Skip to content

[None][perf] Fold q/k/v quantization into qknorm_rope_fused kernel & remove contiguous - #17093

Merged
brb-nv merged 3 commits into
NVIDIA:mainfrom
brb-nv:user/brb/fold-qkv-quant-qknorm-rope-main
Aug 8, 2026
Merged

brb-nv merged 3 commits into
NVIDIA:mainfrom
brb-nv:user/brb/fold-qkv-quant-qknorm-rope-main