
 
         if output_type == "pil":

diffusers/pipelines/deprecated/alt_diffusion/pipeline_alt_diffusion.py

@@ -210,7 +210,7 @@ class AltDiffusionPipeline(
     ):
         super().__init__()
 
-        if hasattr(scheduler.config, "steps_offset") and scheduler.config.steps_offset != 1:
+        if scheduler is not None and getattr(scheduler.config, "steps_offset", 1) != 1:
             deprecation_message = (
                 f"The configuration file of this scheduler: {scheduler} is outdated. `steps_offset`"
                 f" should be set to 1 instead of {scheduler.config.steps_offset}. Please make sure "
@@ -224,7 +224,7 @@ class AltDiffusionPipeline(
             new_config["steps_offset"] = 1
             scheduler._internal_dict = FrozenDict(new_config)
 
-        if hasattr(scheduler.config, "clip_sample") and scheduler.config.clip_sample is True:
+        if scheduler is not None and getattr(scheduler.config, "clip_sample", False) is True:
             deprecation_message = (
                 f"The configuration file of this scheduler: {scheduler} has not set the configuration `clip_sample`."
                 " `clip_sample` should be set to False in the configuration file. Please make sure to update the"
@@ -253,10 +253,14 @@ class AltDiffusionPipeline(
                 " checker. If you do not want to use the safety checker, you can pass `'safety_checker=None'` instead."
             )
 
-        is_unet_version_less_0_9_0 = hasattr(unet.config, "_diffusers_version") and version.parse(
-            version.parse(unet.config._diffusers_version).base_version
-        ) < version.parse("0.9.0.dev0")
-        is_unet_sample_size_less_64 = hasattr(unet.config, "sample_size") and unet.config.sample_size < 64
+        is_unet_version_less_0_9_0 = (
+            unet is not None
+            and hasattr(unet.config, "_diffusers_version")
+            and version.parse(version.parse(unet.config._diffusers_version).base_version) < version.parse("0.9.0.dev0")
+        )
+        is_unet_sample_size_less_64 = (
+            unet is not None and hasattr(unet.config, "sample_size") and unet.config.sample_size < 64
+        )
         if is_unet_version_less_0_9_0 and is_unet_sample_size_less_64:
             deprecation_message = (
                 "The configuration file of the unet has set the default `sample_size` to smaller than"
@@ -284,7 +288,7 @@ class AltDiffusionPipeline(
             feature_extractor=feature_extractor,
             image_encoder=image_encoder,
         )
-        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
+        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 

diffusers/pipelines/deprecated/alt_diffusion/pipeline_alt_diffusion_img2img.py

@@ -238,7 +238,7 @@ class AltDiffusionImg2ImgPipeline(
     ):
         super().__init__()
 
-        if hasattr(scheduler.config, "steps_offset") and scheduler.config.steps_offset != 1:
+        if scheduler is not None and getattr(scheduler.config, "steps_offset", 1) != 1:
             deprecation_message = (
                 f"The configuration file of this scheduler: {scheduler} is outdated. `steps_offset`"
                 f" should be set to 1 instead of {scheduler.config.steps_offset}. Please make sure "
@@ -252,7 +252,7 @@ class AltDiffusionImg2ImgPipeline(
             new_config["steps_offset"] = 1
             scheduler._internal_dict = FrozenDict(new_config)
 
-        if hasattr(scheduler.config, "clip_sample") and scheduler.config.clip_sample is True:
+        if scheduler is not None and getattr(scheduler.config, "clip_sample", False) is True:
             deprecation_message = (
                 f"The configuration file of this scheduler: {scheduler} has not set the configuration `clip_sample`."
                 " `clip_sample` should be set to False in the configuration file. Please make sure to update the"
@@ -281,10 +281,14 @@ class AltDiffusionImg2ImgPipeline(
                 " checker. If you do not want to use the safety checker, you can pass `'safety_checker=None'` instead."
             )
 
-        is_unet_version_less_0_9_0 = hasattr(unet.config, "_diffusers_version") and version.parse(
-            version.parse(unet.config._diffusers_version).base_version
-        ) < version.parse("0.9.0.dev0")
-        is_unet_sample_size_less_64 = hasattr(unet.config, "sample_size") and unet.config.sample_size < 64
+        is_unet_version_less_0_9_0 = (
+            unet is not None
+            and hasattr(unet.config, "_diffusers_version")
+            and version.parse(version.parse(unet.config._diffusers_version).base_version) < version.parse("0.9.0.dev0")
+        )
+        is_unet_sample_size_less_64 = (
+            unet is not None and hasattr(unet.config, "sample_size") and unet.config.sample_size < 64
+        )
         if is_unet_version_less_0_9_0 and is_unet_sample_size_less_64:
             deprecation_message = (
                 "The configuration file of the unet has set the default `sample_size` to smaller than"
@@ -312,7 +316,7 @@ class AltDiffusionImg2ImgPipeline(
             feature_extractor=feature_extractor,
             image_encoder=image_encoder,
         )
-        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
+        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 

diffusers/pipelines/deprecated/repaint/pipeline_repaint.py

@@ -91,7 +91,7 @@ class RePaintPipeline(DiffusionPipeline):
     scheduler: RePaintScheduler
     model_cpu_offload_seq = "unet"
 
-    def __init__(self, unet, scheduler):
+    def __init__(self, unet: UNet2DModel, scheduler: RePaintScheduler):
         super().__init__()
         self.register_modules(unet=unet, scheduler=scheduler)
 

diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_cycle_diffusion.py

@@ -184,7 +184,7 @@ class CycleDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Sta
     ):
         super().__init__()
 
-        if hasattr(scheduler.config, "steps_offset") and scheduler.config.steps_offset != 1:
+        if scheduler is not None and getattr(scheduler.config, "steps_offset", 1) != 1:
             deprecation_message = (
                 f"The configuration file of this scheduler: {scheduler} is outdated. `steps_offset`"
                 f" should be set to 1 instead of {scheduler.config.steps_offset}. Please make sure "
@@ -213,10 +213,14 @@ class CycleDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Sta
                 "Make sure to define a feature extractor when loading {self.__class__} if you want to use the safety"
                 " checker. If you do not want to use the safety checker, you can pass `'safety_checker=None'` instead."
             )
-        is_unet_version_less_0_9_0 = hasattr(unet.config, "_diffusers_version") and version.parse(
-            version.parse(unet.config._diffusers_version).base_version
-        ) < version.parse("0.9.0.dev0")
-        is_unet_sample_size_less_64 = hasattr(unet.config, "sample_size") and unet.config.sample_size < 64
+        is_unet_version_less_0_9_0 = (
+            unet is not None
+            and hasattr(unet.config, "_diffusers_version")
+            and version.parse(version.parse(unet.config._diffusers_version).base_version) < version.parse("0.9.0.dev0")
+        )
+        is_unet_sample_size_less_64 = (
+            unet is not None and hasattr(unet.config, "sample_size") and unet.config.sample_size < 64
+        )
         if is_unet_version_less_0_9_0 and is_unet_sample_size_less_64:
             deprecation_message = (
                 "The configuration file of the unet has set the default `sample_size` to smaller than"
@@ -243,7 +247,7 @@ class CycleDiffusionPipeline(DiffusionPipeline, TextualInversionLoaderMixin, Sta
             safety_checker=safety_checker,
             feature_extractor=feature_extractor,
         )
-        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
+        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 

diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_onnx_stable_diffusion_inpaint_legacy.py

@@ -93,7 +93,7 @@ class OnnxStableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
     ):
         super().__init__()
 
-        if hasattr(scheduler.config, "steps_offset") and scheduler.config.steps_offset != 1:
+        if scheduler is not None and getattr(scheduler.config, "steps_offset", 1) != 1:
             deprecation_message = (
                 f"The configuration file of this scheduler: {scheduler} is outdated. `steps_offset`"
                 f" should be set to 1 instead of {scheduler.config.steps_offset}. Please make sure "
@@ -107,7 +107,7 @@ class OnnxStableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
             new_config["steps_offset"] = 1
             scheduler._internal_dict = FrozenDict(new_config)
 
-        if hasattr(scheduler.config, "clip_sample") and scheduler.config.clip_sample is True:
+        if scheduler is not None and getattr(scheduler.config, "clip_sample", False) is True:
             deprecation_message = (
                 f"The configuration file of this scheduler: {scheduler} has not set the configuration `clip_sample`."
                 " `clip_sample` should be set to False in the configuration file. Please make sure to update the"

diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_inpaint_legacy.py

@@ -140,7 +140,7 @@ class StableDiffusionInpaintPipelineLegacy(
         )
         deprecate("legacy is outdated", "1.0.0", deprecation_message, standard_warn=False)
 
-        if hasattr(scheduler.config, "steps_offset") and scheduler.config.steps_offset != 1:
+        if scheduler is not None and getattr(scheduler.config, "steps_offset", 1) != 1:
             deprecation_message = (
                 f"The configuration file of this scheduler: {scheduler} is outdated. `steps_offset`"
                 f" should be set to 1 instead of {scheduler.config.steps_offset}. Please make sure "
@@ -154,7 +154,7 @@ class StableDiffusionInpaintPipelineLegacy(
             new_config["steps_offset"] = 1
             scheduler._internal_dict = FrozenDict(new_config)
 
-        if hasattr(scheduler.config, "clip_sample") and scheduler.config.clip_sample is True:
+        if scheduler is not None and getattr(scheduler.config, "clip_sample", False) is True:
             deprecation_message = (
                 f"The configuration file of this scheduler: {scheduler} has not set the configuration `clip_sample`."
                 " `clip_sample` should be set to False in the configuration file. Please make sure to update the"
@@ -183,10 +183,14 @@ class StableDiffusionInpaintPipelineLegacy(
                 " checker. If you do not want to use the safety checker, you can pass `'safety_checker=None'` instead."
             )
 
-        is_unet_version_less_0_9_0 = hasattr(unet.config, "_diffusers_version") and version.parse(
-            version.parse(unet.config._diffusers_version).base_version
-        ) < version.parse("0.9.0.dev0")
-        is_unet_sample_size_less_64 = hasattr(unet.config, "sample_size") and unet.config.sample_size < 64
+        is_unet_version_less_0_9_0 = (
+            unet is not None
+            and hasattr(unet.config, "_diffusers_version")
+            and version.parse(version.parse(unet.config._diffusers_version).base_version) < version.parse("0.9.0.dev0")
+        )
+        is_unet_sample_size_less_64 = (
+            unet is not None and hasattr(unet.config, "sample_size") and unet.config.sample_size < 64
+        )
         if is_unet_version_less_0_9_0 and is_unet_sample_size_less_64:
             deprecation_message = (
                 "The configuration file of the unet has set the default `sample_size` to smaller than"
@@ -213,7 +217,7 @@ class StableDiffusionInpaintPipelineLegacy(
             safety_checker=safety_checker,
             feature_extractor=feature_extractor,
         )
-        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
+        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 

diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_model_editing.py

@@ -121,7 +121,7 @@ class StableDiffusionModelEditingPipeline(
             safety_checker=safety_checker,
             feature_extractor=feature_extractor,
         )
-        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
+        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 

diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_paradigms.py

@@ -143,7 +143,7 @@ class StableDiffusionParadigmsPipeline(
             safety_checker=safety_checker,
             feature_extractor=feature_extractor,
         )
-        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
+        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)
 

diffusers/pipelines/deprecated/stable_diffusion_variants/pipeline_stable_diffusion_pix2pix_zero.py

@@ -365,7 +365,7 @@ class StableDiffusionPix2PixZeroPipeline(DiffusionPipeline, StableDiffusionMixin
             caption_generator=caption_generator,
             inverse_scheduler=inverse_scheduler,
         )
-        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
+        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
         self.register_to_config(requires_safety_checker=requires_safety_checker)