Removing _use_sdpa attributes and their fix-copies

src/transformers/models/altclip/modeling_altclip.py

@@ -1214,7 +1214,6 @@ class AltRobertaModel(AltCLIPPreTrainedModel):
 
     config_class = AltCLIPTextConfig
 
-    # Copied from transformers.models.bert.modeling_bert.BertModel.__init__ with Bert->AltRoberta
     def __init__(self, config, add_pooling_layer=True):
         super().__init__(config)
         self.config = config
@@ -1224,9 +1223,6 @@ def __init__(self, config, add_pooling_layer=True):
 
         self.pooler = AltRobertaPooler(config) if add_pooling_layer else None
 
-        self._use_sdpa = config._attn_implementation == "sdpa"
-        self.position_embedding_type = config.position_embedding_type
-
         # Initialize weights and apply final processing
         self.post_init()
 

src/transformers/models/camembert/modeling_camembert.py

@@ -755,7 +755,6 @@ class CamembertModel(CamembertPreTrainedModel):
 
     _no_split_modules = []
 
-    # Copied from transformers.models.bert.modeling_bert.BertModel.__init__ with Bert->Camembert
     def __init__(self, config, add_pooling_layer=True):
         super().__init__(config)
         self.config = config
@@ -765,9 +764,6 @@ def __init__(self, config, add_pooling_layer=True):
 
         self.pooler = CamembertPooler(config) if add_pooling_layer else None
 
-        self._use_sdpa = config._attn_implementation == "sdpa"
-        self.position_embedding_type = config.position_embedding_type
-
         # Initialize weights and apply final processing
         self.post_init()
 

src/transformers/models/clap/modeling_clap.py

@@ -1773,7 +1773,6 @@ class ClapTextModel(ClapPreTrainedModel):
 
     config_class = ClapTextConfig
 
-    # Copied from transformers.models.bert.modeling_bert.BertModel.__init__ with Bert->ClapText
     def __init__(self, config, add_pooling_layer=True):
         super().__init__(config)
         self.config = config
@@ -1783,9 +1782,6 @@ def __init__(self, config, add_pooling_layer=True):
 
         self.pooler = ClapTextPooler(config) if add_pooling_layer else None
 
-        self._use_sdpa = config._attn_implementation == "sdpa"
-        self.position_embedding_type = config.position_embedding_type
-
         # Initialize weights and apply final processing
         self.post_init()
 

src/transformers/models/ernie/modeling_ernie.py

@@ -810,7 +810,6 @@ class ErnieModel(ErniePreTrainedModel):
     `add_cross_attention` set to `True`; an `encoder_hidden_states` is then expected as an input to the forward pass.
     """
 
-    # Copied from transformers.models.bert.modeling_bert.BertModel.__init__ with Bert->Ernie
     def __init__(self, config, add_pooling_layer=True):
         super().__init__(config)
         self.config = config
@@ -820,9 +819,6 @@ def __init__(self, config, add_pooling_layer=True):
 
         self.pooler = ErniePooler(config) if add_pooling_layer else None
 
-        self._use_sdpa = config._attn_implementation == "sdpa"
-        self.position_embedding_type = config.position_embedding_type
-
         # Initialize weights and apply final processing
         self.post_init()
 

src/transformers/models/markuplm/modeling_markuplm.py

@@ -811,7 +811,6 @@ def from_pretrained(cls, pretrained_model_name_or_path: Optional[Union[str, os.P
     MARKUPLM_START_DOCSTRING,
 )
 class MarkupLMModel(MarkupLMPreTrainedModel):
-    # Copied from transformers.models.bert.modeling_bert.BertModel.__init__ with Bert->MarkupLM
     def __init__(self, config, add_pooling_layer=True):
         super().__init__(config)
         self.config = config
@@ -821,9 +820,6 @@ def __init__(self, config, add_pooling_layer=True):
 
         self.pooler = MarkupLMPooler(config) if add_pooling_layer else None
 
-        self._use_sdpa = config._attn_implementation == "sdpa"
-        self.position_embedding_type = config.position_embedding_type
-
         # Initialize weights and apply final processing
         self.post_init()
 

src/transformers/models/roberta/modeling_roberta.py

@@ -702,7 +702,6 @@ class RobertaModel(RobertaPreTrainedModel):
 
     """
 
-    # Copied from transformers.models.bert.modeling_bert.BertModel.__init__ with Bert->Roberta
     def __init__(self, config, add_pooling_layer=True):
         super().__init__(config)
         self.config = config
@@ -712,9 +711,6 @@ def __init__(self, config, add_pooling_layer=True):
 
         self.pooler = RobertaPooler(config) if add_pooling_layer else None
 
-        self._use_sdpa = config._attn_implementation == "sdpa"
-        self.position_embedding_type = config.position_embedding_type
-
         # Initialize weights and apply final processing
         self.post_init()
 

src/transformers/models/roc_bert/modeling_roc_bert.py

@@ -892,7 +892,6 @@ class RoCBertModel(RoCBertPreTrainedModel):
     `add_cross_attention` set to `True`; an `encoder_hidden_states` is then expected as an input to the forward pass.
     """
 
-    # Copied from transformers.models.bert.modeling_bert.BertModel.__init__ with Bert->RoCBert
     def __init__(self, config, add_pooling_layer=True):
         super().__init__(config)
         self.config = config
@@ -902,9 +901,6 @@ def __init__(self, config, add_pooling_layer=True):
 
         self.pooler = RoCBertPooler(config) if add_pooling_layer else None
 
-        self._use_sdpa = config._attn_implementation == "sdpa"
-        self.position_embedding_type = config.position_embedding_type
-
         # Initialize weights and apply final processing
         self.post_init()
 

src/transformers/models/xlm_roberta/modeling_xlm_roberta.py

@@ -704,7 +704,6 @@ class XLMRobertaModel(XLMRobertaPreTrainedModel):
 
     """
 
-    # Copied from transformers.models.bert.modeling_bert.BertModel.__init__ with Bert->XLMRoberta
     def __init__(self, config, add_pooling_layer=True):
         super().__init__(config)
         self.config = config
@@ -714,9 +713,6 @@ def __init__(self, config, add_pooling_layer=True):
 
         self.pooler = XLMRobertaPooler(config) if add_pooling_layer else None
 
-        self._use_sdpa = config._attn_implementation == "sdpa"
-        self.position_embedding_type = config.position_embedding_type
-
         # Initialize weights and apply final processing
         self.post_init()
 

src/transformers/models/xlm_roberta_xl/modeling_xlm_roberta_xl.py

@@ -667,7 +667,6 @@ class XLMRobertaXLModel(XLMRobertaXLPreTrainedModel):
     an input to the forward pass. .. _*Attention is all you need*: https://arxiv.org/abs/1706.03762
     """
 
-    # Copied from transformers.models.bert.modeling_bert.BertModel.__init__ with Bert->XLMRobertaXL
     def __init__(self, config, add_pooling_layer=True):
         super().__init__(config)
         self.config = config
@@ -677,9 +676,6 @@ def __init__(self, config, add_pooling_layer=True):
 
         self.pooler = XLMRobertaXLPooler(config) if add_pooling_layer else None
 
-        self._use_sdpa = config._attn_implementation == "sdpa"
-        self.position_embedding_type = config.position_embedding_type
-
         # Initialize weights and apply final processing
         self.post_init()
 

src/transformers/models/xmod/modeling_xmod.py

@@ -793,7 +793,6 @@ class XmodModel(XmodPreTrainedModel):
 
     """
 
-    # Copied from transformers.models.bert.modeling_bert.BertModel.__init__ with Bert->Xmod
     def __init__(self, config, add_pooling_layer=True):
         super().__init__(config)
         self.config = config
@@ -803,9 +802,6 @@ def __init__(self, config, add_pooling_layer=True):
 
         self.pooler = XmodPooler(config) if add_pooling_layer else None
 
-        self._use_sdpa = config._attn_implementation == "sdpa"
-        self.position_embedding_type = config.position_embedding_type
-
         # Initialize weights and apply final processing
         self.post_init()