Files
MeltMind/AGENTS.md
T

65 lines
21 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# MeltMind project collaboration
- Before moving on from a major result or entering a new stage, follow `docs/collaboration-stage-review-policy.md`: consolidate evidence across the team, revisit the user's research goal and prior work, discuss disagreements, and record the next-stage decision.
- Continue the existing shared design and its fixed commits. Check for an existing draft or running task before starting another. Claude integrates and reviews; MeltAstra supplies execution evidence; MoldAstra supplies mathematical/grouping boundaries; Grok performs only needed targeted checks.
- Keep topology/local-type partitioning and true connectivity as independent open hypotheses. A failed input-coverage mechanism is not a failure of all partition or topology routes.
- Design/document review and information sharing do not authorize new scientific computation or training. Carry out the scope already authorized; record explicit scope and budgets before expanding it.
- The user has authorized project information sharing in their controlled Drive, Git and Slack spaces; see `docs/sharing-authorization-2026-11-14.md`. Do not ask for the same sharing authorization again. Git identity discussion is not a priority.
- Preserve original data, failures, review limitations and immutable evidence. Do not label proposals as executed or reviewed results.
- Current E1 scope follows the standing continue-work authorization relayed in Slack `1794651141.035699` and Claude caps `1794651169.320839`: bounded CPU-only label-free preparation and synthetic checks may proceed after the count checkpoint, without another user approval loop. Total 16 CPU-hours / 25 GiB, at most 8 processes, per-job 30 min / 1.25 GiB tree RSS, at most 2000 registration pairs; the 10-part stage additionally stays sequential within 3 h / 5 GiB. Never read labels, fit models, use GPU, run pilot/full matrix, or exceed caps in this scope. See `docs/s2-e1-budget-checkpoint-2026-11-14.md`.
- Post-witness update (2026-11-14 21:12) supersedes treating the preceding E1-only scope as the entire remaining project scope. Follow the latest section of `docs/collaboration-stage-review-policy.md` and Claude Slack `1794661944.308269`: custody join, label-free copy calibration, k=3 per-route candidate count, bounded registration (up to 4000 pairs / 100 worker CPU-h / 16 workers), band review, component freeze, then coordinated S2 continuation. No repeated approval gates or reruns of completed review; preserve concrete execution limits and distinguish future stages from completed work. User preference: tens of CPU-hours alone are not a reason to block progress. No additional scientific task is implied by this preference.
- S2 execution update (2026-11-15): components/splits and the fixed ten-part measurement are complete; ten-part 10/10 PASS is in `docs/s2-ten-part-results-2026-11-15.md`. The 499 non-test preparation queue has started. Read `docs/s2-pilot-live-handoff-2026-11-15.md` and existing receipts before any launch; never restart from chat idle. Claude accepted implementation fixes at `1794683016.134929`, accepted the frozen ID-order dispatch at `1794683049.057259`, and raised the first-50 continuation ruling to 80 worker CPU-h / 100 GiB at `1794682712.505709`. The running original supervisor still enforces its pinned 60-hour cap; any required continuation may contain only never-attempted parts and must preserve cumulative accounting. After all 499 terminal attempts and required evidence, the authorized single-GPU pilot includes an E256 timing probe and four runs within 2 GPU-hours total, per-node train/validation extrapolation and a frozen common epoch limit. Keep 125 test labels sealed; no repeated user-go or main-review round is required for these accepted fixes. This does not authorize the full training matrix.
- Infrastructure-recovery update (2026-11-15 05:16 UTC+8) supersedes the original live sessions above: original preparation16682 exited INFRA_ABORTED and idle waiter13519 is retired; never restart either or manufacture original batch completion. Active recovery82149 and chain42022 use `artifacts/s2-pilot-recovery-inputs-2026-11-15` and `artifacts/s2-pilot-recovery-continuation-2026-11-15`. Read the leading section of `docs/s2-pilot-live-handoff-2026-11-15.md`. Fixed134 =16 explicitly authorized infrastructure recoveries +118 never attempted; original362 PASS and3 real timeouts are retained. Cumulative60 worker CPU-h/100GiB includes old aborted CPU once; no80h extension, no new first50 barrier or automatic retry. Preserve active source pins. Three-source verification precedes the existing bounded probe/four-run pilot; no main matrix.
- Terminal update (2026-11-15 06:14 UTC+8): recovery82149 and chain42022 have finished. Recovery134=131 PASS+3 wall_limit; canonical499=493 PASS+6 wall_limit. E256 probe passed; N128-seed0 failed at epoch10 with nonfinite encoder output. Pilot stopped without retry/width selection after296.023387s of7200s; other three runs never started. Read `docs/s2-recovery-pilot-stage-result-2026-11-15.md`; preserve all evidence,125-test seal,and source pins. Do not restart from stale progress. Claude received stop notice1794694508.930959; next changes require the existing collaboration review, not a repeat sharing approval.
- Precision-v2 terminal correction (2026-11-15 06:30 UTC+8): diagnostic41846 completed and reproduced fp16 slice overflow; fixed8a548cc preserves original sources and implements the approved shared FP32 aggregation. New v2 session26792 has already exited: probe PASS at187345 nodes, VRAM12586057728B, E_max42, but SELECTED_GPU_BUSY_BEFORE_RUN stopped before any training run (runs=[]), total11.143491s. Earlier inference from pilot-started to N128 running was incorrect and retracted. Do not restart or rerun probe without the engineering continuation ruling; read latest handoff and raw complete receipts.125 tests remain sealed.
- Latest engineering continuation (2026-11-15): Claude1794695421.177269 authorized reuse of v2 probe after its zero-run infrastructure stop. Fixedea5ac05; sole active session34632/root `artifacts/s2-pilot-v2-continuation-2026-11-15`, entryPID209380 and N128 workerPID209415 verified running. Three idle polls passed over20.285703s, sameGPU. ReuseE_max42/probe; original11.143491s plus all new waits/setup/runs share7200s. Keep v2 zero-run stop and original numerical failure immutable. Inspect real worker/start/progress/terminal receipts before any action; never duplicate or modify pinned code.125 tests remain sealed.
- Latest terminal update (2026-11-15 07:14:59 UTC+8): session34632 and all pilot workers have exited. N128/E128 completed42 epochs with guardPASS; N256 failed at epoch2 on4/382 AMP skips exceeding1%, E256 never started. PILOT_INCOMPLETE/no width; cumulative2439.912972s/7200. N256 history/progress also records reserved16416505856B above15GiB; the failed-result path did not inspect recorded peaks. Preserve this resource-boundary gap. Read `docs/s2-v2-pilot-stop-results-2026-11-15.md`; no restart, reprobe, threshold/source change or full matrix is authorized by a polling request.125 tests sealed; evidence submitted toClaude.
- Claude1794698179.161599 records a separate feasibility decision: main-matrix width128, width effect not tested, no use of validation scores to justify selection. Original pilot selected_global_width staysnull. Before any main-matrix execution: fix recorded progress/result VRAM-peak enforcement with synthetic test, allocator configuration per ruling, unchangedAMP1% andE_max42, and budget/execution plan. This read-only stage audit has not implemented or launched that work.
- Main engineering plan update: independent `s2_main_memory_guard.py` checks all progress/result/history peak records during and after failed workers;7 CPU synthetic checks passed. No old pinned source changed. Allocatorexpandable_segments parsedtrue withV100 VMM support, no GPU tensors/workload. Claude1794698734.967459 supersedes pilot-cell reuse: separately prepare6 timeouts at60min and125 fold0 test geometries without label decoding, freeze one final roster, then all90 neural fits fresh; pilot weights remain historical only. Read `docs/s2-main-engineering-and-budget-2026-11-15.md`. Proposed costs/caps are plans; no131 queue or main matrix launched in this engineering turn.
- Latest actual launch (2026-11-15 07:51 UTC+8): Claude1794699027.366469 approved budgets/execution. Sole active131 preparation is session32694/coordinatorPID220180/root `artifacts/s2-main-preparation-2026-11-15`, fixedimplementationbf5203c, contract0a3c628c81c008eab53e9337406b84ec2063edcd49529bcc7a82e3e4a3d719a3. Host verified16guards+16geometryworkers. Do not restart or mutate pinned sources. 493successes reuse only; all131 label-free;6×60min+125×30min/1.25GiB;16workers;new20workerCPU-h with old34.75965h separate; cumulative100GiB.10syntheticchecksPASS. Automatic roster freeze only after131terminal withoutglobalstop. Then90fresh width128/Emax42 with46GPUh includingidle/setup,30min/run,up to4V100,N/Eadjacent,first8projection;R2singlefit30CPUmin,total100CPUh,leaveGPUfeedersfree. Currentmatrix/R2notstarted;all90+15terminal/checkpointsfrozen beforeunifiedcustodypredictions/scoring. Read leading handoff and actual-start-audit; older pending/running entries are historical.
- Latest terminal (09:25): preparation32694/PID220180 exited,131=124PASS+7wall_limit; no globalstop,monitor_errors0. Final617 roster SHA fb0c984f2e553343a598ce2e3c7012d3017053dc0fef58b5aaf46dd3c67a207e automatically frozen;7globallyexcluded,493oldsuccesses reused. Newworker13.8436103492h/20h; cumulativehistory48.6032611436h. Do not restart preparation. Read final-roster/roster-receipt/terminal-audit/final-fold-pool-counts in main-preparation root. Matrix implementation and synthetic acceptance now proceeding under Claude1794699027; no training launch yet.
- Latest matrix launch (09:36,09:37 actual worker evidence): sole session92182/coordinatorPID223006/root artifacts/s2-main-matrix-2026-11-15; implementatione3fa0af, contract198e5711ed2e7485167068769652c866867229e09ecf0d09b21612debc3b7a56,617rosterfb0c984f. Four V100 training workers223127/223141/223131/223145 observed epoch1; initialN f0pool25seed0/1/2 and f0pool50seed0, each followed byE.17syntheticchecksPASS. First4blocks/8runs barrier and46GPUhprojection automatic; allidle/setup/failure charged,30min/run,no retries,pins immutable. CPUaffinity0-7 GPUfeeders;8-31availableforR2. R2 notyetimplemented/started; all90+15terminal/checkpointsfrozen beforecustody. Check actual-start-audit and live receipts; do not duplicate matrix or restart completed131 preparation.
- Latest R2 launch: implementationf560a3d;10newtestsPASS; soleR2session51525/coordinatorPID237829/root artifacts/s2-r2-main-2026-11-15; contractedec86197769f44bf98f234795950f36c9fa42084ccb3c68448b63b440aa126e. FirstHPOguardPASS worker21.929337CPU-s/fit7.789711311CPU-s; first-fitprojection18685.543199s<=360000 permits automatic4worker continuation onCPU8/10/12/14,coord30-31. Claude1794707190: finalrefit samelearningonly; failedHPOconfigsNA,choosebestfinished,fixedgridtie; all12fail orfinalfail=>cellNA. Max180HPO+15final,1800CPUseconds/fit,total100workerCPUh,no retries. Neural92182/PID223006 stillactive,sourcepinsunchanged; first8 gatepending as oflaunch. Do not duplicate either queue or restart preparation. Bothallterminal/checkpointsfrozen beforecustody/testpredictions/scoring. Read livehandoff and R2actual-start-audit.
- Latest first-eight milestone (10:04): same neural92182/PID223006 first8 all42epochs/PASS,4pairs valid, checkpoints hashed; automatic budget checkpointpassed, projected83657.259515GPU-s=23.238128h<=46h, chargedfirst8 4813.381680s. Existingqueue continuedblocks4-7; no newlaunch. R2same51525/PID237829 had48HPO PASS/4active withnoerrors. Read first-eight-audit/checkpoint and liveprogress, retainall90+15terminal custodygate.
- Latest custody launch (16:13): training all terminal frozen6903c07 (89neuralPASS/1NA,15R2finalPASS). Claude1794730198 main-review conditions closed in522ee84;14testsPASS/617columnmetadataPASS. Sole prediction session42606/coordinator455501/worker455512, root artifacts/s2-custody-evaluation-2026-11-15, contracte8b081b31d0f8975da1c2f1e92843181ee788979d98e61d2a8571e5e5d94850c. Firstfold3/121 observed, GPU0active; others wait firstfoldcost. No labels decoded. Follow leading handoff; never restart oldtraining or duplicate prediction. All predictions hashed then scoring code/config Git freeze before custody. Approved4GPUh/45minfold/15GiBVRAM/16GiBRSS/20GiBfold, custody10CPUh/3h/16GiB,100GiBtotal,0retries.
- Latest custody COMPLETE (2026-11-15,终态核验): prediction42606 and custody87444 exited0, allguardsPASS.13449predictions frozen+codelocked c47dd6a before one decode of617parts; results1dea89d/report docs/s2-custody-results-2026-11-15.md. Main100pct209components/617parts: Delta1-E/N=-0.01292367747,95CI[-0.02122332192,-0.00563340237],harmful;R2bestprimaryerror,N usefulness vsR2 fails. Allnewmetricsvalid,undefinedbootstrap0. Prediction410.070561GPU-s,custody72.962534CPU-s. Do not restart/redecode/retrain/change frozen stats; all original NA preserved. Next team stage review ledClaude; no O1/organisation inference from this result. Read newest handoff, older running entries historical.
- Latest posthoc D1-D4 complete (2026-11-15): S2 acceptedClaude1794731320, stage47259fee/correctiond2d69cac. Authorized read-only diagnostics fixeddaed8e873afa7bf56fd3a58f6701f78ce566c5b0; report docs/s2-posthoc-diagnostics-2026-11-15.md; completion.json is latest. D1242.573114CPU-s,D317.133535CPU-s,13449D3rows,0undefined,2lowpoolE signed-oracle reversals retained; no newlabels/model/training/prediction change. S3-A/B not launched; require stage discussion and explicit execution contract. Do not replay earlier queues or treat oracle as deployable calibration.
- Latest S3-A preparation (2026-11-15): Claude1794732043 approved direction, requires one main review before running. Contract263b0b42b832d958ccf5843215403d629cf9d037084951d10f2af40bb89348d8 at artifacts/s3a-planning-2026-11-15;17syntheticchecksPASS. NewR2_E4/R2_G2,360HPO+30final planned, old15R2fits/predictions reused. CPU20h including18hfit+2hpostfit,0retry,noGPU,newdecode0. Scoring/data/budget helpers frozen; actual queue implementation and launch pending main review. Mold1794732176 final D3 read accepted, no recomputation; H not pure ordering, D2 cannot decide longer training success/failure. S3-B not launched/O1open. Do not restart S2.
- Latest S3-A actual launch (2026-11-15 17:01 UTC+8): Claude1794732829.567629 approved implementation+launch, no further main review before results. Sole session5217/coordinatorPID458991/root artifacts/s3a-execution-2026-11-15; implementation6142ad6/contract263b0b42 unchanged. FirstE4job0+G2job12 PASS47.590461CPU-s; projection43360.732672s=12.044648h<=18h, then4guards+4workers CPU8/10/12/14 verified. Original15R2 no refit;390maxnewfits,20workerCPUh incl2hpostfit,0retry,noGPU/newdecode. 8newqueuechecksPASS. Read actual-start-audit/first-fit-cost-check/liveprogress; do not duplicate/restart/mutate sources. Queue stops at all-fit terminal freeze; predictions/scoring not yet started. Parallel stage/literature review does not expand compute scope.
- Latest naive-ordering launch (2026-11-15 17:23 UTC+8): Claude1794733395 permits one617-full-node Fill_time custody;1794733619 corrected Dw fallback alreadyB0, c/ties/alpha accepted, fixed contract then run with no further review. Sole49189/PID460410/root artifacts/naive-ordering-2026-11-15, d2adb6b/contractc49313e.13checksPASS. FirstpartPASS1.148115CPU-s/projection885.196665s<=8h,4workers16/18/20/22;10workerCPUh incl2hpost,0retry/noGPU/no newmodels. S3-A5217/PID458991 unchanged, workers8/10/12/14, sourcepinsverified; neverduplicate/restart either. Full617PASS before clocks/maps/score; originalVTU/Tet/querymax distinct, strictCDF withties, PPSmultiplicityonce/alphaestimate, eikonalnot tested. Checklive receipts and latesthandoff.
- Latest naive and LOG terminal update (2026-11-15): Naive ordering complete: original610PASS+7failures retained, seven cache-only continuations PASS; canonical617,23938maps frozen,score PASS. Worker cumulative836.697435206s/36000; no new fits or decodes in continuation. Original49189,continuation2411,map63422,score10304 all exited; do not rerun. Results9b7121a/docs/naive-ordering-results-2026-11-15.md. LOG92800 complete624/624,411within1pp,611zero-printed-flow undefined,13central-window printed-flow within5%; see docs/log-time-map-results-2026-11-15.md. Both retain complete raw evidence. S3-A original5217/PID458991 still active: 228 receipts/0 failures,progress228,4active,sourcepinsPASS,worker12282.190599s/64800fit cap;20h includes2hpostfit. No terminal freeze, prediction/scoring not started. Do not restart or modify running pins; after all390terminal follow existing approved postfit freeze/predict/score scope. Earlier two-queue running entries are historical.
- Latest S3-A terminal correction (2026-11-15): S3-A original5217/PID458991 has now exited1: S3A_STOPPED_INCOMPLETE/S3A_COORDINATOR_INTERRUPTED.235attempted=231PASS+4interrupted(230,232,233,234;returncode-9),155never-attempted235..389,all30finalnotrun. SourcepinsPASS,no live queue/fit processes,accounting_complete true,worker12507.336661CPU-s; original failures and attempts immutable. Signal origin unknown; handler merges SIGINT/SIGTERM, do not call this numerical failure. No restart/retry authorized by polling; Claude notified1794736881.943279 for infrastructure continuation ruling. Read artifacts/s3a-interruption-audit-2026-11-15/audit.json. Naive/LOG complete; prior S3-A running snapshots are historical.
- Latest S3-A infrastructure continuation actual launch (2026-11-15): S3-A authorized159 continuation now active: session7921/PID464772/root artifacts/s3a-recovery-2026-11-15, orchestration259d1d685bdced1920eb3d483019cd18fedc1bf5 (local fixed; remote publication queued behind087f63b full raw upload). Scientific contract263b0b42 and original6142ad6 workers unchanged. Claude1794736905 explicit4infra recoveries230/232/233/234+155never attempted; reuse231PASS, prior12507.336661CPU-s counted once within20h/18hfit+2hpostfit. Six scope/accounting/barrier testsPASS, actual sourcepinsPASS and4guards+4workers observed; old458991 absent. All360HPO terminal then30final,390canonical thenprediction/scoring. Original partial failures immutable; do not duplicate/restart or mutate either pinned implementation. Read fresh receipts, recovery-scope/code-freeze and actual-start-audit; older stopped/running entries are historical.
- Latest S3-A/naive handoff (2026-11-15): Sole S3-A recovery7921/PID464772 remains active: 287 canonical terminal (56 new receipts, 0 new failures), prior231PASS reused; cumulative15954.571443workerCPU-s includes old12507.336661 once. Source259d1d6, old scientific6142ad6/contract263b0b42 unchanged. Four explicit recovered jobs230/232/233/234 allPASS. Sole postfit chain81572/PID465797 waits at WAIT_EXISTING_RECOVERY; fixed58e4e4c,4newsyntheticchecksPASS. It starts nofits, waits390complete/workerexit,Gitfreezes models,predicts once,Gitfreezes exactmanifest,scores using original frozen scorer within cumulative20h and postfit2h/combinedwall3h; failure stops without retry. Original full2.2GB publication58603/PID464160 still uploading087f63b,followup91597 waits then publishes fixed4d76d35 afterremoteverification. Do not duplicate uploads or queues. Latest head may be locally fixed before remote publication; do not claim remote completion. SmallreviewZIPF0CHEAFDB4P is shared and received/reviewed byMold/Claude. Claude1794737961 withdrew Spearman request; no correlations/bootstrap were run. Read artifacts/s3a-current-handoff-2026-11-15.json and real receipts.
- Latest final-fit process-count policy and actual waiter replacement (2026-11-15): Claude1794738923.689619 final-fit process_count_limit policy is enforced by sole replacement postfit45375/PID466888/root artifacts/s3a-postfit-final-policy-chain-2026-11-15, fixed/pushedcb786589448447be1fba86baed9020b8b6c52456,4testsPASS. Old idle81572/PID465797 intentionally retired(exit143), no children or predictions at retirement; never restart it. Recovery7921/PID464772 unchanged, sourcesPASS; 364canonical terminal at audit. If anyFINAL process_count_limit occurs, new chain writes awaiting-final-infra-ruling and stops before checkpoint adoption/prediction/scoring; no auto recovery, raw worker/batch NA is not adopted as scientific cellNA. HPO323 remains unavailable config11, other11cell26configurations actualPASS with result/modelhash verified. Continue existing360HPO→30final→390freeze workflow under original budget when no final-policy hold. Existing raw087f63b/small4d76d35 publication complete; do not duplicate uploads. New final-result publication expects remote basecb786589.
- Latest S3-A terminal: 390logical terminal,359HPOPASS+1configNA+30finalPASS; sole45375 exited0 and result1e49c1572227116e821d3cd414ebcecd7c430753 published with independent remote readback. No restart/retry/re-score. See leading handoff and docs/s3a-results-2026-11-15.md. Mold1794740467 actual10JSON review and Claude1794740503 acceptance recorded. S3-B(a) remains a proposed <=10workerCPUh CPU-only/no-learning contract subject to one review; no automatic launch, no S3-B(b)/O1 training authorization.
- Latest contract-only S3-B(a) update: see docs/s3b-a-contract-2026-11-15.md. Claude1794740834 fixes both paired-CI gates, total-volume denominator and exact midpoint rule; Mold1794740985 clarifies zero wall edges/UNKNOWN failures. Fixed-ten cost checkpoint and possible20workerCPUh cap supersede early10h-only proposal, but contract awaits single main review and this task authorizes documents only. No implementation, tests or real workload launched. S3-A report delivered bd2473c with original results1e49c157 independently HTTP/hash verified.