2 Commits

Include renames

Author	SHA1	Message	Date
Niko1221	465cd61dfb	Formats cost nothing unless used: Q6_K gate/up experts in the normal build; the K-quant MMQ prompt kernels on by default, on CUDA as their own library (strata_kq) opened only for a model with these experts and STRATA_MMQ_KQ=1	2026-10-12 21:04:17 +02:00
Niko1221	0d3553fc8f	Quant yardstick: tools/quant_quality.py (teacher-forced top-1 agreement and KL against Q8_0 on a fixed 5x1000-token set) and tools/quant_formats.py (per-role tensor formats from GGUF headers, local or HF range requests)	2026-10-12 20:48:37 +02:00