diff options
| author | YurenHao0426 <Blackhao0426@gmail.com> | 2026-06-14 04:06:32 -0500 |
|---|---|---|
| committer | YurenHao0426 <Blackhao0426@gmail.com> | 2026-06-14 04:06:32 -0500 |
| commit | aa73718eb6427d7da3b9cb416275802d90c4b2ed (patch) | |
| tree | b68b0a664fb650744ef934a1c22abd740a7b62a6 /results/cifar_depth_scan_s42 | |
| parent | 827c658fa9a750f3c6ebdb87703762f10f69f6ff (diff) | |
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Diffstat (limited to 'results/cifar_depth_scan_s42')
| -rw-r--r-- | results/cifar_depth_scan_s42/d512_L12_s42.json | 1421 | ||||
| -rw-r--r-- | results/cifar_depth_scan_s42/d512_L2_s42.json | 1151 | ||||
| -rw-r--r-- | results/cifar_depth_scan_s42/d512_L4_s42.json | 1205 | ||||
| -rw-r--r-- | results/cifar_depth_scan_s42/d512_L6_s42.json | 1259 | ||||
| -rw-r--r-- | results/cifar_depth_scan_s42/d512_L8_s42.json | 1313 | ||||
| -rw-r--r-- | results/cifar_depth_scan_s42/summary.json | 186 |
6 files changed, 6535 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 diff --git a/results/cifar_depth_scan_s42/d512_L2_s42.json b/results/cifar_depth_scan_s42/d512_L2_s42.json new file mode 100644 index 0000000..5b1a950 --- /dev/null +++ b/results/cifar_depth_scan_s42/d512_L2_s42.json @@ -0,0 +1,1151 @@ +{ + "bp": { + "log": { + "train_loss": [ + 1.9354850341033936, + 1.7026564234542847, + 1.6196284805679322, + 1.5522706447601318, + 1.5103712742996216, + 1.4760794692230224, + 1.4480401036453248, + 1.4238936393356323, + 1.3998644092941284, + 1.3888274967193603, + 1.370794108505249, + 1.3523651966094972, + 1.3417378455352784, + 1.32940377494812, + 1.3123712255859374, + 1.3025306620025634, + 1.2916712799072265, + 1.277412696762085, + 1.2670877025604248, + 1.2579533868026733, + 1.250452767906189, + 1.2388696469116212, + 1.2288407207870484, + 1.218021198196411, + 1.2081415139007567, + 1.2038527836227417, + 1.1919150150299072, + 1.1843040127944946, + 1.1809553996276856, + 1.1748971781539916, + 1.1629332243728638, + 1.1559197056579589, + 1.1501819496154786, + 1.1386108488464355, + 1.1359955614471435, + 1.1256131409072876, + 1.1200060926818847, + 1.114403334388733, + 1.1042931074905395, + 1.1033966439437866, + 1.0955395514297486, + 1.086461291847229, + 1.076456173362732, + 1.0776306005477905, + 1.0669021976089477, + 1.0603959639167786, + 1.0562950397491455, + 1.0486062710762023, + 1.037363448638916, + 1.0395437144470214, + 1.0320540841293335, + 1.0248265104293823, + 1.0186307384872437, + 1.01219017370224, + 1.0086890588760375, + 1.0036477170944214, + 0.9918886923408509, + 0.9959744847869874, + 0.983810892829895, + 0.9749429069328308, + 0.9761281036186218, + 0.9669857545852661, + 0.9646618202209473, + 0.9585257618713379, + 0.951733235244751, + 0.9470721603393555, + 0.9449367933654785, + 0.9399459014511108, + 0.9332768785476685, + 0.9295426531982421, + 0.9249779984283447, + 0.9206048580551147, + 0.9163932028198242, + 0.9137818552780151, + 0.9054667686653137, + 0.8992140997695923, + 0.8981155136108399, + 0.9007723778915405, + 0.8907698515319824, + 0.8869161108398438, + 0.8879233457946777, + 0.8870189405632019, + 0.8807708535385131, + 0.875256696395874, + 0.8748971775817871, + 0.8757942477035523, + 0.8740769153022766, + 0.8715347571182251, + 0.8681148331069947, + 0.8674819102478027, + 0.863512946510315, + 0.8696056005477906, + 0.8647798666381836, + 0.8631498766708374, + 0.866932133693695, + 0.8609341511917115, + 0.8584218189620971, + 0.8627110484886169, + 0.8589248524093628, + 0.8619690019416809 + ], + "train_acc": [ + 0.31324, + 0.3856, + 0.41254, + 0.43722, + 0.4538, + 0.46698, + 0.47636, + 0.48598, + 0.49418, + 0.49864, + 0.50738, + 0.51122, + 0.51678, + 0.52288, + 0.5272, + 0.53212, + 0.53512, + 0.53952, + 0.5452, + 0.54694, + 0.54924, + 0.55342, + 0.559, + 0.56202, + 0.56542, + 0.56722, + 0.57134, + 0.57538, + 0.57742, + 0.57586, + 0.58118, + 0.58464, + 0.5859, + 0.59246, + 0.59194, + 0.59674, + 0.59898, + 0.60034, + 0.6066, + 0.60422, + 0.60806, + 0.61212, + 0.61618, + 0.61372, + 0.61884, + 0.6198, + 0.62164, + 0.62388, + 0.6294, + 0.62744, + 0.63112, + 0.63436, + 0.63332, + 0.63632, + 0.6381, + 0.6401, + 0.6463, + 0.64188, + 0.64658, + 0.65074, + 0.65084, + 0.65332, + 0.65314, + 0.65706, + 0.65862, + 0.66132, + 0.66338, + 0.66192, + 0.6661, + 0.6698, + 0.66802, + 0.66974, + 0.67114, + 0.67274, + 0.6753, + 0.67902, + 0.6777, + 0.67944, + 0.68172, + 0.68452, + 0.68178, + 0.68314, + 0.68652, + 0.68794, + 0.688, + 0.6866, + 0.68708, + 0.6896, + 0.69086, + 0.69196, + 0.69182, + 0.68984, + 0.69014, + 0.69196, + 0.69008, + 0.69314, + 0.69342, + 0.69112, + 0.6941, + 0.69128 + ], + "test_acc": [ + 0.379, + 0.42, + 0.4414, + 0.468, + 0.4821, + 0.4867, + 0.5048, + 0.5076, + 0.5032, + 0.5071, + 0.5251, + 0.5385, + 0.5339, + 0.5395, + 0.5359, + 0.5373, + 0.5411, + 0.5429, + 0.544, + 0.5558, + 0.5584, + 0.553, + 0.5587, + 0.557, + 0.5617, + 0.5595, + 0.5661, + 0.5732, + 0.5671, + 0.5738, + 0.5715, + 0.5762, + 0.5734, + 0.5772, + 0.5787, + 0.5735, + 0.5753, + 0.5788, + 0.5757, + 0.5779, + 0.5818, + 0.5857, + 0.5877, + 0.5853, + 0.5807, + 0.5926, + 0.5907, + 0.5921, + 0.5863, + 0.5905, + 0.5941, + 0.5959, + 0.5907, + 0.5931, + 0.5922, + 0.5982, + 0.5976, + 0.5888, + 0.595, + 0.5987, + 0.6, + 0.594, + 0.5975, + 0.599, + 0.5987, + 0.5921, + 0.5957, + 0.5921, + 0.5999, + 0.5947, + 0.5996, + 0.5976, + 0.5949, + 0.5992, + 0.5982, + 0.5956, + 0.6015, + 0.6011, + 0.5993, + 0.601, + 0.5974, + 0.5976, + 0.5997, + 0.599, + 0.5988, + 0.6011, + 0.5974, + 0.5985, + 0.6018, + 0.598, + 0.5993, + 0.5991, + 0.599, + 0.5992, + 0.5991, + 0.6004, + 0.5994, + 0.5993, + 0.5994, + 0.5993 + ] + }, + "diagnostics": { + "bp_cosine": [ + 1.0, + 1.0 + ], + "perturbation_rho": [ + 0.9868886470794678, + 0.9831097722053528 + ], + "nudging": { + "0.001": [ + -0.0010212662164121866, + -0.0008843992254696786 + ], + "0.003": [ + -0.0030618617311120033, + -0.0026517128571867943 + ], + "0.01": [ + -0.010182719677686691, + -0.008822238072752953 + ] + } + }, + "drift": { + "embed.weight": 8.850352053801725, + "embed.bias": 26.80721203618006, + "blocks.0.ln.weight": 0.20347274443339167, + "blocks.0.w1.weight": 4.629616202784868, + "blocks.0.w1.bias": 7.3007028332357145, + "blocks.0.w2.weight": 12.573673444728787, + "blocks.1.ln.weight": 0.15135983015831284, + "blocks.1.w1.weight": 4.835127700607666, + "blocks.1.w1.bias": 4.074910897523138, + "blocks.1.w2.weight": 15.468163387197967, + "out_ln.weight": 0.43266956541456525, + "out_head.weight": 2.288095707738499, + "out_head.bias": 1.5696642748538316 + } + }, + "dfa": { + "log": { + "train_loss": [ + 2.0506517180633543, + 2.037630374984741, + 2.0375006256103516, + 2.0347165871810913, + 2.0280541705322266, + 2.0247886113739013, + 2.023957159957886, + 2.017765091934204, + 2.013640316925049, + 2.0121789574813844, + 2.012694746131897, + 2.0094316165542603, + 2.010430689163208, + 2.0080650563812257, + 2.004681011276245, + 2.0067793479156495, + 2.002835368652344, + 2.0035605745697023, + 2.005288227157593, + 2.0020996487045286, + 2.0037673551940918, + 2.0026556689453123, + 2.0015291195678713, + 1.9998969723129272, + 2.001053342666626, + 1.9988280670166017, + 1.998517660293579, + 2.000125762634277, + 1.9984248222351073, + 2.0001446187973024, + 1.997779655456543, + 1.9962129680633545, + 1.996638311729431, + 1.9934515882873536, + 1.9944385040664674, + 1.9950137895965576, + 1.9948566608428955, + 1.9947254082489014, + 1.9938683570861817, + 1.993011798019409, + 1.993586630821228, + 1.9928539782333374, + 1.99163732421875, + 1.9935448779296876, + 1.992500286026001, + 1.9926587213134765, + 1.9939856634140014, + 1.9904723690795898, + 1.99148487449646, + 1.989919489364624, + 1.9883176875305175, + 1.9909399053192138, + 1.9917929566955566, + 1.9895172406005859, + 1.9915735732650757, + 1.98892810256958, + 1.9902667702484131, + 1.9884299784851074, + 1.9891859377288819, + 1.9871826519012452, + 1.9888171925354003, + 1.9876050499725342, + 1.9885715426635742, + 1.9879224799346924, + 1.9875414797210693, + 1.9883081774902345, + 1.9874149275970459, + 1.9873363061904907, + 1.9844847861480712, + 1.9853876978302003, + 1.9853303070449828, + 1.986023975753784, + 1.9871168149566651, + 1.9860221238708495, + 1.9863739596176146, + 1.9852040929412842, + 1.9818872883605958, + 1.9874637450408936, + 1.9860291708374023, + 1.9853426343154907, + 1.9848574941253663, + 1.985159556503296, + 1.9853443636322021, + 1.9854941542053222, + 1.9842009844970703, + 1.9843328964614868, + 1.9830964395141601, + 1.9839789056777954, + 1.9838400217437744, + 1.9850177837371825, + 1.9836825035858154, + 1.9848054162597657, + 1.9832133966827392, + 1.9829493893432617, + 1.9810011239242553, + 1.9834887173843383, + 1.9844633283233644, + 1.9830566509628296, + 1.983135018005371, + 1.9840513249969483 + ], + "train_acc": [ + 0.24914, + 0.2591, + 0.25864, + 0.25854, + 0.26322, + 0.26478, + 0.2666, + 0.2671, + 0.27048, + 0.26588, + 0.26632, + 0.27096, + 0.26978, + 0.2698, + 0.27074, + 0.27022, + 0.27292, + 0.27118, + 0.27308, + 0.27424, + 0.27086, + 0.27322, + 0.27464, + 0.2738, + 0.27538, + 0.27228, + 0.2751, + 0.27742, + 0.27544, + 0.2723, + 0.27602, + 0.27634, + 0.27544, + 0.27646, + 0.27732, + 0.27708, + 0.27726, + 0.2762, + 0.27976, + 0.27826, + 0.27656, + 0.2791, + 0.28084, + 0.27786, + 0.27708, + 0.27874, + 0.27834, + 0.28096, + 0.27838, + 0.28342, + 0.27998, + 0.27926, + 0.27846, + 0.2797, + 0.27926, + 0.28266, + 0.2793, + 0.28242, + 0.27896, + 0.27938, + 0.2807, + 0.28146, + 0.27946, + 0.279, + 0.28374, + 0.28138, + 0.28202, + 0.28122, + 0.2837, + 0.28122, + 0.28228, + 0.2817, + 0.27958, + 0.283, + 0.2825, + 0.28446, + 0.28486, + 0.28054, + 0.28292, + 0.28036, + 0.28328, + 0.28302, + 0.28552, + 0.2844, + 0.28334, + 0.28374, + 0.28518, + 0.28306, + 0.2825, + 0.28426, + 0.28316, + 0.28418, + 0.28372, + 0.285, + 0.28484, + 0.2843, + 0.28284, + 0.28358, + 0.28314, + 0.28086 + ], + "test_acc": [ + 0.2818, + 0.2871, + 0.2791, + 0.2853, + 0.2856, + 0.2861, + 0.3039, + 0.2842, + 0.2723, + 0.2875, + 0.2911, + 0.2934, + 0.3073, + 0.3023, + 0.2854, + 0.2884, + 0.3049, + 0.2901, + 0.2934, + 0.3139, + 0.3034, + 0.3021, + 0.321, + 0.2958, + 0.3122, + 0.2967, + 0.3089, + 0.3045, + 0.2876, + 0.2966, + 0.3042, + 0.2978, + 0.3161, + 0.3091, + 0.3033, + 0.3149, + 0.3106, + 0.2978, + 0.3124, + 0.3057, + 0.3115, + 0.2996, + 0.3098, + 0.303, + 0.2889, + 0.3177, + 0.3011, + 0.2992, + 0.3019, + 0.3108, + 0.3212, + 0.3085, + 0.3138, + 0.3106, + 0.296, + 0.3106, + 0.3058, + 0.3123, + 0.3076, + 0.3118, + 0.3063, + 0.3124, + 0.3079, + 0.3077, + 0.3123, + 0.3141, + 0.3109, + 0.319, + 0.3017, + 0.3049, + 0.3113, + 0.318, + 0.3157, + 0.3079, + 0.3101, + 0.3137, + 0.3058, + 0.3153, + 0.3094, + 0.3139, + 0.3087, + 0.3128, + 0.3162, + 0.3133, + 0.3087, + 0.3134, + 0.3167, + 0.3121, + 0.3118, + 0.3111, + 0.3131, + 0.3113, + 0.3128, + 0.3125, + 0.3131, + 0.3123, + 0.3118, + 0.312, + 0.3121, + 0.3122 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.396158367395401, + -0.005014405585825443 + ], + "perturbation_rho": [ + 0.0009753962513059378, + 0.0 + ], + "nudging": { + "0.001": [ + -4.3818727135658264e-07, + 0.0 + ], + "0.003": [ + -1.3783574104309082e-06, + 9.313225746154785e-10 + ], + "0.01": [ + -4.818197339773178e-06, + 3.725290298461914e-09 + ] + } + }, + "drift": { + "embed.weight": 323.8257144939098, + "embed.bias": 248.21108220164808, + "blocks.0.ln.weight": 8.42464670192897, + "blocks.0.w1.weight": 206.7713721666247, + "blocks.0.w1.bias": 171.8292028267183, + "blocks.0.w2.weight": 323.69930361812675, + "blocks.1.ln.weight": 6.518412820464622, + "blocks.1.w1.weight": 229.69399525127193, + "blocks.1.w1.bias": 219.11418017905737, + "blocks.1.w2.weight": 202.46482559231447, + "out_ln.weight": 0.4266951230953523, + "out_head.weight": 6.867727427000321, + "out_head.bias": 3.719577118675905 + } + }, + "credit_bridge": { + "log": { + "train_loss": [ + 2.0466229999542236, + 2.0390515814208983, + 2.0387567960357664, + 2.035990318336487, + 2.0319369410705566, + 2.0270682551574706, + 2.0257353674316407, + 2.01995243888855, + 2.014456915893555, + 2.014794263648987, + 2.0092950971221923, + 2.0075508432769777, + 2.0089861209869384, + 2.009624204673767, + 2.0054416754150393, + 2.004343302459717, + 2.004835032348633, + 2.004392629699707, + 2.0029932656097413, + 2.002116594848633, + 2.0016485495758056, + 1.9995216164398193, + 1.9959283925628661, + 1.9984067166900634, + 1.9955141537475587, + 1.9943039474105835, + 1.9905847966003418, + 1.9943868572235108, + 1.9933354425811767, + 1.9950886684799194, + 1.9958439822769165, + 1.9941565301132202, + 1.992669308166504, + 1.9904355131530762, + 1.9945945461273193, + 1.9895463079452516, + 1.9893892168807983, + 1.9876361457061769, + 1.9826778482055665, + 1.988116795387268, + 1.9903033963012695, + 1.9945041528320313, + 2.003135726776123, + 2.0090655849838255, + 2.003611517944336, + 1.993031780052185, + 1.9829731557846069, + 1.9795342245483398, + 1.9742745175933838, + 1.9710580517196656, + 1.9662886399841308, + 1.9650878366470337, + 1.9633581380462646, + 1.9602308530426025, + 1.9592647447967528, + 1.9569754809570312, + 1.9564002374267577, + 1.9596573688125611, + 1.9562416759490966, + 1.9577629917907715, + 1.9564975534057618, + 1.9539075463485718, + 1.9531170993804932, + 1.9562737216949464, + 1.9533874069213868, + 1.9565455058288574, + 1.9550479524993896, + 1.9579495740509034, + 1.9561651504898072, + 1.9566914569091798, + 1.9560811403656007, + 1.9559252410888672, + 1.954912179260254, + 1.9543887220001221, + 1.9562247213363648, + 1.9564899509429932, + 1.9564726518249511, + 1.9530779626464845, + 1.954867910079956, + 1.953048494796753, + 1.9528486930084228, + 1.9534204055404663, + 1.9533309042358398, + 1.952870100631714, + 1.9535446406555175, + 1.9520780282592773, + 1.9539363479614258, + 1.9533953338241576, + 1.9520681017303467, + 1.9507475804901122, + 1.9521085049819946, + 1.9542150582122804, + 1.9536484621429444, + 1.9531653139877319, + 1.9510113228988648, + 1.9543765464782714, + 1.9518434842681884, + 1.9509774511337281, + 1.9520590380859375, + 1.9515080471801758 + ], + "train_acc": [ + 0.25224, + 0.25856, + 0.25666, + 0.2584, + 0.26252, + 0.26562, + 0.26418, + 0.2664, + 0.26886, + 0.27, + 0.27072, + 0.27296, + 0.26938, + 0.27142, + 0.27324, + 0.27214, + 0.27118, + 0.27374, + 0.274, + 0.27416, + 0.27408, + 0.27516, + 0.27406, + 0.2768, + 0.27462, + 0.27758, + 0.2832, + 0.27908, + 0.27992, + 0.27762, + 0.27726, + 0.27918, + 0.2775, + 0.2774, + 0.2756, + 0.27898, + 0.27944, + 0.27822, + 0.282, + 0.27834, + 0.28466, + 0.28584, + 0.28626, + 0.28466, + 0.2816, + 0.28212, + 0.28232, + 0.28346, + 0.28134, + 0.28332, + 0.2858, + 0.28386, + 0.28558, + 0.28612, + 0.28698, + 0.28932, + 0.2885, + 0.28432, + 0.2864, + 0.28792, + 0.28548, + 0.28804, + 0.28786, + 0.28798, + 0.2883, + 0.28572, + 0.2894, + 0.28642, + 0.28876, + 0.28766, + 0.28924, + 0.28718, + 0.28808, + 0.28528, + 0.2878, + 0.28852, + 0.2875, + 0.28816, + 0.28838, + 0.28702, + 0.29014, + 0.29108, + 0.28906, + 0.29216, + 0.29062, + 0.28836, + 0.29078, + 0.28994, + 0.28852, + 0.29152, + 0.29096, + 0.29036, + 0.29132, + 0.29078, + 0.29352, + 0.28894, + 0.29248, + 0.2919, + 0.2922, + 0.28814 + ], + "test_acc": [ + 0.2783, + 0.2698, + 0.2993, + 0.282, + 0.3008, + 0.2636, + 0.302, + 0.2948, + 0.2815, + 0.2878, + 0.2823, + 0.3067, + 0.3037, + 0.2988, + 0.2874, + 0.2945, + 0.2916, + 0.2958, + 0.3024, + 0.2933, + 0.2879, + 0.3095, + 0.3108, + 0.3052, + 0.3078, + 0.304, + 0.299, + 0.2887, + 0.2945, + 0.2859, + 0.2947, + 0.3076, + 0.287, + 0.3072, + 0.3058, + 0.3008, + 0.2991, + 0.3036, + 0.3086, + 0.2928, + 0.3031, + 0.3113, + 0.307, + 0.3049, + 0.3033, + 0.3003, + 0.3063, + 0.307, + 0.2942, + 0.312, + 0.3043, + 0.3027, + 0.3029, + 0.3049, + 0.297, + 0.3117, + 0.2993, + 0.3076, + 0.2972, + 0.3089, + 0.3051, + 0.31, + 0.3078, + 0.2928, + 0.3153, + 0.3035, + 0.3138, + 0.3058, + 0.312, + 0.3027, + 0.3098, + 0.3066, + 0.3106, + 0.3118, + 0.3022, + 0.3123, + 0.3051, + 0.3089, + 0.3057, + 0.3095, + 0.3091, + 0.3075, + 0.3085, + 0.3078, + 0.3102, + 0.3103, + 0.3092, + 0.3038, + 0.3111, + 0.31, + 0.3089, + 0.3099, + 0.3095, + 0.3099, + 0.3096, + 0.3092, + 0.3102, + 0.3106, + 0.3106, + 0.3105 + ], + "value_loss": [ + 0.5981157046318054, + 0.23145676259040832, + 0.1629143718481064, + 0.13112473169445993, + 0.1034177236700058, + 0.0916099097251892, + 0.09448877233743667, + 0.07975816262960433, + 0.07028926397323608, + 0.06960906606912613, + 0.0663377025961876, + 0.06277827244520187, + 0.06609309570908546, + 0.05464596090316772, + 0.05347000701904297, + 0.055175553441047666, + 0.05404586620092392, + 0.050111453244686126, + 0.04170590556740761, + 0.042121976542472836, + 0.04078092483282089, + 0.04033791535496712, + 0.041389339334964755, + 0.04141822567462921, + 0.039917332406044004, + 0.036470176842212675, + 0.03570402141869068, + 0.0356453166270256, + 0.0348940605866909, + 0.03294261076688766, + 0.030363075378537177, + 0.031874004955291746, + 0.028748285691142083, + 0.02522519760966301, + 0.025916270488500596, + 0.02379499412894249, + 0.0268276374822855, + 0.021691660759449005, + 0.019947531086206435, + 0.021523804327845574, + 0.02691959607720375, + 0.02570752409338951, + 0.02876643155694008, + 0.0335148137229681, + 0.031904469298124315, + 0.02778498649060726, + 0.02370047950387001, + 0.02384015331029892, + 0.022202999491095544, + 0.02014792114138603, + 0.01951033084630966, + 0.0173978901296854, + 0.015886960296034813, + 0.014518849447965622, + 0.016183216240406036, + 0.016505281180143355, + 0.014820445327758789, + 0.014580657460987568, + 0.012194996480941772, + 0.014215488128364087, + 0.013283061341047286, + 0.011145545057058334, + 0.011805196754932403, + 0.009617018352746963, + 0.01116202095746994, + 0.009037217251360417, + 0.0090115091329813, + 0.011461790952682495, + 0.008820395555198193, + 0.007558346377015114, + 0.00825535637140274, + 0.007499891985356807, + 0.006166567915007472, + 0.006294274886846542, + 0.006129814375638962, + 0.005847277462333441, + 0.004945684289932251, + 0.004713377166241407, + 0.0033198376420140265, + 0.0037709764985740185, + 0.002341279806494713, + 0.002568720788434148, + 0.002159126052595675, + 0.001921226680725813, + 0.0018878655811026693, + 0.0016211944927275181, + 0.0013243718447163702, + 0.001169309064410627, + 0.0010681760500371455, + 0.0008630752410739661, + 0.0008580225262790919, + 0.0008040335815213621, + 0.0007892505507543683, + 0.0006356554404087364, + 0.0006604202647879719, + 0.0006046383780613541, + 0.0006052350796759129, + 0.0005226440225169063, + 0.0005789817444421351, + 0.0006147388235665858 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.33032962679862976, + 0.01970662735402584 + ], + "perturbation_rho": [ + 0.06130741536617279, + 0.0 + ], + "nudging": { + "0.001": [ + -5.778856575489044e-07, + 0.0 + ], + "0.003": [ + -1.7262063920497894e-06, + 0.0 + ], + "0.01": [ + -5.792360752820969e-06, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 175.24499314321244, + "embed.bias": 87.67212401922586, + "blocks.0.ln.weight": 8.05369238852348, + "blocks.0.w1.weight": 157.60363681616167, + "blocks.0.w1.bias": 81.08793311128552, + "blocks.0.w2.weight": 221.88820128646415, + "blocks.1.ln.weight": 5.44449828938583, + "blocks.1.w1.weight": 244.47570610251066, + "blocks.1.w1.bias": 227.1893178317289, + "blocks.1.w2.weight": 212.32680563546222, + "out_ln.weight": 0.33305352742575556, + "out_head.weight": 4.339797161618964, + "out_head.bias": 2.8094622055285003 + } + } +}
\ No newline at end of file diff --git a/results/cifar_depth_scan_s42/d512_L4_s42.json b/results/cifar_depth_scan_s42/d512_L4_s42.json new file mode 100644 index 0000000..3fb7a64 --- /dev/null +++ b/results/cifar_depth_scan_s42/d512_L4_s42.json @@ -0,0 +1,1205 @@ +{ + "bp": { + "log": { + "train_loss": [ + 1.9128315224838257, + 1.6880022193145752, + 1.6019164403533936, + 1.542231183242798, + 1.4950828089141845, + 1.4559073908996583, + 1.4284738327407838, + 1.397783087120056, + 1.3783822360992433, + 1.3537940772247314, + 1.3363390048980712, + 1.3175183240509034, + 1.301716223526001, + 1.286239836654663, + 1.2730019290542602, + 1.2588182968902588, + 1.2396523715209962, + 1.2286129103851318, + 1.216484442100525, + 1.2020161497688293, + 1.191397441482544, + 1.1803369410705566, + 1.1678005861663818, + 1.158150842514038, + 1.1437132249832154, + 1.1321278622055053, + 1.122210889930725, + 1.1123670346450805, + 1.1020959008026123, + 1.0898543254470825, + 1.0772244114494325, + 1.0695367335128785, + 1.0618911952209473, + 1.0524220000839233, + 1.0425953432464599, + 1.0304253259277343, + 1.023384765663147, + 1.0077225874519349, + 0.9960999109649659, + 0.9944095385360717, + 0.983660818862915, + 0.968802798652649, + 0.9584540496826172, + 0.9542556776809692, + 0.9423338845062256, + 0.9271652110481262, + 0.9240609010696411, + 0.9173964101791382, + 0.9032901007080079, + 0.8992292018318176, + 0.8873294083976746, + 0.8718722639083862, + 0.8695090994262695, + 0.858162225074768, + 0.8499397317695617, + 0.8388333418083191, + 0.8261937767410278, + 0.8185464579582215, + 0.810011429862976, + 0.7982917448425293, + 0.7877091710281372, + 0.7813588368415832, + 0.7747608847618103, + 0.7637265836715699, + 0.7545594720649719, + 0.7478000127983093, + 0.7360342462348938, + 0.7309902267074585, + 0.7200408243942261, + 0.7086022729682923, + 0.7048035074615479, + 0.6956654514884949, + 0.6874777590370178, + 0.682139354915619, + 0.6745446991729737, + 0.6719911893653869, + 0.6605463418197632, + 0.6587527468109131, + 0.653552459449768, + 0.6490319774246216, + 0.6399379921340942, + 0.6357991346931458, + 0.6328972053527832, + 0.6210838864898681, + 0.6178263852119446, + 0.615306571187973, + 0.6115417842102051, + 0.6158308602905274, + 0.6098391858100891, + 0.6050217900466919, + 0.6065702855396271, + 0.6006206908416748, + 0.5975759600830078, + 0.5960665979194641, + 0.5940444506263733, + 0.5969071801757813, + 0.5900075828170777, + 0.5945732257270813, + 0.5974702124786377, + 0.590236732711792 + ], + "train_acc": [ + 0.31564, + 0.39178, + 0.42, + 0.44194, + 0.45874, + 0.47242, + 0.48206, + 0.49302, + 0.50368, + 0.5108, + 0.51956, + 0.52752, + 0.531, + 0.5371, + 0.54046, + 0.54568, + 0.55426, + 0.55788, + 0.56382, + 0.56942, + 0.57036, + 0.5758, + 0.5802, + 0.58474, + 0.58888, + 0.5928, + 0.59766, + 0.6015, + 0.60714, + 0.6108, + 0.61308, + 0.61564, + 0.61944, + 0.6234, + 0.62626, + 0.62754, + 0.63226, + 0.63652, + 0.64376, + 0.6433, + 0.64368, + 0.65028, + 0.65508, + 0.65658, + 0.66194, + 0.6674, + 0.6677, + 0.67076, + 0.67436, + 0.67902, + 0.68228, + 0.6872, + 0.68646, + 0.6916, + 0.6952, + 0.69856, + 0.70518, + 0.70516, + 0.7078, + 0.71188, + 0.72018, + 0.71728, + 0.72058, + 0.72528, + 0.72892, + 0.73086, + 0.73516, + 0.73802, + 0.74118, + 0.74528, + 0.7469, + 0.7505, + 0.7517, + 0.75458, + 0.76024, + 0.76044, + 0.76314, + 0.76424, + 0.76568, + 0.76864, + 0.77176, + 0.7746, + 0.7747, + 0.77746, + 0.7786, + 0.77756, + 0.78208, + 0.78172, + 0.78232, + 0.78392, + 0.78542, + 0.78776, + 0.78702, + 0.78846, + 0.78602, + 0.78574, + 0.78886, + 0.78892, + 0.78904, + 0.79178 + ], + "test_acc": [ + 0.3932, + 0.4305, + 0.4599, + 0.4575, + 0.4965, + 0.5052, + 0.5072, + 0.513, + 0.5146, + 0.5331, + 0.5293, + 0.5323, + 0.5389, + 0.545, + 0.5545, + 0.5444, + 0.5494, + 0.5539, + 0.5652, + 0.5702, + 0.5638, + 0.5713, + 0.5729, + 0.5719, + 0.5722, + 0.5763, + 0.5807, + 0.5794, + 0.5855, + 0.5804, + 0.5844, + 0.59, + 0.5892, + 0.59, + 0.5955, + 0.5931, + 0.5877, + 0.5906, + 0.5944, + 0.6001, + 0.5976, + 0.5995, + 0.6036, + 0.6032, + 0.6069, + 0.6, + 0.6025, + 0.6029, + 0.5956, + 0.603, + 0.6055, + 0.6021, + 0.5997, + 0.6045, + 0.608, + 0.6077, + 0.6047, + 0.6033, + 0.6049, + 0.6026, + 0.6064, + 0.6089, + 0.6065, + 0.6025, + 0.6069, + 0.6068, + 0.6032, + 0.6036, + 0.6062, + 0.6023, + 0.6024, + 0.6011, + 0.603, + 0.6052, + 0.602, + 0.6042, + 0.6033, + 0.6026, + 0.6018, + 0.6026, + 0.6035, + 0.6033, + 0.6028, + 0.6044, + 0.6005, + 0.6037, + 0.6018, + 0.6031, + 0.6027, + 0.6043, + 0.604, + 0.6035, + 0.6024, + 0.604, + 0.6038, + 0.6031, + 0.6021, + 0.6027, + 0.603, + 0.6032 + ] + }, + "diagnostics": { + "bp_cosine": [ + 1.0, + 1.0, + 1.0, + 1.0 + ], + "perturbation_rho": [ + 0.98465895652771, + 0.9885985851287842, + 0.9880813360214233, + 0.9883240461349487 + ], + "nudging": { + "0.001": [ + -0.001627826364710927, + -0.0016166457207873464, + -0.0015995900612324476, + -0.0014260262250900269 + ], + "0.003": [ + -0.004879107233136892, + -0.004846072755753994, + -0.004795195069164038, + -0.004275224171578884 + ], + "0.01": [ + -0.016217362135648727, + -0.01610748842358589, + -0.015940139070153236, + -0.014218071475625038 + ] + } + }, + "drift": { + "embed.weight": 8.8243080894614, + "embed.bias": 16.94096306102109, + "blocks.0.ln.weight": 0.27074600681784294, + "blocks.0.w1.weight": 4.412644958801875, + "blocks.0.w1.bias": 7.117085807367152, + "blocks.0.w2.weight": 11.508053240186102, + "blocks.1.ln.weight": 0.16732973494773748, + "blocks.1.w1.weight": 4.742762067497133, + "blocks.1.w1.bias": 4.5510383530906955, + "blocks.1.w2.weight": 13.171214559405497, + "blocks.2.ln.weight": 0.17887226073929113, + "blocks.2.w1.weight": 4.753555652638221, + "blocks.2.w1.bias": 4.0288582989197055, + "blocks.2.w2.weight": 14.123960280583763, + "blocks.3.ln.weight": 0.19235362588922475, + "blocks.3.w1.weight": 4.831153218839085, + "blocks.3.w1.bias": 3.727898847051419, + "blocks.3.w2.weight": 15.974843151960325, + "out_ln.weight": 0.4103308080908126, + "out_head.weight": 2.6455778192608332, + "out_head.bias": 1.2917860402449406 + } + }, + "dfa": { + "log": { + "train_loss": [ + 2.0458639289093017, + 2.0289766119384764, + 2.020620018005371, + 2.012297201309204, + 2.0094007360076906, + 2.0115105359649657, + 2.0102763877105714, + 2.006047547874451, + 2.0053908419799806, + 2.006039619178772, + 2.0009452716064455, + 2.0004891522216797, + 1.997944189224243, + 1.9965615715026857, + 1.9999804154205323, + 1.9981115759277344, + 1.9969525494766236, + 1.9967278913879394, + 1.9977753674316405, + 1.9970432605361939, + 1.9937860690307616, + 1.9955713275909424, + 1.994003203048706, + 1.9966396224975587, + 1.9941929375457763, + 1.9933527754974365, + 1.9948605881500243, + 1.99625807762146, + 1.9941814679718017, + 1.9941422396850585, + 1.99151097530365, + 1.9934743473052978, + 1.9893354850769043, + 1.989959321861267, + 1.9919966284942627, + 1.9909703044128417, + 1.9899194409179688, + 1.9901701150512696, + 1.9903043267822265, + 1.9902096912384033, + 1.9900437306976317, + 1.9880607610321044, + 1.9873294051742554, + 1.9879234255981446, + 1.9870288726043701, + 1.9906337995910643, + 1.98670257522583, + 1.986429832382202, + 1.9861535691070558, + 1.9875364014053345, + 1.9880395639801025, + 1.9859618353652955, + 1.9838686182403564, + 1.9850774764251709, + 1.984304948348999, + 1.9849093035125733, + 1.9841952557373046, + 1.9846441001129151, + 1.984178702659607, + 1.9851755959320068, + 1.9839427744293212, + 1.9841270837020875, + 1.9833368181991577, + 1.9830518650817872, + 1.9827289093780518, + 1.9815367695617676, + 1.9821726945495606, + 1.9826502087020874, + 1.9826591117858887, + 1.9820565001678467, + 1.9842380196762084, + 1.9824368360137938, + 1.9822880676269532, + 1.9798370397949219, + 1.9824768118286134, + 1.982346687965393, + 1.9822141482543945, + 1.9804879917907714, + 1.9800791321182252, + 1.9813715744018554, + 1.9813639123535156, + 1.9790491080093384, + 1.980683646774292, + 1.9790133563232422, + 1.9805930490112305, + 1.980672128944397, + 1.981869688796997, + 1.9792622634124757, + 1.9801762536239624, + 1.9814309076690675, + 1.9808863278579711, + 1.9783899629211426, + 1.980429323272705, + 1.9808515368652344, + 1.9805352714538573, + 1.978631600112915, + 1.9817959378814698, + 1.9797530443954467, + 1.9781674542999268, + 1.9783192428970338 + ], + "train_acc": [ + 0.2491, + 0.25984, + 0.2652, + 0.26518, + 0.26756, + 0.26822, + 0.26646, + 0.26766, + 0.26952, + 0.26786, + 0.2692, + 0.27078, + 0.2744, + 0.27082, + 0.27058, + 0.26992, + 0.27464, + 0.2732, + 0.2712, + 0.27484, + 0.27366, + 0.27528, + 0.27524, + 0.27466, + 0.2739, + 0.27726, + 0.2754, + 0.27326, + 0.27062, + 0.2751, + 0.27784, + 0.27426, + 0.27818, + 0.28086, + 0.27724, + 0.27694, + 0.27696, + 0.27858, + 0.2784, + 0.27956, + 0.27774, + 0.28002, + 0.28038, + 0.2798, + 0.2806, + 0.27946, + 0.27928, + 0.28056, + 0.28084, + 0.28006, + 0.2811, + 0.28282, + 0.2807, + 0.28176, + 0.2819, + 0.28218, + 0.28188, + 0.28362, + 0.28256, + 0.28266, + 0.28368, + 0.28256, + 0.28366, + 0.28098, + 0.28328, + 0.28394, + 0.28346, + 0.28286, + 0.28388, + 0.28246, + 0.28286, + 0.28164, + 0.28372, + 0.28418, + 0.28458, + 0.28382, + 0.28406, + 0.2839, + 0.28352, + 0.28454, + 0.28204, + 0.28542, + 0.28458, + 0.2855, + 0.28488, + 0.28726, + 0.28564, + 0.2841, + 0.2856, + 0.28356, + 0.2863, + 0.28382, + 0.2859, + 0.28326, + 0.28584, + 0.28438, + 0.28494, + 0.2853, + 0.28502, + 0.28404 + ], + "test_acc": [ + 0.29, + 0.3005, + 0.2688, + 0.2879, + 0.2899, + 0.2866, + 0.2708, + 0.2917, + 0.2939, + 0.2888, + 0.3039, + 0.3014, + 0.2872, + 0.2932, + 0.3039, + 0.2983, + 0.3074, + 0.2815, + 0.3073, + 0.3121, + 0.3117, + 0.2972, + 0.2956, + 0.2898, + 0.3126, + 0.3014, + 0.3087, + 0.2844, + 0.2987, + 0.3162, + 0.2839, + 0.3007, + 0.299, + 0.2998, + 0.2977, + 0.3051, + 0.304, + 0.3093, + 0.3049, + 0.3037, + 0.3143, + 0.3147, + 0.2974, + 0.3146, + 0.3051, + 0.3068, + 0.3035, + 0.3095, + 0.3052, + 0.3107, + 0.3102, + 0.3022, + 0.3059, + 0.315, + 0.3074, + 0.3028, + 0.3144, + 0.3099, + 0.3026, + 0.3032, + 0.3081, + 0.3198, + 0.3133, + 0.3066, + 0.314, + 0.3126, + 0.3091, + 0.3119, + 0.3124, + 0.3184, + 0.3069, + 0.3125, + 0.3005, + 0.3117, + 0.3089, + 0.3089, + 0.3169, + 0.3106, + 0.3187, + 0.3165, + 0.3151, + 0.3145, + 0.3146, + 0.3104, + 0.3147, + 0.3125, + 0.3156, + 0.3169, + 0.3141, + 0.316, + 0.3152, + 0.3129, + 0.3126, + 0.3141, + 0.3143, + 0.3147, + 0.3146, + 0.3142, + 0.314, + 0.314 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.4003449082374573, + 0.001211420283652842, + -0.00036704502417705953, + -0.001560344360768795 + ], + "perturbation_rho": [ + 0.011487731710076332, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -4.852190613746643e-07, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -1.4952383935451508e-06, + -9.313225746154785e-10, + 0.0, + 0.0 + ], + "0.01": [ + -4.9872323870658875e-06, + -3.725290298461914e-09, + 2.3283064365386963e-09, + -1.862645149230957e-09 + ] + } + }, + "drift": { + "embed.weight": 315.1599803678731, + "embed.bias": 244.58720774886493, + "blocks.0.ln.weight": 8.197430677494586, + "blocks.0.w1.weight": 199.45530019449558, + "blocks.0.w1.bias": 169.48912351090533, + "blocks.0.w2.weight": 327.4002685876291, + "blocks.1.ln.weight": 6.599592372532966, + "blocks.1.w1.weight": 229.02521385968495, + "blocks.1.w1.bias": 216.20223057872374, + "blocks.1.w2.weight": 201.88759050630722, + "blocks.2.ln.weight": 7.947358922483768, + "blocks.2.w1.weight": 321.18706085148796, + "blocks.2.w1.bias": 277.92923879713146, + "blocks.2.w2.weight": 246.2703649903627, + "blocks.3.ln.weight": 5.776195341542472, + "blocks.3.w1.weight": 200.0071643423771, + "blocks.3.w1.bias": 167.9311499839471, + "blocks.3.w2.weight": 151.89636822620278, + "out_ln.weight": 0.5328500972285665, + "out_head.weight": 8.52020076404927, + "out_head.bias": 1.2327914968426916 + } + }, + "credit_bridge": { + "log": { + "train_loss": [ + 2.0418811988830567, + 2.0244707107162476, + 2.015521148300171, + 2.0061337772369385, + 2.0037599030303954, + 2.0030078707885743, + 1.999246537513733, + 1.9986579943847655, + 1.9972485411834717, + 1.9962204788970948, + 1.9977344912338257, + 1.9994980056381226, + 1.9950928838729858, + 1.9959543724060058, + 1.9950839403533935, + 1.995928331451416, + 1.9931118923950195, + 1.9951996542739867, + 1.9903301047134399, + 1.99565612575531, + 1.9905316178131103, + 1.9879318959045411, + 1.9921765214538574, + 1.988416219329834, + 1.988950412826538, + 1.9850122336578369, + 1.9853902856445313, + 1.985061616744995, + 1.9886472620391846, + 1.9883863719940185, + 1.9913309915161133, + 1.9895416070175171, + 1.987194644241333, + 1.9894078858947755, + 1.9878019535827636, + 1.9846340017700195, + 1.9846721477508544, + 1.9833970055389405, + 1.97837587184906, + 1.9834315983581543, + 1.976168726158142, + 1.976138740310669, + 1.9748867916488648, + 1.9714933393096923, + 1.9723309854888915, + 1.9734577295303344, + 1.9773753978729247, + 1.9739041761779785, + 1.9783673559570312, + 1.9767930798339843, + 1.9739136279678344, + 1.970040503616333, + 1.969602509536743, + 1.970613325881958, + 1.9665965628051758, + 1.9671646138763428, + 1.9675314767074585, + 1.9716551809692382, + 1.971007603111267, + 1.976275298461914, + 1.9748961191558838, + 1.9742881618881225, + 1.9736152155303954, + 1.9752269494628907, + 1.973985399093628, + 1.9731344959259034, + 1.9733161668395995, + 1.9738907193756103, + 1.9722116097640991, + 1.9699555205535888, + 1.972003019142151, + 1.9740287133407592, + 1.9716516065979004, + 1.969599935836792, + 1.9734370888900756, + 1.9703946390533447, + 1.9703987968444825, + 1.96968883518219, + 1.969948963279724, + 1.9703090355682373, + 1.969407484512329, + 1.9689243365097047, + 1.9691953955078125, + 1.9695471990966797, + 1.9693656219482423, + 1.9675928352355958, + 1.9688836163711547, + 1.9679821907043458, + 1.9700432720184327, + 1.9663442990493774, + 1.9663598616790772, + 1.9673082627868652, + 1.9678210963439942, + 1.96654816696167, + 1.9666101715087891, + 1.9660805171203612, + 1.966873331642151, + 1.966620960006714, + 1.9675770812606812, + 1.964846303024292 + ], + "train_acc": [ + 0.2533, + 0.2611, + 0.26526, + 0.26344, + 0.26812, + 0.2687, + 0.27104, + 0.27094, + 0.27094, + 0.27538, + 0.27276, + 0.27328, + 0.2756, + 0.27502, + 0.274, + 0.27276, + 0.2765, + 0.275, + 0.27584, + 0.27372, + 0.27766, + 0.27726, + 0.27552, + 0.2776, + 0.27428, + 0.2813, + 0.27736, + 0.2803, + 0.27882, + 0.27776, + 0.27758, + 0.2801, + 0.28084, + 0.27702, + 0.2819, + 0.27934, + 0.28016, + 0.2815, + 0.28464, + 0.2808, + 0.28334, + 0.28568, + 0.2852, + 0.2873, + 0.28756, + 0.287, + 0.29052, + 0.28836, + 0.28402, + 0.28472, + 0.28438, + 0.28548, + 0.28464, + 0.28252, + 0.2821, + 0.27936, + 0.28156, + 0.27808, + 0.27828, + 0.277, + 0.2753, + 0.27688, + 0.27532, + 0.27604, + 0.27414, + 0.27342, + 0.27416, + 0.27398, + 0.2779, + 0.2781, + 0.276, + 0.27548, + 0.2788, + 0.2794, + 0.27964, + 0.28044, + 0.28268, + 0.2817, + 0.28064, + 0.27928, + 0.28, + 0.28164, + 0.28246, + 0.28102, + 0.28358, + 0.28138, + 0.28232, + 0.28474, + 0.28286, + 0.28322, + 0.28316, + 0.28318, + 0.28198, + 0.28422, + 0.28366, + 0.28454, + 0.28538, + 0.2829, + 0.2831, + 0.28552 + ], + "test_acc": [ + 0.2888, + 0.2864, + 0.2858, + 0.2885, + 0.2871, + 0.3073, + 0.2926, + 0.3191, + 0.3113, + 0.3016, + 0.3124, + 0.2995, + 0.3046, + 0.2943, + 0.3111, + 0.2989, + 0.2944, + 0.2888, + 0.3006, + 0.2952, + 0.2985, + 0.3075, + 0.2969, + 0.3113, + 0.3152, + 0.3054, + 0.3137, + 0.3075, + 0.3116, + 0.3207, + 0.3132, + 0.3203, + 0.3138, + 0.2868, + 0.3006, + 0.3085, + 0.3173, + 0.3158, + 0.3112, + 0.3166, + 0.3082, + 0.2936, + 0.3047, + 0.3188, + 0.3158, + 0.3186, + 0.3109, + 0.2969, + 0.3073, + 0.3111, + 0.3051, + 0.2869, + 0.3112, + 0.2917, + 0.282, + 0.2801, + 0.3098, + 0.3097, + 0.2986, + 0.2964, + 0.2884, + 0.2738, + 0.2961, + 0.2958, + 0.2773, + 0.297, + 0.3029, + 0.2907, + 0.2935, + 0.2936, + 0.2939, + 0.2991, + 0.2958, + 0.2873, + 0.2939, + 0.2961, + 0.2942, + 0.2871, + 0.3067, + 0.293, + 0.3001, + 0.2932, + 0.2985, + 0.2896, + 0.3012, + 0.298, + 0.2904, + 0.2971, + 0.3012, + 0.291, + 0.2965, + 0.2961, + 0.2997, + 0.2997, + 0.3002, + 0.2997, + 0.2982, + 0.2984, + 0.2979, + 0.2983 + ], + "value_loss": [ + 0.5735433899068832, + 0.19343321340560912, + 0.1459548154258728, + 0.11554339327573776, + 0.10444803307533264, + 0.09662322332382202, + 0.07835375997543335, + 0.07846195635318756, + 0.07231992372989654, + 0.06300295606732369, + 0.0541220169878006, + 0.061311349160671234, + 0.054080903697013856, + 0.04913855415701866, + 0.053713992041945456, + 0.053991433643102646, + 0.042828602213859555, + 0.04867008077502251, + 0.036165564210414884, + 0.04136963748335838, + 0.03747292908787728, + 0.036461174008846284, + 0.03738185323476791, + 0.036185857799053193, + 0.037782702654600146, + 0.03236810895383358, + 0.028799428634643555, + 0.026735566897988318, + 0.028131072289943694, + 0.025900923478007317, + 0.02817378466963768, + 0.02747203625589609, + 0.022810295909643175, + 0.02715033429145813, + 0.023775706954598425, + 0.024240291431546213, + 0.02618250633716583, + 0.018726080359220503, + 0.019515147968530655, + 0.020435102397203444, + 0.02499964476108551, + 0.025505309803485872, + 0.029445164866447448, + 0.025735821738243105, + 0.02567795764923096, + 0.026399999004602433, + 0.02691687547147274, + 0.029062440598011018, + 0.030562642726898192, + 0.030370158289074897, + 0.030683111140727998, + 0.026187384767532348, + 0.024638052703738214, + 0.023440595749616624, + 0.020625758872032166, + 0.02062276856124401, + 0.01832583192050457, + 0.019966158434748648, + 0.018222353776693345, + 0.01633087551832199, + 0.01604582875967026, + 0.017060256469249725, + 0.016002053155303002, + 0.012738182610273362, + 0.012326533913612365, + 0.011105860123932361, + 0.009771750206947327, + 0.01024535400569439, + 0.009756301180124282, + 0.010442935926914216, + 0.008487312002182007, + 0.007283874700143933, + 0.007183250502049923, + 0.006603042964339256, + 0.006635444448143244, + 0.0059348720067739485, + 0.005370370441824198, + 0.004778812771886587, + 0.004565887214839458, + 0.0041534857338666915, + 0.003742397453635931, + 0.002904790796712041, + 0.0025213647907227276, + 0.002474344636797905, + 0.0021933346104249358, + 0.0016953631690889597, + 0.0015760234126448631, + 0.0016894616746902466, + 0.0012066031998768448, + 0.0011484015756100417, + 0.0009841965650953352, + 0.0010899798352271318, + 0.0011266277173534035, + 0.000903698658272624, + 0.0006944430249370635, + 0.000758418973647058, + 0.0006867282877117396, + 0.0007330659104511142, + 0.0006502793713379652, + 0.0009213672729767859 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.4021158814430237, + 0.01842920482158661, + 0.0360899493098259, + 0.03449413180351257 + ], + "perturbation_rho": [ + -0.009267053566873074, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -7.189810276031494e-07, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -2.16485932469368e-06, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -7.1660615503787994e-06, + 0.0, + 0.0, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 172.83875443763608, + "embed.bias": 95.3607064520029, + "blocks.0.ln.weight": 7.63160821502783, + "blocks.0.w1.weight": 152.1378494011768, + "blocks.0.w1.bias": 96.77665305560761, + "blocks.0.w2.weight": 233.87471998777892, + "blocks.1.ln.weight": 5.632208554748294, + "blocks.1.w1.weight": 278.87861207801535, + "blocks.1.w1.bias": 289.80305690116165, + "blocks.1.w2.weight": 229.62867282953147, + "blocks.2.ln.weight": 6.009254605154408, + "blocks.2.w1.weight": 273.52084419582525, + "blocks.2.w1.bias": 229.66597331157521, + "blocks.2.w2.weight": 178.21767131714722, + "blocks.3.ln.weight": 4.147781351924253, + "blocks.3.w1.weight": 131.08945490495597, + "blocks.3.w1.bias": 109.99062036071389, + "blocks.3.w2.weight": 129.72460456542194, + "out_ln.weight": 0.3547089789563311, + "out_head.weight": 5.58566439429333, + "out_head.bias": 2.120221083274003 + } + } +}
\ No newline at end of file diff --git a/results/cifar_depth_scan_s42/d512_L6_s42.json b/results/cifar_depth_scan_s42/d512_L6_s42.json new file mode 100644 index 0000000..307248f --- /dev/null +++ b/results/cifar_depth_scan_s42/d512_L6_s42.json @@ -0,0 +1,1259 @@ +{ + "bp": { + "log": { + "train_loss": [ + 1.9181581103515626, + 1.6830080932235718, + 1.5920702320480347, + 1.5328609270477296, + 1.4799005626678468, + 1.443107092781067, + 1.4138772471618652, + 1.3873612524032592, + 1.3651276648712158, + 1.3390819979095459, + 1.3217692392349243, + 1.305022211074829, + 1.2817235651779175, + 1.2682050260925293, + 1.2521741982269288, + 1.2378110882568358, + 1.215573044204712, + 1.2035384091949464, + 1.1932872578430176, + 1.1754147102737427, + 1.1641269732284545, + 1.1506621165275575, + 1.1339097010803223, + 1.1212864548492432, + 1.1109269207000732, + 1.0991801218032837, + 1.085493840484619, + 1.068724670715332, + 1.0622537975883484, + 1.0497802551078796, + 1.0335843839645387, + 1.0245532283782959, + 1.0139307324409486, + 1.003597820968628, + 0.9886502099990845, + 0.9805063750457763, + 0.9631068154907226, + 0.9566182117843628, + 0.9359624940299988, + 0.9330330731201172, + 0.9160753777313232, + 0.9056328008651734, + 0.8898537806510926, + 0.8787901188278199, + 0.8694718393707276, + 0.8506128107833862, + 0.8468854373931884, + 0.8328564296150207, + 0.8235082776260376, + 0.8091592510604858, + 0.7912276449775696, + 0.7819558120727539, + 0.7648613720703125, + 0.7557380415725709, + 0.7431310074996949, + 0.7325722367095947, + 0.7191381365394592, + 0.7073116994094849, + 0.6935642263793945, + 0.6829617013549805, + 0.6721186479949951, + 0.6607764552879334, + 0.6520131350326538, + 0.6347136948776245, + 0.6286079556655884, + 0.6146584116744995, + 0.6037020747566223, + 0.5897762714195252, + 0.5843867614936828, + 0.57075146068573, + 0.5668514904212951, + 0.5526476261138916, + 0.5426406643104553, + 0.5337510031700134, + 0.524534646692276, + 0.5152219123649597, + 0.508841796541214, + 0.5018864378547668, + 0.49528331829071043, + 0.4755337562465668, + 0.47835412044525144, + 0.47154994369506836, + 0.4639086106681824, + 0.4569279855918884, + 0.45118384454727173, + 0.4504211899662018, + 0.44511490434646606, + 0.44647758895874023, + 0.4362948834133148, + 0.43578986066818237, + 0.42930271067619324, + 0.4292936919593811, + 0.42567773310661317, + 0.4261182806301117, + 0.4239413748073578, + 0.42333943349838254, + 0.42293740759849546, + 0.4217962563323975, + 0.4224112439918518, + 0.41819300340652465 + ], + "train_acc": [ + 0.3148, + 0.3896, + 0.4219, + 0.4456, + 0.4649, + 0.47824, + 0.48992, + 0.49962, + 0.50886, + 0.51698, + 0.5202, + 0.52942, + 0.5382, + 0.5446, + 0.55176, + 0.5529, + 0.56062, + 0.56588, + 0.57028, + 0.577, + 0.58132, + 0.58658, + 0.59166, + 0.5987, + 0.59994, + 0.60574, + 0.6103, + 0.61766, + 0.61624, + 0.62304, + 0.62624, + 0.63198, + 0.63774, + 0.6413, + 0.64644, + 0.64766, + 0.65282, + 0.6553, + 0.66614, + 0.66296, + 0.67122, + 0.67416, + 0.68156, + 0.68518, + 0.6881, + 0.69356, + 0.69534, + 0.70084, + 0.70508, + 0.7084, + 0.71508, + 0.72034, + 0.72434, + 0.72938, + 0.73464, + 0.73592, + 0.7404, + 0.74552, + 0.75048, + 0.75408, + 0.75802, + 0.76272, + 0.76562, + 0.77362, + 0.77548, + 0.77904, + 0.78336, + 0.78934, + 0.79158, + 0.79514, + 0.79768, + 0.80414, + 0.80546, + 0.80986, + 0.81272, + 0.8163, + 0.81728, + 0.82112, + 0.82318, + 0.83366, + 0.83042, + 0.83306, + 0.836, + 0.83884, + 0.84078, + 0.83852, + 0.84302, + 0.8422, + 0.84388, + 0.84512, + 0.84892, + 0.85022, + 0.84986, + 0.85088, + 0.85284, + 0.8506, + 0.85018, + 0.85162, + 0.85258, + 0.85238 + ], + "test_acc": [ + 0.3995, + 0.428, + 0.4626, + 0.4828, + 0.4996, + 0.4961, + 0.5122, + 0.5214, + 0.5273, + 0.5318, + 0.5434, + 0.5431, + 0.5502, + 0.5495, + 0.5558, + 0.561, + 0.5593, + 0.5638, + 0.5727, + 0.5612, + 0.5661, + 0.5729, + 0.5769, + 0.5734, + 0.5798, + 0.5882, + 0.5883, + 0.5755, + 0.5828, + 0.5873, + 0.597, + 0.5833, + 0.5976, + 0.5915, + 0.5937, + 0.5918, + 0.5933, + 0.5934, + 0.5907, + 0.5972, + 0.5938, + 0.5974, + 0.5929, + 0.5938, + 0.6001, + 0.5933, + 0.6036, + 0.5987, + 0.5999, + 0.6037, + 0.594, + 0.5996, + 0.598, + 0.6017, + 0.597, + 0.6042, + 0.6017, + 0.5988, + 0.6004, + 0.6015, + 0.5984, + 0.6042, + 0.6041, + 0.6052, + 0.5989, + 0.605, + 0.6022, + 0.6055, + 0.6057, + 0.6064, + 0.6051, + 0.6055, + 0.6057, + 0.6056, + 0.6008, + 0.6045, + 0.6039, + 0.6033, + 0.6037, + 0.6022, + 0.6002, + 0.6018, + 0.5997, + 0.6029, + 0.6038, + 0.601, + 0.6044, + 0.6028, + 0.603, + 0.6031, + 0.6026, + 0.6029, + 0.6016, + 0.6033, + 0.6014, + 0.6027, + 0.6025, + 0.602, + 0.6014, + 0.6015 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.9925569295883179, + 0.9925950765609741, + 0.9925851225852966, + 0.9926368594169617, + 0.992660403251648, + 0.9924540519714355 + ], + "perturbation_rho": [ + 0.9874800443649292, + 0.9890985488891602, + 0.991584300994873, + 0.9924823641777039, + 0.9907411336898804, + 0.9890323877334595 + ], + "nudging": { + "0.001": [ + -0.0019930435810238123, + -0.002028408693149686, + -0.002009383402764797, + -0.001996344421058893, + -0.0018836274975910783, + -0.001594579080119729 + ], + "0.003": [ + -0.00597399240359664, + -0.006079169921576977, + -0.006022875662893057, + -0.005983929615467787, + -0.005646158009767532, + -0.004780753515660763 + ], + "0.01": [ + -0.01984817534685135, + -0.020196668803691864, + -0.020013831555843353, + -0.019886385649442673, + -0.018769418820738792, + -0.015900740399956703 + ] + } + }, + "drift": { + "embed.weight": 8.830199479874222, + "embed.bias": 13.652383066527548, + "blocks.0.ln.weight": 0.3209315667509576, + "blocks.0.w1.weight": 4.304189279431552, + "blocks.0.w1.bias": 6.874941583705022, + "blocks.0.w2.weight": 11.040131085404404, + "blocks.1.ln.weight": 0.19033462322360514, + "blocks.1.w1.weight": 4.611319361803835, + "blocks.1.w1.bias": 4.589828072495789, + "blocks.1.w2.weight": 12.410841455300883, + "blocks.2.ln.weight": 0.17767497411972724, + "blocks.2.w1.weight": 4.66461449969762, + "blocks.2.w1.bias": 4.221528366446113, + "blocks.2.w2.weight": 13.02617784277652, + "blocks.3.ln.weight": 0.18648530933637847, + "blocks.3.w1.weight": 4.703468866414805, + "blocks.3.w1.bias": 3.845782047024037, + "blocks.3.w2.weight": 13.668238845356765, + "blocks.4.ln.weight": 0.20878906370184375, + "blocks.4.w1.weight": 4.6585385757744255, + "blocks.4.w1.bias": 3.740256675221265, + "blocks.4.w2.weight": 14.188102590431034, + "blocks.5.ln.weight": 0.21537598403409916, + "blocks.5.w1.weight": 4.700384882175978, + "blocks.5.w1.bias": 3.7260920466734655, + "blocks.5.w2.weight": 15.8432281824125, + "out_ln.weight": 0.3676580083270373, + "out_head.weight": 2.951732465896682, + "out_head.bias": 1.3797369883266803 + } + }, + "dfa": { + "log": { + "train_loss": [ + 2.0429981646728517, + 2.0311185948181154, + 2.025969026145935, + 2.0156248236846923, + 2.012451182785034, + 2.0101683282852174, + 2.0090535586547853, + 2.0049661489105226, + 2.0042781643676757, + 2.007534557952881, + 2.005720973892212, + 2.0033944567108155, + 2.0026611005401613, + 2.0042629623413086, + 2.0030657132339478, + 2.0019637285614014, + 2.0002238693237304, + 2.0020113787841796, + 2.002593652191162, + 1.999959507446289, + 1.99915257938385, + 1.996764479637146, + 1.99686444190979, + 1.9965196546936035, + 1.9979205041503907, + 1.9956562859344482, + 1.9973748725128173, + 1.9964571334075927, + 1.9960263302993775, + 1.9924271702575684, + 1.992831787261963, + 1.99499480758667, + 1.9891729098892212, + 1.9939897597503662, + 1.9930099172210693, + 1.993308459815979, + 1.9917978497314452, + 1.9914746307754516, + 1.9900447463989257, + 1.992962121963501, + 1.9900524688720702, + 1.9900013018798828, + 1.990018187599182, + 1.9914000466537476, + 1.9891149829864503, + 1.9894212562561036, + 1.99083076171875, + 1.9897665852355957, + 1.98873883934021, + 1.990058041343689, + 1.987586608581543, + 1.987379070777893, + 1.9863710124206544, + 1.988468872642517, + 1.9863781386947632, + 1.9846030602264404, + 1.986063505783081, + 1.9875353760147094, + 1.9865796419525146, + 1.9875003507995606, + 1.9842604379272462, + 1.9855676065444947, + 1.9839949224472047, + 1.985175651473999, + 1.9853779614639282, + 1.9842977197265625, + 1.9852548834228516, + 1.9850790433120729, + 1.983184485321045, + 1.9850950769805908, + 1.9860374071502687, + 1.985449473876953, + 1.986313660736084, + 1.9834488523864746, + 1.9854728340911865, + 1.9834792975616455, + 1.9839858917999267, + 1.983295297241211, + 1.9832000505065919, + 1.9842015452575683, + 1.9817747521209716, + 1.9825036289215088, + 1.983466981277466, + 1.9826897898864746, + 1.9832522836303712, + 1.9837559247589112, + 1.9846134539031983, + 1.980646923828125, + 1.9809106316375733, + 1.9838191221618653, + 1.981862756576538, + 1.9802230101013183, + 1.9814140343475342, + 1.9822363115692139, + 1.982738264541626, + 1.9809983926773072, + 1.9829246353912353, + 1.981134694480896, + 1.9824947088241578, + 1.9818510884094238 + ], + "train_acc": [ + 0.25012, + 0.2569, + 0.25934, + 0.26228, + 0.26296, + 0.26604, + 0.2663, + 0.26786, + 0.26776, + 0.26536, + 0.26794, + 0.2679, + 0.27008, + 0.26912, + 0.27114, + 0.27122, + 0.27314, + 0.27074, + 0.26774, + 0.26882, + 0.2714, + 0.27276, + 0.27218, + 0.2744, + 0.27288, + 0.27368, + 0.27384, + 0.27596, + 0.27198, + 0.27772, + 0.27426, + 0.27528, + 0.27714, + 0.2757, + 0.27328, + 0.27562, + 0.27496, + 0.27752, + 0.27654, + 0.27576, + 0.27678, + 0.2766, + 0.27842, + 0.2765, + 0.28062, + 0.2769, + 0.27744, + 0.27504, + 0.27708, + 0.2794, + 0.27856, + 0.27826, + 0.28152, + 0.2805, + 0.27904, + 0.28034, + 0.28012, + 0.27994, + 0.27844, + 0.27932, + 0.28254, + 0.28096, + 0.28152, + 0.2795, + 0.28278, + 0.28194, + 0.28228, + 0.28134, + 0.28312, + 0.28172, + 0.2811, + 0.28118, + 0.2794, + 0.28338, + 0.2802, + 0.2814, + 0.28282, + 0.28256, + 0.2834, + 0.28248, + 0.28152, + 0.28336, + 0.28338, + 0.28288, + 0.28162, + 0.28292, + 0.28282, + 0.2848, + 0.28472, + 0.28266, + 0.28232, + 0.2875, + 0.28232, + 0.28272, + 0.28286, + 0.28446, + 0.28254, + 0.28228, + 0.28414, + 0.28394 + ], + "test_acc": [ + 0.2768, + 0.2776, + 0.2962, + 0.2815, + 0.2956, + 0.2956, + 0.2961, + 0.2963, + 0.3082, + 0.2895, + 0.2713, + 0.2959, + 0.299, + 0.2984, + 0.2889, + 0.3083, + 0.2967, + 0.2921, + 0.2979, + 0.3049, + 0.2897, + 0.2956, + 0.308, + 0.2866, + 0.2924, + 0.2989, + 0.2996, + 0.3051, + 0.2983, + 0.3066, + 0.2983, + 0.3082, + 0.3046, + 0.2948, + 0.3114, + 0.2894, + 0.3002, + 0.298, + 0.3016, + 0.2936, + 0.3076, + 0.3011, + 0.2884, + 0.2918, + 0.3046, + 0.2889, + 0.3154, + 0.3042, + 0.2921, + 0.3109, + 0.283, + 0.3003, + 0.2988, + 0.3024, + 0.2983, + 0.295, + 0.304, + 0.3112, + 0.3015, + 0.3044, + 0.314, + 0.3058, + 0.3094, + 0.3042, + 0.3112, + 0.3133, + 0.3064, + 0.2975, + 0.2975, + 0.3004, + 0.3054, + 0.3134, + 0.3093, + 0.3082, + 0.3084, + 0.3096, + 0.3041, + 0.3089, + 0.3033, + 0.3073, + 0.311, + 0.3131, + 0.3036, + 0.3127, + 0.3087, + 0.3117, + 0.3102, + 0.3056, + 0.3102, + 0.3087, + 0.3107, + 0.3099, + 0.3092, + 0.3094, + 0.3093, + 0.3102, + 0.3092, + 0.31, + 0.3098, + 0.3098 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.38683176040649414, + 0.0029573289211839437, + -0.0009689683793112636, + -0.0011474918574094772, + 0.0005293102003633976, + -0.0015941932797431946 + ], + "perturbation_rho": [ + -0.0035741720348596573, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -3.972090780735016e-07, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -1.2861564755439758e-06, + -1.862645149230957e-09, + 3.725290298461914e-09, + 0.0, + 0.0, + -1.862645149230957e-09 + ], + "0.01": [ + -4.382338374853134e-06, + -4.6566128730773926e-09, + 3.725290298461914e-09, + 0.0, + 9.313225746154785e-10, + -1.862645149230957e-09 + ] + } + }, + "drift": { + "embed.weight": 317.7495659655673, + "embed.bias": 246.61272235367505, + "blocks.0.ln.weight": 8.188531285985507, + "blocks.0.w1.weight": 202.7185165753587, + "blocks.0.w1.bias": 170.46943772700087, + "blocks.0.w2.weight": 332.85691144395827, + "blocks.1.ln.weight": 6.620851917635279, + "blocks.1.w1.weight": 230.61620075220995, + "blocks.1.w1.bias": 215.19134278855063, + "blocks.1.w2.weight": 203.35112374907072, + "blocks.2.ln.weight": 7.98453379220575, + "blocks.2.w1.weight": 315.5497282597786, + "blocks.2.w1.bias": 271.50424171303735, + "blocks.2.w2.weight": 248.3335414952481, + "blocks.3.ln.weight": 5.77177430564541, + "blocks.3.w1.weight": 199.54069059972556, + "blocks.3.w1.bias": 166.19848046359377, + "blocks.3.w2.weight": 156.3431036649652, + "blocks.4.ln.weight": 6.358256142081412, + "blocks.4.w1.weight": 230.11944086882187, + "blocks.4.w1.bias": 194.42511513927937, + "blocks.4.w2.weight": 171.6648008658166, + "blocks.5.ln.weight": 5.859230703676208, + "blocks.5.w1.weight": 209.3533985279986, + "blocks.5.w1.bias": 181.2514045255841, + "blocks.5.w2.weight": 161.81010288736894, + "out_ln.weight": 0.520622874797811, + "out_head.weight": 8.076281865086848, + "out_head.bias": 1.2313841339767553 + } + }, + "credit_bridge": { + "log": { + "train_loss": [ + 2.0568496820831297, + 2.0357495067596436, + 2.0265325052642824, + 2.0255675648498537, + 2.0211820729827883, + 2.0130653564453125, + 2.0164043712615967, + 2.011985154876709, + 2.0152608364868163, + 2.0114740381622314, + 2.0100359706878663, + 2.0067753753662108, + 2.0120022153091432, + 2.009950690765381, + 2.0081368783187865, + 2.012003229904175, + 2.0099324744415283, + 2.010634113998413, + 2.0069234494400026, + 2.003235507774353, + 2.0058242790985106, + 2.0072046648406983, + 2.00108302066803, + 2.0016218561553956, + 1.9978092765045166, + 1.995793134841919, + 1.994058724822998, + 1.9918723776245117, + 1.98977053981781, + 1.989516119003296, + 1.9898863918304444, + 1.9870066958236694, + 1.9896904999160767, + 1.9879004209136963, + 1.9867612840270996, + 1.9895719361877442, + 1.9847314866638184, + 1.9836825936508178, + 1.9799864123535156, + 1.978990509414673, + 1.97883055809021, + 1.9726349740219116, + 1.9712072018432618, + 1.978619043121338, + 1.9836511195373536, + 1.9836528234100341, + 1.9868938672256469, + 1.993108935470581, + 1.9985688564300537, + 2.0061389767074584, + 2.0023673818969727, + 1.9980067819595337, + 1.990028013458252, + 1.9925069525146484, + 1.9965549280548096, + 2.0004678047943116, + 2.0091910595321654, + 2.0153573443603516, + 2.0439902848815916, + 2.087756749229431, + 2.1218214894104004, + 2.133594408798218, + 2.1355165560913085, + 2.1193470074462892, + 2.070676408615112, + 2.0247885906219483, + 2.002980169754028, + 2.001423611793518, + 2.011682106933594, + 2.0117178466796877, + 2.0037788080596926, + 1.9931315280914306, + 1.986942767906189, + 1.9871479096984863, + 1.9841199736785888, + 1.9854342544555663, + 1.984901067276001, + 1.9828116521835326, + 1.987117846031189, + 1.980504002685547, + 1.9845846620178222, + 1.9854082400131225, + 1.9839192948150635, + 1.9855693399429322, + 1.9832484759521485, + 1.9859395289611816, + 1.9864675799942018, + 1.98523220413208, + 1.9856301559066774, + 1.9843646090698241, + 1.983153977355957, + 1.984056966934204, + 1.9850010103607179, + 1.986421587867737, + 1.9846718848037719, + 1.9840921591186524, + 1.9866302879333495, + 1.9842277576065064, + 1.9866961249160766, + 1.9863304554748535 + ], + "train_acc": [ + 0.24782, + 0.25582, + 0.25558, + 0.25876, + 0.2594, + 0.26644, + 0.26466, + 0.2638, + 0.26388, + 0.26448, + 0.26608, + 0.26734, + 0.26408, + 0.26614, + 0.26748, + 0.26298, + 0.2657, + 0.26878, + 0.26736, + 0.27, + 0.26918, + 0.2684, + 0.27234, + 0.27186, + 0.26998, + 0.27096, + 0.27406, + 0.27414, + 0.27628, + 0.27624, + 0.27302, + 0.27606, + 0.27518, + 0.27628, + 0.27844, + 0.27712, + 0.2805, + 0.27752, + 0.27616, + 0.27868, + 0.27654, + 0.28048, + 0.27978, + 0.2762, + 0.26926, + 0.27272, + 0.2737, + 0.26804, + 0.26878, + 0.26534, + 0.26856, + 0.27126, + 0.27694, + 0.28054, + 0.2827, + 0.2824, + 0.28504, + 0.28396, + 0.2755, + 0.26492, + 0.25088, + 0.24706, + 0.24506, + 0.25206, + 0.26578, + 0.27702, + 0.28328, + 0.27956, + 0.26926, + 0.26452, + 0.2656, + 0.27034, + 0.2789, + 0.2823, + 0.28274, + 0.28074, + 0.28288, + 0.28368, + 0.28134, + 0.28298, + 0.28284, + 0.28366, + 0.28172, + 0.28472, + 0.28516, + 0.2844, + 0.2836, + 0.28454, + 0.28676, + 0.28268, + 0.28488, + 0.28714, + 0.28404, + 0.286, + 0.28564, + 0.28672, + 0.28564, + 0.28652, + 0.2853, + 0.28532 + ], + "test_acc": [ + 0.2802, + 0.284, + 0.3008, + 0.2741, + 0.286, + 0.2865, + 0.2852, + 0.2925, + 0.286, + 0.3018, + 0.3012, + 0.2835, + 0.3037, + 0.2877, + 0.3014, + 0.3066, + 0.2908, + 0.2958, + 0.2946, + 0.2903, + 0.2807, + 0.2935, + 0.3134, + 0.3113, + 0.3034, + 0.2931, + 0.3001, + 0.311, + 0.3128, + 0.3116, + 0.3099, + 0.2976, + 0.2893, + 0.291, + 0.2988, + 0.2959, + 0.2992, + 0.3119, + 0.2906, + 0.2967, + 0.3031, + 0.3076, + 0.2874, + 0.282, + 0.289, + 0.2638, + 0.2745, + 0.2872, + 0.2764, + 0.2711, + 0.2732, + 0.2744, + 0.3014, + 0.302, + 0.2931, + 0.3077, + 0.2884, + 0.2883, + 0.2814, + 0.2816, + 0.2618, + 0.2679, + 0.2625, + 0.269, + 0.2844, + 0.2972, + 0.2945, + 0.2752, + 0.276, + 0.2833, + 0.2859, + 0.2809, + 0.3011, + 0.2909, + 0.3046, + 0.3047, + 0.2977, + 0.3012, + 0.3024, + 0.2816, + 0.3022, + 0.2915, + 0.2965, + 0.3004, + 0.3024, + 0.299, + 0.2977, + 0.3026, + 0.295, + 0.3036, + 0.2982, + 0.3023, + 0.2953, + 0.3008, + 0.3, + 0.3001, + 0.299, + 0.2995, + 0.2985, + 0.2989 + ], + "value_loss": [ + 0.5431940293550491, + 0.19173510383605957, + 0.12268259384393693, + 0.10493469659805298, + 0.10201048459529877, + 0.0708414131128788, + 0.06408844601988792, + 0.0720686790394783, + 0.06697397160291672, + 0.06196294789075851, + 0.05502780232191086, + 0.04616951223134994, + 0.05243575494527817, + 0.04471795217871666, + 0.044362507430315015, + 0.045647807506322864, + 0.04227230613827705, + 0.04649371607661247, + 0.037143924347162245, + 0.03761421796441078, + 0.0374193194770813, + 0.03865032393813133, + 0.03604385726451874, + 0.0348061216533184, + 0.034909065720438955, + 0.03548428529679775, + 0.03034081528186798, + 0.030923059933185577, + 0.03219221098482609, + 0.028869735319614412, + 0.02921472702085972, + 0.025344824229478837, + 0.03069723398923874, + 0.023911022167205812, + 0.02546191979944706, + 0.0234756334400177, + 0.021548947972655295, + 0.024292344799041746, + 0.021204249440431595, + 0.023217988926172255, + 0.032371759564876555, + 0.027472765368819235, + 0.03582359938144684, + 0.040054642882347107, + 0.0429941078042984, + 0.04646095623970032, + 0.05297355647087097, + 0.07127084519863129, + 0.0587913763999939, + 0.0855120901298523, + 0.5175958820724488, + 2.393956252441406, + 11.913925989227295, + 21.20466460494995, + 122.39695868041993, + 517.7628999365235, + 530.3028672167969, + 404.2268926171875, + 258.459100078125, + 283.38131254882813, + 102.50320305175781, + 27.287051020507814, + 12.905007503051758, + 4.956491999969482, + 1.5449322800445557, + 0.5012244256973266, + 0.26081433165550233, + 0.14692765646934508, + 0.10439578160524368, + 0.07197002289533615, + 0.05530310579776764, + 0.04514324131727219, + 0.031837492654323576, + 0.024001008454561233, + 0.019366809893846513, + 0.01406323720932007, + 0.010792444348335267, + 0.008763177410960197, + 0.007306897183656693, + 0.009082207316458226, + 0.005549766648113728, + 0.005990845524594187, + 0.004990600987970829, + 0.00429255537956953, + 0.005432345700562001, + 0.003525814834535122, + 0.004226339301913976, + 0.003729854447096586, + 0.004537549587935209, + 0.0023256922054290773, + 0.003219376254081726, + 0.0022955145540088416, + 0.00276749915830791, + 0.003175384805947542, + 0.0025999026219546795, + 0.0021026480976492168, + 0.0022611236164718864, + 0.0022266844733059405, + 0.002546760700196028, + 0.0024256033681333065 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.3041490912437439, + 0.03387444466352463, + 0.05120290070772171, + 0.0619620680809021, + 0.06251867115497589, + 0.06249140202999115 + ], + "perturbation_rho": [ + -0.005746336653828621, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -7.874332368373871e-07, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -2.527143806219101e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -8.463859558105469e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 150.13851572417417, + "embed.bias": 128.41508525356625, + "blocks.0.ln.weight": 5.836368229603896, + "blocks.0.w1.weight": 137.83692167652973, + "blocks.0.w1.bias": 120.59820508387907, + "blocks.0.w2.weight": 191.89941055268133, + "blocks.1.ln.weight": 5.7364977300128475, + "blocks.1.w1.weight": 223.49362295122356, + "blocks.1.w1.bias": 194.26884532235184, + "blocks.1.w2.weight": 197.19643675985097, + "blocks.2.ln.weight": 5.407984956609397, + "blocks.2.w1.weight": 224.96419292004586, + "blocks.2.w1.bias": 183.18926393313106, + "blocks.2.w2.weight": 163.77095460658097, + "blocks.3.ln.weight": 3.788219163278833, + "blocks.3.w1.weight": 178.78795349787814, + "blocks.3.w1.bias": 158.67792032092242, + "blocks.3.w2.weight": 127.01894259639586, + "blocks.4.ln.weight": 3.65585411727934, + "blocks.4.w1.weight": 138.57885809433404, + "blocks.4.w1.bias": 99.2214242231741, + "blocks.4.w2.weight": 101.99300788405023, + "blocks.5.ln.weight": 3.409437674208733, + "blocks.5.w1.weight": 110.32921272933086, + "blocks.5.w1.bias": 71.51527802753176, + "blocks.5.w2.weight": 83.53550127994924, + "out_ln.weight": 0.3630675420960623, + "out_head.weight": 5.826911511271652, + "out_head.bias": 3.507322897493987 + } + } +}
\ No newline at end of file diff --git a/results/cifar_depth_scan_s42/d512_L8_s42.json b/results/cifar_depth_scan_s42/d512_L8_s42.json new file mode 100644 index 0000000..1b9e3ac --- /dev/null +++ b/results/cifar_depth_scan_s42/d512_L8_s42.json @@ -0,0 +1,1313 @@ +{ + "bp": { + "log": { + "train_loss": [ + 1.915883038673401, + 1.6826532165908814, + 1.5945291757965088, + 1.5274945846557617, + 1.4820034421539308, + 1.442767191696167, + 1.4153068727493285, + 1.3821979986190795, + 1.3577272303390502, + 1.3394956853485107, + 1.317076534767151, + 1.2955091687774658, + 1.2746722560501098, + 1.2573306383514404, + 1.2403777645874023, + 1.2230817644500733, + 1.2058218476486207, + 1.1939524814605713, + 1.1745232657241822, + 1.157467629776001, + 1.1439746014022827, + 1.1305697510528565, + 1.1148247271347045, + 1.1023933839797975, + 1.086283505592346, + 1.073430366706848, + 1.0622342100524902, + 1.0457359760665894, + 1.0301126390266417, + 1.0175300416183473, + 1.0012673997306825, + 0.9858699301910401, + 0.978185770187378, + 0.9613556467437744, + 0.945394755973816, + 0.9297811899185181, + 0.9155407496261597, + 0.8999193870353699, + 0.883978879852295, + 0.8744391927337647, + 0.8499023978042602, + 0.8385310165596008, + 0.8242085730552673, + 0.8106124278068543, + 0.7989885342788696, + 0.7789599366378784, + 0.7652538424110412, + 0.7498692364501953, + 0.7314479119682312, + 0.7191568143844604, + 0.7038094125175476, + 0.6889407563781739, + 0.6683795366287232, + 0.6543509451675416, + 0.6412768173599244, + 0.6269769005203247, + 0.6097362549209595, + 0.5970014540195465, + 0.5797126161766052, + 0.5652852652168274, + 0.5507066846847534, + 0.5374195718574524, + 0.5179209604072571, + 0.5085140320968627, + 0.49228226098060607, + 0.47772936537742616, + 0.46773862747192385, + 0.4499609982967377, + 0.437783351650238, + 0.4272708469581604, + 0.414694590549469, + 0.40083733788490294, + 0.3944969776344299, + 0.3800938123130798, + 0.3698486288642883, + 0.359144784078598, + 0.35212729773521423, + 0.3423941581821442, + 0.33239298017501834, + 0.32468383754730223, + 0.3140012940311432, + 0.30757560186386107, + 0.306006680727005, + 0.2992121668767929, + 0.2903935454463959, + 0.2856505290937424, + 0.28447353170394896, + 0.27721282821655274, + 0.27670302483081816, + 0.2717075553417206, + 0.26587641248703003, + 0.2625578575849533, + 0.25637605533599855, + 0.26298797123908996, + 0.25818991552352905, + 0.25911176864147184, + 0.2525204331064224, + 0.25532632726669313, + 0.2519308549976349, + 0.25220084325790404 + ], + "train_acc": [ + 0.31264, + 0.39142, + 0.4222, + 0.44868, + 0.46492, + 0.47794, + 0.48996, + 0.50224, + 0.51198, + 0.51518, + 0.52792, + 0.5311, + 0.53912, + 0.54762, + 0.55402, + 0.56126, + 0.56546, + 0.56916, + 0.58018, + 0.58254, + 0.59062, + 0.59414, + 0.59952, + 0.60316, + 0.60978, + 0.6147, + 0.61912, + 0.62562, + 0.63032, + 0.63576, + 0.64308, + 0.64636, + 0.64882, + 0.65384, + 0.6618, + 0.66354, + 0.67342, + 0.67882, + 0.68372, + 0.68614, + 0.69758, + 0.69736, + 0.70374, + 0.70786, + 0.7118, + 0.7184, + 0.72552, + 0.72828, + 0.73742, + 0.74336, + 0.74852, + 0.75294, + 0.76092, + 0.76556, + 0.76962, + 0.77602, + 0.7833, + 0.78558, + 0.7929, + 0.79948, + 0.80264, + 0.80768, + 0.8158, + 0.81734, + 0.82548, + 0.8299, + 0.834, + 0.83984, + 0.8439, + 0.84742, + 0.85312, + 0.85774, + 0.85856, + 0.8659, + 0.8691, + 0.87372, + 0.87712, + 0.88168, + 0.88188, + 0.885, + 0.88966, + 0.89152, + 0.89464, + 0.89494, + 0.8985, + 0.90096, + 0.90124, + 0.9037, + 0.90518, + 0.90646, + 0.9091, + 0.90814, + 0.91108, + 0.91036, + 0.91186, + 0.91076, + 0.91388, + 0.91184, + 0.91348, + 0.91236 + ], + "test_acc": [ + 0.3856, + 0.4346, + 0.4501, + 0.478, + 0.4875, + 0.5022, + 0.5082, + 0.5204, + 0.5286, + 0.526, + 0.5408, + 0.5448, + 0.5497, + 0.554, + 0.5586, + 0.5582, + 0.5704, + 0.5716, + 0.5706, + 0.5786, + 0.5784, + 0.5793, + 0.5749, + 0.5803, + 0.5861, + 0.5853, + 0.5914, + 0.5853, + 0.5857, + 0.5931, + 0.5927, + 0.5896, + 0.5884, + 0.5893, + 0.592, + 0.5967, + 0.596, + 0.5986, + 0.5954, + 0.5947, + 0.5926, + 0.5925, + 0.5975, + 0.5989, + 0.596, + 0.5905, + 0.5967, + 0.5981, + 0.6016, + 0.5987, + 0.5932, + 0.5941, + 0.6009, + 0.5969, + 0.6018, + 0.5988, + 0.5972, + 0.5994, + 0.5961, + 0.6002, + 0.5958, + 0.5927, + 0.5963, + 0.5966, + 0.5983, + 0.5938, + 0.5983, + 0.5933, + 0.5971, + 0.5994, + 0.5915, + 0.5951, + 0.5928, + 0.5914, + 0.595, + 0.5951, + 0.5918, + 0.5922, + 0.594, + 0.5895, + 0.5919, + 0.5959, + 0.5924, + 0.59, + 0.5911, + 0.5889, + 0.5895, + 0.5893, + 0.5895, + 0.5895, + 0.5888, + 0.5889, + 0.5883, + 0.5884, + 0.589, + 0.5886, + 0.5887, + 0.5891, + 0.5891, + 0.5893 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.9652402400970459, + 0.9650944471359253, + 0.9653107523918152, + 0.9652330875396729, + 0.9652847051620483, + 0.9657888412475586, + 0.9645384550094604, + 0.9626421928405762 + ], + "perturbation_rho": [ + 0.988792896270752, + 0.9915838837623596, + 0.9924975037574768, + 0.9927228689193726, + 0.9931389093399048, + 0.9920060038566589, + 0.9927037358283997, + 0.9916384816169739 + ], + "nudging": { + "0.001": [ + -0.002859140280634165, + -0.0028454181738197803, + -0.0028910627588629723, + -0.002915592398494482, + -0.002882363274693489, + -0.0027308787684887648, + -0.0025113134179264307, + -0.002093898830935359 + ], + "0.003": [ + -0.008569219149649143, + -0.008527351543307304, + -0.00866447202861309, + -0.008737519383430481, + -0.00863889791071415, + -0.00818516593426466, + -0.007527736481279135, + -0.00627755094319582 + ], + "0.01": [ + -0.028463756665587425, + -0.02832203544676304, + -0.028776202350854874, + -0.02901865355670452, + -0.028697077184915543, + -0.02719692885875702, + -0.025021735578775406, + -0.02087603136897087 + ] + } + }, + "drift": { + "embed.weight": 8.768116164246894, + "embed.bias": 14.842866852450323, + "blocks.0.ln.weight": 0.3243736999277519, + "blocks.0.w1.weight": 4.248609055700286, + "blocks.0.w1.bias": 6.812297519559175, + "blocks.0.w2.weight": 10.91013034026142, + "blocks.1.ln.weight": 0.2264375721290641, + "blocks.1.w1.weight": 4.530696239577078, + "blocks.1.w1.bias": 4.691704909019707, + "blocks.1.w2.weight": 12.033360841473895, + "blocks.2.ln.weight": 0.19514233538894757, + "blocks.2.w1.weight": 4.641536271142397, + "blocks.2.w1.bias": 3.9592492063653415, + "blocks.2.w2.weight": 12.825064196706213, + "blocks.3.ln.weight": 0.20608382603702532, + "blocks.3.w1.weight": 4.675452058260372, + "blocks.3.w1.bias": 3.8959342701768565, + "blocks.3.w2.weight": 13.295810386821595, + "blocks.4.ln.weight": 0.21010986166308263, + "blocks.4.w1.weight": 4.643995186005312, + "blocks.4.w1.bias": 3.768514281224811, + "blocks.4.w2.weight": 13.646370291700444, + "blocks.5.ln.weight": 0.23069735418096324, + "blocks.5.w1.weight": 4.626200002743113, + "blocks.5.w1.bias": 3.8236708931328205, + "blocks.5.w2.weight": 14.086269621793239, + "blocks.6.ln.weight": 0.2603177383528657, + "blocks.6.w1.weight": 4.602646477526439, + "blocks.6.w1.bias": 3.7825134675181187, + "blocks.6.w2.weight": 14.472250927890926, + "blocks.7.ln.weight": 0.261911500383157, + "blocks.7.w1.weight": 4.5487829119704335, + "blocks.7.w1.bias": 3.638463299582741, + "blocks.7.w2.weight": 15.316485617563446, + "out_ln.weight": 0.33173738669135683, + "out_head.weight": 3.2824882859100533, + "out_head.bias": 1.1419967578910684 + } + }, + "dfa": { + "log": { + "train_loss": [ + 2.0608128107070924, + 2.0412321961975097, + 2.0255805563354494, + 2.0184878905487063, + 2.01624364112854, + 2.012751336364746, + 2.013942890319824, + 2.013219351119995, + 2.011828511428833, + 2.0141313094329836, + 2.0115797592163087, + 2.0122245865631103, + 2.0109272706604004, + 2.0090200772476194, + 2.009412646636963, + 2.008880652618408, + 2.0089728942489624, + 2.0055529666900633, + 2.0086945219421386, + 2.0047824951171873, + 2.0031694742202757, + 2.005182927017212, + 2.0044863636779784, + 2.004345258331299, + 2.0049652186584472, + 2.000570617828369, + 2.004866870994568, + 2.0028822119522096, + 2.003401609535217, + 2.001731799697876, + 2.0005709008789063, + 1.9988768460464477, + 1.9997380918884278, + 2.002846463394165, + 1.9990734589385986, + 1.9994710061264038, + 1.9974897908782958, + 1.9986941064453125, + 1.997244268875122, + 1.9990774517822265, + 1.9992041358184816, + 1.998780822906494, + 1.9990862586975098, + 1.9967566333770752, + 1.9988449997329711, + 1.995755309448242, + 1.996486500854492, + 1.9974545507049561, + 1.9962934982681275, + 1.995275443725586, + 1.9972379402923583, + 1.9948255879974366, + 1.994635205116272, + 1.995422209739685, + 1.9959965203857422, + 1.9941707052993773, + 1.9957991414642333, + 1.995743550643921, + 1.9927717569732666, + 1.9928340990829467, + 1.9935791075134277, + 1.9936188889312745, + 1.9914473522949219, + 1.99154779296875, + 1.993619135131836, + 1.9917610000228883, + 1.9916434611511231, + 1.993833284072876, + 1.9928128030395509, + 1.991187938156128, + 1.9918081129455567, + 1.9942408345794678, + 1.9907841648101807, + 1.9906754878234862, + 1.9921169677352906, + 1.990103793334961, + 1.9908813669204712, + 1.9915815322113037, + 1.991861714782715, + 1.9920157837295531, + 1.9913273377227783, + 1.9915946127319335, + 1.9907670964050292, + 1.991461629562378, + 1.9904960403442382, + 1.9905955597686769, + 1.9918609113311767, + 1.991680837097168, + 1.9905535266494752, + 1.9902639037322998, + 1.991429512939453, + 1.9898409436798095, + 1.9906655899047851, + 1.9915429690170288, + 1.98768330619812, + 1.988915832824707, + 1.9896351364135743, + 1.990216370239258, + 1.9885352130889893, + 1.9870271291351318 + ], + "train_acc": [ + 0.243, + 0.25202, + 0.25564, + 0.26482, + 0.26238, + 0.26154, + 0.26364, + 0.26322, + 0.26436, + 0.26482, + 0.2633, + 0.26298, + 0.26476, + 0.2654, + 0.26356, + 0.26538, + 0.26644, + 0.26838, + 0.26656, + 0.26864, + 0.27314, + 0.27042, + 0.26898, + 0.26672, + 0.26908, + 0.2701, + 0.26998, + 0.27056, + 0.26842, + 0.27026, + 0.27222, + 0.27208, + 0.27172, + 0.27086, + 0.27158, + 0.2727, + 0.27244, + 0.27268, + 0.27344, + 0.27336, + 0.27228, + 0.27516, + 0.27298, + 0.2762, + 0.27606, + 0.27508, + 0.27436, + 0.27444, + 0.27638, + 0.27438, + 0.27472, + 0.27788, + 0.27694, + 0.27622, + 0.27612, + 0.2758, + 0.2748, + 0.2753, + 0.27684, + 0.27852, + 0.2789, + 0.27824, + 0.27708, + 0.27816, + 0.27774, + 0.2779, + 0.27866, + 0.27744, + 0.28004, + 0.2772, + 0.27834, + 0.27546, + 0.279, + 0.27854, + 0.27808, + 0.28114, + 0.27966, + 0.27976, + 0.27822, + 0.2767, + 0.28038, + 0.27794, + 0.2817, + 0.27958, + 0.27992, + 0.27988, + 0.28026, + 0.28032, + 0.28058, + 0.27946, + 0.2797, + 0.2796, + 0.28088, + 0.27574, + 0.28058, + 0.27854, + 0.27872, + 0.28038, + 0.28194, + 0.28066 + ], + "test_acc": [ + 0.2682, + 0.2742, + 0.2881, + 0.2861, + 0.2979, + 0.294, + 0.291, + 0.2966, + 0.2988, + 0.2988, + 0.2844, + 0.2888, + 0.2806, + 0.2961, + 0.2761, + 0.2908, + 0.2853, + 0.2763, + 0.2759, + 0.2962, + 0.3049, + 0.2905, + 0.2945, + 0.2803, + 0.2978, + 0.2973, + 0.3048, + 0.2627, + 0.2941, + 0.2926, + 0.3019, + 0.2993, + 0.2753, + 0.2948, + 0.3083, + 0.3031, + 0.2891, + 0.3018, + 0.3049, + 0.2925, + 0.3133, + 0.2912, + 0.3036, + 0.2953, + 0.3011, + 0.2927, + 0.2895, + 0.3024, + 0.3009, + 0.3024, + 0.3012, + 0.3091, + 0.2932, + 0.2974, + 0.2911, + 0.2949, + 0.2978, + 0.3017, + 0.3025, + 0.3032, + 0.3051, + 0.3055, + 0.3024, + 0.3093, + 0.2987, + 0.3004, + 0.2907, + 0.3094, + 0.296, + 0.3036, + 0.302, + 0.3117, + 0.3083, + 0.3012, + 0.3019, + 0.3025, + 0.2994, + 0.3071, + 0.3043, + 0.3032, + 0.3018, + 0.3078, + 0.3044, + 0.3039, + 0.3067, + 0.3035, + 0.3002, + 0.3073, + 0.3035, + 0.3045, + 0.3054, + 0.3075, + 0.3032, + 0.3055, + 0.305, + 0.3059, + 0.3053, + 0.3057, + 0.3058, + 0.3057 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.37686973810195923, + 0.0015315066557377577, + -0.0013667643070220947, + 7.698773697484285e-05, + 5.11965190526098e-05, + -0.0007562801474705338, + -0.00026300083845853806, + -4.065445682499558e-05 + ], + "perturbation_rho": [ + 0.015484973788261414, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -3.7671998143196106e-07, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -1.1455267667770386e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -3.912951797246933e-06, + -4.6566128730773926e-09, + 0.0, + 0.0, + 1.862645149230957e-09, + 0.0, + 0.0, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 322.7518389055776, + "embed.bias": 247.2993592394181, + "blocks.0.ln.weight": 7.905005047928133, + "blocks.0.w1.weight": 212.9905007399479, + "blocks.0.w1.bias": 182.14933416642216, + "blocks.0.w2.weight": 330.2619044995618, + "blocks.1.ln.weight": 6.755700255991533, + "blocks.1.w1.weight": 232.57652888590908, + "blocks.1.w1.bias": 216.016264679947, + "blocks.1.w2.weight": 201.84872258219994, + "blocks.2.ln.weight": 8.01709273899982, + "blocks.2.w1.weight": 320.8453952706143, + "blocks.2.w1.bias": 276.695608451841, + "blocks.2.w2.weight": 249.60378814337417, + "blocks.3.ln.weight": 5.7943043253709785, + "blocks.3.w1.weight": 199.17683771092914, + "blocks.3.w1.bias": 166.1495411478002, + "blocks.3.w2.weight": 154.7202840885994, + "blocks.4.ln.weight": 6.436414607996792, + "blocks.4.w1.weight": 234.06786452519447, + "blocks.4.w1.bias": 198.33331425257262, + "blocks.4.w2.weight": 172.58039651693562, + "blocks.5.ln.weight": 5.904929352186912, + "blocks.5.w1.weight": 213.62873358910434, + "blocks.5.w1.bias": 184.85933031281215, + "blocks.5.w2.weight": 162.5444110371982, + "blocks.6.ln.weight": 5.922534596873395, + "blocks.6.w1.weight": 212.35774255360528, + "blocks.6.w1.bias": 187.6428413347025, + "blocks.6.w2.weight": 168.7920563845852, + "blocks.7.ln.weight": 6.9752563434590575, + "blocks.7.w1.weight": 251.09533847711697, + "blocks.7.w1.bias": 212.42015960200564, + "blocks.7.w2.weight": 182.17183040015334, + "out_ln.weight": 0.5380156572173149, + "out_head.weight": 8.131814182552638, + "out_head.bias": 0.8699895784813219 + } + }, + "credit_bridge": { + "log": { + "train_loss": [ + 2.0511436333847044, + 2.0277864767456055, + 2.02313751663208, + 2.0225361193847657, + 2.0088740489196777, + 2.0053706857681273, + 2.0050094537353518, + 2.0043996282196046, + 1.9971280477905273, + 2.0012310653686525, + 1.9964948630142212, + 1.994698377227783, + 1.9948096728134155, + 1.9947182064437867, + 1.9971965953826905, + 1.9937935817718506, + 1.9970728733825684, + 1.9935891329956055, + 1.9934549239730834, + 1.9943741856384278, + 1.9916830521392823, + 1.987718446044922, + 1.9891010809326173, + 1.9910999655532837, + 1.9870087446594238, + 1.9892022339248656, + 1.9874243151855469, + 1.9917996590423583, + 1.9928570490264892, + 1.9904958669281005, + 1.9926761375427247, + 1.9937843286895751, + 1.9929048935317992, + 1.9935734990692138, + 1.9881847104263306, + 1.9887749886322021, + 1.9857899047851562, + 1.9843308640289308, + 1.9787338375091552, + 1.9824666664886474, + 1.978441342086792, + 1.9784807219696046, + 1.975589619178772, + 1.9833358307647706, + 1.9823905195617675, + 2.005378373565674, + 2.044235708847046, + 2.053414641113281, + 2.062485707550049, + 2.082823034057617, + 2.0909555229187013, + 2.080506026992798, + 2.068187072105408, + 2.0602312271881105, + 2.051138618621826, + 2.048516896209717, + 2.0438889092254637, + 2.0391429753875734, + 2.030038472366333, + 2.025220480957031, + 2.0230712211608886, + 2.0209190673828124, + 2.0193369206237795, + 2.016298021621704, + 2.0145631092071534, + 2.007977565994263, + 2.007634442100525, + 2.00878977973938, + 2.0094445184707643, + 2.01266563041687, + 2.017705497894287, + 2.022621645774841, + 2.032514253387451, + 2.0355931142425536, + 2.0368993505859376, + 2.0395946674346925, + 2.0404893283081056, + 2.0423217945861816, + 2.0412329326629637, + 2.037104062347412, + 2.0325731353759764, + 2.0280113526916503, + 2.0226307469177245, + 2.0187665083312987, + 2.0180757246398926, + 2.016418752593994, + 2.013713173828125, + 2.013864770889282, + 2.0139804064941407, + 2.012384439239502, + 2.0120543003082276, + 2.013385525588989, + 2.0126590882873536, + 2.0125275871276855, + 2.0113051068878174, + 2.012384732666016, + 2.0117678832626344, + 2.0116403619766237, + 2.0123612815093996, + 2.0132470719146727 + ], + "train_acc": [ + 0.24252, + 0.25588, + 0.261, + 0.25814, + 0.26604, + 0.26736, + 0.26858, + 0.26616, + 0.27332, + 0.2707, + 0.27226, + 0.27346, + 0.27344, + 0.271, + 0.2724, + 0.27562, + 0.2716, + 0.27578, + 0.27566, + 0.27588, + 0.2756, + 0.2763, + 0.27646, + 0.2747, + 0.2769, + 0.27584, + 0.27532, + 0.27358, + 0.27436, + 0.27466, + 0.2748, + 0.27282, + 0.27324, + 0.2744, + 0.27718, + 0.27714, + 0.27682, + 0.2795, + 0.28304, + 0.28044, + 0.28062, + 0.28496, + 0.2849, + 0.28414, + 0.28304, + 0.27624, + 0.2686, + 0.26552, + 0.26574, + 0.2669, + 0.2667, + 0.26856, + 0.26866, + 0.27098, + 0.2677, + 0.26784, + 0.26828, + 0.2664, + 0.27158, + 0.27212, + 0.26908, + 0.27222, + 0.2712, + 0.27086, + 0.2689, + 0.27072, + 0.26998, + 0.27202, + 0.26646, + 0.26438, + 0.26242, + 0.26064, + 0.25896, + 0.25936, + 0.25754, + 0.26352, + 0.26094, + 0.263, + 0.26368, + 0.26518, + 0.2691, + 0.2693, + 0.26912, + 0.27422, + 0.2738, + 0.27242, + 0.27296, + 0.27486, + 0.27162, + 0.27384, + 0.27526, + 0.27172, + 0.27436, + 0.27512, + 0.27536, + 0.2737, + 0.27316, + 0.27306, + 0.27564, + 0.27502 + ], + "test_acc": [ + 0.2611, + 0.2728, + 0.2907, + 0.2809, + 0.2926, + 0.308, + 0.3005, + 0.3036, + 0.3052, + 0.2847, + 0.2973, + 0.2946, + 0.3093, + 0.2907, + 0.2893, + 0.3111, + 0.3023, + 0.2891, + 0.2992, + 0.295, + 0.2972, + 0.2999, + 0.3136, + 0.305, + 0.2994, + 0.3022, + 0.308, + 0.2936, + 0.2988, + 0.3132, + 0.299, + 0.2989, + 0.2874, + 0.2961, + 0.3114, + 0.3062, + 0.2921, + 0.3129, + 0.2986, + 0.2975, + 0.3083, + 0.2973, + 0.2912, + 0.3022, + 0.294, + 0.2862, + 0.264, + 0.2576, + 0.2455, + 0.2604, + 0.2623, + 0.2694, + 0.2661, + 0.2696, + 0.2726, + 0.265, + 0.2603, + 0.2717, + 0.2617, + 0.2692, + 0.2806, + 0.2848, + 0.2879, + 0.271, + 0.2743, + 0.2673, + 0.279, + 0.2598, + 0.2612, + 0.2694, + 0.2531, + 0.2553, + 0.254, + 0.2426, + 0.2479, + 0.2594, + 0.2546, + 0.2605, + 0.2661, + 0.27, + 0.2717, + 0.2842, + 0.2891, + 0.2864, + 0.2804, + 0.2904, + 0.2905, + 0.2865, + 0.2837, + 0.2868, + 0.2888, + 0.2865, + 0.2922, + 0.2874, + 0.2855, + 0.2892, + 0.2892, + 0.2887, + 0.2881, + 0.2881 + ], + "value_loss": [ + 0.42954374360084535, + 0.15756836537361146, + 0.12134378553867341, + 0.11469457547187806, + 0.09599290944099426, + 0.0877463513469696, + 0.0803228804397583, + 0.0667632784330845, + 0.0672521750831604, + 0.0677217069363594, + 0.06062764830827713, + 0.05060176089763641, + 0.05358793372392654, + 0.0549625644493103, + 0.052779753321409224, + 0.046068096545934675, + 0.060794142736196516, + 0.045549960198402406, + 0.04104518217980862, + 0.046112847526073454, + 0.04252848902463913, + 0.03344858367443085, + 0.03625167763471603, + 0.04116482120513916, + 0.03612018641352654, + 0.03742522090911865, + 0.03000010380625725, + 0.03407002721071243, + 0.03310543386161327, + 0.02845939307689667, + 0.02669600613832474, + 0.029440363923311234, + 0.025257505105137825, + 0.02826717380940914, + 0.024276180317401887, + 0.021678789145350458, + 0.02275293359041214, + 0.024361893134117126, + 0.019806503029465676, + 0.028984218854904174, + 0.025155036890506743, + 0.028825275114774705, + 0.035153553171157835, + 0.04898144502878189, + 0.053181294503211976, + 0.25244187658309936, + 9.008439293670655, + 244.4939577636719, + 1492.590131953125, + 5003.36131859375, + 15302.2068553125, + 42179.784040625, + 40569.28429, + 55602.10701875, + 58168.276421875, + 47172.94552875, + 32400.684385, + 36457.8510815625, + 55178.61985875, + 33309.7381096875, + 11320.94740828125, + 20943.124659375, + 19742.98079, + 7133.2663896875, + 7511.98685734375, + 4198.7234078125, + 5113.87740875, + 3748.421473125, + 2814.479872578125, + 2090.86994609375, + 1224.1438057421874, + 568.31011984375, + 272.7561458203125, + 120.44688937988282, + 92.17080795898437, + 58.42213998718262, + 22.001411871643068, + 13.230551771850585, + 7.299799581985473, + 1.8038615107917786, + 0.6261436622428894, + 0.23487731760025024, + 0.09752666207790375, + 0.05221867031812668, + 0.03414971536874771, + 0.02229098296046257, + 0.014439315105676651, + 0.01196531643152237, + 0.009077809438109399, + 0.008066386908292771, + 0.006000318608433008, + 0.006079038715660572, + 0.0053833009558916095, + 0.004241169492304325, + 0.003964086800217629, + 0.0037191870298981665, + 0.004121905582398176, + 0.0030519997741281987, + 0.002969844557568431, + 0.00373484493046999 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.20516762137413025, + 0.010197397321462631, + 0.012989584356546402, + 0.01799590140581131, + 0.026415985077619553, + 0.028529509902000427, + 0.02881331369280815, + 0.029378943145275116 + ], + "perturbation_rho": [ + 0.03650364279747009, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -5.299225449562073e-07, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -1.5022233128547668e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -5.0407834351062775e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + } + }, + "drift": { + "embed.weight": 144.386565398537, + "embed.bias": 134.32612651237073, + "blocks.0.ln.weight": 6.748346473737498, + "blocks.0.w1.weight": 132.91428698293475, + "blocks.0.w1.bias": 119.60864597492228, + "blocks.0.w2.weight": 212.526020150235, + "blocks.1.ln.weight": 4.414642051445624, + "blocks.1.w1.weight": 225.12165033097793, + "blocks.1.w1.bias": 234.43494513588323, + "blocks.1.w2.weight": 178.536762788961, + "blocks.2.ln.weight": 5.970302149919209, + "blocks.2.w1.weight": 271.1166277423572, + "blocks.2.w1.bias": 227.3762025819469, + "blocks.2.w2.weight": 203.14523981621292, + "blocks.3.ln.weight": 5.796867528153885, + "blocks.3.w1.weight": 255.0825523445021, + "blocks.3.w1.bias": 231.33899213301072, + "blocks.3.w2.weight": 158.37683433796386, + "blocks.4.ln.weight": 4.529609889606443, + "blocks.4.w1.weight": 162.41528397286507, + "blocks.4.w1.bias": 142.635739824408, + "blocks.4.w2.weight": 120.99022470237554, + "blocks.5.ln.weight": 4.160579158399913, + "blocks.5.w1.weight": 154.8997138115475, + "blocks.5.w1.bias": 138.47252168248815, + "blocks.5.w2.weight": 128.9164645106335, + "blocks.6.ln.weight": 3.6901613153795623, + "blocks.6.w1.weight": 111.81360738270082, + "blocks.6.w1.bias": 84.19462418565291, + "blocks.6.w2.weight": 97.14752351362607, + "blocks.7.ln.weight": 4.210487097037135, + "blocks.7.w1.weight": 93.35554428980572, + "blocks.7.w1.bias": 43.21054718035769, + "blocks.7.w2.weight": 87.7426729890803, + "out_ln.weight": 0.3258527803812823, + "out_head.weight": 5.969889930683836, + "out_head.bias": 3.3113123218582157 + } + } +}
\ No newline at end of file diff --git a/results/cifar_depth_scan_s42/summary.json b/results/cifar_depth_scan_s42/summary.json new file mode 100644 index 0000000..b16273d --- /dev/null +++ b/results/cifar_depth_scan_s42/summary.json @@ -0,0 +1,186 @@ +{ + "d512_L8_s42": { + "bp": { + "test_acc": 0.5893, + "mean_bp_cosine": 0.9648915901780128, + "mean_rho": 0.9918855354189873, + "mean_nudge_01": -0.027046552626416087, + "bp_cosine_per_layer": [ + 0.9652402400970459, + 0.9650944471359253, + 0.9653107523918152, + 0.9652330875396729, + 0.9652847051620483, + 0.9657888412475586, + 0.9645384550094604, + 0.9626421928405762 + ], + "rho_per_layer": [ + 0.988792896270752, + 0.9915838837623596, + 0.9924975037574768, + 0.9927228689193726, + 0.9931389093399048, + 0.9920060038566589, + 0.9927037358283997, + 0.9916384816169739 + ] + }, + "dfa": { + "test_acc": 0.3057, + "mean_bp_cosine": 0.047012841157993535, + "mean_rho": 0.0019356217235326767, + "mean_nudge_01": -4.894682206213474e-07, + "bp_cosine_per_layer": [ + 0.37686973810195923, + 0.0015315066557377577, + -0.0013667643070220947, + 7.698773697484285e-05, + 5.11965190526098e-05, + -0.0007562801474705338, + -0.00026300083845853806, + -4.065445682499558e-05 + ], + "rho_per_layer": [ + 0.015484973788261414, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + }, + "credit_bridge": { + "test_acc": 0.2881, + "mean_bp_cosine": 0.04493603203445673, + "mean_rho": 0.004562955349683762, + "mean_nudge_01": -6.300979293882847e-07, + "bp_cosine_per_layer": [ + 0.20516762137413025, + 0.010197397321462631, + 0.012989584356546402, + 0.01799590140581131, + 0.026415985077619553, + 0.028529509902000427, + 0.02881331369280815, + 0.029378943145275116 + ], + "rho_per_layer": [ + 0.03650364279747009, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + } + }, + "d512_L12_s42": { + "bp": { + "test_acc": 0.594, + "mean_bp_cosine": 0.9404260218143463, + "mean_rho": 0.9899803350369135, + "mean_nudge_01": -0.026733027460674446, + "bp_cosine_per_layer": [ + 0.9423961043357849, + 0.9423063397407532, + 0.9428478479385376, + 0.9422410726547241, + 0.9424552917480469, + 0.9417660236358643, + 0.9408372640609741, + 0.9404677152633667, + 0.9403290748596191, + 0.9396632313728333, + 0.935971736907959, + 0.9338305592536926 + ], + "rho_per_layer": [ + 0.987454891204834, + 0.9897942543029785, + 0.990141749382019, + 0.9901803731918335, + 0.9920680522918701, + 0.9906818866729736, + 0.9903223514556885, + 0.9926531910896301, + 0.9914379119873047, + 0.9897843599319458, + 0.9882909059524536, + 0.9869540929794312 + ] + }, + "dfa": { + "test_acc": 0.309, + "mean_bp_cosine": 0.032277349038243607, + "mean_rho": -0.003907783577839534, + "mean_nudge_01": -3.9018535365660984e-07, + "bp_cosine_per_layer": [ + 0.38828110694885254, + 0.0037502094637602568, + -0.0013616610085591674, + -0.0004177851078566164, + 0.0008564171148464084, + 0.00033698498737066984, + 0.0006134084542281926, + -0.00040099132456816733, + -0.00013616093201562762, + -0.0008071845513768494, + -0.0019259440014138818, + -0.0014602115843445063 + ], + "rho_per_layer": [ + -0.0468934029340744, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + }, + "credit_bridge": { + "test_acc": 0.239, + "mean_bp_cosine": 0.032352436101064086, + "mean_rho": 0.0008616718308379253, + "mean_nudge_01": -3.4909074505170185e-07, + "bp_cosine_per_layer": [ + 0.20759254693984985, + 0.005885659717023373, + 0.014463091269135475, + 0.013757804408669472, + 0.01659775897860527, + 0.0177934467792511, + 0.018012138083577156, + 0.019089948385953903, + 0.019284311681985855, + 0.01838410645723343, + 0.01891487091779709, + 0.018453549593687057 + ], + "rho_per_layer": [ + 0.010340061970055103, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + } + } +}
\ No newline at end of file |
