Fix annotations (#24582)

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

* fix annotations

src/transformers/models/autoformer/modeling_autoformer.py

@@ -735,7 +735,7 @@ class AutoformerEncoderLayer(nn.Module):
     ) -> Tuple[torch.FloatTensor, Optional[torch.FloatTensor]]:
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`torch.FloatTensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size

src/transformers/models/bart/modeling_tf_bart.py

@@ -321,7 +321,7 @@ class TFBartEncoderLayer(tf.keras.layers.Layer):
     ) -> tf.Tensor:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`tf.Tensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`tf.Tensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size
@@ -394,11 +394,11 @@ class TFBartDecoderLayer(tf.keras.layers.Layer):
     ) -> Tuple[tf.Tensor, tf.Tensor, Tuple[Tuple[tf.Tensor]]]:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`tf.Tensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`tf.Tensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             encoder_hidden_states (`tf.Tensor`):
-                cross attention input to the layer of shape `(seq_len, batch, embed_dim)`
+                cross attention input to the layer of shape `(batch, seq_len, embed_dim)`
             encoder_attention_mask (`tf.Tensor`): encoder attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size

src/transformers/models/bigbird_pegasus/modeling_bigbird_pegasus.py

@@ -1388,7 +1388,7 @@ class BigBirdPegasusEncoderLayer(nn.Module):
     ):
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`torch.FloatTensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             output_attentions (`bool`, *optional*):

src/transformers/models/blenderbot/modeling_tf_blenderbot.py

@@ -317,7 +317,7 @@ class TFBlenderbotEncoderLayer(tf.keras.layers.Layer):
     ):
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`tf.Tensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`tf.Tensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size
@@ -391,11 +391,11 @@ class TFBlenderbotDecoderLayer(tf.keras.layers.Layer):
     ) -> Tuple[tf.Tensor, tf.Tensor, Tuple[Tuple[tf.Tensor]]]:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`tf.Tensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`tf.Tensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             encoder_hidden_states (`tf.Tensor`):
-                cross attention input to the layer of shape *(seq_len, batch, embed_dim)*
+                cross attention input to the layer of shape *(batch, seq_len, embed_dim)*
             encoder_attention_mask (`tf.Tensor`): encoder attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size

src/transformers/models/blenderbot_small/modeling_tf_blenderbot_small.py

@@ -317,7 +317,7 @@ class TFBlenderbotSmallEncoderLayer(tf.keras.layers.Layer):
     ) -> tf.Tensor:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`tf.Tensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`tf.Tensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size
@@ -391,11 +391,11 @@ class TFBlenderbotSmallDecoderLayer(tf.keras.layers.Layer):
     ) -> Tuple[tf.Tensor, tf.Tensor, Tuple[Tuple[tf.Tensor]]]:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`tf.Tensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`tf.Tensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             encoder_hidden_states (`tf.Tensor`):
-                cross attention input to the layer of shape `(seq_len, batch, embed_dim)`
+                cross attention input to the layer of shape `(batch, seq_len, embed_dim)`
             encoder_attention_mask (`tf.Tensor`): encoder attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size

src/transformers/models/conditional_detr/modeling_conditional_detr.py

@@ -795,7 +795,7 @@ class ConditionalDetrEncoderLayer(nn.Module):
     ):
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`torch.FloatTensor`): attention mask of size
                 `(batch, 1, target_len, source_len)` where padding elements are indicated by very large negative
                 values.

src/transformers/models/deta/modeling_deta.py

@@ -851,7 +851,7 @@ class DetaDecoderLayer(nn.Module):
         """
         Args:
             hidden_states (`torch.FloatTensor`):
-                Input to the layer of shape `(seq_len, batch, embed_dim)`.
+                Input to the layer of shape `(batch, seq_len, embed_dim)`.
             position_embeddings (`torch.FloatTensor`, *optional*):
                 Position embeddings that are added to the queries and keys in the self-attention layer.
             reference_points (`torch.FloatTensor`, *optional*):
@@ -861,7 +861,7 @@ class DetaDecoderLayer(nn.Module):
             level_start_index (`torch.LongTensor`, *optional*):
                 Level start index.
             encoder_hidden_states (`torch.FloatTensor`):
-                cross attention input to the layer of shape `(seq_len, batch, embed_dim)`
+                cross attention input to the layer of shape `(batch, seq_len, embed_dim)`
             encoder_attention_mask (`torch.FloatTensor`): encoder attention mask of size
                 `(batch, 1, target_len, source_len)` where padding elements are indicated by very large negative
                 values.

src/transformers/models/detr/modeling_detr.py

@@ -642,7 +642,7 @@ class DetrEncoderLayer(nn.Module):
     ):
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`torch.FloatTensor`): attention mask of size
                 `(batch, 1, target_len, source_len)` where padding elements are indicated by very large negative
                 values.
@@ -723,7 +723,7 @@ class DetrDecoderLayer(nn.Module):
     ):
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`torch.FloatTensor`): attention mask of size
                 `(batch, 1, target_len, source_len)` where padding elements are indicated by very large negative
                 values.
@@ -734,7 +734,7 @@ class DetrDecoderLayer(nn.Module):
                 position embeddings that are added to the queries and keys
             in the self-attention layer.
             encoder_hidden_states (`torch.FloatTensor`):
-                cross attention input to the layer of shape `(seq_len, batch, embed_dim)`
+                cross attention input to the layer of shape `(batch, seq_len, embed_dim)`
             encoder_attention_mask (`torch.FloatTensor`): encoder attention mask of size
                 `(batch, 1, target_len, source_len)` where padding elements are indicated by very large negative
                 values.

src/transformers/models/informer/modeling_informer.py

@@ -738,7 +738,7 @@ class InformerEncoderLayer(nn.Module):
     ) -> Tuple[torch.FloatTensor, Optional[torch.FloatTensor]]:
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`torch.FloatTensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size

src/transformers/models/led/modeling_led.py

@@ -962,7 +962,7 @@ class LEDEncoderLayer(nn.Module):
     ):
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`torch.FloatTensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`torch.FloatTensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size
@@ -1040,11 +1040,11 @@ class LEDDecoderLayer(nn.Module):
     ):
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`torch.FloatTensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`torch.FloatTensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             encoder_hidden_states (`torch.FloatTensor`):
-                cross attention input to the layer of shape *(seq_len, batch, embed_dim)*
+                cross attention input to the layer of shape *(batch, seq_len, embed_dim)*
             encoder_attention_mask (`torch.FloatTensor`): encoder attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size

src/transformers/models/led/modeling_tf_led.py

@@ -1181,7 +1181,7 @@ class TFLEDEncoderLayer(tf.keras.layers.Layer):
     ):
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`tf.Tensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`tf.Tensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size
@@ -1256,11 +1256,11 @@ class TFLEDDecoderLayer(tf.keras.layers.Layer):
     ) -> Tuple[tf.Tensor, tf.Tensor, tf.Tensor, Tuple[Tuple[tf.Tensor]]]:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`tf.Tensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`tf.Tensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             encoder_hidden_states (`tf.Tensor`):
-                cross attention input to the layer of shape *(seq_len, batch, embed_dim)*
+                cross attention input to the layer of shape *(batch, seq_len, embed_dim)*
             encoder_attention_mask (`tf.Tensor`): encoder attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size

src/transformers/models/marian/modeling_tf_marian.py

@@ -354,7 +354,7 @@ class TFMarianEncoderLayer(tf.keras.layers.Layer):
     ) -> tf.Tensor:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`tf.Tensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`tf.Tensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size
@@ -428,11 +428,11 @@ class TFMarianDecoderLayer(tf.keras.layers.Layer):
     ) -> Tuple[tf.Tensor, tf.Tensor, Tuple[Tuple[tf.Tensor]]]:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`tf.Tensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`tf.Tensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             encoder_hidden_states (`tf.Tensor`):
-                cross attention input to the layer of shape `(seq_len, batch, embed_dim)`
+                cross attention input to the layer of shape `(batch, seq_len, embed_dim)`
             encoder_attention_mask (`tf.Tensor`): encoder attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size

src/transformers/models/maskformer/modeling_maskformer.py

@@ -571,7 +571,7 @@ class DetrDecoderLayer(nn.Module):
     ):
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`torch.FloatTensor`): attention mask of size
                 `(batch, 1, target_len, source_len)` where padding elements are indicated by very large negative
                 values.
@@ -582,7 +582,7 @@ class DetrDecoderLayer(nn.Module):
                 position embeddings that are added to the queries and keys
             in the self-attention layer.
             encoder_hidden_states (`torch.FloatTensor`):
-                cross attention input to the layer of shape `(seq_len, batch, embed_dim)`
+                cross attention input to the layer of shape `(batch, seq_len, embed_dim)`
             encoder_attention_mask (`torch.FloatTensor`): encoder attention mask of size
                 `(batch, 1, target_len, source_len)` where padding elements are indicated by very large negative
                 values.

src/transformers/models/mbart/modeling_tf_mbart.py

@@ -322,7 +322,7 @@ class TFMBartEncoderLayer(tf.keras.layers.Layer):
     ):
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`tf.Tensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`tf.Tensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size
@@ -395,11 +395,11 @@ class TFMBartDecoderLayer(tf.keras.layers.Layer):
     ) -> Tuple[tf.Tensor, tf.Tensor, Tuple[Tuple[tf.Tensor]]]:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`tf.Tensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`tf.Tensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             encoder_hidden_states (`tf.Tensor`):
-                cross attention input to the layer of shape *(seq_len, batch, embed_dim)*
+                cross attention input to the layer of shape *(batch, seq_len, embed_dim)*
             encoder_attention_mask (`tf.Tensor`): encoder attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size

src/transformers/models/mvp/modeling_mvp.py

@@ -327,7 +327,7 @@ class MvpEncoderLayer(nn.Module):
     ) -> Tuple[torch.FloatTensor, Optional[torch.FloatTensor]]:
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`torch.FloatTensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size

src/transformers/models/nllb_moe/modeling_nllb_moe.py

@@ -672,7 +672,7 @@ class NllbMoeEncoderLayer(nn.Module):
         """
         Args:
             hidden_states (`torch.FloatTensor`):
-                input to the layer of shape `(seq_len, batch, embed_dim)`
+                input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`torch.FloatTensor`):
                 attention mask of size `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very
                 large negative values.

src/transformers/models/opt/modeling_tf_opt.py

@@ -303,7 +303,7 @@ class TFOPTDecoderLayer(tf.keras.layers.Layer):
     ) -> Tuple[tf.Tensor, tf.Tensor, Tuple[Tuple[tf.Tensor]]]:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`tf.Tensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`tf.Tensor`, *optional*): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`, *optional*): mask for attention heads in a given layer of size

src/transformers/models/pegasus/modeling_tf_pegasus.py

@@ -356,7 +356,7 @@ class TFPegasusEncoderLayer(tf.keras.layers.Layer):
     ):
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`tf.Tensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`tf.Tensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size
@@ -430,11 +430,11 @@ class TFPegasusDecoderLayer(tf.keras.layers.Layer):
     ) -> Tuple[tf.Tensor, tf.Tensor, Tuple[Tuple[tf.Tensor]]]:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`tf.Tensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`tf.Tensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             encoder_hidden_states (`tf.Tensor`):
-                cross attention input to the layer of shape *(seq_len, batch, embed_dim)*
+                cross attention input to the layer of shape *(batch, seq_len, embed_dim)*
             encoder_attention_mask (`tf.Tensor`): encoder attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size

src/transformers/models/speech_to_text/modeling_tf_speech_to_text.py

@@ -400,7 +400,7 @@ class TFSpeech2TextEncoderLayer(tf.keras.layers.Layer):
     ):
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`tf.Tensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`tf.Tensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size
@@ -477,11 +477,11 @@ class TFSpeech2TextDecoderLayer(tf.keras.layers.Layer):
     ) -> Tuple[tf.Tensor, tf.Tensor, Tuple[Tuple[tf.Tensor]]]:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`tf.Tensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`tf.Tensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             encoder_hidden_states (`tf.Tensor`):
-                cross attention input to the layer of shape `(seq_len, batch, embed_dim)`
+                cross attention input to the layer of shape `(batch, seq_len, embed_dim)`
             encoder_attention_mask (`tf.Tensor`): encoder attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size

src/transformers/models/speech_to_text_2/modeling_speech_to_text_2.py

@@ -345,11 +345,11 @@ class Speech2Text2DecoderLayer(nn.Module):
     ):
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`torch.FloatTensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             encoder_hidden_states (`torch.FloatTensor`):
-                cross attention input to the layer of shape `(seq_len, batch, embed_dim)`
+                cross attention input to the layer of shape `(batch, seq_len, embed_dim)`
             encoder_attention_mask (`torch.FloatTensor`): encoder attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size

src/transformers/models/table_transformer/modeling_table_transformer.py

@@ -584,7 +584,7 @@ class TableTransformerEncoderLayer(nn.Module):
     ):
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`torch.FloatTensor`): attention mask of size
                 `(batch, 1, target_len, source_len)` where padding elements are indicated by very large negative
                 values.
@@ -668,7 +668,7 @@ class TableTransformerDecoderLayer(nn.Module):
     ):
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`torch.FloatTensor`): attention mask of size
                 `(batch, 1, target_len, source_len)` where padding elements are indicated by very large negative
                 values.
@@ -679,7 +679,7 @@ class TableTransformerDecoderLayer(nn.Module):
                 position embeddings that are added to the queries and keys
             in the self-attention layer.
             encoder_hidden_states (`torch.FloatTensor`):
-                cross attention input to the layer of shape `(seq_len, batch, embed_dim)`
+                cross attention input to the layer of shape `(batch, seq_len, embed_dim)`
             encoder_attention_mask (`torch.FloatTensor`): encoder attention mask of size
                 `(batch, 1, target_len, source_len)` where padding elements are indicated by very large negative
                 values.

src/transformers/models/trocr/modeling_trocr.py

@@ -358,11 +358,11 @@ class TrOCRDecoderLayer(nn.Module):
     ):
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`torch.FloatTensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             encoder_hidden_states (`torch.FloatTensor`):
-                cross attention input to the layer of shape `(seq_len, batch, embed_dim)`
+                cross attention input to the layer of shape `(batch, seq_len, embed_dim)`
             encoder_attention_mask (`torch.FloatTensor`): encoder attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size

src/transformers/models/whisper/modeling_tf_whisper.py

@@ -313,7 +313,7 @@ class TFWhisperEncoderLayer(tf.keras.layers.Layer):
     ):
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`tf.Tensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`tf.Tensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size
@@ -391,11 +391,11 @@ class TFWhisperDecoderLayer(tf.keras.layers.Layer):
     ) -> Tuple[tf.Tensor, tf.Tensor, Tuple[Tuple[tf.Tensor]]]:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
+            hidden_states (`tf.Tensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
             attention_mask (`tf.Tensor`): attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             encoder_hidden_states (`tf.Tensor`):
-                cross attention input to the layer of shape `(seq_len, batch, embed_dim)`
+                cross attention input to the layer of shape `(batch, seq_len, embed_dim)`
             encoder_attention_mask (`tf.Tensor`): encoder attention mask of size
                 `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size

src/transformers/models/xglm/modeling_tf_xglm.py

@@ -348,11 +348,11 @@ class TFXGLMDecoderLayer(tf.keras.layers.Layer):
     ) -> Tuple[tf.Tensor, tf.Tensor, Tuple[Tuple[tf.Tensor]]]:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`tf.Tensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`tf.Tensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             encoder_hidden_states (`tf.Tensor`):
-                cross attention input to the layer of shape *(seq_len, batch, embed_dim)*
+                cross attention input to the layer of shape *(batch, seq_len, embed_dim)*
             encoder_attention_mask (`tf.Tensor`): encoder attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/modeling_tf_{{cookiecutter.lowercase_modelname}}.py

@@ -1794,7 +1794,7 @@ class TF{{cookiecutter.camelcase_modelname}}EncoderLayer(tf.keras.layers.Layer):
     def call(self, hidden_states: tf.Tensor, attention_mask: tf.Tensor, layer_head_mask: tf.Tensor, training=False):
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`tf.Tensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`tf.Tensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size
@@ -1867,10 +1867,10 @@ class TF{{cookiecutter.camelcase_modelname}}DecoderLayer(tf.keras.layers.Layer):
     ) -> Tuple[tf.Tensor, tf.Tensor, Tuple[Tuple[tf.Tensor]]]:
         """
         Args:
-            hidden_states (`tf.Tensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`tf.Tensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`tf.Tensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
-            encoder_hidden_states (`tf.Tensor`): cross attention input to the layer of shape *(seq_len, batch, embed_dim)*
+            encoder_hidden_states (`tf.Tensor`): cross attention input to the layer of shape *(batch, seq_len, embed_dim)*
             encoder_attention_mask (`tf.Tensor`): encoder attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`tf.Tensor`): mask for attention heads in a given layer of size

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/modeling_{{cookiecutter.lowercase_modelname}}.py

@@ -1826,7 +1826,7 @@ class {{cookiecutter.camelcase_modelname}}EncoderLayer(nn.Module):
     ):
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`torch.FloatTensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`torch.FloatTensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size
@@ -1907,10 +1907,10 @@ class {{cookiecutter.camelcase_modelname}}DecoderLayer(nn.Module):
     ):
         """
         Args:
-            hidden_states (`torch.FloatTensor`): input to the layer of shape *(seq_len, batch, embed_dim)*
+            hidden_states (`torch.FloatTensor`): input to the layer of shape *(batch, seq_len, embed_dim)*
             attention_mask (`torch.FloatTensor`): attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
-            encoder_hidden_states (`torch.FloatTensor`): cross attention input to the layer of shape *(seq_len, batch, embed_dim)*
+            encoder_hidden_states (`torch.FloatTensor`): cross attention input to the layer of shape *(batch, seq_len, embed_dim)*
             encoder_attention_mask (`torch.FloatTensor`): encoder attention mask of size
                 *(batch, 1, tgt_len, src_len)* where padding elements are indicated by very large negative values.
             layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size