Document LatentMoE sigmoid routing, sparse dispatch, and K3 block figures

Ledger C10–C12 match the permute-pad-bmm path and Switch aux/z-loss.
Section 8 adds overview and component TikZ; MoE capacity is C_moe so it
does not collide with KDA chunk size.
This commit is contained in:
dela
2026-08-26 14:43:58 +08:00
parent ea7167b3f7
commit a2c4217dae
6 changed files with 239 additions and 18 deletions
+1
View File
@@ -10,6 +10,7 @@
\usepackage{subcaption}
\usepackage{float}
\usepackage{tikz}
\usetikzlibrary{positioning, arrows.meta, decorations.pathreplacing, calc}
\usepackage{hyperref}
\usepackage{xcolor}
\usepackage{multicol}