🚀🚀 CS336-From-Scratch Spring 2026🚀🚀

The NoteBook and Assignments implemention via Learning CS336 Spring 2026！

“And in the end, the love you take is equal to the love you make.” —— The Beatles

I'm mungeryang, a master's student from the University of Chinese Academy of Sciences(UCAS-iie). In this Repo, I have open-sourced all of my study notes, implementation details for assignments, and results.

Class HomePage：https://cs336.stanford.edu/

图解大模型架构参见：Hand-Drawn-LLM

🧐50 $QA_{s}$ About LLM - 大模型面试50问

通过整理 Standford CS336 Spring26 课堂笔记，总结大模型算法经典面试50问

参考资料: CS336课堂笔记、李博杰老师 - 大模型面试题 200 问、百面大模型、大模型技术30讲

⭐️⭐️⭐️ 50问整理

P.S. 由于精力和能力有限，仅整理出我本人认为较为经典的50问。问题与作答全部开源，欢迎任何感兴趣的人fork更新～ 👏👏👏

💻 Assignments

Assignment 1: Basics

• Implement all of the components (tokenizer, model architecture, optimizer) necessary to train a standard Transformer language model
• Train a minimal language model

Assignment1	Status	Link
train_bpe	✅	BPE Implementation
BPETokenizer	✅	Tiny_BPETokenizer Class Implementation
Linear	✅	Linear Class
Emebdding	✅	EMbedding Class
RMSNorm	✅	RMSNorm
Swiglu	✅	SwiGLU FFN
RoPE	✅	RoPE Class
softmax	✅	softmax funcion
attention	✅	Scaled_Dot_Attn
mul-attn	✅	MultiHeadAttn Class
LM block	✅	Transformer Block
cross-entropy	✅	train function
AdamW	✅	Adamw optimizer

⌨️ Assignment 1: Results

Assignment 2: Systems

• Profile and benchmark the model and layers from Assignment 1 using advanced tools, optimize Attention with your own Triton implementation of FlashAttention2

• Build a memory-efficient, distributed version of the Assignment 1 model training code

Assignment2	Status	Link

Assignment 3: Scaling

• Understand the function of each component of the Transformer
• Query a training API to fit a scaling law to project model scaling

Assignment 4: Data

• Convert raw Common Crawl dumps into usable pretraining data
• Perform filtering and deduplication to improve model performance

Assignment 5: Alignment

• Apply supervised finetuning and reinforcement learning to train LMs to reason when solving math problems
• Optional Part 2: implement and apply safety alignment methods such as DPO

⭐️ Star History