Glm4 mtp optimizations#4

Draft

SamuelOliveirads wants to merge 5 commits intoF1LM1:glm4-moe-mtpfrom

SamuelOliveirads:glm4-mtp-optimizations

Commits on Oct 12, 2025

mtp-graph(feat): Reactivate graph reuse only for main model path
SamuelOliveirads
committed

Commits on Oct 16, 2025

Merge branch 'glm4-mtp-batch' of https://github.com/SamuelOliveirads/llama.cpp into glm4-mtp-graph-cache
SamuelOliveirads
committed

Commits on Oct 23, 2025

mtp-graph (wip): testing different ways to allow graph reuse
SamuelOliveirads
committed
mtp-op (feat): merge update kv and draft into one operation
SamuelOliveirads
committed

Commits on Oct 24, 2025

mtp-op (refactor): fix language for log
SamuelOliveirads
committed