Fix: update type hints for Tuple parameters across multiple files to support variable-length tuples (#12544)

* Fix: update type hints for Tuple parameters across multiple files to support variable-length tuples

* Apply style fixes

---------

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

docs/source/en/modular_diffusers/loop_sequential_pipeline_blocks.md

@@ -12,7 +12,7 @@ specific language governing permissions and limitations under the License.
 
 # LoopSequentialPipelineBlocks
 
-[`~modular_pipelines.LoopSequentialPipelineBlocks`] are a multi-block type that composes other [`~modular_pipelines.ModularPipelineBlocks`] together in a loop. Data flows circularly, using `intermediate_inputs` and `intermediate_outputs`, and each block is run iteratively. This is typically used to create a denoising loop which is iterative by default.
+[`~modular_pipelines.LoopSequentialPipelineBlocks`] are a multi-block type that composes other [`~modular_pipelines.ModularPipelineBlocks`] together in a loop. Data flows circularly, using `inputs` and `intermediate_outputs`, and each block is run iteratively. This is typically used to create a denoising loop which is iterative by default.
 
 This guide shows you how to create [`~modular_pipelines.LoopSequentialPipelineBlocks`].
 
@@ -21,7 +21,6 @@ This guide shows you how to create [`~modular_pipelines.LoopSequentialPipelineBl
 [`~modular_pipelines.LoopSequentialPipelineBlocks`], is also known as the *loop wrapper* because it defines the loop structure, iteration variables, and configuration. Within the loop wrapper, you need the following variables.
 
 - `loop_inputs` are user provided values and equivalent to [`~modular_pipelines.ModularPipelineBlocks.inputs`].
-- `loop_intermediate_inputs` are intermediate variables from the [`~modular_pipelines.PipelineState`] and equivalent to [`~modular_pipelines.ModularPipelineBlocks.intermediate_inputs`].
 - `loop_intermediate_outputs` are new intermediate variables created by the block and added to the [`~modular_pipelines.PipelineState`]. It is equivalent to [`~modular_pipelines.ModularPipelineBlocks.intermediate_outputs`].
 - `__call__` method defines the loop structure and iteration logic.
 
@@ -90,4 +89,4 @@ Add more loop blocks to run within each iteration with [`~modular_pipelines.Loop
 
 ```py
 loop = LoopWrapper.from_blocks_dict({"block1": LoopBlock(), "block2": LoopBlock})
-```
+```

docs/source/en/modular_diffusers/pipeline_block.md

@@ -37,17 +37,7 @@ A [`~modular_pipelines.ModularPipelineBlocks`] requires `inputs`, and `intermedi
     ]
     ```
 
-- `intermediate_inputs` are values typically created from a previous block but it can also be directly provided if no preceding block generates them. Unlike `inputs`, `intermediate_inputs` can be modified.
-
-    Use `InputParam` to define `intermediate_inputs`.
-
-    ```py
-    user_intermediate_inputs = [
-        InputParam(name="processed_image", type_hint="torch.Tensor", description="image that has been preprocessed and normalized"),
-    ]
-    ```
-
-- `intermediate_outputs` are new values created by a block and added to the [`~modular_pipelines.PipelineState`]. The `intermediate_outputs` are available as `intermediate_inputs` for subsequent blocks or available as the final output from running the pipeline.
+- `intermediate_outputs` are new values created by a block and added to the [`~modular_pipelines.PipelineState`]. The `intermediate_outputs` are available as `inputs` for subsequent blocks or available as the final output from running the pipeline.
 
     Use `OutputParam` to define `intermediate_outputs`.
 
@@ -65,8 +55,8 @@ The intermediate inputs and outputs share data to connect blocks. They are acces
 
 The computation a block performs is defined in the `__call__` method and it follows a specific structure.
 
-1. Retrieve the [`~modular_pipelines.BlockState`] to get a local view of the `inputs` and `intermediate_inputs`.
-2. Implement the computation logic on the `inputs` and `intermediate_inputs`.
+1. Retrieve the [`~modular_pipelines.BlockState`] to get a local view of the `inputs`
+2. Implement the computation logic on the `inputs`.
 3. Update [`~modular_pipelines.PipelineState`] to push changes from the local [`~modular_pipelines.BlockState`] back to the global [`~modular_pipelines.PipelineState`].
 4. Return the components and state which becomes available to the next block.
 
@@ -76,7 +66,7 @@ def __call__(self, components, state):
     block_state = self.get_block_state(state)
 
     # Your computation logic here
-    # block_state contains all your inputs and intermediate_inputs
+    # block_state contains all your inputs
     # Access them like: block_state.image, block_state.processed_image
 
     # Update the pipeline state with your updated block_states
@@ -112,4 +102,4 @@ def __call__(self, components, state):
     unet = components.unet
     vae = components.vae
     scheduler = components.scheduler
-```
+```

docs/source/en/modular_diffusers/quickstart.md

@@ -183,7 +183,7 @@ from diffusers.modular_pipelines import ComponentsManager
 components = ComponentManager()
 
 dd_pipeline = dd_blocks.init_pipeline("YiYiXu/modular-demo-auto", components_manager=components, collection="diffdiff")
-dd_pipeline.load_default_componenets(torch_dtype=torch.float16)
+dd_pipeline.load_componenets(torch_dtype=torch.float16)
 dd_pipeline.to("cuda")
 ```
 

docs/source/en/modular_diffusers/sequential_pipeline_blocks.md

@@ -12,11 +12,11 @@ specific language governing permissions and limitations under the License.
 
 # SequentialPipelineBlocks
 
-[`~modular_pipelines.SequentialPipelineBlocks`] are a multi-block type that composes other [`~modular_pipelines.ModularPipelineBlocks`] together in a sequence. Data flows linearly from one block to the next using `intermediate_inputs` and `intermediate_outputs`. Each block in [`~modular_pipelines.SequentialPipelineBlocks`] usually represents a step in the pipeline, and by combining them, you gradually build a pipeline.
+[`~modular_pipelines.SequentialPipelineBlocks`] are a multi-block type that composes other [`~modular_pipelines.ModularPipelineBlocks`] together in a sequence. Data flows linearly from one block to the next using `inputs` and `intermediate_outputs`. Each block in [`~modular_pipelines.SequentialPipelineBlocks`] usually represents a step in the pipeline, and by combining them, you gradually build a pipeline.
 
 This guide shows you how to connect two blocks into a [`~modular_pipelines.SequentialPipelineBlocks`].
 
-Create two [`~modular_pipelines.ModularPipelineBlocks`]. The first block, `InputBlock`, outputs a `batch_size` value and the second block, `ImageEncoderBlock` uses `batch_size` as `intermediate_inputs`.
+Create two [`~modular_pipelines.ModularPipelineBlocks`]. The first block, `InputBlock`, outputs a `batch_size` value and the second block, `ImageEncoderBlock` uses `batch_size` as `inputs`.
 
 <hfoptions id="sequential">
 <hfoption id="InputBlock">
@@ -110,4 +110,4 @@ Inspect the sub-blocks in [`~modular_pipelines.SequentialPipelineBlocks`] by cal
 ```py
 print(blocks)
 print(blocks.doc)
-```
+```

examples/community/img2img_inpainting.py

@@ -45,7 +45,7 @@ def check_size(image, height, width):
         raise ValueError(f"Image size should be {height}x{width}, but got {h}x{w}")
 
 
-def overlay_inner_image(image, inner_image, paste_offset: Tuple[int] = (0, 0)):
+def overlay_inner_image(image, inner_image, paste_offset: Tuple[int, ...] = (0, 0)):
     inner_image = inner_image.convert("RGBA")
     image = image.convert("RGB")
 

examples/community/matryoshka.py

@@ -1966,16 +1966,21 @@ class MatryoshkaUNet2DConditionModel(
         center_input_sample: bool = False,
         flip_sin_to_cos: bool = True,
         freq_shift: int = 0,
-        down_block_types: Tuple[str] = (
+        down_block_types: Tuple[str, ...] = (
             "CrossAttnDownBlock2D",
             "CrossAttnDownBlock2D",
             "CrossAttnDownBlock2D",
             "DownBlock2D",
         ),
         mid_block_type: Optional[str] = "UNetMidBlock2DCrossAttn",
-        up_block_types: Tuple[str] = ("UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D"),
+        up_block_types: Tuple[str, ...] = (
+            "UpBlock2D",
+            "CrossAttnUpBlock2D",
+            "CrossAttnUpBlock2D",
+            "CrossAttnUpBlock2D",
+        ),
         only_cross_attention: Union[bool, Tuple[bool]] = False,
-        block_out_channels: Tuple[int] = (320, 640, 1280, 1280),
+        block_out_channels: Tuple[int, ...] = (320, 640, 1280, 1280),
         layers_per_block: Union[int, Tuple[int]] = 2,
         downsample_padding: int = 1,
         mid_block_scale_factor: float = 1,
@@ -2294,10 +2299,10 @@ class MatryoshkaUNet2DConditionModel(
 
     def _check_config(
         self,
-        down_block_types: Tuple[str],
-        up_block_types: Tuple[str],
+        down_block_types: Tuple[str, ...],
+        up_block_types: Tuple[str, ...],
         only_cross_attention: Union[bool, Tuple[bool]],
-        block_out_channels: Tuple[int],
+        block_out_channels: Tuple[int, ...],
         layers_per_block: Union[int, Tuple[int]],
         cross_attention_dim: Union[int, Tuple[int]],
         transformer_layers_per_block: Union[int, Tuple[int], Tuple[Tuple[int]]],

examples/community/pipeline_faithdiff_stable_diffusion_xl.py

@@ -438,16 +438,21 @@ class UNet2DConditionModel(OriginalUNet2DConditionModel, ConfigMixin, UNet2DCond
         center_input_sample: bool = False,
         flip_sin_to_cos: bool = True,
         freq_shift: int = 0,
-        down_block_types: Tuple[str] = (
+        down_block_types: Tuple[str, ...] = (
             "CrossAttnDownBlock2D",
             "CrossAttnDownBlock2D",
             "CrossAttnDownBlock2D",
             "DownBlock2D",
         ),
         mid_block_type: Optional[str] = "UNetMidBlock2DCrossAttn",
-        up_block_types: Tuple[str] = ("UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D"),
+        up_block_types: Tuple[str, ...] = (
+            "UpBlock2D",
+            "CrossAttnUpBlock2D",
+            "CrossAttnUpBlock2D",
+            "CrossAttnUpBlock2D",
+        ),
         only_cross_attention: Union[bool, Tuple[bool]] = False,
-        block_out_channels: Tuple[int] = (320, 640, 1280, 1280),
+        block_out_channels: Tuple[int, ...] = (320, 640, 1280, 1280),
         layers_per_block: Union[int, Tuple[int]] = 2,
         downsample_padding: int = 1,
         mid_block_scale_factor: float = 1,

src/diffusers/models/_modeling_parallel.py

@@ -44,11 +44,16 @@ class ContextParallelConfig:
 
     Args:
         ring_degree (`int`, *optional*, defaults to `1`):
-            Number of devices to use for ring attention within a context parallel region. Must be a divisor of the
-            total number of devices in the context parallel mesh.
+            Number of devices to use for Ring Attention. Sequence is split across devices. Each device computes
+            attention between its local Q and KV chunks passed sequentially around ring. Lower memory (only holds 1/N
+            of KV at a time), overlaps compute with communication, but requires N iterations to see all tokens. Best
+            for long sequences with limited memory/bandwidth. Number of devices to use for ring attention within a
+            context parallel region. Must be a divisor of the total number of devices in the context parallel mesh.
         ulysses_degree (`int`, *optional*, defaults to `1`):
-            Number of devices to use for ulysses attention within a context parallel region. Must be a divisor of the
-            total number of devices in the context parallel mesh.
+            Number of devices to use for Ulysses Attention. Sequence split is across devices. Each device computes
+            local QKV, then all-gathers all KV chunks to compute full attention in one pass. Higher memory (stores all
+            KV), requires high-bandwidth all-to-all communication, but lower latency. Best for moderate sequences with
+            good interconnect bandwidth.
         convert_to_fp32 (`bool`, *optional*, defaults to `True`):
             Whether to convert output and LSE to float32 for ring attention numerical stability.
         rotate_method (`str`, *optional*, defaults to `"allgather"`):
@@ -79,29 +84,46 @@ class ContextParallelConfig:
         if self.ulysses_degree is None:
             self.ulysses_degree = 1
 
+        if self.ring_degree == 1 and self.ulysses_degree == 1:
+            raise ValueError(
+                "Either ring_degree or ulysses_degree must be greater than 1 in order to use context parallel inference"
+            )
+        if self.ring_degree < 1 or self.ulysses_degree < 1:
+            raise ValueError("`ring_degree` and `ulysses_degree` must be greater than or equal to 1.")
+        if self.ring_degree > 1 and self.ulysses_degree > 1:
+            raise ValueError(
+                "Unified Ulysses-Ring attention is not yet supported. Please set either `ring_degree` or `ulysses_degree` to 1."
+            )
+        if self.rotate_method != "allgather":
+            raise NotImplementedError(
+                f"Only rotate_method='allgather' is supported for now, but got {self.rotate_method}."
+            )
+
+    @property
+    def mesh_shape(self) -> Tuple[int, int]:
+        return (self.ring_degree, self.ulysses_degree)
+
+    @property
+    def mesh_dim_names(self) -> Tuple[str, str]:
+        """Dimension names for the device mesh."""
+        return ("ring", "ulysses")
+
     def setup(self, rank: int, world_size: int, device: torch.device, mesh: torch.distributed.device_mesh.DeviceMesh):
         self._rank = rank
         self._world_size = world_size
         self._device = device
         self._mesh = mesh
-        if self.ring_degree is None:
-            self.ring_degree = 1
-        if self.ulysses_degree is None:
-            self.ulysses_degree = 1
-        if self.rotate_method != "allgather":
-            raise NotImplementedError(
-                f"Only rotate_method='allgather' is supported for now, but got {self.rotate_method}."
+
+        if self.ulysses_degree * self.ring_degree > world_size:
+            raise ValueError(
+                f"The product of `ring_degree` ({self.ring_degree}) and `ulysses_degree` ({self.ulysses_degree}) must not exceed the world size ({world_size})."
             )
-        if self._flattened_mesh is None:
-            self._flattened_mesh = self._mesh._flatten()
-        if self._ring_mesh is None:
-            self._ring_mesh = self._mesh["ring"]
-        if self._ulysses_mesh is None:
-            self._ulysses_mesh = self._mesh["ulysses"]
-        if self._ring_local_rank is None:
-            self._ring_local_rank = self._ring_mesh.get_local_rank()
-        if self._ulysses_local_rank is None:
-            self._ulysses_local_rank = self._ulysses_mesh.get_local_rank()
+
+        self._flattened_mesh = self._mesh._flatten()
+        self._ring_mesh = self._mesh["ring"]
+        self._ulysses_mesh = self._mesh["ulysses"]
+        self._ring_local_rank = self._ring_mesh.get_local_rank()
+        self._ulysses_local_rank = self._ulysses_mesh.get_local_rank()
 
 
 @dataclass
@@ -119,7 +141,7 @@ class ParallelConfig:
     _rank: int = None
     _world_size: int = None
     _device: torch.device = None
-    _cp_mesh: torch.distributed.device_mesh.DeviceMesh = None
+    _mesh: torch.distributed.device_mesh.DeviceMesh = None
 
     def setup(
         self,
@@ -127,14 +149,14 @@ class ParallelConfig:
         world_size: int,
         device: torch.device,
         *,
-        cp_mesh: Optional[torch.distributed.device_mesh.DeviceMesh] = None,
+        mesh: Optional[torch.distributed.device_mesh.DeviceMesh] = None,
     ):
         self._rank = rank
         self._world_size = world_size
         self._device = device
-        self._cp_mesh = cp_mesh
+        self._mesh = mesh
         if self.context_parallel_config is not None:
-            self.context_parallel_config.setup(rank, world_size, device, cp_mesh)
+            self.context_parallel_config.setup(rank, world_size, device, mesh)
 
 
 @dataclass(frozen=True)

src/diffusers/models/attention_dispatch.py

@@ -220,7 +220,7 @@ class _AttentionBackendRegistry:
     _backends = {}
     _constraints = {}
     _supported_arg_names = {}
-    _supports_context_parallel = {}
+    _supports_context_parallel = set()
     _active_backend = AttentionBackendName(DIFFUSERS_ATTN_BACKEND)
     _checks_enabled = DIFFUSERS_ATTN_CHECKS
 
@@ -237,7 +237,9 @@ class _AttentionBackendRegistry:
             cls._backends[backend] = func
             cls._constraints[backend] = constraints or []
             cls._supported_arg_names[backend] = set(inspect.signature(func).parameters.keys())
-            cls._supports_context_parallel[backend] = supports_context_parallel
+            if supports_context_parallel:
+                cls._supports_context_parallel.add(backend.value)
+
             return func
 
         return decorator
@@ -251,15 +253,12 @@ class _AttentionBackendRegistry:
         return list(cls._backends.keys())
 
     @classmethod
-    def _is_context_parallel_enabled(
-        cls, backend: AttentionBackendName, parallel_config: Optional["ParallelConfig"]
+    def _is_context_parallel_available(
+        cls,
+        backend: AttentionBackendName,
     ) -> bool:
-        supports_context_parallel = backend in cls._supports_context_parallel
-        is_degree_greater_than_1 = parallel_config is not None and (
-            parallel_config.context_parallel_config.ring_degree > 1
-            or parallel_config.context_parallel_config.ulysses_degree > 1
-        )
-        return supports_context_parallel and is_degree_greater_than_1
+        supports_context_parallel = backend.value in cls._supports_context_parallel
+        return supports_context_parallel
 
 
 @contextlib.contextmanager
@@ -306,14 +305,6 @@ def dispatch_attention_fn(
         backend_name = AttentionBackendName(backend)
         backend_fn = _AttentionBackendRegistry._backends.get(backend_name)
 
-    if parallel_config is not None and not _AttentionBackendRegistry._is_context_parallel_enabled(
-        backend_name, parallel_config
-    ):
-        raise ValueError(
-            f"Backend {backend_name} either does not support context parallelism or context parallelism "
-            f"was enabled with a world size of 1."
-        )
-
     kwargs = {
         "query": query,
         "key": key,

src/diffusers/models/autoencoders/autoencoder_dc.py

@@ -102,7 +102,7 @@ def get_block(
     attention_head_dim: int,
     norm_type: str,
     act_fn: str,
-    qkv_mutliscales: Tuple[int] = (),
+    qkv_mutliscales: Tuple[int, ...] = (),
 ):
     if block_type == "ResBlock":
         block = ResBlock(in_channels, out_channels, norm_type, act_fn)
@@ -206,8 +206,8 @@ class Encoder(nn.Module):
         latent_channels: int,
         attention_head_dim: int = 32,
         block_type: Union[str, Tuple[str]] = "ResBlock",
-        block_out_channels: Tuple[int] = (128, 256, 512, 512, 1024, 1024),
-        layers_per_block: Tuple[int] = (2, 2, 2, 2, 2, 2),
+        block_out_channels: Tuple[int, ...] = (128, 256, 512, 512, 1024, 1024),
+        layers_per_block: Tuple[int, ...] = (2, 2, 2, 2, 2, 2),
         qkv_multiscales: Tuple[Tuple[int, ...], ...] = ((), (), (), (5,), (5,), (5,)),
         downsample_block_type: str = "pixel_unshuffle",
         out_shortcut: bool = True,
@@ -292,8 +292,8 @@ class Decoder(nn.Module):
         latent_channels: int,
         attention_head_dim: int = 32,
         block_type: Union[str, Tuple[str]] = "ResBlock",
-        block_out_channels: Tuple[int] = (128, 256, 512, 512, 1024, 1024),
-        layers_per_block: Tuple[int] = (2, 2, 2, 2, 2, 2),
+        block_out_channels: Tuple[int, ...] = (128, 256, 512, 512, 1024, 1024),
+        layers_per_block: Tuple[int, ...] = (2, 2, 2, 2, 2, 2),
         qkv_multiscales: Tuple[Tuple[int, ...], ...] = ((), (), (), (5,), (5,), (5,)),
         norm_type: Union[str, Tuple[str]] = "rms_norm",
         act_fn: Union[str, Tuple[str]] = "silu",
@@ -440,8 +440,8 @@ class AutoencoderDC(ModelMixin, AutoencoderMixin, ConfigMixin, FromOriginalModel
         decoder_block_types: Union[str, Tuple[str]] = "ResBlock",
         encoder_block_out_channels: Tuple[int, ...] = (128, 256, 512, 512, 1024, 1024),
         decoder_block_out_channels: Tuple[int, ...] = (128, 256, 512, 512, 1024, 1024),
-        encoder_layers_per_block: Tuple[int] = (2, 2, 2, 3, 3, 3),
-        decoder_layers_per_block: Tuple[int] = (3, 3, 3, 3, 3, 3),
+        encoder_layers_per_block: Tuple[int, ...] = (2, 2, 2, 3, 3, 3),
+        decoder_layers_per_block: Tuple[int, ...] = (3, 3, 3, 3, 3, 3),
         encoder_qkv_multiscales: Tuple[Tuple[int, ...], ...] = ((), (), (), (5,), (5,), (5,)),
         decoder_qkv_multiscales: Tuple[Tuple[int, ...], ...] = ((), (), (), (5,), (5,), (5,)),
         upsample_block_type: str = "pixel_shuffle",

src/diffusers/models/autoencoders/autoencoder_kl.py

@@ -78,9 +78,9 @@ class AutoencoderKL(ModelMixin, AutoencoderMixin, ConfigMixin, FromOriginalModel
         self,
         in_channels: int = 3,
         out_channels: int = 3,
-        down_block_types: Tuple[str] = ("DownEncoderBlock2D",),
-        up_block_types: Tuple[str] = ("UpDecoderBlock2D",),
-        block_out_channels: Tuple[int] = (64,),
+        down_block_types: Tuple[str, ...] = ("DownEncoderBlock2D",),
+        up_block_types: Tuple[str, ...] = ("UpDecoderBlock2D",),
+        block_out_channels: Tuple[int, ...] = (64,),
         layers_per_block: int = 1,
         act_fn: str = "silu",
         latent_channels: int = 4,

src/diffusers/models/autoencoders/autoencoder_kl_cogvideox.py

@@ -995,19 +995,19 @@ class AutoencoderKLCogVideoX(ModelMixin, AutoencoderMixin, ConfigMixin, FromOrig
         self,
         in_channels: int = 3,
         out_channels: int = 3,
-        down_block_types: Tuple[str] = (
+        down_block_types: Tuple[str, ...] = (
             "CogVideoXDownBlock3D",
             "CogVideoXDownBlock3D",
             "CogVideoXDownBlock3D",
             "CogVideoXDownBlock3D",
         ),
-        up_block_types: Tuple[str] = (
+        up_block_types: Tuple[str, ...] = (
             "CogVideoXUpBlock3D",
             "CogVideoXUpBlock3D",
             "CogVideoXUpBlock3D",
             "CogVideoXUpBlock3D",
         ),
-        block_out_channels: Tuple[int] = (128, 256, 256, 512),
+        block_out_channels: Tuple[int, ...] = (128, 256, 256, 512),
         latent_channels: int = 16,
         layers_per_block: int = 3,
         act_fn: str = "silu",

src/diffusers/models/autoencoders/autoencoder_kl_hunyuan_video.py

@@ -653,7 +653,7 @@ class AutoencoderKLHunyuanVideo(ModelMixin, AutoencoderMixin, ConfigMixin):
             "HunyuanVideoUpBlock3D",
             "HunyuanVideoUpBlock3D",
         ),
-        block_out_channels: Tuple[int] = (128, 256, 512, 512),
+        block_out_channels: Tuple[int, ...] = (128, 256, 512, 512),
         layers_per_block: int = 2,
         act_fn: str = "silu",
         norm_num_groups: int = 32,

src/diffusers/models/autoencoders/autoencoder_kl_hunyuanimage_refiner.py

@@ -601,7 +601,7 @@ class AutoencoderKLHunyuanImageRefiner(ModelMixin, ConfigMixin):
         in_channels: int = 3,
         out_channels: int = 3,
         latent_channels: int = 32,
-        block_out_channels: Tuple[int] = (128, 256, 512, 1024, 1024),
+        block_out_channels: Tuple[int, ...] = (128, 256, 512, 1024, 1024),
         layers_per_block: int = 2,
         spatial_compression_ratio: int = 16,
         temporal_compression_ratio: int = 4,

src/diffusers/models/autoencoders/autoencoder_kl_mochi.py

@@ -688,8 +688,8 @@ class AutoencoderKLMochi(ModelMixin, AutoencoderMixin, ConfigMixin):
         self,
         in_channels: int = 15,
         out_channels: int = 3,
-        encoder_block_out_channels: Tuple[int] = (64, 128, 256, 384),
-        decoder_block_out_channels: Tuple[int] = (128, 256, 512, 768),
+        encoder_block_out_channels: Tuple[int, ...] = (64, 128, 256, 384),
+        decoder_block_out_channels: Tuple[int, ...] = (128, 256, 512, 768),
         latent_channels: int = 12,
         layers_per_block: Tuple[int, ...] = (3, 3, 4, 6, 3),
         act_fn: str = "silu",

src/diffusers/models/autoencoders/autoencoder_kl_qwenimage.py

@@ -679,7 +679,7 @@ class AutoencoderKLQwenImage(ModelMixin, AutoencoderMixin, ConfigMixin, FromOrig
         self,
         base_dim: int = 96,
         z_dim: int = 16,
-        dim_mult: Tuple[int] = [1, 2, 4, 4],
+        dim_mult: Tuple[int, ...] = (1, 2, 4, 4),
         num_res_blocks: int = 2,
         attn_scales: List[float] = [],
         temperal_downsample: List[bool] = [False, True, True],

src/diffusers/models/autoencoders/autoencoder_kl_temporal_decoder.py

@@ -31,7 +31,7 @@ class TemporalDecoder(nn.Module):
         self,
         in_channels: int = 4,
         out_channels: int = 3,
-        block_out_channels: Tuple[int] = (128, 256, 512, 512),
+        block_out_channels: Tuple[int, ...] = (128, 256, 512, 512),
         layers_per_block: int = 2,
     ):
         super().__init__()
@@ -172,8 +172,8 @@ class AutoencoderKLTemporalDecoder(ModelMixin, AutoencoderMixin, ConfigMixin):
         self,
         in_channels: int = 3,
         out_channels: int = 3,
-        down_block_types: Tuple[str] = ("DownEncoderBlock2D",),
-        block_out_channels: Tuple[int] = (64,),
+        down_block_types: Tuple[str, ...] = ("DownEncoderBlock2D",),
+        block_out_channels: Tuple[int, ...] = (64,),
         layers_per_block: int = 1,
         latent_channels: int = 4,
         sample_size: int = 32,

src/diffusers/models/autoencoders/autoencoder_kl_wan.py

@@ -971,7 +971,7 @@ class AutoencoderKLWan(ModelMixin, AutoencoderMixin, ConfigMixin, FromOriginalMo
         base_dim: int = 96,
         decoder_base_dim: Optional[int] = None,
         z_dim: int = 16,
-        dim_mult: Tuple[int] = [1, 2, 4, 4],
+        dim_mult: Tuple[int, ...] = (1, 2, 4, 4),
         num_res_blocks: int = 2,
         attn_scales: List[float] = [],
         temperal_downsample: List[bool] = [False, True, True],

src/diffusers/models/controlnets/controlnet_xs.py

@@ -293,14 +293,14 @@ class ControlNetXSAdapter(ModelMixin, ConfigMixin):
         self,
         conditioning_channels: int = 3,
         conditioning_channel_order: str = "rgb",
-        conditioning_embedding_out_channels: Tuple[int] = (16, 32, 96, 256),
+        conditioning_embedding_out_channels: Tuple[int, ...] = (16, 32, 96, 256),
         time_embedding_mix: float = 1.0,
         learn_time_embedding: bool = False,
         num_attention_heads: Union[int, Tuple[int]] = 4,
-        block_out_channels: Tuple[int] = (4, 8, 16, 16),
-        base_block_out_channels: Tuple[int] = (320, 640, 1280, 1280),
+        block_out_channels: Tuple[int, ...] = (4, 8, 16, 16),
+        base_block_out_channels: Tuple[int, ...] = (320, 640, 1280, 1280),
         cross_attention_dim: int = 1024,
-        down_block_types: Tuple[str] = (
+        down_block_types: Tuple[str, ...] = (
             "CrossAttnDownBlock2D",
             "CrossAttnDownBlock2D",
             "CrossAttnDownBlock2D",
@@ -436,7 +436,7 @@ class ControlNetXSAdapter(ModelMixin, ConfigMixin):
         time_embedding_mix: int = 1.0,
         conditioning_channels: int = 3,
         conditioning_channel_order: str = "rgb",
-        conditioning_embedding_out_channels: Tuple[int] = (16, 32, 96, 256),
+        conditioning_embedding_out_channels: Tuple[int, ...] = (16, 32, 96, 256),
     ):
         r"""
         Instantiate a [`ControlNetXSAdapter`] from a [`UNet2DConditionModel`].
@@ -529,14 +529,19 @@ class UNetControlNetXSModel(ModelMixin, ConfigMixin):
         self,
         # unet configs
         sample_size: Optional[int] = 96,
-        down_block_types: Tuple[str] = (
+        down_block_types: Tuple[str, ...] = (
             "CrossAttnDownBlock2D",
             "CrossAttnDownBlock2D",
             "CrossAttnDownBlock2D",
             "DownBlock2D",
         ),
-        up_block_types: Tuple[str] = ("UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D"),
-        block_out_channels: Tuple[int] = (320, 640, 1280, 1280),
+        up_block_types: Tuple[str, ...] = (
+            "UpBlock2D",
+            "CrossAttnUpBlock2D",
+            "CrossAttnUpBlock2D",
+            "CrossAttnUpBlock2D",
+        ),
+        block_out_channels: Tuple[int, ...] = (320, 640, 1280, 1280),
         norm_num_groups: Optional[int] = 32,
         cross_attention_dim: Union[int, Tuple[int]] = 1024,
         transformer_layers_per_block: Union[int, Tuple[int]] = 1,
@@ -550,10 +555,10 @@ class UNetControlNetXSModel(ModelMixin, ConfigMixin):
         # additional controlnet configs
         time_embedding_mix: float = 1.0,
         ctrl_conditioning_channels: int = 3,
-        ctrl_conditioning_embedding_out_channels: Tuple[int] = (16, 32, 96, 256),
+        ctrl_conditioning_embedding_out_channels: Tuple[int, ...] = (16, 32, 96, 256),
         ctrl_conditioning_channel_order: str = "rgb",
         ctrl_learn_time_embedding: bool = False,
-        ctrl_block_out_channels: Tuple[int] = (4, 8, 16, 16),
+        ctrl_block_out_channels: Tuple[int, ...] = (4, 8, 16, 16),
         ctrl_num_attention_heads: Union[int, Tuple[int]] = 4,
         ctrl_max_norm_num_groups: int = 32,
     ):

src/diffusers/models/modeling_utils.py

@@ -1484,59 +1484,71 @@ class ModelMixin(torch.nn.Module, PushToHubMixin):
         config: Union[ParallelConfig, ContextParallelConfig],
         cp_plan: Optional[Dict[str, ContextParallelModelPlan]] = None,
     ):
-        from ..hooks.context_parallel import apply_context_parallel
-        from .attention import AttentionModuleMixin
-        from .attention_processor import Attention, MochiAttention
-
         logger.warning(
             "`enable_parallelism` is an experimental feature. The API may change in the future and breaking changes may be introduced at any time without warning."
         )
 
+        if not torch.distributed.is_available() and not torch.distributed.is_initialized():
+            raise RuntimeError(
+                "torch.distributed must be available and initialized before calling `enable_parallelism`."
+            )
+
+        from ..hooks.context_parallel import apply_context_parallel
+        from .attention import AttentionModuleMixin
+        from .attention_dispatch import AttentionBackendName, _AttentionBackendRegistry
+        from .attention_processor import Attention, MochiAttention
+
         if isinstance(config, ContextParallelConfig):
             config = ParallelConfig(context_parallel_config=config)
 
-        if not torch.distributed.is_initialized():
-            raise RuntimeError("torch.distributed must be initialized before calling `enable_parallelism`.")
-
         rank = torch.distributed.get_rank()
         world_size = torch.distributed.get_world_size()
         device_type = torch._C._get_accelerator().type
         device_module = torch.get_device_module(device_type)
         device = torch.device(device_type, rank % device_module.device_count())
 
-        cp_mesh = None
+        attention_classes = (Attention, MochiAttention, AttentionModuleMixin)
+
+        if config.context_parallel_config is not None:
+            for module in self.modules():
+                if not isinstance(module, attention_classes):
+                    continue
+
+                processor = module.processor
+                if processor is None or not hasattr(processor, "_attention_backend"):
+                    continue
+
+                attention_backend = processor._attention_backend
+                if attention_backend is None:
+                    attention_backend, _ = _AttentionBackendRegistry.get_active_backend()
+                else:
+                    attention_backend = AttentionBackendName(attention_backend)
+
+                if not _AttentionBackendRegistry._is_context_parallel_available(attention_backend):
+                    compatible_backends = sorted(_AttentionBackendRegistry._supports_context_parallel)
+                    raise ValueError(
+                        f"Context parallelism is enabled but the attention processor '{processor.__class__.__name__}' "
+                        f"is using backend '{attention_backend.value}' which does not support context parallelism. "
+                        f"Please set a compatible attention backend: {compatible_backends} using `model.set_attention_backend()` before "
+                        f"calling `enable_parallelism()`."
+                    )
+
+                # All modules use the same attention processor and backend. We don't need to
+                # iterate over all modules after checking the first processor
+                break
+
+        mesh = None
         if config.context_parallel_config is not None:
             cp_config = config.context_parallel_config
-            if cp_config.ring_degree < 1 or cp_config.ulysses_degree < 1:
-                raise ValueError("`ring_degree` and `ulysses_degree` must be greater than or equal to 1.")
-            if cp_config.ring_degree > 1 and cp_config.ulysses_degree > 1:
-                raise ValueError(
-                    "Unified Ulysses-Ring attention is not yet supported. Please set either `ring_degree` or `ulysses_degree` to 1."
-                )
-            if cp_config.ring_degree * cp_config.ulysses_degree > world_size:
-                raise ValueError(
-                    f"The product of `ring_degree` ({cp_config.ring_degree}) and `ulysses_degree` ({cp_config.ulysses_degree}) must not exceed the world size ({world_size})."
-                )
-            cp_mesh = torch.distributed.device_mesh.init_device_mesh(
+            mesh = torch.distributed.device_mesh.init_device_mesh(
                 device_type=device_type,
-                mesh_shape=(cp_config.ring_degree, cp_config.ulysses_degree),
-                mesh_dim_names=("ring", "ulysses"),
+                mesh_shape=cp_config.mesh_shape,
+                mesh_dim_names=cp_config.mesh_dim_names,
             )
 
-        config.setup(rank, world_size, device, cp_mesh=cp_mesh)
-
-        if cp_plan is None and self._cp_plan is None:
-            raise ValueError(
-                "`cp_plan` must be provided either as an argument or set in the model's `_cp_plan` attribute."
-            )
-        cp_plan = cp_plan if cp_plan is not None else self._cp_plan
-
-        if config.context_parallel_config is not None:
-            apply_context_parallel(self, config.context_parallel_config, cp_plan)
-
+        config.setup(rank, world_size, device, mesh=mesh)
         self._parallel_config = config
 
-        attention_classes = (Attention, MochiAttention, AttentionModuleMixin)
         for module in self.modules():
             if not isinstance(module, attention_classes):
                 continue
@@ -1545,6 +1557,14 @@ class ModelMixin(torch.nn.Module, PushToHubMixin):
                 continue
             processor._parallel_config = config
 
+        if config.context_parallel_config is not None:
+            if cp_plan is None and self._cp_plan is None:
+                raise ValueError(
+                    "`cp_plan` must be provided either as an argument or set in the model's `_cp_plan` attribute."
+                )
+            cp_plan = cp_plan if cp_plan is not None else self._cp_plan
+            apply_context_parallel(self, config.context_parallel_config, cp_plan)
+
     @classmethod
     def _load_pretrained_model(
         cls,

src/diffusers/models/transformers/transformer_hunyuan_video.py

@@ -914,7 +914,7 @@ class HunyuanVideoTransformer3DModel(ModelMixin, ConfigMixin, PeftAdapterMixin,
         text_embed_dim: int = 4096,
         pooled_projection_dim: int = 768,
         rope_theta: float = 256.0,
-        rope_axes_dim: Tuple[int] = (16, 56, 56),
+        rope_axes_dim: Tuple[int, ...] = (16, 56, 56),
         image_condition_type: Optional[str] = None,
     ) -> None:
         super().__init__()

src/diffusers/models/transformers/transformer_hunyuan_video_framepack.py

@@ -139,7 +139,7 @@ class HunyuanVideoFramepackTransformer3DModel(
         text_embed_dim: int = 4096,
         pooled_projection_dim: int = 768,
         rope_theta: float = 256.0,
-        rope_axes_dim: Tuple[int] = (16, 56, 56),
+        rope_axes_dim: Tuple[int, ...] = (16, 56, 56),
         image_condition_type: Optional[str] = None,
         has_image_proj: int = False,
         image_proj_dim: int = 1152,

src/diffusers/models/transformers/transformer_hunyuanimage.py

@@ -689,7 +689,7 @@ class HunyuanImageTransformer2DModel(ModelMixin, ConfigMixin, PeftAdapterMixin,
         text_embed_dim: int = 3584,
         text_embed_2_dim: Optional[int] = None,
         rope_theta: float = 256.0,
-        rope_axes_dim: Tuple[int] = (64, 64),
+        rope_axes_dim: Tuple[int, ...] = (64, 64),
         use_meanflow: bool = False,
     ) -> None:
         super().__init__()

src/diffusers/models/transformers/transformer_skyreels_v2.py

@@ -570,7 +570,7 @@ class SkyReelsV2Transformer3DModel(
     @register_to_config
     def __init__(
         self,
-        patch_size: Tuple[int] = (1, 2, 2),
+        patch_size: Tuple[int, ...] = (1, 2, 2),
         num_attention_heads: int = 16,
         attention_head_dim: int = 128,
         in_channels: int = 16,

src/diffusers/models/transformers/transformer_wan.py

@@ -563,7 +563,7 @@ class WanTransformer3DModel(
     @register_to_config
     def __init__(
         self,
-        patch_size: Tuple[int] = (1, 2, 2),
+        patch_size: Tuple[int, ...] = (1, 2, 2),
         num_attention_heads: int = 40,
         attention_head_dim: int = 128,
         in_channels: int = 16,

src/diffusers/models/transformers/transformer_wan_vace.py

@@ -182,7 +182,7 @@ class WanVACETransformer3DModel(
     @register_to_config
     def __init__(
         self,
-        patch_size: Tuple[int] = (1, 2, 2),
+        patch_size: Tuple[int, ...] = (1, 2, 2),
         num_attention_heads: int = 40,
         attention_head_dim: int = 128,
         in_channels: int = 16,

src/diffusers/models/unets/unet_1d.py

@@ -86,11 +86,11 @@ class UNet1DModel(ModelMixin, ConfigMixin):
         flip_sin_to_cos: bool = True,
         use_timestep_embedding: bool = False,
         freq_shift: float = 0.0,
-        down_block_types: Tuple[str] = ("DownBlock1DNoSkip", "DownBlock1D", "AttnDownBlock1D"),
-        up_block_types: Tuple[str] = ("AttnUpBlock1D", "UpBlock1D", "UpBlock1DNoSkip"),
-        mid_block_type: Tuple[str] = "UNetMidBlock1D",
+        down_block_types: Tuple[str, ...] = ("DownBlock1DNoSkip", "DownBlock1D", "AttnDownBlock1D"),
+        up_block_types: Tuple[str, ...] = ("AttnUpBlock1D", "UpBlock1D", "UpBlock1DNoSkip"),
+        mid_block_type: str = "UNetMidBlock1D",
         out_block_type: str = None,
-        block_out_channels: Tuple[int] = (32, 32, 64),
+        block_out_channels: Tuple[int, ...] = (32, 32, 64),
         act_fn: str = None,
         norm_num_groups: int = 8,
         layers_per_block: int = 1,

src/diffusers/models/unets/unet_2d_condition.py

@@ -177,16 +177,21 @@ class UNet2DConditionModel(
         center_input_sample: bool = False,
         flip_sin_to_cos: bool = True,
         freq_shift: int = 0,
-        down_block_types: Tuple[str] = (
+        down_block_types: Tuple[str, ...] = (
             "CrossAttnDownBlock2D",
             "CrossAttnDownBlock2D",
             "CrossAttnDownBlock2D",
             "DownBlock2D",
         ),
         mid_block_type: Optional[str] = "UNetMidBlock2DCrossAttn",
-        up_block_types: Tuple[str] = ("UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D"),
+        up_block_types: Tuple[str, ...] = (
+            "UpBlock2D",
+            "CrossAttnUpBlock2D",
+            "CrossAttnUpBlock2D",
+            "CrossAttnUpBlock2D",
+        ),
         only_cross_attention: Union[bool, Tuple[bool]] = False,
-        block_out_channels: Tuple[int] = (320, 640, 1280, 1280),
+        block_out_channels: Tuple[int, ...] = (320, 640, 1280, 1280),
         layers_per_block: Union[int, Tuple[int]] = 2,
         downsample_padding: int = 1,
         mid_block_scale_factor: float = 1,
@@ -486,10 +491,10 @@ class UNet2DConditionModel(
 
     def _check_config(
         self,
-        down_block_types: Tuple[str],
-        up_block_types: Tuple[str],
+        down_block_types: Tuple[str, ...],
+        up_block_types: Tuple[str, ...],
         only_cross_attention: Union[bool, Tuple[bool]],
-        block_out_channels: Tuple[int],
+        block_out_channels: Tuple[int, ...],
         layers_per_block: Union[int, Tuple[int]],
         cross_attention_dim: Union[int, Tuple[int]],
         transformer_layers_per_block: Union[int, Tuple[int], Tuple[Tuple[int]]],

src/diffusers/models/unets/unet_kandinsky3.py

@@ -54,7 +54,7 @@ class Kandinsky3UNet(ModelMixin, ConfigMixin):
         groups: int = 32,
         attention_head_dim: int = 64,
         layers_per_block: Union[int, Tuple[int]] = 3,
-        block_out_channels: Tuple[int] = (384, 768, 1536, 3072),
+        block_out_channels: Tuple[int, ...] = (384, 768, 1536, 3072),
         cross_attention_dim: Union[int, Tuple[int]] = 4096,
         encoder_hid_dim: int = 4096,
     ):

src/diffusers/models/unets/unet_spatio_temporal_condition.py

@@ -73,25 +73,25 @@ class UNetSpatioTemporalConditionModel(ModelMixin, ConfigMixin, UNet2DConditionL
         sample_size: Optional[int] = None,
         in_channels: int = 8,
         out_channels: int = 4,
-        down_block_types: Tuple[str] = (
+        down_block_types: Tuple[str, ...] = (
             "CrossAttnDownBlockSpatioTemporal",
             "CrossAttnDownBlockSpatioTemporal",
             "CrossAttnDownBlockSpatioTemporal",
             "DownBlockSpatioTemporal",
         ),
-        up_block_types: Tuple[str] = (
+        up_block_types: Tuple[str, ...] = (
             "UpBlockSpatioTemporal",
             "CrossAttnUpBlockSpatioTemporal",
             "CrossAttnUpBlockSpatioTemporal",
             "CrossAttnUpBlockSpatioTemporal",
         ),
-        block_out_channels: Tuple[int] = (320, 640, 1280, 1280),
+        block_out_channels: Tuple[int, ...] = (320, 640, 1280, 1280),
         addition_time_embed_dim: int = 256,
         projection_class_embeddings_input_dim: int = 768,
         layers_per_block: Union[int, Tuple[int]] = 2,
         cross_attention_dim: Union[int, Tuple[int]] = 1024,
         transformer_layers_per_block: Union[int, Tuple[int], Tuple[Tuple]] = 1,
-        num_attention_heads: Union[int, Tuple[int]] = (5, 10, 20, 20),
+        num_attention_heads: Union[int, Tuple[int, ...]] = (5, 10, 20, 20),
         num_frames: int = 25,
     ):
         super().__init__()

src/diffusers/models/unets/unet_stable_cascade.py

@@ -145,10 +145,10 @@ class StableCascadeUNet(ModelMixin, ConfigMixin, FromOriginalModelMixin):
         timestep_ratio_embedding_dim: int = 64,
         patch_size: int = 1,
         conditioning_dim: int = 2048,
-        block_out_channels: Tuple[int] = (2048, 2048),
-        num_attention_heads: Tuple[int] = (32, 32),
-        down_num_layers_per_block: Tuple[int] = (8, 24),
-        up_num_layers_per_block: Tuple[int] = (24, 8),
+        block_out_channels: Tuple[int, ...] = (2048, 2048),
+        num_attention_heads: Tuple[int, ...] = (32, 32),
+        down_num_layers_per_block: Tuple[int, ...] = (8, 24),
+        up_num_layers_per_block: Tuple[int, ...] = (24, 8),
         down_blocks_repeat_mappers: Optional[Tuple[int]] = (
             1,
             1,
@@ -167,7 +167,7 @@ class StableCascadeUNet(ModelMixin, ConfigMixin, FromOriginalModelMixin):
         kernel_size=3,
         dropout: Union[float, Tuple[float]] = (0.1, 0.1),
         self_attn: Union[bool, Tuple[bool]] = True,
-        timestep_conditioning_type: Tuple[str] = ("sca", "crp"),
+        timestep_conditioning_type: Tuple[str, ...] = ("sca", "crp"),
         switch_level: Optional[Tuple[bool]] = None,
     ):
         """

src/diffusers/models/vae_flax.py

@@ -532,8 +532,8 @@ class FlaxEncoder(nn.Module):
 
     in_channels: int = 3
     out_channels: int = 3
-    down_block_types: Tuple[str] = ("DownEncoderBlock2D",)
-    block_out_channels: Tuple[int] = (64,)
+    down_block_types: Tuple[str, ...] = ("DownEncoderBlock2D",)
+    block_out_channels: Tuple[int, ...] = (64,)
     layers_per_block: int = 2
     norm_num_groups: int = 32
     act_fn: str = "silu"
@@ -650,8 +650,8 @@ class FlaxDecoder(nn.Module):
 
     in_channels: int = 3
     out_channels: int = 3
-    up_block_types: Tuple[str] = ("UpDecoderBlock2D",)
-    block_out_channels: int = (64,)
+    up_block_types: Tuple[str, ...] = ("UpDecoderBlock2D",)
+    block_out_channels: Tuple[int, ...] = (64,)
     layers_per_block: int = 2
     norm_num_groups: int = 32
     act_fn: str = "silu"
@@ -823,9 +823,9 @@ class FlaxAutoencoderKL(nn.Module, FlaxModelMixin, ConfigMixin):
 
     in_channels: int = 3
     out_channels: int = 3
-    down_block_types: Tuple[str] = ("DownEncoderBlock2D",)
-    up_block_types: Tuple[str] = ("UpDecoderBlock2D",)
-    block_out_channels: Tuple[int] = (64,)
+    down_block_types: Tuple[str, ...] = ("DownEncoderBlock2D",)
+    up_block_types: Tuple[str, ...] = ("UpDecoderBlock2D",)
+    block_out_channels: Tuple[int, ...] = (64,)
     layers_per_block: int = 1
     act_fn: str = "silu"
     latent_channels: int = 4

src/diffusers/pipelines/audioldm2/modeling_audioldm2.py

@@ -245,16 +245,21 @@ class AudioLDM2UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoad
         out_channels: int = 4,
         flip_sin_to_cos: bool = True,
         freq_shift: int = 0,
-        down_block_types: Tuple[str] = (
+        down_block_types: Tuple[str, ...] = (
             "CrossAttnDownBlock2D",
             "CrossAttnDownBlock2D",
             "CrossAttnDownBlock2D",
             "DownBlock2D",
         ),
         mid_block_type: Optional[str] = "UNetMidBlock2DCrossAttn",
-        up_block_types: Tuple[str] = ("UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D"),
+        up_block_types: Tuple[str, ...] = (
+            "UpBlock2D",
+            "CrossAttnUpBlock2D",
+            "CrossAttnUpBlock2D",
+            "CrossAttnUpBlock2D",
+        ),
         only_cross_attention: Union[bool, Tuple[bool]] = False,
-        block_out_channels: Tuple[int] = (320, 640, 1280, 1280),
+        block_out_channels: Tuple[int, ...] = (320, 640, 1280, 1280),
         layers_per_block: Union[int, Tuple[int]] = 2,
         downsample_padding: int = 1,
         mid_block_scale_factor: float = 1,

src/diffusers/pipelines/deprecated/versatile_diffusion/modeling_text_unet.py

@@ -374,21 +374,21 @@ class UNetFlatConditionModel(ModelMixin, ConfigMixin):
         center_input_sample: bool = False,
         flip_sin_to_cos: bool = True,
         freq_shift: int = 0,
-        down_block_types: Tuple[str] = (
+        down_block_types: Tuple[str, ...] = (
             "CrossAttnDownBlockFlat",
             "CrossAttnDownBlockFlat",
             "CrossAttnDownBlockFlat",
             "DownBlockFlat",
         ),
         mid_block_type: Optional[str] = "UNetMidBlockFlatCrossAttn",
-        up_block_types: Tuple[str] = (
+        up_block_types: Tuple[str, ...] = (
             "UpBlockFlat",
             "CrossAttnUpBlockFlat",
             "CrossAttnUpBlockFlat",
             "CrossAttnUpBlockFlat",
         ),
         only_cross_attention: Union[bool, Tuple[bool]] = False,
-        block_out_channels: Tuple[int] = (320, 640, 1280, 1280),
+        block_out_channels: Tuple[int, ...] = (320, 640, 1280, 1280),
         layers_per_block: Union[int, Tuple[int]] = 2,
         downsample_padding: int = 1,
         mid_block_scale_factor: float = 1,

src/diffusers/pipelines/shap_e/renderer.py

@@ -742,7 +742,7 @@ class ShapEParamsProjModel(ModelMixin, ConfigMixin):
     def __init__(
         self,
         *,
-        param_names: Tuple[str] = (
+        param_names: Tuple[str, ...] = (
             "nerstf.mlp.0.weight",
             "nerstf.mlp.1.weight",
             "nerstf.mlp.2.weight",
@@ -786,13 +786,13 @@ class ShapERenderer(ModelMixin, ConfigMixin):
     def __init__(
         self,
         *,
-        param_names: Tuple[str] = (
+        param_names: Tuple[str, ...] = (
             "nerstf.mlp.0.weight",
             "nerstf.mlp.1.weight",
             "nerstf.mlp.2.weight",
             "nerstf.mlp.3.weight",
         ),
-        param_shapes: Tuple[Tuple[int]] = (
+        param_shapes: Tuple[Tuple[int, int], ...] = (
             (256, 93),
             (256, 256),
             (256, 256),
@@ -804,7 +804,7 @@ class ShapERenderer(ModelMixin, ConfigMixin):
         n_hidden_layers: int = 6,
         act_fn: str = "swish",
         insert_direction_at: int = 4,
-        background: Tuple[float] = (
+        background: Tuple[float, ...] = (
             255.0,
             255.0,
             255.0,