Joel Farthing 3c58ae373a loader: add `--defer-ple` to keep per-layer token embedding out of resident memory (#2389) ... * loader: disable expert deferral on huge-page mappings * loader: add --defer-ple for the per-layer token embedding		2026-08-31 18:49:44 +02:00
..
llama.h	loader: add `--defer-ple` to keep per-layer token embedding out of resident memory (#2389)	2026-08-31 18:49:44 +02:00