Bert学习小记

BERT: Pre-training of Deep Bidirectional Transformers forLanguage Understanding embedding input embedding = token embedding + segmentation embedding + position embedding segment embedding 对于句子对来说（两个句子

>>阅读原文<<