diff options
| author | YurenHao0426 <blackhao0426@gmail.com> | 2026-06-13 12:35:36 -0500 |
|---|---|---|
| committer | YurenHao0426 <blackhao0426@gmail.com> | 2026-06-13 12:35:36 -0500 |
| commit | 66e0d8b9fd4d0f7a2231d689c055e26fdf1cf04a (patch) | |
| tree | c29cba61124018755a19b02c9d33e3ad5f2e05cc /research/flossing/problem_tracks/trm_learning_events.csv | |
Curated export for clone-and-run Maze training (2x A6000) + diagnostics.
trm/hrm pretrain.py carry trajectory-augmentation code (backward-compatible).
Heavy artifacts (checkpoints/wandb/npz) gitignored; see PROVENANCE.md.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Diffstat (limited to 'research/flossing/problem_tracks/trm_learning_events.csv')
| -rw-r--r-- | research/flossing/problem_tracks/trm_learning_events.csv | 37 |
1 files changed, 37 insertions, 0 deletions
diff --git a/research/flossing/problem_tracks/trm_learning_events.csv b/research/flossing/problem_tracks/trm_learning_events.csv new file mode 100644 index 0000000..717f939 --- /dev/null +++ b/research/flossing/problem_tracks/trm_learning_events.csv @@ -0,0 +1,37 @@ +kind,from_step,to_step,from_epoch,to_epoch,event,n,from_token_acc_mean,to_token_acc_mean,lambda_max_before,lambda_max_after,lambda_max_delta,mean8_before,mean8_after,mean8_delta,tail_mean_5_8_before,tail_mean_5_8_after,tail_mean_5_8_delta,positive_count_before,positive_count_after,positive_count_delta,positive_sum_before,positive_sum_after,positive_sum_delta
+TRM,26041,52082,5000,10000,learned_fail_to_success,61,0.6844768270117337,1.0,0.04095346362856751,0.01130250356299421,-0.0296509600655733,0.015569461502346697,-0.0039388805430984365,-0.019508342045445135,0.00450889594701978,-0.009700431184056902,-0.014209327131076683,5.934426229508197,2.0163934426229506,-3.918032786885246,0.14043893001830018,0.027653619372805954,-0.11278531064549423
+TRM,26041,52082,5000,10000,lost_success_to_fail,24,1.0,0.7021604975064596,-0.017263541842112318,0.06876462713504831,0.08602816897716063,-0.03177595781517084,0.03872489225962757,0.07050085007479841,-0.03782351667450712,0.025608245707492944,0.06343176238200006,0.7916666666666666,7.875,7.083333333333333,0.00850467175647888,0.31008046285569435,0.30157579109921545
+TRM,26041,52082,5000,10000,stayed_failure,156,0.6374643859572899,0.6431623934171139,0.031975116116555936,0.05863720388151705,0.02666208776496112,0.010370061944551404,0.03239625788568773,0.022026195941136324,0.0011294046405276813,0.02098523936449335,0.01985583472396567,5.384615384615385,7.923076923076923,2.5384615384615383,0.10369389380041237,0.2593554582230214,0.15566156442260903
+TRM,26041,52082,5000,10000,stayed_success,271,1.0,1.0,-0.03639817160659047,-0.0026813522725212696,0.0337168193340692,-0.04572985614410462,-0.010540279390967392,0.03518957675313723,-0.0497526878224062,-0.01376034084844453,0.03599234697396168,0.11439114391143912,0.7564575645756457,0.6420664206642066,0.0016417544531873067,0.007957042507742226,0.006315288054554919
+TRM,52082,78123,10000,15000,learned_fail_to_success,45,0.6886145393053691,1.0,0.06373828107284175,0.01469134650590907,-0.04904693456693268,0.036088030263393495,-0.01778541501991842,-0.053873445283311915,0.0238580148266111,-0.026259723155009043,-0.05011773798162014,8.0,1.7333333333333334,-6.266666666666667,0.28870424210714796,0.0387454634739293,-0.24995877863321866
+TRM,52082,78123,10000,15000,lost_success_to_fail,25,1.0,0.6765432119369507,0.013729429183294997,0.08733505010604858,0.07360562092275359,-0.0023064174062164965,0.051225730990990995,0.05353214839720749,-0.008355203102109954,0.036106420699507,0.04446162380161695,2.24,8.0,5.76,0.04125833526602946,0.40980584792792796,0.3685475126618985
+TRM,52082,78123,10000,15000,stayed_failure,135,0.6385002299591347,0.63950617313385,0.05873727561836993,0.08107733505743521,0.02234005943906528,0.03229075764848622,0.04857104088231507,0.01628028323382885,0.020849515338098475,0.03474471943832382,0.013895204100225345,7.888888888888889,8.0,0.1111111111111111,0.25859030886300993,0.38856832705852057,0.12997801819551064
+TRM,52082,78123,10000,15000,stayed_success,307,1.0,1.0,-0.0012391839677296163,-0.0045661997550452684,-0.003327015787315652,-0.009899110726142527,-0.032767963419847765,-0.022868852693705238,-0.013393806497078794,-0.03904287366627904,-0.025649067169200245,0.8859934853420195,0.504885993485342,-0.3811074918566775,0.00915886293058166,0.007548392093377316,-0.001610470837204344
+TRM,78123,104164,15000,20000,learned_fail_to_success,42,0.6748971215316227,1.0,0.08431322872638702,0.032027611919190915,-0.05228561680719611,0.05053495965521073,0.006543990983897986,-0.04399096867131275,0.03605126524122343,-0.0033311979622383313,-0.03938246320346176,8.0,3.6904761904761907,-4.309523809523809,0.40427967724168584,0.10858904316375542,-0.29569063407793045
+TRM,78123,104164,15000,20000,lost_success_to_fail,30,1.0,0.6847736616929372,0.011231523806297142,0.09582423393925031,0.08459271013295316,-0.020423608111771803,0.05470764508160452,0.07513125319337632,-0.028769394504100394,0.03790804650634527,0.06667744101044566,1.4333333333333333,8.0,6.566666666666666,0.028053055804533263,0.43766116065283617,0.4096081048483029
+TRM,78123,104164,15000,20000,stayed_failure,118,0.6347562251454693,0.6352793450577784,0.08125136338048063,0.0946442298591137,0.013392866478633072,0.048434453291258124,0.05644692583463423,0.008012472543376102,0.034568173741779715,0.04056167824912059,0.00599350450734087,8.0,8.0,0.0,0.387475626330065,0.45157540667707385,0.06409978034700882
+TRM,78123,104164,15000,20000,stayed_success,322,1.0,1.0,-0.0033467653609375882,-0.0028673611494872695,0.000479404211450319,-0.03182422423116906,-0.01583997491176961,0.01598424931939945,-0.038213564976397704,-0.020403333244893535,0.017810231731504166,0.5900621118012422,0.7204968944099379,0.13043478260869565,0.009997858866017569,0.013265862348969405,0.0032680034829518376
+TRM,104164,130205,20000,25000,learned_fail_to_success,50,0.6804938280582428,1.0,0.09401120468974114,0.03770299475596403,-0.05630820993377711,0.055930792740546166,0.012810704448684191,-0.04312008829186197,0.03986294572241604,0.0028754143451988055,-0.03698753137721724,8.0,5.06,-2.94,0.44744634192436933,0.1199192828458763,-0.32752705907849305
+TRM,104164,130205,20000,25000,lost_success_to_fail,27,1.0,0.6575217269085072,0.0314528225765874,0.10427529282040066,0.07282247024381326,0.0068112800104750525,0.05774478058123754,0.050933500570762485,-0.002804141968630979,0.039058988379245554,0.04186313034787653,3.5185185185185186,8.0,4.481481481481482,0.11630381328804004,0.4619582446499003,0.3456544313618603
+TRM,104164,130205,20000,25000,stayed_failure,98,0.6273620527009575,0.6210632409368243,0.0953284276231211,0.10753870945499867,0.012210281831877572,0.05617782616232313,0.0605833601475009,0.004405533985177778,0.040105838188426376,0.04148375476967087,0.001377916581244493,8.0,8.0,0.0,0.449422609298585,0.4846668811800072,0.03524427188142222
+TRM,104164,130205,20000,25000,stayed_success,337,1.0,1.0,-0.0012681210655689679,0.010519292579277294,0.011787413644846262,-0.014865070802815804,-0.0011317419686626743,0.01373332883415313,-0.01968567295583588,-0.00581690880209499,0.013868764153740892,0.8664688427299704,2.0178041543026706,1.1513353115727003,0.016890636588928174,0.028340769676497318,0.011450133087569142
+TRM,130205,156246,25000,30000,learned_fail_to_success,36,0.6615226318438848,1.0,0.10779489700992902,0.0496451583728938,-0.05814973863703522,0.060211614615076944,0.0020279816988590937,-0.05818363291621785,0.040751232791485056,-0.014924997131149413,-0.05567622992263447,8.0,3.9166666666666665,-4.083333333333333,0.48169291692061555,0.11456112468239452,-0.36713179223822107
+TRM,130205,156246,25000,30000,lost_success_to_fail,35,1.0,0.6684303368840899,0.04063011546220098,0.11092394015618733,0.07029382469398635,0.013878692329477157,0.06066181986326618,0.04678312753378902,0.0034880245831248303,0.04024732473439404,0.03675930015126921,4.771428571428571,8.0,3.2285714285714286,0.13045002815852058,0.48529455890612944,0.35484453074760886
+TRM,130205,156246,25000,30000,stayed_failure,89,0.6157580841793103,0.6254681650172459,0.1064450577571151,0.10696759694412852,0.0005225391870134332,0.05987258701130036,0.059057195764427396,-0.0008153912468729645,0.04104445250757206,0.03980186048848031,-0.0012425920190917475,8.0,8.0,0.0,0.4789806960904029,0.47245756611541917,-0.006523129974983716
+TRM,130205,156246,25000,30000,stayed_success,352,1.0,1.0,0.011386640044993224,0.03174139531593441,0.020354755270941185,-0.0006437956037977619,-0.019161262890415157,-0.018517467286617396,-0.005507410254134774,-0.03439320480515045,-0.028885794551015677,2.1761363636363638,1.5028409090909092,-0.6732954545454546,0.031196171982173837,0.04352239211735051,0.012326220135176673
+TRM,156246,182287,30000,35000,learned_fail_to_success,34,0.691721136079115,1.0,0.10662735976717051,0.040389080076505816,-0.0662382796906647,0.05807228958206799,-0.005224973876813558,-0.06329726345888155,0.03843013702563065,-0.01886841586255126,-0.05729855288818191,8.0,2.3529411764705883,-5.647058823529412,0.4645783166565439,0.0783735490986146,-0.3862047675579293
+TRM,156246,182287,30000,35000,lost_success_to_fail,42,1.0,0.6590241066047123,0.05099954224472943,0.10153728688047045,0.050537744635741025,0.001467981777669333,0.04981639054208894,0.048348408764419606,-0.015194870930394072,0.02946084769584039,0.044655718626234464,3.619047619047619,7.976190476190476,4.357142857142857,0.10997597763036451,0.3985460633134824,0.2885700856831179
+TRM,156246,182287,30000,35000,stayed_failure,90,0.617146776119868,0.6224965714746051,0.10863470890455776,0.1012288083632787,-0.007405900541279051,0.06005329191620048,0.049981245352602045,-0.010072046563598431,0.040493303225634415,0.029898996461027613,-0.010594306764606801,8.0,8.0,0.0,0.4804263353296038,0.39984996282081636,-0.08057637250878745
+TRM,156246,182287,30000,35000,stayed_success,346,1.0,1.0,0.03126652045767183,0.005291063447186988,-0.025975457010484843,-0.019460729569159885,-0.020014630322294655,-0.0005539007531347702,-0.03469804453484909,-0.027679566077947228,0.007018478456901866,1.4971098265895955,0.7543352601156069,-0.7427745664739884,0.042847085125428536,0.022412010842762845,-0.020435074282665687
+TRM,182287,208328,35000,40000,learned_fail_to_success,38,0.6591942835795251,1.0,0.1029400615707824,0.0570142870082667,-0.04592577456251571,0.050415613402851465,0.011271269248601655,-0.03914434415424981,0.02967732811904218,-0.004436708284342067,-0.03411403640338425,8.0,4.657894736842105,-3.3421052631578947,0.4033249072228117,0.13436546987328543,-0.26895943734952626
+TRM,182287,208328,35000,40000,lost_success_to_fail,34,1.0,0.670297747149187,0.04045343530039164,0.10570072382688522,0.06524728852649359,-0.0032343704180963118,0.05597326002658948,0.059207630444685785,-0.017199626846430713,0.035191281082327754,0.052390907928758466,2.6176470588235294,8.0,5.382352941176471,0.08064539896464767,0.4477860802127158,0.36714068124806815
+TRM,182287,208328,35000,40000,stayed_failure,94,0.6239821397243662,0.6187286605226233,0.10067485597856501,0.10716580845257069,0.006490952474005679,0.04973199122546345,0.05547681558775676,0.005744824362293312,0.029792838470150837,0.03494465546162324,0.005151816991472399,7.98936170212766,8.0,0.010638297872340425,0.39786260466566903,0.44381452470205407,0.045951920036385034
+TRM,182287,208328,35000,40000,stayed_success,346,1.0,1.0,0.005284739523452543,0.03552449149677848,0.030239751973325932,-0.020210238754654154,-0.004939072638686333,0.01527116611596782,-0.027843551125774104,-0.016626303564166032,0.011217247561608072,0.7283236994219653,2.0838150289017343,1.3554913294797688,0.022188765480204684,0.050417376088078246,0.02822861060787356
+TRM,208328,234369,40000,45000,learned_fail_to_success,45,0.6496570653385586,1.0,0.10920602844821083,0.04854822551779863,-0.060657802930412195,0.05672489060088992,0.008596817297721928,-0.04812807330316799,0.03611175337185462,-0.004522915925513694,-0.040634669297368316,8.0,4.177777777777778,-3.8222222222222224,0.45379912480711937,0.12477201784753965,-0.3290271069595797
+TRM,208328,234369,40000,45000,lost_success_to_fail,33,1.0,0.6352413098017374,0.062286600188323944,0.10571520346583742,0.043428603277513474,0.01137912593394779,0.052789654194687806,0.041410528260740015,-0.005373023553560911,0.031694172147774334,0.037067195701335244,4.212121212121212,8.0,3.787878787878788,0.14217090340124824,0.42231723355750245,0.2801463301562542
+TRM,208328,234369,40000,45000,stayed_failure,83,0.6230849343610098,0.6302246002547712,0.10545950993356934,0.10543236639126237,-2.7143542306968965e-05,0.05500351119413411,0.052311619395861424,-0.002691891798272684,0.034412918897087585,0.0311619305242616,-0.003250988372825984,8.0,8.0,0.0,0.4400280895530729,0.4184929551668914,-0.021535134386181473
+TRM,208328,234369,40000,45000,stayed_success,351,1.0,1.0,0.03533492637602506,0.014196671783756859,-0.021138254592268202,-0.004718296459711924,-0.014446210490412838,-0.009727914030700913,-0.016364632965068193,-0.022218270485360785,-0.005853637520292594,2.1623931623931623,1.1111111111111112,-1.0512820512820513,0.050879373702047655,0.02393511410796923,-0.026944259594078424
+TRM,234369,260410,45000,50000,learned_fail_to_success,35,0.6613756656646729,1.0,0.10479770622083119,0.029873117019035687,-0.0749245892017955,0.05238420652957367,0.006787718061972165,-0.045596488467601504,0.03146511843827154,-0.001710809776568567,-0.033175928214840104,8.0,4.057142857142857,-3.942857142857143,0.4190736522365894,0.07165596114105678,-0.3474176910955326
+TRM,234369,260410,45000,50000,lost_success_to_fail,37,1.0,0.6593259943498148,0.04475835560997193,0.10352358201871047,0.05876522640873854,0.0072115822668648265,0.05163027802671029,0.044418695759845465,-0.0057618086022950986,0.030947812927282742,0.03670962152957784,4.081081081081081,7.972972972972973,3.891891891891892,0.11042228619555465,0.4130832626408822,0.3026609764453275
+TRM,234369,260410,45000,50000,stayed_failure,81,0.6188081079795037,0.6168267027831372,0.10582183218664593,0.10270964078329227,-0.003112191403353656,0.05247500950303848,0.050216694190713035,-0.0022583153123254484,0.031247762580873605,0.030064451486665617,-0.001183311094207988,8.0,8.0,0.0,0.41980007602430786,0.4017335535257043,-0.018066522498603587
+TRM,234369,260410,45000,50000,stayed_success,359,1.0,1.0,0.01535276542292656,0.011521664883778544,-0.003831100539148015,-0.01378994887914044,0.0011695116181482481,0.014959460497288688,-0.021696259717896474,-0.002636141539739297,0.019060118178157174,1.1894150417827298,3.097493036211699,1.9080779944289694,0.027661117732036104,0.025961193830398188,-0.0016999239016379185
|
