def __init__(self, config, model_argobj=None):
NLL.__init__(self, model_argobj)
RobertaForSequenceClassification.__init__(self, config)
self.embeddingHead = nn.Linear(config.hidden_size, 768)
self.norm = nn.LayerNorm(768)
self.apply(self._init_weights)
self.sparse_attention_mask_query = None
self.sparse_attention_mask_document = None
self.is_representation_l2_normalization = False # switch for L2 normalization after output
self.is_projection_l2_normalization = False # do l2 normalization on an extra non-linear projection layer
def __init__(self, config, model_argobj=None):
NLL.__init__(self, model_argobj)
RobertaForSequenceClassification.__init__(self, config)
self.embeddingHead = nn.Linear(config.hidden_size, 768)
self.norm = nn.LayerNorm(768)
self.apply(self._init_weights)
