57 posts in total

---

2025

05-12

Optimizer Factory -- 写一个能够按层衰减的优化器工厂

05-12

RLHF -- DPO

05-12

RLHF -- GRPO

05-12

RLHF -- From Zero to PPO 代码篇

05-12

RLHF -- From Zero to PPO 理论篇

05-12

最初的sin/cos编码

05-12

Why model.enable_input_require_grads()?

05-08

Rethinking R1-like Rule-based RL

05-08

05：矩阵分解

2024

01-07

02：信号的分析方法