XNews

Datawhale and RadixArk launch open course to build mini-sglang from scratch

Datawhale and RadixArk launch open course to build mini-sglang from scratch

Datawhale and RadixArk (SGLang) have launched an open-source course that teaches LLM inference by building a mini-sglang from scratch. It covers concepts (prefill/decode, roofline model), then implements KV cache, scheduler, paged memory, RadixAttention, and finally dives into real SGLang code for advanced optimizations. The course includes a chapter on coding ethics and is designed to turn learners into contributors.

Source posts