Update document of WhisperDecoderLayer (#21621)

* Update document of WhisperDecoderLayer

* Update modeling_mbart.py

* Update doc with utils/check_copies.py --fix_and_overwrite

* Update modeling_xlm_prophetnet.py

src/transformers/models/bigbird_pegasus/modeling_bigbird_pegasus.py

@@ -1483,11 +1483,11 @@ class BigBirdPegasusDecoderLayer(nn.Module):
     ) -> torch.Tensor:
         """
         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/blenderbot/modeling_blenderbot.py

@@ -390,11 +390,11 @@ class BlenderbotDecoderLayer(nn.Module):
     ) -> torch.Tensor:
         """
         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/m2m_100/modeling_m2m_100.py

@@ -458,11 +458,11 @@ class M2M100DecoderLayer(nn.Module):
     ) -> torch.Tensor:
         """
         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/mbart/modeling_mbart.py

@@ -397,11 +397,11 @@ class MBartDecoderLayer(nn.Module):
     ) -> torch.Tensor:
         """
         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/pegasus/modeling_pegasus.py

@@ -405,11 +405,11 @@ class PegasusDecoderLayer(nn.Module):
     ) -> torch.Tensor:
         """
         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/speech_to_text/modeling_speech_to_text.py

@@ -465,11 +465,11 @@ class Speech2TextDecoderLayer(nn.Module):
     ) -> torch.Tensor:
         """
         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_whisper.py

@@ -372,11 +372,11 @@ class WhisperDecoderLayer(nn.Module):
     ) -> torch.Tensor:
         """
         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/xglm/modeling_xglm.py

@@ -430,11 +430,11 @@ class XGLMDecoderLayer(nn.Module):
     ) -> torch.Tensor:
         """
         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