▲ 516 today2k starsPython
A Free Course Teaches You to Build an LLM Inference Engine
What is zero-to-sglang?
zero-to-sglang is a hands-on tutorial that walks you through building a miniature version of the SGLang inference engine from scratch and then guides you to read the real source code. It covers KV cache, prefill vs decode, continuous batching and more, with videos in English and Chinese.
Who it is for
If you want to understand LLM serving bottlenecks or contribute to SGLang, this course gives you practical code you can run and a path to your first PR.
Try it in 60 seconds
No one-line install. The README has the setup steps.
No license listedUpdated todayPython5 open issues
New projects are not vetted. Skim the code, and run it in a fresh folder or a container before giving it your files or keys.
The story so far
Early Signal first spotted datawhalechina/zero-to-sglang on November 10, 2026, when it had 1.9k GitHub stars. It has since grown to 2k. On its busiest day it gained 516 stars. Early Signal files it under Inference, Video, LLMs.