diff options
Diffstat (limited to 'results')
| -rw-r--r-- | results/cifar10/results_cifar10.json | 1985 | ||||
| -rw-r--r-- | results/cifar10_seed123/results_cifar10.json | 1985 | ||||
| -rw-r--r-- | results/cifar10_seed456/results_cifar10.json | 1985 | ||||
| -rw-r--r-- | results/toy_lq_frozen/toy_lq_v2_seed123_lam0.1_sig0.1_tgw1.0_fm0.0.json | 266 | ||||
| -rw-r--r-- | results/toy_lq_frozen/toy_lq_v2_seed42_lam0.1_sig0.1_tgw1.0_fm0.0.json | 266 | ||||
| -rw-r--r-- | results/toy_lq_frozen/toy_lq_v2_seed456_lam0.1_sig0.1_tgw1.0_fm0.0.json | 266 |
6 files changed, 6753 insertions, 0 deletions
diff --git a/results/cifar10/results_cifar10.json b/results/cifar10/results_cifar10.json new file mode 100644 index 0000000..b4bd094 --- /dev/null +++ b/results/cifar10/results_cifar10.json @@ -0,0 +1,1985 @@ +{ + "42": { + "bp": { + "log": { + "train_loss": [ + 1.9159408319854736, + 1.6833457696151732, + 1.5935267736053467, + 1.5389775912475585, + 1.4867800998687744, + 1.4524269732284545, + 1.4178237664794922, + 1.3897597776794433, + 1.3664597565460206, + 1.342870984840393, + 1.3235610486221314, + 1.301216788482666, + 1.279724592475891, + 1.2618738833999634, + 1.2407353455352783, + 1.22462412940979, + 1.2109693797302246, + 1.1892420639038086, + 1.1723702849578856, + 1.1557361126327514, + 1.1423172385787963, + 1.1261568475341797, + 1.1062135691070556, + 1.0966129437637329, + 1.0756375621795655, + 1.0581538507270813, + 1.0487863679885865, + 1.0285031553268433, + 1.0112163684082032, + 0.9978593858337402, + 0.9784134501266479, + 0.9645692346191406, + 0.9500791883468628, + 0.9311204533195495, + 0.9137258143806457, + 0.8955737284660339, + 0.8760775566101074, + 0.8636410433959961, + 0.8472155108451843, + 0.8257175942802429, + 0.8081484350395203, + 0.7949162258720398, + 0.7682366105842591, + 0.7535309395217895, + 0.7374585538101196, + 0.7163707330703736, + 0.7002532124710082, + 0.6810121598052978, + 0.6590044124794007, + 0.6416365943050385, + 0.6245126323890686, + 0.6053193593788146, + 0.5843256318569183, + 0.5672504850196839, + 0.5574265190124512, + 0.5282080096054077, + 0.5124851553916931, + 0.49809444856643675, + 0.4815664218711853, + 0.46325312936782836, + 0.44530024614334107, + 0.4232256102848053, + 0.41155227093696595, + 0.3951142164516449, + 0.3779419419193268, + 0.36394513979434967, + 0.3495154238319397, + 0.33435829617500307, + 0.3247876634025574, + 0.31244264134407046, + 0.30034732496261596, + 0.2886586371707916, + 0.2739575230407715, + 0.26273826862812044, + 0.25702780318260193, + 0.23921729731559754, + 0.2356625511932373, + 0.23087995275020598, + 0.21462623807907105, + 0.21162452461242676, + 0.20137000153541565, + 0.1957099277973175, + 0.19024162811279297, + 0.1820225635457039, + 0.17860664487838745, + 0.17204254170417785, + 0.1648883514738083, + 0.1611987146615982, + 0.15891523240566255, + 0.15874515238285064, + 0.15378751336574553, + 0.15460485486030578, + 0.1495811799764633, + 0.14360208065509797, + 0.14562011506080627, + 0.14506465085983275, + 0.1431816494703293, + 0.1397772234392166, + 0.14342590034008026, + 0.142473215675354 + ], + "train_acc": [ + 0.31126, + 0.39118, + 0.41882, + 0.44388, + 0.4617, + 0.4731, + 0.48828, + 0.49882, + 0.50712, + 0.51646, + 0.5211, + 0.53098, + 0.53964, + 0.54616, + 0.55448, + 0.55806, + 0.56566, + 0.57324, + 0.57908, + 0.58582, + 0.58824, + 0.59418, + 0.60314, + 0.60678, + 0.61408, + 0.62094, + 0.62264, + 0.63056, + 0.63836, + 0.64046, + 0.64822, + 0.65186, + 0.65832, + 0.66298, + 0.67222, + 0.67798, + 0.686, + 0.688, + 0.695, + 0.7014, + 0.709, + 0.71344, + 0.7245, + 0.72892, + 0.73524, + 0.7426, + 0.74692, + 0.75382, + 0.76378, + 0.76914, + 0.77694, + 0.78392, + 0.78984, + 0.79764, + 0.79948, + 0.81182, + 0.81536, + 0.82284, + 0.82968, + 0.83584, + 0.84158, + 0.8514, + 0.85392, + 0.86052, + 0.86604, + 0.87092, + 0.87644, + 0.88364, + 0.88486, + 0.88988, + 0.89316, + 0.89914, + 0.9048, + 0.90892, + 0.90944, + 0.91828, + 0.9189, + 0.92122, + 0.92622, + 0.92784, + 0.93044, + 0.9332, + 0.93514, + 0.93836, + 0.93948, + 0.94166, + 0.94506, + 0.9452, + 0.94606, + 0.94738, + 0.9493, + 0.94928, + 0.94976, + 0.95178, + 0.9521, + 0.95234, + 0.95228, + 0.9543, + 0.95236, + 0.9534 + ], + "test_acc": [ + 0.3908, + 0.433, + 0.4419, + 0.4858, + 0.4871, + 0.4878, + 0.5141, + 0.5211, + 0.5188, + 0.5279, + 0.5296, + 0.5402, + 0.5415, + 0.5584, + 0.5559, + 0.5648, + 0.5531, + 0.5686, + 0.5685, + 0.575, + 0.5694, + 0.5749, + 0.5711, + 0.585, + 0.5778, + 0.5875, + 0.5827, + 0.5876, + 0.5868, + 0.5947, + 0.5903, + 0.6001, + 0.5963, + 0.6013, + 0.5943, + 0.5994, + 0.5879, + 0.5925, + 0.6006, + 0.5978, + 0.6007, + 0.6034, + 0.5914, + 0.5958, + 0.5959, + 0.5959, + 0.5928, + 0.5961, + 0.6013, + 0.5978, + 0.5974, + 0.5998, + 0.5919, + 0.5936, + 0.5958, + 0.6019, + 0.5972, + 0.5927, + 0.5998, + 0.5945, + 0.5942, + 0.5984, + 0.5904, + 0.5907, + 0.595, + 0.5928, + 0.5944, + 0.5973, + 0.5924, + 0.5909, + 0.5976, + 0.5938, + 0.5943, + 0.5926, + 0.5969, + 0.5918, + 0.5943, + 0.5881, + 0.5907, + 0.5921, + 0.5934, + 0.5926, + 0.5919, + 0.5923, + 0.5912, + 0.5908, + 0.5888, + 0.5899, + 0.5904, + 0.5911, + 0.5911, + 0.5905, + 0.5894, + 0.5918, + 0.5926, + 0.5931, + 0.5935, + 0.5936, + 0.5939, + 0.594 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.9423961043357849, + 0.9423063397407532, + 0.9428478479385376, + 0.9422410726547241, + 0.9424552917480469, + 0.9417660236358643, + 0.9408372640609741, + 0.9404677152633667, + 0.9403290748596191, + 0.9396632313728333, + 0.935971736907959, + 0.9338305592536926 + ], + "perturbation_rho": [ + 0.987454891204834, + 0.9897942543029785, + 0.990141749382019, + 0.9901803731918335, + 0.9920680522918701, + 0.9906818866729736, + 0.9903223514556885, + 0.9926531910896301, + 0.9914379119873047, + 0.9897843599319458, + 0.9882909059524536, + 0.9869540929794312 + ], + "nudging": { + "0.001": [ + -0.002897711703553796, + -0.002905686851590872, + -0.0029487141873687506, + -0.0030073150992393494, + -0.002988070249557495, + -0.0028954721055924892, + -0.0027900305576622486, + -0.002700705546885729, + -0.0025893584825098515, + -0.0024207504466176033, + -0.002165877493098378, + -0.00189895648509264 + ], + "0.003": [ + -0.008684658445417881, + -0.008708733133971691, + -0.008837150409817696, + -0.009012945927679539, + -0.008955370634794235, + -0.008678246289491653, + -0.008362163789570332, + -0.008094580844044685, + -0.007761792279779911, + -0.007257224526256323, + -0.006493166089057922, + -0.005693747662007809 + ], + "0.01": [ + -0.028850272297859192, + -0.028929298743605614, + -0.029354672878980637, + -0.029939891770482063, + -0.029747068881988525, + -0.028829604387283325, + -0.027784891426563263, + -0.026898572221398354, + -0.025797637179493904, + -0.024130169302225113, + -0.02159387245774269, + -0.018940377980470657 + ] + } + }, + "drift": { + "embed.weight": 8.833701837291958, + "embed.bias": 14.261815769395275, + "blocks.0.ln.weight": 0.34404799666331826, + "blocks.0.w1.weight": 4.175094808562947, + "blocks.0.w1.bias": 6.98398106061982, + "blocks.0.w2.weight": 10.592751719664884, + "blocks.1.ln.weight": 0.2322510556310593, + "blocks.1.w1.weight": 4.523413515316886, + "blocks.1.w1.bias": 4.3749413052218085, + "blocks.1.w2.weight": 11.940958034527862, + "blocks.2.ln.weight": 0.2150404740407862, + "blocks.2.w1.weight": 4.58583359520105, + "blocks.2.w1.bias": 4.159787661269507, + "blocks.2.w2.weight": 12.401201224911981, + "blocks.3.ln.weight": 0.22107801019329978, + "blocks.3.w1.weight": 4.586995215167557, + "blocks.3.w1.bias": 3.881613437324734, + "blocks.3.w2.weight": 12.64024804795375, + "blocks.4.ln.weight": 0.23170407382172373, + "blocks.4.w1.weight": 4.578125313924808, + "blocks.4.w1.bias": 3.7932938751482683, + "blocks.4.w2.weight": 12.963501907885533, + "blocks.5.ln.weight": 0.23628225424614754, + "blocks.5.w1.weight": 4.5240873092186416, + "blocks.5.w1.bias": 3.803760854867017, + "blocks.5.w2.weight": 12.912394584749915, + "blocks.6.ln.weight": 0.2619501911907522, + "blocks.6.w1.weight": 4.461384236716512, + "blocks.6.w1.bias": 3.799496573870174, + "blocks.6.w2.weight": 13.155197671866139, + "blocks.7.ln.weight": 0.2657175505296468, + "blocks.7.w1.weight": 4.422447215896801, + "blocks.7.w1.bias": 3.725743197114046, + "blocks.7.w2.weight": 13.149302405423146, + "blocks.8.ln.weight": 0.28028080422594653, + "blocks.8.w1.weight": 4.320993108213275, + "blocks.8.w1.bias": 3.577600797368634, + "blocks.8.w2.weight": 13.112473965027567, + "blocks.9.ln.weight": 0.3122222575319159, + "blocks.9.w1.weight": 4.3011661234952125, + "blocks.9.w1.bias": 3.5701937101673704, + "blocks.9.w2.weight": 13.20746990540504, + "blocks.10.ln.weight": 0.2907587848995931, + "blocks.10.w1.weight": 4.244185315341837, + "blocks.10.w1.bias": 3.487530574891753, + "blocks.10.w2.weight": 13.342063167376029, + "blocks.11.ln.weight": 0.30672613980463925, + "blocks.11.w1.weight": 4.1911870376224645, + "blocks.11.w1.bias": 3.626239385452188, + "blocks.11.w2.weight": 13.53569165604796, + "out_ln.weight": 0.304987181878743, + "out_head.weight": 3.659824943443211, + "out_head.bias": 0.7848376848440143 + } + }, + "dfa": { + "log": { + "train_loss": [ + 2.060509294204712, + 2.0283543349456785, + 2.0286419232940673, + 2.0282608281707764, + 2.0229165492248535, + 2.0251557587432862, + 2.0200582264709475, + 2.021266340560913, + 2.0176746366119387, + 2.0216268117523195, + 2.016007042388916, + 2.0150310187530516, + 2.0130176863098144, + 2.011090052871704, + 2.0114055884552, + 2.012881319503784, + 2.0109381372070314, + 2.011416566619873, + 2.009194619445801, + 2.008145341949463, + 2.008374558181763, + 2.009201191329956, + 2.0073699200820925, + 2.0068252397155764, + 2.006670061340332, + 2.006027181663513, + 2.0085990719604494, + 2.006233550338745, + 2.007863084945679, + 2.007588200149536, + 2.0059480542755126, + 2.0044747730255126, + 2.006523677520752, + 2.0035686435317994, + 2.0047321628189088, + 2.005557894821167, + 2.003938685531616, + 2.005093680458069, + 2.0022387979507448, + 2.004965279159546, + 2.006689316253662, + 2.0055594663238527, + 2.004364362411499, + 2.0020452467346193, + 2.005489226150513, + 2.003978981628418, + 2.0023497142791746, + 2.003284743270874, + 2.0037755869293212, + 2.001501432533264, + 2.003944476890564, + 2.0025060961914063, + 2.002634817123413, + 2.0017083084869385, + 2.0028567737579346, + 2.000872339820862, + 2.003729987487793, + 2.002261364898682, + 2.0024312939453126, + 2.0008276205444337, + 2.001199483261108, + 2.0006834197235106, + 2.000615832366943, + 1.9993626528930664, + 2.0005247971343993, + 1.9997058053588868, + 1.999853487548828, + 1.9997802280044557, + 1.9991605996704103, + 2.0015039880752563, + 1.9991352321624756, + 1.9994654761505126, + 1.99805854221344, + 1.9966946460723878, + 1.9976105904388428, + 1.9990063135528564, + 1.995104397277832, + 1.9979646807479858, + 1.996112073097229, + 1.999074932899475, + 1.997551460647583, + 1.9962433197784424, + 1.997741110534668, + 1.998971039352417, + 1.9981346895599366, + 1.9993204745864868, + 1.9973036824798585, + 1.9972220791625976, + 1.9969999575042725, + 1.9983418984222412, + 1.9977535138702394, + 1.9974792404174804, + 1.9975066395568848, + 1.9948017989349365, + 1.9976891431427002, + 1.9985219841003419, + 1.9948620611190795, + 1.995753077468872, + 1.998552144088745, + 1.9990386080932616 + ], + "train_acc": [ + 0.23936, + 0.25434, + 0.25346, + 0.25514, + 0.25882, + 0.25722, + 0.25998, + 0.25844, + 0.262, + 0.25976, + 0.26256, + 0.2639, + 0.26116, + 0.26558, + 0.26594, + 0.26562, + 0.26654, + 0.26378, + 0.2653, + 0.26746, + 0.26592, + 0.2681, + 0.26896, + 0.27006, + 0.26992, + 0.26862, + 0.26736, + 0.27126, + 0.26738, + 0.2694, + 0.26836, + 0.26744, + 0.27024, + 0.27184, + 0.26816, + 0.26806, + 0.27192, + 0.26898, + 0.27294, + 0.27124, + 0.2707, + 0.27056, + 0.272, + 0.27288, + 0.27096, + 0.27286, + 0.27098, + 0.27252, + 0.26964, + 0.27208, + 0.27, + 0.2737, + 0.27368, + 0.27134, + 0.27342, + 0.27386, + 0.27258, + 0.27398, + 0.27468, + 0.27518, + 0.27328, + 0.2752, + 0.27404, + 0.27526, + 0.2754, + 0.27588, + 0.27358, + 0.27472, + 0.27548, + 0.27356, + 0.27508, + 0.27656, + 0.27638, + 0.27784, + 0.27648, + 0.27606, + 0.27718, + 0.27632, + 0.27822, + 0.27514, + 0.27692, + 0.27566, + 0.27648, + 0.276, + 0.27642, + 0.27526, + 0.27678, + 0.2763, + 0.27776, + 0.27652, + 0.27578, + 0.27868, + 0.27676, + 0.27782, + 0.27764, + 0.27704, + 0.27938, + 0.27672, + 0.2781, + 0.2784 + ], + "test_acc": [ + 0.2781, + 0.2717, + 0.2892, + 0.2711, + 0.2577, + 0.2687, + 0.2516, + 0.281, + 0.2759, + 0.2822, + 0.2529, + 0.2933, + 0.2908, + 0.3048, + 0.2739, + 0.2751, + 0.2888, + 0.2869, + 0.3078, + 0.2855, + 0.2884, + 0.2932, + 0.2955, + 0.2952, + 0.2919, + 0.2886, + 0.2871, + 0.3017, + 0.295, + 0.3001, + 0.2957, + 0.2866, + 0.2896, + 0.3055, + 0.2928, + 0.3027, + 0.2836, + 0.2903, + 0.2958, + 0.2913, + 0.2966, + 0.287, + 0.2877, + 0.2879, + 0.303, + 0.3006, + 0.2946, + 0.2867, + 0.2973, + 0.2899, + 0.2917, + 0.3002, + 0.2971, + 0.2964, + 0.3037, + 0.2892, + 0.3038, + 0.2898, + 0.2999, + 0.3003, + 0.3047, + 0.2936, + 0.2972, + 0.2905, + 0.3016, + 0.2933, + 0.295, + 0.2998, + 0.3, + 0.3008, + 0.3028, + 0.3003, + 0.303, + 0.2946, + 0.2915, + 0.2974, + 0.2901, + 0.2978, + 0.2991, + 0.2973, + 0.2948, + 0.3032, + 0.3049, + 0.3017, + 0.2999, + 0.3019, + 0.2958, + 0.2981, + 0.3009, + 0.2968, + 0.2987, + 0.297, + 0.2988, + 0.2996, + 0.3002, + 0.3003, + 0.3001, + 0.3003, + 0.3001, + 0.3001 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.3634233772754669, + 0.001164254266768694, + -0.0006461217999458313, + -0.00019622840045485646, + 0.0002198266884079203, + -9.971270628739148e-06, + 0.00014979815750848502, + -0.0004021040804218501, + -5.425223207566887e-05, + -0.00024170493998099118, + -0.00025007897056639194, + -0.00018576104776002467 + ], + "perturbation_rho": [ + 0.061064913868904114, + 0.0, + 0.0, + 0.0, + 0.00392885273322463, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -3.8743019104003906e-07, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -1.1324882507324219e-06, + 0.0, + 2.7939677238464355e-09, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -3.606081008911133e-06, + 0.0, + 1.3504177331924438e-08, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 333.70584841410414, + "embed.bias": 239.49966422804454, + "blocks.0.ln.weight": 10.156513336148182, + "blocks.0.w1.weight": 293.93859688585053, + "blocks.0.w1.bias": 281.7468078045666, + "blocks.0.w2.weight": 492.64207719718667, + "blocks.1.ln.weight": 8.920722535237505, + "blocks.1.w1.weight": 311.4826752480239, + "blocks.1.w1.bias": 310.86370160689785, + "blocks.1.w2.weight": 324.5405256645745, + "blocks.2.ln.weight": 10.152850269187711, + "blocks.2.w1.weight": 422.5618024966304, + "blocks.2.w1.bias": 406.03916399589247, + "blocks.2.w2.weight": 427.4195122419038, + "blocks.3.ln.weight": 6.982139429700868, + "blocks.3.w1.weight": 239.67203375670587, + "blocks.3.w1.bias": 218.06238930657585, + "blocks.3.w2.weight": 238.92558897030935, + "blocks.4.ln.weight": 7.833071503056953, + "blocks.4.w1.weight": 298.1966678874803, + "blocks.4.w1.bias": 278.16504637194174, + "blocks.4.w2.weight": 274.11280875965394, + "blocks.5.ln.weight": 7.242266411414109, + "blocks.5.w1.weight": 264.2525555458606, + "blocks.5.w1.bias": 252.72325009085455, + "blocks.5.w2.weight": 257.616366671581, + "blocks.6.ln.weight": 7.548699630616584, + "blocks.6.w1.weight": 262.2510067819824, + "blocks.6.w1.bias": 251.81627023731494, + "blocks.6.w2.weight": 273.41300895523614, + "blocks.7.ln.weight": 8.476523067012732, + "blocks.7.w1.weight": 318.7369678667611, + "blocks.7.w1.bias": 299.5996411120576, + "blocks.7.w2.weight": 294.5553970248432, + "blocks.8.ln.weight": 10.089688663753533, + "blocks.8.w1.weight": 404.9516419003623, + "blocks.8.w1.bias": 379.7449486217603, + "blocks.8.w2.weight": 384.6524371767172, + "blocks.9.ln.weight": 10.376434921092248, + "blocks.9.w1.weight": 410.9149275889754, + "blocks.9.w1.bias": 383.0074711781066, + "blocks.9.w2.weight": 392.8539842927831, + "blocks.10.ln.weight": 6.445966095557191, + "blocks.10.w1.weight": 233.37502830905152, + "blocks.10.w1.bias": 219.30705311523053, + "blocks.10.w2.weight": 224.7808320380166, + "blocks.11.ln.weight": 9.439442353351934, + "blocks.11.w1.weight": 374.5461893118647, + "blocks.11.w1.bias": 354.2106427892852, + "blocks.11.w2.weight": 349.272673385976, + "out_ln.weight": 0.6481155342918145, + "out_head.weight": 9.174388328815116, + "out_head.bias": 0.47817510393040696 + } + }, + "state_bridge": { + "log": { + "train_loss": [ + 2.249122398071289, + 2.143466188659668, + 2.182890202178955, + 2.316445587387085, + 2.3062049280548096, + 2.266785619735718, + 2.2987561087799073, + 2.2884853198242188, + 2.240994506072998, + 2.2790724418640136, + 2.2708915201568605, + 2.266933690185547, + 2.2886832971954347, + 2.2700843003845215, + 2.1964657810211183, + 2.129935463180542, + 2.1168104545593263, + 2.12433757270813, + 2.113173648223877, + 2.1130354350280762, + 2.1059150744628905, + 2.091750974731445, + 2.092220969924927, + 2.124610143814087, + 2.096196698989868, + 2.089525361709595, + 2.0865603291320802, + 2.0807934146118163, + 2.069572374725342, + 2.0693005324554443, + 2.064417702713013, + 2.061735754241943, + 2.0726303718566896, + 2.068593667602539, + 2.0655730045318603, + 2.0626395158386233, + 2.063313564758301, + 2.0622880627441407, + 2.058299614639282, + 2.059487857131958, + 2.0631128742980955, + 2.0610568396759033, + 2.0551611611938476, + 2.056872489852905, + 2.0546858750915526, + 2.056325934524536, + 2.053202699737549, + 2.0516857704162597, + 2.0511376399230956, + 2.052292654342651, + 2.0525611149597167, + 2.052803971786499, + 2.0563543517684937, + 2.058361290130615, + 2.059238384399414, + 2.060622958755493, + 2.063846826324463, + 2.0644858879852297, + 2.0641055697631834, + 2.062063186683655, + 2.0649280325317383, + 2.065730286026001, + 2.0664522916412356, + 2.066531411819458, + 2.0671734712982177, + 2.066738704376221, + 2.06800027633667, + 2.0698299375915528, + 2.0699130378723143, + 2.069541824951172, + 2.069294273452759, + 2.068808924255371, + 2.071917738342285, + 2.0731585116577147, + 2.0724312242126466, + 2.073472029876709, + 2.0752377517700196, + 2.0765309967041015, + 2.0784705491638182, + 2.079111976928711, + 2.079522602157593, + 2.080129363708496, + 2.0800775938415526, + 2.0806046531677245, + 2.080093946685791, + 2.082602798614502, + 2.0823856858062744, + 2.083605904006958, + 2.082737918624878, + 2.0829258354187012, + 2.082134012680054, + 2.081413818206787, + 2.080659938735962, + 2.079646938858032, + 2.079785990867615, + 2.079912822265625, + 2.0787852781677247, + 2.080009171295166, + 2.0796959645080566, + 2.07859284072876 + ], + "train_acc": [ + 0.14508, + 0.16856, + 0.15632, + 0.11096, + 0.12388, + 0.14108, + 0.12374, + 0.1273, + 0.14434, + 0.12372, + 0.13196, + 0.13698, + 0.1278, + 0.13406, + 0.15046, + 0.17556, + 0.18282, + 0.17838, + 0.18174, + 0.1771, + 0.1835, + 0.18812, + 0.1935, + 0.17326, + 0.18206, + 0.1833, + 0.18656, + 0.19114, + 0.19432, + 0.19874, + 0.19688, + 0.19564, + 0.19196, + 0.19548, + 0.1934, + 0.19826, + 0.19794, + 0.19896, + 0.20276, + 0.20258, + 0.20152, + 0.20098, + 0.2025, + 0.20058, + 0.20388, + 0.20206, + 0.2039, + 0.20444, + 0.2054, + 0.20728, + 0.2042, + 0.20478, + 0.204, + 0.1998, + 0.20252, + 0.2013, + 0.1988, + 0.2002, + 0.20158, + 0.2012, + 0.20384, + 0.20102, + 0.1987, + 0.20136, + 0.19972, + 0.20122, + 0.19876, + 0.19792, + 0.20008, + 0.20138, + 0.1992, + 0.19956, + 0.20076, + 0.19942, + 0.20054, + 0.20192, + 0.20256, + 0.2009, + 0.1972, + 0.20028, + 0.19826, + 0.1992, + 0.19678, + 0.19736, + 0.19872, + 0.20076, + 0.20126, + 0.20262, + 0.20088, + 0.20116, + 0.2009, + 0.20188, + 0.20326, + 0.20284, + 0.20404, + 0.20164, + 0.20386, + 0.20324, + 0.20414, + 0.2032 + ], + "test_acc": [ + 0.1701, + 0.1589, + 0.0964, + 0.0977, + 0.147, + 0.1388, + 0.1311, + 0.1208, + 0.1018, + 0.1576, + 0.117, + 0.1127, + 0.1215, + 0.1579, + 0.1648, + 0.1561, + 0.179, + 0.1723, + 0.1962, + 0.1791, + 0.1863, + 0.1757, + 0.161, + 0.166, + 0.1768, + 0.1795, + 0.18, + 0.1802, + 0.1787, + 0.1767, + 0.1754, + 0.1604, + 0.1761, + 0.1675, + 0.1741, + 0.171, + 0.1697, + 0.1797, + 0.1836, + 0.1869, + 0.19, + 0.1893, + 0.1907, + 0.1866, + 0.1877, + 0.1889, + 0.1871, + 0.1949, + 0.1817, + 0.1912, + 0.1923, + 0.1955, + 0.1958, + 0.1893, + 0.1961, + 0.1915, + 0.1966, + 0.2031, + 0.1959, + 0.1915, + 0.1753, + 0.1929, + 0.18, + 0.1823, + 0.194, + 0.1869, + 0.1969, + 0.1912, + 0.1946, + 0.1919, + 0.1971, + 0.1907, + 0.2003, + 0.1926, + 0.1998, + 0.1947, + 0.1862, + 0.1997, + 0.1874, + 0.1841, + 0.188, + 0.1869, + 0.1828, + 0.1856, + 0.1888, + 0.1734, + 0.1826, + 0.1813, + 0.1816, + 0.1868, + 0.1879, + 0.1846, + 0.1837, + 0.182, + 0.1821, + 0.1836, + 0.1838, + 0.1833, + 0.1832, + 0.1836 + ], + "state_pred_error": [ + 0.3305018243932724, + 0.3240715994644165, + 0.2682553328800201, + 0.13893559893369675, + 0.07450400803923607, + 0.05266491491317749, + 0.035537137892246246, + 0.08557363073348999, + 0.0663624380338192, + 0.05056835491657257, + 0.057957575828433035, + 0.03669406015276909, + 0.044723540180921555, + 0.05952853068828583, + 0.06200094776391983, + 0.034456856629252436, + 0.02363166331708431, + 0.030915076084136964, + 0.027922455305457117, + 0.022749129890203477, + 0.01869435190498829, + 0.016911904378533363, + 0.012851448996663094, + 0.009603057160079479, + 0.0068074954861402515, + 0.004831086874008178, + 0.002716479171216488, + 0.0017772294883430003, + 0.0011674952478520572, + 0.0009641216253489256, + 0.000903967960383743, + 0.0008178514664620161, + 0.0009606654725596308, + 0.0010464790955930948, + 0.0009382549912855029, + 0.0008025767210498452, + 0.0006237311275862158, + 0.0005590206385217607, + 0.0006743966669589281, + 0.0007392789867147803, + 0.0007174553440324963, + 0.0006993773563578725, + 0.000645849712882191, + 0.0005546821858268231, + 0.00046942476788535716, + 0.0003906826237775385, + 0.0003712719864305109, + 0.0003606485006958246, + 0.0003457060167193413, + 0.0003746068775840104, + 0.00038963068067096173, + 0.0003513397928327322, + 0.00032962770885787906, + 0.00033959761667996645, + 0.0003122131848614663, + 0.00028035666743293406, + 0.00026064857746940106, + 0.00022029355119913816, + 0.00019001943845301867, + 0.00020799835772719233, + 0.0002275417461199686, + 0.00020783223733305932, + 0.00018264570890925825, + 0.00015985633003059774, + 0.0001389945726841688, + 0.00011476213769754394, + 8.967853816691787e-05, + 7.122783363331109e-05, + 5.970104366540909e-05, + 5.121438584523275e-05, + 4.536606192123145e-05, + 4.0154456216841935e-05, + 3.647202713880688e-05, + 3.4129825913114475e-05, + 3.2702885311446155e-05, + 3.160392617806792e-05, + 3.0352398289833216e-05, + 2.940795918868389e-05, + 2.8640556085156276e-05, + 2.736381722963415e-05, + 2.6395615809597075e-05, + 2.5282455111155287e-05, + 2.4516119782347232e-05, + 2.388914132665377e-05, + 2.2789849505061285e-05, + 2.1843732089619152e-05, + 2.0776926304679363e-05, + 1.9760204966296442e-05, + 1.8386752204969525e-05, + 1.7638299430254845e-05, + 1.702676533255726e-05, + 1.6189092184067705e-05, + 1.5407667033723554e-05, + 1.4915904570952988e-05, + 1.4520096725318581e-05, + 1.3764947652525734e-05, + 1.333884218824096e-05, + 1.3029495921800845e-05, + 1.262850838684244e-05, + 1.2368245947291143e-05 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.15691214799880981, + 0.015022292733192444, + 0.006593926809728146, + 0.007205793634057045, + 0.008038382977247238, + 0.0067636389285326, + 0.005863174330443144, + 0.0026869114954024553, + 0.00880957767367363, + 0.009782599285244942, + 0.00875927321612835, + 0.015338878147304058 + ], + "perturbation_rho": [ + 0.046312831342220306, + 0.002848614938557148, + -7.617553637828678e-05, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -4.103407263755798e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -1.2268312275409698e-05, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -4.0831975638866425e-05, + 7.450580596923828e-09, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 9.313225746154785e-10, + 0.0, + 0.0, + 0.0, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 20.26978490963512, + "embed.bias": 97.32930339256669, + "blocks.0.ln.weight": 1.9308135867617757, + "blocks.0.w1.weight": 66.3483568365869, + "blocks.0.w1.bias": 53.284724707801, + "blocks.0.w2.weight": 177.08907259952375, + "blocks.1.ln.weight": 3.057512831818983, + "blocks.1.w1.weight": 170.34177401016777, + "blocks.1.w1.bias": 213.60524387953882, + "blocks.1.w2.weight": 200.26342063170046, + "blocks.2.ln.weight": 2.5608747595416883, + "blocks.2.w1.weight": 103.95319184812183, + "blocks.2.w1.bias": 101.21405935076268, + "blocks.2.w2.weight": 134.8362490743747, + "blocks.3.ln.weight": 2.3109975543869266, + "blocks.3.w1.weight": 116.99989433662596, + "blocks.3.w1.bias": 120.91968382837668, + "blocks.3.w2.weight": 142.44744178171467, + "blocks.4.ln.weight": 2.718708135007822, + "blocks.4.w1.weight": 124.26248150423525, + "blocks.4.w1.bias": 121.48105264933947, + "blocks.4.w2.weight": 118.7892683407888, + "blocks.5.ln.weight": 1.8158878954501887, + "blocks.5.w1.weight": 56.878036794490896, + "blocks.5.w1.bias": 36.57930796279972, + "blocks.5.w2.weight": 85.47839722655326, + "blocks.6.ln.weight": 3.0875197025983447, + "blocks.6.w1.weight": 100.99150638794737, + "blocks.6.w1.bias": 97.42240055489651, + "blocks.6.w2.weight": 87.13862478741176, + "blocks.7.ln.weight": 7.772248541992971, + "blocks.7.w1.weight": 297.4998489055548, + "blocks.7.w1.bias": 282.6199479818741, + "blocks.7.w2.weight": 193.40369417394206, + "blocks.8.ln.weight": 3.0697438474661696, + "blocks.8.w1.weight": 143.40991382362168, + "blocks.8.w1.bias": 132.0178424400649, + "blocks.8.w2.weight": 126.31745589603334, + "blocks.9.ln.weight": 3.3171246104472094, + "blocks.9.w1.weight": 150.22737087992024, + "blocks.9.w1.bias": 147.1140425995551, + "blocks.9.w2.weight": 118.0665737085708, + "blocks.10.ln.weight": 6.064775829759772, + "blocks.10.w1.weight": 281.65394416567705, + "blocks.10.w1.bias": 267.43803892028814, + "blocks.10.w2.weight": 206.89673602536328, + "blocks.11.ln.weight": 2.0710718946854763, + "blocks.11.w1.weight": 93.44069995553728, + "blocks.11.w1.bias": 93.69984582471675, + "blocks.11.w2.weight": 110.16073882917532, + "out_ln.weight": 0.4517673564729089, + "out_head.weight": 5.403719578492724, + "out_head.bias": 2.04427795310877 + } + }, + "credit_bridge": { + "log": { + "train_loss": [ + 2.06320516204834, + 2.0399101606750487, + 2.033929338531494, + 2.0309921591186524, + 2.0298601413726804, + 2.0217224433135987, + 2.024232417259216, + 2.019040475921631, + 2.0207128286361695, + 2.019753430519104, + 2.016178896408081, + 2.011948891983032, + 2.0130044828796385, + 2.0113372691345215, + 2.012964423980713, + 2.0107208766937257, + 2.0117168256378175, + 2.0097561017227172, + 2.0089866023254395, + 2.009570707740784, + 2.011733226318359, + 2.007577508773804, + 2.0085436083602906, + 2.0043470149230957, + 2.0070890647888184, + 2.0043670698547364, + 2.0048324042129515, + 2.0042362908935547, + 2.0050302404403686, + 2.004891293258667, + 2.005185372543335, + 2.003529516067505, + 2.0045382270050047, + 2.005508458251953, + 2.00552710067749, + 2.002495421791077, + 1.9999042555999755, + 1.998048825454712, + 1.9961405738067628, + 1.989240593185425, + 1.9859777431488037, + 1.9825943405532838, + 1.9781343281555175, + 1.9761270376968383, + 1.9706906524658203, + 1.9723773523712158, + 1.9700259783172607, + 1.9686059323883056, + 1.9655450037384032, + 1.965288708114624, + 1.9665511059570313, + 1.9662485723114014, + 1.9659714967346191, + 1.9650201892852783, + 1.962706196937561, + 1.9626981174468994, + 1.962720913772583, + 1.9627574170684814, + 1.9608846031951905, + 1.9610804913711548, + 1.9616359865570068, + 1.9621461741638184, + 1.9603107889556886, + 1.9571026586914062, + 1.9596751629257203, + 1.9583080463409424, + 1.9566787594604491, + 1.9575193979644776, + 1.9559337323760986, + 1.9559677182006836, + 1.9548421646499634, + 1.9558928606414796, + 1.9595830995941161, + 1.9542909908676147, + 1.9583268686676025, + 1.9573874398040771, + 1.9573734212875367, + 1.956038475189209, + 1.9572998455047608, + 1.95643936958313, + 1.9555049303436278, + 1.9551573079681397, + 1.9559287934112548, + 1.9554652740859986, + 1.9546309524536132, + 1.9547708898925782, + 1.9570305030059814, + 1.9546662924194336, + 1.953575323562622, + 1.9553636624526978, + 1.954565380935669, + 1.9541568119812012, + 1.9543512627410888, + 1.954592315673828, + 1.9545559118652345, + 1.9531384017562867, + 1.9550698448944093, + 1.955762207107544, + 1.9552061297988892, + 1.957028095779419 + ], + "train_acc": [ + 0.23984, + 0.24842, + 0.2523, + 0.25414, + 0.25272, + 0.25882, + 0.257, + 0.26044, + 0.25976, + 0.26, + 0.26116, + 0.2649, + 0.26124, + 0.2642, + 0.26532, + 0.26592, + 0.26512, + 0.26658, + 0.26512, + 0.2671, + 0.2687, + 0.26848, + 0.26662, + 0.26946, + 0.26794, + 0.26952, + 0.26786, + 0.26836, + 0.26706, + 0.27072, + 0.26694, + 0.26872, + 0.2661, + 0.26742, + 0.26802, + 0.26896, + 0.26998, + 0.27168, + 0.2705, + 0.27562, + 0.27362, + 0.27702, + 0.27414, + 0.2807, + 0.28288, + 0.2806, + 0.28036, + 0.28098, + 0.28132, + 0.28288, + 0.28464, + 0.28428, + 0.28284, + 0.28638, + 0.28584, + 0.28538, + 0.28668, + 0.28688, + 0.28718, + 0.2852, + 0.28592, + 0.28522, + 0.28774, + 0.29002, + 0.28946, + 0.28842, + 0.28964, + 0.29052, + 0.29124, + 0.29164, + 0.29108, + 0.29114, + 0.28798, + 0.28992, + 0.28808, + 0.29178, + 0.28762, + 0.28974, + 0.29016, + 0.28958, + 0.29256, + 0.29012, + 0.29266, + 0.29038, + 0.2914, + 0.2924, + 0.2919, + 0.2951, + 0.2932, + 0.29244, + 0.29184, + 0.2924, + 0.29446, + 0.2932, + 0.29322, + 0.29452, + 0.29264, + 0.29084, + 0.29366, + 0.2917 + ], + "test_acc": [ + 0.2379, + 0.2678, + 0.2756, + 0.2987, + 0.2573, + 0.2947, + 0.2621, + 0.2983, + 0.2971, + 0.2616, + 0.278, + 0.2983, + 0.2803, + 0.2915, + 0.2714, + 0.2835, + 0.2837, + 0.2794, + 0.2915, + 0.2982, + 0.2854, + 0.2997, + 0.3019, + 0.2841, + 0.2873, + 0.2874, + 0.2676, + 0.3023, + 0.3033, + 0.2965, + 0.2888, + 0.279, + 0.3008, + 0.2752, + 0.2953, + 0.3039, + 0.3075, + 0.3057, + 0.2836, + 0.2859, + 0.303, + 0.2871, + 0.3052, + 0.3119, + 0.2982, + 0.2954, + 0.3026, + 0.303, + 0.3007, + 0.3087, + 0.3026, + 0.301, + 0.3082, + 0.3071, + 0.3137, + 0.3006, + 0.3102, + 0.3033, + 0.3079, + 0.3136, + 0.3092, + 0.3083, + 0.3181, + 0.3098, + 0.312, + 0.312, + 0.3052, + 0.3092, + 0.3124, + 0.3069, + 0.3038, + 0.3079, + 0.3117, + 0.3054, + 0.3065, + 0.3056, + 0.3086, + 0.3103, + 0.3038, + 0.3132, + 0.3157, + 0.3131, + 0.3092, + 0.313, + 0.3106, + 0.3118, + 0.3085, + 0.3115, + 0.3113, + 0.3129, + 0.3098, + 0.3146, + 0.3102, + 0.3109, + 0.3104, + 0.3117, + 0.3113, + 0.3106, + 0.3104, + 0.3104 + ], + "value_loss": [ + 0.4835468523311615, + 0.17319115218162537, + 0.12466768116235732, + 0.1044780854511261, + 0.11361225022792816, + 0.09506864285826683, + 0.08982749298095703, + 0.06943095141410828, + 0.06597960061311722, + 0.0689908292388916, + 0.05427408220410347, + 0.048282792123556136, + 0.04676814959049225, + 0.04863041600465774, + 0.0471925854742527, + 0.043705004994869234, + 0.0460531382226944, + 0.03692107924818993, + 0.03888409217596054, + 0.03788371595621109, + 0.03733079287290573, + 0.0377624527490139, + 0.036395574208498004, + 0.03193316010415554, + 0.03243833178460598, + 0.027760677633285522, + 0.029560993740558626, + 0.029699042382240297, + 0.02779862470269203, + 0.02897036738872528, + 0.025522839539051056, + 0.02311282460749149, + 0.020680676515102386, + 0.023304677747488022, + 0.022652746698856353, + 0.019273730003237725, + 0.018258913558125495, + 0.02010319753885269, + 0.01653825303465128, + 0.02135470446884632, + 0.018013720120489598, + 0.016730768181085588, + 0.01824077806353569, + 0.013809284281134605, + 0.01404231985926628, + 0.016477530472278595, + 0.01449635619878769, + 0.013006495736837388, + 0.013415186036229133, + 0.011523299672305583, + 0.01087947401329875, + 0.0120885151720047, + 0.011659734555482864, + 0.009783919822871684, + 0.009495561037063598, + 0.01004114615738392, + 0.009555040076076984, + 0.009653655390739441, + 0.009292597230225802, + 0.009270159545093774, + 0.008550122949182988, + 0.00753715006917715, + 0.007719495616853237, + 0.006912159933894873, + 0.006747392172813415, + 0.007531150514781475, + 0.006493680753260851, + 0.006225158845484257, + 0.006196837233901024, + 0.006545155131220818, + 0.005813424973487854, + 0.0052813203772902485, + 0.0056783805817365645, + 0.005296767835021019, + 0.005148515439927578, + 0.004331958843171596, + 0.00367335773833096, + 0.003545896329060197, + 0.0034597759203612802, + 0.003181714376732707, + 0.0032860505221784116, + 0.002376726396679878, + 0.002331429560147226, + 0.002127379362359643, + 0.0018355507389456033, + 0.0019746782282367347, + 0.0013750521056354046, + 0.0016369795854389668, + 0.0013883132124319672, + 0.0014942838795110583, + 0.0011334093083441258, + 0.0013581191193312407, + 0.0012651530192419886, + 0.0013948661609459669, + 0.00119701892176643, + 0.0009705750445276499, + 0.0009873379099927843, + 0.0013789990733936428, + 0.0015264763329923153, + 0.0007223523693159222 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.442160040140152, + 0.011405838653445244, + 0.021783415228128433, + 0.02257826365530491, + 0.022579727694392204, + 0.022490674629807472, + 0.022444019094109535, + 0.02243717387318611, + 0.02238454297184944, + 0.021799881011247635, + 0.021308548748493195, + 0.021358368918299675 + ], + "perturbation_rho": [ + -0.0024002324789762497, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -9.359791874885559e-07, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -3.080349415540695e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -1.0516028851270676e-05, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 185.7504766018605, + "embed.bias": 113.97655387234977, + "blocks.0.ln.weight": 8.006394857140304, + "blocks.0.w1.weight": 201.9426375330318, + "blocks.0.w1.bias": 163.53267733285352, + "blocks.0.w2.weight": 338.84776840881347, + "blocks.1.ln.weight": 6.295072957460427, + "blocks.1.w1.weight": 219.06553999907572, + "blocks.1.w1.bias": 226.32772829898, + "blocks.1.w2.weight": 242.1925329344591, + "blocks.2.ln.weight": 7.320912432081406, + "blocks.2.w1.weight": 311.3035934820304, + "blocks.2.w1.bias": 298.6852088590304, + "blocks.2.w2.weight": 316.80472460197296, + "blocks.3.ln.weight": 4.940329586675122, + "blocks.3.w1.weight": 172.30509352305455, + "blocks.3.w1.bias": 152.69976843471107, + "blocks.3.w2.weight": 171.54613400302898, + "blocks.4.ln.weight": 5.398528552340865, + "blocks.4.w1.weight": 194.3066729117384, + "blocks.4.w1.bias": 177.2111282722944, + "blocks.4.w2.weight": 181.75464322675703, + "blocks.5.ln.weight": 5.0186360723250045, + "blocks.5.w1.weight": 177.28809608368013, + "blocks.5.w1.bias": 166.1128575027424, + "blocks.5.w2.weight": 177.87652026549083, + "blocks.6.ln.weight": 5.11718383519975, + "blocks.6.w1.weight": 175.68155404770292, + "blocks.6.w1.bias": 165.18102396382997, + "blocks.6.w2.weight": 188.26739446960036, + "blocks.7.ln.weight": 5.749714981932067, + "blocks.7.w1.weight": 211.62845167555633, + "blocks.7.w1.bias": 194.987439334531, + "blocks.7.w2.weight": 197.40821780414132, + "blocks.8.ln.weight": 7.020310313763905, + "blocks.8.w1.weight": 281.07720524059584, + "blocks.8.w1.bias": 259.36087916311686, + "blocks.8.w2.weight": 264.53573649142305, + "blocks.9.ln.weight": 7.277383166194383, + "blocks.9.w1.weight": 285.7959615127006, + "blocks.9.w1.bias": 261.89392533705353, + "blocks.9.w2.weight": 274.12202590162974, + "blocks.10.ln.weight": 4.304915932630116, + "blocks.10.w1.weight": 146.38313888426148, + "blocks.10.w1.bias": 134.29068175155572, + "blocks.10.w2.weight": 146.4223780813692, + "blocks.11.ln.weight": 6.661402581393363, + "blocks.11.w1.weight": 265.5729059266825, + "blocks.11.w1.bias": 247.27216580071305, + "blocks.11.w2.weight": 242.98119602092984, + "out_ln.weight": 0.5446617517258924, + "out_head.weight": 7.799632034433011, + "out_head.bias": 1.2214285543387151 + } + } + }, + "config": { + "dataset": "cifar10", + "d_hidden": 512, + "num_blocks": 12, + "batch_size": 128, + "epochs": 100, + "lr": 0.001, + "lr_fb": 0.001, + "wd": 0.01, + "lam": 0.1, + "K": 4, + "sigma_bridge": 0.05, + "ema_momentum": 0.995, + "term_grad_weight": 1.0, + "seeds": [ + 42 + ], + "gpu": 0, + "output_dir": "results/cifar10", + "num_classes": 10 + } +}
\ No newline at end of file diff --git a/results/cifar10_seed123/results_cifar10.json b/results/cifar10_seed123/results_cifar10.json new file mode 100644 index 0000000..fd8e585 --- /dev/null +++ b/results/cifar10_seed123/results_cifar10.json @@ -0,0 +1,1985 @@ +{ + "123": { + "bp": { + "log": { + "train_loss": [ + 1.9175124251556397, + 1.6847565829849243, + 1.5875056425857543, + 1.5320447119140626, + 1.4794927943801879, + 1.4405573056030274, + 1.4091039984893798, + 1.3804642078399658, + 1.3562547276306152, + 1.3386102033233642, + 1.3113672603607178, + 1.286414957962036, + 1.2713352434539795, + 1.2514976085662841, + 1.2332095288848877, + 1.2136467219924927, + 1.2016698885345458, + 1.1760129442214966, + 1.1615581885910033, + 1.1460976659965516, + 1.1286324452972412, + 1.1125253838348388, + 1.0944270223999024, + 1.0830238839530946, + 1.0608218252563477, + 1.0437702729034424, + 1.034650561027527, + 1.017014478340149, + 0.9988842744445801, + 0.9852388265800476, + 0.9635365026664734, + 0.9439176512908936, + 0.9298832908248902, + 0.9137922927856446, + 0.895711201210022, + 0.8814193804550171, + 0.8632927577590942, + 0.84351964012146, + 0.827347836894989, + 0.807900242767334, + 0.7877964169311523, + 0.7770450174331665, + 0.7534891272926331, + 0.7365407562255859, + 0.7147518105125428, + 0.6962322827911377, + 0.6742078248596192, + 0.6564239510726929, + 0.6403258143234253, + 0.6197586893653869, + 0.6014538905525207, + 0.583438274936676, + 0.5667518968963623, + 0.543176082239151, + 0.5258174272155761, + 0.5108640935707093, + 0.4897696580505371, + 0.47116634141922, + 0.4534935041809082, + 0.4354778845024109, + 0.4233958884239197, + 0.4017178301334381, + 0.3859938202762604, + 0.37423386340141296, + 0.365246470451355, + 0.3411486894798279, + 0.32471628838539124, + 0.31114961336135866, + 0.29552998408317566, + 0.28871005929946897, + 0.27408093184471133, + 0.26901549862384794, + 0.2596221783256531, + 0.24521093573570252, + 0.2299592537355423, + 0.22053928468227388, + 0.21593001784324645, + 0.1976856986284256, + 0.19874527307510376, + 0.19333338455200194, + 0.18259587281227113, + 0.17466963967323304, + 0.1689953342151642, + 0.16150579230308532, + 0.15866700973510742, + 0.15368373103141786, + 0.1482925960588455, + 0.145876254529953, + 0.14030234585762025, + 0.13952740331172944, + 0.1361180233001709, + 0.1323960340666771, + 0.12677059319972991, + 0.1287061265563965, + 0.12679034420013427, + 0.12853138402462005, + 0.12579439376831056, + 0.1280787334895134, + 0.12339243181705475, + 0.12544435278892518 + ], + "train_acc": [ + 0.31434, + 0.39008, + 0.42376, + 0.44654, + 0.46472, + 0.47872, + 0.49158, + 0.5009, + 0.51142, + 0.51944, + 0.52706, + 0.53622, + 0.54168, + 0.55052, + 0.5568, + 0.56322, + 0.56372, + 0.57526, + 0.58298, + 0.58918, + 0.59476, + 0.6007, + 0.6078, + 0.61058, + 0.61816, + 0.62692, + 0.62922, + 0.63428, + 0.64188, + 0.64736, + 0.6547, + 0.65824, + 0.6673, + 0.67046, + 0.67872, + 0.68352, + 0.68988, + 0.69742, + 0.70294, + 0.7099, + 0.71534, + 0.72286, + 0.72898, + 0.73446, + 0.74268, + 0.75076, + 0.75844, + 0.76398, + 0.76912, + 0.77878, + 0.78504, + 0.78992, + 0.79944, + 0.80564, + 0.81262, + 0.81906, + 0.82502, + 0.83208, + 0.83826, + 0.84468, + 0.8507, + 0.8589, + 0.86356, + 0.86656, + 0.87174, + 0.87936, + 0.88406, + 0.8908, + 0.89792, + 0.89858, + 0.90396, + 0.90662, + 0.90894, + 0.91456, + 0.9191, + 0.92516, + 0.92578, + 0.93306, + 0.93206, + 0.93384, + 0.9388, + 0.94064, + 0.94322, + 0.94482, + 0.9463, + 0.94904, + 0.94962, + 0.95136, + 0.95286, + 0.9532, + 0.95498, + 0.95586, + 0.9589, + 0.95728, + 0.95824, + 0.95676, + 0.95896, + 0.95834, + 0.95978, + 0.95964 + ], + "test_acc": [ + 0.3978, + 0.4369, + 0.4561, + 0.4846, + 0.4943, + 0.5129, + 0.5144, + 0.5192, + 0.5235, + 0.5332, + 0.53, + 0.5382, + 0.5478, + 0.5507, + 0.5534, + 0.5638, + 0.5725, + 0.5626, + 0.572, + 0.5735, + 0.5778, + 0.5778, + 0.583, + 0.5789, + 0.5885, + 0.589, + 0.5849, + 0.5957, + 0.5916, + 0.5872, + 0.5864, + 0.589, + 0.599, + 0.6012, + 0.5912, + 0.5967, + 0.5973, + 0.5923, + 0.5983, + 0.5947, + 0.5966, + 0.5996, + 0.6007, + 0.5935, + 0.6005, + 0.6016, + 0.5972, + 0.5976, + 0.596, + 0.5938, + 0.5933, + 0.598, + 0.5987, + 0.5998, + 0.5914, + 0.5916, + 0.5988, + 0.5984, + 0.5936, + 0.6002, + 0.5917, + 0.5921, + 0.5944, + 0.5968, + 0.5925, + 0.5974, + 0.5946, + 0.5979, + 0.5923, + 0.5956, + 0.5968, + 0.5956, + 0.5974, + 0.5934, + 0.5943, + 0.5954, + 0.5982, + 0.5923, + 0.5921, + 0.5937, + 0.5937, + 0.5941, + 0.5921, + 0.5957, + 0.5953, + 0.5939, + 0.5939, + 0.5965, + 0.5952, + 0.5957, + 0.5951, + 0.5976, + 0.5967, + 0.5968, + 0.5974, + 0.5965, + 0.5973, + 0.596, + 0.5961, + 0.5962 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.9620110392570496, + 0.9621291160583496, + 0.9621908664703369, + 0.9623005986213684, + 0.9625044465065002, + 0.9625605344772339, + 0.9631243348121643, + 0.9631022214889526, + 0.9631184935569763, + 0.9626430869102478, + 0.9622764587402344, + 0.9619318246841431 + ], + "perturbation_rho": [ + 0.9879884719848633, + 0.9869263172149658, + 0.9891279935836792, + 0.9888099431991577, + 0.991609513759613, + 0.9902887344360352, + 0.9909973740577698, + 0.9915982484817505, + 0.9906468987464905, + 0.9890569448471069, + 0.9870684146881104, + 0.9818294644355774 + ], + "nudging": { + "0.001": [ + -0.003427116898819804, + -0.003351739142090082, + -0.0033903690055012703, + -0.003365215379744768, + -0.0033891985658556223, + -0.003362787188962102, + -0.0032895321492105722, + -0.0031716772355139256, + -0.0029750647954642773, + -0.002764828037470579, + -0.0024368567392230034, + -0.0021113622933626175 + ], + "0.003": [ + -0.010271656326949596, + -0.010045798495411873, + -0.010160606354475021, + -0.010086143389344215, + -0.01015765406191349, + -0.010078245773911476, + -0.009858867153525352, + -0.009506190195679665, + -0.008917586877942085, + -0.008288294076919556, + -0.007305148057639599, + -0.006330306641757488 + ], + "0.01": [ + -0.03411904349923134, + -0.033372972160577774, + -0.033751748502254486, + -0.03350646048784256, + -0.0337451696395874, + -0.03347958251833916, + -0.03275400772690773, + -0.03158621862530708, + -0.029637902975082397, + -0.02755369432270527, + -0.024292215704917908, + -0.02105848491191864 + ] + } + }, + "drift": { + "embed.weight": 8.763637326480993, + "embed.bias": 15.277378973045789, + "blocks.0.ln.weight": 0.34908970813559786, + "blocks.0.w1.weight": 4.164846741111791, + "blocks.0.w1.bias": 6.735275454452179, + "blocks.0.w2.weight": 10.512885267241659, + "blocks.1.ln.weight": 0.2548678557191968, + "blocks.1.w1.weight": 4.428344802990091, + "blocks.1.w1.bias": 4.431127174572938, + "blocks.1.w2.weight": 11.609675967743058, + "blocks.2.ln.weight": 0.20993716494835368, + "blocks.2.w1.weight": 4.563301300136555, + "blocks.2.w1.bias": 4.053222361596439, + "blocks.2.w2.weight": 12.283684374504883, + "blocks.3.ln.weight": 0.22639993499269975, + "blocks.3.w1.weight": 4.586109188420391, + "blocks.3.w1.bias": 3.9811075138013527, + "blocks.3.w2.weight": 12.651716892358182, + "blocks.4.ln.weight": 0.23813753747592467, + "blocks.4.w1.weight": 4.55512749423403, + "blocks.4.w1.bias": 3.7036288780635465, + "blocks.4.w2.weight": 12.900508521442045, + "blocks.5.ln.weight": 0.24851322774648468, + "blocks.5.w1.weight": 4.519407405633592, + "blocks.5.w1.bias": 3.7828748042863825, + "blocks.5.w2.weight": 12.877614228932648, + "blocks.6.ln.weight": 0.24789524956961745, + "blocks.6.w1.weight": 4.510086973031788, + "blocks.6.w1.bias": 3.7520206765756337, + "blocks.6.w2.weight": 13.18270185196777, + "blocks.7.ln.weight": 0.27277168975906585, + "blocks.7.w1.weight": 4.471720887903336, + "blocks.7.w1.bias": 3.7457909142015353, + "blocks.7.w2.weight": 13.30644187374482, + "blocks.8.ln.weight": 0.29086480529883674, + "blocks.8.w1.weight": 4.434570868617996, + "blocks.8.w1.bias": 3.7342218578126882, + "blocks.8.w2.weight": 13.49877934675731, + "blocks.9.ln.weight": 0.2919660181755758, + "blocks.9.w1.weight": 4.429555710823986, + "blocks.9.w1.bias": 3.634447923096686, + "blocks.9.w2.weight": 13.75719577660119, + "blocks.10.ln.weight": 0.32005145624529063, + "blocks.10.w1.weight": 4.353451334572149, + "blocks.10.w1.bias": 3.4614471694540865, + "blocks.10.w2.weight": 13.887069102514094, + "blocks.11.ln.weight": 0.29895213239033414, + "blocks.11.w1.weight": 4.374616657143753, + "blocks.11.w1.bias": 3.645530912656648, + "blocks.11.w2.weight": 14.561155498919536, + "out_ln.weight": 0.283104664197929, + "out_head.weight": 3.694886816637194, + "out_head.bias": 1.105623778518022 + } + }, + "dfa": { + "log": { + "train_loss": [ + 2.0684899793243408, + 2.0445925352478027, + 2.03087772151947, + 2.027939418258667, + 2.0253660607528685, + 2.0250345896148683, + 2.022372142868042, + 2.019951826019287, + 2.017632607574463, + 2.015866649131775, + 2.017579344139099, + 2.01485851020813, + 2.012323605117798, + 2.009420676612854, + 2.0132182874298095, + 2.007821100616455, + 2.0097865366363528, + 2.005040230484009, + 2.004450204849243, + 2.0059849935150145, + 2.005384602432251, + 2.001610414581299, + 2.002629147567749, + 2.0022453118896486, + 1.9994831539154052, + 1.9995977224349974, + 2.0003068398284913, + 2.0012258628082273, + 1.9992249011611938, + 1.9970151906967164, + 1.9978694770812988, + 1.997133782272339, + 1.9935067112731935, + 1.9952505539321899, + 1.992918910446167, + 1.9958944741058349, + 1.9945134730529785, + 1.993255687599182, + 1.9926469972991943, + 1.9950547982406617, + 1.9904610315322877, + 1.9907950922393798, + 1.991843765411377, + 1.9908243858337402, + 1.9923807875823976, + 1.9917906607055664, + 1.992043903465271, + 1.9904995797348022, + 1.9916813136291505, + 1.98937647026062, + 1.9889434642791748, + 1.9899712393188476, + 1.9896988627624512, + 1.9895934002304076, + 1.9901937147521973, + 1.9906840317535401, + 1.9852267303466797, + 1.989434090652466, + 1.9889313174438477, + 1.9886722135925292, + 1.9877896743392944, + 1.9861792268371583, + 1.986029208984375, + 1.9876557461547852, + 1.985245605545044, + 1.9858075296401978, + 1.9867199410247802, + 1.9870598261260985, + 1.9854085816955567, + 1.9867712688827515, + 1.9855045788955688, + 1.9881172052764893, + 1.9837462771987915, + 1.98582817527771, + 1.9842846584320069, + 1.9848911337280273, + 1.9850436270904541, + 1.9835600254821777, + 1.9840570623779297, + 1.9839281851196289, + 1.984888575820923, + 1.9857918256378173, + 1.9845330112075805, + 1.984420718612671, + 1.9853396697998047, + 1.9839463347625732, + 1.9833207806396485, + 1.9841535271453858, + 1.9810106130218506, + 1.9846471352386474, + 1.9849586167144775, + 1.984921664199829, + 1.984304642715454, + 1.9863978981781005, + 1.9833829919052124, + 1.984193635787964, + 1.983270346107483, + 1.9827691938018799, + 1.9846319453430177, + 1.9829401163101197 + ], + "train_acc": [ + 0.2366, + 0.24504, + 0.25212, + 0.25468, + 0.25604, + 0.2569, + 0.25714, + 0.2588, + 0.25558, + 0.2631, + 0.2608, + 0.26104, + 0.26492, + 0.26478, + 0.26264, + 0.26596, + 0.26544, + 0.267, + 0.26848, + 0.26756, + 0.26702, + 0.26954, + 0.27278, + 0.26886, + 0.26902, + 0.27146, + 0.272, + 0.2698, + 0.2728, + 0.27384, + 0.2713, + 0.27288, + 0.27422, + 0.2767, + 0.2742, + 0.2748, + 0.27392, + 0.27568, + 0.27816, + 0.27558, + 0.27752, + 0.27658, + 0.27714, + 0.277, + 0.27766, + 0.27532, + 0.27622, + 0.27806, + 0.27688, + 0.27908, + 0.27896, + 0.27908, + 0.27972, + 0.28032, + 0.27848, + 0.27914, + 0.2816, + 0.27856, + 0.27972, + 0.28028, + 0.28068, + 0.27928, + 0.28124, + 0.2814, + 0.28118, + 0.28066, + 0.2815, + 0.28218, + 0.28352, + 0.2822, + 0.28114, + 0.28234, + 0.28204, + 0.28104, + 0.28, + 0.2824, + 0.28114, + 0.28372, + 0.282, + 0.28204, + 0.28208, + 0.28162, + 0.28212, + 0.28526, + 0.28178, + 0.28234, + 0.28358, + 0.28328, + 0.28346, + 0.28396, + 0.28374, + 0.28334, + 0.28204, + 0.28406, + 0.28382, + 0.28466, + 0.28292, + 0.28436, + 0.28206, + 0.28276 + ], + "test_acc": [ + 0.2589, + 0.2617, + 0.2631, + 0.2816, + 0.2556, + 0.2607, + 0.2876, + 0.2818, + 0.2763, + 0.2847, + 0.3005, + 0.2738, + 0.283, + 0.2729, + 0.2859, + 0.2782, + 0.3055, + 0.2874, + 0.2796, + 0.3027, + 0.2977, + 0.3013, + 0.2869, + 0.2663, + 0.2917, + 0.2707, + 0.2784, + 0.3006, + 0.2812, + 0.2801, + 0.2755, + 0.2935, + 0.3091, + 0.3041, + 0.2967, + 0.3054, + 0.3107, + 0.2963, + 0.2956, + 0.2927, + 0.2941, + 0.2943, + 0.2948, + 0.3057, + 0.3015, + 0.3116, + 0.2928, + 0.301, + 0.288, + 0.296, + 0.2909, + 0.2957, + 0.3145, + 0.3124, + 0.2994, + 0.307, + 0.3047, + 0.3026, + 0.2989, + 0.2978, + 0.3023, + 0.2971, + 0.3062, + 0.3025, + 0.2919, + 0.309, + 0.301, + 0.2947, + 0.3086, + 0.2976, + 0.3106, + 0.3013, + 0.3109, + 0.3012, + 0.3019, + 0.2969, + 0.3022, + 0.3038, + 0.3035, + 0.3087, + 0.2972, + 0.2974, + 0.2952, + 0.3032, + 0.3034, + 0.3027, + 0.3024, + 0.3055, + 0.3031, + 0.3003, + 0.3044, + 0.3046, + 0.3032, + 0.3029, + 0.303, + 0.3028, + 0.3038, + 0.3032, + 0.3035, + 0.3034 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.40911683440208435, + 0.0002619648876134306, + -0.00015980367606971413, + 0.0005440536187961698, + 0.0002772972220554948, + 7.781770545989275e-05, + -0.00020515912910923362, + -0.00044000157504342496, + -0.0008366729598492384, + 0.00015155941946431994, + 0.00032340025063604116, + -0.00015023337618913502 + ], + "perturbation_rho": [ + 0.0528201088309288, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -4.898756742477417e-07, + 1.862645149230957e-09, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + -1.862645149230957e-09, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -1.2246891856193542e-06, + 1.862645149230957e-09, + -9.313225746154785e-10, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + -1.862645149230957e-09, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -4.1155144572257996e-06, + -1.862645149230957e-09, + -9.313225746154785e-10, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + -2.7939677238464355e-09, + 0.0, + -9.313225746154785e-10, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 323.8982252804048, + "embed.bias": 232.81638531211948, + "blocks.0.ln.weight": 10.852100889778994, + "blocks.0.w1.weight": 272.36679039121384, + "blocks.0.w1.bias": 238.87995630640614, + "blocks.0.w2.weight": 445.2406120934057, + "blocks.1.ln.weight": 9.73650143893558, + "blocks.1.w1.weight": 390.9712215089673, + "blocks.1.w1.bias": 374.17715191606453, + "blocks.1.w2.weight": 403.2488589095519, + "blocks.2.ln.weight": 6.710448026907898, + "blocks.2.w1.weight": 227.28731670207446, + "blocks.2.w1.bias": 202.7493867156122, + "blocks.2.w2.weight": 223.99250691076898, + "blocks.3.ln.weight": 7.215269500019514, + "blocks.3.w1.weight": 231.09319791624182, + "blocks.3.w1.bias": 206.9745974350102, + "blocks.3.w2.weight": 238.74614111179253, + "blocks.4.ln.weight": 9.257100922232066, + "blocks.4.w1.weight": 374.01823924044805, + "blocks.4.w1.bias": 343.71540244431617, + "blocks.4.w2.weight": 350.8111852358107, + "blocks.5.ln.weight": 6.914568252312619, + "blocks.5.w1.weight": 266.9118711467473, + "blocks.5.w1.bias": 256.9588103323234, + "blocks.5.w2.weight": 254.31906752172955, + "blocks.6.ln.weight": 8.82549425820829, + "blocks.6.w1.weight": 354.28665902336076, + "blocks.6.w1.bias": 332.44091471694884, + "blocks.6.w2.weight": 326.29279154408147, + "blocks.7.ln.weight": 10.262108389389345, + "blocks.7.w1.weight": 411.01500781893094, + "blocks.7.w1.bias": 383.25852137267697, + "blocks.7.w2.weight": 398.26834704743106, + "blocks.8.ln.weight": 9.486443170916516, + "blocks.8.w1.weight": 379.7970569993528, + "blocks.8.w1.bias": 353.44640446436694, + "blocks.8.w2.weight": 354.75408312717866, + "blocks.9.ln.weight": 7.225265383958218, + "blocks.9.w1.weight": 280.98274227898526, + "blocks.9.w1.bias": 264.21754537134655, + "blocks.9.w2.weight": 250.42097452359232, + "blocks.10.ln.weight": 9.12118778587681, + "blocks.10.w1.weight": 363.4177719662398, + "blocks.10.w1.bias": 333.2020904512465, + "blocks.10.w2.weight": 343.55519901127, + "blocks.11.ln.weight": 9.146084772655259, + "blocks.11.w1.weight": 367.21625652465485, + "blocks.11.w1.bias": 343.39432049385726, + "blocks.11.w2.weight": 338.5639684764078, + "out_ln.weight": 0.5993398286123665, + "out_head.weight": 9.522011867407477, + "out_head.bias": 0.44134062139959057 + } + }, + "state_bridge": { + "log": { + "train_loss": [ + 2.27535658454895, + 2.2637348059844973, + 2.23140262840271, + 2.13100928276062, + 2.1194422898101806, + 2.113937710952759, + 2.123109260482788, + 2.123372416381836, + 2.118385257644653, + 2.123976169281006, + 2.143634383087158, + 2.15997088432312, + 2.1503283737182617, + 2.2361232402801514, + 2.31340713142395, + 2.312125205230713, + 2.3117453997802735, + 2.3094616239929198, + 2.3084144255065917, + 2.3107836222076417, + 2.3105288831329345, + 2.312015315246582, + 2.311362181777954, + 2.3104846384429933, + 2.3100794078826903, + 2.308620408859253, + 2.2244751792907715, + 2.1188239234924318, + 2.099667515487671, + 2.1012451152038576, + 2.101906224517822, + 2.1015794395446776, + 2.102679093170166, + 2.104404850845337, + 2.105138458328247, + 2.1025040253448486, + 2.09926293258667, + 2.0996690463256837, + 2.0955256491851806, + 2.0939024459075926, + 2.0932679664611817, + 2.0943772303771975, + 2.087596756668091, + 2.0909996266937254, + 2.087642866668701, + 2.089086704559326, + 2.0858034885406496, + 2.084843352355957, + 2.081013270111084, + 2.0763836125183106, + 2.068111530685425, + 2.0633422270965576, + 2.059932788619995, + 2.0572856521987917, + 2.0567763134765626, + 2.055301329345703, + 2.0559660177230836, + 2.0547456675720217, + 2.0552883127593993, + 2.052630202102661, + 2.050533991241455, + 2.0526472858428955, + 2.051843546676636, + 2.0524809063720704, + 2.052277666931152, + 2.049621379356384, + 2.0498316369628906, + 2.049651178665161, + 2.0472222846221926, + 2.0483968115997313, + 2.04511710357666, + 2.0473428889465333, + 2.0471922763061525, + 2.050198796348572, + 2.0515007726287844, + 2.051603550682068, + 2.0523804613494874, + 2.054355347518921, + 2.055121799697876, + 2.0543520391082764, + 2.0561765496063233, + 2.056192929840088, + 2.054548904418945, + 2.05620756149292, + 2.055558791809082, + 2.056474009552002, + 2.055143079071045, + 2.054688854446411, + 2.0569355255126953, + 2.0542632417297364, + 2.054791463470459, + 2.054189739532471, + 2.0533234220123293, + 2.0529643350982667, + 2.053456902542114, + 2.0533332372283937, + 2.0545412349700927, + 2.053972573928833, + 2.0536360316467284, + 2.053008867416382 + ], + "train_acc": [ + 0.13698, + 0.13998, + 0.1447, + 0.17058, + 0.17036, + 0.1725, + 0.1677, + 0.16766, + 0.1698, + 0.1709, + 0.16482, + 0.1638, + 0.16334, + 0.14304, + 0.10524, + 0.11206, + 0.10998, + 0.11288, + 0.11528, + 0.10822, + 0.10712, + 0.10462, + 0.1037, + 0.10238, + 0.10366, + 0.10334, + 0.15168, + 0.17212, + 0.17798, + 0.17688, + 0.17846, + 0.17562, + 0.17644, + 0.17718, + 0.175, + 0.17492, + 0.1766, + 0.17688, + 0.17674, + 0.18138, + 0.18304, + 0.17868, + 0.18296, + 0.18172, + 0.18454, + 0.18428, + 0.18226, + 0.18568, + 0.18754, + 0.19132, + 0.2004, + 0.20496, + 0.207, + 0.20808, + 0.2054, + 0.20596, + 0.20646, + 0.20616, + 0.20466, + 0.20608, + 0.20848, + 0.20616, + 0.20746, + 0.20588, + 0.20528, + 0.20778, + 0.20856, + 0.2078, + 0.2101, + 0.20888, + 0.21296, + 0.211, + 0.20934, + 0.20746, + 0.20674, + 0.20832, + 0.20954, + 0.20736, + 0.20522, + 0.20496, + 0.20606, + 0.20302, + 0.20594, + 0.20828, + 0.2064, + 0.20466, + 0.20378, + 0.20496, + 0.20534, + 0.20862, + 0.20546, + 0.20594, + 0.20688, + 0.20658, + 0.20712, + 0.20672, + 0.20568, + 0.20664, + 0.20646, + 0.20888 + ], + "test_acc": [ + 0.1514, + 0.1391, + 0.1828, + 0.1457, + 0.1765, + 0.186, + 0.1602, + 0.1723, + 0.1754, + 0.1714, + 0.1724, + 0.1634, + 0.1615, + 0.0959, + 0.1252, + 0.1087, + 0.135, + 0.1366, + 0.1028, + 0.1085, + 0.1257, + 0.114, + 0.119, + 0.1, + 0.0848, + 0.1194, + 0.1607, + 0.1795, + 0.1859, + 0.1635, + 0.1571, + 0.1682, + 0.1748, + 0.1717, + 0.1789, + 0.1658, + 0.1673, + 0.1783, + 0.1799, + 0.1761, + 0.1812, + 0.173, + 0.1716, + 0.164, + 0.1657, + 0.1769, + 0.1728, + 0.1726, + 0.1801, + 0.1758, + 0.1873, + 0.1862, + 0.1948, + 0.1935, + 0.2009, + 0.1933, + 0.2102, + 0.2094, + 0.2153, + 0.2097, + 0.2054, + 0.211, + 0.2109, + 0.1975, + 0.2107, + 0.2099, + 0.2109, + 0.202, + 0.2055, + 0.2103, + 0.2068, + 0.2068, + 0.2085, + 0.2085, + 0.206, + 0.203, + 0.1998, + 0.2021, + 0.2092, + 0.2039, + 0.2034, + 0.1976, + 0.2064, + 0.2098, + 0.2078, + 0.2045, + 0.2078, + 0.2055, + 0.2068, + 0.2065, + 0.2075, + 0.2084, + 0.206, + 0.2056, + 0.2061, + 0.2063, + 0.2066, + 0.2067, + 0.2068, + 0.2069 + ], + "state_pred_error": [ + 0.3455641258764267, + 0.1890164827811718, + 0.06542741327524185, + 0.15099643572807311, + 0.15767600553512573, + 0.14595494849205018, + 0.09619720283508301, + 0.05848943448066712, + 0.04476255656838417, + 0.03768930072784424, + 0.034250264310240745, + 0.02679455811023712, + 0.024154067809581757, + 0.020689564958810806, + 0.015427032306194306, + 0.01494347603648901, + 0.012853515672385693, + 0.00926094017356634, + 0.00883663154706359, + 0.00866620466351509, + 0.008630412181615829, + 0.009436789927184582, + 0.010587773686945438, + 0.006365314861536026, + 0.0058888214336335655, + 0.004059859301149845, + 0.0059570299243927, + 0.005231297912150622, + 0.004334697146564722, + 0.005451656972020865, + 0.005537090619504452, + 0.006261329069137574, + 0.006663460256308317, + 0.0068712751638889315, + 0.006620877954363823, + 0.0059924150297045705, + 0.004774981212168932, + 0.0035280540695041417, + 0.003671990767121315, + 0.0035207420617341997, + 0.0030331813983619214, + 0.0031412731628119943, + 0.0028461132158339023, + 0.002806654215082526, + 0.0028670129170268776, + 0.0030720302928984164, + 0.0032383741122484207, + 0.002952979260608554, + 0.002582101986259222, + 0.0024391376911848785, + 0.002423995024561882, + 0.002365549064427614, + 0.002312273128144443, + 0.0020981324925273656, + 0.0018509870688617229, + 0.001701984341070056, + 0.001557768930643797, + 0.001504055576622486, + 0.0015004488579928876, + 0.0016905955363065005, + 0.001851444918513298, + 0.002072783012315631, + 0.0021090806840360164, + 0.0020937418549507856, + 0.001958419643007219, + 0.0018501618901640177, + 0.0017675739562511444, + 0.001684609704464674, + 0.0016207635071873664, + 0.0015579476598650216, + 0.0014976035876572132, + 0.0014524117913842201, + 0.0013800759248435498, + 0.001322413182966411, + 0.0012982203678414226, + 0.0012594390138238667, + 0.0012120740934088827, + 0.0011591399506293237, + 0.001117924986295402, + 0.0010595228926092387, + 0.0010102490696310997, + 0.0009533731169998645, + 0.0009156427671574056, + 0.0008670865286141634, + 0.0008243481399863959, + 0.0007827390401065349, + 0.0007445751608535648, + 0.0007058267378434539, + 0.0006572319846227765, + 0.0006192173832282424, + 0.0005905027001723647, + 0.0005582181675359607, + 0.0005225332551077009, + 0.0004948109107278287, + 0.0004635604086332023, + 0.0004460584534611553, + 0.00041995075650513175, + 0.00039428566612303257, + 0.0003732312531303614, + 0.0003609425350651145 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.13039325177669525, + -0.0001226231688633561, + -6.530806422233582e-05, + 0.0008607044583186507, + 0.00019595431513153017, + 0.0003264691913500428, + 0.0007734490791335702, + 0.00040342434658668935, + 0.0006691448506899178, + 0.0006843262817710638, + 0.0009011077927425504, + 0.0007364039774984121 + ], + "perturbation_rho": [ + 0.06105021387338638, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -8.485279977321625e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -2.5329645723104477e-05, + 9.313225746154785e-09, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -8.417712524533272e-05, + -4.190951585769653e-09, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 18.244614908467394, + "embed.bias": 91.4065704680882, + "blocks.0.ln.weight": 2.7002334851153766, + "blocks.0.w1.weight": 134.61543991177393, + "blocks.0.w1.bias": 122.71752984768428, + "blocks.0.w2.weight": 276.02165008885817, + "blocks.1.ln.weight": 6.996325721977635, + "blocks.1.w1.weight": 316.6573873211569, + "blocks.1.w1.bias": 288.7266985364082, + "blocks.1.w2.weight": 232.96436903686606, + "blocks.2.ln.weight": 3.005190384482946, + "blocks.2.w1.weight": 113.28816941109099, + "blocks.2.w1.bias": 102.79639141280903, + "blocks.2.w2.weight": 154.79558550064107, + "blocks.3.ln.weight": 3.854659793583277, + "blocks.3.w1.weight": 136.45977626032953, + "blocks.3.w1.bias": 125.02248154351406, + "blocks.3.w2.weight": 160.12595576311708, + "blocks.4.ln.weight": 5.321574474789146, + "blocks.4.w1.weight": 200.03582097403066, + "blocks.4.w1.bias": 185.1705216854366, + "blocks.4.w2.weight": 197.8192680660012, + "blocks.5.ln.weight": 3.689939791539998, + "blocks.5.w1.weight": 121.91986650696485, + "blocks.5.w1.bias": 112.80564875251632, + "blocks.5.w2.weight": 175.23886752618034, + "blocks.6.ln.weight": 5.257637367329094, + "blocks.6.w1.weight": 159.5448554465857, + "blocks.6.w1.bias": 148.9053783966513, + "blocks.6.w2.weight": 146.5770282072606, + "blocks.7.ln.weight": 4.684853931243148, + "blocks.7.w1.weight": 133.95190141354294, + "blocks.7.w1.bias": 120.52600626182299, + "blocks.7.w2.weight": 137.015420718898, + "blocks.8.ln.weight": 4.553687541635819, + "blocks.8.w1.weight": 134.4598258206248, + "blocks.8.w1.bias": 124.54115675222299, + "blocks.8.w2.weight": 138.93702924685354, + "blocks.9.ln.weight": 4.188852950738847, + "blocks.9.w1.weight": 128.35707494719534, + "blocks.9.w1.bias": 120.498165463794, + "blocks.9.w2.weight": 149.134498305582, + "blocks.10.ln.weight": 3.3037616990170426, + "blocks.10.w1.weight": 108.25092049032806, + "blocks.10.w1.bias": 101.78620102240454, + "blocks.10.w2.weight": 127.47031512506157, + "blocks.11.ln.weight": 1.6054786857904342, + "blocks.11.w1.weight": 50.83956015087797, + "blocks.11.w1.bias": 49.1835893514986, + "blocks.11.w2.weight": 90.71530778227628, + "out_ln.weight": 0.5849072301486865, + "out_head.weight": 3.217267238682831, + "out_head.bias": 0.8024946363335762 + } + }, + "credit_bridge": { + "log": { + "train_loss": [ + 2.07422056892395, + 2.055305141906738, + 2.0377359949493408, + 2.0334118033599853, + 2.034562978172302, + 2.0315364552307127, + 2.0276041582489013, + 2.029994778060913, + 2.025310488204956, + 2.0206511740875244, + 2.024435912742615, + 2.017680696372986, + 2.014923624572754, + 2.015591604614258, + 2.0193257177734374, + 2.0126227207946776, + 2.013033943634033, + 2.0076499520111084, + 2.0094668504333497, + 2.0096006536865234, + 2.0050730683135987, + 2.006946638031006, + 2.0071442542266844, + 2.004831351776123, + 2.0035909980010986, + 2.004108156204224, + 2.0006491820907595, + 2.0016160436248778, + 1.9999982108306884, + 1.9992566723251344, + 1.9979173023986816, + 1.996204802093506, + 1.9980539007568359, + 1.995016636199951, + 1.9975717867279053, + 1.9963324966812135, + 1.9947675510406495, + 1.9942235395812988, + 1.995222140235901, + 1.9917342311096191, + 1.9899134133148193, + 1.988487755279541, + 1.9872248657608031, + 1.9865933574676513, + 1.9852813260650635, + 1.9867496614837648, + 1.9831518542480469, + 1.9813233506011962, + 1.9796981923675536, + 1.9797288468933105, + 1.9774669593811036, + 1.9774521997833252, + 1.9714781398773193, + 1.9715750045013427, + 1.9715392060089112, + 1.970840647201538, + 1.9684521988296508, + 1.9670102074813842, + 1.9633512356948852, + 1.9662449013900758, + 1.9640194422149657, + 1.9607671911239624, + 1.9630818260955811, + 1.9602737934112549, + 1.960026322631836, + 1.961152883758545, + 1.9590496487426758, + 1.9572670712280273, + 1.9577932699966432, + 1.957121894493103, + 1.9570679434585572, + 1.954615853881836, + 1.9549290727996826, + 1.9529239632415771, + 1.9538047943115235, + 1.9561965101623535, + 1.9530966287994385, + 1.9537969312286376, + 1.9531251779174805, + 1.9546874990463257, + 1.9529049239730836, + 1.9550184294891357, + 1.9533232469940185, + 1.9538247448730468, + 1.9541607789611817, + 1.954350821533203, + 1.9529511869049072, + 1.9544397570037841, + 1.9555162839126587, + 1.9529233142089844, + 1.9550124865722656, + 1.954115503501892, + 1.9515360607910157, + 1.9521209592056274, + 1.9523113599014281, + 1.9537422107696534, + 1.953715394668579, + 1.952323772354126, + 1.9519077111434937, + 1.9538698761367799 + ], + "train_acc": [ + 0.2349, + 0.23648, + 0.24714, + 0.25066, + 0.24922, + 0.2532, + 0.25626, + 0.25148, + 0.2532, + 0.26078, + 0.25636, + 0.2586, + 0.26336, + 0.26146, + 0.25874, + 0.26418, + 0.2629, + 0.26778, + 0.26784, + 0.26732, + 0.26768, + 0.26668, + 0.2681, + 0.26688, + 0.26936, + 0.26936, + 0.2677, + 0.2705, + 0.2677, + 0.2709, + 0.27164, + 0.27174, + 0.27098, + 0.27192, + 0.27154, + 0.26932, + 0.27164, + 0.26854, + 0.26818, + 0.27076, + 0.26892, + 0.26664, + 0.27098, + 0.2697, + 0.2714, + 0.2725, + 0.27024, + 0.27174, + 0.27278, + 0.26878, + 0.27302, + 0.27324, + 0.27562, + 0.2744, + 0.27374, + 0.2738, + 0.27576, + 0.27626, + 0.2782, + 0.27834, + 0.27942, + 0.27984, + 0.28104, + 0.27978, + 0.28036, + 0.281, + 0.28012, + 0.28318, + 0.28276, + 0.28262, + 0.28174, + 0.28244, + 0.28384, + 0.28378, + 0.28154, + 0.2809, + 0.28316, + 0.2827, + 0.2864, + 0.2831, + 0.2849, + 0.28526, + 0.28446, + 0.28448, + 0.28456, + 0.28378, + 0.28358, + 0.28354, + 0.2834, + 0.2829, + 0.28564, + 0.28302, + 0.28616, + 0.28478, + 0.2849, + 0.28396, + 0.28816, + 0.28346, + 0.28456, + 0.28572 + ], + "test_acc": [ + 0.2422, + 0.2524, + 0.2511, + 0.2709, + 0.2522, + 0.2695, + 0.2654, + 0.2622, + 0.282, + 0.2861, + 0.259, + 0.2868, + 0.2668, + 0.2736, + 0.2902, + 0.2962, + 0.3035, + 0.2747, + 0.296, + 0.2894, + 0.2905, + 0.2902, + 0.3018, + 0.2788, + 0.2989, + 0.2832, + 0.3054, + 0.2928, + 0.2921, + 0.2723, + 0.29, + 0.281, + 0.2819, + 0.2924, + 0.3021, + 0.291, + 0.2721, + 0.2887, + 0.2827, + 0.2766, + 0.2938, + 0.2976, + 0.2883, + 0.2888, + 0.3015, + 0.2923, + 0.2893, + 0.2859, + 0.2876, + 0.2915, + 0.2857, + 0.2848, + 0.2936, + 0.2842, + 0.2815, + 0.2766, + 0.2905, + 0.2933, + 0.2846, + 0.2876, + 0.2872, + 0.2876, + 0.293, + 0.2942, + 0.2917, + 0.2816, + 0.2931, + 0.2874, + 0.2906, + 0.2935, + 0.2918, + 0.2838, + 0.2872, + 0.29, + 0.2891, + 0.2893, + 0.2908, + 0.2873, + 0.293, + 0.292, + 0.286, + 0.286, + 0.2872, + 0.2867, + 0.2876, + 0.2865, + 0.2841, + 0.2826, + 0.2844, + 0.2853, + 0.2866, + 0.2887, + 0.2856, + 0.2867, + 0.2882, + 0.2895, + 0.2885, + 0.2885, + 0.2885, + 0.2885 + ], + "value_loss": [ + 0.46219210285663603, + 0.16240292426109315, + 0.11675834127902984, + 0.10173126588344573, + 0.10192921797513962, + 0.08491676461458206, + 0.08081110213756561, + 0.10098561144590378, + 0.086660576441288, + 0.07095537901163101, + 0.07209414057731628, + 0.06507166877388953, + 0.05680496859550476, + 0.051555730621814726, + 0.05897924602746964, + 0.0477625340873003, + 0.050076086919903756, + 0.04791341652750969, + 0.043811181032657624, + 0.048990424308776856, + 0.041765395416021346, + 0.04788978707790375, + 0.038938788844347, + 0.038173083332777025, + 0.037605566675663, + 0.03472667615830898, + 0.029743025472164154, + 0.03321148899495602, + 0.031504894320964816, + 0.028381420345902443, + 0.02973802535891533, + 0.025553143047094346, + 0.028812745112776757, + 0.026098545892238617, + 0.02787227361112833, + 0.022305913635492326, + 0.02363199061870575, + 0.021854550403952597, + 0.02323986495792866, + 0.018420420945882796, + 0.018242734627723693, + 0.019795424582660197, + 0.01764130767583847, + 0.01667629898786545, + 0.016429091787040235, + 0.015772414491176605, + 0.014359855057001113, + 0.012779874252080917, + 0.012927391214668751, + 0.014349255759716033, + 0.013130034263134003, + 0.013914088915586471, + 0.010768152612149716, + 0.012323264547288418, + 0.012266730073690415, + 0.012296598086953162, + 0.011092334405183792, + 0.008943616542220116, + 0.009557433066666126, + 0.008033960376381873, + 0.008238149175047874, + 0.0075173489648103714, + 0.009243258455693721, + 0.008397357228696346, + 0.008250873034894466, + 0.009054240943193436, + 0.0061908810329437255, + 0.007204210565388202, + 0.006787694528698921, + 0.006576583880782127, + 0.0068760585397481915, + 0.00587797952502966, + 0.006058499993383884, + 0.004864574173986912, + 0.005336316513419152, + 0.004476593678146601, + 0.00504198596701026, + 0.00414283695474267, + 0.003692763058245182, + 0.0036398935145139696, + 0.0031052616691589355, + 0.003147630747780204, + 0.002547671852298081, + 0.0023914932864904405, + 0.0022162304655462504, + 0.0017571020238101483, + 0.0018844479297846555, + 0.0015003561322391033, + 0.001468994250819087, + 0.0015521602664142846, + 0.0010236410408467054, + 0.000977659215182066, + 0.001276740662381053, + 0.0008685083091259003, + 0.0012216014919430019, + 0.001092172961011529, + 0.0008087569988891483, + 0.0009069699543341994, + 0.0013558788930997253, + 0.0006715716527029872 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.37664300203323364, + 0.00668198149651289, + 0.02027757838368416, + 0.02049040049314499, + 0.020580384880304337, + 0.020434698089957237, + 0.020465239882469177, + 0.021316751837730408, + 0.021797576919198036, + 0.021945470944046974, + 0.021979298442602158, + 0.021852022036910057 + ], + "perturbation_rho": [ + 0.009359830990433693, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -9.862706065177917e-07, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -2.905726432800293e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -9.72812995314598e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 179.58500900515028, + "embed.bias": 134.71300218044718, + "blocks.0.ln.weight": 9.192845521766614, + "blocks.0.w1.weight": 201.0064270139136, + "blocks.0.w1.bias": 136.56099913744478, + "blocks.0.w2.weight": 299.7118104200996, + "blocks.1.ln.weight": 6.775393287001583, + "blocks.1.w1.weight": 298.99303772053656, + "blocks.1.w1.bias": 288.6803183648879, + "blocks.1.w2.weight": 299.96832262513726, + "blocks.2.ln.weight": 4.442030759950682, + "blocks.2.w1.weight": 156.3224513959878, + "blocks.2.w1.bias": 134.77200050699193, + "blocks.2.w2.weight": 164.77485249426002, + "blocks.3.ln.weight": 4.854857462306176, + "blocks.3.w1.weight": 165.45425459739124, + "blocks.3.w1.bias": 143.7040311765167, + "blocks.3.w2.weight": 178.2893036667054, + "blocks.4.ln.weight": 6.433765762816246, + "blocks.4.w1.weight": 262.9945171233052, + "blocks.4.w1.bias": 237.2497554637253, + "blocks.4.w2.weight": 247.70202715443386, + "blocks.5.ln.weight": 4.669370190359616, + "blocks.5.w1.weight": 179.65430058156437, + "blocks.5.w1.bias": 167.17352257399838, + "blocks.5.w2.weight": 176.56809805245783, + "blocks.6.ln.weight": 5.836574580577738, + "blocks.6.w1.weight": 236.71076201812485, + "blocks.6.w1.bias": 216.59863578081908, + "blocks.6.w2.weight": 222.1449202786447, + "blocks.7.ln.weight": 7.017034323466784, + "blocks.7.w1.weight": 283.45578137085835, + "blocks.7.w1.bias": 258.2162159056869, + "blocks.7.w2.weight": 275.54620177391195, + "blocks.8.ln.weight": 6.320927520653432, + "blocks.8.w1.weight": 252.95263320055057, + "blocks.8.w1.bias": 230.26130735012416, + "blocks.8.w2.weight": 239.07712197948655, + "blocks.9.ln.weight": 4.771737132124388, + "blocks.9.w1.weight": 184.2226928948631, + "blocks.9.w1.bias": 168.12677483943122, + "blocks.9.w2.weight": 169.7918496964596, + "blocks.10.ln.weight": 6.193554347477001, + "blocks.10.w1.weight": 248.40930737157998, + "blocks.10.w1.bias": 221.7518821584096, + "blocks.10.w2.weight": 238.1444968107213, + "blocks.11.ln.weight": 5.976649802588373, + "blocks.11.w1.weight": 240.99115354219526, + "blocks.11.w1.bias": 220.01049332622438, + "blocks.11.w2.weight": 226.72197924893874, + "out_ln.weight": 0.47882108182026845, + "out_head.weight": 7.4188430237353336, + "out_head.bias": 1.3782394013786212 + } + } + }, + "config": { + "dataset": "cifar10", + "d_hidden": 512, + "num_blocks": 12, + "batch_size": 128, + "epochs": 100, + "lr": 0.001, + "lr_fb": 0.001, + "wd": 0.01, + "lam": 0.1, + "K": 4, + "sigma_bridge": 0.05, + "ema_momentum": 0.995, + "term_grad_weight": 1.0, + "seeds": [ + 123 + ], + "gpu": 0, + "output_dir": "results/cifar10_seed123", + "num_classes": 10 + } +}
\ No newline at end of file diff --git a/results/cifar10_seed456/results_cifar10.json b/results/cifar10_seed456/results_cifar10.json new file mode 100644 index 0000000..06a3107 --- /dev/null +++ b/results/cifar10_seed456/results_cifar10.json @@ -0,0 +1,1985 @@ +{ + "456": { + "bp": { + "log": { + "train_loss": [ + 1.932577501449585, + 1.6900249602508546, + 1.6011367318725587, + 1.5466754390716553, + 1.4920579125213622, + 1.4546358205413819, + 1.4231391117858887, + 1.3938725402450562, + 1.3629425603866576, + 1.342505719642639, + 1.3237205046844482, + 1.300122989501953, + 1.2825427363967896, + 1.2623162283325196, + 1.2426594482040405, + 1.227365916481018, + 1.206711524925232, + 1.190613479347229, + 1.1721268761444092, + 1.1622920285797118, + 1.1426683759689331, + 1.1272672763824463, + 1.1123097320747375, + 1.095245484790802, + 1.0796047261238098, + 1.0664704471588136, + 1.0544639612579345, + 1.0334472966003418, + 1.0216301371002197, + 1.0061453499603272, + 0.9892120691299439, + 0.9690513248062134, + 0.9597649993705749, + 0.9393893640518188, + 0.9262461110305786, + 0.9142684574699402, + 0.8982764770317078, + 0.882060844745636, + 0.8589154948616028, + 0.8484143830871582, + 0.8298149911689758, + 0.8117531860733033, + 0.7923431532669067, + 0.7819426927185058, + 0.7630360532951355, + 0.7432640241622925, + 0.725852869091034, + 0.7111906971359253, + 0.6900048795890809, + 0.6697983954048157, + 0.6541705364608764, + 0.6370760979270935, + 0.6198028992843628, + 0.5972342865371704, + 0.5801713646697998, + 0.5637391392326355, + 0.5492425328063965, + 0.5351220473480225, + 0.517744523601532, + 0.4910259468269348, + 0.48044670740127565, + 0.4666313941383362, + 0.4454871296596527, + 0.4287984812068939, + 0.4167753814029694, + 0.4019090121555328, + 0.3892905419158936, + 0.3745353739929199, + 0.3589991714382172, + 0.34674671503067017, + 0.3343308783531189, + 0.32089566977500916, + 0.3101300811481476, + 0.2954042288208008, + 0.2848380702114105, + 0.2828438286781311, + 0.264346894235611, + 0.26094838644981383, + 0.25063199663162233, + 0.24315124837875365, + 0.2345473566555977, + 0.22507079369306565, + 0.2194176360464096, + 0.21162880234241485, + 0.20445122440338134, + 0.2012126348876953, + 0.19840873272895812, + 0.19022230600357057, + 0.1889617627120018, + 0.18509810159683227, + 0.18254041098594664, + 0.1836097541332245, + 0.17741597022533417, + 0.1746750330734253, + 0.17486595999717713, + 0.17480743077754973, + 0.17309590301036834, + 0.17227657488822937, + 0.1740615609550476, + 0.16915745746612548 + ], + "train_acc": [ + 0.30628, + 0.38592, + 0.42152, + 0.4394, + 0.45908, + 0.47766, + 0.48482, + 0.49718, + 0.51004, + 0.51544, + 0.52028, + 0.53156, + 0.53764, + 0.54346, + 0.55342, + 0.55874, + 0.56678, + 0.571, + 0.57738, + 0.57968, + 0.59098, + 0.59448, + 0.59972, + 0.60702, + 0.61296, + 0.61686, + 0.62066, + 0.629, + 0.63218, + 0.63706, + 0.6458, + 0.65254, + 0.6557, + 0.66064, + 0.66696, + 0.67284, + 0.67792, + 0.68442, + 0.6895, + 0.69484, + 0.69998, + 0.70706, + 0.71588, + 0.71756, + 0.72552, + 0.7343, + 0.74024, + 0.74394, + 0.75236, + 0.76022, + 0.76464, + 0.77054, + 0.7777, + 0.78372, + 0.79538, + 0.7972, + 0.80364, + 0.8092, + 0.81504, + 0.82654, + 0.82918, + 0.83316, + 0.84092, + 0.84854, + 0.85006, + 0.85778, + 0.8616, + 0.8676, + 0.87464, + 0.87672, + 0.88258, + 0.88796, + 0.89154, + 0.89712, + 0.90188, + 0.90184, + 0.90858, + 0.90758, + 0.9118, + 0.91542, + 0.9187, + 0.92286, + 0.9233, + 0.92722, + 0.93012, + 0.93172, + 0.93268, + 0.936, + 0.93634, + 0.9374, + 0.93696, + 0.93802, + 0.94054, + 0.94164, + 0.9409, + 0.9415, + 0.94154, + 0.94296, + 0.94158, + 0.94336 + ], + "test_acc": [ + 0.3915, + 0.4289, + 0.4595, + 0.4775, + 0.4915, + 0.4936, + 0.5132, + 0.5139, + 0.5291, + 0.5317, + 0.5407, + 0.5449, + 0.5481, + 0.5499, + 0.5545, + 0.5589, + 0.5696, + 0.5674, + 0.5737, + 0.5776, + 0.573, + 0.583, + 0.5773, + 0.572, + 0.5836, + 0.5774, + 0.5838, + 0.5796, + 0.5884, + 0.5942, + 0.5925, + 0.5927, + 0.5974, + 0.5934, + 0.5962, + 0.5972, + 0.5995, + 0.595, + 0.5921, + 0.5848, + 0.5907, + 0.5934, + 0.5935, + 0.5921, + 0.5969, + 0.5936, + 0.5961, + 0.5948, + 0.5992, + 0.5954, + 0.5978, + 0.5933, + 0.5934, + 0.5987, + 0.5987, + 0.591, + 0.5889, + 0.5924, + 0.5937, + 0.593, + 0.5925, + 0.5944, + 0.5938, + 0.5924, + 0.5946, + 0.5905, + 0.5928, + 0.5899, + 0.5891, + 0.587, + 0.5891, + 0.5846, + 0.5867, + 0.586, + 0.5865, + 0.585, + 0.5851, + 0.5824, + 0.5821, + 0.5855, + 0.5825, + 0.5861, + 0.5857, + 0.5869, + 0.5855, + 0.589, + 0.5887, + 0.5872, + 0.5863, + 0.5857, + 0.5865, + 0.5868, + 0.5867, + 0.5874, + 0.5872, + 0.5876, + 0.5869, + 0.5879, + 0.5871, + 0.587 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.9672282934188843, + 0.9684309363365173, + 0.9675205945968628, + 0.9669639468193054, + 0.9665698409080505, + 0.967890739440918, + 0.9664402008056641, + 0.9660602807998657, + 0.9652657508850098, + 0.9614400863647461, + 0.9602784514427185, + 0.9578725099563599 + ], + "perturbation_rho": [ + 0.9848342537879944, + 0.9891709685325623, + 0.9909648895263672, + 0.9893825054168701, + 0.9924193024635315, + 0.9927305579185486, + 0.9941403865814209, + 0.9934987425804138, + 0.9928717017173767, + 0.9915444254875183, + 0.9898866415023804, + 0.9873980283737183 + ], + "nudging": { + "0.001": [ + -0.0028492475394159555, + -0.0028252778574824333, + -0.0028643091209232807, + -0.0029328675009310246, + -0.002945749554783106, + -0.0029587389435619116, + -0.00288290623575449, + -0.0028087471146136522, + -0.0026325639337301254, + -0.002420567674562335, + -0.002228461904451251, + -0.0018789597088471055 + ], + "0.003": [ + -0.008540419861674309, + -0.008468939922749996, + -0.008585674688220024, + -0.008790891617536545, + -0.008829280734062195, + -0.008868295699357986, + -0.00864122249186039, + -0.008419105783104897, + -0.007891927845776081, + -0.007256433367729187, + -0.006680862978100777, + -0.0056340815499424934 + ], + "0.01": [ + -0.028379585593938828, + -0.028143716976046562, + -0.028530558571219444, + -0.02920939214527607, + -0.02933822199702263, + -0.029468875378370285, + -0.028714874759316444, + -0.027982797473669052, + -0.026236172765493393, + -0.024128224700689316, + -0.022218093276023865, + -0.018745578825473785 + ] + } + }, + "drift": { + "embed.weight": 8.834297843591171, + "embed.bias": 15.413548194165312, + "blocks.0.ln.weight": 0.3280907780755125, + "blocks.0.w1.weight": 4.220952328018163, + "blocks.0.w1.bias": 6.7337611045609105, + "blocks.0.w2.weight": 10.671510043086428, + "blocks.1.ln.weight": 0.22467724575060716, + "blocks.1.w1.weight": 4.523437520518719, + "blocks.1.w1.bias": 4.4537544678846235, + "blocks.1.w2.weight": 11.922866714123732, + "blocks.2.ln.weight": 0.19572857698834295, + "blocks.2.w1.weight": 4.598553685356083, + "blocks.2.w1.bias": 4.125724114621679, + "blocks.2.w2.weight": 12.310215404426595, + "blocks.3.ln.weight": 0.21390126577946425, + "blocks.3.w1.weight": 4.584201339373962, + "blocks.3.w1.bias": 3.7580021741696004, + "blocks.3.w2.weight": 12.601184193203743, + "blocks.4.ln.weight": 0.24161530581374482, + "blocks.4.w1.weight": 4.518574188984943, + "blocks.4.w1.bias": 3.8030594737653343, + "blocks.4.w2.weight": 12.669272886895255, + "blocks.5.ln.weight": 0.25422945739387653, + "blocks.5.w1.weight": 4.461826473966918, + "blocks.5.w1.bias": 3.6451227277370744, + "blocks.5.w2.weight": 12.680759003338304, + "blocks.6.ln.weight": 0.25745358599301726, + "blocks.6.w1.weight": 4.47036651988774, + "blocks.6.w1.bias": 3.706163791416811, + "blocks.6.w2.weight": 13.119190581706045, + "blocks.7.ln.weight": 0.263116520949555, + "blocks.7.w1.weight": 4.434269376799208, + "blocks.7.w1.bias": 3.6960182135036708, + "blocks.7.w2.weight": 13.098144437096936, + "blocks.8.ln.weight": 0.27514230251440513, + "blocks.8.w1.weight": 4.344487791539744, + "blocks.8.w1.bias": 3.648654542354885, + "blocks.8.w2.weight": 13.035629633632155, + "blocks.9.ln.weight": 0.29172129249550904, + "blocks.9.w1.weight": 4.255456615430207, + "blocks.9.w1.bias": 3.6968063474376733, + "blocks.9.w2.weight": 13.086873727145367, + "blocks.10.ln.weight": 0.3000076159356563, + "blocks.10.w1.weight": 4.2734266858824945, + "blocks.10.w1.bias": 3.6213422959644603, + "blocks.10.w2.weight": 13.200754189456744, + "blocks.11.ln.weight": 0.2902311063472059, + "blocks.11.w1.weight": 4.144956242341598, + "blocks.11.w1.bias": 3.5713523930610735, + "blocks.11.w2.weight": 13.254643701043216, + "out_ln.weight": 0.3155177407468068, + "out_head.weight": 3.523856665191762, + "out_head.bias": 1.0793116101743525 + } + }, + "dfa": { + "log": { + "train_loss": [ + 2.07058688041687, + 2.0397231711578367, + 2.040001149024963, + 2.036939584312439, + 2.035007101402283, + 2.0326563871002197, + 2.0309880726623537, + 2.028131689300537, + 2.0242818844223023, + 2.021325262145996, + 2.01944966506958, + 2.0185741081237794, + 2.0193587062072753, + 2.0140840717315673, + 2.01566628868103, + 2.0162557276153565, + 2.01352592338562, + 2.0122515713882447, + 2.007948448524475, + 2.0104500115966797, + 2.008844228897095, + 2.007831936798096, + 2.0087617783355713, + 2.0067150954437256, + 2.0060685678100585, + 2.0067182321548462, + 2.004691079711914, + 2.004318997192383, + 2.006708196105957, + 2.0059428287506105, + 2.002060938720703, + 2.0030168786239626, + 2.0013856718444822, + 2.001412846107483, + 2.0029963275909424, + 2.001156611251831, + 2.0020238697052, + 2.0013174156951905, + 2.001038536148071, + 1.9974690299224853, + 1.9983380174255372, + 2.00068127532959, + 2.002073029937744, + 1.9994858475494384, + 1.999331954574585, + 1.9974100791168212, + 1.9981875708389283, + 1.9986576483917236, + 1.9978187020492553, + 1.9974584699249267, + 1.9957377856063843, + 1.995005799217224, + 1.9989547797012328, + 1.997181711654663, + 1.995425957107544, + 1.9960723597717285, + 1.996761396522522, + 1.9958302701187134, + 1.9968557986450195, + 1.9969658164215087, + 1.9949297325134276, + 1.9950204270172118, + 1.9955382054901123, + 1.9956761499786377, + 1.9941519859313965, + 1.9952024503326415, + 1.995510894241333, + 1.9926798336791993, + 1.9947078677749635, + 1.9926326573181152, + 1.9923986828231812, + 1.993803144607544, + 1.9924064002227784, + 1.9935845022583008, + 1.993097350692749, + 1.9917376801300049, + 1.9939163592147826, + 1.9913390029144287, + 1.9931303585052491, + 1.992877590637207, + 1.9936017736434937, + 1.9915303264999389, + 1.9928937450408934, + 1.9911755990600586, + 1.9931563230133056, + 1.992949102935791, + 1.9930351052856445, + 1.9907214881134032, + 1.9905328282165526, + 1.9933595889282227, + 1.9914652145385743, + 1.99171005859375, + 1.9924902798080444, + 1.992575800704956, + 1.99016278049469, + 1.9921005247497559, + 1.9918115587615968, + 1.9918856573867798, + 1.988608092727661, + 1.9906916750717163 + ], + "train_acc": [ + 0.23438, + 0.24682, + 0.24856, + 0.24744, + 0.2493, + 0.25268, + 0.25096, + 0.25544, + 0.25558, + 0.25914, + 0.2613, + 0.2586, + 0.26004, + 0.26336, + 0.2627, + 0.26136, + 0.26178, + 0.26482, + 0.26748, + 0.26462, + 0.26374, + 0.26504, + 0.26682, + 0.26656, + 0.26808, + 0.26636, + 0.26534, + 0.26986, + 0.26678, + 0.26804, + 0.27068, + 0.27112, + 0.2728, + 0.27114, + 0.27098, + 0.27242, + 0.27134, + 0.269, + 0.27124, + 0.27356, + 0.27462, + 0.27154, + 0.27048, + 0.27332, + 0.27462, + 0.2742, + 0.2748, + 0.27136, + 0.27366, + 0.27372, + 0.2756, + 0.2745, + 0.27688, + 0.2753, + 0.2738, + 0.27568, + 0.27534, + 0.2752, + 0.2746, + 0.27426, + 0.2754, + 0.27652, + 0.27636, + 0.27514, + 0.27602, + 0.2772, + 0.27466, + 0.27792, + 0.27762, + 0.27772, + 0.2748, + 0.27744, + 0.27738, + 0.2778, + 0.27648, + 0.27882, + 0.27752, + 0.2773, + 0.27792, + 0.27792, + 0.27564, + 0.2783, + 0.2765, + 0.27966, + 0.27638, + 0.27776, + 0.27614, + 0.27868, + 0.27806, + 0.27834, + 0.27876, + 0.2785, + 0.27914, + 0.27822, + 0.27976, + 0.2771, + 0.2782, + 0.27906, + 0.27974, + 0.278 + ], + "test_acc": [ + 0.267, + 0.2833, + 0.2633, + 0.2623, + 0.2722, + 0.2805, + 0.2629, + 0.2946, + 0.2836, + 0.2778, + 0.2841, + 0.2871, + 0.2812, + 0.2675, + 0.2747, + 0.2994, + 0.2768, + 0.2969, + 0.2894, + 0.3029, + 0.2895, + 0.296, + 0.2693, + 0.2842, + 0.2805, + 0.2916, + 0.2823, + 0.2769, + 0.2952, + 0.2914, + 0.2952, + 0.2872, + 0.2933, + 0.2701, + 0.2877, + 0.2897, + 0.2824, + 0.2808, + 0.2978, + 0.2775, + 0.2873, + 0.2938, + 0.2942, + 0.2895, + 0.2924, + 0.2896, + 0.2957, + 0.2915, + 0.294, + 0.2914, + 0.2868, + 0.2964, + 0.2861, + 0.2952, + 0.2823, + 0.3038, + 0.2965, + 0.3008, + 0.2928, + 0.2991, + 0.2963, + 0.295, + 0.2983, + 0.2957, + 0.2997, + 0.2953, + 0.2891, + 0.2982, + 0.3011, + 0.2934, + 0.2948, + 0.305, + 0.2971, + 0.3022, + 0.2961, + 0.2952, + 0.3052, + 0.2981, + 0.2997, + 0.2923, + 0.2967, + 0.299, + 0.2967, + 0.2969, + 0.2986, + 0.2949, + 0.2994, + 0.2954, + 0.2994, + 0.2989, + 0.2964, + 0.2977, + 0.2943, + 0.2971, + 0.296, + 0.2965, + 0.296, + 0.2968, + 0.2968, + 0.2967 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.39808189868927, + -0.0005002215621061623, + -0.00010176157957175747, + -0.00011934398207813501, + -0.000299518636893481, + 0.0004310115473344922, + -0.0001575749192852527, + -0.00013478680921252817, + -0.00013778718130197376, + -0.0002350577706238255, + -7.549257134087384e-06, + -3.051616658922285e-05 + ], + "perturbation_rho": [ + 0.026208851486444473, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -3.781169652938843e-07, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -9.867362678050995e-07, + 0.0, + -1.1175870895385742e-08, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + -1.862645149230957e-09, + 0.0 + ], + "0.01": [ + -3.6079436540603638e-06, + 0.0, + -1.30385160446167e-08, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + -1.862645149230957e-09, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 330.294227499211, + "embed.bias": 259.4924618202614, + "blocks.0.ln.weight": 10.952609998429608, + "blocks.0.w1.weight": 280.8978034197629, + "blocks.0.w1.bias": 260.7432178686956, + "blocks.0.w2.weight": 459.75981344197356, + "blocks.1.ln.weight": 8.338033915232403, + "blocks.1.w1.weight": 303.52507755893413, + "blocks.1.w1.bias": 305.583927901598, + "blocks.1.w2.weight": 333.8270370300826, + "blocks.2.ln.weight": 10.70947460327803, + "blocks.2.w1.weight": 423.89849941751544, + "blocks.2.w1.bias": 420.3062621668287, + "blocks.2.w2.weight": 429.4082983364606, + "blocks.3.ln.weight": 8.290240736902255, + "blocks.3.w1.weight": 317.09537244180814, + "blocks.3.w1.bias": 287.7440748790634, + "blocks.3.w2.weight": 298.4352829162209, + "blocks.4.ln.weight": 9.625651853706938, + "blocks.4.w1.weight": 380.24334326208, + "blocks.4.w1.bias": 350.1027945318976, + "blocks.4.w2.weight": 351.4361032986557, + "blocks.5.ln.weight": 9.64751089559273, + "blocks.5.w1.weight": 381.41558228811317, + "blocks.5.w1.bias": 348.41973459641565, + "blocks.5.w2.weight": 358.39732437217214, + "blocks.6.ln.weight": 9.747382743882953, + "blocks.6.w1.weight": 390.26899720474756, + "blocks.6.w1.bias": 371.3396255333559, + "blocks.6.w2.weight": 383.2513961535749, + "blocks.7.ln.weight": 8.825122691585458, + "blocks.7.w1.weight": 346.68999846263586, + "blocks.7.w1.bias": 322.5261718218226, + "blocks.7.w2.weight": 338.7459123423596, + "blocks.8.ln.weight": 8.098539333504899, + "blocks.8.w1.weight": 318.2267389133952, + "blocks.8.w1.bias": 297.59966764417567, + "blocks.8.w2.weight": 295.4097011432017, + "blocks.9.ln.weight": 7.141109927159286, + "blocks.9.w1.weight": 262.58513708371515, + "blocks.9.w1.bias": 251.1580247152912, + "blocks.9.w2.weight": 243.52122031424597, + "blocks.10.ln.weight": 10.631727490611157, + "blocks.10.w1.weight": 421.8663669312304, + "blocks.10.w1.bias": 393.793856832655, + "blocks.10.w2.weight": 426.1507190170269, + "blocks.11.ln.weight": 11.204882830494729, + "blocks.11.w1.weight": 446.3738461267888, + "blocks.11.w1.bias": 412.5154467592355, + "blocks.11.w2.weight": 426.65837546735577, + "out_ln.weight": 0.6667026319780115, + "out_head.weight": 9.48630928590299, + "out_head.bias": 0.44493816450857304 + } + }, + "state_bridge": { + "log": { + "train_loss": [ + 2.3018941720581054, + 2.2664404193878176, + 2.241582181625366, + 2.2375785102844237, + 2.134529279785156, + 2.119604638824463, + 2.112871742706299, + 2.1424258409881594, + 2.158468070068359, + 2.143250986175537, + 2.1479166007995607, + 2.141409687652588, + 2.1494772535705566, + 2.1696451356506348, + 2.2107076875305176, + 2.216366109008789, + 2.2074287937927246, + 2.1900475843811034, + 2.1331632147216797, + 2.130491197357178, + 2.1317693274688723, + 2.1269900135040283, + 2.1228091914367675, + 2.099304881210327, + 2.085384643859863, + 2.085484384918213, + 2.087498538970947, + 2.0871505451202395, + 2.0823512071990966, + 2.0860876009368896, + 2.086019044647217, + 2.0951096993255613, + 2.0864567431259156, + 2.078469701538086, + 2.072162138824463, + 2.0764309199523927, + 2.0748440283203125, + 2.0789254627990723, + 2.0826576986694336, + 2.0871692906188963, + 2.086170325393677, + 2.084945154571533, + 2.0844455377197266, + 2.085197698135376, + 2.0900804764556886, + 2.082403769226074, + 2.082503260040283, + 2.0882599214935302, + 2.0945081477355956, + 2.0928621436309816, + 2.091693960571289, + 2.101140612335205, + 2.095073269882202, + 2.0920650315856935, + 2.091407716522217, + 2.092963116760254, + 2.102848686065674, + 2.10209814743042, + 2.100728988342285, + 2.1085473413085936, + 2.1115043045043946, + 2.1238191069030763, + 2.1308215964508057, + 2.1344855577087403, + 2.1272054124450683, + 2.1218687773132325, + 2.12194469039917, + 2.127710237426758, + 2.1268114820861816, + 2.1254331665039063, + 2.1245197431182863, + 2.125965220489502, + 2.12684192817688, + 2.1273793972015382, + 2.128570475311279, + 2.1321925525665284, + 2.13282696762085, + 2.1370943976593018, + 2.1397155253601072, + 2.143300206756592, + 2.1476288512420654, + 2.152116884613037, + 2.1553422878265383, + 2.154549729385376, + 2.155469595031738, + 2.155468685760498, + 2.15419854057312, + 2.1527544927978517, + 2.151362074584961, + 2.1510060634613035, + 2.1480546588134763, + 2.1468420223999023, + 2.1457022499084473, + 2.1455462963104246, + 2.144765656814575, + 2.146289216003418, + 2.142656379470825, + 2.1456896996307373, + 2.1448883264923095, + 2.144138221282959 + ], + "train_acc": [ + 0.13994, + 0.14018, + 0.1448, + 0.1448, + 0.16884, + 0.17316, + 0.1748, + 0.1631, + 0.15902, + 0.1599, + 0.15558, + 0.16126, + 0.16064, + 0.15736, + 0.1483, + 0.14638, + 0.15088, + 0.15526, + 0.16726, + 0.16864, + 0.1672, + 0.17562, + 0.1804, + 0.1857, + 0.19408, + 0.19066, + 0.1875, + 0.19146, + 0.19404, + 0.18652, + 0.1868, + 0.18316, + 0.1862, + 0.19086, + 0.19266, + 0.18878, + 0.19202, + 0.18684, + 0.18708, + 0.18076, + 0.18768, + 0.18408, + 0.18372, + 0.18428, + 0.18508, + 0.18442, + 0.1863, + 0.18434, + 0.18142, + 0.17946, + 0.18168, + 0.18058, + 0.1805, + 0.18198, + 0.18268, + 0.1812, + 0.17942, + 0.17946, + 0.18054, + 0.17744, + 0.17834, + 0.17516, + 0.17416, + 0.17206, + 0.17546, + 0.17682, + 0.17458, + 0.17444, + 0.1757, + 0.17654, + 0.1778, + 0.17268, + 0.1757, + 0.1751, + 0.17666, + 0.1775, + 0.17746, + 0.17622, + 0.17406, + 0.17472, + 0.17496, + 0.17596, + 0.17324, + 0.17394, + 0.17302, + 0.1713, + 0.17072, + 0.17288, + 0.1718, + 0.17242, + 0.1725, + 0.17408, + 0.1736, + 0.17198, + 0.17344, + 0.17254, + 0.17392, + 0.17376, + 0.17466, + 0.1739 + ], + "test_acc": [ + 0.0608, + 0.1367, + 0.1165, + 0.1638, + 0.1577, + 0.176, + 0.1529, + 0.136, + 0.1352, + 0.1245, + 0.1322, + 0.137, + 0.1201, + 0.1178, + 0.1187, + 0.1118, + 0.1187, + 0.1373, + 0.1421, + 0.1151, + 0.1412, + 0.1284, + 0.1374, + 0.1796, + 0.1927, + 0.1818, + 0.1859, + 0.1762, + 0.1865, + 0.1827, + 0.1708, + 0.1826, + 0.1859, + 0.191, + 0.1833, + 0.1858, + 0.1875, + 0.1902, + 0.1859, + 0.1872, + 0.1894, + 0.193, + 0.1882, + 0.1783, + 0.1792, + 0.179, + 0.1836, + 0.1618, + 0.1749, + 0.1881, + 0.1879, + 0.1757, + 0.1705, + 0.1868, + 0.1871, + 0.1811, + 0.1727, + 0.16, + 0.1799, + 0.1524, + 0.1648, + 0.1445, + 0.1739, + 0.1721, + 0.1828, + 0.1857, + 0.1719, + 0.1722, + 0.1716, + 0.1694, + 0.1512, + 0.1719, + 0.1704, + 0.1747, + 0.1718, + 0.1712, + 0.173, + 0.175, + 0.1733, + 0.1749, + 0.165, + 0.1672, + 0.1708, + 0.161, + 0.155, + 0.1605, + 0.1629, + 0.1598, + 0.1625, + 0.1628, + 0.1625, + 0.1609, + 0.1603, + 0.1616, + 0.1607, + 0.1644, + 0.1622, + 0.1621, + 0.1639, + 0.1637 + ], + "state_pred_error": [ + 0.4747012776184082, + 0.8295380383682252, + 0.21469590087890625, + 0.12448142728805542, + 0.08570515047073364, + 0.08470476587295532, + 0.12634123967647554, + 0.09733817555308343, + 0.0353814222663641, + 0.03443111445069313, + 0.03137257686555386, + 0.029351851464509963, + 0.03382749609231949, + 0.03600793175697327, + 0.028298149601221085, + 0.02450420175075531, + 0.025809087952375413, + 0.033925389347076416, + 0.039575903968811034, + 0.03546395682752133, + 0.03896381706476212, + 0.035592052783966065, + 0.020547916915416718, + 0.027013523270487786, + 0.024955789625644684, + 0.01776892081618309, + 0.010803042232394219, + 0.008741003857254981, + 0.009176803540587425, + 0.00826898938536644, + 0.007538356708586216, + 0.0065028233194351194, + 0.006696901108622551, + 0.006364495917409658, + 0.004348094679266215, + 0.0027382008469104767, + 0.001712799895554781, + 0.001062507140468806, + 0.0006733776653371752, + 0.0004975279747322202, + 0.0005460627448558807, + 0.0007091301748156548, + 0.0008156654949113727, + 0.0005741439528390766, + 0.0004632441856153309, + 0.0004077366246469319, + 0.00037149273524992167, + 0.00032670746902935206, + 0.0003252365568652749, + 0.0002759585575200617, + 0.00019474619667977095, + 0.00015623675198294222, + 0.00016898068460635842, + 0.00014236534488853067, + 0.00010944471210241317, + 8.656780364457518e-05, + 8.127526033436879e-05, + 7.092417092993855e-05, + 7.672614848706871e-05, + 6.656669022282586e-05, + 6.597209698753432e-05, + 6.993545684730634e-05, + 7.107698564883322e-05, + 7.049443874042482e-05, + 6.95053919684142e-05, + 6.41730681899935e-05, + 6.155454550869764e-05, + 6.015078729018569e-05, + 5.4422928835265336e-05, + 4.830317224375904e-05, + 4.409612072864547e-05, + 3.9640857164049524e-05, + 3.758983974112198e-05, + 3.4887928910320625e-05, + 3.156908231205307e-05, + 2.8848007169435732e-05, + 2.6238742939895017e-05, + 2.2889468831708655e-05, + 2.021120209712535e-05, + 1.8343317908002065e-05, + 1.5500704940350262e-05, + 1.386145944125019e-05, + 1.3288812919927295e-05, + 1.3635638843989e-05, + 1.4378732023760676e-05, + 1.5836126625945327e-05, + 1.6757641757139935e-05, + 1.6754487288417295e-05, + 1.670292949071154e-05, + 1.5964577784761785e-05, + 1.5669193891226315e-05, + 1.4928079603705555e-05, + 1.4118930798722432e-05, + 1.3290679465862922e-05, + 1.259699832968181e-05, + 1.169607178599108e-05, + 1.0905416535097175e-05, + 9.99030330684036e-06, + 9.378302599070595e-06, + 8.833123445947422e-06 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.07415470480918884, + 0.005267417058348656, + 0.006899588741362095, + 0.009260526858270168, + 0.014165668748319149, + 0.007070831023156643, + 0.003572859801352024, + 0.0013162376126274467, + 0.01693638414144516, + 0.011615054681897163, + 0.01297210343182087, + 0.019879184663295746 + ], + "perturbation_rho": [ + 0.06713278591632843, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -3.7383288145065308e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -1.1199153959751129e-05, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -3.7682242691516876e-05, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 16.995931312362433, + "embed.bias": 107.50455214257622, + "blocks.0.ln.weight": 1.1747809607171977, + "blocks.0.w1.weight": 87.48144222271058, + "blocks.0.w1.bias": 108.71150020387606, + "blocks.0.w2.weight": 231.25407676027953, + "blocks.1.ln.weight": 2.5949076659471078, + "blocks.1.w1.weight": 171.7945773092806, + "blocks.1.w1.bias": 169.71996864021074, + "blocks.1.w2.weight": 213.0622571876725, + "blocks.2.ln.weight": 3.47458321612997, + "blocks.2.w1.weight": 89.72539771461543, + "blocks.2.w1.bias": 64.43223038220326, + "blocks.2.w2.weight": 130.15474280855415, + "blocks.3.ln.weight": 3.2490652039574544, + "blocks.3.w1.weight": 125.1742859456183, + "blocks.3.w1.bias": 103.82426516529186, + "blocks.3.w2.weight": 122.83893982946046, + "blocks.4.ln.weight": 3.143539271549031, + "blocks.4.w1.weight": 124.0384309768722, + "blocks.4.w1.bias": 100.70299850760337, + "blocks.4.w2.weight": 149.19128092745697, + "blocks.5.ln.weight": 3.8369736269462678, + "blocks.5.w1.weight": 70.0473609489046, + "blocks.5.w1.bias": 55.88760753160443, + "blocks.5.w2.weight": 86.95978496337405, + "blocks.6.ln.weight": 2.788527931096308, + "blocks.6.w1.weight": 68.53157012655858, + "blocks.6.w1.bias": 60.09690095738387, + "blocks.6.w2.weight": 89.81505023786389, + "blocks.7.ln.weight": 6.399420463964318, + "blocks.7.w1.weight": 211.5238472865523, + "blocks.7.w1.bias": 193.8907626146654, + "blocks.7.w2.weight": 198.88821232136138, + "blocks.8.ln.weight": 3.718593242696815, + "blocks.8.w1.weight": 131.87660631337303, + "blocks.8.w1.bias": 117.13821442329933, + "blocks.8.w2.weight": 175.03912755183418, + "blocks.9.ln.weight": 2.399620577204472, + "blocks.9.w1.weight": 62.45119981781294, + "blocks.9.w1.bias": 48.049978443761816, + "blocks.9.w2.weight": 88.05873456328695, + "blocks.10.ln.weight": 4.012307554325392, + "blocks.10.w1.weight": 150.00438163426782, + "blocks.10.w1.bias": 134.3950208692614, + "blocks.10.w2.weight": 137.82521611752918, + "blocks.11.ln.weight": 2.857283266360585, + "blocks.11.w1.weight": 108.3646369610001, + "blocks.11.w1.bias": 100.36750299157434, + "blocks.11.w2.weight": 117.58094348226459, + "out_ln.weight": 0.45424689776008886, + "out_head.weight": 5.496813200883646, + "out_head.bias": 3.0092640143628198 + } + }, + "credit_bridge": { + "log": { + "train_loss": [ + 2.080038804626465, + 2.0486898488616943, + 2.046282894744873, + 2.046656800842285, + 2.039552648086548, + 2.0396751835632325, + 2.037381534881592, + 2.035384338226318, + 2.035809118041992, + 2.0331781942749023, + 2.0284877285385132, + 2.027495804901123, + 2.026125898208618, + 2.020480899810791, + 2.0207955992126463, + 2.0185290005874634, + 2.017150591125488, + 2.0164879455566407, + 2.0157563932037355, + 2.0127871045684813, + 2.0112040578460695, + 2.010815860519409, + 2.009171763458252, + 2.007995718460083, + 2.0075271183013914, + 2.006510923538208, + 2.003063318405151, + 2.004764850692749, + 2.0030145164489745, + 2.001830623397827, + 2.0015824402618407, + 2.0029561541748047, + 2.0005920767593386, + 1.9996838995361328, + 2.0028357711029052, + 2.0006233686065675, + 2.002561650390625, + 2.0042890097808836, + 2.0027454135513305, + 2.0027183213806152, + 2.0006818413543703, + 2.0015592734146117, + 1.9975921591949464, + 1.9961159555053711, + 1.9909970508575439, + 1.9920473384857178, + 1.9899114678192138, + 1.9886371549987794, + 1.9901284057998658, + 1.9869096327972413, + 1.9846498122406007, + 1.986077251586914, + 1.9829593098449707, + 1.9857685099792481, + 1.9820374918365478, + 1.9831326490783692, + 1.9809515007781981, + 1.9830830813980103, + 1.9793231730651855, + 1.9804900155639649, + 1.9807616325378419, + 1.9796239783096314, + 1.9790023094177247, + 1.976966079940796, + 1.978210344467163, + 1.9777095768737794, + 1.9761795248794556, + 1.9757038269042968, + 1.9747423070907593, + 1.9738406778717041, + 1.9735038549041748, + 1.97461932346344, + 1.9734378964233399, + 1.9701263136291505, + 1.9708808054351807, + 1.974146450958252, + 1.9722303295135497, + 1.9706154293823241, + 1.968834009361267, + 1.970496709022522, + 1.9716806288909912, + 1.9713043495941163, + 1.9710750988769532, + 1.9707429261016847, + 1.9708539364242554, + 1.9677796157455445, + 1.969852265663147, + 1.9710520673370362, + 1.970287728881836, + 1.9704584929656983, + 1.9702265856552124, + 1.9696108251571656, + 1.9689367693328856, + 1.9707266571044921, + 1.967924552001953, + 1.9694965203857422, + 1.9685778493881225, + 1.9698442002868652, + 1.9704585931015015, + 1.9684682028961182 + ], + "train_acc": [ + 0.23092, + 0.24236, + 0.24134, + 0.2427, + 0.24468, + 0.24838, + 0.25104, + 0.24966, + 0.24992, + 0.2533, + 0.25824, + 0.25604, + 0.25956, + 0.25866, + 0.25884, + 0.26064, + 0.26036, + 0.26042, + 0.26344, + 0.26316, + 0.26514, + 0.26496, + 0.26788, + 0.26702, + 0.26586, + 0.2673, + 0.27052, + 0.26806, + 0.26714, + 0.26944, + 0.26906, + 0.26936, + 0.27068, + 0.27078, + 0.26864, + 0.26786, + 0.26666, + 0.26666, + 0.26544, + 0.26578, + 0.26854, + 0.26384, + 0.26708, + 0.26708, + 0.2699, + 0.26658, + 0.2711, + 0.26876, + 0.26902, + 0.27082, + 0.27148, + 0.266, + 0.27216, + 0.26798, + 0.27106, + 0.26994, + 0.27048, + 0.26962, + 0.2742, + 0.27182, + 0.27056, + 0.27532, + 0.27016, + 0.2717, + 0.2714, + 0.27288, + 0.274, + 0.27434, + 0.27378, + 0.27492, + 0.27316, + 0.27746, + 0.27542, + 0.27628, + 0.27614, + 0.27594, + 0.27926, + 0.27974, + 0.28028, + 0.27698, + 0.27918, + 0.28146, + 0.27674, + 0.27706, + 0.2764, + 0.27804, + 0.27652, + 0.27794, + 0.27888, + 0.27896, + 0.27918, + 0.2798, + 0.27814, + 0.27842, + 0.2812, + 0.28084, + 0.27868, + 0.2795, + 0.27768, + 0.28226 + ], + "test_acc": [ + 0.258, + 0.2756, + 0.2609, + 0.2627, + 0.2583, + 0.2669, + 0.2661, + 0.2481, + 0.2517, + 0.278, + 0.2729, + 0.2822, + 0.2748, + 0.2742, + 0.2787, + 0.2778, + 0.2707, + 0.2794, + 0.2583, + 0.2854, + 0.2723, + 0.2857, + 0.2747, + 0.2855, + 0.2884, + 0.2972, + 0.3, + 0.2951, + 0.278, + 0.3009, + 0.2933, + 0.2909, + 0.2962, + 0.2866, + 0.2697, + 0.2857, + 0.2673, + 0.2944, + 0.2914, + 0.2886, + 0.2744, + 0.2794, + 0.286, + 0.2811, + 0.2905, + 0.2843, + 0.2743, + 0.2756, + 0.279, + 0.2945, + 0.2845, + 0.2711, + 0.2852, + 0.2824, + 0.2842, + 0.2934, + 0.2924, + 0.287, + 0.2801, + 0.2987, + 0.282, + 0.2911, + 0.2793, + 0.2753, + 0.2963, + 0.2823, + 0.2872, + 0.2971, + 0.2958, + 0.2753, + 0.285, + 0.2899, + 0.2999, + 0.2834, + 0.2931, + 0.2934, + 0.2869, + 0.2886, + 0.2916, + 0.2905, + 0.2822, + 0.2883, + 0.2927, + 0.3001, + 0.295, + 0.2954, + 0.2931, + 0.2848, + 0.2906, + 0.2907, + 0.291, + 0.2922, + 0.2929, + 0.2921, + 0.2908, + 0.2913, + 0.29, + 0.2895, + 0.2894, + 0.2893 + ], + "value_loss": [ + 0.4853640641069412, + 0.13146528634786606, + 0.10278774485111236, + 0.10451403197526932, + 0.09186984123706818, + 0.09725872883081436, + 0.08755186069607734, + 0.07744781212091446, + 0.08400812898874283, + 0.0792426191854477, + 0.07378534253954887, + 0.06448388241052627, + 0.06442826303482056, + 0.062765171585083, + 0.05268099952340126, + 0.046547150465250015, + 0.04977830931901932, + 0.046939543528556824, + 0.04594481763601303, + 0.04108059468030929, + 0.04083769999027252, + 0.04402205483198166, + 0.04000264639854431, + 0.03796569113850594, + 0.035944590538740155, + 0.03808709916114807, + 0.03455143286049366, + 0.038163829424381256, + 0.03331702991783619, + 0.033281468888521196, + 0.02748794436097145, + 0.030440516521930694, + 0.028207264231443405, + 0.02704140617221594, + 0.026097478928565978, + 0.025327355717718603, + 0.024099903733730315, + 0.027659756012558937, + 0.022987231121063233, + 0.020967545130252837, + 0.021579559311568738, + 0.020051995359063147, + 0.01838304534137249, + 0.02002642317146063, + 0.017698023419380187, + 0.018389571142196654, + 0.013522476999759675, + 0.01383736302614212, + 0.014502173573970795, + 0.016614098087251186, + 0.013939322325885297, + 0.014496620377600192, + 0.010598833594471216, + 0.012348202850222588, + 0.011290375513732433, + 0.009235630955398083, + 0.011637351908683776, + 0.00849102252393961, + 0.009930187560021878, + 0.0093223687428236, + 0.008464423409998418, + 0.007654866740107537, + 0.007954280150234699, + 0.00844777645289898, + 0.008321732766479253, + 0.007419437230825424, + 0.008567752386033535, + 0.007364120395183563, + 0.006373357061594725, + 0.00595162553101778, + 0.006675690564513207, + 0.0059091412272304295, + 0.0058950428107380866, + 0.005543788010179997, + 0.004646000351905823, + 0.004232099659517407, + 0.00391991232007742, + 0.003104630061686039, + 0.0029366533225774765, + 0.0028203235547989607, + 0.0027620538933947683, + 0.002386382454186678, + 0.0019722532033920287, + 0.001968352645486593, + 0.001672238507643342, + 0.0015232293362915515, + 0.0017160395857319236, + 0.0012631227846443653, + 0.0011414727784320711, + 0.0009458377653546632, + 0.0011035545341856778, + 0.0009393164394423365, + 0.001123441938534379, + 0.0008937438951432705, + 0.0009866366090252995, + 0.0011745557901449502, + 0.0006501818283647299, + 0.0009478944011032582, + 0.001031599847562611, + 0.0008783576951175928 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.4526920020580292, + 0.006497844588011503, + 0.01631792262196541, + 0.01694033294916153, + 0.017034951597452164, + 0.01775282248854637, + 0.01775078848004341, + 0.017723549157381058, + 0.01778416708111763, + 0.01786961778998375, + 0.017997052520513535, + 0.017978299409151077 + ], + "perturbation_rho": [ + -0.054250095039606094, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -8.353963494300842e-07, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -2.439599484205246e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -7.801689207553864e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 193.53208266596525, + "embed.bias": 115.78380404718655, + "blocks.0.ln.weight": 8.867635038294205, + "blocks.0.w1.weight": 196.51094816310516, + "blocks.0.w1.bias": 74.82582896315267, + "blocks.0.w2.weight": 317.40620630343955, + "blocks.1.ln.weight": 5.764116391727315, + "blocks.1.w1.weight": 210.05221683875882, + "blocks.1.w1.bias": 200.7126628347933, + "blocks.1.w2.weight": 235.05000374612118, + "blocks.2.ln.weight": 7.735844453849748, + "blocks.2.w1.weight": 328.46033970353545, + "blocks.2.w1.bias": 322.5548152476072, + "blocks.2.w2.weight": 334.35863169713303, + "blocks.3.ln.weight": 5.5455094513045, + "blocks.3.w1.weight": 213.2636693536328, + "blocks.3.w1.bias": 189.3065053401033, + "blocks.3.w2.weight": 201.9218429504037, + "blocks.4.ln.weight": 6.963102540839665, + "blocks.4.w1.weight": 281.1829353200361, + "blocks.4.w1.bias": 254.14827278048986, + "blocks.4.w2.weight": 252.72823154554072, + "blocks.5.ln.weight": 6.651053338270696, + "blocks.5.w1.weight": 265.808423173078, + "blocks.5.w1.bias": 237.94000901426764, + "blocks.5.w2.weight": 246.35312384551574, + "blocks.6.ln.weight": 6.796230689261986, + "blocks.6.w1.weight": 277.37913499866477, + "blocks.6.w1.bias": 258.5087504829616, + "blocks.6.w2.weight": 266.3321474533456, + "blocks.7.ln.weight": 6.190933150629552, + "blocks.7.w1.weight": 247.84752961611798, + "blocks.7.w1.bias": 225.69508996072392, + "blocks.7.w2.weight": 237.71747748473527, + "blocks.8.ln.weight": 5.572316870464946, + "blocks.8.w1.weight": 220.49723448227533, + "blocks.8.w1.bias": 201.84500185584645, + "blocks.8.w2.weight": 203.8926634145943, + "blocks.9.ln.weight": 4.956248620217523, + "blocks.9.w1.weight": 176.87866597616605, + "blocks.9.w1.bias": 165.25106759218716, + "blocks.9.w2.weight": 166.2448646005026, + "blocks.10.ln.weight": 7.452869833755125, + "blocks.10.w1.weight": 298.53765102170973, + "blocks.10.w1.bias": 273.411202411899, + "blocks.10.w2.weight": 295.16401679878913, + "blocks.11.ln.weight": 7.774196063602293, + "blocks.11.w1.weight": 313.80618162900083, + "blocks.11.w1.bias": 283.94177751762567, + "blocks.11.w2.weight": 294.8394911964541, + "out_ln.weight": 0.5766241941627794, + "out_head.weight": 8.360089920565844, + "out_head.bias": 1.5207819852306046 + } + } + }, + "config": { + "dataset": "cifar10", + "d_hidden": 512, + "num_blocks": 12, + "batch_size": 128, + "epochs": 100, + "lr": 0.001, + "lr_fb": 0.001, + "wd": 0.01, + "lam": 0.1, + "K": 4, + "sigma_bridge": 0.05, + "ema_momentum": 0.995, + "term_grad_weight": 1.0, + "seeds": [ + 456 + ], + "gpu": 0, + "output_dir": "results/cifar10_seed456", + "num_classes": 10 + } +}
\ No newline at end of file diff --git a/results/toy_lq_frozen/toy_lq_v2_seed123_lam0.1_sig0.1_tgw1.0_fm0.0.json b/results/toy_lq_frozen/toy_lq_v2_seed123_lam0.1_sig0.1_tgw1.0_fm0.0.json new file mode 100644 index 0000000..50eed4c --- /dev/null +++ b/results/toy_lq_frozen/toy_lq_v2_seed123_lam0.1_sig0.1_tgw1.0_fm0.0.json @@ -0,0 +1,266 @@ +{ + "config": { + "d_hidden": 64, + "output_dim": 10, + "num_layers": 12, + "sigma": 0.03, + "batch_size": 256, + "num_steps": 8000, + "lr_fb": 0.001, + "lam": 0.1, + "K": 8, + "ema_momentum": 0.995, + "sigma_bridge": 0.1, + "eval_every": 2000, + "seed": 123, + "gpu": 0, + "output_dir": "results/toy_lq_frozen", + "vnet_hidden": 256, + "vnet_layers": 3, + "term_grad_weight": 1.0, + "fm_weight": 0.0 + }, + "log": { + "steps": [ + 1, + 2000, + 4000, + 6000, + 8000 + ], + "dfa_costate_cos": [ + 0.0061469420325011015, + 0.005306758529817064, + 0.0032278880632172027, + 0.010428216774016619, + 0.005160536073769133 + ], + "state_costate_cos": [ + 0.04923563698927561, + 0.9359788099924723, + 0.93496473133564, + 0.9391002754370371, + 0.9422274927298228 + ], + "credit_costate_cos": [ + 0.005350367398932576, + 0.9094806561867396, + 0.933747818072637, + 0.939527784784635, + 0.9428367614746094 + ], + "dfa_rho": [ + 0.014851124413932363, + -0.0009741405956447124, + 0.010637809444839755, + 0.01579231140203774, + 0.0006830055887500445 + ], + "state_rho": [ + 0.05303209243963162, + 0.9233627518018087, + 0.9250594973564148, + 0.9248303671677908, + 0.9291473726431528 + ], + "credit_rho": [ + 8.900166722014546e-05, + 0.8767146865526835, + 0.9206246634324392, + 0.9250840544700623, + 0.9295186003049215 + ], + "dfa_nudge": [ + -0.0020856610499322414, + -0.0016959290175388257, + -0.0004252120852470398, + -0.00336669214690725, + -0.0014625868449608486 + ], + "state_nudge": [ + -0.017791483240822952, + -0.32088013738393784, + -0.3230740974346797, + -0.318187120060126, + -0.324822299182415 + ], + "credit_nudge": [ + -0.0004618208234508832, + -0.31195973853270215, + -0.3216395452618599, + -0.3171723609169324, + -0.3237730637192726 + ], + "bridge_residual": [], + "state_bridge_loss": [ + 1.0102367401123047, + 0.029022935777902603, + 0.031384147703647614, + 0.02959899604320526, + 0.031896352767944336 + ], + "credit_bridge_loss": [ + 129.2601776123047, + 8.807141304016113, + 8.771026611328125, + 7.9213643074035645, + 8.751266479492188 + ], + "term_loss": [ + 109.68403625488281, + 4.207324028015137, + 4.038488388061523, + 3.279881477355957, + 3.9708642959594727 + ], + "bridge_loss": [ + 5.943464884694549e-07, + 0.11741068214178085, + 0.08691056072711945, + 0.10882671922445297, + 0.12579748034477234 + ], + "term_grad_loss": [ + 19.57614517211914, + 4.4824066162109375, + 4.645627975463867, + 4.532656192779541, + 4.654605388641357 + ], + "fm_loss": [ + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + }, + "final_per_layer": { + "dfa_costate_cos": [ + 0.04813992977142334, + -0.042106956243515015, + 0.004819205962121487, + 0.06406223028898239, + 0.027982598170638084, + -0.024616792798042297, + 0.03184102475643158, + -0.03284747526049614, + -0.031094228848814964, + -0.04093015938997269, + 0.012475000694394112, + 0.0442020557820797 + ], + "state_costate_cos": [ + 0.9390542507171631, + 0.9402485489845276, + 0.9412673711776733, + 0.9416499137878418, + 0.9424928426742554, + 0.9428901672363281, + 0.9427830576896667, + 0.9428685903549194, + 0.9430990219116211, + 0.9434692859649658, + 0.9433966875076294, + 0.9435101747512817 + ], + "credit_costate_cos": [ + 0.9391632676124573, + 0.9400047659873962, + 0.9408236742019653, + 0.9413831233978271, + 0.9420697689056396, + 0.942909836769104, + 0.9431338906288147, + 0.943551778793335, + 0.9441136121749878, + 0.9448702335357666, + 0.9456644654273987, + 0.9463527202606201 + ], + "dfa_rho": [ + 0.0134794432669878, + -0.042476922273635864, + 0.04209040105342865, + 0.06388744711875916, + -0.012325622141361237, + -0.025234051048755646, + -0.020182523876428604, + -0.02071167342364788, + -0.049487367272377014, + -0.05043957754969597, + 0.04616815596818924, + 0.0634283572435379 + ], + "state_rho": [ + 0.9241164922714233, + 0.9311657547950745, + 0.926814079284668, + 0.9336568117141724, + 0.9266120791435242, + 0.9352933764457703, + 0.9264969825744629, + 0.9290511012077332, + 0.9249886870384216, + 0.9313560128211975, + 0.9301149845123291, + 0.9301021099090576 + ], + "credit_rho": [ + 0.9276995658874512, + 0.9273138046264648, + 0.9248120784759521, + 0.9268039464950562, + 0.92667555809021, + 0.9278458952903748, + 0.9272637367248535, + 0.9345529079437256, + 0.9366906881332397, + 0.9292744398117065, + 0.9313026070594788, + 0.9339879751205444 + ], + "dfa_nudge": [ + -0.017539501190185547, + 0.01593128778040409, + -0.0030720066279172897, + -0.020152313634753227, + -0.009087346494197845, + 0.007799159735441208, + -0.010878749191761017, + 0.013225046917796135, + 0.011669810861349106, + 0.017475932836532593, + -0.006174879148602486, + -0.016747483983635902 + ], + "state_nudge": [ + -0.32301613688468933, + -0.3241724371910095, + -0.32460731267929077, + -0.32475656270980835, + -0.325019896030426, + -0.3253987431526184, + -0.32468554377555847, + -0.3243948817253113, + -0.3248230516910553, + -0.32554829120635986, + -0.3256106674671173, + -0.3258340656757355 + ], + "credit_nudge": [ + -0.32093775272369385, + -0.32208389043807983, + -0.32268720865249634, + -0.3229762613773346, + -0.3233621120452881, + -0.324078768491745, + -0.32360684871673584, + -0.32358717918395996, + -0.3243335783481598, + -0.32536792755126953, + -0.3258592486381531, + -0.32639598846435547 + ] + } +}
\ No newline at end of file diff --git a/results/toy_lq_frozen/toy_lq_v2_seed42_lam0.1_sig0.1_tgw1.0_fm0.0.json b/results/toy_lq_frozen/toy_lq_v2_seed42_lam0.1_sig0.1_tgw1.0_fm0.0.json new file mode 100644 index 0000000..bbcd5c4 --- /dev/null +++ b/results/toy_lq_frozen/toy_lq_v2_seed42_lam0.1_sig0.1_tgw1.0_fm0.0.json @@ -0,0 +1,266 @@ +{ + "config": { + "d_hidden": 64, + "output_dim": 10, + "num_layers": 12, + "sigma": 0.03, + "batch_size": 256, + "num_steps": 8000, + "lr_fb": 0.001, + "lam": 0.1, + "K": 8, + "ema_momentum": 0.995, + "sigma_bridge": 0.1, + "eval_every": 2000, + "seed": 42, + "gpu": 0, + "output_dir": "results/toy_lq_frozen", + "vnet_hidden": 256, + "vnet_layers": 3, + "term_grad_weight": 1.0, + "fm_weight": 0.0 + }, + "log": { + "steps": [ + 1, + 2000, + 4000, + 6000, + 8000 + ], + "dfa_costate_cos": [ + 0.001022340264171362, + 0.001144224622597297, + 0.0035498210539420447, + 0.0025704174380128584, + 0.0012041970621794462 + ], + "state_costate_cos": [ + 0.009041048353537917, + 0.9367498010396957, + 0.9443944742282232, + 0.9488719999790192, + 0.9465398589769999 + ], + "credit_costate_cos": [ + 0.024892715892444055, + 0.9009914000829061, + 0.9385832945505778, + 0.947337324420611, + 0.9453924099604288 + ], + "dfa_rho": [ + 0.015879416760678094, + 0.003749566695963343, + 0.0025427089616035423, + 0.0041017483454197645, + -0.006658251901778082 + ], + "state_rho": [ + 0.0029661240599428615, + 0.9205784201622009, + 0.9331746151049932, + 0.9367670218149821, + 0.9348721752564112 + ], + "credit_rho": [ + 0.02234963719577839, + 0.8549301077922186, + 0.9257398645083109, + 0.9335627655188242, + 0.9330280274152756 + ], + "dfa_nudge": [ + -0.0003799900102118651, + -0.0001569713931530714, + -0.0012285423775513966, + -0.00047482581188281375, + -0.0003635381193210681 + ], + "state_nudge": [ + -0.002246703254058957, + -0.3365800951917966, + -0.3378218387564023, + -0.34102949251731235, + -0.3362140009800593 + ], + "credit_nudge": [ + -0.0079942528779308, + -0.32416996111472446, + -0.3345658630132675, + -0.3391987532377243, + -0.33437975496053696 + ], + "bridge_residual": [], + "state_bridge_loss": [ + 1.0073249340057373, + 0.03208087012171745, + 0.02979956567287445, + 0.03251900523900986, + 0.03256681188941002 + ], + "credit_bridge_loss": [ + 132.09298706054688, + 10.374980926513672, + 10.61994743347168, + 10.229816436767578, + 9.931343078613281 + ], + "term_loss": [ + 111.63633728027344, + 5.137801647186279, + 5.388574600219727, + 5.321071147918701, + 4.950850486755371 + ], + "bridge_loss": [ + 6.45359421014291e-07, + 0.2699485719203949, + 0.12590564787387848, + 0.13655179738998413, + 0.15239471197128296 + ], + "term_grad_loss": [ + 20.456655502319336, + 4.967230796813965, + 5.105466842651367, + 4.772193431854248, + 4.828097343444824 + ], + "fm_loss": [ + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + }, + "final_per_layer": { + "dfa_costate_cos": [ + -0.053928110748529434, + -0.03558344021439552, + -0.002497725188732147, + -0.025605209171772003, + -0.0015886849723756313, + -0.004757361486554146, + 0.038880251348018646, + 0.04505593329668045, + 0.044440463185310364, + 0.049856118857860565, + -0.036305028945207596, + -0.0035168412141501904 + ], + "state_costate_cos": [ + 0.943811297416687, + 0.9446508884429932, + 0.9451277256011963, + 0.9456416368484497, + 0.9462511539459229, + 0.9469271898269653, + 0.9473466873168945, + 0.947996973991394, + 0.9481196999549866, + 0.9477524757385254, + 0.9474242925643921, + 0.947428286075592 + ], + "credit_costate_cos": [ + 0.941976010799408, + 0.9424219727516174, + 0.94321209192276, + 0.9437947273254395, + 0.9448409080505371, + 0.9454640746116638, + 0.9459144473075867, + 0.9465770721435547, + 0.9470815062522888, + 0.9474484920501709, + 0.947708010673523, + 0.9482696056365967 + ], + "dfa_rho": [ + -0.04425683990120888, + -0.06072389334440231, + -0.02133849635720253, + -0.009745828807353973, + 0.012875140644609928, + -0.06230748072266579, + -0.011334436014294624, + 0.06567166745662689, + 0.05127769336104393, + 0.02585187926888466, + -0.02806694805622101, + 0.0021985196508467197 + ], + "state_rho": [ + 0.9318734407424927, + 0.9358463883399963, + 0.9371962547302246, + 0.932350754737854, + 0.9360877871513367, + 0.9400110840797424, + 0.9344450235366821, + 0.9253479242324829, + 0.9369418621063232, + 0.9354310035705566, + 0.9379286170005798, + 0.9350059628486633 + ], + "credit_rho": [ + 0.925601065158844, + 0.9282341599464417, + 0.9294254779815674, + 0.9282456040382385, + 0.9352800846099854, + 0.9381647109985352, + 0.9306818246841431, + 0.9283095598220825, + 0.9359391331672668, + 0.9364811182022095, + 0.940323531627655, + 0.9396500587463379 + ], + "dfa_nudge": [ + 0.018354831263422966, + 0.012472891248762608, + 0.0010786037892103195, + 0.008631331846117973, + -0.000161781907081604, + 0.0025226473808288574, + -0.013589534908533096, + -0.017021682113409042, + -0.01592499390244484, + -0.017948923632502556, + 0.015213390812277794, + 0.0020107626914978027 + ], + "state_nudge": [ + -0.3381166160106659, + -0.33782684803009033, + -0.33681023120880127, + -0.3365652561187744, + -0.3364033102989197, + -0.336093544960022, + -0.33575671911239624, + -0.3358075022697449, + -0.33578699827194214, + -0.3356783986091614, + -0.3347625136375427, + -0.33496007323265076 + ], + "credit_nudge": [ + -0.3348296880722046, + -0.33472180366516113, + -0.33402615785598755, + -0.3340165615081787, + -0.33416393399238586, + -0.33410122990608215, + -0.33391931653022766, + -0.3341727554798126, + -0.334547221660614, + -0.33483123779296875, + -0.3342774510383606, + -0.33494970202445984 + ] + } +}
\ No newline at end of file diff --git a/results/toy_lq_frozen/toy_lq_v2_seed456_lam0.1_sig0.1_tgw1.0_fm0.0.json b/results/toy_lq_frozen/toy_lq_v2_seed456_lam0.1_sig0.1_tgw1.0_fm0.0.json new file mode 100644 index 0000000..7684971 --- /dev/null +++ b/results/toy_lq_frozen/toy_lq_v2_seed456_lam0.1_sig0.1_tgw1.0_fm0.0.json @@ -0,0 +1,266 @@ +{ + "config": { + "d_hidden": 64, + "output_dim": 10, + "num_layers": 12, + "sigma": 0.03, + "batch_size": 256, + "num_steps": 8000, + "lr_fb": 0.001, + "lam": 0.1, + "K": 8, + "ema_momentum": 0.995, + "sigma_bridge": 0.1, + "eval_every": 2000, + "seed": 456, + "gpu": 0, + "output_dir": "results/toy_lq_frozen", + "vnet_hidden": 256, + "vnet_layers": 3, + "term_grad_weight": 1.0, + "fm_weight": 0.0 + }, + "log": { + "steps": [ + 1, + 2000, + 4000, + 6000, + 8000 + ], + "dfa_costate_cos": [ + -0.008305357536301017, + -0.008392414000506202, + -0.0038773335108999163, + -0.005692236203079422, + -0.0027285031198213496 + ], + "state_costate_cos": [ + 0.010795300283158818, + 0.9460556507110596, + 0.9393202712138494, + 0.9483461380004883, + 0.9449077347914377 + ], + "credit_costate_cos": [ + 0.010942678588132063, + 0.9265013436476389, + 0.9366317639748255, + 0.947089821100235, + 0.9429272959629694 + ], + "dfa_rho": [ + -0.0028248391657446823, + -0.0069896553953488665, + 0.006519075793524583, + -0.0016956791902581851, + 0.009361718820097545 + ], + "state_rho": [ + 0.02673721546307206, + 0.9346788177887598, + 0.9264062345027924, + 0.9364438454310099, + 0.928318440914154 + ], + "credit_rho": [ + 0.015292729716748, + 0.8930107355117798, + 0.9212059676647186, + 0.9316875636577606, + 0.9271238495906194 + ], + "dfa_nudge": [ + 0.004384364855165283, + 0.004451872392868002, + 0.0023875126304725804, + 0.0029402777242163816, + 0.0018359140182534854 + ], + "state_nudge": [ + -0.005124294043829043, + -0.34853988885879517, + -0.36369936913251877, + -0.3535121629635493, + -0.3701842874288559 + ], + "credit_nudge": [ + -0.003232262640570601, + -0.3407190019885699, + -0.36137687663237256, + -0.35145895431439084, + -0.3679164672891299 + ], + "bridge_residual": [], + "state_bridge_loss": [ + 1.0078721046447754, + 0.029677797108888626, + 0.03204238414764404, + 0.0327574647963047, + 0.033611781895160675 + ], + "credit_bridge_loss": [ + 158.73072814941406, + 11.051497459411621, + 9.754555702209473, + 10.701624870300293, + 9.23304557800293 + ], + "term_loss": [ + 132.93673706054688, + 5.0680437088012695, + 4.583424091339111, + 5.3644843101501465, + 4.062508583068848 + ], + "bridge_loss": [ + 7.166463547036983e-07, + 0.25247713923454285, + 0.14709019660949707, + 0.1511959433555603, + 0.12251871824264526 + ], + "term_grad_loss": [ + 25.793991088867188, + 5.730976581573486, + 5.024041175842285, + 5.185944557189941, + 5.048018932342529 + ], + "fm_loss": [ + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + }, + "final_per_layer": { + "dfa_costate_cos": [ + 0.005115414969623089, + -0.0511106476187706, + 0.026345418766140938, + -0.06129831820726395, + 0.023028161376714706, + 0.043917812407016754, + -0.02071939967572689, + 0.020611021667718887, + -0.025742489844560623, + -0.052854984998703, + 0.004462959244847298, + 0.05550301447510719 + ], + "state_costate_cos": [ + 0.9431805610656738, + 0.9437761306762695, + 0.9443435072898865, + 0.9446961879730225, + 0.9448176622390747, + 0.9449890851974487, + 0.945353627204895, + 0.9453505277633667, + 0.9454029202461243, + 0.9456251859664917, + 0.945793867111206, + 0.945563554763794 + ], + "credit_costate_cos": [ + 0.9382357001304626, + 0.9392683506011963, + 0.9404538869857788, + 0.9414603114128113, + 0.9420560598373413, + 0.9428950548171997, + 0.9436970353126526, + 0.9444013833999634, + 0.9445676803588867, + 0.9454110860824585, + 0.9460842609405518, + 0.9465967416763306 + ], + "dfa_rho": [ + 0.0005743983201682568, + -0.06480635702610016, + 0.04855071008205414, + -0.09910713136196136, + 0.0678277313709259, + 0.07211608439683914, + -0.024130607023835182, + 0.001894976943731308, + -0.009040179662406445, + -0.02709592692553997, + 0.047522202134132385, + 0.09803472459316254 + ], + "state_rho": [ + 0.9245786666870117, + 0.9240697026252747, + 0.9293171167373657, + 0.9301248788833618, + 0.9324039220809937, + 0.9262233376502991, + 0.9288182854652405, + 0.9328013062477112, + 0.926672101020813, + 0.928871750831604, + 0.9281368255615234, + 0.9278033971786499 + ], + "credit_rho": [ + 0.9264800548553467, + 0.9184226989746094, + 0.9257187247276306, + 0.9245353937149048, + 0.9274380803108215, + 0.9219262599945068, + 0.9251310229301453, + 0.9257712960243225, + 0.9327481985092163, + 0.9326720833778381, + 0.9336825013160706, + 0.9309598803520203 + ], + "dfa_nudge": [ + -0.0008005127310752869, + 0.01917375810444355, + -0.009830990806221962, + 0.02491075173020363, + -0.007071066647768021, + -0.017435496672987938, + 0.009654557332396507, + -0.005623402073979378, + 0.01048743724822998, + 0.022874275222420692, + -0.0009297188371419907, + -0.02337862364947796 + ], + "state_nudge": [ + -0.3704739809036255, + -0.3704678416252136, + -0.36917200684547424, + -0.36994534730911255, + -0.3694732189178467, + -0.37023210525512695, + -0.3705016076564789, + -0.37048155069351196, + -0.36992865800857544, + -0.37050265073776245, + -0.37076336145401, + -0.37026911973953247 + ], + "credit_nudge": [ + -0.3661743402481079, + -0.3663756549358368, + -0.3655211329460144, + -0.3667157292366028, + -0.3666227161884308, + -0.36782804131507874, + -0.36846989393234253, + -0.3688392639160156, + -0.3685477077960968, + -0.36957746744155884, + -0.37019991874694824, + -0.37012574076652527 + ] + } +}
\ No newline at end of file |
