diff options
Diffstat (limited to 'results/cifar_depth_scan_s42/d512_L6_s42.json')
| -rw-r--r-- | results/cifar_depth_scan_s42/d512_L6_s42.json | 1259 |
1 files changed, 1259 insertions, 0 deletions
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 |
