CS 336: Language Modeling from Scratch

cs336
stanford
language-models
Stanford CS 336 (Spring 2026). Building language models end-to-end: tokenization, architectures, training systems, scaling, data, and alignment.
Author

Rohan Rajput

Published

March 30, 2026

Stanford CS 336 — Language Modeling from Scratch, Spring 2026. Instructors: Tatsunori Hashimoto and Percy Liang.

The course walks through building a modern language model end-to-end: tokenization, architecture, optimization, parallel training, inference, scaling laws, data curation, and post-training alignment (SFT, RLHF, DPO, RL for reasoning).

My write-ups for each assignment live below. Lecture notes will appear here as I finish them.

Assignments

# Due Title
1 Apr 15, 2026 Assignment 1: Basics
No matching items
# Date Title
No matching items