| from transformers import BertModel | |
| from torch import nn | |
| class BERTClassifier(nn.Module): | |
| def __init__(self, bert_model_name, num_classes): | |
| super(BERTClassifier, self).__init__() | |
| self.bert = BertModel.from_pretrained(bert_model_name) | |
| self.dropout = nn.Dropout(0.1) | |
| self.fc = nn.Linear(self.bert.config.hidden_size, num_classes) | |
| def forward(self, input_ids, attention_mask): | |
| outputs = self.bert(input_ids=input_ids, attention_mask=attention_mask) | |
| pooled_output = outputs.pooler_output | |
| x = self.dropout(pooled_output) | |
| logits = self.fc(x) | |
| return logits | |