Transformer&QuestionAnsweringThangLuong.pdf
TheTransformernetworkconsistsofmultiplelayers,eachwithseveralAttentionHeads(andadditionallayers),usedtolearndifferentrelationshipsbetweentokens.AsinmanyNLPmodels,theinputtokensarefirstembeddedintovectorsThisslideThangLuongfromgooglebraintotalkaboutdetailtransfomernetwork
下载地址
用户评论