Skip to content

Cut inference and AX hot-path waste: P-core threads, gated logprob, halved KV, geometry cache#667

Merged
FuJacob merged 1 commit into
mainfrom
perf-inference-hot-paths
Jun 11, 2026
Merged

Cut inference and AX hot-path waste: P-core threads, gated logprob, halved KV, geometry cache#667
FuJacob merged 1 commit into
mainfrom
perf-inference-hot-paths

Skip discarded per-token logprob, cache display geometry, hoist norma…

0469a14
Select commit
Loading
Failed to load commit list.