diff options
Diffstat (limited to 'results/cifar_depth_scan_s42/d512_L12_s42.json')
| -rw-r--r-- | results/cifar_depth_scan_s42/d512_L12_s42.json | 1421 |
1 files changed, 1421 insertions, 0 deletions
diff --git a/results/cifar_depth_scan_s42/d512_L12_s42.json b/results/cifar_depth_scan_s42/d512_L12_s42.json new file mode 100644 index 0000000..e551271 --- /dev/null +++ b/results/cifar_depth_scan_s42/d512_L12_s42.json @@ -0,0 +1,1421 @@ +{ + "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": [ + 2.057183644256592, + 2.0219752458190916, + 2.014074349899292, + 2.009915184326172, + 2.003742285079956, + 2.0064792041397093, + 2.001210139312744, + 2.0027513356781004, + 1.9994710051727296, + 2.0039267013549806, + 1.9983517964935302, + 1.9981007271575928, + 1.9967711097717284, + 1.9957333689880372, + 1.9961917282485961, + 1.997950509262085, + 1.9958872592163086, + 1.9964322351837158, + 1.9945315238189698, + 1.993023081741333, + 1.9933346648788453, + 1.9943037507629395, + 1.9923084811019898, + 1.9912947034454345, + 1.9911876581573487, + 1.9902342232513428, + 1.9931360609436035, + 1.9907362972640992, + 1.9919358252334596, + 1.9918178892517089, + 1.99015292427063, + 1.9888984406280517, + 1.990421195602417, + 1.9874123781585693, + 1.9881913418197632, + 1.9892149132537842, + 1.9870399390411377, + 1.988616922302246, + 1.9857829595947265, + 1.9879872678375243, + 1.9910465408325195, + 1.9890870415878297, + 1.9886888786315917, + 1.9860569588470458, + 1.9893753149414062, + 1.9879478802490234, + 1.986084238319397, + 1.9871028468322753, + 1.987663519668579, + 1.9854809813690186, + 1.987514366722107, + 1.9865355487442016, + 1.9870760778808594, + 1.9852242317199706, + 1.9868980590438843, + 1.9847661309432982, + 1.9880792671203613, + 1.9864630212402343, + 1.9864557014465332, + 1.9846630696105958, + 1.9851684265518188, + 1.9845601000976563, + 1.984690249786377, + 1.9832856945037842, + 1.9843514255523682, + 1.98395031955719, + 1.9839791688537598, + 1.9839756365203858, + 1.9832659860992432, + 1.985604306640625, + 1.9832666548156739, + 1.983997225265503, + 1.9818758429718017, + 1.9806820659637452, + 1.9815394610595702, + 1.9834156256103515, + 1.9787700980377196, + 1.982170662612915, + 1.9797498672485352, + 1.9827086754608154, + 1.9817713089370728, + 1.9803067615127563, + 1.9819608657073975, + 1.9829605780410766, + 1.9819707880401611, + 1.9838256735229491, + 1.9817811890411376, + 1.9816452063369752, + 1.9808973921203614, + 1.98217180267334, + 1.9815813362884522, + 1.981448392868042, + 1.9818018493652343, + 1.978947674217224, + 1.981713777770996, + 1.9825912845611573, + 1.9784613372421265, + 1.9797205515289307, + 1.9829137060546875, + 1.9835036209869386 + ], + "train_acc": [ + 0.24212, + 0.26076, + 0.26024, + 0.26518, + 0.26958, + 0.26636, + 0.26982, + 0.26922, + 0.27358, + 0.26904, + 0.27336, + 0.27268, + 0.27124, + 0.27456, + 0.27402, + 0.27524, + 0.27538, + 0.27344, + 0.27274, + 0.27408, + 0.27474, + 0.27582, + 0.27818, + 0.27952, + 0.27862, + 0.27686, + 0.27516, + 0.27816, + 0.2769, + 0.27774, + 0.27712, + 0.2764, + 0.27884, + 0.27974, + 0.27744, + 0.27654, + 0.28156, + 0.27792, + 0.28188, + 0.28016, + 0.27916, + 0.27818, + 0.28024, + 0.28084, + 0.27922, + 0.28046, + 0.27866, + 0.2805, + 0.27788, + 0.2798, + 0.27828, + 0.28152, + 0.2793, + 0.28102, + 0.28142, + 0.28174, + 0.2784, + 0.28092, + 0.28108, + 0.2819, + 0.27962, + 0.28176, + 0.2825, + 0.2829, + 0.283, + 0.28336, + 0.28122, + 0.28198, + 0.2826, + 0.2796, + 0.28184, + 0.28408, + 0.28324, + 0.28434, + 0.28316, + 0.28344, + 0.28512, + 0.283, + 0.28622, + 0.28278, + 0.28352, + 0.28448, + 0.28224, + 0.2846, + 0.28276, + 0.28316, + 0.28462, + 0.28342, + 0.28424, + 0.28358, + 0.2832, + 0.28652, + 0.28394, + 0.28506, + 0.2852, + 0.28386, + 0.28612, + 0.28354, + 0.28348, + 0.28488 + ], + "test_acc": [ + 0.2796, + 0.2818, + 0.3016, + 0.2892, + 0.2743, + 0.2847, + 0.2712, + 0.2939, + 0.2897, + 0.2995, + 0.2732, + 0.305, + 0.3018, + 0.3117, + 0.2831, + 0.2864, + 0.3018, + 0.2984, + 0.3147, + 0.298, + 0.2968, + 0.3081, + 0.3062, + 0.305, + 0.2994, + 0.3014, + 0.2985, + 0.315, + 0.3063, + 0.31, + 0.3086, + 0.2951, + 0.3006, + 0.3114, + 0.3053, + 0.3148, + 0.2984, + 0.3013, + 0.3048, + 0.3018, + 0.3083, + 0.2985, + 0.3067, + 0.2993, + 0.3114, + 0.3088, + 0.3051, + 0.3005, + 0.3107, + 0.3014, + 0.3002, + 0.3077, + 0.3098, + 0.3078, + 0.3139, + 0.3013, + 0.3122, + 0.2999, + 0.3094, + 0.3123, + 0.3146, + 0.3057, + 0.309, + 0.3068, + 0.3141, + 0.3044, + 0.3053, + 0.311, + 0.3116, + 0.3115, + 0.3105, + 0.3096, + 0.3132, + 0.3069, + 0.3019, + 0.3068, + 0.3023, + 0.3101, + 0.3091, + 0.3059, + 0.3075, + 0.3129, + 0.3145, + 0.3114, + 0.311, + 0.3123, + 0.307, + 0.3076, + 0.3106, + 0.3098, + 0.3089, + 0.3079, + 0.3111, + 0.3096, + 0.3102, + 0.311, + 0.3088, + 0.3092, + 0.3091, + 0.309 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.38828110694885254, + 0.0037502094637602568, + -0.0013616610085591674, + -0.0004177851078566164, + 0.0008564171148464084, + 0.00033698498737066984, + 0.0006134084542281926, + -0.00040099132456816733, + -0.00013616093201562762, + -0.0008071845513768494, + -0.0019259440014138818, + -0.0014602115843445063 + ], + "perturbation_rho": [ + -0.0468934029340744, + 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": [ + -3.948807716369629e-07, + 0.0, + -1.862645149230957e-09, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + -9.313225746154785e-10, + 0.0, + 0.0 + ], + "0.003": [ + -1.2898817658424377e-06, + -3.725290298461914e-09, + -1.862645149230957e-09, + 0.0, + 0.0, + 0.0, + -9.313225746154785e-10, + 0.0, + 0.0, + -9.313225746154785e-10, + 0.0, + 3.725290298461914e-09 + ], + "0.01": [ + -4.674773663282394e-06, + -3.725290298461914e-09, + -1.862645149230957e-09, + -3.725290298461914e-09, + 0.0, + 9.313225746154785e-10, + -9.313225746154785e-10, + -1.862645149230957e-09, + 0.0, + -9.313225746154785e-10, + 9.313225746154785e-10, + 3.725290298461914e-09 + ] + } + }, + "drift": { + "embed.weight": 319.6256145809284, + "embed.bias": 241.39131706930516, + "blocks.0.ln.weight": 8.451152686823706, + "blocks.0.w1.weight": 198.0653736517787, + "blocks.0.w1.bias": 172.12729591798578, + "blocks.0.w2.weight": 339.00018245731724, + "blocks.1.ln.weight": 7.000199186009471, + "blocks.1.w1.weight": 213.18535798636668, + "blocks.1.w1.bias": 202.05532367109083, + "blocks.1.w2.weight": 208.8135070796582, + "blocks.2.ln.weight": 7.935838334127239, + "blocks.2.w1.weight": 314.5106044563863, + "blocks.2.w1.bias": 271.6455147079551, + "blocks.2.w2.weight": 247.9045165300641, + "blocks.3.ln.weight": 5.795920066983145, + "blocks.3.w1.weight": 182.91564428516534, + "blocks.3.w1.bias": 152.5137308265898, + "blocks.3.w2.weight": 151.050379705954, + "blocks.4.ln.weight": 6.3276885496198645, + "blocks.4.w1.weight": 215.98859872586803, + "blocks.4.w1.bias": 183.97920728444788, + "blocks.4.w2.weight": 165.74279725091483, + "blocks.5.ln.weight": 5.792878750342549, + "blocks.5.w1.weight": 193.44922307145256, + "blocks.5.w1.bias": 172.0595502919292, + "blocks.5.w2.weight": 156.55012054903048, + "blocks.6.ln.weight": 5.98080042430076, + "blocks.6.w1.weight": 182.32068889300982, + "blocks.6.w1.bias": 161.74264639168877, + "blocks.6.w2.weight": 161.51005233041852, + "blocks.7.ln.weight": 6.798251040626781, + "blocks.7.w1.weight": 235.9970011878696, + "blocks.7.w1.bias": 202.73306946131746, + "blocks.7.w2.weight": 176.16251425264502, + "blocks.8.ln.weight": 8.05032468668243, + "blocks.8.w1.weight": 304.9749512658957, + "blocks.8.w1.bias": 254.78399637200565, + "blocks.8.w2.weight": 226.8419755905067, + "blocks.9.ln.weight": 8.324689211241104, + "blocks.9.w1.weight": 304.64714323529006, + "blocks.9.w1.bias": 254.27332736169836, + "blocks.9.w2.weight": 226.94604254635792, + "blocks.10.ln.weight": 5.162597908184438, + "blocks.10.w1.weight": 159.94571322633212, + "blocks.10.w1.bias": 138.12937914272587, + "blocks.10.w2.weight": 129.89821404142262, + "blocks.11.ln.weight": 7.516332199105425, + "blocks.11.w1.weight": 279.9271343014864, + "blocks.11.w1.bias": 236.42300656616862, + "blocks.11.w2.weight": 202.50408675479144, + "out_ln.weight": 0.6271483618826256, + "out_head.weight": 8.718326486393146, + "out_head.bias": 0.5753880166404655 + } + }, + "credit_bridge": { + "log": { + "train_loss": [ + 2.0568576118469237, + 2.033485134429932, + 2.0218416510772705, + 2.0142132845306397, + 2.0108241705703733, + 2.002156679153442, + 2.0038612014770507, + 1.9997895935821532, + 2.0023641468048097, + 2.0011226501846315, + 1.9980047481536864, + 1.9944430965423583, + 1.9959918328094481, + 1.9943659350585938, + 1.9955686949157714, + 1.9932625021743775, + 1.9940180532073974, + 1.992023607826233, + 1.990704457244873, + 1.9915499368286134, + 1.9944144274139404, + 1.9895493435668945, + 1.9895022438812255, + 1.9851936352157593, + 1.9862560325622558, + 1.9824435089874268, + 1.9816281591796876, + 1.9810021967315674, + 1.9812662688446046, + 1.9824039008331298, + 1.9829267932891845, + 1.9802660181427, + 1.9798483668518065, + 1.979789539527893, + 1.9795299366760253, + 1.9757647130584717, + 1.972223770904541, + 1.9694228728485108, + 1.9702133034896852, + 1.9706256021499633, + 1.9694662646102905, + 1.9712112203979493, + 1.9850462644958495, + 2.002737700653076, + 2.0086276149749756, + 2.0277000471496582, + 2.0281700329589842, + 2.0319336601257323, + 2.023782074584961, + 2.0170300160217285, + 2.0127930587005616, + 2.0128867221832274, + 2.01850300453186, + 2.0133023000335695, + 2.0104183267593383, + 2.0138787005996703, + 2.021632391052246, + 2.029548404159546, + 2.0301968224334717, + 2.0342870476531982, + 2.0400729270172118, + 2.0447248320770264, + 2.0492120973205568, + 2.0500868199157716, + 2.0525983279418947, + 2.053036574554443, + 2.055363027687073, + 2.0557615393066406, + 2.060003659210205, + 2.059733818283081, + 2.063136697921753, + 2.063323797454834, + 2.0652847254180906, + 2.065591119003296, + 2.0690882136535644, + 2.07202782333374, + 2.0749201114654543, + 2.0776294555664063, + 2.079117084503174, + 2.078432328796387, + 2.078386816864014, + 2.076197295227051, + 2.075240274810791, + 2.071873186187744, + 2.0715130477142334, + 2.070623317947388, + 2.0708017208862306, + 2.0684812643432617, + 2.067380738296509, + 2.065728325653076, + 2.0663505850982666, + 2.0655458202362063, + 2.0657374589538575, + 2.0645635641479494, + 2.0653134523010253, + 2.0636908391571045, + 2.063971264190674, + 2.0637667852783204, + 2.064012709197998, + 2.064872869949341 + ], + "train_acc": [ + 0.24184, + 0.2521, + 0.25882, + 0.26394, + 0.26408, + 0.26908, + 0.2699, + 0.2708, + 0.27198, + 0.27044, + 0.27198, + 0.27592, + 0.27358, + 0.27498, + 0.2747, + 0.27506, + 0.2745, + 0.27506, + 0.27742, + 0.2764, + 0.27738, + 0.2759, + 0.2767, + 0.27868, + 0.27842, + 0.27882, + 0.27966, + 0.27812, + 0.2803, + 0.27992, + 0.27782, + 0.28038, + 0.27874, + 0.28038, + 0.28104, + 0.28038, + 0.28444, + 0.28556, + 0.28248, + 0.28622, + 0.2816, + 0.27992, + 0.2709, + 0.26912, + 0.264, + 0.254, + 0.2571, + 0.25402, + 0.25446, + 0.25542, + 0.25666, + 0.25594, + 0.25068, + 0.25624, + 0.2585, + 0.25766, + 0.25522, + 0.24842, + 0.25234, + 0.25176, + 0.24892, + 0.24678, + 0.24418, + 0.24632, + 0.24382, + 0.2433, + 0.24304, + 0.24668, + 0.24448, + 0.24628, + 0.24476, + 0.24558, + 0.2455, + 0.25012, + 0.24546, + 0.24462, + 0.24508, + 0.2428, + 0.23992, + 0.23902, + 0.23556, + 0.23456, + 0.23548, + 0.23628, + 0.2383, + 0.23514, + 0.23668, + 0.2391, + 0.2391, + 0.23886, + 0.24188, + 0.24092, + 0.24206, + 0.24588, + 0.24146, + 0.24336, + 0.24526, + 0.24318, + 0.24306, + 0.241 + ], + "test_acc": [ + 0.2607, + 0.2707, + 0.2859, + 0.3016, + 0.277, + 0.3077, + 0.2813, + 0.309, + 0.3096, + 0.283, + 0.2983, + 0.3135, + 0.2938, + 0.3079, + 0.2923, + 0.3001, + 0.2908, + 0.2904, + 0.3037, + 0.3131, + 0.3022, + 0.3126, + 0.3113, + 0.299, + 0.2999, + 0.304, + 0.2825, + 0.3158, + 0.3181, + 0.3042, + 0.3046, + 0.2901, + 0.3147, + 0.286, + 0.3127, + 0.3144, + 0.3237, + 0.3242, + 0.3072, + 0.2959, + 0.2996, + 0.2938, + 0.2808, + 0.2722, + 0.2706, + 0.2656, + 0.2651, + 0.2755, + 0.2528, + 0.2641, + 0.2557, + 0.27, + 0.2477, + 0.2502, + 0.2324, + 0.2355, + 0.2419, + 0.2416, + 0.2607, + 0.2678, + 0.275, + 0.2598, + 0.262, + 0.2613, + 0.2586, + 0.252, + 0.26, + 0.2567, + 0.2422, + 0.2632, + 0.2589, + 0.2585, + 0.268, + 0.2485, + 0.2499, + 0.2309, + 0.2442, + 0.2451, + 0.2248, + 0.2271, + 0.2286, + 0.2357, + 0.2347, + 0.2262, + 0.2355, + 0.2376, + 0.2363, + 0.2374, + 0.2359, + 0.2392, + 0.2359, + 0.2366, + 0.2387, + 0.2367, + 0.2364, + 0.2389, + 0.2405, + 0.2399, + 0.2392, + 0.239 + ], + "value_loss": [ + 0.45757097380638123, + 0.20398492289066314, + 0.14403926629781724, + 0.12142798287391662, + 0.12356067106246948, + 0.10323704772472382, + 0.09025520797491074, + 0.07278211853384972, + 0.0721493047785759, + 0.07462056404113769, + 0.058030206863880156, + 0.051864567153453826, + 0.05143641318678856, + 0.049471625967025756, + 0.048022189247608184, + 0.04581756822347641, + 0.048392335476875305, + 0.039053659332990646, + 0.03958373228907585, + 0.039008795657157896, + 0.03813901361584664, + 0.03919893253147602, + 0.03719111958146095, + 0.033129610191583635, + 0.03316450059056282, + 0.029290390903949737, + 0.030135655524730683, + 0.03032812103867531, + 0.028851064978837968, + 0.030176216821670532, + 0.027550049550533296, + 0.024665569834709168, + 0.02101791775226593, + 0.025735651454925536, + 0.0251218857216835, + 0.02088318651020527, + 0.019842562421560288, + 0.021842277721762657, + 0.019696957327723502, + 0.039216412217617035, + 0.044192443869113925, + 0.04437062278270722, + 0.10603765468597412, + 0.5277386970996857, + 4.009285344390869, + 23.118207588806154, + 144.34815091796875, + 518.8025195117187, + 1318.1295115625, + 1762.37145984375, + 6916.118271875, + 4658.513114609375, + 2876.65883546875, + 1116.1913117382812, + 672.6694246875, + 704.66234140625, + 552.1133823925782, + 171.6432964111328, + 135.70525376220704, + 134.33308825195311, + 132.31754293945312, + 72.1968391418457, + 41.47797743041992, + 43.47026284912109, + 36.734252014160155, + 25.379087185058594, + 18.40212633392334, + 29.28388342590332, + 10.164241724700927, + 2.897604249019623, + 1.047686642589569, + 0.325347878112793, + 0.14909634278774261, + 0.07973611199140548, + 0.06035885235905647, + 0.04197409330606461, + 0.028365342289209367, + 0.017914904502928257, + 0.01303940503180027, + 0.009001612865626812, + 0.006971689679771661, + 0.005598475634455681, + 0.005475701079368591, + 0.004648658817410469, + 0.0038136353914439676, + 0.0030602712966501713, + 0.0028339786121994257, + 0.0029074203512072564, + 0.0026963330183178186, + 0.0023759404706954956, + 0.002058511268571019, + 0.002492244097441435, + 0.0014459992367774248, + 0.0019114786966145038, + 0.0025138171672821046, + 0.00124135938256979, + 0.0020614514429681003, + 0.0013046426071599126, + 0.0029164830996282398, + 0.0010984712824225426 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.20759254693984985, + 0.005885659717023373, + 0.014463091269135475, + 0.013757804408669472, + 0.01659775897860527, + 0.0177934467792511, + 0.018012138083577156, + 0.019089948385953903, + 0.019284311681985855, + 0.01838410645723343, + 0.01891487091779709, + 0.018453549593687057 + ], + "perturbation_rho": [ + 0.010340061970055103, + 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": [ + -4.516914486885071e-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": [ + -1.214444637298584e-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": [ + -4.189088940620422e-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": 150.67057498445888, + "embed.bias": 134.41454679929575, + "blocks.0.ln.weight": 5.823816560727751, + "blocks.0.w1.weight": 121.67399642558134, + "blocks.0.w1.bias": 113.51970790036184, + "blocks.0.w2.weight": 204.33462559926585, + "blocks.1.ln.weight": 6.501127217482716, + "blocks.1.w1.weight": 285.45091938635363, + "blocks.1.w1.bias": 289.6326970920111, + "blocks.1.w2.weight": 243.81894015436046, + "blocks.2.ln.weight": 6.326138219917707, + "blocks.2.w1.weight": 324.84264337850595, + "blocks.2.w1.bias": 293.5437947059724, + "blocks.2.w2.weight": 224.099887924036, + "blocks.3.ln.weight": 5.602491317116855, + "blocks.3.w1.weight": 245.36011562278688, + "blocks.3.w1.bias": 220.93326847278095, + "blocks.3.w2.weight": 200.84652256624017, + "blocks.4.ln.weight": 5.329743882869523, + "blocks.4.w1.weight": 182.97700815621855, + "blocks.4.w1.bias": 140.02593230649856, + "blocks.4.w2.weight": 140.60772145327962, + "blocks.5.ln.weight": 5.286696439762922, + "blocks.5.w1.weight": 155.33931234812562, + "blocks.5.w1.bias": 93.8828284250912, + "blocks.5.w2.weight": 133.6255221433454, + "blocks.6.ln.weight": 5.3698026086033686, + "blocks.6.w1.weight": 153.69676226206676, + "blocks.6.w1.bias": 104.50919556116708, + "blocks.6.w2.weight": 146.8099489527109, + "blocks.7.ln.weight": 5.580818732892067, + "blocks.7.w1.weight": 159.38306611743945, + "blocks.7.w1.bias": 101.31052511081181, + "blocks.7.w2.weight": 104.61639494957436, + "blocks.8.ln.weight": 6.117239553334436, + "blocks.8.w1.weight": 351.20193371314235, + "blocks.8.w1.bias": 322.90500949025466, + "blocks.8.w2.weight": 205.93771553542751, + "blocks.9.ln.weight": 5.428160484749458, + "blocks.9.w1.weight": 151.20229062994952, + "blocks.9.w1.bias": 86.27513547236327, + "blocks.9.w2.weight": 119.94171519954084, + "blocks.10.ln.weight": 3.187901359176882, + "blocks.10.w1.weight": 100.37314795265215, + "blocks.10.w1.bias": 86.90571689186352, + "blocks.10.w2.weight": 116.36697077856738, + "blocks.11.ln.weight": 4.855792448009543, + "blocks.11.w1.weight": 290.16810994044624, + "blocks.11.w1.bias": 269.48650679370706, + "blocks.11.w2.weight": 207.9959909476014, + "out_ln.weight": 0.44311953950689703, + "out_head.weight": 7.617038267805974, + "out_head.bias": 2.5026283939923624 + } + } +}
\ No newline at end of file |
