// the find
afshinea/stanford-cme-295-transformers-large-language-models
VIP cheatsheet for Stanford's CME 295 Transformers and Large Language Models
A single-PDF cheatsheet covering Stanford's CME 295 course on transformers and LLMs: attention mechanisms, finetuning (LoRA, RLHF, DPO), distributed training tricks like KV caching and speculative decoding, and agent/eval topics. It's for someone who wants a dense, exam-style reference to the current LLM stack rather than a working codebase — there's no code in this repo at all.
The authors are the same pair behind the widely-used Stanford CS229/CS230 cheatsheets, so the content is dense and technically accurate rather than a marketing summary of buzzwords. It's translated into 15 languages, which is unusual for this kind of technical reference and clearly has real maintenance effort behind it. The scope is genuinely broad for a single document — SFT, RLHF, DPO, RLVR, speculative decoding, and diffusion LLMs all get a page, which is a decent index of what to go read papers on.
There is zero code — no notebooks, no runnable examples, nothing you can clone and execute, so this isn't something you 'use' in a project, only something you read. The whole repo is a wrapper around one PDF per language; commit history on a PDF is opaque, so you can't diff what changed between versions the way you could with a real doc-as-code setup. It also functions as a funnel to a paid book (superstudy.guide) — the free cheatsheet is explicitly described as a teaser for the ~250-page book, so don't expect the depth implied by the topic list to actually be in the free PDF.