Bert學習小記

BERT: Pre-training of Deep Bidirectional Transformers forLanguage Understanding embedding input embedding = token embedding + segmentation embedding + position embedding segment embedding 對於句子對來說（兩個句子

>>阅读原文<<