Remove more unused attributes in config classes (#21000)

* Remove gradient_checkpointing from MarkupLMConfig

* Remove predict_special_tokens from OpenAIGPTConfig

* Remove enable_cls from RoCBertConfig

* Remove batch_size from TrajectoryTransformerConfig

* Remove searcher_seq_len from RealmConfig

* Remove feat_quantizer_dropout from WavLMConfig

* Remove position_biased_input from SEWDConfig

* Remove max_source_positions from Speech2Text2Config
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

src/transformers/models/markuplm/configuration_markuplm.py

@@ -65,8 +65,6 @@ class MarkupLMConfig(PretrainedConfig):
             The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
         layer_norm_eps (`float`, *optional*, defaults to 1e-12):
             The epsilon used by the layer normalization layers.
-        gradient_checkpointing (`bool`, *optional*, defaults to `False`):
-            If True, use gradient checkpointing to save memory at the expense of slower backward pass.
         max_tree_id_unit_embeddings (`int`, *optional*, defaults to 1024):
             The maximum value that the tree id unit embedding might ever use. Typically set this to something large
             just in case (e.g., 1024).
@@ -119,7 +117,6 @@ class MarkupLMConfig(PretrainedConfig):
         pad_token_id=0,
         bos_token_id=0,
         eos_token_id=2,
-        gradient_checkpointing=False,
         max_xpath_tag_unit_embeddings=256,
         max_xpath_subs_unit_embeddings=1024,
         tag_pad_id=216,
@@ -135,7 +132,6 @@ class MarkupLMConfig(PretrainedConfig):
             pad_token_id=pad_token_id,
             bos_token_id=bos_token_id,
             eos_token_id=eos_token_id,
-            gradient_checkpointing=gradient_checkpointing,
             **kwargs,
         )
         self.vocab_size = vocab_size

src/transformers/models/openai/configuration_openai.py

@@ -60,8 +60,6 @@ class OpenAIGPTConfig(PretrainedConfig):
             The epsilon to use in the layer normalization layers
         initializer_range (`float`, *optional*, defaults to 0.02):
             The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
-        predict_special_tokens (`bool`, *optional*, defaults to `True`):
-            Whether or not special tokens should be predicted when the model has a language modeling head.
         summary_type (`str`, *optional*, defaults to `"cls_index"`):
             Argument used when doing sequence summary, used in the models [`OpenAIGPTDoubleHeadsModel`] and
             [`OpenAIGPTDoubleHeadsModel`].
@@ -133,7 +131,6 @@ class OpenAIGPTConfig(PretrainedConfig):
         attn_pdrop=0.1,
         layer_norm_epsilon=1e-5,
         initializer_range=0.02,
-        predict_special_tokens=True,
         summary_type="cls_index",
         summary_use_proj=True,
         summary_activation=None,
@@ -152,7 +149,6 @@ class OpenAIGPTConfig(PretrainedConfig):
         self.attn_pdrop = attn_pdrop
         self.layer_norm_epsilon = layer_norm_epsilon
         self.initializer_range = initializer_range
-        self.predict_special_tokens = predict_special_tokens
         self.summary_type = summary_type
         self.summary_use_proj = summary_use_proj
         self.summary_activation = summary_activation

src/transformers/models/realm/configuration_realm.py

@@ -110,8 +110,6 @@ class RealmConfig(PretrainedConfig):
         searcher_beam_size (`int`, *optional*, defaults to 5000):
             Beam size of the searcher. Note that when eval mode is enabled, *searcher_beam_size* will be the same as
             *reader_beam_size*.
-        searcher_seq_len (`int`, *optional*, defaults to 64):
-            Maximum sequence length of the searcher.
 
     Example:
 
@@ -152,7 +150,6 @@ class RealmConfig(PretrainedConfig):
         reader_seq_len=320,  # 288 + 32
         num_block_records=13353718,
         searcher_beam_size=5000,
-        searcher_seq_len=64,
         pad_token_id=1,
         bos_token_id=0,
         eos_token_id=2,
@@ -186,4 +183,3 @@ class RealmConfig(PretrainedConfig):
         # Retrieval config
         self.num_block_records = num_block_records
         self.searcher_beam_size = searcher_beam_size
-        self.searcher_seq_len = searcher_seq_len

src/transformers/models/roc_bert/configuration_roc_bert.py

@@ -77,8 +77,6 @@ class RoCBertConfig(PretrainedConfig):
             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.
-        enable_cls (`bool`, *optional*, defaults to `True`):
-            Whether or not the model use cls loss when pretrained.
         enable_pronunciation (`bool`, *optional*, defaults to `True`):
             Whether or not the model use pronunciation embed when training.
         enable_shape (`bool`, *optional*, defaults to `True`):
@@ -131,7 +129,6 @@ class RoCBertConfig(PretrainedConfig):
         pad_token_id=0,
         position_embedding_type="absolute",
         classifier_dropout=None,
-        enable_cls=True,
         enable_pronunciation=True,
         enable_shape=True,
         pronunciation_embed_dim=768,
@@ -154,7 +151,6 @@ class RoCBertConfig(PretrainedConfig):
         self.type_vocab_size = type_vocab_size
         self.layer_norm_eps = layer_norm_eps
         self.use_cache = use_cache
-        self.enable_cls = enable_cls
         self.enable_pronunciation = enable_pronunciation
         self.enable_shape = enable_shape
         self.pronunciation_embed_dim = pronunciation_embed_dim

src/transformers/models/sew_d/configuration_sew_d.py

@@ -63,8 +63,6 @@ class SEWDConfig(PretrainedConfig):
             Whether to share attention key with c2p and p2c.
         relative_attention (`bool`, *optional*, defaults to `True`):
             Whether to use relative position encoding.
-        position_biased_input (`bool`, *optional*, defaults to `False`):
-            Whether to add absolute position embedding to content embedding.
         pos_att_type (`Tuple[str]`, *optional*, defaults to `("p2c", "c2p")`):
             The type of relative position attention, it can be a combination of `("p2c", "c2p")`, e.g. `("p2c")`,
             `("p2c", "c2p")`, `("p2c", "c2p")`.
@@ -183,7 +181,6 @@ class SEWDConfig(PretrainedConfig):
         position_buckets=256,
         share_att_key=True,
         relative_attention=True,
-        position_biased_input=False,
         pos_att_type=("p2c", "c2p"),
         norm_rel_ebd="layer_norm",
         hidden_act="gelu_python",
@@ -239,7 +236,6 @@ class SEWDConfig(PretrainedConfig):
         self.share_att_key = share_att_key
         self.relative_attention = relative_attention
         self.norm_rel_ebd = norm_rel_ebd
-        self.position_biased_input = position_biased_input
         self.pos_att_type = list(pos_att_type)
         self.hidden_act = hidden_act
         self.num_attention_heads = num_attention_heads

src/transformers/models/speech_to_text_2/configuration_speech_to_text_2.py

@@ -68,8 +68,6 @@ class Speech2Text2Config(PretrainedConfig):
             for more details.
         use_cache (`bool`, *optional*, defaults to `True`):
             Whether or not the model should return the last key/values attentions (not used by all models).
-        max_source_positions (`int`, *optional*, defaults to 6000):
-            The maximum sequence length of log-mel filter-bank features that this model might ever be used with.
         max_target_positions (`int`, *optional*, defaults to 1024):
             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).
@@ -111,7 +109,6 @@ class Speech2Text2Config(PretrainedConfig):
         pad_token_id=1,
         bos_token_id=0,
         eos_token_id=2,
-        max_source_positions=6000,
         max_target_positions=1024,
         **kwargs
     ):
@@ -129,7 +126,6 @@ class Speech2Text2Config(PretrainedConfig):
         self.use_cache = use_cache
         self.num_hidden_layers = decoder_layers
         self.scale_embedding = scale_embedding  # scale factor will be sqrt(d_model) if True
-        self.max_source_positions = max_source_positions
         self.max_target_positions = max_target_positions
 
         super().__init__(

src/transformers/models/trajectory_transformer/configuration_trajectory_transformer.py

@@ -45,8 +45,6 @@ class TrajectoryTransformerConfig(PretrainedConfig):
         vocab_size (`int`, *optional*, defaults to 100):
             Vocabulary size of the TrajectoryTransformer model. Defines the number of different tokens that can be
             represented by the `trajectories` passed when calling [`TrajectoryTransformerModel`]
-        batch_size (`int`, *optional*, defaults to 256):
-            Size of the batch of trajectories passed to the model.
         action_weight (`int`, *optional*, defaults to 5):
             Weight of the action in the loss function
         reward_weight (`int`, *optional*, defaults to 1):
@@ -115,7 +113,6 @@ class TrajectoryTransformerConfig(PretrainedConfig):
     def __init__(
         self,
         vocab_size=100,
-        batch_size=256,
         action_weight=5,
         reward_weight=1,
         value_weight=1,
@@ -142,7 +139,6 @@ class TrajectoryTransformerConfig(PretrainedConfig):
         **kwargs
     ):
         self.vocab_size = vocab_size
-        self.batch_size = batch_size
         self.action_weight = action_weight
         self.reward_weight = reward_weight
         self.value_weight = value_weight

src/transformers/models/wavlm/configuration_wavlm.py

@@ -75,8 +75,6 @@ class WavLMConfig(PretrainedConfig):
         feat_extract_activation (`str, `optional`, defaults to `"gelu"`):
             The non-linear activation function (function or string) in the 1D convolutional layers of the feature
             extractor. If string, `"gelu"`, `"relu"`, `"selu"` and `"gelu_new"` are supported.
-        feat_quantizer_dropout (`float`, *optional*, defaults to 0.0):
-            The dropout probabilitiy for quantized feature encoder states.
         conv_dim (`Tuple[int]` or `List[int]`, *optional*, defaults to `(512, 512, 512, 512, 512, 512, 512)`):
             A tuple of integers defining the number of input and output channels of each 1D convolutional layer in the
             feature encoder. The length of *conv_dim* defines the number of 1D convolutional layers.
@@ -124,8 +122,6 @@ class WavLMConfig(PretrainedConfig):
             Number of codevector groups for product codevector quantization.
         contrastive_logits_temperature (`float`, *optional*, defaults to 0.1):
             The temperature *kappa* in the contrastive loss.
-        feat_quantizer_dropout (`float`, *optional*, defaults to 0.0):
-            The dropout probabilitiy for the output of the feature encoder that's used by the quantizer.
         num_negatives (`int`, *optional*, defaults to 100):
             Number of negative samples for the contrastive loss.
         codevector_dim (`int`, *optional*, defaults to 256):
@@ -205,7 +201,6 @@ class WavLMConfig(PretrainedConfig):
         activation_dropout=0.1,
         attention_dropout=0.1,
         feat_proj_dropout=0.0,
-        feat_quantizer_dropout=0.0,
         final_dropout=0.1,
         layerdrop=0.1,
         initializer_range=0.02,
@@ -308,7 +303,6 @@ class WavLMConfig(PretrainedConfig):
         self.num_codevectors_per_group = num_codevectors_per_group
         self.num_codevector_groups = num_codevector_groups
         self.contrastive_logits_temperature = contrastive_logits_temperature
-        self.feat_quantizer_dropout = feat_quantizer_dropout
         self.num_negatives = num_negatives
         self.codevector_dim = codevector_dim
         self.proj_codevector_dim = proj_codevector_dim