qiaoyu-tan/sine

sine's Issues

代码与论文不一致 Code does not match the paper

1. X_hat^u 加上X^u的残差为什么没在论文中提及?
2. C^u_apt 缺少LayerNorm4以及后续多出来的运算?
3. (7)的τ去哪了
4. (3)里面为什么有τ
5. (3)的W3去哪了
6. (5)的LayerNorm3在哪
7. 很多地方加上了position_embedding, 论文为什么不提及?

---

1. why is the residual of X_hat^u plus X^u not mentioned in the paper?
2. C^u_apt is missing LayerNorm4 and the subsequent extra operations?
3. where is the τ of (7)?
4. why is there τ in (3)?
5. where did the W3 of (3) go?
6. where is the LayerNorm3 of (5)?
7. many places added position_embedding, why not mentioned in the paper?