如何选取Bert输出层的维度
bert
TensorFlow code and pre-trained models for BERT
项目地址:https://gitcode.com/gh_mirrors/be/bert

·
尝试了bert的-2层每个sentence被转化为128768 = 98304维的向量;-1层,每个sentence 被转化为7 768维向量
尝试了后接轻量网络的维度设置 ,50维和10维
bert
TensorFlow code and pre-trained models for BERT
项目地址:https://gitcode.com/gh_mirrors/be/bert




TensorFlow code and pre-trained models for BERT
最近提交(Master分支:7 个月前 )
eedf5716
Add links to 24 smaller BERT models. 5 年前
8028c045 - 5 年前
更多推荐
所有评论(0)