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

..

agent.txt

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

code.txt

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

docs.txt

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

english.txt

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

multilingual.txt

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

set.ids.json

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