9 days ago
-
results/hashing: SPIRA Hashing scripts, reports, artefacts (final push) (part 7)
-
results/hashing: SPIRA Hashing scripts, reports, artefacts (final push) (part 6)
-
results/hashing: SPIRA Hashing scripts, reports, artefacts (final push) (part 5)
-
results/hashing: SPIRA Hashing scripts, reports, artefacts (final push) (part 4)
-
results/hashing: SPIRA Hashing scripts, reports, artefacts (final push) (part 3)
-
results/hashing: SPIRA Hashing scripts, reports, artefacts (final push) (part 2)
-
results/hashing: SPIRA Hashing scripts, reports, artefacts (final push)
10 days ago
-
hashing ngram8_overlap_v1c ctx64_cp1: r3 redacted README + 2 scripts, Checker leak audit + push PASS 03:46 (Toronto)
-
hashing ngram8_overlap_v1c ctx64_cp1: 4 JSON (selfcheck sanitized x2, selftest, results), Checker leak audit + push PASS 02:55 (Toronto)
-
hashing v1c_ctl_data: DIAG + gap (DESCRIPTIF, NON MESURABLE scale 0.025764), Checker leak audit + push PASS 01:45 (Toronto)
-
Checker: re-audit v1c_ctl_data r2 PASS + code_prompt_r1 recount (no eval data)
-
Checker: redact prompt fragments (sha256+bytes only) in m8v1c_code_audit
-
Checker: M8 v1c audit code/post/scale recount
11 days ago
-
M8cp1 verdict NON MESURABLE (figé, Checker PASS 19h16)
-
v3_scale/checker: audit verdict M8cp1 — PASS (NON MESURABLE, scale ctl 0.02978 < 0.03) ; recompte db3314e5, agrégats seulement
-
v3_scale/checker: recompte dédup 64 o corpus_diverse_v1 (REPORT, agrégats assainis sans plages ni texte, dedup64_div1.py 5230e995, run.sh) — raw 186812/687/231506 ; norm 149315/454/221174
16 days ago
-
v3_scale/checker: recompte M8cp1 FIGÉ (recount_m8cp1.py sha db3314e5, README, tests synthétiques 25/25, waiter) — aucune sortie, aucun npz/donnée
-
v3_scale/checker: génération old_ref vs bg1_m8 + vérif copie du train (copie ASSAINIE, agrégats seulement : extraits, offsets et texte retirés)
-
v3_scale/checker: re-audit Checker v2_resume try 4 (AUDIT.md + scripts de test et logs, aucun ckpt/npz/donnée)
-
v3_scale/checker: recompte Checker bench_test20k_old bg1_m8 + écart seen-fresh propre (agrégats seulement)
-
v3_scale: verdict v4.2 (segmentation commune figée) bg1_m8 vs old_ref / old_ref v2b / m8_broad14
-
v3_scale: checkpoint bg1_m8 terminé et calibré (0.06328236016662679 OK) + banc v4 à jour
-
Checker: rapport creux MemAvailable 21h05 (fork VM + ballon virtio) + extraits texte (ps assaini)
-
Checker: recompte independant M7sem (verdict OUI confirme) + audit sha, agregats et script seulement
17 days ago
-
Checker: recount M8 v1b ctl (NON MESURABLE regle B) + audit ckpt cov b (PASS avec reserves)
-
v3_scale: audit Checker incident mémoire 17h26-17h29 (blocker_1730 + incident_1740), aucune donnée d'éval
-
v3_scale: Checker — recompte indépendant 64 o brut+normalisé code_prose_v1 (train 91f6950c + sample 8c29bfc4 : 0 partout, PASS) + self-test + watcher + audit des 3 écarts (sans données)
-
v3_scale: Checker — recette code_prose_v1 (code delta CS1 + prose EN/FR Gutenberg pool A, dédup 64 o brut+normalisé) — proposition, sans données
-
v3_scale: Checker — manifeste multi_lang_v1 (58 dépôts, licences permissives, dédup 64 o 0) — PASS (sans données)
-
v3_scale: Checker — gap_bg1d : écart seen (tranche SNN 32, trained_frac>=0,99) − fresh PROPRE pour bg1_m8, bootstrap blocs 512 o B=10000 — code + README (sans données)
-
v3_scale: Checker — dédup 64 o corpus de diversité (multi_lang_v1/v2 PASS, corpus_diverse_v1 FAIL test20k 1405/6596) + préparation banc/gen M8 v1b ctl (non lancé)
-
v3_scale: Checker CS1 r2 — audit verdict_CS1 r2 (tests synthétiques r2) — scripts + agrégats
-
v3_scale: Checker CS1 — recompte 64 o mix (PASS), audit verdict_CS1/PREDECL_CS1 (FAIL 1 point : dénominateur appariement), relecture make_sample_cs1 (PASS) — scripts + agrégats
-
v3_scale: scan 64 o code_scale_v1 (Checker) — agrégats + script, dédup PASS
-
v3_scale: pré-audit Checker r3 verdict M8 v1b (1244 + 1252) — scripts + agrégats json/md seulement (ni npz, ni masques, ni échantillons)
-
v3_scale: banc test20k ancien (old_local_v1/test.bin ce0526c6) strict propre — script, watcher bg1d, agrégats m8_v3/old_ref (m8_broad14 SKIP scale 0,0296)
-
Checker: code_scale_v1 recompte léger indépendant (script + agrégats, add-only)
-
hashing m8_v1b: verdict_M8v1b.py r3B frozen (rule B, sha b3db0875), Checker freeze-audit PASS 13:04 + push PASS 13:07 (Toronto)
-
Checker: gen_quality_metrics.py + watch_gen_quality_bg1.sh (scripts only, add-only)
-
hashing m8_v1b: train_m8v1b.py frozen (sha b2b12c8c), Checker push PASS 12:20; verdict script NOT pushed (r3 pending)
-
SPIRA Checker 12h2x: recompte smoke M8 v1b vtest (3000 ticks, NON MESURABLE, non décisionnel, agrégats seulement)
-
SPIRA Checker 11h4x: recompte indépendant 64 o corpus_diverse_v1b dédup (PASS, agrégats seulement)
-
Add Hashing verdict_SP2 (REJETÉ, Checker PASS 11h41)
-
hashing lever2_sparse verdict_SP1 (k-WTA k=51, REJETÉ): md+json+log.txt, aggregates only (Checker push PASS 11:17)
-
SPIRA Checker 10h5x: recompte indépendant verdict lever2 SP1 (REJETÉ confirmé, agrégats seulement) + verify_1028
-
SPIRA Checker 10h18: seen32 recount cov a (indicatif, test20k brut masqué), review v1b fix1 + ancrage (code 9121fdef), dashboard v4 + ligne seen32 indicative, gen broad_cova 5p (indicatif, scale HORS)
-
SPIRA Hashing: covb report_covb r1 (broad+v3), COPIES r1, verdict_SP r1 — PASS Checker 10h05 (add-only)
-
hashing covb_eval (broad + v3): eval/report/gate scripts + verdict v4.2 wiring (verdict_v42.py byte-identical copy) + lever2_sparse verdict_SP v4.2 (26 files, Checker PASS 09h45)
-
v3_scale: m8_v3 verdict CORRIGÉ v4.2 (apparié, segmentation commune figée) NON RETENU — corrige 60c39893 (taux non appariés) ; banc v4 + ligne m8_v3 ; gén. 5p×200 o fused (indicatif) ; recomptage ; UNK ; checkpoint NON poussé — Checker
-
v3_scale: verdict v4.1 m8_v3 vs old_ref
