Home › Evidence › Records › cfg-0147
⚠ This notebook has stopped taking notes: newest record is 23 days old (eng-0280), against an expected cadence of 14 days.

cfg-0147

backfilledfork
citable URL: https://halobench.com/records/cfg-0147/ — this address never moves; the anchor /records/#cfg-0147 keeps resolving
modelQwen3.8-27B-Q8_0.gguf · Q8_0 · rev 57484e3196aaff8dbbd71c666158b00206baa2bc134e4f16ae90fc8cbdfbeea2
engineggml-org/llama.cpp 2586f6eddae19bef3dd21a1a0b109cca7bcf1c32 · Vulkan0/Strix-Halo · host aihydra (igpu)
flagsHO-004-DIAG v0.6.10 stock-MTP Stage-A sentinel fingerprint: identical target/runtime/backend/KV/batch/load-mode/-c 32768/--temp 0/--seed 4242 as cfg-0146 plus the native draft-mtp path with --spec-type draft-mtp --spec-draft-n-max 3 (HARD ceiling 3; n_max>=4 corrupts generations on this family, clm-0055). No DFlash2 drafter, no ngram/proxy/llama-swap/ROCm, no n_max>=4, no KV quant, no alternate target quant.
templatenot recorded at test time
treefork — carrying con-0002, con-0009
aged evidence — reconstructed from the archive. Recorded from the runner command line at review time. Sentinal outcome: stock-MTP n_max=3 PASS Stage-A (empty-assistant 0, empty-tool-call 0, single-token-EOS 0, code-valid 3/3, tool-correct 3/3) plus the 4/4 house guard, with log-backed draft activation (draft acceptance 0.8218/0.9531). This config is ONLY a Stage-A sentinel/gate record: no throughput, no Stage-B production-depth performance, and no MTP safety claim beyond n_max=3.

Capability basis

measured on this config

Runs on this config (1)

rundatekindsuitemetrics
run-05732026-08-23guardstageA-c32768-varied-promptstasks_passed=4 · tasks_total=4

Cited by — computed at build time, never stored

claims clm-0109