qwen4exp: compute QSA cache-layout inputs once per ubatch, not per layer - #150

qwen4exp: compute QSA cache-layout inputs once per ubatch, not per layer#150

talexeyft wants to merge 2 commits into

unslothai:qwen4exp/qwen3.8-flash-nextunslothai/llama.cpp:qwen4exp/qwen3.8-flash-nextfrom

talexeyft:qwen4exp-qsa-perftalexeyft/llama.cpp:qwen4exp-qsa-perfCopy head branch name to clipboard