Someone wanted to learn this too, so Grasp built them a personal learning path.
You can keep exploring this one or create your own
大模型后训练与智能体
Recommended resources
The freely available resources this course will direct you to.
- PyTorch Course (2022), Part 1: Tensorswww.youtube.com
- Nonsquare matrices as transformations between dimensions | Chapter 8, Essence of linear algebrawww.youtube.com
- Linear — PyTorch 2.9 documentationdocs.pytorch.org
- 2.3. Linear Algebra — Dive into Deep Learning 1.0.3 documentationd2l.ai
- Multiplying a matrix by a matrix | Matrices | Precalculus | Khan Academywww.youtube.com
- Vector dot product and vector length | Vectors and spaces | Linear Algebra | Khan Academywww.youtube.com
Module 1
后训练所需的数学语言
Module 2
统计学习、信息论与优化基础
Module 3
Transformer 与自回归语言模型
Module 4
监督微调的数据与训练机制
Module 5
人类反馈数据与奖励模型
Module 6
强化学习与策略梯度基础
Module 7
基于 PPO 的 RLHF 训练
Module 8
直接偏好优化与 GRPO
Module 9
后训练评测、数据飞轮与实验设计
Module 10
训练系统、显存与分布式基础
Module 11
训练框架定位与源码修改
Module 12