MCPcopy Create free account
hub / github.com/THUDM/GLM / forward

Method forward

model/modeling_bert.py:1132–1143  ·  view source on GitHub ↗
(self, input_ids, token_type_ids=None, attention_mask=None, next_sentence_label=None,
                checkpoint_activations=False)

Source from the content-addressed store, hash-verified

1130 self.apply(self.init_bert_weights)
1131
1132 def forward(self, input_ids, token_type_ids=None, attention_mask=None, next_sentence_label=None,
1133 checkpoint_activations=False):
1134 _, pooled_output = self.bert(input_ids, token_type_ids, attention_mask,
1135 output_all_encoded_layers=False, checkpoint_activations=checkpoint_activations)
1136 seq_relationship_score = self.cls(pooled_output)
1137
1138 if next_sentence_label is not None:
1139 loss_fct = CrossEntropyLoss(ignore_index=-1)
1140 next_sentence_loss = loss_fct(seq_relationship_score.view(-1, 2), next_sentence_label.view(-1))
1141 return next_sentence_loss
1142 else:
1143 return seq_relationship_score
1144
1145
1146class BertForSequenceClassification(PreTrainedBertModel):

Callers

nothing calls this directly

Calls

no outgoing calls

Tested by

no test coverage detected