Merge branch 'main' into ltxv-0.9.1-integration

Author: a-r-r-o-w

docs/source/en/api/models/autoencoder_kl_hunyuan_video.md

@@ -18,7 +18,7 @@ The model can be loaded with the following code snippet.
 ```python
 from diffusers import AutoencoderKLHunyuanVideo
 
-vae = AutoencoderKLHunyuanVideo.from_pretrained("tencent/HunyuanVideo", torch_dtype=torch.float16)
+vae = AutoencoderKLHunyuanVideo.from_pretrained("hunyuanvideo-community/HunyuanVideo", subfolder="vae", torch_dtype=torch.float16)
 ```
 
 ## AutoencoderKLHunyuanVideo

docs/source/en/api/models/hunyuan_video_transformer_3d.md

@@ -18,7 +18,7 @@ The model can be loaded with the following code snippet.
 ```python
 from diffusers import HunyuanVideoTransformer3DModel
 
-transformer = HunyuanVideoTransformer3DModel.from_pretrained("tencent/HunyuanVideo", torch_dtype=torch.bfloat16)
+transformer = HunyuanVideoTransformer3DModel.from_pretrained("hunyuanvideo-community/HunyuanVideo", subfolder="transformer", torch_dtype=torch.bfloat16)
 ```
 
 ## HunyuanVideoTransformer3DModel

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

@@ -29,7 +29,7 @@ Recommendations for inference:
 - Transformer should be in `torch.bfloat16`.
 - VAE should be in `torch.float16`.
 - `num_frames` should be of the form `4 * k + 1`, for example `49` or `129`.
-- For smaller resolution images, try lower values of `shift` (between `2.0` to `5.0`) in the [Scheduler](https://huggingface.co/docs/diffusers/main/en/api/schedulers/flow_match_euler_discrete#diffusers.FlowMatchEulerDiscreteScheduler.shift). For larger resolution images, try higher values (between `7.0` and `12.0`). The default value is `7.0` for HunyuanVideo.
+- For smaller resolution videos, try lower values of `shift` (between `2.0` to `5.0`) in the [Scheduler](https://huggingface.co/docs/diffusers/main/en/api/schedulers/flow_match_euler_discrete#diffusers.FlowMatchEulerDiscreteScheduler.shift). For larger resolution images, try higher values (between `7.0` and `12.0`). The default value is `7.0` for HunyuanVideo.
 - For more information about supported resolutions and other details, please refer to the original repository [here](https://github.com/Tencent/HunyuanVideo/).
 
 ## HunyuanVideoPipeline

src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py

@@ -39,7 +39,7 @@
         >>> from diffusers import HunyuanVideoPipeline, HunyuanVideoTransformer3DModel
         >>> from diffusers.utils import export_to_video
 
-        >>> model_id = "tencent/HunyuanVideo"
+        >>> model_id = "hunyuanvideo-community/HunyuanVideo"
         >>> transformer = HunyuanVideoTransformer3DModel.from_pretrained(
         ...     model_id, subfolder="transformer", torch_dtype=torch.bfloat16
         ... )

src/diffusers/pipelines/kandinsky/pipeline_kandinsky_combined.py

@@ -193,15 +193,15 @@ def __init__(
     def enable_xformers_memory_efficient_attention(self, attention_op: Optional[Callable] = None):
         self.decoder_pipe.enable_xformers_memory_efficient_attention(attention_op)
 
-    def enable_sequential_cpu_offload(self, gpu_id=0):
+    def enable_sequential_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models (`unet`, `text_encoder`, `vae`, and `safety checker` state dicts) to CPU using 🤗
         Accelerate, significantly reducing memory usage. Models are moved to a `torch.device('meta')` and loaded on a
         GPU only when their specific submodule's `forward` method is called. Offloading happens on a submodule basis.
         Memory savings are higher than using `enable_model_cpu_offload`, but performance is lower.
         """
-        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
-        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
+        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
 
     def progress_bar(self, iterable=None, total=None):
         self.prior_pipe.progress_bar(iterable=iterable, total=total)
@@ -411,16 +411,16 @@ def __init__(
     def enable_xformers_memory_efficient_attention(self, attention_op: Optional[Callable] = None):
         self.decoder_pipe.enable_xformers_memory_efficient_attention(attention_op)
 
-    def enable_sequential_cpu_offload(self, gpu_id=0):
+    def enable_sequential_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
         text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
         `torch.device('meta') and loaded to GPU only when their specific submodule has its `forward` method called.
         Note that offloading happens on a submodule basis. Memory savings are higher than with
         `enable_model_cpu_offload`, but performance is lower.
         """
-        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
-        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
+        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
 
     def progress_bar(self, iterable=None, total=None):
         self.prior_pipe.progress_bar(iterable=iterable, total=total)
@@ -652,16 +652,16 @@ def __init__(
     def enable_xformers_memory_efficient_attention(self, attention_op: Optional[Callable] = None):
         self.decoder_pipe.enable_xformers_memory_efficient_attention(attention_op)
 
-    def enable_sequential_cpu_offload(self, gpu_id=0):
+    def enable_sequential_cpu_offload(self, gpu_id: Optional[int] = None, device: Union[torch.device, str] = "cuda"):
         r"""
         Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
         text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
         `torch.device('meta') and loaded to GPU only when their specific submodule has its `forward` method called.
         Note that offloading happens on a submodule basis. Memory savings are higher than with
         `enable_model_cpu_offload`, but performance is lower.
         """
-        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
-        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id)
+        self.prior_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
+        self.decoder_pipe.enable_sequential_cpu_offload(gpu_id=gpu_id, device=device)
 
     def progress_bar(self, iterable=None, total=None):
         self.prior_pipe.progress_bar(iterable=iterable, total=total)