Set default name of Bert instance to "backbone"

[jbischof]

Current default name is "bert", which looks pretty funny in the layers of a task model (see below).

Having "bert_classifier" as the default for a BertClassifier looks pretty good though.

classifier = keras_nlp.models.BertClassifier(
    "bert_tiny_uncased_en",
    num_classes=2,
)
classifier.summary()

>>>
Model: "bert_classifier"
__________________________________________________________________________________________________
 Layer (type)                   Output Shape         Param #     Connected to                     
==================================================================================================
 padding_mask (InputLayer)      [(None, None)]       0           []                               
                                                                                                  
 segment_ids (InputLayer)       [(None, None)]       0           []                               
                                                                                                  
 token_ids (InputLayer)         [(None, None)]       0           []                               
                                                                                                  
 bert (Bert)                    {'sequence_output':  4385920     ['padding_mask[0][0]',           
                                 (None, None, 128),               'segment_ids[0][0]',            
                                 'pooled_output': (               'token_ids[0][0]']              
                                None, 128)}                                                       
                                                                                                  
 logits (Dense)                 (None, 2)            258         ['bert[0][0]']                   
                                                                                                  
==================================================================================================
Total params: 4,386,178
Trainable params: 4,386,178
Non-trainable params: 0

[mattdangerw]

Thanks!