Merge branch 'main' into flux

Author: hlky

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

@@ -20,7 +20,7 @@
 
 <Tip>
 
-Make sure to check out the Schedulers [guide](../../using-diffusers/schedulers.md) to learn how to explore the tradeoff between scheduler speed and quality, and see the [reuse components across pipelines](../../using-diffusers/loading.md#reuse-a-pipeline) section to learn how to efficiently load the same components into multiple pipelines.
+Make sure to check out the Schedulers [guide](https://huggingface.co/docs/diffusers/main/en/using-diffusers/schedulers) to learn how to explore the tradeoff between scheduler speed and quality, and see the [reuse components across pipelines](https://huggingface.co/docs/diffusers/main/en/using-diffusers/loading#reuse-a-pipeline) section to learn how to efficiently load the same components into multiple pipelines.
 
 </Tip>
 

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

@@ -30,7 +30,7 @@ HunyuanDiT has the following components:
 
 <Tip>
 
-Make sure to check out the Schedulers [guide](../../using-diffusers/schedulers.md) to learn how to explore the tradeoff between scheduler speed and quality, and see the [reuse components across pipelines](../../using-diffusers/loading.md#reuse-a-pipeline) section to learn how to efficiently load the same components into multiple pipelines.
+Make sure to check out the Schedulers [guide](https://huggingface.co/docs/diffusers/main/en/using-diffusers/schedulers) to learn how to explore the tradeoff between scheduler speed and quality, and see the [reuse components across pipelines](https://huggingface.co/docs/diffusers/main/en/using-diffusers/loading#reuse-a-pipeline) section to learn how to efficiently load the same components into multiple pipelines.
 
 </Tip>
 

src/diffusers/pipelines/aura_flow/pipeline_aura_flow.py

@@ -21,11 +21,18 @@
 from ...models import AuraFlowTransformer2DModel, AutoencoderKL
 from ...models.attention_processor import AttnProcessor2_0, FusedAttnProcessor2_0, XFormersAttnProcessor
 from ...schedulers import FlowMatchEulerDiscreteScheduler
-from ...utils import 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, ImagePipelineOutput
 
 
+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
 
 
@@ -564,6 +571,9 @@ def __call__(
                 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()
+
         if output_type == "latent":
             image = latents
         else:

src/diffusers/pipelines/sana/pipeline_sana.py

@@ -31,6 +31,7 @@
     USE_PEFT_BACKEND,
     is_bs4_available,
     is_ftfy_available,
+    is_torch_xla_available,
     logging,
     replace_example_docstring,
     scale_lora_layers,
@@ -46,6 +47,13 @@
 from .pipeline_output import SanaPipelineOutput
 
 
+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():
@@ -864,6 +872,9 @@ def __call__(
                 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()
+
         if output_type == "latent":
             image = latents
         else:

src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3_img2img.py

@@ -240,12 +240,21 @@ def __init__(
             transformer=transformer,
             scheduler=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 hasattr(self, "vae") and self.vae is not None else 8
+        )
+        latent_channels = self.vae.config.latent_channels if hasattr(self, "vae") and self.vae is not None else 16
         self.image_processor = VaeImageProcessor(
-            vae_scale_factor=self.vae_scale_factor, vae_latent_channels=self.vae.config.latent_channels
+            vae_scale_factor=self.vae_scale_factor, vae_latent_channels=latent_channels
+        )
+        self.tokenizer_max_length = (
+            self.tokenizer.model_max_length if hasattr(self, "tokenizer") and self.tokenizer is not None else 77
+        )
+        self.default_sample_size = (
+            self.transformer.config.sample_size
+            if hasattr(self, "transformer") and self.transformer is not None
+            else 128
         )
-        self.tokenizer_max_length = self.tokenizer.model_max_length
-        self.default_sample_size = self.transformer.config.sample_size
         self.patch_size = (
             self.transformer.config.patch_size if hasattr(self, "transformer") and self.transformer is not None else 2
         )

src/diffusers/pipelines/stable_diffusion_3/pipeline_stable_diffusion_3_inpaint.py

@@ -239,19 +239,28 @@ def __init__(
             transformer=transformer,
             scheduler=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 hasattr(self, "vae") and self.vae is not None else 8
+        )
+        latent_channels = self.vae.config.latent_channels if hasattr(self, "vae") and self.vae is not None else 16
         self.image_processor = VaeImageProcessor(
-            vae_scale_factor=self.vae_scale_factor, vae_latent_channels=self.vae.config.latent_channels
+            vae_scale_factor=self.vae_scale_factor, vae_latent_channels=latent_channels
         )
         self.mask_processor = VaeImageProcessor(
             vae_scale_factor=self.vae_scale_factor,
-            vae_latent_channels=self.vae.config.latent_channels,
+            vae_latent_channels=latent_channels,
             do_normalize=False,
             do_binarize=True,
             do_convert_grayscale=True,
         )
-        self.tokenizer_max_length = self.tokenizer.model_max_length
-        self.default_sample_size = self.transformer.config.sample_size
+        self.tokenizer_max_length = (
+            self.tokenizer.model_max_length if hasattr(self, "tokenizer") and self.tokenizer is not None else 77
+        )
+        self.default_sample_size = (
+            self.transformer.config.sample_size
+            if hasattr(self, "transformer") and self.transformer is not None
+            else 128
+        )
         self.patch_size = (
             self.transformer.config.patch_size if hasattr(self, "transformer") and self.transformer is not None else 2
         )