def __init__(self, config, weight=None):
super(ElectraForSequenceClassification, self).__init__(config)
self.num_labels = config.num_labels
self.bert = ElectraModel(config)
self.classifier = ElectraClassificationHead(config)
self.init_weights()
def __init__(self, config, weight=None):
super().__init__(config)
self.num_labels = config.num_labels
self.electra = ElectraModel(config)
self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)
self.init_weights()
def __init__(self, config):
super().__init__()
self._mask_prob = config.mask_prob
self._mask_token_id = config.mask_token_id
self._masking_strategy = config.masking_strategy
self.backbone = ElectraModel(config)
self.generator_head = GeneratorHead(config)
def __init__(self, config, weight=None):
super().__init__(config)
self.num_labels = config.num_labels
self.electra = ElectraModel(config)
self.pooler = ElectraPooler(config)
self.dropout = nn.Dropout(config.hidden_dropout_prob)
self.classifier = nn.Linear(config.hidden_size, self.config.num_labels)
self.weight = weight
def __init__(self, config, pos_weight=None):
super(ElectraForMultiLabelSequenceClassification,
self).__init__(config)
self.num_labels = config.num_labels
self.pos_weight = pos_weight
self.electra = ElectraModel(config)
self.pooler = ElectraPooler(config)
self.dropout = nn.Dropout(config.hidden_dropout_prob)
self.classifier = nn.Linear(config.hidden_size, self.config.num_labels)
def __init__(self, config, pos_weight=None, regression=False, loss_fct=None):
super(ElectraForMultiLabelSequenceClassification, self).__init__(config)
self.num_labels = config.num_labels
self.pos_weight = pos_weight
self.electra = ElectraModel(config)
self.pooler = ElectraPooler(config)
self.dropout = nn.Dropout(config.hidden_dropout_prob)
self.classifier = nn.Linear(config.hidden_size, self.config.num_labels)
self.regression = regression
if loss_fct is None:
if self.regression:
self.loss_fct = torch.nn.MSELoss()
else:
self.loss_fct = torch.nn.BCEWithLogitsLoss(pos_weight=self.pos_weight)
elif loss_fct == "MAELoss":
self.loss_fct = torch.nn.L1Loss()
elif loss_fct == "MSELoss":
self.loss_fct = torch.nn.MSELoss()
self.init_weights()