Upload roberta_layers.py

roberta_layers.py

@@ -739,9 +739,9 @@ class RobertaForMaskedLM(RobertaPreTrainedModel):
                 'If you want to use `RobertaForMaskedLM` make sure `config.is_decoder=False` for '
                 'bi-directional self-attention.')
 
-        self.bert = RobertaModel(config, add_pooling_layer=False)
-        self.cls = RobertaOnlyMLMHead(config,
-                                   self.bert.embeddings.word_embeddings.weight)
+        self.roberta  = RobertaModel(config, add_pooling_layer=False)
+        self.lm_head  = RobertaOnlyMLMHead(config,
+                                   self.roberta.embeddings.word_embeddings.weight)
 
         # Initialize weights and apply final processing
         self.post_init()
@@ -820,7 +820,7 @@ class RobertaForMaskedLM(RobertaPreTrainedModel):
 
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
-        outputs = self.bert(
+        outputs = self.roberta(
             input_ids,
             attention_mask=attention_mask,
             token_type_ids=token_type_ids,
@@ -905,7 +905,7 @@ class RobertaForSequenceClassification(RobertaPreTrainedModel):
         self.num_labels = config.num_labels
         self.config = config
 
-        self.bert = RobertaModel(config)
+        self.roberta = RobertaModel(config)
         classifier_dropout = (config.classifier_dropout
                               if config.classifier_dropout is not None else
                               config.hidden_dropout_prob)
@@ -969,7 +969,7 @@ class RobertaForSequenceClassification(RobertaPreTrainedModel):
 
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
-        outputs = self.bert(
+        outputs = self.roberta(
             input_ids,
             attention_mask=attention_mask,
             token_type_ids=token_type_ids,