dela a2c4217dae 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.

2026-08-26 14:43:58 +08:00

..

sections

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

2026-08-26 14:43:58 +08:00

ledger.yaml

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

2026-08-26 14:43:58 +08:00

notes-macros.tex

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

2026-08-26 14:43:58 +08:00

notes.pdf

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

2026-08-26 14:43:58 +08:00

notes.tex

Initial K3 snapshot: 0.5B KDA/MLA/MoE train path

2026-08-25 14:43:17 +08:00

verify_gref_overflow.py

Initial K3 snapshot: 0.5B KDA/MLA/MoE train path

2026-08-25 14:43:17 +08:00

verify_l2norm_stability.py

Initial K3 snapshot: 0.5B KDA/MLA/MoE train path

2026-08-25 14:43:17 +08:00

verify_severity_sweep.py

Initial K3 snapshot: 0.5B KDA/MLA/MoE train path

2026-08-25 14:43:17 +08:00