fix many type hint errors

Author: DefTruth

src/diffusers/models/transformers/auraflow_transformer_2d.py

@@ -472,7 +472,7 @@ def forward(
         timestep: torch.LongTensor = None,
         attention_kwargs: Optional[Dict[str, Any]] = None,
         return_dict: bool = True,
-    ) -> Union[torch.Tensor, Transformer2DModelOutput]:
+    ) -> Union[Tuple[torch.Tensor], Transformer2DModelOutput]:
         if attention_kwargs is not None:
             attention_kwargs = attention_kwargs.copy()
             lora_scale = attention_kwargs.pop("scale", 1.0)

src/diffusers/models/transformers/cogvideox_transformer_3d.py

@@ -441,7 +441,7 @@ def forward(
         image_rotary_emb: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
         attention_kwargs: Optional[Dict[str, Any]] = None,
         return_dict: bool = True,
-    ) -> Union[torch.Tensor, Transformer2DModelOutput]:
+    ) -> Union[Tuple[torch.Tensor], Transformer2DModelOutput]:
         if attention_kwargs is not None:
             attention_kwargs = attention_kwargs.copy()
             lora_scale = attention_kwargs.pop("scale", 1.0)

src/diffusers/models/transformers/consisid_transformer_3d.py

@@ -691,7 +691,7 @@ def forward(
         id_cond: Optional[torch.Tensor] = None,
         id_vit_hidden: Optional[torch.Tensor] = None,
         return_dict: bool = True,
-    ) -> Union[torch.Tensor, Transformer2DModelOutput]:
+    ) -> Union[Tuple[torch.Tensor], Transformer2DModelOutput]:
         if attention_kwargs is not None:
             attention_kwargs = attention_kwargs.copy()
             lora_scale = attention_kwargs.pop("scale", 1.0)

src/diffusers/models/transformers/lumina_nextdit2d.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import Any, Dict, Optional, Union
+from typing import Any, Dict, Optional, Union, Tuple
 
 import torch
 import torch.nn as nn
@@ -297,7 +297,7 @@ def forward(
         image_rotary_emb: torch.Tensor,
         cross_attention_kwargs: Dict[str, Any] = None,
         return_dict=True,
-    ) -> Union[torch.Tensor, Transformer2DModelOutput]:
+    ) -> Union[Tuple[torch.Tensor], Transformer2DModelOutput]:
         """
         Forward pass of LuminaNextDiT.
 

src/diffusers/models/transformers/transformer_bria.py

@@ -588,7 +588,7 @@ def forward(
         return_dict: bool = True,
         controlnet_block_samples=None,
         controlnet_single_block_samples=None,
-    ) -> Union[torch.FloatTensor, Transformer2DModelOutput]:
+    ) -> Union[Tuple[torch.Tensor], Transformer2DModelOutput]:
         """
         The [`BriaTransformer2DModel`] forward method.
 

src/diffusers/models/transformers/transformer_cogview3plus.py

@@ -293,7 +293,7 @@ def forward(
         target_size: torch.Tensor,
         crop_coords: torch.Tensor,
         return_dict: bool = True,
-    ) -> Union[torch.Tensor, Transformer2DModelOutput]:
+    ) -> Union[Tuple[torch.Tensor], Transformer2DModelOutput]:
         """
         The [`CogView3PlusTransformer2DModel`] forward method.
 

src/diffusers/models/transformers/transformer_cogview4.py

@@ -717,7 +717,7 @@ def forward(
         image_rotary_emb: Optional[
             Union[Tuple[torch.Tensor, torch.Tensor], List[Tuple[torch.Tensor, torch.Tensor]]]
         ] = None,
-    ) -> Union[torch.Tensor, Transformer2DModelOutput]:
+    ) -> Union[Tuple[torch.Tensor], Transformer2DModelOutput]:
         if attention_kwargs is not None:
             attention_kwargs = attention_kwargs.copy()
             lora_scale = attention_kwargs.pop("scale", 1.0)

src/diffusers/models/transformers/transformer_hidream_image.py

@@ -786,7 +786,7 @@ def forward(
         attention_kwargs: Optional[Dict[str, Any]] = None,
         return_dict: bool = True,
         **kwargs,
-    ) -> Union[torch.Tensor, Transformer2DModelOutput]:
+    ) -> Union[Tuple[torch.Tensor], Transformer2DModelOutput]:
         encoder_hidden_states = kwargs.get("encoder_hidden_states", None)
 
         if encoder_hidden_states is not None:

src/diffusers/models/transformers/transformer_hunyuan_video_framepack.py

@@ -216,7 +216,7 @@ def forward(
         indices_latents_history_4x: Optional[torch.Tensor] = None,
         attention_kwargs: Optional[Dict[str, Any]] = None,
         return_dict: bool = True,
-    ) -> Union[torch.Tensor, Transformer2DModelOutput]:
+    ) -> Union[Tuple[torch.Tensor], Transformer2DModelOutput]:
         if attention_kwargs is not None:
             attention_kwargs = attention_kwargs.copy()
             lora_scale = attention_kwargs.pop("scale", 1.0)