Prepare amended phase3 item1 six-fit entry with complete schedules and120000GPU-s cap

This commit is contained in:
MeltAstra committed 2026-11-18 04:55:33 +08:00
1 parent 30c34b0e66
commit ae089067a1
23 files changed
+3129 -1

No files matched your search

@@ -3716,9 +3716,17 @@
"http_readback": "4/4PASS"
},
"phase3_item1": {
"status": "PREPARED_EXACT_REVIEW_PENDING_NOT_STARTED",
"status": "SUPERSEDED_UNSTARTED_BY_V2",
"root": "/home/frank14f/MeltMind-runs/phase3-data-sufficiency-2026-11-18",
"implementation": "c8e2ee4ca3d08cf0080ac18a1d17bfb42fb1b8a6",
"contract_sha256": "a84e5e8269269951a2abbde89f8497c6918222cb8cfe7dd240300d7eba6a656f"
},
"phase3_item1_v2": {
"status": "PREPARED_NEW_REVIEW_PENDING",
"root": "/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18",
"implementation": "73f94bbc77e5ed3156ec85ff4a32ec2c834fdb07",
"contract_sha256": "1bdf2331c5127e03bec160835eda9d04ff333c3f70d5fe12a0629ffa82ec5826",
"item_cap": 120000,
"phase3_cap": 612000
}
}
@@ -0,0 +1,56 @@
{
"authority": "1794948643.973609",
"old_entry": "c8e2ee4ca3d08cf0080ac18a1d17bfb42fb1b8a6",
"old_never_started": true,
"new_fits": 6,
"old_fits": 4,
"cells": [
{
"id": 0,
"pool": 25,
"mode": "epoch",
"steps": 22200,
"projected_gpu_seconds": 9225.593911228025
},
{
"id": 1,
"pool": 50,
"mode": "epoch",
"steps": 38000,
"projected_gpu_seconds": 15443.120772559705
},
{
"id": 2,
"pool": 50,
"mode": "step",
"steps": 32000,
"projected_gpu_seconds": 13082.034622686915
},
{
"id": 3,
"pool": 50,
"mode": "step",
"steps": 64000,
"projected_gpu_seconds": 25674.49408867513
},
{
"id": 4,
"pool": 100,
"mode": "step",
"steps": 32000,
"projected_gpu_seconds": 13082.034622686915
},
{
"id": 5,
"pool": 100,
"mode": "step",
"steps": 64000,
"projected_gpu_seconds": 25674.49408867513
}
],
"postfit_reserve": 7200,
"projected_total_gpu_seconds": 109381.77210651181,
"provisional27h_seconds": 97200,
"status": "BUDGET_CLARIFICATION_PENDING_NO_EXECUTION",
"interpretation": "Each U and2U is a fresh fully decayed schedule; update effect includes changed LR schedule; supersedes original e24 common-trajectory design"
}
@@ -534,3 +534,8 @@ S2四fold训练及postfit已COMPLETE,不能重启。TRON依Claude1794946447一
Assist1794947961独立终态审计PASS原件d4999e4a完整接收,保留target日志非syscall追踪、CPU起始resource计数未留存等限制。Melt重核662当前哈希+552预测。terminal4b2a710d/resultc43a6ab2/complete2a552871;总98816.327737793GPU-s/144000,postCPU492.447055337/3600,审计和描述表成本另列。Claude1794946961已触发304f5771计划;下一步构建item1 e24fff7c/24GPUh(170GPUh总计划内)及新的Assist审阅,当前尚未启动。gate-v2失败保留,fresh-family仍关闭。下方旧S2运行及phase3未触发快照均为历史。
## item1 v2六fit已固定待审,v1未运行作废(2026-11-18)
Claude1794948643将e24共同轨迹改为U/2U独立fresh完整日程,8710最终帽120000GPU-s含postfit(phase3总170h不变)。新根/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18,implementation73f94bbc/contract1bdf2331/bindingd5ccfdd4;16CPU合成PASS,总投影109072.196950。六fit最多4GPU,单fit首epoch门及每5%重估,training112800+postfit7200;g_n联合更新预算/完整LR日程,非纯续训。原c8e2ee4c/a84e5e82从未启动,保留源及准备稿,不能签旧PASS或启动。新入口须实际Assist五字段PASS,训练末6checkpoint固定后停,后处理另有checkpoint-bound新审阅;当前无训练。S2已完成/发布2338dfe9,不能重启。
+4
View File
@@ -1,3 +1,7 @@
## item1 v2六fit已固定待审,v1未运行作废(2026-11-18)
Claude1794948643将e24共同轨迹改为U/2U独立fresh完整日程,8710最终帽120000GPU-s含postfit(phase3总170h不变)。新根/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18,implementation73f94bbc/contract1bdf2331/bindingd5ccfdd4;16CPU合成PASS,总投影109072.196950。六fit最多4GPU,单fit首epoch门及每5%重估,training112800+postfit7200;g_n联合更新预算/完整LR日程,非纯续训。原c8e2ee4c/a84e5e82从未启动,保留源及准备稿,不能签旧PASS或启动。新入口须实际Assist五字段PASS,训练末6checkpoint固定后停,后处理另有checkpoint-bound新审阅;当前无训练。S2已完成/发布2338dfe9,不能重启。
## phase3 item1入口已固定待新审阅,未启动(2026-11-18)
根/home/frank14f/MeltMind-runs/phase3-data-sufficiency-2026-11-18,implementationc8e2ee4c/contracta84e5e82;13CPU合成PASS。四新fit/6checkpoint,原S1同配方100端点复用,34/65/132组件;逐格成本投影83217.702861/86400GPU-s,training79200+postfit7200,phase3总170GPUh。详见docs/phase3-data-sufficiency-entry-2026-11-18.md;须精确新Assist PASS才唯一启动,停TRAIN_COMPLETE_POSTFIT_PENDING,后处理另有checkpoint-bound审阅。没有真实训练/test预测/target读取。S2已发布2338dfe9/4HTTP通过,不重启。
@@ -0,0 +1,15 @@
# Phase3 item1 v2 — separate complete update-budget schedules
Claude1794948643.973609 amends e24fff7c's common-trajectory design; Claude1794948710.366709 fixes the total item cap at120000 GPU-s within phase3's612000 GPU-s. The earlier c8e2ee4c/a84e5e82 entry is preserved, never started, superseded. This is a labelled design revision, not unchanged execution of e24.
Original fold0 pools remain nested:34/65/132 learning components,111/190/320 parts; validation111/test121 unchanged. Same-recipe25/50 pools get fresh seed0,200epochs with the S1 epoch schedule. Same-recipe100 endpoint reuses S1 final checkpoint and existing external predictions/metrics by exact identity; no refit. Each pool reuses its own previously frozen learning-only clock on exactly the same list, recomputes learning-only standardization, and retains the same zero-residual initialization/full-node FP32 graph model/FP64 skip/node-weighted objective/AdamW/clip1. Geometry-category coverage is unavailable because no frozen category ontology exists; component/source-exposure and node/edge coverage are reported without inventing categories.
Step-matched50%-U,50%-2U,100%-U,100%-2U are four separate fresh seed0 fits. U=32000,2U=64000. Each has3200 warmup steps from1e-4 to1e-3, then cosine to1e-5 at its own endpoint. Formula: for s<=3200,1e-4+(s-1)*9e-4/3199; afterward1e-5+(.001-.00001)*(1+cos(pi*(s-3200)/(S-3200)))/2 with S32000 or64000. Only the terminal checkpoint is adopted per fit. No common-trajectory U checkpoint reuse. g_n now measures the joint change in update budget and complete LR schedule, not pure additional updates along the same trajectory.
There are six new fits/checkpoints. IDs0/1 are epoch25/epoch50; IDs2/3 are50%-U/2U; IDs4/5 are100%-U/2U. FixedGPU assignments repeat0/1 only after prior jobs release them; at most4 simultaneous workers. Each traverses its fixed list with SeedSequence([0,traversal]); partial final traversals are retained. Node/edge exposure and repetitions are reported. Validation diagnostics every10epochs for same-recipe or3200steps for update arms never select weights. The online loss is not a frozen-model training-domain error. New test predictions and target scoring are absent from this executable entry.
The comparison contract fixes same-recipe and fixed-budget component-equal field/last/tau, separate cF-only baselines for all three clocks, training-own-domain and common validation/test evaluation, and jointly paired2000 bootstrap replicates(seed2026111408). Family slope divides bylog2(132/65), delta=.001. Update effect includes schedule changes. Wide/undefined intervals are not equivalence; one fold/seed excludes training/pool sampling uncertainty. S1 test reuse is disclosed. No3U, newfold/seed, fresh-family admission or CAD-count extrapolation.
Budget: training112800 plus postfit reserve7200 =120000 GPU-s, phase3 total612000 unchanged. Per-jobGPU andCPU caps11000/17000/14000/28400/14000/28400;15GiB VRAM,16GiB RSS,20GiB output each, aggregate64GiB RSS/80GiB. Setup/import/idle/waits/failure time charged. Every fit pauses after its first complete epoch until a cell-specific projection passes; the total includes settled cells once, current measured projections and unstarted-cell conservative projections plus postfit reserve. Refresh at each5-percent checkpoint. Stop on individual/aggregate excess, monitor/identity failure, no retries or cap changes. No dependency on all six being simultaneously active.
After six terminal passes, hash and Git-freeze six checkpoints, then stop TRAIN_COMPLETE_POSTFIT_PENDING. A new exact Assist PASS on this implementation/contract/comparison/full source map is mandatory before training launch. Postfit is subsequently implemented and bound to these checkpoints, with another exact pre-score PASS; old S1/S2 reviews never transfer. Existing user continuation authorization suffices; these are engineering identity reviews, not new user-go loops.
@@ -0,0 +1,7 @@
{
"status": "PASS",
"tests": 16,
"scope": "CPU schedule/dispatch/GPU ownership/partial traversal/cost/no-double-count/stop-gate checks",
"test_targets_read": false,
"GPU_workload": false
}
@@ -0,0 +1,11 @@
{
"science_ruling": "1794948643.973609",
"budget_ruling": "1794948710.366709",
"old_entry": "/home/frank14f/MeltMind-runs/phase3-data-sufficiency-2026-11-18",
"old_implementation": "c8e2ee4ca3d08cf0080ac18a1d17bfb42fb1b8a6",
"old_started": false,
"old_common_trajectory_preserved": true,
"new_complete_schedules": true,
"new_fits": 6,
"original_e24_spec_not_unchanged": true
}
@@ -0,0 +1,306 @@
{
"spec_sha256": "e24fff7c5cb91dff7acf1532f502daec60e4b7fec21b2d4d4dcde1e793be7fb4",
"fold": 0,
"seed": 0,
"pools": [
25,
50,
100
],
"common_validation": [
"00000111_ce24e0cbec8e45c89735d148_step_003",
"00000115_ce24e0cbec8e45c89735d148_step_007",
"00000116_ce24e0cbec8e45c89735d148_step_008",
"00000117_ce24e0cbec8e45c89735d148_step_009",
"00000119_ce24e0cbec8e45c89735d148_step_011",
"00000120_ce24e0cbec8e45c89735d148_step_012",
"00000131_945244874ed84e599987daa9_step_000",
"00000132_438507a38e404bfdbd51a16b_step_000",
"00000133_08b739e0511f4305b161cd05_step_000",
"00000190_185ecbc783f34408aef52119_step_000",
"00000196_c9b71ecdd0654716891b0db8_step_000",
"00000206_27417003e762409185cb2b69_step_000",
"00000238_4829bc0645d6414d9b5cec31_step_001",
"00000241_4829bc0645d6414d9b5cec31_step_004",
"00000243_4829bc0645d6414d9b5cec31_step_006",
"00000248_e87554b09cdc48c3add39db9_step_000",
"00000257_845fb06d7f2f4802a806a600_step_002",
"00000258_845fb06d7f2f4802a806a600_step_003",
"00000259_845fb06d7f2f4802a806a600_step_004",
"00000261_845fb06d7f2f4802a806a600_step_006",
"00000262_845fb06d7f2f4802a806a600_step_007",
"00000264_845fb06d7f2f4802a806a600_step_009",
"00000265_845fb06d7f2f4802a806a600_step_010",
"00000266_845fb06d7f2f4802a806a600_step_011",
"00000267_845fb06d7f2f4802a806a600_step_012",
"00000269_845fb06d7f2f4802a806a600_step_014",
"00000391_fd490b564a44413884fab719_step_000",
"00000405_f3759bbb4682423caf207e8e_step_002",
"00000406_f3759bbb4682423caf207e8e_step_003",
"00000444_4ed4c78d6d754aac90876fc2_step_003",
"00000445_4ed4c78d6d754aac90876fc2_step_004",
"00000501_5499914343bd4f549b38046d_step_002",
"00000502_5499914343bd4f549b38046d_step_003",
"00000574_5a4cef68211d4706b1ec8586_step_003",
"00000578_5a4cef68211d4706b1ec8586_step_007",
"00000579_5a4cef68211d4706b1ec8586_step_008",
"00000580_5a4cef68211d4706b1ec8586_step_009",
"00000582_5a4cef68211d4706b1ec8586_step_011",
"00000583_5a4cef68211d4706b1ec8586_step_012",
"00000597_7fe894812c4244e18241aa08_step_003",
"00000601_7fe894812c4244e18241aa08_step_007",
"00000602_7fe894812c4244e18241aa08_step_008",
"00000603_7fe894812c4244e18241aa08_step_009",
"00000605_7fe894812c4244e18241aa08_step_011",
"00000606_7fe894812c4244e18241aa08_step_012",
"00000654_d87d11d9c3ca4ee5a88797ae_step_002",
"00000758_8ab50540e9804365a8717ec0_step_005",
"00000759_8ab50540e9804365a8717ec0_step_006",
"00000760_8ab50540e9804365a8717ec0_step_007",
"00000763_8ab50540e9804365a8717ec0_step_010",
"00000816_36e6258346004cabb4c091b2_step_004",
"00000817_36e6258346004cabb4c091b2_step_005",
"00000892_be26dc83f78947738b2bfabd_step_001",
"00000894_be26dc83f78947738b2bfabd_step_003",
"00000897_be26dc83f78947738b2bfabd_step_006",
"00000919_cdcebf32644443b9812097ba_step_003",
"00000923_cdcebf32644443b9812097ba_step_007",
"00000927_cdcebf32644443b9812097ba_step_011",
"00000928_cdcebf32644443b9812097ba_step_012",
"00000955_a741c7966d2c40a8a511263e_step_001",
"00000984_0dd1386597104265ab67e6cf_step_007",
"00000986_0dd1386597104265ab67e6cf_step_009",
"00000987_0dd1386597104265ab67e6cf_step_010",
"00000988_0dd1386597104265ab67e6cf_step_011",
"00000996_6d71695ba3fd475c927a00a7_step_000",
"00001019_ca4cf9693d6d4a98b515ba99_step_000",
"00001020_ca4cf9693d6d4a98b515ba99_step_001",
"00001021_ca4cf9693d6d4a98b515ba99_step_002",
"00001026_0f133d8e7fd448708a554e28_step_000",
"00001027_0f133d8e7fd448708a554e28_step_001",
"00001040_74c42207fa764831b03f7658_step_000",
"00001052_2fd6155fa34f418083d0bd0f_step_001",
"00001053_2fd6155fa34f418083d0bd0f_step_002",
"00001097_256ee7ec3a4b4972a8b87c8d_step_001",
"00001099_256ee7ec3a4b4972a8b87c8d_step_003",
"00001102_256ee7ec3a4b4972a8b87c8d_step_006",
"00001107_78566795c3144719804f8d34_step_000",
"00001167_107f61d0402f4c9a9d9d1342_step_000",
"00001168_107f61d0402f4c9a9d9d1342_step_001",
"00001217_e517c61f1c49433bb81cf980_step_000",
"00001238_96503b6a26c64074ad598927_step_000",
"00001239_96503b6a26c64074ad598927_step_001",
"00001240_96503b6a26c64074ad598927_step_002",
"00001241_96503b6a26c64074ad598927_step_003",
"00001242_96503b6a26c64074ad598927_step_004",
"00001243_96503b6a26c64074ad598927_step_005",
"00001244_96503b6a26c64074ad598927_step_006",
"00001246_96503b6a26c64074ad598927_step_008",
"00001247_96503b6a26c64074ad598927_step_009",
"00001249_96503b6a26c64074ad598927_step_011",
"00001252_96503b6a26c64074ad598927_step_014",
"00001253_96503b6a26c64074ad598927_step_015",
"00001314_a9ed5e56a6a8414f91b39bdb_step_000",
"00001333_beab79941e884455a62af40c_step_001",
"00001334_beab79941e884455a62af40c_step_002",
"00001335_beab79941e884455a62af40c_step_003",
"00001336_beab79941e884455a62af40c_step_004",
"00001338_beab79941e884455a62af40c_step_006",
"00001339_beab79941e884455a62af40c_step_007",
"00001340_beab79941e884455a62af40c_step_008",
"00001352_4e1d56bc2cef49ea84626b11_step_000",
"00001353_4e1d56bc2cef49ea84626b11_step_001",
"00001354_4e1d56bc2cef49ea84626b11_step_002",
"00001379_5b1d6e2599ea47d498a04256_step_000",
"00001387_60cf3ad636dd42e0bcec6946_step_000",
"00001511_5cbe58077e4b4885a5c213b3_step_004",
"00001512_5cbe58077e4b4885a5c213b3_step_005",
"00001525_e0d35a03b46745259cf8878f_step_000",
"00001532_9f0a4a641e8f4cc588eb4315_step_000",
"00001552_b0a9c4d836294846891615fe_step_000",
"00001553_adfa1c33ae4d4d0aa441ddc3_step_000"
],
"common_test": [
"00000006_d4fe04f0f5f84b52bd4f10e4_step_001",
"00000008_9b3d6a97e8de4aa193b81000_step_000",
"00000009_9b3d6a97e8de4aa193b81000_step_001",
"00000011_e909f412cda24521865fac0f_step_000",
"00000102_5ed74bccca6f4e89829bcb5e_step_002",
"00000183_6c4172e8a17140f89d3b1c0a_step_000",
"00000220_5324c79b172f41e8a49afc6f_step_007",
"00000221_5324c79b172f41e8a49afc6f_step_008",
"00000249_07a2da801a60492fb1e5eeb3_step_000",
"00000408_36bcb4030e36417b8dc74b87_step_000",
"00000410_30afd12eede44fa8a1def049_step_000",
"00000411_305e45c9479f43979b0d7b53_step_000",
"00000413_94f421b1b33d47c592c9fe1d_step_000",
"00000421_bd0972033f464f8f961a150a_step_002",
"00000508_d1c0a44331fd41faa91d4c2e_step_000",
"00000509_d1c0a44331fd41faa91d4c2e_step_001",
"00000510_d1c0a44331fd41faa91d4c2e_step_002",
"00000522_a56dd4152b7e4ccf91212d7b_step_000",
"00000527_db75f3a38d9746a1bd66168d_step_000",
"00000542_4b1756f6e72b45c29a8e4eb4_step_000",
"00000549_41a24c0b74a34217832f5af0_step_000",
"00000551_41a24c0b74a34217832f5af0_step_002",
"00000562_1724088724bf4e53a36e363e_step_000",
"00000566_2f2b015ef4904c1caef56154_step_000",
"00000637_5ae45a997c76423aba2fcc11_step_002",
"00000638_5ae45a997c76423aba2fcc11_step_003",
"00000643_bb55eaa4e9e24c2ea9d18ee0_step_000",
"00000659_6fae5591b3034f949b7e8eae_step_000",
"00000695_3f46763af0744ab2a90e6b48_step_000",
"00000746_6cb3ddbb49974d7db0e97135_step_001",
"00000796_dec80701adc84239b485449d_step_002",
"00000806_05cadeb104be4c8aa75ccc71_step_000",
"00000807_05cadeb104be4c8aa75ccc71_step_001",
"00000808_05cadeb104be4c8aa75ccc71_step_002",
"00000809_05cadeb104be4c8aa75ccc71_step_003",
"00000840_2c1efcc32d094e9fb9e1a88e_step_000",
"00000842_2c1efcc32d094e9fb9e1a88e_step_002",
"00000843_2c1efcc32d094e9fb9e1a88e_step_003",
"00000844_2c1efcc32d094e9fb9e1a88e_step_004",
"00000849_2c1efcc32d094e9fb9e1a88e_step_009",
"00000850_2c1efcc32d094e9fb9e1a88e_step_010",
"00000851_2c1efcc32d094e9fb9e1a88e_step_011",
"00000855_2c1efcc32d094e9fb9e1a88e_step_015",
"00000856_2c1efcc32d094e9fb9e1a88e_step_016",
"00000859_58ac6a347bbe4cf5ae07a565_step_000",
"00000863_029c2575dd474650ba1b8d21_step_003",
"00000864_029c2575dd474650ba1b8d21_step_004",
"00000866_029c2575dd474650ba1b8d21_step_006",
"00000867_029c2575dd474650ba1b8d21_step_007",
"00000868_029c2575dd474650ba1b8d21_step_008",
"00000870_029c2575dd474650ba1b8d21_step_010",
"00000871_029c2575dd474650ba1b8d21_step_011",
"00000872_029c2575dd474650ba1b8d21_step_012",
"00000873_029c2575dd474650ba1b8d21_step_013",
"00000874_029c2575dd474650ba1b8d21_step_014",
"00000877_029c2575dd474650ba1b8d21_step_017",
"00000879_029c2575dd474650ba1b8d21_step_019",
"00000880_029c2575dd474650ba1b8d21_step_020",
"00000882_029c2575dd474650ba1b8d21_step_022",
"00000883_029c2575dd474650ba1b8d21_step_023",
"00000884_029c2575dd474650ba1b8d21_step_024",
"00000885_029c2575dd474650ba1b8d21_step_025",
"00000886_029c2575dd474650ba1b8d21_step_026",
"00000887_029c2575dd474650ba1b8d21_step_027",
"00000938_4f351b53bfd84bb4a2f7809c_step_003",
"00000970_edacb215ec6949d8af6c8c68_step_000",
"00000993_ac4ba032d4b5431ca04f542c_step_000",
"00000994_ac4ba032d4b5431ca04f542c_step_001",
"00000998_6962bc8aad2c4e27b046d5b4_step_000",
"00001000_6962bc8aad2c4e27b046d5b4_step_002",
"00001002_6962bc8aad2c4e27b046d5b4_step_004",
"00001003_6962bc8aad2c4e27b046d5b4_step_005",
"00001004_6962bc8aad2c4e27b046d5b4_step_006",
"00001005_6962bc8aad2c4e27b046d5b4_step_007",
"00001006_6962bc8aad2c4e27b046d5b4_step_008",
"00001009_6962bc8aad2c4e27b046d5b4_step_011",
"00001010_6962bc8aad2c4e27b046d5b4_step_012",
"00001011_6962bc8aad2c4e27b046d5b4_step_013",
"00001013_6962bc8aad2c4e27b046d5b4_step_015",
"00001014_6962bc8aad2c4e27b046d5b4_step_016",
"00001015_6962bc8aad2c4e27b046d5b4_step_017",
"00001034_085828cd621d404a90ccd883_step_000",
"00001063_c6fda10dc58448d1951c5cc3_step_000",
"00001064_c6fda10dc58448d1951c5cc3_step_001",
"00001065_c6fda10dc58448d1951c5cc3_step_002",
"00001067_8a2edcf36bcd435092f4c8ae_step_001",
"00001069_8a2edcf36bcd435092f4c8ae_step_003",
"00001071_8a2edcf36bcd435092f4c8ae_step_005",
"00001112_3bd4a008a3a0483d86ccab3f_step_000",
"00001177_9f1ce06d5e334fe88e1581aa_step_001",
"00001178_515d7e1aeb364a2d88083cf3_step_000",
"00001262_157e5348bf6545e282472856_step_000",
"00001263_157e5348bf6545e282472856_step_001",
"00001305_3f3a389e9d5a4370b952bdce_step_001",
"00001310_3f3a389e9d5a4370b952bdce_step_006",
"00001358_2e314c6e080b40fcb6a6db96_step_001",
"00001360_2e314c6e080b40fcb6a6db96_step_003",
"00001368_2e314c6e080b40fcb6a6db96_step_011",
"00001369_2e314c6e080b40fcb6a6db96_step_012",
"00001371_2e314c6e080b40fcb6a6db96_step_014",
"00001373_2e314c6e080b40fcb6a6db96_step_016",
"00001375_2e314c6e080b40fcb6a6db96_step_018",
"00001377_2e314c6e080b40fcb6a6db96_step_020",
"00001398_49414db2dd594dc6a3907cf3_step_001",
"00001399_49414db2dd594dc6a3907cf3_step_002",
"00001400_49414db2dd594dc6a3907cf3_step_003",
"00001401_49414db2dd594dc6a3907cf3_step_004",
"00001402_49414db2dd594dc6a3907cf3_step_005",
"00001403_49414db2dd594dc6a3907cf3_step_006",
"00001404_49414db2dd594dc6a3907cf3_step_007",
"00001405_49414db2dd594dc6a3907cf3_step_008",
"00001406_49414db2dd594dc6a3907cf3_step_009",
"00001407_49414db2dd594dc6a3907cf3_step_010",
"00001408_49414db2dd594dc6a3907cf3_step_011",
"00001409_49414db2dd594dc6a3907cf3_step_012",
"00001410_49414db2dd594dc6a3907cf3_step_013",
"00001411_49414db2dd594dc6a3907cf3_step_014",
"00001412_49414db2dd594dc6a3907cf3_step_015",
"00001413_4642a707b7894949ba816427_step_000",
"00001522_7079a898746d41ba8ae13d69_step_003",
"00001598_c118664e9f6f4943855293fa_step_000"
],
"external_fold_previously_observed": true,
"epochs": 200,
"U": 32000,
"twoU": 64000,
"delta": 0.001,
"bootstrap": {
"replicates": 2000,
"seed": 2026111408,
"unit": "paired evaluation component",
"joint_resampling_all_contrasts": true
},
"endpoints": [
"epoch25",
"epoch50",
"S1_epoch100_reuse",
"step50_U",
"step50_2U",
"step100_U",
"step100_2U"
],
"clock_baselines": [
"c25F",
"c50F",
"c100F"
],
"domains": [
"own_learning",
"common_validation",
"common_test"
],
"metrics": [
"relative_l2",
"last_location",
"kendall_tau_b"
],
"component_counts": {
"25": 34,
"50": 65,
"100": 132
},
"slope": "(E50_H-E100_H)/log2(n100/n50), H=U and2U",
"gain": "E_n_U-E_n_2U for n50,n100",
"interaction": "s_2U-s_U; joint component resampling",
"classifications": {
"data_limited": "s2U low>delta and g100 entireCI in[-delta,delta]",
"updates_limited": "s2U entireCI in[-delta,delta] and g100 low>delta",
"both": "both low>delta",
"local_plateau": "both entireCI in[-delta,delta]",
"otherwise": "UNIDENTIFIED_OR_OVERFIT_SIGNS"
},
"undefined": "Report missing/undefined support; do not call wide/cross-zero CI plateau",
"not_training_randomness": true,
"no_3U_or_extrapolation": true,
"predictions_frozen_before_external_targets": true,
"new_exact_postfit_review_required": true,
"revision": "Claude1794948643 complete-schedule amendment to e24fff7c",
"schedule": "Four separate fresh step-budget fits, own complete warmup3200+cosine endpoint",
"update_effect": "joint update budget plus complete LR schedule, not pure continuation",
"new_fits": 6,
"old_entry_never_started": true
}
File diff suppressed because it is too large. Load diff
@@ -0,0 +1,34 @@
{
"reviewed_commit": "73f94bbc77e5ed3156ec85ff4a32ec2c834fdb07",
"contract_sha256": "1bdf2331c5127e03bec160835eda9d04ff333c3f70d5fe12a0629ffa82ec5826",
"implementation_proof_sha256": "fef00e34d1d8463c293ebf3c80b321a71d639d5213d6f63412c87c21a0e07422",
"comparison_contract_sha256": "332f2515911c99aadf2486086ddff82bd195d889faa8391296443e7aeacdfee9",
"code_sha256": {
"/home/frank14f/optane/MeltMind/tools/test_phase3_data_sufficiency_v2.py": "4f4c5238e14c33a536359ba581efbe73b0157afa3ff29bb9c0843467b34d4f84",
"/home/frank14f/optane/MeltMind/tools/phase3_data_sufficiency_v2_common.py": "1236bb02d3b3b7f31965bbeb2eba25aeaea1c920fa4031e26e6c5998e1c5b7fa",
"/home/frank14f/optane/MeltMind/tools/phase2_home_freeze_io.py": "bf09204b85ed2a7038acdd202257f618474d1110a9d6d772f105a3ffb334d3ff",
"/home/frank14f/optane/MeltMind/tools/phase2_local_freeze_io.py": "60adf3382577613c5234f5cf30779336b9951b66e90c075cb7cd0adab9cc42dd",
"/home/frank14f/optane/MeltMind/tools/phase2_h1_worker_v2.py": "1543df85b615bb69b79bacd572074a7c77ae19dab110197601573a7e5805ed20",
"/home/frank14f/optane/MeltMind/tools/phase2_residual_math.py": "ffbd9624dec79bdaabf95244fe87209055908f31a977e73ea4377045d6f8fe62",
"/home/frank14f/optane/MeltMind/tools/gre24_models.py": "53b6fee8b700b9b982b06d29908ce2ce23e0c57512c46dbf74b2aade79e3ac3a",
"/home/frank14f/optane/MeltMind/tools/phase2_fixed_io.py": "ea17fdfc7508f3f980e3cdf69534b1eb8ceb8a8e469c32451b2d5e16d0ada575",
"/home/frank14f/optane/MeltMind/tools/prepare_phase3_data_sufficiency_v2.py": "5529f7cfd2de6ab7ec85e3c6e589e17592cb03e86537b092ce42c4e9ee9ba5ff",
"/home/frank14f/optane/MeltMind/tools/phase2_h2_prepare.py": "3a31af0766f628a80f38e0b7286beb11ca3109188eabc4adfdb2ecb0eef3dc27",
"/home/frank14f/optane/MeltMind/tools/phase3_data_sufficiency_v2_queue.py": "ac99cab59152519fdcceddaad1e5e4cdd79f9f8dbedbab90ab04ad6171bcfe69",
"/home/frank14f/optane/MeltMind/tools/s2_recovery_gpu_supervisor.py": "689c0f3179e4fcd362e546d90214829a9e4f8739abdbdf74b560763e5a392f15",
"/home/frank14f/optane/MeltMind/tools/s2_recovery_monitoring.py": "442c2a9c75e8c00b30a8f2e91cb52d87c466b19a80aea9128b10c021e61b922e",
"/home/frank14f/optane/MeltMind/tools/phase2_gate_outcome_review.py": "915617aeb8292d03c654e637bc34a014c254fd178a773f135d1806c4b41372a8",
"/home/frank14f/optane/MeltMind/tools/phase3_data_sufficiency_v2_guard.py": "b978cde4bff64490141527cf7e084b7240aee8ff73d60172585585137a3ace44",
"/home/frank14f/optane/MeltMind/tools/s2_precision_continuation_guard.py": "01235d897c7c03e5ffa109c58b14f5d24d128e8f1ba873e1f76cfb14e02f7a7f",
"/home/frank14f/optane/MeltMind/tools/s2_gpu_idle_gate.py": "e2b1613e7e6ea398a8d555220a9365a1ad998ab4ef4b2bf8ec7cc4e31f8b4b9f",
"/home/frank14f/optane/MeltMind/tools/s2_pilot_train.py": "e9cceead51a5cc092aad23409ad98909c84a62e7e10571e750d1e9be1ad3b948",
"/home/frank14f/optane/MeltMind/tools/s2_models/network.py": "5137e845d81d1223afc13bf87e03ccba1132a3baa75d5189a8f320b2752be071",
"/home/frank14f/optane/MeltMind/tools/freeze_g1_components.py": "d0cee60d95243ac2c59ef3f2244ba818315718c4e3d1895f2f9f064f1f7f5ecd",
"/home/frank14f/optane/MeltMind/tools/s2_main_memory_guard.py": "e094f5443e11518efe9018215c11718a8e18aa462a977bf43e974b2d0b2aebb1",
"/home/frank14f/optane/MeltMind/tools/phase3_data_sufficiency_v2_worker.py": "851bc0710858f7af79339e56473d4dfb9b18fcb0961cca72c35654325be0a522",
"/home/frank14f/optane/MeltMind/tools/phase2_h2_pilot_worker.py": "14651ca402e6d02466d8afdb77bc9e62fb2e752f0c0c3d20d98593157d5ebe81",
"/home/frank14f/optane/MeltMind/tools/phase2_h2_model.py": "fab2470dd0654ca74c3b7d893f35ad1f3ae6c5c08ffe334ef6312cf6ec87892a",
"/home/frank14f/optane/MeltMind/tools/phase2_h2_data.py": "7e03dd7a94e7bb43593277f183c222473d7cb7aaf2f60e53b45ab8d9499fcb09",
"/home/frank14f/optane/MeltMind/docs/phase3-data-sufficiency-v2-entry-2026-11-18.md": "9c11a07fc40cbbc426b607ac1317ea58bcf8eb579816a41e61ddacfe01326c1d"
}
}
@@ -0,0 +1,4 @@
{
"contract_sha256": "1bdf2331c5127e03bec160835eda9d04ff333c3f70d5fe12a0629ffa82ec5826",
"proof_sha256": "fef00e34d1d8463c293ebf3c80b321a71d639d5213d6f63412c87c21a0e07422"
}
@@ -0,0 +1,75 @@
{
"commit": "73f94bbc77e5ed3156ec85ff4a32ec2c834fdb07",
"files": {
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/contract.json": "1bdf2331c5127e03bec160835eda9d04ff333c3f70d5fe12a0629ffa82ec5826",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/comparison-contract.json": "332f2515911c99aadf2486086ddff82bd195d889faa8391296443e7aeacdfee9",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/acceptance.json": "822c4e1c3012d4f89af7c89c091ba7f6341685ef76d3d21789ad26671125d690",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/prelaunch-projection.json": "0bf92288bc20b0c655f78d64f452fc3e85c30abb68d59b3fe53e3b5fd45e0bae",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/amendment.json": "dbb8f5ca1f941fd0410d5c4b27bc99628b159cd2e256d0a98df29395baaf195e",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/preparation-audit.json": "f78d9d01b052060d3875948343b173108b27b436962a2092c23feab21f5134a9",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/synthetic-attempt1.log": "c14966a3112a71f261f0eddd3071e1b2264946caaa01291e95557d0a0d5bc289",
"/home/frank14f/optane/MeltMind/tools/test_phase3_data_sufficiency_v2.py": "4f4c5238e14c33a536359ba581efbe73b0157afa3ff29bb9c0843467b34d4f84",
"/home/frank14f/optane/MeltMind/tools/phase3_data_sufficiency_v2_common.py": "1236bb02d3b3b7f31965bbeb2eba25aeaea1c920fa4031e26e6c5998e1c5b7fa",
"/home/frank14f/optane/MeltMind/tools/phase2_home_freeze_io.py": "bf09204b85ed2a7038acdd202257f618474d1110a9d6d772f105a3ffb334d3ff",
"/home/frank14f/optane/MeltMind/tools/phase2_local_freeze_io.py": "60adf3382577613c5234f5cf30779336b9951b66e90c075cb7cd0adab9cc42dd",
"/home/frank14f/optane/MeltMind/tools/phase2_h1_worker_v2.py": "1543df85b615bb69b79bacd572074a7c77ae19dab110197601573a7e5805ed20",
"/home/frank14f/optane/MeltMind/tools/phase2_residual_math.py": "ffbd9624dec79bdaabf95244fe87209055908f31a977e73ea4377045d6f8fe62",
"/home/frank14f/optane/MeltMind/tools/gre24_models.py": "53b6fee8b700b9b982b06d29908ce2ce23e0c57512c46dbf74b2aade79e3ac3a",
"/home/frank14f/optane/MeltMind/tools/phase2_fixed_io.py": "ea17fdfc7508f3f980e3cdf69534b1eb8ceb8a8e469c32451b2d5e16d0ada575",
"/home/frank14f/optane/MeltMind/tools/prepare_phase3_data_sufficiency_v2.py": "5529f7cfd2de6ab7ec85e3c6e589e17592cb03e86537b092ce42c4e9ee9ba5ff",
"/home/frank14f/optane/MeltMind/tools/phase2_h2_prepare.py": "3a31af0766f628a80f38e0b7286beb11ca3109188eabc4adfdb2ecb0eef3dc27",
"/home/frank14f/optane/MeltMind/tools/phase3_data_sufficiency_v2_queue.py": "ac99cab59152519fdcceddaad1e5e4cdd79f9f8dbedbab90ab04ad6171bcfe69",
"/home/frank14f/optane/MeltMind/tools/s2_recovery_gpu_supervisor.py": "689c0f3179e4fcd362e546d90214829a9e4f8739abdbdf74b560763e5a392f15",
"/home/frank14f/optane/MeltMind/tools/s2_recovery_monitoring.py": "442c2a9c75e8c00b30a8f2e91cb52d87c466b19a80aea9128b10c021e61b922e",
"/home/frank14f/optane/MeltMind/tools/phase2_gate_outcome_review.py": "915617aeb8292d03c654e637bc34a014c254fd178a773f135d1806c4b41372a8",
"/home/frank14f/optane/MeltMind/tools/phase3_data_sufficiency_v2_guard.py": "b978cde4bff64490141527cf7e084b7240aee8ff73d60172585585137a3ace44",
"/home/frank14f/optane/MeltMind/tools/s2_precision_continuation_guard.py": "01235d897c7c03e5ffa109c58b14f5d24d128e8f1ba873e1f76cfb14e02f7a7f",
"/home/frank14f/optane/MeltMind/tools/s2_gpu_idle_gate.py": "e2b1613e7e6ea398a8d555220a9365a1ad998ab4ef4b2bf8ec7cc4e31f8b4b9f",
"/home/frank14f/optane/MeltMind/tools/s2_pilot_train.py": "e9cceead51a5cc092aad23409ad98909c84a62e7e10571e750d1e9be1ad3b948",
"/home/frank14f/optane/MeltMind/tools/s2_models/network.py": "5137e845d81d1223afc13bf87e03ccba1132a3baa75d5189a8f320b2752be071",
"/home/frank14f/optane/MeltMind/tools/freeze_g1_components.py": "d0cee60d95243ac2c59ef3f2244ba818315718c4e3d1895f2f9f064f1f7f5ecd",
"/home/frank14f/optane/MeltMind/tools/s2_main_memory_guard.py": "e094f5443e11518efe9018215c11718a8e18aa462a977bf43e974b2d0b2aebb1",
"/home/frank14f/optane/MeltMind/tools/phase3_data_sufficiency_v2_worker.py": "851bc0710858f7af79339e56473d4dfb9b18fcb0961cca72c35654325be0a522",
"/home/frank14f/optane/MeltMind/tools/phase2_h2_pilot_worker.py": "14651ca402e6d02466d8afdb77bc9e62fb2e752f0c0c3d20d98593157d5ebe81",
"/home/frank14f/optane/MeltMind/tools/phase2_h2_model.py": "fab2470dd0654ca74c3b7d893f35ad1f3ae6c5c08ffe334ef6312cf6ec87892a",
"/home/frank14f/optane/MeltMind/tools/phase2_h2_data.py": "7e03dd7a94e7bb43593277f183c222473d7cb7aaf2f60e53b45ab8d9499fcb09",
"/home/frank14f/optane/MeltMind/docs/phase3-data-sufficiency-v2-entry-2026-11-18.md": "9c11a07fc40cbbc426b607ac1317ea58bcf8eb579816a41e61ddacfe01326c1d"
},
"git_paths": {
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/contract.json": "home-runs/phase3-data-sufficiency-v2-2026-11-18/contract.json",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/comparison-contract.json": "home-runs/phase3-data-sufficiency-v2-2026-11-18/comparison-contract.json",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/acceptance.json": "home-runs/phase3-data-sufficiency-v2-2026-11-18/acceptance.json",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/prelaunch-projection.json": "home-runs/phase3-data-sufficiency-v2-2026-11-18/prelaunch-projection.json",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/amendment.json": "home-runs/phase3-data-sufficiency-v2-2026-11-18/amendment.json",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/preparation-audit.json": "home-runs/phase3-data-sufficiency-v2-2026-11-18/preparation-audit.json",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/synthetic-attempt1.log": "home-runs/phase3-data-sufficiency-v2-2026-11-18/synthetic-attempt1.log",
"/home/frank14f/optane/MeltMind/tools/test_phase3_data_sufficiency_v2.py": "tools/test_phase3_data_sufficiency_v2.py",
"/home/frank14f/optane/MeltMind/tools/phase3_data_sufficiency_v2_common.py": "tools/phase3_data_sufficiency_v2_common.py",
"/home/frank14f/optane/MeltMind/tools/phase2_home_freeze_io.py": "tools/phase2_home_freeze_io.py",
"/home/frank14f/optane/MeltMind/tools/phase2_local_freeze_io.py": "tools/phase2_local_freeze_io.py",
"/home/frank14f/optane/MeltMind/tools/phase2_h1_worker_v2.py": "tools/phase2_h1_worker_v2.py",
"/home/frank14f/optane/MeltMind/tools/phase2_residual_math.py": "tools/phase2_residual_math.py",
"/home/frank14f/optane/MeltMind/tools/gre24_models.py": "tools/gre24_models.py",
"/home/frank14f/optane/MeltMind/tools/phase2_fixed_io.py": "tools/phase2_fixed_io.py",
"/home/frank14f/optane/MeltMind/tools/prepare_phase3_data_sufficiency_v2.py": "tools/prepare_phase3_data_sufficiency_v2.py",
"/home/frank14f/optane/MeltMind/tools/phase2_h2_prepare.py": "tools/phase2_h2_prepare.py",
"/home/frank14f/optane/MeltMind/tools/phase3_data_sufficiency_v2_queue.py": "tools/phase3_data_sufficiency_v2_queue.py",
"/home/frank14f/optane/MeltMind/tools/s2_recovery_gpu_supervisor.py": "tools/s2_recovery_gpu_supervisor.py",
"/home/frank14f/optane/MeltMind/tools/s2_recovery_monitoring.py": "tools/s2_recovery_monitoring.py",
"/home/frank14f/optane/MeltMind/tools/phase2_gate_outcome_review.py": "tools/phase2_gate_outcome_review.py",
"/home/frank14f/optane/MeltMind/tools/phase3_data_sufficiency_v2_guard.py": "tools/phase3_data_sufficiency_v2_guard.py",
"/home/frank14f/optane/MeltMind/tools/s2_precision_continuation_guard.py": "tools/s2_precision_continuation_guard.py",
"/home/frank14f/optane/MeltMind/tools/s2_gpu_idle_gate.py": "tools/s2_gpu_idle_gate.py",
"/home/frank14f/optane/MeltMind/tools/s2_pilot_train.py": "tools/s2_pilot_train.py",
"/home/frank14f/optane/MeltMind/tools/s2_models/network.py": "tools/s2_models/network.py",
"/home/frank14f/optane/MeltMind/tools/freeze_g1_components.py": "tools/freeze_g1_components.py",
"/home/frank14f/optane/MeltMind/tools/s2_main_memory_guard.py": "tools/s2_main_memory_guard.py",
"/home/frank14f/optane/MeltMind/tools/phase3_data_sufficiency_v2_worker.py": "tools/phase3_data_sufficiency_v2_worker.py",
"/home/frank14f/optane/MeltMind/tools/phase2_h2_pilot_worker.py": "tools/phase2_h2_pilot_worker.py",
"/home/frank14f/optane/MeltMind/tools/phase2_h2_model.py": "tools/phase2_h2_model.py",
"/home/frank14f/optane/MeltMind/tools/phase2_h2_data.py": "tools/phase2_h2_data.py",
"/home/frank14f/optane/MeltMind/docs/phase3-data-sufficiency-v2-entry-2026-11-18.md": "docs/phase3-data-sufficiency-v2-entry-2026-11-18.md"
},
"remote_published": false,
"local_repository": "/tmp/meltmind-phase2-local-fixed-20261117"
}
@@ -0,0 +1,72 @@
{
"cells": [
{
"id": 0,
"pool": 25,
"mode": "epoch",
"steps": 22200,
"expected_new_gpu_seconds": 9225.593911228025,
"job_cap": 11000,
"within_cap": true,
"validation_checks": 20
},
{
"id": 1,
"pool": 50,
"mode": "epoch",
"steps": 38000,
"expected_new_gpu_seconds": 15443.120772559705,
"job_cap": 17000,
"within_cap": true,
"validation_checks": 20
},
{
"id": 2,
"pool": 50,
"mode": "step",
"steps": 32000,
"expected_new_gpu_seconds": 12927.247044337564,
"job_cap": 14000,
"within_cap": true,
"validation_checks": 10
},
{
"id": 3,
"pool": 50,
"mode": "step",
"steps": 64000,
"expected_new_gpu_seconds": 25674.49408867513,
"job_cap": 28400,
"within_cap": true,
"validation_checks": 20
},
{
"id": 4,
"pool": 100,
"mode": "step",
"steps": 32000,
"expected_new_gpu_seconds": 12927.247044337564,
"job_cap": 14000,
"within_cap": true,
"validation_checks": 10
},
{
"id": 5,
"pool": 100,
"mode": "step",
"steps": 64000,
"expected_new_gpu_seconds": 25674.49408867513,
"job_cap": 28400,
"within_cap": true,
"validation_checks": 20
}
],
"total_gpu_seconds": 109072.19694981311,
"postfit_reserve": 7200,
"item_cap": 120000,
"phase3_cap": 612000,
"source_max_epoch": 100.7396757279057,
"source_max_validation": 12.383006267948076,
"safety_factor": 1.25,
"actual_run": false
}
@@ -0,0 +1,8 @@
{
"status": "PREPARED_NOT_STARTED",
"metadata_cpu_seconds": 0.4903385389999997,
"source_pins_verified": true,
"old_root_unstarted": true,
"no_target_arrays": true,
"no_GPU_workload": true
}
@@ -0,0 +1,4 @@
{
"contract_sha256": "1bdf2331c5127e03bec160835eda9d04ff333c3f70d5fe12a0629ffa82ec5826",
"proof_sha256": "91e70065f505470e1fcb91fa86794b17ab372be0c322545c2b6e08d23d793893"
}
@@ -0,0 +1,17 @@
{
"commit": "49f738f225b5706cb361ce09e91552e5e5b7c920",
"files": {
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/contract.json": "1bdf2331c5127e03bec160835eda9d04ff333c3f70d5fe12a0629ffa82ec5826",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/comparison-contract.json": "332f2515911c99aadf2486086ddff82bd195d889faa8391296443e7aeacdfee9",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/expected-review-binding.json": "d5ccfdd4edac51838b25d14dc801aeae4d62d5a6d9ce48c12a6cc5846bbf2b3c",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/implementation-git-freeze.json": "fef00e34d1d8463c293ebf3c80b321a71d639d5213d6f63412c87c21a0e07422"
},
"git_paths": {
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/contract.json": "home-runs/phase3-data-sufficiency-v2-2026-11-18/contract.json",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/comparison-contract.json": "home-runs/phase3-data-sufficiency-v2-2026-11-18/comparison-contract.json",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/expected-review-binding.json": "home-runs/phase3-data-sufficiency-v2-2026-11-18/expected-review-binding.json",
"/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18/implementation-git-freeze.json": "home-runs/phase3-data-sufficiency-v2-2026-11-18/implementation-git-freeze.json"
},
"remote_published": false,
"local_repository": "/tmp/meltmind-phase2-local-fixed-20261117"
}
@@ -0,0 +1,21 @@
test_cell_cap (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_cell_specific_gate (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_common_warmup (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_complete_schedule_both_budgets (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_distinct_U_schedule (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_global_stop (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_gpu_ownership (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_initial_four_dispatch (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_invalid_step (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_nan_cost (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_no_duplicate_gpu (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_no_more_than_four (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_partial_epoch (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_pending_projection_charged (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_terminal_not_double_counted (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
test_total_cap (tools.test_phase3_data_sufficiency_v2.V2Tests) ... ok
----------------------------------------------------------------------
Ran 16 tests in 0.003s
OK
@@ -0,0 +1,70 @@
"""Item1 frozen roles, schedules and costs. No model execution or target access."""
import json,math
from pathlib import Path
from phase2_fixed_io import BASE
from phase2_h1_worker_v2 import checked,sha
from phase2_home_freeze_io import verify
def read(p):return json.loads(Path(p).read_text())
def lr_step(step,total=64000,warmup=3200):
if not 1<=step<=total or not 1<warmup<total:raise ValueError('1-based fixed update domain')
if step<=warmup:return 1e-4+(step-1)*9e-4/(warmup-1)
return 1e-5+(1e-3-1e-5)*(1+math.cos(math.pi*(step-warmup)/(total-warmup)))/2
def execution_command(args):
return any(Path(x).name in {'phase3_data_sufficiency_v2_queue.py','phase3_data_sufficiency_v2_guard.py','phase3_data_sufficiency_v2_worker.py'} for x in args)
def config(root):
c=read(root/'contract.json');verify(root,'implementation')
if c['approval']!='1794948643.973609' or c['new_fits']!=6 or c['item_gpu_cap']!=120000 or c['budget_approval']!='1794948710.366709' or c['phase3_gpu_cap']!=612000 or c['train_gpu_cap']+c['postfit_gpu_reserve']!=c['item_gpu_cap']:raise PermissionError('Frozen item1 scope')
if c['U']!=32000 or c['delta']!=.001 or c['seed']!=0 or c['fold']!=0 or not c['no_retry']:raise ValueError('Frozen science identity')
for p,h in {**c['source_sha256'],**c['code_sha256']}.items():checked(p,h)
pilot=read(c['pilot_contract'])
import torch,numpy as np
if torch.__version__!=pilot['versions']['torch'] or np.__version__!=pilot['versions']['numpy']:raise ValueError('Original library identity')
if len({j['gpu_uuid'] for j in c['jobs']})!=4 or sum(j['gpu_seconds'] for j in c['jobs'])!=c['train_gpu_cap']:raise ValueError('Job resources')
pools=c['pools'];sets=[set(pools[str(p)]['learning']) for p in [25,50,100]]
if not sets[0]<sets[1]<sets[2] or sets[2]!=set(pilot['folds'][0]['learning']):raise ValueError('Original nested roles')
for p in [25,50,100]:
q=pools[str(p)]
if set(q['learning'])&(set(q['validation'])|set(q['test'])) or q['validation']!=pilot['folds'][0]['validation'] or q['test']!=pilot['folds'][0]['test']:raise ValueError('Role separation')
if read(q['clock_path'])['parts']!=q['learning'] or q['learning_nodes']!=sum(pilot['parts'][x]['nodes'] for x in q['learning']):raise ValueError('Learning clock/node identity')
for j in c['jobs']:
q=pools[str(j['pool'])];target=200*len(q['learning']) if j['mode']=='epoch' else j['step_budget']
if j['target_steps']!=target or j['learning_parts']!=len(q['learning']) or j['learning_nodes']!=q['learning_nodes']:raise ValueError('Cell identity')
return c,pilot
def view_for(c,pilot,job):return dict(pilot,folds=[c['pools'][str(job['pool'])]])
def project_cell(job,observation):
v=observation;steps=v['steps'];per=v['max_epoch_seconds']/job['learning_parts'];val=max(v['max_validation_seconds'],job['validation_seconds_bound'])
if not 0<steps<=job['target_steps'] or not all(math.isfinite(x) and x>=0 for x in [per,val,v['charged']]) or per<=0:raise ValueError('Invalid cost evidence')
checks=20 if job['mode']=='epoch' else job['target_steps']//3200
future=1.25*((job['target_steps']-steps)*per+checks*val)+120
return dict(id=job['id'],pool=job['pool'],mode=job['mode'],projected_gpu_seconds=v['charged']+future,job_cap=job['gpu_seconds'],within_cap=v['charged']+future<=job['gpu_seconds'])
def project_total(c,estimates,terminal):
if any(i not in range(6) for i in estimates) or any(i not in range(6) for i in terminal):raise ValueError('Unknown cell')
total=c['postfit_gpu_reserve'];rows=[]
for j in c['jobs']:
i=j['id'];cost=terminal[i]['gpu_seconds_charged'] if i in terminal else estimates.get(i,{}).get('projected_gpu_seconds',j['prelaunch_projected_gpu_seconds'])
if not math.isfinite(cost) or cost<0:raise ValueError('Invalid ledger')
rows.append(dict(id=i,projected_gpu_seconds=cost,within_cap=cost<=j['gpu_seconds']));total+=cost
return dict(status='PASS' if total<=c['item_gpu_cap'] and all(x['within_cap'] for x in rows) else 'STOP_BUDGET_PROJECTION',rows=rows,projected_item_gpu_seconds=total,postfit_reserve=c['postfit_gpu_reserve'],phase3_cap=c['phase3_gpu_cap'])
def first_epoch_release(root,jobid):
if (root/'global-stop.json').exists():raise RuntimeError('Global item1 stop')
p=root/'jobs'/str(jobid)/'first-epoch-projection.json'
if not p.exists():return False
d=read(p)
if d['status']!='PASS' or d['contract_sha256']!=sha(root/'contract.json'):raise RuntimeError('First-epoch projection rejected')
return True
def dispatchable(pending,active_jobs,limit=4):
occupied={j['gpu_uuid'] for j in active_jobs};slots=limit-len(active_jobs);selected=[]
if slots<0 or len(occupied)!=len(active_jobs):raise ValueError('Active GPU collision/cap')
for j in pending:
if not slots:break
if j['gpu_uuid'] in occupied:continue
selected.append(j);occupied.add(j['gpu_uuid']);slots-=1
return selected
+86
View File
@@ -0,0 +1,86 @@
"""Independent bounded ITEM1 fold guard; shared global stop and immutable S1."""
import fcntl
import json
import os
from pathlib import Path
import signal
import sys
import threading
import time
import psutil
from phase3_data_sufficiency_v2_common import config
from phase2_h1_worker_v2 import save,sha
from s2_main_memory_guard import PeakRecords
from s2_precision_continuation_guard import device_usage
from s2_recovery_gpu_supervisor import supervise,tree_usage,stop_group,disk_usage
def run(queue,fold_id):
c,_=config(queue);job=c['jobs'][fold_id];c=dict(c,**job);root=queue/'jobs'/str(fold_id);os.sched_setaffinity(0,c['guard_cpus'])
with (root/'fold.lock').open('a') as lock:
fcntl.flock(lock,fcntl.LOCK_EX|fcntl.LOCK_NB)
if (root/'started.json').exists():raise FileExistsError('Single ITEM1 fold launch, no retry')
start=time.monotonic();cpu=time.process_time();requested=[];reasons=[];errors=[];samples=[];done=threading.Event();tracker=PeakRecords();receipt=None
save(root/'started.json',dict(pid=os.getpid(),unix=time.time(),contract_sha256=sha(queue/'contract.json'),gpu_uuid=c['gpu_uuid'],setup_gpu_seconds=c['setup_gpu_seconds']))
def elapsed():return c['setup_gpu_seconds']+time.monotonic()-start
def reason(x):
if x not in reasons:reasons.append(x)
command=[sys.executable,'tools/phase3_data_sufficiency_v2_worker.py',str(queue),str(fold_id)]
def kill_worker():
for p in psutil.Process().children(recursive=True):
try:
if p.cmdline()==command:stop_group(p.pid)
except psutil.NoSuchProcess:pass
old={s:signal.signal(s,lambda signum,frame:requested.append('COORDINATOR_SIGNAL_'+str(signum))) for s in [signal.SIGINT,signal.SIGTERM]}
def scan():
if tracker.scan(root/'worker'):reason('RECORDED_VRAM_LIMIT')
def watch():
while not done.is_set():
try:
scan();mem,util=device_usage(c['gpu_uuid']);_,rss,_=tree_usage(os.getpid())
if mem>c['vram_bytes']:reason('DEVICE_VRAM_LIMIT')
if rss>c['rss_bytes']:reason('WHOLE_TREE_RSS_LIMIT')
if disk_usage(root)>c['disk_bytes']:reason('OUTPUT_DISK_LIMIT')
if elapsed()>=c['gpu_seconds']-5:reason('ITEM1_FOLD_TIME_LIMIT')
if (queue/'global-stop.json').exists():reason('ITEM1_GLOBAL_STOP')
for x in requested:reason(x)
save(root/'live-progress.json',dict(stage='ITEM1_TRAINING',fold=fold_id,elapsed_gpu_seconds=elapsed(),device_bytes=mem,device_utilization=util,tree_rss_bytes=rss,recorded_peaks=tracker.maxima,reasons=reasons,unix=time.time()))
except Exception as exc:reason('MONITOR_ERROR');errors.append(dict(type=type(exc).__name__,reason=str(exc)))
if reasons:kill_worker()
done.wait(.5)
thread=None
try:
for i in range(3):
mem,util=device_usage(c['gpu_uuid']);samples.append(dict(unix=time.time(),memory_bytes=mem,utilization=util))
if mem>128*2**20 or util!=0:raise RuntimeError('SELECTED_GPU_BUSY')
if i<2:time.sleep(2)
save(root/'idle-polls.json',samples)
thread=threading.Thread(target=watch,daemon=True);thread.start()
remaining=c['gpu_seconds']-elapsed()-10
receipt=supervise(command,root/'guards','train',dict(total_cpu_seconds=c['cpu_seconds']-60,job_cpu_seconds=c['cpu_seconds']-60,job_wall_seconds=remaining,tree_rss_bytes=c['rss_bytes'],disk_bytes=c['disk_bytes'],max_processes=1),poll=.2,gpu_uuid=c['gpu_uuid'])
except BaseException as exc:reason(type(exc).__name__+': '+str(exc));kill_worker()
finally:
done.set()
if thread:thread.join(timeout=4)
if thread and thread.is_alive():reason('MONITOR_DID_NOT_EXIT')
try:scan()
except Exception as exc:reason('FINAL_PEAK_SCAN_ERROR');errors.append(dict(reason=str(exc)))
for s,handler in old.items():signal.signal(s,handler)
value=None;p=root/'worker/result.json'
try:
if p.exists():value=json.loads(p.read_text())
except Exception as exc:reason('RESULT_READ_ERROR');errors.append(dict(reason=str(exc)))
if not receipt or receipt['status']!='PASS':reason('WORKER_NOT_PASS')
if receipt and receipt.get('monitor_errors'):reason('WORKER_MONITOR_ERRORS')
if not value or value.get('status')!='COMPLETE':reason('NO_COMPLETE_TRAIN_RESULT')
if not all(k in tracker.witnesses for k in ['peak_allocated_bytes','peak_reserved_bytes']):reason('MISSING_CUDA_PEAK_RECORDS')
totalcpu=60+(receipt['cpu_seconds'] if receipt else 0)+time.process_time()-cpu
if elapsed()>c['gpu_seconds'] or totalcpu>c['cpu_seconds']:reason('FINAL_RESOURCE_LIMIT')
try:config(queue)
except Exception as exc:reason('FINAL_SOURCE_IDENTITY: '+str(exc))
if value and (value.get('optimizer_steps')!=c['target_steps'] or value.get('job')!=fold_id):reason('INCOMPLETE_EPOCH_OR_FOLD_COUNT')
save(root/'complete.json',dict(status='PASS' if not reasons else 'STOPPED_OR_FAILED',decision='TRAIN_COMPLETE_CHECKPOINT_FREEZE_PENDING' if not reasons else 'ITEM1_FOLD_INCOMPLETE',reasons=reasons,monitor_errors=errors,
worker=receipt,recorded_memory=tracker.as_dict(),gpu_seconds_charged=elapsed(),fold=fold_id,cpu_seconds_charged=totalcpu,
worker_result_sha256=sha(p) if p.exists() else None,prediction_started=False,scoring_started=False,s2_started=True,no_retry=True))
return 0 if not reasons else 1
if __name__=='__main__':raise SystemExit(run(Path(sys.argv[1]),int(sys.argv[2])))
+95
View File
@@ -0,0 +1,95 @@
"""Single six-cell queue, at most four GPU workers; first-epoch and periodic cost gates; checkpoint-only terminal."""
import fcntl,os,signal,subprocess,sys,time
from pathlib import Path
import psutil
from phase3_data_sufficiency_v2_common import config,read,project_cell,project_total,execution_command,dispatchable
from phase2_h1_worker_v2 import save,sha,checked
from phase2_home_freeze_io import freeze
from phase2_gate_outcome_review import validate_review
from s2_recovery_gpu_supervisor import tree_usage,disk_usage
def run(root):
c,pilot=config(root);validate_review(root);lock=(root/'queue.lock').open('a');fcntl.flock(lock,fcntl.LOCK_EX|fcntl.LOCK_NB)
if (root/'started.json').exists():raise FileExistsError('Single item1 attempt; no retry')
for p in psutil.process_iter(['pid','cmdline']):
if p.pid!=os.getpid() and execution_command(p.info['cmdline'] or []):raise RuntimeError('Existing item1 execution')
os.sched_setaffinity(0,c['coordinator_cpus']);os.environ['PYTORCH_CUDA_ALLOC_CONF']='expandable_segments:True';start=time.monotonic();cpu=time.process_time();active={};terminal={};requested=[];last_projection={};estimates={};pending=list(c['jobs'])
def stop(reason):
if not (root/'global-stop.json').exists():save(root/'global-stop.json',dict(reason=reason,unix=time.time(),no_retry=True))
def charged():return sum(v['gpu_seconds_charged'] for v in terminal.values())+sum(60+time.monotonic()-v['start'] for v in active.values())
def terminate(v):
if v.get('stop_sent') is None:v['process'].send_signal(signal.SIGTERM);v['stop_sent']=time.monotonic()
elif time.monotonic()-v['stop_sent']>20:
try:
parent=psutil.Process(v['process'].pid)
for p in reversed(parent.children(recursive=True)):
try:p.kill()
except psutil.NoSuchProcess:pass
parent.kill()
except psutil.NoSuchProcess:pass
def collect(i,v):
rc=v['process'].poll()
if rc is None:return False
p=root/'jobs'/str(i)/'complete.json';r=read(p) if p.exists() else None;observed=60+time.monotonic()-v['start'];ok=rc==0 and r is not None and r['status']=='PASS'
terminal[i]=dict(id=i,status='PASS' if ok else 'STOPPED_OR_FAILED',returncode=rc,gpu_seconds_charged=max(observed,r['gpu_seconds_charged'] if r else 0),guard_complete_sha256=sha(p) if r else None);v['log'].close();del active[i]
save(root/'terminal-ledger.json',dict(entries=list(terminal.values()),settled_gpu_seconds=sum(x['gpu_seconds_charged'] for x in terminal.values())))
if not ok:stop('CELL_'+str(i)+'_FAILED')
return True
def observe(j):
i=j['id'];rows=[read(p) for p in sorted((root/'jobs'/str(i)/'worker/history').glob('*.json'))];full=[r for r in rows if r['full_epoch']]
if not full:return None
return dict(steps=rows[-1]['optimizer_steps'],charged=terminal[i]['gpu_seconds_charged'] if i in terminal else 60+time.monotonic()-active[i]['start'],max_epoch_seconds=max(r['train_epoch_seconds'] for r in full),max_validation_seconds=max(r['validation_seconds'] for r in rows))
def dispatch():
for j in dispatchable(pending,[c['jobs'][i] for i in active]):
i=j['id'];d=root/'jobs'/str(i);d.mkdir(parents=True,exist_ok=False);log=(d/'entry.log').open('x');begin=time.monotonic();cmd=[sys.executable,'tools/phase3_data_sufficiency_v2_guard.py',str(root),str(i)];p=subprocess.Popen(cmd,stdout=log,stderr=subprocess.STDOUT,start_new_session=True);active[i]=dict(process=p,log=log,start=begin,stop_sent=None);save(d/'dispatch.json',dict(pid=p.pid,command=cmd,unix=time.time(),gpu_uuid=j['gpu_uuid']));pending.remove(j)
handlers={s:signal.signal(s,lambda n,f:requested.append('SIGNAL_'+str(n))) for s in [signal.SIGTERM,signal.SIGINT]}
save(root/'started.json',dict(pid=os.getpid(),unix=time.time(),contract_sha256=sha(root/'contract.json'),review_sha256=sha(root/'review.json'),new_fits=6,S1_retrained=False,item_gpu_cap=120000,postfit_started=False))
try:
while active or pending:
if not (root/'global-stop.json').exists():dispatch()
for i,v in list(active.items()):collect(i,v)
for x in requested:stop(x)
amount=charged();_,rss,_=tree_usage(os.getpid())
if amount>=c['train_gpu_cap']-80:stop('CUMULATIVE_TRAIN_GPU_RESERVE_BOUNDARY')
if rss>64*2**30 or disk_usage(root)>80*2**30:stop('AGGREGATE_MEMORY_DISK_LIMIT')
if not (root/'global-stop.json').exists():
for j in c['jobs']:
i=j['id']
if i not in active:continue
obs=observe(j)
if obs is None:continue
stage=int(obs['steps']/j['target_steps']*20)
if stage>last_projection.get(i,-1):
cell=project_cell(j,obs);estimates[i]=cell;total=project_total(c,estimates,terminal);p=dict(**total,cell=cell,contract_sha256=sha(root/'contract.json'),unix=time.time())
if not cell['within_cap']:p['status']='STOP_BUDGET_PROJECTION'
save(root/'jobs'/str(i)/('first-epoch-projection.json' if i not in last_projection else 'latest-cost-projection.json'),p);save(root/'latest-cost-projection.json',p);last_projection[i]=stage
if p['status']!='PASS':stop('PROJECTED_ITEM_GPU_BUDGET')
if (root/'global-stop.json').exists():
for v in active.values():terminate(v)
save(root/'live-progress.json',dict(stage='STOPPING' if (root/'global-stop.json').exists() else 'ITEM1_TRAINING',unix=time.time(),active=[dict(id=i,guard_pid=v['process'].pid) for i,v in active.items()],terminal=list(terminal.values()),pending=[j['id'] for j in pending],charged_training_gpu_seconds=amount,postfit_gpu_reserve=c['postfit_gpu_reserve'],item_cap=120000,phase3_cap=612000))
if (root/'global-stop.json').exists() and not active:break
if active:time.sleep(1)
if (root/'global-stop.json').exists() or len(terminal)!=6 or any(x['status']!='PASS' for x in terminal.values()):raise RuntimeError('Item1 train incomplete')
config(root);models=[]
for j in c['jobs']:
w=root/'jobs'/str(j['id'])/'worker';r=read(w/'result.json');stats=read(checked(w/'standardization.json',r['standardization_sha256']))
if r['optimizer_steps']!=j['target_steps'] or stats['parts']!=c['pools'][str(j['pool'])]['learning'] or r['test_targets_read']:raise ValueError('Terminal cell identity')
for m in r['checkpoints']:checked(m['path'],m['sha256']);models.append(dict(job=j['id'],pool=j['pool'],mode=j['mode'],**m))
if len(models)!=6 or charged()>c['train_gpu_cap']:raise ValueError('Checkpoint or cumulative cost')
save(root/'models-terminal.json',dict(models=models,reused_S1=c['S1_reuse'],new_fits=6,new_checkpoints=6,test_prediction_started=False))
freeze(root,'models',[root/'contract.json',root/'models-terminal.json',root/'terminal-ledger.json',*[Path(m['path']) for m in models],*[Path(m['standardization']) for m in models]])
save(root/'complete.json',dict(status='TRAIN_COMPLETE_POSTFIT_PENDING',entries=list(terminal.values()),training_gpu_seconds=charged(),postfit_gpu_reserve=c['postfit_gpu_reserve'],item_gpu_cap=120000,coordinator_cpu_seconds=time.process_time()-cpu,wall_seconds=time.monotonic()-start,no_retry=True,test_targets_read=False));freeze(root,'terminal',[root/'contract.json',root/'complete.json',root/'models-terminal.json']);return 0
except BaseException as exc:
stop(type(exc).__name__+': '+str(exc))
while active:
for i,v in list(active.items()):
try:
if not collect(i,v):terminate(v)
except Exception:
terminate(v)
if v['process'].poll() is not None:terminal[i]=dict(id=i,status='MALFORMED_TERMINAL',gpu_seconds_charged=60+time.monotonic()-v['start']);v['log'].close();del active[i]
if active:time.sleep(.5)
save(root/'complete.json',dict(status='STOPPED',reason=str(exc),entries=list(terminal.values()),never_attempted=[j['id'] for j in pending],training_gpu_seconds=charged(),postfit_started=False,no_retry=True));return 1
finally:
for s,h in handlers.items():signal.signal(s,h)
if __name__=='__main__':raise SystemExit(run(Path(sys.argv[1])))
@@ -0,0 +1,71 @@
"""Six fixed fresh fits; each update budget has a complete schedule; no test prediction/target access."""
import os,sys,time,traceback,collections
from pathlib import Path
import numpy as np
import torch
from phase3_data_sufficiency_v2_common import config,view_for,lr_step,first_epoch_release
from phase2_h2_model import H2Geo,optimizer,lr_epoch
from phase2_h2_data import standardization,load_part,to_device,arguments,cached
from phase2_h2_pilot_worker import seed,peaks,train_step
from phase2_residual_math import residual_target
from phase2_h1_worker_v2 import save,sha
def run(root,jobid):
c,pilot=config(root);j=c['jobs'][jobid];view=view_for(c,pilot,j);f=view['folds'][0];out=root/'jobs'/str(jobid)/'worker';out.mkdir(exist_ok=False);(out/'history').mkdir()
os.sched_setaffinity(0,j['worker_cpus']);torch.set_num_threads(1);torch.set_num_interop_threads(1)
start=time.monotonic();cpu=time.process_time();steps=0;epoch=0;stage='SETUP';exposures=collections.Counter();nodes_exposed=0;edges_exposed=0;checkpoints=[];val_checks=0
def record(extra,final=False):
d=dict(stage=stage,job=jobid,pool=j['pool'],mode=j['mode'],epoch=epoch,optimizer_steps=steps,nodes_exposed=nodes_exposed,edges_exposed=edges_exposed,wall_seconds=time.monotonic()-start,cpu_seconds=time.process_time()-cpu,**extra,**peaks());save(out/'progress.json',d)
if final:save(out/'history'/f'{epoch:04d}.json',d)
return d
def checkpoint(tag):
path=out/(tag+'.pt');torch.save(dict(model=model.state_dict(),optimizer=opt.state_dict(),completed_epochs=epoch,optimizer_steps=steps,seed=0,fold=0,pool=j['pool'],mode=j['mode'],contract_sha256=sha(root/'contract.json'),role=tag),path)
checkpoints.append(dict(tag=tag,steps=steps,path=str(path),sha256=sha(path),standardization=str(out/'standardization.json'),standardization_sha256=sha(out/'standardization.json')))
def validation():
nonlocal val_checks
begin=time.monotonic();model.eval();loss=0.;n=0
with torch.no_grad():
for case in f['validation']:
d=to_device(load_part(view,case,stats,'validation',out),'cuda');p=model(*arguments(d),checkpoint_blocks=False)
if not bool(torch.isfinite(p).all()):raise ValueError('Nonfinite validation output')
with (out/'validation-target-access.jsonl').open('a') as log:
import json
log.write(json.dumps(dict(case=case,steps=steps,role='validation_diagnostic_only',target_sha256=view['parts'][case]['output_sha256']['target.npy']))+'\n')
target=residual_target(cached(view,case,'target'),d['F'],f['c'],fit_dtype='float32');delta=p.cpu().numpy().astype(np.float64)-target;loss+=float(delta@delta);n+=d['nodes'];del d,p
torch.cuda.synchronize()
if n!=f['validation_nodes']:raise ValueError('Validation coverage')
val_checks+=1;v=dict(steps=steps,validation_mse=loss/n,validation_nodes=n,validation_seconds=time.monotonic()-begin,validation_used_for_selection=False);save(out/f'validation-{steps:06d}.json',v);return v
try:
if os.environ.get('CUDA_VISIBLE_DEVICES')!=j['gpu_uuid'] or os.environ.get('PYTORCH_CUDA_ALLOC_CONF')!='expandable_segments:True':raise ValueError('GPU allocator identity')
stats=standardization(view,out)
torch.backends.cuda.matmul.allow_tf32=False;torch.backends.cudnn.allow_tf32=False;torch.cuda.set_device(0);torch.cuda.init();torch.cuda.reset_peak_memory_stats();seed();model=H2Geo().cuda();opt=optimizer(model)
save(out/'environment.json',dict(torch=torch.__version__,gpu_uuid=j['gpu_uuid'],allocator_settings=str(torch.cuda.memory._snapshot().get('allocator_settings',{})),seed=0,fresh_initialization=True,model_dtype='FP32',autocast=False))
while steps<j['target_steps']:
epoch+=1;stage='TRAIN';begin=time.monotonic();seen=0;enodes=0;loss_sum=0.;validation_seconds=0.
order=np.random.default_rng(np.random.SeedSequence([0,epoch])).permutation(len(f['learning']))
for index in order:
case=f['learning'][int(index)];lr=lr_epoch(epoch) if j['mode']=='epoch' else lr_step(steps+1,total=j['target_steps'])
for group in opt.param_groups:group['lr']=lr
d=to_device(load_part(view,case,stats,'learning',out),'cuda');res=train_step(model,opt,d,len(f['learning']),f['learning_nodes']);steps+=1;seen+=1;enodes+=d['nodes'];nodes_exposed+=d['nodes'];edges_exposed+=d['directed_edges'];exposures[case]+=1;loss_sum+=res['loss'];del d
record(dict(case=case,epoch_parts=seen,epoch_nodes=enodes,lr=lr,loss=res['loss']))
if j['mode']=='step' and steps==j['target_steps']:checkpoint('U' if steps==c['U'] else '2U')
if j['mode']=='step' and steps%3200==0:validation_seconds+=validation()['validation_seconds']
if steps==j['target_steps']:break
train_seconds=time.monotonic()-begin-validation_seconds
if j['mode']=='epoch' and epoch%10==0:validation_seconds+=validation()['validation_seconds']
full_epoch=seen==len(f['learning'])
if full_epoch and enodes!=f['learning_nodes']:raise ValueError('Epoch full-node identity')
stage='EPOCH_COMPLETE';record(dict(epoch_parts=seen,epoch_nodes=enodes,full_epoch=full_epoch,train_epoch_seconds=train_seconds,validation_seconds=validation_seconds,train_online_mean_weighted_loss=loss_sum/seen,not_frozen_model_training_domain_error=True),final=True)
if epoch==1:
stage='WAIT_FIRST_EPOCH_PROJECTION'
while not first_epoch_release(root,jobid):
record(dict(waiting=True));time.sleep(1)
if (root/'global-stop.json').exists():raise RuntimeError('Global item1 stop')
if j['mode']=='epoch':checkpoint('epoch200')
expected=['epoch200'] if j['mode']=='epoch' else ['U' if j['target_steps']==c['U'] else '2U']
if [x['tag'] for x in checkpoints]!=expected:raise ValueError('Checkpoint coverage')
save(out/'exposures.json',dict(per_part=dict(exposures),nodes_exposed=nodes_exposed,edges_exposed=edges_exposed,steps=steps,common_steps_not_equal_compute=True))
save(out/'result.json',dict(status='COMPLETE',job=jobid,pool=j['pool'],mode=j['mode'],completed_epochs=epoch,optimizer_steps=steps,checkpoints=checkpoints,validation_checks=val_checks,standardization_sha256=sha(out/'standardization.json'),exposures_sha256=sha(out/'exposures.json'),test_targets_read=False,test_prediction_started=False,worker_wall_seconds=time.monotonic()-start,worker_cpu_seconds=time.process_time()-cpu,**peaks()))
except BaseException as exc:
save(out/'failure.json',dict(status='FAILED',job=jobid,stage=stage,epoch=epoch,optimizer_steps=steps,reason=str(exc),traceback=traceback.format_exc(),**(peaks() if torch.cuda.is_initialized() else {})));raise
if __name__=='__main__':run(Path(sys.argv[1]),int(sys.argv[2]))
@@ -0,0 +1,34 @@
"""Metadata-only preparation of Claude8643/8710 amendment, preserving unrun v1."""
import json,time,shutil,copy
from pathlib import Path
from phase2_h1_worker_v2 import checked,sha,save
from phase2_h2_prepare import closure
from phase2_home_freeze_io import verify,freeze
from phase3_data_sufficiency_v2_common import config,execution_command
B=Path('/home/frank14f/optane/MeltMind');R=Path('/home/frank14f/MeltMind-runs/phase3-data-sufficiency-v2-2026-11-18');O=R.parent/'phase3-data-sufficiency-2026-11-18'
def read(p):return json.loads(Path(p).read_text())
def main():
import psutil
start=time.process_time()
if R.exists() or (O/'started.json').exists():raise FileExistsError('Existing newroot or started oldroot')
if any(execution_command(p.info['cmdline'] or []) for p in psutil.process_iter(['cmdline'])):raise RuntimeError('Existing v2 task')
verify(O,'implementation');old=read(O/'contract.json');c=copy.deepcopy(old)
for p,h in old['source_sha256'].items():checked(p,h)
p0=read(O/'prelaunch-projection.json');t=p0['source_max_train_epoch_seconds'];v=p0['source_max_validation_seconds'];jobs=[];projections=[]
for i,(pool,mode,steps,cap) in enumerate([(25,'epoch',22200,11000),(50,'epoch',38000,17000),(50,'step',32000,14000),(50,'step',64000,28400),(100,'step',32000,14000),(100,'step',64000,28400)]):
f=c['pools'][str(pool)];n=len(f['learning']);f0=c['pools']['100'];ratio=max(f['learning_nodes']/f0['learning_nodes'],f['learning_edges']/f0['learning_edges'],n/320);checks=20 if mode=='epoch' else steps//3200;est=1.25*(t*ratio/n*steps+checks*v)+180;j=copy.deepcopy(old['jobs'][i%4]);j.update(id=i,pool=pool,mode=mode,step_budget=steps if mode=='step' else None,target_steps=steps,learning_parts=n,learning_nodes=f['learning_nodes'],learning_edges=f['learning_edges'],gpu_seconds=cap,cpu_seconds=cap,prelaunch_projected_gpu_seconds=est);jobs.append(j);projections.append(dict(id=i,pool=pool,mode=mode,steps=steps,expected_new_gpu_seconds=est,job_cap=cap,within_cap=est<=cap,validation_checks=checks))
total=sum(x['expected_new_gpu_seconds'] for x in projections)+7200
assert all(x['within_cap'] for x in projections) and total<=120000
c.update(version='phase3-item1-training-v2',approval='1794948643.973609',budget_approval='1794948710.366709',new_fits=6,jobs=jobs,item_gpu_cap=120000,approved_item_gpu_cap=120000,train_gpu_cap=112800,postfit_gpu_reserve=7200,supersedes_unstarted_root=str(O),supersedes_implementation='c8e2ee4ca3d08cf0080ac18a1d17bfb42fb1b8a6',schedule='Each step fit freshly initialized; warmup3200 then cosine to its own U/2U endpoint',interpretation='Update-budget effect includes complete schedule change; not pure common-trajectory continuation')
c['source_sha256'][str(O/'contract.json')]=sha(O/'contract.json');c['source_sha256'][str(O/'implementation-git-freeze.json')]=sha(O/'implementation-git-freeze.json')
comp=read(O/'comparison-contract.json');comp.update(revision='Claude1794948643 complete-schedule amendment to e24fff7c',schedule='Four separate fresh step-budget fits, own complete warmup3200+cosine endpoint',update_effect='joint update budget plus complete LR schedule, not pure continuation',new_fits=6,old_entry_never_started=True)
R.mkdir();save(R/'comparison-contract.json',comp);c['comparison_contract']=str(R/'comparison-contract.json');c['comparison_contract_sha256']=sha(R/'comparison-contract.json')
codes={str(B/p):d for p,d in closure(['phase3_data_sufficiency_v2_common','phase3_data_sufficiency_v2_worker','phase3_data_sufficiency_v2_guard','phase3_data_sufficiency_v2_queue','prepare_phase3_data_sufficiency_v2','test_phase3_data_sufficiency_v2']).items()};doc=B/'docs/phase3-data-sufficiency-v2-entry-2026-11-18.md';codes[str(doc)]=sha(doc);c['code_sha256']=codes;save(R/'contract.json',c)
shutil.copyfile('/tmp/phase3-data-sufficiency-v2-tests1.log',R/'synthetic-attempt1.log');assert (R/'synthetic-attempt1.log').read_text().rstrip().endswith('OK');save(R/'acceptance.json',dict(status='PASS',tests=16,scope='CPU schedule/dispatch/GPU ownership/partial traversal/cost/no-double-count/stop-gate checks',test_targets_read=False,GPU_workload=False))
save(R/'prelaunch-projection.json',dict(cells=projections,total_gpu_seconds=total,postfit_reserve=7200,item_cap=120000,phase3_cap=612000,source_max_epoch=t,source_max_validation=v,safety_factor=1.25,actual_run=False))
save(R/'amendment.json',dict(science_ruling='1794948643.973609',budget_ruling='1794948710.366709',old_entry=str(O),old_implementation=c['supersedes_implementation'],old_started=False,old_common_trajectory_preserved=True,new_complete_schedules=True,new_fits=6,original_e24_spec_not_unchanged=True))
save(R/'preparation-audit.json',dict(status='PREPARED_NOT_STARTED',metadata_cpu_seconds=time.process_time()-start,source_pins_verified=True,old_root_unstarted=True,no_target_arrays=True,no_GPU_workload=True))
freeze(R,'implementation',[R/'contract.json',R/'comparison-contract.json',R/'acceptance.json',R/'prelaunch-projection.json',R/'amendment.json',R/'preparation-audit.json',R/'synthetic-attempt1.log',*[Path(p) for p in codes]])
binding=dict(reviewed_commit=verify(R,'implementation')['commit'],contract_sha256=sha(R/'contract.json'),implementation_proof_sha256=sha(R/'implementation-git-freeze.json'),comparison_contract_sha256=c['comparison_contract_sha256'],code_sha256=codes);save(R/'expected-review-binding.json',binding);freeze(R,'review-binding',[R/'contract.json',R/'comparison-contract.json',R/'expected-review-binding.json',R/'implementation-git-freeze.json']);config(R)
print(json.dumps(dict(root=str(R),implementation=binding['reviewed_commit'],contract_sha256=binding['contract_sha256'],binding_sha256=sha(R/'expected-review-binding.json'),projected_gpu_seconds=total,new_fits=6,started=False)))
if __name__=='__main__':main()
+35
View File
@@ -0,0 +1,35 @@
import unittest,tempfile
from pathlib import Path
from phase3_data_sufficiency_v2_common import lr_step,project_cell,project_total,first_epoch_release,dispatchable
class V2Tests(unittest.TestCase):
def test_complete_schedule_both_budgets(self):
for end in [32000,64000]:
self.assertAlmostEqual(lr_step(1,end),1e-4);self.assertAlmostEqual(lr_step(3200,end),1e-3);self.assertAlmostEqual(lr_step(end,end),1e-5)
def test_distinct_U_schedule(self):self.assertNotEqual(lr_step(32000,32000),lr_step(32000,64000))
def test_common_warmup(self):self.assertEqual(lr_step(1600,32000),lr_step(1600,64000))
def test_partial_epoch(self):self.assertEqual(divmod(32000,190),(168,80));self.assertEqual(divmod(64000,190),(336,160))
def test_invalid_step(self):
for end in [32000,64000]:
with self.assertRaises(ValueError):lr_step(end+1,end)
def jobs(self):return [dict(id=i,pool=50,mode='step',target_steps=32000,learning_parts=190,validation_seconds_bound=10,gpu_seconds=14000,prelaunch_projected_gpu_seconds=12000,gpu_uuid=str(i%4)) for i in range(6)]
def c(self):return dict(jobs=self.jobs(),postfit_gpu_reserve=7200,item_gpu_cap=120000,phase3_gpu_cap=612000)
def test_initial_four_dispatch(self):self.assertEqual([j['id'] for j in dispatchable(self.jobs(),[])],[0,1,2,3])
def test_gpu_ownership(self):self.assertEqual([j['id'] for j in dispatchable(self.jobs()[4:],[self.jobs()[1],self.jobs()[2],self.jobs()[3]])],[4])
def test_no_duplicate_gpu(self):
with self.assertRaises(ValueError):dispatchable([], [self.jobs()[0],self.jobs()[4]])
def test_no_more_than_four(self):self.assertEqual(dispatchable(self.jobs()[4:],self.jobs()[:4]),[])
def test_pending_projection_charged(self):self.assertEqual(project_total(self.c(),{}, {})['projected_item_gpu_seconds'],79200)
def test_terminal_not_double_counted(self):self.assertEqual(project_total(self.c(),{0:dict(projected_gpu_seconds=13500)},{0:dict(gpu_seconds_charged=10000)})['projected_item_gpu_seconds'],77200)
def test_cell_cap(self):
p=project_cell(self.jobs()[0],dict(steps=190,max_epoch_seconds=200,max_validation_seconds=0,charged=100));self.assertFalse(p['within_cap'])
def test_total_cap(self):
c=self.c();c['item_gpu_cap']=79000;self.assertEqual(project_total(c,{}, {})['status'],'STOP_BUDGET_PROJECTION')
def test_nan_cost(self):
with self.assertRaises(ValueError):project_cell(self.jobs()[0],dict(steps=190,max_epoch_seconds=float('nan'),max_validation_seconds=0,charged=100))
def test_cell_specific_gate(self):
with tempfile.TemporaryDirectory() as p:self.assertFalse(first_epoch_release(Path(p),4))
def test_global_stop(self):
with tempfile.TemporaryDirectory() as p:
r=Path(p);(r/'global-stop.json').write_text('{}')
with self.assertRaises(RuntimeError):first_epoch_release(r,0)
if __name__=='__main__':unittest.main()