HYF Blog

Thinking will not overcome fear but action will.

Gpt

GPT-1 Transformer整体结构 Transformer Decoder结构说明 1.Sequence mask sequence mask是为了使得decoder不能看见未来的信息。也就是对于一个序列，在time_step为t的时刻，我们的解码输出应该只能依赖于t时刻之前的输出，而不能依赖t之后的输出。因此我们需要想一个办法，把t之后的信息给...

Posted by HYF Blog on October 2, 2020

Xgboost

layout: post title: XgBoost subtitle: BY Blog 添加 Gitalk 的评论插件了 date: 2020-08-24 author: BY header-img: img/post-bg-universe.jpg catalog: true tags: - Blog —

Posted by HYF Blog on September 19, 2020

优化器

layout: post title: Optimizer subtitle: BY Blog 添加 Gitalk 的评论插件了 date: 2020-08-24 author: BY header-img: img/post-bg-universe.jpg catalog: true tags: - Blog — 优化器算法Optim...

Posted by HYF Blog on September 16, 2020

激活函数

BY Blog 添加 Gitalk 的评论插件了

Posted by BY on September 11, 2020

熵

BY Blog 添加 Gitalk 的评论插件了

Posted by BY on September 11, 2020

Bert,attention机制原理

Bert 模型结构图 第一部分：Bert Embedding class BERTEmbedding(nn.Module): Bert Embedding 由三部分组成， TokenEmbedding是单词embedding的结果， PositionalEmbedding是位置编码，用sin、cos公式计算 SegmentEmbedding是用来区分段落的（不知道为什么初始化...

Posted by HYF Blog on September 8, 2020

Word2vec

BY Blog 添加 Gitalk 的评论插件了

Posted by BY on September 7, 2020

softmax与sigmoid

BY Blog 添加 Gitalk 的评论插件了

Posted by BY on September 7, 2020

迭代器生成器装饰器

BY Blog 添加 Gitalk 的评论插件了

Posted by BY on September 7, 2020

BN层的作用和原理

BY Blog 添加 Gitalk 的评论插件了

Posted by BY on September 7, 2020

FEATURED TAGS
Blog
ABOUT ME

Goals determine what you going to be!

✉️ huiyanfei.buaa@gmail.com

FRIENDS