diff options
Diffstat (limited to 'results/cifar_depth_scan_s42/d512_L4_s42.json')
| -rw-r--r-- | results/cifar_depth_scan_s42/d512_L4_s42.json | 1205 |
1 files changed, 1205 insertions, 0 deletions
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 |
