Perf/qwen3 moe graph runtime - #516

Open

wooway777 wants to merge 3 commits into

mainfrom

perf/qwen3-moe-graph-runtime

Commits on Jul 28, 2026

perf(graph): bound paged captures by batch capacity

surprisely
authored and
wooway777
committed
4f080a2

Commits on Jul 29, 2026

perf(graph): skip empty non-quant runtime resets

surprisely
authored and
wooway777
committed
0b595a3
fix(graph): preserve runtime resets for other models

surprisely
authored and
wooway777
committed
ecee461