Stanford's build-an-LLM-from-scratch course CS336 publishes full lectures, slides and code

stanfordnlp · x · 2026-11-16

A retweet amplified by Stanford NLP highlights CS336, a full course on building an LLM from scratch. The schedule covers tokenization, PyTorch and FLOPs/memory accounting, architectures and hyperparameters, attention variants and MoE, GPUs/TPUs, Triton kernels, parallelism, scaling laws, and inference — with all lecture notes and code files publicly available, taught by Percy Liang and Tatsunori Hashimoto.

Original post →

More from Companies & People

Companies & People channel →