3 people 46d4e75314 docs: Gemma 2 investigation — ISWA dual-cache likely root cause, not head_dim ...

Gemma 2 K/V head_dim IS 128 (not 256 as initially assumed). ISWA
(interleaved sliding window attention) creates dual KV caches.
q8_0 works fine. turbo3/turbo4 catastrophic. Likely ISWA interaction.
Quick bench parsing fixed. Llama 70B test pending.

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

2026-03-28 18:30:12 -05:00

..

papers

docs: add KLD nuance to resurrection — q4_0 lower despite worse PPL

2026-03-28 15:49:21 -05:00

threshold-ablation-logs

docs: Gemma 2 investigation — ISWA dual-cache likely root cause, not head_dim

2026-03-28 18:30:12 -05:00

attention-gated-optimizations.md

docs: f16 sparse V experiment — not worth it, reverted

2026-03-27 13:13:35 -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

cross-model-validation.md

docs: Mixtral skipped (bad GGUF), Llama 70B running

2026-03-28 18:04:33 -05:00

decode-speed-hardware-analysis.md

docs: add QC precompute result (#16) + ILP root cause analysis

2026-03-28 18:25:37 -05:00

decode-speed-status.md

docs: Mario PDF test confirms decode scaling — 0.66x q8_0 at 48K context

2026-03-25 22:08:23 -05:00

experiment-context-scaling-fix.md

docs: group-32 rotation failed (PPL 7.06)

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

experiment-layer-adaptive-extended.md

docs: layer-adaptive mode 2 verified 2K-32K — q8_0 quality at 3.5x compression

2026-03-26 00:17:35 -05:00

experiment-layer-adaptive.md

docs: layer-adaptive results — last 8 layers at q8_0 matches q8_0 quality at 3.5x compression

2026-03-25 18:44:37 -05:00

hardware-comparison-matrix.md

docs: update comparison matrix with 10 community testers from X post

2026-03-26 12:03:57 -05:00

kl-divergence-results.md

docs: KL divergence benchmarks — MoE and dense, 3 cache types each

2026-03-27 12:29:07 -05:00

long-context-sparse-v-validation.md

docs: add q4_0 baseline to 50-chunk 32K validation

2026-03-27 21:35:48 -05:00

niah-results-m5max-multikey.json

docs: NIAH multi-key distractor results — 100% both cache types through 32K

2026-03-26 15:33:30 -05:00

niah-results-m5max-multikey.md

docs: NIAH multi-key distractor results — 100% both cache types through 32K

2026-03-26 15:33:30 -05:00

niah-results-m5max-single.json

docs: NIAH v2 single-needle results — M5 Max baseline

2026-03-26 15:30:11 -05:00

niah-results-m5max-single.md

docs: NIAH v2 single-needle results — M5 Max baseline

2026-03-26 15:30:11 -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: add q4_0 baseline to 50-chunk 32K validation

2026-03-27 21:35:48 -05:00

smem-dequant-experiment.md

docs: add QC precompute result (#16) + ILP root cause analysis

2026-03-28 18:25:37 -05:00

sparse-v-upstream-validation.md

docs: turbo4 full characterization — PPL, prefill, decode, NIAH, KLD, PDF

2026-03-28 14:19:35 -05:00

speed-experiments.md

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

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

test-suite-definition.md

docs: final polish + test suite formalized

2026-03-27 23:48:59 -05:00

threshold-ablation.md

docs: threshold ablation sweep (τ=1e-4 to 1e-8) for sparse V paper

2026-03-27 07:07:07 -05:00

turbo4-investigation.md

docs: turbo4 full characterization — PPL, prefill, decode, NIAH, KLD, PDF

2026-03-28 14:19:35 -05:00

turbo4-next-steps.md

docs: turbo4 full characterization — PPL, prefill, decode, NIAH, KLD, PDF

2026-03-28 14:19:35 -05:00

turbo4-profiling-plan.md

docs: turbo4 full characterization — PPL, prefill, decode, NIAH, KLD, PDF

2026-03-28 14:19:35 -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: comprehensive experiment status update — all 2026-03-26 results

2026-03-26 18:25:59 -05:00

upstream-pr-plan.md

fix: address Codex + Gemini review findings

2026-03-26 09:34:29 -05:00