Fix many type hint errors

src/diffusers/models/attention.py

@@ -674,7 +674,7 @@ def forward(
         encoder_hidden_states: torch.FloatTensor,
         temb: torch.FloatTensor,
         joint_attention_kwargs: Optional[Dict[str, Any]] = None,
-    ):
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
         joint_attention_kwargs = joint_attention_kwargs or {}
         if self.use_dual_attention:
             norm_hidden_states, gate_msa, shift_mlp, scale_mlp, gate_mlp, norm_hidden_states2, gate_msa2 = self.norm1(

src/diffusers/models/transformers/auraflow_transformer_2d.py

@@ -173,7 +173,7 @@ def forward(
         hidden_states: torch.FloatTensor,
         temb: torch.FloatTensor,
         attention_kwargs: Optional[Dict[str, Any]] = None,
-    ):
+    ) -> torch.Tensor:
         residual = hidden_states
         attention_kwargs = attention_kwargs or {}
 

src/diffusers/models/transformers/cogvideox_transformer_3d.py

@@ -122,7 +122,7 @@ def forward(
         temb: torch.Tensor,
         image_rotary_emb: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
         attention_kwargs: Optional[Dict[str, Any]] = None,
-    ) -> torch.Tensor:
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
         text_seq_length = encoder_hidden_states.size(1)
         attention_kwargs = attention_kwargs or {}
 

src/diffusers/models/transformers/consisid_transformer_3d.py

@@ -315,7 +315,7 @@ def forward(
         encoder_hidden_states: torch.Tensor,
         temb: torch.Tensor,
         image_rotary_emb: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
-    ) -> torch.Tensor:
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
         text_seq_length = encoder_hidden_states.size(1)
 
         # norm & modulate

src/diffusers/models/transformers/lumina_nextdit2d.py

@@ -124,7 +124,7 @@ def forward(
         encoder_mask: torch.Tensor,
         temb: torch.Tensor,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
-    ):
+    ) -> torch.Tensor:
         """
         Perform a forward pass through the LuminaNextDiTBlock.
 

src/diffusers/models/transformers/transformer_bria.py

@@ -472,7 +472,7 @@ def forward(
         temb: torch.Tensor,
         image_rotary_emb: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
         attention_kwargs: Optional[Dict[str, Any]] = None,
-    ) -> torch.Tensor:
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
         text_seq_len = encoder_hidden_states.shape[1]
         hidden_states = torch.cat([encoder_hidden_states, hidden_states], dim=1)
 

src/diffusers/models/transformers/transformer_cogview3plus.py

@@ -13,7 +13,7 @@
 # limitations under the License.
 
 
-from typing import Dict, Union
+from typing import Dict, Union, Tuple
 
 import torch
 import torch.nn as nn
@@ -79,7 +79,7 @@ def forward(
         hidden_states: torch.Tensor,
         encoder_hidden_states: torch.Tensor,
         emb: torch.Tensor,
-    ) -> torch.Tensor:
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
         text_seq_length = encoder_hidden_states.size(1)
 
         # norm & modulate

src/diffusers/models/transformers/transformer_cogview4.py

@@ -494,7 +494,7 @@ def forward(
         ] = None,
         attention_mask: Optional[Dict[str, torch.Tensor]] = None,
         attention_kwargs: Optional[Dict[str, Any]] = None,
-    ) -> torch.Tensor:
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
         # 1. Timestep conditioning
         (
             norm_hidden_states,

src/diffusers/models/transformers/transformer_hidream_image.py

@@ -534,7 +534,7 @@ def forward(
         encoder_hidden_states: Optional[torch.Tensor] = None,
         temb: Optional[torch.Tensor] = None,
         image_rotary_emb: torch.Tensor = None,
-    ) -> torch.Tensor:
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
         wtype = hidden_states.dtype
         (
             shift_msa_i,
@@ -592,7 +592,7 @@ def forward(
         encoder_hidden_states: Optional[torch.Tensor] = None,
         temb: Optional[torch.Tensor] = None,
         image_rotary_emb: torch.Tensor = None,
-    ) -> torch.Tensor:
+    ) -> Union[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]]:
         return self.block(
             hidden_states=hidden_states,
             hidden_states_masks=hidden_states_masks,

src/diffusers/models/transformers/transformer_hunyuan_video.py

@@ -684,7 +684,7 @@ def forward(
         image_rotary_emb: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
         token_replace_emb: torch.Tensor = None,
         num_tokens: int = None,
-    ) -> torch.Tensor:
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
         text_seq_length = encoder_hidden_states.shape[1]
         hidden_states = torch.cat([hidden_states, encoder_hidden_states], dim=1)