{ "bp": { "log": { "train_loss": [ 1.9354850341033936, 1.7026564234542847, 1.6196284805679322, 1.5522706447601318, 1.5103712742996216, 1.4760794692230224, 1.4480401036453248, 1.4238936393356323, 1.3998644092941284, 1.3888274967193603, 1.370794108505249, 1.3523651966094972, 1.3417378455352784, 1.32940377494812, 1.3123712255859374, 1.3025306620025634, 1.2916712799072265, 1.277412696762085, 1.2670877025604248, 1.2579533868026733, 1.250452767906189, 1.2388696469116212, 1.2288407207870484, 1.218021198196411, 1.2081415139007567, 1.2038527836227417, 1.1919150150299072, 1.1843040127944946, 1.1809553996276856, 1.1748971781539916, 1.1629332243728638, 1.1559197056579589, 1.1501819496154786, 1.1386108488464355, 1.1359955614471435, 1.1256131409072876, 1.1200060926818847, 1.114403334388733, 1.1042931074905395, 1.1033966439437866, 1.0955395514297486, 1.086461291847229, 1.076456173362732, 1.0776306005477905, 1.0669021976089477, 1.0603959639167786, 1.0562950397491455, 1.0486062710762023, 1.037363448638916, 1.0395437144470214, 1.0320540841293335, 1.0248265104293823, 1.0186307384872437, 1.01219017370224, 1.0086890588760375, 1.0036477170944214, 0.9918886923408509, 0.9959744847869874, 0.983810892829895, 0.9749429069328308, 0.9761281036186218, 0.9669857545852661, 0.9646618202209473, 0.9585257618713379, 0.951733235244751, 0.9470721603393555, 0.9449367933654785, 0.9399459014511108, 0.9332768785476685, 0.9295426531982421, 0.9249779984283447, 0.9206048580551147, 0.9163932028198242, 0.9137818552780151, 0.9054667686653137, 0.8992140997695923, 0.8981155136108399, 0.9007723778915405, 0.8907698515319824, 0.8869161108398438, 0.8879233457946777, 0.8870189405632019, 0.8807708535385131, 0.875256696395874, 0.8748971775817871, 0.8757942477035523, 0.8740769153022766, 0.8715347571182251, 0.8681148331069947, 0.8674819102478027, 0.863512946510315, 0.8696056005477906, 0.8647798666381836, 0.8631498766708374, 0.866932133693695, 0.8609341511917115, 0.8584218189620971, 0.8627110484886169, 0.8589248524093628, 0.8619690019416809 ], "train_acc": [ 0.31324, 0.3856, 0.41254, 0.43722, 0.4538, 0.46698, 0.47636, 0.48598, 0.49418, 0.49864, 0.50738, 0.51122, 0.51678, 0.52288, 0.5272, 0.53212, 0.53512, 0.53952, 0.5452, 0.54694, 0.54924, 0.55342, 0.559, 0.56202, 0.56542, 0.56722, 0.57134, 0.57538, 0.57742, 0.57586, 0.58118, 0.58464, 0.5859, 0.59246, 0.59194, 0.59674, 0.59898, 0.60034, 0.6066, 0.60422, 0.60806, 0.61212, 0.61618, 0.61372, 0.61884, 0.6198, 0.62164, 0.62388, 0.6294, 0.62744, 0.63112, 0.63436, 0.63332, 0.63632, 0.6381, 0.6401, 0.6463, 0.64188, 0.64658, 0.65074, 0.65084, 0.65332, 0.65314, 0.65706, 0.65862, 0.66132, 0.66338, 0.66192, 0.6661, 0.6698, 0.66802, 0.66974, 0.67114, 0.67274, 0.6753, 0.67902, 0.6777, 0.67944, 0.68172, 0.68452, 0.68178, 0.68314, 0.68652, 0.68794, 0.688, 0.6866, 0.68708, 0.6896, 0.69086, 0.69196, 0.69182, 0.68984, 0.69014, 0.69196, 0.69008, 0.69314, 0.69342, 0.69112, 0.6941, 0.69128 ], "test_acc": [ 0.379, 0.42, 0.4414, 0.468, 0.4821, 0.4867, 0.5048, 0.5076, 0.5032, 0.5071, 0.5251, 0.5385, 0.5339, 0.5395, 0.5359, 0.5373, 0.5411, 0.5429, 0.544, 0.5558, 0.5584, 0.553, 0.5587, 0.557, 0.5617, 0.5595, 0.5661, 0.5732, 0.5671, 0.5738, 0.5715, 0.5762, 0.5734, 0.5772, 0.5787, 0.5735, 0.5753, 0.5788, 0.5757, 0.5779, 0.5818, 0.5857, 0.5877, 0.5853, 0.5807, 0.5926, 0.5907, 0.5921, 0.5863, 0.5905, 0.5941, 0.5959, 0.5907, 0.5931, 0.5922, 0.5982, 0.5976, 0.5888, 0.595, 0.5987, 0.6, 0.594, 0.5975, 0.599, 0.5987, 0.5921, 0.5957, 0.5921, 0.5999, 0.5947, 0.5996, 0.5976, 0.5949, 0.5992, 0.5982, 0.5956, 0.6015, 0.6011, 0.5993, 0.601, 0.5974, 0.5976, 0.5997, 0.599, 0.5988, 0.6011, 0.5974, 0.5985, 0.6018, 0.598, 0.5993, 0.5991, 0.599, 0.5992, 0.5991, 0.6004, 0.5994, 0.5993, 0.5994, 0.5993 ] }, "diagnostics": { "bp_cosine": [ 1.0, 1.0 ], "perturbation_rho": [ 0.9868886470794678, 0.9831097722053528 ], "nudging": { "0.001": [ -0.0010212662164121866, -0.0008843992254696786 ], "0.003": [ -0.0030618617311120033, -0.0026517128571867943 ], "0.01": [ -0.010182719677686691, -0.008822238072752953 ] } }, "drift": { "embed.weight": 8.850352053801725, "embed.bias": 26.80721203618006, "blocks.0.ln.weight": 0.20347274443339167, "blocks.0.w1.weight": 4.629616202784868, "blocks.0.w1.bias": 7.3007028332357145, "blocks.0.w2.weight": 12.573673444728787, "blocks.1.ln.weight": 0.15135983015831284, "blocks.1.w1.weight": 4.835127700607666, "blocks.1.w1.bias": 4.074910897523138, "blocks.1.w2.weight": 15.468163387197967, "out_ln.weight": 0.43266956541456525, "out_head.weight": 2.288095707738499, "out_head.bias": 1.5696642748538316 } }, "dfa": { "log": { "train_loss": [ 2.0506517180633543, 2.037630374984741, 2.0375006256103516, 2.0347165871810913, 2.0280541705322266, 2.0247886113739013, 2.023957159957886, 2.017765091934204, 2.013640316925049, 2.0121789574813844, 2.012694746131897, 2.0094316165542603, 2.010430689163208, 2.0080650563812257, 2.004681011276245, 2.0067793479156495, 2.002835368652344, 2.0035605745697023, 2.005288227157593, 2.0020996487045286, 2.0037673551940918, 2.0026556689453123, 2.0015291195678713, 1.9998969723129272, 2.001053342666626, 1.9988280670166017, 1.998517660293579, 2.000125762634277, 1.9984248222351073, 2.0001446187973024, 1.997779655456543, 1.9962129680633545, 1.996638311729431, 1.9934515882873536, 1.9944385040664674, 1.9950137895965576, 1.9948566608428955, 1.9947254082489014, 1.9938683570861817, 1.993011798019409, 1.993586630821228, 1.9928539782333374, 1.99163732421875, 1.9935448779296876, 1.992500286026001, 1.9926587213134765, 1.9939856634140014, 1.9904723690795898, 1.99148487449646, 1.989919489364624, 1.9883176875305175, 1.9909399053192138, 1.9917929566955566, 1.9895172406005859, 1.9915735732650757, 1.98892810256958, 1.9902667702484131, 1.9884299784851074, 1.9891859377288819, 1.9871826519012452, 1.9888171925354003, 1.9876050499725342, 1.9885715426635742, 1.9879224799346924, 1.9875414797210693, 1.9883081774902345, 1.9874149275970459, 1.9873363061904907, 1.9844847861480712, 1.9853876978302003, 1.9853303070449828, 1.986023975753784, 1.9871168149566651, 1.9860221238708495, 1.9863739596176146, 1.9852040929412842, 1.9818872883605958, 1.9874637450408936, 1.9860291708374023, 1.9853426343154907, 1.9848574941253663, 1.985159556503296, 1.9853443636322021, 1.9854941542053222, 1.9842009844970703, 1.9843328964614868, 1.9830964395141601, 1.9839789056777954, 1.9838400217437744, 1.9850177837371825, 1.9836825035858154, 1.9848054162597657, 1.9832133966827392, 1.9829493893432617, 1.9810011239242553, 1.9834887173843383, 1.9844633283233644, 1.9830566509628296, 1.983135018005371, 1.9840513249969483 ], "train_acc": [ 0.24914, 0.2591, 0.25864, 0.25854, 0.26322, 0.26478, 0.2666, 0.2671, 0.27048, 0.26588, 0.26632, 0.27096, 0.26978, 0.2698, 0.27074, 0.27022, 0.27292, 0.27118, 0.27308, 0.27424, 0.27086, 0.27322, 0.27464, 0.2738, 0.27538, 0.27228, 0.2751, 0.27742, 0.27544, 0.2723, 0.27602, 0.27634, 0.27544, 0.27646, 0.27732, 0.27708, 0.27726, 0.2762, 0.27976, 0.27826, 0.27656, 0.2791, 0.28084, 0.27786, 0.27708, 0.27874, 0.27834, 0.28096, 0.27838, 0.28342, 0.27998, 0.27926, 0.27846, 0.2797, 0.27926, 0.28266, 0.2793, 0.28242, 0.27896, 0.27938, 0.2807, 0.28146, 0.27946, 0.279, 0.28374, 0.28138, 0.28202, 0.28122, 0.2837, 0.28122, 0.28228, 0.2817, 0.27958, 0.283, 0.2825, 0.28446, 0.28486, 0.28054, 0.28292, 0.28036, 0.28328, 0.28302, 0.28552, 0.2844, 0.28334, 0.28374, 0.28518, 0.28306, 0.2825, 0.28426, 0.28316, 0.28418, 0.28372, 0.285, 0.28484, 0.2843, 0.28284, 0.28358, 0.28314, 0.28086 ], "test_acc": [ 0.2818, 0.2871, 0.2791, 0.2853, 0.2856, 0.2861, 0.3039, 0.2842, 0.2723, 0.2875, 0.2911, 0.2934, 0.3073, 0.3023, 0.2854, 0.2884, 0.3049, 0.2901, 0.2934, 0.3139, 0.3034, 0.3021, 0.321, 0.2958, 0.3122, 0.2967, 0.3089, 0.3045, 0.2876, 0.2966, 0.3042, 0.2978, 0.3161, 0.3091, 0.3033, 0.3149, 0.3106, 0.2978, 0.3124, 0.3057, 0.3115, 0.2996, 0.3098, 0.303, 0.2889, 0.3177, 0.3011, 0.2992, 0.3019, 0.3108, 0.3212, 0.3085, 0.3138, 0.3106, 0.296, 0.3106, 0.3058, 0.3123, 0.3076, 0.3118, 0.3063, 0.3124, 0.3079, 0.3077, 0.3123, 0.3141, 0.3109, 0.319, 0.3017, 0.3049, 0.3113, 0.318, 0.3157, 0.3079, 0.3101, 0.3137, 0.3058, 0.3153, 0.3094, 0.3139, 0.3087, 0.3128, 0.3162, 0.3133, 0.3087, 0.3134, 0.3167, 0.3121, 0.3118, 0.3111, 0.3131, 0.3113, 0.3128, 0.3125, 0.3131, 0.3123, 0.3118, 0.312, 0.3121, 0.3122 ] }, "diagnostics": { "bp_cosine": [ 0.396158367395401, -0.005014405585825443 ], "perturbation_rho": [ 0.0009753962513059378, 0.0 ], "nudging": { "0.001": [ -4.3818727135658264e-07, 0.0 ], "0.003": [ -1.3783574104309082e-06, 9.313225746154785e-10 ], "0.01": [ -4.818197339773178e-06, 3.725290298461914e-09 ] } }, "drift": { "embed.weight": 323.8257144939098, "embed.bias": 248.21108220164808, "blocks.0.ln.weight": 8.42464670192897, "blocks.0.w1.weight": 206.7713721666247, "blocks.0.w1.bias": 171.8292028267183, "blocks.0.w2.weight": 323.69930361812675, "blocks.1.ln.weight": 6.518412820464622, "blocks.1.w1.weight": 229.69399525127193, "blocks.1.w1.bias": 219.11418017905737, "blocks.1.w2.weight": 202.46482559231447, "out_ln.weight": 0.4266951230953523, "out_head.weight": 6.867727427000321, "out_head.bias": 3.719577118675905 } }, "credit_bridge": { "log": { "train_loss": [ 2.0466229999542236, 2.0390515814208983, 2.0387567960357664, 2.035990318336487, 2.0319369410705566, 2.0270682551574706, 2.0257353674316407, 2.01995243888855, 2.014456915893555, 2.014794263648987, 2.0092950971221923, 2.0075508432769777, 2.0089861209869384, 2.009624204673767, 2.0054416754150393, 2.004343302459717, 2.004835032348633, 2.004392629699707, 2.0029932656097413, 2.002116594848633, 2.0016485495758056, 1.9995216164398193, 1.9959283925628661, 1.9984067166900634, 1.9955141537475587, 1.9943039474105835, 1.9905847966003418, 1.9943868572235108, 1.9933354425811767, 1.9950886684799194, 1.9958439822769165, 1.9941565301132202, 1.992669308166504, 1.9904355131530762, 1.9945945461273193, 1.9895463079452516, 1.9893892168807983, 1.9876361457061769, 1.9826778482055665, 1.988116795387268, 1.9903033963012695, 1.9945041528320313, 2.003135726776123, 2.0090655849838255, 2.003611517944336, 1.993031780052185, 1.9829731557846069, 1.9795342245483398, 1.9742745175933838, 1.9710580517196656, 1.9662886399841308, 1.9650878366470337, 1.9633581380462646, 1.9602308530426025, 1.9592647447967528, 1.9569754809570312, 1.9564002374267577, 1.9596573688125611, 1.9562416759490966, 1.9577629917907715, 1.9564975534057618, 1.9539075463485718, 1.9531170993804932, 1.9562737216949464, 1.9533874069213868, 1.9565455058288574, 1.9550479524993896, 1.9579495740509034, 1.9561651504898072, 1.9566914569091798, 1.9560811403656007, 1.9559252410888672, 1.954912179260254, 1.9543887220001221, 1.9562247213363648, 1.9564899509429932, 1.9564726518249511, 1.9530779626464845, 1.954867910079956, 1.953048494796753, 1.9528486930084228, 1.9534204055404663, 1.9533309042358398, 1.952870100631714, 1.9535446406555175, 1.9520780282592773, 1.9539363479614258, 1.9533953338241576, 1.9520681017303467, 1.9507475804901122, 1.9521085049819946, 1.9542150582122804, 1.9536484621429444, 1.9531653139877319, 1.9510113228988648, 1.9543765464782714, 1.9518434842681884, 1.9509774511337281, 1.9520590380859375, 1.9515080471801758 ], "train_acc": [ 0.25224, 0.25856, 0.25666, 0.2584, 0.26252, 0.26562, 0.26418, 0.2664, 0.26886, 0.27, 0.27072, 0.27296, 0.26938, 0.27142, 0.27324, 0.27214, 0.27118, 0.27374, 0.274, 0.27416, 0.27408, 0.27516, 0.27406, 0.2768, 0.27462, 0.27758, 0.2832, 0.27908, 0.27992, 0.27762, 0.27726, 0.27918, 0.2775, 0.2774, 0.2756, 0.27898, 0.27944, 0.27822, 0.282, 0.27834, 0.28466, 0.28584, 0.28626, 0.28466, 0.2816, 0.28212, 0.28232, 0.28346, 0.28134, 0.28332, 0.2858, 0.28386, 0.28558, 0.28612, 0.28698, 0.28932, 0.2885, 0.28432, 0.2864, 0.28792, 0.28548, 0.28804, 0.28786, 0.28798, 0.2883, 0.28572, 0.2894, 0.28642, 0.28876, 0.28766, 0.28924, 0.28718, 0.28808, 0.28528, 0.2878, 0.28852, 0.2875, 0.28816, 0.28838, 0.28702, 0.29014, 0.29108, 0.28906, 0.29216, 0.29062, 0.28836, 0.29078, 0.28994, 0.28852, 0.29152, 0.29096, 0.29036, 0.29132, 0.29078, 0.29352, 0.28894, 0.29248, 0.2919, 0.2922, 0.28814 ], "test_acc": [ 0.2783, 0.2698, 0.2993, 0.282, 0.3008, 0.2636, 0.302, 0.2948, 0.2815, 0.2878, 0.2823, 0.3067, 0.3037, 0.2988, 0.2874, 0.2945, 0.2916, 0.2958, 0.3024, 0.2933, 0.2879, 0.3095, 0.3108, 0.3052, 0.3078, 0.304, 0.299, 0.2887, 0.2945, 0.2859, 0.2947, 0.3076, 0.287, 0.3072, 0.3058, 0.3008, 0.2991, 0.3036, 0.3086, 0.2928, 0.3031, 0.3113, 0.307, 0.3049, 0.3033, 0.3003, 0.3063, 0.307, 0.2942, 0.312, 0.3043, 0.3027, 0.3029, 0.3049, 0.297, 0.3117, 0.2993, 0.3076, 0.2972, 0.3089, 0.3051, 0.31, 0.3078, 0.2928, 0.3153, 0.3035, 0.3138, 0.3058, 0.312, 0.3027, 0.3098, 0.3066, 0.3106, 0.3118, 0.3022, 0.3123, 0.3051, 0.3089, 0.3057, 0.3095, 0.3091, 0.3075, 0.3085, 0.3078, 0.3102, 0.3103, 0.3092, 0.3038, 0.3111, 0.31, 0.3089, 0.3099, 0.3095, 0.3099, 0.3096, 0.3092, 0.3102, 0.3106, 0.3106, 0.3105 ], "value_loss": [ 0.5981157046318054, 0.23145676259040832, 0.1629143718481064, 0.13112473169445993, 0.1034177236700058, 0.0916099097251892, 0.09448877233743667, 0.07975816262960433, 0.07028926397323608, 0.06960906606912613, 0.0663377025961876, 0.06277827244520187, 0.06609309570908546, 0.05464596090316772, 0.05347000701904297, 0.055175553441047666, 0.05404586620092392, 0.050111453244686126, 0.04170590556740761, 0.042121976542472836, 0.04078092483282089, 0.04033791535496712, 0.041389339334964755, 0.04141822567462921, 0.039917332406044004, 0.036470176842212675, 0.03570402141869068, 0.0356453166270256, 0.0348940605866909, 0.03294261076688766, 0.030363075378537177, 0.031874004955291746, 0.028748285691142083, 0.02522519760966301, 0.025916270488500596, 0.02379499412894249, 0.0268276374822855, 0.021691660759449005, 0.019947531086206435, 0.021523804327845574, 0.02691959607720375, 0.02570752409338951, 0.02876643155694008, 0.0335148137229681, 0.031904469298124315, 0.02778498649060726, 0.02370047950387001, 0.02384015331029892, 0.022202999491095544, 0.02014792114138603, 0.01951033084630966, 0.0173978901296854, 0.015886960296034813, 0.014518849447965622, 0.016183216240406036, 0.016505281180143355, 0.014820445327758789, 0.014580657460987568, 0.012194996480941772, 0.014215488128364087, 0.013283061341047286, 0.011145545057058334, 0.011805196754932403, 0.009617018352746963, 0.01116202095746994, 0.009037217251360417, 0.0090115091329813, 0.011461790952682495, 0.008820395555198193, 0.007558346377015114, 0.00825535637140274, 0.007499891985356807, 0.006166567915007472, 0.006294274886846542, 0.006129814375638962, 0.005847277462333441, 0.004945684289932251, 0.004713377166241407, 0.0033198376420140265, 0.0037709764985740185, 0.002341279806494713, 0.002568720788434148, 0.002159126052595675, 0.001921226680725813, 0.0018878655811026693, 0.0016211944927275181, 0.0013243718447163702, 0.001169309064410627, 0.0010681760500371455, 0.0008630752410739661, 0.0008580225262790919, 0.0008040335815213621, 0.0007892505507543683, 0.0006356554404087364, 0.0006604202647879719, 0.0006046383780613541, 0.0006052350796759129, 0.0005226440225169063, 0.0005789817444421351, 0.0006147388235665858 ] }, "diagnostics": { "bp_cosine": [ 0.33032962679862976, 0.01970662735402584 ], "perturbation_rho": [ 0.06130741536617279, 0.0 ], "nudging": { "0.001": [ -5.778856575489044e-07, 0.0 ], "0.003": [ -1.7262063920497894e-06, 0.0 ], "0.01": [ -5.792360752820969e-06, 0.0 ] } }, "drift": { "embed.weight": 175.24499314321244, "embed.bias": 87.67212401922586, "blocks.0.ln.weight": 8.05369238852348, "blocks.0.w1.weight": 157.60363681616167, "blocks.0.w1.bias": 81.08793311128552, "blocks.0.w2.weight": 221.88820128646415, "blocks.1.ln.weight": 5.44449828938583, "blocks.1.w1.weight": 244.47570610251066, "blocks.1.w1.bias": 227.1893178317289, "blocks.1.w2.weight": 212.32680563546222, "out_ln.weight": 0.33305352742575556, "out_head.weight": 4.339797161618964, "out_head.bias": 2.8094622055285003 } } }