Full-access lesson
Context Windows
> Under this chapter's 32-layer, 32-KV-head, 128-dimensional, fp16 MHA configuration, 32 fully occupied 8,192-token sequences require 128 GiB of raw KV payload. GQA, MQA, lower-precision caches, paging, fragmentation, metadata, and workspace change actual allocation.
The complete curriculum
All 12 volumes, companion resources, interviews, and architecture reviews.
Executable engineering practice
Subscriber-only Python, Java, TypeScript, and SQL labs in the isolated runner.