article reading note

【ing】

attention sink

知乎：深度解析 Attention Sinks 究竟为啥有效？
《How Attention Sinks Keep Language Models Stable》

1. 模型会将大量的注意力“倾倒”在最初的几个token上
2. 两种实现：
   • StreamingLLM方法是在序列开头设置一个专门的、可学习的池token $k_0$，其注意力计算方式为：$attn(q_i,K)=softmax(q_i^T{K}_c,q_i^T{k}_0)$ ，其中$K_c$是内容token。
   • OpenAI则采用了一个更简洁的通用标量方法： $attn(q_i,K)=softmax(q_i^T{K}_c,b)$

AI即将进入下半场

算法->环境（解决什么问题，怎么评估）

The N implementation details of rlhf with ppo

lm-human-preference的一些代码实现细节

• local_seed = seed + rank * 100003 The seed is going to make the model produce different responses and get different scores

The 37 Implementation Details of PPO

An Initial Exploration of Theoretical Support for Language Model Data Engineering. Part 1: Pretraining

• grokking: the speed of learning a skill
• skill may have different grokking point during training
• it is not the case that the model cannot learn — it is the speed that is being different, and the model learns faster for some formats.
• the more detailed the COT is, the faster the model learn.
• Training on a particular ordering of data can give faster learning speed than training on skill-specific data.
• Different mix ratio may result in different speed of learning
• results of data engineering on model scale les