Remove more unused attributes in config classes (#21327)

* remove unused classifier_dropout

* remove unused dropout

* remove unused pooler_fn

* remove unnecessary is_encoder_decoder

* remove unnecessary drop_rate

* remove unused classifier_dropout

* remove unused classifier_dropout

* remove unused dropout

* remove unused dropout

* remove unused summary_* attributes

* remove unused tie_word_embeddings

* remove unused summary_* attributes

* fix

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

src/transformers/models/altclip/configuration_altclip.py

@@ -77,12 +77,8 @@ class AltCLIPTextConfig(PretrainedConfig):
         use_cache (`bool`, *optional*, defaults to `True`):
             Whether or not the model should return the last key/values attentions (not used by all models). Only
             relevant if `config.is_decoder=True`.
-        classifier_dropout (`float`, *optional*):
-            The dropout ratio for the classification head.
         project_dim (`int`, *optional*, defaults to 768):
             The dimentions of the teacher model before the mapping layer.
-        pooler_fn (`str`, *optional*, defaults to `"cls"`):
-            Type of pooler we use. We take the first token as pooled output.
 
     Examples:
 
@@ -120,9 +116,7 @@ class AltCLIPTextConfig(PretrainedConfig):
         eos_token_id=2,
         position_embedding_type="absolute",
         use_cache=True,
-        classifier_dropout=None,
         project_dim=768,
-        pooler_fn="cls",
         **kwargs
     ):
         super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)
@@ -142,9 +136,7 @@ class AltCLIPTextConfig(PretrainedConfig):
         self.layer_norm_eps = layer_norm_eps
         self.position_embedding_type = position_embedding_type
         self.use_cache = use_cache
-        self.classifier_dropout = classifier_dropout
         self.project_dim = project_dim
-        self.pooler_fn = pooler_fn
 
 
 class AltCLIPVisionConfig(PretrainedConfig):
@@ -176,8 +168,6 @@ class AltCLIPVisionConfig(PretrainedConfig):
             `"relu"`, `"selu"` and `"gelu_new"` ``"quick_gelu"` are supported.
         layer_norm_eps (`float`, *optional*, defaults to 1e-5):
             The epsilon used by the layer normalization layers.
-        dropout (`float`, *optional*, defaults to 0.0):
-            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
         attention_dropout (`float`, *optional*, defaults to 0.0):
             The dropout ratio for the attention probabilities.
         initializer_range (`float`, *optional*, defaults to 0.02):
@@ -215,7 +205,6 @@ class AltCLIPVisionConfig(PretrainedConfig):
         patch_size=32,
         hidden_act="quick_gelu",
         layer_norm_eps=1e-5,
-        dropout=0.0,
         attention_dropout=0.0,
         initializer_range=0.02,
         initializer_factor=1.0,
@@ -226,7 +215,6 @@ class AltCLIPVisionConfig(PretrainedConfig):
         self.hidden_size = hidden_size
         self.intermediate_size = intermediate_size
         self.projection_dim = projection_dim
-        self.dropout = dropout
         self.num_hidden_layers = num_hidden_layers
         self.num_attention_heads = num_attention_heads
         self.num_channels = num_channels

src/transformers/models/biogpt/configuration_biogpt.py

@@ -68,8 +68,6 @@ class BioGptConfig(PretrainedConfig):
         use_cache (`bool`, *optional*, defaults to `True`):
             Whether or not the model should return the last key/values attentions (not used by all models). Only
             relevant if `config.is_decoder=True`.
-        is_encoder_decoder (`bool`, *optional*, defaults to `False`):
-            Whether this is an encoder/decoder model.
         layerdrop (`float`, *optional*, defaults to 0.0):
             Please refer to the paper about LayerDrop: https://arxiv.org/abs/1909.11556 for further details
         activation_dropout (`float`, *optional*, defaults to 0.0):
@@ -111,7 +109,6 @@ class BioGptConfig(PretrainedConfig):
         layer_norm_eps=1e-12,
         scale_embedding=True,
         use_cache=True,
-        is_encoder_decoder=False,
         layerdrop=0.0,
         activation_dropout=0.0,
         pad_token_id=1,
@@ -132,7 +129,6 @@ class BioGptConfig(PretrainedConfig):
         self.layer_norm_eps = layer_norm_eps
         self.scale_embedding = scale_embedding
         self.use_cache = use_cache
-        self.is_encoder_decoder = is_encoder_decoder
         self.layerdrop = layerdrop
         self.activation_dropout = activation_dropout
         super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)

src/transformers/models/blip/configuration_blip.py

@@ -77,10 +77,10 @@ class BlipTextConfig(PretrainedConfig):
             `"relu"`, `"selu"` and `"gelu_new"` ``"gelu"` are supported.
         layer_norm_eps (`float`, *optional*, defaults to 1e-12):
             The epsilon used by the layer normalization layers.
+        hidden_dropout_prob (`float`, *optional*, defaults to 0.0):
+            The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
         attention_dropout (`float`, *optional*, defaults to 0.0):
             The dropout ratio for the attention probabilities.
-        dropout (`float`, *optional*, defaults to 0.0):
-            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
         initializer_range (`float`, *optional*, defaults to 0.02):
             The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
         initializer_factor (`float``, *optional*, defaults to 1):
@@ -211,8 +211,6 @@ class BlipVisionConfig(PretrainedConfig):
             `"relu"`, `"selu"` and `"gelu_new"` ``"gelu"` are supported.
         layer_norm_eps (`float`, *optional*, defaults to 1e-5):
             The epsilon used by the layer normalization layers.
-        dropout (`float`, *optional*, defaults to 0.0):
-            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
         attention_dropout (`float`, *optional*, defaults to 0.0):
             The dropout ratio for the attention probabilities.
         initializer_range (`float`, *optional*, defaults to 0.02):
@@ -250,7 +248,6 @@ class BlipVisionConfig(PretrainedConfig):
         patch_size=16,
         hidden_act="gelu",
         layer_norm_eps=1e-5,
-        dropout=0.0,
         attention_dropout=0.0,
         initializer_range=1e-10,
         initializer_factor=1.0,
@@ -261,7 +258,6 @@ class BlipVisionConfig(PretrainedConfig):
         self.hidden_size = hidden_size
         self.intermediate_size = intermediate_size
         self.projection_dim = projection_dim
-        self.dropout = dropout
         self.num_hidden_layers = num_hidden_layers
         self.num_attention_heads = num_attention_heads
         self.num_channels = num_channels

src/transformers/models/bridgetower/configuration_bridgetower.py

@@ -260,8 +260,6 @@ class BridgeTowerConfig(PretrainedConfig):
     Args:
         share_cross_modal_transformer_layers (`bool`, *optional*, defaults to `True`):
             Whether cross modal transformer layers are shared.
-        drop_rate (`float`, *optional*, defaults to 0.1):
-            Drop out probability.
         head_hidden_scale (`int`, *optional*, defaults to 2):
             Scale of hidden layers head.
         hidden_act (`str` or `function`, *optional*, defaults to `"gelu"`):
@@ -271,8 +269,6 @@ class BridgeTowerConfig(PretrainedConfig):
         initializer_factor (`float``, *optional*, defaults to 1):
             A factor for initializing all weight matrices (should be kept to 1, used internally for initialization
             testing).
-        is_encoder_decoder (`bool`, *optional*, defaults to `False`):
-            Whether this is an encoder/decoder model
         layer_norm_eps (`float`, *optional*, defaults to 1e-05):
             The epsilon used by the layer normalization layers.
         share_link_tower_layers (`bool`, *optional*, defaults to `False`):
@@ -311,12 +307,10 @@ class BridgeTowerConfig(PretrainedConfig):
     def __init__(
         self,
         share_cross_modal_transformer_layers=True,
-        drop_rate=0.1,
         head_hidden_scale=2,
         hidden_act="gelu",
         hidden_size=768,
         initializer_factor=1,
-        is_encoder_decoder=False,
         layer_norm_eps=1e-05,
         share_link_tower_layers=False,
         link_tower_type="add",
@@ -330,12 +324,10 @@ class BridgeTowerConfig(PretrainedConfig):
     ):
         super().__init__(**kwargs)
         self.share_cross_modal_transformer_layers = share_cross_modal_transformer_layers
-        self.drop_rate = drop_rate
         self.head_hidden_scale = head_hidden_scale
         self.hidden_act = hidden_act
         self.hidden_size = hidden_size
         self.initializer_factor = initializer_factor
-        self.is_encoder_decoder = is_encoder_decoder
         self.layer_norm_eps = layer_norm_eps
         self.share_link_tower_layers = share_link_tower_layers
         self.link_tower_type = link_tower_type

src/transformers/models/chinese_clip/configuration_chinese_clip.py

@@ -190,8 +190,6 @@ class ChineseCLIPVisionConfig(PretrainedConfig):
             `"relu"`, `"selu"` and `"gelu_new"` ``"quick_gelu"` are supported.
         layer_norm_eps (`float`, *optional*, defaults to 1e-5):
             The epsilon used by the layer normalization layers.
-        dropout (`float`, *optional*, defaults to 0.0):
-            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
         attention_dropout (`float`, *optional*, defaults to 0.0):
             The dropout ratio for the attention probabilities.
         initializer_range (`float`, *optional*, defaults to 0.02):
@@ -227,7 +225,6 @@ class ChineseCLIPVisionConfig(PretrainedConfig):
         patch_size=32,
         hidden_act="quick_gelu",
         layer_norm_eps=1e-5,
-        dropout=0.0,
         attention_dropout=0.0,
         initializer_range=0.02,
         initializer_factor=1.0,
@@ -238,7 +235,6 @@ class ChineseCLIPVisionConfig(PretrainedConfig):
         self.hidden_size = hidden_size
         self.intermediate_size = intermediate_size
         self.projection_dim = projection_dim
-        self.dropout = dropout
         self.num_hidden_layers = num_hidden_layers
         self.num_attention_heads = num_attention_heads
         self.num_channels = num_channels

src/transformers/models/clip/configuration_clip.py

@@ -69,8 +69,6 @@ class CLIPTextConfig(PretrainedConfig):
             The epsilon used by the layer normalization layers.
         attention_dropout (`float`, *optional*, defaults to 0.0):
             The dropout ratio for the attention probabilities.
-        dropout (`float`, *optional*, defaults to 0.0):
-            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
         initializer_range (`float`, *optional*, defaults to 0.02):
             The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
         initializer_factor (`float`, *optional*, defaults to 1):
@@ -104,7 +102,6 @@ class CLIPTextConfig(PretrainedConfig):
         max_position_embeddings=77,
         hidden_act="quick_gelu",
         layer_norm_eps=1e-5,
-        dropout=0.0,
         attention_dropout=0.0,
         initializer_range=0.02,
         initializer_factor=1.0,
@@ -119,7 +116,6 @@ class CLIPTextConfig(PretrainedConfig):
         self.hidden_size = hidden_size
         self.intermediate_size = intermediate_size
         self.projection_dim = projection_dim
-        self.dropout = dropout
         self.num_hidden_layers = num_hidden_layers
         self.num_attention_heads = num_attention_heads
         self.max_position_embeddings = max_position_embeddings
@@ -175,8 +171,6 @@ class CLIPVisionConfig(PretrainedConfig):
             `"relu"`, `"selu"` and `"gelu_new"` ``"quick_gelu"` are supported.
         layer_norm_eps (`float`, *optional*, defaults to 1e-5):
             The epsilon used by the layer normalization layers.
-        dropout (`float`, *optional*, defaults to 0.0):
-            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
         attention_dropout (`float`, *optional*, defaults to 0.0):
             The dropout ratio for the attention probabilities.
         initializer_range (`float`, *optional*, defaults to 0.02):
@@ -214,7 +208,6 @@ class CLIPVisionConfig(PretrainedConfig):
         patch_size=32,
         hidden_act="quick_gelu",
         layer_norm_eps=1e-5,
-        dropout=0.0,
         attention_dropout=0.0,
         initializer_range=0.02,
         initializer_factor=1.0,
@@ -225,7 +218,6 @@ class CLIPVisionConfig(PretrainedConfig):
         self.hidden_size = hidden_size
         self.intermediate_size = intermediate_size
         self.projection_dim = projection_dim
-        self.dropout = dropout
         self.num_hidden_layers = num_hidden_layers
         self.num_attention_heads = num_attention_heads
         self.num_channels = num_channels

src/transformers/models/clipseg/configuration_clipseg.py

@@ -61,8 +61,6 @@ class CLIPSegTextConfig(PretrainedConfig):
             The epsilon used by the layer normalization layers.
         attention_dropout (`float`, *optional*, defaults to 0.0):
             The dropout ratio for the attention probabilities.
-        dropout (`float`, *optional*, defaults to 0.0):
-            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
         initializer_range (`float`, *optional*, defaults to 0.02):
             The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
         initializer_factor (`float``, *optional*, defaults to 1):
@@ -95,7 +93,6 @@ class CLIPSegTextConfig(PretrainedConfig):
         max_position_embeddings=77,
         hidden_act="quick_gelu",
         layer_norm_eps=1e-5,
-        dropout=0.0,
         attention_dropout=0.0,
         initializer_range=0.02,
         initializer_factor=1.0,
@@ -109,7 +106,6 @@ class CLIPSegTextConfig(PretrainedConfig):
         self.vocab_size = vocab_size
         self.hidden_size = hidden_size
         self.intermediate_size = intermediate_size
-        self.dropout = dropout
         self.num_hidden_layers = num_hidden_layers
         self.num_attention_heads = num_attention_heads
         self.max_position_embeddings = max_position_embeddings
@@ -165,8 +161,6 @@ class CLIPSegVisionConfig(PretrainedConfig):
             `"relu"`, `"selu"` and `"gelu_new"` ``"quick_gelu"` are supported.
         layer_norm_eps (`float`, *optional*, defaults to 1e-5):
             The epsilon used by the layer normalization layers.
-        dropout (`float`, *optional*, defaults to 0.0):
-            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
         attention_dropout (`float`, *optional*, defaults to 0.0):
             The dropout ratio for the attention probabilities.
         initializer_range (`float`, *optional*, defaults to 0.02):
@@ -203,7 +197,6 @@ class CLIPSegVisionConfig(PretrainedConfig):
         patch_size=32,
         hidden_act="quick_gelu",
         layer_norm_eps=1e-5,
-        dropout=0.0,
         attention_dropout=0.0,
         initializer_range=0.02,
         initializer_factor=1.0,
@@ -213,7 +206,6 @@ class CLIPSegVisionConfig(PretrainedConfig):
 
         self.hidden_size = hidden_size
         self.intermediate_size = intermediate_size
-        self.dropout = dropout
         self.num_hidden_layers = num_hidden_layers
         self.num_attention_heads = num_attention_heads
         self.num_channels = num_channels

src/transformers/models/conditional_detr/configuration_conditional_detr.py

@@ -164,7 +164,6 @@ class ConditionalDetrConfig(PretrainedConfig):
         activation_dropout=0.0,
         init_std=0.02,
         init_xavier_std=1.0,
-        classifier_dropout=0.0,
         scale_embedding=False,
         auxiliary_loss=False,
         position_embedding_type="sine",

src/transformers/models/ctrl/configuration_ctrl.py

@@ -96,11 +96,6 @@ class CTRLConfig(PretrainedConfig):
         embd_pdrop=0.1,
         layer_norm_epsilon=1e-6,
         initializer_range=0.02,
-        summary_type="cls_index",
-        summary_use_proj=True,
-        summary_activation=None,
-        summary_proj_to_labels=True,
-        summary_first_dropout=0.1,
         use_cache=True,
         **kwargs
     ):
@@ -115,11 +110,6 @@ class CTRLConfig(PretrainedConfig):
         self.layer_norm_epsilon = layer_norm_epsilon
         self.initializer_range = initializer_range
 
-        self.summary_type = summary_type
-        self.summary_use_proj = summary_use_proj
-        self.summary_activation = summary_activation
-        self.summary_first_dropout = summary_first_dropout
-        self.summary_proj_to_labels = summary_proj_to_labels
         self.use_cache = use_cache
 
         super().__init__(**kwargs)

src/transformers/models/decision_transformer/configuration_decision_transformer.py

@@ -129,11 +129,6 @@ class DecisionTransformerConfig(PretrainedConfig):
         attn_pdrop=0.1,
         layer_norm_epsilon=1e-5,
         initializer_range=0.02,
-        summary_type="cls_index",
-        summary_use_proj=True,
-        summary_activation=None,
-        summary_proj_to_labels=True,
-        summary_first_dropout=0.1,
         scale_attn_weights=True,
         use_cache=True,
         bos_token_id=50256,
@@ -160,11 +155,6 @@ class DecisionTransformerConfig(PretrainedConfig):
         self.attn_pdrop = attn_pdrop
         self.layer_norm_epsilon = layer_norm_epsilon
         self.initializer_range = initializer_range
-        self.summary_type = summary_type
-        self.summary_use_proj = summary_use_proj
-        self.summary_activation = summary_activation
-        self.summary_first_dropout = summary_first_dropout
-        self.summary_proj_to_labels = summary_proj_to_labels
         self.scale_attn_weights = scale_attn_weights
         self.use_cache = use_cache
         self.scale_attn_by_inverse_layer_idx = scale_attn_by_inverse_layer_idx

src/transformers/models/detr/configuration_detr.py

@@ -161,7 +161,6 @@ class DetrConfig(PretrainedConfig):
         activation_dropout=0.0,
         init_std=0.02,
         init_xavier_std=1.0,
-        classifier_dropout=0.0,
         scale_embedding=False,
         auxiliary_loss=False,
         position_embedding_type="sine",

src/transformers/models/esm/configuration_esm.py

@@ -81,8 +81,6 @@ class EsmConfig(PretrainedConfig):
         use_cache (`bool`, *optional*, defaults to `True`):
             Whether or not the model should return the last key/values attentions (not used by all models). Only
             relevant if `config.is_decoder=True`.
-        classifier_dropout (`float`, *optional*):
-            The dropout ratio for the classification head.
         emb_layer_norm_before (`bool`, *optional*):
             Whether to apply layer normalization after embeddings but before the main stem of the network.
         token_dropout (`bool`, defaults to `False`):
@@ -117,7 +115,6 @@ class EsmConfig(PretrainedConfig):
         layer_norm_eps=1e-12,
         position_embedding_type="absolute",
         use_cache=True,
-        classifier_dropout=None,
         emb_layer_norm_before=None,
         token_dropout=False,
         is_folding_model=False,
@@ -139,7 +136,6 @@ class EsmConfig(PretrainedConfig):
         self.layer_norm_eps = layer_norm_eps
         self.position_embedding_type = position_embedding_type
         self.use_cache = use_cache
-        self.classifier_dropout = classifier_dropout
         self.emb_layer_norm_before = emb_layer_norm_before
         self.token_dropout = token_dropout
         self.is_folding_model = is_folding_model

src/transformers/models/git/configuration_git.py

@@ -57,8 +57,6 @@ class GitVisionConfig(PretrainedConfig):
             `"relu"`, `"selu"` and `"gelu_new"` ``"quick_gelu"` are supported.
         layer_norm_eps (`float`, *optional*, defaults to 1e-5):
             The epsilon used by the layer normalization layers.
-        dropout (`float`, *optional*, defaults to 0.0):
-            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
         attention_dropout (`float`, *optional*, defaults to 0.0):
             The dropout ratio for the attention probabilities.
         initializer_range (`float`, *optional*, defaults to 0.02):
@@ -96,7 +94,6 @@ class GitVisionConfig(PretrainedConfig):
         patch_size=16,
         hidden_act="quick_gelu",
         layer_norm_eps=1e-5,
-        dropout=0.0,
         attention_dropout=0.0,
         initializer_range=0.02,
         initializer_factor=1.0,
@@ -107,7 +104,6 @@ class GitVisionConfig(PretrainedConfig):
         self.hidden_size = hidden_size
         self.intermediate_size = intermediate_size
         self.projection_dim = projection_dim
-        self.dropout = dropout
         self.num_hidden_layers = num_hidden_layers
         self.num_attention_heads = num_attention_heads
         self.num_channels = num_channels
@@ -183,8 +179,6 @@ class GitConfig(PretrainedConfig):
             with Better Relative Position Embeddings (Huang et al.)](https://arxiv.org/abs/2009.13658).
         use_cache (`bool`, *optional*, defaults to `True`):
             Whether or not the model should return the last key/values attentions (not used by all models).
-        classifier_dropout (`float`, *optional*):
-            The dropout ratio for the classification head.
         num_image_with_embedding (`int`, *optional*):
             The number of temporal embeddings to add, in case the model is used for video captioning/VQA.
 
@@ -221,7 +215,6 @@ class GitConfig(PretrainedConfig):
         pad_token_id=0,
         position_embedding_type="absolute",
         use_cache=True,
-        classifier_dropout=None,
         tie_word_embeddings=False,
         bos_token_id=101,
         eos_token_id=102,
@@ -248,7 +241,6 @@ class GitConfig(PretrainedConfig):
         self.layer_norm_eps = layer_norm_eps
         self.position_embedding_type = position_embedding_type
         self.use_cache = use_cache
-        self.classifier_dropout = classifier_dropout
         self.tie_word_embeddings = tie_word_embeddings
         self.num_image_with_embedding = num_image_with_embedding
 

src/transformers/models/gpt_neo/configuration_gpt_neo.py

@@ -113,11 +113,6 @@ class GPTNeoConfig(PretrainedConfig):
         attention_dropout=0.0,
         layer_norm_epsilon=1e-5,
         initializer_range=0.02,
-        summary_type="cls_index",
-        summary_use_proj=True,
-        summary_activation=None,
-        summary_proj_to_labels=True,
-        summary_first_dropout=0.1,
         use_cache=True,
         bos_token_id=50256,
         eos_token_id=50256,
@@ -136,11 +131,6 @@ class GPTNeoConfig(PretrainedConfig):
         self.attention_dropout = attention_dropout
         self.layer_norm_epsilon = layer_norm_epsilon
         self.initializer_range = initializer_range
-        self.summary_type = summary_type
-        self.summary_use_proj = summary_use_proj
-        self.summary_activation = summary_activation
-        self.summary_first_dropout = summary_first_dropout
-        self.summary_proj_to_labels = summary_proj_to_labels
         self.use_cache = use_cache
 
         self.bos_token_id = bos_token_id

src/transformers/models/layoutlm/configuration_layoutlm.py

@@ -111,7 +111,6 @@ class LayoutLMConfig(PretrainedConfig):
         pad_token_id=0,
         position_embedding_type="absolute",
         use_cache=True,
-        classifier_dropout=None,
         max_2d_position_embeddings=1024,
         **kwargs
     ):
@@ -130,7 +129,6 @@ class LayoutLMConfig(PretrainedConfig):
         self.layer_norm_eps = layer_norm_eps
         self.position_embedding_type = position_embedding_type
         self.use_cache = use_cache
-        self.classifier_dropout = classifier_dropout
         self.max_2d_position_embeddings = max_2d_position_embeddings
 
 

src/transformers/models/longformer/configuration_longformer.py

@@ -92,8 +92,6 @@ class LongformerConfig(PretrainedConfig):
             [Self-Attention with Relative Position Representations (Shaw et al.)](https://arxiv.org/abs/1803.02155).
             For more information on `"relative_key_query"`, please refer to *Method 4* in [Improve Transformer Models
             with Better Relative Position Embeddings (Huang et al.)](https://arxiv.org/abs/2009.13658).
-        classifier_dropout (`float`, *optional*):
-            The dropout ratio for the classification head.
         attention_window (`int` or `List[int]`, *optional*, defaults to 512):
             Size of an attention window around each token. If an `int`, use the same size for all layers. To specify a
             different window size for each layer, use a `List[int]` where `len(attention_window) == num_hidden_layers`.
@@ -134,7 +132,6 @@ class LongformerConfig(PretrainedConfig):
         initializer_range: float = 0.02,
         layer_norm_eps: float = 1e-12,
         position_embedding_type: str = "absolute",
-        classifier_dropout: float = None,
         onnx_export: bool = False,
         **kwargs
     ):
@@ -158,7 +155,6 @@ class LongformerConfig(PretrainedConfig):
         self.initializer_range = initializer_range
         self.layer_norm_eps = layer_norm_eps
         self.position_embedding_type = position_embedding_type
-        self.classifier_dropout = classifier_dropout
         self.onnx_export = onnx_export
 
 

src/transformers/models/owlvit/configuration_owlvit.py

@@ -71,8 +71,6 @@ class OwlViTTextConfig(PretrainedConfig):
             The epsilon used by the layer normalization layers.
         attention_dropout (`float`, *optional*, defaults to 0.0):
             The dropout ratio for the attention probabilities.
-        dropout (`float`, *optional*, defaults to 0.0):
-            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
         initializer_range (`float`, *optional*, defaults to 0.02):
             The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
         initializer_factor (`float`, *optional*, defaults to 1):
@@ -105,7 +103,6 @@ class OwlViTTextConfig(PretrainedConfig):
         max_position_embeddings=16,
         hidden_act="quick_gelu",
         layer_norm_eps=1e-5,
-        dropout=0.0,
         attention_dropout=0.0,
         initializer_range=0.02,
         initializer_factor=1.0,
@@ -124,7 +121,6 @@ class OwlViTTextConfig(PretrainedConfig):
         self.max_position_embeddings = max_position_embeddings
         self.hidden_act = hidden_act
         self.layer_norm_eps = layer_norm_eps
-        self.dropout = dropout
         self.attention_dropout = attention_dropout
         self.initializer_range = initializer_range
         self.initializer_factor = initializer_factor
@@ -177,8 +173,6 @@ class OwlViTVisionConfig(PretrainedConfig):
             `"relu"`, `"selu"` and `"gelu_new"` ``"quick_gelu"` are supported.
         layer_norm_eps (`float`, *optional*, defaults to 1e-5):
             The epsilon used by the layer normalization layers.
-        dropout (`float`, *optional*, defaults to 0.0):
-            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
         attention_dropout (`float`, *optional*, defaults to 0.0):
             The dropout ratio for the attention probabilities.
         initializer_range (`float`, *optional*, defaults to 0.02):
@@ -215,7 +209,6 @@ class OwlViTVisionConfig(PretrainedConfig):
         patch_size=32,
         hidden_act="quick_gelu",
         layer_norm_eps=1e-5,
-        dropout=0.0,
         attention_dropout=0.0,
         initializer_range=0.02,
         initializer_factor=1.0,
@@ -232,7 +225,6 @@ class OwlViTVisionConfig(PretrainedConfig):
         self.patch_size = patch_size
         self.hidden_act = hidden_act
         self.layer_norm_eps = layer_norm_eps
-        self.dropout = dropout
         self.attention_dropout = attention_dropout
         self.initializer_range = initializer_range
         self.initializer_factor = initializer_factor

src/transformers/models/pegasus_x/configuration_pegasus_x.py

@@ -65,8 +65,6 @@ class PegasusXConfig(PretrainedConfig):
             The dropout ratio for the attention probabilities.
         activation_dropout (`float`, *optional*, defaults to 0.0):
             The dropout ratio for activations inside the fully connected layer.
-        classifier_dropout (`float`, *optional*, defaults to 0.0):
-            The dropout ratio for classifier.
         max_position_embeddings (`int`, *optional*, defaults to 16384):
             The maximum sequence length that this model might ever be used with. Typically set this to something large
             just in case (e.g., 512 or 1024 or 2048).
@@ -130,7 +128,6 @@ class PegasusXConfig(PretrainedConfig):
         activation_dropout=0.0,
         init_std=0.02,
         decoder_start_token_id=0,
-        classifier_dropout=0.0,
         scale_embedding=True,
         pad_token_id=0,
         eos_token_id=1,
@@ -156,7 +153,6 @@ class PegasusXConfig(PretrainedConfig):
         self.init_std = init_std
         self.encoder_layerdrop = encoder_layerdrop
         self.decoder_layerdrop = decoder_layerdrop
-        self.classifier_dropout = classifier_dropout
         self.use_cache = use_cache
         self.num_hidden_layers = encoder_layers
         self.scale_embedding = scale_embedding  # scale factor will be sqrt(d_model) if True

src/transformers/models/speech_to_text/configuration_speech_to_text.py

@@ -66,8 +66,6 @@ class Speech2TextConfig(PretrainedConfig):
             The dropout ratio for the attention probabilities.
         activation_dropout (`float`, *optional*, defaults to 0.0):
             The dropout ratio for activations inside the fully connected layer.
-        classifier_dropout (`float`, *optional*, defaults to 0.0):
-            The dropout ratio for classifier.
         init_std (`float`, *optional*, defaults to 0.02):
             The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
         encoder_layerdrop (`float`, *optional*, defaults to 0.0):
@@ -135,7 +133,6 @@ class Speech2TextConfig(PretrainedConfig):
         activation_dropout=0.0,
         init_std=0.02,
         decoder_start_token_id=2,
-        classifier_dropout=0.0,
         scale_embedding=True,
         pad_token_id=1,
         bos_token_id=0,
@@ -164,7 +161,6 @@ class Speech2TextConfig(PretrainedConfig):
         self.init_std = init_std
         self.encoder_layerdrop = encoder_layerdrop
         self.decoder_layerdrop = decoder_layerdrop
-        self.classifier_dropout = classifier_dropout
         self.use_cache = use_cache
         self.num_hidden_layers = encoder_layers
         self.scale_embedding = scale_embedding  # scale factor will be sqrt(d_model) if True

src/transformers/models/table_transformer/configuration_table_transformer.py

@@ -163,7 +163,6 @@ class TableTransformerConfig(PretrainedConfig):
         activation_dropout=0.0,
         init_std=0.02,
         init_xavier_std=1.0,
-        classifier_dropout=0.0,
         scale_embedding=False,
         auxiliary_loss=False,
         position_embedding_type="sine",