3 people fdedbfe039 docs: variant A (half LUT + float norm) is new best — +8.7% at 48K ...

8 decode experiments tested, 2 winners, 6 dead.
Half LUT + float norm broadcast: 78.7/68.8/39.9 tok/s (short/8K/48K)
vs main: 75.3/59.2/36.7 (+4.5%/+16.2%/+8.7%)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: tturney@psyguard.ai

2026-03-25 22:51:44 -05:00

..

context-scaling-deep-dive.md

docs: extended context scaling verified — 0.99x q8_0 through 32K context

2026-03-25 20:21:22 -05:00

experiment-context-scaling-fix.md

docs: group-32 rotation failed (PPL 7.06)

2026-03-25 19:31:11 -05:00

experiment-decode-fp16-lut-results.md

docs: variant A (half LUT + float norm) is new best — +8.7% at 48K

2026-03-25 22:51:44 -05:00

experiment-decode-speed.md

docs: complete decode investigation — LUT cache thrashing scales with context

2026-03-25 22:01:40 -05:00

pre-rotate-queries-investigation.md

docs: move engineering docs here from llama.cpp fork

2026-03-25 16:16:58 -05:00

pre-rotate-queries-plan.md

docs: move engineering docs here from llama.cpp fork

2026-03-25 16:16:58 -05:00

quality-benchmarks.md

docs: update all docs with q8_0 parity results (2747 tok/s, 4.6x compression)

2026-03-25 18:01:14 -05:00

speed-experiments.md

docs: block-32 breakthrough — q8_0 parity at 4.6x compression

2026-03-25 17:34:27 -05:00

turbo-speed-investigation.md

docs: update all docs with q8_0 parity results (2747 tok/s, 4.6x compression)

2026-03-25 18:01:14 -05:00

turboquant-plus-experiments.md

docs: TurboQuant+ experiments master document — status, findings, dead ends

2026-03-25 20:49:49 -05:00