llama: add pshard runtime for plan switching and streamed weights - #22692

Open

aukarande wants to merge 2 commits into

ggml-org:masterfrom

aukarande:pshard/inference

Commits on May 4, 2026

feat: pipeshard plan switch/decode
aukarande
committed
e14c6cf

Commits on Jun 20, 2026

Fix pshard prefetch tracking, state remaps, and common-weight preload
aukarande
committed
c64c73d