diffusers 0.32.2__py3-none-any.whl → 0.33.0__py3-none-any.whl

diffusers/pipelines/__init__.py

@@ -10,6 +10,7 @@ from ..utils import (
     is_librosa_available,
     is_note_seq_available,
     is_onnx_available,
+    is_opencv_available,
     is_sentencepiece_available,
     is_torch_available,
     is_torch_npu_available,
@@ -154,6 +155,7 @@ else:
         "CogVideoXFunControlPipeline",
     ]
     _import_structure["cogview3"] = ["CogView3PlusPipeline"]
+    _import_structure["cogview4"] = ["CogView4Pipeline", "CogView4ControlPipeline"]
     _import_structure["controlnet"].extend(
         [
             "BlipDiffusionControlNetPipeline",
@@ -214,8 +216,17 @@ else:
         "IFPipeline",
         "IFSuperResolutionPipeline",
     ]
+    _import_structure["easyanimate"] = [
+        "EasyAnimatePipeline",
+        "EasyAnimateInpaintPipeline",
+        "EasyAnimateControlPipeline",
+    ]
     _import_structure["hunyuandit"] = ["HunyuanDiTPipeline"]
-    _import_structure["hunyuan_video"] = ["HunyuanVideoPipeline"]
+    _import_structure["hunyuan_video"] = [
+        "HunyuanVideoPipeline",
+        "HunyuanSkyreelsImageToVideoPipeline",
+        "HunyuanVideoImageToVideoPipeline",
+    ]
     _import_structure["kandinsky"] = [
         "KandinskyCombinedPipeline",
         "KandinskyImg2ImgCombinedPipeline",
@@ -253,20 +264,23 @@ else:
         ]
     )
     _import_structure["latte"] = ["LattePipeline"]
-    _import_structure["ltx"] = ["LTXPipeline", "LTXImageToVideoPipeline"]
-    _import_structure["lumina"] = ["LuminaText2ImgPipeline"]
+    _import_structure["ltx"] = ["LTXPipeline", "LTXImageToVideoPipeline", "LTXConditionPipeline"]
+    _import_structure["lumina"] = ["LuminaPipeline", "LuminaText2ImgPipeline"]
+    _import_structure["lumina2"] = ["Lumina2Pipeline", "Lumina2Text2ImgPipeline"]
     _import_structure["marigold"].extend(
         [
             "MarigoldDepthPipeline",
+            "MarigoldIntrinsicsPipeline",
             "MarigoldNormalsPipeline",
         ]
     )
     _import_structure["mochi"] = ["MochiPipeline"]
     _import_structure["musicldm"] = ["MusicLDMPipeline"]
+    _import_structure["omnigen"] = ["OmniGenPipeline"]
     _import_structure["paint_by_example"] = ["PaintByExamplePipeline"]
     _import_structure["pia"] = ["PIAPipeline"]
     _import_structure["pixart_alpha"] = ["PixArtAlphaPipeline", "PixArtSigmaPipeline"]
-    _import_structure["sana"] = ["SanaPipeline"]
+    _import_structure["sana"] = ["SanaPipeline", "SanaSprintPipeline"]
     _import_structure["semantic_stable_diffusion"] = ["SemanticStableDiffusionPipeline"]
     _import_structure["shap_e"] = ["ShapEImg2ImgPipeline", "ShapEPipeline"]
     _import_structure["stable_audio"] = [
@@ -342,6 +356,7 @@ else:
         "WuerstchenDecoderPipeline",
         "WuerstchenPriorPipeline",
     ]
+    _import_structure["wan"] = ["WanPipeline", "WanImageToVideoPipeline", "WanVideoToVideoPipeline"]
 try:
     if not is_onnx_available():
         raise OptionalDependencyNotAvailable()
@@ -399,6 +414,18 @@ else:
         "KolorsImg2ImgPipeline",
     ]
 
+try:
+    if not (is_torch_available() and is_transformers_available() and is_opencv_available()):
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    from ..utils import (
+        dummy_torch_and_transformers_and_opencv_objects,
+    )
+
+    _dummy_objects.update(get_objects_from_module(dummy_torch_and_transformers_and_opencv_objects))
+else:
+    _import_structure["consisid"] = ["ConsisIDPipeline"]
+
 try:
     if not is_flax_available():
         raise OptionalDependencyNotAvailable()
@@ -496,6 +523,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             CogVideoXVideoToVideoPipeline,
         )
         from .cogview3 import CogView3PlusPipeline
+        from .cogview4 import CogView4ControlPipeline, CogView4Pipeline
         from .controlnet import (
             BlipDiffusionControlNetPipeline,
             StableDiffusionControlNetImg2ImgPipeline,
@@ -538,6 +566,11 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             VersatileDiffusionTextToImagePipeline,
             VQDiffusionPipeline,
         )
+        from .easyanimate import (
+            EasyAnimateControlPipeline,
+            EasyAnimateInpaintPipeline,
+            EasyAnimatePipeline,
+        )
         from .flux import (
             FluxControlImg2ImgPipeline,
             FluxControlInpaintPipeline,
@@ -552,7 +585,11 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             FluxPriorReduxPipeline,
             ReduxImageEncoder,
         )
-        from .hunyuan_video import HunyuanVideoPipeline
+        from .hunyuan_video import (
+            HunyuanSkyreelsImageToVideoPipeline,
+            HunyuanVideoImageToVideoPipeline,
+            HunyuanVideoPipeline,
+        )
         from .hunyuandit import HunyuanDiTPipeline
         from .i2vgen_xl import I2VGenXLPipeline
         from .kandinsky import (
@@ -592,14 +629,17 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             LEditsPPPipelineStableDiffusion,
             LEditsPPPipelineStableDiffusionXL,
         )
-        from .ltx import LTXImageToVideoPipeline, LTXPipeline
-        from .lumina import LuminaText2ImgPipeline
+        from .ltx import LTXConditionPipeline, LTXImageToVideoPipeline, LTXPipeline
+        from .lumina import LuminaPipeline, LuminaText2ImgPipeline
+        from .lumina2 import Lumina2Pipeline, Lumina2Text2ImgPipeline
         from .marigold import (
             MarigoldDepthPipeline,
+            MarigoldIntrinsicsPipeline,
             MarigoldNormalsPipeline,
         )
         from .mochi import MochiPipeline
         from .musicldm import MusicLDMPipeline
+        from .omnigen import OmniGenPipeline
         from .pag import (
             AnimateDiffPAGPipeline,
             HunyuanDiTPAGPipeline,
@@ -622,7 +662,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
         from .paint_by_example import PaintByExamplePipeline
         from .pia import PIAPipeline
         from .pixart_alpha import PixArtAlphaPipeline, PixArtSigmaPipeline
-        from .sana import SanaPipeline
+        from .sana import SanaPipeline, SanaSprintPipeline
         from .semantic_stable_diffusion import SemanticStableDiffusionPipeline
         from .shap_e import ShapEImg2ImgPipeline, ShapEPipeline
         from .stable_audio import StableAudioPipeline, StableAudioProjectionModel
@@ -680,6 +720,7 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
             UniDiffuserPipeline,
             UniDiffuserTextDecoder,
         )
+        from .wan import WanImageToVideoPipeline, WanPipeline, WanVideoToVideoPipeline
         from .wuerstchen import (
             WuerstchenCombinedPipeline,
             WuerstchenDecoderPipeline,
@@ -731,6 +772,14 @@ if TYPE_CHECKING or DIFFUSERS_SLOW_IMPORT:
                 KolorsPipeline,
             )
 
+        try:
+            if not (is_torch_available() and is_transformers_available() and is_opencv_available()):
+                raise OptionalDependencyNotAvailable()
+        except OptionalDependencyNotAvailable:
+            from ..utils.dummy_torch_and_transformers_and_opencv_objects import *
+        else:
+            from .consisid import ConsisIDPipeline
+
         try:
             if not is_flax_available():
                 raise OptionalDependencyNotAvailable()

diffusers/pipelines/allegro/pipeline_allegro.py

@@ -33,6 +33,7 @@ from ...utils import (
     deprecate,
     is_bs4_available,
     is_ftfy_available,
+    is_torch_xla_available,
     logging,
     replace_example_docstring,
 )
@@ -41,6 +42,14 @@ from ...video_processor import VideoProcessor
 from .pipeline_output import AllegroPipelineOutput
 
 
+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__)
 
 if is_bs4_available():
@@ -194,10 +203,10 @@ class AllegroPipeline(DiffusionPipeline):
             tokenizer=tokenizer, text_encoder=text_encoder, vae=vae, transformer=transformer, scheduler=scheduler
         )
         self.vae_scale_factor_spatial = (
-            2 ** (len(self.vae.config.block_out_channels) - 1) if hasattr(self, "vae") and self.vae is not None else 8
+            2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
         )
         self.vae_scale_factor_temporal = (
-            self.vae.config.temporal_compression_ratio if hasattr(self, "vae") and self.vae is not None else 4
+            self.vae.config.temporal_compression_ratio if getattr(self, "vae", None) else 4
         )
 
         self.video_processor = VideoProcessor(vae_scale_factor=self.vae_scale_factor_spatial)
@@ -674,6 +683,10 @@ class AllegroPipeline(DiffusionPipeline):
     def num_timesteps(self):
         return self._num_timesteps
 
+    @property
+    def current_timestep(self):
+        return self._current_timestep
+
     @property
     def interrupt(self):
         return self._interrupt
@@ -806,6 +819,7 @@ class AllegroPipeline(DiffusionPipeline):
             negative_prompt_attention_mask,
         )
         self._guidance_scale = guidance_scale
+        self._current_timestep = None
         self._interrupt = False
 
         # 2. Default height and width to transformer
@@ -883,6 +897,7 @@ class AllegroPipeline(DiffusionPipeline):
                 if self.interrupt:
                     continue
 
+                self._current_timestep = t
                 latent_model_input = torch.cat([latents] * 2) if do_classifier_free_guidance else latents
                 latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
 
@@ -921,6 +936,11 @@ class AllegroPipeline(DiffusionPipeline):
                 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()
+
+        self._current_timestep = None
+
         if not output_type == "latent":
             latents = latents.to(self.vae.dtype)
             video = self.decode_latents(latents)

diffusers/pipelines/amused/pipeline_amused.py

@@ -20,10 +20,18 @@ from transformers import CLIPTextModelWithProjection, CLIPTokenizer
 from ...image_processor import VaeImageProcessor
 from ...models import UVit2DModel, VQModel
 from ...schedulers import AmusedScheduler
-from ...utils import replace_example_docstring
+from ...utils import is_torch_xla_available, replace_example_docstring
 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
+
+
 EXAMPLE_DOC_STRING = """
     Examples:
         ```py
@@ -66,7 +74,9 @@ class AmusedPipeline(DiffusionPipeline):
             transformer=transformer,
             scheduler=scheduler,
         )
-        self.vae_scale_factor = 2 ** (len(self.vqvae.config.block_out_channels) - 1)
+        self.vae_scale_factor = (
+            2 ** (len(self.vqvae.config.block_out_channels) - 1) if getattr(self, "vqvae", None) else 8
+        )
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor, do_normalize=False)
 
     @torch.no_grad()
@@ -297,6 +307,9 @@ class AmusedPipeline(DiffusionPipeline):
                         step_idx = i // getattr(self.scheduler, "order", 1)
                         callback(step_idx, timestep, latents)
 
+                if XLA_AVAILABLE:
+                    xm.mark_step()
+
         if output_type == "latent":
             output = latents
         else:

diffusers/pipelines/amused/pipeline_amused_img2img.py

@@ -20,10 +20,18 @@ from transformers import CLIPTextModelWithProjection, CLIPTokenizer
 from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...models import UVit2DModel, VQModel
 from ...schedulers import AmusedScheduler
-from ...utils import replace_example_docstring
+from ...utils import is_torch_xla_available, replace_example_docstring
 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
+
+
 EXAMPLE_DOC_STRING = """
     Examples:
         ```py
@@ -81,7 +89,9 @@ class AmusedImg2ImgPipeline(DiffusionPipeline):
             transformer=transformer,
             scheduler=scheduler,
         )
-        self.vae_scale_factor = 2 ** (len(self.vqvae.config.block_out_channels) - 1)
+        self.vae_scale_factor = (
+            2 ** (len(self.vqvae.config.block_out_channels) - 1) if getattr(self, "vqvae", None) else 8
+        )
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor, do_normalize=False)
 
     @torch.no_grad()
@@ -323,6 +333,9 @@ class AmusedImg2ImgPipeline(DiffusionPipeline):
                         step_idx = i // getattr(self.scheduler, "order", 1)
                         callback(step_idx, timestep, latents)
 
+                if XLA_AVAILABLE:
+                    xm.mark_step()
+
         if output_type == "latent":
             output = latents
         else:

diffusers/pipelines/amused/pipeline_amused_inpaint.py

@@ -21,10 +21,18 @@ from transformers import CLIPTextModelWithProjection, CLIPTokenizer
 from ...image_processor import PipelineImageInput, VaeImageProcessor
 from ...models import UVit2DModel, VQModel
 from ...schedulers import AmusedScheduler
-from ...utils import replace_example_docstring
+from ...utils import is_torch_xla_available, replace_example_docstring
 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
+
+
 EXAMPLE_DOC_STRING = """
     Examples:
         ```py
@@ -89,7 +97,9 @@ class AmusedInpaintPipeline(DiffusionPipeline):
             transformer=transformer,
             scheduler=scheduler,
         )
-        self.vae_scale_factor = 2 ** (len(self.vqvae.config.block_out_channels) - 1)
+        self.vae_scale_factor = (
+            2 ** (len(self.vqvae.config.block_out_channels) - 1) if getattr(self, "vqvae", None) else 8
+        )
         self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor, do_normalize=False)
         self.mask_processor = VaeImageProcessor(
             vae_scale_factor=self.vae_scale_factor,
@@ -354,6 +364,9 @@ class AmusedInpaintPipeline(DiffusionPipeline):
                         step_idx = i // getattr(self.scheduler, "order", 1)
                         callback(step_idx, timestep, latents)
 
+                if XLA_AVAILABLE:
+                    xm.mark_step()
+
         if output_type == "latent":
             output = latents
         else:

diffusers/pipelines/animatediff/pipeline_animatediff.py

@@ -19,7 +19,7 @@ import torch
 from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
 
 from ...image_processor import PipelineImageInput
-from ...loaders import IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from ...loaders import FromSingleFileMixin, IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel, UNetMotionModel
 from ...models.lora import adjust_lora_scale_text_encoder
 from ...models.unets.unet_motion_model import MotionAdapter
@@ -34,6 +34,7 @@ from ...schedulers import (
 from ...utils import (
     USE_PEFT_BACKEND,
     deprecate,
+    is_torch_xla_available,
     logging,
     replace_example_docstring,
     scale_lora_layers,
@@ -47,8 +48,16 @@ from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from .pipeline_output import AnimateDiffPipelineOutput
 
 
+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
 
+
 EXAMPLE_DOC_STRING = """
     Examples:
         ```py
@@ -74,6 +83,7 @@ class AnimateDiffPipeline(
     StableDiffusionLoraLoaderMixin,
     FreeInitMixin,
     AnimateDiffFreeNoiseMixin,
+    FromSingleFileMixin,
 ):
     r"""
     Pipeline for text-to-video generation.
@@ -139,7 +149,7 @@ class AnimateDiffPipeline(
             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.video_processor = VideoProcessor(do_resize=False, vae_scale_factor=self.vae_scale_factor)
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
@@ -844,6 +854,9 @@ class AnimateDiffPipeline(
                         if callback is not None and i % callback_steps == 0:
                             callback(i, t, latents)
 
+                    if XLA_AVAILABLE:
+                        xm.mark_step()
+
         # 9. Post processing
         if output_type == "latent":
             video = latents

diffusers/pipelines/animatediff/pipeline_animatediff_controlnet.py

@@ -20,7 +20,7 @@ import torch.nn.functional as F
 from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
 
 from ...image_processor import PipelineImageInput
-from ...loaders import IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from ...loaders import FromSingleFileMixin, IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import (
     AutoencoderKL,
     ControlNetModel,
@@ -32,7 +32,7 @@ from ...models import (
 from ...models.lora import adjust_lora_scale_text_encoder
 from ...models.unets.unet_motion_model import MotionAdapter
 from ...schedulers import KarrasDiffusionSchedulers
-from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers, unscale_lora_layers
+from ...utils import USE_PEFT_BACKEND, is_torch_xla_available, logging, scale_lora_layers, unscale_lora_layers
 from ...utils.torch_utils import is_compiled_module, randn_tensor
 from ...video_processor import VideoProcessor
 from ..free_init_utils import FreeInitMixin
@@ -41,8 +41,16 @@ from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from .pipeline_output import AnimateDiffPipelineOutput
 
 
+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
 
+
 EXAMPLE_DOC_STRING = """
     Examples:
         ```py
@@ -117,6 +125,7 @@ class AnimateDiffControlNetPipeline(
     StableDiffusionLoraLoaderMixin,
     FreeInitMixin,
     AnimateDiffFreeNoiseMixin,
+    FromSingleFileMixin,
 ):
     r"""
     Pipeline for text-to-video generation with ControlNet guidance.
@@ -180,7 +189,7 @@ class AnimateDiffControlNetPipeline(
             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.video_processor = VideoProcessor(vae_scale_factor=self.vae_scale_factor)
         self.control_video_processor = VideoProcessor(
             vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True, do_normalize=False
@@ -1090,6 +1099,9 @@ class AnimateDiffControlNetPipeline(
                     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()
+
         # 9. Post processing
         if output_type == "latent":
             video = latents

diffusers/pipelines/animatediff/pipeline_animatediff_sdxl.py

@@ -48,6 +48,7 @@ from ...schedulers import (
 )
 from ...utils import (
     USE_PEFT_BACKEND,
+    is_torch_xla_available,
     logging,
     replace_example_docstring,
     scale_lora_layers,
@@ -60,8 +61,16 @@ from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from .pipeline_output import AnimateDiffPipelineOutput
 
 
+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
 
+
 EXAMPLE_DOC_STRING = """
     Examples:
         ```py
@@ -307,10 +316,14 @@ class AnimateDiffSDXLPipeline(
             feature_extractor=feature_extractor,
         )
         self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)
-        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.video_processor = VideoProcessor(vae_scale_factor=self.vae_scale_factor)
 
-        self.default_sample_size = self.unet.config.sample_size
+        self.default_sample_size = (
+            self.unet.config.sample_size
+            if hasattr(self, "unet") and self.unet is not None and hasattr(self.unet.config, "sample_size")
+            else 128
+        )
 
     # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt with num_images_per_prompt->num_videos_per_prompt
     def encode_prompt(
@@ -438,7 +451,9 @@ class AnimateDiffSDXLPipeline(
                 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:
@@ -497,8 +512,10 @@ class AnimateDiffSDXLPipeline(
                     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)
@@ -1261,6 +1278,9 @@ class AnimateDiffSDXLPipeline(
 
                     progress_bar.update()
 
+                    if XLA_AVAILABLE:
+                        xm.mark_step()
+
         # make sure the VAE is in float32 mode, as it overflows in float16
         needs_upcasting = self.vae.dtype == torch.float16 and self.vae.config.force_upcast
 

diffusers/pipelines/animatediff/pipeline_animatediff_sparsectrl.py

@@ -22,7 +22,7 @@ import torch.nn.functional as F
 from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
 
 from ...image_processor import PipelineImageInput, VaeImageProcessor
-from ...loaders import IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
+from ...loaders import FromSingleFileMixin, IPAdapterMixin, StableDiffusionLoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel, UNetMotionModel
 from ...models.controlnets.controlnet_sparsectrl import SparseControlNetModel
 from ...models.lora import adjust_lora_scale_text_encoder
@@ -30,6 +30,7 @@ from ...models.unets.unet_motion_model import MotionAdapter
 from ...schedulers import KarrasDiffusionSchedulers
 from ...utils import (
     USE_PEFT_BACKEND,
+    is_torch_xla_available,
     logging,
     replace_example_docstring,
     scale_lora_layers,
@@ -42,8 +43,16 @@ from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from .pipeline_output import AnimateDiffPipelineOutput
 
 
+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
 
+
 EXAMPLE_DOC_STRING = """
     Examples:
         ```python
@@ -127,6 +136,7 @@ class AnimateDiffSparseControlNetPipeline(
     IPAdapterMixin,
     StableDiffusionLoraLoaderMixin,
     FreeInitMixin,
+    FromSingleFileMixin,
 ):
     r"""
     Pipeline for controlled text-to-video generation using the method described in [SparseCtrl: Adding Sparse Controls
@@ -188,7 +198,7 @@ class AnimateDiffSparseControlNetPipeline(
             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.video_processor = VideoProcessor(do_resize=False, vae_scale_factor=self.vae_scale_factor)
         self.control_image_processor = VaeImageProcessor(
             vae_scale_factor=self.vae_scale_factor, do_convert_rgb=True, do_normalize=False
@@ -994,6 +1004,9 @@ class AnimateDiffSparseControlNetPipeline(
                     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()
+
         # 11. Post processing
         if output_type == "latent":
             video = latents