iomgaa/Video-Tree-TRM5
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Commit Graph

  • 6ac96dff64 Merge branch 'feat/question-gen-v3' main iomgaa 2026-07-17 06:13:44 -04:00
  • 3dcc7d9899 fix: use SLURM_SUBMIT_DIR to locate repo root under sbatch feat/question-gen-v3 iomgaa 2026-07-17 06:13:28 -04:00
  • 7386e5af57 Merge remote-tracking branch 'origin/main' iomgaa 2026-07-17 06:05:42 -04:00
  • 0414948682 merge: question-gen v3 + preflight fixes + continuous concurrent gate iomgaa 2026-07-17 06:02:39 -04:00
  • 9bb09a44d4 feat: add Slurm gpu-sbatch training entry for 4090-server iomgaa 2026-07-17 06:02:39 -04:00
  • b0be1f1ae5 docs: gate speedup final review and delivery record iomgaa 2026-07-17 05:38:44 -04:00
  • 172b7a8ec4 merge: continuous concurrent gate speedup (algo #6 semantic revision, design v3) iomgaa 2026-07-17 05:29:42 -04:00
  • eb12006d38 fix: idempotent ladder_rank migration for legacy gate_evidence tables iomgaa 2026-07-17 05:29:20 -04:00
  • 1930ad32a4 chore: raise Redis cache TTL to 7 days iomgaa 2026-07-17 04:52:34 -04:00
  • 8958eee11b refactor: remove block-sequential gate path and gate_block knob (algo #6) iomgaa 2026-07-17 04:40:14 -04:00
  • 0b839937df fix: predictions row carries arm run_id under shared gate_log; drain evolve gather on failure (algo #6) iomgaa 2026-07-17 03:52:19 -04:00
  • 23a64042fe feat: parallel evolve + continuous gate wiring in runner (algo #6) iomgaa 2026-07-17 03:32:14 -04:00
  • b3aba7c31d fix: escape all LIKE specials in step-row cleanup iomgaa 2026-07-17 01:17:02 -04:00
  • 16993ed362 style: consolidate inference stub, complete docstrings (algo #6) iomgaa 2026-07-17 01:13:19 -04:00
  • ea6bec5421 fix: clear gate-derived rows on step rerun (idempotency) iomgaa 2026-07-17 00:57:53 -04:00
  • 9e8a254fbb fix: gate orchestrator materialize leak + cancel-drain on abort (algo #6) iomgaa 2026-07-17 00:45:47 -04:00
  • e8b66f85ab feat: continuous concurrent gate orchestrator (algo #6) iomgaa 2026-07-17 00:36:01 -04:00
  • c61a6dac84 docs: soften slot fairness claim, complete docstrings (algo #6) iomgaa 2026-07-17 00:28:25 -04:00
  • 30c1cf10c0 fix: gate slot cancel-safety + post-inference freeze discard (algo #6) iomgaa 2026-07-17 00:25:18 -04:00
  • 232afd525b feat: gate unit-arm tasks with question-slot gate (algo #6) iomgaa 2026-07-17 00:13:51 -04:00
  • 0a8e1ad18b docs: complete four-part docstrings for gate prefix section (algo #6) iomgaa 2026-07-17 00:06:35 -04:00
  • 2ab8088c81 test: strengthen frozen-noop full-state assertion (algo #6) iomgaa 2026-07-17 00:03:17 -04:00
  • 1e92928d4e fix: lock all-INFRA contract, idempotency tests, tuple units (algo #6) iomgaa 2026-07-17 00:00:56 -04:00
  • 21c360bc87 feat: gate prefix-ordered consumption core (algo #6) iomgaa 2026-07-16 23:44:32 -04:00
  • b02db74237 docs: apply Codex plan review fixes (tail-INFRA rejudge, slot deadlock guard) iomgaa 2026-07-16 23:30:56 -04:00
  • 85ccad3b72 docs: implementation plan for continuous concurrent gate iomgaa 2026-07-16 23:12:53 -04:00
  • 0e8069af27 docs: gate speedup v3 - ladder-prefix consumption fixes arrival-order bias (Codex round 2) iomgaa 2026-07-16 22:46:01 -04:00
  • 53a8c4be02 docs: redesign gate speedup v2 (continuous concurrent gate, drop preload) iomgaa 2026-07-16 22:24:28 -04:00
  • 269e0d294c docs: design for gate speedup (preload BaselineCache + parallel arms) iomgaa 2026-07-16 14:23:32 -04:00
  • cc01d5ed62 fix: address Codex review of telemetry fix (mkdir degrade, close, degrade tests) iomgaa 2026-07-16 08:53:50 -04:00
  • 065c8ae1b9 fix: telemetry single persistent connection + lock (kill concurrent write lock) iomgaa 2026-07-16 08:47:23 -04:00
  • 766592d855 docs: design for telemetry SQLite concurrency fix (align HarnessLog pattern) iomgaa 2026-07-16 08:47:23 -04:00
  • 933de6251e chore: add FORCE flag to split script; align train config pool sizes to 0.4 split iomgaa 2026-07-16 07:46:31 -04:00
  • 5fdcd9b5aa feat: add Video-MME 900 training entry (config + self-contained script) iomgaa 2026-07-16 07:33:16 -04:00
  • a1c4e051e7 Merge branch 'feat/preflight-train-fixes': preflight training fixes (4 WPs, 21+ defects) iomgaa 2026-07-16 07:25:50 -04:00
  • 6a9b1ea188 docs: record preflight final review + integration fixes (I-3/I-4/M-2) iomgaa 2026-07-16 07:21:10 -04:00
  • 6911c83d66 refactor: share INFRA stop-reason set from core.evolution iomgaa 2026-07-16 07:18:14 -04:00
  • b3ba11c7a5 fix: count gate INFRA guard numerator by unit not record iomgaa 2026-07-16 07:16:33 -04:00
  • ee69721ea3 fix: filter frozen pools by requested task_types subset iomgaa 2026-07-16 07:13:31 -04:00
  • e1f08dcd3b fix: count units not questions in trainability pre-flight; fail-fast when all types filtered iomgaa 2026-07-16 06:45:16 -04:00
  • ef244c52bd chore: add trainable_min_units to train_ar30 config iomgaa 2026-07-16 06:35:10 -04:00
  • ed59c6091a test: accept cache_salt kwarg in AgentLoop mock LLM stub iomgaa 2026-07-16 06:34:16 -04:00
  • 9be3bdb8eb feat: inject run_id as cache_salt for per-epoch resampling (algo #10) iomgaa 2026-07-16 06:33:09 -04:00
  • 58a0203522 perf: dedup holdout four-way eval (baseline derive, best_hard memo) iomgaa 2026-07-16 06:31:00 -04:00
  • 77fd35830c fix: idempotent _run_step (DELETE stale) + checkpoint after gate save iomgaa 2026-07-16 06:27:38 -04:00
  • efbdeb1647 fix: route None/degraded diagnoses to lapse; abort on high degrade rate iomgaa 2026-07-16 06:24:11 -04:00
  • 65126feada fix: patch checks full target span + marker injection/integrity (algo #8) iomgaa 2026-07-16 06:20:29 -04:00
  • caea9150b3 feat: pre-flight filter of untrainable task types before gate iomgaa 2026-07-16 06:16:02 -04:00
  • c44f6010eb fix: early_stop patience counts epochs not steps iomgaa 2026-07-16 06:09:11 -04:00
  • 25918a73ff fix: normalize non-scalar prediction; harden predictions insert iomgaa 2026-07-16 06:06:58 -04:00
  • 7d02cded99 fix: wrap diagnosis run_log with StepsJsonRunLog (restore algo #7 traces) iomgaa 2026-07-16 06:05:55 -04:00
  • e1cac644c8 docs: address Codex review of WP3 plan (pools passthrough, holdout memo, DELETE table) iomgaa 2026-07-16 06:03:17 -04:00
  • 03337af8f8 fix: gate INFRA isolation edge cases (all-INFRA fail-loud, parse_error in guard) iomgaa 2026-07-16 05:54:02 -04:00
  • b307f51340 feat: --retry-uncertain re-diagnoses uncertain rows; fix docstring iomgaa 2026-07-16 05:41:19 -04:00
  • c68de849da fix: slow-update R2 dual_metric uses slow_candidate kind (not final) iomgaa 2026-07-16 05:38:05 -04:00
  • 1468a53b7a fix: read-only baseline queries skip _runs upsert (register_run flag) iomgaa 2026-07-16 05:36:47 -04:00
  • 96884dd149 fix: atomic writes for manifest/record_run/update_best (tmp+replace) iomgaa 2026-07-16 05:34:45 -04:00
  • d1516bf56b fix: isolate gate baseline-arm INFRA errors from BaselineCache (algo #6) iomgaa 2026-07-16 05:33:21 -04:00
  • 87908b23eb fix: half-open circuit admits single probe (no thundering herd) iomgaa 2026-07-16 05:27:46 -04:00
  • 1cbfa97b8d fix: cover httpx disconnect family in transient error set iomgaa 2026-07-16 05:26:52 -04:00
  • 200053fbdc fix: treat SSE stream without [DONE] as truncated (retry, no cache) iomgaa 2026-07-16 05:26:21 -04:00
  • 7cc6aa5b23 fix: reject REDIS_CACHE_TTL<=0 (kill implicit never-expire) iomgaa 2026-07-16 05:25:36 -04:00
  • 564c92a043 feat: add cache_salt dimension to LLM response cache iomgaa 2026-07-16 05:23:58 -04:00
  • c5c11ea81b docs: address Codex review of WP4 plan (salt key compat, store/protocol, INFRA path) iomgaa 2026-07-16 05:21:18 -04:00
  • 04bda25335 docs: update WP4 plan workspace anchors after WP2 line drift iomgaa 2026-07-16 05:14:12 -04:00
  • 8f349c5c62 fix: make frozen product backup unique and atomic on --force iomgaa 2026-07-16 05:12:08 -04:00
  • a0c7e043e8 fix: validate global frozen pools baseline_run_id + sha256 on load iomgaa 2026-07-16 05:05:15 -04:00
  • 4d60a545f6 feat: seed carries frozen pools.json into training workspace iomgaa 2026-07-16 05:02:16 -04:00
  • a39846af6e feat: guard frozen split products against silent overwrite (--force) iomgaa 2026-07-16 05:00:42 -04:00
  • 5bb8319220 feat: tier-aware diag/val split with val-power repair (design 5.1) iomgaa 2026-07-16 04:58:41 -04:00
  • 6432ffef9a feat: add dedupe_per_question to extract_run_db (canonical 902->900) iomgaa 2026-07-16 04:54:19 -04:00
  • 06c575f0a4 chore: bump video_split val_ratio 0.3->0.4 for terminal-eval power iomgaa 2026-07-16 04:53:14 -04:00
  • 4fd8a6bfef docs: fix WP2 plan GeneratedQuestion import + NULL qid note (Codex review) iomgaa 2026-07-16 04:52:31 -04:00
  • 232a56e826 fix: fail-loud on missing evolve/diagnose templates (no silent empty) iomgaa 2026-07-16 04:43:22 -04:00
  • 1108235818 refactor: drop dead prompt fields consolidate_system/span_eval_user iomgaa 2026-07-16 04:42:26 -04:00
  • d3639a94b8 feat: migrate 5 evolve/momentum templates from TRM4 (algo #8) iomgaa 2026-07-16 04:41:02 -04:00
  • 6cdf05d56b docs: address Codex review of WP1 plan (real loader tests, red-green order) iomgaa 2026-07-16 04:38:51 -04:00
  • 39a9210e4d docs: add 4 work-package implementation plans for preflight fixes iomgaa 2026-07-16 04:32:47 -04:00
  • a4c429b247 chore: snapshot in-progress question-gen work before preflight fixes iomgaa 2026-07-16 04:12:21 -04:00
  • 11a5545f57 docs: address Codex design review (coverage matrix, salt/gate pairing, tier termination) iomgaa 2026-07-16 03:52:50 -04:00
  • d7e965639e docs: add preflight training-fixes design (23 bugs + wiring + split tuning) iomgaa 2026-07-16 03:44:31 -04:00
  • 57f590b57e test: integration for diagnosis tree_data link + core dep direction iomgaa 2026-07-15 23:05:27 -04:00
  • 224bb25bb5 fix: load real tree_data for training-loop diagnosis (runner) iomgaa 2026-07-15 22:53:34 -04:00
  • 8dedd082ff fix: self-contained tree-inject test + fail-loud on missing question (review) iomgaa 2026-07-15 22:48:37 -04:00
  • 83056688cf fix: load real tree_data for offline diagnosis (video_split_cli) iomgaa 2026-07-15 22:38:23 -04:00
  • f1b6865861 fix: include question_id in diagnosis tree-coverage fail-loud message iomgaa 2026-07-15 22:33:23 -04:00
  • c5fff7f8b3 fix: fail loud when diagnosis video tree not covered (algo #7 input) iomgaa 2026-07-15 22:25:30 -04:00
  • d6c595c4a4 fix: fail-loud on node without time info; anchor test store path iomgaa 2026-07-15 22:16:39 -04:00
  • 05294412df test: cover missing-id and non-list roots; drop unreachable empty-nodes guard iomgaa 2026-07-15 22:11:36 -04:00
  • e6adaad8b3 feat: add tree.json flattener for diagnosis ground_truth iomgaa 2026-07-15 22:05:50 -04:00
  • 748a969e0e docs: implementation plan for diagnosis tree_data fix iomgaa 2026-07-15 22:01:20 -04:00
  • 63dfee1b9b docs: design for diagnosis tree_data broken-link fix (TRM4 regression) iomgaa 2026-07-15 21:44:42 -04:00
  • 266a1ab4bc fix: use conda run --no-capture-output for unbuffered real-time logs iomgaa 2026-07-15 14:02:13 -04:00
  • 94728c5220 chore: set video-split diagnosis concurrency to 12 (overridable) iomgaa 2026-07-15 13:57:59 -04:00
  • 26ab536fea chore: migrate diagnosis prompts from TRM4 (rename to TRM5 loader filenames) iomgaa 2026-07-15 13:53:16 -04:00
  • 8fef7ced42 fix: address whole-impl review (INFRA T0 rows, reproducible manifest, evolution_target report, dead config, canonical DRY) iomgaa 2026-07-15 13:39:14 -04:00
  • 02b8145b7f fix: fail loud on diag model fingerprint drift vs .env SEARCH_LLM_MODEL iomgaa 2026-07-15 13:20:51 -04:00
  • 2844732126 refactor: self-contained two-phase video-split CLI entry iomgaa 2026-07-15 13:12:21 -04:00
  • 6a21d80313 feat: add video-split config knobs and reproducible script iomgaa 2026-07-15 12:54:25 -04:00
  • 43d7346526 feat: wire end-to-end results-driven split with defensive asserts iomgaa 2026-07-15 12:43:25 -04:00