diff --git "a/instruct/trainer_state.json" "b/instruct/trainer_state.json" new file mode 100644--- /dev/null +++ "b/instruct/trainer_state.json" @@ -0,0 +1,213816 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 2.999831616759745, + "eval_steps": 500, + "global_step": 35631, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 1.870907390084191e-08, + "loss": 2.1446, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 3.741814780168382e-08, + "loss": 2.2989, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 5.612722170252572e-08, + "loss": 2.1866, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 7.483629560336764e-08, + "loss": 2.271, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 9.354536950420954e-08, + "loss": 2.0438, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 1.1225444340505145e-07, + "loss": 2.037, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 1.3096351730589338e-07, + "loss": 2.069, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 1.4967259120673527e-07, + "loss": 2.1736, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 1.6838166510757722e-07, + "loss": 2.3463, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 1.8709073900841908e-07, + "loss": 2.2016, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 2.0579981290926103e-07, + "loss": 2.1483, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 2.245088868101029e-07, + "loss": 1.9061, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 2.432179607109448e-07, + "loss": 2.1625, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 2.6192703461178676e-07, + "loss": 2.1611, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 2.8063610851262865e-07, + "loss": 2.1402, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 2.9934518241347054e-07, + "loss": 2.2027, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 3.180542563143125e-07, + "loss": 2.0376, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 3.3676333021515443e-07, + "loss": 2.067, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 3.5547240411599627e-07, + "loss": 2.2303, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 3.7418147801683816e-07, + "loss": 2.0024, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 3.928905519176801e-07, + "loss": 2.0839, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 4.1159962581852205e-07, + "loss": 2.2116, + "step": 22 + }, + { + "epoch": 0.0, + "learning_rate": 4.3030869971936394e-07, + "loss": 2.0248, + "step": 23 + }, + { + "epoch": 0.0, + "learning_rate": 4.490177736202058e-07, + "loss": 2.053, + "step": 24 + }, + { + "epoch": 0.0, + "learning_rate": 4.6772684752104773e-07, + "loss": 2.1323, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 4.864359214218896e-07, + "loss": 2.0056, + "step": 26 + }, + { + "epoch": 0.0, + "learning_rate": 5.051449953227316e-07, + "loss": 2.1318, + "step": 27 + }, + { + "epoch": 0.0, + "learning_rate": 5.238540692235735e-07, + "loss": 2.1765, + "step": 28 + }, + { + "epoch": 0.0, + "learning_rate": 5.425631431244154e-07, + "loss": 2.232, + "step": 29 + }, + { + "epoch": 0.0, + "learning_rate": 5.612722170252573e-07, + "loss": 1.9804, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 5.799812909260992e-07, + "loss": 2.2644, + "step": 31 + }, + { + "epoch": 0.0, + "learning_rate": 5.986903648269411e-07, + "loss": 2.0979, + "step": 32 + }, + { + "epoch": 0.0, + "learning_rate": 6.173994387277831e-07, + "loss": 2.0698, + "step": 33 + }, + { + "epoch": 0.0, + "learning_rate": 6.36108512628625e-07, + "loss": 2.1111, + "step": 34 + }, + { + "epoch": 0.0, + "learning_rate": 6.548175865294669e-07, + "loss": 2.1527, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 6.735266604303089e-07, + "loss": 2.2111, + "step": 36 + }, + { + "epoch": 0.0, + "learning_rate": 6.922357343311506e-07, + "loss": 2.1742, + "step": 37 + }, + { + "epoch": 0.0, + "learning_rate": 7.109448082319925e-07, + "loss": 1.955, + "step": 38 + }, + { + "epoch": 0.0, + "learning_rate": 7.296538821328344e-07, + "loss": 2.1732, + "step": 39 + }, + { + "epoch": 0.0, + "learning_rate": 7.483629560336763e-07, + "loss": 2.0598, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 7.670720299345183e-07, + "loss": 2.1183, + "step": 41 + }, + { + "epoch": 0.0, + "learning_rate": 7.857811038353602e-07, + "loss": 2.0442, + "step": 42 + }, + { + "epoch": 0.0, + "learning_rate": 8.044901777362021e-07, + "loss": 1.9987, + "step": 43 + }, + { + "epoch": 0.0, + "learning_rate": 8.231992516370441e-07, + "loss": 1.8915, + "step": 44 + }, + { + "epoch": 0.0, + "learning_rate": 8.41908325537886e-07, + "loss": 2.1371, + "step": 45 + }, + { + "epoch": 0.0, + "learning_rate": 8.606173994387279e-07, + "loss": 2.064, + "step": 46 + }, + { + "epoch": 0.0, + "learning_rate": 8.793264733395698e-07, + "loss": 2.1435, + "step": 47 + }, + { + "epoch": 0.0, + "learning_rate": 8.980355472404116e-07, + "loss": 2.0009, + "step": 48 + }, + { + "epoch": 0.0, + "learning_rate": 9.167446211412536e-07, + "loss": 2.028, + "step": 49 + }, + { + "epoch": 0.0, + "learning_rate": 9.354536950420955e-07, + "loss": 2.1319, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 9.541627689429374e-07, + "loss": 2.0155, + "step": 51 + }, + { + "epoch": 0.0, + "learning_rate": 9.728718428437792e-07, + "loss": 1.8399, + "step": 52 + }, + { + "epoch": 0.0, + "learning_rate": 9.915809167446211e-07, + "loss": 1.896, + "step": 53 + }, + { + "epoch": 0.0, + "learning_rate": 1.0102899906454632e-06, + "loss": 2.0254, + "step": 54 + }, + { + "epoch": 0.0, + "learning_rate": 1.0289990645463051e-06, + "loss": 1.9622, + "step": 55 + }, + { + "epoch": 0.0, + "learning_rate": 1.047708138447147e-06, + "loss": 1.8026, + "step": 56 + }, + { + "epoch": 0.0, + "learning_rate": 1.066417212347989e-06, + "loss": 2.0504, + "step": 57 + }, + { + "epoch": 0.0, + "learning_rate": 1.0851262862488308e-06, + "loss": 1.8856, + "step": 58 + }, + { + "epoch": 0.0, + "learning_rate": 1.1038353601496727e-06, + "loss": 1.8816, + "step": 59 + }, + { + "epoch": 0.01, + "learning_rate": 1.1225444340505146e-06, + "loss": 1.9704, + "step": 60 + }, + { + "epoch": 0.01, + "learning_rate": 1.1412535079513565e-06, + "loss": 1.8911, + "step": 61 + }, + { + "epoch": 0.01, + "learning_rate": 1.1599625818521984e-06, + "loss": 1.9066, + "step": 62 + }, + { + "epoch": 0.01, + "learning_rate": 1.1786716557530403e-06, + "loss": 1.8924, + "step": 63 + }, + { + "epoch": 0.01, + "learning_rate": 1.1973807296538822e-06, + "loss": 1.903, + "step": 64 + }, + { + "epoch": 0.01, + "learning_rate": 1.2160898035547243e-06, + "loss": 1.8723, + "step": 65 + }, + { + "epoch": 0.01, + "learning_rate": 1.2347988774555662e-06, + "loss": 1.8541, + "step": 66 + }, + { + "epoch": 0.01, + "learning_rate": 1.253507951356408e-06, + "loss": 1.9066, + "step": 67 + }, + { + "epoch": 0.01, + "learning_rate": 1.27221702525725e-06, + "loss": 1.8368, + "step": 68 + }, + { + "epoch": 0.01, + "learning_rate": 1.2909260991580918e-06, + "loss": 1.9843, + "step": 69 + }, + { + "epoch": 0.01, + "learning_rate": 1.3096351730589337e-06, + "loss": 1.8762, + "step": 70 + }, + { + "epoch": 0.01, + "learning_rate": 1.3283442469597756e-06, + "loss": 1.821, + "step": 71 + }, + { + "epoch": 0.01, + "learning_rate": 1.3470533208606177e-06, + "loss": 1.901, + "step": 72 + }, + { + "epoch": 0.01, + "learning_rate": 1.3657623947614596e-06, + "loss": 1.8411, + "step": 73 + }, + { + "epoch": 0.01, + "learning_rate": 1.3844714686623013e-06, + "loss": 1.9041, + "step": 74 + }, + { + "epoch": 0.01, + "learning_rate": 1.4031805425631432e-06, + "loss": 1.7856, + "step": 75 + }, + { + "epoch": 0.01, + "learning_rate": 1.421889616463985e-06, + "loss": 1.9139, + "step": 76 + }, + { + "epoch": 0.01, + "learning_rate": 1.440598690364827e-06, + "loss": 1.891, + "step": 77 + }, + { + "epoch": 0.01, + "learning_rate": 1.4593077642656689e-06, + "loss": 1.7643, + "step": 78 + }, + { + "epoch": 0.01, + "learning_rate": 1.4780168381665108e-06, + "loss": 1.9433, + "step": 79 + }, + { + "epoch": 0.01, + "learning_rate": 1.4967259120673526e-06, + "loss": 1.8097, + "step": 80 + }, + { + "epoch": 0.01, + "learning_rate": 1.5154349859681948e-06, + "loss": 1.7479, + "step": 81 + }, + { + "epoch": 0.01, + "learning_rate": 1.5341440598690366e-06, + "loss": 1.7304, + "step": 82 + }, + { + "epoch": 0.01, + "learning_rate": 1.5528531337698785e-06, + "loss": 1.7655, + "step": 83 + }, + { + "epoch": 0.01, + "learning_rate": 1.5715622076707204e-06, + "loss": 1.7246, + "step": 84 + }, + { + "epoch": 0.01, + "learning_rate": 1.5902712815715623e-06, + "loss": 1.7196, + "step": 85 + }, + { + "epoch": 0.01, + "learning_rate": 1.6089803554724042e-06, + "loss": 1.8489, + "step": 86 + }, + { + "epoch": 0.01, + "learning_rate": 1.6276894293732461e-06, + "loss": 1.7815, + "step": 87 + }, + { + "epoch": 0.01, + "learning_rate": 1.6463985032740882e-06, + "loss": 1.7407, + "step": 88 + }, + { + "epoch": 0.01, + "learning_rate": 1.66510757717493e-06, + "loss": 1.8147, + "step": 89 + }, + { + "epoch": 0.01, + "learning_rate": 1.683816651075772e-06, + "loss": 1.7635, + "step": 90 + }, + { + "epoch": 0.01, + "learning_rate": 1.7025257249766139e-06, + "loss": 1.7644, + "step": 91 + }, + { + "epoch": 0.01, + "learning_rate": 1.7212347988774558e-06, + "loss": 1.724, + "step": 92 + }, + { + "epoch": 0.01, + "learning_rate": 1.7399438727782977e-06, + "loss": 1.7468, + "step": 93 + }, + { + "epoch": 0.01, + "learning_rate": 1.7586529466791396e-06, + "loss": 1.7407, + "step": 94 + }, + { + "epoch": 0.01, + "learning_rate": 1.7773620205799812e-06, + "loss": 1.6838, + "step": 95 + }, + { + "epoch": 0.01, + "learning_rate": 1.7960710944808231e-06, + "loss": 1.675, + "step": 96 + }, + { + "epoch": 0.01, + "learning_rate": 1.8147801683816652e-06, + "loss": 1.7736, + "step": 97 + }, + { + "epoch": 0.01, + "learning_rate": 1.8334892422825071e-06, + "loss": 1.8433, + "step": 98 + }, + { + "epoch": 0.01, + "learning_rate": 1.852198316183349e-06, + "loss": 1.7284, + "step": 99 + }, + { + "epoch": 0.01, + "learning_rate": 1.870907390084191e-06, + "loss": 1.6254, + "step": 100 + }, + { + "epoch": 0.01, + "learning_rate": 1.8896164639850328e-06, + "loss": 1.7183, + "step": 101 + }, + { + "epoch": 0.01, + "learning_rate": 1.9083255378858747e-06, + "loss": 1.7971, + "step": 102 + }, + { + "epoch": 0.01, + "learning_rate": 1.927034611786717e-06, + "loss": 1.8042, + "step": 103 + }, + { + "epoch": 0.01, + "learning_rate": 1.9457436856875585e-06, + "loss": 1.9779, + "step": 104 + }, + { + "epoch": 0.01, + "learning_rate": 1.9644527595884006e-06, + "loss": 1.8006, + "step": 105 + }, + { + "epoch": 0.01, + "learning_rate": 1.9831618334892423e-06, + "loss": 1.7822, + "step": 106 + }, + { + "epoch": 0.01, + "learning_rate": 2.0018709073900844e-06, + "loss": 1.6725, + "step": 107 + }, + { + "epoch": 0.01, + "learning_rate": 2.0205799812909265e-06, + "loss": 1.5739, + "step": 108 + }, + { + "epoch": 0.01, + "learning_rate": 2.039289055191768e-06, + "loss": 1.7052, + "step": 109 + }, + { + "epoch": 0.01, + "learning_rate": 2.0579981290926103e-06, + "loss": 1.7935, + "step": 110 + }, + { + "epoch": 0.01, + "learning_rate": 2.076707202993452e-06, + "loss": 1.7179, + "step": 111 + }, + { + "epoch": 0.01, + "learning_rate": 2.095416276894294e-06, + "loss": 1.7047, + "step": 112 + }, + { + "epoch": 0.01, + "learning_rate": 2.1141253507951357e-06, + "loss": 1.6893, + "step": 113 + }, + { + "epoch": 0.01, + "learning_rate": 2.132834424695978e-06, + "loss": 1.8181, + "step": 114 + }, + { + "epoch": 0.01, + "learning_rate": 2.15154349859682e-06, + "loss": 1.7348, + "step": 115 + }, + { + "epoch": 0.01, + "learning_rate": 2.1702525724976616e-06, + "loss": 1.6761, + "step": 116 + }, + { + "epoch": 0.01, + "learning_rate": 2.1889616463985033e-06, + "loss": 1.6752, + "step": 117 + }, + { + "epoch": 0.01, + "learning_rate": 2.2076707202993454e-06, + "loss": 1.8013, + "step": 118 + }, + { + "epoch": 0.01, + "learning_rate": 2.226379794200187e-06, + "loss": 1.6927, + "step": 119 + }, + { + "epoch": 0.01, + "learning_rate": 2.245088868101029e-06, + "loss": 1.632, + "step": 120 + }, + { + "epoch": 0.01, + "learning_rate": 2.263797942001871e-06, + "loss": 1.7707, + "step": 121 + }, + { + "epoch": 0.01, + "learning_rate": 2.282507015902713e-06, + "loss": 1.6991, + "step": 122 + }, + { + "epoch": 0.01, + "learning_rate": 2.301216089803555e-06, + "loss": 1.7245, + "step": 123 + }, + { + "epoch": 0.01, + "learning_rate": 2.3199251637043968e-06, + "loss": 1.6169, + "step": 124 + }, + { + "epoch": 0.01, + "learning_rate": 2.338634237605239e-06, + "loss": 1.6947, + "step": 125 + }, + { + "epoch": 0.01, + "learning_rate": 2.3573433115060805e-06, + "loss": 1.6388, + "step": 126 + }, + { + "epoch": 0.01, + "learning_rate": 2.3760523854069226e-06, + "loss": 1.658, + "step": 127 + }, + { + "epoch": 0.01, + "learning_rate": 2.3947614593077643e-06, + "loss": 1.7099, + "step": 128 + }, + { + "epoch": 0.01, + "learning_rate": 2.4134705332086064e-06, + "loss": 1.6653, + "step": 129 + }, + { + "epoch": 0.01, + "learning_rate": 2.4321796071094485e-06, + "loss": 1.7272, + "step": 130 + }, + { + "epoch": 0.01, + "learning_rate": 2.45088868101029e-06, + "loss": 1.7173, + "step": 131 + }, + { + "epoch": 0.01, + "learning_rate": 2.4695977549111323e-06, + "loss": 1.7293, + "step": 132 + }, + { + "epoch": 0.01, + "learning_rate": 2.488306828811974e-06, + "loss": 1.7062, + "step": 133 + }, + { + "epoch": 0.01, + "learning_rate": 2.507015902712816e-06, + "loss": 1.638, + "step": 134 + }, + { + "epoch": 0.01, + "learning_rate": 2.5257249766136578e-06, + "loss": 1.637, + "step": 135 + }, + { + "epoch": 0.01, + "learning_rate": 2.5444340505145e-06, + "loss": 1.6467, + "step": 136 + }, + { + "epoch": 0.01, + "learning_rate": 2.5631431244153416e-06, + "loss": 1.7813, + "step": 137 + }, + { + "epoch": 0.01, + "learning_rate": 2.5818521983161837e-06, + "loss": 1.5826, + "step": 138 + }, + { + "epoch": 0.01, + "learning_rate": 2.6005612722170253e-06, + "loss": 1.6557, + "step": 139 + }, + { + "epoch": 0.01, + "learning_rate": 2.6192703461178675e-06, + "loss": 1.638, + "step": 140 + }, + { + "epoch": 0.01, + "learning_rate": 2.637979420018709e-06, + "loss": 1.6365, + "step": 141 + }, + { + "epoch": 0.01, + "learning_rate": 2.6566884939195512e-06, + "loss": 1.6194, + "step": 142 + }, + { + "epoch": 0.01, + "learning_rate": 2.675397567820393e-06, + "loss": 1.7593, + "step": 143 + }, + { + "epoch": 0.01, + "learning_rate": 2.6941066417212354e-06, + "loss": 1.7075, + "step": 144 + }, + { + "epoch": 0.01, + "learning_rate": 2.7128157156220767e-06, + "loss": 1.7786, + "step": 145 + }, + { + "epoch": 0.01, + "learning_rate": 2.7315247895229192e-06, + "loss": 1.6366, + "step": 146 + }, + { + "epoch": 0.01, + "learning_rate": 2.750233863423761e-06, + "loss": 1.6153, + "step": 147 + }, + { + "epoch": 0.01, + "learning_rate": 2.7689429373246026e-06, + "loss": 1.7079, + "step": 148 + }, + { + "epoch": 0.01, + "learning_rate": 2.7876520112254447e-06, + "loss": 1.695, + "step": 149 + }, + { + "epoch": 0.01, + "learning_rate": 2.8063610851262864e-06, + "loss": 1.5956, + "step": 150 + }, + { + "epoch": 0.01, + "learning_rate": 2.8250701590271285e-06, + "loss": 1.5486, + "step": 151 + }, + { + "epoch": 0.01, + "learning_rate": 2.84377923292797e-06, + "loss": 1.6426, + "step": 152 + }, + { + "epoch": 0.01, + "learning_rate": 2.8624883068288123e-06, + "loss": 1.714, + "step": 153 + }, + { + "epoch": 0.01, + "learning_rate": 2.881197380729654e-06, + "loss": 1.6618, + "step": 154 + }, + { + "epoch": 0.01, + "learning_rate": 2.899906454630496e-06, + "loss": 1.7183, + "step": 155 + }, + { + "epoch": 0.01, + "learning_rate": 2.9186155285313377e-06, + "loss": 1.7116, + "step": 156 + }, + { + "epoch": 0.01, + "learning_rate": 2.93732460243218e-06, + "loss": 1.7404, + "step": 157 + }, + { + "epoch": 0.01, + "learning_rate": 2.9560336763330215e-06, + "loss": 1.6718, + "step": 158 + }, + { + "epoch": 0.01, + "learning_rate": 2.974742750233864e-06, + "loss": 1.7387, + "step": 159 + }, + { + "epoch": 0.01, + "learning_rate": 2.9934518241347053e-06, + "loss": 1.5916, + "step": 160 + }, + { + "epoch": 0.01, + "learning_rate": 3.012160898035548e-06, + "loss": 1.6435, + "step": 161 + }, + { + "epoch": 0.01, + "learning_rate": 3.0308699719363895e-06, + "loss": 1.554, + "step": 162 + }, + { + "epoch": 0.01, + "learning_rate": 3.0495790458372316e-06, + "loss": 1.6345, + "step": 163 + }, + { + "epoch": 0.01, + "learning_rate": 3.0682881197380733e-06, + "loss": 1.5527, + "step": 164 + }, + { + "epoch": 0.01, + "learning_rate": 3.0869971936389154e-06, + "loss": 1.6323, + "step": 165 + }, + { + "epoch": 0.01, + "learning_rate": 3.105706267539757e-06, + "loss": 1.5481, + "step": 166 + }, + { + "epoch": 0.01, + "learning_rate": 3.124415341440599e-06, + "loss": 1.7189, + "step": 167 + }, + { + "epoch": 0.01, + "learning_rate": 3.143124415341441e-06, + "loss": 1.6354, + "step": 168 + }, + { + "epoch": 0.01, + "learning_rate": 3.1618334892422825e-06, + "loss": 1.5776, + "step": 169 + }, + { + "epoch": 0.01, + "learning_rate": 3.1805425631431246e-06, + "loss": 1.6347, + "step": 170 + }, + { + "epoch": 0.01, + "learning_rate": 3.1992516370439663e-06, + "loss": 1.5706, + "step": 171 + }, + { + "epoch": 0.01, + "learning_rate": 3.2179607109448084e-06, + "loss": 1.6812, + "step": 172 + }, + { + "epoch": 0.01, + "learning_rate": 3.23666978484565e-06, + "loss": 1.6564, + "step": 173 + }, + { + "epoch": 0.01, + "learning_rate": 3.2553788587464922e-06, + "loss": 1.6807, + "step": 174 + }, + { + "epoch": 0.01, + "learning_rate": 3.274087932647334e-06, + "loss": 1.672, + "step": 175 + }, + { + "epoch": 0.01, + "learning_rate": 3.2927970065481764e-06, + "loss": 1.5667, + "step": 176 + }, + { + "epoch": 0.01, + "learning_rate": 3.311506080449018e-06, + "loss": 1.6547, + "step": 177 + }, + { + "epoch": 0.01, + "learning_rate": 3.33021515434986e-06, + "loss": 1.6896, + "step": 178 + }, + { + "epoch": 0.02, + "learning_rate": 3.348924228250702e-06, + "loss": 1.6217, + "step": 179 + }, + { + "epoch": 0.02, + "learning_rate": 3.367633302151544e-06, + "loss": 1.7164, + "step": 180 + }, + { + "epoch": 0.02, + "learning_rate": 3.3863423760523857e-06, + "loss": 1.6719, + "step": 181 + }, + { + "epoch": 0.02, + "learning_rate": 3.4050514499532278e-06, + "loss": 1.6394, + "step": 182 + }, + { + "epoch": 0.02, + "learning_rate": 3.4237605238540695e-06, + "loss": 1.5038, + "step": 183 + }, + { + "epoch": 0.02, + "learning_rate": 3.4424695977549116e-06, + "loss": 1.6622, + "step": 184 + }, + { + "epoch": 0.02, + "learning_rate": 3.4611786716557532e-06, + "loss": 1.5962, + "step": 185 + }, + { + "epoch": 0.02, + "learning_rate": 3.4798877455565953e-06, + "loss": 1.5568, + "step": 186 + }, + { + "epoch": 0.02, + "learning_rate": 3.498596819457437e-06, + "loss": 1.4548, + "step": 187 + }, + { + "epoch": 0.02, + "learning_rate": 3.517305893358279e-06, + "loss": 1.544, + "step": 188 + }, + { + "epoch": 0.02, + "learning_rate": 3.536014967259121e-06, + "loss": 1.5813, + "step": 189 + }, + { + "epoch": 0.02, + "learning_rate": 3.5547240411599625e-06, + "loss": 1.6289, + "step": 190 + }, + { + "epoch": 0.02, + "learning_rate": 3.573433115060805e-06, + "loss": 1.5993, + "step": 191 + }, + { + "epoch": 0.02, + "learning_rate": 3.5921421889616463e-06, + "loss": 1.5617, + "step": 192 + }, + { + "epoch": 0.02, + "learning_rate": 3.610851262862489e-06, + "loss": 1.5487, + "step": 193 + }, + { + "epoch": 0.02, + "learning_rate": 3.6295603367633305e-06, + "loss": 1.6523, + "step": 194 + }, + { + "epoch": 0.02, + "learning_rate": 3.6482694106641726e-06, + "loss": 1.7069, + "step": 195 + }, + { + "epoch": 0.02, + "learning_rate": 3.6669784845650143e-06, + "loss": 1.6394, + "step": 196 + }, + { + "epoch": 0.02, + "learning_rate": 3.6856875584658564e-06, + "loss": 1.636, + "step": 197 + }, + { + "epoch": 0.02, + "learning_rate": 3.704396632366698e-06, + "loss": 1.5758, + "step": 198 + }, + { + "epoch": 0.02, + "learning_rate": 3.72310570626754e-06, + "loss": 1.6268, + "step": 199 + }, + { + "epoch": 0.02, + "learning_rate": 3.741814780168382e-06, + "loss": 1.5913, + "step": 200 + }, + { + "epoch": 0.02, + "learning_rate": 3.760523854069224e-06, + "loss": 1.6472, + "step": 201 + }, + { + "epoch": 0.02, + "learning_rate": 3.7792329279700656e-06, + "loss": 1.5865, + "step": 202 + }, + { + "epoch": 0.02, + "learning_rate": 3.7979420018709077e-06, + "loss": 1.5344, + "step": 203 + }, + { + "epoch": 0.02, + "learning_rate": 3.816651075771749e-06, + "loss": 1.5284, + "step": 204 + }, + { + "epoch": 0.02, + "learning_rate": 3.8353601496725915e-06, + "loss": 1.5597, + "step": 205 + }, + { + "epoch": 0.02, + "learning_rate": 3.854069223573434e-06, + "loss": 1.5334, + "step": 206 + }, + { + "epoch": 0.02, + "learning_rate": 3.872778297474276e-06, + "loss": 1.591, + "step": 207 + }, + { + "epoch": 0.02, + "learning_rate": 3.891487371375117e-06, + "loss": 1.5911, + "step": 208 + }, + { + "epoch": 0.02, + "learning_rate": 3.910196445275959e-06, + "loss": 1.6435, + "step": 209 + }, + { + "epoch": 0.02, + "learning_rate": 3.928905519176801e-06, + "loss": 1.529, + "step": 210 + }, + { + "epoch": 0.02, + "learning_rate": 3.9476145930776424e-06, + "loss": 1.6517, + "step": 211 + }, + { + "epoch": 0.02, + "learning_rate": 3.9663236669784845e-06, + "loss": 1.6402, + "step": 212 + }, + { + "epoch": 0.02, + "learning_rate": 3.985032740879327e-06, + "loss": 1.5779, + "step": 213 + }, + { + "epoch": 0.02, + "learning_rate": 4.003741814780169e-06, + "loss": 1.5909, + "step": 214 + }, + { + "epoch": 0.02, + "learning_rate": 4.02245088868101e-06, + "loss": 1.6196, + "step": 215 + }, + { + "epoch": 0.02, + "learning_rate": 4.041159962581853e-06, + "loss": 1.5928, + "step": 216 + }, + { + "epoch": 0.02, + "learning_rate": 4.059869036482694e-06, + "loss": 1.6255, + "step": 217 + }, + { + "epoch": 0.02, + "learning_rate": 4.078578110383536e-06, + "loss": 1.5945, + "step": 218 + }, + { + "epoch": 0.02, + "learning_rate": 4.097287184284378e-06, + "loss": 1.6401, + "step": 219 + }, + { + "epoch": 0.02, + "learning_rate": 4.1159962581852205e-06, + "loss": 1.6468, + "step": 220 + }, + { + "epoch": 0.02, + "learning_rate": 4.134705332086062e-06, + "loss": 1.5973, + "step": 221 + }, + { + "epoch": 0.02, + "learning_rate": 4.153414405986904e-06, + "loss": 1.6128, + "step": 222 + }, + { + "epoch": 0.02, + "learning_rate": 4.172123479887746e-06, + "loss": 1.6975, + "step": 223 + }, + { + "epoch": 0.02, + "learning_rate": 4.190832553788588e-06, + "loss": 1.5582, + "step": 224 + }, + { + "epoch": 0.02, + "learning_rate": 4.209541627689429e-06, + "loss": 1.6158, + "step": 225 + }, + { + "epoch": 0.02, + "learning_rate": 4.2282507015902715e-06, + "loss": 1.5866, + "step": 226 + }, + { + "epoch": 0.02, + "learning_rate": 4.2469597754911136e-06, + "loss": 1.6538, + "step": 227 + }, + { + "epoch": 0.02, + "learning_rate": 4.265668849391956e-06, + "loss": 1.5829, + "step": 228 + }, + { + "epoch": 0.02, + "learning_rate": 4.284377923292797e-06, + "loss": 1.5564, + "step": 229 + }, + { + "epoch": 0.02, + "learning_rate": 4.30308699719364e-06, + "loss": 1.4608, + "step": 230 + }, + { + "epoch": 0.02, + "learning_rate": 4.321796071094481e-06, + "loss": 1.4806, + "step": 231 + }, + { + "epoch": 0.02, + "learning_rate": 4.340505144995323e-06, + "loss": 1.5999, + "step": 232 + }, + { + "epoch": 0.02, + "learning_rate": 4.359214218896165e-06, + "loss": 1.5976, + "step": 233 + }, + { + "epoch": 0.02, + "learning_rate": 4.377923292797007e-06, + "loss": 1.5114, + "step": 234 + }, + { + "epoch": 0.02, + "learning_rate": 4.396632366697849e-06, + "loss": 1.5107, + "step": 235 + }, + { + "epoch": 0.02, + "learning_rate": 4.415341440598691e-06, + "loss": 1.4515, + "step": 236 + }, + { + "epoch": 0.02, + "learning_rate": 4.434050514499533e-06, + "loss": 1.5189, + "step": 237 + }, + { + "epoch": 0.02, + "learning_rate": 4.452759588400374e-06, + "loss": 1.5427, + "step": 238 + }, + { + "epoch": 0.02, + "learning_rate": 4.471468662301216e-06, + "loss": 1.5397, + "step": 239 + }, + { + "epoch": 0.02, + "learning_rate": 4.490177736202058e-06, + "loss": 1.515, + "step": 240 + }, + { + "epoch": 0.02, + "learning_rate": 4.5088868101029005e-06, + "loss": 1.4614, + "step": 241 + }, + { + "epoch": 0.02, + "learning_rate": 4.527595884003742e-06, + "loss": 1.5577, + "step": 242 + }, + { + "epoch": 0.02, + "learning_rate": 4.546304957904585e-06, + "loss": 1.5188, + "step": 243 + }, + { + "epoch": 0.02, + "learning_rate": 4.565014031805426e-06, + "loss": 1.5655, + "step": 244 + }, + { + "epoch": 0.02, + "learning_rate": 4.583723105706268e-06, + "loss": 1.5576, + "step": 245 + }, + { + "epoch": 0.02, + "learning_rate": 4.60243217960711e-06, + "loss": 1.6968, + "step": 246 + }, + { + "epoch": 0.02, + "learning_rate": 4.621141253507952e-06, + "loss": 1.509, + "step": 247 + }, + { + "epoch": 0.02, + "learning_rate": 4.6398503274087935e-06, + "loss": 1.5299, + "step": 248 + }, + { + "epoch": 0.02, + "learning_rate": 4.658559401309636e-06, + "loss": 1.5334, + "step": 249 + }, + { + "epoch": 0.02, + "learning_rate": 4.677268475210478e-06, + "loss": 1.5051, + "step": 250 + }, + { + "epoch": 0.02, + "learning_rate": 4.69597754911132e-06, + "loss": 1.5549, + "step": 251 + }, + { + "epoch": 0.02, + "learning_rate": 4.714686623012161e-06, + "loss": 1.5121, + "step": 252 + }, + { + "epoch": 0.02, + "learning_rate": 4.733395696913003e-06, + "loss": 1.4929, + "step": 253 + }, + { + "epoch": 0.02, + "learning_rate": 4.752104770813845e-06, + "loss": 1.5601, + "step": 254 + }, + { + "epoch": 0.02, + "learning_rate": 4.7708138447146865e-06, + "loss": 1.5546, + "step": 255 + }, + { + "epoch": 0.02, + "learning_rate": 4.789522918615529e-06, + "loss": 1.609, + "step": 256 + }, + { + "epoch": 0.02, + "learning_rate": 4.808231992516371e-06, + "loss": 1.5229, + "step": 257 + }, + { + "epoch": 0.02, + "learning_rate": 4.826941066417213e-06, + "loss": 1.5324, + "step": 258 + }, + { + "epoch": 0.02, + "learning_rate": 4.845650140318054e-06, + "loss": 1.5238, + "step": 259 + }, + { + "epoch": 0.02, + "learning_rate": 4.864359214218897e-06, + "loss": 1.5932, + "step": 260 + }, + { + "epoch": 0.02, + "learning_rate": 4.883068288119738e-06, + "loss": 1.4929, + "step": 261 + }, + { + "epoch": 0.02, + "learning_rate": 4.90177736202058e-06, + "loss": 1.5203, + "step": 262 + }, + { + "epoch": 0.02, + "learning_rate": 4.9204864359214225e-06, + "loss": 1.523, + "step": 263 + }, + { + "epoch": 0.02, + "learning_rate": 4.939195509822265e-06, + "loss": 1.5615, + "step": 264 + }, + { + "epoch": 0.02, + "learning_rate": 4.957904583723106e-06, + "loss": 1.5465, + "step": 265 + }, + { + "epoch": 0.02, + "learning_rate": 4.976613657623948e-06, + "loss": 1.4972, + "step": 266 + }, + { + "epoch": 0.02, + "learning_rate": 4.99532273152479e-06, + "loss": 1.5591, + "step": 267 + }, + { + "epoch": 0.02, + "learning_rate": 5.014031805425632e-06, + "loss": 1.5067, + "step": 268 + }, + { + "epoch": 0.02, + "learning_rate": 5.032740879326474e-06, + "loss": 1.4587, + "step": 269 + }, + { + "epoch": 0.02, + "learning_rate": 5.0514499532273156e-06, + "loss": 1.5246, + "step": 270 + }, + { + "epoch": 0.02, + "learning_rate": 5.070159027128158e-06, + "loss": 1.5632, + "step": 271 + }, + { + "epoch": 0.02, + "learning_rate": 5.088868101029e-06, + "loss": 1.4836, + "step": 272 + }, + { + "epoch": 0.02, + "learning_rate": 5.107577174929841e-06, + "loss": 1.5629, + "step": 273 + }, + { + "epoch": 0.02, + "learning_rate": 5.126286248830683e-06, + "loss": 1.6649, + "step": 274 + }, + { + "epoch": 0.02, + "learning_rate": 5.144995322731525e-06, + "loss": 1.6517, + "step": 275 + }, + { + "epoch": 0.02, + "learning_rate": 5.163704396632367e-06, + "loss": 1.5324, + "step": 276 + }, + { + "epoch": 0.02, + "learning_rate": 5.182413470533209e-06, + "loss": 1.529, + "step": 277 + }, + { + "epoch": 0.02, + "learning_rate": 5.201122544434051e-06, + "loss": 1.6024, + "step": 278 + }, + { + "epoch": 0.02, + "learning_rate": 5.219831618334893e-06, + "loss": 1.5287, + "step": 279 + }, + { + "epoch": 0.02, + "learning_rate": 5.238540692235735e-06, + "loss": 1.5564, + "step": 280 + }, + { + "epoch": 0.02, + "learning_rate": 5.257249766136576e-06, + "loss": 1.5218, + "step": 281 + }, + { + "epoch": 0.02, + "learning_rate": 5.275958840037418e-06, + "loss": 1.428, + "step": 282 + }, + { + "epoch": 0.02, + "learning_rate": 5.29466791393826e-06, + "loss": 1.569, + "step": 283 + }, + { + "epoch": 0.02, + "learning_rate": 5.3133769878391025e-06, + "loss": 1.5178, + "step": 284 + }, + { + "epoch": 0.02, + "learning_rate": 5.332086061739944e-06, + "loss": 1.488, + "step": 285 + }, + { + "epoch": 0.02, + "learning_rate": 5.350795135640786e-06, + "loss": 1.4842, + "step": 286 + }, + { + "epoch": 0.02, + "learning_rate": 5.369504209541628e-06, + "loss": 1.5285, + "step": 287 + }, + { + "epoch": 0.02, + "learning_rate": 5.388213283442471e-06, + "loss": 1.5341, + "step": 288 + }, + { + "epoch": 0.02, + "learning_rate": 5.406922357343311e-06, + "loss": 1.5429, + "step": 289 + }, + { + "epoch": 0.02, + "learning_rate": 5.425631431244153e-06, + "loss": 1.4935, + "step": 290 + }, + { + "epoch": 0.02, + "learning_rate": 5.444340505144996e-06, + "loss": 1.6386, + "step": 291 + }, + { + "epoch": 0.02, + "learning_rate": 5.4630495790458385e-06, + "loss": 1.4528, + "step": 292 + }, + { + "epoch": 0.02, + "learning_rate": 5.48175865294668e-06, + "loss": 1.544, + "step": 293 + }, + { + "epoch": 0.02, + "learning_rate": 5.500467726847522e-06, + "loss": 1.6428, + "step": 294 + }, + { + "epoch": 0.02, + "learning_rate": 5.519176800748364e-06, + "loss": 1.543, + "step": 295 + }, + { + "epoch": 0.02, + "learning_rate": 5.537885874649205e-06, + "loss": 1.5062, + "step": 296 + }, + { + "epoch": 0.03, + "learning_rate": 5.556594948550047e-06, + "loss": 1.5437, + "step": 297 + }, + { + "epoch": 0.03, + "learning_rate": 5.575304022450889e-06, + "loss": 1.5781, + "step": 298 + }, + { + "epoch": 0.03, + "learning_rate": 5.5940130963517315e-06, + "loss": 1.5289, + "step": 299 + }, + { + "epoch": 0.03, + "learning_rate": 5.612722170252573e-06, + "loss": 1.4553, + "step": 300 + }, + { + "epoch": 0.03, + "learning_rate": 5.631431244153415e-06, + "loss": 1.4762, + "step": 301 + }, + { + "epoch": 0.03, + "learning_rate": 5.650140318054257e-06, + "loss": 1.5723, + "step": 302 + }, + { + "epoch": 0.03, + "learning_rate": 5.668849391955099e-06, + "loss": 1.4661, + "step": 303 + }, + { + "epoch": 0.03, + "learning_rate": 5.68755846585594e-06, + "loss": 1.5501, + "step": 304 + }, + { + "epoch": 0.03, + "learning_rate": 5.706267539756782e-06, + "loss": 1.5415, + "step": 305 + }, + { + "epoch": 0.03, + "learning_rate": 5.7249766136576245e-06, + "loss": 1.5, + "step": 306 + }, + { + "epoch": 0.03, + "learning_rate": 5.743685687558467e-06, + "loss": 1.4951, + "step": 307 + }, + { + "epoch": 0.03, + "learning_rate": 5.762394761459308e-06, + "loss": 1.5193, + "step": 308 + }, + { + "epoch": 0.03, + "learning_rate": 5.78110383536015e-06, + "loss": 1.5084, + "step": 309 + }, + { + "epoch": 0.03, + "learning_rate": 5.799812909260992e-06, + "loss": 1.5008, + "step": 310 + }, + { + "epoch": 0.03, + "learning_rate": 5.818521983161834e-06, + "loss": 1.4402, + "step": 311 + }, + { + "epoch": 0.03, + "learning_rate": 5.8372310570626755e-06, + "loss": 1.577, + "step": 312 + }, + { + "epoch": 0.03, + "learning_rate": 5.8559401309635176e-06, + "loss": 1.4923, + "step": 313 + }, + { + "epoch": 0.03, + "learning_rate": 5.87464920486436e-06, + "loss": 1.5023, + "step": 314 + }, + { + "epoch": 0.03, + "learning_rate": 5.893358278765201e-06, + "loss": 1.4437, + "step": 315 + }, + { + "epoch": 0.03, + "learning_rate": 5.912067352666043e-06, + "loss": 1.442, + "step": 316 + }, + { + "epoch": 0.03, + "learning_rate": 5.930776426566885e-06, + "loss": 1.4798, + "step": 317 + }, + { + "epoch": 0.03, + "learning_rate": 5.949485500467728e-06, + "loss": 1.5724, + "step": 318 + }, + { + "epoch": 0.03, + "learning_rate": 5.9681945743685685e-06, + "loss": 1.5251, + "step": 319 + }, + { + "epoch": 0.03, + "learning_rate": 5.986903648269411e-06, + "loss": 1.4784, + "step": 320 + }, + { + "epoch": 0.03, + "learning_rate": 6.0056127221702535e-06, + "loss": 1.5184, + "step": 321 + }, + { + "epoch": 0.03, + "learning_rate": 6.024321796071096e-06, + "loss": 1.4868, + "step": 322 + }, + { + "epoch": 0.03, + "learning_rate": 6.043030869971936e-06, + "loss": 1.5227, + "step": 323 + }, + { + "epoch": 0.03, + "learning_rate": 6.061739943872779e-06, + "loss": 1.5864, + "step": 324 + }, + { + "epoch": 0.03, + "learning_rate": 6.080449017773621e-06, + "loss": 1.5195, + "step": 325 + }, + { + "epoch": 0.03, + "learning_rate": 6.099158091674463e-06, + "loss": 1.5446, + "step": 326 + }, + { + "epoch": 0.03, + "learning_rate": 6.1178671655753045e-06, + "loss": 1.6046, + "step": 327 + }, + { + "epoch": 0.03, + "learning_rate": 6.136576239476147e-06, + "loss": 1.3602, + "step": 328 + }, + { + "epoch": 0.03, + "learning_rate": 6.155285313376989e-06, + "loss": 1.5306, + "step": 329 + }, + { + "epoch": 0.03, + "learning_rate": 6.173994387277831e-06, + "loss": 1.5261, + "step": 330 + }, + { + "epoch": 0.03, + "learning_rate": 6.192703461178672e-06, + "loss": 1.4634, + "step": 331 + }, + { + "epoch": 0.03, + "learning_rate": 6.211412535079514e-06, + "loss": 1.4756, + "step": 332 + }, + { + "epoch": 0.03, + "learning_rate": 6.230121608980356e-06, + "loss": 1.6169, + "step": 333 + }, + { + "epoch": 0.03, + "learning_rate": 6.248830682881198e-06, + "loss": 1.5211, + "step": 334 + }, + { + "epoch": 0.03, + "learning_rate": 6.26753975678204e-06, + "loss": 1.5917, + "step": 335 + }, + { + "epoch": 0.03, + "learning_rate": 6.286248830682882e-06, + "loss": 1.515, + "step": 336 + }, + { + "epoch": 0.03, + "learning_rate": 6.304957904583724e-06, + "loss": 1.4717, + "step": 337 + }, + { + "epoch": 0.03, + "learning_rate": 6.323666978484565e-06, + "loss": 1.5532, + "step": 338 + }, + { + "epoch": 0.03, + "learning_rate": 6.342376052385407e-06, + "loss": 1.5061, + "step": 339 + }, + { + "epoch": 0.03, + "learning_rate": 6.361085126286249e-06, + "loss": 1.4849, + "step": 340 + }, + { + "epoch": 0.03, + "learning_rate": 6.379794200187091e-06, + "loss": 1.4761, + "step": 341 + }, + { + "epoch": 0.03, + "learning_rate": 6.398503274087933e-06, + "loss": 1.4567, + "step": 342 + }, + { + "epoch": 0.03, + "learning_rate": 6.417212347988775e-06, + "loss": 1.4137, + "step": 343 + }, + { + "epoch": 0.03, + "learning_rate": 6.435921421889617e-06, + "loss": 1.4779, + "step": 344 + }, + { + "epoch": 0.03, + "learning_rate": 6.454630495790459e-06, + "loss": 1.5159, + "step": 345 + }, + { + "epoch": 0.03, + "learning_rate": 6.4733395696913e-06, + "loss": 1.357, + "step": 346 + }, + { + "epoch": 0.03, + "learning_rate": 6.492048643592142e-06, + "loss": 1.494, + "step": 347 + }, + { + "epoch": 0.03, + "learning_rate": 6.5107577174929844e-06, + "loss": 1.5151, + "step": 348 + }, + { + "epoch": 0.03, + "learning_rate": 6.529466791393827e-06, + "loss": 1.441, + "step": 349 + }, + { + "epoch": 0.03, + "learning_rate": 6.548175865294668e-06, + "loss": 1.4676, + "step": 350 + }, + { + "epoch": 0.03, + "learning_rate": 6.56688493919551e-06, + "loss": 1.4213, + "step": 351 + }, + { + "epoch": 0.03, + "learning_rate": 6.585594013096353e-06, + "loss": 1.4398, + "step": 352 + }, + { + "epoch": 0.03, + "learning_rate": 6.604303086997195e-06, + "loss": 1.4668, + "step": 353 + }, + { + "epoch": 0.03, + "learning_rate": 6.623012160898036e-06, + "loss": 1.5124, + "step": 354 + }, + { + "epoch": 0.03, + "learning_rate": 6.641721234798878e-06, + "loss": 1.4926, + "step": 355 + }, + { + "epoch": 0.03, + "learning_rate": 6.66043030869972e-06, + "loss": 1.4255, + "step": 356 + }, + { + "epoch": 0.03, + "learning_rate": 6.679139382600562e-06, + "loss": 1.5373, + "step": 357 + }, + { + "epoch": 0.03, + "learning_rate": 6.697848456501404e-06, + "loss": 1.489, + "step": 358 + }, + { + "epoch": 0.03, + "learning_rate": 6.716557530402246e-06, + "loss": 1.4995, + "step": 359 + }, + { + "epoch": 0.03, + "learning_rate": 6.735266604303088e-06, + "loss": 1.5306, + "step": 360 + }, + { + "epoch": 0.03, + "learning_rate": 6.753975678203929e-06, + "loss": 1.4396, + "step": 361 + }, + { + "epoch": 0.03, + "learning_rate": 6.772684752104771e-06, + "loss": 1.5138, + "step": 362 + }, + { + "epoch": 0.03, + "learning_rate": 6.7913938260056134e-06, + "loss": 1.4912, + "step": 363 + }, + { + "epoch": 0.03, + "learning_rate": 6.8101028999064555e-06, + "loss": 1.6047, + "step": 364 + }, + { + "epoch": 0.03, + "learning_rate": 6.828811973807297e-06, + "loss": 1.333, + "step": 365 + }, + { + "epoch": 0.03, + "learning_rate": 6.847521047708139e-06, + "loss": 1.3892, + "step": 366 + }, + { + "epoch": 0.03, + "learning_rate": 6.866230121608981e-06, + "loss": 1.4207, + "step": 367 + }, + { + "epoch": 0.03, + "learning_rate": 6.884939195509823e-06, + "loss": 1.4554, + "step": 368 + }, + { + "epoch": 0.03, + "learning_rate": 6.903648269410664e-06, + "loss": 1.4573, + "step": 369 + }, + { + "epoch": 0.03, + "learning_rate": 6.9223573433115065e-06, + "loss": 1.4928, + "step": 370 + }, + { + "epoch": 0.03, + "learning_rate": 6.941066417212349e-06, + "loss": 1.4236, + "step": 371 + }, + { + "epoch": 0.03, + "learning_rate": 6.959775491113191e-06, + "loss": 1.5038, + "step": 372 + }, + { + "epoch": 0.03, + "learning_rate": 6.978484565014032e-06, + "loss": 1.4777, + "step": 373 + }, + { + "epoch": 0.03, + "learning_rate": 6.997193638914874e-06, + "loss": 1.524, + "step": 374 + }, + { + "epoch": 0.03, + "learning_rate": 7.015902712815716e-06, + "loss": 1.449, + "step": 375 + }, + { + "epoch": 0.03, + "learning_rate": 7.034611786716558e-06, + "loss": 1.4591, + "step": 376 + }, + { + "epoch": 0.03, + "learning_rate": 7.0533208606173995e-06, + "loss": 1.4037, + "step": 377 + }, + { + "epoch": 0.03, + "learning_rate": 7.072029934518242e-06, + "loss": 1.5396, + "step": 378 + }, + { + "epoch": 0.03, + "learning_rate": 7.090739008419084e-06, + "loss": 1.4717, + "step": 379 + }, + { + "epoch": 0.03, + "learning_rate": 7.109448082319925e-06, + "loss": 1.5271, + "step": 380 + }, + { + "epoch": 0.03, + "learning_rate": 7.128157156220767e-06, + "loss": 1.5319, + "step": 381 + }, + { + "epoch": 0.03, + "learning_rate": 7.14686623012161e-06, + "loss": 1.47, + "step": 382 + }, + { + "epoch": 0.03, + "learning_rate": 7.165575304022452e-06, + "loss": 1.477, + "step": 383 + }, + { + "epoch": 0.03, + "learning_rate": 7.1842843779232925e-06, + "loss": 1.5313, + "step": 384 + }, + { + "epoch": 0.03, + "learning_rate": 7.2029934518241355e-06, + "loss": 1.5691, + "step": 385 + }, + { + "epoch": 0.03, + "learning_rate": 7.221702525724978e-06, + "loss": 1.5078, + "step": 386 + }, + { + "epoch": 0.03, + "learning_rate": 7.24041159962582e-06, + "loss": 1.4535, + "step": 387 + }, + { + "epoch": 0.03, + "learning_rate": 7.259120673526661e-06, + "loss": 1.3301, + "step": 388 + }, + { + "epoch": 0.03, + "learning_rate": 7.277829747427503e-06, + "loss": 1.4688, + "step": 389 + }, + { + "epoch": 0.03, + "learning_rate": 7.296538821328345e-06, + "loss": 1.545, + "step": 390 + }, + { + "epoch": 0.03, + "learning_rate": 7.315247895229187e-06, + "loss": 1.4386, + "step": 391 + }, + { + "epoch": 0.03, + "learning_rate": 7.3339569691300285e-06, + "loss": 1.4994, + "step": 392 + }, + { + "epoch": 0.03, + "learning_rate": 7.352666043030871e-06, + "loss": 1.5172, + "step": 393 + }, + { + "epoch": 0.03, + "learning_rate": 7.371375116931713e-06, + "loss": 1.3704, + "step": 394 + }, + { + "epoch": 0.03, + "learning_rate": 7.390084190832555e-06, + "loss": 1.5456, + "step": 395 + }, + { + "epoch": 0.03, + "learning_rate": 7.408793264733396e-06, + "loss": 1.4766, + "step": 396 + }, + { + "epoch": 0.03, + "learning_rate": 7.427502338634238e-06, + "loss": 1.4078, + "step": 397 + }, + { + "epoch": 0.03, + "learning_rate": 7.44621141253508e-06, + "loss": 1.4741, + "step": 398 + }, + { + "epoch": 0.03, + "learning_rate": 7.4649204864359216e-06, + "loss": 1.4731, + "step": 399 + }, + { + "epoch": 0.03, + "learning_rate": 7.483629560336764e-06, + "loss": 1.5046, + "step": 400 + }, + { + "epoch": 0.03, + "learning_rate": 7.502338634237606e-06, + "loss": 1.5039, + "step": 401 + }, + { + "epoch": 0.03, + "learning_rate": 7.521047708138448e-06, + "loss": 1.4073, + "step": 402 + }, + { + "epoch": 0.03, + "learning_rate": 7.539756782039289e-06, + "loss": 1.4643, + "step": 403 + }, + { + "epoch": 0.03, + "learning_rate": 7.558465855940131e-06, + "loss": 1.3705, + "step": 404 + }, + { + "epoch": 0.03, + "learning_rate": 7.577174929840973e-06, + "loss": 1.5405, + "step": 405 + }, + { + "epoch": 0.03, + "learning_rate": 7.5958840037418154e-06, + "loss": 1.4846, + "step": 406 + }, + { + "epoch": 0.03, + "learning_rate": 7.614593077642657e-06, + "loss": 1.4937, + "step": 407 + }, + { + "epoch": 0.03, + "learning_rate": 7.633302151543499e-06, + "loss": 1.5272, + "step": 408 + }, + { + "epoch": 0.03, + "learning_rate": 7.652011225444342e-06, + "loss": 1.4392, + "step": 409 + }, + { + "epoch": 0.03, + "learning_rate": 7.670720299345183e-06, + "loss": 1.4629, + "step": 410 + }, + { + "epoch": 0.03, + "learning_rate": 7.689429373246024e-06, + "loss": 1.4415, + "step": 411 + }, + { + "epoch": 0.03, + "learning_rate": 7.708138447146867e-06, + "loss": 1.4745, + "step": 412 + }, + { + "epoch": 0.03, + "learning_rate": 7.726847521047708e-06, + "loss": 1.3969, + "step": 413 + }, + { + "epoch": 0.03, + "learning_rate": 7.745556594948551e-06, + "loss": 1.5633, + "step": 414 + }, + { + "epoch": 0.03, + "learning_rate": 7.764265668849393e-06, + "loss": 1.4348, + "step": 415 + }, + { + "epoch": 0.04, + "learning_rate": 7.782974742750234e-06, + "loss": 1.4071, + "step": 416 + }, + { + "epoch": 0.04, + "learning_rate": 7.801683816651077e-06, + "loss": 1.5096, + "step": 417 + }, + { + "epoch": 0.04, + "learning_rate": 7.820392890551918e-06, + "loss": 1.4652, + "step": 418 + }, + { + "epoch": 0.04, + "learning_rate": 7.83910196445276e-06, + "loss": 1.4242, + "step": 419 + }, + { + "epoch": 0.04, + "learning_rate": 7.857811038353602e-06, + "loss": 1.5101, + "step": 420 + }, + { + "epoch": 0.04, + "learning_rate": 7.876520112254444e-06, + "loss": 1.473, + "step": 421 + }, + { + "epoch": 0.04, + "learning_rate": 7.895229186155285e-06, + "loss": 1.4013, + "step": 422 + }, + { + "epoch": 0.04, + "learning_rate": 7.913938260056128e-06, + "loss": 1.5188, + "step": 423 + }, + { + "epoch": 0.04, + "learning_rate": 7.932647333956969e-06, + "loss": 1.464, + "step": 424 + }, + { + "epoch": 0.04, + "learning_rate": 7.951356407857812e-06, + "loss": 1.4754, + "step": 425 + }, + { + "epoch": 0.04, + "learning_rate": 7.970065481758653e-06, + "loss": 1.4754, + "step": 426 + }, + { + "epoch": 0.04, + "learning_rate": 7.988774555659495e-06, + "loss": 1.3936, + "step": 427 + }, + { + "epoch": 0.04, + "learning_rate": 8.007483629560337e-06, + "loss": 1.4679, + "step": 428 + }, + { + "epoch": 0.04, + "learning_rate": 8.02619270346118e-06, + "loss": 1.4254, + "step": 429 + }, + { + "epoch": 0.04, + "learning_rate": 8.04490177736202e-06, + "loss": 1.5345, + "step": 430 + }, + { + "epoch": 0.04, + "learning_rate": 8.063610851262863e-06, + "loss": 1.4943, + "step": 431 + }, + { + "epoch": 0.04, + "learning_rate": 8.082319925163706e-06, + "loss": 1.3369, + "step": 432 + }, + { + "epoch": 0.04, + "learning_rate": 8.101028999064547e-06, + "loss": 1.424, + "step": 433 + }, + { + "epoch": 0.04, + "learning_rate": 8.119738072965388e-06, + "loss": 1.4223, + "step": 434 + }, + { + "epoch": 0.04, + "learning_rate": 8.138447146866231e-06, + "loss": 1.3512, + "step": 435 + }, + { + "epoch": 0.04, + "learning_rate": 8.157156220767073e-06, + "loss": 1.4415, + "step": 436 + }, + { + "epoch": 0.04, + "learning_rate": 8.175865294667916e-06, + "loss": 1.5064, + "step": 437 + }, + { + "epoch": 0.04, + "learning_rate": 8.194574368568757e-06, + "loss": 1.3734, + "step": 438 + }, + { + "epoch": 0.04, + "learning_rate": 8.213283442469598e-06, + "loss": 1.5322, + "step": 439 + }, + { + "epoch": 0.04, + "learning_rate": 8.231992516370441e-06, + "loss": 1.5067, + "step": 440 + }, + { + "epoch": 0.04, + "learning_rate": 8.250701590271282e-06, + "loss": 1.5014, + "step": 441 + }, + { + "epoch": 0.04, + "learning_rate": 8.269410664172124e-06, + "loss": 1.4607, + "step": 442 + }, + { + "epoch": 0.04, + "learning_rate": 8.288119738072967e-06, + "loss": 1.4436, + "step": 443 + }, + { + "epoch": 0.04, + "learning_rate": 8.306828811973808e-06, + "loss": 1.5251, + "step": 444 + }, + { + "epoch": 0.04, + "learning_rate": 8.325537885874649e-06, + "loss": 1.4914, + "step": 445 + }, + { + "epoch": 0.04, + "learning_rate": 8.344246959775492e-06, + "loss": 1.5804, + "step": 446 + }, + { + "epoch": 0.04, + "learning_rate": 8.362956033676333e-06, + "loss": 1.5989, + "step": 447 + }, + { + "epoch": 0.04, + "learning_rate": 8.381665107577176e-06, + "loss": 1.407, + "step": 448 + }, + { + "epoch": 0.04, + "learning_rate": 8.400374181478017e-06, + "loss": 1.4745, + "step": 449 + }, + { + "epoch": 0.04, + "learning_rate": 8.419083255378859e-06, + "loss": 1.5069, + "step": 450 + }, + { + "epoch": 0.04, + "learning_rate": 8.437792329279702e-06, + "loss": 1.4974, + "step": 451 + }, + { + "epoch": 0.04, + "learning_rate": 8.456501403180543e-06, + "loss": 1.431, + "step": 452 + }, + { + "epoch": 0.04, + "learning_rate": 8.475210477081384e-06, + "loss": 1.4441, + "step": 453 + }, + { + "epoch": 0.04, + "learning_rate": 8.493919550982227e-06, + "loss": 1.5066, + "step": 454 + }, + { + "epoch": 0.04, + "learning_rate": 8.512628624883068e-06, + "loss": 1.459, + "step": 455 + }, + { + "epoch": 0.04, + "learning_rate": 8.531337698783911e-06, + "loss": 1.4871, + "step": 456 + }, + { + "epoch": 0.04, + "learning_rate": 8.550046772684753e-06, + "loss": 1.4383, + "step": 457 + }, + { + "epoch": 0.04, + "learning_rate": 8.568755846585594e-06, + "loss": 1.4953, + "step": 458 + }, + { + "epoch": 0.04, + "learning_rate": 8.587464920486437e-06, + "loss": 1.4495, + "step": 459 + }, + { + "epoch": 0.04, + "learning_rate": 8.60617399438728e-06, + "loss": 1.5321, + "step": 460 + }, + { + "epoch": 0.04, + "learning_rate": 8.624883068288121e-06, + "loss": 1.4824, + "step": 461 + }, + { + "epoch": 0.04, + "learning_rate": 8.643592142188962e-06, + "loss": 1.4416, + "step": 462 + }, + { + "epoch": 0.04, + "learning_rate": 8.662301216089805e-06, + "loss": 1.4778, + "step": 463 + }, + { + "epoch": 0.04, + "learning_rate": 8.681010289990646e-06, + "loss": 1.4098, + "step": 464 + }, + { + "epoch": 0.04, + "learning_rate": 8.699719363891488e-06, + "loss": 1.5615, + "step": 465 + }, + { + "epoch": 0.04, + "learning_rate": 8.71842843779233e-06, + "loss": 1.5087, + "step": 466 + }, + { + "epoch": 0.04, + "learning_rate": 8.737137511693172e-06, + "loss": 1.4599, + "step": 467 + }, + { + "epoch": 0.04, + "learning_rate": 8.755846585594013e-06, + "loss": 1.5709, + "step": 468 + }, + { + "epoch": 0.04, + "learning_rate": 8.774555659494856e-06, + "loss": 1.4338, + "step": 469 + }, + { + "epoch": 0.04, + "learning_rate": 8.793264733395697e-06, + "loss": 1.4382, + "step": 470 + }, + { + "epoch": 0.04, + "learning_rate": 8.81197380729654e-06, + "loss": 1.42, + "step": 471 + }, + { + "epoch": 0.04, + "learning_rate": 8.830682881197382e-06, + "loss": 1.5261, + "step": 472 + }, + { + "epoch": 0.04, + "learning_rate": 8.849391955098223e-06, + "loss": 1.4447, + "step": 473 + }, + { + "epoch": 0.04, + "learning_rate": 8.868101028999066e-06, + "loss": 1.4208, + "step": 474 + }, + { + "epoch": 0.04, + "learning_rate": 8.886810102899907e-06, + "loss": 1.4902, + "step": 475 + }, + { + "epoch": 0.04, + "learning_rate": 8.905519176800748e-06, + "loss": 1.4461, + "step": 476 + }, + { + "epoch": 0.04, + "learning_rate": 8.924228250701591e-06, + "loss": 1.4726, + "step": 477 + }, + { + "epoch": 0.04, + "learning_rate": 8.942937324602433e-06, + "loss": 1.3446, + "step": 478 + }, + { + "epoch": 0.04, + "learning_rate": 8.961646398503275e-06, + "loss": 1.6174, + "step": 479 + }, + { + "epoch": 0.04, + "learning_rate": 8.980355472404117e-06, + "loss": 1.359, + "step": 480 + }, + { + "epoch": 0.04, + "learning_rate": 8.999064546304958e-06, + "loss": 1.4489, + "step": 481 + }, + { + "epoch": 0.04, + "learning_rate": 9.017773620205801e-06, + "loss": 1.4736, + "step": 482 + }, + { + "epoch": 0.04, + "learning_rate": 9.036482694106642e-06, + "loss": 1.4734, + "step": 483 + }, + { + "epoch": 0.04, + "learning_rate": 9.055191768007483e-06, + "loss": 1.4848, + "step": 484 + }, + { + "epoch": 0.04, + "learning_rate": 9.073900841908326e-06, + "loss": 1.5038, + "step": 485 + }, + { + "epoch": 0.04, + "learning_rate": 9.09260991580917e-06, + "loss": 1.6397, + "step": 486 + }, + { + "epoch": 0.04, + "learning_rate": 9.111318989710009e-06, + "loss": 1.3728, + "step": 487 + }, + { + "epoch": 0.04, + "learning_rate": 9.130028063610852e-06, + "loss": 1.5714, + "step": 488 + }, + { + "epoch": 0.04, + "learning_rate": 9.148737137511695e-06, + "loss": 1.425, + "step": 489 + }, + { + "epoch": 0.04, + "learning_rate": 9.167446211412536e-06, + "loss": 1.4343, + "step": 490 + }, + { + "epoch": 0.04, + "learning_rate": 9.186155285313377e-06, + "loss": 1.3972, + "step": 491 + }, + { + "epoch": 0.04, + "learning_rate": 9.20486435921422e-06, + "loss": 1.4271, + "step": 492 + }, + { + "epoch": 0.04, + "learning_rate": 9.223573433115062e-06, + "loss": 1.4414, + "step": 493 + }, + { + "epoch": 0.04, + "learning_rate": 9.242282507015905e-06, + "loss": 1.4991, + "step": 494 + }, + { + "epoch": 0.04, + "learning_rate": 9.260991580916746e-06, + "loss": 1.5106, + "step": 495 + }, + { + "epoch": 0.04, + "learning_rate": 9.279700654817587e-06, + "loss": 1.4086, + "step": 496 + }, + { + "epoch": 0.04, + "learning_rate": 9.29840972871843e-06, + "loss": 1.4286, + "step": 497 + }, + { + "epoch": 0.04, + "learning_rate": 9.317118802619271e-06, + "loss": 1.5129, + "step": 498 + }, + { + "epoch": 0.04, + "learning_rate": 9.335827876520112e-06, + "loss": 1.3698, + "step": 499 + }, + { + "epoch": 0.04, + "learning_rate": 9.354536950420955e-06, + "loss": 1.4821, + "step": 500 + }, + { + "epoch": 0.04, + "learning_rate": 9.373246024321797e-06, + "loss": 1.4679, + "step": 501 + }, + { + "epoch": 0.04, + "learning_rate": 9.39195509822264e-06, + "loss": 1.5184, + "step": 502 + }, + { + "epoch": 0.04, + "learning_rate": 9.410664172123481e-06, + "loss": 1.3548, + "step": 503 + }, + { + "epoch": 0.04, + "learning_rate": 9.429373246024322e-06, + "loss": 1.4396, + "step": 504 + }, + { + "epoch": 0.04, + "learning_rate": 9.448082319925165e-06, + "loss": 1.5201, + "step": 505 + }, + { + "epoch": 0.04, + "learning_rate": 9.466791393826006e-06, + "loss": 1.351, + "step": 506 + }, + { + "epoch": 0.04, + "learning_rate": 9.485500467726848e-06, + "loss": 1.4594, + "step": 507 + }, + { + "epoch": 0.04, + "learning_rate": 9.50420954162769e-06, + "loss": 1.4837, + "step": 508 + }, + { + "epoch": 0.04, + "learning_rate": 9.522918615528532e-06, + "loss": 1.3873, + "step": 509 + }, + { + "epoch": 0.04, + "learning_rate": 9.541627689429373e-06, + "loss": 1.4634, + "step": 510 + }, + { + "epoch": 0.04, + "learning_rate": 9.560336763330216e-06, + "loss": 1.5455, + "step": 511 + }, + { + "epoch": 0.04, + "learning_rate": 9.579045837231057e-06, + "loss": 1.5208, + "step": 512 + }, + { + "epoch": 0.04, + "learning_rate": 9.5977549111319e-06, + "loss": 1.5475, + "step": 513 + }, + { + "epoch": 0.04, + "learning_rate": 9.616463985032741e-06, + "loss": 1.5029, + "step": 514 + }, + { + "epoch": 0.04, + "learning_rate": 9.635173058933583e-06, + "loss": 1.5222, + "step": 515 + }, + { + "epoch": 0.04, + "learning_rate": 9.653882132834426e-06, + "loss": 1.4043, + "step": 516 + }, + { + "epoch": 0.04, + "learning_rate": 9.672591206735269e-06, + "loss": 1.4754, + "step": 517 + }, + { + "epoch": 0.04, + "learning_rate": 9.691300280636108e-06, + "loss": 1.3862, + "step": 518 + }, + { + "epoch": 0.04, + "learning_rate": 9.710009354536951e-06, + "loss": 1.4559, + "step": 519 + }, + { + "epoch": 0.04, + "learning_rate": 9.728718428437794e-06, + "loss": 1.4575, + "step": 520 + }, + { + "epoch": 0.04, + "learning_rate": 9.747427502338635e-06, + "loss": 1.4915, + "step": 521 + }, + { + "epoch": 0.04, + "learning_rate": 9.766136576239477e-06, + "loss": 1.4132, + "step": 522 + }, + { + "epoch": 0.04, + "learning_rate": 9.78484565014032e-06, + "loss": 1.4209, + "step": 523 + }, + { + "epoch": 0.04, + "learning_rate": 9.80355472404116e-06, + "loss": 1.5318, + "step": 524 + }, + { + "epoch": 0.04, + "learning_rate": 9.822263797942002e-06, + "loss": 1.4689, + "step": 525 + }, + { + "epoch": 0.04, + "learning_rate": 9.840972871842845e-06, + "loss": 1.536, + "step": 526 + }, + { + "epoch": 0.04, + "learning_rate": 9.859681945743686e-06, + "loss": 1.4468, + "step": 527 + }, + { + "epoch": 0.04, + "learning_rate": 9.87839101964453e-06, + "loss": 1.4094, + "step": 528 + }, + { + "epoch": 0.04, + "learning_rate": 9.89710009354537e-06, + "loss": 1.5186, + "step": 529 + }, + { + "epoch": 0.04, + "learning_rate": 9.915809167446212e-06, + "loss": 1.4964, + "step": 530 + }, + { + "epoch": 0.04, + "learning_rate": 9.934518241347055e-06, + "loss": 1.559, + "step": 531 + }, + { + "epoch": 0.04, + "learning_rate": 9.953227315247896e-06, + "loss": 1.4268, + "step": 532 + }, + { + "epoch": 0.04, + "learning_rate": 9.971936389148737e-06, + "loss": 1.4289, + "step": 533 + }, + { + "epoch": 0.04, + "learning_rate": 9.99064546304958e-06, + "loss": 1.4725, + "step": 534 + }, + { + "epoch": 0.05, + "learning_rate": 1.0009354536950421e-05, + "loss": 1.3898, + "step": 535 + }, + { + "epoch": 0.05, + "learning_rate": 1.0028063610851264e-05, + "loss": 1.3263, + "step": 536 + }, + { + "epoch": 0.05, + "learning_rate": 1.0046772684752106e-05, + "loss": 1.5003, + "step": 537 + }, + { + "epoch": 0.05, + "learning_rate": 1.0065481758652949e-05, + "loss": 1.4377, + "step": 538 + }, + { + "epoch": 0.05, + "learning_rate": 1.0084190832553788e-05, + "loss": 1.3887, + "step": 539 + }, + { + "epoch": 0.05, + "learning_rate": 1.0102899906454631e-05, + "loss": 1.4585, + "step": 540 + }, + { + "epoch": 0.05, + "learning_rate": 1.0121608980355472e-05, + "loss": 1.4611, + "step": 541 + }, + { + "epoch": 0.05, + "learning_rate": 1.0140318054256315e-05, + "loss": 1.4042, + "step": 542 + }, + { + "epoch": 0.05, + "learning_rate": 1.0159027128157157e-05, + "loss": 1.5036, + "step": 543 + }, + { + "epoch": 0.05, + "learning_rate": 1.0177736202058e-05, + "loss": 1.4366, + "step": 544 + }, + { + "epoch": 0.05, + "learning_rate": 1.0196445275958842e-05, + "loss": 1.5978, + "step": 545 + }, + { + "epoch": 0.05, + "learning_rate": 1.0215154349859682e-05, + "loss": 1.3662, + "step": 546 + }, + { + "epoch": 0.05, + "learning_rate": 1.0233863423760523e-05, + "loss": 1.4343, + "step": 547 + }, + { + "epoch": 0.05, + "learning_rate": 1.0252572497661366e-05, + "loss": 1.4057, + "step": 548 + }, + { + "epoch": 0.05, + "learning_rate": 1.0271281571562208e-05, + "loss": 1.369, + "step": 549 + }, + { + "epoch": 0.05, + "learning_rate": 1.028999064546305e-05, + "loss": 1.4121, + "step": 550 + }, + { + "epoch": 0.05, + "learning_rate": 1.0308699719363893e-05, + "loss": 1.5102, + "step": 551 + }, + { + "epoch": 0.05, + "learning_rate": 1.0327408793264735e-05, + "loss": 1.4511, + "step": 552 + }, + { + "epoch": 0.05, + "learning_rate": 1.0346117867165578e-05, + "loss": 1.4536, + "step": 553 + }, + { + "epoch": 0.05, + "learning_rate": 1.0364826941066417e-05, + "loss": 1.4403, + "step": 554 + }, + { + "epoch": 0.05, + "learning_rate": 1.0383536014967258e-05, + "loss": 1.4287, + "step": 555 + }, + { + "epoch": 0.05, + "learning_rate": 1.0402245088868101e-05, + "loss": 1.5551, + "step": 556 + }, + { + "epoch": 0.05, + "learning_rate": 1.0420954162768944e-05, + "loss": 1.3888, + "step": 557 + }, + { + "epoch": 0.05, + "learning_rate": 1.0439663236669786e-05, + "loss": 1.353, + "step": 558 + }, + { + "epoch": 0.05, + "learning_rate": 1.0458372310570629e-05, + "loss": 1.4208, + "step": 559 + }, + { + "epoch": 0.05, + "learning_rate": 1.047708138447147e-05, + "loss": 1.3885, + "step": 560 + }, + { + "epoch": 0.05, + "learning_rate": 1.0495790458372313e-05, + "loss": 1.4301, + "step": 561 + }, + { + "epoch": 0.05, + "learning_rate": 1.0514499532273152e-05, + "loss": 1.5172, + "step": 562 + }, + { + "epoch": 0.05, + "learning_rate": 1.0533208606173995e-05, + "loss": 1.4758, + "step": 563 + }, + { + "epoch": 0.05, + "learning_rate": 1.0551917680074837e-05, + "loss": 1.4015, + "step": 564 + }, + { + "epoch": 0.05, + "learning_rate": 1.057062675397568e-05, + "loss": 1.4451, + "step": 565 + }, + { + "epoch": 0.05, + "learning_rate": 1.058933582787652e-05, + "loss": 1.4223, + "step": 566 + }, + { + "epoch": 0.05, + "learning_rate": 1.0608044901777364e-05, + "loss": 1.4249, + "step": 567 + }, + { + "epoch": 0.05, + "learning_rate": 1.0626753975678205e-05, + "loss": 1.4983, + "step": 568 + }, + { + "epoch": 0.05, + "learning_rate": 1.0645463049579046e-05, + "loss": 1.3257, + "step": 569 + }, + { + "epoch": 0.05, + "learning_rate": 1.0664172123479887e-05, + "loss": 1.5196, + "step": 570 + }, + { + "epoch": 0.05, + "learning_rate": 1.068288119738073e-05, + "loss": 1.4347, + "step": 571 + }, + { + "epoch": 0.05, + "learning_rate": 1.0701590271281572e-05, + "loss": 1.4758, + "step": 572 + }, + { + "epoch": 0.05, + "learning_rate": 1.0720299345182415e-05, + "loss": 1.4692, + "step": 573 + }, + { + "epoch": 0.05, + "learning_rate": 1.0739008419083256e-05, + "loss": 1.5244, + "step": 574 + }, + { + "epoch": 0.05, + "learning_rate": 1.0757717492984099e-05, + "loss": 1.4468, + "step": 575 + }, + { + "epoch": 0.05, + "learning_rate": 1.0776426566884942e-05, + "loss": 1.3983, + "step": 576 + }, + { + "epoch": 0.05, + "learning_rate": 1.0795135640785781e-05, + "loss": 1.4863, + "step": 577 + }, + { + "epoch": 0.05, + "learning_rate": 1.0813844714686623e-05, + "loss": 1.5132, + "step": 578 + }, + { + "epoch": 0.05, + "learning_rate": 1.0832553788587466e-05, + "loss": 1.58, + "step": 579 + }, + { + "epoch": 0.05, + "learning_rate": 1.0851262862488307e-05, + "loss": 1.4145, + "step": 580 + }, + { + "epoch": 0.05, + "learning_rate": 1.086997193638915e-05, + "loss": 1.438, + "step": 581 + }, + { + "epoch": 0.05, + "learning_rate": 1.0888681010289993e-05, + "loss": 1.4144, + "step": 582 + }, + { + "epoch": 0.05, + "learning_rate": 1.0907390084190834e-05, + "loss": 1.5176, + "step": 583 + }, + { + "epoch": 0.05, + "learning_rate": 1.0926099158091677e-05, + "loss": 1.4347, + "step": 584 + }, + { + "epoch": 0.05, + "learning_rate": 1.0944808231992516e-05, + "loss": 1.4212, + "step": 585 + }, + { + "epoch": 0.05, + "learning_rate": 1.096351730589336e-05, + "loss": 1.4445, + "step": 586 + }, + { + "epoch": 0.05, + "learning_rate": 1.09822263797942e-05, + "loss": 1.4241, + "step": 587 + }, + { + "epoch": 0.05, + "learning_rate": 1.1000935453695044e-05, + "loss": 1.4726, + "step": 588 + }, + { + "epoch": 0.05, + "learning_rate": 1.1019644527595885e-05, + "loss": 1.4103, + "step": 589 + }, + { + "epoch": 0.05, + "learning_rate": 1.1038353601496728e-05, + "loss": 1.4385, + "step": 590 + }, + { + "epoch": 0.05, + "learning_rate": 1.1057062675397569e-05, + "loss": 1.4941, + "step": 591 + }, + { + "epoch": 0.05, + "learning_rate": 1.107577174929841e-05, + "loss": 1.4405, + "step": 592 + }, + { + "epoch": 0.05, + "learning_rate": 1.1094480823199252e-05, + "loss": 1.6119, + "step": 593 + }, + { + "epoch": 0.05, + "learning_rate": 1.1113189897100095e-05, + "loss": 1.3589, + "step": 594 + }, + { + "epoch": 0.05, + "learning_rate": 1.1131898971000936e-05, + "loss": 1.4254, + "step": 595 + }, + { + "epoch": 0.05, + "learning_rate": 1.1150608044901779e-05, + "loss": 1.4098, + "step": 596 + }, + { + "epoch": 0.05, + "learning_rate": 1.116931711880262e-05, + "loss": 1.4624, + "step": 597 + }, + { + "epoch": 0.05, + "learning_rate": 1.1188026192703463e-05, + "loss": 1.5239, + "step": 598 + }, + { + "epoch": 0.05, + "learning_rate": 1.1206735266604304e-05, + "loss": 1.5085, + "step": 599 + }, + { + "epoch": 0.05, + "learning_rate": 1.1225444340505145e-05, + "loss": 1.4969, + "step": 600 + }, + { + "epoch": 0.05, + "learning_rate": 1.1244153414405987e-05, + "loss": 1.39, + "step": 601 + }, + { + "epoch": 0.05, + "learning_rate": 1.126286248830683e-05, + "loss": 1.4591, + "step": 602 + }, + { + "epoch": 0.05, + "learning_rate": 1.1281571562207671e-05, + "loss": 1.5168, + "step": 603 + }, + { + "epoch": 0.05, + "learning_rate": 1.1300280636108514e-05, + "loss": 1.3915, + "step": 604 + }, + { + "epoch": 0.05, + "learning_rate": 1.1318989710009355e-05, + "loss": 1.5399, + "step": 605 + }, + { + "epoch": 0.05, + "learning_rate": 1.1337698783910198e-05, + "loss": 1.4786, + "step": 606 + }, + { + "epoch": 0.05, + "learning_rate": 1.1356407857811041e-05, + "loss": 1.4251, + "step": 607 + }, + { + "epoch": 0.05, + "learning_rate": 1.137511693171188e-05, + "loss": 1.373, + "step": 608 + }, + { + "epoch": 0.05, + "learning_rate": 1.1393826005612722e-05, + "loss": 1.4628, + "step": 609 + }, + { + "epoch": 0.05, + "learning_rate": 1.1412535079513565e-05, + "loss": 1.5035, + "step": 610 + }, + { + "epoch": 0.05, + "learning_rate": 1.1431244153414408e-05, + "loss": 1.3237, + "step": 611 + }, + { + "epoch": 0.05, + "learning_rate": 1.1449953227315249e-05, + "loss": 1.4391, + "step": 612 + }, + { + "epoch": 0.05, + "learning_rate": 1.1468662301216092e-05, + "loss": 1.3418, + "step": 613 + }, + { + "epoch": 0.05, + "learning_rate": 1.1487371375116933e-05, + "loss": 1.3128, + "step": 614 + }, + { + "epoch": 0.05, + "learning_rate": 1.1506080449017773e-05, + "loss": 1.4412, + "step": 615 + }, + { + "epoch": 0.05, + "learning_rate": 1.1524789522918616e-05, + "loss": 1.4876, + "step": 616 + }, + { + "epoch": 0.05, + "learning_rate": 1.1543498596819459e-05, + "loss": 1.4746, + "step": 617 + }, + { + "epoch": 0.05, + "learning_rate": 1.15622076707203e-05, + "loss": 1.4286, + "step": 618 + }, + { + "epoch": 0.05, + "learning_rate": 1.1580916744621143e-05, + "loss": 1.5369, + "step": 619 + }, + { + "epoch": 0.05, + "learning_rate": 1.1599625818521984e-05, + "loss": 1.4169, + "step": 620 + }, + { + "epoch": 0.05, + "learning_rate": 1.1618334892422827e-05, + "loss": 1.398, + "step": 621 + }, + { + "epoch": 0.05, + "learning_rate": 1.1637043966323668e-05, + "loss": 1.4566, + "step": 622 + }, + { + "epoch": 0.05, + "learning_rate": 1.165575304022451e-05, + "loss": 1.3645, + "step": 623 + }, + { + "epoch": 0.05, + "learning_rate": 1.1674462114125351e-05, + "loss": 1.3992, + "step": 624 + }, + { + "epoch": 0.05, + "learning_rate": 1.1693171188026194e-05, + "loss": 1.464, + "step": 625 + }, + { + "epoch": 0.05, + "learning_rate": 1.1711880261927035e-05, + "loss": 1.4973, + "step": 626 + }, + { + "epoch": 0.05, + "learning_rate": 1.1730589335827878e-05, + "loss": 1.4297, + "step": 627 + }, + { + "epoch": 0.05, + "learning_rate": 1.174929840972872e-05, + "loss": 1.4683, + "step": 628 + }, + { + "epoch": 0.05, + "learning_rate": 1.1768007483629562e-05, + "loss": 1.5133, + "step": 629 + }, + { + "epoch": 0.05, + "learning_rate": 1.1786716557530402e-05, + "loss": 1.4797, + "step": 630 + }, + { + "epoch": 0.05, + "learning_rate": 1.1805425631431245e-05, + "loss": 1.4955, + "step": 631 + }, + { + "epoch": 0.05, + "learning_rate": 1.1824134705332086e-05, + "loss": 1.5524, + "step": 632 + }, + { + "epoch": 0.05, + "learning_rate": 1.1842843779232929e-05, + "loss": 1.46, + "step": 633 + }, + { + "epoch": 0.05, + "learning_rate": 1.186155285313377e-05, + "loss": 1.479, + "step": 634 + }, + { + "epoch": 0.05, + "learning_rate": 1.1880261927034613e-05, + "loss": 1.4696, + "step": 635 + }, + { + "epoch": 0.05, + "learning_rate": 1.1898971000935456e-05, + "loss": 1.5472, + "step": 636 + }, + { + "epoch": 0.05, + "learning_rate": 1.1917680074836297e-05, + "loss": 1.4954, + "step": 637 + }, + { + "epoch": 0.05, + "learning_rate": 1.1936389148737137e-05, + "loss": 1.4195, + "step": 638 + }, + { + "epoch": 0.05, + "learning_rate": 1.195509822263798e-05, + "loss": 1.5197, + "step": 639 + }, + { + "epoch": 0.05, + "learning_rate": 1.1973807296538821e-05, + "loss": 1.3601, + "step": 640 + }, + { + "epoch": 0.05, + "learning_rate": 1.1992516370439664e-05, + "loss": 1.2924, + "step": 641 + }, + { + "epoch": 0.05, + "learning_rate": 1.2011225444340507e-05, + "loss": 1.4647, + "step": 642 + }, + { + "epoch": 0.05, + "learning_rate": 1.2029934518241348e-05, + "loss": 1.3746, + "step": 643 + }, + { + "epoch": 0.05, + "learning_rate": 1.2048643592142191e-05, + "loss": 1.46, + "step": 644 + }, + { + "epoch": 0.05, + "learning_rate": 1.2067352666043033e-05, + "loss": 1.4221, + "step": 645 + }, + { + "epoch": 0.05, + "learning_rate": 1.2086061739943872e-05, + "loss": 1.531, + "step": 646 + }, + { + "epoch": 0.05, + "learning_rate": 1.2104770813844715e-05, + "loss": 1.5153, + "step": 647 + }, + { + "epoch": 0.05, + "learning_rate": 1.2123479887745558e-05, + "loss": 1.4908, + "step": 648 + }, + { + "epoch": 0.05, + "learning_rate": 1.21421889616464e-05, + "loss": 1.4628, + "step": 649 + }, + { + "epoch": 0.05, + "learning_rate": 1.2160898035547242e-05, + "loss": 1.4631, + "step": 650 + }, + { + "epoch": 0.05, + "learning_rate": 1.2179607109448083e-05, + "loss": 1.3402, + "step": 651 + }, + { + "epoch": 0.05, + "learning_rate": 1.2198316183348926e-05, + "loss": 1.4829, + "step": 652 + }, + { + "epoch": 0.05, + "learning_rate": 1.2217025257249766e-05, + "loss": 1.5214, + "step": 653 + }, + { + "epoch": 0.06, + "learning_rate": 1.2235734331150609e-05, + "loss": 1.4751, + "step": 654 + }, + { + "epoch": 0.06, + "learning_rate": 1.225444340505145e-05, + "loss": 1.4657, + "step": 655 + }, + { + "epoch": 0.06, + "learning_rate": 1.2273152478952293e-05, + "loss": 1.3718, + "step": 656 + }, + { + "epoch": 0.06, + "learning_rate": 1.2291861552853134e-05, + "loss": 1.4539, + "step": 657 + }, + { + "epoch": 0.06, + "learning_rate": 1.2310570626753977e-05, + "loss": 1.3987, + "step": 658 + }, + { + "epoch": 0.06, + "learning_rate": 1.2329279700654819e-05, + "loss": 1.3914, + "step": 659 + }, + { + "epoch": 0.06, + "learning_rate": 1.2347988774555662e-05, + "loss": 1.4116, + "step": 660 + }, + { + "epoch": 0.06, + "learning_rate": 1.2366697848456501e-05, + "loss": 1.3755, + "step": 661 + }, + { + "epoch": 0.06, + "learning_rate": 1.2385406922357344e-05, + "loss": 1.5418, + "step": 662 + }, + { + "epoch": 0.06, + "learning_rate": 1.2404115996258185e-05, + "loss": 1.469, + "step": 663 + }, + { + "epoch": 0.06, + "learning_rate": 1.2422825070159028e-05, + "loss": 1.3877, + "step": 664 + }, + { + "epoch": 0.06, + "learning_rate": 1.244153414405987e-05, + "loss": 1.4396, + "step": 665 + }, + { + "epoch": 0.06, + "learning_rate": 1.2460243217960713e-05, + "loss": 1.4897, + "step": 666 + }, + { + "epoch": 0.06, + "learning_rate": 1.2478952291861555e-05, + "loss": 1.4454, + "step": 667 + }, + { + "epoch": 0.06, + "learning_rate": 1.2497661365762397e-05, + "loss": 1.4648, + "step": 668 + }, + { + "epoch": 0.06, + "learning_rate": 1.2516370439663236e-05, + "loss": 1.4412, + "step": 669 + }, + { + "epoch": 0.06, + "learning_rate": 1.253507951356408e-05, + "loss": 1.4813, + "step": 670 + }, + { + "epoch": 0.06, + "learning_rate": 1.255378858746492e-05, + "loss": 1.4197, + "step": 671 + }, + { + "epoch": 0.06, + "learning_rate": 1.2572497661365763e-05, + "loss": 1.4247, + "step": 672 + }, + { + "epoch": 0.06, + "learning_rate": 1.2591206735266606e-05, + "loss": 1.4114, + "step": 673 + }, + { + "epoch": 0.06, + "learning_rate": 1.2609915809167448e-05, + "loss": 1.4416, + "step": 674 + }, + { + "epoch": 0.06, + "learning_rate": 1.262862488306829e-05, + "loss": 1.425, + "step": 675 + }, + { + "epoch": 0.06, + "learning_rate": 1.264733395696913e-05, + "loss": 1.4179, + "step": 676 + }, + { + "epoch": 0.06, + "learning_rate": 1.2666043030869971e-05, + "loss": 1.4003, + "step": 677 + }, + { + "epoch": 0.06, + "learning_rate": 1.2684752104770814e-05, + "loss": 1.4333, + "step": 678 + }, + { + "epoch": 0.06, + "learning_rate": 1.2703461178671657e-05, + "loss": 1.4453, + "step": 679 + }, + { + "epoch": 0.06, + "learning_rate": 1.2722170252572499e-05, + "loss": 1.5393, + "step": 680 + }, + { + "epoch": 0.06, + "learning_rate": 1.2740879326473342e-05, + "loss": 1.4356, + "step": 681 + }, + { + "epoch": 0.06, + "learning_rate": 1.2759588400374183e-05, + "loss": 1.4039, + "step": 682 + }, + { + "epoch": 0.06, + "learning_rate": 1.2778297474275026e-05, + "loss": 1.4489, + "step": 683 + }, + { + "epoch": 0.06, + "learning_rate": 1.2797006548175865e-05, + "loss": 1.5006, + "step": 684 + }, + { + "epoch": 0.06, + "learning_rate": 1.2815715622076708e-05, + "loss": 1.4543, + "step": 685 + }, + { + "epoch": 0.06, + "learning_rate": 1.283442469597755e-05, + "loss": 1.4154, + "step": 686 + }, + { + "epoch": 0.06, + "learning_rate": 1.2853133769878392e-05, + "loss": 1.4465, + "step": 687 + }, + { + "epoch": 0.06, + "learning_rate": 1.2871842843779234e-05, + "loss": 1.4899, + "step": 688 + }, + { + "epoch": 0.06, + "learning_rate": 1.2890551917680077e-05, + "loss": 1.4878, + "step": 689 + }, + { + "epoch": 0.06, + "learning_rate": 1.2909260991580918e-05, + "loss": 1.5354, + "step": 690 + }, + { + "epoch": 0.06, + "learning_rate": 1.2927970065481761e-05, + "loss": 1.4291, + "step": 691 + }, + { + "epoch": 0.06, + "learning_rate": 1.29466791393826e-05, + "loss": 1.473, + "step": 692 + }, + { + "epoch": 0.06, + "learning_rate": 1.2965388213283443e-05, + "loss": 1.5014, + "step": 693 + }, + { + "epoch": 0.06, + "learning_rate": 1.2984097287184285e-05, + "loss": 1.4313, + "step": 694 + }, + { + "epoch": 0.06, + "learning_rate": 1.3002806361085128e-05, + "loss": 1.4369, + "step": 695 + }, + { + "epoch": 0.06, + "learning_rate": 1.3021515434985969e-05, + "loss": 1.4544, + "step": 696 + }, + { + "epoch": 0.06, + "learning_rate": 1.3040224508886812e-05, + "loss": 1.4723, + "step": 697 + }, + { + "epoch": 0.06, + "learning_rate": 1.3058933582787655e-05, + "loss": 1.4637, + "step": 698 + }, + { + "epoch": 0.06, + "learning_rate": 1.3077642656688494e-05, + "loss": 1.3888, + "step": 699 + }, + { + "epoch": 0.06, + "learning_rate": 1.3096351730589336e-05, + "loss": 1.3778, + "step": 700 + }, + { + "epoch": 0.06, + "learning_rate": 1.3115060804490179e-05, + "loss": 1.4276, + "step": 701 + }, + { + "epoch": 0.06, + "learning_rate": 1.313376987839102e-05, + "loss": 1.48, + "step": 702 + }, + { + "epoch": 0.06, + "learning_rate": 1.3152478952291863e-05, + "loss": 1.5253, + "step": 703 + }, + { + "epoch": 0.06, + "learning_rate": 1.3171188026192706e-05, + "loss": 1.4438, + "step": 704 + }, + { + "epoch": 0.06, + "learning_rate": 1.3189897100093547e-05, + "loss": 1.4416, + "step": 705 + }, + { + "epoch": 0.06, + "learning_rate": 1.320860617399439e-05, + "loss": 1.5126, + "step": 706 + }, + { + "epoch": 0.06, + "learning_rate": 1.322731524789523e-05, + "loss": 1.4114, + "step": 707 + }, + { + "epoch": 0.06, + "learning_rate": 1.3246024321796072e-05, + "loss": 1.4789, + "step": 708 + }, + { + "epoch": 0.06, + "learning_rate": 1.3264733395696914e-05, + "loss": 1.4607, + "step": 709 + }, + { + "epoch": 0.06, + "learning_rate": 1.3283442469597757e-05, + "loss": 1.4348, + "step": 710 + }, + { + "epoch": 0.06, + "learning_rate": 1.3302151543498598e-05, + "loss": 1.3789, + "step": 711 + }, + { + "epoch": 0.06, + "learning_rate": 1.332086061739944e-05, + "loss": 1.4375, + "step": 712 + }, + { + "epoch": 0.06, + "learning_rate": 1.3339569691300282e-05, + "loss": 1.5054, + "step": 713 + }, + { + "epoch": 0.06, + "learning_rate": 1.3358278765201123e-05, + "loss": 1.4495, + "step": 714 + }, + { + "epoch": 0.06, + "learning_rate": 1.3376987839101965e-05, + "loss": 1.4142, + "step": 715 + }, + { + "epoch": 0.06, + "learning_rate": 1.3395696913002808e-05, + "loss": 1.4882, + "step": 716 + }, + { + "epoch": 0.06, + "learning_rate": 1.3414405986903649e-05, + "loss": 1.4114, + "step": 717 + }, + { + "epoch": 0.06, + "learning_rate": 1.3433115060804492e-05, + "loss": 1.3582, + "step": 718 + }, + { + "epoch": 0.06, + "learning_rate": 1.3451824134705333e-05, + "loss": 1.4882, + "step": 719 + }, + { + "epoch": 0.06, + "learning_rate": 1.3470533208606176e-05, + "loss": 1.4436, + "step": 720 + }, + { + "epoch": 0.06, + "learning_rate": 1.3489242282507017e-05, + "loss": 1.4077, + "step": 721 + }, + { + "epoch": 0.06, + "learning_rate": 1.3507951356407858e-05, + "loss": 1.4301, + "step": 722 + }, + { + "epoch": 0.06, + "learning_rate": 1.35266604303087e-05, + "loss": 1.4376, + "step": 723 + }, + { + "epoch": 0.06, + "learning_rate": 1.3545369504209543e-05, + "loss": 1.431, + "step": 724 + }, + { + "epoch": 0.06, + "learning_rate": 1.3564078578110384e-05, + "loss": 1.4539, + "step": 725 + }, + { + "epoch": 0.06, + "learning_rate": 1.3582787652011227e-05, + "loss": 1.3592, + "step": 726 + }, + { + "epoch": 0.06, + "learning_rate": 1.3601496725912068e-05, + "loss": 1.3695, + "step": 727 + }, + { + "epoch": 0.06, + "learning_rate": 1.3620205799812911e-05, + "loss": 1.4988, + "step": 728 + }, + { + "epoch": 0.06, + "learning_rate": 1.3638914873713754e-05, + "loss": 1.3783, + "step": 729 + }, + { + "epoch": 0.06, + "learning_rate": 1.3657623947614594e-05, + "loss": 1.3999, + "step": 730 + }, + { + "epoch": 0.06, + "learning_rate": 1.3676333021515435e-05, + "loss": 1.3575, + "step": 731 + }, + { + "epoch": 0.06, + "learning_rate": 1.3695042095416278e-05, + "loss": 1.412, + "step": 732 + }, + { + "epoch": 0.06, + "learning_rate": 1.3713751169317119e-05, + "loss": 1.4061, + "step": 733 + }, + { + "epoch": 0.06, + "learning_rate": 1.3732460243217962e-05, + "loss": 1.3851, + "step": 734 + }, + { + "epoch": 0.06, + "learning_rate": 1.3751169317118805e-05, + "loss": 1.5606, + "step": 735 + }, + { + "epoch": 0.06, + "learning_rate": 1.3769878391019646e-05, + "loss": 1.4219, + "step": 736 + }, + { + "epoch": 0.06, + "learning_rate": 1.3788587464920486e-05, + "loss": 1.3523, + "step": 737 + }, + { + "epoch": 0.06, + "learning_rate": 1.3807296538821329e-05, + "loss": 1.4507, + "step": 738 + }, + { + "epoch": 0.06, + "learning_rate": 1.3826005612722172e-05, + "loss": 1.4444, + "step": 739 + }, + { + "epoch": 0.06, + "learning_rate": 1.3844714686623013e-05, + "loss": 1.431, + "step": 740 + }, + { + "epoch": 0.06, + "learning_rate": 1.3863423760523856e-05, + "loss": 1.4784, + "step": 741 + }, + { + "epoch": 0.06, + "learning_rate": 1.3882132834424697e-05, + "loss": 1.4765, + "step": 742 + }, + { + "epoch": 0.06, + "learning_rate": 1.390084190832554e-05, + "loss": 1.3307, + "step": 743 + }, + { + "epoch": 0.06, + "learning_rate": 1.3919550982226381e-05, + "loss": 1.5084, + "step": 744 + }, + { + "epoch": 0.06, + "learning_rate": 1.3938260056127223e-05, + "loss": 1.4878, + "step": 745 + }, + { + "epoch": 0.06, + "learning_rate": 1.3956969130028064e-05, + "loss": 1.4508, + "step": 746 + }, + { + "epoch": 0.06, + "learning_rate": 1.3975678203928907e-05, + "loss": 1.3984, + "step": 747 + }, + { + "epoch": 0.06, + "learning_rate": 1.3994387277829748e-05, + "loss": 1.4168, + "step": 748 + }, + { + "epoch": 0.06, + "learning_rate": 1.4013096351730591e-05, + "loss": 1.4717, + "step": 749 + }, + { + "epoch": 0.06, + "learning_rate": 1.4031805425631432e-05, + "loss": 1.4627, + "step": 750 + }, + { + "epoch": 0.06, + "learning_rate": 1.4050514499532275e-05, + "loss": 1.4757, + "step": 751 + }, + { + "epoch": 0.06, + "learning_rate": 1.4069223573433117e-05, + "loss": 1.4004, + "step": 752 + }, + { + "epoch": 0.06, + "learning_rate": 1.4087932647333958e-05, + "loss": 1.339, + "step": 753 + }, + { + "epoch": 0.06, + "learning_rate": 1.4106641721234799e-05, + "loss": 1.2957, + "step": 754 + }, + { + "epoch": 0.06, + "learning_rate": 1.4125350795135642e-05, + "loss": 1.3658, + "step": 755 + }, + { + "epoch": 0.06, + "learning_rate": 1.4144059869036483e-05, + "loss": 1.4363, + "step": 756 + }, + { + "epoch": 0.06, + "learning_rate": 1.4162768942937326e-05, + "loss": 1.4905, + "step": 757 + }, + { + "epoch": 0.06, + "learning_rate": 1.4181478016838167e-05, + "loss": 1.4455, + "step": 758 + }, + { + "epoch": 0.06, + "learning_rate": 1.420018709073901e-05, + "loss": 1.521, + "step": 759 + }, + { + "epoch": 0.06, + "learning_rate": 1.421889616463985e-05, + "loss": 1.3712, + "step": 760 + }, + { + "epoch": 0.06, + "learning_rate": 1.4237605238540693e-05, + "loss": 1.4405, + "step": 761 + }, + { + "epoch": 0.06, + "learning_rate": 1.4256314312441534e-05, + "loss": 1.4568, + "step": 762 + }, + { + "epoch": 0.06, + "learning_rate": 1.4275023386342377e-05, + "loss": 1.4568, + "step": 763 + }, + { + "epoch": 0.06, + "learning_rate": 1.429373246024322e-05, + "loss": 1.4516, + "step": 764 + }, + { + "epoch": 0.06, + "learning_rate": 1.4312441534144061e-05, + "loss": 1.5685, + "step": 765 + }, + { + "epoch": 0.06, + "learning_rate": 1.4331150608044904e-05, + "loss": 1.4178, + "step": 766 + }, + { + "epoch": 0.06, + "learning_rate": 1.4349859681945746e-05, + "loss": 1.4899, + "step": 767 + }, + { + "epoch": 0.06, + "learning_rate": 1.4368568755846585e-05, + "loss": 1.4394, + "step": 768 + }, + { + "epoch": 0.06, + "learning_rate": 1.4387277829747428e-05, + "loss": 1.3934, + "step": 769 + }, + { + "epoch": 0.06, + "learning_rate": 1.4405986903648271e-05, + "loss": 1.3985, + "step": 770 + }, + { + "epoch": 0.06, + "learning_rate": 1.4424695977549112e-05, + "loss": 1.4525, + "step": 771 + }, + { + "epoch": 0.06, + "learning_rate": 1.4443405051449955e-05, + "loss": 1.3891, + "step": 772 + }, + { + "epoch": 0.07, + "learning_rate": 1.4462114125350796e-05, + "loss": 1.567, + "step": 773 + }, + { + "epoch": 0.07, + "learning_rate": 1.448082319925164e-05, + "loss": 1.4196, + "step": 774 + }, + { + "epoch": 0.07, + "learning_rate": 1.449953227315248e-05, + "loss": 1.36, + "step": 775 + }, + { + "epoch": 0.07, + "learning_rate": 1.4518241347053322e-05, + "loss": 1.3919, + "step": 776 + }, + { + "epoch": 0.07, + "learning_rate": 1.4536950420954163e-05, + "loss": 1.4252, + "step": 777 + }, + { + "epoch": 0.07, + "learning_rate": 1.4555659494855006e-05, + "loss": 1.3608, + "step": 778 + }, + { + "epoch": 0.07, + "learning_rate": 1.4574368568755847e-05, + "loss": 1.3656, + "step": 779 + }, + { + "epoch": 0.07, + "learning_rate": 1.459307764265669e-05, + "loss": 1.3131, + "step": 780 + }, + { + "epoch": 0.07, + "learning_rate": 1.4611786716557532e-05, + "loss": 1.4542, + "step": 781 + }, + { + "epoch": 0.07, + "learning_rate": 1.4630495790458375e-05, + "loss": 1.4932, + "step": 782 + }, + { + "epoch": 0.07, + "learning_rate": 1.4649204864359214e-05, + "loss": 1.4899, + "step": 783 + }, + { + "epoch": 0.07, + "learning_rate": 1.4667913938260057e-05, + "loss": 1.3366, + "step": 784 + }, + { + "epoch": 0.07, + "learning_rate": 1.4686623012160898e-05, + "loss": 1.4071, + "step": 785 + }, + { + "epoch": 0.07, + "learning_rate": 1.4705332086061741e-05, + "loss": 1.4389, + "step": 786 + }, + { + "epoch": 0.07, + "learning_rate": 1.4724041159962583e-05, + "loss": 1.4643, + "step": 787 + }, + { + "epoch": 0.07, + "learning_rate": 1.4742750233863425e-05, + "loss": 1.4701, + "step": 788 + }, + { + "epoch": 0.07, + "learning_rate": 1.4761459307764268e-05, + "loss": 1.4713, + "step": 789 + }, + { + "epoch": 0.07, + "learning_rate": 1.478016838166511e-05, + "loss": 1.4727, + "step": 790 + }, + { + "epoch": 0.07, + "learning_rate": 1.479887745556595e-05, + "loss": 1.4456, + "step": 791 + }, + { + "epoch": 0.07, + "learning_rate": 1.4817586529466792e-05, + "loss": 1.4418, + "step": 792 + }, + { + "epoch": 0.07, + "learning_rate": 1.4836295603367633e-05, + "loss": 1.4232, + "step": 793 + }, + { + "epoch": 0.07, + "learning_rate": 1.4855004677268476e-05, + "loss": 1.5144, + "step": 794 + }, + { + "epoch": 0.07, + "learning_rate": 1.487371375116932e-05, + "loss": 1.2849, + "step": 795 + }, + { + "epoch": 0.07, + "learning_rate": 1.489242282507016e-05, + "loss": 1.4709, + "step": 796 + }, + { + "epoch": 0.07, + "learning_rate": 1.4911131898971004e-05, + "loss": 1.4388, + "step": 797 + }, + { + "epoch": 0.07, + "learning_rate": 1.4929840972871843e-05, + "loss": 1.4128, + "step": 798 + }, + { + "epoch": 0.07, + "learning_rate": 1.4948550046772684e-05, + "loss": 1.3469, + "step": 799 + }, + { + "epoch": 0.07, + "learning_rate": 1.4967259120673527e-05, + "loss": 1.4291, + "step": 800 + }, + { + "epoch": 0.07, + "learning_rate": 1.498596819457437e-05, + "loss": 1.3717, + "step": 801 + }, + { + "epoch": 0.07, + "learning_rate": 1.5004677268475212e-05, + "loss": 1.48, + "step": 802 + }, + { + "epoch": 0.07, + "learning_rate": 1.5023386342376054e-05, + "loss": 1.3294, + "step": 803 + }, + { + "epoch": 0.07, + "learning_rate": 1.5042095416276896e-05, + "loss": 1.3621, + "step": 804 + }, + { + "epoch": 0.07, + "learning_rate": 1.5060804490177739e-05, + "loss": 1.4841, + "step": 805 + }, + { + "epoch": 0.07, + "learning_rate": 1.5079513564078578e-05, + "loss": 1.4252, + "step": 806 + }, + { + "epoch": 0.07, + "learning_rate": 1.5098222637979421e-05, + "loss": 1.483, + "step": 807 + }, + { + "epoch": 0.07, + "learning_rate": 1.5116931711880262e-05, + "loss": 1.3893, + "step": 808 + }, + { + "epoch": 0.07, + "learning_rate": 1.5135640785781105e-05, + "loss": 1.4011, + "step": 809 + }, + { + "epoch": 0.07, + "learning_rate": 1.5154349859681947e-05, + "loss": 1.3463, + "step": 810 + }, + { + "epoch": 0.07, + "learning_rate": 1.517305893358279e-05, + "loss": 1.4925, + "step": 811 + }, + { + "epoch": 0.07, + "learning_rate": 1.5191768007483631e-05, + "loss": 1.4968, + "step": 812 + }, + { + "epoch": 0.07, + "learning_rate": 1.5210477081384474e-05, + "loss": 1.3873, + "step": 813 + }, + { + "epoch": 0.07, + "learning_rate": 1.5229186155285313e-05, + "loss": 1.4178, + "step": 814 + }, + { + "epoch": 0.07, + "learning_rate": 1.5247895229186156e-05, + "loss": 1.5056, + "step": 815 + }, + { + "epoch": 0.07, + "learning_rate": 1.5266604303086998e-05, + "loss": 1.4813, + "step": 816 + }, + { + "epoch": 0.07, + "learning_rate": 1.528531337698784e-05, + "loss": 1.4556, + "step": 817 + }, + { + "epoch": 0.07, + "learning_rate": 1.5304022450888684e-05, + "loss": 1.4231, + "step": 818 + }, + { + "epoch": 0.07, + "learning_rate": 1.5322731524789525e-05, + "loss": 1.4594, + "step": 819 + }, + { + "epoch": 0.07, + "learning_rate": 1.5341440598690366e-05, + "loss": 1.4106, + "step": 820 + }, + { + "epoch": 0.07, + "learning_rate": 1.5360149672591207e-05, + "loss": 1.4192, + "step": 821 + }, + { + "epoch": 0.07, + "learning_rate": 1.537885874649205e-05, + "loss": 1.4533, + "step": 822 + }, + { + "epoch": 0.07, + "learning_rate": 1.539756782039289e-05, + "loss": 1.3866, + "step": 823 + }, + { + "epoch": 0.07, + "learning_rate": 1.5416276894293734e-05, + "loss": 1.5019, + "step": 824 + }, + { + "epoch": 0.07, + "learning_rate": 1.5434985968194576e-05, + "loss": 1.3881, + "step": 825 + }, + { + "epoch": 0.07, + "learning_rate": 1.5453695042095417e-05, + "loss": 1.3003, + "step": 826 + }, + { + "epoch": 0.07, + "learning_rate": 1.547240411599626e-05, + "loss": 1.4162, + "step": 827 + }, + { + "epoch": 0.07, + "learning_rate": 1.5491113189897103e-05, + "loss": 1.394, + "step": 828 + }, + { + "epoch": 0.07, + "learning_rate": 1.550982226379794e-05, + "loss": 1.4033, + "step": 829 + }, + { + "epoch": 0.07, + "learning_rate": 1.5528531337698785e-05, + "loss": 1.432, + "step": 830 + }, + { + "epoch": 0.07, + "learning_rate": 1.5547240411599627e-05, + "loss": 1.4312, + "step": 831 + }, + { + "epoch": 0.07, + "learning_rate": 1.5565949485500468e-05, + "loss": 1.3521, + "step": 832 + }, + { + "epoch": 0.07, + "learning_rate": 1.5584658559401313e-05, + "loss": 1.4603, + "step": 833 + }, + { + "epoch": 0.07, + "learning_rate": 1.5603367633302154e-05, + "loss": 1.4968, + "step": 834 + }, + { + "epoch": 0.07, + "learning_rate": 1.5622076707202995e-05, + "loss": 1.4009, + "step": 835 + }, + { + "epoch": 0.07, + "learning_rate": 1.5640785781103836e-05, + "loss": 1.4507, + "step": 836 + }, + { + "epoch": 0.07, + "learning_rate": 1.5659494855004678e-05, + "loss": 1.446, + "step": 837 + }, + { + "epoch": 0.07, + "learning_rate": 1.567820392890552e-05, + "loss": 1.5017, + "step": 838 + }, + { + "epoch": 0.07, + "learning_rate": 1.5696913002806363e-05, + "loss": 1.2885, + "step": 839 + }, + { + "epoch": 0.07, + "learning_rate": 1.5715622076707205e-05, + "loss": 1.3881, + "step": 840 + }, + { + "epoch": 0.07, + "learning_rate": 1.5734331150608046e-05, + "loss": 1.425, + "step": 841 + }, + { + "epoch": 0.07, + "learning_rate": 1.5753040224508887e-05, + "loss": 1.4138, + "step": 842 + }, + { + "epoch": 0.07, + "learning_rate": 1.5771749298409732e-05, + "loss": 1.3692, + "step": 843 + }, + { + "epoch": 0.07, + "learning_rate": 1.579045837231057e-05, + "loss": 1.3707, + "step": 844 + }, + { + "epoch": 0.07, + "learning_rate": 1.5809167446211414e-05, + "loss": 1.4023, + "step": 845 + }, + { + "epoch": 0.07, + "learning_rate": 1.5827876520112256e-05, + "loss": 1.4878, + "step": 846 + }, + { + "epoch": 0.07, + "learning_rate": 1.5846585594013097e-05, + "loss": 1.4506, + "step": 847 + }, + { + "epoch": 0.07, + "learning_rate": 1.5865294667913938e-05, + "loss": 1.4196, + "step": 848 + }, + { + "epoch": 0.07, + "learning_rate": 1.5884003741814783e-05, + "loss": 1.537, + "step": 849 + }, + { + "epoch": 0.07, + "learning_rate": 1.5902712815715624e-05, + "loss": 1.4858, + "step": 850 + }, + { + "epoch": 0.07, + "learning_rate": 1.5921421889616465e-05, + "loss": 1.3899, + "step": 851 + }, + { + "epoch": 0.07, + "learning_rate": 1.5940130963517307e-05, + "loss": 1.3482, + "step": 852 + }, + { + "epoch": 0.07, + "learning_rate": 1.5958840037418148e-05, + "loss": 1.5142, + "step": 853 + }, + { + "epoch": 0.07, + "learning_rate": 1.597754911131899e-05, + "loss": 1.3265, + "step": 854 + }, + { + "epoch": 0.07, + "learning_rate": 1.5996258185219834e-05, + "loss": 1.4382, + "step": 855 + }, + { + "epoch": 0.07, + "learning_rate": 1.6014967259120675e-05, + "loss": 1.4431, + "step": 856 + }, + { + "epoch": 0.07, + "learning_rate": 1.6033676333021516e-05, + "loss": 1.4781, + "step": 857 + }, + { + "epoch": 0.07, + "learning_rate": 1.605238540692236e-05, + "loss": 1.3619, + "step": 858 + }, + { + "epoch": 0.07, + "learning_rate": 1.6071094480823202e-05, + "loss": 1.4366, + "step": 859 + }, + { + "epoch": 0.07, + "learning_rate": 1.608980355472404e-05, + "loss": 1.3532, + "step": 860 + }, + { + "epoch": 0.07, + "learning_rate": 1.6108512628624885e-05, + "loss": 1.4998, + "step": 861 + }, + { + "epoch": 0.07, + "learning_rate": 1.6127221702525726e-05, + "loss": 1.3227, + "step": 862 + }, + { + "epoch": 0.07, + "learning_rate": 1.6145930776426567e-05, + "loss": 1.4746, + "step": 863 + }, + { + "epoch": 0.07, + "learning_rate": 1.6164639850327412e-05, + "loss": 1.4633, + "step": 864 + }, + { + "epoch": 0.07, + "learning_rate": 1.6183348924228253e-05, + "loss": 1.3409, + "step": 865 + }, + { + "epoch": 0.07, + "learning_rate": 1.6202057998129094e-05, + "loss": 1.3759, + "step": 866 + }, + { + "epoch": 0.07, + "learning_rate": 1.6220767072029936e-05, + "loss": 1.4614, + "step": 867 + }, + { + "epoch": 0.07, + "learning_rate": 1.6239476145930777e-05, + "loss": 1.3963, + "step": 868 + }, + { + "epoch": 0.07, + "learning_rate": 1.6258185219831618e-05, + "loss": 1.3688, + "step": 869 + }, + { + "epoch": 0.07, + "learning_rate": 1.6276894293732463e-05, + "loss": 1.424, + "step": 870 + }, + { + "epoch": 0.07, + "learning_rate": 1.6295603367633304e-05, + "loss": 1.4429, + "step": 871 + }, + { + "epoch": 0.07, + "learning_rate": 1.6314312441534145e-05, + "loss": 1.41, + "step": 872 + }, + { + "epoch": 0.07, + "learning_rate": 1.6333021515434987e-05, + "loss": 1.5298, + "step": 873 + }, + { + "epoch": 0.07, + "learning_rate": 1.635173058933583e-05, + "loss": 1.4756, + "step": 874 + }, + { + "epoch": 0.07, + "learning_rate": 1.637043966323667e-05, + "loss": 1.4967, + "step": 875 + }, + { + "epoch": 0.07, + "learning_rate": 1.6389148737137514e-05, + "loss": 1.3678, + "step": 876 + }, + { + "epoch": 0.07, + "learning_rate": 1.6407857811038355e-05, + "loss": 1.4258, + "step": 877 + }, + { + "epoch": 0.07, + "learning_rate": 1.6426566884939196e-05, + "loss": 1.407, + "step": 878 + }, + { + "epoch": 0.07, + "learning_rate": 1.6445275958840037e-05, + "loss": 1.4116, + "step": 879 + }, + { + "epoch": 0.07, + "learning_rate": 1.6463985032740882e-05, + "loss": 1.4041, + "step": 880 + }, + { + "epoch": 0.07, + "learning_rate": 1.6482694106641723e-05, + "loss": 1.4426, + "step": 881 + }, + { + "epoch": 0.07, + "learning_rate": 1.6501403180542565e-05, + "loss": 1.3653, + "step": 882 + }, + { + "epoch": 0.07, + "learning_rate": 1.6520112254443406e-05, + "loss": 1.3263, + "step": 883 + }, + { + "epoch": 0.07, + "learning_rate": 1.6538821328344247e-05, + "loss": 1.3689, + "step": 884 + }, + { + "epoch": 0.07, + "learning_rate": 1.655753040224509e-05, + "loss": 1.4315, + "step": 885 + }, + { + "epoch": 0.07, + "learning_rate": 1.6576239476145933e-05, + "loss": 1.3963, + "step": 886 + }, + { + "epoch": 0.07, + "learning_rate": 1.6594948550046774e-05, + "loss": 1.4817, + "step": 887 + }, + { + "epoch": 0.07, + "learning_rate": 1.6613657623947616e-05, + "loss": 1.3452, + "step": 888 + }, + { + "epoch": 0.07, + "learning_rate": 1.663236669784846e-05, + "loss": 1.4107, + "step": 889 + }, + { + "epoch": 0.07, + "learning_rate": 1.6651075771749298e-05, + "loss": 1.4652, + "step": 890 + }, + { + "epoch": 0.08, + "learning_rate": 1.666978484565014e-05, + "loss": 1.38, + "step": 891 + }, + { + "epoch": 0.08, + "learning_rate": 1.6688493919550984e-05, + "loss": 1.4181, + "step": 892 + }, + { + "epoch": 0.08, + "learning_rate": 1.6707202993451825e-05, + "loss": 1.3231, + "step": 893 + }, + { + "epoch": 0.08, + "learning_rate": 1.6725912067352666e-05, + "loss": 1.4004, + "step": 894 + }, + { + "epoch": 0.08, + "learning_rate": 1.674462114125351e-05, + "loss": 1.4102, + "step": 895 + }, + { + "epoch": 0.08, + "learning_rate": 1.6763330215154352e-05, + "loss": 1.404, + "step": 896 + }, + { + "epoch": 0.08, + "learning_rate": 1.6782039289055194e-05, + "loss": 1.2962, + "step": 897 + }, + { + "epoch": 0.08, + "learning_rate": 1.6800748362956035e-05, + "loss": 1.4669, + "step": 898 + }, + { + "epoch": 0.08, + "learning_rate": 1.6819457436856876e-05, + "loss": 1.4423, + "step": 899 + }, + { + "epoch": 0.08, + "learning_rate": 1.6838166510757717e-05, + "loss": 1.3927, + "step": 900 + }, + { + "epoch": 0.08, + "learning_rate": 1.6856875584658562e-05, + "loss": 1.4042, + "step": 901 + }, + { + "epoch": 0.08, + "learning_rate": 1.6875584658559403e-05, + "loss": 1.4332, + "step": 902 + }, + { + "epoch": 0.08, + "learning_rate": 1.6894293732460245e-05, + "loss": 1.4063, + "step": 903 + }, + { + "epoch": 0.08, + "learning_rate": 1.6913002806361086e-05, + "loss": 1.4072, + "step": 904 + }, + { + "epoch": 0.08, + "learning_rate": 1.6931711880261927e-05, + "loss": 1.3567, + "step": 905 + }, + { + "epoch": 0.08, + "learning_rate": 1.695042095416277e-05, + "loss": 1.4398, + "step": 906 + }, + { + "epoch": 0.08, + "learning_rate": 1.6969130028063613e-05, + "loss": 1.3622, + "step": 907 + }, + { + "epoch": 0.08, + "learning_rate": 1.6987839101964454e-05, + "loss": 1.5128, + "step": 908 + }, + { + "epoch": 0.08, + "learning_rate": 1.7006548175865295e-05, + "loss": 1.4018, + "step": 909 + }, + { + "epoch": 0.08, + "learning_rate": 1.7025257249766137e-05, + "loss": 1.4799, + "step": 910 + }, + { + "epoch": 0.08, + "learning_rate": 1.704396632366698e-05, + "loss": 1.4067, + "step": 911 + }, + { + "epoch": 0.08, + "learning_rate": 1.7062675397567823e-05, + "loss": 1.3413, + "step": 912 + }, + { + "epoch": 0.08, + "learning_rate": 1.7081384471468664e-05, + "loss": 1.3934, + "step": 913 + }, + { + "epoch": 0.08, + "learning_rate": 1.7100093545369505e-05, + "loss": 1.5173, + "step": 914 + }, + { + "epoch": 0.08, + "learning_rate": 1.7118802619270346e-05, + "loss": 1.38, + "step": 915 + }, + { + "epoch": 0.08, + "learning_rate": 1.7137511693171188e-05, + "loss": 1.3913, + "step": 916 + }, + { + "epoch": 0.08, + "learning_rate": 1.7156220767072032e-05, + "loss": 1.4034, + "step": 917 + }, + { + "epoch": 0.08, + "learning_rate": 1.7174929840972874e-05, + "loss": 1.3735, + "step": 918 + }, + { + "epoch": 0.08, + "learning_rate": 1.7193638914873715e-05, + "loss": 1.4261, + "step": 919 + }, + { + "epoch": 0.08, + "learning_rate": 1.721234798877456e-05, + "loss": 1.4258, + "step": 920 + }, + { + "epoch": 0.08, + "learning_rate": 1.7231057062675397e-05, + "loss": 1.3566, + "step": 921 + }, + { + "epoch": 0.08, + "learning_rate": 1.7249766136576242e-05, + "loss": 1.4478, + "step": 922 + }, + { + "epoch": 0.08, + "learning_rate": 1.7268475210477083e-05, + "loss": 1.4342, + "step": 923 + }, + { + "epoch": 0.08, + "learning_rate": 1.7287184284377925e-05, + "loss": 1.3906, + "step": 924 + }, + { + "epoch": 0.08, + "learning_rate": 1.7305893358278766e-05, + "loss": 1.3801, + "step": 925 + }, + { + "epoch": 0.08, + "learning_rate": 1.732460243217961e-05, + "loss": 1.4324, + "step": 926 + }, + { + "epoch": 0.08, + "learning_rate": 1.734331150608045e-05, + "loss": 1.3908, + "step": 927 + }, + { + "epoch": 0.08, + "learning_rate": 1.7362020579981293e-05, + "loss": 1.3511, + "step": 928 + }, + { + "epoch": 0.08, + "learning_rate": 1.7380729653882134e-05, + "loss": 1.313, + "step": 929 + }, + { + "epoch": 0.08, + "learning_rate": 1.7399438727782975e-05, + "loss": 1.3571, + "step": 930 + }, + { + "epoch": 0.08, + "learning_rate": 1.7418147801683817e-05, + "loss": 1.3339, + "step": 931 + }, + { + "epoch": 0.08, + "learning_rate": 1.743685687558466e-05, + "loss": 1.4729, + "step": 932 + }, + { + "epoch": 0.08, + "learning_rate": 1.7455565949485503e-05, + "loss": 1.4058, + "step": 933 + }, + { + "epoch": 0.08, + "learning_rate": 1.7474275023386344e-05, + "loss": 1.5053, + "step": 934 + }, + { + "epoch": 0.08, + "learning_rate": 1.7492984097287185e-05, + "loss": 1.3101, + "step": 935 + }, + { + "epoch": 0.08, + "learning_rate": 1.7511693171188026e-05, + "loss": 1.3748, + "step": 936 + }, + { + "epoch": 0.08, + "learning_rate": 1.7530402245088868e-05, + "loss": 1.3709, + "step": 937 + }, + { + "epoch": 0.08, + "learning_rate": 1.7549111318989712e-05, + "loss": 1.385, + "step": 938 + }, + { + "epoch": 0.08, + "learning_rate": 1.7567820392890554e-05, + "loss": 1.3871, + "step": 939 + }, + { + "epoch": 0.08, + "learning_rate": 1.7586529466791395e-05, + "loss": 1.4594, + "step": 940 + }, + { + "epoch": 0.08, + "learning_rate": 1.7605238540692236e-05, + "loss": 1.4802, + "step": 941 + }, + { + "epoch": 0.08, + "learning_rate": 1.762394761459308e-05, + "loss": 1.4854, + "step": 942 + }, + { + "epoch": 0.08, + "learning_rate": 1.7642656688493922e-05, + "loss": 1.4065, + "step": 943 + }, + { + "epoch": 0.08, + "learning_rate": 1.7661365762394763e-05, + "loss": 1.3661, + "step": 944 + }, + { + "epoch": 0.08, + "learning_rate": 1.7680074836295604e-05, + "loss": 1.4509, + "step": 945 + }, + { + "epoch": 0.08, + "learning_rate": 1.7698783910196446e-05, + "loss": 1.4044, + "step": 946 + }, + { + "epoch": 0.08, + "learning_rate": 1.771749298409729e-05, + "loss": 1.4146, + "step": 947 + }, + { + "epoch": 0.08, + "learning_rate": 1.773620205799813e-05, + "loss": 1.4567, + "step": 948 + }, + { + "epoch": 0.08, + "learning_rate": 1.7754911131898973e-05, + "loss": 1.486, + "step": 949 + }, + { + "epoch": 0.08, + "learning_rate": 1.7773620205799814e-05, + "loss": 1.3192, + "step": 950 + }, + { + "epoch": 0.08, + "learning_rate": 1.7792329279700655e-05, + "loss": 1.4621, + "step": 951 + }, + { + "epoch": 0.08, + "learning_rate": 1.7811038353601497e-05, + "loss": 1.4594, + "step": 952 + }, + { + "epoch": 0.08, + "learning_rate": 1.782974742750234e-05, + "loss": 1.4818, + "step": 953 + }, + { + "epoch": 0.08, + "learning_rate": 1.7848456501403183e-05, + "loss": 1.5281, + "step": 954 + }, + { + "epoch": 0.08, + "learning_rate": 1.7867165575304024e-05, + "loss": 1.4126, + "step": 955 + }, + { + "epoch": 0.08, + "learning_rate": 1.7885874649204865e-05, + "loss": 1.4116, + "step": 956 + }, + { + "epoch": 0.08, + "learning_rate": 1.790458372310571e-05, + "loss": 1.4304, + "step": 957 + }, + { + "epoch": 0.08, + "learning_rate": 1.792329279700655e-05, + "loss": 1.4888, + "step": 958 + }, + { + "epoch": 0.08, + "learning_rate": 1.7942001870907392e-05, + "loss": 1.3888, + "step": 959 + }, + { + "epoch": 0.08, + "learning_rate": 1.7960710944808233e-05, + "loss": 1.4362, + "step": 960 + }, + { + "epoch": 0.08, + "learning_rate": 1.7979420018709075e-05, + "loss": 1.4154, + "step": 961 + }, + { + "epoch": 0.08, + "learning_rate": 1.7998129092609916e-05, + "loss": 1.3945, + "step": 962 + }, + { + "epoch": 0.08, + "learning_rate": 1.801683816651076e-05, + "loss": 1.3527, + "step": 963 + }, + { + "epoch": 0.08, + "learning_rate": 1.8035547240411602e-05, + "loss": 1.3427, + "step": 964 + }, + { + "epoch": 0.08, + "learning_rate": 1.8054256314312443e-05, + "loss": 1.4289, + "step": 965 + }, + { + "epoch": 0.08, + "learning_rate": 1.8072965388213284e-05, + "loss": 1.4092, + "step": 966 + }, + { + "epoch": 0.08, + "learning_rate": 1.8091674462114126e-05, + "loss": 1.3376, + "step": 967 + }, + { + "epoch": 0.08, + "learning_rate": 1.8110383536014967e-05, + "loss": 1.3688, + "step": 968 + }, + { + "epoch": 0.08, + "learning_rate": 1.812909260991581e-05, + "loss": 1.5387, + "step": 969 + }, + { + "epoch": 0.08, + "learning_rate": 1.8147801683816653e-05, + "loss": 1.5368, + "step": 970 + }, + { + "epoch": 0.08, + "learning_rate": 1.8166510757717494e-05, + "loss": 1.3764, + "step": 971 + }, + { + "epoch": 0.08, + "learning_rate": 1.818521983161834e-05, + "loss": 1.3717, + "step": 972 + }, + { + "epoch": 0.08, + "learning_rate": 1.820392890551918e-05, + "loss": 1.4625, + "step": 973 + }, + { + "epoch": 0.08, + "learning_rate": 1.8222637979420018e-05, + "loss": 1.327, + "step": 974 + }, + { + "epoch": 0.08, + "learning_rate": 1.8241347053320862e-05, + "loss": 1.4143, + "step": 975 + }, + { + "epoch": 0.08, + "learning_rate": 1.8260056127221704e-05, + "loss": 1.4276, + "step": 976 + }, + { + "epoch": 0.08, + "learning_rate": 1.8278765201122545e-05, + "loss": 1.3892, + "step": 977 + }, + { + "epoch": 0.08, + "learning_rate": 1.829747427502339e-05, + "loss": 1.4024, + "step": 978 + }, + { + "epoch": 0.08, + "learning_rate": 1.831618334892423e-05, + "loss": 1.4966, + "step": 979 + }, + { + "epoch": 0.08, + "learning_rate": 1.8334892422825072e-05, + "loss": 1.4896, + "step": 980 + }, + { + "epoch": 0.08, + "learning_rate": 1.8353601496725913e-05, + "loss": 1.4115, + "step": 981 + }, + { + "epoch": 0.08, + "learning_rate": 1.8372310570626755e-05, + "loss": 1.4272, + "step": 982 + }, + { + "epoch": 0.08, + "learning_rate": 1.8391019644527596e-05, + "loss": 1.5083, + "step": 983 + }, + { + "epoch": 0.08, + "learning_rate": 1.840972871842844e-05, + "loss": 1.4429, + "step": 984 + }, + { + "epoch": 0.08, + "learning_rate": 1.8428437792329282e-05, + "loss": 1.3988, + "step": 985 + }, + { + "epoch": 0.08, + "learning_rate": 1.8447146866230123e-05, + "loss": 1.3399, + "step": 986 + }, + { + "epoch": 0.08, + "learning_rate": 1.8465855940130964e-05, + "loss": 1.5205, + "step": 987 + }, + { + "epoch": 0.08, + "learning_rate": 1.848456501403181e-05, + "loss": 1.5084, + "step": 988 + }, + { + "epoch": 0.08, + "learning_rate": 1.8503274087932647e-05, + "loss": 1.4468, + "step": 989 + }, + { + "epoch": 0.08, + "learning_rate": 1.852198316183349e-05, + "loss": 1.447, + "step": 990 + }, + { + "epoch": 0.08, + "learning_rate": 1.8540692235734333e-05, + "loss": 1.5033, + "step": 991 + }, + { + "epoch": 0.08, + "learning_rate": 1.8559401309635174e-05, + "loss": 1.3698, + "step": 992 + }, + { + "epoch": 0.08, + "learning_rate": 1.8578110383536015e-05, + "loss": 1.4487, + "step": 993 + }, + { + "epoch": 0.08, + "learning_rate": 1.859681945743686e-05, + "loss": 1.581, + "step": 994 + }, + { + "epoch": 0.08, + "learning_rate": 1.86155285313377e-05, + "loss": 1.416, + "step": 995 + }, + { + "epoch": 0.08, + "learning_rate": 1.8634237605238542e-05, + "loss": 1.485, + "step": 996 + }, + { + "epoch": 0.08, + "learning_rate": 1.8652946679139384e-05, + "loss": 1.4398, + "step": 997 + }, + { + "epoch": 0.08, + "learning_rate": 1.8671655753040225e-05, + "loss": 1.4277, + "step": 998 + }, + { + "epoch": 0.08, + "learning_rate": 1.8690364826941066e-05, + "loss": 1.4326, + "step": 999 + }, + { + "epoch": 0.08, + "learning_rate": 1.870907390084191e-05, + "loss": 1.4811, + "step": 1000 + }, + { + "epoch": 0.08, + "learning_rate": 1.8727782974742752e-05, + "loss": 1.4204, + "step": 1001 + }, + { + "epoch": 0.08, + "learning_rate": 1.8746492048643593e-05, + "loss": 1.3608, + "step": 1002 + }, + { + "epoch": 0.08, + "learning_rate": 1.8765201122544438e-05, + "loss": 1.3512, + "step": 1003 + }, + { + "epoch": 0.08, + "learning_rate": 1.878391019644528e-05, + "loss": 1.4096, + "step": 1004 + }, + { + "epoch": 0.08, + "learning_rate": 1.8802619270346117e-05, + "loss": 1.4451, + "step": 1005 + }, + { + "epoch": 0.08, + "learning_rate": 1.8821328344246962e-05, + "loss": 1.4811, + "step": 1006 + }, + { + "epoch": 0.08, + "learning_rate": 1.8840037418147803e-05, + "loss": 1.3811, + "step": 1007 + }, + { + "epoch": 0.08, + "learning_rate": 1.8858746492048644e-05, + "loss": 1.3997, + "step": 1008 + }, + { + "epoch": 0.08, + "learning_rate": 1.887745556594949e-05, + "loss": 1.2888, + "step": 1009 + }, + { + "epoch": 0.09, + "learning_rate": 1.889616463985033e-05, + "loss": 1.3597, + "step": 1010 + }, + { + "epoch": 0.09, + "learning_rate": 1.891487371375117e-05, + "loss": 1.4325, + "step": 1011 + }, + { + "epoch": 0.09, + "learning_rate": 1.8933582787652013e-05, + "loss": 1.433, + "step": 1012 + }, + { + "epoch": 0.09, + "learning_rate": 1.8952291861552854e-05, + "loss": 1.458, + "step": 1013 + }, + { + "epoch": 0.09, + "learning_rate": 1.8971000935453695e-05, + "loss": 1.35, + "step": 1014 + }, + { + "epoch": 0.09, + "learning_rate": 1.898971000935454e-05, + "loss": 1.4185, + "step": 1015 + }, + { + "epoch": 0.09, + "learning_rate": 1.900841908325538e-05, + "loss": 1.45, + "step": 1016 + }, + { + "epoch": 0.09, + "learning_rate": 1.9027128157156222e-05, + "loss": 1.4848, + "step": 1017 + }, + { + "epoch": 0.09, + "learning_rate": 1.9045837231057064e-05, + "loss": 1.4239, + "step": 1018 + }, + { + "epoch": 0.09, + "learning_rate": 1.9064546304957908e-05, + "loss": 1.3794, + "step": 1019 + }, + { + "epoch": 0.09, + "learning_rate": 1.9083255378858746e-05, + "loss": 1.47, + "step": 1020 + }, + { + "epoch": 0.09, + "learning_rate": 1.910196445275959e-05, + "loss": 1.4589, + "step": 1021 + }, + { + "epoch": 0.09, + "learning_rate": 1.9120673526660432e-05, + "loss": 1.5056, + "step": 1022 + }, + { + "epoch": 0.09, + "learning_rate": 1.9139382600561273e-05, + "loss": 1.3328, + "step": 1023 + }, + { + "epoch": 0.09, + "learning_rate": 1.9158091674462115e-05, + "loss": 1.3804, + "step": 1024 + }, + { + "epoch": 0.09, + "learning_rate": 1.917680074836296e-05, + "loss": 1.403, + "step": 1025 + }, + { + "epoch": 0.09, + "learning_rate": 1.91955098222638e-05, + "loss": 1.4801, + "step": 1026 + }, + { + "epoch": 0.09, + "learning_rate": 1.9214218896164642e-05, + "loss": 1.3841, + "step": 1027 + }, + { + "epoch": 0.09, + "learning_rate": 1.9232927970065483e-05, + "loss": 1.3507, + "step": 1028 + }, + { + "epoch": 0.09, + "learning_rate": 1.9251637043966324e-05, + "loss": 1.4452, + "step": 1029 + }, + { + "epoch": 0.09, + "learning_rate": 1.9270346117867166e-05, + "loss": 1.4402, + "step": 1030 + }, + { + "epoch": 0.09, + "learning_rate": 1.928905519176801e-05, + "loss": 1.4694, + "step": 1031 + }, + { + "epoch": 0.09, + "learning_rate": 1.930776426566885e-05, + "loss": 1.4377, + "step": 1032 + }, + { + "epoch": 0.09, + "learning_rate": 1.9326473339569693e-05, + "loss": 1.4167, + "step": 1033 + }, + { + "epoch": 0.09, + "learning_rate": 1.9345182413470537e-05, + "loss": 1.3582, + "step": 1034 + }, + { + "epoch": 0.09, + "learning_rate": 1.9363891487371375e-05, + "loss": 1.3904, + "step": 1035 + }, + { + "epoch": 0.09, + "learning_rate": 1.9382600561272216e-05, + "loss": 1.3033, + "step": 1036 + }, + { + "epoch": 0.09, + "learning_rate": 1.940130963517306e-05, + "loss": 1.2905, + "step": 1037 + }, + { + "epoch": 0.09, + "learning_rate": 1.9420018709073902e-05, + "loss": 1.4305, + "step": 1038 + }, + { + "epoch": 0.09, + "learning_rate": 1.9438727782974744e-05, + "loss": 1.4335, + "step": 1039 + }, + { + "epoch": 0.09, + "learning_rate": 1.9457436856875588e-05, + "loss": 1.3623, + "step": 1040 + }, + { + "epoch": 0.09, + "learning_rate": 1.947614593077643e-05, + "loss": 1.4198, + "step": 1041 + }, + { + "epoch": 0.09, + "learning_rate": 1.949485500467727e-05, + "loss": 1.5029, + "step": 1042 + }, + { + "epoch": 0.09, + "learning_rate": 1.9513564078578112e-05, + "loss": 1.4202, + "step": 1043 + }, + { + "epoch": 0.09, + "learning_rate": 1.9532273152478953e-05, + "loss": 1.4768, + "step": 1044 + }, + { + "epoch": 0.09, + "learning_rate": 1.9550982226379795e-05, + "loss": 1.4531, + "step": 1045 + }, + { + "epoch": 0.09, + "learning_rate": 1.956969130028064e-05, + "loss": 1.3032, + "step": 1046 + }, + { + "epoch": 0.09, + "learning_rate": 1.958840037418148e-05, + "loss": 1.4193, + "step": 1047 + }, + { + "epoch": 0.09, + "learning_rate": 1.960710944808232e-05, + "loss": 1.3837, + "step": 1048 + }, + { + "epoch": 0.09, + "learning_rate": 1.9625818521983163e-05, + "loss": 1.4594, + "step": 1049 + }, + { + "epoch": 0.09, + "learning_rate": 1.9644527595884004e-05, + "loss": 1.4858, + "step": 1050 + }, + { + "epoch": 0.09, + "learning_rate": 1.9663236669784845e-05, + "loss": 1.2982, + "step": 1051 + }, + { + "epoch": 0.09, + "learning_rate": 1.968194574368569e-05, + "loss": 1.4729, + "step": 1052 + }, + { + "epoch": 0.09, + "learning_rate": 1.970065481758653e-05, + "loss": 1.3053, + "step": 1053 + }, + { + "epoch": 0.09, + "learning_rate": 1.9719363891487373e-05, + "loss": 1.409, + "step": 1054 + }, + { + "epoch": 0.09, + "learning_rate": 1.9738072965388214e-05, + "loss": 1.3993, + "step": 1055 + }, + { + "epoch": 0.09, + "learning_rate": 1.975678203928906e-05, + "loss": 1.3313, + "step": 1056 + }, + { + "epoch": 0.09, + "learning_rate": 1.97754911131899e-05, + "loss": 1.4007, + "step": 1057 + }, + { + "epoch": 0.09, + "learning_rate": 1.979420018709074e-05, + "loss": 1.4982, + "step": 1058 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812909260991582e-05, + "loss": 1.4633, + "step": 1059 + }, + { + "epoch": 0.09, + "learning_rate": 1.9831618334892424e-05, + "loss": 1.3942, + "step": 1060 + }, + { + "epoch": 0.09, + "learning_rate": 1.9850327408793265e-05, + "loss": 1.4659, + "step": 1061 + }, + { + "epoch": 0.09, + "learning_rate": 1.986903648269411e-05, + "loss": 1.4033, + "step": 1062 + }, + { + "epoch": 0.09, + "learning_rate": 1.988774555659495e-05, + "loss": 1.4423, + "step": 1063 + }, + { + "epoch": 0.09, + "learning_rate": 1.9906454630495792e-05, + "loss": 1.3793, + "step": 1064 + }, + { + "epoch": 0.09, + "learning_rate": 1.9925163704396637e-05, + "loss": 1.4384, + "step": 1065 + }, + { + "epoch": 0.09, + "learning_rate": 1.9943872778297474e-05, + "loss": 1.4785, + "step": 1066 + }, + { + "epoch": 0.09, + "learning_rate": 1.9962581852198316e-05, + "loss": 1.3466, + "step": 1067 + }, + { + "epoch": 0.09, + "learning_rate": 1.998129092609916e-05, + "loss": 1.3154, + "step": 1068 + }, + { + "epoch": 0.09, + "learning_rate": 2e-05, + "loss": 1.597, + "step": 1069 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999999958688402e-05, + "loss": 1.4114, + "step": 1070 + }, + { + "epoch": 0.09, + "learning_rate": 1.99999998347536e-05, + "loss": 1.4433, + "step": 1071 + }, + { + "epoch": 0.09, + "learning_rate": 1.99999996281956e-05, + "loss": 1.4558, + "step": 1072 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999999339014406e-05, + "loss": 1.3863, + "step": 1073 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999998967210015e-05, + "loss": 1.4207, + "step": 1074 + }, + { + "epoch": 0.09, + "learning_rate": 1.999999851278243e-05, + "loss": 1.4052, + "step": 1075 + }, + { + "epoch": 0.09, + "learning_rate": 1.999999797573166e-05, + "loss": 1.4269, + "step": 1076 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999997356057703e-05, + "loss": 1.4189, + "step": 1077 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999996653760568e-05, + "loss": 1.4429, + "step": 1078 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999995868840262e-05, + "loss": 1.3982, + "step": 1079 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999995001296792e-05, + "loss": 1.5311, + "step": 1080 + }, + { + "epoch": 0.09, + "learning_rate": 1.999999405113016e-05, + "loss": 1.4177, + "step": 1081 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999993018340373e-05, + "loss": 1.4032, + "step": 1082 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999991902927446e-05, + "loss": 1.4375, + "step": 1083 + }, + { + "epoch": 0.09, + "learning_rate": 1.999999070489139e-05, + "loss": 1.3876, + "step": 1084 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999989424232208e-05, + "loss": 1.3646, + "step": 1085 + }, + { + "epoch": 0.09, + "learning_rate": 1.999998806094991e-05, + "loss": 1.4517, + "step": 1086 + }, + { + "epoch": 0.09, + "learning_rate": 1.999998661504451e-05, + "loss": 1.3811, + "step": 1087 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999985086516024e-05, + "loss": 1.4613, + "step": 1088 + }, + { + "epoch": 0.09, + "learning_rate": 1.999998347536446e-05, + "loss": 1.4613, + "step": 1089 + }, + { + "epoch": 0.09, + "learning_rate": 1.999998178158983e-05, + "loss": 1.4334, + "step": 1090 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999980005192153e-05, + "loss": 1.3342, + "step": 1091 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999978146171437e-05, + "loss": 1.4633, + "step": 1092 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999976204527705e-05, + "loss": 1.4059, + "step": 1093 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999974180260965e-05, + "loss": 1.4561, + "step": 1094 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999972073371242e-05, + "loss": 1.3499, + "step": 1095 + }, + { + "epoch": 0.09, + "learning_rate": 1.999996988385855e-05, + "loss": 1.3856, + "step": 1096 + }, + { + "epoch": 0.09, + "learning_rate": 1.99999676117229e-05, + "loss": 1.4465, + "step": 1097 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999965256964323e-05, + "loss": 1.3955, + "step": 1098 + }, + { + "epoch": 0.09, + "learning_rate": 1.999996281958283e-05, + "loss": 1.4623, + "step": 1099 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999960299578446e-05, + "loss": 1.4574, + "step": 1100 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999957696951188e-05, + "loss": 1.495, + "step": 1101 + }, + { + "epoch": 0.09, + "learning_rate": 1.999995501170108e-05, + "loss": 1.44, + "step": 1102 + }, + { + "epoch": 0.09, + "learning_rate": 1.999995224382814e-05, + "loss": 1.3497, + "step": 1103 + }, + { + "epoch": 0.09, + "learning_rate": 1.99999493933324e-05, + "loss": 1.4113, + "step": 1104 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999946460213872e-05, + "loss": 1.3053, + "step": 1105 + }, + { + "epoch": 0.09, + "learning_rate": 1.999994344447259e-05, + "loss": 1.3802, + "step": 1106 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999940346108576e-05, + "loss": 1.4714, + "step": 1107 + }, + { + "epoch": 0.09, + "learning_rate": 1.999993716512185e-05, + "loss": 1.3574, + "step": 1108 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999933901512445e-05, + "loss": 1.3867, + "step": 1109 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999930555280387e-05, + "loss": 1.4377, + "step": 1110 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999927126425697e-05, + "loss": 1.4266, + "step": 1111 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999923614948414e-05, + "loss": 1.477, + "step": 1112 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999920020848563e-05, + "loss": 1.4258, + "step": 1113 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999916344126168e-05, + "loss": 1.4481, + "step": 1114 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999912584781265e-05, + "loss": 1.4294, + "step": 1115 + }, + { + "epoch": 0.09, + "learning_rate": 1.999990874281388e-05, + "loss": 1.3849, + "step": 1116 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999904818224055e-05, + "loss": 1.468, + "step": 1117 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999900811011815e-05, + "loss": 1.4102, + "step": 1118 + }, + { + "epoch": 0.09, + "learning_rate": 1.999989672117719e-05, + "loss": 1.3806, + "step": 1119 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999892548720225e-05, + "loss": 1.4263, + "step": 1120 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999888293640944e-05, + "loss": 1.4339, + "step": 1121 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999883955939383e-05, + "loss": 1.3758, + "step": 1122 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999879535615583e-05, + "loss": 1.3857, + "step": 1123 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999875032669576e-05, + "loss": 1.3632, + "step": 1124 + }, + { + "epoch": 0.09, + "learning_rate": 1.99998704471014e-05, + "loss": 1.352, + "step": 1125 + }, + { + "epoch": 0.09, + "learning_rate": 1.99998657789111e-05, + "loss": 1.3944, + "step": 1126 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999861028098703e-05, + "loss": 1.4198, + "step": 1127 + }, + { + "epoch": 0.09, + "learning_rate": 1.9999856194664256e-05, + "loss": 1.4416, + "step": 1128 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999851278607795e-05, + "loss": 1.3756, + "step": 1129 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999846279929363e-05, + "loss": 1.4823, + "step": 1130 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999841198629003e-05, + "loss": 1.3727, + "step": 1131 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999836034706755e-05, + "loss": 1.3333, + "step": 1132 + }, + { + "epoch": 0.1, + "learning_rate": 1.999983078816266e-05, + "loss": 1.3651, + "step": 1133 + }, + { + "epoch": 0.1, + "learning_rate": 1.999982545899676e-05, + "loss": 1.4457, + "step": 1134 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999820047209103e-05, + "loss": 1.4263, + "step": 1135 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999814552799735e-05, + "loss": 1.4076, + "step": 1136 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999808975768697e-05, + "loss": 1.4343, + "step": 1137 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999803316116036e-05, + "loss": 1.352, + "step": 1138 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999797573841802e-05, + "loss": 1.434, + "step": 1139 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999791748946037e-05, + "loss": 1.386, + "step": 1140 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999785841428794e-05, + "loss": 1.3926, + "step": 1141 + }, + { + "epoch": 0.1, + "learning_rate": 1.999977985129012e-05, + "loss": 1.3882, + "step": 1142 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999773778530063e-05, + "loss": 1.3954, + "step": 1143 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999767623148676e-05, + "loss": 1.5065, + "step": 1144 + }, + { + "epoch": 0.1, + "learning_rate": 1.999976138514601e-05, + "loss": 1.4425, + "step": 1145 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999755064522116e-05, + "loss": 1.4031, + "step": 1146 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999748661277042e-05, + "loss": 1.4888, + "step": 1147 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999742175410843e-05, + "loss": 1.451, + "step": 1148 + }, + { + "epoch": 0.1, + "learning_rate": 1.999973560692358e-05, + "loss": 1.4834, + "step": 1149 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999728955815296e-05, + "loss": 1.3932, + "step": 1150 + }, + { + "epoch": 0.1, + "learning_rate": 1.999972222208605e-05, + "loss": 1.4042, + "step": 1151 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999715405735902e-05, + "loss": 1.425, + "step": 1152 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999708506764904e-05, + "loss": 1.3682, + "step": 1153 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999701525173113e-05, + "loss": 1.3961, + "step": 1154 + }, + { + "epoch": 0.1, + "learning_rate": 1.999969446096059e-05, + "loss": 1.3413, + "step": 1155 + }, + { + "epoch": 0.1, + "learning_rate": 1.999968731412739e-05, + "loss": 1.3953, + "step": 1156 + }, + { + "epoch": 0.1, + "learning_rate": 1.999968008467357e-05, + "loss": 1.4912, + "step": 1157 + }, + { + "epoch": 0.1, + "learning_rate": 1.99996727725992e-05, + "loss": 1.4697, + "step": 1158 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999665377904327e-05, + "loss": 1.4396, + "step": 1159 + }, + { + "epoch": 0.1, + "learning_rate": 1.999965790058902e-05, + "loss": 1.4455, + "step": 1160 + }, + { + "epoch": 0.1, + "learning_rate": 1.999965034065334e-05, + "loss": 1.377, + "step": 1161 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999642698097347e-05, + "loss": 1.3633, + "step": 1162 + }, + { + "epoch": 0.1, + "learning_rate": 1.999963497292111e-05, + "loss": 1.4526, + "step": 1163 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999627165124683e-05, + "loss": 1.4185, + "step": 1164 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999619274708136e-05, + "loss": 1.3426, + "step": 1165 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999611301671537e-05, + "loss": 1.3719, + "step": 1166 + }, + { + "epoch": 0.1, + "learning_rate": 1.999960324601495e-05, + "loss": 1.3962, + "step": 1167 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999595107738437e-05, + "loss": 1.4116, + "step": 1168 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999586886842072e-05, + "loss": 1.2356, + "step": 1169 + }, + { + "epoch": 0.1, + "learning_rate": 1.999957858332592e-05, + "loss": 1.351, + "step": 1170 + }, + { + "epoch": 0.1, + "learning_rate": 1.999957019719005e-05, + "loss": 1.3135, + "step": 1171 + }, + { + "epoch": 0.1, + "learning_rate": 1.999956172843453e-05, + "loss": 1.3861, + "step": 1172 + }, + { + "epoch": 0.1, + "learning_rate": 1.999955317705943e-05, + "loss": 1.4002, + "step": 1173 + }, + { + "epoch": 0.1, + "learning_rate": 1.999954454306482e-05, + "loss": 1.4167, + "step": 1174 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999535826450775e-05, + "loss": 1.3473, + "step": 1175 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999527027217365e-05, + "loss": 1.4323, + "step": 1176 + }, + { + "epoch": 0.1, + "learning_rate": 1.999951814536466e-05, + "loss": 1.3993, + "step": 1177 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999509180892738e-05, + "loss": 1.3384, + "step": 1178 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999500133801672e-05, + "loss": 1.4627, + "step": 1179 + }, + { + "epoch": 0.1, + "learning_rate": 1.999949100409153e-05, + "loss": 1.4145, + "step": 1180 + }, + { + "epoch": 0.1, + "learning_rate": 1.99994817917624e-05, + "loss": 1.4233, + "step": 1181 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999472496814345e-05, + "loss": 1.4184, + "step": 1182 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999463119247452e-05, + "loss": 1.3967, + "step": 1183 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999453659061793e-05, + "loss": 1.4448, + "step": 1184 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999444116257448e-05, + "loss": 1.4238, + "step": 1185 + }, + { + "epoch": 0.1, + "learning_rate": 1.99994344908345e-05, + "loss": 1.3967, + "step": 1186 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999424782793016e-05, + "loss": 1.4416, + "step": 1187 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999414992133088e-05, + "loss": 1.3117, + "step": 1188 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999405118854793e-05, + "loss": 1.4958, + "step": 1189 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999395162958212e-05, + "loss": 1.5066, + "step": 1190 + }, + { + "epoch": 0.1, + "learning_rate": 1.999938512444343e-05, + "loss": 1.4465, + "step": 1191 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999375003310525e-05, + "loss": 1.421, + "step": 1192 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999364799559585e-05, + "loss": 1.44, + "step": 1193 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999354513190694e-05, + "loss": 1.4453, + "step": 1194 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999344144203934e-05, + "loss": 1.4429, + "step": 1195 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999333692599393e-05, + "loss": 1.4402, + "step": 1196 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999323158377156e-05, + "loss": 1.4459, + "step": 1197 + }, + { + "epoch": 0.1, + "learning_rate": 1.999931254153731e-05, + "loss": 1.3811, + "step": 1198 + }, + { + "epoch": 0.1, + "learning_rate": 1.999930184207995e-05, + "loss": 1.494, + "step": 1199 + }, + { + "epoch": 0.1, + "learning_rate": 1.999929106000515e-05, + "loss": 1.4316, + "step": 1200 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999280195313013e-05, + "loss": 1.3682, + "step": 1201 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999269248003622e-05, + "loss": 1.4608, + "step": 1202 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999258218077064e-05, + "loss": 1.3859, + "step": 1203 + }, + { + "epoch": 0.1, + "learning_rate": 1.999924710553344e-05, + "loss": 1.4452, + "step": 1204 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999235910372833e-05, + "loss": 1.423, + "step": 1205 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999224632595337e-05, + "loss": 1.457, + "step": 1206 + }, + { + "epoch": 0.1, + "learning_rate": 1.999921327220105e-05, + "loss": 1.3174, + "step": 1207 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999201829190062e-05, + "loss": 1.4178, + "step": 1208 + }, + { + "epoch": 0.1, + "learning_rate": 1.999919030356247e-05, + "loss": 1.3446, + "step": 1209 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999178695318365e-05, + "loss": 1.4566, + "step": 1210 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999167004457846e-05, + "loss": 1.445, + "step": 1211 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999155230981006e-05, + "loss": 1.4643, + "step": 1212 + }, + { + "epoch": 0.1, + "learning_rate": 1.999914337488795e-05, + "loss": 1.4923, + "step": 1213 + }, + { + "epoch": 0.1, + "learning_rate": 1.999913143617877e-05, + "loss": 1.4162, + "step": 1214 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999119414853564e-05, + "loss": 1.4897, + "step": 1215 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999107310912434e-05, + "loss": 1.5032, + "step": 1216 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999095124355485e-05, + "loss": 1.4513, + "step": 1217 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999082855182804e-05, + "loss": 1.3146, + "step": 1218 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999070503394504e-05, + "loss": 1.4161, + "step": 1219 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999058068990686e-05, + "loss": 1.3279, + "step": 1220 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999045551971445e-05, + "loss": 1.4098, + "step": 1221 + }, + { + "epoch": 0.1, + "learning_rate": 1.9999032952336893e-05, + "loss": 1.4873, + "step": 1222 + }, + { + "epoch": 0.1, + "learning_rate": 1.999902027008713e-05, + "loss": 1.3794, + "step": 1223 + }, + { + "epoch": 0.1, + "learning_rate": 1.999900750522226e-05, + "loss": 1.4757, + "step": 1224 + }, + { + "epoch": 0.1, + "learning_rate": 1.999899465774239e-05, + "loss": 1.383, + "step": 1225 + }, + { + "epoch": 0.1, + "learning_rate": 1.9998981727647625e-05, + "loss": 1.3973, + "step": 1226 + }, + { + "epoch": 0.1, + "learning_rate": 1.9998968714938076e-05, + "loss": 1.4417, + "step": 1227 + }, + { + "epoch": 0.1, + "learning_rate": 1.9998955619613847e-05, + "loss": 1.3901, + "step": 1228 + }, + { + "epoch": 0.1, + "learning_rate": 1.9998942441675045e-05, + "loss": 1.454, + "step": 1229 + }, + { + "epoch": 0.1, + "learning_rate": 1.999892918112178e-05, + "loss": 1.4454, + "step": 1230 + }, + { + "epoch": 0.1, + "learning_rate": 1.9998915837954164e-05, + "loss": 1.3112, + "step": 1231 + }, + { + "epoch": 0.1, + "learning_rate": 1.9998902412172305e-05, + "loss": 1.4571, + "step": 1232 + }, + { + "epoch": 0.1, + "learning_rate": 1.999888890377631e-05, + "loss": 1.4689, + "step": 1233 + }, + { + "epoch": 0.1, + "learning_rate": 1.99988753127663e-05, + "loss": 1.4206, + "step": 1234 + }, + { + "epoch": 0.1, + "learning_rate": 1.999886163914238e-05, + "loss": 1.2963, + "step": 1235 + }, + { + "epoch": 0.1, + "learning_rate": 1.9998847882904665e-05, + "loss": 1.3323, + "step": 1236 + }, + { + "epoch": 0.1, + "learning_rate": 1.999883404405327e-05, + "loss": 1.3551, + "step": 1237 + }, + { + "epoch": 0.1, + "learning_rate": 1.999882012258831e-05, + "loss": 1.3424, + "step": 1238 + }, + { + "epoch": 0.1, + "learning_rate": 1.9998806118509896e-05, + "loss": 1.4111, + "step": 1239 + }, + { + "epoch": 0.1, + "learning_rate": 1.9998792031818143e-05, + "loss": 1.4166, + "step": 1240 + }, + { + "epoch": 0.1, + "learning_rate": 1.9998777862513175e-05, + "loss": 1.341, + "step": 1241 + }, + { + "epoch": 0.1, + "learning_rate": 1.99987636105951e-05, + "loss": 1.4674, + "step": 1242 + }, + { + "epoch": 0.1, + "learning_rate": 1.9998749276064046e-05, + "loss": 1.3768, + "step": 1243 + }, + { + "epoch": 0.1, + "learning_rate": 1.9998734858920118e-05, + "loss": 1.4563, + "step": 1244 + }, + { + "epoch": 0.1, + "learning_rate": 1.9998720359163447e-05, + "loss": 1.407, + "step": 1245 + }, + { + "epoch": 0.1, + "learning_rate": 1.999870577679415e-05, + "loss": 1.4737, + "step": 1246 + }, + { + "epoch": 0.1, + "learning_rate": 1.999869111181234e-05, + "loss": 1.3696, + "step": 1247 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998676364218146e-05, + "loss": 1.4213, + "step": 1248 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998661534011693e-05, + "loss": 1.4173, + "step": 1249 + }, + { + "epoch": 0.11, + "learning_rate": 1.999864662119309e-05, + "loss": 1.4218, + "step": 1250 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998631625762476e-05, + "loss": 1.2914, + "step": 1251 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998616547719963e-05, + "loss": 1.3378, + "step": 1252 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998601387065683e-05, + "loss": 1.3372, + "step": 1253 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998586143799757e-05, + "loss": 1.382, + "step": 1254 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998570817922315e-05, + "loss": 1.4117, + "step": 1255 + }, + { + "epoch": 0.11, + "learning_rate": 1.999855540943348e-05, + "loss": 1.4276, + "step": 1256 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998539918333374e-05, + "loss": 1.3651, + "step": 1257 + }, + { + "epoch": 0.11, + "learning_rate": 1.999852434462214e-05, + "loss": 1.4576, + "step": 1258 + }, + { + "epoch": 0.11, + "learning_rate": 1.999850868829989e-05, + "loss": 1.5058, + "step": 1259 + }, + { + "epoch": 0.11, + "learning_rate": 1.999849294936677e-05, + "loss": 1.5059, + "step": 1260 + }, + { + "epoch": 0.11, + "learning_rate": 1.999847712782289e-05, + "loss": 1.4122, + "step": 1261 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998461223668403e-05, + "loss": 1.4337, + "step": 1262 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998445236903424e-05, + "loss": 1.3125, + "step": 1263 + }, + { + "epoch": 0.11, + "learning_rate": 1.999842916752809e-05, + "loss": 1.2737, + "step": 1264 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998413015542537e-05, + "loss": 1.3911, + "step": 1265 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998396780946895e-05, + "loss": 1.4194, + "step": 1266 + }, + { + "epoch": 0.11, + "learning_rate": 1.99983804637413e-05, + "loss": 1.463, + "step": 1267 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998364063925885e-05, + "loss": 1.339, + "step": 1268 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998347581500783e-05, + "loss": 1.4389, + "step": 1269 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998331016466137e-05, + "loss": 1.4094, + "step": 1270 + }, + { + "epoch": 0.11, + "learning_rate": 1.999831436882208e-05, + "loss": 1.3734, + "step": 1271 + }, + { + "epoch": 0.11, + "learning_rate": 1.999829763856875e-05, + "loss": 1.3362, + "step": 1272 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998280825706284e-05, + "loss": 1.3632, + "step": 1273 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998263930234822e-05, + "loss": 1.4363, + "step": 1274 + }, + { + "epoch": 0.11, + "learning_rate": 1.99982469521545e-05, + "loss": 1.387, + "step": 1275 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998229891465467e-05, + "loss": 1.4555, + "step": 1276 + }, + { + "epoch": 0.11, + "learning_rate": 1.999821274816786e-05, + "loss": 1.3088, + "step": 1277 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998195522261814e-05, + "loss": 1.4191, + "step": 1278 + }, + { + "epoch": 0.11, + "learning_rate": 1.999817821374748e-05, + "loss": 1.3825, + "step": 1279 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998160822624997e-05, + "loss": 1.3738, + "step": 1280 + }, + { + "epoch": 0.11, + "learning_rate": 1.999814334889451e-05, + "loss": 1.3531, + "step": 1281 + }, + { + "epoch": 0.11, + "learning_rate": 1.999812579255616e-05, + "loss": 1.4026, + "step": 1282 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998108153610102e-05, + "loss": 1.3988, + "step": 1283 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998090432056467e-05, + "loss": 1.3749, + "step": 1284 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998072627895415e-05, + "loss": 1.3958, + "step": 1285 + }, + { + "epoch": 0.11, + "learning_rate": 1.999805474112708e-05, + "loss": 1.3955, + "step": 1286 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998036771751623e-05, + "loss": 1.3984, + "step": 1287 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998018719769184e-05, + "loss": 1.4111, + "step": 1288 + }, + { + "epoch": 0.11, + "learning_rate": 1.9998000585179915e-05, + "loss": 1.363, + "step": 1289 + }, + { + "epoch": 0.11, + "learning_rate": 1.999798236798397e-05, + "loss": 1.4189, + "step": 1290 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997964068181486e-05, + "loss": 1.3926, + "step": 1291 + }, + { + "epoch": 0.11, + "learning_rate": 1.999794568577263e-05, + "loss": 1.3646, + "step": 1292 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997927220757543e-05, + "loss": 1.4102, + "step": 1293 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997908673136384e-05, + "loss": 1.3838, + "step": 1294 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997890042909303e-05, + "loss": 1.301, + "step": 1295 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997871330076455e-05, + "loss": 1.4073, + "step": 1296 + }, + { + "epoch": 0.11, + "learning_rate": 1.999785253463799e-05, + "loss": 1.3254, + "step": 1297 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997833656594076e-05, + "loss": 1.4657, + "step": 1298 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997814695944857e-05, + "loss": 1.3036, + "step": 1299 + }, + { + "epoch": 0.11, + "learning_rate": 1.999779565269049e-05, + "loss": 1.4055, + "step": 1300 + }, + { + "epoch": 0.11, + "learning_rate": 1.999777652683114e-05, + "loss": 1.3331, + "step": 1301 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997757318366958e-05, + "loss": 1.2921, + "step": 1302 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997738027298107e-05, + "loss": 1.4754, + "step": 1303 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997718653624743e-05, + "loss": 1.3145, + "step": 1304 + }, + { + "epoch": 0.11, + "learning_rate": 1.999769919734703e-05, + "loss": 1.4292, + "step": 1305 + }, + { + "epoch": 0.11, + "learning_rate": 1.999767965846512e-05, + "loss": 1.4767, + "step": 1306 + }, + { + "epoch": 0.11, + "learning_rate": 1.999766003697919e-05, + "loss": 1.4647, + "step": 1307 + }, + { + "epoch": 0.11, + "learning_rate": 1.999764033288939e-05, + "loss": 1.4459, + "step": 1308 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997620546195885e-05, + "loss": 1.3542, + "step": 1309 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997600676898838e-05, + "loss": 1.4681, + "step": 1310 + }, + { + "epoch": 0.11, + "learning_rate": 1.999758072499842e-05, + "loss": 1.308, + "step": 1311 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997560690494784e-05, + "loss": 1.3994, + "step": 1312 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997540573388104e-05, + "loss": 1.4456, + "step": 1313 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997520373678543e-05, + "loss": 1.398, + "step": 1314 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997500091366272e-05, + "loss": 1.4938, + "step": 1315 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997479726451453e-05, + "loss": 1.4545, + "step": 1316 + }, + { + "epoch": 0.11, + "learning_rate": 1.999745927893426e-05, + "loss": 1.3968, + "step": 1317 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997438748814856e-05, + "loss": 1.362, + "step": 1318 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997418136093416e-05, + "loss": 1.2993, + "step": 1319 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997397440770106e-05, + "loss": 1.464, + "step": 1320 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997376662845098e-05, + "loss": 1.4643, + "step": 1321 + }, + { + "epoch": 0.11, + "learning_rate": 1.999735580231857e-05, + "loss": 1.4454, + "step": 1322 + }, + { + "epoch": 0.11, + "learning_rate": 1.999733485919068e-05, + "loss": 1.3677, + "step": 1323 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997313833461614e-05, + "loss": 1.5248, + "step": 1324 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997292725131544e-05, + "loss": 1.516, + "step": 1325 + }, + { + "epoch": 0.11, + "learning_rate": 1.999727153420064e-05, + "loss": 1.4632, + "step": 1326 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997250260669076e-05, + "loss": 1.4306, + "step": 1327 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997228904537037e-05, + "loss": 1.4415, + "step": 1328 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997207465804686e-05, + "loss": 1.3627, + "step": 1329 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997185944472214e-05, + "loss": 1.4757, + "step": 1330 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997164340539788e-05, + "loss": 1.3483, + "step": 1331 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997142654007593e-05, + "loss": 1.3058, + "step": 1332 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997120884875804e-05, + "loss": 1.405, + "step": 1333 + }, + { + "epoch": 0.11, + "learning_rate": 1.9997099033144605e-05, + "loss": 1.4586, + "step": 1334 + }, + { + "epoch": 0.11, + "learning_rate": 1.999707709881417e-05, + "loss": 1.3827, + "step": 1335 + }, + { + "epoch": 0.11, + "learning_rate": 1.999705508188469e-05, + "loss": 1.3947, + "step": 1336 + }, + { + "epoch": 0.11, + "learning_rate": 1.999703298235634e-05, + "loss": 1.4258, + "step": 1337 + }, + { + "epoch": 0.11, + "learning_rate": 1.99970108002293e-05, + "loss": 1.4302, + "step": 1338 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996988535503762e-05, + "loss": 1.2587, + "step": 1339 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996966188179905e-05, + "loss": 1.4422, + "step": 1340 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996943758257914e-05, + "loss": 1.4451, + "step": 1341 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996921245737972e-05, + "loss": 1.3968, + "step": 1342 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996898650620267e-05, + "loss": 1.4114, + "step": 1343 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996875972904987e-05, + "loss": 1.3725, + "step": 1344 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996853212592317e-05, + "loss": 1.3251, + "step": 1345 + }, + { + "epoch": 0.11, + "learning_rate": 1.999683036968245e-05, + "loss": 1.4313, + "step": 1346 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996807444175567e-05, + "loss": 1.4277, + "step": 1347 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996784436071864e-05, + "loss": 1.3763, + "step": 1348 + }, + { + "epoch": 0.11, + "learning_rate": 1.999676134537153e-05, + "loss": 1.4215, + "step": 1349 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996738172074752e-05, + "loss": 1.3947, + "step": 1350 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996714916181725e-05, + "loss": 1.3466, + "step": 1351 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996691577692638e-05, + "loss": 1.4365, + "step": 1352 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996668156607688e-05, + "loss": 1.367, + "step": 1353 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996644652927065e-05, + "loss": 1.4043, + "step": 1354 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996621066650964e-05, + "loss": 1.4, + "step": 1355 + }, + { + "epoch": 0.11, + "learning_rate": 1.999659739777958e-05, + "loss": 1.38, + "step": 1356 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996573646313112e-05, + "loss": 1.4127, + "step": 1357 + }, + { + "epoch": 0.11, + "learning_rate": 1.999654981225175e-05, + "loss": 1.4515, + "step": 1358 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996525895595697e-05, + "loss": 1.4147, + "step": 1359 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996501896345146e-05, + "loss": 1.3988, + "step": 1360 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996477814500297e-05, + "loss": 1.329, + "step": 1361 + }, + { + "epoch": 0.11, + "learning_rate": 1.999645365006135e-05, + "loss": 1.4483, + "step": 1362 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996429403028505e-05, + "loss": 1.2786, + "step": 1363 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996405073401955e-05, + "loss": 1.4597, + "step": 1364 + }, + { + "epoch": 0.11, + "learning_rate": 1.9996380661181917e-05, + "loss": 1.3653, + "step": 1365 + }, + { + "epoch": 0.12, + "learning_rate": 1.9996356166368574e-05, + "loss": 1.4552, + "step": 1366 + }, + { + "epoch": 0.12, + "learning_rate": 1.999633158896214e-05, + "loss": 1.4357, + "step": 1367 + }, + { + "epoch": 0.12, + "learning_rate": 1.9996306928962816e-05, + "loss": 1.367, + "step": 1368 + }, + { + "epoch": 0.12, + "learning_rate": 1.9996282186370804e-05, + "loss": 1.3856, + "step": 1369 + }, + { + "epoch": 0.12, + "learning_rate": 1.999625736118631e-05, + "loss": 1.4358, + "step": 1370 + }, + { + "epoch": 0.12, + "learning_rate": 1.9996232453409538e-05, + "loss": 1.319, + "step": 1371 + }, + { + "epoch": 0.12, + "learning_rate": 1.9996207463040694e-05, + "loss": 1.3662, + "step": 1372 + }, + { + "epoch": 0.12, + "learning_rate": 1.9996182390079984e-05, + "loss": 1.5298, + "step": 1373 + }, + { + "epoch": 0.12, + "learning_rate": 1.9996157234527617e-05, + "loss": 1.4523, + "step": 1374 + }, + { + "epoch": 0.12, + "learning_rate": 1.9996131996383797e-05, + "loss": 1.4825, + "step": 1375 + }, + { + "epoch": 0.12, + "learning_rate": 1.999610667564874e-05, + "loss": 1.4082, + "step": 1376 + }, + { + "epoch": 0.12, + "learning_rate": 1.9996081272322646e-05, + "loss": 1.4479, + "step": 1377 + }, + { + "epoch": 0.12, + "learning_rate": 1.9996055786405734e-05, + "loss": 1.4865, + "step": 1378 + }, + { + "epoch": 0.12, + "learning_rate": 1.9996030217898208e-05, + "loss": 1.4076, + "step": 1379 + }, + { + "epoch": 0.12, + "learning_rate": 1.9996004566800283e-05, + "loss": 1.434, + "step": 1380 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995978833112164e-05, + "loss": 1.4933, + "step": 1381 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995953016834073e-05, + "loss": 1.4906, + "step": 1382 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995927117966218e-05, + "loss": 1.4148, + "step": 1383 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995901136508815e-05, + "loss": 1.472, + "step": 1384 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995875072462074e-05, + "loss": 1.3713, + "step": 1385 + }, + { + "epoch": 0.12, + "learning_rate": 1.999584892582622e-05, + "loss": 1.4379, + "step": 1386 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995822696601457e-05, + "loss": 1.385, + "step": 1387 + }, + { + "epoch": 0.12, + "learning_rate": 1.999579638478801e-05, + "loss": 1.4488, + "step": 1388 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995769990386095e-05, + "loss": 1.3341, + "step": 1389 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995743513395932e-05, + "loss": 1.4656, + "step": 1390 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995716953817733e-05, + "loss": 1.3499, + "step": 1391 + }, + { + "epoch": 0.12, + "learning_rate": 1.999569031165172e-05, + "loss": 1.3622, + "step": 1392 + }, + { + "epoch": 0.12, + "learning_rate": 1.999566358689812e-05, + "loss": 1.3468, + "step": 1393 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995636779557145e-05, + "loss": 1.4469, + "step": 1394 + }, + { + "epoch": 0.12, + "learning_rate": 1.999560988962902e-05, + "loss": 1.3103, + "step": 1395 + }, + { + "epoch": 0.12, + "learning_rate": 1.999558291711397e-05, + "loss": 1.4201, + "step": 1396 + }, + { + "epoch": 0.12, + "learning_rate": 1.999555586201221e-05, + "loss": 1.3836, + "step": 1397 + }, + { + "epoch": 0.12, + "learning_rate": 1.999552872432397e-05, + "loss": 1.3264, + "step": 1398 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995501504049474e-05, + "loss": 1.5105, + "step": 1399 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995474201188947e-05, + "loss": 1.3484, + "step": 1400 + }, + { + "epoch": 0.12, + "learning_rate": 1.999544681574261e-05, + "loss": 1.333, + "step": 1401 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995419347710695e-05, + "loss": 1.508, + "step": 1402 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995391797093427e-05, + "loss": 1.4452, + "step": 1403 + }, + { + "epoch": 0.12, + "learning_rate": 1.999536416389103e-05, + "loss": 1.4976, + "step": 1404 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995336448103736e-05, + "loss": 1.2792, + "step": 1405 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995308649731774e-05, + "loss": 1.3367, + "step": 1406 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995280768775373e-05, + "loss": 1.444, + "step": 1407 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995252805234765e-05, + "loss": 1.274, + "step": 1408 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995224759110175e-05, + "loss": 1.4319, + "step": 1409 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995196630401845e-05, + "loss": 1.3585, + "step": 1410 + }, + { + "epoch": 0.12, + "learning_rate": 1.999516841911e-05, + "loss": 1.4173, + "step": 1411 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995140125234872e-05, + "loss": 1.5208, + "step": 1412 + }, + { + "epoch": 0.12, + "learning_rate": 1.99951117487767e-05, + "loss": 1.4164, + "step": 1413 + }, + { + "epoch": 0.12, + "learning_rate": 1.999508328973571e-05, + "loss": 1.3697, + "step": 1414 + }, + { + "epoch": 0.12, + "learning_rate": 1.999505474811215e-05, + "loss": 1.2805, + "step": 1415 + }, + { + "epoch": 0.12, + "learning_rate": 1.9995026123906248e-05, + "loss": 1.3545, + "step": 1416 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994997417118237e-05, + "loss": 1.4108, + "step": 1417 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994968627748362e-05, + "loss": 1.4642, + "step": 1418 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994939755796857e-05, + "loss": 1.3886, + "step": 1419 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994910801263964e-05, + "loss": 1.3597, + "step": 1420 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994881764149915e-05, + "loss": 1.4278, + "step": 1421 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994852644454954e-05, + "loss": 1.3626, + "step": 1422 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994823442179323e-05, + "loss": 1.3992, + "step": 1423 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994794157323262e-05, + "loss": 1.344, + "step": 1424 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994764789887014e-05, + "loss": 1.3532, + "step": 1425 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994735339870817e-05, + "loss": 1.4085, + "step": 1426 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994705807274922e-05, + "loss": 1.4983, + "step": 1427 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994676192099566e-05, + "loss": 1.3165, + "step": 1428 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994646494345e-05, + "loss": 1.4701, + "step": 1429 + }, + { + "epoch": 0.12, + "learning_rate": 1.999461671401146e-05, + "loss": 1.3499, + "step": 1430 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994586851099202e-05, + "loss": 1.378, + "step": 1431 + }, + { + "epoch": 0.12, + "learning_rate": 1.999455690560847e-05, + "loss": 1.3474, + "step": 1432 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994526877539507e-05, + "loss": 1.3448, + "step": 1433 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994496766892567e-05, + "loss": 1.4401, + "step": 1434 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994466573667893e-05, + "loss": 1.3932, + "step": 1435 + }, + { + "epoch": 0.12, + "learning_rate": 1.999443629786574e-05, + "loss": 1.3873, + "step": 1436 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994405939486356e-05, + "loss": 1.3886, + "step": 1437 + }, + { + "epoch": 0.12, + "learning_rate": 1.999437549852999e-05, + "loss": 1.3344, + "step": 1438 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994344974996894e-05, + "loss": 1.3168, + "step": 1439 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994314368887325e-05, + "loss": 1.4491, + "step": 1440 + }, + { + "epoch": 0.12, + "learning_rate": 1.999428368020153e-05, + "loss": 1.3935, + "step": 1441 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994252908939764e-05, + "loss": 1.3977, + "step": 1442 + }, + { + "epoch": 0.12, + "learning_rate": 1.999422205510228e-05, + "loss": 1.3986, + "step": 1443 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994191118689338e-05, + "loss": 1.427, + "step": 1444 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994160099701192e-05, + "loss": 1.3698, + "step": 1445 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994128998138094e-05, + "loss": 1.4075, + "step": 1446 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994097814000303e-05, + "loss": 1.448, + "step": 1447 + }, + { + "epoch": 0.12, + "learning_rate": 1.999406654728808e-05, + "loss": 1.3409, + "step": 1448 + }, + { + "epoch": 0.12, + "learning_rate": 1.999403519800168e-05, + "loss": 1.3306, + "step": 1449 + }, + { + "epoch": 0.12, + "learning_rate": 1.9994003766141362e-05, + "loss": 1.4013, + "step": 1450 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993972251707386e-05, + "loss": 1.3868, + "step": 1451 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993940654700013e-05, + "loss": 1.3898, + "step": 1452 + }, + { + "epoch": 0.12, + "learning_rate": 1.99939089751195e-05, + "loss": 1.3872, + "step": 1453 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993877212966118e-05, + "loss": 1.308, + "step": 1454 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993845368240125e-05, + "loss": 1.3875, + "step": 1455 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993813440941778e-05, + "loss": 1.4306, + "step": 1456 + }, + { + "epoch": 0.12, + "learning_rate": 1.999378143107135e-05, + "loss": 1.3554, + "step": 1457 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993749338629097e-05, + "loss": 1.482, + "step": 1458 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993717163615294e-05, + "loss": 1.4612, + "step": 1459 + }, + { + "epoch": 0.12, + "learning_rate": 1.99936849060302e-05, + "loss": 1.3832, + "step": 1460 + }, + { + "epoch": 0.12, + "learning_rate": 1.999365256587408e-05, + "loss": 1.4121, + "step": 1461 + }, + { + "epoch": 0.12, + "learning_rate": 1.999362014314721e-05, + "loss": 1.4677, + "step": 1462 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993587637849842e-05, + "loss": 1.4497, + "step": 1463 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993555049982264e-05, + "loss": 1.4132, + "step": 1464 + }, + { + "epoch": 0.12, + "learning_rate": 1.999352237954473e-05, + "loss": 1.2907, + "step": 1465 + }, + { + "epoch": 0.12, + "learning_rate": 1.999348962653752e-05, + "loss": 1.4061, + "step": 1466 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993456790960898e-05, + "loss": 1.4764, + "step": 1467 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993423872815136e-05, + "loss": 1.3614, + "step": 1468 + }, + { + "epoch": 0.12, + "learning_rate": 1.999339087210051e-05, + "loss": 1.4324, + "step": 1469 + }, + { + "epoch": 0.12, + "learning_rate": 1.999335778881729e-05, + "loss": 1.4034, + "step": 1470 + }, + { + "epoch": 0.12, + "learning_rate": 1.999332462296575e-05, + "loss": 1.3543, + "step": 1471 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993291374546164e-05, + "loss": 1.4661, + "step": 1472 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993258043558804e-05, + "loss": 1.3694, + "step": 1473 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993224630003952e-05, + "loss": 1.391, + "step": 1474 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993191133881878e-05, + "loss": 1.3702, + "step": 1475 + }, + { + "epoch": 0.12, + "learning_rate": 1.999315755519286e-05, + "loss": 1.462, + "step": 1476 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993123893937176e-05, + "loss": 1.3309, + "step": 1477 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993090150115105e-05, + "loss": 1.396, + "step": 1478 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993056323726922e-05, + "loss": 1.3409, + "step": 1479 + }, + { + "epoch": 0.12, + "learning_rate": 1.9993022414772915e-05, + "loss": 1.3846, + "step": 1480 + }, + { + "epoch": 0.12, + "learning_rate": 1.9992988423253355e-05, + "loss": 1.4004, + "step": 1481 + }, + { + "epoch": 0.12, + "learning_rate": 1.9992954349168527e-05, + "loss": 1.3601, + "step": 1482 + }, + { + "epoch": 0.12, + "learning_rate": 1.999292019251871e-05, + "loss": 1.41, + "step": 1483 + }, + { + "epoch": 0.12, + "learning_rate": 1.999288595330419e-05, + "loss": 1.4435, + "step": 1484 + }, + { + "epoch": 0.13, + "learning_rate": 1.999285163152525e-05, + "loss": 1.3623, + "step": 1485 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992817227182167e-05, + "loss": 1.3937, + "step": 1486 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992782740275232e-05, + "loss": 1.4876, + "step": 1487 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992748170804733e-05, + "loss": 1.4543, + "step": 1488 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992713518770944e-05, + "loss": 1.4327, + "step": 1489 + }, + { + "epoch": 0.13, + "learning_rate": 1.999267878417416e-05, + "loss": 1.3202, + "step": 1490 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992643967014666e-05, + "loss": 1.3852, + "step": 1491 + }, + { + "epoch": 0.13, + "learning_rate": 1.999260906729275e-05, + "loss": 1.4123, + "step": 1492 + }, + { + "epoch": 0.13, + "learning_rate": 1.99925740850087e-05, + "loss": 1.3911, + "step": 1493 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992539020162805e-05, + "loss": 1.4049, + "step": 1494 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992503872755353e-05, + "loss": 1.3018, + "step": 1495 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992468642786637e-05, + "loss": 1.4562, + "step": 1496 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992433330256946e-05, + "loss": 1.3686, + "step": 1497 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992397935166573e-05, + "loss": 1.445, + "step": 1498 + }, + { + "epoch": 0.13, + "learning_rate": 1.999236245751581e-05, + "loss": 1.3602, + "step": 1499 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992326897304952e-05, + "loss": 1.286, + "step": 1500 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992291254534292e-05, + "loss": 1.3277, + "step": 1501 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992255529204124e-05, + "loss": 1.4272, + "step": 1502 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992219721314742e-05, + "loss": 1.3932, + "step": 1503 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992183830866437e-05, + "loss": 1.479, + "step": 1504 + }, + { + "epoch": 0.13, + "learning_rate": 1.999214785785952e-05, + "loss": 1.4727, + "step": 1505 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992111802294274e-05, + "loss": 1.4761, + "step": 1506 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992075664171005e-05, + "loss": 1.4938, + "step": 1507 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992039443490006e-05, + "loss": 1.409, + "step": 1508 + }, + { + "epoch": 0.13, + "learning_rate": 1.9992003140251582e-05, + "loss": 1.3931, + "step": 1509 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991966754456032e-05, + "loss": 1.2905, + "step": 1510 + }, + { + "epoch": 0.13, + "learning_rate": 1.999193028610365e-05, + "loss": 1.472, + "step": 1511 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991893735194744e-05, + "loss": 1.3494, + "step": 1512 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991857101729616e-05, + "loss": 1.3559, + "step": 1513 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991820385708567e-05, + "loss": 1.3353, + "step": 1514 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991783587131896e-05, + "loss": 1.3925, + "step": 1515 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991746705999918e-05, + "loss": 1.3298, + "step": 1516 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991709742312927e-05, + "loss": 1.4909, + "step": 1517 + }, + { + "epoch": 0.13, + "learning_rate": 1.999167269607123e-05, + "loss": 1.352, + "step": 1518 + }, + { + "epoch": 0.13, + "learning_rate": 1.999163556727514e-05, + "loss": 1.336, + "step": 1519 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991598355924953e-05, + "loss": 1.412, + "step": 1520 + }, + { + "epoch": 0.13, + "learning_rate": 1.999156106202099e-05, + "loss": 1.382, + "step": 1521 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991523685563544e-05, + "loss": 1.4182, + "step": 1522 + }, + { + "epoch": 0.13, + "learning_rate": 1.999148622655294e-05, + "loss": 1.3442, + "step": 1523 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991448684989474e-05, + "loss": 1.3985, + "step": 1524 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991411060873463e-05, + "loss": 1.3993, + "step": 1525 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991373354205213e-05, + "loss": 1.3343, + "step": 1526 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991335564985045e-05, + "loss": 1.3513, + "step": 1527 + }, + { + "epoch": 0.13, + "learning_rate": 1.999129769321326e-05, + "loss": 1.355, + "step": 1528 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991259738890178e-05, + "loss": 1.4327, + "step": 1529 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991221702016106e-05, + "loss": 1.4053, + "step": 1530 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991183582591368e-05, + "loss": 1.4896, + "step": 1531 + }, + { + "epoch": 0.13, + "learning_rate": 1.999114538061627e-05, + "loss": 1.5044, + "step": 1532 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991107096091133e-05, + "loss": 1.3268, + "step": 1533 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991068729016275e-05, + "loss": 1.4052, + "step": 1534 + }, + { + "epoch": 0.13, + "learning_rate": 1.9991030279392003e-05, + "loss": 1.3277, + "step": 1535 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990991747218643e-05, + "loss": 1.4118, + "step": 1536 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990953132496516e-05, + "loss": 1.3115, + "step": 1537 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990914435225935e-05, + "loss": 1.3734, + "step": 1538 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990875655407217e-05, + "loss": 1.3951, + "step": 1539 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990836793040692e-05, + "loss": 1.3401, + "step": 1540 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990797848126674e-05, + "loss": 1.3882, + "step": 1541 + }, + { + "epoch": 0.13, + "learning_rate": 1.999075882066549e-05, + "loss": 1.416, + "step": 1542 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990719710657455e-05, + "loss": 1.3413, + "step": 1543 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990680518102895e-05, + "loss": 1.4817, + "step": 1544 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990641243002138e-05, + "loss": 1.4087, + "step": 1545 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990601885355504e-05, + "loss": 1.5011, + "step": 1546 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990562445163323e-05, + "loss": 1.3614, + "step": 1547 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990522922425915e-05, + "loss": 1.3948, + "step": 1548 + }, + { + "epoch": 0.13, + "learning_rate": 1.999048331714361e-05, + "loss": 1.5, + "step": 1549 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990443629316733e-05, + "loss": 1.3545, + "step": 1550 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990403858945618e-05, + "loss": 1.3608, + "step": 1551 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990364006030582e-05, + "loss": 1.2953, + "step": 1552 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990324070571966e-05, + "loss": 1.3641, + "step": 1553 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990284052570094e-05, + "loss": 1.3977, + "step": 1554 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990243952025298e-05, + "loss": 1.5373, + "step": 1555 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990203768937907e-05, + "loss": 1.3862, + "step": 1556 + }, + { + "epoch": 0.13, + "learning_rate": 1.999016350330826e-05, + "loss": 1.4178, + "step": 1557 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990123155136683e-05, + "loss": 1.4431, + "step": 1558 + }, + { + "epoch": 0.13, + "learning_rate": 1.999008272442351e-05, + "loss": 1.3302, + "step": 1559 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990042211169076e-05, + "loss": 1.3802, + "step": 1560 + }, + { + "epoch": 0.13, + "learning_rate": 1.9990001615373718e-05, + "loss": 1.3576, + "step": 1561 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989960937037768e-05, + "loss": 1.333, + "step": 1562 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989920176161563e-05, + "loss": 1.4231, + "step": 1563 + }, + { + "epoch": 0.13, + "learning_rate": 1.998987933274544e-05, + "loss": 1.3971, + "step": 1564 + }, + { + "epoch": 0.13, + "learning_rate": 1.998983840678974e-05, + "loss": 1.4421, + "step": 1565 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989797398294795e-05, + "loss": 1.3244, + "step": 1566 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989756307260947e-05, + "loss": 1.4174, + "step": 1567 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989715133688535e-05, + "loss": 1.3454, + "step": 1568 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989673877577903e-05, + "loss": 1.3344, + "step": 1569 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989632538929385e-05, + "loss": 1.39, + "step": 1570 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989591117743327e-05, + "loss": 1.4087, + "step": 1571 + }, + { + "epoch": 0.13, + "learning_rate": 1.998954961402007e-05, + "loss": 1.3916, + "step": 1572 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989508027759954e-05, + "loss": 1.3014, + "step": 1573 + }, + { + "epoch": 0.13, + "learning_rate": 1.998946635896333e-05, + "loss": 1.4353, + "step": 1574 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989424607630538e-05, + "loss": 1.3559, + "step": 1575 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989382773761917e-05, + "loss": 1.3737, + "step": 1576 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989340857357823e-05, + "loss": 1.4267, + "step": 1577 + }, + { + "epoch": 0.13, + "learning_rate": 1.99892988584186e-05, + "loss": 1.3606, + "step": 1578 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989256776944588e-05, + "loss": 1.3906, + "step": 1579 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989214612936144e-05, + "loss": 1.3445, + "step": 1580 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989172366393607e-05, + "loss": 1.4777, + "step": 1581 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989130037317334e-05, + "loss": 1.4226, + "step": 1582 + }, + { + "epoch": 0.13, + "learning_rate": 1.9989087625707672e-05, + "loss": 1.3153, + "step": 1583 + }, + { + "epoch": 0.13, + "learning_rate": 1.998904513156497e-05, + "loss": 1.3963, + "step": 1584 + }, + { + "epoch": 0.13, + "learning_rate": 1.998900255488958e-05, + "loss": 1.3482, + "step": 1585 + }, + { + "epoch": 0.13, + "learning_rate": 1.9988959895681853e-05, + "loss": 1.3999, + "step": 1586 + }, + { + "epoch": 0.13, + "learning_rate": 1.9988917153942147e-05, + "loss": 1.4309, + "step": 1587 + }, + { + "epoch": 0.13, + "learning_rate": 1.998887432967081e-05, + "loss": 1.4686, + "step": 1588 + }, + { + "epoch": 0.13, + "learning_rate": 1.998883142286819e-05, + "loss": 1.4141, + "step": 1589 + }, + { + "epoch": 0.13, + "learning_rate": 1.9988788433534655e-05, + "loss": 1.3804, + "step": 1590 + }, + { + "epoch": 0.13, + "learning_rate": 1.9988745361670546e-05, + "loss": 1.3495, + "step": 1591 + }, + { + "epoch": 0.13, + "learning_rate": 1.9988702207276232e-05, + "loss": 1.4045, + "step": 1592 + }, + { + "epoch": 0.13, + "learning_rate": 1.9988658970352063e-05, + "loss": 1.3944, + "step": 1593 + }, + { + "epoch": 0.13, + "learning_rate": 1.9988615650898394e-05, + "loss": 1.3931, + "step": 1594 + }, + { + "epoch": 0.13, + "learning_rate": 1.998857224891559e-05, + "loss": 1.2873, + "step": 1595 + }, + { + "epoch": 0.13, + "learning_rate": 1.9988528764404002e-05, + "loss": 1.2969, + "step": 1596 + }, + { + "epoch": 0.13, + "learning_rate": 1.9988485197363992e-05, + "loss": 1.3994, + "step": 1597 + }, + { + "epoch": 0.13, + "learning_rate": 1.998844154779592e-05, + "loss": 1.3568, + "step": 1598 + }, + { + "epoch": 0.13, + "learning_rate": 1.998839781570015e-05, + "loss": 1.3964, + "step": 1599 + }, + { + "epoch": 0.13, + "learning_rate": 1.998835400107704e-05, + "loss": 1.5319, + "step": 1600 + }, + { + "epoch": 0.13, + "learning_rate": 1.9988310103926952e-05, + "loss": 1.388, + "step": 1601 + }, + { + "epoch": 0.13, + "learning_rate": 1.998826612425025e-05, + "loss": 1.3046, + "step": 1602 + }, + { + "epoch": 0.13, + "learning_rate": 1.9988222062047294e-05, + "loss": 1.4222, + "step": 1603 + }, + { + "epoch": 0.14, + "learning_rate": 1.9988177917318457e-05, + "loss": 1.4136, + "step": 1604 + }, + { + "epoch": 0.14, + "learning_rate": 1.9988133690064094e-05, + "loss": 1.3926, + "step": 1605 + }, + { + "epoch": 0.14, + "learning_rate": 1.9988089380284572e-05, + "loss": 1.3275, + "step": 1606 + }, + { + "epoch": 0.14, + "learning_rate": 1.998804498798026e-05, + "loss": 1.329, + "step": 1607 + }, + { + "epoch": 0.14, + "learning_rate": 1.9988000513151525e-05, + "loss": 1.4643, + "step": 1608 + }, + { + "epoch": 0.14, + "learning_rate": 1.998795595579873e-05, + "loss": 1.474, + "step": 1609 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987911315922248e-05, + "loss": 1.3777, + "step": 1610 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987866593522443e-05, + "loss": 1.4297, + "step": 1611 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987821788599694e-05, + "loss": 1.3166, + "step": 1612 + }, + { + "epoch": 0.14, + "learning_rate": 1.998777690115436e-05, + "loss": 1.4683, + "step": 1613 + }, + { + "epoch": 0.14, + "learning_rate": 1.998773193118682e-05, + "loss": 1.4496, + "step": 1614 + }, + { + "epoch": 0.14, + "learning_rate": 1.998768687869744e-05, + "loss": 1.3049, + "step": 1615 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987641743686597e-05, + "loss": 1.3944, + "step": 1616 + }, + { + "epoch": 0.14, + "learning_rate": 1.998759652615466e-05, + "loss": 1.3258, + "step": 1617 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987551226102005e-05, + "loss": 1.3494, + "step": 1618 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987505843529004e-05, + "loss": 1.3105, + "step": 1619 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987460378436034e-05, + "loss": 1.4556, + "step": 1620 + }, + { + "epoch": 0.14, + "learning_rate": 1.998741483082347e-05, + "loss": 1.3598, + "step": 1621 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987369200691687e-05, + "loss": 1.4365, + "step": 1622 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987323488041062e-05, + "loss": 1.4327, + "step": 1623 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987277692871976e-05, + "loss": 1.3789, + "step": 1624 + }, + { + "epoch": 0.14, + "learning_rate": 1.998723181518481e-05, + "loss": 1.3927, + "step": 1625 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987185854979932e-05, + "loss": 1.3171, + "step": 1626 + }, + { + "epoch": 0.14, + "learning_rate": 1.998713981225773e-05, + "loss": 1.3017, + "step": 1627 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987093687018582e-05, + "loss": 1.3091, + "step": 1628 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987047479262872e-05, + "loss": 1.4783, + "step": 1629 + }, + { + "epoch": 0.14, + "learning_rate": 1.9987001188990977e-05, + "loss": 1.4834, + "step": 1630 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986954816203283e-05, + "loss": 1.4324, + "step": 1631 + }, + { + "epoch": 0.14, + "learning_rate": 1.998690836090017e-05, + "loss": 1.3093, + "step": 1632 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986861823082027e-05, + "loss": 1.3541, + "step": 1633 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986815202749236e-05, + "loss": 1.4119, + "step": 1634 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986768499902178e-05, + "loss": 1.3671, + "step": 1635 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986721714541243e-05, + "loss": 1.4133, + "step": 1636 + }, + { + "epoch": 0.14, + "learning_rate": 1.998667484666682e-05, + "loss": 1.4188, + "step": 1637 + }, + { + "epoch": 0.14, + "learning_rate": 1.998662789627929e-05, + "loss": 1.4219, + "step": 1638 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986580863379048e-05, + "loss": 1.3979, + "step": 1639 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986533747966477e-05, + "loss": 1.2592, + "step": 1640 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986486550041967e-05, + "loss": 1.3906, + "step": 1641 + }, + { + "epoch": 0.14, + "learning_rate": 1.998643926960591e-05, + "loss": 1.4641, + "step": 1642 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986391906658694e-05, + "loss": 1.3075, + "step": 1643 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986344461200714e-05, + "loss": 1.4027, + "step": 1644 + }, + { + "epoch": 0.14, + "learning_rate": 1.998629693323236e-05, + "loss": 1.4831, + "step": 1645 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986249322754023e-05, + "loss": 1.3979, + "step": 1646 + }, + { + "epoch": 0.14, + "learning_rate": 1.99862016297661e-05, + "loss": 1.3161, + "step": 1647 + }, + { + "epoch": 0.14, + "learning_rate": 1.998615385426898e-05, + "loss": 1.4066, + "step": 1648 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986105996263065e-05, + "loss": 1.3914, + "step": 1649 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986058055748744e-05, + "loss": 1.3476, + "step": 1650 + }, + { + "epoch": 0.14, + "learning_rate": 1.9986010032726418e-05, + "loss": 1.3879, + "step": 1651 + }, + { + "epoch": 0.14, + "learning_rate": 1.998596192719648e-05, + "loss": 1.4545, + "step": 1652 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985913739159327e-05, + "loss": 1.4404, + "step": 1653 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985865468615356e-05, + "loss": 1.4002, + "step": 1654 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985817115564975e-05, + "loss": 1.3967, + "step": 1655 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985768680008574e-05, + "loss": 1.3692, + "step": 1656 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985720161946556e-05, + "loss": 1.4341, + "step": 1657 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985671561379324e-05, + "loss": 1.3687, + "step": 1658 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985622878307275e-05, + "loss": 1.3984, + "step": 1659 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985574112730815e-05, + "loss": 1.3379, + "step": 1660 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985525264650347e-05, + "loss": 1.2753, + "step": 1661 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985476334066272e-05, + "loss": 1.4171, + "step": 1662 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985427320979e-05, + "loss": 1.3162, + "step": 1663 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985378225388925e-05, + "loss": 1.3531, + "step": 1664 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985329047296466e-05, + "loss": 1.4019, + "step": 1665 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985279786702015e-05, + "loss": 1.391, + "step": 1666 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985230443605993e-05, + "loss": 1.4002, + "step": 1667 + }, + { + "epoch": 0.14, + "learning_rate": 1.99851810180088e-05, + "loss": 1.2991, + "step": 1668 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985131509910843e-05, + "loss": 1.431, + "step": 1669 + }, + { + "epoch": 0.14, + "learning_rate": 1.998508191931254e-05, + "loss": 1.4399, + "step": 1670 + }, + { + "epoch": 0.14, + "learning_rate": 1.9985032246214287e-05, + "loss": 1.3585, + "step": 1671 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984982490616506e-05, + "loss": 1.4875, + "step": 1672 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984932652519603e-05, + "loss": 1.3853, + "step": 1673 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984882731923992e-05, + "loss": 1.3783, + "step": 1674 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984832728830085e-05, + "loss": 1.4191, + "step": 1675 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984782643238294e-05, + "loss": 1.4088, + "step": 1676 + }, + { + "epoch": 0.14, + "learning_rate": 1.998473247514903e-05, + "loss": 1.4637, + "step": 1677 + }, + { + "epoch": 0.14, + "learning_rate": 1.998468222456271e-05, + "loss": 1.392, + "step": 1678 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984631891479758e-05, + "loss": 1.4063, + "step": 1679 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984581475900575e-05, + "loss": 1.3633, + "step": 1680 + }, + { + "epoch": 0.14, + "learning_rate": 1.998453097782559e-05, + "loss": 1.3534, + "step": 1681 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984480397255207e-05, + "loss": 1.4907, + "step": 1682 + }, + { + "epoch": 0.14, + "learning_rate": 1.998442973418986e-05, + "loss": 1.4494, + "step": 1683 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984378988629952e-05, + "loss": 1.4439, + "step": 1684 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984328160575913e-05, + "loss": 1.3438, + "step": 1685 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984277250028156e-05, + "loss": 1.3665, + "step": 1686 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984226256987108e-05, + "loss": 1.3705, + "step": 1687 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984175181453186e-05, + "loss": 1.4423, + "step": 1688 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984124023426817e-05, + "loss": 1.3931, + "step": 1689 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984072782908414e-05, + "loss": 1.3103, + "step": 1690 + }, + { + "epoch": 0.14, + "learning_rate": 1.9984021459898407e-05, + "loss": 1.3588, + "step": 1691 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983970054397223e-05, + "loss": 1.3371, + "step": 1692 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983918566405283e-05, + "loss": 1.3865, + "step": 1693 + }, + { + "epoch": 0.14, + "learning_rate": 1.998386699592301e-05, + "loss": 1.4133, + "step": 1694 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983815342950838e-05, + "loss": 1.3004, + "step": 1695 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983763607489183e-05, + "loss": 1.3111, + "step": 1696 + }, + { + "epoch": 0.14, + "learning_rate": 1.998371178953848e-05, + "loss": 1.4492, + "step": 1697 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983659889099152e-05, + "loss": 1.4663, + "step": 1698 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983607906171634e-05, + "loss": 1.4002, + "step": 1699 + }, + { + "epoch": 0.14, + "learning_rate": 1.998355584075635e-05, + "loss": 1.3616, + "step": 1700 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983503692853734e-05, + "loss": 1.4045, + "step": 1701 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983451462464216e-05, + "loss": 1.3667, + "step": 1702 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983399149588227e-05, + "loss": 1.4223, + "step": 1703 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983346754226195e-05, + "loss": 1.3581, + "step": 1704 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983294276378562e-05, + "loss": 1.3442, + "step": 1705 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983241716045754e-05, + "loss": 1.4463, + "step": 1706 + }, + { + "epoch": 0.14, + "learning_rate": 1.998318907322821e-05, + "loss": 1.4838, + "step": 1707 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983136347926363e-05, + "loss": 1.3476, + "step": 1708 + }, + { + "epoch": 0.14, + "learning_rate": 1.9983083540140647e-05, + "loss": 1.319, + "step": 1709 + }, + { + "epoch": 0.14, + "learning_rate": 1.99830306498715e-05, + "loss": 1.3391, + "step": 1710 + }, + { + "epoch": 0.14, + "learning_rate": 1.9982977677119356e-05, + "loss": 1.4103, + "step": 1711 + }, + { + "epoch": 0.14, + "learning_rate": 1.998292462188466e-05, + "loss": 1.3553, + "step": 1712 + }, + { + "epoch": 0.14, + "learning_rate": 1.9982871484167843e-05, + "loss": 1.3816, + "step": 1713 + }, + { + "epoch": 0.14, + "learning_rate": 1.9982818263969348e-05, + "loss": 1.3341, + "step": 1714 + }, + { + "epoch": 0.14, + "learning_rate": 1.9982764961289614e-05, + "loss": 1.2968, + "step": 1715 + }, + { + "epoch": 0.14, + "learning_rate": 1.998271157612908e-05, + "loss": 1.3351, + "step": 1716 + }, + { + "epoch": 0.14, + "learning_rate": 1.998265810848819e-05, + "loss": 1.4496, + "step": 1717 + }, + { + "epoch": 0.14, + "learning_rate": 1.998260455836738e-05, + "loss": 1.3235, + "step": 1718 + }, + { + "epoch": 0.14, + "learning_rate": 1.99825509257671e-05, + "loss": 1.4447, + "step": 1719 + }, + { + "epoch": 0.14, + "learning_rate": 1.998249721068779e-05, + "loss": 1.382, + "step": 1720 + }, + { + "epoch": 0.14, + "learning_rate": 1.998244341312989e-05, + "loss": 1.4499, + "step": 1721 + }, + { + "epoch": 0.14, + "learning_rate": 1.998238953309385e-05, + "loss": 1.3428, + "step": 1722 + }, + { + "epoch": 0.15, + "learning_rate": 1.998233557058011e-05, + "loss": 1.2459, + "step": 1723 + }, + { + "epoch": 0.15, + "learning_rate": 1.998228152558912e-05, + "loss": 1.3623, + "step": 1724 + }, + { + "epoch": 0.15, + "learning_rate": 1.998222739812133e-05, + "loss": 1.3447, + "step": 1725 + }, + { + "epoch": 0.15, + "learning_rate": 1.9982173188177178e-05, + "loss": 1.3741, + "step": 1726 + }, + { + "epoch": 0.15, + "learning_rate": 1.9982118895757114e-05, + "loss": 1.3505, + "step": 1727 + }, + { + "epoch": 0.15, + "learning_rate": 1.9982064520861594e-05, + "loss": 1.3264, + "step": 1728 + }, + { + "epoch": 0.15, + "learning_rate": 1.998201006349106e-05, + "loss": 1.333, + "step": 1729 + }, + { + "epoch": 0.15, + "learning_rate": 1.9981955523645963e-05, + "loss": 1.429, + "step": 1730 + }, + { + "epoch": 0.15, + "learning_rate": 1.9981900901326756e-05, + "loss": 1.3266, + "step": 1731 + }, + { + "epoch": 0.15, + "learning_rate": 1.998184619653389e-05, + "loss": 1.4079, + "step": 1732 + }, + { + "epoch": 0.15, + "learning_rate": 1.9981791409267815e-05, + "loss": 1.3024, + "step": 1733 + }, + { + "epoch": 0.15, + "learning_rate": 1.9981736539528987e-05, + "loss": 1.3247, + "step": 1734 + }, + { + "epoch": 0.15, + "learning_rate": 1.9981681587317853e-05, + "loss": 1.327, + "step": 1735 + }, + { + "epoch": 0.15, + "learning_rate": 1.9981626552634872e-05, + "loss": 1.4798, + "step": 1736 + }, + { + "epoch": 0.15, + "learning_rate": 1.99815714354805e-05, + "loss": 1.2855, + "step": 1737 + }, + { + "epoch": 0.15, + "learning_rate": 1.998151623585519e-05, + "loss": 1.3786, + "step": 1738 + }, + { + "epoch": 0.15, + "learning_rate": 1.99814609537594e-05, + "loss": 1.4465, + "step": 1739 + }, + { + "epoch": 0.15, + "learning_rate": 1.9981405589193582e-05, + "loss": 1.4075, + "step": 1740 + }, + { + "epoch": 0.15, + "learning_rate": 1.9981350142158198e-05, + "loss": 1.4353, + "step": 1741 + }, + { + "epoch": 0.15, + "learning_rate": 1.9981294612653704e-05, + "loss": 1.4009, + "step": 1742 + }, + { + "epoch": 0.15, + "learning_rate": 1.9981239000680565e-05, + "loss": 1.3504, + "step": 1743 + }, + { + "epoch": 0.15, + "learning_rate": 1.9981183306239227e-05, + "loss": 1.3323, + "step": 1744 + }, + { + "epoch": 0.15, + "learning_rate": 1.998112752933016e-05, + "loss": 1.2907, + "step": 1745 + }, + { + "epoch": 0.15, + "learning_rate": 1.9981071669953826e-05, + "loss": 1.3541, + "step": 1746 + }, + { + "epoch": 0.15, + "learning_rate": 1.9981015728110682e-05, + "loss": 1.3497, + "step": 1747 + }, + { + "epoch": 0.15, + "learning_rate": 1.998095970380119e-05, + "loss": 1.3217, + "step": 1748 + }, + { + "epoch": 0.15, + "learning_rate": 1.9980903597025815e-05, + "loss": 1.3664, + "step": 1749 + }, + { + "epoch": 0.15, + "learning_rate": 1.998084740778502e-05, + "loss": 1.3763, + "step": 1750 + }, + { + "epoch": 0.15, + "learning_rate": 1.9980791136079274e-05, + "loss": 1.4319, + "step": 1751 + }, + { + "epoch": 0.15, + "learning_rate": 1.9980734781909032e-05, + "loss": 1.3775, + "step": 1752 + }, + { + "epoch": 0.15, + "learning_rate": 1.998067834527477e-05, + "loss": 1.3992, + "step": 1753 + }, + { + "epoch": 0.15, + "learning_rate": 1.9980621826176943e-05, + "loss": 1.2823, + "step": 1754 + }, + { + "epoch": 0.15, + "learning_rate": 1.9980565224616028e-05, + "loss": 1.3236, + "step": 1755 + }, + { + "epoch": 0.15, + "learning_rate": 1.9980508540592488e-05, + "loss": 1.4285, + "step": 1756 + }, + { + "epoch": 0.15, + "learning_rate": 1.9980451774106794e-05, + "loss": 1.3067, + "step": 1757 + }, + { + "epoch": 0.15, + "learning_rate": 1.9980394925159414e-05, + "loss": 1.3855, + "step": 1758 + }, + { + "epoch": 0.15, + "learning_rate": 1.9980337993750814e-05, + "loss": 1.292, + "step": 1759 + }, + { + "epoch": 0.15, + "learning_rate": 1.9980280979881467e-05, + "loss": 1.2625, + "step": 1760 + }, + { + "epoch": 0.15, + "learning_rate": 1.9980223883551846e-05, + "loss": 1.4413, + "step": 1761 + }, + { + "epoch": 0.15, + "learning_rate": 1.998016670476242e-05, + "loss": 1.3252, + "step": 1762 + }, + { + "epoch": 0.15, + "learning_rate": 1.9980109443513664e-05, + "loss": 1.3315, + "step": 1763 + }, + { + "epoch": 0.15, + "learning_rate": 1.9980052099806052e-05, + "loss": 1.3946, + "step": 1764 + }, + { + "epoch": 0.15, + "learning_rate": 1.997999467364005e-05, + "loss": 1.3612, + "step": 1765 + }, + { + "epoch": 0.15, + "learning_rate": 1.9979937165016143e-05, + "loss": 1.3195, + "step": 1766 + }, + { + "epoch": 0.15, + "learning_rate": 1.9979879573934798e-05, + "loss": 1.3122, + "step": 1767 + }, + { + "epoch": 0.15, + "learning_rate": 1.9979821900396496e-05, + "loss": 1.3509, + "step": 1768 + }, + { + "epoch": 0.15, + "learning_rate": 1.997976414440171e-05, + "loss": 1.4804, + "step": 1769 + }, + { + "epoch": 0.15, + "learning_rate": 1.997970630595092e-05, + "loss": 1.4884, + "step": 1770 + }, + { + "epoch": 0.15, + "learning_rate": 1.99796483850446e-05, + "loss": 1.4576, + "step": 1771 + }, + { + "epoch": 0.15, + "learning_rate": 1.997959038168323e-05, + "loss": 1.3543, + "step": 1772 + }, + { + "epoch": 0.15, + "learning_rate": 1.9979532295867295e-05, + "loss": 1.4639, + "step": 1773 + }, + { + "epoch": 0.15, + "learning_rate": 1.997947412759727e-05, + "loss": 1.3776, + "step": 1774 + }, + { + "epoch": 0.15, + "learning_rate": 1.9979415876873635e-05, + "loss": 1.3855, + "step": 1775 + }, + { + "epoch": 0.15, + "learning_rate": 1.9979357543696872e-05, + "loss": 1.3097, + "step": 1776 + }, + { + "epoch": 0.15, + "learning_rate": 1.9979299128067462e-05, + "loss": 1.3367, + "step": 1777 + }, + { + "epoch": 0.15, + "learning_rate": 1.997924062998589e-05, + "loss": 1.4349, + "step": 1778 + }, + { + "epoch": 0.15, + "learning_rate": 1.997918204945264e-05, + "loss": 1.4161, + "step": 1779 + }, + { + "epoch": 0.15, + "learning_rate": 1.997912338646819e-05, + "loss": 1.4305, + "step": 1780 + }, + { + "epoch": 0.15, + "learning_rate": 1.997906464103303e-05, + "loss": 1.3212, + "step": 1781 + }, + { + "epoch": 0.15, + "learning_rate": 1.9979005813147645e-05, + "loss": 1.4162, + "step": 1782 + }, + { + "epoch": 0.15, + "learning_rate": 1.9978946902812522e-05, + "loss": 1.4179, + "step": 1783 + }, + { + "epoch": 0.15, + "learning_rate": 1.9978887910028143e-05, + "loss": 1.4066, + "step": 1784 + }, + { + "epoch": 0.15, + "learning_rate": 1.9978828834795e-05, + "loss": 1.4137, + "step": 1785 + }, + { + "epoch": 0.15, + "learning_rate": 1.9978769677113582e-05, + "loss": 1.4013, + "step": 1786 + }, + { + "epoch": 0.15, + "learning_rate": 1.997871043698437e-05, + "loss": 1.394, + "step": 1787 + }, + { + "epoch": 0.15, + "learning_rate": 1.997865111440786e-05, + "loss": 1.2789, + "step": 1788 + }, + { + "epoch": 0.15, + "learning_rate": 1.9978591709384544e-05, + "loss": 1.2831, + "step": 1789 + }, + { + "epoch": 0.15, + "learning_rate": 1.9978532221914907e-05, + "loss": 1.4516, + "step": 1790 + }, + { + "epoch": 0.15, + "learning_rate": 1.9978472651999446e-05, + "loss": 1.3749, + "step": 1791 + }, + { + "epoch": 0.15, + "learning_rate": 1.9978412999638646e-05, + "loss": 1.3498, + "step": 1792 + }, + { + "epoch": 0.15, + "learning_rate": 1.997835326483301e-05, + "loss": 1.3788, + "step": 1793 + }, + { + "epoch": 0.15, + "learning_rate": 1.9978293447583023e-05, + "loss": 1.3222, + "step": 1794 + }, + { + "epoch": 0.15, + "learning_rate": 1.9978233547889182e-05, + "loss": 1.4236, + "step": 1795 + }, + { + "epoch": 0.15, + "learning_rate": 1.997817356575198e-05, + "loss": 1.3105, + "step": 1796 + }, + { + "epoch": 0.15, + "learning_rate": 1.997811350117192e-05, + "loss": 1.3762, + "step": 1797 + }, + { + "epoch": 0.15, + "learning_rate": 1.997805335414949e-05, + "loss": 1.3928, + "step": 1798 + }, + { + "epoch": 0.15, + "learning_rate": 1.997799312468519e-05, + "loss": 1.38, + "step": 1799 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977932812779516e-05, + "loss": 1.4091, + "step": 1800 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977872418432965e-05, + "loss": 1.3513, + "step": 1801 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977811941646043e-05, + "loss": 1.4562, + "step": 1802 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977751382419244e-05, + "loss": 1.3544, + "step": 1803 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977690740753073e-05, + "loss": 1.4592, + "step": 1804 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977630016648025e-05, + "loss": 1.4484, + "step": 1805 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977569210104603e-05, + "loss": 1.3938, + "step": 1806 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977508321123314e-05, + "loss": 1.488, + "step": 1807 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977447349704655e-05, + "loss": 1.3832, + "step": 1808 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977386295849138e-05, + "loss": 1.3696, + "step": 1809 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977325159557258e-05, + "loss": 1.3918, + "step": 1810 + }, + { + "epoch": 0.15, + "learning_rate": 1.997726394082952e-05, + "loss": 1.3466, + "step": 1811 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977202639666445e-05, + "loss": 1.4266, + "step": 1812 + }, + { + "epoch": 0.15, + "learning_rate": 1.997714125606852e-05, + "loss": 1.4496, + "step": 1813 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977079790036264e-05, + "loss": 1.3117, + "step": 1814 + }, + { + "epoch": 0.15, + "learning_rate": 1.9977018241570183e-05, + "loss": 1.3688, + "step": 1815 + }, + { + "epoch": 0.15, + "learning_rate": 1.9976956610670783e-05, + "loss": 1.4152, + "step": 1816 + }, + { + "epoch": 0.15, + "learning_rate": 1.9976894897338575e-05, + "loss": 1.3621, + "step": 1817 + }, + { + "epoch": 0.15, + "learning_rate": 1.997683310157407e-05, + "loss": 1.3237, + "step": 1818 + }, + { + "epoch": 0.15, + "learning_rate": 1.9976771223377773e-05, + "loss": 1.4013, + "step": 1819 + }, + { + "epoch": 0.15, + "learning_rate": 1.99767092627502e-05, + "loss": 1.4036, + "step": 1820 + }, + { + "epoch": 0.15, + "learning_rate": 1.9976647219691862e-05, + "loss": 1.3821, + "step": 1821 + }, + { + "epoch": 0.15, + "learning_rate": 1.9976585094203274e-05, + "loss": 1.3258, + "step": 1822 + }, + { + "epoch": 0.15, + "learning_rate": 1.9976522886284943e-05, + "loss": 1.3759, + "step": 1823 + }, + { + "epoch": 0.15, + "learning_rate": 1.997646059593739e-05, + "loss": 1.3303, + "step": 1824 + }, + { + "epoch": 0.15, + "learning_rate": 1.9976398223161125e-05, + "loss": 1.3886, + "step": 1825 + }, + { + "epoch": 0.15, + "learning_rate": 1.9976335767956668e-05, + "loss": 1.3779, + "step": 1826 + }, + { + "epoch": 0.15, + "learning_rate": 1.997627323032453e-05, + "loss": 1.4361, + "step": 1827 + }, + { + "epoch": 0.15, + "learning_rate": 1.9976210610265234e-05, + "loss": 1.3505, + "step": 1828 + }, + { + "epoch": 0.15, + "learning_rate": 1.9976147907779286e-05, + "loss": 1.4591, + "step": 1829 + }, + { + "epoch": 0.15, + "learning_rate": 1.9976085122867214e-05, + "loss": 1.3378, + "step": 1830 + }, + { + "epoch": 0.15, + "learning_rate": 1.9976022255529535e-05, + "loss": 1.4165, + "step": 1831 + }, + { + "epoch": 0.15, + "learning_rate": 1.997595930576677e-05, + "loss": 1.4156, + "step": 1832 + }, + { + "epoch": 0.15, + "learning_rate": 1.9975896273579433e-05, + "loss": 1.3948, + "step": 1833 + }, + { + "epoch": 0.15, + "learning_rate": 1.9975833158968046e-05, + "loss": 1.3604, + "step": 1834 + }, + { + "epoch": 0.15, + "learning_rate": 1.9975769961933134e-05, + "loss": 1.3812, + "step": 1835 + }, + { + "epoch": 0.15, + "learning_rate": 1.997570668247522e-05, + "loss": 1.3935, + "step": 1836 + }, + { + "epoch": 0.15, + "learning_rate": 1.9975643320594825e-05, + "loss": 1.2613, + "step": 1837 + }, + { + "epoch": 0.15, + "learning_rate": 1.997557987629247e-05, + "loss": 1.3926, + "step": 1838 + }, + { + "epoch": 0.15, + "learning_rate": 1.997551634956868e-05, + "loss": 1.4927, + "step": 1839 + }, + { + "epoch": 0.15, + "learning_rate": 1.997545274042399e-05, + "loss": 1.3551, + "step": 1840 + }, + { + "epoch": 0.15, + "learning_rate": 1.997538904885891e-05, + "loss": 1.4011, + "step": 1841 + }, + { + "epoch": 0.16, + "learning_rate": 1.9975325274873972e-05, + "loss": 1.4462, + "step": 1842 + }, + { + "epoch": 0.16, + "learning_rate": 1.997526141846971e-05, + "loss": 1.4097, + "step": 1843 + }, + { + "epoch": 0.16, + "learning_rate": 1.997519747964664e-05, + "loss": 1.3271, + "step": 1844 + }, + { + "epoch": 0.16, + "learning_rate": 1.9975133458405298e-05, + "loss": 1.4408, + "step": 1845 + }, + { + "epoch": 0.16, + "learning_rate": 1.997506935474621e-05, + "loss": 1.3915, + "step": 1846 + }, + { + "epoch": 0.16, + "learning_rate": 1.9975005168669906e-05, + "loss": 1.3529, + "step": 1847 + }, + { + "epoch": 0.16, + "learning_rate": 1.997494090017692e-05, + "loss": 1.3993, + "step": 1848 + }, + { + "epoch": 0.16, + "learning_rate": 1.997487654926778e-05, + "loss": 1.2867, + "step": 1849 + }, + { + "epoch": 0.16, + "learning_rate": 1.9974812115943015e-05, + "loss": 1.4142, + "step": 1850 + }, + { + "epoch": 0.16, + "learning_rate": 1.9974747600203158e-05, + "loss": 1.5404, + "step": 1851 + }, + { + "epoch": 0.16, + "learning_rate": 1.997468300204875e-05, + "loss": 1.4236, + "step": 1852 + }, + { + "epoch": 0.16, + "learning_rate": 1.9974618321480316e-05, + "loss": 1.4895, + "step": 1853 + }, + { + "epoch": 0.16, + "learning_rate": 1.9974553558498392e-05, + "loss": 1.3605, + "step": 1854 + }, + { + "epoch": 0.16, + "learning_rate": 1.9974488713103513e-05, + "loss": 1.3476, + "step": 1855 + }, + { + "epoch": 0.16, + "learning_rate": 1.997442378529622e-05, + "loss": 1.4081, + "step": 1856 + }, + { + "epoch": 0.16, + "learning_rate": 1.997435877507704e-05, + "loss": 1.4369, + "step": 1857 + }, + { + "epoch": 0.16, + "learning_rate": 1.997429368244652e-05, + "loss": 1.4784, + "step": 1858 + }, + { + "epoch": 0.16, + "learning_rate": 1.9974228507405193e-05, + "loss": 1.363, + "step": 1859 + }, + { + "epoch": 0.16, + "learning_rate": 1.9974163249953597e-05, + "loss": 1.3974, + "step": 1860 + }, + { + "epoch": 0.16, + "learning_rate": 1.997409791009227e-05, + "loss": 1.3882, + "step": 1861 + }, + { + "epoch": 0.16, + "learning_rate": 1.9974032487821758e-05, + "loss": 1.3782, + "step": 1862 + }, + { + "epoch": 0.16, + "learning_rate": 1.9973966983142596e-05, + "loss": 1.3985, + "step": 1863 + }, + { + "epoch": 0.16, + "learning_rate": 1.9973901396055325e-05, + "loss": 1.3312, + "step": 1864 + }, + { + "epoch": 0.16, + "learning_rate": 1.9973835726560492e-05, + "loss": 1.3524, + "step": 1865 + }, + { + "epoch": 0.16, + "learning_rate": 1.997376997465863e-05, + "loss": 1.3601, + "step": 1866 + }, + { + "epoch": 0.16, + "learning_rate": 1.9973704140350296e-05, + "loss": 1.3595, + "step": 1867 + }, + { + "epoch": 0.16, + "learning_rate": 1.9973638223636023e-05, + "loss": 1.3824, + "step": 1868 + }, + { + "epoch": 0.16, + "learning_rate": 1.997357222451636e-05, + "loss": 1.3489, + "step": 1869 + }, + { + "epoch": 0.16, + "learning_rate": 1.997350614299185e-05, + "loss": 1.3946, + "step": 1870 + }, + { + "epoch": 0.16, + "learning_rate": 1.997343997906304e-05, + "loss": 1.264, + "step": 1871 + }, + { + "epoch": 0.16, + "learning_rate": 1.9973373732730476e-05, + "loss": 1.4211, + "step": 1872 + }, + { + "epoch": 0.16, + "learning_rate": 1.9973307403994712e-05, + "loss": 1.4086, + "step": 1873 + }, + { + "epoch": 0.16, + "learning_rate": 1.9973240992856283e-05, + "loss": 1.3454, + "step": 1874 + }, + { + "epoch": 0.16, + "learning_rate": 1.997317449931575e-05, + "loss": 1.4382, + "step": 1875 + }, + { + "epoch": 0.16, + "learning_rate": 1.9973107923373654e-05, + "loss": 1.36, + "step": 1876 + }, + { + "epoch": 0.16, + "learning_rate": 1.997304126503055e-05, + "loss": 1.3828, + "step": 1877 + }, + { + "epoch": 0.16, + "learning_rate": 1.9972974524286986e-05, + "loss": 1.3669, + "step": 1878 + }, + { + "epoch": 0.16, + "learning_rate": 1.9972907701143522e-05, + "loss": 1.3868, + "step": 1879 + }, + { + "epoch": 0.16, + "learning_rate": 1.9972840795600696e-05, + "loss": 1.3888, + "step": 1880 + }, + { + "epoch": 0.16, + "learning_rate": 1.997277380765907e-05, + "loss": 1.4079, + "step": 1881 + }, + { + "epoch": 0.16, + "learning_rate": 1.997270673731919e-05, + "loss": 1.4587, + "step": 1882 + }, + { + "epoch": 0.16, + "learning_rate": 1.997263958458162e-05, + "loss": 1.32, + "step": 1883 + }, + { + "epoch": 0.16, + "learning_rate": 1.997257234944691e-05, + "loss": 1.348, + "step": 1884 + }, + { + "epoch": 0.16, + "learning_rate": 1.9972505031915618e-05, + "loss": 1.3498, + "step": 1885 + }, + { + "epoch": 0.16, + "learning_rate": 1.9972437631988294e-05, + "loss": 1.3472, + "step": 1886 + }, + { + "epoch": 0.16, + "learning_rate": 1.9972370149665497e-05, + "loss": 1.3944, + "step": 1887 + }, + { + "epoch": 0.16, + "learning_rate": 1.997230258494779e-05, + "loss": 1.4243, + "step": 1888 + }, + { + "epoch": 0.16, + "learning_rate": 1.9972234937835728e-05, + "loss": 1.4295, + "step": 1889 + }, + { + "epoch": 0.16, + "learning_rate": 1.9972167208329866e-05, + "loss": 1.5028, + "step": 1890 + }, + { + "epoch": 0.16, + "learning_rate": 1.9972099396430768e-05, + "loss": 1.3494, + "step": 1891 + }, + { + "epoch": 0.16, + "learning_rate": 1.9972031502138996e-05, + "loss": 1.2704, + "step": 1892 + }, + { + "epoch": 0.16, + "learning_rate": 1.9971963525455103e-05, + "loss": 1.2882, + "step": 1893 + }, + { + "epoch": 0.16, + "learning_rate": 1.9971895466379662e-05, + "loss": 1.4034, + "step": 1894 + }, + { + "epoch": 0.16, + "learning_rate": 1.9971827324913224e-05, + "loss": 1.3022, + "step": 1895 + }, + { + "epoch": 0.16, + "learning_rate": 1.9971759101056362e-05, + "loss": 1.3732, + "step": 1896 + }, + { + "epoch": 0.16, + "learning_rate": 1.9971690794809632e-05, + "loss": 1.3467, + "step": 1897 + }, + { + "epoch": 0.16, + "learning_rate": 1.9971622406173603e-05, + "loss": 1.3956, + "step": 1898 + }, + { + "epoch": 0.16, + "learning_rate": 1.9971553935148838e-05, + "loss": 1.3535, + "step": 1899 + }, + { + "epoch": 0.16, + "learning_rate": 1.99714853817359e-05, + "loss": 1.2977, + "step": 1900 + }, + { + "epoch": 0.16, + "learning_rate": 1.997141674593536e-05, + "loss": 1.3832, + "step": 1901 + }, + { + "epoch": 0.16, + "learning_rate": 1.9971348027747784e-05, + "loss": 1.5146, + "step": 1902 + }, + { + "epoch": 0.16, + "learning_rate": 1.997127922717374e-05, + "loss": 1.381, + "step": 1903 + }, + { + "epoch": 0.16, + "learning_rate": 1.9971210344213797e-05, + "loss": 1.3735, + "step": 1904 + }, + { + "epoch": 0.16, + "learning_rate": 1.997114137886852e-05, + "loss": 1.3938, + "step": 1905 + }, + { + "epoch": 0.16, + "learning_rate": 1.9971072331138487e-05, + "loss": 1.4621, + "step": 1906 + }, + { + "epoch": 0.16, + "learning_rate": 1.9971003201024256e-05, + "loss": 1.3685, + "step": 1907 + }, + { + "epoch": 0.16, + "learning_rate": 1.9970933988526412e-05, + "loss": 1.3567, + "step": 1908 + }, + { + "epoch": 0.16, + "learning_rate": 1.9970864693645516e-05, + "loss": 1.4924, + "step": 1909 + }, + { + "epoch": 0.16, + "learning_rate": 1.9970795316382146e-05, + "loss": 1.3378, + "step": 1910 + }, + { + "epoch": 0.16, + "learning_rate": 1.9970725856736872e-05, + "loss": 1.2846, + "step": 1911 + }, + { + "epoch": 0.16, + "learning_rate": 1.997065631471027e-05, + "loss": 1.4401, + "step": 1912 + }, + { + "epoch": 0.16, + "learning_rate": 1.997058669030292e-05, + "loss": 1.3007, + "step": 1913 + }, + { + "epoch": 0.16, + "learning_rate": 1.9970516983515388e-05, + "loss": 1.3333, + "step": 1914 + }, + { + "epoch": 0.16, + "learning_rate": 1.9970447194348254e-05, + "loss": 1.4167, + "step": 1915 + }, + { + "epoch": 0.16, + "learning_rate": 1.9970377322802096e-05, + "loss": 1.3517, + "step": 1916 + }, + { + "epoch": 0.16, + "learning_rate": 1.997030736887749e-05, + "loss": 1.341, + "step": 1917 + }, + { + "epoch": 0.16, + "learning_rate": 1.997023733257501e-05, + "loss": 1.4993, + "step": 1918 + }, + { + "epoch": 0.16, + "learning_rate": 1.9970167213895237e-05, + "loss": 1.3307, + "step": 1919 + }, + { + "epoch": 0.16, + "learning_rate": 1.9970097012838755e-05, + "loss": 1.3539, + "step": 1920 + }, + { + "epoch": 0.16, + "learning_rate": 1.9970026729406144e-05, + "loss": 1.4514, + "step": 1921 + }, + { + "epoch": 0.16, + "learning_rate": 1.9969956363597976e-05, + "loss": 1.3811, + "step": 1922 + }, + { + "epoch": 0.16, + "learning_rate": 1.9969885915414838e-05, + "loss": 1.4565, + "step": 1923 + }, + { + "epoch": 0.16, + "learning_rate": 1.996981538485731e-05, + "loss": 1.4779, + "step": 1924 + }, + { + "epoch": 0.16, + "learning_rate": 1.9969744771925983e-05, + "loss": 1.4331, + "step": 1925 + }, + { + "epoch": 0.16, + "learning_rate": 1.996967407662143e-05, + "loss": 1.311, + "step": 1926 + }, + { + "epoch": 0.16, + "learning_rate": 1.996960329894424e-05, + "loss": 1.3276, + "step": 1927 + }, + { + "epoch": 0.16, + "learning_rate": 1.9969532438894994e-05, + "loss": 1.2911, + "step": 1928 + }, + { + "epoch": 0.16, + "learning_rate": 1.9969461496474283e-05, + "loss": 1.3708, + "step": 1929 + }, + { + "epoch": 0.16, + "learning_rate": 1.9969390471682692e-05, + "loss": 1.2179, + "step": 1930 + }, + { + "epoch": 0.16, + "learning_rate": 1.9969319364520806e-05, + "loss": 1.416, + "step": 1931 + }, + { + "epoch": 0.16, + "learning_rate": 1.996924817498921e-05, + "loss": 1.3869, + "step": 1932 + }, + { + "epoch": 0.16, + "learning_rate": 1.9969176903088496e-05, + "loss": 1.3569, + "step": 1933 + }, + { + "epoch": 0.16, + "learning_rate": 1.9969105548819252e-05, + "loss": 1.4096, + "step": 1934 + }, + { + "epoch": 0.16, + "learning_rate": 1.9969034112182068e-05, + "loss": 1.4091, + "step": 1935 + }, + { + "epoch": 0.16, + "learning_rate": 1.9968962593177533e-05, + "loss": 1.4435, + "step": 1936 + }, + { + "epoch": 0.16, + "learning_rate": 1.9968890991806237e-05, + "loss": 1.3649, + "step": 1937 + }, + { + "epoch": 0.16, + "learning_rate": 1.9968819308068776e-05, + "loss": 1.3294, + "step": 1938 + }, + { + "epoch": 0.16, + "learning_rate": 1.996874754196574e-05, + "loss": 1.4027, + "step": 1939 + }, + { + "epoch": 0.16, + "learning_rate": 1.9968675693497718e-05, + "loss": 1.4309, + "step": 1940 + }, + { + "epoch": 0.16, + "learning_rate": 1.996860376266531e-05, + "loss": 1.3004, + "step": 1941 + }, + { + "epoch": 0.16, + "learning_rate": 1.9968531749469106e-05, + "loss": 1.385, + "step": 1942 + }, + { + "epoch": 0.16, + "learning_rate": 1.9968459653909704e-05, + "loss": 1.3165, + "step": 1943 + }, + { + "epoch": 0.16, + "learning_rate": 1.9968387475987696e-05, + "loss": 1.392, + "step": 1944 + }, + { + "epoch": 0.16, + "learning_rate": 1.9968315215703683e-05, + "loss": 1.3655, + "step": 1945 + }, + { + "epoch": 0.16, + "learning_rate": 1.9968242873058256e-05, + "loss": 1.3406, + "step": 1946 + }, + { + "epoch": 0.16, + "learning_rate": 1.9968170448052017e-05, + "loss": 1.4585, + "step": 1947 + }, + { + "epoch": 0.16, + "learning_rate": 1.9968097940685563e-05, + "loss": 1.3861, + "step": 1948 + }, + { + "epoch": 0.16, + "learning_rate": 1.9968025350959497e-05, + "loss": 1.434, + "step": 1949 + }, + { + "epoch": 0.16, + "learning_rate": 1.9967952678874415e-05, + "loss": 1.279, + "step": 1950 + }, + { + "epoch": 0.16, + "learning_rate": 1.9967879924430913e-05, + "loss": 1.4422, + "step": 1951 + }, + { + "epoch": 0.16, + "learning_rate": 1.9967807087629602e-05, + "loss": 1.368, + "step": 1952 + }, + { + "epoch": 0.16, + "learning_rate": 1.9967734168471077e-05, + "loss": 1.4678, + "step": 1953 + }, + { + "epoch": 0.16, + "learning_rate": 1.9967661166955942e-05, + "loss": 1.4858, + "step": 1954 + }, + { + "epoch": 0.16, + "learning_rate": 1.9967588083084797e-05, + "loss": 1.4133, + "step": 1955 + }, + { + "epoch": 0.16, + "learning_rate": 1.9967514916858254e-05, + "loss": 1.3255, + "step": 1956 + }, + { + "epoch": 0.16, + "learning_rate": 1.996744166827691e-05, + "loss": 1.3856, + "step": 1957 + }, + { + "epoch": 0.16, + "learning_rate": 1.9967368337341375e-05, + "loss": 1.3911, + "step": 1958 + }, + { + "epoch": 0.16, + "learning_rate": 1.9967294924052252e-05, + "loss": 1.457, + "step": 1959 + }, + { + "epoch": 0.17, + "learning_rate": 1.9967221428410147e-05, + "loss": 1.3578, + "step": 1960 + }, + { + "epoch": 0.17, + "learning_rate": 1.996714785041567e-05, + "loss": 1.4467, + "step": 1961 + }, + { + "epoch": 0.17, + "learning_rate": 1.996707419006943e-05, + "loss": 1.3368, + "step": 1962 + }, + { + "epoch": 0.17, + "learning_rate": 1.9967000447372032e-05, + "loss": 1.3227, + "step": 1963 + }, + { + "epoch": 0.17, + "learning_rate": 1.9966926622324083e-05, + "loss": 1.3622, + "step": 1964 + }, + { + "epoch": 0.17, + "learning_rate": 1.9966852714926198e-05, + "loss": 1.4626, + "step": 1965 + }, + { + "epoch": 0.17, + "learning_rate": 1.9966778725178987e-05, + "loss": 1.4305, + "step": 1966 + }, + { + "epoch": 0.17, + "learning_rate": 1.996670465308306e-05, + "loss": 1.4216, + "step": 1967 + }, + { + "epoch": 0.17, + "learning_rate": 1.996663049863903e-05, + "loss": 1.3692, + "step": 1968 + }, + { + "epoch": 0.17, + "learning_rate": 1.996655626184751e-05, + "loss": 1.3948, + "step": 1969 + }, + { + "epoch": 0.17, + "learning_rate": 1.996648194270911e-05, + "loss": 1.3068, + "step": 1970 + }, + { + "epoch": 0.17, + "learning_rate": 1.996640754122445e-05, + "loss": 1.3667, + "step": 1971 + }, + { + "epoch": 0.17, + "learning_rate": 1.9966333057394135e-05, + "loss": 1.3288, + "step": 1972 + }, + { + "epoch": 0.17, + "learning_rate": 1.996625849121879e-05, + "loss": 1.417, + "step": 1973 + }, + { + "epoch": 0.17, + "learning_rate": 1.996618384269903e-05, + "loss": 1.4055, + "step": 1974 + }, + { + "epoch": 0.17, + "learning_rate": 1.9966109111835467e-05, + "loss": 1.3842, + "step": 1975 + }, + { + "epoch": 0.17, + "learning_rate": 1.996603429862872e-05, + "loss": 1.2981, + "step": 1976 + }, + { + "epoch": 0.17, + "learning_rate": 1.996595940307941e-05, + "loss": 1.3123, + "step": 1977 + }, + { + "epoch": 0.17, + "learning_rate": 1.9965884425188155e-05, + "loss": 1.3555, + "step": 1978 + }, + { + "epoch": 0.17, + "learning_rate": 1.996580936495557e-05, + "loss": 1.4117, + "step": 1979 + }, + { + "epoch": 0.17, + "learning_rate": 1.9965734222382284e-05, + "loss": 1.4369, + "step": 1980 + }, + { + "epoch": 0.17, + "learning_rate": 1.996565899746891e-05, + "loss": 1.3884, + "step": 1981 + }, + { + "epoch": 0.17, + "learning_rate": 1.9965583690216067e-05, + "loss": 1.3349, + "step": 1982 + }, + { + "epoch": 0.17, + "learning_rate": 1.996550830062439e-05, + "loss": 1.4093, + "step": 1983 + }, + { + "epoch": 0.17, + "learning_rate": 1.996543282869449e-05, + "loss": 1.3916, + "step": 1984 + }, + { + "epoch": 0.17, + "learning_rate": 1.9965357274426992e-05, + "loss": 1.4211, + "step": 1985 + }, + { + "epoch": 0.17, + "learning_rate": 1.9965281637822526e-05, + "loss": 1.3755, + "step": 1986 + }, + { + "epoch": 0.17, + "learning_rate": 1.9965205918881714e-05, + "loss": 1.5113, + "step": 1987 + }, + { + "epoch": 0.17, + "learning_rate": 1.996513011760518e-05, + "loss": 1.3617, + "step": 1988 + }, + { + "epoch": 0.17, + "learning_rate": 1.9965054233993554e-05, + "loss": 1.4163, + "step": 1989 + }, + { + "epoch": 0.17, + "learning_rate": 1.9964978268047458e-05, + "loss": 1.4275, + "step": 1990 + }, + { + "epoch": 0.17, + "learning_rate": 1.9964902219767524e-05, + "loss": 1.4248, + "step": 1991 + }, + { + "epoch": 0.17, + "learning_rate": 1.9964826089154375e-05, + "loss": 1.3132, + "step": 1992 + }, + { + "epoch": 0.17, + "learning_rate": 1.9964749876208648e-05, + "loss": 1.2786, + "step": 1993 + }, + { + "epoch": 0.17, + "learning_rate": 1.996467358093097e-05, + "loss": 1.2836, + "step": 1994 + }, + { + "epoch": 0.17, + "learning_rate": 1.9964597203321965e-05, + "loss": 1.2902, + "step": 1995 + }, + { + "epoch": 0.17, + "learning_rate": 1.9964520743382274e-05, + "loss": 1.3804, + "step": 1996 + }, + { + "epoch": 0.17, + "learning_rate": 1.9964444201112522e-05, + "loss": 1.3182, + "step": 1997 + }, + { + "epoch": 0.17, + "learning_rate": 1.996436757651334e-05, + "loss": 1.4774, + "step": 1998 + }, + { + "epoch": 0.17, + "learning_rate": 1.9964290869585366e-05, + "loss": 1.3542, + "step": 1999 + }, + { + "epoch": 0.17, + "learning_rate": 1.996421408032923e-05, + "loss": 1.3049, + "step": 2000 + }, + { + "epoch": 0.17, + "learning_rate": 1.996413720874557e-05, + "loss": 1.3479, + "step": 2001 + }, + { + "epoch": 0.17, + "learning_rate": 1.9964060254835017e-05, + "loss": 1.429, + "step": 2002 + }, + { + "epoch": 0.17, + "learning_rate": 1.996398321859821e-05, + "loss": 1.3009, + "step": 2003 + }, + { + "epoch": 0.17, + "learning_rate": 1.9963906100035787e-05, + "loss": 1.4041, + "step": 2004 + }, + { + "epoch": 0.17, + "learning_rate": 1.9963828899148384e-05, + "loss": 1.4153, + "step": 2005 + }, + { + "epoch": 0.17, + "learning_rate": 1.9963751615936636e-05, + "loss": 1.431, + "step": 2006 + }, + { + "epoch": 0.17, + "learning_rate": 1.9963674250401183e-05, + "loss": 1.4238, + "step": 2007 + }, + { + "epoch": 0.17, + "learning_rate": 1.9963596802542662e-05, + "loss": 1.4063, + "step": 2008 + }, + { + "epoch": 0.17, + "learning_rate": 1.996351927236172e-05, + "loss": 1.3746, + "step": 2009 + }, + { + "epoch": 0.17, + "learning_rate": 1.9963441659858993e-05, + "loss": 1.2697, + "step": 2010 + }, + { + "epoch": 0.17, + "learning_rate": 1.996336396503512e-05, + "loss": 1.3991, + "step": 2011 + }, + { + "epoch": 0.17, + "learning_rate": 1.9963286187890747e-05, + "loss": 1.3211, + "step": 2012 + }, + { + "epoch": 0.17, + "learning_rate": 1.9963208328426513e-05, + "loss": 1.3609, + "step": 2013 + }, + { + "epoch": 0.17, + "learning_rate": 1.9963130386643063e-05, + "loss": 1.3702, + "step": 2014 + }, + { + "epoch": 0.17, + "learning_rate": 1.9963052362541046e-05, + "loss": 1.3086, + "step": 2015 + }, + { + "epoch": 0.17, + "learning_rate": 1.9962974256121093e-05, + "loss": 1.5371, + "step": 2016 + }, + { + "epoch": 0.17, + "learning_rate": 1.9962896067383867e-05, + "loss": 1.3859, + "step": 2017 + }, + { + "epoch": 0.17, + "learning_rate": 1.996281779633e-05, + "loss": 1.3766, + "step": 2018 + }, + { + "epoch": 0.17, + "learning_rate": 1.9962739442960143e-05, + "loss": 1.413, + "step": 2019 + }, + { + "epoch": 0.17, + "learning_rate": 1.9962661007274947e-05, + "loss": 1.4817, + "step": 2020 + }, + { + "epoch": 0.17, + "learning_rate": 1.9962582489275058e-05, + "loss": 1.3899, + "step": 2021 + }, + { + "epoch": 0.17, + "learning_rate": 1.996250388896112e-05, + "loss": 1.3405, + "step": 2022 + }, + { + "epoch": 0.17, + "learning_rate": 1.996242520633379e-05, + "loss": 1.3136, + "step": 2023 + }, + { + "epoch": 0.17, + "learning_rate": 1.9962346441393713e-05, + "loss": 1.3693, + "step": 2024 + }, + { + "epoch": 0.17, + "learning_rate": 1.9962267594141544e-05, + "loss": 1.4294, + "step": 2025 + }, + { + "epoch": 0.17, + "learning_rate": 1.996218866457793e-05, + "loss": 1.4192, + "step": 2026 + }, + { + "epoch": 0.17, + "learning_rate": 1.9962109652703522e-05, + "loss": 1.4188, + "step": 2027 + }, + { + "epoch": 0.17, + "learning_rate": 1.996203055851898e-05, + "loss": 1.429, + "step": 2028 + }, + { + "epoch": 0.17, + "learning_rate": 1.996195138202495e-05, + "loss": 1.3523, + "step": 2029 + }, + { + "epoch": 0.17, + "learning_rate": 1.9961872123222094e-05, + "loss": 1.3199, + "step": 2030 + }, + { + "epoch": 0.17, + "learning_rate": 1.9961792782111057e-05, + "loss": 1.3908, + "step": 2031 + }, + { + "epoch": 0.17, + "learning_rate": 1.99617133586925e-05, + "loss": 1.3145, + "step": 2032 + }, + { + "epoch": 0.17, + "learning_rate": 1.9961633852967086e-05, + "loss": 1.4074, + "step": 2033 + }, + { + "epoch": 0.17, + "learning_rate": 1.996155426493546e-05, + "loss": 1.4111, + "step": 2034 + }, + { + "epoch": 0.17, + "learning_rate": 1.9961474594598285e-05, + "loss": 1.3245, + "step": 2035 + }, + { + "epoch": 0.17, + "learning_rate": 1.996139484195622e-05, + "loss": 1.4067, + "step": 2036 + }, + { + "epoch": 0.17, + "learning_rate": 1.9961315007009922e-05, + "loss": 1.3672, + "step": 2037 + }, + { + "epoch": 0.17, + "learning_rate": 1.9961235089760048e-05, + "loss": 1.2667, + "step": 2038 + }, + { + "epoch": 0.17, + "learning_rate": 1.9961155090207265e-05, + "loss": 1.4379, + "step": 2039 + }, + { + "epoch": 0.17, + "learning_rate": 1.9961075008352232e-05, + "loss": 1.4005, + "step": 2040 + }, + { + "epoch": 0.17, + "learning_rate": 1.9960994844195608e-05, + "loss": 1.3743, + "step": 2041 + }, + { + "epoch": 0.17, + "learning_rate": 1.9960914597738055e-05, + "loss": 1.3735, + "step": 2042 + }, + { + "epoch": 0.17, + "learning_rate": 1.996083426898024e-05, + "loss": 1.4297, + "step": 2043 + }, + { + "epoch": 0.17, + "learning_rate": 1.996075385792282e-05, + "loss": 1.4248, + "step": 2044 + }, + { + "epoch": 0.17, + "learning_rate": 1.9960673364566467e-05, + "loss": 1.4132, + "step": 2045 + }, + { + "epoch": 0.17, + "learning_rate": 1.9960592788911845e-05, + "loss": 1.4208, + "step": 2046 + }, + { + "epoch": 0.17, + "learning_rate": 1.9960512130959614e-05, + "loss": 1.4271, + "step": 2047 + }, + { + "epoch": 0.17, + "learning_rate": 1.9960431390710445e-05, + "loss": 1.4041, + "step": 2048 + }, + { + "epoch": 0.17, + "learning_rate": 1.9960350568165004e-05, + "loss": 1.4794, + "step": 2049 + }, + { + "epoch": 0.17, + "learning_rate": 1.996026966332396e-05, + "loss": 1.3556, + "step": 2050 + }, + { + "epoch": 0.17, + "learning_rate": 1.996018867618798e-05, + "loss": 1.3082, + "step": 2051 + }, + { + "epoch": 0.17, + "learning_rate": 1.9960107606757735e-05, + "loss": 1.3512, + "step": 2052 + }, + { + "epoch": 0.17, + "learning_rate": 1.9960026455033892e-05, + "loss": 1.3642, + "step": 2053 + }, + { + "epoch": 0.17, + "learning_rate": 1.995994522101712e-05, + "loss": 1.5717, + "step": 2054 + }, + { + "epoch": 0.17, + "learning_rate": 1.99598639047081e-05, + "loss": 1.4545, + "step": 2055 + }, + { + "epoch": 0.17, + "learning_rate": 1.995978250610749e-05, + "loss": 1.3597, + "step": 2056 + }, + { + "epoch": 0.17, + "learning_rate": 1.995970102521597e-05, + "loss": 1.3143, + "step": 2057 + }, + { + "epoch": 0.17, + "learning_rate": 1.9959619462034216e-05, + "loss": 1.3761, + "step": 2058 + }, + { + "epoch": 0.17, + "learning_rate": 1.9959537816562897e-05, + "loss": 1.351, + "step": 2059 + }, + { + "epoch": 0.17, + "learning_rate": 1.9959456088802687e-05, + "loss": 1.3939, + "step": 2060 + }, + { + "epoch": 0.17, + "learning_rate": 1.9959374278754267e-05, + "loss": 1.3873, + "step": 2061 + }, + { + "epoch": 0.17, + "learning_rate": 1.9959292386418306e-05, + "loss": 1.3789, + "step": 2062 + }, + { + "epoch": 0.17, + "learning_rate": 1.995921041179549e-05, + "loss": 1.4268, + "step": 2063 + }, + { + "epoch": 0.17, + "learning_rate": 1.9959128354886483e-05, + "loss": 1.422, + "step": 2064 + }, + { + "epoch": 0.17, + "learning_rate": 1.9959046215691972e-05, + "loss": 1.4187, + "step": 2065 + }, + { + "epoch": 0.17, + "learning_rate": 1.9958963994212635e-05, + "loss": 1.3569, + "step": 2066 + }, + { + "epoch": 0.17, + "learning_rate": 1.995888169044915e-05, + "loss": 1.4112, + "step": 2067 + }, + { + "epoch": 0.17, + "learning_rate": 1.9958799304402198e-05, + "loss": 1.3406, + "step": 2068 + }, + { + "epoch": 0.17, + "learning_rate": 1.995871683607246e-05, + "loss": 1.3751, + "step": 2069 + }, + { + "epoch": 0.17, + "learning_rate": 1.9958634285460615e-05, + "loss": 1.4352, + "step": 2070 + }, + { + "epoch": 0.17, + "learning_rate": 1.9958551652567344e-05, + "loss": 1.4184, + "step": 2071 + }, + { + "epoch": 0.17, + "learning_rate": 1.9958468937393332e-05, + "loss": 1.3683, + "step": 2072 + }, + { + "epoch": 0.17, + "learning_rate": 1.9958386139939266e-05, + "loss": 1.3517, + "step": 2073 + }, + { + "epoch": 0.17, + "learning_rate": 1.9958303260205824e-05, + "loss": 1.3849, + "step": 2074 + }, + { + "epoch": 0.17, + "learning_rate": 1.9958220298193696e-05, + "loss": 1.3233, + "step": 2075 + }, + { + "epoch": 0.17, + "learning_rate": 1.9958137253903563e-05, + "loss": 1.4452, + "step": 2076 + }, + { + "epoch": 0.17, + "learning_rate": 1.995805412733611e-05, + "loss": 1.4387, + "step": 2077 + }, + { + "epoch": 0.17, + "learning_rate": 1.9957970918492034e-05, + "loss": 1.4282, + "step": 2078 + }, + { + "epoch": 0.18, + "learning_rate": 1.995788762737201e-05, + "loss": 1.3666, + "step": 2079 + }, + { + "epoch": 0.18, + "learning_rate": 1.9957804253976732e-05, + "loss": 1.3785, + "step": 2080 + }, + { + "epoch": 0.18, + "learning_rate": 1.995772079830689e-05, + "loss": 1.3953, + "step": 2081 + }, + { + "epoch": 0.18, + "learning_rate": 1.995763726036317e-05, + "loss": 1.3398, + "step": 2082 + }, + { + "epoch": 0.18, + "learning_rate": 1.9957553640146266e-05, + "loss": 1.4614, + "step": 2083 + }, + { + "epoch": 0.18, + "learning_rate": 1.995746993765687e-05, + "loss": 1.2326, + "step": 2084 + }, + { + "epoch": 0.18, + "learning_rate": 1.9957386152895662e-05, + "loss": 1.2995, + "step": 2085 + }, + { + "epoch": 0.18, + "learning_rate": 1.995730228586335e-05, + "loss": 1.325, + "step": 2086 + }, + { + "epoch": 0.18, + "learning_rate": 1.995721833656062e-05, + "loss": 1.4324, + "step": 2087 + }, + { + "epoch": 0.18, + "learning_rate": 1.995713430498816e-05, + "loss": 1.3638, + "step": 2088 + }, + { + "epoch": 0.18, + "learning_rate": 1.9957050191146675e-05, + "loss": 1.3621, + "step": 2089 + }, + { + "epoch": 0.18, + "learning_rate": 1.995696599503685e-05, + "loss": 1.3512, + "step": 2090 + }, + { + "epoch": 0.18, + "learning_rate": 1.9956881716659388e-05, + "loss": 1.4367, + "step": 2091 + }, + { + "epoch": 0.18, + "learning_rate": 1.9956797356014982e-05, + "loss": 1.3951, + "step": 2092 + }, + { + "epoch": 0.18, + "learning_rate": 1.995671291310433e-05, + "loss": 1.3483, + "step": 2093 + }, + { + "epoch": 0.18, + "learning_rate": 1.9956628387928128e-05, + "loss": 1.4926, + "step": 2094 + }, + { + "epoch": 0.18, + "learning_rate": 1.9956543780487076e-05, + "loss": 1.3766, + "step": 2095 + }, + { + "epoch": 0.18, + "learning_rate": 1.9956459090781877e-05, + "loss": 1.4014, + "step": 2096 + }, + { + "epoch": 0.18, + "learning_rate": 1.9956374318813223e-05, + "loss": 1.4009, + "step": 2097 + }, + { + "epoch": 0.18, + "learning_rate": 1.9956289464581817e-05, + "loss": 1.3549, + "step": 2098 + }, + { + "epoch": 0.18, + "learning_rate": 1.995620452808836e-05, + "loss": 1.4323, + "step": 2099 + }, + { + "epoch": 0.18, + "learning_rate": 1.995611950933356e-05, + "loss": 1.4181, + "step": 2100 + }, + { + "epoch": 0.18, + "learning_rate": 1.995603440831811e-05, + "loss": 1.439, + "step": 2101 + }, + { + "epoch": 0.18, + "learning_rate": 1.995594922504272e-05, + "loss": 1.3517, + "step": 2102 + }, + { + "epoch": 0.18, + "learning_rate": 1.995586395950809e-05, + "loss": 1.4477, + "step": 2103 + }, + { + "epoch": 0.18, + "learning_rate": 1.9955778611714925e-05, + "loss": 1.427, + "step": 2104 + }, + { + "epoch": 0.18, + "learning_rate": 1.9955693181663932e-05, + "loss": 1.3621, + "step": 2105 + }, + { + "epoch": 0.18, + "learning_rate": 1.995560766935581e-05, + "loss": 1.4456, + "step": 2106 + }, + { + "epoch": 0.18, + "learning_rate": 1.9955522074791277e-05, + "loss": 1.4506, + "step": 2107 + }, + { + "epoch": 0.18, + "learning_rate": 1.9955436397971035e-05, + "loss": 1.436, + "step": 2108 + }, + { + "epoch": 0.18, + "learning_rate": 1.995535063889579e-05, + "loss": 1.4022, + "step": 2109 + }, + { + "epoch": 0.18, + "learning_rate": 1.995526479756625e-05, + "loss": 1.4276, + "step": 2110 + }, + { + "epoch": 0.18, + "learning_rate": 1.995517887398313e-05, + "loss": 1.3405, + "step": 2111 + }, + { + "epoch": 0.18, + "learning_rate": 1.9955092868147134e-05, + "loss": 1.3441, + "step": 2112 + }, + { + "epoch": 0.18, + "learning_rate": 1.9955006780058978e-05, + "loss": 1.3687, + "step": 2113 + }, + { + "epoch": 0.18, + "learning_rate": 1.9954920609719366e-05, + "loss": 1.3632, + "step": 2114 + }, + { + "epoch": 0.18, + "learning_rate": 1.9954834357129013e-05, + "loss": 1.4265, + "step": 2115 + }, + { + "epoch": 0.18, + "learning_rate": 1.9954748022288634e-05, + "loss": 1.3695, + "step": 2116 + }, + { + "epoch": 0.18, + "learning_rate": 1.9954661605198943e-05, + "loss": 1.4444, + "step": 2117 + }, + { + "epoch": 0.18, + "learning_rate": 1.9954575105860648e-05, + "loss": 1.4553, + "step": 2118 + }, + { + "epoch": 0.18, + "learning_rate": 1.9954488524274472e-05, + "loss": 1.3221, + "step": 2119 + }, + { + "epoch": 0.18, + "learning_rate": 1.9954401860441122e-05, + "loss": 1.3736, + "step": 2120 + }, + { + "epoch": 0.18, + "learning_rate": 1.995431511436132e-05, + "loss": 1.3434, + "step": 2121 + }, + { + "epoch": 0.18, + "learning_rate": 1.995422828603578e-05, + "loss": 1.2974, + "step": 2122 + }, + { + "epoch": 0.18, + "learning_rate": 1.995414137546522e-05, + "loss": 1.4161, + "step": 2123 + }, + { + "epoch": 0.18, + "learning_rate": 1.995405438265036e-05, + "loss": 1.3806, + "step": 2124 + }, + { + "epoch": 0.18, + "learning_rate": 1.9953967307591916e-05, + "loss": 1.2425, + "step": 2125 + }, + { + "epoch": 0.18, + "learning_rate": 1.995388015029061e-05, + "loss": 1.2962, + "step": 2126 + }, + { + "epoch": 0.18, + "learning_rate": 1.995379291074716e-05, + "loss": 1.3284, + "step": 2127 + }, + { + "epoch": 0.18, + "learning_rate": 1.9953705588962285e-05, + "loss": 1.2931, + "step": 2128 + }, + { + "epoch": 0.18, + "learning_rate": 1.9953618184936706e-05, + "loss": 1.279, + "step": 2129 + }, + { + "epoch": 0.18, + "learning_rate": 1.9953530698671156e-05, + "loss": 1.3729, + "step": 2130 + }, + { + "epoch": 0.18, + "learning_rate": 1.9953443130166344e-05, + "loss": 1.483, + "step": 2131 + }, + { + "epoch": 0.18, + "learning_rate": 1.9953355479423e-05, + "loss": 1.3014, + "step": 2132 + }, + { + "epoch": 0.18, + "learning_rate": 1.995326774644185e-05, + "loss": 1.3667, + "step": 2133 + }, + { + "epoch": 0.18, + "learning_rate": 1.9953179931223614e-05, + "loss": 1.3798, + "step": 2134 + }, + { + "epoch": 0.18, + "learning_rate": 1.995309203376902e-05, + "loss": 1.2616, + "step": 2135 + }, + { + "epoch": 0.18, + "learning_rate": 1.9953004054078796e-05, + "loss": 1.3197, + "step": 2136 + }, + { + "epoch": 0.18, + "learning_rate": 1.9952915992153667e-05, + "loss": 1.3039, + "step": 2137 + }, + { + "epoch": 0.18, + "learning_rate": 1.995282784799436e-05, + "loss": 1.4191, + "step": 2138 + }, + { + "epoch": 0.18, + "learning_rate": 1.9952739621601606e-05, + "loss": 1.3182, + "step": 2139 + }, + { + "epoch": 0.18, + "learning_rate": 1.995265131297613e-05, + "loss": 1.3847, + "step": 2140 + }, + { + "epoch": 0.18, + "learning_rate": 1.9952562922118663e-05, + "loss": 1.3682, + "step": 2141 + }, + { + "epoch": 0.18, + "learning_rate": 1.9952474449029937e-05, + "loss": 1.345, + "step": 2142 + }, + { + "epoch": 0.18, + "learning_rate": 1.995238589371068e-05, + "loss": 1.3863, + "step": 2143 + }, + { + "epoch": 0.18, + "learning_rate": 1.995229725616163e-05, + "loss": 1.382, + "step": 2144 + }, + { + "epoch": 0.18, + "learning_rate": 1.995220853638351e-05, + "loss": 1.4654, + "step": 2145 + }, + { + "epoch": 0.18, + "learning_rate": 1.995211973437706e-05, + "loss": 1.4735, + "step": 2146 + }, + { + "epoch": 0.18, + "learning_rate": 1.9952030850143012e-05, + "loss": 1.3305, + "step": 2147 + }, + { + "epoch": 0.18, + "learning_rate": 1.9951941883682098e-05, + "loss": 1.4027, + "step": 2148 + }, + { + "epoch": 0.18, + "learning_rate": 1.9951852834995057e-05, + "loss": 1.2988, + "step": 2149 + }, + { + "epoch": 0.18, + "learning_rate": 1.9951763704082622e-05, + "loss": 1.3538, + "step": 2150 + }, + { + "epoch": 0.18, + "learning_rate": 1.995167449094553e-05, + "loss": 1.4409, + "step": 2151 + }, + { + "epoch": 0.18, + "learning_rate": 1.9951585195584525e-05, + "loss": 1.4171, + "step": 2152 + }, + { + "epoch": 0.18, + "learning_rate": 1.995149581800033e-05, + "loss": 1.3561, + "step": 2153 + }, + { + "epoch": 0.18, + "learning_rate": 1.9951406358193696e-05, + "loss": 1.4404, + "step": 2154 + }, + { + "epoch": 0.18, + "learning_rate": 1.9951316816165356e-05, + "loss": 1.2309, + "step": 2155 + }, + { + "epoch": 0.18, + "learning_rate": 1.995122719191605e-05, + "loss": 1.441, + "step": 2156 + }, + { + "epoch": 0.18, + "learning_rate": 1.9951137485446523e-05, + "loss": 1.3281, + "step": 2157 + }, + { + "epoch": 0.18, + "learning_rate": 1.9951047696757513e-05, + "loss": 1.4185, + "step": 2158 + }, + { + "epoch": 0.18, + "learning_rate": 1.995095782584976e-05, + "loss": 1.337, + "step": 2159 + }, + { + "epoch": 0.18, + "learning_rate": 1.9950867872724008e-05, + "loss": 1.3969, + "step": 2160 + }, + { + "epoch": 0.18, + "learning_rate": 1.9950777837381006e-05, + "loss": 1.4158, + "step": 2161 + }, + { + "epoch": 0.18, + "learning_rate": 1.995068771982149e-05, + "loss": 1.3884, + "step": 2162 + }, + { + "epoch": 0.18, + "learning_rate": 1.9950597520046205e-05, + "loss": 1.3136, + "step": 2163 + }, + { + "epoch": 0.18, + "learning_rate": 1.9950507238055903e-05, + "loss": 1.2897, + "step": 2164 + }, + { + "epoch": 0.18, + "learning_rate": 1.995041687385132e-05, + "loss": 1.3751, + "step": 2165 + }, + { + "epoch": 0.18, + "learning_rate": 1.9950326427433214e-05, + "loss": 1.3965, + "step": 2166 + }, + { + "epoch": 0.18, + "learning_rate": 1.9950235898802327e-05, + "loss": 1.5057, + "step": 2167 + }, + { + "epoch": 0.18, + "learning_rate": 1.9950145287959405e-05, + "loss": 1.3931, + "step": 2168 + }, + { + "epoch": 0.18, + "learning_rate": 1.9950054594905196e-05, + "loss": 1.3787, + "step": 2169 + }, + { + "epoch": 0.18, + "learning_rate": 1.994996381964045e-05, + "loss": 1.3849, + "step": 2170 + }, + { + "epoch": 0.18, + "learning_rate": 1.9949872962165925e-05, + "loss": 1.4094, + "step": 2171 + }, + { + "epoch": 0.18, + "learning_rate": 1.994978202248236e-05, + "loss": 1.4177, + "step": 2172 + }, + { + "epoch": 0.18, + "learning_rate": 1.9949691000590513e-05, + "loss": 1.4465, + "step": 2173 + }, + { + "epoch": 0.18, + "learning_rate": 1.9949599896491133e-05, + "loss": 1.3627, + "step": 2174 + }, + { + "epoch": 0.18, + "learning_rate": 1.9949508710184977e-05, + "loss": 1.4196, + "step": 2175 + }, + { + "epoch": 0.18, + "learning_rate": 1.9949417441672793e-05, + "loss": 1.3958, + "step": 2176 + }, + { + "epoch": 0.18, + "learning_rate": 1.994932609095534e-05, + "loss": 1.386, + "step": 2177 + }, + { + "epoch": 0.18, + "learning_rate": 1.9949234658033373e-05, + "loss": 1.3345, + "step": 2178 + }, + { + "epoch": 0.18, + "learning_rate": 1.994914314290764e-05, + "loss": 1.4164, + "step": 2179 + }, + { + "epoch": 0.18, + "learning_rate": 1.9949051545578906e-05, + "loss": 1.4124, + "step": 2180 + }, + { + "epoch": 0.18, + "learning_rate": 1.9948959866047925e-05, + "loss": 1.3577, + "step": 2181 + }, + { + "epoch": 0.18, + "learning_rate": 1.994886810431545e-05, + "loss": 1.3209, + "step": 2182 + }, + { + "epoch": 0.18, + "learning_rate": 1.9948776260382243e-05, + "loss": 1.3859, + "step": 2183 + }, + { + "epoch": 0.18, + "learning_rate": 1.994868433424906e-05, + "loss": 1.3288, + "step": 2184 + }, + { + "epoch": 0.18, + "learning_rate": 1.994859232591667e-05, + "loss": 1.3953, + "step": 2185 + }, + { + "epoch": 0.18, + "learning_rate": 1.9948500235385825e-05, + "loss": 1.361, + "step": 2186 + }, + { + "epoch": 0.18, + "learning_rate": 1.994840806265728e-05, + "loss": 1.35, + "step": 2187 + }, + { + "epoch": 0.18, + "learning_rate": 1.9948315807731812e-05, + "loss": 1.3632, + "step": 2188 + }, + { + "epoch": 0.18, + "learning_rate": 1.994822347061017e-05, + "loss": 1.3693, + "step": 2189 + }, + { + "epoch": 0.18, + "learning_rate": 1.994813105129312e-05, + "loss": 1.3671, + "step": 2190 + }, + { + "epoch": 0.18, + "learning_rate": 1.9948038549781436e-05, + "loss": 1.3158, + "step": 2191 + }, + { + "epoch": 0.18, + "learning_rate": 1.9947945966075868e-05, + "loss": 1.3283, + "step": 2192 + }, + { + "epoch": 0.18, + "learning_rate": 1.9947853300177185e-05, + "loss": 1.4443, + "step": 2193 + }, + { + "epoch": 0.18, + "learning_rate": 1.9947760552086158e-05, + "loss": 1.3581, + "step": 2194 + }, + { + "epoch": 0.18, + "learning_rate": 1.9947667721803546e-05, + "loss": 1.3717, + "step": 2195 + }, + { + "epoch": 0.18, + "learning_rate": 1.9947574809330122e-05, + "loss": 1.3611, + "step": 2196 + }, + { + "epoch": 0.18, + "learning_rate": 1.9947481814666653e-05, + "loss": 1.4078, + "step": 2197 + }, + { + "epoch": 0.19, + "learning_rate": 1.9947388737813902e-05, + "loss": 1.4061, + "step": 2198 + }, + { + "epoch": 0.19, + "learning_rate": 1.9947295578772647e-05, + "loss": 1.4139, + "step": 2199 + }, + { + "epoch": 0.19, + "learning_rate": 1.9947202337543646e-05, + "loss": 1.3743, + "step": 2200 + }, + { + "epoch": 0.19, + "learning_rate": 1.9947109014127678e-05, + "loss": 1.3595, + "step": 2201 + }, + { + "epoch": 0.19, + "learning_rate": 1.9947015608525513e-05, + "loss": 1.3804, + "step": 2202 + }, + { + "epoch": 0.19, + "learning_rate": 1.994692212073792e-05, + "loss": 1.3309, + "step": 2203 + }, + { + "epoch": 0.19, + "learning_rate": 1.9946828550765674e-05, + "loss": 1.3322, + "step": 2204 + }, + { + "epoch": 0.19, + "learning_rate": 1.994673489860955e-05, + "loss": 1.4152, + "step": 2205 + }, + { + "epoch": 0.19, + "learning_rate": 1.9946641164270312e-05, + "loss": 1.3429, + "step": 2206 + }, + { + "epoch": 0.19, + "learning_rate": 1.9946547347748744e-05, + "loss": 1.2384, + "step": 2207 + }, + { + "epoch": 0.19, + "learning_rate": 1.994645344904562e-05, + "loss": 1.4537, + "step": 2208 + }, + { + "epoch": 0.19, + "learning_rate": 1.9946359468161717e-05, + "loss": 1.3877, + "step": 2209 + }, + { + "epoch": 0.19, + "learning_rate": 1.9946265405097803e-05, + "loss": 1.2962, + "step": 2210 + }, + { + "epoch": 0.19, + "learning_rate": 1.9946171259854664e-05, + "loss": 1.4189, + "step": 2211 + }, + { + "epoch": 0.19, + "learning_rate": 1.9946077032433077e-05, + "loss": 1.373, + "step": 2212 + }, + { + "epoch": 0.19, + "learning_rate": 1.9945982722833817e-05, + "loss": 1.2739, + "step": 2213 + }, + { + "epoch": 0.19, + "learning_rate": 1.9945888331057663e-05, + "loss": 1.3371, + "step": 2214 + }, + { + "epoch": 0.19, + "learning_rate": 1.9945793857105398e-05, + "loss": 1.3251, + "step": 2215 + }, + { + "epoch": 0.19, + "learning_rate": 1.99456993009778e-05, + "loss": 1.3969, + "step": 2216 + }, + { + "epoch": 0.19, + "learning_rate": 1.9945604662675652e-05, + "loss": 1.2834, + "step": 2217 + }, + { + "epoch": 0.19, + "learning_rate": 1.9945509942199733e-05, + "loss": 1.2773, + "step": 2218 + }, + { + "epoch": 0.19, + "learning_rate": 1.994541513955083e-05, + "loss": 1.3144, + "step": 2219 + }, + { + "epoch": 0.19, + "learning_rate": 1.9945320254729728e-05, + "loss": 1.385, + "step": 2220 + }, + { + "epoch": 0.19, + "learning_rate": 1.99452252877372e-05, + "loss": 1.3972, + "step": 2221 + }, + { + "epoch": 0.19, + "learning_rate": 1.994513023857404e-05, + "loss": 1.3711, + "step": 2222 + }, + { + "epoch": 0.19, + "learning_rate": 1.9945035107241034e-05, + "loss": 1.363, + "step": 2223 + }, + { + "epoch": 0.19, + "learning_rate": 1.9944939893738963e-05, + "loss": 1.333, + "step": 2224 + }, + { + "epoch": 0.19, + "learning_rate": 1.9944844598068615e-05, + "loss": 1.4455, + "step": 2225 + }, + { + "epoch": 0.19, + "learning_rate": 1.9944749220230782e-05, + "loss": 1.3887, + "step": 2226 + }, + { + "epoch": 0.19, + "learning_rate": 1.9944653760226243e-05, + "loss": 1.3289, + "step": 2227 + }, + { + "epoch": 0.19, + "learning_rate": 1.9944558218055795e-05, + "loss": 1.4194, + "step": 2228 + }, + { + "epoch": 0.19, + "learning_rate": 1.9944462593720227e-05, + "loss": 1.396, + "step": 2229 + }, + { + "epoch": 0.19, + "learning_rate": 1.9944366887220322e-05, + "loss": 1.3114, + "step": 2230 + }, + { + "epoch": 0.19, + "learning_rate": 1.9944271098556876e-05, + "loss": 1.3822, + "step": 2231 + }, + { + "epoch": 0.19, + "learning_rate": 1.994417522773068e-05, + "loss": 1.3753, + "step": 2232 + }, + { + "epoch": 0.19, + "learning_rate": 1.9944079274742526e-05, + "loss": 1.3625, + "step": 2233 + }, + { + "epoch": 0.19, + "learning_rate": 1.9943983239593204e-05, + "loss": 1.3367, + "step": 2234 + }, + { + "epoch": 0.19, + "learning_rate": 1.9943887122283512e-05, + "loss": 1.4247, + "step": 2235 + }, + { + "epoch": 0.19, + "learning_rate": 1.9943790922814244e-05, + "loss": 1.3826, + "step": 2236 + }, + { + "epoch": 0.19, + "learning_rate": 1.994369464118619e-05, + "loss": 1.3649, + "step": 2237 + }, + { + "epoch": 0.19, + "learning_rate": 1.994359827740015e-05, + "loss": 1.3427, + "step": 2238 + }, + { + "epoch": 0.19, + "learning_rate": 1.994350183145692e-05, + "loss": 1.3365, + "step": 2239 + }, + { + "epoch": 0.19, + "learning_rate": 1.9943405303357294e-05, + "loss": 1.4346, + "step": 2240 + }, + { + "epoch": 0.19, + "learning_rate": 1.994330869310207e-05, + "loss": 1.4218, + "step": 2241 + }, + { + "epoch": 0.19, + "learning_rate": 1.994321200069205e-05, + "loss": 1.3804, + "step": 2242 + }, + { + "epoch": 0.19, + "learning_rate": 1.994311522612803e-05, + "loss": 1.4363, + "step": 2243 + }, + { + "epoch": 0.19, + "learning_rate": 1.9943018369410805e-05, + "loss": 1.4212, + "step": 2244 + }, + { + "epoch": 0.19, + "learning_rate": 1.9942921430541188e-05, + "loss": 1.4003, + "step": 2245 + }, + { + "epoch": 0.19, + "learning_rate": 1.9942824409519966e-05, + "loss": 1.3646, + "step": 2246 + }, + { + "epoch": 0.19, + "learning_rate": 1.9942727306347946e-05, + "loss": 1.3812, + "step": 2247 + }, + { + "epoch": 0.19, + "learning_rate": 1.9942630121025934e-05, + "loss": 1.3435, + "step": 2248 + }, + { + "epoch": 0.19, + "learning_rate": 1.9942532853554727e-05, + "loss": 1.29, + "step": 2249 + }, + { + "epoch": 0.19, + "learning_rate": 1.9942435503935137e-05, + "loss": 1.3082, + "step": 2250 + }, + { + "epoch": 0.19, + "learning_rate": 1.9942338072167957e-05, + "loss": 1.3907, + "step": 2251 + }, + { + "epoch": 0.19, + "learning_rate": 1.9942240558254e-05, + "loss": 1.3625, + "step": 2252 + }, + { + "epoch": 0.19, + "learning_rate": 1.994214296219407e-05, + "loss": 1.4093, + "step": 2253 + }, + { + "epoch": 0.19, + "learning_rate": 1.9942045283988974e-05, + "loss": 1.4233, + "step": 2254 + }, + { + "epoch": 0.19, + "learning_rate": 1.9941947523639517e-05, + "loss": 1.414, + "step": 2255 + }, + { + "epoch": 0.19, + "learning_rate": 1.9941849681146508e-05, + "loss": 1.3624, + "step": 2256 + }, + { + "epoch": 0.19, + "learning_rate": 1.9941751756510755e-05, + "loss": 1.3649, + "step": 2257 + }, + { + "epoch": 0.19, + "learning_rate": 1.9941653749733064e-05, + "loss": 1.3581, + "step": 2258 + }, + { + "epoch": 0.19, + "learning_rate": 1.9941555660814254e-05, + "loss": 1.3511, + "step": 2259 + }, + { + "epoch": 0.19, + "learning_rate": 1.9941457489755125e-05, + "loss": 1.4136, + "step": 2260 + }, + { + "epoch": 0.19, + "learning_rate": 1.9941359236556495e-05, + "loss": 1.373, + "step": 2261 + }, + { + "epoch": 0.19, + "learning_rate": 1.9941260901219172e-05, + "loss": 1.4436, + "step": 2262 + }, + { + "epoch": 0.19, + "learning_rate": 1.994116248374397e-05, + "loss": 1.3602, + "step": 2263 + }, + { + "epoch": 0.19, + "learning_rate": 1.99410639841317e-05, + "loss": 1.3713, + "step": 2264 + }, + { + "epoch": 0.19, + "learning_rate": 1.994096540238318e-05, + "loss": 1.4853, + "step": 2265 + }, + { + "epoch": 0.19, + "learning_rate": 1.9940866738499225e-05, + "loss": 1.3211, + "step": 2266 + }, + { + "epoch": 0.19, + "learning_rate": 1.9940767992480644e-05, + "loss": 1.3791, + "step": 2267 + }, + { + "epoch": 0.19, + "learning_rate": 1.9940669164328253e-05, + "loss": 1.3872, + "step": 2268 + }, + { + "epoch": 0.19, + "learning_rate": 1.994057025404288e-05, + "loss": 1.2989, + "step": 2269 + }, + { + "epoch": 0.19, + "learning_rate": 1.994047126162533e-05, + "loss": 1.3197, + "step": 2270 + }, + { + "epoch": 0.19, + "learning_rate": 1.9940372187076425e-05, + "loss": 1.3513, + "step": 2271 + }, + { + "epoch": 0.19, + "learning_rate": 1.9940273030396982e-05, + "loss": 1.4042, + "step": 2272 + }, + { + "epoch": 0.19, + "learning_rate": 1.9940173791587826e-05, + "loss": 1.2252, + "step": 2273 + }, + { + "epoch": 0.19, + "learning_rate": 1.994007447064977e-05, + "loss": 1.4293, + "step": 2274 + }, + { + "epoch": 0.19, + "learning_rate": 1.9939975067583638e-05, + "loss": 1.4013, + "step": 2275 + }, + { + "epoch": 0.19, + "learning_rate": 1.9939875582390253e-05, + "loss": 1.2993, + "step": 2276 + }, + { + "epoch": 0.19, + "learning_rate": 1.993977601507043e-05, + "loss": 1.3467, + "step": 2277 + }, + { + "epoch": 0.19, + "learning_rate": 1.9939676365625e-05, + "loss": 1.4291, + "step": 2278 + }, + { + "epoch": 0.19, + "learning_rate": 1.9939576634054782e-05, + "loss": 1.4431, + "step": 2279 + }, + { + "epoch": 0.19, + "learning_rate": 1.99394768203606e-05, + "loss": 1.3132, + "step": 2280 + }, + { + "epoch": 0.19, + "learning_rate": 1.993937692454328e-05, + "loss": 1.3391, + "step": 2281 + }, + { + "epoch": 0.19, + "learning_rate": 1.9939276946603647e-05, + "loss": 1.3439, + "step": 2282 + }, + { + "epoch": 0.19, + "learning_rate": 1.993917688654253e-05, + "loss": 1.4048, + "step": 2283 + }, + { + "epoch": 0.19, + "learning_rate": 1.9939076744360748e-05, + "loss": 1.3335, + "step": 2284 + }, + { + "epoch": 0.19, + "learning_rate": 1.9938976520059135e-05, + "loss": 1.3572, + "step": 2285 + }, + { + "epoch": 0.19, + "learning_rate": 1.9938876213638515e-05, + "loss": 1.4206, + "step": 2286 + }, + { + "epoch": 0.19, + "learning_rate": 1.993877582509972e-05, + "loss": 1.3534, + "step": 2287 + }, + { + "epoch": 0.19, + "learning_rate": 1.993867535444358e-05, + "loss": 1.4693, + "step": 2288 + }, + { + "epoch": 0.19, + "learning_rate": 1.9938574801670922e-05, + "loss": 1.3949, + "step": 2289 + }, + { + "epoch": 0.19, + "learning_rate": 1.993847416678258e-05, + "loss": 1.4574, + "step": 2290 + }, + { + "epoch": 0.19, + "learning_rate": 1.993837344977938e-05, + "loss": 1.3866, + "step": 2291 + }, + { + "epoch": 0.19, + "learning_rate": 1.9938272650662164e-05, + "loss": 1.3923, + "step": 2292 + }, + { + "epoch": 0.19, + "learning_rate": 1.993817176943175e-05, + "loss": 1.316, + "step": 2293 + }, + { + "epoch": 0.19, + "learning_rate": 1.9938070806088987e-05, + "loss": 1.3465, + "step": 2294 + }, + { + "epoch": 0.19, + "learning_rate": 1.9937969760634702e-05, + "loss": 1.4083, + "step": 2295 + }, + { + "epoch": 0.19, + "learning_rate": 1.9937868633069728e-05, + "loss": 1.3402, + "step": 2296 + }, + { + "epoch": 0.19, + "learning_rate": 1.9937767423394905e-05, + "loss": 1.3108, + "step": 2297 + }, + { + "epoch": 0.19, + "learning_rate": 1.9937666131611065e-05, + "loss": 1.3272, + "step": 2298 + }, + { + "epoch": 0.19, + "learning_rate": 1.9937564757719044e-05, + "loss": 1.3858, + "step": 2299 + }, + { + "epoch": 0.19, + "learning_rate": 1.9937463301719687e-05, + "loss": 1.4781, + "step": 2300 + }, + { + "epoch": 0.19, + "learning_rate": 1.993736176361383e-05, + "loss": 1.3911, + "step": 2301 + }, + { + "epoch": 0.19, + "learning_rate": 1.9937260143402306e-05, + "loss": 1.4174, + "step": 2302 + }, + { + "epoch": 0.19, + "learning_rate": 1.993715844108596e-05, + "loss": 1.377, + "step": 2303 + }, + { + "epoch": 0.19, + "learning_rate": 1.993705665666563e-05, + "loss": 1.3205, + "step": 2304 + }, + { + "epoch": 0.19, + "learning_rate": 1.9936954790142155e-05, + "loss": 1.2918, + "step": 2305 + }, + { + "epoch": 0.19, + "learning_rate": 1.9936852841516382e-05, + "loss": 1.398, + "step": 2306 + }, + { + "epoch": 0.19, + "learning_rate": 1.9936750810789148e-05, + "loss": 1.3739, + "step": 2307 + }, + { + "epoch": 0.19, + "learning_rate": 1.9936648697961304e-05, + "loss": 1.4542, + "step": 2308 + }, + { + "epoch": 0.19, + "learning_rate": 1.9936546503033684e-05, + "loss": 1.3421, + "step": 2309 + }, + { + "epoch": 0.19, + "learning_rate": 1.993644422600714e-05, + "loss": 1.3326, + "step": 2310 + }, + { + "epoch": 0.19, + "learning_rate": 1.993634186688251e-05, + "loss": 1.3787, + "step": 2311 + }, + { + "epoch": 0.19, + "learning_rate": 1.9936239425660643e-05, + "loss": 1.4625, + "step": 2312 + }, + { + "epoch": 0.19, + "learning_rate": 1.9936136902342392e-05, + "loss": 1.4519, + "step": 2313 + }, + { + "epoch": 0.19, + "learning_rate": 1.9936034296928592e-05, + "loss": 1.2579, + "step": 2314 + }, + { + "epoch": 0.19, + "learning_rate": 1.99359316094201e-05, + "loss": 1.3713, + "step": 2315 + }, + { + "epoch": 0.19, + "learning_rate": 1.9935828839817758e-05, + "loss": 1.3424, + "step": 2316 + }, + { + "epoch": 0.2, + "learning_rate": 1.993572598812242e-05, + "loss": 1.3647, + "step": 2317 + }, + { + "epoch": 0.2, + "learning_rate": 1.9935623054334936e-05, + "loss": 1.3839, + "step": 2318 + }, + { + "epoch": 0.2, + "learning_rate": 1.993552003845615e-05, + "loss": 1.3506, + "step": 2319 + }, + { + "epoch": 0.2, + "learning_rate": 1.9935416940486922e-05, + "loss": 1.3641, + "step": 2320 + }, + { + "epoch": 0.2, + "learning_rate": 1.9935313760428095e-05, + "loss": 1.3198, + "step": 2321 + }, + { + "epoch": 0.2, + "learning_rate": 1.993521049828053e-05, + "loss": 1.3936, + "step": 2322 + }, + { + "epoch": 0.2, + "learning_rate": 1.9935107154045075e-05, + "loss": 1.3624, + "step": 2323 + }, + { + "epoch": 0.2, + "learning_rate": 1.9935003727722586e-05, + "loss": 1.4282, + "step": 2324 + }, + { + "epoch": 0.2, + "learning_rate": 1.9934900219313914e-05, + "loss": 1.3824, + "step": 2325 + }, + { + "epoch": 0.2, + "learning_rate": 1.993479662881992e-05, + "loss": 1.3474, + "step": 2326 + }, + { + "epoch": 0.2, + "learning_rate": 1.9934692956241456e-05, + "loss": 1.4414, + "step": 2327 + }, + { + "epoch": 0.2, + "learning_rate": 1.9934589201579378e-05, + "loss": 1.4288, + "step": 2328 + }, + { + "epoch": 0.2, + "learning_rate": 1.9934485364834545e-05, + "loss": 1.3091, + "step": 2329 + }, + { + "epoch": 0.2, + "learning_rate": 1.9934381446007812e-05, + "loss": 1.3965, + "step": 2330 + }, + { + "epoch": 0.2, + "learning_rate": 1.9934277445100046e-05, + "loss": 1.3094, + "step": 2331 + }, + { + "epoch": 0.2, + "learning_rate": 1.9934173362112093e-05, + "loss": 1.4017, + "step": 2332 + }, + { + "epoch": 0.2, + "learning_rate": 1.993406919704483e-05, + "loss": 1.3832, + "step": 2333 + }, + { + "epoch": 0.2, + "learning_rate": 1.99339649498991e-05, + "loss": 1.3111, + "step": 2334 + }, + { + "epoch": 0.2, + "learning_rate": 1.9933860620675772e-05, + "loss": 1.3992, + "step": 2335 + }, + { + "epoch": 0.2, + "learning_rate": 1.993375620937571e-05, + "loss": 1.3552, + "step": 2336 + }, + { + "epoch": 0.2, + "learning_rate": 1.9933651715999777e-05, + "loss": 1.4496, + "step": 2337 + }, + { + "epoch": 0.2, + "learning_rate": 1.9933547140548834e-05, + "loss": 1.4024, + "step": 2338 + }, + { + "epoch": 0.2, + "learning_rate": 1.9933442483023743e-05, + "loss": 1.3296, + "step": 2339 + }, + { + "epoch": 0.2, + "learning_rate": 1.9933337743425374e-05, + "loss": 1.3822, + "step": 2340 + }, + { + "epoch": 0.2, + "learning_rate": 1.9933232921754587e-05, + "loss": 1.4299, + "step": 2341 + }, + { + "epoch": 0.2, + "learning_rate": 1.993312801801225e-05, + "loss": 1.3331, + "step": 2342 + }, + { + "epoch": 0.2, + "learning_rate": 1.9933023032199236e-05, + "loss": 1.36, + "step": 2343 + }, + { + "epoch": 0.2, + "learning_rate": 1.9932917964316402e-05, + "loss": 1.3679, + "step": 2344 + }, + { + "epoch": 0.2, + "learning_rate": 1.993281281436462e-05, + "loss": 1.367, + "step": 2345 + }, + { + "epoch": 0.2, + "learning_rate": 1.9932707582344762e-05, + "loss": 1.4104, + "step": 2346 + }, + { + "epoch": 0.2, + "learning_rate": 1.993260226825769e-05, + "loss": 1.3567, + "step": 2347 + }, + { + "epoch": 0.2, + "learning_rate": 1.9932496872104287e-05, + "loss": 1.4181, + "step": 2348 + }, + { + "epoch": 0.2, + "learning_rate": 1.9932391393885412e-05, + "loss": 1.4397, + "step": 2349 + }, + { + "epoch": 0.2, + "learning_rate": 1.9932285833601943e-05, + "loss": 1.3976, + "step": 2350 + }, + { + "epoch": 0.2, + "learning_rate": 1.9932180191254748e-05, + "loss": 1.3661, + "step": 2351 + }, + { + "epoch": 0.2, + "learning_rate": 1.99320744668447e-05, + "loss": 1.4312, + "step": 2352 + }, + { + "epoch": 0.2, + "learning_rate": 1.9931968660372677e-05, + "loss": 1.393, + "step": 2353 + }, + { + "epoch": 0.2, + "learning_rate": 1.9931862771839546e-05, + "loss": 1.3928, + "step": 2354 + }, + { + "epoch": 0.2, + "learning_rate": 1.993175680124619e-05, + "loss": 1.3902, + "step": 2355 + }, + { + "epoch": 0.2, + "learning_rate": 1.993165074859348e-05, + "loss": 1.4301, + "step": 2356 + }, + { + "epoch": 0.2, + "learning_rate": 1.993154461388229e-05, + "loss": 1.3971, + "step": 2357 + }, + { + "epoch": 0.2, + "learning_rate": 1.9931438397113502e-05, + "loss": 1.3797, + "step": 2358 + }, + { + "epoch": 0.2, + "learning_rate": 1.9931332098287993e-05, + "loss": 1.3805, + "step": 2359 + }, + { + "epoch": 0.2, + "learning_rate": 1.993122571740664e-05, + "loss": 1.3582, + "step": 2360 + }, + { + "epoch": 0.2, + "learning_rate": 1.993111925447032e-05, + "loss": 1.2968, + "step": 2361 + }, + { + "epoch": 0.2, + "learning_rate": 1.9931012709479915e-05, + "loss": 1.3482, + "step": 2362 + }, + { + "epoch": 0.2, + "learning_rate": 1.9930906082436305e-05, + "loss": 1.464, + "step": 2363 + }, + { + "epoch": 0.2, + "learning_rate": 1.993079937334037e-05, + "loss": 1.3451, + "step": 2364 + }, + { + "epoch": 0.2, + "learning_rate": 1.9930692582192992e-05, + "loss": 1.288, + "step": 2365 + }, + { + "epoch": 0.2, + "learning_rate": 1.9930585708995055e-05, + "loss": 1.4112, + "step": 2366 + }, + { + "epoch": 0.2, + "learning_rate": 1.9930478753747438e-05, + "loss": 1.3468, + "step": 2367 + }, + { + "epoch": 0.2, + "learning_rate": 1.993037171645103e-05, + "loss": 1.3183, + "step": 2368 + }, + { + "epoch": 0.2, + "learning_rate": 1.993026459710671e-05, + "loss": 1.4207, + "step": 2369 + }, + { + "epoch": 0.2, + "learning_rate": 1.993015739571537e-05, + "loss": 1.394, + "step": 2370 + }, + { + "epoch": 0.2, + "learning_rate": 1.9930050112277887e-05, + "loss": 1.3822, + "step": 2371 + }, + { + "epoch": 0.2, + "learning_rate": 1.992994274679516e-05, + "loss": 1.397, + "step": 2372 + }, + { + "epoch": 0.2, + "learning_rate": 1.992983529926806e-05, + "loss": 1.4907, + "step": 2373 + }, + { + "epoch": 0.2, + "learning_rate": 1.992972776969748e-05, + "loss": 1.359, + "step": 2374 + }, + { + "epoch": 0.2, + "learning_rate": 1.9929620158084318e-05, + "loss": 1.4495, + "step": 2375 + }, + { + "epoch": 0.2, + "learning_rate": 1.9929512464429455e-05, + "loss": 1.2851, + "step": 2376 + }, + { + "epoch": 0.2, + "learning_rate": 1.9929404688733786e-05, + "loss": 1.3838, + "step": 2377 + }, + { + "epoch": 0.2, + "learning_rate": 1.992929683099819e-05, + "loss": 1.391, + "step": 2378 + }, + { + "epoch": 0.2, + "learning_rate": 1.992918889122357e-05, + "loss": 1.3499, + "step": 2379 + }, + { + "epoch": 0.2, + "learning_rate": 1.9929080869410813e-05, + "loss": 1.4007, + "step": 2380 + }, + { + "epoch": 0.2, + "learning_rate": 1.992897276556081e-05, + "loss": 1.3503, + "step": 2381 + }, + { + "epoch": 0.2, + "learning_rate": 1.9928864579674458e-05, + "loss": 1.4082, + "step": 2382 + }, + { + "epoch": 0.2, + "learning_rate": 1.9928756311752652e-05, + "loss": 1.3978, + "step": 2383 + }, + { + "epoch": 0.2, + "learning_rate": 1.9928647961796282e-05, + "loss": 1.3128, + "step": 2384 + }, + { + "epoch": 0.2, + "learning_rate": 1.9928539529806245e-05, + "loss": 1.3797, + "step": 2385 + }, + { + "epoch": 0.2, + "learning_rate": 1.9928431015783434e-05, + "loss": 1.2802, + "step": 2386 + }, + { + "epoch": 0.2, + "learning_rate": 1.9928322419728753e-05, + "loss": 1.3554, + "step": 2387 + }, + { + "epoch": 0.2, + "learning_rate": 1.992821374164309e-05, + "loss": 1.3817, + "step": 2388 + }, + { + "epoch": 0.2, + "learning_rate": 1.992810498152735e-05, + "loss": 1.3717, + "step": 2389 + }, + { + "epoch": 0.2, + "learning_rate": 1.9927996139382428e-05, + "loss": 1.3428, + "step": 2390 + }, + { + "epoch": 0.2, + "learning_rate": 1.9927887215209226e-05, + "loss": 1.3497, + "step": 2391 + }, + { + "epoch": 0.2, + "learning_rate": 1.992777820900864e-05, + "loss": 1.4947, + "step": 2392 + }, + { + "epoch": 0.2, + "learning_rate": 1.992766912078158e-05, + "loss": 1.4354, + "step": 2393 + }, + { + "epoch": 0.2, + "learning_rate": 1.9927559950528935e-05, + "loss": 1.3421, + "step": 2394 + }, + { + "epoch": 0.2, + "learning_rate": 1.9927450698251612e-05, + "loss": 1.4084, + "step": 2395 + }, + { + "epoch": 0.2, + "learning_rate": 1.9927341363950515e-05, + "loss": 1.4536, + "step": 2396 + }, + { + "epoch": 0.2, + "learning_rate": 1.992723194762655e-05, + "loss": 1.3927, + "step": 2397 + }, + { + "epoch": 0.2, + "learning_rate": 1.992712244928061e-05, + "loss": 1.4447, + "step": 2398 + }, + { + "epoch": 0.2, + "learning_rate": 1.9927012868913614e-05, + "loss": 1.4335, + "step": 2399 + }, + { + "epoch": 0.2, + "learning_rate": 1.992690320652646e-05, + "loss": 1.4506, + "step": 2400 + }, + { + "epoch": 0.2, + "learning_rate": 1.9926793462120052e-05, + "loss": 1.4069, + "step": 2401 + }, + { + "epoch": 0.2, + "learning_rate": 1.99266836356953e-05, + "loss": 1.4041, + "step": 2402 + }, + { + "epoch": 0.2, + "learning_rate": 1.992657372725311e-05, + "loss": 1.4202, + "step": 2403 + }, + { + "epoch": 0.2, + "learning_rate": 1.9926463736794393e-05, + "loss": 1.2966, + "step": 2404 + }, + { + "epoch": 0.2, + "learning_rate": 1.9926353664320054e-05, + "loss": 1.3561, + "step": 2405 + }, + { + "epoch": 0.2, + "learning_rate": 1.9926243509831008e-05, + "loss": 1.4427, + "step": 2406 + }, + { + "epoch": 0.2, + "learning_rate": 1.9926133273328156e-05, + "loss": 1.5026, + "step": 2407 + }, + { + "epoch": 0.2, + "learning_rate": 1.9926022954812417e-05, + "loss": 1.3417, + "step": 2408 + }, + { + "epoch": 0.2, + "learning_rate": 1.9925912554284698e-05, + "loss": 1.403, + "step": 2409 + }, + { + "epoch": 0.2, + "learning_rate": 1.9925802071745914e-05, + "loss": 1.4605, + "step": 2410 + }, + { + "epoch": 0.2, + "learning_rate": 1.992569150719698e-05, + "loss": 1.281, + "step": 2411 + }, + { + "epoch": 0.2, + "learning_rate": 1.99255808606388e-05, + "loss": 1.3499, + "step": 2412 + }, + { + "epoch": 0.2, + "learning_rate": 1.9925470132072297e-05, + "loss": 1.3713, + "step": 2413 + }, + { + "epoch": 0.2, + "learning_rate": 1.9925359321498384e-05, + "loss": 1.3928, + "step": 2414 + }, + { + "epoch": 0.2, + "learning_rate": 1.9925248428917974e-05, + "loss": 1.4415, + "step": 2415 + }, + { + "epoch": 0.2, + "learning_rate": 1.9925137454331988e-05, + "loss": 1.3562, + "step": 2416 + }, + { + "epoch": 0.2, + "learning_rate": 1.9925026397741337e-05, + "loss": 1.3827, + "step": 2417 + }, + { + "epoch": 0.2, + "learning_rate": 1.9924915259146942e-05, + "loss": 1.3582, + "step": 2418 + }, + { + "epoch": 0.2, + "learning_rate": 1.992480403854972e-05, + "loss": 1.3106, + "step": 2419 + }, + { + "epoch": 0.2, + "learning_rate": 1.9924692735950594e-05, + "loss": 1.3927, + "step": 2420 + }, + { + "epoch": 0.2, + "learning_rate": 1.992458135135048e-05, + "loss": 1.3076, + "step": 2421 + }, + { + "epoch": 0.2, + "learning_rate": 1.992446988475029e-05, + "loss": 1.3491, + "step": 2422 + }, + { + "epoch": 0.2, + "learning_rate": 1.9924358336150965e-05, + "loss": 1.3425, + "step": 2423 + }, + { + "epoch": 0.2, + "learning_rate": 1.992424670555341e-05, + "loss": 1.3816, + "step": 2424 + }, + { + "epoch": 0.2, + "learning_rate": 1.992413499295855e-05, + "loss": 1.3698, + "step": 2425 + }, + { + "epoch": 0.2, + "learning_rate": 1.9924023198367317e-05, + "loss": 1.3906, + "step": 2426 + }, + { + "epoch": 0.2, + "learning_rate": 1.9923911321780625e-05, + "loss": 1.4202, + "step": 2427 + }, + { + "epoch": 0.2, + "learning_rate": 1.99237993631994e-05, + "loss": 1.3851, + "step": 2428 + }, + { + "epoch": 0.2, + "learning_rate": 1.9923687322624568e-05, + "loss": 1.4, + "step": 2429 + }, + { + "epoch": 0.2, + "learning_rate": 1.9923575200057058e-05, + "loss": 1.317, + "step": 2430 + }, + { + "epoch": 0.2, + "learning_rate": 1.9923462995497795e-05, + "loss": 1.3701, + "step": 2431 + }, + { + "epoch": 0.2, + "learning_rate": 1.9923350708947705e-05, + "loss": 1.3382, + "step": 2432 + }, + { + "epoch": 0.2, + "learning_rate": 1.992323834040771e-05, + "loss": 1.3743, + "step": 2433 + }, + { + "epoch": 0.2, + "learning_rate": 1.992312588987875e-05, + "loss": 1.403, + "step": 2434 + }, + { + "epoch": 0.21, + "learning_rate": 1.9923013357361745e-05, + "loss": 1.301, + "step": 2435 + }, + { + "epoch": 0.21, + "learning_rate": 1.992290074285763e-05, + "loss": 1.3007, + "step": 2436 + }, + { + "epoch": 0.21, + "learning_rate": 1.992278804636733e-05, + "loss": 1.204, + "step": 2437 + }, + { + "epoch": 0.21, + "learning_rate": 1.992267526789178e-05, + "loss": 1.3274, + "step": 2438 + }, + { + "epoch": 0.21, + "learning_rate": 1.9922562407431915e-05, + "loss": 1.3199, + "step": 2439 + }, + { + "epoch": 0.21, + "learning_rate": 1.9922449464988662e-05, + "loss": 1.4358, + "step": 2440 + }, + { + "epoch": 0.21, + "learning_rate": 1.9922336440562953e-05, + "loss": 1.4551, + "step": 2441 + }, + { + "epoch": 0.21, + "learning_rate": 1.992222333415573e-05, + "loss": 1.315, + "step": 2442 + }, + { + "epoch": 0.21, + "learning_rate": 1.9922110145767917e-05, + "loss": 1.4754, + "step": 2443 + }, + { + "epoch": 0.21, + "learning_rate": 1.992199687540046e-05, + "loss": 1.3383, + "step": 2444 + }, + { + "epoch": 0.21, + "learning_rate": 1.9921883523054283e-05, + "loss": 1.3466, + "step": 2445 + }, + { + "epoch": 0.21, + "learning_rate": 1.9921770088730334e-05, + "loss": 1.4057, + "step": 2446 + }, + { + "epoch": 0.21, + "learning_rate": 1.9921656572429544e-05, + "loss": 1.3838, + "step": 2447 + }, + { + "epoch": 0.21, + "learning_rate": 1.992154297415285e-05, + "loss": 1.3405, + "step": 2448 + }, + { + "epoch": 0.21, + "learning_rate": 1.9921429293901196e-05, + "loss": 1.3903, + "step": 2449 + }, + { + "epoch": 0.21, + "learning_rate": 1.9921315531675516e-05, + "loss": 1.3983, + "step": 2450 + }, + { + "epoch": 0.21, + "learning_rate": 1.992120168747675e-05, + "loss": 1.4604, + "step": 2451 + }, + { + "epoch": 0.21, + "learning_rate": 1.9921087761305843e-05, + "loss": 1.2322, + "step": 2452 + }, + { + "epoch": 0.21, + "learning_rate": 1.9920973753163733e-05, + "loss": 1.4372, + "step": 2453 + }, + { + "epoch": 0.21, + "learning_rate": 1.9920859663051362e-05, + "loss": 1.3836, + "step": 2454 + }, + { + "epoch": 0.21, + "learning_rate": 1.992074549096967e-05, + "loss": 1.3597, + "step": 2455 + }, + { + "epoch": 0.21, + "learning_rate": 1.9920631236919613e-05, + "loss": 1.3934, + "step": 2456 + }, + { + "epoch": 0.21, + "learning_rate": 1.992051690090212e-05, + "loss": 1.4481, + "step": 2457 + }, + { + "epoch": 0.21, + "learning_rate": 1.992040248291814e-05, + "loss": 1.2911, + "step": 2458 + }, + { + "epoch": 0.21, + "learning_rate": 1.9920287982968624e-05, + "loss": 1.3456, + "step": 2459 + }, + { + "epoch": 0.21, + "learning_rate": 1.992017340105451e-05, + "loss": 1.3556, + "step": 2460 + }, + { + "epoch": 0.21, + "learning_rate": 1.992005873717675e-05, + "loss": 1.3132, + "step": 2461 + }, + { + "epoch": 0.21, + "learning_rate": 1.991994399133629e-05, + "loss": 1.3173, + "step": 2462 + }, + { + "epoch": 0.21, + "learning_rate": 1.9919829163534077e-05, + "loss": 1.3171, + "step": 2463 + }, + { + "epoch": 0.21, + "learning_rate": 1.9919714253771064e-05, + "loss": 1.3248, + "step": 2464 + }, + { + "epoch": 0.21, + "learning_rate": 1.991959926204819e-05, + "loss": 1.3409, + "step": 2465 + }, + { + "epoch": 0.21, + "learning_rate": 1.991948418836642e-05, + "loss": 1.2485, + "step": 2466 + }, + { + "epoch": 0.21, + "learning_rate": 1.9919369032726693e-05, + "loss": 1.3288, + "step": 2467 + }, + { + "epoch": 0.21, + "learning_rate": 1.9919253795129966e-05, + "loss": 1.3796, + "step": 2468 + }, + { + "epoch": 0.21, + "learning_rate": 1.991913847557719e-05, + "loss": 1.3805, + "step": 2469 + }, + { + "epoch": 0.21, + "learning_rate": 1.9919023074069318e-05, + "loss": 1.3385, + "step": 2470 + }, + { + "epoch": 0.21, + "learning_rate": 1.9918907590607298e-05, + "loss": 1.3636, + "step": 2471 + }, + { + "epoch": 0.21, + "learning_rate": 1.9918792025192093e-05, + "loss": 1.3395, + "step": 2472 + }, + { + "epoch": 0.21, + "learning_rate": 1.9918676377824654e-05, + "loss": 1.2676, + "step": 2473 + }, + { + "epoch": 0.21, + "learning_rate": 1.9918560648505934e-05, + "loss": 1.3006, + "step": 2474 + }, + { + "epoch": 0.21, + "learning_rate": 1.9918444837236892e-05, + "loss": 1.3922, + "step": 2475 + }, + { + "epoch": 0.21, + "learning_rate": 1.991832894401849e-05, + "loss": 1.3687, + "step": 2476 + }, + { + "epoch": 0.21, + "learning_rate": 1.9918212968851672e-05, + "loss": 1.3498, + "step": 2477 + }, + { + "epoch": 0.21, + "learning_rate": 1.991809691173741e-05, + "loss": 1.3002, + "step": 2478 + }, + { + "epoch": 0.21, + "learning_rate": 1.991798077267665e-05, + "loss": 1.2593, + "step": 2479 + }, + { + "epoch": 0.21, + "learning_rate": 1.9917864551670366e-05, + "loss": 1.3023, + "step": 2480 + }, + { + "epoch": 0.21, + "learning_rate": 1.9917748248719507e-05, + "loss": 1.3959, + "step": 2481 + }, + { + "epoch": 0.21, + "learning_rate": 1.991763186382504e-05, + "loss": 1.4304, + "step": 2482 + }, + { + "epoch": 0.21, + "learning_rate": 1.9917515396987925e-05, + "loss": 1.3367, + "step": 2483 + }, + { + "epoch": 0.21, + "learning_rate": 1.991739884820912e-05, + "loss": 1.3156, + "step": 2484 + }, + { + "epoch": 0.21, + "learning_rate": 1.9917282217489598e-05, + "loss": 1.4776, + "step": 2485 + }, + { + "epoch": 0.21, + "learning_rate": 1.9917165504830312e-05, + "loss": 1.3577, + "step": 2486 + }, + { + "epoch": 0.21, + "learning_rate": 1.991704871023223e-05, + "loss": 1.3573, + "step": 2487 + }, + { + "epoch": 0.21, + "learning_rate": 1.991693183369632e-05, + "loss": 1.4007, + "step": 2488 + }, + { + "epoch": 0.21, + "learning_rate": 1.9916814875223546e-05, + "loss": 1.4038, + "step": 2489 + }, + { + "epoch": 0.21, + "learning_rate": 1.991669783481487e-05, + "loss": 1.3782, + "step": 2490 + }, + { + "epoch": 0.21, + "learning_rate": 1.9916580712471267e-05, + "loss": 1.3603, + "step": 2491 + }, + { + "epoch": 0.21, + "learning_rate": 1.9916463508193697e-05, + "loss": 1.3263, + "step": 2492 + }, + { + "epoch": 0.21, + "learning_rate": 1.9916346221983134e-05, + "loss": 1.4441, + "step": 2493 + }, + { + "epoch": 0.21, + "learning_rate": 1.991622885384054e-05, + "loss": 1.3641, + "step": 2494 + }, + { + "epoch": 0.21, + "learning_rate": 1.9916111403766898e-05, + "loss": 1.3652, + "step": 2495 + }, + { + "epoch": 0.21, + "learning_rate": 1.9915993871763163e-05, + "loss": 1.3966, + "step": 2496 + }, + { + "epoch": 0.21, + "learning_rate": 1.991587625783032e-05, + "loss": 1.3486, + "step": 2497 + }, + { + "epoch": 0.21, + "learning_rate": 1.9915758561969327e-05, + "loss": 1.3856, + "step": 2498 + }, + { + "epoch": 0.21, + "learning_rate": 1.9915640784181163e-05, + "loss": 1.2946, + "step": 2499 + }, + { + "epoch": 0.21, + "learning_rate": 1.9915522924466808e-05, + "loss": 1.4319, + "step": 2500 + }, + { + "epoch": 0.21, + "learning_rate": 1.9915404982827223e-05, + "loss": 1.3785, + "step": 2501 + }, + { + "epoch": 0.21, + "learning_rate": 1.9915286959263394e-05, + "loss": 1.3141, + "step": 2502 + }, + { + "epoch": 0.21, + "learning_rate": 1.9915168853776286e-05, + "loss": 1.3531, + "step": 2503 + }, + { + "epoch": 0.21, + "learning_rate": 1.9915050666366883e-05, + "loss": 1.4138, + "step": 2504 + }, + { + "epoch": 0.21, + "learning_rate": 1.9914932397036156e-05, + "loss": 1.4059, + "step": 2505 + }, + { + "epoch": 0.21, + "learning_rate": 1.9914814045785088e-05, + "loss": 1.3864, + "step": 2506 + }, + { + "epoch": 0.21, + "learning_rate": 1.991469561261465e-05, + "loss": 1.424, + "step": 2507 + }, + { + "epoch": 0.21, + "learning_rate": 1.9914577097525823e-05, + "loss": 1.3031, + "step": 2508 + }, + { + "epoch": 0.21, + "learning_rate": 1.9914458500519587e-05, + "loss": 1.3449, + "step": 2509 + }, + { + "epoch": 0.21, + "learning_rate": 1.9914339821596927e-05, + "loss": 1.36, + "step": 2510 + }, + { + "epoch": 0.21, + "learning_rate": 1.9914221060758812e-05, + "loss": 1.3482, + "step": 2511 + }, + { + "epoch": 0.21, + "learning_rate": 1.991410221800623e-05, + "loss": 1.3281, + "step": 2512 + }, + { + "epoch": 0.21, + "learning_rate": 1.9913983293340167e-05, + "loss": 1.3883, + "step": 2513 + }, + { + "epoch": 0.21, + "learning_rate": 1.99138642867616e-05, + "loss": 1.3564, + "step": 2514 + }, + { + "epoch": 0.21, + "learning_rate": 1.9913745198271512e-05, + "loss": 1.3777, + "step": 2515 + }, + { + "epoch": 0.21, + "learning_rate": 1.991362602787089e-05, + "loss": 1.3225, + "step": 2516 + }, + { + "epoch": 0.21, + "learning_rate": 1.9913506775560715e-05, + "loss": 1.4252, + "step": 2517 + }, + { + "epoch": 0.21, + "learning_rate": 1.9913387441341973e-05, + "loss": 1.2955, + "step": 2518 + }, + { + "epoch": 0.21, + "learning_rate": 1.9913268025215653e-05, + "loss": 1.3408, + "step": 2519 + }, + { + "epoch": 0.21, + "learning_rate": 1.991314852718274e-05, + "loss": 1.3599, + "step": 2520 + }, + { + "epoch": 0.21, + "learning_rate": 1.9913028947244223e-05, + "loss": 1.2292, + "step": 2521 + }, + { + "epoch": 0.21, + "learning_rate": 1.9912909285401085e-05, + "loss": 1.3438, + "step": 2522 + }, + { + "epoch": 0.21, + "learning_rate": 1.991278954165432e-05, + "loss": 1.4018, + "step": 2523 + }, + { + "epoch": 0.21, + "learning_rate": 1.9912669716004914e-05, + "loss": 1.409, + "step": 2524 + }, + { + "epoch": 0.21, + "learning_rate": 1.991254980845386e-05, + "loss": 1.4297, + "step": 2525 + }, + { + "epoch": 0.21, + "learning_rate": 1.9912429819002146e-05, + "loss": 1.3351, + "step": 2526 + }, + { + "epoch": 0.21, + "learning_rate": 1.9912309747650765e-05, + "loss": 1.2878, + "step": 2527 + }, + { + "epoch": 0.21, + "learning_rate": 1.9912189594400707e-05, + "loss": 1.3508, + "step": 2528 + }, + { + "epoch": 0.21, + "learning_rate": 1.991206935925297e-05, + "loss": 1.3804, + "step": 2529 + }, + { + "epoch": 0.21, + "learning_rate": 1.9911949042208542e-05, + "loss": 1.437, + "step": 2530 + }, + { + "epoch": 0.21, + "learning_rate": 1.9911828643268416e-05, + "loss": 1.3774, + "step": 2531 + }, + { + "epoch": 0.21, + "learning_rate": 1.991170816243359e-05, + "loss": 1.386, + "step": 2532 + }, + { + "epoch": 0.21, + "learning_rate": 1.991158759970506e-05, + "loss": 1.2943, + "step": 2533 + }, + { + "epoch": 0.21, + "learning_rate": 1.9911466955083822e-05, + "loss": 1.3149, + "step": 2534 + }, + { + "epoch": 0.21, + "learning_rate": 1.9911346228570875e-05, + "loss": 1.3983, + "step": 2535 + }, + { + "epoch": 0.21, + "learning_rate": 1.991122542016721e-05, + "loss": 1.3891, + "step": 2536 + }, + { + "epoch": 0.21, + "learning_rate": 1.9911104529873828e-05, + "loss": 1.203, + "step": 2537 + }, + { + "epoch": 0.21, + "learning_rate": 1.9910983557691727e-05, + "loss": 1.3281, + "step": 2538 + }, + { + "epoch": 0.21, + "learning_rate": 1.991086250362191e-05, + "loss": 1.3247, + "step": 2539 + }, + { + "epoch": 0.21, + "learning_rate": 1.9910741367665375e-05, + "loss": 1.4264, + "step": 2540 + }, + { + "epoch": 0.21, + "learning_rate": 1.9910620149823123e-05, + "loss": 1.346, + "step": 2541 + }, + { + "epoch": 0.21, + "learning_rate": 1.9910498850096157e-05, + "loss": 1.3612, + "step": 2542 + }, + { + "epoch": 0.21, + "learning_rate": 1.9910377468485475e-05, + "loss": 1.3748, + "step": 2543 + }, + { + "epoch": 0.21, + "learning_rate": 1.9910256004992085e-05, + "loss": 1.3599, + "step": 2544 + }, + { + "epoch": 0.21, + "learning_rate": 1.9910134459616985e-05, + "loss": 1.3879, + "step": 2545 + }, + { + "epoch": 0.21, + "learning_rate": 1.9910012832361185e-05, + "loss": 1.3616, + "step": 2546 + }, + { + "epoch": 0.21, + "learning_rate": 1.990989112322569e-05, + "loss": 1.258, + "step": 2547 + }, + { + "epoch": 0.21, + "learning_rate": 1.9909769332211498e-05, + "loss": 1.2902, + "step": 2548 + }, + { + "epoch": 0.21, + "learning_rate": 1.990964745931962e-05, + "loss": 1.3687, + "step": 2549 + }, + { + "epoch": 0.21, + "learning_rate": 1.9909525504551067e-05, + "loss": 1.3117, + "step": 2550 + }, + { + "epoch": 0.21, + "learning_rate": 1.990940346790684e-05, + "loss": 1.4152, + "step": 2551 + }, + { + "epoch": 0.21, + "learning_rate": 1.9909281349387952e-05, + "loss": 1.4421, + "step": 2552 + }, + { + "epoch": 0.21, + "learning_rate": 1.990915914899541e-05, + "loss": 1.3363, + "step": 2553 + }, + { + "epoch": 0.22, + "learning_rate": 1.9909036866730225e-05, + "loss": 1.3676, + "step": 2554 + }, + { + "epoch": 0.22, + "learning_rate": 1.9908914502593403e-05, + "loss": 1.3494, + "step": 2555 + }, + { + "epoch": 0.22, + "learning_rate": 1.9908792056585958e-05, + "loss": 1.3804, + "step": 2556 + }, + { + "epoch": 0.22, + "learning_rate": 1.9908669528708902e-05, + "loss": 1.4329, + "step": 2557 + }, + { + "epoch": 0.22, + "learning_rate": 1.990854691896325e-05, + "loss": 1.4522, + "step": 2558 + }, + { + "epoch": 0.22, + "learning_rate": 1.9908424227350012e-05, + "loss": 1.3293, + "step": 2559 + }, + { + "epoch": 0.22, + "learning_rate": 1.99083014538702e-05, + "loss": 1.4224, + "step": 2560 + }, + { + "epoch": 0.22, + "learning_rate": 1.990817859852483e-05, + "loss": 1.4414, + "step": 2561 + }, + { + "epoch": 0.22, + "learning_rate": 1.990805566131492e-05, + "loss": 1.3426, + "step": 2562 + }, + { + "epoch": 0.22, + "learning_rate": 1.9907932642241482e-05, + "loss": 1.4732, + "step": 2563 + }, + { + "epoch": 0.22, + "learning_rate": 1.9907809541305533e-05, + "loss": 1.432, + "step": 2564 + }, + { + "epoch": 0.22, + "learning_rate": 1.990768635850809e-05, + "loss": 1.3697, + "step": 2565 + }, + { + "epoch": 0.22, + "learning_rate": 1.990756309385017e-05, + "loss": 1.3376, + "step": 2566 + }, + { + "epoch": 0.22, + "learning_rate": 1.9907439747332797e-05, + "loss": 1.4131, + "step": 2567 + }, + { + "epoch": 0.22, + "learning_rate": 1.9907316318956984e-05, + "loss": 1.3559, + "step": 2568 + }, + { + "epoch": 0.22, + "learning_rate": 1.990719280872375e-05, + "loss": 1.3408, + "step": 2569 + }, + { + "epoch": 0.22, + "learning_rate": 1.9907069216634124e-05, + "loss": 1.4198, + "step": 2570 + }, + { + "epoch": 0.22, + "learning_rate": 1.9906945542689117e-05, + "loss": 1.2947, + "step": 2571 + }, + { + "epoch": 0.22, + "learning_rate": 1.9906821786889757e-05, + "loss": 1.346, + "step": 2572 + }, + { + "epoch": 0.22, + "learning_rate": 1.9906697949237066e-05, + "loss": 1.2835, + "step": 2573 + }, + { + "epoch": 0.22, + "learning_rate": 1.9906574029732065e-05, + "loss": 1.367, + "step": 2574 + }, + { + "epoch": 0.22, + "learning_rate": 1.9906450028375776e-05, + "loss": 1.3116, + "step": 2575 + }, + { + "epoch": 0.22, + "learning_rate": 1.990632594516923e-05, + "loss": 1.2893, + "step": 2576 + }, + { + "epoch": 0.22, + "learning_rate": 1.9906201780113447e-05, + "loss": 1.3494, + "step": 2577 + }, + { + "epoch": 0.22, + "learning_rate": 1.9906077533209456e-05, + "loss": 1.2492, + "step": 2578 + }, + { + "epoch": 0.22, + "learning_rate": 1.990595320445828e-05, + "loss": 1.4439, + "step": 2579 + }, + { + "epoch": 0.22, + "learning_rate": 1.990582879386095e-05, + "loss": 1.414, + "step": 2580 + }, + { + "epoch": 0.22, + "learning_rate": 1.9905704301418493e-05, + "loss": 1.2564, + "step": 2581 + }, + { + "epoch": 0.22, + "learning_rate": 1.9905579727131938e-05, + "loss": 1.4596, + "step": 2582 + }, + { + "epoch": 0.22, + "learning_rate": 1.9905455071002308e-05, + "loss": 1.3896, + "step": 2583 + }, + { + "epoch": 0.22, + "learning_rate": 1.990533033303064e-05, + "loss": 1.3668, + "step": 2584 + }, + { + "epoch": 0.22, + "learning_rate": 1.9905205513217965e-05, + "loss": 1.2861, + "step": 2585 + }, + { + "epoch": 0.22, + "learning_rate": 1.990508061156531e-05, + "loss": 1.2384, + "step": 2586 + }, + { + "epoch": 0.22, + "learning_rate": 1.9904955628073708e-05, + "loss": 1.3319, + "step": 2587 + }, + { + "epoch": 0.22, + "learning_rate": 1.9904830562744192e-05, + "loss": 1.3708, + "step": 2588 + }, + { + "epoch": 0.22, + "learning_rate": 1.99047054155778e-05, + "loss": 1.3438, + "step": 2589 + }, + { + "epoch": 0.22, + "learning_rate": 1.990458018657556e-05, + "loss": 1.4065, + "step": 2590 + }, + { + "epoch": 0.22, + "learning_rate": 1.9904454875738507e-05, + "loss": 1.3394, + "step": 2591 + }, + { + "epoch": 0.22, + "learning_rate": 1.9904329483067682e-05, + "loss": 1.3362, + "step": 2592 + }, + { + "epoch": 0.22, + "learning_rate": 1.9904204008564112e-05, + "loss": 1.3257, + "step": 2593 + }, + { + "epoch": 0.22, + "learning_rate": 1.9904078452228842e-05, + "loss": 1.4119, + "step": 2594 + }, + { + "epoch": 0.22, + "learning_rate": 1.9903952814062905e-05, + "loss": 1.3075, + "step": 2595 + }, + { + "epoch": 0.22, + "learning_rate": 1.990382709406734e-05, + "loss": 1.3804, + "step": 2596 + }, + { + "epoch": 0.22, + "learning_rate": 1.990370129224319e-05, + "loss": 1.3354, + "step": 2597 + }, + { + "epoch": 0.22, + "learning_rate": 1.9903575408591484e-05, + "loss": 1.4304, + "step": 2598 + }, + { + "epoch": 0.22, + "learning_rate": 1.990344944311327e-05, + "loss": 1.4296, + "step": 2599 + }, + { + "epoch": 0.22, + "learning_rate": 1.990332339580959e-05, + "loss": 1.37, + "step": 2600 + }, + { + "epoch": 0.22, + "learning_rate": 1.990319726668148e-05, + "loss": 1.39, + "step": 2601 + }, + { + "epoch": 0.22, + "learning_rate": 1.9903071055729987e-05, + "loss": 1.4044, + "step": 2602 + }, + { + "epoch": 0.22, + "learning_rate": 1.9902944762956148e-05, + "loss": 1.3008, + "step": 2603 + }, + { + "epoch": 0.22, + "learning_rate": 1.9902818388361013e-05, + "loss": 1.3439, + "step": 2604 + }, + { + "epoch": 0.22, + "learning_rate": 1.9902691931945622e-05, + "loss": 1.4232, + "step": 2605 + }, + { + "epoch": 0.22, + "learning_rate": 1.9902565393711023e-05, + "loss": 1.4228, + "step": 2606 + }, + { + "epoch": 0.22, + "learning_rate": 1.9902438773658256e-05, + "loss": 1.4633, + "step": 2607 + }, + { + "epoch": 0.22, + "learning_rate": 1.9902312071788374e-05, + "loss": 1.3081, + "step": 2608 + }, + { + "epoch": 0.22, + "learning_rate": 1.990218528810242e-05, + "loss": 1.4799, + "step": 2609 + }, + { + "epoch": 0.22, + "learning_rate": 1.990205842260144e-05, + "loss": 1.3702, + "step": 2610 + }, + { + "epoch": 0.22, + "learning_rate": 1.9901931475286485e-05, + "loss": 1.4109, + "step": 2611 + }, + { + "epoch": 0.22, + "learning_rate": 1.9901804446158604e-05, + "loss": 1.3114, + "step": 2612 + }, + { + "epoch": 0.22, + "learning_rate": 1.9901677335218845e-05, + "loss": 1.4071, + "step": 2613 + }, + { + "epoch": 0.22, + "learning_rate": 1.9901550142468258e-05, + "loss": 1.3949, + "step": 2614 + }, + { + "epoch": 0.22, + "learning_rate": 1.9901422867907897e-05, + "loss": 1.3693, + "step": 2615 + }, + { + "epoch": 0.22, + "learning_rate": 1.990129551153881e-05, + "loss": 1.3254, + "step": 2616 + }, + { + "epoch": 0.22, + "learning_rate": 1.9901168073362053e-05, + "loss": 1.3585, + "step": 2617 + }, + { + "epoch": 0.22, + "learning_rate": 1.9901040553378673e-05, + "loss": 1.2961, + "step": 2618 + }, + { + "epoch": 0.22, + "learning_rate": 1.990091295158973e-05, + "loss": 1.3615, + "step": 2619 + }, + { + "epoch": 0.22, + "learning_rate": 1.990078526799627e-05, + "loss": 1.3486, + "step": 2620 + }, + { + "epoch": 0.22, + "learning_rate": 1.9900657502599362e-05, + "loss": 1.3771, + "step": 2621 + }, + { + "epoch": 0.22, + "learning_rate": 1.9900529655400047e-05, + "loss": 1.2784, + "step": 2622 + }, + { + "epoch": 0.22, + "learning_rate": 1.990040172639939e-05, + "loss": 1.4549, + "step": 2623 + }, + { + "epoch": 0.22, + "learning_rate": 1.9900273715598444e-05, + "loss": 1.4109, + "step": 2624 + }, + { + "epoch": 0.22, + "learning_rate": 1.9900145622998268e-05, + "loss": 1.2936, + "step": 2625 + }, + { + "epoch": 0.22, + "learning_rate": 1.9900017448599924e-05, + "loss": 1.3851, + "step": 2626 + }, + { + "epoch": 0.22, + "learning_rate": 1.989988919240446e-05, + "loss": 1.3496, + "step": 2627 + }, + { + "epoch": 0.22, + "learning_rate": 1.989976085441295e-05, + "loss": 1.3421, + "step": 2628 + }, + { + "epoch": 0.22, + "learning_rate": 1.9899632434626445e-05, + "loss": 1.3627, + "step": 2629 + }, + { + "epoch": 0.22, + "learning_rate": 1.989950393304601e-05, + "loss": 1.3142, + "step": 2630 + }, + { + "epoch": 0.22, + "learning_rate": 1.9899375349672703e-05, + "loss": 1.3664, + "step": 2631 + }, + { + "epoch": 0.22, + "learning_rate": 1.989924668450759e-05, + "loss": 1.381, + "step": 2632 + }, + { + "epoch": 0.22, + "learning_rate": 1.989911793755173e-05, + "loss": 1.3115, + "step": 2633 + }, + { + "epoch": 0.22, + "learning_rate": 1.9898989108806194e-05, + "loss": 1.3149, + "step": 2634 + }, + { + "epoch": 0.22, + "learning_rate": 1.989886019827204e-05, + "loss": 1.4951, + "step": 2635 + }, + { + "epoch": 0.22, + "learning_rate": 1.9898731205950333e-05, + "loss": 1.4672, + "step": 2636 + }, + { + "epoch": 0.22, + "learning_rate": 1.9898602131842144e-05, + "loss": 1.3485, + "step": 2637 + }, + { + "epoch": 0.22, + "learning_rate": 1.9898472975948536e-05, + "loss": 1.3113, + "step": 2638 + }, + { + "epoch": 0.22, + "learning_rate": 1.9898343738270573e-05, + "loss": 1.4765, + "step": 2639 + }, + { + "epoch": 0.22, + "learning_rate": 1.989821441880933e-05, + "loss": 1.3067, + "step": 2640 + }, + { + "epoch": 0.22, + "learning_rate": 1.9898085017565868e-05, + "loss": 1.3783, + "step": 2641 + }, + { + "epoch": 0.22, + "learning_rate": 1.989795553454126e-05, + "loss": 1.403, + "step": 2642 + }, + { + "epoch": 0.22, + "learning_rate": 1.989782596973658e-05, + "loss": 1.3777, + "step": 2643 + }, + { + "epoch": 0.22, + "learning_rate": 1.989769632315289e-05, + "loss": 1.3394, + "step": 2644 + }, + { + "epoch": 0.22, + "learning_rate": 1.989756659479127e-05, + "loss": 1.378, + "step": 2645 + }, + { + "epoch": 0.22, + "learning_rate": 1.989743678465278e-05, + "loss": 1.286, + "step": 2646 + }, + { + "epoch": 0.22, + "learning_rate": 1.9897306892738506e-05, + "loss": 1.2871, + "step": 2647 + }, + { + "epoch": 0.22, + "learning_rate": 1.9897176919049512e-05, + "loss": 1.4133, + "step": 2648 + }, + { + "epoch": 0.22, + "learning_rate": 1.9897046863586877e-05, + "loss": 1.3843, + "step": 2649 + }, + { + "epoch": 0.22, + "learning_rate": 1.989691672635167e-05, + "loss": 1.3272, + "step": 2650 + }, + { + "epoch": 0.22, + "learning_rate": 1.989678650734497e-05, + "loss": 1.3875, + "step": 2651 + }, + { + "epoch": 0.22, + "learning_rate": 1.9896656206567854e-05, + "loss": 1.3354, + "step": 2652 + }, + { + "epoch": 0.22, + "learning_rate": 1.9896525824021397e-05, + "loss": 1.39, + "step": 2653 + }, + { + "epoch": 0.22, + "learning_rate": 1.9896395359706677e-05, + "loss": 1.366, + "step": 2654 + }, + { + "epoch": 0.22, + "learning_rate": 1.989626481362477e-05, + "loss": 1.2973, + "step": 2655 + }, + { + "epoch": 0.22, + "learning_rate": 1.9896134185776755e-05, + "loss": 1.365, + "step": 2656 + }, + { + "epoch": 0.22, + "learning_rate": 1.9896003476163717e-05, + "loss": 1.3966, + "step": 2657 + }, + { + "epoch": 0.22, + "learning_rate": 1.9895872684786724e-05, + "loss": 1.3597, + "step": 2658 + }, + { + "epoch": 0.22, + "learning_rate": 1.989574181164687e-05, + "loss": 1.4544, + "step": 2659 + }, + { + "epoch": 0.22, + "learning_rate": 1.9895610856745227e-05, + "loss": 1.4722, + "step": 2660 + }, + { + "epoch": 0.22, + "learning_rate": 1.9895479820082882e-05, + "loss": 1.3369, + "step": 2661 + }, + { + "epoch": 0.22, + "learning_rate": 1.989534870166091e-05, + "loss": 1.3688, + "step": 2662 + }, + { + "epoch": 0.22, + "learning_rate": 1.9895217501480407e-05, + "loss": 1.3943, + "step": 2663 + }, + { + "epoch": 0.22, + "learning_rate": 1.989508621954245e-05, + "loss": 1.3406, + "step": 2664 + }, + { + "epoch": 0.22, + "learning_rate": 1.989495485584812e-05, + "loss": 1.4097, + "step": 2665 + }, + { + "epoch": 0.22, + "learning_rate": 1.9894823410398506e-05, + "loss": 1.3185, + "step": 2666 + }, + { + "epoch": 0.22, + "learning_rate": 1.9894691883194698e-05, + "loss": 1.3713, + "step": 2667 + }, + { + "epoch": 0.22, + "learning_rate": 1.9894560274237777e-05, + "loss": 1.2486, + "step": 2668 + }, + { + "epoch": 0.22, + "learning_rate": 1.989442858352883e-05, + "loss": 1.2689, + "step": 2669 + }, + { + "epoch": 0.22, + "learning_rate": 1.989429681106895e-05, + "loss": 1.3456, + "step": 2670 + }, + { + "epoch": 0.22, + "learning_rate": 1.9894164956859224e-05, + "loss": 1.3761, + "step": 2671 + }, + { + "epoch": 0.22, + "learning_rate": 1.9894033020900738e-05, + "loss": 1.3175, + "step": 2672 + }, + { + "epoch": 0.23, + "learning_rate": 1.989390100319459e-05, + "loss": 1.4024, + "step": 2673 + }, + { + "epoch": 0.23, + "learning_rate": 1.989376890374186e-05, + "loss": 1.3721, + "step": 2674 + }, + { + "epoch": 0.23, + "learning_rate": 1.9893636722543645e-05, + "loss": 1.2685, + "step": 2675 + }, + { + "epoch": 0.23, + "learning_rate": 1.9893504459601042e-05, + "loss": 1.3452, + "step": 2676 + }, + { + "epoch": 0.23, + "learning_rate": 1.9893372114915132e-05, + "loss": 1.2726, + "step": 2677 + }, + { + "epoch": 0.23, + "learning_rate": 1.989323968848702e-05, + "loss": 1.4162, + "step": 2678 + }, + { + "epoch": 0.23, + "learning_rate": 1.9893107180317794e-05, + "loss": 1.3393, + "step": 2679 + }, + { + "epoch": 0.23, + "learning_rate": 1.989297459040855e-05, + "loss": 1.4172, + "step": 2680 + }, + { + "epoch": 0.23, + "learning_rate": 1.9892841918760384e-05, + "loss": 1.38, + "step": 2681 + }, + { + "epoch": 0.23, + "learning_rate": 1.9892709165374392e-05, + "loss": 1.3622, + "step": 2682 + }, + { + "epoch": 0.23, + "learning_rate": 1.989257633025167e-05, + "loss": 1.3991, + "step": 2683 + }, + { + "epoch": 0.23, + "learning_rate": 1.989244341339332e-05, + "loss": 1.297, + "step": 2684 + }, + { + "epoch": 0.23, + "learning_rate": 1.9892310414800433e-05, + "loss": 1.3456, + "step": 2685 + }, + { + "epoch": 0.23, + "learning_rate": 1.989217733447411e-05, + "loss": 1.3598, + "step": 2686 + }, + { + "epoch": 0.23, + "learning_rate": 1.9892044172415456e-05, + "loss": 1.4006, + "step": 2687 + }, + { + "epoch": 0.23, + "learning_rate": 1.9891910928625563e-05, + "loss": 1.3996, + "step": 2688 + }, + { + "epoch": 0.23, + "learning_rate": 1.989177760310554e-05, + "loss": 1.3525, + "step": 2689 + }, + { + "epoch": 0.23, + "learning_rate": 1.9891644195856482e-05, + "loss": 1.3404, + "step": 2690 + }, + { + "epoch": 0.23, + "learning_rate": 1.9891510706879496e-05, + "loss": 1.4296, + "step": 2691 + }, + { + "epoch": 0.23, + "learning_rate": 1.9891377136175682e-05, + "loss": 1.3478, + "step": 2692 + }, + { + "epoch": 0.23, + "learning_rate": 1.9891243483746145e-05, + "loss": 1.3862, + "step": 2693 + }, + { + "epoch": 0.23, + "learning_rate": 1.9891109749591986e-05, + "loss": 1.3253, + "step": 2694 + }, + { + "epoch": 0.23, + "learning_rate": 1.9890975933714316e-05, + "loss": 1.3772, + "step": 2695 + }, + { + "epoch": 0.23, + "learning_rate": 1.9890842036114237e-05, + "loss": 1.3521, + "step": 2696 + }, + { + "epoch": 0.23, + "learning_rate": 1.9890708056792852e-05, + "loss": 1.397, + "step": 2697 + }, + { + "epoch": 0.23, + "learning_rate": 1.9890573995751277e-05, + "loss": 1.396, + "step": 2698 + }, + { + "epoch": 0.23, + "learning_rate": 1.989043985299061e-05, + "loss": 1.4177, + "step": 2699 + }, + { + "epoch": 0.23, + "learning_rate": 1.9890305628511964e-05, + "loss": 1.3191, + "step": 2700 + }, + { + "epoch": 0.23, + "learning_rate": 1.989017132231645e-05, + "loss": 1.4164, + "step": 2701 + }, + { + "epoch": 0.23, + "learning_rate": 1.9890036934405174e-05, + "loss": 1.4333, + "step": 2702 + }, + { + "epoch": 0.23, + "learning_rate": 1.9889902464779246e-05, + "loss": 1.4393, + "step": 2703 + }, + { + "epoch": 0.23, + "learning_rate": 1.9889767913439782e-05, + "loss": 1.4439, + "step": 2704 + }, + { + "epoch": 0.23, + "learning_rate": 1.988963328038789e-05, + "loss": 1.3509, + "step": 2705 + }, + { + "epoch": 0.23, + "learning_rate": 1.988949856562468e-05, + "loss": 1.3959, + "step": 2706 + }, + { + "epoch": 0.23, + "learning_rate": 1.988936376915127e-05, + "loss": 1.3276, + "step": 2707 + }, + { + "epoch": 0.23, + "learning_rate": 1.988922889096877e-05, + "loss": 1.3236, + "step": 2708 + }, + { + "epoch": 0.23, + "learning_rate": 1.9889093931078297e-05, + "loss": 1.3371, + "step": 2709 + }, + { + "epoch": 0.23, + "learning_rate": 1.9888958889480964e-05, + "loss": 1.4746, + "step": 2710 + }, + { + "epoch": 0.23, + "learning_rate": 1.988882376617789e-05, + "loss": 1.2917, + "step": 2711 + }, + { + "epoch": 0.23, + "learning_rate": 1.9888688561170185e-05, + "loss": 1.3839, + "step": 2712 + }, + { + "epoch": 0.23, + "learning_rate": 1.9888553274458976e-05, + "loss": 1.3623, + "step": 2713 + }, + { + "epoch": 0.23, + "learning_rate": 1.988841790604537e-05, + "loss": 1.2929, + "step": 2714 + }, + { + "epoch": 0.23, + "learning_rate": 1.9888282455930495e-05, + "loss": 1.2891, + "step": 2715 + }, + { + "epoch": 0.23, + "learning_rate": 1.9888146924115462e-05, + "loss": 1.3441, + "step": 2716 + }, + { + "epoch": 0.23, + "learning_rate": 1.9888011310601395e-05, + "loss": 1.2951, + "step": 2717 + }, + { + "epoch": 0.23, + "learning_rate": 1.9887875615389417e-05, + "loss": 1.3755, + "step": 2718 + }, + { + "epoch": 0.23, + "learning_rate": 1.988773983848064e-05, + "loss": 1.3231, + "step": 2719 + }, + { + "epoch": 0.23, + "learning_rate": 1.98876039798762e-05, + "loss": 1.4184, + "step": 2720 + }, + { + "epoch": 0.23, + "learning_rate": 1.9887468039577206e-05, + "loss": 1.4334, + "step": 2721 + }, + { + "epoch": 0.23, + "learning_rate": 1.988733201758479e-05, + "loss": 1.3928, + "step": 2722 + }, + { + "epoch": 0.23, + "learning_rate": 1.9887195913900067e-05, + "loss": 1.3649, + "step": 2723 + }, + { + "epoch": 0.23, + "learning_rate": 1.988705972852417e-05, + "loss": 1.4668, + "step": 2724 + }, + { + "epoch": 0.23, + "learning_rate": 1.9886923461458224e-05, + "loss": 1.3602, + "step": 2725 + }, + { + "epoch": 0.23, + "learning_rate": 1.988678711270335e-05, + "loss": 1.4299, + "step": 2726 + }, + { + "epoch": 0.23, + "learning_rate": 1.988665068226068e-05, + "loss": 1.3916, + "step": 2727 + }, + { + "epoch": 0.23, + "learning_rate": 1.988651417013133e-05, + "loss": 1.4208, + "step": 2728 + }, + { + "epoch": 0.23, + "learning_rate": 1.9886377576316444e-05, + "loss": 1.3793, + "step": 2729 + }, + { + "epoch": 0.23, + "learning_rate": 1.9886240900817136e-05, + "loss": 1.4008, + "step": 2730 + }, + { + "epoch": 0.23, + "learning_rate": 1.9886104143634545e-05, + "loss": 1.3447, + "step": 2731 + }, + { + "epoch": 0.23, + "learning_rate": 1.98859673047698e-05, + "loss": 1.3817, + "step": 2732 + }, + { + "epoch": 0.23, + "learning_rate": 1.9885830384224025e-05, + "loss": 1.3874, + "step": 2733 + }, + { + "epoch": 0.23, + "learning_rate": 1.9885693381998357e-05, + "loss": 1.3883, + "step": 2734 + }, + { + "epoch": 0.23, + "learning_rate": 1.9885556298093928e-05, + "loss": 1.3886, + "step": 2735 + }, + { + "epoch": 0.23, + "learning_rate": 1.9885419132511863e-05, + "loss": 1.3231, + "step": 2736 + }, + { + "epoch": 0.23, + "learning_rate": 1.9885281885253305e-05, + "loss": 1.3816, + "step": 2737 + }, + { + "epoch": 0.23, + "learning_rate": 1.988514455631939e-05, + "loss": 1.42, + "step": 2738 + }, + { + "epoch": 0.23, + "learning_rate": 1.9885007145711242e-05, + "loss": 1.3611, + "step": 2739 + }, + { + "epoch": 0.23, + "learning_rate": 1.988486965343e-05, + "loss": 1.3167, + "step": 2740 + }, + { + "epoch": 0.23, + "learning_rate": 1.9884732079476804e-05, + "loss": 1.367, + "step": 2741 + }, + { + "epoch": 0.23, + "learning_rate": 1.9884594423852786e-05, + "loss": 1.3397, + "step": 2742 + }, + { + "epoch": 0.23, + "learning_rate": 1.9884456686559085e-05, + "loss": 1.378, + "step": 2743 + }, + { + "epoch": 0.23, + "learning_rate": 1.9884318867596845e-05, + "loss": 1.3543, + "step": 2744 + }, + { + "epoch": 0.23, + "learning_rate": 1.9884180966967192e-05, + "loss": 1.3712, + "step": 2745 + }, + { + "epoch": 0.23, + "learning_rate": 1.9884042984671278e-05, + "loss": 1.3989, + "step": 2746 + }, + { + "epoch": 0.23, + "learning_rate": 1.9883904920710233e-05, + "loss": 1.4428, + "step": 2747 + }, + { + "epoch": 0.23, + "learning_rate": 1.9883766775085204e-05, + "loss": 1.4462, + "step": 2748 + }, + { + "epoch": 0.23, + "learning_rate": 1.9883628547797332e-05, + "loss": 1.2715, + "step": 2749 + }, + { + "epoch": 0.23, + "learning_rate": 1.9883490238847756e-05, + "loss": 1.3107, + "step": 2750 + }, + { + "epoch": 0.23, + "learning_rate": 1.988335184823762e-05, + "loss": 1.3537, + "step": 2751 + }, + { + "epoch": 0.23, + "learning_rate": 1.988321337596807e-05, + "loss": 1.2515, + "step": 2752 + }, + { + "epoch": 0.23, + "learning_rate": 1.9883074822040246e-05, + "loss": 1.4282, + "step": 2753 + }, + { + "epoch": 0.23, + "learning_rate": 1.9882936186455297e-05, + "loss": 1.3368, + "step": 2754 + }, + { + "epoch": 0.23, + "learning_rate": 1.988279746921437e-05, + "loss": 1.3285, + "step": 2755 + }, + { + "epoch": 0.23, + "learning_rate": 1.98826586703186e-05, + "loss": 1.3533, + "step": 2756 + }, + { + "epoch": 0.23, + "learning_rate": 1.9882519789769143e-05, + "loss": 1.3422, + "step": 2757 + }, + { + "epoch": 0.23, + "learning_rate": 1.9882380827567147e-05, + "loss": 1.3634, + "step": 2758 + }, + { + "epoch": 0.23, + "learning_rate": 1.9882241783713754e-05, + "loss": 1.3737, + "step": 2759 + }, + { + "epoch": 0.23, + "learning_rate": 1.988210265821012e-05, + "loss": 1.3587, + "step": 2760 + }, + { + "epoch": 0.23, + "learning_rate": 1.988196345105739e-05, + "loss": 1.4105, + "step": 2761 + }, + { + "epoch": 0.23, + "learning_rate": 1.988182416225672e-05, + "loss": 1.2746, + "step": 2762 + }, + { + "epoch": 0.23, + "learning_rate": 1.9881684791809248e-05, + "loss": 1.3866, + "step": 2763 + }, + { + "epoch": 0.23, + "learning_rate": 1.988154533971614e-05, + "loss": 1.337, + "step": 2764 + }, + { + "epoch": 0.23, + "learning_rate": 1.988140580597854e-05, + "loss": 1.3754, + "step": 2765 + }, + { + "epoch": 0.23, + "learning_rate": 1.98812661905976e-05, + "loss": 1.3519, + "step": 2766 + }, + { + "epoch": 0.23, + "learning_rate": 1.988112649357448e-05, + "loss": 1.4271, + "step": 2767 + }, + { + "epoch": 0.23, + "learning_rate": 1.988098671491033e-05, + "loss": 1.323, + "step": 2768 + }, + { + "epoch": 0.23, + "learning_rate": 1.9880846854606303e-05, + "loss": 1.3058, + "step": 2769 + }, + { + "epoch": 0.23, + "learning_rate": 1.9880706912663558e-05, + "loss": 1.3859, + "step": 2770 + }, + { + "epoch": 0.23, + "learning_rate": 1.9880566889083252e-05, + "loss": 1.2955, + "step": 2771 + }, + { + "epoch": 0.23, + "learning_rate": 1.9880426783866537e-05, + "loss": 1.3738, + "step": 2772 + }, + { + "epoch": 0.23, + "learning_rate": 1.9880286597014576e-05, + "loss": 1.2992, + "step": 2773 + }, + { + "epoch": 0.23, + "learning_rate": 1.9880146328528527e-05, + "loss": 1.3981, + "step": 2774 + }, + { + "epoch": 0.23, + "learning_rate": 1.9880005978409546e-05, + "loss": 1.357, + "step": 2775 + }, + { + "epoch": 0.23, + "learning_rate": 1.987986554665879e-05, + "loss": 1.3962, + "step": 2776 + }, + { + "epoch": 0.23, + "learning_rate": 1.9879725033277425e-05, + "loss": 1.2344, + "step": 2777 + }, + { + "epoch": 0.23, + "learning_rate": 1.9879584438266612e-05, + "loss": 1.4773, + "step": 2778 + }, + { + "epoch": 0.23, + "learning_rate": 1.987944376162751e-05, + "loss": 1.3161, + "step": 2779 + }, + { + "epoch": 0.23, + "learning_rate": 1.987930300336128e-05, + "loss": 1.4245, + "step": 2780 + }, + { + "epoch": 0.23, + "learning_rate": 1.9879162163469085e-05, + "loss": 1.3404, + "step": 2781 + }, + { + "epoch": 0.23, + "learning_rate": 1.9879021241952093e-05, + "loss": 1.273, + "step": 2782 + }, + { + "epoch": 0.23, + "learning_rate": 1.9878880238811465e-05, + "loss": 1.3111, + "step": 2783 + }, + { + "epoch": 0.23, + "learning_rate": 1.987873915404837e-05, + "loss": 1.278, + "step": 2784 + }, + { + "epoch": 0.23, + "learning_rate": 1.9878597987663966e-05, + "loss": 1.3042, + "step": 2785 + }, + { + "epoch": 0.23, + "learning_rate": 1.9878456739659425e-05, + "loss": 1.4044, + "step": 2786 + }, + { + "epoch": 0.23, + "learning_rate": 1.9878315410035914e-05, + "loss": 1.4158, + "step": 2787 + }, + { + "epoch": 0.23, + "learning_rate": 1.9878173998794602e-05, + "loss": 1.4256, + "step": 2788 + }, + { + "epoch": 0.23, + "learning_rate": 1.987803250593665e-05, + "loss": 1.3545, + "step": 2789 + }, + { + "epoch": 0.23, + "learning_rate": 1.9877890931463234e-05, + "loss": 1.3846, + "step": 2790 + }, + { + "epoch": 0.23, + "learning_rate": 1.9877749275375522e-05, + "loss": 1.382, + "step": 2791 + }, + { + "epoch": 0.24, + "learning_rate": 1.9877607537674685e-05, + "loss": 1.3062, + "step": 2792 + }, + { + "epoch": 0.24, + "learning_rate": 1.987746571836189e-05, + "loss": 1.3719, + "step": 2793 + }, + { + "epoch": 0.24, + "learning_rate": 1.9877323817438317e-05, + "loss": 1.3845, + "step": 2794 + }, + { + "epoch": 0.24, + "learning_rate": 1.9877181834905134e-05, + "loss": 1.321, + "step": 2795 + }, + { + "epoch": 0.24, + "learning_rate": 1.987703977076351e-05, + "loss": 1.3315, + "step": 2796 + }, + { + "epoch": 0.24, + "learning_rate": 1.9876897625014622e-05, + "loss": 1.3275, + "step": 2797 + }, + { + "epoch": 0.24, + "learning_rate": 1.9876755397659647e-05, + "loss": 1.3122, + "step": 2798 + }, + { + "epoch": 0.24, + "learning_rate": 1.9876613088699756e-05, + "loss": 1.3542, + "step": 2799 + }, + { + "epoch": 0.24, + "learning_rate": 1.987647069813613e-05, + "loss": 1.4358, + "step": 2800 + }, + { + "epoch": 0.24, + "learning_rate": 1.987632822596994e-05, + "loss": 1.2226, + "step": 2801 + }, + { + "epoch": 0.24, + "learning_rate": 1.9876185672202362e-05, + "loss": 1.3663, + "step": 2802 + }, + { + "epoch": 0.24, + "learning_rate": 1.987604303683458e-05, + "loss": 1.3766, + "step": 2803 + }, + { + "epoch": 0.24, + "learning_rate": 1.987590031986777e-05, + "loss": 1.4013, + "step": 2804 + }, + { + "epoch": 0.24, + "learning_rate": 1.9875757521303113e-05, + "loss": 1.3454, + "step": 2805 + }, + { + "epoch": 0.24, + "learning_rate": 1.9875614641141784e-05, + "loss": 1.3837, + "step": 2806 + }, + { + "epoch": 0.24, + "learning_rate": 1.9875471679384965e-05, + "loss": 1.2857, + "step": 2807 + }, + { + "epoch": 0.24, + "learning_rate": 1.987532863603384e-05, + "loss": 1.3859, + "step": 2808 + }, + { + "epoch": 0.24, + "learning_rate": 1.9875185511089592e-05, + "loss": 1.3864, + "step": 2809 + }, + { + "epoch": 0.24, + "learning_rate": 1.9875042304553397e-05, + "loss": 1.3996, + "step": 2810 + }, + { + "epoch": 0.24, + "learning_rate": 1.9874899016426445e-05, + "loss": 1.3439, + "step": 2811 + }, + { + "epoch": 0.24, + "learning_rate": 1.9874755646709915e-05, + "loss": 1.3743, + "step": 2812 + }, + { + "epoch": 0.24, + "learning_rate": 1.9874612195404996e-05, + "loss": 1.331, + "step": 2813 + }, + { + "epoch": 0.24, + "learning_rate": 1.9874468662512868e-05, + "loss": 1.1751, + "step": 2814 + }, + { + "epoch": 0.24, + "learning_rate": 1.987432504803472e-05, + "loss": 1.2518, + "step": 2815 + }, + { + "epoch": 0.24, + "learning_rate": 1.987418135197174e-05, + "loss": 1.3689, + "step": 2816 + }, + { + "epoch": 0.24, + "learning_rate": 1.987403757432511e-05, + "loss": 1.3156, + "step": 2817 + }, + { + "epoch": 0.24, + "learning_rate": 1.9873893715096023e-05, + "loss": 1.4319, + "step": 2818 + }, + { + "epoch": 0.24, + "learning_rate": 1.9873749774285668e-05, + "loss": 1.3269, + "step": 2819 + }, + { + "epoch": 0.24, + "learning_rate": 1.987360575189523e-05, + "loss": 1.4578, + "step": 2820 + }, + { + "epoch": 0.24, + "learning_rate": 1.9873461647925902e-05, + "loss": 1.2816, + "step": 2821 + }, + { + "epoch": 0.24, + "learning_rate": 1.9873317462378875e-05, + "loss": 1.3448, + "step": 2822 + }, + { + "epoch": 0.24, + "learning_rate": 1.987317319525534e-05, + "loss": 1.3292, + "step": 2823 + }, + { + "epoch": 0.24, + "learning_rate": 1.9873028846556485e-05, + "loss": 1.2444, + "step": 2824 + }, + { + "epoch": 0.24, + "learning_rate": 1.9872884416283507e-05, + "loss": 1.3871, + "step": 2825 + }, + { + "epoch": 0.24, + "learning_rate": 1.98727399044376e-05, + "loss": 1.3342, + "step": 2826 + }, + { + "epoch": 0.24, + "learning_rate": 1.9872595311019954e-05, + "loss": 1.3446, + "step": 2827 + }, + { + "epoch": 0.24, + "learning_rate": 1.9872450636031768e-05, + "loss": 1.3665, + "step": 2828 + }, + { + "epoch": 0.24, + "learning_rate": 1.9872305879474234e-05, + "loss": 1.3866, + "step": 2829 + }, + { + "epoch": 0.24, + "learning_rate": 1.987216104134855e-05, + "loss": 1.2593, + "step": 2830 + }, + { + "epoch": 0.24, + "learning_rate": 1.987201612165591e-05, + "loss": 1.3836, + "step": 2831 + }, + { + "epoch": 0.24, + "learning_rate": 1.9871871120397516e-05, + "loss": 1.2868, + "step": 2832 + }, + { + "epoch": 0.24, + "learning_rate": 1.9871726037574564e-05, + "loss": 1.3349, + "step": 2833 + }, + { + "epoch": 0.24, + "learning_rate": 1.987158087318825e-05, + "loss": 1.3588, + "step": 2834 + }, + { + "epoch": 0.24, + "learning_rate": 1.9871435627239776e-05, + "loss": 1.3567, + "step": 2835 + }, + { + "epoch": 0.24, + "learning_rate": 1.987129029973034e-05, + "loss": 1.4423, + "step": 2836 + }, + { + "epoch": 0.24, + "learning_rate": 1.987114489066115e-05, + "loss": 1.3655, + "step": 2837 + }, + { + "epoch": 0.24, + "learning_rate": 1.9870999400033394e-05, + "loss": 1.2856, + "step": 2838 + }, + { + "epoch": 0.24, + "learning_rate": 1.9870853827848286e-05, + "loss": 1.2844, + "step": 2839 + }, + { + "epoch": 0.24, + "learning_rate": 1.9870708174107024e-05, + "loss": 1.3867, + "step": 2840 + }, + { + "epoch": 0.24, + "learning_rate": 1.987056243881081e-05, + "loss": 1.3128, + "step": 2841 + }, + { + "epoch": 0.24, + "learning_rate": 1.9870416621960856e-05, + "loss": 1.3569, + "step": 2842 + }, + { + "epoch": 0.24, + "learning_rate": 1.9870270723558356e-05, + "loss": 1.384, + "step": 2843 + }, + { + "epoch": 0.24, + "learning_rate": 1.9870124743604523e-05, + "loss": 1.4192, + "step": 2844 + }, + { + "epoch": 0.24, + "learning_rate": 1.9869978682100556e-05, + "loss": 1.3578, + "step": 2845 + }, + { + "epoch": 0.24, + "learning_rate": 1.986983253904767e-05, + "loss": 1.3502, + "step": 2846 + }, + { + "epoch": 0.24, + "learning_rate": 1.986968631444707e-05, + "loss": 1.3478, + "step": 2847 + }, + { + "epoch": 0.24, + "learning_rate": 1.986954000829996e-05, + "loss": 1.3849, + "step": 2848 + }, + { + "epoch": 0.24, + "learning_rate": 1.9869393620607553e-05, + "loss": 1.3497, + "step": 2849 + }, + { + "epoch": 0.24, + "learning_rate": 1.9869247151371057e-05, + "loss": 1.2845, + "step": 2850 + }, + { + "epoch": 0.24, + "learning_rate": 1.9869100600591685e-05, + "loss": 1.3563, + "step": 2851 + }, + { + "epoch": 0.24, + "learning_rate": 1.9868953968270644e-05, + "loss": 1.3213, + "step": 2852 + }, + { + "epoch": 0.24, + "learning_rate": 1.9868807254409145e-05, + "loss": 1.3314, + "step": 2853 + }, + { + "epoch": 0.24, + "learning_rate": 1.9868660459008403e-05, + "loss": 1.3941, + "step": 2854 + }, + { + "epoch": 0.24, + "learning_rate": 1.9868513582069627e-05, + "loss": 1.3632, + "step": 2855 + }, + { + "epoch": 0.24, + "learning_rate": 1.9868366623594036e-05, + "loss": 1.3523, + "step": 2856 + }, + { + "epoch": 0.24, + "learning_rate": 1.9868219583582844e-05, + "loss": 1.3283, + "step": 2857 + }, + { + "epoch": 0.24, + "learning_rate": 1.9868072462037263e-05, + "loss": 1.4588, + "step": 2858 + }, + { + "epoch": 0.24, + "learning_rate": 1.9867925258958506e-05, + "loss": 1.3901, + "step": 2859 + }, + { + "epoch": 0.24, + "learning_rate": 1.9867777974347796e-05, + "loss": 1.4057, + "step": 2860 + }, + { + "epoch": 0.24, + "learning_rate": 1.9867630608206344e-05, + "loss": 1.3743, + "step": 2861 + }, + { + "epoch": 0.24, + "learning_rate": 1.986748316053537e-05, + "loss": 1.4016, + "step": 2862 + }, + { + "epoch": 0.24, + "learning_rate": 1.9867335631336093e-05, + "loss": 1.4336, + "step": 2863 + }, + { + "epoch": 0.24, + "learning_rate": 1.986718802060973e-05, + "loss": 1.4099, + "step": 2864 + }, + { + "epoch": 0.24, + "learning_rate": 1.98670403283575e-05, + "loss": 1.4283, + "step": 2865 + }, + { + "epoch": 0.24, + "learning_rate": 1.986689255458063e-05, + "loss": 1.4123, + "step": 2866 + }, + { + "epoch": 0.24, + "learning_rate": 1.9866744699280333e-05, + "loss": 1.2492, + "step": 2867 + }, + { + "epoch": 0.24, + "learning_rate": 1.9866596762457832e-05, + "loss": 1.4213, + "step": 2868 + }, + { + "epoch": 0.24, + "learning_rate": 1.986644874411435e-05, + "loss": 1.3299, + "step": 2869 + }, + { + "epoch": 0.24, + "learning_rate": 1.9866300644251115e-05, + "loss": 1.3871, + "step": 2870 + }, + { + "epoch": 0.24, + "learning_rate": 1.9866152462869345e-05, + "loss": 1.3611, + "step": 2871 + }, + { + "epoch": 0.24, + "learning_rate": 1.9866004199970266e-05, + "loss": 1.3141, + "step": 2872 + }, + { + "epoch": 0.24, + "learning_rate": 1.9865855855555104e-05, + "loss": 1.3215, + "step": 2873 + }, + { + "epoch": 0.24, + "learning_rate": 1.986570742962508e-05, + "loss": 1.3188, + "step": 2874 + }, + { + "epoch": 0.24, + "learning_rate": 1.9865558922181426e-05, + "loss": 1.427, + "step": 2875 + }, + { + "epoch": 0.24, + "learning_rate": 1.9865410333225362e-05, + "loss": 1.3082, + "step": 2876 + }, + { + "epoch": 0.24, + "learning_rate": 1.986526166275813e-05, + "loss": 1.2404, + "step": 2877 + }, + { + "epoch": 0.24, + "learning_rate": 1.986511291078094e-05, + "loss": 1.4417, + "step": 2878 + }, + { + "epoch": 0.24, + "learning_rate": 1.9864964077295035e-05, + "loss": 1.399, + "step": 2879 + }, + { + "epoch": 0.24, + "learning_rate": 1.9864815162301637e-05, + "loss": 1.4573, + "step": 2880 + }, + { + "epoch": 0.24, + "learning_rate": 1.986466616580198e-05, + "loss": 1.3621, + "step": 2881 + }, + { + "epoch": 0.24, + "learning_rate": 1.9864517087797295e-05, + "loss": 1.3435, + "step": 2882 + }, + { + "epoch": 0.24, + "learning_rate": 1.9864367928288813e-05, + "loss": 1.3705, + "step": 2883 + }, + { + "epoch": 0.24, + "learning_rate": 1.9864218687277764e-05, + "loss": 1.4262, + "step": 2884 + }, + { + "epoch": 0.24, + "learning_rate": 1.9864069364765384e-05, + "loss": 1.3036, + "step": 2885 + }, + { + "epoch": 0.24, + "learning_rate": 1.9863919960752907e-05, + "loss": 1.2989, + "step": 2886 + }, + { + "epoch": 0.24, + "learning_rate": 1.9863770475241563e-05, + "loss": 1.394, + "step": 2887 + }, + { + "epoch": 0.24, + "learning_rate": 1.9863620908232594e-05, + "loss": 1.4205, + "step": 2888 + }, + { + "epoch": 0.24, + "learning_rate": 1.986347125972723e-05, + "loss": 1.2948, + "step": 2889 + }, + { + "epoch": 0.24, + "learning_rate": 1.9863321529726715e-05, + "loss": 1.411, + "step": 2890 + }, + { + "epoch": 0.24, + "learning_rate": 1.9863171718232274e-05, + "loss": 1.3766, + "step": 2891 + }, + { + "epoch": 0.24, + "learning_rate": 1.9863021825245156e-05, + "loss": 1.2788, + "step": 2892 + }, + { + "epoch": 0.24, + "learning_rate": 1.9862871850766594e-05, + "loss": 1.4006, + "step": 2893 + }, + { + "epoch": 0.24, + "learning_rate": 1.9862721794797828e-05, + "loss": 1.3611, + "step": 2894 + }, + { + "epoch": 0.24, + "learning_rate": 1.98625716573401e-05, + "loss": 1.3343, + "step": 2895 + }, + { + "epoch": 0.24, + "learning_rate": 1.9862421438394643e-05, + "loss": 1.2694, + "step": 2896 + }, + { + "epoch": 0.24, + "learning_rate": 1.986227113796271e-05, + "loss": 1.3927, + "step": 2897 + }, + { + "epoch": 0.24, + "learning_rate": 1.9862120756045534e-05, + "loss": 1.3441, + "step": 2898 + }, + { + "epoch": 0.24, + "learning_rate": 1.986197029264436e-05, + "loss": 1.3388, + "step": 2899 + }, + { + "epoch": 0.24, + "learning_rate": 1.986181974776043e-05, + "loss": 1.3306, + "step": 2900 + }, + { + "epoch": 0.24, + "learning_rate": 1.9861669121394992e-05, + "loss": 1.3003, + "step": 2901 + }, + { + "epoch": 0.24, + "learning_rate": 1.9861518413549287e-05, + "loss": 1.3767, + "step": 2902 + }, + { + "epoch": 0.24, + "learning_rate": 1.9861367624224558e-05, + "loss": 1.4158, + "step": 2903 + }, + { + "epoch": 0.24, + "learning_rate": 1.9861216753422053e-05, + "loss": 1.3428, + "step": 2904 + }, + { + "epoch": 0.24, + "learning_rate": 1.9861065801143023e-05, + "loss": 1.3775, + "step": 2905 + }, + { + "epoch": 0.24, + "learning_rate": 1.986091476738871e-05, + "loss": 1.3649, + "step": 2906 + }, + { + "epoch": 0.24, + "learning_rate": 1.9860763652160363e-05, + "loss": 1.3606, + "step": 2907 + }, + { + "epoch": 0.24, + "learning_rate": 1.9860612455459228e-05, + "loss": 1.3937, + "step": 2908 + }, + { + "epoch": 0.24, + "learning_rate": 1.986046117728656e-05, + "loss": 1.2892, + "step": 2909 + }, + { + "epoch": 0.24, + "learning_rate": 1.986030981764361e-05, + "loss": 1.3462, + "step": 2910 + }, + { + "epoch": 0.25, + "learning_rate": 1.986015837653162e-05, + "loss": 1.3169, + "step": 2911 + }, + { + "epoch": 0.25, + "learning_rate": 1.9860006853951844e-05, + "loss": 1.3122, + "step": 2912 + }, + { + "epoch": 0.25, + "learning_rate": 1.985985524990554e-05, + "loss": 1.2383, + "step": 2913 + }, + { + "epoch": 0.25, + "learning_rate": 1.9859703564393954e-05, + "loss": 1.4753, + "step": 2914 + }, + { + "epoch": 0.25, + "learning_rate": 1.9859551797418343e-05, + "loss": 1.4482, + "step": 2915 + }, + { + "epoch": 0.25, + "learning_rate": 1.9859399948979958e-05, + "loss": 1.3809, + "step": 2916 + }, + { + "epoch": 0.25, + "learning_rate": 1.9859248019080057e-05, + "loss": 1.3689, + "step": 2917 + }, + { + "epoch": 0.25, + "learning_rate": 1.9859096007719892e-05, + "loss": 1.4605, + "step": 2918 + }, + { + "epoch": 0.25, + "learning_rate": 1.985894391490072e-05, + "loss": 1.3236, + "step": 2919 + }, + { + "epoch": 0.25, + "learning_rate": 1.98587917406238e-05, + "loss": 1.4117, + "step": 2920 + }, + { + "epoch": 0.25, + "learning_rate": 1.9858639484890387e-05, + "loss": 1.4202, + "step": 2921 + }, + { + "epoch": 0.25, + "learning_rate": 1.9858487147701737e-05, + "loss": 1.3666, + "step": 2922 + }, + { + "epoch": 0.25, + "learning_rate": 1.9858334729059115e-05, + "loss": 1.4095, + "step": 2923 + }, + { + "epoch": 0.25, + "learning_rate": 1.9858182228963773e-05, + "loss": 1.4962, + "step": 2924 + }, + { + "epoch": 0.25, + "learning_rate": 1.9858029647416976e-05, + "loss": 1.415, + "step": 2925 + }, + { + "epoch": 0.25, + "learning_rate": 1.9857876984419983e-05, + "loss": 1.3491, + "step": 2926 + }, + { + "epoch": 0.25, + "learning_rate": 1.9857724239974057e-05, + "loss": 1.3677, + "step": 2927 + }, + { + "epoch": 0.25, + "learning_rate": 1.9857571414080454e-05, + "loss": 1.5084, + "step": 2928 + }, + { + "epoch": 0.25, + "learning_rate": 1.9857418506740446e-05, + "loss": 1.3485, + "step": 2929 + }, + { + "epoch": 0.25, + "learning_rate": 1.985726551795529e-05, + "loss": 1.3935, + "step": 2930 + }, + { + "epoch": 0.25, + "learning_rate": 1.985711244772625e-05, + "loss": 1.4021, + "step": 2931 + }, + { + "epoch": 0.25, + "learning_rate": 1.985695929605459e-05, + "loss": 1.3704, + "step": 2932 + }, + { + "epoch": 0.25, + "learning_rate": 1.9856806062941584e-05, + "loss": 1.2824, + "step": 2933 + }, + { + "epoch": 0.25, + "learning_rate": 1.9856652748388488e-05, + "loss": 1.3327, + "step": 2934 + }, + { + "epoch": 0.25, + "learning_rate": 1.9856499352396575e-05, + "loss": 1.3966, + "step": 2935 + }, + { + "epoch": 0.25, + "learning_rate": 1.9856345874967106e-05, + "loss": 1.3643, + "step": 2936 + }, + { + "epoch": 0.25, + "learning_rate": 1.9856192316101355e-05, + "loss": 1.3858, + "step": 2937 + }, + { + "epoch": 0.25, + "learning_rate": 1.985603867580059e-05, + "loss": 1.3554, + "step": 2938 + }, + { + "epoch": 0.25, + "learning_rate": 1.9855884954066077e-05, + "loss": 1.293, + "step": 2939 + }, + { + "epoch": 0.25, + "learning_rate": 1.985573115089909e-05, + "loss": 1.3847, + "step": 2940 + }, + { + "epoch": 0.25, + "learning_rate": 1.9855577266300894e-05, + "loss": 1.5023, + "step": 2941 + }, + { + "epoch": 0.25, + "learning_rate": 1.985542330027277e-05, + "loss": 1.3777, + "step": 2942 + }, + { + "epoch": 0.25, + "learning_rate": 1.9855269252815982e-05, + "loss": 1.3668, + "step": 2943 + }, + { + "epoch": 0.25, + "learning_rate": 1.9855115123931806e-05, + "loss": 1.3627, + "step": 2944 + }, + { + "epoch": 0.25, + "learning_rate": 1.9854960913621513e-05, + "loss": 1.3472, + "step": 2945 + }, + { + "epoch": 0.25, + "learning_rate": 1.985480662188638e-05, + "loss": 1.4186, + "step": 2946 + }, + { + "epoch": 0.25, + "learning_rate": 1.985465224872768e-05, + "loss": 1.4173, + "step": 2947 + }, + { + "epoch": 0.25, + "learning_rate": 1.985449779414669e-05, + "loss": 1.3969, + "step": 2948 + }, + { + "epoch": 0.25, + "learning_rate": 1.9854343258144687e-05, + "loss": 1.3544, + "step": 2949 + }, + { + "epoch": 0.25, + "learning_rate": 1.9854188640722946e-05, + "loss": 1.4465, + "step": 2950 + }, + { + "epoch": 0.25, + "learning_rate": 1.985403394188274e-05, + "loss": 1.4088, + "step": 2951 + }, + { + "epoch": 0.25, + "learning_rate": 1.9853879161625354e-05, + "loss": 1.3636, + "step": 2952 + }, + { + "epoch": 0.25, + "learning_rate": 1.9853724299952068e-05, + "loss": 1.4241, + "step": 2953 + }, + { + "epoch": 0.25, + "learning_rate": 1.9853569356864157e-05, + "loss": 1.279, + "step": 2954 + }, + { + "epoch": 0.25, + "learning_rate": 1.98534143323629e-05, + "loss": 1.3738, + "step": 2955 + }, + { + "epoch": 0.25, + "learning_rate": 1.9853259226449582e-05, + "loss": 1.405, + "step": 2956 + }, + { + "epoch": 0.25, + "learning_rate": 1.9853104039125484e-05, + "loss": 1.3672, + "step": 2957 + }, + { + "epoch": 0.25, + "learning_rate": 1.9852948770391887e-05, + "loss": 1.3288, + "step": 2958 + }, + { + "epoch": 0.25, + "learning_rate": 1.985279342025007e-05, + "loss": 1.3308, + "step": 2959 + }, + { + "epoch": 0.25, + "learning_rate": 1.9852637988701323e-05, + "loss": 1.3538, + "step": 2960 + }, + { + "epoch": 0.25, + "learning_rate": 1.9852482475746928e-05, + "loss": 1.3584, + "step": 2961 + }, + { + "epoch": 0.25, + "learning_rate": 1.985232688138817e-05, + "loss": 1.3699, + "step": 2962 + }, + { + "epoch": 0.25, + "learning_rate": 1.9852171205626335e-05, + "loss": 1.4588, + "step": 2963 + }, + { + "epoch": 0.25, + "learning_rate": 1.9852015448462705e-05, + "loss": 1.3566, + "step": 2964 + }, + { + "epoch": 0.25, + "learning_rate": 1.9851859609898572e-05, + "loss": 1.357, + "step": 2965 + }, + { + "epoch": 0.25, + "learning_rate": 1.9851703689935224e-05, + "loss": 1.3197, + "step": 2966 + }, + { + "epoch": 0.25, + "learning_rate": 1.9851547688573943e-05, + "loss": 1.389, + "step": 2967 + }, + { + "epoch": 0.25, + "learning_rate": 1.9851391605816026e-05, + "loss": 1.3512, + "step": 2968 + }, + { + "epoch": 0.25, + "learning_rate": 1.9851235441662758e-05, + "loss": 1.4083, + "step": 2969 + }, + { + "epoch": 0.25, + "learning_rate": 1.985107919611543e-05, + "loss": 1.3204, + "step": 2970 + }, + { + "epoch": 0.25, + "learning_rate": 1.9850922869175327e-05, + "loss": 1.398, + "step": 2971 + }, + { + "epoch": 0.25, + "learning_rate": 1.9850766460843752e-05, + "loss": 1.3376, + "step": 2972 + }, + { + "epoch": 0.25, + "learning_rate": 1.9850609971121994e-05, + "loss": 1.2521, + "step": 2973 + }, + { + "epoch": 0.25, + "learning_rate": 1.985045340001134e-05, + "loss": 1.3023, + "step": 2974 + }, + { + "epoch": 0.25, + "learning_rate": 1.9850296747513092e-05, + "loss": 1.3635, + "step": 2975 + }, + { + "epoch": 0.25, + "learning_rate": 1.9850140013628538e-05, + "loss": 1.244, + "step": 2976 + }, + { + "epoch": 0.25, + "learning_rate": 1.984998319835897e-05, + "loss": 1.3408, + "step": 2977 + }, + { + "epoch": 0.25, + "learning_rate": 1.984982630170569e-05, + "loss": 1.3395, + "step": 2978 + }, + { + "epoch": 0.25, + "learning_rate": 1.9849669323669994e-05, + "loss": 1.3187, + "step": 2979 + }, + { + "epoch": 0.25, + "learning_rate": 1.9849512264253178e-05, + "loss": 1.3233, + "step": 2980 + }, + { + "epoch": 0.25, + "learning_rate": 1.9849355123456543e-05, + "loss": 1.3144, + "step": 2981 + }, + { + "epoch": 0.25, + "learning_rate": 1.9849197901281377e-05, + "loss": 1.3686, + "step": 2982 + }, + { + "epoch": 0.25, + "learning_rate": 1.984904059772899e-05, + "loss": 1.4102, + "step": 2983 + }, + { + "epoch": 0.25, + "learning_rate": 1.984888321280068e-05, + "loss": 1.3576, + "step": 2984 + }, + { + "epoch": 0.25, + "learning_rate": 1.984872574649774e-05, + "loss": 1.2764, + "step": 2985 + }, + { + "epoch": 0.25, + "learning_rate": 1.984856819882148e-05, + "loss": 1.3461, + "step": 2986 + }, + { + "epoch": 0.25, + "learning_rate": 1.9848410569773194e-05, + "loss": 1.3531, + "step": 2987 + }, + { + "epoch": 0.25, + "learning_rate": 1.9848252859354186e-05, + "loss": 1.2714, + "step": 2988 + }, + { + "epoch": 0.25, + "learning_rate": 1.9848095067565765e-05, + "loss": 1.4486, + "step": 2989 + }, + { + "epoch": 0.25, + "learning_rate": 1.9847937194409228e-05, + "loss": 1.4479, + "step": 2990 + }, + { + "epoch": 0.25, + "learning_rate": 1.9847779239885886e-05, + "loss": 1.2255, + "step": 2991 + }, + { + "epoch": 0.25, + "learning_rate": 1.9847621203997037e-05, + "loss": 1.4205, + "step": 2992 + }, + { + "epoch": 0.25, + "learning_rate": 1.984746308674399e-05, + "loss": 1.4274, + "step": 2993 + }, + { + "epoch": 0.25, + "learning_rate": 1.9847304888128055e-05, + "loss": 1.277, + "step": 2994 + }, + { + "epoch": 0.25, + "learning_rate": 1.9847146608150534e-05, + "loss": 1.4067, + "step": 2995 + }, + { + "epoch": 0.25, + "learning_rate": 1.9846988246812735e-05, + "loss": 1.2811, + "step": 2996 + }, + { + "epoch": 0.25, + "learning_rate": 1.9846829804115967e-05, + "loss": 1.3013, + "step": 2997 + }, + { + "epoch": 0.25, + "learning_rate": 1.984667128006154e-05, + "loss": 1.3848, + "step": 2998 + }, + { + "epoch": 0.25, + "learning_rate": 1.9846512674650765e-05, + "loss": 1.3396, + "step": 2999 + }, + { + "epoch": 0.25, + "learning_rate": 1.9846353987884946e-05, + "loss": 1.3374, + "step": 3000 + }, + { + "epoch": 0.25, + "learning_rate": 1.9846195219765408e-05, + "loss": 1.3898, + "step": 3001 + }, + { + "epoch": 0.25, + "learning_rate": 1.9846036370293448e-05, + "loss": 1.3082, + "step": 3002 + }, + { + "epoch": 0.25, + "learning_rate": 1.9845877439470386e-05, + "loss": 1.3656, + "step": 3003 + }, + { + "epoch": 0.25, + "learning_rate": 1.9845718427297532e-05, + "loss": 1.3853, + "step": 3004 + }, + { + "epoch": 0.25, + "learning_rate": 1.9845559333776202e-05, + "loss": 1.3677, + "step": 3005 + }, + { + "epoch": 0.25, + "learning_rate": 1.984540015890771e-05, + "loss": 1.3857, + "step": 3006 + }, + { + "epoch": 0.25, + "learning_rate": 1.9845240902693373e-05, + "loss": 1.3078, + "step": 3007 + }, + { + "epoch": 0.25, + "learning_rate": 1.9845081565134503e-05, + "loss": 1.293, + "step": 3008 + }, + { + "epoch": 0.25, + "learning_rate": 1.9844922146232416e-05, + "loss": 1.4633, + "step": 3009 + }, + { + "epoch": 0.25, + "learning_rate": 1.9844762645988433e-05, + "loss": 1.3545, + "step": 3010 + }, + { + "epoch": 0.25, + "learning_rate": 1.9844603064403874e-05, + "loss": 1.3789, + "step": 3011 + }, + { + "epoch": 0.25, + "learning_rate": 1.9844443401480048e-05, + "loss": 1.2966, + "step": 3012 + }, + { + "epoch": 0.25, + "learning_rate": 1.9844283657218284e-05, + "loss": 1.2875, + "step": 3013 + }, + { + "epoch": 0.25, + "learning_rate": 1.9844123831619897e-05, + "loss": 1.381, + "step": 3014 + }, + { + "epoch": 0.25, + "learning_rate": 1.9843963924686208e-05, + "loss": 1.3887, + "step": 3015 + }, + { + "epoch": 0.25, + "learning_rate": 1.9843803936418538e-05, + "loss": 1.3797, + "step": 3016 + }, + { + "epoch": 0.25, + "learning_rate": 1.9843643866818212e-05, + "loss": 1.3587, + "step": 3017 + }, + { + "epoch": 0.25, + "learning_rate": 1.9843483715886545e-05, + "loss": 1.3484, + "step": 3018 + }, + { + "epoch": 0.25, + "learning_rate": 1.984332348362487e-05, + "loss": 1.4271, + "step": 3019 + }, + { + "epoch": 0.25, + "learning_rate": 1.9843163170034504e-05, + "loss": 1.3675, + "step": 3020 + }, + { + "epoch": 0.25, + "learning_rate": 1.9843002775116775e-05, + "loss": 1.3858, + "step": 3021 + }, + { + "epoch": 0.25, + "learning_rate": 1.9842842298873008e-05, + "loss": 1.3989, + "step": 3022 + }, + { + "epoch": 0.25, + "learning_rate": 1.9842681741304523e-05, + "loss": 1.2517, + "step": 3023 + }, + { + "epoch": 0.25, + "learning_rate": 1.9842521102412652e-05, + "loss": 1.3248, + "step": 3024 + }, + { + "epoch": 0.25, + "learning_rate": 1.9842360382198724e-05, + "loss": 1.2924, + "step": 3025 + }, + { + "epoch": 0.25, + "learning_rate": 1.984219958066407e-05, + "loss": 1.3376, + "step": 3026 + }, + { + "epoch": 0.25, + "learning_rate": 1.9842038697810002e-05, + "loss": 1.3206, + "step": 3027 + }, + { + "epoch": 0.25, + "learning_rate": 1.984187773363787e-05, + "loss": 1.3115, + "step": 3028 + }, + { + "epoch": 0.26, + "learning_rate": 1.9841716688148988e-05, + "loss": 1.3637, + "step": 3029 + }, + { + "epoch": 0.26, + "learning_rate": 1.9841555561344696e-05, + "loss": 1.2334, + "step": 3030 + }, + { + "epoch": 0.26, + "learning_rate": 1.9841394353226322e-05, + "loss": 1.3189, + "step": 3031 + }, + { + "epoch": 0.26, + "learning_rate": 1.9841233063795197e-05, + "loss": 1.324, + "step": 3032 + }, + { + "epoch": 0.26, + "learning_rate": 1.9841071693052653e-05, + "loss": 1.3164, + "step": 3033 + }, + { + "epoch": 0.26, + "learning_rate": 1.984091024100003e-05, + "loss": 1.2931, + "step": 3034 + }, + { + "epoch": 0.26, + "learning_rate": 1.9840748707638652e-05, + "loss": 1.3525, + "step": 3035 + }, + { + "epoch": 0.26, + "learning_rate": 1.9840587092969864e-05, + "loss": 1.3431, + "step": 3036 + }, + { + "epoch": 0.26, + "learning_rate": 1.9840425396994994e-05, + "loss": 1.27, + "step": 3037 + }, + { + "epoch": 0.26, + "learning_rate": 1.984026361971538e-05, + "loss": 1.2972, + "step": 3038 + }, + { + "epoch": 0.26, + "learning_rate": 1.9840101761132357e-05, + "loss": 1.3079, + "step": 3039 + }, + { + "epoch": 0.26, + "learning_rate": 1.9839939821247265e-05, + "loss": 1.3665, + "step": 3040 + }, + { + "epoch": 0.26, + "learning_rate": 1.9839777800061443e-05, + "loss": 1.3926, + "step": 3041 + }, + { + "epoch": 0.26, + "learning_rate": 1.9839615697576223e-05, + "loss": 1.4466, + "step": 3042 + }, + { + "epoch": 0.26, + "learning_rate": 1.9839453513792954e-05, + "loss": 1.4395, + "step": 3043 + }, + { + "epoch": 0.26, + "learning_rate": 1.983929124871297e-05, + "loss": 1.4361, + "step": 3044 + }, + { + "epoch": 0.26, + "learning_rate": 1.983912890233761e-05, + "loss": 1.4544, + "step": 3045 + }, + { + "epoch": 0.26, + "learning_rate": 1.983896647466822e-05, + "loss": 1.3745, + "step": 3046 + }, + { + "epoch": 0.26, + "learning_rate": 1.9838803965706138e-05, + "loss": 1.2698, + "step": 3047 + }, + { + "epoch": 0.26, + "learning_rate": 1.9838641375452713e-05, + "loss": 1.391, + "step": 3048 + }, + { + "epoch": 0.26, + "learning_rate": 1.983847870390928e-05, + "loss": 1.3174, + "step": 3049 + }, + { + "epoch": 0.26, + "learning_rate": 1.9838315951077193e-05, + "loss": 1.3878, + "step": 3050 + }, + { + "epoch": 0.26, + "learning_rate": 1.9838153116957785e-05, + "loss": 1.3448, + "step": 3051 + }, + { + "epoch": 0.26, + "learning_rate": 1.983799020155241e-05, + "loss": 1.4065, + "step": 3052 + }, + { + "epoch": 0.26, + "learning_rate": 1.9837827204862412e-05, + "loss": 1.3736, + "step": 3053 + }, + { + "epoch": 0.26, + "learning_rate": 1.9837664126889137e-05, + "loss": 1.3996, + "step": 3054 + }, + { + "epoch": 0.26, + "learning_rate": 1.983750096763393e-05, + "loss": 1.2761, + "step": 3055 + }, + { + "epoch": 0.26, + "learning_rate": 1.9837337727098142e-05, + "loss": 1.2933, + "step": 3056 + }, + { + "epoch": 0.26, + "learning_rate": 1.9837174405283123e-05, + "loss": 1.3695, + "step": 3057 + }, + { + "epoch": 0.26, + "learning_rate": 1.983701100219022e-05, + "loss": 1.4269, + "step": 3058 + }, + { + "epoch": 0.26, + "learning_rate": 1.9836847517820783e-05, + "loss": 1.3542, + "step": 3059 + }, + { + "epoch": 0.26, + "learning_rate": 1.9836683952176166e-05, + "loss": 1.349, + "step": 3060 + }, + { + "epoch": 0.26, + "learning_rate": 1.9836520305257715e-05, + "loss": 1.3886, + "step": 3061 + }, + { + "epoch": 0.26, + "learning_rate": 1.9836356577066788e-05, + "loss": 1.2563, + "step": 3062 + }, + { + "epoch": 0.26, + "learning_rate": 1.9836192767604733e-05, + "loss": 1.4118, + "step": 3063 + }, + { + "epoch": 0.26, + "learning_rate": 1.983602887687291e-05, + "loss": 1.3731, + "step": 3064 + }, + { + "epoch": 0.26, + "learning_rate": 1.9835864904872662e-05, + "loss": 1.3484, + "step": 3065 + }, + { + "epoch": 0.26, + "learning_rate": 1.9835700851605354e-05, + "loss": 1.3308, + "step": 3066 + }, + { + "epoch": 0.26, + "learning_rate": 1.9835536717072335e-05, + "loss": 1.2917, + "step": 3067 + }, + { + "epoch": 0.26, + "learning_rate": 1.983537250127497e-05, + "loss": 1.3126, + "step": 3068 + }, + { + "epoch": 0.26, + "learning_rate": 1.9835208204214603e-05, + "loss": 1.4349, + "step": 3069 + }, + { + "epoch": 0.26, + "learning_rate": 1.98350438258926e-05, + "loss": 1.4059, + "step": 3070 + }, + { + "epoch": 0.26, + "learning_rate": 1.983487936631032e-05, + "loss": 1.3071, + "step": 3071 + }, + { + "epoch": 0.26, + "learning_rate": 1.983471482546911e-05, + "loss": 1.3051, + "step": 3072 + }, + { + "epoch": 0.26, + "learning_rate": 1.983455020337035e-05, + "loss": 1.335, + "step": 3073 + }, + { + "epoch": 0.26, + "learning_rate": 1.9834385500015384e-05, + "loss": 1.3546, + "step": 3074 + }, + { + "epoch": 0.26, + "learning_rate": 1.9834220715405578e-05, + "loss": 1.258, + "step": 3075 + }, + { + "epoch": 0.26, + "learning_rate": 1.983405584954229e-05, + "loss": 1.4388, + "step": 3076 + }, + { + "epoch": 0.26, + "learning_rate": 1.983389090242689e-05, + "loss": 1.3429, + "step": 3077 + }, + { + "epoch": 0.26, + "learning_rate": 1.9833725874060733e-05, + "loss": 1.3923, + "step": 3078 + }, + { + "epoch": 0.26, + "learning_rate": 1.983356076444519e-05, + "loss": 1.4317, + "step": 3079 + }, + { + "epoch": 0.26, + "learning_rate": 1.9833395573581617e-05, + "loss": 1.3829, + "step": 3080 + }, + { + "epoch": 0.26, + "learning_rate": 1.9833230301471383e-05, + "loss": 1.4197, + "step": 3081 + }, + { + "epoch": 0.26, + "learning_rate": 1.9833064948115853e-05, + "loss": 1.4141, + "step": 3082 + }, + { + "epoch": 0.26, + "learning_rate": 1.9832899513516392e-05, + "loss": 1.4245, + "step": 3083 + }, + { + "epoch": 0.26, + "learning_rate": 1.9832733997674373e-05, + "loss": 1.3688, + "step": 3084 + }, + { + "epoch": 0.26, + "learning_rate": 1.9832568400591154e-05, + "loss": 1.2991, + "step": 3085 + }, + { + "epoch": 0.26, + "learning_rate": 1.983240272226811e-05, + "loss": 1.3906, + "step": 3086 + }, + { + "epoch": 0.26, + "learning_rate": 1.983223696270661e-05, + "loss": 1.3233, + "step": 3087 + }, + { + "epoch": 0.26, + "learning_rate": 1.9832071121908018e-05, + "loss": 1.2778, + "step": 3088 + }, + { + "epoch": 0.26, + "learning_rate": 1.983190519987371e-05, + "loss": 1.3369, + "step": 3089 + }, + { + "epoch": 0.26, + "learning_rate": 1.9831739196605053e-05, + "loss": 1.3057, + "step": 3090 + }, + { + "epoch": 0.26, + "learning_rate": 1.9831573112103423e-05, + "loss": 1.3253, + "step": 3091 + }, + { + "epoch": 0.26, + "learning_rate": 1.9831406946370187e-05, + "loss": 1.3299, + "step": 3092 + }, + { + "epoch": 0.26, + "learning_rate": 1.983124069940672e-05, + "loss": 1.2939, + "step": 3093 + }, + { + "epoch": 0.26, + "learning_rate": 1.9831074371214398e-05, + "loss": 1.3904, + "step": 3094 + }, + { + "epoch": 0.26, + "learning_rate": 1.983090796179459e-05, + "loss": 1.3583, + "step": 3095 + }, + { + "epoch": 0.26, + "learning_rate": 1.983074147114868e-05, + "loss": 1.4828, + "step": 3096 + }, + { + "epoch": 0.26, + "learning_rate": 1.983057489927803e-05, + "loss": 1.367, + "step": 3097 + }, + { + "epoch": 0.26, + "learning_rate": 1.983040824618403e-05, + "loss": 1.3191, + "step": 3098 + }, + { + "epoch": 0.26, + "learning_rate": 1.983024151186805e-05, + "loss": 1.3487, + "step": 3099 + }, + { + "epoch": 0.26, + "learning_rate": 1.9830074696331465e-05, + "loss": 1.4363, + "step": 3100 + }, + { + "epoch": 0.26, + "learning_rate": 1.9829907799575657e-05, + "loss": 1.2516, + "step": 3101 + }, + { + "epoch": 0.26, + "learning_rate": 1.982974082160201e-05, + "loss": 1.3523, + "step": 3102 + }, + { + "epoch": 0.26, + "learning_rate": 1.9829573762411893e-05, + "loss": 1.2349, + "step": 3103 + }, + { + "epoch": 0.26, + "learning_rate": 1.9829406622006698e-05, + "loss": 1.4031, + "step": 3104 + }, + { + "epoch": 0.26, + "learning_rate": 1.9829239400387795e-05, + "loss": 1.4412, + "step": 3105 + }, + { + "epoch": 0.26, + "learning_rate": 1.982907209755657e-05, + "loss": 1.3804, + "step": 3106 + }, + { + "epoch": 0.26, + "learning_rate": 1.9828904713514405e-05, + "loss": 1.3669, + "step": 3107 + }, + { + "epoch": 0.26, + "learning_rate": 1.9828737248262687e-05, + "loss": 1.414, + "step": 3108 + }, + { + "epoch": 0.26, + "learning_rate": 1.9828569701802793e-05, + "loss": 1.3416, + "step": 3109 + }, + { + "epoch": 0.26, + "learning_rate": 1.9828402074136112e-05, + "loss": 1.3228, + "step": 3110 + }, + { + "epoch": 0.26, + "learning_rate": 1.982823436526403e-05, + "loss": 1.3542, + "step": 3111 + }, + { + "epoch": 0.26, + "learning_rate": 1.982806657518793e-05, + "loss": 1.3285, + "step": 3112 + }, + { + "epoch": 0.26, + "learning_rate": 1.9827898703909195e-05, + "loss": 1.3575, + "step": 3113 + }, + { + "epoch": 0.26, + "learning_rate": 1.9827730751429218e-05, + "loss": 1.4486, + "step": 3114 + }, + { + "epoch": 0.26, + "learning_rate": 1.982756271774938e-05, + "loss": 1.3061, + "step": 3115 + }, + { + "epoch": 0.26, + "learning_rate": 1.982739460287108e-05, + "loss": 1.3278, + "step": 3116 + }, + { + "epoch": 0.26, + "learning_rate": 1.9827226406795698e-05, + "loss": 1.3572, + "step": 3117 + }, + { + "epoch": 0.26, + "learning_rate": 1.9827058129524626e-05, + "loss": 1.3443, + "step": 3118 + }, + { + "epoch": 0.26, + "learning_rate": 1.9826889771059253e-05, + "loss": 1.451, + "step": 3119 + }, + { + "epoch": 0.26, + "learning_rate": 1.982672133140097e-05, + "loss": 1.2729, + "step": 3120 + }, + { + "epoch": 0.26, + "learning_rate": 1.9826552810551177e-05, + "loss": 1.3378, + "step": 3121 + }, + { + "epoch": 0.26, + "learning_rate": 1.9826384208511258e-05, + "loss": 1.2986, + "step": 3122 + }, + { + "epoch": 0.26, + "learning_rate": 1.9826215525282604e-05, + "loss": 1.5, + "step": 3123 + }, + { + "epoch": 0.26, + "learning_rate": 1.9826046760866615e-05, + "loss": 1.3104, + "step": 3124 + }, + { + "epoch": 0.26, + "learning_rate": 1.9825877915264683e-05, + "loss": 1.4135, + "step": 3125 + }, + { + "epoch": 0.26, + "learning_rate": 1.9825708988478204e-05, + "loss": 1.3978, + "step": 3126 + }, + { + "epoch": 0.26, + "learning_rate": 1.982553998050857e-05, + "loss": 1.3759, + "step": 3127 + }, + { + "epoch": 0.26, + "learning_rate": 1.982537089135718e-05, + "loss": 1.3982, + "step": 3128 + }, + { + "epoch": 0.26, + "learning_rate": 1.9825201721025437e-05, + "loss": 1.416, + "step": 3129 + }, + { + "epoch": 0.26, + "learning_rate": 1.9825032469514728e-05, + "loss": 1.3738, + "step": 3130 + }, + { + "epoch": 0.26, + "learning_rate": 1.9824863136826456e-05, + "loss": 1.3417, + "step": 3131 + }, + { + "epoch": 0.26, + "learning_rate": 1.9824693722962022e-05, + "loss": 1.3723, + "step": 3132 + }, + { + "epoch": 0.26, + "learning_rate": 1.9824524227922822e-05, + "loss": 1.3593, + "step": 3133 + }, + { + "epoch": 0.26, + "learning_rate": 1.982435465171026e-05, + "loss": 1.3815, + "step": 3134 + }, + { + "epoch": 0.26, + "learning_rate": 1.9824184994325738e-05, + "loss": 1.3448, + "step": 3135 + }, + { + "epoch": 0.26, + "learning_rate": 1.982401525577065e-05, + "loss": 1.5007, + "step": 3136 + }, + { + "epoch": 0.26, + "learning_rate": 1.9823845436046406e-05, + "loss": 1.3649, + "step": 3137 + }, + { + "epoch": 0.26, + "learning_rate": 1.9823675535154413e-05, + "loss": 1.3838, + "step": 3138 + }, + { + "epoch": 0.26, + "learning_rate": 1.9823505553096063e-05, + "loss": 1.3068, + "step": 3139 + }, + { + "epoch": 0.26, + "learning_rate": 1.9823335489872767e-05, + "loss": 1.2934, + "step": 3140 + }, + { + "epoch": 0.26, + "learning_rate": 1.9823165345485926e-05, + "loss": 1.3293, + "step": 3141 + }, + { + "epoch": 0.26, + "learning_rate": 1.9822995119936954e-05, + "loss": 1.4281, + "step": 3142 + }, + { + "epoch": 0.26, + "learning_rate": 1.982282481322725e-05, + "loss": 1.3366, + "step": 3143 + }, + { + "epoch": 0.26, + "learning_rate": 1.9822654425358228e-05, + "loss": 1.3576, + "step": 3144 + }, + { + "epoch": 0.26, + "learning_rate": 1.9822483956331286e-05, + "loss": 1.2829, + "step": 3145 + }, + { + "epoch": 0.26, + "learning_rate": 1.9822313406147843e-05, + "loss": 1.3107, + "step": 3146 + }, + { + "epoch": 0.26, + "learning_rate": 1.98221427748093e-05, + "loss": 1.375, + "step": 3147 + }, + { + "epoch": 0.27, + "learning_rate": 1.982197206231707e-05, + "loss": 1.2528, + "step": 3148 + }, + { + "epoch": 0.27, + "learning_rate": 1.982180126867256e-05, + "loss": 1.3487, + "step": 3149 + }, + { + "epoch": 0.27, + "learning_rate": 1.982163039387719e-05, + "loss": 1.2884, + "step": 3150 + }, + { + "epoch": 0.27, + "learning_rate": 1.9821459437932362e-05, + "loss": 1.411, + "step": 3151 + }, + { + "epoch": 0.27, + "learning_rate": 1.9821288400839494e-05, + "loss": 1.3282, + "step": 3152 + }, + { + "epoch": 0.27, + "learning_rate": 1.9821117282600002e-05, + "loss": 1.3393, + "step": 3153 + }, + { + "epoch": 0.27, + "learning_rate": 1.9820946083215293e-05, + "loss": 1.4126, + "step": 3154 + }, + { + "epoch": 0.27, + "learning_rate": 1.9820774802686785e-05, + "loss": 1.3066, + "step": 3155 + }, + { + "epoch": 0.27, + "learning_rate": 1.9820603441015893e-05, + "loss": 1.4939, + "step": 3156 + }, + { + "epoch": 0.27, + "learning_rate": 1.982043199820403e-05, + "loss": 1.2229, + "step": 3157 + }, + { + "epoch": 0.27, + "learning_rate": 1.9820260474252613e-05, + "loss": 1.3722, + "step": 3158 + }, + { + "epoch": 0.27, + "learning_rate": 1.9820088869163066e-05, + "loss": 1.2964, + "step": 3159 + }, + { + "epoch": 0.27, + "learning_rate": 1.9819917182936802e-05, + "loss": 1.3283, + "step": 3160 + }, + { + "epoch": 0.27, + "learning_rate": 1.9819745415575237e-05, + "loss": 1.3548, + "step": 3161 + }, + { + "epoch": 0.27, + "learning_rate": 1.981957356707979e-05, + "loss": 1.2629, + "step": 3162 + }, + { + "epoch": 0.27, + "learning_rate": 1.981940163745189e-05, + "loss": 1.3839, + "step": 3163 + }, + { + "epoch": 0.27, + "learning_rate": 1.981922962669295e-05, + "loss": 1.411, + "step": 3164 + }, + { + "epoch": 0.27, + "learning_rate": 1.981905753480439e-05, + "loss": 1.3733, + "step": 3165 + }, + { + "epoch": 0.27, + "learning_rate": 1.9818885361787633e-05, + "loss": 1.3756, + "step": 3166 + }, + { + "epoch": 0.27, + "learning_rate": 1.9818713107644104e-05, + "loss": 1.3238, + "step": 3167 + }, + { + "epoch": 0.27, + "learning_rate": 1.9818540772375226e-05, + "loss": 1.3719, + "step": 3168 + }, + { + "epoch": 0.27, + "learning_rate": 1.981836835598242e-05, + "loss": 1.4345, + "step": 3169 + }, + { + "epoch": 0.27, + "learning_rate": 1.9818195858467115e-05, + "loss": 1.2438, + "step": 3170 + }, + { + "epoch": 0.27, + "learning_rate": 1.981802327983073e-05, + "loss": 1.3754, + "step": 3171 + }, + { + "epoch": 0.27, + "learning_rate": 1.9817850620074696e-05, + "loss": 1.4097, + "step": 3172 + }, + { + "epoch": 0.27, + "learning_rate": 1.981767787920044e-05, + "loss": 1.3304, + "step": 3173 + }, + { + "epoch": 0.27, + "learning_rate": 1.9817505057209386e-05, + "loss": 1.3509, + "step": 3174 + }, + { + "epoch": 0.27, + "learning_rate": 1.9817332154102962e-05, + "loss": 1.4191, + "step": 3175 + }, + { + "epoch": 0.27, + "learning_rate": 1.9817159169882597e-05, + "loss": 1.3659, + "step": 3176 + }, + { + "epoch": 0.27, + "learning_rate": 1.9816986104549722e-05, + "loss": 1.319, + "step": 3177 + }, + { + "epoch": 0.27, + "learning_rate": 1.981681295810577e-05, + "loss": 1.4104, + "step": 3178 + }, + { + "epoch": 0.27, + "learning_rate": 1.9816639730552164e-05, + "loss": 1.3928, + "step": 3179 + }, + { + "epoch": 0.27, + "learning_rate": 1.9816466421890336e-05, + "loss": 1.3295, + "step": 3180 + }, + { + "epoch": 0.27, + "learning_rate": 1.9816293032121725e-05, + "loss": 1.3661, + "step": 3181 + }, + { + "epoch": 0.27, + "learning_rate": 1.981611956124776e-05, + "loss": 1.3521, + "step": 3182 + }, + { + "epoch": 0.27, + "learning_rate": 1.981594600926987e-05, + "loss": 1.3766, + "step": 3183 + }, + { + "epoch": 0.27, + "learning_rate": 1.9815772376189495e-05, + "loss": 1.2695, + "step": 3184 + }, + { + "epoch": 0.27, + "learning_rate": 1.9815598662008068e-05, + "loss": 1.3987, + "step": 3185 + }, + { + "epoch": 0.27, + "learning_rate": 1.981542486672702e-05, + "loss": 1.3077, + "step": 3186 + }, + { + "epoch": 0.27, + "learning_rate": 1.981525099034779e-05, + "loss": 1.4837, + "step": 3187 + }, + { + "epoch": 0.27, + "learning_rate": 1.981507703287182e-05, + "loss": 1.3611, + "step": 3188 + }, + { + "epoch": 0.27, + "learning_rate": 1.981490299430054e-05, + "loss": 1.3361, + "step": 3189 + }, + { + "epoch": 0.27, + "learning_rate": 1.9814728874635388e-05, + "loss": 1.3046, + "step": 3190 + }, + { + "epoch": 0.27, + "learning_rate": 1.981455467387781e-05, + "loss": 1.4234, + "step": 3191 + }, + { + "epoch": 0.27, + "learning_rate": 1.9814380392029237e-05, + "loss": 1.3685, + "step": 3192 + }, + { + "epoch": 0.27, + "learning_rate": 1.9814206029091113e-05, + "loss": 1.2275, + "step": 3193 + }, + { + "epoch": 0.27, + "learning_rate": 1.9814031585064876e-05, + "loss": 1.3242, + "step": 3194 + }, + { + "epoch": 0.27, + "learning_rate": 1.981385705995197e-05, + "loss": 1.4773, + "step": 3195 + }, + { + "epoch": 0.27, + "learning_rate": 1.9813682453753836e-05, + "loss": 1.4312, + "step": 3196 + }, + { + "epoch": 0.27, + "learning_rate": 1.9813507766471916e-05, + "loss": 1.3782, + "step": 3197 + }, + { + "epoch": 0.27, + "learning_rate": 1.9813332998107656e-05, + "loss": 1.3941, + "step": 3198 + }, + { + "epoch": 0.27, + "learning_rate": 1.9813158148662495e-05, + "loss": 1.312, + "step": 3199 + }, + { + "epoch": 0.27, + "learning_rate": 1.9812983218137887e-05, + "loss": 1.3995, + "step": 3200 + }, + { + "epoch": 0.27, + "learning_rate": 1.9812808206535265e-05, + "loss": 1.396, + "step": 3201 + }, + { + "epoch": 0.27, + "learning_rate": 1.9812633113856086e-05, + "loss": 1.3962, + "step": 3202 + }, + { + "epoch": 0.27, + "learning_rate": 1.981245794010179e-05, + "loss": 1.3836, + "step": 3203 + }, + { + "epoch": 0.27, + "learning_rate": 1.9812282685273826e-05, + "loss": 1.3479, + "step": 3204 + }, + { + "epoch": 0.27, + "learning_rate": 1.981210734937364e-05, + "loss": 1.4049, + "step": 3205 + }, + { + "epoch": 0.27, + "learning_rate": 1.9811931932402685e-05, + "loss": 1.3435, + "step": 3206 + }, + { + "epoch": 0.27, + "learning_rate": 1.9811756434362407e-05, + "loss": 1.3633, + "step": 3207 + }, + { + "epoch": 0.27, + "learning_rate": 1.9811580855254256e-05, + "loss": 1.3583, + "step": 3208 + }, + { + "epoch": 0.27, + "learning_rate": 1.9811405195079686e-05, + "loss": 1.3325, + "step": 3209 + }, + { + "epoch": 0.27, + "learning_rate": 1.9811229453840145e-05, + "loss": 1.3627, + "step": 3210 + }, + { + "epoch": 0.27, + "learning_rate": 1.981105363153709e-05, + "loss": 1.3437, + "step": 3211 + }, + { + "epoch": 0.27, + "learning_rate": 1.9810877728171965e-05, + "loss": 1.3495, + "step": 3212 + }, + { + "epoch": 0.27, + "learning_rate": 1.9810701743746227e-05, + "loss": 1.3518, + "step": 3213 + }, + { + "epoch": 0.27, + "learning_rate": 1.9810525678261336e-05, + "loss": 1.3176, + "step": 3214 + }, + { + "epoch": 0.27, + "learning_rate": 1.9810349531718742e-05, + "loss": 1.2735, + "step": 3215 + }, + { + "epoch": 0.27, + "learning_rate": 1.98101733041199e-05, + "loss": 1.3674, + "step": 3216 + }, + { + "epoch": 0.27, + "learning_rate": 1.9809996995466263e-05, + "loss": 1.3368, + "step": 3217 + }, + { + "epoch": 0.27, + "learning_rate": 1.9809820605759294e-05, + "loss": 1.4664, + "step": 3218 + }, + { + "epoch": 0.27, + "learning_rate": 1.9809644135000448e-05, + "loss": 1.2862, + "step": 3219 + }, + { + "epoch": 0.27, + "learning_rate": 1.9809467583191183e-05, + "loss": 1.339, + "step": 3220 + }, + { + "epoch": 0.27, + "learning_rate": 1.9809290950332955e-05, + "loss": 1.3278, + "step": 3221 + }, + { + "epoch": 0.27, + "learning_rate": 1.9809114236427226e-05, + "loss": 1.3245, + "step": 3222 + }, + { + "epoch": 0.27, + "learning_rate": 1.980893744147546e-05, + "loss": 1.3505, + "step": 3223 + }, + { + "epoch": 0.27, + "learning_rate": 1.980876056547911e-05, + "loss": 1.3709, + "step": 3224 + }, + { + "epoch": 0.27, + "learning_rate": 1.980858360843964e-05, + "loss": 1.4319, + "step": 3225 + }, + { + "epoch": 0.27, + "learning_rate": 1.9808406570358517e-05, + "loss": 1.4599, + "step": 3226 + }, + { + "epoch": 0.27, + "learning_rate": 1.9808229451237196e-05, + "loss": 1.3684, + "step": 3227 + }, + { + "epoch": 0.27, + "learning_rate": 1.9808052251077147e-05, + "loss": 1.456, + "step": 3228 + }, + { + "epoch": 0.27, + "learning_rate": 1.980787496987983e-05, + "loss": 1.3986, + "step": 3229 + }, + { + "epoch": 0.27, + "learning_rate": 1.9807697607646713e-05, + "loss": 1.3614, + "step": 3230 + }, + { + "epoch": 0.27, + "learning_rate": 1.9807520164379257e-05, + "loss": 1.3067, + "step": 3231 + }, + { + "epoch": 0.27, + "learning_rate": 1.980734264007893e-05, + "loss": 1.3417, + "step": 3232 + }, + { + "epoch": 0.27, + "learning_rate": 1.9807165034747202e-05, + "loss": 1.3405, + "step": 3233 + }, + { + "epoch": 0.27, + "learning_rate": 1.9806987348385535e-05, + "loss": 1.3152, + "step": 3234 + }, + { + "epoch": 0.27, + "learning_rate": 1.9806809580995402e-05, + "loss": 1.2695, + "step": 3235 + }, + { + "epoch": 0.27, + "learning_rate": 1.980663173257827e-05, + "loss": 1.3185, + "step": 3236 + }, + { + "epoch": 0.27, + "learning_rate": 1.980645380313561e-05, + "loss": 1.3265, + "step": 3237 + }, + { + "epoch": 0.27, + "learning_rate": 1.9806275792668884e-05, + "loss": 1.3176, + "step": 3238 + }, + { + "epoch": 0.27, + "learning_rate": 1.9806097701179574e-05, + "loss": 1.3468, + "step": 3239 + }, + { + "epoch": 0.27, + "learning_rate": 1.9805919528669144e-05, + "loss": 1.2898, + "step": 3240 + }, + { + "epoch": 0.27, + "learning_rate": 1.980574127513907e-05, + "loss": 1.346, + "step": 3241 + }, + { + "epoch": 0.27, + "learning_rate": 1.980556294059082e-05, + "loss": 1.3897, + "step": 3242 + }, + { + "epoch": 0.27, + "learning_rate": 1.9805384525025876e-05, + "loss": 1.3279, + "step": 3243 + }, + { + "epoch": 0.27, + "learning_rate": 1.9805206028445702e-05, + "loss": 1.3866, + "step": 3244 + }, + { + "epoch": 0.27, + "learning_rate": 1.980502745085178e-05, + "loss": 1.2988, + "step": 3245 + }, + { + "epoch": 0.27, + "learning_rate": 1.9804848792245584e-05, + "loss": 1.3017, + "step": 3246 + }, + { + "epoch": 0.27, + "learning_rate": 1.9804670052628588e-05, + "loss": 1.4547, + "step": 3247 + }, + { + "epoch": 0.27, + "learning_rate": 1.9804491232002272e-05, + "loss": 1.3911, + "step": 3248 + }, + { + "epoch": 0.27, + "learning_rate": 1.980431233036811e-05, + "loss": 1.3566, + "step": 3249 + }, + { + "epoch": 0.27, + "learning_rate": 1.9804133347727583e-05, + "loss": 1.3405, + "step": 3250 + }, + { + "epoch": 0.27, + "learning_rate": 1.980395428408217e-05, + "loss": 1.4229, + "step": 3251 + }, + { + "epoch": 0.27, + "learning_rate": 1.9803775139433346e-05, + "loss": 1.226, + "step": 3252 + }, + { + "epoch": 0.27, + "learning_rate": 1.9803595913782594e-05, + "loss": 1.3511, + "step": 3253 + }, + { + "epoch": 0.27, + "learning_rate": 1.9803416607131395e-05, + "loss": 1.388, + "step": 3254 + }, + { + "epoch": 0.27, + "learning_rate": 1.980323721948123e-05, + "loss": 1.3883, + "step": 3255 + }, + { + "epoch": 0.27, + "learning_rate": 1.9803057750833585e-05, + "loss": 1.304, + "step": 3256 + }, + { + "epoch": 0.27, + "learning_rate": 1.980287820118994e-05, + "loss": 1.3635, + "step": 3257 + }, + { + "epoch": 0.27, + "learning_rate": 1.9802698570551775e-05, + "loss": 1.3139, + "step": 3258 + }, + { + "epoch": 0.27, + "learning_rate": 1.9802518858920575e-05, + "loss": 1.3599, + "step": 3259 + }, + { + "epoch": 0.27, + "learning_rate": 1.980233906629783e-05, + "loss": 1.3068, + "step": 3260 + }, + { + "epoch": 0.27, + "learning_rate": 1.9802159192685023e-05, + "loss": 1.3449, + "step": 3261 + }, + { + "epoch": 0.27, + "learning_rate": 1.980197923808364e-05, + "loss": 1.3143, + "step": 3262 + }, + { + "epoch": 0.27, + "learning_rate": 1.9801799202495164e-05, + "loss": 1.4016, + "step": 3263 + }, + { + "epoch": 0.27, + "learning_rate": 1.980161908592109e-05, + "loss": 1.3138, + "step": 3264 + }, + { + "epoch": 0.27, + "learning_rate": 1.98014388883629e-05, + "loss": 1.3909, + "step": 3265 + }, + { + "epoch": 0.27, + "learning_rate": 1.9801258609822083e-05, + "loss": 1.3266, + "step": 3266 + }, + { + "epoch": 0.28, + "learning_rate": 1.9801078250300136e-05, + "loss": 1.2706, + "step": 3267 + }, + { + "epoch": 0.28, + "learning_rate": 1.980089780979854e-05, + "loss": 1.4117, + "step": 3268 + }, + { + "epoch": 0.28, + "learning_rate": 1.9800717288318792e-05, + "loss": 1.2673, + "step": 3269 + }, + { + "epoch": 0.28, + "learning_rate": 1.980053668586238e-05, + "loss": 1.3216, + "step": 3270 + }, + { + "epoch": 0.28, + "learning_rate": 1.9800356002430795e-05, + "loss": 1.3677, + "step": 3271 + }, + { + "epoch": 0.28, + "learning_rate": 1.9800175238025537e-05, + "loss": 1.263, + "step": 3272 + }, + { + "epoch": 0.28, + "learning_rate": 1.979999439264809e-05, + "loss": 1.4073, + "step": 3273 + }, + { + "epoch": 0.28, + "learning_rate": 1.9799813466299954e-05, + "loss": 1.4002, + "step": 3274 + }, + { + "epoch": 0.28, + "learning_rate": 1.9799632458982624e-05, + "loss": 1.3438, + "step": 3275 + }, + { + "epoch": 0.28, + "learning_rate": 1.9799451370697595e-05, + "loss": 1.3035, + "step": 3276 + }, + { + "epoch": 0.28, + "learning_rate": 1.979927020144636e-05, + "loss": 1.3595, + "step": 3277 + }, + { + "epoch": 0.28, + "learning_rate": 1.979908895123042e-05, + "loss": 1.3645, + "step": 3278 + }, + { + "epoch": 0.28, + "learning_rate": 1.9798907620051272e-05, + "loss": 1.3435, + "step": 3279 + }, + { + "epoch": 0.28, + "learning_rate": 1.9798726207910412e-05, + "loss": 1.4029, + "step": 3280 + }, + { + "epoch": 0.28, + "learning_rate": 1.979854471480934e-05, + "loss": 1.3062, + "step": 3281 + }, + { + "epoch": 0.28, + "learning_rate": 1.9798363140749556e-05, + "loss": 1.3465, + "step": 3282 + }, + { + "epoch": 0.28, + "learning_rate": 1.979818148573256e-05, + "loss": 1.4144, + "step": 3283 + }, + { + "epoch": 0.28, + "learning_rate": 1.979799974975985e-05, + "loss": 1.3957, + "step": 3284 + }, + { + "epoch": 0.28, + "learning_rate": 1.9797817932832935e-05, + "loss": 1.3516, + "step": 3285 + }, + { + "epoch": 0.28, + "learning_rate": 1.979763603495331e-05, + "loss": 1.4032, + "step": 3286 + }, + { + "epoch": 0.28, + "learning_rate": 1.9797454056122478e-05, + "loss": 1.3431, + "step": 3287 + }, + { + "epoch": 0.28, + "learning_rate": 1.979727199634195e-05, + "loss": 1.2535, + "step": 3288 + }, + { + "epoch": 0.28, + "learning_rate": 1.9797089855613222e-05, + "loss": 1.4388, + "step": 3289 + }, + { + "epoch": 0.28, + "learning_rate": 1.9796907633937804e-05, + "loss": 1.3467, + "step": 3290 + }, + { + "epoch": 0.28, + "learning_rate": 1.9796725331317196e-05, + "loss": 1.3359, + "step": 3291 + }, + { + "epoch": 0.28, + "learning_rate": 1.9796542947752914e-05, + "loss": 1.3132, + "step": 3292 + }, + { + "epoch": 0.28, + "learning_rate": 1.9796360483246455e-05, + "loss": 1.3518, + "step": 3293 + }, + { + "epoch": 0.28, + "learning_rate": 1.979617793779933e-05, + "loss": 1.2947, + "step": 3294 + }, + { + "epoch": 0.28, + "learning_rate": 1.9795995311413047e-05, + "loss": 1.369, + "step": 3295 + }, + { + "epoch": 0.28, + "learning_rate": 1.979581260408912e-05, + "loss": 1.3426, + "step": 3296 + }, + { + "epoch": 0.28, + "learning_rate": 1.9795629815829053e-05, + "loss": 1.4, + "step": 3297 + }, + { + "epoch": 0.28, + "learning_rate": 1.9795446946634355e-05, + "loss": 1.4024, + "step": 3298 + }, + { + "epoch": 0.28, + "learning_rate": 1.979526399650654e-05, + "loss": 1.2484, + "step": 3299 + }, + { + "epoch": 0.28, + "learning_rate": 1.979508096544712e-05, + "loss": 1.3887, + "step": 3300 + }, + { + "epoch": 0.28, + "learning_rate": 1.9794897853457605e-05, + "loss": 1.3983, + "step": 3301 + }, + { + "epoch": 0.28, + "learning_rate": 1.9794714660539508e-05, + "loss": 1.2878, + "step": 3302 + }, + { + "epoch": 0.28, + "learning_rate": 1.9794531386694345e-05, + "loss": 1.3118, + "step": 3303 + }, + { + "epoch": 0.28, + "learning_rate": 1.979434803192363e-05, + "loss": 1.3095, + "step": 3304 + }, + { + "epoch": 0.28, + "learning_rate": 1.9794164596228876e-05, + "loss": 1.4415, + "step": 3305 + }, + { + "epoch": 0.28, + "learning_rate": 1.9793981079611602e-05, + "loss": 1.3921, + "step": 3306 + }, + { + "epoch": 0.28, + "learning_rate": 1.979379748207332e-05, + "loss": 1.4102, + "step": 3307 + }, + { + "epoch": 0.28, + "learning_rate": 1.9793613803615548e-05, + "loss": 1.3193, + "step": 3308 + }, + { + "epoch": 0.28, + "learning_rate": 1.9793430044239804e-05, + "loss": 1.3113, + "step": 3309 + }, + { + "epoch": 0.28, + "learning_rate": 1.9793246203947612e-05, + "loss": 1.3175, + "step": 3310 + }, + { + "epoch": 0.28, + "learning_rate": 1.979306228274048e-05, + "loss": 1.4786, + "step": 3311 + }, + { + "epoch": 0.28, + "learning_rate": 1.9792878280619933e-05, + "loss": 1.4228, + "step": 3312 + }, + { + "epoch": 0.28, + "learning_rate": 1.9792694197587496e-05, + "loss": 1.2239, + "step": 3313 + }, + { + "epoch": 0.28, + "learning_rate": 1.9792510033644682e-05, + "loss": 1.3076, + "step": 3314 + }, + { + "epoch": 0.28, + "learning_rate": 1.9792325788793017e-05, + "loss": 1.4505, + "step": 3315 + }, + { + "epoch": 0.28, + "learning_rate": 1.9792141463034022e-05, + "loss": 1.3335, + "step": 3316 + }, + { + "epoch": 0.28, + "learning_rate": 1.9791957056369218e-05, + "loss": 1.3222, + "step": 3317 + }, + { + "epoch": 0.28, + "learning_rate": 1.9791772568800133e-05, + "loss": 1.3267, + "step": 3318 + }, + { + "epoch": 0.28, + "learning_rate": 1.979158800032829e-05, + "loss": 1.3306, + "step": 3319 + }, + { + "epoch": 0.28, + "learning_rate": 1.9791403350955213e-05, + "loss": 1.3396, + "step": 3320 + }, + { + "epoch": 0.28, + "learning_rate": 1.9791218620682427e-05, + "loss": 1.3299, + "step": 3321 + }, + { + "epoch": 0.28, + "learning_rate": 1.9791033809511462e-05, + "loss": 1.3059, + "step": 3322 + }, + { + "epoch": 0.28, + "learning_rate": 1.979084891744384e-05, + "loss": 1.4186, + "step": 3323 + }, + { + "epoch": 0.28, + "learning_rate": 1.979066394448109e-05, + "loss": 1.3477, + "step": 3324 + }, + { + "epoch": 0.28, + "learning_rate": 1.9790478890624745e-05, + "loss": 1.3226, + "step": 3325 + }, + { + "epoch": 0.28, + "learning_rate": 1.9790293755876325e-05, + "loss": 1.2519, + "step": 3326 + }, + { + "epoch": 0.28, + "learning_rate": 1.9790108540237367e-05, + "loss": 1.4335, + "step": 3327 + }, + { + "epoch": 0.28, + "learning_rate": 1.97899232437094e-05, + "loss": 1.3494, + "step": 3328 + }, + { + "epoch": 0.28, + "learning_rate": 1.978973786629395e-05, + "loss": 1.3508, + "step": 3329 + }, + { + "epoch": 0.28, + "learning_rate": 1.978955240799256e-05, + "loss": 1.3265, + "step": 3330 + }, + { + "epoch": 0.28, + "learning_rate": 1.9789366868806747e-05, + "loss": 1.3387, + "step": 3331 + }, + { + "epoch": 0.28, + "learning_rate": 1.9789181248738054e-05, + "loss": 1.3837, + "step": 3332 + }, + { + "epoch": 0.28, + "learning_rate": 1.9788995547788014e-05, + "loss": 1.4084, + "step": 3333 + }, + { + "epoch": 0.28, + "learning_rate": 1.9788809765958163e-05, + "loss": 1.3653, + "step": 3334 + }, + { + "epoch": 0.28, + "learning_rate": 1.9788623903250024e-05, + "loss": 1.3954, + "step": 3335 + }, + { + "epoch": 0.28, + "learning_rate": 1.978843795966515e-05, + "loss": 1.382, + "step": 3336 + }, + { + "epoch": 0.28, + "learning_rate": 1.9788251935205066e-05, + "loss": 1.3233, + "step": 3337 + }, + { + "epoch": 0.28, + "learning_rate": 1.9788065829871313e-05, + "loss": 1.4399, + "step": 3338 + }, + { + "epoch": 0.28, + "learning_rate": 1.9787879643665426e-05, + "loss": 1.2726, + "step": 3339 + }, + { + "epoch": 0.28, + "learning_rate": 1.9787693376588945e-05, + "loss": 1.2058, + "step": 3340 + }, + { + "epoch": 0.28, + "learning_rate": 1.9787507028643407e-05, + "loss": 1.3464, + "step": 3341 + }, + { + "epoch": 0.28, + "learning_rate": 1.978732059983036e-05, + "loss": 1.3516, + "step": 3342 + }, + { + "epoch": 0.28, + "learning_rate": 1.978713409015133e-05, + "loss": 1.3143, + "step": 3343 + }, + { + "epoch": 0.28, + "learning_rate": 1.9786947499607873e-05, + "loss": 1.391, + "step": 3344 + }, + { + "epoch": 0.28, + "learning_rate": 1.978676082820152e-05, + "loss": 1.3822, + "step": 3345 + }, + { + "epoch": 0.28, + "learning_rate": 1.9786574075933817e-05, + "loss": 1.3721, + "step": 3346 + }, + { + "epoch": 0.28, + "learning_rate": 1.9786387242806303e-05, + "loss": 1.3125, + "step": 3347 + }, + { + "epoch": 0.28, + "learning_rate": 1.9786200328820532e-05, + "loss": 1.3792, + "step": 3348 + }, + { + "epoch": 0.28, + "learning_rate": 1.978601333397804e-05, + "loss": 1.3914, + "step": 3349 + }, + { + "epoch": 0.28, + "learning_rate": 1.978582625828037e-05, + "loss": 1.4075, + "step": 3350 + }, + { + "epoch": 0.28, + "learning_rate": 1.9785639101729076e-05, + "loss": 1.3034, + "step": 3351 + }, + { + "epoch": 0.28, + "learning_rate": 1.97854518643257e-05, + "loss": 1.3671, + "step": 3352 + }, + { + "epoch": 0.28, + "learning_rate": 1.9785264546071787e-05, + "loss": 1.3784, + "step": 3353 + }, + { + "epoch": 0.28, + "learning_rate": 1.9785077146968885e-05, + "loss": 1.3413, + "step": 3354 + }, + { + "epoch": 0.28, + "learning_rate": 1.978488966701855e-05, + "loss": 1.3704, + "step": 3355 + }, + { + "epoch": 0.28, + "learning_rate": 1.978470210622232e-05, + "loss": 1.3552, + "step": 3356 + }, + { + "epoch": 0.28, + "learning_rate": 1.9784514464581747e-05, + "loss": 1.3052, + "step": 3357 + }, + { + "epoch": 0.28, + "learning_rate": 1.978432674209839e-05, + "loss": 1.3778, + "step": 3358 + }, + { + "epoch": 0.28, + "learning_rate": 1.978413893877379e-05, + "loss": 1.4429, + "step": 3359 + }, + { + "epoch": 0.28, + "learning_rate": 1.9783951054609507e-05, + "loss": 1.2471, + "step": 3360 + }, + { + "epoch": 0.28, + "learning_rate": 1.9783763089607088e-05, + "loss": 1.2918, + "step": 3361 + }, + { + "epoch": 0.28, + "learning_rate": 1.9783575043768083e-05, + "loss": 1.4013, + "step": 3362 + }, + { + "epoch": 0.28, + "learning_rate": 1.9783386917094054e-05, + "loss": 1.3562, + "step": 3363 + }, + { + "epoch": 0.28, + "learning_rate": 1.978319870958655e-05, + "loss": 1.2925, + "step": 3364 + }, + { + "epoch": 0.28, + "learning_rate": 1.9783010421247128e-05, + "loss": 1.3973, + "step": 3365 + }, + { + "epoch": 0.28, + "learning_rate": 1.978282205207734e-05, + "loss": 1.349, + "step": 3366 + }, + { + "epoch": 0.28, + "learning_rate": 1.978263360207875e-05, + "loss": 1.385, + "step": 3367 + }, + { + "epoch": 0.28, + "learning_rate": 1.9782445071252908e-05, + "loss": 1.3637, + "step": 3368 + }, + { + "epoch": 0.28, + "learning_rate": 1.9782256459601373e-05, + "loss": 1.3908, + "step": 3369 + }, + { + "epoch": 0.28, + "learning_rate": 1.9782067767125707e-05, + "loss": 1.344, + "step": 3370 + }, + { + "epoch": 0.28, + "learning_rate": 1.9781878993827465e-05, + "loss": 1.3692, + "step": 3371 + }, + { + "epoch": 0.28, + "learning_rate": 1.9781690139708205e-05, + "loss": 1.356, + "step": 3372 + }, + { + "epoch": 0.28, + "learning_rate": 1.9781501204769493e-05, + "loss": 1.2955, + "step": 3373 + }, + { + "epoch": 0.28, + "learning_rate": 1.9781312189012887e-05, + "loss": 1.2783, + "step": 3374 + }, + { + "epoch": 0.28, + "learning_rate": 1.9781123092439954e-05, + "loss": 1.3389, + "step": 3375 + }, + { + "epoch": 0.28, + "learning_rate": 1.9780933915052247e-05, + "loss": 1.3935, + "step": 3376 + }, + { + "epoch": 0.28, + "learning_rate": 1.9780744656851335e-05, + "loss": 1.386, + "step": 3377 + }, + { + "epoch": 0.28, + "learning_rate": 1.9780555317838778e-05, + "loss": 1.3139, + "step": 3378 + }, + { + "epoch": 0.28, + "learning_rate": 1.9780365898016147e-05, + "loss": 1.4488, + "step": 3379 + }, + { + "epoch": 0.28, + "learning_rate": 1.9780176397385e-05, + "loss": 1.2785, + "step": 3380 + }, + { + "epoch": 0.28, + "learning_rate": 1.9779986815946905e-05, + "loss": 1.3673, + "step": 3381 + }, + { + "epoch": 0.28, + "learning_rate": 1.977979715370343e-05, + "loss": 1.4086, + "step": 3382 + }, + { + "epoch": 0.28, + "learning_rate": 1.9779607410656144e-05, + "loss": 1.2691, + "step": 3383 + }, + { + "epoch": 0.28, + "learning_rate": 1.9779417586806607e-05, + "loss": 1.237, + "step": 3384 + }, + { + "epoch": 0.28, + "learning_rate": 1.977922768215639e-05, + "loss": 1.3398, + "step": 3385 + }, + { + "epoch": 0.29, + "learning_rate": 1.9779037696707072e-05, + "loss": 1.357, + "step": 3386 + }, + { + "epoch": 0.29, + "learning_rate": 1.9778847630460212e-05, + "loss": 1.3546, + "step": 3387 + }, + { + "epoch": 0.29, + "learning_rate": 1.977865748341738e-05, + "loss": 1.386, + "step": 3388 + }, + { + "epoch": 0.29, + "learning_rate": 1.977846725558015e-05, + "loss": 1.3114, + "step": 3389 + }, + { + "epoch": 0.29, + "learning_rate": 1.97782769469501e-05, + "loss": 1.3233, + "step": 3390 + }, + { + "epoch": 0.29, + "learning_rate": 1.9778086557528793e-05, + "loss": 1.321, + "step": 3391 + }, + { + "epoch": 0.29, + "learning_rate": 1.9777896087317806e-05, + "loss": 1.3362, + "step": 3392 + }, + { + "epoch": 0.29, + "learning_rate": 1.977770553631871e-05, + "loss": 1.3444, + "step": 3393 + }, + { + "epoch": 0.29, + "learning_rate": 1.9777514904533084e-05, + "loss": 1.3262, + "step": 3394 + }, + { + "epoch": 0.29, + "learning_rate": 1.9777324191962498e-05, + "loss": 1.3064, + "step": 3395 + }, + { + "epoch": 0.29, + "learning_rate": 1.9777133398608536e-05, + "loss": 1.4629, + "step": 3396 + }, + { + "epoch": 0.29, + "learning_rate": 1.977694252447276e-05, + "loss": 1.351, + "step": 3397 + }, + { + "epoch": 0.29, + "learning_rate": 1.977675156955676e-05, + "loss": 1.3227, + "step": 3398 + }, + { + "epoch": 0.29, + "learning_rate": 1.9776560533862112e-05, + "loss": 1.4247, + "step": 3399 + }, + { + "epoch": 0.29, + "learning_rate": 1.9776369417390394e-05, + "loss": 1.4257, + "step": 3400 + }, + { + "epoch": 0.29, + "learning_rate": 1.9776178220143177e-05, + "loss": 1.2566, + "step": 3401 + }, + { + "epoch": 0.29, + "learning_rate": 1.977598694212205e-05, + "loss": 1.276, + "step": 3402 + }, + { + "epoch": 0.29, + "learning_rate": 1.9775795583328586e-05, + "loss": 1.3255, + "step": 3403 + }, + { + "epoch": 0.29, + "learning_rate": 1.9775604143764376e-05, + "loss": 1.3936, + "step": 3404 + }, + { + "epoch": 0.29, + "learning_rate": 1.9775412623430994e-05, + "loss": 1.3693, + "step": 3405 + }, + { + "epoch": 0.29, + "learning_rate": 1.9775221022330022e-05, + "loss": 1.3406, + "step": 3406 + }, + { + "epoch": 0.29, + "learning_rate": 1.977502934046305e-05, + "loss": 1.3324, + "step": 3407 + }, + { + "epoch": 0.29, + "learning_rate": 1.977483757783165e-05, + "loss": 1.2681, + "step": 3408 + }, + { + "epoch": 0.29, + "learning_rate": 1.9774645734437422e-05, + "loss": 1.378, + "step": 3409 + }, + { + "epoch": 0.29, + "learning_rate": 1.9774453810281936e-05, + "loss": 1.2535, + "step": 3410 + }, + { + "epoch": 0.29, + "learning_rate": 1.977426180536679e-05, + "loss": 1.4015, + "step": 3411 + }, + { + "epoch": 0.29, + "learning_rate": 1.977406971969356e-05, + "loss": 1.3605, + "step": 3412 + }, + { + "epoch": 0.29, + "learning_rate": 1.977387755326384e-05, + "loss": 1.3256, + "step": 3413 + }, + { + "epoch": 0.29, + "learning_rate": 1.9773685306079217e-05, + "loss": 1.3167, + "step": 3414 + }, + { + "epoch": 0.29, + "learning_rate": 1.9773492978141275e-05, + "loss": 1.3533, + "step": 3415 + }, + { + "epoch": 0.29, + "learning_rate": 1.9773300569451607e-05, + "loss": 1.4536, + "step": 3416 + }, + { + "epoch": 0.29, + "learning_rate": 1.9773108080011804e-05, + "loss": 1.3713, + "step": 3417 + }, + { + "epoch": 0.29, + "learning_rate": 1.9772915509823452e-05, + "loss": 1.3188, + "step": 3418 + }, + { + "epoch": 0.29, + "learning_rate": 1.9772722858888144e-05, + "loss": 1.3422, + "step": 3419 + }, + { + "epoch": 0.29, + "learning_rate": 1.9772530127207475e-05, + "loss": 1.4048, + "step": 3420 + }, + { + "epoch": 0.29, + "learning_rate": 1.977233731478303e-05, + "loss": 1.2959, + "step": 3421 + }, + { + "epoch": 0.29, + "learning_rate": 1.977214442161641e-05, + "loss": 1.3581, + "step": 3422 + }, + { + "epoch": 0.29, + "learning_rate": 1.9771951447709204e-05, + "loss": 1.3028, + "step": 3423 + }, + { + "epoch": 0.29, + "learning_rate": 1.977175839306301e-05, + "loss": 1.3764, + "step": 3424 + }, + { + "epoch": 0.29, + "learning_rate": 1.9771565257679416e-05, + "loss": 1.458, + "step": 3425 + }, + { + "epoch": 0.29, + "learning_rate": 1.9771372041560026e-05, + "loss": 1.3135, + "step": 3426 + }, + { + "epoch": 0.29, + "learning_rate": 1.9771178744706435e-05, + "loss": 1.2085, + "step": 3427 + }, + { + "epoch": 0.29, + "learning_rate": 1.9770985367120234e-05, + "loss": 1.3445, + "step": 3428 + }, + { + "epoch": 0.29, + "learning_rate": 1.9770791908803027e-05, + "loss": 1.3945, + "step": 3429 + }, + { + "epoch": 0.29, + "learning_rate": 1.9770598369756405e-05, + "loss": 1.3194, + "step": 3430 + }, + { + "epoch": 0.29, + "learning_rate": 1.977040474998198e-05, + "loss": 1.4099, + "step": 3431 + }, + { + "epoch": 0.29, + "learning_rate": 1.977021104948134e-05, + "loss": 1.3114, + "step": 3432 + }, + { + "epoch": 0.29, + "learning_rate": 1.977001726825609e-05, + "loss": 1.422, + "step": 3433 + }, + { + "epoch": 0.29, + "learning_rate": 1.9769823406307828e-05, + "loss": 1.3925, + "step": 3434 + }, + { + "epoch": 0.29, + "learning_rate": 1.976962946363816e-05, + "loss": 1.3443, + "step": 3435 + }, + { + "epoch": 0.29, + "learning_rate": 1.976943544024869e-05, + "loss": 1.2729, + "step": 3436 + }, + { + "epoch": 0.29, + "learning_rate": 1.9769241336141016e-05, + "loss": 1.3672, + "step": 3437 + }, + { + "epoch": 0.29, + "learning_rate": 1.9769047151316742e-05, + "loss": 1.3466, + "step": 3438 + }, + { + "epoch": 0.29, + "learning_rate": 1.9768852885777473e-05, + "loss": 1.4457, + "step": 3439 + }, + { + "epoch": 0.29, + "learning_rate": 1.976865853952482e-05, + "loss": 1.2872, + "step": 3440 + }, + { + "epoch": 0.29, + "learning_rate": 1.976846411256038e-05, + "loss": 1.3444, + "step": 3441 + }, + { + "epoch": 0.29, + "learning_rate": 1.9768269604885764e-05, + "loss": 1.3491, + "step": 3442 + }, + { + "epoch": 0.29, + "learning_rate": 1.976807501650258e-05, + "loss": 1.3206, + "step": 3443 + }, + { + "epoch": 0.29, + "learning_rate": 1.976788034741243e-05, + "loss": 1.3094, + "step": 3444 + }, + { + "epoch": 0.29, + "learning_rate": 1.976768559761693e-05, + "loss": 1.4173, + "step": 3445 + }, + { + "epoch": 0.29, + "learning_rate": 1.9767490767117682e-05, + "loss": 1.3042, + "step": 3446 + }, + { + "epoch": 0.29, + "learning_rate": 1.9767295855916303e-05, + "loss": 1.3988, + "step": 3447 + }, + { + "epoch": 0.29, + "learning_rate": 1.97671008640144e-05, + "loss": 1.3692, + "step": 3448 + }, + { + "epoch": 0.29, + "learning_rate": 1.9766905791413578e-05, + "loss": 1.2833, + "step": 3449 + }, + { + "epoch": 0.29, + "learning_rate": 1.976671063811546e-05, + "loss": 1.3917, + "step": 3450 + }, + { + "epoch": 0.29, + "learning_rate": 1.976651540412165e-05, + "loss": 1.3236, + "step": 3451 + }, + { + "epoch": 0.29, + "learning_rate": 1.9766320089433766e-05, + "loss": 1.4112, + "step": 3452 + }, + { + "epoch": 0.29, + "learning_rate": 1.976612469405342e-05, + "loss": 1.4108, + "step": 3453 + }, + { + "epoch": 0.29, + "learning_rate": 1.9765929217982224e-05, + "loss": 1.3626, + "step": 3454 + }, + { + "epoch": 0.29, + "learning_rate": 1.9765733661221796e-05, + "loss": 1.2827, + "step": 3455 + }, + { + "epoch": 0.29, + "learning_rate": 1.9765538023773755e-05, + "loss": 1.2465, + "step": 3456 + }, + { + "epoch": 0.29, + "learning_rate": 1.9765342305639706e-05, + "loss": 1.3329, + "step": 3457 + }, + { + "epoch": 0.29, + "learning_rate": 1.976514650682128e-05, + "loss": 1.4288, + "step": 3458 + }, + { + "epoch": 0.29, + "learning_rate": 1.9764950627320086e-05, + "loss": 1.3499, + "step": 3459 + }, + { + "epoch": 0.29, + "learning_rate": 1.9764754667137747e-05, + "loss": 1.3048, + "step": 3460 + }, + { + "epoch": 0.29, + "learning_rate": 1.9764558626275875e-05, + "loss": 1.3155, + "step": 3461 + }, + { + "epoch": 0.29, + "learning_rate": 1.9764362504736097e-05, + "loss": 1.2827, + "step": 3462 + }, + { + "epoch": 0.29, + "learning_rate": 1.9764166302520033e-05, + "loss": 1.3267, + "step": 3463 + }, + { + "epoch": 0.29, + "learning_rate": 1.9763970019629298e-05, + "loss": 1.3225, + "step": 3464 + }, + { + "epoch": 0.29, + "learning_rate": 1.9763773656065522e-05, + "loss": 1.3379, + "step": 3465 + }, + { + "epoch": 0.29, + "learning_rate": 1.976357721183032e-05, + "loss": 1.4016, + "step": 3466 + }, + { + "epoch": 0.29, + "learning_rate": 1.976338068692532e-05, + "loss": 1.3562, + "step": 3467 + }, + { + "epoch": 0.29, + "learning_rate": 1.9763184081352145e-05, + "loss": 1.379, + "step": 3468 + }, + { + "epoch": 0.29, + "learning_rate": 1.9762987395112418e-05, + "loss": 1.3672, + "step": 3469 + }, + { + "epoch": 0.29, + "learning_rate": 1.976279062820776e-05, + "loss": 1.3442, + "step": 3470 + }, + { + "epoch": 0.29, + "learning_rate": 1.9762593780639808e-05, + "loss": 1.2845, + "step": 3471 + }, + { + "epoch": 0.29, + "learning_rate": 1.9762396852410178e-05, + "loss": 1.3449, + "step": 3472 + }, + { + "epoch": 0.29, + "learning_rate": 1.97621998435205e-05, + "loss": 1.3628, + "step": 3473 + }, + { + "epoch": 0.29, + "learning_rate": 1.9762002753972406e-05, + "loss": 1.3633, + "step": 3474 + }, + { + "epoch": 0.29, + "learning_rate": 1.976180558376752e-05, + "loss": 1.3352, + "step": 3475 + }, + { + "epoch": 0.29, + "learning_rate": 1.976160833290747e-05, + "loss": 1.2244, + "step": 3476 + }, + { + "epoch": 0.29, + "learning_rate": 1.976141100139389e-05, + "loss": 1.3078, + "step": 3477 + }, + { + "epoch": 0.29, + "learning_rate": 1.976121358922841e-05, + "loss": 1.2968, + "step": 3478 + }, + { + "epoch": 0.29, + "learning_rate": 1.9761016096412654e-05, + "loss": 1.3786, + "step": 3479 + }, + { + "epoch": 0.29, + "learning_rate": 1.9760818522948263e-05, + "loss": 1.3066, + "step": 3480 + }, + { + "epoch": 0.29, + "learning_rate": 1.9760620868836863e-05, + "loss": 1.3494, + "step": 3481 + }, + { + "epoch": 0.29, + "learning_rate": 1.9760423134080092e-05, + "loss": 1.2888, + "step": 3482 + }, + { + "epoch": 0.29, + "learning_rate": 1.976022531867958e-05, + "loss": 1.4381, + "step": 3483 + }, + { + "epoch": 0.29, + "learning_rate": 1.9760027422636967e-05, + "loss": 1.3359, + "step": 3484 + }, + { + "epoch": 0.29, + "learning_rate": 1.9759829445953878e-05, + "loss": 1.4117, + "step": 3485 + }, + { + "epoch": 0.29, + "learning_rate": 1.9759631388631958e-05, + "loss": 1.357, + "step": 3486 + }, + { + "epoch": 0.29, + "learning_rate": 1.9759433250672837e-05, + "loss": 1.4193, + "step": 3487 + }, + { + "epoch": 0.29, + "learning_rate": 1.9759235032078155e-05, + "loss": 1.3349, + "step": 3488 + }, + { + "epoch": 0.29, + "learning_rate": 1.9759036732849552e-05, + "loss": 1.3626, + "step": 3489 + }, + { + "epoch": 0.29, + "learning_rate": 1.9758838352988667e-05, + "loss": 1.2859, + "step": 3490 + }, + { + "epoch": 0.29, + "learning_rate": 1.975863989249713e-05, + "loss": 1.3062, + "step": 3491 + }, + { + "epoch": 0.29, + "learning_rate": 1.975844135137659e-05, + "loss": 1.3728, + "step": 3492 + }, + { + "epoch": 0.29, + "learning_rate": 1.9758242729628683e-05, + "loss": 1.3276, + "step": 3493 + }, + { + "epoch": 0.29, + "learning_rate": 1.9758044027255057e-05, + "loss": 1.3153, + "step": 3494 + }, + { + "epoch": 0.29, + "learning_rate": 1.975784524425734e-05, + "loss": 1.3265, + "step": 3495 + }, + { + "epoch": 0.29, + "learning_rate": 1.9757646380637187e-05, + "loss": 1.264, + "step": 3496 + }, + { + "epoch": 0.29, + "learning_rate": 1.975744743639624e-05, + "loss": 1.4686, + "step": 3497 + }, + { + "epoch": 0.29, + "learning_rate": 1.9757248411536137e-05, + "loss": 1.4069, + "step": 3498 + }, + { + "epoch": 0.29, + "learning_rate": 1.975704930605852e-05, + "loss": 1.2884, + "step": 3499 + }, + { + "epoch": 0.29, + "learning_rate": 1.9756850119965046e-05, + "loss": 1.3447, + "step": 3500 + }, + { + "epoch": 0.29, + "learning_rate": 1.975665085325735e-05, + "loss": 1.3539, + "step": 3501 + }, + { + "epoch": 0.29, + "learning_rate": 1.9756451505937085e-05, + "loss": 1.3594, + "step": 3502 + }, + { + "epoch": 0.29, + "learning_rate": 1.975625207800589e-05, + "loss": 1.2562, + "step": 3503 + }, + { + "epoch": 0.3, + "learning_rate": 1.975605256946542e-05, + "loss": 1.3389, + "step": 3504 + }, + { + "epoch": 0.3, + "learning_rate": 1.9755852980317326e-05, + "loss": 1.3863, + "step": 3505 + }, + { + "epoch": 0.3, + "learning_rate": 1.9755653310563247e-05, + "loss": 1.3181, + "step": 3506 + }, + { + "epoch": 0.3, + "learning_rate": 1.975545356020484e-05, + "loss": 1.2828, + "step": 3507 + }, + { + "epoch": 0.3, + "learning_rate": 1.9755253729243755e-05, + "loss": 1.3362, + "step": 3508 + }, + { + "epoch": 0.3, + "learning_rate": 1.975505381768164e-05, + "loss": 1.3098, + "step": 3509 + }, + { + "epoch": 0.3, + "learning_rate": 1.9754853825520148e-05, + "loss": 1.3321, + "step": 3510 + }, + { + "epoch": 0.3, + "learning_rate": 1.9754653752760932e-05, + "loss": 1.2291, + "step": 3511 + }, + { + "epoch": 0.3, + "learning_rate": 1.9754453599405643e-05, + "loss": 1.416, + "step": 3512 + }, + { + "epoch": 0.3, + "learning_rate": 1.975425336545594e-05, + "loss": 1.299, + "step": 3513 + }, + { + "epoch": 0.3, + "learning_rate": 1.9754053050913473e-05, + "loss": 1.4507, + "step": 3514 + }, + { + "epoch": 0.3, + "learning_rate": 1.9753852655779897e-05, + "loss": 1.3517, + "step": 3515 + }, + { + "epoch": 0.3, + "learning_rate": 1.975365218005687e-05, + "loss": 1.3855, + "step": 3516 + }, + { + "epoch": 0.3, + "learning_rate": 1.9753451623746046e-05, + "loss": 1.4462, + "step": 3517 + }, + { + "epoch": 0.3, + "learning_rate": 1.975325098684908e-05, + "loss": 1.2787, + "step": 3518 + }, + { + "epoch": 0.3, + "learning_rate": 1.9753050269367637e-05, + "loss": 1.3095, + "step": 3519 + }, + { + "epoch": 0.3, + "learning_rate": 1.975284947130337e-05, + "loss": 1.3671, + "step": 3520 + }, + { + "epoch": 0.3, + "learning_rate": 1.9752648592657944e-05, + "loss": 1.4052, + "step": 3521 + }, + { + "epoch": 0.3, + "learning_rate": 1.9752447633433006e-05, + "loss": 1.3896, + "step": 3522 + }, + { + "epoch": 0.3, + "learning_rate": 1.975224659363023e-05, + "loss": 1.3338, + "step": 3523 + }, + { + "epoch": 0.3, + "learning_rate": 1.975204547325127e-05, + "loss": 1.3181, + "step": 3524 + }, + { + "epoch": 0.3, + "learning_rate": 1.975184427229779e-05, + "loss": 1.3932, + "step": 3525 + }, + { + "epoch": 0.3, + "learning_rate": 1.975164299077145e-05, + "loss": 1.2307, + "step": 3526 + }, + { + "epoch": 0.3, + "learning_rate": 1.9751441628673916e-05, + "loss": 1.3915, + "step": 3527 + }, + { + "epoch": 0.3, + "learning_rate": 1.975124018600685e-05, + "loss": 1.307, + "step": 3528 + }, + { + "epoch": 0.3, + "learning_rate": 1.975103866277192e-05, + "loss": 1.3392, + "step": 3529 + }, + { + "epoch": 0.3, + "learning_rate": 1.975083705897078e-05, + "loss": 1.4081, + "step": 3530 + }, + { + "epoch": 0.3, + "learning_rate": 1.975063537460511e-05, + "loss": 1.3875, + "step": 3531 + }, + { + "epoch": 0.3, + "learning_rate": 1.975043360967657e-05, + "loss": 1.4103, + "step": 3532 + }, + { + "epoch": 0.3, + "learning_rate": 1.9750231764186824e-05, + "loss": 1.3893, + "step": 3533 + }, + { + "epoch": 0.3, + "learning_rate": 1.9750029838137544e-05, + "loss": 1.347, + "step": 3534 + }, + { + "epoch": 0.3, + "learning_rate": 1.9749827831530397e-05, + "loss": 1.4317, + "step": 3535 + }, + { + "epoch": 0.3, + "learning_rate": 1.974962574436705e-05, + "loss": 1.3897, + "step": 3536 + }, + { + "epoch": 0.3, + "learning_rate": 1.974942357664918e-05, + "loss": 1.3145, + "step": 3537 + }, + { + "epoch": 0.3, + "learning_rate": 1.9749221328378446e-05, + "loss": 1.4912, + "step": 3538 + }, + { + "epoch": 0.3, + "learning_rate": 1.974901899955653e-05, + "loss": 1.3598, + "step": 3539 + }, + { + "epoch": 0.3, + "learning_rate": 1.9748816590185097e-05, + "loss": 1.2973, + "step": 3540 + }, + { + "epoch": 0.3, + "learning_rate": 1.974861410026582e-05, + "loss": 1.3233, + "step": 3541 + }, + { + "epoch": 0.3, + "learning_rate": 1.9748411529800374e-05, + "loss": 1.3147, + "step": 3542 + }, + { + "epoch": 0.3, + "learning_rate": 1.9748208878790435e-05, + "loss": 1.3641, + "step": 3543 + }, + { + "epoch": 0.3, + "learning_rate": 1.974800614723767e-05, + "loss": 1.3599, + "step": 3544 + }, + { + "epoch": 0.3, + "learning_rate": 1.9747803335143763e-05, + "loss": 1.3296, + "step": 3545 + }, + { + "epoch": 0.3, + "learning_rate": 1.9747600442510385e-05, + "loss": 1.3904, + "step": 3546 + }, + { + "epoch": 0.3, + "learning_rate": 1.9747397469339206e-05, + "loss": 1.3382, + "step": 3547 + }, + { + "epoch": 0.3, + "learning_rate": 1.9747194415631915e-05, + "loss": 1.3227, + "step": 3548 + }, + { + "epoch": 0.3, + "learning_rate": 1.9746991281390184e-05, + "loss": 1.3508, + "step": 3549 + }, + { + "epoch": 0.3, + "learning_rate": 1.974678806661569e-05, + "loss": 1.3869, + "step": 3550 + }, + { + "epoch": 0.3, + "learning_rate": 1.9746584771310114e-05, + "loss": 1.3034, + "step": 3551 + }, + { + "epoch": 0.3, + "learning_rate": 1.974638139547514e-05, + "loss": 1.4049, + "step": 3552 + }, + { + "epoch": 0.3, + "learning_rate": 1.9746177939112436e-05, + "loss": 1.3791, + "step": 3553 + }, + { + "epoch": 0.3, + "learning_rate": 1.9745974402223694e-05, + "loss": 1.3575, + "step": 3554 + }, + { + "epoch": 0.3, + "learning_rate": 1.9745770784810593e-05, + "loss": 1.3639, + "step": 3555 + }, + { + "epoch": 0.3, + "learning_rate": 1.9745567086874815e-05, + "loss": 1.3234, + "step": 3556 + }, + { + "epoch": 0.3, + "learning_rate": 1.974536330841804e-05, + "loss": 1.3865, + "step": 3557 + }, + { + "epoch": 0.3, + "learning_rate": 1.9745159449441956e-05, + "loss": 1.395, + "step": 3558 + }, + { + "epoch": 0.3, + "learning_rate": 1.9744955509948246e-05, + "loss": 1.3425, + "step": 3559 + }, + { + "epoch": 0.3, + "learning_rate": 1.9744751489938595e-05, + "loss": 1.3131, + "step": 3560 + }, + { + "epoch": 0.3, + "learning_rate": 1.974454738941469e-05, + "loss": 1.29, + "step": 3561 + }, + { + "epoch": 0.3, + "learning_rate": 1.9744343208378214e-05, + "loss": 1.245, + "step": 3562 + }, + { + "epoch": 0.3, + "learning_rate": 1.9744138946830857e-05, + "loss": 1.3833, + "step": 3563 + }, + { + "epoch": 0.3, + "learning_rate": 1.9743934604774303e-05, + "loss": 1.3368, + "step": 3564 + }, + { + "epoch": 0.3, + "learning_rate": 1.9743730182210242e-05, + "loss": 1.332, + "step": 3565 + }, + { + "epoch": 0.3, + "learning_rate": 1.9743525679140368e-05, + "loss": 1.3346, + "step": 3566 + }, + { + "epoch": 0.3, + "learning_rate": 1.9743321095566366e-05, + "loss": 1.3066, + "step": 3567 + }, + { + "epoch": 0.3, + "learning_rate": 1.9743116431489925e-05, + "loss": 1.3682, + "step": 3568 + }, + { + "epoch": 0.3, + "learning_rate": 1.974291168691274e-05, + "loss": 1.3559, + "step": 3569 + }, + { + "epoch": 0.3, + "learning_rate": 1.9742706861836498e-05, + "loss": 1.2361, + "step": 3570 + }, + { + "epoch": 0.3, + "learning_rate": 1.9742501956262893e-05, + "loss": 1.3507, + "step": 3571 + }, + { + "epoch": 0.3, + "learning_rate": 1.9742296970193623e-05, + "loss": 1.4653, + "step": 3572 + }, + { + "epoch": 0.3, + "learning_rate": 1.9742091903630375e-05, + "loss": 1.3289, + "step": 3573 + }, + { + "epoch": 0.3, + "learning_rate": 1.9741886756574846e-05, + "loss": 1.2628, + "step": 3574 + }, + { + "epoch": 0.3, + "learning_rate": 1.9741681529028728e-05, + "loss": 1.3097, + "step": 3575 + }, + { + "epoch": 0.3, + "learning_rate": 1.9741476220993723e-05, + "loss": 1.3417, + "step": 3576 + }, + { + "epoch": 0.3, + "learning_rate": 1.9741270832471524e-05, + "loss": 1.414, + "step": 3577 + }, + { + "epoch": 0.3, + "learning_rate": 1.9741065363463827e-05, + "loss": 1.3674, + "step": 3578 + }, + { + "epoch": 0.3, + "learning_rate": 1.974085981397233e-05, + "loss": 1.3631, + "step": 3579 + }, + { + "epoch": 0.3, + "learning_rate": 1.974065418399873e-05, + "loss": 1.2523, + "step": 3580 + }, + { + "epoch": 0.3, + "learning_rate": 1.9740448473544733e-05, + "loss": 1.4071, + "step": 3581 + }, + { + "epoch": 0.3, + "learning_rate": 1.9740242682612027e-05, + "loss": 1.3427, + "step": 3582 + }, + { + "epoch": 0.3, + "learning_rate": 1.9740036811202325e-05, + "loss": 1.3047, + "step": 3583 + }, + { + "epoch": 0.3, + "learning_rate": 1.9739830859317316e-05, + "loss": 1.3597, + "step": 3584 + }, + { + "epoch": 0.3, + "learning_rate": 1.973962482695871e-05, + "loss": 1.4263, + "step": 3585 + }, + { + "epoch": 0.3, + "learning_rate": 1.9739418714128208e-05, + "loss": 1.4026, + "step": 3586 + }, + { + "epoch": 0.3, + "learning_rate": 1.973921252082751e-05, + "loss": 1.3549, + "step": 3587 + }, + { + "epoch": 0.3, + "learning_rate": 1.973900624705832e-05, + "loss": 1.2693, + "step": 3588 + }, + { + "epoch": 0.3, + "learning_rate": 1.973879989282234e-05, + "loss": 1.5216, + "step": 3589 + }, + { + "epoch": 0.3, + "learning_rate": 1.973859345812129e-05, + "loss": 1.3899, + "step": 3590 + }, + { + "epoch": 0.3, + "learning_rate": 1.9738386942956853e-05, + "loss": 1.3559, + "step": 3591 + }, + { + "epoch": 0.3, + "learning_rate": 1.9738180347330755e-05, + "loss": 1.3367, + "step": 3592 + }, + { + "epoch": 0.3, + "learning_rate": 1.9737973671244685e-05, + "loss": 1.4531, + "step": 3593 + }, + { + "epoch": 0.3, + "learning_rate": 1.9737766914700366e-05, + "loss": 1.3195, + "step": 3594 + }, + { + "epoch": 0.3, + "learning_rate": 1.97375600776995e-05, + "loss": 1.3367, + "step": 3595 + }, + { + "epoch": 0.3, + "learning_rate": 1.9737353160243793e-05, + "loss": 1.2887, + "step": 3596 + }, + { + "epoch": 0.3, + "learning_rate": 1.9737146162334962e-05, + "loss": 1.3458, + "step": 3597 + }, + { + "epoch": 0.3, + "learning_rate": 1.973693908397471e-05, + "loss": 1.3658, + "step": 3598 + }, + { + "epoch": 0.3, + "learning_rate": 1.9736731925164755e-05, + "loss": 1.3668, + "step": 3599 + }, + { + "epoch": 0.3, + "learning_rate": 1.97365246859068e-05, + "loss": 1.2905, + "step": 3600 + }, + { + "epoch": 0.3, + "learning_rate": 1.9736317366202564e-05, + "loss": 1.3302, + "step": 3601 + }, + { + "epoch": 0.3, + "learning_rate": 1.973610996605376e-05, + "loss": 1.2946, + "step": 3602 + }, + { + "epoch": 0.3, + "learning_rate": 1.9735902485462098e-05, + "loss": 1.3684, + "step": 3603 + }, + { + "epoch": 0.3, + "learning_rate": 1.973569492442929e-05, + "loss": 1.3638, + "step": 3604 + }, + { + "epoch": 0.3, + "learning_rate": 1.9735487282957057e-05, + "loss": 1.2965, + "step": 3605 + }, + { + "epoch": 0.3, + "learning_rate": 1.9735279561047112e-05, + "loss": 1.331, + "step": 3606 + }, + { + "epoch": 0.3, + "learning_rate": 1.9735071758701178e-05, + "loss": 1.2771, + "step": 3607 + }, + { + "epoch": 0.3, + "learning_rate": 1.9734863875920957e-05, + "loss": 1.3241, + "step": 3608 + }, + { + "epoch": 0.3, + "learning_rate": 1.9734655912708178e-05, + "loss": 1.3514, + "step": 3609 + }, + { + "epoch": 0.3, + "learning_rate": 1.973444786906456e-05, + "loss": 1.3082, + "step": 3610 + }, + { + "epoch": 0.3, + "learning_rate": 1.9734239744991815e-05, + "loss": 1.3656, + "step": 3611 + }, + { + "epoch": 0.3, + "learning_rate": 1.9734031540491662e-05, + "loss": 1.4803, + "step": 3612 + }, + { + "epoch": 0.3, + "learning_rate": 1.9733823255565832e-05, + "loss": 1.3483, + "step": 3613 + }, + { + "epoch": 0.3, + "learning_rate": 1.9733614890216034e-05, + "loss": 1.3497, + "step": 3614 + }, + { + "epoch": 0.3, + "learning_rate": 1.9733406444443998e-05, + "loss": 1.3523, + "step": 3615 + }, + { + "epoch": 0.3, + "learning_rate": 1.973319791825144e-05, + "loss": 1.4108, + "step": 3616 + }, + { + "epoch": 0.3, + "learning_rate": 1.973298931164009e-05, + "loss": 1.3231, + "step": 3617 + }, + { + "epoch": 0.3, + "learning_rate": 1.9732780624611664e-05, + "loss": 1.3894, + "step": 3618 + }, + { + "epoch": 0.3, + "learning_rate": 1.9732571857167892e-05, + "loss": 1.3928, + "step": 3619 + }, + { + "epoch": 0.3, + "learning_rate": 1.973236300931049e-05, + "loss": 1.2892, + "step": 3620 + }, + { + "epoch": 0.3, + "learning_rate": 1.9732154081041197e-05, + "loss": 1.4034, + "step": 3621 + }, + { + "epoch": 0.3, + "learning_rate": 1.973194507236173e-05, + "loss": 1.3955, + "step": 3622 + }, + { + "epoch": 0.31, + "learning_rate": 1.9731735983273818e-05, + "loss": 1.286, + "step": 3623 + }, + { + "epoch": 0.31, + "learning_rate": 1.973152681377919e-05, + "loss": 1.3418, + "step": 3624 + }, + { + "epoch": 0.31, + "learning_rate": 1.9731317563879572e-05, + "loss": 1.3553, + "step": 3625 + }, + { + "epoch": 0.31, + "learning_rate": 1.9731108233576694e-05, + "loss": 1.3102, + "step": 3626 + }, + { + "epoch": 0.31, + "learning_rate": 1.9730898822872287e-05, + "loss": 1.4405, + "step": 3627 + }, + { + "epoch": 0.31, + "learning_rate": 1.9730689331768074e-05, + "loss": 1.3622, + "step": 3628 + }, + { + "epoch": 0.31, + "learning_rate": 1.97304797602658e-05, + "loss": 1.2575, + "step": 3629 + }, + { + "epoch": 0.31, + "learning_rate": 1.973027010836718e-05, + "loss": 1.3627, + "step": 3630 + }, + { + "epoch": 0.31, + "learning_rate": 1.9730060376073958e-05, + "loss": 1.3505, + "step": 3631 + }, + { + "epoch": 0.31, + "learning_rate": 1.9729850563387863e-05, + "loss": 1.3096, + "step": 3632 + }, + { + "epoch": 0.31, + "learning_rate": 1.972964067031063e-05, + "loss": 1.3288, + "step": 3633 + }, + { + "epoch": 0.31, + "learning_rate": 1.972943069684399e-05, + "loss": 1.2241, + "step": 3634 + }, + { + "epoch": 0.31, + "learning_rate": 1.9729220642989682e-05, + "loss": 1.2938, + "step": 3635 + }, + { + "epoch": 0.31, + "learning_rate": 1.9729010508749436e-05, + "loss": 1.3311, + "step": 3636 + }, + { + "epoch": 0.31, + "learning_rate": 1.972880029412499e-05, + "loss": 1.3742, + "step": 3637 + }, + { + "epoch": 0.31, + "learning_rate": 1.972858999911809e-05, + "loss": 1.3194, + "step": 3638 + }, + { + "epoch": 0.31, + "learning_rate": 1.972837962373046e-05, + "loss": 1.3486, + "step": 3639 + }, + { + "epoch": 0.31, + "learning_rate": 1.9728169167963845e-05, + "loss": 1.3044, + "step": 3640 + }, + { + "epoch": 0.31, + "learning_rate": 1.9727958631819985e-05, + "loss": 1.3073, + "step": 3641 + }, + { + "epoch": 0.31, + "learning_rate": 1.9727748015300616e-05, + "loss": 1.2652, + "step": 3642 + }, + { + "epoch": 0.31, + "learning_rate": 1.9727537318407476e-05, + "loss": 1.3997, + "step": 3643 + }, + { + "epoch": 0.31, + "learning_rate": 1.9727326541142314e-05, + "loss": 1.3214, + "step": 3644 + }, + { + "epoch": 0.31, + "learning_rate": 1.9727115683506864e-05, + "loss": 1.429, + "step": 3645 + }, + { + "epoch": 0.31, + "learning_rate": 1.972690474550288e-05, + "loss": 1.3595, + "step": 3646 + }, + { + "epoch": 0.31, + "learning_rate": 1.9726693727132086e-05, + "loss": 1.2665, + "step": 3647 + }, + { + "epoch": 0.31, + "learning_rate": 1.9726482628396236e-05, + "loss": 1.2256, + "step": 3648 + }, + { + "epoch": 0.31, + "learning_rate": 1.9726271449297075e-05, + "loss": 1.2686, + "step": 3649 + }, + { + "epoch": 0.31, + "learning_rate": 1.972606018983635e-05, + "loss": 1.3346, + "step": 3650 + }, + { + "epoch": 0.31, + "learning_rate": 1.9725848850015802e-05, + "loss": 1.2624, + "step": 3651 + }, + { + "epoch": 0.31, + "learning_rate": 1.9725637429837173e-05, + "loss": 1.2143, + "step": 3652 + }, + { + "epoch": 0.31, + "learning_rate": 1.9725425929302223e-05, + "loss": 1.4239, + "step": 3653 + }, + { + "epoch": 0.31, + "learning_rate": 1.9725214348412685e-05, + "loss": 1.3899, + "step": 3654 + }, + { + "epoch": 0.31, + "learning_rate": 1.9725002687170316e-05, + "loss": 1.4029, + "step": 3655 + }, + { + "epoch": 0.31, + "learning_rate": 1.9724790945576863e-05, + "loss": 1.3264, + "step": 3656 + }, + { + "epoch": 0.31, + "learning_rate": 1.9724579123634074e-05, + "loss": 1.2855, + "step": 3657 + }, + { + "epoch": 0.31, + "learning_rate": 1.97243672213437e-05, + "loss": 1.2423, + "step": 3658 + }, + { + "epoch": 0.31, + "learning_rate": 1.9724155238707488e-05, + "loss": 1.3154, + "step": 3659 + }, + { + "epoch": 0.31, + "learning_rate": 1.97239431757272e-05, + "loss": 1.3138, + "step": 3660 + }, + { + "epoch": 0.31, + "learning_rate": 1.9723731032404577e-05, + "loss": 1.3537, + "step": 3661 + }, + { + "epoch": 0.31, + "learning_rate": 1.972351880874138e-05, + "loss": 1.3432, + "step": 3662 + }, + { + "epoch": 0.31, + "learning_rate": 1.9723306504739355e-05, + "loss": 1.3675, + "step": 3663 + }, + { + "epoch": 0.31, + "learning_rate": 1.9723094120400263e-05, + "loss": 1.3289, + "step": 3664 + }, + { + "epoch": 0.31, + "learning_rate": 1.9722881655725855e-05, + "loss": 1.4194, + "step": 3665 + }, + { + "epoch": 0.31, + "learning_rate": 1.972266911071788e-05, + "loss": 1.3715, + "step": 3666 + }, + { + "epoch": 0.31, + "learning_rate": 1.972245648537811e-05, + "loss": 1.3707, + "step": 3667 + }, + { + "epoch": 0.31, + "learning_rate": 1.972224377970829e-05, + "loss": 1.4347, + "step": 3668 + }, + { + "epoch": 0.31, + "learning_rate": 1.972203099371018e-05, + "loss": 1.3074, + "step": 3669 + }, + { + "epoch": 0.31, + "learning_rate": 1.972181812738554e-05, + "loss": 1.3467, + "step": 3670 + }, + { + "epoch": 0.31, + "learning_rate": 1.9721605180736124e-05, + "loss": 1.4147, + "step": 3671 + }, + { + "epoch": 0.31, + "learning_rate": 1.97213921537637e-05, + "loss": 1.3528, + "step": 3672 + }, + { + "epoch": 0.31, + "learning_rate": 1.972117904647002e-05, + "loss": 1.3785, + "step": 3673 + }, + { + "epoch": 0.31, + "learning_rate": 1.9720965858856847e-05, + "loss": 1.3847, + "step": 3674 + }, + { + "epoch": 0.31, + "learning_rate": 1.972075259092594e-05, + "loss": 1.3271, + "step": 3675 + }, + { + "epoch": 0.31, + "learning_rate": 1.9720539242679068e-05, + "loss": 1.4016, + "step": 3676 + }, + { + "epoch": 0.31, + "learning_rate": 1.972032581411799e-05, + "loss": 1.4269, + "step": 3677 + }, + { + "epoch": 0.31, + "learning_rate": 1.9720112305244464e-05, + "loss": 1.4539, + "step": 3678 + }, + { + "epoch": 0.31, + "learning_rate": 1.9719898716060266e-05, + "loss": 1.4057, + "step": 3679 + }, + { + "epoch": 0.31, + "learning_rate": 1.9719685046567153e-05, + "loss": 1.3241, + "step": 3680 + }, + { + "epoch": 0.31, + "learning_rate": 1.9719471296766887e-05, + "loss": 1.3275, + "step": 3681 + }, + { + "epoch": 0.31, + "learning_rate": 1.9719257466661244e-05, + "loss": 1.4, + "step": 3682 + }, + { + "epoch": 0.31, + "learning_rate": 1.971904355625198e-05, + "loss": 1.3163, + "step": 3683 + }, + { + "epoch": 0.31, + "learning_rate": 1.971882956554087e-05, + "loss": 1.3745, + "step": 3684 + }, + { + "epoch": 0.31, + "learning_rate": 1.9718615494529682e-05, + "loss": 1.3219, + "step": 3685 + }, + { + "epoch": 0.31, + "learning_rate": 1.971840134322018e-05, + "loss": 1.4019, + "step": 3686 + }, + { + "epoch": 0.31, + "learning_rate": 1.9718187111614136e-05, + "loss": 1.3479, + "step": 3687 + }, + { + "epoch": 0.31, + "learning_rate": 1.971797279971332e-05, + "loss": 1.3917, + "step": 3688 + }, + { + "epoch": 0.31, + "learning_rate": 1.9717758407519502e-05, + "loss": 1.3751, + "step": 3689 + }, + { + "epoch": 0.31, + "learning_rate": 1.9717543935034456e-05, + "loss": 1.3069, + "step": 3690 + }, + { + "epoch": 0.31, + "learning_rate": 1.971732938225995e-05, + "loss": 1.3347, + "step": 3691 + }, + { + "epoch": 0.31, + "learning_rate": 1.971711474919776e-05, + "loss": 1.3745, + "step": 3692 + }, + { + "epoch": 0.31, + "learning_rate": 1.9716900035849657e-05, + "loss": 1.3729, + "step": 3693 + }, + { + "epoch": 0.31, + "learning_rate": 1.9716685242217416e-05, + "loss": 1.391, + "step": 3694 + }, + { + "epoch": 0.31, + "learning_rate": 1.9716470368302815e-05, + "loss": 1.2845, + "step": 3695 + }, + { + "epoch": 0.31, + "learning_rate": 1.971625541410762e-05, + "loss": 1.3252, + "step": 3696 + }, + { + "epoch": 0.31, + "learning_rate": 1.971604037963362e-05, + "loss": 1.3264, + "step": 3697 + }, + { + "epoch": 0.31, + "learning_rate": 1.9715825264882578e-05, + "loss": 1.4449, + "step": 3698 + }, + { + "epoch": 0.31, + "learning_rate": 1.9715610069856283e-05, + "loss": 1.3744, + "step": 3699 + }, + { + "epoch": 0.31, + "learning_rate": 1.971539479455651e-05, + "loss": 1.2954, + "step": 3700 + }, + { + "epoch": 0.31, + "learning_rate": 1.971517943898503e-05, + "loss": 1.3908, + "step": 3701 + }, + { + "epoch": 0.31, + "learning_rate": 1.9714964003143632e-05, + "loss": 1.2793, + "step": 3702 + }, + { + "epoch": 0.31, + "learning_rate": 1.971474848703409e-05, + "loss": 1.4071, + "step": 3703 + }, + { + "epoch": 0.31, + "learning_rate": 1.971453289065819e-05, + "loss": 1.3795, + "step": 3704 + }, + { + "epoch": 0.31, + "learning_rate": 1.9714317214017706e-05, + "loss": 1.3472, + "step": 3705 + }, + { + "epoch": 0.31, + "learning_rate": 1.9714101457114426e-05, + "loss": 1.3027, + "step": 3706 + }, + { + "epoch": 0.31, + "learning_rate": 1.9713885619950133e-05, + "loss": 1.2891, + "step": 3707 + }, + { + "epoch": 0.31, + "learning_rate": 1.9713669702526605e-05, + "loss": 1.3617, + "step": 3708 + }, + { + "epoch": 0.31, + "learning_rate": 1.971345370484563e-05, + "loss": 1.345, + "step": 3709 + }, + { + "epoch": 0.31, + "learning_rate": 1.9713237626908994e-05, + "loss": 1.338, + "step": 3710 + }, + { + "epoch": 0.31, + "learning_rate": 1.9713021468718478e-05, + "loss": 1.3851, + "step": 3711 + }, + { + "epoch": 0.31, + "learning_rate": 1.971280523027587e-05, + "loss": 1.4107, + "step": 3712 + }, + { + "epoch": 0.31, + "learning_rate": 1.971258891158296e-05, + "loss": 1.2342, + "step": 3713 + }, + { + "epoch": 0.31, + "learning_rate": 1.9712372512641526e-05, + "loss": 1.2818, + "step": 3714 + }, + { + "epoch": 0.31, + "learning_rate": 1.9712156033453367e-05, + "loss": 1.3115, + "step": 3715 + }, + { + "epoch": 0.31, + "learning_rate": 1.9711939474020264e-05, + "loss": 1.237, + "step": 3716 + }, + { + "epoch": 0.31, + "learning_rate": 1.971172283434401e-05, + "loss": 1.2737, + "step": 3717 + }, + { + "epoch": 0.31, + "learning_rate": 1.9711506114426395e-05, + "loss": 1.3363, + "step": 3718 + }, + { + "epoch": 0.31, + "learning_rate": 1.9711289314269205e-05, + "loss": 1.3382, + "step": 3719 + }, + { + "epoch": 0.31, + "learning_rate": 1.9711072433874235e-05, + "loss": 1.426, + "step": 3720 + }, + { + "epoch": 0.31, + "learning_rate": 1.971085547324328e-05, + "loss": 1.3267, + "step": 3721 + }, + { + "epoch": 0.31, + "learning_rate": 1.9710638432378127e-05, + "loss": 1.3352, + "step": 3722 + }, + { + "epoch": 0.31, + "learning_rate": 1.971042131128057e-05, + "loss": 1.3416, + "step": 3723 + }, + { + "epoch": 0.31, + "learning_rate": 1.971020410995241e-05, + "loss": 1.3577, + "step": 3724 + }, + { + "epoch": 0.31, + "learning_rate": 1.970998682839543e-05, + "loss": 1.3748, + "step": 3725 + }, + { + "epoch": 0.31, + "learning_rate": 1.9709769466611434e-05, + "loss": 1.3397, + "step": 3726 + }, + { + "epoch": 0.31, + "learning_rate": 1.9709552024602215e-05, + "loss": 1.3955, + "step": 3727 + }, + { + "epoch": 0.31, + "learning_rate": 1.9709334502369567e-05, + "loss": 1.4168, + "step": 3728 + }, + { + "epoch": 0.31, + "learning_rate": 1.9709116899915296e-05, + "loss": 1.3499, + "step": 3729 + }, + { + "epoch": 0.31, + "learning_rate": 1.9708899217241187e-05, + "loss": 1.338, + "step": 3730 + }, + { + "epoch": 0.31, + "learning_rate": 1.970868145434905e-05, + "loss": 1.3616, + "step": 3731 + }, + { + "epoch": 0.31, + "learning_rate": 1.9708463611240677e-05, + "loss": 1.2862, + "step": 3732 + }, + { + "epoch": 0.31, + "learning_rate": 1.970824568791787e-05, + "loss": 1.3426, + "step": 3733 + }, + { + "epoch": 0.31, + "learning_rate": 1.9708027684382434e-05, + "loss": 1.3498, + "step": 3734 + }, + { + "epoch": 0.31, + "learning_rate": 1.9707809600636165e-05, + "loss": 1.2999, + "step": 3735 + }, + { + "epoch": 0.31, + "learning_rate": 1.9707591436680863e-05, + "loss": 1.3555, + "step": 3736 + }, + { + "epoch": 0.31, + "learning_rate": 1.9707373192518337e-05, + "loss": 1.2942, + "step": 3737 + }, + { + "epoch": 0.31, + "learning_rate": 1.9707154868150384e-05, + "loss": 1.3653, + "step": 3738 + }, + { + "epoch": 0.31, + "learning_rate": 1.9706936463578816e-05, + "loss": 1.3492, + "step": 3739 + }, + { + "epoch": 0.31, + "learning_rate": 1.9706717978805428e-05, + "loss": 1.4016, + "step": 3740 + }, + { + "epoch": 0.31, + "learning_rate": 1.970649941383203e-05, + "loss": 1.2921, + "step": 3741 + }, + { + "epoch": 0.32, + "learning_rate": 1.9706280768660428e-05, + "loss": 1.4488, + "step": 3742 + }, + { + "epoch": 0.32, + "learning_rate": 1.9706062043292427e-05, + "loss": 1.4031, + "step": 3743 + }, + { + "epoch": 0.32, + "learning_rate": 1.9705843237729837e-05, + "loss": 1.3239, + "step": 3744 + }, + { + "epoch": 0.32, + "learning_rate": 1.970562435197446e-05, + "loss": 1.3555, + "step": 3745 + }, + { + "epoch": 0.32, + "learning_rate": 1.9705405386028113e-05, + "loss": 1.4053, + "step": 3746 + }, + { + "epoch": 0.32, + "learning_rate": 1.9705186339892598e-05, + "loss": 1.281, + "step": 3747 + }, + { + "epoch": 0.32, + "learning_rate": 1.9704967213569727e-05, + "loss": 1.3052, + "step": 3748 + }, + { + "epoch": 0.32, + "learning_rate": 1.9704748007061313e-05, + "loss": 1.3121, + "step": 3749 + }, + { + "epoch": 0.32, + "learning_rate": 1.9704528720369166e-05, + "loss": 1.3586, + "step": 3750 + }, + { + "epoch": 0.32, + "learning_rate": 1.9704309353495094e-05, + "loss": 1.4063, + "step": 3751 + }, + { + "epoch": 0.32, + "learning_rate": 1.9704089906440914e-05, + "loss": 1.3966, + "step": 3752 + }, + { + "epoch": 0.32, + "learning_rate": 1.9703870379208435e-05, + "loss": 1.3452, + "step": 3753 + }, + { + "epoch": 0.32, + "learning_rate": 1.9703650771799476e-05, + "loss": 1.2994, + "step": 3754 + }, + { + "epoch": 0.32, + "learning_rate": 1.9703431084215847e-05, + "loss": 1.4239, + "step": 3755 + }, + { + "epoch": 0.32, + "learning_rate": 1.9703211316459368e-05, + "loss": 1.2415, + "step": 3756 + }, + { + "epoch": 0.32, + "learning_rate": 1.9702991468531846e-05, + "loss": 1.3545, + "step": 3757 + }, + { + "epoch": 0.32, + "learning_rate": 1.970277154043511e-05, + "loss": 1.3634, + "step": 3758 + }, + { + "epoch": 0.32, + "learning_rate": 1.9702551532170963e-05, + "loss": 1.2903, + "step": 3759 + }, + { + "epoch": 0.32, + "learning_rate": 1.9702331443741236e-05, + "loss": 1.4029, + "step": 3760 + }, + { + "epoch": 0.32, + "learning_rate": 1.970211127514774e-05, + "loss": 1.3506, + "step": 3761 + }, + { + "epoch": 0.32, + "learning_rate": 1.9701891026392294e-05, + "loss": 1.2662, + "step": 3762 + }, + { + "epoch": 0.32, + "learning_rate": 1.970167069747672e-05, + "loss": 1.2703, + "step": 3763 + }, + { + "epoch": 0.32, + "learning_rate": 1.9701450288402837e-05, + "loss": 1.3797, + "step": 3764 + }, + { + "epoch": 0.32, + "learning_rate": 1.970122979917247e-05, + "loss": 1.4101, + "step": 3765 + }, + { + "epoch": 0.32, + "learning_rate": 1.9701009229787433e-05, + "loss": 1.4004, + "step": 3766 + }, + { + "epoch": 0.32, + "learning_rate": 1.9700788580249556e-05, + "loss": 1.3585, + "step": 3767 + }, + { + "epoch": 0.32, + "learning_rate": 1.970056785056066e-05, + "loss": 1.3991, + "step": 3768 + }, + { + "epoch": 0.32, + "learning_rate": 1.9700347040722566e-05, + "loss": 1.3166, + "step": 3769 + }, + { + "epoch": 0.32, + "learning_rate": 1.97001261507371e-05, + "loss": 1.287, + "step": 3770 + }, + { + "epoch": 0.32, + "learning_rate": 1.9699905180606087e-05, + "loss": 1.3648, + "step": 3771 + }, + { + "epoch": 0.32, + "learning_rate": 1.9699684130331357e-05, + "loss": 1.3647, + "step": 3772 + }, + { + "epoch": 0.32, + "learning_rate": 1.9699462999914726e-05, + "loss": 1.322, + "step": 3773 + }, + { + "epoch": 0.32, + "learning_rate": 1.969924178935803e-05, + "loss": 1.4015, + "step": 3774 + }, + { + "epoch": 0.32, + "learning_rate": 1.9699020498663096e-05, + "loss": 1.3156, + "step": 3775 + }, + { + "epoch": 0.32, + "learning_rate": 1.9698799127831752e-05, + "loss": 1.345, + "step": 3776 + }, + { + "epoch": 0.32, + "learning_rate": 1.9698577676865822e-05, + "loss": 1.4168, + "step": 3777 + }, + { + "epoch": 0.32, + "learning_rate": 1.969835614576714e-05, + "loss": 1.3788, + "step": 3778 + }, + { + "epoch": 0.32, + "learning_rate": 1.9698134534537536e-05, + "loss": 1.356, + "step": 3779 + }, + { + "epoch": 0.32, + "learning_rate": 1.969791284317884e-05, + "loss": 1.3221, + "step": 3780 + }, + { + "epoch": 0.32, + "learning_rate": 1.9697691071692887e-05, + "loss": 1.2442, + "step": 3781 + }, + { + "epoch": 0.32, + "learning_rate": 1.9697469220081502e-05, + "loss": 1.3276, + "step": 3782 + }, + { + "epoch": 0.32, + "learning_rate": 1.9697247288346526e-05, + "loss": 1.3539, + "step": 3783 + }, + { + "epoch": 0.32, + "learning_rate": 1.969702527648979e-05, + "loss": 1.3721, + "step": 3784 + }, + { + "epoch": 0.32, + "learning_rate": 1.9696803184513126e-05, + "loss": 1.292, + "step": 3785 + }, + { + "epoch": 0.32, + "learning_rate": 1.9696581012418372e-05, + "loss": 1.3664, + "step": 3786 + }, + { + "epoch": 0.32, + "learning_rate": 1.9696358760207362e-05, + "loss": 1.3398, + "step": 3787 + }, + { + "epoch": 0.32, + "learning_rate": 1.969613642788193e-05, + "loss": 1.3203, + "step": 3788 + }, + { + "epoch": 0.32, + "learning_rate": 1.969591401544392e-05, + "loss": 1.3506, + "step": 3789 + }, + { + "epoch": 0.32, + "learning_rate": 1.969569152289516e-05, + "loss": 1.4454, + "step": 3790 + }, + { + "epoch": 0.32, + "learning_rate": 1.96954689502375e-05, + "loss": 1.3918, + "step": 3791 + }, + { + "epoch": 0.32, + "learning_rate": 1.969524629747277e-05, + "loss": 1.3273, + "step": 3792 + }, + { + "epoch": 0.32, + "learning_rate": 1.969502356460281e-05, + "loss": 1.3512, + "step": 3793 + }, + { + "epoch": 0.32, + "learning_rate": 1.9694800751629465e-05, + "loss": 1.3322, + "step": 3794 + }, + { + "epoch": 0.32, + "learning_rate": 1.9694577858554574e-05, + "loss": 1.317, + "step": 3795 + }, + { + "epoch": 0.32, + "learning_rate": 1.9694354885379975e-05, + "loss": 1.3302, + "step": 3796 + }, + { + "epoch": 0.32, + "learning_rate": 1.969413183210752e-05, + "loss": 1.3471, + "step": 3797 + }, + { + "epoch": 0.32, + "learning_rate": 1.9693908698739037e-05, + "loss": 1.3285, + "step": 3798 + }, + { + "epoch": 0.32, + "learning_rate": 1.969368548527638e-05, + "loss": 1.363, + "step": 3799 + }, + { + "epoch": 0.32, + "learning_rate": 1.9693462191721394e-05, + "loss": 1.3261, + "step": 3800 + }, + { + "epoch": 0.32, + "learning_rate": 1.969323881807592e-05, + "loss": 1.392, + "step": 3801 + }, + { + "epoch": 0.32, + "learning_rate": 1.9693015364341804e-05, + "loss": 1.398, + "step": 3802 + }, + { + "epoch": 0.32, + "learning_rate": 1.9692791830520895e-05, + "loss": 1.3391, + "step": 3803 + }, + { + "epoch": 0.32, + "learning_rate": 1.9692568216615033e-05, + "loss": 1.3797, + "step": 3804 + }, + { + "epoch": 0.32, + "learning_rate": 1.9692344522626072e-05, + "loss": 1.4108, + "step": 3805 + }, + { + "epoch": 0.32, + "learning_rate": 1.9692120748555862e-05, + "loss": 1.3692, + "step": 3806 + }, + { + "epoch": 0.32, + "learning_rate": 1.9691896894406246e-05, + "loss": 1.2298, + "step": 3807 + }, + { + "epoch": 0.32, + "learning_rate": 1.9691672960179076e-05, + "loss": 1.3535, + "step": 3808 + }, + { + "epoch": 0.32, + "learning_rate": 1.9691448945876204e-05, + "loss": 1.3262, + "step": 3809 + }, + { + "epoch": 0.32, + "learning_rate": 1.969122485149948e-05, + "loss": 1.3508, + "step": 3810 + }, + { + "epoch": 0.32, + "learning_rate": 1.969100067705075e-05, + "loss": 1.3424, + "step": 3811 + }, + { + "epoch": 0.32, + "learning_rate": 1.969077642253187e-05, + "loss": 1.3638, + "step": 3812 + }, + { + "epoch": 0.32, + "learning_rate": 1.96905520879447e-05, + "loss": 1.3578, + "step": 3813 + }, + { + "epoch": 0.32, + "learning_rate": 1.9690327673291085e-05, + "loss": 1.3145, + "step": 3814 + }, + { + "epoch": 0.32, + "learning_rate": 1.969010317857288e-05, + "loss": 1.3599, + "step": 3815 + }, + { + "epoch": 0.32, + "learning_rate": 1.968987860379194e-05, + "loss": 1.412, + "step": 3816 + }, + { + "epoch": 0.32, + "learning_rate": 1.9689653948950127e-05, + "loss": 1.4344, + "step": 3817 + }, + { + "epoch": 0.32, + "learning_rate": 1.968942921404929e-05, + "loss": 1.4035, + "step": 3818 + }, + { + "epoch": 0.32, + "learning_rate": 1.9689204399091285e-05, + "loss": 1.3666, + "step": 3819 + }, + { + "epoch": 0.32, + "learning_rate": 1.9688979504077977e-05, + "loss": 1.4343, + "step": 3820 + }, + { + "epoch": 0.32, + "learning_rate": 1.9688754529011213e-05, + "loss": 1.4224, + "step": 3821 + }, + { + "epoch": 0.32, + "learning_rate": 1.9688529473892863e-05, + "loss": 1.3615, + "step": 3822 + }, + { + "epoch": 0.32, + "learning_rate": 1.9688304338724783e-05, + "loss": 1.4573, + "step": 3823 + }, + { + "epoch": 0.32, + "learning_rate": 1.9688079123508826e-05, + "loss": 1.3356, + "step": 3824 + }, + { + "epoch": 0.32, + "learning_rate": 1.9687853828246864e-05, + "loss": 1.3077, + "step": 3825 + }, + { + "epoch": 0.32, + "learning_rate": 1.9687628452940753e-05, + "loss": 1.3547, + "step": 3826 + }, + { + "epoch": 0.32, + "learning_rate": 1.968740299759235e-05, + "loss": 1.3188, + "step": 3827 + }, + { + "epoch": 0.32, + "learning_rate": 1.9687177462203532e-05, + "loss": 1.3247, + "step": 3828 + }, + { + "epoch": 0.32, + "learning_rate": 1.9686951846776146e-05, + "loss": 1.347, + "step": 3829 + }, + { + "epoch": 0.32, + "learning_rate": 1.9686726151312065e-05, + "loss": 1.186, + "step": 3830 + }, + { + "epoch": 0.32, + "learning_rate": 1.9686500375813152e-05, + "loss": 1.3653, + "step": 3831 + }, + { + "epoch": 0.32, + "learning_rate": 1.9686274520281277e-05, + "loss": 1.3511, + "step": 3832 + }, + { + "epoch": 0.32, + "learning_rate": 1.9686048584718298e-05, + "loss": 1.3464, + "step": 3833 + }, + { + "epoch": 0.32, + "learning_rate": 1.9685822569126086e-05, + "loss": 1.3352, + "step": 3834 + }, + { + "epoch": 0.32, + "learning_rate": 1.968559647350651e-05, + "loss": 1.3067, + "step": 3835 + }, + { + "epoch": 0.32, + "learning_rate": 1.9685370297861434e-05, + "loss": 1.3763, + "step": 3836 + }, + { + "epoch": 0.32, + "learning_rate": 1.968514404219273e-05, + "loss": 1.2934, + "step": 3837 + }, + { + "epoch": 0.32, + "learning_rate": 1.968491770650227e-05, + "loss": 1.3761, + "step": 3838 + }, + { + "epoch": 0.32, + "learning_rate": 1.9684691290791913e-05, + "loss": 1.3652, + "step": 3839 + }, + { + "epoch": 0.32, + "learning_rate": 1.968446479506354e-05, + "loss": 1.2892, + "step": 3840 + }, + { + "epoch": 0.32, + "learning_rate": 1.968423821931902e-05, + "loss": 1.3619, + "step": 3841 + }, + { + "epoch": 0.32, + "learning_rate": 1.9684011563560226e-05, + "loss": 1.3726, + "step": 3842 + }, + { + "epoch": 0.32, + "learning_rate": 1.9683784827789025e-05, + "loss": 1.3228, + "step": 3843 + }, + { + "epoch": 0.32, + "learning_rate": 1.96835580120073e-05, + "loss": 1.2791, + "step": 3844 + }, + { + "epoch": 0.32, + "learning_rate": 1.9683331116216915e-05, + "loss": 1.3077, + "step": 3845 + }, + { + "epoch": 0.32, + "learning_rate": 1.9683104140419752e-05, + "loss": 1.3279, + "step": 3846 + }, + { + "epoch": 0.32, + "learning_rate": 1.968287708461768e-05, + "loss": 1.3546, + "step": 3847 + }, + { + "epoch": 0.32, + "learning_rate": 1.9682649948812583e-05, + "loss": 1.3436, + "step": 3848 + }, + { + "epoch": 0.32, + "learning_rate": 1.968242273300633e-05, + "loss": 1.3628, + "step": 3849 + }, + { + "epoch": 0.32, + "learning_rate": 1.9682195437200805e-05, + "loss": 1.2832, + "step": 3850 + }, + { + "epoch": 0.32, + "learning_rate": 1.968196806139788e-05, + "loss": 1.3552, + "step": 3851 + }, + { + "epoch": 0.32, + "learning_rate": 1.9681740605599437e-05, + "loss": 1.4038, + "step": 3852 + }, + { + "epoch": 0.32, + "learning_rate": 1.9681513069807353e-05, + "loss": 1.3922, + "step": 3853 + }, + { + "epoch": 0.32, + "learning_rate": 1.968128545402351e-05, + "loss": 1.26, + "step": 3854 + }, + { + "epoch": 0.32, + "learning_rate": 1.9681057758249792e-05, + "loss": 1.2589, + "step": 3855 + }, + { + "epoch": 0.32, + "learning_rate": 1.9680829982488073e-05, + "loss": 1.3783, + "step": 3856 + }, + { + "epoch": 0.32, + "learning_rate": 1.9680602126740237e-05, + "loss": 1.3277, + "step": 3857 + }, + { + "epoch": 0.32, + "learning_rate": 1.968037419100817e-05, + "loss": 1.2962, + "step": 3858 + }, + { + "epoch": 0.32, + "learning_rate": 1.9680146175293754e-05, + "loss": 1.3503, + "step": 3859 + }, + { + "epoch": 0.32, + "learning_rate": 1.967991807959887e-05, + "loss": 1.3933, + "step": 3860 + }, + { + "epoch": 0.33, + "learning_rate": 1.967968990392541e-05, + "loss": 1.3866, + "step": 3861 + }, + { + "epoch": 0.33, + "learning_rate": 1.967946164827525e-05, + "loss": 1.3039, + "step": 3862 + }, + { + "epoch": 0.33, + "learning_rate": 1.9679233312650283e-05, + "loss": 1.3059, + "step": 3863 + }, + { + "epoch": 0.33, + "learning_rate": 1.967900489705239e-05, + "loss": 1.3058, + "step": 3864 + }, + { + "epoch": 0.33, + "learning_rate": 1.967877640148346e-05, + "loss": 1.446, + "step": 3865 + }, + { + "epoch": 0.33, + "learning_rate": 1.967854782594538e-05, + "loss": 1.2786, + "step": 3866 + }, + { + "epoch": 0.33, + "learning_rate": 1.9678319170440047e-05, + "loss": 1.3168, + "step": 3867 + }, + { + "epoch": 0.33, + "learning_rate": 1.9678090434969338e-05, + "loss": 1.3585, + "step": 3868 + }, + { + "epoch": 0.33, + "learning_rate": 1.967786161953515e-05, + "loss": 1.375, + "step": 3869 + }, + { + "epoch": 0.33, + "learning_rate": 1.9677632724139373e-05, + "loss": 1.2946, + "step": 3870 + }, + { + "epoch": 0.33, + "learning_rate": 1.9677403748783895e-05, + "loss": 1.3221, + "step": 3871 + }, + { + "epoch": 0.33, + "learning_rate": 1.9677174693470615e-05, + "loss": 1.3351, + "step": 3872 + }, + { + "epoch": 0.33, + "learning_rate": 1.9676945558201415e-05, + "loss": 1.4177, + "step": 3873 + }, + { + "epoch": 0.33, + "learning_rate": 1.9676716342978194e-05, + "loss": 1.2882, + "step": 3874 + }, + { + "epoch": 0.33, + "learning_rate": 1.967648704780285e-05, + "loss": 1.3448, + "step": 3875 + }, + { + "epoch": 0.33, + "learning_rate": 1.967625767267727e-05, + "loss": 1.3262, + "step": 3876 + }, + { + "epoch": 0.33, + "learning_rate": 1.9676028217603353e-05, + "loss": 1.3654, + "step": 3877 + }, + { + "epoch": 0.33, + "learning_rate": 1.9675798682582995e-05, + "loss": 1.3062, + "step": 3878 + }, + { + "epoch": 0.33, + "learning_rate": 1.9675569067618088e-05, + "loss": 1.4404, + "step": 3879 + }, + { + "epoch": 0.33, + "learning_rate": 1.9675339372710537e-05, + "loss": 1.2875, + "step": 3880 + }, + { + "epoch": 0.33, + "learning_rate": 1.9675109597862233e-05, + "loss": 1.3649, + "step": 3881 + }, + { + "epoch": 0.33, + "learning_rate": 1.967487974307508e-05, + "loss": 1.3696, + "step": 3882 + }, + { + "epoch": 0.33, + "learning_rate": 1.967464980835097e-05, + "loss": 1.3223, + "step": 3883 + }, + { + "epoch": 0.33, + "learning_rate": 1.967441979369181e-05, + "loss": 1.4053, + "step": 3884 + }, + { + "epoch": 0.33, + "learning_rate": 1.96741896990995e-05, + "loss": 1.4513, + "step": 3885 + }, + { + "epoch": 0.33, + "learning_rate": 1.9673959524575933e-05, + "loss": 1.3508, + "step": 3886 + }, + { + "epoch": 0.33, + "learning_rate": 1.967372927012302e-05, + "loss": 1.2932, + "step": 3887 + }, + { + "epoch": 0.33, + "learning_rate": 1.9673498935742655e-05, + "loss": 1.2982, + "step": 3888 + }, + { + "epoch": 0.33, + "learning_rate": 1.967326852143675e-05, + "loss": 1.3096, + "step": 3889 + }, + { + "epoch": 0.33, + "learning_rate": 1.9673038027207207e-05, + "loss": 1.3939, + "step": 3890 + }, + { + "epoch": 0.33, + "learning_rate": 1.9672807453055923e-05, + "loss": 1.3695, + "step": 3891 + }, + { + "epoch": 0.33, + "learning_rate": 1.967257679898481e-05, + "loss": 1.3877, + "step": 3892 + }, + { + "epoch": 0.33, + "learning_rate": 1.9672346064995773e-05, + "loss": 1.3714, + "step": 3893 + }, + { + "epoch": 0.33, + "learning_rate": 1.967211525109072e-05, + "loss": 1.396, + "step": 3894 + }, + { + "epoch": 0.33, + "learning_rate": 1.9671884357271548e-05, + "loss": 1.3865, + "step": 3895 + }, + { + "epoch": 0.33, + "learning_rate": 1.9671653383540176e-05, + "loss": 1.3813, + "step": 3896 + }, + { + "epoch": 0.33, + "learning_rate": 1.9671422329898506e-05, + "loss": 1.3758, + "step": 3897 + }, + { + "epoch": 0.33, + "learning_rate": 1.9671191196348452e-05, + "loss": 1.3478, + "step": 3898 + }, + { + "epoch": 0.33, + "learning_rate": 1.9670959982891922e-05, + "loss": 1.3443, + "step": 3899 + }, + { + "epoch": 0.33, + "learning_rate": 1.9670728689530824e-05, + "loss": 1.3442, + "step": 3900 + }, + { + "epoch": 0.33, + "learning_rate": 1.9670497316267072e-05, + "loss": 1.3111, + "step": 3901 + }, + { + "epoch": 0.33, + "learning_rate": 1.9670265863102573e-05, + "loss": 1.323, + "step": 3902 + }, + { + "epoch": 0.33, + "learning_rate": 1.9670034330039245e-05, + "loss": 1.3451, + "step": 3903 + }, + { + "epoch": 0.33, + "learning_rate": 1.9669802717079e-05, + "loss": 1.4115, + "step": 3904 + }, + { + "epoch": 0.33, + "learning_rate": 1.9669571024223746e-05, + "loss": 1.294, + "step": 3905 + }, + { + "epoch": 0.33, + "learning_rate": 1.9669339251475407e-05, + "loss": 1.4601, + "step": 3906 + }, + { + "epoch": 0.33, + "learning_rate": 1.966910739883589e-05, + "loss": 1.4306, + "step": 3907 + }, + { + "epoch": 0.33, + "learning_rate": 1.966887546630711e-05, + "loss": 1.3302, + "step": 3908 + }, + { + "epoch": 0.33, + "learning_rate": 1.966864345389099e-05, + "loss": 1.3256, + "step": 3909 + }, + { + "epoch": 0.33, + "learning_rate": 1.966841136158944e-05, + "loss": 1.398, + "step": 3910 + }, + { + "epoch": 0.33, + "learning_rate": 1.9668179189404385e-05, + "loss": 1.331, + "step": 3911 + }, + { + "epoch": 0.33, + "learning_rate": 1.966794693733774e-05, + "loss": 1.4252, + "step": 3912 + }, + { + "epoch": 0.33, + "learning_rate": 1.9667714605391418e-05, + "loss": 1.3158, + "step": 3913 + }, + { + "epoch": 0.33, + "learning_rate": 1.9667482193567345e-05, + "loss": 1.2582, + "step": 3914 + }, + { + "epoch": 0.33, + "learning_rate": 1.9667249701867446e-05, + "loss": 1.2807, + "step": 3915 + }, + { + "epoch": 0.33, + "learning_rate": 1.966701713029363e-05, + "loss": 1.436, + "step": 3916 + }, + { + "epoch": 0.33, + "learning_rate": 1.9666784478847826e-05, + "loss": 1.3404, + "step": 3917 + }, + { + "epoch": 0.33, + "learning_rate": 1.9666551747531958e-05, + "loss": 1.3578, + "step": 3918 + }, + { + "epoch": 0.33, + "learning_rate": 1.9666318936347944e-05, + "loss": 1.2518, + "step": 3919 + }, + { + "epoch": 0.33, + "learning_rate": 1.966608604529771e-05, + "loss": 1.2581, + "step": 3920 + }, + { + "epoch": 0.33, + "learning_rate": 1.9665853074383173e-05, + "loss": 1.3314, + "step": 3921 + }, + { + "epoch": 0.33, + "learning_rate": 1.966562002360627e-05, + "loss": 1.2847, + "step": 3922 + }, + { + "epoch": 0.33, + "learning_rate": 1.966538689296892e-05, + "loss": 1.2433, + "step": 3923 + }, + { + "epoch": 0.33, + "learning_rate": 1.9665153682473054e-05, + "loss": 1.4065, + "step": 3924 + }, + { + "epoch": 0.33, + "learning_rate": 1.966492039212059e-05, + "loss": 1.2801, + "step": 3925 + }, + { + "epoch": 0.33, + "learning_rate": 1.9664687021913463e-05, + "loss": 1.3264, + "step": 3926 + }, + { + "epoch": 0.33, + "learning_rate": 1.96644535718536e-05, + "loss": 1.3557, + "step": 3927 + }, + { + "epoch": 0.33, + "learning_rate": 1.9664220041942925e-05, + "loss": 1.422, + "step": 3928 + }, + { + "epoch": 0.33, + "learning_rate": 1.9663986432183374e-05, + "loss": 1.2828, + "step": 3929 + }, + { + "epoch": 0.33, + "learning_rate": 1.9663752742576873e-05, + "loss": 1.4025, + "step": 3930 + }, + { + "epoch": 0.33, + "learning_rate": 1.9663518973125353e-05, + "loss": 1.3668, + "step": 3931 + }, + { + "epoch": 0.33, + "learning_rate": 1.9663285123830746e-05, + "loss": 1.2878, + "step": 3932 + }, + { + "epoch": 0.33, + "learning_rate": 1.9663051194694986e-05, + "loss": 1.3205, + "step": 3933 + }, + { + "epoch": 0.33, + "learning_rate": 1.9662817185720008e-05, + "loss": 1.3758, + "step": 3934 + }, + { + "epoch": 0.33, + "learning_rate": 1.9662583096907736e-05, + "loss": 1.2588, + "step": 3935 + }, + { + "epoch": 0.33, + "learning_rate": 1.9662348928260115e-05, + "loss": 1.3107, + "step": 3936 + }, + { + "epoch": 0.33, + "learning_rate": 1.9662114679779073e-05, + "loss": 1.3575, + "step": 3937 + }, + { + "epoch": 0.33, + "learning_rate": 1.9661880351466544e-05, + "loss": 1.4338, + "step": 3938 + }, + { + "epoch": 0.33, + "learning_rate": 1.966164594332447e-05, + "loss": 1.2775, + "step": 3939 + }, + { + "epoch": 0.33, + "learning_rate": 1.9661411455354783e-05, + "loss": 1.3054, + "step": 3940 + }, + { + "epoch": 0.33, + "learning_rate": 1.9661176887559427e-05, + "loss": 1.4221, + "step": 3941 + }, + { + "epoch": 0.33, + "learning_rate": 1.9660942239940332e-05, + "loss": 1.3129, + "step": 3942 + }, + { + "epoch": 0.33, + "learning_rate": 1.9660707512499445e-05, + "loss": 1.4433, + "step": 3943 + }, + { + "epoch": 0.33, + "learning_rate": 1.9660472705238693e-05, + "loss": 1.2905, + "step": 3944 + }, + { + "epoch": 0.33, + "learning_rate": 1.966023781816003e-05, + "loss": 1.2953, + "step": 3945 + }, + { + "epoch": 0.33, + "learning_rate": 1.9660002851265388e-05, + "loss": 1.4209, + "step": 3946 + }, + { + "epoch": 0.33, + "learning_rate": 1.9659767804556712e-05, + "loss": 1.392, + "step": 3947 + }, + { + "epoch": 0.33, + "learning_rate": 1.9659532678035944e-05, + "loss": 1.4032, + "step": 3948 + }, + { + "epoch": 0.33, + "learning_rate": 1.965929747170502e-05, + "loss": 1.3729, + "step": 3949 + }, + { + "epoch": 0.33, + "learning_rate": 1.9659062185565893e-05, + "loss": 1.4694, + "step": 3950 + }, + { + "epoch": 0.33, + "learning_rate": 1.96588268196205e-05, + "loss": 1.256, + "step": 3951 + }, + { + "epoch": 0.33, + "learning_rate": 1.9658591373870796e-05, + "loss": 1.3438, + "step": 3952 + }, + { + "epoch": 0.33, + "learning_rate": 1.9658355848318714e-05, + "loss": 1.3532, + "step": 3953 + }, + { + "epoch": 0.33, + "learning_rate": 1.9658120242966206e-05, + "loss": 1.3408, + "step": 3954 + }, + { + "epoch": 0.33, + "learning_rate": 1.9657884557815214e-05, + "loss": 1.3663, + "step": 3955 + }, + { + "epoch": 0.33, + "learning_rate": 1.965764879286769e-05, + "loss": 1.3653, + "step": 3956 + }, + { + "epoch": 0.33, + "learning_rate": 1.9657412948125586e-05, + "loss": 1.3186, + "step": 3957 + }, + { + "epoch": 0.33, + "learning_rate": 1.9657177023590843e-05, + "loss": 1.2493, + "step": 3958 + }, + { + "epoch": 0.33, + "learning_rate": 1.965694101926541e-05, + "loss": 1.5343, + "step": 3959 + }, + { + "epoch": 0.33, + "learning_rate": 1.9656704935151237e-05, + "loss": 1.324, + "step": 3960 + }, + { + "epoch": 0.33, + "learning_rate": 1.9656468771250285e-05, + "loss": 1.3852, + "step": 3961 + }, + { + "epoch": 0.33, + "learning_rate": 1.965623252756449e-05, + "loss": 1.3334, + "step": 3962 + }, + { + "epoch": 0.33, + "learning_rate": 1.9655996204095816e-05, + "loss": 1.3528, + "step": 3963 + }, + { + "epoch": 0.33, + "learning_rate": 1.9655759800846213e-05, + "loss": 1.2822, + "step": 3964 + }, + { + "epoch": 0.33, + "learning_rate": 1.965552331781763e-05, + "loss": 1.3948, + "step": 3965 + }, + { + "epoch": 0.33, + "learning_rate": 1.9655286755012023e-05, + "loss": 1.3654, + "step": 3966 + }, + { + "epoch": 0.33, + "learning_rate": 1.9655050112431346e-05, + "loss": 1.2961, + "step": 3967 + }, + { + "epoch": 0.33, + "learning_rate": 1.9654813390077557e-05, + "loss": 1.2562, + "step": 3968 + }, + { + "epoch": 0.33, + "learning_rate": 1.9654576587952607e-05, + "loss": 1.3483, + "step": 3969 + }, + { + "epoch": 0.33, + "learning_rate": 1.9654339706058458e-05, + "loss": 1.3063, + "step": 3970 + }, + { + "epoch": 0.33, + "learning_rate": 1.965410274439706e-05, + "loss": 1.3096, + "step": 3971 + }, + { + "epoch": 0.33, + "learning_rate": 1.9653865702970382e-05, + "loss": 1.3584, + "step": 3972 + }, + { + "epoch": 0.33, + "learning_rate": 1.965362858178037e-05, + "loss": 1.361, + "step": 3973 + }, + { + "epoch": 0.33, + "learning_rate": 1.9653391380828994e-05, + "loss": 1.3788, + "step": 3974 + }, + { + "epoch": 0.33, + "learning_rate": 1.9653154100118204e-05, + "loss": 1.2824, + "step": 3975 + }, + { + "epoch": 0.33, + "learning_rate": 1.965291673964997e-05, + "loss": 1.241, + "step": 3976 + }, + { + "epoch": 0.33, + "learning_rate": 1.9652679299426247e-05, + "loss": 1.2395, + "step": 3977 + }, + { + "epoch": 0.33, + "learning_rate": 1.9652441779449002e-05, + "loss": 1.3513, + "step": 3978 + }, + { + "epoch": 0.33, + "learning_rate": 1.9652204179720187e-05, + "loss": 1.3076, + "step": 3979 + }, + { + "epoch": 0.34, + "learning_rate": 1.9651966500241776e-05, + "loss": 1.4041, + "step": 3980 + }, + { + "epoch": 0.34, + "learning_rate": 1.9651728741015725e-05, + "loss": 1.3432, + "step": 3981 + }, + { + "epoch": 0.34, + "learning_rate": 1.965149090204401e-05, + "loss": 1.3354, + "step": 3982 + }, + { + "epoch": 0.34, + "learning_rate": 1.965125298332858e-05, + "loss": 1.2997, + "step": 3983 + }, + { + "epoch": 0.34, + "learning_rate": 1.9651014984871412e-05, + "loss": 1.2963, + "step": 3984 + }, + { + "epoch": 0.34, + "learning_rate": 1.9650776906674468e-05, + "loss": 1.4468, + "step": 3985 + }, + { + "epoch": 0.34, + "learning_rate": 1.9650538748739722e-05, + "loss": 1.2291, + "step": 3986 + }, + { + "epoch": 0.34, + "learning_rate": 1.965030051106913e-05, + "loss": 1.3906, + "step": 3987 + }, + { + "epoch": 0.34, + "learning_rate": 1.965006219366467e-05, + "loss": 1.2929, + "step": 3988 + }, + { + "epoch": 0.34, + "learning_rate": 1.9649823796528305e-05, + "loss": 1.349, + "step": 3989 + }, + { + "epoch": 0.34, + "learning_rate": 1.9649585319662013e-05, + "loss": 1.3818, + "step": 3990 + }, + { + "epoch": 0.34, + "learning_rate": 1.9649346763067755e-05, + "loss": 1.3557, + "step": 3991 + }, + { + "epoch": 0.34, + "learning_rate": 1.9649108126747506e-05, + "loss": 1.3183, + "step": 3992 + }, + { + "epoch": 0.34, + "learning_rate": 1.9648869410703236e-05, + "loss": 1.3106, + "step": 3993 + }, + { + "epoch": 0.34, + "learning_rate": 1.964863061493692e-05, + "loss": 1.3446, + "step": 3994 + }, + { + "epoch": 0.34, + "learning_rate": 1.9648391739450532e-05, + "loss": 1.4423, + "step": 3995 + }, + { + "epoch": 0.34, + "learning_rate": 1.9648152784246042e-05, + "loss": 1.3339, + "step": 3996 + }, + { + "epoch": 0.34, + "learning_rate": 1.9647913749325424e-05, + "loss": 1.2374, + "step": 3997 + }, + { + "epoch": 0.34, + "learning_rate": 1.9647674634690658e-05, + "loss": 1.331, + "step": 3998 + }, + { + "epoch": 0.34, + "learning_rate": 1.9647435440343715e-05, + "loss": 1.3864, + "step": 3999 + }, + { + "epoch": 0.34, + "learning_rate": 1.9647196166286574e-05, + "loss": 1.2954, + "step": 4000 + }, + { + "epoch": 0.34, + "learning_rate": 1.964695681252121e-05, + "loss": 1.2584, + "step": 4001 + }, + { + "epoch": 0.34, + "learning_rate": 1.96467173790496e-05, + "loss": 1.2585, + "step": 4002 + }, + { + "epoch": 0.34, + "learning_rate": 1.9646477865873724e-05, + "loss": 1.3436, + "step": 4003 + }, + { + "epoch": 0.34, + "learning_rate": 1.9646238272995564e-05, + "loss": 1.4179, + "step": 4004 + }, + { + "epoch": 0.34, + "learning_rate": 1.9645998600417095e-05, + "loss": 1.274, + "step": 4005 + }, + { + "epoch": 0.34, + "learning_rate": 1.9645758848140296e-05, + "loss": 1.4196, + "step": 4006 + }, + { + "epoch": 0.34, + "learning_rate": 1.964551901616715e-05, + "loss": 1.3593, + "step": 4007 + }, + { + "epoch": 0.34, + "learning_rate": 1.9645279104499643e-05, + "loss": 1.2812, + "step": 4008 + }, + { + "epoch": 0.34, + "learning_rate": 1.964503911313975e-05, + "loss": 1.2781, + "step": 4009 + }, + { + "epoch": 0.34, + "learning_rate": 1.9644799042089456e-05, + "loss": 1.359, + "step": 4010 + }, + { + "epoch": 0.34, + "learning_rate": 1.9644558891350748e-05, + "loss": 1.3668, + "step": 4011 + }, + { + "epoch": 0.34, + "learning_rate": 1.964431866092561e-05, + "loss": 1.3311, + "step": 4012 + }, + { + "epoch": 0.34, + "learning_rate": 1.964407835081602e-05, + "loss": 1.35, + "step": 4013 + }, + { + "epoch": 0.34, + "learning_rate": 1.964383796102397e-05, + "loss": 1.2698, + "step": 4014 + }, + { + "epoch": 0.34, + "learning_rate": 1.9643597491551442e-05, + "loss": 1.3135, + "step": 4015 + }, + { + "epoch": 0.34, + "learning_rate": 1.964335694240043e-05, + "loss": 1.3276, + "step": 4016 + }, + { + "epoch": 0.34, + "learning_rate": 1.9643116313572914e-05, + "loss": 1.3561, + "step": 4017 + }, + { + "epoch": 0.34, + "learning_rate": 1.9642875605070884e-05, + "loss": 1.2963, + "step": 4018 + }, + { + "epoch": 0.34, + "learning_rate": 1.9642634816896332e-05, + "loss": 1.3646, + "step": 4019 + }, + { + "epoch": 0.34, + "learning_rate": 1.9642393949051243e-05, + "loss": 1.3231, + "step": 4020 + }, + { + "epoch": 0.34, + "learning_rate": 1.964215300153761e-05, + "loss": 1.3803, + "step": 4021 + }, + { + "epoch": 0.34, + "learning_rate": 1.964191197435742e-05, + "loss": 1.3295, + "step": 4022 + }, + { + "epoch": 0.34, + "learning_rate": 1.9641670867512672e-05, + "loss": 1.3893, + "step": 4023 + }, + { + "epoch": 0.34, + "learning_rate": 1.9641429681005352e-05, + "loss": 1.3377, + "step": 4024 + }, + { + "epoch": 0.34, + "learning_rate": 1.9641188414837457e-05, + "loss": 1.2958, + "step": 4025 + }, + { + "epoch": 0.34, + "learning_rate": 1.964094706901097e-05, + "loss": 1.3451, + "step": 4026 + }, + { + "epoch": 0.34, + "learning_rate": 1.9640705643527898e-05, + "loss": 1.2824, + "step": 4027 + }, + { + "epoch": 0.34, + "learning_rate": 1.9640464138390233e-05, + "loss": 1.3175, + "step": 4028 + }, + { + "epoch": 0.34, + "learning_rate": 1.9640222553599964e-05, + "loss": 1.4198, + "step": 4029 + }, + { + "epoch": 0.34, + "learning_rate": 1.963998088915909e-05, + "loss": 1.3318, + "step": 4030 + }, + { + "epoch": 0.34, + "learning_rate": 1.9639739145069613e-05, + "loss": 1.2829, + "step": 4031 + }, + { + "epoch": 0.34, + "learning_rate": 1.9639497321333525e-05, + "loss": 1.3473, + "step": 4032 + }, + { + "epoch": 0.34, + "learning_rate": 1.9639255417952826e-05, + "loss": 1.3973, + "step": 4033 + }, + { + "epoch": 0.34, + "learning_rate": 1.963901343492951e-05, + "loss": 1.2709, + "step": 4034 + }, + { + "epoch": 0.34, + "learning_rate": 1.9638771372265583e-05, + "loss": 1.2205, + "step": 4035 + }, + { + "epoch": 0.34, + "learning_rate": 1.9638529229963043e-05, + "loss": 1.4086, + "step": 4036 + }, + { + "epoch": 0.34, + "learning_rate": 1.9638287008023888e-05, + "loss": 1.3426, + "step": 4037 + }, + { + "epoch": 0.34, + "learning_rate": 1.9638044706450118e-05, + "loss": 1.3575, + "step": 4038 + }, + { + "epoch": 0.34, + "learning_rate": 1.963780232524374e-05, + "loss": 1.2836, + "step": 4039 + }, + { + "epoch": 0.34, + "learning_rate": 1.963755986440676e-05, + "loss": 1.345, + "step": 4040 + }, + { + "epoch": 0.34, + "learning_rate": 1.963731732394117e-05, + "loss": 1.2621, + "step": 4041 + }, + { + "epoch": 0.34, + "learning_rate": 1.963707470384898e-05, + "loss": 1.3175, + "step": 4042 + }, + { + "epoch": 0.34, + "learning_rate": 1.96368320041322e-05, + "loss": 1.3165, + "step": 4043 + }, + { + "epoch": 0.34, + "learning_rate": 1.9636589224792825e-05, + "loss": 1.3912, + "step": 4044 + }, + { + "epoch": 0.34, + "learning_rate": 1.9636346365832865e-05, + "loss": 1.4139, + "step": 4045 + }, + { + "epoch": 0.34, + "learning_rate": 1.963610342725433e-05, + "loss": 1.3886, + "step": 4046 + }, + { + "epoch": 0.34, + "learning_rate": 1.963586040905922e-05, + "loss": 1.3091, + "step": 4047 + }, + { + "epoch": 0.34, + "learning_rate": 1.9635617311249553e-05, + "loss": 1.3405, + "step": 4048 + }, + { + "epoch": 0.34, + "learning_rate": 1.963537413382733e-05, + "loss": 1.3128, + "step": 4049 + }, + { + "epoch": 0.34, + "learning_rate": 1.9635130876794564e-05, + "loss": 1.3778, + "step": 4050 + }, + { + "epoch": 0.34, + "learning_rate": 1.963488754015326e-05, + "loss": 1.2724, + "step": 4051 + }, + { + "epoch": 0.34, + "learning_rate": 1.9634644123905433e-05, + "loss": 1.3661, + "step": 4052 + }, + { + "epoch": 0.34, + "learning_rate": 1.9634400628053093e-05, + "loss": 1.2924, + "step": 4053 + }, + { + "epoch": 0.34, + "learning_rate": 1.9634157052598254e-05, + "loss": 1.2432, + "step": 4054 + }, + { + "epoch": 0.34, + "learning_rate": 1.963391339754292e-05, + "loss": 1.3552, + "step": 4055 + }, + { + "epoch": 0.34, + "learning_rate": 1.9633669662889113e-05, + "loss": 1.3677, + "step": 4056 + }, + { + "epoch": 0.34, + "learning_rate": 1.9633425848638846e-05, + "loss": 1.2826, + "step": 4057 + }, + { + "epoch": 0.34, + "learning_rate": 1.963318195479413e-05, + "loss": 1.4164, + "step": 4058 + }, + { + "epoch": 0.34, + "learning_rate": 1.9632937981356985e-05, + "loss": 1.4268, + "step": 4059 + }, + { + "epoch": 0.34, + "learning_rate": 1.9632693928329418e-05, + "loss": 1.3333, + "step": 4060 + }, + { + "epoch": 0.34, + "learning_rate": 1.9632449795713453e-05, + "loss": 1.3911, + "step": 4061 + }, + { + "epoch": 0.34, + "learning_rate": 1.9632205583511106e-05, + "loss": 1.4314, + "step": 4062 + }, + { + "epoch": 0.34, + "learning_rate": 1.9631961291724392e-05, + "loss": 1.3652, + "step": 4063 + }, + { + "epoch": 0.34, + "learning_rate": 1.9631716920355332e-05, + "loss": 1.2671, + "step": 4064 + }, + { + "epoch": 0.34, + "learning_rate": 1.9631472469405944e-05, + "loss": 1.3576, + "step": 4065 + }, + { + "epoch": 0.34, + "learning_rate": 1.963122793887825e-05, + "loss": 1.2919, + "step": 4066 + }, + { + "epoch": 0.34, + "learning_rate": 1.9630983328774265e-05, + "loss": 1.3194, + "step": 4067 + }, + { + "epoch": 0.34, + "learning_rate": 1.9630738639096017e-05, + "loss": 1.3405, + "step": 4068 + }, + { + "epoch": 0.34, + "learning_rate": 1.9630493869845525e-05, + "loss": 1.3537, + "step": 4069 + }, + { + "epoch": 0.34, + "learning_rate": 1.9630249021024806e-05, + "loss": 1.3444, + "step": 4070 + }, + { + "epoch": 0.34, + "learning_rate": 1.963000409263589e-05, + "loss": 1.341, + "step": 4071 + }, + { + "epoch": 0.34, + "learning_rate": 1.96297590846808e-05, + "loss": 1.403, + "step": 4072 + }, + { + "epoch": 0.34, + "learning_rate": 1.9629513997161553e-05, + "loss": 1.2417, + "step": 4073 + }, + { + "epoch": 0.34, + "learning_rate": 1.9629268830080184e-05, + "loss": 1.3758, + "step": 4074 + }, + { + "epoch": 0.34, + "learning_rate": 1.962902358343871e-05, + "loss": 1.3284, + "step": 4075 + }, + { + "epoch": 0.34, + "learning_rate": 1.9628778257239167e-05, + "loss": 1.3972, + "step": 4076 + }, + { + "epoch": 0.34, + "learning_rate": 1.962853285148357e-05, + "loss": 1.3611, + "step": 4077 + }, + { + "epoch": 0.34, + "learning_rate": 1.962828736617396e-05, + "loss": 1.3213, + "step": 4078 + }, + { + "epoch": 0.34, + "learning_rate": 1.9628041801312353e-05, + "loss": 1.3586, + "step": 4079 + }, + { + "epoch": 0.34, + "learning_rate": 1.9627796156900786e-05, + "loss": 1.2318, + "step": 4080 + }, + { + "epoch": 0.34, + "learning_rate": 1.9627550432941285e-05, + "loss": 1.3346, + "step": 4081 + }, + { + "epoch": 0.34, + "learning_rate": 1.962730462943588e-05, + "loss": 1.3244, + "step": 4082 + }, + { + "epoch": 0.34, + "learning_rate": 1.9627058746386608e-05, + "loss": 1.3447, + "step": 4083 + }, + { + "epoch": 0.34, + "learning_rate": 1.9626812783795493e-05, + "loss": 1.2862, + "step": 4084 + }, + { + "epoch": 0.34, + "learning_rate": 1.9626566741664566e-05, + "loss": 1.2642, + "step": 4085 + }, + { + "epoch": 0.34, + "learning_rate": 1.9626320619995866e-05, + "loss": 1.3595, + "step": 4086 + }, + { + "epoch": 0.34, + "learning_rate": 1.9626074418791426e-05, + "loss": 1.3218, + "step": 4087 + }, + { + "epoch": 0.34, + "learning_rate": 1.9625828138053276e-05, + "loss": 1.3235, + "step": 4088 + }, + { + "epoch": 0.34, + "learning_rate": 1.9625581777783456e-05, + "loss": 1.389, + "step": 4089 + }, + { + "epoch": 0.34, + "learning_rate": 1.9625335337983994e-05, + "loss": 1.2784, + "step": 4090 + }, + { + "epoch": 0.34, + "learning_rate": 1.9625088818656937e-05, + "loss": 1.2154, + "step": 4091 + }, + { + "epoch": 0.34, + "learning_rate": 1.962484221980431e-05, + "loss": 1.3754, + "step": 4092 + }, + { + "epoch": 0.34, + "learning_rate": 1.962459554142816e-05, + "loss": 1.3999, + "step": 4093 + }, + { + "epoch": 0.34, + "learning_rate": 1.9624348783530524e-05, + "loss": 1.3708, + "step": 4094 + }, + { + "epoch": 0.34, + "learning_rate": 1.9624101946113434e-05, + "loss": 1.3865, + "step": 4095 + }, + { + "epoch": 0.34, + "learning_rate": 1.9623855029178934e-05, + "loss": 1.3911, + "step": 4096 + }, + { + "epoch": 0.34, + "learning_rate": 1.9623608032729063e-05, + "loss": 1.3795, + "step": 4097 + }, + { + "epoch": 0.35, + "learning_rate": 1.9623360956765865e-05, + "loss": 1.3154, + "step": 4098 + }, + { + "epoch": 0.35, + "learning_rate": 1.9623113801291377e-05, + "loss": 1.3277, + "step": 4099 + }, + { + "epoch": 0.35, + "learning_rate": 1.9622866566307643e-05, + "loss": 1.403, + "step": 4100 + }, + { + "epoch": 0.35, + "learning_rate": 1.9622619251816705e-05, + "loss": 1.3016, + "step": 4101 + }, + { + "epoch": 0.35, + "learning_rate": 1.9622371857820607e-05, + "loss": 1.2536, + "step": 4102 + }, + { + "epoch": 0.35, + "learning_rate": 1.9622124384321398e-05, + "loss": 1.2584, + "step": 4103 + }, + { + "epoch": 0.35, + "learning_rate": 1.9621876831321113e-05, + "loss": 1.3491, + "step": 4104 + }, + { + "epoch": 0.35, + "learning_rate": 1.9621629198821802e-05, + "loss": 1.4125, + "step": 4105 + }, + { + "epoch": 0.35, + "learning_rate": 1.962138148682551e-05, + "loss": 1.3484, + "step": 4106 + }, + { + "epoch": 0.35, + "learning_rate": 1.9621133695334287e-05, + "loss": 1.3134, + "step": 4107 + }, + { + "epoch": 0.35, + "learning_rate": 1.9620885824350184e-05, + "loss": 1.3283, + "step": 4108 + }, + { + "epoch": 0.35, + "learning_rate": 1.9620637873875236e-05, + "loss": 1.2599, + "step": 4109 + }, + { + "epoch": 0.35, + "learning_rate": 1.96203898439115e-05, + "loss": 1.4192, + "step": 4110 + }, + { + "epoch": 0.35, + "learning_rate": 1.9620141734461026e-05, + "loss": 1.3089, + "step": 4111 + }, + { + "epoch": 0.35, + "learning_rate": 1.961989354552586e-05, + "loss": 1.3239, + "step": 4112 + }, + { + "epoch": 0.35, + "learning_rate": 1.961964527710806e-05, + "loss": 1.264, + "step": 4113 + }, + { + "epoch": 0.35, + "learning_rate": 1.9619396929209668e-05, + "loss": 1.3516, + "step": 4114 + }, + { + "epoch": 0.35, + "learning_rate": 1.9619148501832738e-05, + "loss": 1.3418, + "step": 4115 + }, + { + "epoch": 0.35, + "learning_rate": 1.9618899994979327e-05, + "loss": 1.3812, + "step": 4116 + }, + { + "epoch": 0.35, + "learning_rate": 1.9618651408651484e-05, + "loss": 1.4243, + "step": 4117 + }, + { + "epoch": 0.35, + "learning_rate": 1.9618402742851266e-05, + "loss": 1.3096, + "step": 4118 + }, + { + "epoch": 0.35, + "learning_rate": 1.9618153997580727e-05, + "loss": 1.2766, + "step": 4119 + }, + { + "epoch": 0.35, + "learning_rate": 1.9617905172841922e-05, + "loss": 1.3008, + "step": 4120 + }, + { + "epoch": 0.35, + "learning_rate": 1.9617656268636906e-05, + "loss": 1.2914, + "step": 4121 + }, + { + "epoch": 0.35, + "learning_rate": 1.9617407284967734e-05, + "loss": 1.3796, + "step": 4122 + }, + { + "epoch": 0.35, + "learning_rate": 1.9617158221836466e-05, + "loss": 1.3537, + "step": 4123 + }, + { + "epoch": 0.35, + "learning_rate": 1.961690907924516e-05, + "loss": 1.4121, + "step": 4124 + }, + { + "epoch": 0.35, + "learning_rate": 1.961665985719587e-05, + "loss": 1.293, + "step": 4125 + }, + { + "epoch": 0.35, + "learning_rate": 1.961641055569066e-05, + "loss": 1.3552, + "step": 4126 + }, + { + "epoch": 0.35, + "learning_rate": 1.961616117473159e-05, + "loss": 1.296, + "step": 4127 + }, + { + "epoch": 0.35, + "learning_rate": 1.961591171432072e-05, + "loss": 1.3194, + "step": 4128 + }, + { + "epoch": 0.35, + "learning_rate": 1.9615662174460105e-05, + "loss": 1.3767, + "step": 4129 + }, + { + "epoch": 0.35, + "learning_rate": 1.9615412555151816e-05, + "loss": 1.439, + "step": 4130 + }, + { + "epoch": 0.35, + "learning_rate": 1.961516285639791e-05, + "loss": 1.3038, + "step": 4131 + }, + { + "epoch": 0.35, + "learning_rate": 1.9614913078200453e-05, + "loss": 1.3388, + "step": 4132 + }, + { + "epoch": 0.35, + "learning_rate": 1.9614663220561504e-05, + "loss": 1.2854, + "step": 4133 + }, + { + "epoch": 0.35, + "learning_rate": 1.9614413283483132e-05, + "loss": 1.3408, + "step": 4134 + }, + { + "epoch": 0.35, + "learning_rate": 1.96141632669674e-05, + "loss": 1.3472, + "step": 4135 + }, + { + "epoch": 0.35, + "learning_rate": 1.9613913171016373e-05, + "loss": 1.3027, + "step": 4136 + }, + { + "epoch": 0.35, + "learning_rate": 1.961366299563212e-05, + "loss": 1.3193, + "step": 4137 + }, + { + "epoch": 0.35, + "learning_rate": 1.9613412740816703e-05, + "loss": 1.4436, + "step": 4138 + }, + { + "epoch": 0.35, + "learning_rate": 1.96131624065722e-05, + "loss": 1.2987, + "step": 4139 + }, + { + "epoch": 0.35, + "learning_rate": 1.961291199290067e-05, + "loss": 1.306, + "step": 4140 + }, + { + "epoch": 0.35, + "learning_rate": 1.961266149980418e-05, + "loss": 1.3236, + "step": 4141 + }, + { + "epoch": 0.35, + "learning_rate": 1.961241092728481e-05, + "loss": 1.296, + "step": 4142 + }, + { + "epoch": 0.35, + "learning_rate": 1.9612160275344623e-05, + "loss": 1.2876, + "step": 4143 + }, + { + "epoch": 0.35, + "learning_rate": 1.961190954398569e-05, + "loss": 1.3398, + "step": 4144 + }, + { + "epoch": 0.35, + "learning_rate": 1.9611658733210085e-05, + "loss": 1.3511, + "step": 4145 + }, + { + "epoch": 0.35, + "learning_rate": 1.9611407843019877e-05, + "loss": 1.349, + "step": 4146 + }, + { + "epoch": 0.35, + "learning_rate": 1.9611156873417145e-05, + "loss": 1.2824, + "step": 4147 + }, + { + "epoch": 0.35, + "learning_rate": 1.961090582440396e-05, + "loss": 1.3604, + "step": 4148 + }, + { + "epoch": 0.35, + "learning_rate": 1.961065469598239e-05, + "loss": 1.3363, + "step": 4149 + }, + { + "epoch": 0.35, + "learning_rate": 1.961040348815452e-05, + "loss": 1.4159, + "step": 4150 + }, + { + "epoch": 0.35, + "learning_rate": 1.9610152200922418e-05, + "loss": 1.3218, + "step": 4151 + }, + { + "epoch": 0.35, + "learning_rate": 1.960990083428816e-05, + "loss": 1.3277, + "step": 4152 + }, + { + "epoch": 0.35, + "learning_rate": 1.960964938825383e-05, + "loss": 1.389, + "step": 4153 + }, + { + "epoch": 0.35, + "learning_rate": 1.9609397862821503e-05, + "loss": 1.2864, + "step": 4154 + }, + { + "epoch": 0.35, + "learning_rate": 1.9609146257993253e-05, + "loss": 1.3437, + "step": 4155 + }, + { + "epoch": 0.35, + "learning_rate": 1.9608894573771162e-05, + "loss": 1.3712, + "step": 4156 + }, + { + "epoch": 0.35, + "learning_rate": 1.960864281015731e-05, + "loss": 1.4004, + "step": 4157 + }, + { + "epoch": 0.35, + "learning_rate": 1.9608390967153772e-05, + "loss": 1.4461, + "step": 4158 + }, + { + "epoch": 0.35, + "learning_rate": 1.9608139044762635e-05, + "loss": 1.3588, + "step": 4159 + }, + { + "epoch": 0.35, + "learning_rate": 1.9607887042985977e-05, + "loss": 1.2704, + "step": 4160 + }, + { + "epoch": 0.35, + "learning_rate": 1.9607634961825883e-05, + "loss": 1.3634, + "step": 4161 + }, + { + "epoch": 0.35, + "learning_rate": 1.9607382801284432e-05, + "loss": 1.2925, + "step": 4162 + }, + { + "epoch": 0.35, + "learning_rate": 1.9607130561363713e-05, + "loss": 1.3681, + "step": 4163 + }, + { + "epoch": 0.35, + "learning_rate": 1.9606878242065804e-05, + "loss": 1.294, + "step": 4164 + }, + { + "epoch": 0.35, + "learning_rate": 1.9606625843392794e-05, + "loss": 1.3542, + "step": 4165 + }, + { + "epoch": 0.35, + "learning_rate": 1.9606373365346766e-05, + "loss": 1.4074, + "step": 4166 + }, + { + "epoch": 0.35, + "learning_rate": 1.9606120807929807e-05, + "loss": 1.3871, + "step": 4167 + }, + { + "epoch": 0.35, + "learning_rate": 1.9605868171144004e-05, + "loss": 1.2996, + "step": 4168 + }, + { + "epoch": 0.35, + "learning_rate": 1.960561545499144e-05, + "loss": 1.321, + "step": 4169 + }, + { + "epoch": 0.35, + "learning_rate": 1.960536265947421e-05, + "loss": 1.2705, + "step": 4170 + }, + { + "epoch": 0.35, + "learning_rate": 1.96051097845944e-05, + "loss": 1.3295, + "step": 4171 + }, + { + "epoch": 0.35, + "learning_rate": 1.96048568303541e-05, + "loss": 1.2912, + "step": 4172 + }, + { + "epoch": 0.35, + "learning_rate": 1.9604603796755396e-05, + "loss": 1.4355, + "step": 4173 + }, + { + "epoch": 0.35, + "learning_rate": 1.9604350683800382e-05, + "loss": 1.3288, + "step": 4174 + }, + { + "epoch": 0.35, + "learning_rate": 1.960409749149115e-05, + "loss": 1.2715, + "step": 4175 + }, + { + "epoch": 0.35, + "learning_rate": 1.9603844219829794e-05, + "loss": 1.3425, + "step": 4176 + }, + { + "epoch": 0.35, + "learning_rate": 1.96035908688184e-05, + "loss": 1.3405, + "step": 4177 + }, + { + "epoch": 0.35, + "learning_rate": 1.9603337438459064e-05, + "loss": 1.3197, + "step": 4178 + }, + { + "epoch": 0.35, + "learning_rate": 1.960308392875388e-05, + "loss": 1.342, + "step": 4179 + }, + { + "epoch": 0.35, + "learning_rate": 1.9602830339704946e-05, + "loss": 1.3158, + "step": 4180 + }, + { + "epoch": 0.35, + "learning_rate": 1.960257667131435e-05, + "loss": 1.3455, + "step": 4181 + }, + { + "epoch": 0.35, + "learning_rate": 1.9602322923584198e-05, + "loss": 1.4421, + "step": 4182 + }, + { + "epoch": 0.35, + "learning_rate": 1.9602069096516577e-05, + "loss": 1.4662, + "step": 4183 + }, + { + "epoch": 0.35, + "learning_rate": 1.9601815190113587e-05, + "loss": 1.2988, + "step": 4184 + }, + { + "epoch": 0.35, + "learning_rate": 1.9601561204377333e-05, + "loss": 1.2937, + "step": 4185 + }, + { + "epoch": 0.35, + "learning_rate": 1.9601307139309904e-05, + "loss": 1.4099, + "step": 4186 + }, + { + "epoch": 0.35, + "learning_rate": 1.96010529949134e-05, + "loss": 1.3665, + "step": 4187 + }, + { + "epoch": 0.35, + "learning_rate": 1.9600798771189927e-05, + "loss": 1.3845, + "step": 4188 + }, + { + "epoch": 0.35, + "learning_rate": 1.960054446814158e-05, + "loss": 1.3994, + "step": 4189 + }, + { + "epoch": 0.35, + "learning_rate": 1.9600290085770463e-05, + "loss": 1.2424, + "step": 4190 + }, + { + "epoch": 0.35, + "learning_rate": 1.9600035624078674e-05, + "loss": 1.2807, + "step": 4191 + }, + { + "epoch": 0.35, + "learning_rate": 1.9599781083068324e-05, + "loss": 1.3687, + "step": 4192 + }, + { + "epoch": 0.35, + "learning_rate": 1.9599526462741508e-05, + "loss": 1.3786, + "step": 4193 + }, + { + "epoch": 0.35, + "learning_rate": 1.959927176310033e-05, + "loss": 1.3235, + "step": 4194 + }, + { + "epoch": 0.35, + "learning_rate": 1.9599016984146897e-05, + "loss": 1.3722, + "step": 4195 + }, + { + "epoch": 0.35, + "learning_rate": 1.9598762125883315e-05, + "loss": 1.3315, + "step": 4196 + }, + { + "epoch": 0.35, + "learning_rate": 1.959850718831169e-05, + "loss": 1.3709, + "step": 4197 + }, + { + "epoch": 0.35, + "learning_rate": 1.9598252171434122e-05, + "loss": 1.4068, + "step": 4198 + }, + { + "epoch": 0.35, + "learning_rate": 1.9597997075252724e-05, + "loss": 1.3375, + "step": 4199 + }, + { + "epoch": 0.35, + "learning_rate": 1.9597741899769604e-05, + "loss": 1.3793, + "step": 4200 + }, + { + "epoch": 0.35, + "learning_rate": 1.9597486644986866e-05, + "loss": 1.3386, + "step": 4201 + }, + { + "epoch": 0.35, + "learning_rate": 1.9597231310906626e-05, + "loss": 1.4318, + "step": 4202 + }, + { + "epoch": 0.35, + "learning_rate": 1.959697589753099e-05, + "loss": 1.2307, + "step": 4203 + }, + { + "epoch": 0.35, + "learning_rate": 1.9596720404862063e-05, + "loss": 1.3797, + "step": 4204 + }, + { + "epoch": 0.35, + "learning_rate": 1.9596464832901964e-05, + "loss": 1.3731, + "step": 4205 + }, + { + "epoch": 0.35, + "learning_rate": 1.95962091816528e-05, + "loss": 1.2842, + "step": 4206 + }, + { + "epoch": 0.35, + "learning_rate": 1.959595345111668e-05, + "loss": 1.35, + "step": 4207 + }, + { + "epoch": 0.35, + "learning_rate": 1.9595697641295727e-05, + "loss": 1.3469, + "step": 4208 + }, + { + "epoch": 0.35, + "learning_rate": 1.959544175219205e-05, + "loss": 1.2957, + "step": 4209 + }, + { + "epoch": 0.35, + "learning_rate": 1.959518578380776e-05, + "loss": 1.3818, + "step": 4210 + }, + { + "epoch": 0.35, + "learning_rate": 1.9594929736144978e-05, + "loss": 1.3478, + "step": 4211 + }, + { + "epoch": 0.35, + "learning_rate": 1.959467360920581e-05, + "loss": 1.3341, + "step": 4212 + }, + { + "epoch": 0.35, + "learning_rate": 1.959441740299238e-05, + "loss": 1.3761, + "step": 4213 + }, + { + "epoch": 0.35, + "learning_rate": 1.9594161117506803e-05, + "loss": 1.4025, + "step": 4214 + }, + { + "epoch": 0.35, + "learning_rate": 1.9593904752751197e-05, + "loss": 1.3527, + "step": 4215 + }, + { + "epoch": 0.35, + "learning_rate": 1.959364830872768e-05, + "loss": 1.361, + "step": 4216 + }, + { + "epoch": 0.36, + "learning_rate": 1.9593391785438368e-05, + "loss": 1.3796, + "step": 4217 + }, + { + "epoch": 0.36, + "learning_rate": 1.9593135182885384e-05, + "loss": 1.3337, + "step": 4218 + }, + { + "epoch": 0.36, + "learning_rate": 1.9592878501070847e-05, + "loss": 1.2827, + "step": 4219 + }, + { + "epoch": 0.36, + "learning_rate": 1.959262173999688e-05, + "loss": 1.2888, + "step": 4220 + }, + { + "epoch": 0.36, + "learning_rate": 1.9592364899665596e-05, + "loss": 1.4385, + "step": 4221 + }, + { + "epoch": 0.36, + "learning_rate": 1.959210798007913e-05, + "loss": 1.24, + "step": 4222 + }, + { + "epoch": 0.36, + "learning_rate": 1.9591850981239596e-05, + "loss": 1.449, + "step": 4223 + }, + { + "epoch": 0.36, + "learning_rate": 1.959159390314912e-05, + "loss": 1.3638, + "step": 4224 + }, + { + "epoch": 0.36, + "learning_rate": 1.959133674580982e-05, + "loss": 1.4041, + "step": 4225 + }, + { + "epoch": 0.36, + "learning_rate": 1.9591079509223833e-05, + "loss": 1.39, + "step": 4226 + }, + { + "epoch": 0.36, + "learning_rate": 1.9590822193393274e-05, + "loss": 1.3384, + "step": 4227 + }, + { + "epoch": 0.36, + "learning_rate": 1.9590564798320272e-05, + "loss": 1.2633, + "step": 4228 + }, + { + "epoch": 0.36, + "learning_rate": 1.9590307324006955e-05, + "loss": 1.3856, + "step": 4229 + }, + { + "epoch": 0.36, + "learning_rate": 1.9590049770455454e-05, + "loss": 1.2952, + "step": 4230 + }, + { + "epoch": 0.36, + "learning_rate": 1.9589792137667887e-05, + "loss": 1.3828, + "step": 4231 + }, + { + "epoch": 0.36, + "learning_rate": 1.958953442564639e-05, + "loss": 1.4, + "step": 4232 + }, + { + "epoch": 0.36, + "learning_rate": 1.958927663439309e-05, + "loss": 1.2352, + "step": 4233 + }, + { + "epoch": 0.36, + "learning_rate": 1.958901876391012e-05, + "loss": 1.3957, + "step": 4234 + }, + { + "epoch": 0.36, + "learning_rate": 1.9588760814199604e-05, + "loss": 1.2932, + "step": 4235 + }, + { + "epoch": 0.36, + "learning_rate": 1.958850278526368e-05, + "loss": 1.2255, + "step": 4236 + }, + { + "epoch": 0.36, + "learning_rate": 1.9588244677104477e-05, + "loss": 1.2901, + "step": 4237 + }, + { + "epoch": 0.36, + "learning_rate": 1.9587986489724128e-05, + "loss": 1.3668, + "step": 4238 + }, + { + "epoch": 0.36, + "learning_rate": 1.9587728223124764e-05, + "loss": 1.3569, + "step": 4239 + }, + { + "epoch": 0.36, + "learning_rate": 1.9587469877308524e-05, + "loss": 1.3437, + "step": 4240 + }, + { + "epoch": 0.36, + "learning_rate": 1.9587211452277536e-05, + "loss": 1.3019, + "step": 4241 + }, + { + "epoch": 0.36, + "learning_rate": 1.958695294803394e-05, + "loss": 1.3037, + "step": 4242 + }, + { + "epoch": 0.36, + "learning_rate": 1.9586694364579874e-05, + "loss": 1.3078, + "step": 4243 + }, + { + "epoch": 0.36, + "learning_rate": 1.958643570191747e-05, + "loss": 1.3233, + "step": 4244 + }, + { + "epoch": 0.36, + "learning_rate": 1.958617696004886e-05, + "loss": 1.3187, + "step": 4245 + }, + { + "epoch": 0.36, + "learning_rate": 1.9585918138976193e-05, + "loss": 1.2882, + "step": 4246 + }, + { + "epoch": 0.36, + "learning_rate": 1.9585659238701605e-05, + "loss": 1.3091, + "step": 4247 + }, + { + "epoch": 0.36, + "learning_rate": 1.958540025922723e-05, + "loss": 1.2909, + "step": 4248 + }, + { + "epoch": 0.36, + "learning_rate": 1.958514120055521e-05, + "loss": 1.2791, + "step": 4249 + }, + { + "epoch": 0.36, + "learning_rate": 1.9584882062687686e-05, + "loss": 1.3457, + "step": 4250 + }, + { + "epoch": 0.36, + "learning_rate": 1.9584622845626797e-05, + "loss": 1.3313, + "step": 4251 + }, + { + "epoch": 0.36, + "learning_rate": 1.9584363549374687e-05, + "loss": 1.3752, + "step": 4252 + }, + { + "epoch": 0.36, + "learning_rate": 1.95841041739335e-05, + "loss": 1.3452, + "step": 4253 + }, + { + "epoch": 0.36, + "learning_rate": 1.9583844719305377e-05, + "loss": 1.438, + "step": 4254 + }, + { + "epoch": 0.36, + "learning_rate": 1.958358518549246e-05, + "loss": 1.3048, + "step": 4255 + }, + { + "epoch": 0.36, + "learning_rate": 1.9583325572496895e-05, + "loss": 1.3745, + "step": 4256 + }, + { + "epoch": 0.36, + "learning_rate": 1.9583065880320827e-05, + "loss": 1.4102, + "step": 4257 + }, + { + "epoch": 0.36, + "learning_rate": 1.9582806108966402e-05, + "loss": 1.3139, + "step": 4258 + }, + { + "epoch": 0.36, + "learning_rate": 1.958254625843577e-05, + "loss": 1.3039, + "step": 4259 + }, + { + "epoch": 0.36, + "learning_rate": 1.9582286328731067e-05, + "loss": 1.3332, + "step": 4260 + }, + { + "epoch": 0.36, + "learning_rate": 1.958202631985445e-05, + "loss": 1.3132, + "step": 4261 + }, + { + "epoch": 0.36, + "learning_rate": 1.958176623180807e-05, + "loss": 1.2379, + "step": 4262 + }, + { + "epoch": 0.36, + "learning_rate": 1.9581506064594067e-05, + "loss": 1.3347, + "step": 4263 + }, + { + "epoch": 0.36, + "learning_rate": 1.9581245818214594e-05, + "loss": 1.2301, + "step": 4264 + }, + { + "epoch": 0.36, + "learning_rate": 1.95809854926718e-05, + "loss": 1.3967, + "step": 4265 + }, + { + "epoch": 0.36, + "learning_rate": 1.958072508796784e-05, + "loss": 1.3393, + "step": 4266 + }, + { + "epoch": 0.36, + "learning_rate": 1.9580464604104862e-05, + "loss": 1.3043, + "step": 4267 + }, + { + "epoch": 0.36, + "learning_rate": 1.958020404108502e-05, + "loss": 1.4071, + "step": 4268 + }, + { + "epoch": 0.36, + "learning_rate": 1.9579943398910468e-05, + "loss": 1.4574, + "step": 4269 + }, + { + "epoch": 0.36, + "learning_rate": 1.9579682677583355e-05, + "loss": 1.3629, + "step": 4270 + }, + { + "epoch": 0.36, + "learning_rate": 1.957942187710584e-05, + "loss": 1.3841, + "step": 4271 + }, + { + "epoch": 0.36, + "learning_rate": 1.9579160997480074e-05, + "loss": 1.3096, + "step": 4272 + }, + { + "epoch": 0.36, + "learning_rate": 1.9578900038708216e-05, + "loss": 1.3407, + "step": 4273 + }, + { + "epoch": 0.36, + "learning_rate": 1.9578639000792415e-05, + "loss": 1.3586, + "step": 4274 + }, + { + "epoch": 0.36, + "learning_rate": 1.957837788373484e-05, + "loss": 1.4464, + "step": 4275 + }, + { + "epoch": 0.36, + "learning_rate": 1.9578116687537635e-05, + "loss": 1.3866, + "step": 4276 + }, + { + "epoch": 0.36, + "learning_rate": 1.957785541220297e-05, + "loss": 1.3581, + "step": 4277 + }, + { + "epoch": 0.36, + "learning_rate": 1.9577594057732997e-05, + "loss": 1.3348, + "step": 4278 + }, + { + "epoch": 0.36, + "learning_rate": 1.9577332624129875e-05, + "loss": 1.3883, + "step": 4279 + }, + { + "epoch": 0.36, + "learning_rate": 1.9577071111395766e-05, + "loss": 1.3337, + "step": 4280 + }, + { + "epoch": 0.36, + "learning_rate": 1.957680951953283e-05, + "loss": 1.2488, + "step": 4281 + }, + { + "epoch": 0.36, + "learning_rate": 1.9576547848543227e-05, + "loss": 1.3211, + "step": 4282 + }, + { + "epoch": 0.36, + "learning_rate": 1.9576286098429124e-05, + "loss": 1.2815, + "step": 4283 + }, + { + "epoch": 0.36, + "learning_rate": 1.957602426919268e-05, + "loss": 1.3235, + "step": 4284 + }, + { + "epoch": 0.36, + "learning_rate": 1.9575762360836062e-05, + "loss": 1.3102, + "step": 4285 + }, + { + "epoch": 0.36, + "learning_rate": 1.9575500373361426e-05, + "loss": 1.3461, + "step": 4286 + }, + { + "epoch": 0.36, + "learning_rate": 1.957523830677094e-05, + "loss": 1.2912, + "step": 4287 + }, + { + "epoch": 0.36, + "learning_rate": 1.9574976161066773e-05, + "loss": 1.3204, + "step": 4288 + }, + { + "epoch": 0.36, + "learning_rate": 1.957471393625109e-05, + "loss": 1.3094, + "step": 4289 + }, + { + "epoch": 0.36, + "learning_rate": 1.9574451632326056e-05, + "loss": 1.3969, + "step": 4290 + }, + { + "epoch": 0.36, + "learning_rate": 1.9574189249293837e-05, + "loss": 1.2733, + "step": 4291 + }, + { + "epoch": 0.36, + "learning_rate": 1.9573926787156602e-05, + "loss": 1.2972, + "step": 4292 + }, + { + "epoch": 0.36, + "learning_rate": 1.957366424591652e-05, + "loss": 1.3242, + "step": 4293 + }, + { + "epoch": 0.36, + "learning_rate": 1.9573401625575764e-05, + "loss": 1.2966, + "step": 4294 + }, + { + "epoch": 0.36, + "learning_rate": 1.9573138926136495e-05, + "loss": 1.314, + "step": 4295 + }, + { + "epoch": 0.36, + "learning_rate": 1.957287614760089e-05, + "loss": 1.2906, + "step": 4296 + }, + { + "epoch": 0.36, + "learning_rate": 1.957261328997112e-05, + "loss": 1.3178, + "step": 4297 + }, + { + "epoch": 0.36, + "learning_rate": 1.9572350353249354e-05, + "loss": 1.3672, + "step": 4298 + }, + { + "epoch": 0.36, + "learning_rate": 1.9572087337437763e-05, + "loss": 1.3221, + "step": 4299 + }, + { + "epoch": 0.36, + "learning_rate": 1.9571824242538527e-05, + "loss": 1.3341, + "step": 4300 + }, + { + "epoch": 0.36, + "learning_rate": 1.9571561068553816e-05, + "loss": 1.322, + "step": 4301 + }, + { + "epoch": 0.36, + "learning_rate": 1.95712978154858e-05, + "loss": 1.3528, + "step": 4302 + }, + { + "epoch": 0.36, + "learning_rate": 1.957103448333666e-05, + "loss": 1.3495, + "step": 4303 + }, + { + "epoch": 0.36, + "learning_rate": 1.9570771072108572e-05, + "loss": 1.2822, + "step": 4304 + }, + { + "epoch": 0.36, + "learning_rate": 1.9570507581803707e-05, + "loss": 1.3061, + "step": 4305 + }, + { + "epoch": 0.36, + "learning_rate": 1.957024401242425e-05, + "loss": 1.3568, + "step": 4306 + }, + { + "epoch": 0.36, + "learning_rate": 1.956998036397237e-05, + "loss": 1.3618, + "step": 4307 + }, + { + "epoch": 0.36, + "learning_rate": 1.956971663645025e-05, + "loss": 1.2953, + "step": 4308 + }, + { + "epoch": 0.36, + "learning_rate": 1.956945282986007e-05, + "loss": 1.2914, + "step": 4309 + }, + { + "epoch": 0.36, + "learning_rate": 1.956918894420401e-05, + "loss": 1.3724, + "step": 4310 + }, + { + "epoch": 0.36, + "learning_rate": 1.9568924979484245e-05, + "loss": 1.2906, + "step": 4311 + }, + { + "epoch": 0.36, + "learning_rate": 1.956866093570296e-05, + "loss": 1.2828, + "step": 4312 + }, + { + "epoch": 0.36, + "learning_rate": 1.9568396812862337e-05, + "loss": 1.3284, + "step": 4313 + }, + { + "epoch": 0.36, + "learning_rate": 1.9568132610964557e-05, + "loss": 1.3465, + "step": 4314 + }, + { + "epoch": 0.36, + "learning_rate": 1.9567868330011804e-05, + "loss": 1.448, + "step": 4315 + }, + { + "epoch": 0.36, + "learning_rate": 1.9567603970006258e-05, + "loss": 1.3948, + "step": 4316 + }, + { + "epoch": 0.36, + "learning_rate": 1.9567339530950106e-05, + "loss": 1.2458, + "step": 4317 + }, + { + "epoch": 0.36, + "learning_rate": 1.9567075012845535e-05, + "loss": 1.3204, + "step": 4318 + }, + { + "epoch": 0.36, + "learning_rate": 1.956681041569473e-05, + "loss": 1.2936, + "step": 4319 + }, + { + "epoch": 0.36, + "learning_rate": 1.956654573949987e-05, + "loss": 1.3601, + "step": 4320 + }, + { + "epoch": 0.36, + "learning_rate": 1.9566280984263152e-05, + "loss": 1.3013, + "step": 4321 + }, + { + "epoch": 0.36, + "learning_rate": 1.9566016149986756e-05, + "loss": 1.2982, + "step": 4322 + }, + { + "epoch": 0.36, + "learning_rate": 1.9565751236672875e-05, + "loss": 1.3132, + "step": 4323 + }, + { + "epoch": 0.36, + "learning_rate": 1.95654862443237e-05, + "loss": 1.3446, + "step": 4324 + }, + { + "epoch": 0.36, + "learning_rate": 1.9565221172941408e-05, + "loss": 1.2655, + "step": 4325 + }, + { + "epoch": 0.36, + "learning_rate": 1.95649560225282e-05, + "loss": 1.3753, + "step": 4326 + }, + { + "epoch": 0.36, + "learning_rate": 1.9564690793086268e-05, + "loss": 1.3584, + "step": 4327 + }, + { + "epoch": 0.36, + "learning_rate": 1.95644254846178e-05, + "loss": 1.2776, + "step": 4328 + }, + { + "epoch": 0.36, + "learning_rate": 1.956416009712498e-05, + "loss": 1.4249, + "step": 4329 + }, + { + "epoch": 0.36, + "learning_rate": 1.9563894630610013e-05, + "loss": 1.3276, + "step": 4330 + }, + { + "epoch": 0.36, + "learning_rate": 1.9563629085075087e-05, + "loss": 1.3527, + "step": 4331 + }, + { + "epoch": 0.36, + "learning_rate": 1.9563363460522396e-05, + "loss": 1.306, + "step": 4332 + }, + { + "epoch": 0.36, + "learning_rate": 1.956309775695414e-05, + "loss": 1.41, + "step": 4333 + }, + { + "epoch": 0.36, + "learning_rate": 1.9562831974372506e-05, + "loss": 1.3724, + "step": 4334 + }, + { + "epoch": 0.36, + "learning_rate": 1.9562566112779694e-05, + "loss": 1.3205, + "step": 4335 + }, + { + "epoch": 0.37, + "learning_rate": 1.9562300172177902e-05, + "loss": 1.3932, + "step": 4336 + }, + { + "epoch": 0.37, + "learning_rate": 1.9562034152569326e-05, + "loss": 1.3964, + "step": 4337 + }, + { + "epoch": 0.37, + "learning_rate": 1.9561768053956157e-05, + "loss": 1.2531, + "step": 4338 + }, + { + "epoch": 0.37, + "learning_rate": 1.9561501876340607e-05, + "loss": 1.3677, + "step": 4339 + }, + { + "epoch": 0.37, + "learning_rate": 1.9561235619724867e-05, + "loss": 1.4024, + "step": 4340 + }, + { + "epoch": 0.37, + "learning_rate": 1.9560969284111137e-05, + "loss": 1.3224, + "step": 4341 + }, + { + "epoch": 0.37, + "learning_rate": 1.956070286950162e-05, + "loss": 1.4454, + "step": 4342 + }, + { + "epoch": 0.37, + "learning_rate": 1.9560436375898516e-05, + "loss": 1.2983, + "step": 4343 + }, + { + "epoch": 0.37, + "learning_rate": 1.956016980330403e-05, + "loss": 1.4155, + "step": 4344 + }, + { + "epoch": 0.37, + "learning_rate": 1.9559903151720356e-05, + "loss": 1.3201, + "step": 4345 + }, + { + "epoch": 0.37, + "learning_rate": 1.9559636421149704e-05, + "loss": 1.3677, + "step": 4346 + }, + { + "epoch": 0.37, + "learning_rate": 1.9559369611594278e-05, + "loss": 1.2662, + "step": 4347 + }, + { + "epoch": 0.37, + "learning_rate": 1.9559102723056278e-05, + "loss": 1.3513, + "step": 4348 + }, + { + "epoch": 0.37, + "learning_rate": 1.9558835755537914e-05, + "loss": 1.2553, + "step": 4349 + }, + { + "epoch": 0.37, + "learning_rate": 1.955856870904139e-05, + "loss": 1.3105, + "step": 4350 + }, + { + "epoch": 0.37, + "learning_rate": 1.9558301583568914e-05, + "loss": 1.3267, + "step": 4351 + }, + { + "epoch": 0.37, + "learning_rate": 1.955803437912269e-05, + "loss": 1.3235, + "step": 4352 + }, + { + "epoch": 0.37, + "learning_rate": 1.9557767095704925e-05, + "loss": 1.3596, + "step": 4353 + }, + { + "epoch": 0.37, + "learning_rate": 1.955749973331783e-05, + "loss": 1.3574, + "step": 4354 + }, + { + "epoch": 0.37, + "learning_rate": 1.9557232291963614e-05, + "loss": 1.3469, + "step": 4355 + }, + { + "epoch": 0.37, + "learning_rate": 1.9556964771644488e-05, + "loss": 1.2914, + "step": 4356 + }, + { + "epoch": 0.37, + "learning_rate": 1.9556697172362657e-05, + "loss": 1.3396, + "step": 4357 + }, + { + "epoch": 0.37, + "learning_rate": 1.9556429494120336e-05, + "loss": 1.265, + "step": 4358 + }, + { + "epoch": 0.37, + "learning_rate": 1.955616173691974e-05, + "loss": 1.2677, + "step": 4359 + }, + { + "epoch": 0.37, + "learning_rate": 1.955589390076307e-05, + "loss": 1.3319, + "step": 4360 + }, + { + "epoch": 0.37, + "learning_rate": 1.9555625985652553e-05, + "loss": 1.3882, + "step": 4361 + }, + { + "epoch": 0.37, + "learning_rate": 1.9555357991590394e-05, + "loss": 1.3226, + "step": 4362 + }, + { + "epoch": 0.37, + "learning_rate": 1.955508991857881e-05, + "loss": 1.3618, + "step": 4363 + }, + { + "epoch": 0.37, + "learning_rate": 1.9554821766620012e-05, + "loss": 1.379, + "step": 4364 + }, + { + "epoch": 0.37, + "learning_rate": 1.9554553535716222e-05, + "loss": 1.3319, + "step": 4365 + }, + { + "epoch": 0.37, + "learning_rate": 1.955428522586965e-05, + "loss": 1.4241, + "step": 4366 + }, + { + "epoch": 0.37, + "learning_rate": 1.9554016837082516e-05, + "loss": 1.2907, + "step": 4367 + }, + { + "epoch": 0.37, + "learning_rate": 1.9553748369357038e-05, + "loss": 1.3112, + "step": 4368 + }, + { + "epoch": 0.37, + "learning_rate": 1.9553479822695435e-05, + "loss": 1.2899, + "step": 4369 + }, + { + "epoch": 0.37, + "learning_rate": 1.9553211197099923e-05, + "loss": 1.3092, + "step": 4370 + }, + { + "epoch": 0.37, + "learning_rate": 1.955294249257272e-05, + "loss": 1.3697, + "step": 4371 + }, + { + "epoch": 0.37, + "learning_rate": 1.9552673709116054e-05, + "loss": 1.3855, + "step": 4372 + }, + { + "epoch": 0.37, + "learning_rate": 1.955240484673214e-05, + "loss": 1.3356, + "step": 4373 + }, + { + "epoch": 0.37, + "learning_rate": 1.95521359054232e-05, + "loss": 1.4318, + "step": 4374 + }, + { + "epoch": 0.37, + "learning_rate": 1.9551866885191448e-05, + "loss": 1.3566, + "step": 4375 + }, + { + "epoch": 0.37, + "learning_rate": 1.9551597786039124e-05, + "loss": 1.388, + "step": 4376 + }, + { + "epoch": 0.37, + "learning_rate": 1.955132860796844e-05, + "loss": 1.3172, + "step": 4377 + }, + { + "epoch": 0.37, + "learning_rate": 1.9551059350981618e-05, + "loss": 1.3064, + "step": 4378 + }, + { + "epoch": 0.37, + "learning_rate": 1.955079001508089e-05, + "loss": 1.3008, + "step": 4379 + }, + { + "epoch": 0.37, + "learning_rate": 1.955052060026848e-05, + "loss": 1.3808, + "step": 4380 + }, + { + "epoch": 0.37, + "learning_rate": 1.955025110654661e-05, + "loss": 1.3715, + "step": 4381 + }, + { + "epoch": 0.37, + "learning_rate": 1.9549981533917507e-05, + "loss": 1.2483, + "step": 4382 + }, + { + "epoch": 0.37, + "learning_rate": 1.9549711882383407e-05, + "loss": 1.2916, + "step": 4383 + }, + { + "epoch": 0.37, + "learning_rate": 1.9549442151946524e-05, + "loss": 1.2508, + "step": 4384 + }, + { + "epoch": 0.37, + "learning_rate": 1.95491723426091e-05, + "loss": 1.2866, + "step": 4385 + }, + { + "epoch": 0.37, + "learning_rate": 1.9548902454373353e-05, + "loss": 1.3725, + "step": 4386 + }, + { + "epoch": 0.37, + "learning_rate": 1.954863248724152e-05, + "loss": 1.4202, + "step": 4387 + }, + { + "epoch": 0.37, + "learning_rate": 1.954836244121583e-05, + "loss": 1.387, + "step": 4388 + }, + { + "epoch": 0.37, + "learning_rate": 1.954809231629851e-05, + "loss": 1.2898, + "step": 4389 + }, + { + "epoch": 0.37, + "learning_rate": 1.95478221124918e-05, + "loss": 1.3655, + "step": 4390 + }, + { + "epoch": 0.37, + "learning_rate": 1.9547551829797922e-05, + "loss": 1.2681, + "step": 4391 + }, + { + "epoch": 0.37, + "learning_rate": 1.954728146821912e-05, + "loss": 1.4313, + "step": 4392 + }, + { + "epoch": 0.37, + "learning_rate": 1.9547011027757623e-05, + "loss": 1.3712, + "step": 4393 + }, + { + "epoch": 0.37, + "learning_rate": 1.9546740508415663e-05, + "loss": 1.3552, + "step": 4394 + }, + { + "epoch": 0.37, + "learning_rate": 1.954646991019548e-05, + "loss": 1.2427, + "step": 4395 + }, + { + "epoch": 0.37, + "learning_rate": 1.9546199233099306e-05, + "loss": 1.3007, + "step": 4396 + }, + { + "epoch": 0.37, + "learning_rate": 1.9545928477129382e-05, + "loss": 1.3602, + "step": 4397 + }, + { + "epoch": 0.37, + "learning_rate": 1.9545657642287937e-05, + "loss": 1.3122, + "step": 4398 + }, + { + "epoch": 0.37, + "learning_rate": 1.9545386728577215e-05, + "loss": 1.4167, + "step": 4399 + }, + { + "epoch": 0.37, + "learning_rate": 1.9545115735999455e-05, + "loss": 1.4158, + "step": 4400 + }, + { + "epoch": 0.37, + "learning_rate": 1.9544844664556894e-05, + "loss": 1.4252, + "step": 4401 + }, + { + "epoch": 0.37, + "learning_rate": 1.9544573514251766e-05, + "loss": 1.3611, + "step": 4402 + }, + { + "epoch": 0.37, + "learning_rate": 1.9544302285086324e-05, + "loss": 1.3345, + "step": 4403 + }, + { + "epoch": 0.37, + "learning_rate": 1.9544030977062798e-05, + "loss": 1.4209, + "step": 4404 + }, + { + "epoch": 0.37, + "learning_rate": 1.9543759590183435e-05, + "loss": 1.4234, + "step": 4405 + }, + { + "epoch": 0.37, + "learning_rate": 1.9543488124450476e-05, + "loss": 1.3055, + "step": 4406 + }, + { + "epoch": 0.37, + "learning_rate": 1.9543216579866162e-05, + "loss": 1.4335, + "step": 4407 + }, + { + "epoch": 0.37, + "learning_rate": 1.954294495643274e-05, + "loss": 1.3218, + "step": 4408 + }, + { + "epoch": 0.37, + "learning_rate": 1.9542673254152454e-05, + "loss": 1.3693, + "step": 4409 + }, + { + "epoch": 0.37, + "learning_rate": 1.9542401473027546e-05, + "loss": 1.3487, + "step": 4410 + }, + { + "epoch": 0.37, + "learning_rate": 1.9542129613060265e-05, + "loss": 1.4847, + "step": 4411 + }, + { + "epoch": 0.37, + "learning_rate": 1.954185767425285e-05, + "loss": 1.429, + "step": 4412 + }, + { + "epoch": 0.37, + "learning_rate": 1.954158565660756e-05, + "loss": 1.3488, + "step": 4413 + }, + { + "epoch": 0.37, + "learning_rate": 1.954131356012663e-05, + "loss": 1.3387, + "step": 4414 + }, + { + "epoch": 0.37, + "learning_rate": 1.9541041384812316e-05, + "loss": 1.4313, + "step": 4415 + }, + { + "epoch": 0.37, + "learning_rate": 1.9540769130666863e-05, + "loss": 1.3994, + "step": 4416 + }, + { + "epoch": 0.37, + "learning_rate": 1.9540496797692525e-05, + "loss": 1.3327, + "step": 4417 + }, + { + "epoch": 0.37, + "learning_rate": 1.9540224385891547e-05, + "loss": 1.4067, + "step": 4418 + }, + { + "epoch": 0.37, + "learning_rate": 1.953995189526618e-05, + "loss": 1.3253, + "step": 4419 + }, + { + "epoch": 0.37, + "learning_rate": 1.953967932581868e-05, + "loss": 1.4073, + "step": 4420 + }, + { + "epoch": 0.37, + "learning_rate": 1.9539406677551298e-05, + "loss": 1.3698, + "step": 4421 + }, + { + "epoch": 0.37, + "learning_rate": 1.9539133950466283e-05, + "loss": 1.3813, + "step": 4422 + }, + { + "epoch": 0.37, + "learning_rate": 1.953886114456589e-05, + "loss": 1.25, + "step": 4423 + }, + { + "epoch": 0.37, + "learning_rate": 1.9538588259852375e-05, + "loss": 1.3695, + "step": 4424 + }, + { + "epoch": 0.37, + "learning_rate": 1.9538315296327988e-05, + "loss": 1.2939, + "step": 4425 + }, + { + "epoch": 0.37, + "learning_rate": 1.953804225399499e-05, + "loss": 1.2853, + "step": 4426 + }, + { + "epoch": 0.37, + "learning_rate": 1.9537769132855634e-05, + "loss": 1.3792, + "step": 4427 + }, + { + "epoch": 0.37, + "learning_rate": 1.9537495932912175e-05, + "loss": 1.4646, + "step": 4428 + }, + { + "epoch": 0.37, + "learning_rate": 1.9537222654166876e-05, + "loss": 1.2934, + "step": 4429 + }, + { + "epoch": 0.37, + "learning_rate": 1.9536949296621987e-05, + "loss": 1.3319, + "step": 4430 + }, + { + "epoch": 0.37, + "learning_rate": 1.9536675860279774e-05, + "loss": 1.3407, + "step": 4431 + }, + { + "epoch": 0.37, + "learning_rate": 1.9536402345142487e-05, + "loss": 1.3273, + "step": 4432 + }, + { + "epoch": 0.37, + "learning_rate": 1.9536128751212397e-05, + "loss": 1.3954, + "step": 4433 + }, + { + "epoch": 0.37, + "learning_rate": 1.953585507849176e-05, + "loss": 1.4146, + "step": 4434 + }, + { + "epoch": 0.37, + "learning_rate": 1.9535581326982834e-05, + "loss": 1.3758, + "step": 4435 + }, + { + "epoch": 0.37, + "learning_rate": 1.9535307496687884e-05, + "loss": 1.3268, + "step": 4436 + }, + { + "epoch": 0.37, + "learning_rate": 1.953503358760917e-05, + "loss": 1.3559, + "step": 4437 + }, + { + "epoch": 0.37, + "learning_rate": 1.9534759599748962e-05, + "loss": 1.3126, + "step": 4438 + }, + { + "epoch": 0.37, + "learning_rate": 1.9534485533109516e-05, + "loss": 1.3951, + "step": 4439 + }, + { + "epoch": 0.37, + "learning_rate": 1.9534211387693096e-05, + "loss": 1.2819, + "step": 4440 + }, + { + "epoch": 0.37, + "learning_rate": 1.9533937163501974e-05, + "loss": 1.3827, + "step": 4441 + }, + { + "epoch": 0.37, + "learning_rate": 1.953366286053841e-05, + "loss": 1.4203, + "step": 4442 + }, + { + "epoch": 0.37, + "learning_rate": 1.9533388478804674e-05, + "loss": 1.3801, + "step": 4443 + }, + { + "epoch": 0.37, + "learning_rate": 1.9533114018303033e-05, + "loss": 1.3838, + "step": 4444 + }, + { + "epoch": 0.37, + "learning_rate": 1.9532839479035747e-05, + "loss": 1.3475, + "step": 4445 + }, + { + "epoch": 0.37, + "learning_rate": 1.9532564861005096e-05, + "loss": 1.3878, + "step": 4446 + }, + { + "epoch": 0.37, + "learning_rate": 1.953229016421334e-05, + "loss": 1.3433, + "step": 4447 + }, + { + "epoch": 0.37, + "learning_rate": 1.953201538866275e-05, + "loss": 1.4183, + "step": 4448 + }, + { + "epoch": 0.37, + "learning_rate": 1.9531740534355604e-05, + "loss": 1.2929, + "step": 4449 + }, + { + "epoch": 0.37, + "learning_rate": 1.9531465601294163e-05, + "loss": 1.2609, + "step": 4450 + }, + { + "epoch": 0.37, + "learning_rate": 1.9531190589480707e-05, + "loss": 1.3332, + "step": 4451 + }, + { + "epoch": 0.37, + "learning_rate": 1.9530915498917497e-05, + "loss": 1.301, + "step": 4452 + }, + { + "epoch": 0.37, + "learning_rate": 1.9530640329606817e-05, + "loss": 1.3733, + "step": 4453 + }, + { + "epoch": 0.37, + "learning_rate": 1.9530365081550936e-05, + "loss": 1.3539, + "step": 4454 + }, + { + "epoch": 0.38, + "learning_rate": 1.953008975475213e-05, + "loss": 1.4307, + "step": 4455 + }, + { + "epoch": 0.38, + "learning_rate": 1.9529814349212672e-05, + "loss": 1.2612, + "step": 4456 + }, + { + "epoch": 0.38, + "learning_rate": 1.9529538864934837e-05, + "loss": 1.3458, + "step": 4457 + }, + { + "epoch": 0.38, + "learning_rate": 1.95292633019209e-05, + "loss": 1.4019, + "step": 4458 + }, + { + "epoch": 0.38, + "learning_rate": 1.9528987660173146e-05, + "loss": 1.3776, + "step": 4459 + }, + { + "epoch": 0.38, + "learning_rate": 1.952871193969384e-05, + "loss": 1.3665, + "step": 4460 + }, + { + "epoch": 0.38, + "learning_rate": 1.952843614048527e-05, + "loss": 1.3402, + "step": 4461 + }, + { + "epoch": 0.38, + "learning_rate": 1.952816026254971e-05, + "loss": 1.3871, + "step": 4462 + }, + { + "epoch": 0.38, + "learning_rate": 1.9527884305889438e-05, + "loss": 1.3465, + "step": 4463 + }, + { + "epoch": 0.38, + "learning_rate": 1.952760827050674e-05, + "loss": 1.3347, + "step": 4464 + }, + { + "epoch": 0.38, + "learning_rate": 1.9527332156403895e-05, + "loss": 1.3704, + "step": 4465 + }, + { + "epoch": 0.38, + "learning_rate": 1.9527055963583178e-05, + "loss": 1.3496, + "step": 4466 + }, + { + "epoch": 0.38, + "learning_rate": 1.9526779692046877e-05, + "loss": 1.3211, + "step": 4467 + }, + { + "epoch": 0.38, + "learning_rate": 1.9526503341797274e-05, + "loss": 1.4003, + "step": 4468 + }, + { + "epoch": 0.38, + "learning_rate": 1.952622691283665e-05, + "loss": 1.2679, + "step": 4469 + }, + { + "epoch": 0.38, + "learning_rate": 1.9525950405167293e-05, + "loss": 1.4572, + "step": 4470 + }, + { + "epoch": 0.38, + "learning_rate": 1.952567381879148e-05, + "loss": 1.351, + "step": 4471 + }, + { + "epoch": 0.38, + "learning_rate": 1.9525397153711506e-05, + "loss": 1.437, + "step": 4472 + }, + { + "epoch": 0.38, + "learning_rate": 1.9525120409929654e-05, + "loss": 1.3176, + "step": 4473 + }, + { + "epoch": 0.38, + "learning_rate": 1.9524843587448202e-05, + "loss": 1.3378, + "step": 4474 + }, + { + "epoch": 0.38, + "learning_rate": 1.952456668626945e-05, + "loss": 1.316, + "step": 4475 + }, + { + "epoch": 0.38, + "learning_rate": 1.9524289706395678e-05, + "loss": 1.3636, + "step": 4476 + }, + { + "epoch": 0.38, + "learning_rate": 1.9524012647829176e-05, + "loss": 1.293, + "step": 4477 + }, + { + "epoch": 0.38, + "learning_rate": 1.9523735510572233e-05, + "loss": 1.2838, + "step": 4478 + }, + { + "epoch": 0.38, + "learning_rate": 1.952345829462714e-05, + "loss": 1.3142, + "step": 4479 + }, + { + "epoch": 0.38, + "learning_rate": 1.9523180999996186e-05, + "loss": 1.3828, + "step": 4480 + }, + { + "epoch": 0.38, + "learning_rate": 1.9522903626681663e-05, + "loss": 1.3271, + "step": 4481 + }, + { + "epoch": 0.38, + "learning_rate": 1.9522626174685864e-05, + "loss": 1.2411, + "step": 4482 + }, + { + "epoch": 0.38, + "learning_rate": 1.9522348644011075e-05, + "loss": 1.2819, + "step": 4483 + }, + { + "epoch": 0.38, + "learning_rate": 1.95220710346596e-05, + "loss": 1.3268, + "step": 4484 + }, + { + "epoch": 0.38, + "learning_rate": 1.9521793346633723e-05, + "loss": 1.3907, + "step": 4485 + }, + { + "epoch": 0.38, + "learning_rate": 1.9521515579935744e-05, + "loss": 1.456, + "step": 4486 + }, + { + "epoch": 0.38, + "learning_rate": 1.9521237734567954e-05, + "loss": 1.4222, + "step": 4487 + }, + { + "epoch": 0.38, + "learning_rate": 1.9520959810532654e-05, + "loss": 1.2366, + "step": 4488 + }, + { + "epoch": 0.38, + "learning_rate": 1.9520681807832133e-05, + "loss": 1.3563, + "step": 4489 + }, + { + "epoch": 0.38, + "learning_rate": 1.9520403726468692e-05, + "loss": 1.3466, + "step": 4490 + }, + { + "epoch": 0.38, + "learning_rate": 1.9520125566444628e-05, + "loss": 1.3676, + "step": 4491 + }, + { + "epoch": 0.38, + "learning_rate": 1.9519847327762246e-05, + "loss": 1.3487, + "step": 4492 + }, + { + "epoch": 0.38, + "learning_rate": 1.9519569010423832e-05, + "loss": 1.327, + "step": 4493 + }, + { + "epoch": 0.38, + "learning_rate": 1.9519290614431697e-05, + "loss": 1.4168, + "step": 4494 + }, + { + "epoch": 0.38, + "learning_rate": 1.9519012139788134e-05, + "loss": 1.398, + "step": 4495 + }, + { + "epoch": 0.38, + "learning_rate": 1.9518733586495445e-05, + "loss": 1.3699, + "step": 4496 + }, + { + "epoch": 0.38, + "learning_rate": 1.9518454954555933e-05, + "loss": 1.3538, + "step": 4497 + }, + { + "epoch": 0.38, + "learning_rate": 1.9518176243971903e-05, + "loss": 1.281, + "step": 4498 + }, + { + "epoch": 0.38, + "learning_rate": 1.951789745474565e-05, + "loss": 1.3397, + "step": 4499 + }, + { + "epoch": 0.38, + "learning_rate": 1.9517618586879487e-05, + "loss": 1.3973, + "step": 4500 + }, + { + "epoch": 0.38, + "learning_rate": 1.951733964037571e-05, + "loss": 1.2688, + "step": 4501 + }, + { + "epoch": 0.38, + "learning_rate": 1.951706061523663e-05, + "loss": 1.3353, + "step": 4502 + }, + { + "epoch": 0.38, + "learning_rate": 1.9516781511464545e-05, + "loss": 1.2804, + "step": 4503 + }, + { + "epoch": 0.38, + "learning_rate": 1.9516502329061767e-05, + "loss": 1.3628, + "step": 4504 + }, + { + "epoch": 0.38, + "learning_rate": 1.95162230680306e-05, + "loss": 1.3595, + "step": 4505 + }, + { + "epoch": 0.38, + "learning_rate": 1.9515943728373353e-05, + "loss": 1.2912, + "step": 4506 + }, + { + "epoch": 0.38, + "learning_rate": 1.9515664310092336e-05, + "loss": 1.27, + "step": 4507 + }, + { + "epoch": 0.38, + "learning_rate": 1.9515384813189853e-05, + "loss": 1.3805, + "step": 4508 + }, + { + "epoch": 0.38, + "learning_rate": 1.9515105237668215e-05, + "loss": 1.2399, + "step": 4509 + }, + { + "epoch": 0.38, + "learning_rate": 1.9514825583529732e-05, + "loss": 1.3372, + "step": 4510 + }, + { + "epoch": 0.38, + "learning_rate": 1.951454585077672e-05, + "loss": 1.3848, + "step": 4511 + }, + { + "epoch": 0.38, + "learning_rate": 1.951426603941148e-05, + "loss": 1.3003, + "step": 4512 + }, + { + "epoch": 0.38, + "learning_rate": 1.951398614943633e-05, + "loss": 1.3638, + "step": 4513 + }, + { + "epoch": 0.38, + "learning_rate": 1.951370618085358e-05, + "loss": 1.3248, + "step": 4514 + }, + { + "epoch": 0.38, + "learning_rate": 1.951342613366555e-05, + "loss": 1.3446, + "step": 4515 + }, + { + "epoch": 0.38, + "learning_rate": 1.951314600787454e-05, + "loss": 1.3189, + "step": 4516 + }, + { + "epoch": 0.38, + "learning_rate": 1.9512865803482882e-05, + "loss": 1.397, + "step": 4517 + }, + { + "epoch": 0.38, + "learning_rate": 1.951258552049288e-05, + "loss": 1.2946, + "step": 4518 + }, + { + "epoch": 0.38, + "learning_rate": 1.9512305158906847e-05, + "loss": 1.4023, + "step": 4519 + }, + { + "epoch": 0.38, + "learning_rate": 1.9512024718727108e-05, + "loss": 1.3096, + "step": 4520 + }, + { + "epoch": 0.38, + "learning_rate": 1.9511744199955976e-05, + "loss": 1.3667, + "step": 4521 + }, + { + "epoch": 0.38, + "learning_rate": 1.951146360259577e-05, + "loss": 1.2699, + "step": 4522 + }, + { + "epoch": 0.38, + "learning_rate": 1.9511182926648808e-05, + "loss": 1.31, + "step": 4523 + }, + { + "epoch": 0.38, + "learning_rate": 1.951090217211741e-05, + "loss": 1.4123, + "step": 4524 + }, + { + "epoch": 0.38, + "learning_rate": 1.951062133900389e-05, + "loss": 1.2948, + "step": 4525 + }, + { + "epoch": 0.38, + "learning_rate": 1.9510340427310576e-05, + "loss": 1.2823, + "step": 4526 + }, + { + "epoch": 0.38, + "learning_rate": 1.9510059437039784e-05, + "loss": 1.3184, + "step": 4527 + }, + { + "epoch": 0.38, + "learning_rate": 1.950977836819384e-05, + "loss": 1.368, + "step": 4528 + }, + { + "epoch": 0.38, + "learning_rate": 1.950949722077506e-05, + "loss": 1.3952, + "step": 4529 + }, + { + "epoch": 0.38, + "learning_rate": 1.9509215994785774e-05, + "loss": 1.3752, + "step": 4530 + }, + { + "epoch": 0.38, + "learning_rate": 1.95089346902283e-05, + "loss": 1.3383, + "step": 4531 + }, + { + "epoch": 0.38, + "learning_rate": 1.9508653307104965e-05, + "loss": 1.3751, + "step": 4532 + }, + { + "epoch": 0.38, + "learning_rate": 1.950837184541809e-05, + "loss": 1.4368, + "step": 4533 + }, + { + "epoch": 0.38, + "learning_rate": 1.950809030517001e-05, + "loss": 1.3107, + "step": 4534 + }, + { + "epoch": 0.38, + "learning_rate": 1.950780868636304e-05, + "loss": 1.3425, + "step": 4535 + }, + { + "epoch": 0.38, + "learning_rate": 1.9507526988999515e-05, + "loss": 1.362, + "step": 4536 + }, + { + "epoch": 0.38, + "learning_rate": 1.9507245213081755e-05, + "loss": 1.3089, + "step": 4537 + }, + { + "epoch": 0.38, + "learning_rate": 1.9506963358612093e-05, + "loss": 1.4504, + "step": 4538 + }, + { + "epoch": 0.38, + "learning_rate": 1.9506681425592862e-05, + "loss": 1.408, + "step": 4539 + }, + { + "epoch": 0.38, + "learning_rate": 1.9506399414026385e-05, + "loss": 1.3376, + "step": 4540 + }, + { + "epoch": 0.38, + "learning_rate": 1.950611732391499e-05, + "loss": 1.2312, + "step": 4541 + }, + { + "epoch": 0.38, + "learning_rate": 1.9505835155261015e-05, + "loss": 1.275, + "step": 4542 + }, + { + "epoch": 0.38, + "learning_rate": 1.9505552908066784e-05, + "loss": 1.3236, + "step": 4543 + }, + { + "epoch": 0.38, + "learning_rate": 1.9505270582334635e-05, + "loss": 1.2953, + "step": 4544 + }, + { + "epoch": 0.38, + "learning_rate": 1.95049881780669e-05, + "loss": 1.3299, + "step": 4545 + }, + { + "epoch": 0.38, + "learning_rate": 1.9504705695265907e-05, + "loss": 1.3865, + "step": 4546 + }, + { + "epoch": 0.38, + "learning_rate": 1.9504423133933997e-05, + "loss": 1.3198, + "step": 4547 + }, + { + "epoch": 0.38, + "learning_rate": 1.95041404940735e-05, + "loss": 1.3281, + "step": 4548 + }, + { + "epoch": 0.38, + "learning_rate": 1.9503857775686753e-05, + "loss": 1.2856, + "step": 4549 + }, + { + "epoch": 0.38, + "learning_rate": 1.950357497877609e-05, + "loss": 1.3549, + "step": 4550 + }, + { + "epoch": 0.38, + "learning_rate": 1.950329210334385e-05, + "loss": 1.3187, + "step": 4551 + }, + { + "epoch": 0.38, + "learning_rate": 1.950300914939237e-05, + "loss": 1.2493, + "step": 4552 + }, + { + "epoch": 0.38, + "learning_rate": 1.9502726116923984e-05, + "loss": 1.4179, + "step": 4553 + }, + { + "epoch": 0.38, + "learning_rate": 1.9502443005941037e-05, + "loss": 1.2589, + "step": 4554 + }, + { + "epoch": 0.38, + "learning_rate": 1.950215981644586e-05, + "loss": 1.308, + "step": 4555 + }, + { + "epoch": 0.38, + "learning_rate": 1.9501876548440804e-05, + "loss": 1.2935, + "step": 4556 + }, + { + "epoch": 0.38, + "learning_rate": 1.9501593201928196e-05, + "loss": 1.2997, + "step": 4557 + }, + { + "epoch": 0.38, + "learning_rate": 1.950130977691039e-05, + "loss": 1.3384, + "step": 4558 + }, + { + "epoch": 0.38, + "learning_rate": 1.950102627338972e-05, + "loss": 1.3278, + "step": 4559 + }, + { + "epoch": 0.38, + "learning_rate": 1.950074269136853e-05, + "loss": 1.3331, + "step": 4560 + }, + { + "epoch": 0.38, + "learning_rate": 1.9500459030849162e-05, + "loss": 1.3584, + "step": 4561 + }, + { + "epoch": 0.38, + "learning_rate": 1.9500175291833962e-05, + "loss": 1.2779, + "step": 4562 + }, + { + "epoch": 0.38, + "learning_rate": 1.9499891474325276e-05, + "loss": 1.3109, + "step": 4563 + }, + { + "epoch": 0.38, + "learning_rate": 1.9499607578325444e-05, + "loss": 1.3127, + "step": 4564 + }, + { + "epoch": 0.38, + "learning_rate": 1.9499323603836813e-05, + "loss": 1.2838, + "step": 4565 + }, + { + "epoch": 0.38, + "learning_rate": 1.949903955086173e-05, + "loss": 1.3402, + "step": 4566 + }, + { + "epoch": 0.38, + "learning_rate": 1.949875541940254e-05, + "loss": 1.3907, + "step": 4567 + }, + { + "epoch": 0.38, + "learning_rate": 1.94984712094616e-05, + "loss": 1.3951, + "step": 4568 + }, + { + "epoch": 0.38, + "learning_rate": 1.9498186921041246e-05, + "loss": 1.2856, + "step": 4569 + }, + { + "epoch": 0.38, + "learning_rate": 1.9497902554143832e-05, + "loss": 1.2787, + "step": 4570 + }, + { + "epoch": 0.38, + "learning_rate": 1.949761810877171e-05, + "loss": 1.3784, + "step": 4571 + }, + { + "epoch": 0.38, + "learning_rate": 1.9497333584927222e-05, + "loss": 1.3424, + "step": 4572 + }, + { + "epoch": 0.39, + "learning_rate": 1.9497048982612728e-05, + "loss": 1.3252, + "step": 4573 + }, + { + "epoch": 0.39, + "learning_rate": 1.9496764301830576e-05, + "loss": 1.4268, + "step": 4574 + }, + { + "epoch": 0.39, + "learning_rate": 1.9496479542583117e-05, + "loss": 1.3418, + "step": 4575 + }, + { + "epoch": 0.39, + "learning_rate": 1.9496194704872706e-05, + "loss": 1.2746, + "step": 4576 + }, + { + "epoch": 0.39, + "learning_rate": 1.9495909788701697e-05, + "loss": 1.3776, + "step": 4577 + }, + { + "epoch": 0.39, + "learning_rate": 1.949562479407244e-05, + "loss": 1.3984, + "step": 4578 + }, + { + "epoch": 0.39, + "learning_rate": 1.949533972098729e-05, + "loss": 1.4032, + "step": 4579 + }, + { + "epoch": 0.39, + "learning_rate": 1.9495054569448605e-05, + "loss": 1.2974, + "step": 4580 + }, + { + "epoch": 0.39, + "learning_rate": 1.9494769339458744e-05, + "loss": 1.4142, + "step": 4581 + }, + { + "epoch": 0.39, + "learning_rate": 1.9494484031020055e-05, + "loss": 1.3027, + "step": 4582 + }, + { + "epoch": 0.39, + "learning_rate": 1.9494198644134904e-05, + "loss": 1.3852, + "step": 4583 + }, + { + "epoch": 0.39, + "learning_rate": 1.9493913178805642e-05, + "loss": 1.2282, + "step": 4584 + }, + { + "epoch": 0.39, + "learning_rate": 1.9493627635034634e-05, + "loss": 1.3311, + "step": 4585 + }, + { + "epoch": 0.39, + "learning_rate": 1.9493342012824238e-05, + "loss": 1.3363, + "step": 4586 + }, + { + "epoch": 0.39, + "learning_rate": 1.949305631217681e-05, + "loss": 1.3804, + "step": 4587 + }, + { + "epoch": 0.39, + "learning_rate": 1.9492770533094706e-05, + "loss": 1.238, + "step": 4588 + }, + { + "epoch": 0.39, + "learning_rate": 1.9492484675580302e-05, + "loss": 1.2851, + "step": 4589 + }, + { + "epoch": 0.39, + "learning_rate": 1.9492198739635947e-05, + "loss": 1.2882, + "step": 4590 + }, + { + "epoch": 0.39, + "learning_rate": 1.9491912725264014e-05, + "loss": 1.3461, + "step": 4591 + }, + { + "epoch": 0.39, + "learning_rate": 1.9491626632466857e-05, + "loss": 1.4691, + "step": 4592 + }, + { + "epoch": 0.39, + "learning_rate": 1.949134046124684e-05, + "loss": 1.2535, + "step": 4593 + }, + { + "epoch": 0.39, + "learning_rate": 1.9491054211606335e-05, + "loss": 1.3289, + "step": 4594 + }, + { + "epoch": 0.39, + "learning_rate": 1.94907678835477e-05, + "loss": 1.2901, + "step": 4595 + }, + { + "epoch": 0.39, + "learning_rate": 1.9490481477073304e-05, + "loss": 1.4078, + "step": 4596 + }, + { + "epoch": 0.39, + "learning_rate": 1.9490194992185513e-05, + "loss": 1.3974, + "step": 4597 + }, + { + "epoch": 0.39, + "learning_rate": 1.948990842888669e-05, + "loss": 1.1974, + "step": 4598 + }, + { + "epoch": 0.39, + "learning_rate": 1.9489621787179212e-05, + "loss": 1.3539, + "step": 4599 + }, + { + "epoch": 0.39, + "learning_rate": 1.948933506706544e-05, + "loss": 1.2996, + "step": 4600 + }, + { + "epoch": 0.39, + "learning_rate": 1.9489048268547744e-05, + "loss": 1.4007, + "step": 4601 + }, + { + "epoch": 0.39, + "learning_rate": 1.948876139162849e-05, + "loss": 1.2496, + "step": 4602 + }, + { + "epoch": 0.39, + "learning_rate": 1.948847443631006e-05, + "loss": 1.3213, + "step": 4603 + }, + { + "epoch": 0.39, + "learning_rate": 1.9488187402594816e-05, + "loss": 1.2306, + "step": 4604 + }, + { + "epoch": 0.39, + "learning_rate": 1.9487900290485128e-05, + "loss": 1.3636, + "step": 4605 + }, + { + "epoch": 0.39, + "learning_rate": 1.9487613099983375e-05, + "loss": 1.3035, + "step": 4606 + }, + { + "epoch": 0.39, + "learning_rate": 1.9487325831091923e-05, + "loss": 1.2693, + "step": 4607 + }, + { + "epoch": 0.39, + "learning_rate": 1.948703848381315e-05, + "loss": 1.2135, + "step": 4608 + }, + { + "epoch": 0.39, + "learning_rate": 1.948675105814943e-05, + "loss": 1.33, + "step": 4609 + }, + { + "epoch": 0.39, + "learning_rate": 1.9486463554103136e-05, + "loss": 1.3146, + "step": 4610 + }, + { + "epoch": 0.39, + "learning_rate": 1.948617597167664e-05, + "loss": 1.3643, + "step": 4611 + }, + { + "epoch": 0.39, + "learning_rate": 1.9485888310872326e-05, + "loss": 1.4396, + "step": 4612 + }, + { + "epoch": 0.39, + "learning_rate": 1.9485600571692567e-05, + "loss": 1.2397, + "step": 4613 + }, + { + "epoch": 0.39, + "learning_rate": 1.948531275413974e-05, + "loss": 1.3817, + "step": 4614 + }, + { + "epoch": 0.39, + "learning_rate": 1.948502485821622e-05, + "loss": 1.4398, + "step": 4615 + }, + { + "epoch": 0.39, + "learning_rate": 1.948473688392439e-05, + "loss": 1.2554, + "step": 4616 + }, + { + "epoch": 0.39, + "learning_rate": 1.9484448831266632e-05, + "loss": 1.2772, + "step": 4617 + }, + { + "epoch": 0.39, + "learning_rate": 1.9484160700245322e-05, + "loss": 1.3254, + "step": 4618 + }, + { + "epoch": 0.39, + "learning_rate": 1.9483872490862842e-05, + "loss": 1.2875, + "step": 4619 + }, + { + "epoch": 0.39, + "learning_rate": 1.9483584203121568e-05, + "loss": 1.2476, + "step": 4620 + }, + { + "epoch": 0.39, + "learning_rate": 1.9483295837023888e-05, + "loss": 1.3136, + "step": 4621 + }, + { + "epoch": 0.39, + "learning_rate": 1.948300739257218e-05, + "loss": 1.3584, + "step": 4622 + }, + { + "epoch": 0.39, + "learning_rate": 1.9482718869768834e-05, + "loss": 1.4104, + "step": 4623 + }, + { + "epoch": 0.39, + "learning_rate": 1.9482430268616232e-05, + "loss": 1.3705, + "step": 4624 + }, + { + "epoch": 0.39, + "learning_rate": 1.948214158911675e-05, + "loss": 1.3619, + "step": 4625 + }, + { + "epoch": 0.39, + "learning_rate": 1.9481852831272784e-05, + "loss": 1.2637, + "step": 4626 + }, + { + "epoch": 0.39, + "learning_rate": 1.9481563995086712e-05, + "loss": 1.3069, + "step": 4627 + }, + { + "epoch": 0.39, + "learning_rate": 1.9481275080560925e-05, + "loss": 1.3197, + "step": 4628 + }, + { + "epoch": 0.39, + "learning_rate": 1.948098608769781e-05, + "loss": 1.3321, + "step": 4629 + }, + { + "epoch": 0.39, + "learning_rate": 1.9480697016499754e-05, + "loss": 1.3212, + "step": 4630 + }, + { + "epoch": 0.39, + "learning_rate": 1.9480407866969145e-05, + "loss": 1.2762, + "step": 4631 + }, + { + "epoch": 0.39, + "learning_rate": 1.948011863910837e-05, + "loss": 1.2997, + "step": 4632 + }, + { + "epoch": 0.39, + "learning_rate": 1.947982933291982e-05, + "loss": 1.2895, + "step": 4633 + }, + { + "epoch": 0.39, + "learning_rate": 1.947953994840589e-05, + "loss": 1.2581, + "step": 4634 + }, + { + "epoch": 0.39, + "learning_rate": 1.9479250485568967e-05, + "loss": 1.3584, + "step": 4635 + }, + { + "epoch": 0.39, + "learning_rate": 1.947896094441144e-05, + "loss": 1.3381, + "step": 4636 + }, + { + "epoch": 0.39, + "learning_rate": 1.9478671324935704e-05, + "loss": 1.4483, + "step": 4637 + }, + { + "epoch": 0.39, + "learning_rate": 1.9478381627144154e-05, + "loss": 1.3095, + "step": 4638 + }, + { + "epoch": 0.39, + "learning_rate": 1.947809185103918e-05, + "loss": 1.3018, + "step": 4639 + }, + { + "epoch": 0.39, + "learning_rate": 1.947780199662318e-05, + "loss": 1.3814, + "step": 4640 + }, + { + "epoch": 0.39, + "learning_rate": 1.9477512063898543e-05, + "loss": 1.311, + "step": 4641 + }, + { + "epoch": 0.39, + "learning_rate": 1.947722205286767e-05, + "loss": 1.3299, + "step": 4642 + }, + { + "epoch": 0.39, + "learning_rate": 1.9476931963532955e-05, + "loss": 1.2517, + "step": 4643 + }, + { + "epoch": 0.39, + "learning_rate": 1.9476641795896794e-05, + "loss": 1.2749, + "step": 4644 + }, + { + "epoch": 0.39, + "learning_rate": 1.9476351549961587e-05, + "loss": 1.3819, + "step": 4645 + }, + { + "epoch": 0.39, + "learning_rate": 1.947606122572973e-05, + "loss": 1.3425, + "step": 4646 + }, + { + "epoch": 0.39, + "learning_rate": 1.9475770823203626e-05, + "loss": 1.2842, + "step": 4647 + }, + { + "epoch": 0.39, + "learning_rate": 1.9475480342385668e-05, + "loss": 1.328, + "step": 4648 + }, + { + "epoch": 0.39, + "learning_rate": 1.9475189783278257e-05, + "loss": 1.3142, + "step": 4649 + }, + { + "epoch": 0.39, + "learning_rate": 1.94748991458838e-05, + "loss": 1.2433, + "step": 4650 + }, + { + "epoch": 0.39, + "learning_rate": 1.9474608430204687e-05, + "loss": 1.329, + "step": 4651 + }, + { + "epoch": 0.39, + "learning_rate": 1.947431763624333e-05, + "loss": 1.2797, + "step": 4652 + }, + { + "epoch": 0.39, + "learning_rate": 1.947402676400213e-05, + "loss": 1.2991, + "step": 4653 + }, + { + "epoch": 0.39, + "learning_rate": 1.947373581348349e-05, + "loss": 1.227, + "step": 4654 + }, + { + "epoch": 0.39, + "learning_rate": 1.9473444784689814e-05, + "loss": 1.2712, + "step": 4655 + }, + { + "epoch": 0.39, + "learning_rate": 1.9473153677623498e-05, + "loss": 1.3564, + "step": 4656 + }, + { + "epoch": 0.39, + "learning_rate": 1.947286249228696e-05, + "loss": 1.3765, + "step": 4657 + }, + { + "epoch": 0.39, + "learning_rate": 1.94725712286826e-05, + "loss": 1.3217, + "step": 4658 + }, + { + "epoch": 0.39, + "learning_rate": 1.9472279886812822e-05, + "loss": 1.3712, + "step": 4659 + }, + { + "epoch": 0.39, + "learning_rate": 1.947198846668004e-05, + "loss": 1.4052, + "step": 4660 + }, + { + "epoch": 0.39, + "learning_rate": 1.9471696968286657e-05, + "loss": 1.1824, + "step": 4661 + }, + { + "epoch": 0.39, + "learning_rate": 1.947140539163508e-05, + "loss": 1.2974, + "step": 4662 + }, + { + "epoch": 0.39, + "learning_rate": 1.9471113736727722e-05, + "loss": 1.3127, + "step": 4663 + }, + { + "epoch": 0.39, + "learning_rate": 1.947082200356699e-05, + "loss": 1.2668, + "step": 4664 + }, + { + "epoch": 0.39, + "learning_rate": 1.9470530192155297e-05, + "loss": 1.2735, + "step": 4665 + }, + { + "epoch": 0.39, + "learning_rate": 1.9470238302495055e-05, + "loss": 1.2879, + "step": 4666 + }, + { + "epoch": 0.39, + "learning_rate": 1.946994633458867e-05, + "loss": 1.3767, + "step": 4667 + }, + { + "epoch": 0.39, + "learning_rate": 1.9469654288438556e-05, + "loss": 1.3371, + "step": 4668 + }, + { + "epoch": 0.39, + "learning_rate": 1.946936216404713e-05, + "loss": 1.3288, + "step": 4669 + }, + { + "epoch": 0.39, + "learning_rate": 1.9469069961416803e-05, + "loss": 1.3729, + "step": 4670 + }, + { + "epoch": 0.39, + "learning_rate": 1.946877768054999e-05, + "loss": 1.3376, + "step": 4671 + }, + { + "epoch": 0.39, + "learning_rate": 1.9468485321449106e-05, + "loss": 1.3365, + "step": 4672 + }, + { + "epoch": 0.39, + "learning_rate": 1.9468192884116564e-05, + "loss": 1.2461, + "step": 4673 + }, + { + "epoch": 0.39, + "learning_rate": 1.9467900368554783e-05, + "loss": 1.2521, + "step": 4674 + }, + { + "epoch": 0.39, + "learning_rate": 1.946760777476618e-05, + "loss": 1.3235, + "step": 4675 + }, + { + "epoch": 0.39, + "learning_rate": 1.9467315102753167e-05, + "loss": 1.4112, + "step": 4676 + }, + { + "epoch": 0.39, + "learning_rate": 1.946702235251817e-05, + "loss": 1.203, + "step": 4677 + }, + { + "epoch": 0.39, + "learning_rate": 1.9466729524063605e-05, + "loss": 1.2752, + "step": 4678 + }, + { + "epoch": 0.39, + "learning_rate": 1.9466436617391892e-05, + "loss": 1.2837, + "step": 4679 + }, + { + "epoch": 0.39, + "learning_rate": 1.946614363250545e-05, + "loss": 1.3708, + "step": 4680 + }, + { + "epoch": 0.39, + "learning_rate": 1.9465850569406697e-05, + "loss": 1.3989, + "step": 4681 + }, + { + "epoch": 0.39, + "learning_rate": 1.946555742809806e-05, + "loss": 1.2966, + "step": 4682 + }, + { + "epoch": 0.39, + "learning_rate": 1.9465264208581954e-05, + "loss": 1.3099, + "step": 4683 + }, + { + "epoch": 0.39, + "learning_rate": 1.946497091086081e-05, + "loss": 1.3251, + "step": 4684 + }, + { + "epoch": 0.39, + "learning_rate": 1.9464677534937045e-05, + "loss": 1.2634, + "step": 4685 + }, + { + "epoch": 0.39, + "learning_rate": 1.9464384080813085e-05, + "loss": 1.381, + "step": 4686 + }, + { + "epoch": 0.39, + "learning_rate": 1.9464090548491355e-05, + "loss": 1.2819, + "step": 4687 + }, + { + "epoch": 0.39, + "learning_rate": 1.946379693797428e-05, + "loss": 1.3782, + "step": 4688 + }, + { + "epoch": 0.39, + "learning_rate": 1.946350324926429e-05, + "loss": 1.3609, + "step": 4689 + }, + { + "epoch": 0.39, + "learning_rate": 1.9463209482363805e-05, + "loss": 1.3819, + "step": 4690 + }, + { + "epoch": 0.39, + "learning_rate": 1.9462915637275254e-05, + "loss": 1.5435, + "step": 4691 + }, + { + "epoch": 0.4, + "learning_rate": 1.9462621714001063e-05, + "loss": 1.2877, + "step": 4692 + }, + { + "epoch": 0.4, + "learning_rate": 1.9462327712543664e-05, + "loss": 1.2605, + "step": 4693 + }, + { + "epoch": 0.4, + "learning_rate": 1.9462033632905486e-05, + "loss": 1.3148, + "step": 4694 + }, + { + "epoch": 0.4, + "learning_rate": 1.946173947508896e-05, + "loss": 1.3223, + "step": 4695 + }, + { + "epoch": 0.4, + "learning_rate": 1.946144523909651e-05, + "loss": 1.2717, + "step": 4696 + }, + { + "epoch": 0.4, + "learning_rate": 1.9461150924930575e-05, + "loss": 1.3726, + "step": 4697 + }, + { + "epoch": 0.4, + "learning_rate": 1.946085653259358e-05, + "loss": 1.2697, + "step": 4698 + }, + { + "epoch": 0.4, + "learning_rate": 1.9460562062087963e-05, + "loss": 1.3705, + "step": 4699 + }, + { + "epoch": 0.4, + "learning_rate": 1.9460267513416153e-05, + "loss": 1.3217, + "step": 4700 + }, + { + "epoch": 0.4, + "learning_rate": 1.9459972886580587e-05, + "loss": 1.2925, + "step": 4701 + }, + { + "epoch": 0.4, + "learning_rate": 1.9459678181583697e-05, + "loss": 1.3057, + "step": 4702 + }, + { + "epoch": 0.4, + "learning_rate": 1.9459383398427916e-05, + "loss": 1.342, + "step": 4703 + }, + { + "epoch": 0.4, + "learning_rate": 1.9459088537115686e-05, + "loss": 1.3111, + "step": 4704 + }, + { + "epoch": 0.4, + "learning_rate": 1.9458793597649434e-05, + "loss": 1.3319, + "step": 4705 + }, + { + "epoch": 0.4, + "learning_rate": 1.9458498580031606e-05, + "loss": 1.3352, + "step": 4706 + }, + { + "epoch": 0.4, + "learning_rate": 1.9458203484264634e-05, + "loss": 1.4044, + "step": 4707 + }, + { + "epoch": 0.4, + "learning_rate": 1.945790831035096e-05, + "loss": 1.3685, + "step": 4708 + }, + { + "epoch": 0.4, + "learning_rate": 1.9457613058293018e-05, + "loss": 1.3089, + "step": 4709 + }, + { + "epoch": 0.4, + "learning_rate": 1.945731772809325e-05, + "loss": 1.3664, + "step": 4710 + }, + { + "epoch": 0.4, + "learning_rate": 1.9457022319754098e-05, + "loss": 1.4201, + "step": 4711 + }, + { + "epoch": 0.4, + "learning_rate": 1.9456726833278e-05, + "loss": 1.3286, + "step": 4712 + }, + { + "epoch": 0.4, + "learning_rate": 1.94564312686674e-05, + "loss": 1.3386, + "step": 4713 + }, + { + "epoch": 0.4, + "learning_rate": 1.9456135625924735e-05, + "loss": 1.3572, + "step": 4714 + }, + { + "epoch": 0.4, + "learning_rate": 1.9455839905052452e-05, + "loss": 1.348, + "step": 4715 + }, + { + "epoch": 0.4, + "learning_rate": 1.9455544106052998e-05, + "loss": 1.2994, + "step": 4716 + }, + { + "epoch": 0.4, + "learning_rate": 1.945524822892881e-05, + "loss": 1.2221, + "step": 4717 + }, + { + "epoch": 0.4, + "learning_rate": 1.945495227368233e-05, + "loss": 1.2787, + "step": 4718 + }, + { + "epoch": 0.4, + "learning_rate": 1.9454656240316012e-05, + "loss": 1.3562, + "step": 4719 + }, + { + "epoch": 0.4, + "learning_rate": 1.94543601288323e-05, + "loss": 1.3132, + "step": 4720 + }, + { + "epoch": 0.4, + "learning_rate": 1.9454063939233635e-05, + "loss": 1.2962, + "step": 4721 + }, + { + "epoch": 0.4, + "learning_rate": 1.9453767671522472e-05, + "loss": 1.3342, + "step": 4722 + }, + { + "epoch": 0.4, + "learning_rate": 1.9453471325701255e-05, + "loss": 1.341, + "step": 4723 + }, + { + "epoch": 0.4, + "learning_rate": 1.945317490177243e-05, + "loss": 1.4059, + "step": 4724 + }, + { + "epoch": 0.4, + "learning_rate": 1.9452878399738447e-05, + "loss": 1.4172, + "step": 4725 + }, + { + "epoch": 0.4, + "learning_rate": 1.945258181960176e-05, + "loss": 1.2576, + "step": 4726 + }, + { + "epoch": 0.4, + "learning_rate": 1.945228516136482e-05, + "loss": 1.3801, + "step": 4727 + }, + { + "epoch": 0.4, + "learning_rate": 1.945198842503007e-05, + "loss": 1.378, + "step": 4728 + }, + { + "epoch": 0.4, + "learning_rate": 1.9451691610599967e-05, + "loss": 1.4452, + "step": 4729 + }, + { + "epoch": 0.4, + "learning_rate": 1.9451394718076963e-05, + "loss": 1.2382, + "step": 4730 + }, + { + "epoch": 0.4, + "learning_rate": 1.9451097747463512e-05, + "loss": 1.3447, + "step": 4731 + }, + { + "epoch": 0.4, + "learning_rate": 1.9450800698762066e-05, + "loss": 1.2566, + "step": 4732 + }, + { + "epoch": 0.4, + "learning_rate": 1.9450503571975082e-05, + "loss": 1.3213, + "step": 4733 + }, + { + "epoch": 0.4, + "learning_rate": 1.9450206367105013e-05, + "loss": 1.2821, + "step": 4734 + }, + { + "epoch": 0.4, + "learning_rate": 1.9449909084154312e-05, + "loss": 1.3483, + "step": 4735 + }, + { + "epoch": 0.4, + "learning_rate": 1.9449611723125442e-05, + "loss": 1.2902, + "step": 4736 + }, + { + "epoch": 0.4, + "learning_rate": 1.944931428402085e-05, + "loss": 1.3733, + "step": 4737 + }, + { + "epoch": 0.4, + "learning_rate": 1.9449016766843e-05, + "loss": 1.3439, + "step": 4738 + }, + { + "epoch": 0.4, + "learning_rate": 1.944871917159435e-05, + "loss": 1.3155, + "step": 4739 + }, + { + "epoch": 0.4, + "learning_rate": 1.9448421498277362e-05, + "loss": 1.222, + "step": 4740 + }, + { + "epoch": 0.4, + "learning_rate": 1.944812374689449e-05, + "loss": 1.414, + "step": 4741 + }, + { + "epoch": 0.4, + "learning_rate": 1.9447825917448194e-05, + "loss": 1.2737, + "step": 4742 + }, + { + "epoch": 0.4, + "learning_rate": 1.9447528009940938e-05, + "loss": 1.28, + "step": 4743 + }, + { + "epoch": 0.4, + "learning_rate": 1.944723002437518e-05, + "loss": 1.3506, + "step": 4744 + }, + { + "epoch": 0.4, + "learning_rate": 1.944693196075339e-05, + "loss": 1.3524, + "step": 4745 + }, + { + "epoch": 0.4, + "learning_rate": 1.944663381907802e-05, + "loss": 1.313, + "step": 4746 + }, + { + "epoch": 0.4, + "learning_rate": 1.9446335599351534e-05, + "loss": 1.2972, + "step": 4747 + }, + { + "epoch": 0.4, + "learning_rate": 1.9446037301576405e-05, + "loss": 1.2757, + "step": 4748 + }, + { + "epoch": 0.4, + "learning_rate": 1.9445738925755095e-05, + "loss": 1.4253, + "step": 4749 + }, + { + "epoch": 0.4, + "learning_rate": 1.9445440471890062e-05, + "loss": 1.3086, + "step": 4750 + }, + { + "epoch": 0.4, + "learning_rate": 1.9445141939983778e-05, + "loss": 1.2921, + "step": 4751 + }, + { + "epoch": 0.4, + "learning_rate": 1.944484333003871e-05, + "loss": 1.3336, + "step": 4752 + }, + { + "epoch": 0.4, + "learning_rate": 1.9444544642057327e-05, + "loss": 1.3341, + "step": 4753 + }, + { + "epoch": 0.4, + "learning_rate": 1.9444245876042087e-05, + "loss": 1.3926, + "step": 4754 + }, + { + "epoch": 0.4, + "learning_rate": 1.9443947031995467e-05, + "loss": 1.299, + "step": 4755 + }, + { + "epoch": 0.4, + "learning_rate": 1.9443648109919936e-05, + "loss": 1.3064, + "step": 4756 + }, + { + "epoch": 0.4, + "learning_rate": 1.944334910981796e-05, + "loss": 1.3485, + "step": 4757 + }, + { + "epoch": 0.4, + "learning_rate": 1.9443050031692012e-05, + "loss": 1.3684, + "step": 4758 + }, + { + "epoch": 0.4, + "learning_rate": 1.9442750875544564e-05, + "loss": 1.3658, + "step": 4759 + }, + { + "epoch": 0.4, + "learning_rate": 1.9442451641378082e-05, + "loss": 1.3591, + "step": 4760 + }, + { + "epoch": 0.4, + "learning_rate": 1.944215232919505e-05, + "loss": 1.4858, + "step": 4761 + }, + { + "epoch": 0.4, + "learning_rate": 1.9441852938997926e-05, + "loss": 1.3165, + "step": 4762 + }, + { + "epoch": 0.4, + "learning_rate": 1.9441553470789194e-05, + "loss": 1.3638, + "step": 4763 + }, + { + "epoch": 0.4, + "learning_rate": 1.9441253924571327e-05, + "loss": 1.3785, + "step": 4764 + }, + { + "epoch": 0.4, + "learning_rate": 1.9440954300346797e-05, + "loss": 1.3533, + "step": 4765 + }, + { + "epoch": 0.4, + "learning_rate": 1.944065459811808e-05, + "loss": 1.4231, + "step": 4766 + }, + { + "epoch": 0.4, + "learning_rate": 1.9440354817887653e-05, + "loss": 1.3771, + "step": 4767 + }, + { + "epoch": 0.4, + "learning_rate": 1.9440054959658e-05, + "loss": 1.2958, + "step": 4768 + }, + { + "epoch": 0.4, + "learning_rate": 1.9439755023431584e-05, + "loss": 1.355, + "step": 4769 + }, + { + "epoch": 0.4, + "learning_rate": 1.9439455009210893e-05, + "loss": 1.3076, + "step": 4770 + }, + { + "epoch": 0.4, + "learning_rate": 1.9439154916998407e-05, + "loss": 1.3161, + "step": 4771 + }, + { + "epoch": 0.4, + "learning_rate": 1.9438854746796595e-05, + "loss": 1.2986, + "step": 4772 + }, + { + "epoch": 0.4, + "learning_rate": 1.943855449860795e-05, + "loss": 1.403, + "step": 4773 + }, + { + "epoch": 0.4, + "learning_rate": 1.9438254172434946e-05, + "loss": 1.3765, + "step": 4774 + }, + { + "epoch": 0.4, + "learning_rate": 1.9437953768280063e-05, + "loss": 1.3556, + "step": 4775 + }, + { + "epoch": 0.4, + "learning_rate": 1.9437653286145792e-05, + "loss": 1.3358, + "step": 4776 + }, + { + "epoch": 0.4, + "learning_rate": 1.9437352726034602e-05, + "loss": 1.3038, + "step": 4777 + }, + { + "epoch": 0.4, + "learning_rate": 1.9437052087948985e-05, + "loss": 1.3522, + "step": 4778 + }, + { + "epoch": 0.4, + "learning_rate": 1.9436751371891428e-05, + "loss": 1.3532, + "step": 4779 + }, + { + "epoch": 0.4, + "learning_rate": 1.9436450577864406e-05, + "loss": 1.3949, + "step": 4780 + }, + { + "epoch": 0.4, + "learning_rate": 1.943614970587041e-05, + "loss": 1.3502, + "step": 4781 + }, + { + "epoch": 0.4, + "learning_rate": 1.9435848755911927e-05, + "loss": 1.3297, + "step": 4782 + }, + { + "epoch": 0.4, + "learning_rate": 1.943554772799144e-05, + "loss": 1.1975, + "step": 4783 + }, + { + "epoch": 0.4, + "learning_rate": 1.9435246622111438e-05, + "loss": 1.1986, + "step": 4784 + }, + { + "epoch": 0.4, + "learning_rate": 1.943494543827441e-05, + "loss": 1.3045, + "step": 4785 + }, + { + "epoch": 0.4, + "learning_rate": 1.943464417648284e-05, + "loss": 1.3405, + "step": 4786 + }, + { + "epoch": 0.4, + "learning_rate": 1.9434342836739226e-05, + "loss": 1.2882, + "step": 4787 + }, + { + "epoch": 0.4, + "learning_rate": 1.9434041419046047e-05, + "loss": 1.3964, + "step": 4788 + }, + { + "epoch": 0.4, + "learning_rate": 1.94337399234058e-05, + "loss": 1.329, + "step": 4789 + }, + { + "epoch": 0.4, + "learning_rate": 1.9433438349820975e-05, + "loss": 1.345, + "step": 4790 + }, + { + "epoch": 0.4, + "learning_rate": 1.9433136698294064e-05, + "loss": 1.3271, + "step": 4791 + }, + { + "epoch": 0.4, + "learning_rate": 1.9432834968827558e-05, + "loss": 1.3217, + "step": 4792 + }, + { + "epoch": 0.4, + "learning_rate": 1.943253316142395e-05, + "loss": 1.3554, + "step": 4793 + }, + { + "epoch": 0.4, + "learning_rate": 1.9432231276085735e-05, + "loss": 1.3946, + "step": 4794 + }, + { + "epoch": 0.4, + "learning_rate": 1.943192931281541e-05, + "loss": 1.3707, + "step": 4795 + }, + { + "epoch": 0.4, + "learning_rate": 1.943162727161546e-05, + "loss": 1.3573, + "step": 4796 + }, + { + "epoch": 0.4, + "learning_rate": 1.9431325152488387e-05, + "loss": 1.3259, + "step": 4797 + }, + { + "epoch": 0.4, + "learning_rate": 1.9431022955436692e-05, + "loss": 1.2489, + "step": 4798 + }, + { + "epoch": 0.4, + "learning_rate": 1.9430720680462866e-05, + "loss": 1.2904, + "step": 4799 + }, + { + "epoch": 0.4, + "learning_rate": 1.9430418327569405e-05, + "loss": 1.2903, + "step": 4800 + }, + { + "epoch": 0.4, + "learning_rate": 1.943011589675881e-05, + "loss": 1.2404, + "step": 4801 + }, + { + "epoch": 0.4, + "learning_rate": 1.942981338803358e-05, + "loss": 1.3383, + "step": 4802 + }, + { + "epoch": 0.4, + "learning_rate": 1.9429510801396216e-05, + "loss": 1.4096, + "step": 4803 + }, + { + "epoch": 0.4, + "learning_rate": 1.9429208136849213e-05, + "loss": 1.2548, + "step": 4804 + }, + { + "epoch": 0.4, + "learning_rate": 1.9428905394395078e-05, + "loss": 1.3466, + "step": 4805 + }, + { + "epoch": 0.4, + "learning_rate": 1.9428602574036306e-05, + "loss": 1.2954, + "step": 4806 + }, + { + "epoch": 0.4, + "learning_rate": 1.94282996757754e-05, + "loss": 1.3795, + "step": 4807 + }, + { + "epoch": 0.4, + "learning_rate": 1.942799669961487e-05, + "loss": 1.3406, + "step": 4808 + }, + { + "epoch": 0.4, + "learning_rate": 1.942769364555721e-05, + "loss": 1.3767, + "step": 4809 + }, + { + "epoch": 0.4, + "learning_rate": 1.942739051360493e-05, + "loss": 1.3076, + "step": 4810 + }, + { + "epoch": 0.41, + "learning_rate": 1.9427087303760533e-05, + "loss": 1.3693, + "step": 4811 + }, + { + "epoch": 0.41, + "learning_rate": 1.9426784016026522e-05, + "loss": 1.2701, + "step": 4812 + }, + { + "epoch": 0.41, + "learning_rate": 1.9426480650405405e-05, + "loss": 1.4783, + "step": 4813 + }, + { + "epoch": 0.41, + "learning_rate": 1.9426177206899687e-05, + "loss": 1.3151, + "step": 4814 + }, + { + "epoch": 0.41, + "learning_rate": 1.9425873685511875e-05, + "loss": 1.3114, + "step": 4815 + }, + { + "epoch": 0.41, + "learning_rate": 1.9425570086244482e-05, + "loss": 1.407, + "step": 4816 + }, + { + "epoch": 0.41, + "learning_rate": 1.942526640910001e-05, + "loss": 1.285, + "step": 4817 + }, + { + "epoch": 0.41, + "learning_rate": 1.942496265408097e-05, + "loss": 1.2949, + "step": 4818 + }, + { + "epoch": 0.41, + "learning_rate": 1.9424658821189873e-05, + "loss": 1.3473, + "step": 4819 + }, + { + "epoch": 0.41, + "learning_rate": 1.9424354910429227e-05, + "loss": 1.2365, + "step": 4820 + }, + { + "epoch": 0.41, + "learning_rate": 1.9424050921801548e-05, + "loss": 1.3249, + "step": 4821 + }, + { + "epoch": 0.41, + "learning_rate": 1.942374685530934e-05, + "loss": 1.2988, + "step": 4822 + }, + { + "epoch": 0.41, + "learning_rate": 1.942344271095512e-05, + "loss": 1.3689, + "step": 4823 + }, + { + "epoch": 0.41, + "learning_rate": 1.9423138488741403e-05, + "loss": 1.4124, + "step": 4824 + }, + { + "epoch": 0.41, + "learning_rate": 1.9422834188670703e-05, + "loss": 1.4041, + "step": 4825 + }, + { + "epoch": 0.41, + "learning_rate": 1.9422529810745526e-05, + "loss": 1.3847, + "step": 4826 + }, + { + "epoch": 0.41, + "learning_rate": 1.9422225354968395e-05, + "loss": 1.2274, + "step": 4827 + }, + { + "epoch": 0.41, + "learning_rate": 1.942192082134182e-05, + "loss": 1.3276, + "step": 4828 + }, + { + "epoch": 0.41, + "learning_rate": 1.9421616209868318e-05, + "loss": 1.4241, + "step": 4829 + }, + { + "epoch": 0.41, + "learning_rate": 1.9421311520550413e-05, + "loss": 1.3951, + "step": 4830 + }, + { + "epoch": 0.41, + "learning_rate": 1.9421006753390615e-05, + "loss": 1.3438, + "step": 4831 + }, + { + "epoch": 0.41, + "learning_rate": 1.9420701908391443e-05, + "loss": 1.2914, + "step": 4832 + }, + { + "epoch": 0.41, + "learning_rate": 1.9420396985555418e-05, + "loss": 1.3277, + "step": 4833 + }, + { + "epoch": 0.41, + "learning_rate": 1.9420091984885058e-05, + "loss": 1.3965, + "step": 4834 + }, + { + "epoch": 0.41, + "learning_rate": 1.9419786906382884e-05, + "loss": 1.3479, + "step": 4835 + }, + { + "epoch": 0.41, + "learning_rate": 1.9419481750051413e-05, + "loss": 1.3461, + "step": 4836 + }, + { + "epoch": 0.41, + "learning_rate": 1.941917651589317e-05, + "loss": 1.2149, + "step": 4837 + }, + { + "epoch": 0.41, + "learning_rate": 1.941887120391068e-05, + "loss": 1.397, + "step": 4838 + }, + { + "epoch": 0.41, + "learning_rate": 1.9418565814106457e-05, + "loss": 1.3465, + "step": 4839 + }, + { + "epoch": 0.41, + "learning_rate": 1.941826034648303e-05, + "loss": 1.4209, + "step": 4840 + }, + { + "epoch": 0.41, + "learning_rate": 1.941795480104292e-05, + "loss": 1.3412, + "step": 4841 + }, + { + "epoch": 0.41, + "learning_rate": 1.9417649177788657e-05, + "loss": 1.2754, + "step": 4842 + }, + { + "epoch": 0.41, + "learning_rate": 1.9417343476722762e-05, + "loss": 1.2868, + "step": 4843 + }, + { + "epoch": 0.41, + "learning_rate": 1.941703769784776e-05, + "loss": 1.2697, + "step": 4844 + }, + { + "epoch": 0.41, + "learning_rate": 1.9416731841166178e-05, + "loss": 1.309, + "step": 4845 + }, + { + "epoch": 0.41, + "learning_rate": 1.9416425906680544e-05, + "loss": 1.2352, + "step": 4846 + }, + { + "epoch": 0.41, + "learning_rate": 1.941611989439339e-05, + "loss": 1.4041, + "step": 4847 + }, + { + "epoch": 0.41, + "learning_rate": 1.9415813804307233e-05, + "loss": 1.2265, + "step": 4848 + }, + { + "epoch": 0.41, + "learning_rate": 1.9415507636424612e-05, + "loss": 1.3135, + "step": 4849 + }, + { + "epoch": 0.41, + "learning_rate": 1.9415201390748055e-05, + "loss": 1.3739, + "step": 4850 + }, + { + "epoch": 0.41, + "learning_rate": 1.9414895067280087e-05, + "loss": 1.3773, + "step": 4851 + }, + { + "epoch": 0.41, + "learning_rate": 1.9414588666023246e-05, + "loss": 1.3298, + "step": 4852 + }, + { + "epoch": 0.41, + "learning_rate": 1.9414282186980062e-05, + "loss": 1.3001, + "step": 4853 + }, + { + "epoch": 0.41, + "learning_rate": 1.9413975630153064e-05, + "loss": 1.3262, + "step": 4854 + }, + { + "epoch": 0.41, + "learning_rate": 1.9413668995544785e-05, + "loss": 1.3481, + "step": 4855 + }, + { + "epoch": 0.41, + "learning_rate": 1.941336228315776e-05, + "loss": 1.3968, + "step": 4856 + }, + { + "epoch": 0.41, + "learning_rate": 1.9413055492994526e-05, + "loss": 1.3191, + "step": 4857 + }, + { + "epoch": 0.41, + "learning_rate": 1.9412748625057612e-05, + "loss": 1.3201, + "step": 4858 + }, + { + "epoch": 0.41, + "learning_rate": 1.9412441679349558e-05, + "loss": 1.3832, + "step": 4859 + }, + { + "epoch": 0.41, + "learning_rate": 1.9412134655872895e-05, + "loss": 1.3544, + "step": 4860 + }, + { + "epoch": 0.41, + "learning_rate": 1.9411827554630167e-05, + "loss": 1.283, + "step": 4861 + }, + { + "epoch": 0.41, + "learning_rate": 1.9411520375623906e-05, + "loss": 1.3089, + "step": 4862 + }, + { + "epoch": 0.41, + "learning_rate": 1.9411213118856653e-05, + "loss": 1.2148, + "step": 4863 + }, + { + "epoch": 0.41, + "learning_rate": 1.9410905784330945e-05, + "loss": 1.3217, + "step": 4864 + }, + { + "epoch": 0.41, + "learning_rate": 1.941059837204932e-05, + "loss": 1.2496, + "step": 4865 + }, + { + "epoch": 0.41, + "learning_rate": 1.941029088201432e-05, + "loss": 1.44, + "step": 4866 + }, + { + "epoch": 0.41, + "learning_rate": 1.9409983314228487e-05, + "loss": 1.2582, + "step": 4867 + }, + { + "epoch": 0.41, + "learning_rate": 1.9409675668694358e-05, + "loss": 1.3712, + "step": 4868 + }, + { + "epoch": 0.41, + "learning_rate": 1.9409367945414476e-05, + "loss": 1.3643, + "step": 4869 + }, + { + "epoch": 0.41, + "learning_rate": 1.940906014439139e-05, + "loss": 1.313, + "step": 4870 + }, + { + "epoch": 0.41, + "learning_rate": 1.9408752265627633e-05, + "loss": 1.2785, + "step": 4871 + }, + { + "epoch": 0.41, + "learning_rate": 1.9408444309125754e-05, + "loss": 1.296, + "step": 4872 + }, + { + "epoch": 0.41, + "learning_rate": 1.9408136274888297e-05, + "loss": 1.3477, + "step": 4873 + }, + { + "epoch": 0.41, + "learning_rate": 1.940782816291781e-05, + "loss": 1.3386, + "step": 4874 + }, + { + "epoch": 0.41, + "learning_rate": 1.9407519973216833e-05, + "loss": 1.2965, + "step": 4875 + }, + { + "epoch": 0.41, + "learning_rate": 1.9407211705787916e-05, + "loss": 1.3044, + "step": 4876 + }, + { + "epoch": 0.41, + "learning_rate": 1.9406903360633605e-05, + "loss": 1.3115, + "step": 4877 + }, + { + "epoch": 0.41, + "learning_rate": 1.9406594937756453e-05, + "loss": 1.2114, + "step": 4878 + }, + { + "epoch": 0.41, + "learning_rate": 1.9406286437158996e-05, + "loss": 1.3703, + "step": 4879 + }, + { + "epoch": 0.41, + "learning_rate": 1.940597785884379e-05, + "loss": 1.2534, + "step": 4880 + }, + { + "epoch": 0.41, + "learning_rate": 1.9405669202813387e-05, + "loss": 1.3859, + "step": 4881 + }, + { + "epoch": 0.41, + "learning_rate": 1.9405360469070335e-05, + "loss": 1.338, + "step": 4882 + }, + { + "epoch": 0.41, + "learning_rate": 1.9405051657617187e-05, + "loss": 1.4233, + "step": 4883 + }, + { + "epoch": 0.41, + "learning_rate": 1.9404742768456488e-05, + "loss": 1.4122, + "step": 4884 + }, + { + "epoch": 0.41, + "learning_rate": 1.9404433801590796e-05, + "loss": 1.2957, + "step": 4885 + }, + { + "epoch": 0.41, + "learning_rate": 1.940412475702266e-05, + "loss": 1.3032, + "step": 4886 + }, + { + "epoch": 0.41, + "learning_rate": 1.9403815634754642e-05, + "loss": 1.4268, + "step": 4887 + }, + { + "epoch": 0.41, + "learning_rate": 1.9403506434789284e-05, + "loss": 1.3247, + "step": 4888 + }, + { + "epoch": 0.41, + "learning_rate": 1.9403197157129144e-05, + "loss": 1.3488, + "step": 4889 + }, + { + "epoch": 0.41, + "learning_rate": 1.9402887801776788e-05, + "loss": 1.347, + "step": 4890 + }, + { + "epoch": 0.41, + "learning_rate": 1.9402578368734756e-05, + "loss": 1.3567, + "step": 4891 + }, + { + "epoch": 0.41, + "learning_rate": 1.9402268858005617e-05, + "loss": 1.3508, + "step": 4892 + }, + { + "epoch": 0.41, + "learning_rate": 1.940195926959192e-05, + "loss": 1.4502, + "step": 4893 + }, + { + "epoch": 0.41, + "learning_rate": 1.9401649603496228e-05, + "loss": 1.3434, + "step": 4894 + }, + { + "epoch": 0.41, + "learning_rate": 1.94013398597211e-05, + "loss": 1.3195, + "step": 4895 + }, + { + "epoch": 0.41, + "learning_rate": 1.9401030038269092e-05, + "loss": 1.2503, + "step": 4896 + }, + { + "epoch": 0.41, + "learning_rate": 1.9400720139142763e-05, + "loss": 1.3877, + "step": 4897 + }, + { + "epoch": 0.41, + "learning_rate": 1.940041016234468e-05, + "loss": 1.3726, + "step": 4898 + }, + { + "epoch": 0.41, + "learning_rate": 1.94001001078774e-05, + "loss": 1.3236, + "step": 4899 + }, + { + "epoch": 0.41, + "learning_rate": 1.9399789975743476e-05, + "loss": 1.3272, + "step": 4900 + }, + { + "epoch": 0.41, + "learning_rate": 1.9399479765945486e-05, + "loss": 1.3273, + "step": 4901 + }, + { + "epoch": 0.41, + "learning_rate": 1.9399169478485985e-05, + "loss": 1.3714, + "step": 4902 + }, + { + "epoch": 0.41, + "learning_rate": 1.9398859113367536e-05, + "loss": 1.2503, + "step": 4903 + }, + { + "epoch": 0.41, + "learning_rate": 1.9398548670592707e-05, + "loss": 1.3449, + "step": 4904 + }, + { + "epoch": 0.41, + "learning_rate": 1.939823815016406e-05, + "loss": 1.3861, + "step": 4905 + }, + { + "epoch": 0.41, + "learning_rate": 1.9397927552084162e-05, + "loss": 1.3781, + "step": 4906 + }, + { + "epoch": 0.41, + "learning_rate": 1.9397616876355577e-05, + "loss": 1.2994, + "step": 4907 + }, + { + "epoch": 0.41, + "learning_rate": 1.9397306122980877e-05, + "loss": 1.23, + "step": 4908 + }, + { + "epoch": 0.41, + "learning_rate": 1.9396995291962625e-05, + "loss": 1.3388, + "step": 4909 + }, + { + "epoch": 0.41, + "learning_rate": 1.939668438330339e-05, + "loss": 1.3134, + "step": 4910 + }, + { + "epoch": 0.41, + "learning_rate": 1.939637339700574e-05, + "loss": 1.4236, + "step": 4911 + }, + { + "epoch": 0.41, + "learning_rate": 1.939606233307225e-05, + "loss": 1.2822, + "step": 4912 + }, + { + "epoch": 0.41, + "learning_rate": 1.939575119150548e-05, + "loss": 1.3173, + "step": 4913 + }, + { + "epoch": 0.41, + "learning_rate": 1.9395439972308013e-05, + "loss": 1.2988, + "step": 4914 + }, + { + "epoch": 0.41, + "learning_rate": 1.939512867548241e-05, + "loss": 1.3572, + "step": 4915 + }, + { + "epoch": 0.41, + "learning_rate": 1.9394817301031246e-05, + "loss": 1.2912, + "step": 4916 + }, + { + "epoch": 0.41, + "learning_rate": 1.9394505848957097e-05, + "loss": 1.3287, + "step": 4917 + }, + { + "epoch": 0.41, + "learning_rate": 1.9394194319262532e-05, + "loss": 1.2846, + "step": 4918 + }, + { + "epoch": 0.41, + "learning_rate": 1.939388271195013e-05, + "loss": 1.3402, + "step": 4919 + }, + { + "epoch": 0.41, + "learning_rate": 1.9393571027022463e-05, + "loss": 1.2745, + "step": 4920 + }, + { + "epoch": 0.41, + "learning_rate": 1.93932592644821e-05, + "loss": 1.2683, + "step": 4921 + }, + { + "epoch": 0.41, + "learning_rate": 1.9392947424331628e-05, + "loss": 1.3277, + "step": 4922 + }, + { + "epoch": 0.41, + "learning_rate": 1.9392635506573614e-05, + "loss": 1.333, + "step": 4923 + }, + { + "epoch": 0.41, + "learning_rate": 1.9392323511210645e-05, + "loss": 1.4204, + "step": 4924 + }, + { + "epoch": 0.41, + "learning_rate": 1.939201143824529e-05, + "loss": 1.2848, + "step": 4925 + }, + { + "epoch": 0.41, + "learning_rate": 1.939169928768013e-05, + "loss": 1.3658, + "step": 4926 + }, + { + "epoch": 0.41, + "learning_rate": 1.939138705951775e-05, + "loss": 1.3734, + "step": 4927 + }, + { + "epoch": 0.41, + "learning_rate": 1.939107475376072e-05, + "loss": 1.2064, + "step": 4928 + }, + { + "epoch": 0.41, + "learning_rate": 1.9390762370411625e-05, + "loss": 1.3198, + "step": 4929 + }, + { + "epoch": 0.42, + "learning_rate": 1.9390449909473042e-05, + "loss": 1.3462, + "step": 4930 + }, + { + "epoch": 0.42, + "learning_rate": 1.9390137370947563e-05, + "loss": 1.3182, + "step": 4931 + }, + { + "epoch": 0.42, + "learning_rate": 1.9389824754837762e-05, + "loss": 1.2474, + "step": 4932 + }, + { + "epoch": 0.42, + "learning_rate": 1.938951206114622e-05, + "loss": 1.3443, + "step": 4933 + }, + { + "epoch": 0.42, + "learning_rate": 1.938919928987553e-05, + "loss": 1.37, + "step": 4934 + }, + { + "epoch": 0.42, + "learning_rate": 1.9388886441028268e-05, + "loss": 1.3812, + "step": 4935 + }, + { + "epoch": 0.42, + "learning_rate": 1.9388573514607022e-05, + "loss": 1.3404, + "step": 4936 + }, + { + "epoch": 0.42, + "learning_rate": 1.9388260510614377e-05, + "loss": 1.3125, + "step": 4937 + }, + { + "epoch": 0.42, + "learning_rate": 1.9387947429052918e-05, + "loss": 1.3855, + "step": 4938 + }, + { + "epoch": 0.42, + "learning_rate": 1.9387634269925236e-05, + "loss": 1.4356, + "step": 4939 + }, + { + "epoch": 0.42, + "learning_rate": 1.938732103323391e-05, + "loss": 1.3045, + "step": 4940 + }, + { + "epoch": 0.42, + "learning_rate": 1.9387007718981537e-05, + "loss": 1.2431, + "step": 4941 + }, + { + "epoch": 0.42, + "learning_rate": 1.93866943271707e-05, + "loss": 1.3361, + "step": 4942 + }, + { + "epoch": 0.42, + "learning_rate": 1.938638085780399e-05, + "loss": 1.3373, + "step": 4943 + }, + { + "epoch": 0.42, + "learning_rate": 1.9386067310884e-05, + "loss": 1.308, + "step": 4944 + }, + { + "epoch": 0.42, + "learning_rate": 1.9385753686413316e-05, + "loss": 1.3167, + "step": 4945 + }, + { + "epoch": 0.42, + "learning_rate": 1.938543998439453e-05, + "loss": 1.3957, + "step": 4946 + }, + { + "epoch": 0.42, + "learning_rate": 1.938512620483024e-05, + "loss": 1.25, + "step": 4947 + }, + { + "epoch": 0.42, + "learning_rate": 1.938481234772303e-05, + "loss": 1.4017, + "step": 4948 + }, + { + "epoch": 0.42, + "learning_rate": 1.9384498413075497e-05, + "loss": 1.4024, + "step": 4949 + }, + { + "epoch": 0.42, + "learning_rate": 1.9384184400890234e-05, + "loss": 1.2348, + "step": 4950 + }, + { + "epoch": 0.42, + "learning_rate": 1.938387031116984e-05, + "loss": 1.3072, + "step": 4951 + }, + { + "epoch": 0.42, + "learning_rate": 1.9383556143916903e-05, + "loss": 1.418, + "step": 4952 + }, + { + "epoch": 0.42, + "learning_rate": 1.9383241899134018e-05, + "loss": 1.3694, + "step": 4953 + }, + { + "epoch": 0.42, + "learning_rate": 1.9382927576823793e-05, + "loss": 1.3264, + "step": 4954 + }, + { + "epoch": 0.42, + "learning_rate": 1.9382613176988814e-05, + "loss": 1.3048, + "step": 4955 + }, + { + "epoch": 0.42, + "learning_rate": 1.938229869963168e-05, + "loss": 1.2965, + "step": 4956 + }, + { + "epoch": 0.42, + "learning_rate": 1.9381984144754995e-05, + "loss": 1.3932, + "step": 4957 + }, + { + "epoch": 0.42, + "learning_rate": 1.9381669512361353e-05, + "loss": 1.3099, + "step": 4958 + }, + { + "epoch": 0.42, + "learning_rate": 1.9381354802453357e-05, + "loss": 1.2029, + "step": 4959 + }, + { + "epoch": 0.42, + "learning_rate": 1.9381040015033604e-05, + "loss": 1.2981, + "step": 4960 + }, + { + "epoch": 0.42, + "learning_rate": 1.9380725150104695e-05, + "loss": 1.303, + "step": 4961 + }, + { + "epoch": 0.42, + "learning_rate": 1.938041020766923e-05, + "loss": 1.2808, + "step": 4962 + }, + { + "epoch": 0.42, + "learning_rate": 1.9380095187729817e-05, + "loss": 1.3191, + "step": 4963 + }, + { + "epoch": 0.42, + "learning_rate": 1.9379780090289056e-05, + "loss": 1.3818, + "step": 4964 + }, + { + "epoch": 0.42, + "learning_rate": 1.9379464915349547e-05, + "loss": 1.2698, + "step": 4965 + }, + { + "epoch": 0.42, + "learning_rate": 1.9379149662913903e-05, + "loss": 1.3231, + "step": 4966 + }, + { + "epoch": 0.42, + "learning_rate": 1.9378834332984716e-05, + "loss": 1.3119, + "step": 4967 + }, + { + "epoch": 0.42, + "learning_rate": 1.93785189255646e-05, + "loss": 1.2217, + "step": 4968 + }, + { + "epoch": 0.42, + "learning_rate": 1.9378203440656163e-05, + "loss": 1.3762, + "step": 4969 + }, + { + "epoch": 0.42, + "learning_rate": 1.9377887878262006e-05, + "loss": 1.2762, + "step": 4970 + }, + { + "epoch": 0.42, + "learning_rate": 1.9377572238384737e-05, + "loss": 1.2775, + "step": 4971 + }, + { + "epoch": 0.42, + "learning_rate": 1.9377256521026962e-05, + "loss": 1.3282, + "step": 4972 + }, + { + "epoch": 0.42, + "learning_rate": 1.93769407261913e-05, + "loss": 1.3334, + "step": 4973 + }, + { + "epoch": 0.42, + "learning_rate": 1.9376624853880347e-05, + "loss": 1.3068, + "step": 4974 + }, + { + "epoch": 0.42, + "learning_rate": 1.9376308904096718e-05, + "loss": 1.4059, + "step": 4975 + }, + { + "epoch": 0.42, + "learning_rate": 1.937599287684303e-05, + "loss": 1.3603, + "step": 4976 + }, + { + "epoch": 0.42, + "learning_rate": 1.9375676772121885e-05, + "loss": 1.3855, + "step": 4977 + }, + { + "epoch": 0.42, + "learning_rate": 1.9375360589935894e-05, + "loss": 1.3395, + "step": 4978 + }, + { + "epoch": 0.42, + "learning_rate": 1.9375044330287677e-05, + "loss": 1.3221, + "step": 4979 + }, + { + "epoch": 0.42, + "learning_rate": 1.937472799317984e-05, + "loss": 1.2852, + "step": 4980 + }, + { + "epoch": 0.42, + "learning_rate": 1.9374411578615005e-05, + "loss": 1.3213, + "step": 4981 + }, + { + "epoch": 0.42, + "learning_rate": 1.9374095086595777e-05, + "loss": 1.4069, + "step": 4982 + }, + { + "epoch": 0.42, + "learning_rate": 1.937377851712478e-05, + "loss": 1.4012, + "step": 4983 + }, + { + "epoch": 0.42, + "learning_rate": 1.937346187020462e-05, + "loss": 1.2316, + "step": 4984 + }, + { + "epoch": 0.42, + "learning_rate": 1.937314514583792e-05, + "loss": 1.4223, + "step": 4985 + }, + { + "epoch": 0.42, + "learning_rate": 1.9372828344027295e-05, + "loss": 1.3568, + "step": 4986 + }, + { + "epoch": 0.42, + "learning_rate": 1.9372511464775365e-05, + "loss": 1.2469, + "step": 4987 + }, + { + "epoch": 0.42, + "learning_rate": 1.9372194508084742e-05, + "loss": 1.3261, + "step": 4988 + }, + { + "epoch": 0.42, + "learning_rate": 1.9371877473958052e-05, + "loss": 1.2421, + "step": 4989 + }, + { + "epoch": 0.42, + "learning_rate": 1.9371560362397907e-05, + "loss": 1.2782, + "step": 4990 + }, + { + "epoch": 0.42, + "learning_rate": 1.9371243173406936e-05, + "loss": 1.4012, + "step": 4991 + }, + { + "epoch": 0.42, + "learning_rate": 1.937092590698775e-05, + "loss": 1.2803, + "step": 4992 + }, + { + "epoch": 0.42, + "learning_rate": 1.9370608563142976e-05, + "loss": 1.4136, + "step": 4993 + }, + { + "epoch": 0.42, + "learning_rate": 1.9370291141875236e-05, + "loss": 1.3072, + "step": 4994 + }, + { + "epoch": 0.42, + "learning_rate": 1.9369973643187152e-05, + "loss": 1.3985, + "step": 4995 + }, + { + "epoch": 0.42, + "learning_rate": 1.936965606708135e-05, + "loss": 1.3293, + "step": 4996 + }, + { + "epoch": 0.42, + "learning_rate": 1.9369338413560443e-05, + "loss": 1.3066, + "step": 4997 + }, + { + "epoch": 0.42, + "learning_rate": 1.9369020682627073e-05, + "loss": 1.2814, + "step": 4998 + }, + { + "epoch": 0.42, + "learning_rate": 1.9368702874283846e-05, + "loss": 1.3097, + "step": 4999 + }, + { + "epoch": 0.42, + "learning_rate": 1.9368384988533405e-05, + "loss": 1.3721, + "step": 5000 + }, + { + "epoch": 0.42, + "learning_rate": 1.9368067025378362e-05, + "loss": 1.2789, + "step": 5001 + }, + { + "epoch": 0.42, + "learning_rate": 1.9367748984821358e-05, + "loss": 1.3068, + "step": 5002 + }, + { + "epoch": 0.42, + "learning_rate": 1.936743086686501e-05, + "loss": 1.356, + "step": 5003 + }, + { + "epoch": 0.42, + "learning_rate": 1.9367112671511948e-05, + "loss": 1.2599, + "step": 5004 + }, + { + "epoch": 0.42, + "learning_rate": 1.936679439876481e-05, + "loss": 1.3431, + "step": 5005 + }, + { + "epoch": 0.42, + "learning_rate": 1.9366476048626216e-05, + "loss": 1.3398, + "step": 5006 + }, + { + "epoch": 0.42, + "learning_rate": 1.9366157621098794e-05, + "loss": 1.3356, + "step": 5007 + }, + { + "epoch": 0.42, + "learning_rate": 1.936583911618519e-05, + "loss": 1.4068, + "step": 5008 + }, + { + "epoch": 0.42, + "learning_rate": 1.936552053388802e-05, + "loss": 1.3514, + "step": 5009 + }, + { + "epoch": 0.42, + "learning_rate": 1.936520187420992e-05, + "loss": 1.4021, + "step": 5010 + }, + { + "epoch": 0.42, + "learning_rate": 1.9364883137153528e-05, + "loss": 1.3204, + "step": 5011 + }, + { + "epoch": 0.42, + "learning_rate": 1.9364564322721473e-05, + "loss": 1.2624, + "step": 5012 + }, + { + "epoch": 0.42, + "learning_rate": 1.936424543091639e-05, + "loss": 1.2858, + "step": 5013 + }, + { + "epoch": 0.42, + "learning_rate": 1.936392646174092e-05, + "loss": 1.3798, + "step": 5014 + }, + { + "epoch": 0.42, + "learning_rate": 1.9363607415197686e-05, + "loss": 1.37, + "step": 5015 + }, + { + "epoch": 0.42, + "learning_rate": 1.9363288291289333e-05, + "loss": 1.3178, + "step": 5016 + }, + { + "epoch": 0.42, + "learning_rate": 1.9362969090018497e-05, + "loss": 1.3775, + "step": 5017 + }, + { + "epoch": 0.42, + "learning_rate": 1.936264981138781e-05, + "loss": 1.2809, + "step": 5018 + }, + { + "epoch": 0.42, + "learning_rate": 1.9362330455399917e-05, + "loss": 1.3486, + "step": 5019 + }, + { + "epoch": 0.42, + "learning_rate": 1.9362011022057455e-05, + "loss": 1.2972, + "step": 5020 + }, + { + "epoch": 0.42, + "learning_rate": 1.936169151136306e-05, + "loss": 1.4056, + "step": 5021 + }, + { + "epoch": 0.42, + "learning_rate": 1.9361371923319372e-05, + "loss": 1.4099, + "step": 5022 + }, + { + "epoch": 0.42, + "learning_rate": 1.9361052257929032e-05, + "loss": 1.362, + "step": 5023 + }, + { + "epoch": 0.42, + "learning_rate": 1.9360732515194685e-05, + "loss": 1.3162, + "step": 5024 + }, + { + "epoch": 0.42, + "learning_rate": 1.936041269511897e-05, + "loss": 1.4047, + "step": 5025 + }, + { + "epoch": 0.42, + "learning_rate": 1.936009279770453e-05, + "loss": 1.3856, + "step": 5026 + }, + { + "epoch": 0.42, + "learning_rate": 1.9359772822954007e-05, + "loss": 1.3022, + "step": 5027 + }, + { + "epoch": 0.42, + "learning_rate": 1.9359452770870044e-05, + "loss": 1.354, + "step": 5028 + }, + { + "epoch": 0.42, + "learning_rate": 1.935913264145529e-05, + "loss": 1.393, + "step": 5029 + }, + { + "epoch": 0.42, + "learning_rate": 1.935881243471238e-05, + "loss": 1.3694, + "step": 5030 + }, + { + "epoch": 0.42, + "learning_rate": 1.9358492150643975e-05, + "loss": 1.2926, + "step": 5031 + }, + { + "epoch": 0.42, + "learning_rate": 1.9358171789252708e-05, + "loss": 1.3278, + "step": 5032 + }, + { + "epoch": 0.42, + "learning_rate": 1.9357851350541233e-05, + "loss": 1.327, + "step": 5033 + }, + { + "epoch": 0.42, + "learning_rate": 1.9357530834512195e-05, + "loss": 1.3007, + "step": 5034 + }, + { + "epoch": 0.42, + "learning_rate": 1.935721024116824e-05, + "loss": 1.3334, + "step": 5035 + }, + { + "epoch": 0.42, + "learning_rate": 1.9356889570512022e-05, + "loss": 1.3398, + "step": 5036 + }, + { + "epoch": 0.42, + "learning_rate": 1.9356568822546186e-05, + "loss": 1.3359, + "step": 5037 + }, + { + "epoch": 0.42, + "learning_rate": 1.9356247997273385e-05, + "loss": 1.2371, + "step": 5038 + }, + { + "epoch": 0.42, + "learning_rate": 1.9355927094696267e-05, + "loss": 1.3709, + "step": 5039 + }, + { + "epoch": 0.42, + "learning_rate": 1.9355606114817486e-05, + "loss": 1.3703, + "step": 5040 + }, + { + "epoch": 0.42, + "learning_rate": 1.935528505763969e-05, + "loss": 1.335, + "step": 5041 + }, + { + "epoch": 0.42, + "learning_rate": 1.935496392316554e-05, + "loss": 1.3553, + "step": 5042 + }, + { + "epoch": 0.42, + "learning_rate": 1.9354642711397684e-05, + "loss": 1.3342, + "step": 5043 + }, + { + "epoch": 0.42, + "learning_rate": 1.9354321422338772e-05, + "loss": 1.3908, + "step": 5044 + }, + { + "epoch": 0.42, + "learning_rate": 1.9354000055991468e-05, + "loss": 1.3833, + "step": 5045 + }, + { + "epoch": 0.42, + "learning_rate": 1.9353678612358418e-05, + "loss": 1.3205, + "step": 5046 + }, + { + "epoch": 0.42, + "learning_rate": 1.935335709144228e-05, + "loss": 1.3442, + "step": 5047 + }, + { + "epoch": 0.42, + "learning_rate": 1.9353035493245718e-05, + "loss": 1.3107, + "step": 5048 + }, + { + "epoch": 0.43, + "learning_rate": 1.935271381777138e-05, + "loss": 1.4425, + "step": 5049 + }, + { + "epoch": 0.43, + "learning_rate": 1.9352392065021925e-05, + "loss": 1.3664, + "step": 5050 + }, + { + "epoch": 0.43, + "learning_rate": 1.9352070235000016e-05, + "loss": 1.2465, + "step": 5051 + }, + { + "epoch": 0.43, + "learning_rate": 1.935174832770831e-05, + "loss": 1.3035, + "step": 5052 + }, + { + "epoch": 0.43, + "learning_rate": 1.935142634314947e-05, + "loss": 1.3069, + "step": 5053 + }, + { + "epoch": 0.43, + "learning_rate": 1.9351104281326147e-05, + "loss": 1.4107, + "step": 5054 + }, + { + "epoch": 0.43, + "learning_rate": 1.9350782142241008e-05, + "loss": 1.3872, + "step": 5055 + }, + { + "epoch": 0.43, + "learning_rate": 1.9350459925896717e-05, + "loss": 1.3558, + "step": 5056 + }, + { + "epoch": 0.43, + "learning_rate": 1.935013763229593e-05, + "loss": 1.35, + "step": 5057 + }, + { + "epoch": 0.43, + "learning_rate": 1.9349815261441317e-05, + "loss": 1.4108, + "step": 5058 + }, + { + "epoch": 0.43, + "learning_rate": 1.9349492813335537e-05, + "loss": 1.3015, + "step": 5059 + }, + { + "epoch": 0.43, + "learning_rate": 1.9349170287981253e-05, + "loss": 1.3333, + "step": 5060 + }, + { + "epoch": 0.43, + "learning_rate": 1.9348847685381136e-05, + "loss": 1.379, + "step": 5061 + }, + { + "epoch": 0.43, + "learning_rate": 1.9348525005537845e-05, + "loss": 1.2679, + "step": 5062 + }, + { + "epoch": 0.43, + "learning_rate": 1.934820224845405e-05, + "loss": 1.2998, + "step": 5063 + }, + { + "epoch": 0.43, + "learning_rate": 1.9347879414132413e-05, + "loss": 1.29, + "step": 5064 + }, + { + "epoch": 0.43, + "learning_rate": 1.934755650257561e-05, + "loss": 1.2558, + "step": 5065 + }, + { + "epoch": 0.43, + "learning_rate": 1.9347233513786296e-05, + "loss": 1.3082, + "step": 5066 + }, + { + "epoch": 0.43, + "learning_rate": 1.9346910447767152e-05, + "loss": 1.3939, + "step": 5067 + }, + { + "epoch": 0.43, + "learning_rate": 1.934658730452084e-05, + "loss": 1.3036, + "step": 5068 + }, + { + "epoch": 0.43, + "learning_rate": 1.9346264084050035e-05, + "loss": 1.305, + "step": 5069 + }, + { + "epoch": 0.43, + "learning_rate": 1.9345940786357406e-05, + "loss": 1.3493, + "step": 5070 + }, + { + "epoch": 0.43, + "learning_rate": 1.934561741144562e-05, + "loss": 1.2691, + "step": 5071 + }, + { + "epoch": 0.43, + "learning_rate": 1.9345293959317356e-05, + "loss": 1.3271, + "step": 5072 + }, + { + "epoch": 0.43, + "learning_rate": 1.934497042997528e-05, + "loss": 1.2626, + "step": 5073 + }, + { + "epoch": 0.43, + "learning_rate": 1.9344646823422067e-05, + "loss": 1.2971, + "step": 5074 + }, + { + "epoch": 0.43, + "learning_rate": 1.9344323139660388e-05, + "loss": 1.2632, + "step": 5075 + }, + { + "epoch": 0.43, + "learning_rate": 1.9343999378692927e-05, + "loss": 1.3619, + "step": 5076 + }, + { + "epoch": 0.43, + "learning_rate": 1.934367554052235e-05, + "loss": 1.2677, + "step": 5077 + }, + { + "epoch": 0.43, + "learning_rate": 1.9343351625151333e-05, + "loss": 1.4144, + "step": 5078 + }, + { + "epoch": 0.43, + "learning_rate": 1.9343027632582557e-05, + "loss": 1.4149, + "step": 5079 + }, + { + "epoch": 0.43, + "learning_rate": 1.9342703562818697e-05, + "loss": 1.3709, + "step": 5080 + }, + { + "epoch": 0.43, + "learning_rate": 1.934237941586243e-05, + "loss": 1.2602, + "step": 5081 + }, + { + "epoch": 0.43, + "learning_rate": 1.934205519171643e-05, + "loss": 1.3556, + "step": 5082 + }, + { + "epoch": 0.43, + "learning_rate": 1.9341730890383386e-05, + "loss": 1.3587, + "step": 5083 + }, + { + "epoch": 0.43, + "learning_rate": 1.934140651186597e-05, + "loss": 1.2377, + "step": 5084 + }, + { + "epoch": 0.43, + "learning_rate": 1.934108205616686e-05, + "loss": 1.2559, + "step": 5085 + }, + { + "epoch": 0.43, + "learning_rate": 1.9340757523288747e-05, + "loss": 1.3163, + "step": 5086 + }, + { + "epoch": 0.43, + "learning_rate": 1.9340432913234304e-05, + "loss": 1.3918, + "step": 5087 + }, + { + "epoch": 0.43, + "learning_rate": 1.9340108226006212e-05, + "loss": 1.4137, + "step": 5088 + }, + { + "epoch": 0.43, + "learning_rate": 1.933978346160716e-05, + "loss": 1.3959, + "step": 5089 + }, + { + "epoch": 0.43, + "learning_rate": 1.9339458620039825e-05, + "loss": 1.3563, + "step": 5090 + }, + { + "epoch": 0.43, + "learning_rate": 1.9339133701306895e-05, + "loss": 1.2906, + "step": 5091 + }, + { + "epoch": 0.43, + "learning_rate": 1.9338808705411055e-05, + "loss": 1.327, + "step": 5092 + }, + { + "epoch": 0.43, + "learning_rate": 1.9338483632354986e-05, + "loss": 1.3489, + "step": 5093 + }, + { + "epoch": 0.43, + "learning_rate": 1.9338158482141378e-05, + "loss": 1.3053, + "step": 5094 + }, + { + "epoch": 0.43, + "learning_rate": 1.9337833254772914e-05, + "loss": 1.2786, + "step": 5095 + }, + { + "epoch": 0.43, + "learning_rate": 1.933750795025229e-05, + "loss": 1.3123, + "step": 5096 + }, + { + "epoch": 0.43, + "learning_rate": 1.9337182568582178e-05, + "loss": 1.3056, + "step": 5097 + }, + { + "epoch": 0.43, + "learning_rate": 1.9336857109765282e-05, + "loss": 1.3819, + "step": 5098 + }, + { + "epoch": 0.43, + "learning_rate": 1.933653157380428e-05, + "loss": 1.3448, + "step": 5099 + }, + { + "epoch": 0.43, + "learning_rate": 1.933620596070187e-05, + "loss": 1.3392, + "step": 5100 + }, + { + "epoch": 0.43, + "learning_rate": 1.9335880270460735e-05, + "loss": 1.3082, + "step": 5101 + }, + { + "epoch": 0.43, + "learning_rate": 1.933555450308357e-05, + "loss": 1.2551, + "step": 5102 + }, + { + "epoch": 0.43, + "learning_rate": 1.9335228658573064e-05, + "loss": 1.2872, + "step": 5103 + }, + { + "epoch": 0.43, + "learning_rate": 1.9334902736931914e-05, + "loss": 1.364, + "step": 5104 + }, + { + "epoch": 0.43, + "learning_rate": 1.933457673816281e-05, + "loss": 1.3771, + "step": 5105 + }, + { + "epoch": 0.43, + "learning_rate": 1.9334250662268446e-05, + "loss": 1.2939, + "step": 5106 + }, + { + "epoch": 0.43, + "learning_rate": 1.9333924509251516e-05, + "loss": 1.3022, + "step": 5107 + }, + { + "epoch": 0.43, + "learning_rate": 1.9333598279114712e-05, + "loss": 1.3182, + "step": 5108 + }, + { + "epoch": 0.43, + "learning_rate": 1.9333271971860732e-05, + "loss": 1.2234, + "step": 5109 + }, + { + "epoch": 0.43, + "learning_rate": 1.9332945587492276e-05, + "loss": 1.3269, + "step": 5110 + }, + { + "epoch": 0.43, + "learning_rate": 1.933261912601203e-05, + "loss": 1.2599, + "step": 5111 + }, + { + "epoch": 0.43, + "learning_rate": 1.93322925874227e-05, + "loss": 1.3208, + "step": 5112 + }, + { + "epoch": 0.43, + "learning_rate": 1.9331965971726986e-05, + "loss": 1.3695, + "step": 5113 + }, + { + "epoch": 0.43, + "learning_rate": 1.933163927892758e-05, + "loss": 1.3952, + "step": 5114 + }, + { + "epoch": 0.43, + "learning_rate": 1.9331312509027186e-05, + "loss": 1.2487, + "step": 5115 + }, + { + "epoch": 0.43, + "learning_rate": 1.9330985662028498e-05, + "loss": 1.3265, + "step": 5116 + }, + { + "epoch": 0.43, + "learning_rate": 1.9330658737934222e-05, + "loss": 1.266, + "step": 5117 + }, + { + "epoch": 0.43, + "learning_rate": 1.933033173674706e-05, + "loss": 1.35, + "step": 5118 + }, + { + "epoch": 0.43, + "learning_rate": 1.9330004658469705e-05, + "loss": 1.3039, + "step": 5119 + }, + { + "epoch": 0.43, + "learning_rate": 1.9329677503104868e-05, + "loss": 1.3085, + "step": 5120 + }, + { + "epoch": 0.43, + "learning_rate": 1.932935027065525e-05, + "loss": 1.3369, + "step": 5121 + }, + { + "epoch": 0.43, + "learning_rate": 1.9329022961123556e-05, + "loss": 1.3136, + "step": 5122 + }, + { + "epoch": 0.43, + "learning_rate": 1.9328695574512487e-05, + "loss": 1.3394, + "step": 5123 + }, + { + "epoch": 0.43, + "learning_rate": 1.932836811082475e-05, + "loss": 1.3731, + "step": 5124 + }, + { + "epoch": 0.43, + "learning_rate": 1.9328040570063048e-05, + "loss": 1.3533, + "step": 5125 + }, + { + "epoch": 0.43, + "learning_rate": 1.9327712952230093e-05, + "loss": 1.2893, + "step": 5126 + }, + { + "epoch": 0.43, + "learning_rate": 1.932738525732859e-05, + "loss": 1.3238, + "step": 5127 + }, + { + "epoch": 0.43, + "learning_rate": 1.9327057485361238e-05, + "loss": 1.3176, + "step": 5128 + }, + { + "epoch": 0.43, + "learning_rate": 1.9326729636330756e-05, + "loss": 1.2407, + "step": 5129 + }, + { + "epoch": 0.43, + "learning_rate": 1.9326401710239848e-05, + "loss": 1.4178, + "step": 5130 + }, + { + "epoch": 0.43, + "learning_rate": 1.9326073707091226e-05, + "loss": 1.2999, + "step": 5131 + }, + { + "epoch": 0.43, + "learning_rate": 1.9325745626887597e-05, + "loss": 1.3546, + "step": 5132 + }, + { + "epoch": 0.43, + "learning_rate": 1.932541746963167e-05, + "loss": 1.3684, + "step": 5133 + }, + { + "epoch": 0.43, + "learning_rate": 1.9325089235326165e-05, + "loss": 1.2939, + "step": 5134 + }, + { + "epoch": 0.43, + "learning_rate": 1.9324760923973784e-05, + "loss": 1.3091, + "step": 5135 + }, + { + "epoch": 0.43, + "learning_rate": 1.9324432535577247e-05, + "loss": 1.2893, + "step": 5136 + }, + { + "epoch": 0.43, + "learning_rate": 1.9324104070139262e-05, + "loss": 1.2299, + "step": 5137 + }, + { + "epoch": 0.43, + "learning_rate": 1.9323775527662543e-05, + "loss": 1.4124, + "step": 5138 + }, + { + "epoch": 0.43, + "learning_rate": 1.932344690814981e-05, + "loss": 1.3235, + "step": 5139 + }, + { + "epoch": 0.43, + "learning_rate": 1.9323118211603776e-05, + "loss": 1.3091, + "step": 5140 + }, + { + "epoch": 0.43, + "learning_rate": 1.9322789438027155e-05, + "loss": 1.4569, + "step": 5141 + }, + { + "epoch": 0.43, + "learning_rate": 1.9322460587422662e-05, + "loss": 1.2067, + "step": 5142 + }, + { + "epoch": 0.43, + "learning_rate": 1.9322131659793018e-05, + "loss": 1.2505, + "step": 5143 + }, + { + "epoch": 0.43, + "learning_rate": 1.9321802655140937e-05, + "loss": 1.2711, + "step": 5144 + }, + { + "epoch": 0.43, + "learning_rate": 1.932147357346914e-05, + "loss": 1.3534, + "step": 5145 + }, + { + "epoch": 0.43, + "learning_rate": 1.9321144414780343e-05, + "loss": 1.3382, + "step": 5146 + }, + { + "epoch": 0.43, + "learning_rate": 1.9320815179077274e-05, + "loss": 1.2475, + "step": 5147 + }, + { + "epoch": 0.43, + "learning_rate": 1.932048586636264e-05, + "loss": 1.3036, + "step": 5148 + }, + { + "epoch": 0.43, + "learning_rate": 1.9320156476639173e-05, + "loss": 1.388, + "step": 5149 + }, + { + "epoch": 0.43, + "learning_rate": 1.9319827009909586e-05, + "loss": 1.3871, + "step": 5150 + }, + { + "epoch": 0.43, + "learning_rate": 1.9319497466176606e-05, + "loss": 1.37, + "step": 5151 + }, + { + "epoch": 0.43, + "learning_rate": 1.931916784544296e-05, + "loss": 1.2273, + "step": 5152 + }, + { + "epoch": 0.43, + "learning_rate": 1.9318838147711365e-05, + "loss": 1.3562, + "step": 5153 + }, + { + "epoch": 0.43, + "learning_rate": 1.9318508372984545e-05, + "loss": 1.3518, + "step": 5154 + }, + { + "epoch": 0.43, + "learning_rate": 1.931817852126523e-05, + "loss": 1.338, + "step": 5155 + }, + { + "epoch": 0.43, + "learning_rate": 1.9317848592556136e-05, + "loss": 1.3059, + "step": 5156 + }, + { + "epoch": 0.43, + "learning_rate": 1.931751858686e-05, + "loss": 1.2524, + "step": 5157 + }, + { + "epoch": 0.43, + "learning_rate": 1.931718850417954e-05, + "loss": 1.1988, + "step": 5158 + }, + { + "epoch": 0.43, + "learning_rate": 1.9316858344517488e-05, + "loss": 1.3979, + "step": 5159 + }, + { + "epoch": 0.43, + "learning_rate": 1.931652810787657e-05, + "loss": 1.314, + "step": 5160 + }, + { + "epoch": 0.43, + "learning_rate": 1.931619779425952e-05, + "loss": 1.3455, + "step": 5161 + }, + { + "epoch": 0.43, + "learning_rate": 1.9315867403669054e-05, + "loss": 1.2836, + "step": 5162 + }, + { + "epoch": 0.43, + "learning_rate": 1.9315536936107918e-05, + "loss": 1.3166, + "step": 5163 + }, + { + "epoch": 0.43, + "learning_rate": 1.931520639157883e-05, + "loss": 1.2146, + "step": 5164 + }, + { + "epoch": 0.43, + "learning_rate": 1.9314875770084526e-05, + "loss": 1.3138, + "step": 5165 + }, + { + "epoch": 0.43, + "learning_rate": 1.9314545071627738e-05, + "loss": 1.4375, + "step": 5166 + }, + { + "epoch": 0.44, + "learning_rate": 1.9314214296211197e-05, + "loss": 1.3144, + "step": 5167 + }, + { + "epoch": 0.44, + "learning_rate": 1.9313883443837636e-05, + "loss": 1.4146, + "step": 5168 + }, + { + "epoch": 0.44, + "learning_rate": 1.931355251450979e-05, + "loss": 1.3163, + "step": 5169 + }, + { + "epoch": 0.44, + "learning_rate": 1.9313221508230394e-05, + "loss": 1.3171, + "step": 5170 + }, + { + "epoch": 0.44, + "learning_rate": 1.9312890425002178e-05, + "loss": 1.3174, + "step": 5171 + }, + { + "epoch": 0.44, + "learning_rate": 1.931255926482788e-05, + "loss": 1.3019, + "step": 5172 + }, + { + "epoch": 0.44, + "learning_rate": 1.931222802771024e-05, + "loss": 1.2284, + "step": 5173 + }, + { + "epoch": 0.44, + "learning_rate": 1.9311896713651993e-05, + "loss": 1.3246, + "step": 5174 + }, + { + "epoch": 0.44, + "learning_rate": 1.9311565322655874e-05, + "loss": 1.3738, + "step": 5175 + }, + { + "epoch": 0.44, + "learning_rate": 1.931123385472462e-05, + "loss": 1.4857, + "step": 5176 + }, + { + "epoch": 0.44, + "learning_rate": 1.9310902309860973e-05, + "loss": 1.2907, + "step": 5177 + }, + { + "epoch": 0.44, + "learning_rate": 1.931057068806767e-05, + "loss": 1.308, + "step": 5178 + }, + { + "epoch": 0.44, + "learning_rate": 1.9310238989347456e-05, + "loss": 1.3267, + "step": 5179 + }, + { + "epoch": 0.44, + "learning_rate": 1.9309907213703067e-05, + "loss": 1.3736, + "step": 5180 + }, + { + "epoch": 0.44, + "learning_rate": 1.9309575361137244e-05, + "loss": 1.4005, + "step": 5181 + }, + { + "epoch": 0.44, + "learning_rate": 1.930924343165273e-05, + "loss": 1.3249, + "step": 5182 + }, + { + "epoch": 0.44, + "learning_rate": 1.9308911425252265e-05, + "loss": 1.2901, + "step": 5183 + }, + { + "epoch": 0.44, + "learning_rate": 1.9308579341938598e-05, + "loss": 1.3978, + "step": 5184 + }, + { + "epoch": 0.44, + "learning_rate": 1.9308247181714467e-05, + "loss": 1.3275, + "step": 5185 + }, + { + "epoch": 0.44, + "learning_rate": 1.930791494458262e-05, + "loss": 1.3628, + "step": 5186 + }, + { + "epoch": 0.44, + "learning_rate": 1.93075826305458e-05, + "loss": 1.3983, + "step": 5187 + }, + { + "epoch": 0.44, + "learning_rate": 1.9307250239606756e-05, + "loss": 1.2836, + "step": 5188 + }, + { + "epoch": 0.44, + "learning_rate": 1.930691777176823e-05, + "loss": 1.295, + "step": 5189 + }, + { + "epoch": 0.44, + "learning_rate": 1.930658522703297e-05, + "loss": 1.3433, + "step": 5190 + }, + { + "epoch": 0.44, + "learning_rate": 1.9306252605403725e-05, + "loss": 1.3404, + "step": 5191 + }, + { + "epoch": 0.44, + "learning_rate": 1.930591990688324e-05, + "loss": 1.3332, + "step": 5192 + }, + { + "epoch": 0.44, + "learning_rate": 1.930558713147427e-05, + "loss": 1.2878, + "step": 5193 + }, + { + "epoch": 0.44, + "learning_rate": 1.9305254279179563e-05, + "loss": 1.3544, + "step": 5194 + }, + { + "epoch": 0.44, + "learning_rate": 1.930492135000186e-05, + "loss": 1.329, + "step": 5195 + }, + { + "epoch": 0.44, + "learning_rate": 1.9304588343943927e-05, + "loss": 1.2927, + "step": 5196 + }, + { + "epoch": 0.44, + "learning_rate": 1.93042552610085e-05, + "loss": 1.3783, + "step": 5197 + }, + { + "epoch": 0.44, + "learning_rate": 1.9303922101198343e-05, + "loss": 1.3894, + "step": 5198 + }, + { + "epoch": 0.44, + "learning_rate": 1.9303588864516202e-05, + "loss": 1.3536, + "step": 5199 + }, + { + "epoch": 0.44, + "learning_rate": 1.9303255550964834e-05, + "loss": 1.3305, + "step": 5200 + }, + { + "epoch": 0.44, + "learning_rate": 1.930292216054699e-05, + "loss": 1.4224, + "step": 5201 + }, + { + "epoch": 0.44, + "learning_rate": 1.9302588693265426e-05, + "loss": 1.4124, + "step": 5202 + }, + { + "epoch": 0.44, + "learning_rate": 1.9302255149122895e-05, + "loss": 1.2521, + "step": 5203 + }, + { + "epoch": 0.44, + "learning_rate": 1.930192152812216e-05, + "loss": 1.2739, + "step": 5204 + }, + { + "epoch": 0.44, + "learning_rate": 1.9301587830265967e-05, + "loss": 1.3221, + "step": 5205 + }, + { + "epoch": 0.44, + "learning_rate": 1.930125405555708e-05, + "loss": 1.2739, + "step": 5206 + }, + { + "epoch": 0.44, + "learning_rate": 1.9300920203998254e-05, + "loss": 1.3194, + "step": 5207 + }, + { + "epoch": 0.44, + "learning_rate": 1.930058627559225e-05, + "loss": 1.3137, + "step": 5208 + }, + { + "epoch": 0.44, + "learning_rate": 1.9300252270341828e-05, + "loss": 1.273, + "step": 5209 + }, + { + "epoch": 0.44, + "learning_rate": 1.9299918188249743e-05, + "loss": 1.2298, + "step": 5210 + }, + { + "epoch": 0.44, + "learning_rate": 1.9299584029318755e-05, + "loss": 1.3466, + "step": 5211 + }, + { + "epoch": 0.44, + "learning_rate": 1.929924979355163e-05, + "loss": 1.2846, + "step": 5212 + }, + { + "epoch": 0.44, + "learning_rate": 1.9298915480951132e-05, + "loss": 1.2891, + "step": 5213 + }, + { + "epoch": 0.44, + "learning_rate": 1.9298581091520012e-05, + "loss": 1.3215, + "step": 5214 + }, + { + "epoch": 0.44, + "learning_rate": 1.929824662526104e-05, + "loss": 1.2926, + "step": 5215 + }, + { + "epoch": 0.44, + "learning_rate": 1.9297912082176978e-05, + "loss": 1.2666, + "step": 5216 + }, + { + "epoch": 0.44, + "learning_rate": 1.9297577462270593e-05, + "loss": 1.3664, + "step": 5217 + }, + { + "epoch": 0.44, + "learning_rate": 1.9297242765544644e-05, + "loss": 1.2585, + "step": 5218 + }, + { + "epoch": 0.44, + "learning_rate": 1.92969079920019e-05, + "loss": 1.2722, + "step": 5219 + }, + { + "epoch": 0.44, + "learning_rate": 1.9296573141645132e-05, + "loss": 1.298, + "step": 5220 + }, + { + "epoch": 0.44, + "learning_rate": 1.9296238214477098e-05, + "loss": 1.3378, + "step": 5221 + }, + { + "epoch": 0.44, + "learning_rate": 1.929590321050057e-05, + "loss": 1.3017, + "step": 5222 + }, + { + "epoch": 0.44, + "learning_rate": 1.9295568129718313e-05, + "loss": 1.3562, + "step": 5223 + }, + { + "epoch": 0.44, + "learning_rate": 1.92952329721331e-05, + "loss": 1.3059, + "step": 5224 + }, + { + "epoch": 0.44, + "learning_rate": 1.929489773774769e-05, + "loss": 1.4101, + "step": 5225 + }, + { + "epoch": 0.44, + "learning_rate": 1.9294562426564867e-05, + "loss": 1.3137, + "step": 5226 + }, + { + "epoch": 0.44, + "learning_rate": 1.9294227038587392e-05, + "loss": 1.2604, + "step": 5227 + }, + { + "epoch": 0.44, + "learning_rate": 1.9293891573818042e-05, + "loss": 1.2859, + "step": 5228 + }, + { + "epoch": 0.44, + "learning_rate": 1.9293556032259583e-05, + "loss": 1.3513, + "step": 5229 + }, + { + "epoch": 0.44, + "learning_rate": 1.9293220413914788e-05, + "loss": 1.3339, + "step": 5230 + }, + { + "epoch": 0.44, + "learning_rate": 1.9292884718786432e-05, + "loss": 1.3864, + "step": 5231 + }, + { + "epoch": 0.44, + "learning_rate": 1.929254894687729e-05, + "loss": 1.3233, + "step": 5232 + }, + { + "epoch": 0.44, + "learning_rate": 1.9292213098190136e-05, + "loss": 1.3796, + "step": 5233 + }, + { + "epoch": 0.44, + "learning_rate": 1.929187717272774e-05, + "loss": 1.2684, + "step": 5234 + }, + { + "epoch": 0.44, + "learning_rate": 1.929154117049288e-05, + "loss": 1.2633, + "step": 5235 + }, + { + "epoch": 0.44, + "learning_rate": 1.929120509148834e-05, + "loss": 1.2947, + "step": 5236 + }, + { + "epoch": 0.44, + "learning_rate": 1.9290868935716888e-05, + "loss": 1.3489, + "step": 5237 + }, + { + "epoch": 0.44, + "learning_rate": 1.92905327031813e-05, + "loss": 1.2994, + "step": 5238 + }, + { + "epoch": 0.44, + "learning_rate": 1.9290196393884357e-05, + "loss": 1.3256, + "step": 5239 + }, + { + "epoch": 0.44, + "learning_rate": 1.928986000782884e-05, + "loss": 1.3095, + "step": 5240 + }, + { + "epoch": 0.44, + "learning_rate": 1.9289523545017528e-05, + "loss": 1.2602, + "step": 5241 + }, + { + "epoch": 0.44, + "learning_rate": 1.9289187005453197e-05, + "loss": 1.2759, + "step": 5242 + }, + { + "epoch": 0.44, + "learning_rate": 1.9288850389138633e-05, + "loss": 1.4112, + "step": 5243 + }, + { + "epoch": 0.44, + "learning_rate": 1.9288513696076608e-05, + "loss": 1.3403, + "step": 5244 + }, + { + "epoch": 0.44, + "learning_rate": 1.9288176926269915e-05, + "loss": 1.2268, + "step": 5245 + }, + { + "epoch": 0.44, + "learning_rate": 1.9287840079721333e-05, + "loss": 1.4193, + "step": 5246 + }, + { + "epoch": 0.44, + "learning_rate": 1.928750315643364e-05, + "loss": 1.3185, + "step": 5247 + }, + { + "epoch": 0.44, + "learning_rate": 1.9287166156409625e-05, + "loss": 1.428, + "step": 5248 + }, + { + "epoch": 0.44, + "learning_rate": 1.928682907965207e-05, + "loss": 1.3852, + "step": 5249 + }, + { + "epoch": 0.44, + "learning_rate": 1.9286491926163762e-05, + "loss": 1.3836, + "step": 5250 + }, + { + "epoch": 0.44, + "learning_rate": 1.9286154695947487e-05, + "loss": 1.4086, + "step": 5251 + }, + { + "epoch": 0.44, + "learning_rate": 1.928581738900603e-05, + "loss": 1.2311, + "step": 5252 + }, + { + "epoch": 0.44, + "learning_rate": 1.9285480005342175e-05, + "loss": 1.291, + "step": 5253 + }, + { + "epoch": 0.44, + "learning_rate": 1.9285142544958715e-05, + "loss": 1.2421, + "step": 5254 + }, + { + "epoch": 0.44, + "learning_rate": 1.9284805007858435e-05, + "loss": 1.264, + "step": 5255 + }, + { + "epoch": 0.44, + "learning_rate": 1.9284467394044123e-05, + "loss": 1.2984, + "step": 5256 + }, + { + "epoch": 0.44, + "learning_rate": 1.928412970351857e-05, + "loss": 1.2663, + "step": 5257 + }, + { + "epoch": 0.44, + "learning_rate": 1.9283791936284567e-05, + "loss": 1.3224, + "step": 5258 + }, + { + "epoch": 0.44, + "learning_rate": 1.9283454092344908e-05, + "loss": 1.3479, + "step": 5259 + }, + { + "epoch": 0.44, + "learning_rate": 1.9283116171702373e-05, + "loss": 1.389, + "step": 5260 + }, + { + "epoch": 0.44, + "learning_rate": 1.9282778174359765e-05, + "loss": 1.2829, + "step": 5261 + }, + { + "epoch": 0.44, + "learning_rate": 1.928244010031987e-05, + "loss": 1.437, + "step": 5262 + }, + { + "epoch": 0.44, + "learning_rate": 1.928210194958549e-05, + "loss": 1.2765, + "step": 5263 + }, + { + "epoch": 0.44, + "learning_rate": 1.928176372215941e-05, + "loss": 1.2977, + "step": 5264 + }, + { + "epoch": 0.44, + "learning_rate": 1.9281425418044427e-05, + "loss": 1.2702, + "step": 5265 + }, + { + "epoch": 0.44, + "learning_rate": 1.9281087037243336e-05, + "loss": 1.3126, + "step": 5266 + }, + { + "epoch": 0.44, + "learning_rate": 1.9280748579758936e-05, + "loss": 1.3047, + "step": 5267 + }, + { + "epoch": 0.44, + "learning_rate": 1.9280410045594018e-05, + "loss": 1.2904, + "step": 5268 + }, + { + "epoch": 0.44, + "learning_rate": 1.9280071434751382e-05, + "loss": 1.2553, + "step": 5269 + }, + { + "epoch": 0.44, + "learning_rate": 1.927973274723383e-05, + "loss": 1.4063, + "step": 5270 + }, + { + "epoch": 0.44, + "learning_rate": 1.9279393983044153e-05, + "loss": 1.3932, + "step": 5271 + }, + { + "epoch": 0.44, + "learning_rate": 1.9279055142185156e-05, + "loss": 1.3224, + "step": 5272 + }, + { + "epoch": 0.44, + "learning_rate": 1.927871622465963e-05, + "loss": 1.373, + "step": 5273 + }, + { + "epoch": 0.44, + "learning_rate": 1.9278377230470386e-05, + "loss": 1.3188, + "step": 5274 + }, + { + "epoch": 0.44, + "learning_rate": 1.9278038159620216e-05, + "loss": 1.3865, + "step": 5275 + }, + { + "epoch": 0.44, + "learning_rate": 1.927769901211193e-05, + "loss": 1.3413, + "step": 5276 + }, + { + "epoch": 0.44, + "learning_rate": 1.927735978794832e-05, + "loss": 1.3392, + "step": 5277 + }, + { + "epoch": 0.44, + "learning_rate": 1.92770204871322e-05, + "loss": 1.374, + "step": 5278 + }, + { + "epoch": 0.44, + "learning_rate": 1.9276681109666365e-05, + "loss": 1.269, + "step": 5279 + }, + { + "epoch": 0.44, + "learning_rate": 1.927634165555362e-05, + "loss": 1.3452, + "step": 5280 + }, + { + "epoch": 0.44, + "learning_rate": 1.9276002124796775e-05, + "loss": 1.3843, + "step": 5281 + }, + { + "epoch": 0.44, + "learning_rate": 1.927566251739863e-05, + "loss": 1.3188, + "step": 5282 + }, + { + "epoch": 0.44, + "learning_rate": 1.9275322833361993e-05, + "loss": 1.2451, + "step": 5283 + }, + { + "epoch": 0.44, + "learning_rate": 1.927498307268967e-05, + "loss": 1.3943, + "step": 5284 + }, + { + "epoch": 0.44, + "learning_rate": 1.927464323538447e-05, + "loss": 1.3356, + "step": 5285 + }, + { + "epoch": 0.45, + "learning_rate": 1.9274303321449194e-05, + "loss": 1.3459, + "step": 5286 + }, + { + "epoch": 0.45, + "learning_rate": 1.927396333088666e-05, + "loss": 1.3555, + "step": 5287 + }, + { + "epoch": 0.45, + "learning_rate": 1.9273623263699674e-05, + "loss": 1.3882, + "step": 5288 + }, + { + "epoch": 0.45, + "learning_rate": 1.9273283119891043e-05, + "loss": 1.2369, + "step": 5289 + }, + { + "epoch": 0.45, + "learning_rate": 1.927294289946358e-05, + "loss": 1.3129, + "step": 5290 + }, + { + "epoch": 0.45, + "learning_rate": 1.9272602602420093e-05, + "loss": 1.3946, + "step": 5291 + }, + { + "epoch": 0.45, + "learning_rate": 1.92722622287634e-05, + "loss": 1.3597, + "step": 5292 + }, + { + "epoch": 0.45, + "learning_rate": 1.9271921778496305e-05, + "loss": 1.3556, + "step": 5293 + }, + { + "epoch": 0.45, + "learning_rate": 1.9271581251621625e-05, + "loss": 1.3188, + "step": 5294 + }, + { + "epoch": 0.45, + "learning_rate": 1.9271240648142173e-05, + "loss": 1.3192, + "step": 5295 + }, + { + "epoch": 0.45, + "learning_rate": 1.9270899968060765e-05, + "loss": 1.4341, + "step": 5296 + }, + { + "epoch": 0.45, + "learning_rate": 1.9270559211380212e-05, + "loss": 1.2901, + "step": 5297 + }, + { + "epoch": 0.45, + "learning_rate": 1.9270218378103337e-05, + "loss": 1.356, + "step": 5298 + }, + { + "epoch": 0.45, + "learning_rate": 1.9269877468232946e-05, + "loss": 1.2315, + "step": 5299 + }, + { + "epoch": 0.45, + "learning_rate": 1.9269536481771863e-05, + "loss": 1.3552, + "step": 5300 + }, + { + "epoch": 0.45, + "learning_rate": 1.9269195418722902e-05, + "loss": 1.3122, + "step": 5301 + }, + { + "epoch": 0.45, + "learning_rate": 1.9268854279088884e-05, + "loss": 1.3619, + "step": 5302 + }, + { + "epoch": 0.45, + "learning_rate": 1.9268513062872626e-05, + "loss": 1.3684, + "step": 5303 + }, + { + "epoch": 0.45, + "learning_rate": 1.926817177007694e-05, + "loss": 1.3467, + "step": 5304 + }, + { + "epoch": 0.45, + "learning_rate": 1.926783040070466e-05, + "loss": 1.3933, + "step": 5305 + }, + { + "epoch": 0.45, + "learning_rate": 1.9267488954758596e-05, + "loss": 1.2997, + "step": 5306 + }, + { + "epoch": 0.45, + "learning_rate": 1.9267147432241573e-05, + "loss": 1.3906, + "step": 5307 + }, + { + "epoch": 0.45, + "learning_rate": 1.926680583315641e-05, + "loss": 1.3561, + "step": 5308 + }, + { + "epoch": 0.45, + "learning_rate": 1.9266464157505937e-05, + "loss": 1.3637, + "step": 5309 + }, + { + "epoch": 0.45, + "learning_rate": 1.9266122405292965e-05, + "loss": 1.3688, + "step": 5310 + }, + { + "epoch": 0.45, + "learning_rate": 1.9265780576520327e-05, + "loss": 1.3093, + "step": 5311 + }, + { + "epoch": 0.45, + "learning_rate": 1.9265438671190843e-05, + "loss": 1.2771, + "step": 5312 + }, + { + "epoch": 0.45, + "learning_rate": 1.9265096689307344e-05, + "loss": 1.2779, + "step": 5313 + }, + { + "epoch": 0.45, + "learning_rate": 1.9264754630872646e-05, + "loss": 1.3278, + "step": 5314 + }, + { + "epoch": 0.45, + "learning_rate": 1.9264412495889582e-05, + "loss": 1.292, + "step": 5315 + }, + { + "epoch": 0.45, + "learning_rate": 1.9264070284360977e-05, + "loss": 1.2797, + "step": 5316 + }, + { + "epoch": 0.45, + "learning_rate": 1.926372799628966e-05, + "loss": 1.4128, + "step": 5317 + }, + { + "epoch": 0.45, + "learning_rate": 1.9263385631678454e-05, + "loss": 1.277, + "step": 5318 + }, + { + "epoch": 0.45, + "learning_rate": 1.9263043190530195e-05, + "loss": 1.2705, + "step": 5319 + }, + { + "epoch": 0.45, + "learning_rate": 1.9262700672847704e-05, + "loss": 1.3196, + "step": 5320 + }, + { + "epoch": 0.45, + "learning_rate": 1.9262358078633818e-05, + "loss": 1.3038, + "step": 5321 + }, + { + "epoch": 0.45, + "learning_rate": 1.9262015407891366e-05, + "loss": 1.3521, + "step": 5322 + }, + { + "epoch": 0.45, + "learning_rate": 1.9261672660623177e-05, + "loss": 1.3025, + "step": 5323 + }, + { + "epoch": 0.45, + "learning_rate": 1.9261329836832086e-05, + "loss": 1.3259, + "step": 5324 + }, + { + "epoch": 0.45, + "learning_rate": 1.9260986936520924e-05, + "loss": 1.3435, + "step": 5325 + }, + { + "epoch": 0.45, + "learning_rate": 1.926064395969252e-05, + "loss": 1.4506, + "step": 5326 + }, + { + "epoch": 0.45, + "learning_rate": 1.9260300906349715e-05, + "loss": 1.2817, + "step": 5327 + }, + { + "epoch": 0.45, + "learning_rate": 1.925995777649534e-05, + "loss": 1.3036, + "step": 5328 + }, + { + "epoch": 0.45, + "learning_rate": 1.9259614570132235e-05, + "loss": 1.3179, + "step": 5329 + }, + { + "epoch": 0.45, + "learning_rate": 1.9259271287263222e-05, + "loss": 1.2215, + "step": 5330 + }, + { + "epoch": 0.45, + "learning_rate": 1.9258927927891153e-05, + "loss": 1.2815, + "step": 5331 + }, + { + "epoch": 0.45, + "learning_rate": 1.9258584492018856e-05, + "loss": 1.2945, + "step": 5332 + }, + { + "epoch": 0.45, + "learning_rate": 1.9258240979649173e-05, + "loss": 1.2106, + "step": 5333 + }, + { + "epoch": 0.45, + "learning_rate": 1.9257897390784938e-05, + "loss": 1.2983, + "step": 5334 + }, + { + "epoch": 0.45, + "learning_rate": 1.925755372542899e-05, + "loss": 1.3062, + "step": 5335 + }, + { + "epoch": 0.45, + "learning_rate": 1.925720998358417e-05, + "loss": 1.3941, + "step": 5336 + }, + { + "epoch": 0.45, + "learning_rate": 1.925686616525332e-05, + "loss": 1.2644, + "step": 5337 + }, + { + "epoch": 0.45, + "learning_rate": 1.925652227043928e-05, + "loss": 1.3375, + "step": 5338 + }, + { + "epoch": 0.45, + "learning_rate": 1.925617829914489e-05, + "loss": 1.2997, + "step": 5339 + }, + { + "epoch": 0.45, + "learning_rate": 1.9255834251372995e-05, + "loss": 1.3382, + "step": 5340 + }, + { + "epoch": 0.45, + "learning_rate": 1.9255490127126433e-05, + "loss": 1.2598, + "step": 5341 + }, + { + "epoch": 0.45, + "learning_rate": 1.9255145926408047e-05, + "loss": 1.2469, + "step": 5342 + }, + { + "epoch": 0.45, + "learning_rate": 1.9254801649220688e-05, + "loss": 1.2878, + "step": 5343 + }, + { + "epoch": 0.45, + "learning_rate": 1.925445729556719e-05, + "loss": 1.3093, + "step": 5344 + }, + { + "epoch": 0.45, + "learning_rate": 1.925411286545041e-05, + "loss": 1.2519, + "step": 5345 + }, + { + "epoch": 0.45, + "learning_rate": 1.9253768358873187e-05, + "loss": 1.3794, + "step": 5346 + }, + { + "epoch": 0.45, + "learning_rate": 1.9253423775838365e-05, + "loss": 1.3732, + "step": 5347 + }, + { + "epoch": 0.45, + "learning_rate": 1.9253079116348794e-05, + "loss": 1.3019, + "step": 5348 + }, + { + "epoch": 0.45, + "learning_rate": 1.9252734380407324e-05, + "loss": 1.3241, + "step": 5349 + }, + { + "epoch": 0.45, + "learning_rate": 1.92523895680168e-05, + "loss": 1.3977, + "step": 5350 + }, + { + "epoch": 0.45, + "learning_rate": 1.9252044679180075e-05, + "loss": 1.3017, + "step": 5351 + }, + { + "epoch": 0.45, + "learning_rate": 1.9251699713899993e-05, + "loss": 1.3439, + "step": 5352 + }, + { + "epoch": 0.45, + "learning_rate": 1.9251354672179408e-05, + "loss": 1.3198, + "step": 5353 + }, + { + "epoch": 0.45, + "learning_rate": 1.925100955402117e-05, + "loss": 1.2646, + "step": 5354 + }, + { + "epoch": 0.45, + "learning_rate": 1.925066435942813e-05, + "loss": 1.3829, + "step": 5355 + }, + { + "epoch": 0.45, + "learning_rate": 1.925031908840314e-05, + "loss": 1.291, + "step": 5356 + }, + { + "epoch": 0.45, + "learning_rate": 1.9249973740949057e-05, + "loss": 1.2172, + "step": 5357 + }, + { + "epoch": 0.45, + "learning_rate": 1.9249628317068728e-05, + "loss": 1.4039, + "step": 5358 + }, + { + "epoch": 0.45, + "learning_rate": 1.924928281676501e-05, + "loss": 1.4067, + "step": 5359 + }, + { + "epoch": 0.45, + "learning_rate": 1.9248937240040754e-05, + "loss": 1.3953, + "step": 5360 + }, + { + "epoch": 0.45, + "learning_rate": 1.9248591586898822e-05, + "loss": 1.3225, + "step": 5361 + }, + { + "epoch": 0.45, + "learning_rate": 1.9248245857342066e-05, + "loss": 1.3629, + "step": 5362 + }, + { + "epoch": 0.45, + "learning_rate": 1.924790005137334e-05, + "loss": 1.3608, + "step": 5363 + }, + { + "epoch": 0.45, + "learning_rate": 1.9247554168995507e-05, + "loss": 1.4058, + "step": 5364 + }, + { + "epoch": 0.45, + "learning_rate": 1.9247208210211422e-05, + "loss": 1.2709, + "step": 5365 + }, + { + "epoch": 0.45, + "learning_rate": 1.9246862175023942e-05, + "loss": 1.3318, + "step": 5366 + }, + { + "epoch": 0.45, + "learning_rate": 1.9246516063435927e-05, + "loss": 1.329, + "step": 5367 + }, + { + "epoch": 0.45, + "learning_rate": 1.924616987545024e-05, + "loss": 1.3764, + "step": 5368 + }, + { + "epoch": 0.45, + "learning_rate": 1.9245823611069735e-05, + "loss": 1.2991, + "step": 5369 + }, + { + "epoch": 0.45, + "learning_rate": 1.9245477270297278e-05, + "loss": 1.3321, + "step": 5370 + }, + { + "epoch": 0.45, + "learning_rate": 1.9245130853135727e-05, + "loss": 1.3953, + "step": 5371 + }, + { + "epoch": 0.45, + "learning_rate": 1.9244784359587948e-05, + "loss": 1.1669, + "step": 5372 + }, + { + "epoch": 0.45, + "learning_rate": 1.9244437789656802e-05, + "loss": 1.2606, + "step": 5373 + }, + { + "epoch": 0.45, + "learning_rate": 1.924409114334515e-05, + "loss": 1.2711, + "step": 5374 + }, + { + "epoch": 0.45, + "learning_rate": 1.924374442065586e-05, + "loss": 1.2617, + "step": 5375 + }, + { + "epoch": 0.45, + "learning_rate": 1.9243397621591796e-05, + "loss": 1.2628, + "step": 5376 + }, + { + "epoch": 0.45, + "learning_rate": 1.924305074615582e-05, + "loss": 1.3159, + "step": 5377 + }, + { + "epoch": 0.45, + "learning_rate": 1.9242703794350804e-05, + "loss": 1.3164, + "step": 5378 + }, + { + "epoch": 0.45, + "learning_rate": 1.924235676617961e-05, + "loss": 1.2859, + "step": 5379 + }, + { + "epoch": 0.45, + "learning_rate": 1.9242009661645107e-05, + "loss": 1.3228, + "step": 5380 + }, + { + "epoch": 0.45, + "learning_rate": 1.924166248075016e-05, + "loss": 1.3101, + "step": 5381 + }, + { + "epoch": 0.45, + "learning_rate": 1.9241315223497642e-05, + "loss": 1.3528, + "step": 5382 + }, + { + "epoch": 0.45, + "learning_rate": 1.924096788989042e-05, + "loss": 1.3237, + "step": 5383 + }, + { + "epoch": 0.45, + "learning_rate": 1.9240620479931363e-05, + "loss": 1.331, + "step": 5384 + }, + { + "epoch": 0.45, + "learning_rate": 1.9240272993623345e-05, + "loss": 1.3251, + "step": 5385 + }, + { + "epoch": 0.45, + "learning_rate": 1.923992543096923e-05, + "loss": 1.297, + "step": 5386 + }, + { + "epoch": 0.45, + "learning_rate": 1.9239577791971894e-05, + "loss": 1.3564, + "step": 5387 + }, + { + "epoch": 0.45, + "learning_rate": 1.923923007663421e-05, + "loss": 1.3697, + "step": 5388 + }, + { + "epoch": 0.45, + "learning_rate": 1.923888228495905e-05, + "loss": 1.336, + "step": 5389 + }, + { + "epoch": 0.45, + "learning_rate": 1.923853441694929e-05, + "loss": 1.3179, + "step": 5390 + }, + { + "epoch": 0.45, + "learning_rate": 1.9238186472607803e-05, + "loss": 1.2779, + "step": 5391 + }, + { + "epoch": 0.45, + "learning_rate": 1.923783845193746e-05, + "loss": 1.321, + "step": 5392 + }, + { + "epoch": 0.45, + "learning_rate": 1.923749035494114e-05, + "loss": 1.3551, + "step": 5393 + }, + { + "epoch": 0.45, + "learning_rate": 1.9237142181621717e-05, + "loss": 1.298, + "step": 5394 + }, + { + "epoch": 0.45, + "learning_rate": 1.923679393198207e-05, + "loss": 1.3131, + "step": 5395 + }, + { + "epoch": 0.45, + "learning_rate": 1.9236445606025075e-05, + "loss": 1.3074, + "step": 5396 + }, + { + "epoch": 0.45, + "learning_rate": 1.923609720375361e-05, + "loss": 1.4039, + "step": 5397 + }, + { + "epoch": 0.45, + "learning_rate": 1.9235748725170555e-05, + "loss": 1.327, + "step": 5398 + }, + { + "epoch": 0.45, + "learning_rate": 1.9235400170278788e-05, + "loss": 1.3563, + "step": 5399 + }, + { + "epoch": 0.45, + "learning_rate": 1.923505153908119e-05, + "loss": 1.3196, + "step": 5400 + }, + { + "epoch": 0.45, + "learning_rate": 1.923470283158064e-05, + "loss": 1.461, + "step": 5401 + }, + { + "epoch": 0.45, + "learning_rate": 1.9234354047780018e-05, + "loss": 1.2422, + "step": 5402 + }, + { + "epoch": 0.45, + "learning_rate": 1.923400518768221e-05, + "loss": 1.2969, + "step": 5403 + }, + { + "epoch": 0.45, + "learning_rate": 1.9233656251290096e-05, + "loss": 1.2607, + "step": 5404 + }, + { + "epoch": 0.46, + "learning_rate": 1.9233307238606556e-05, + "loss": 1.3104, + "step": 5405 + }, + { + "epoch": 0.46, + "learning_rate": 1.923295814963448e-05, + "loss": 1.3527, + "step": 5406 + }, + { + "epoch": 0.46, + "learning_rate": 1.9232608984376746e-05, + "loss": 1.4616, + "step": 5407 + }, + { + "epoch": 0.46, + "learning_rate": 1.9232259742836243e-05, + "loss": 1.255, + "step": 5408 + }, + { + "epoch": 0.46, + "learning_rate": 1.9231910425015856e-05, + "loss": 1.2997, + "step": 5409 + }, + { + "epoch": 0.46, + "learning_rate": 1.923156103091847e-05, + "loss": 1.2658, + "step": 5410 + }, + { + "epoch": 0.46, + "learning_rate": 1.923121156054697e-05, + "loss": 1.3245, + "step": 5411 + }, + { + "epoch": 0.46, + "learning_rate": 1.9230862013904248e-05, + "loss": 1.3943, + "step": 5412 + }, + { + "epoch": 0.46, + "learning_rate": 1.923051239099319e-05, + "loss": 1.3707, + "step": 5413 + }, + { + "epoch": 0.46, + "learning_rate": 1.9230162691816684e-05, + "loss": 1.3404, + "step": 5414 + }, + { + "epoch": 0.46, + "learning_rate": 1.922981291637762e-05, + "loss": 1.3212, + "step": 5415 + }, + { + "epoch": 0.46, + "learning_rate": 1.9229463064678885e-05, + "loss": 1.2197, + "step": 5416 + }, + { + "epoch": 0.46, + "learning_rate": 1.9229113136723377e-05, + "loss": 1.3396, + "step": 5417 + }, + { + "epoch": 0.46, + "learning_rate": 1.9228763132513978e-05, + "loss": 1.2678, + "step": 5418 + }, + { + "epoch": 0.46, + "learning_rate": 1.922841305205359e-05, + "loss": 1.3111, + "step": 5419 + }, + { + "epoch": 0.46, + "learning_rate": 1.9228062895345093e-05, + "loss": 1.3725, + "step": 5420 + }, + { + "epoch": 0.46, + "learning_rate": 1.9227712662391395e-05, + "loss": 1.3415, + "step": 5421 + }, + { + "epoch": 0.46, + "learning_rate": 1.9227362353195376e-05, + "loss": 1.3353, + "step": 5422 + }, + { + "epoch": 0.46, + "learning_rate": 1.9227011967759937e-05, + "loss": 1.2187, + "step": 5423 + }, + { + "epoch": 0.46, + "learning_rate": 1.9226661506087973e-05, + "loss": 1.2569, + "step": 5424 + }, + { + "epoch": 0.46, + "learning_rate": 1.922631096818238e-05, + "loss": 1.3426, + "step": 5425 + }, + { + "epoch": 0.46, + "learning_rate": 1.922596035404605e-05, + "loss": 1.2803, + "step": 5426 + }, + { + "epoch": 0.46, + "learning_rate": 1.9225609663681884e-05, + "loss": 1.3284, + "step": 5427 + }, + { + "epoch": 0.46, + "learning_rate": 1.9225258897092776e-05, + "loss": 1.3454, + "step": 5428 + }, + { + "epoch": 0.46, + "learning_rate": 1.9224908054281634e-05, + "loss": 1.3686, + "step": 5429 + }, + { + "epoch": 0.46, + "learning_rate": 1.922455713525134e-05, + "loss": 1.3295, + "step": 5430 + }, + { + "epoch": 0.46, + "learning_rate": 1.922420614000481e-05, + "loss": 1.4154, + "step": 5431 + }, + { + "epoch": 0.46, + "learning_rate": 1.9223855068544933e-05, + "loss": 1.2665, + "step": 5432 + }, + { + "epoch": 0.46, + "learning_rate": 1.9223503920874615e-05, + "loss": 1.3794, + "step": 5433 + }, + { + "epoch": 0.46, + "learning_rate": 1.9223152696996756e-05, + "loss": 1.3251, + "step": 5434 + }, + { + "epoch": 0.46, + "learning_rate": 1.9222801396914253e-05, + "loss": 1.3443, + "step": 5435 + }, + { + "epoch": 0.46, + "learning_rate": 1.9222450020630017e-05, + "loss": 1.3196, + "step": 5436 + }, + { + "epoch": 0.46, + "learning_rate": 1.922209856814695e-05, + "loss": 1.2796, + "step": 5437 + }, + { + "epoch": 0.46, + "learning_rate": 1.922174703946795e-05, + "loss": 1.3467, + "step": 5438 + }, + { + "epoch": 0.46, + "learning_rate": 1.9221395434595924e-05, + "loss": 1.3775, + "step": 5439 + }, + { + "epoch": 0.46, + "learning_rate": 1.922104375353378e-05, + "loss": 1.4481, + "step": 5440 + }, + { + "epoch": 0.46, + "learning_rate": 1.922069199628442e-05, + "loss": 1.3363, + "step": 5441 + }, + { + "epoch": 0.46, + "learning_rate": 1.922034016285075e-05, + "loss": 1.3854, + "step": 5442 + }, + { + "epoch": 0.46, + "learning_rate": 1.921998825323568e-05, + "loss": 1.2965, + "step": 5443 + }, + { + "epoch": 0.46, + "learning_rate": 1.9219636267442118e-05, + "loss": 1.307, + "step": 5444 + }, + { + "epoch": 0.46, + "learning_rate": 1.921928420547297e-05, + "loss": 1.432, + "step": 5445 + }, + { + "epoch": 0.46, + "learning_rate": 1.921893206733114e-05, + "loss": 1.2868, + "step": 5446 + }, + { + "epoch": 0.46, + "learning_rate": 1.921857985301955e-05, + "loss": 1.3266, + "step": 5447 + }, + { + "epoch": 0.46, + "learning_rate": 1.92182275625411e-05, + "loss": 1.2613, + "step": 5448 + }, + { + "epoch": 0.46, + "learning_rate": 1.9217875195898706e-05, + "loss": 1.2685, + "step": 5449 + }, + { + "epoch": 0.46, + "learning_rate": 1.9217522753095276e-05, + "loss": 1.3269, + "step": 5450 + }, + { + "epoch": 0.46, + "learning_rate": 1.921717023413372e-05, + "loss": 1.3171, + "step": 5451 + }, + { + "epoch": 0.46, + "learning_rate": 1.9216817639016955e-05, + "loss": 1.3684, + "step": 5452 + }, + { + "epoch": 0.46, + "learning_rate": 1.9216464967747895e-05, + "loss": 1.2847, + "step": 5453 + }, + { + "epoch": 0.46, + "learning_rate": 1.921611222032945e-05, + "loss": 1.2285, + "step": 5454 + }, + { + "epoch": 0.46, + "learning_rate": 1.9215759396764538e-05, + "loss": 1.2819, + "step": 5455 + }, + { + "epoch": 0.46, + "learning_rate": 1.921540649705607e-05, + "loss": 1.2748, + "step": 5456 + }, + { + "epoch": 0.46, + "learning_rate": 1.9215053521206965e-05, + "loss": 1.3607, + "step": 5457 + }, + { + "epoch": 0.46, + "learning_rate": 1.921470046922014e-05, + "loss": 1.363, + "step": 5458 + }, + { + "epoch": 0.46, + "learning_rate": 1.921434734109851e-05, + "loss": 1.375, + "step": 5459 + }, + { + "epoch": 0.46, + "learning_rate": 1.9213994136844993e-05, + "loss": 1.3254, + "step": 5460 + }, + { + "epoch": 0.46, + "learning_rate": 1.921364085646251e-05, + "loss": 1.2382, + "step": 5461 + }, + { + "epoch": 0.46, + "learning_rate": 1.9213287499953973e-05, + "loss": 1.3249, + "step": 5462 + }, + { + "epoch": 0.46, + "learning_rate": 1.9212934067322307e-05, + "loss": 1.3591, + "step": 5463 + }, + { + "epoch": 0.46, + "learning_rate": 1.9212580558570434e-05, + "loss": 1.3716, + "step": 5464 + }, + { + "epoch": 0.46, + "learning_rate": 1.921222697370127e-05, + "loss": 1.2923, + "step": 5465 + }, + { + "epoch": 0.46, + "learning_rate": 1.9211873312717738e-05, + "loss": 1.4093, + "step": 5466 + }, + { + "epoch": 0.46, + "learning_rate": 1.921151957562276e-05, + "loss": 1.4596, + "step": 5467 + }, + { + "epoch": 0.46, + "learning_rate": 1.9211165762419263e-05, + "loss": 1.2288, + "step": 5468 + }, + { + "epoch": 0.46, + "learning_rate": 1.9210811873110164e-05, + "loss": 1.2729, + "step": 5469 + }, + { + "epoch": 0.46, + "learning_rate": 1.9210457907698386e-05, + "loss": 1.3588, + "step": 5470 + }, + { + "epoch": 0.46, + "learning_rate": 1.9210103866186862e-05, + "loss": 1.3465, + "step": 5471 + }, + { + "epoch": 0.46, + "learning_rate": 1.9209749748578514e-05, + "loss": 1.3425, + "step": 5472 + }, + { + "epoch": 0.46, + "learning_rate": 1.920939555487626e-05, + "loss": 1.2752, + "step": 5473 + }, + { + "epoch": 0.46, + "learning_rate": 1.9209041285083034e-05, + "loss": 1.2892, + "step": 5474 + }, + { + "epoch": 0.46, + "learning_rate": 1.9208686939201766e-05, + "loss": 1.3303, + "step": 5475 + }, + { + "epoch": 0.46, + "learning_rate": 1.9208332517235375e-05, + "loss": 1.2358, + "step": 5476 + }, + { + "epoch": 0.46, + "learning_rate": 1.9207978019186793e-05, + "loss": 1.3401, + "step": 5477 + }, + { + "epoch": 0.46, + "learning_rate": 1.9207623445058953e-05, + "loss": 1.3938, + "step": 5478 + }, + { + "epoch": 0.46, + "learning_rate": 1.920726879485478e-05, + "loss": 1.2873, + "step": 5479 + }, + { + "epoch": 0.46, + "learning_rate": 1.9206914068577206e-05, + "loss": 1.3339, + "step": 5480 + }, + { + "epoch": 0.46, + "learning_rate": 1.920655926622916e-05, + "loss": 1.3994, + "step": 5481 + }, + { + "epoch": 0.46, + "learning_rate": 1.9206204387813578e-05, + "loss": 1.2816, + "step": 5482 + }, + { + "epoch": 0.46, + "learning_rate": 1.9205849433333385e-05, + "loss": 1.3544, + "step": 5483 + }, + { + "epoch": 0.46, + "learning_rate": 1.920549440279152e-05, + "loss": 1.3055, + "step": 5484 + }, + { + "epoch": 0.46, + "learning_rate": 1.9205139296190915e-05, + "loss": 1.3229, + "step": 5485 + }, + { + "epoch": 0.46, + "learning_rate": 1.92047841135345e-05, + "loss": 1.334, + "step": 5486 + }, + { + "epoch": 0.46, + "learning_rate": 1.9204428854825215e-05, + "loss": 1.3298, + "step": 5487 + }, + { + "epoch": 0.46, + "learning_rate": 1.9204073520065992e-05, + "loss": 1.315, + "step": 5488 + }, + { + "epoch": 0.46, + "learning_rate": 1.920371810925977e-05, + "loss": 1.2698, + "step": 5489 + }, + { + "epoch": 0.46, + "learning_rate": 1.9203362622409482e-05, + "loss": 1.3017, + "step": 5490 + }, + { + "epoch": 0.46, + "learning_rate": 1.9203007059518066e-05, + "loss": 1.3759, + "step": 5491 + }, + { + "epoch": 0.46, + "learning_rate": 1.9202651420588458e-05, + "loss": 1.4005, + "step": 5492 + }, + { + "epoch": 0.46, + "learning_rate": 1.9202295705623603e-05, + "loss": 1.3602, + "step": 5493 + }, + { + "epoch": 0.46, + "learning_rate": 1.9201939914626434e-05, + "loss": 1.2753, + "step": 5494 + }, + { + "epoch": 0.46, + "learning_rate": 1.9201584047599893e-05, + "loss": 1.2172, + "step": 5495 + }, + { + "epoch": 0.46, + "learning_rate": 1.920122810454692e-05, + "loss": 1.3792, + "step": 5496 + }, + { + "epoch": 0.46, + "learning_rate": 1.9200872085470454e-05, + "loss": 1.2565, + "step": 5497 + }, + { + "epoch": 0.46, + "learning_rate": 1.9200515990373437e-05, + "loss": 1.3543, + "step": 5498 + }, + { + "epoch": 0.46, + "learning_rate": 1.9200159819258816e-05, + "loss": 1.2986, + "step": 5499 + }, + { + "epoch": 0.46, + "learning_rate": 1.919980357212953e-05, + "loss": 1.2958, + "step": 5500 + }, + { + "epoch": 0.46, + "learning_rate": 1.919944724898852e-05, + "loss": 1.3004, + "step": 5501 + }, + { + "epoch": 0.46, + "learning_rate": 1.9199090849838734e-05, + "loss": 1.458, + "step": 5502 + }, + { + "epoch": 0.46, + "learning_rate": 1.9198734374683116e-05, + "loss": 1.3567, + "step": 5503 + }, + { + "epoch": 0.46, + "learning_rate": 1.919837782352461e-05, + "loss": 1.2974, + "step": 5504 + }, + { + "epoch": 0.46, + "learning_rate": 1.919802119636616e-05, + "loss": 1.3159, + "step": 5505 + }, + { + "epoch": 0.46, + "learning_rate": 1.919766449321072e-05, + "loss": 1.4574, + "step": 5506 + }, + { + "epoch": 0.46, + "learning_rate": 1.9197307714061232e-05, + "loss": 1.3598, + "step": 5507 + }, + { + "epoch": 0.46, + "learning_rate": 1.9196950858920637e-05, + "loss": 1.2904, + "step": 5508 + }, + { + "epoch": 0.46, + "learning_rate": 1.9196593927791897e-05, + "loss": 1.2712, + "step": 5509 + }, + { + "epoch": 0.46, + "learning_rate": 1.9196236920677952e-05, + "loss": 1.3066, + "step": 5510 + }, + { + "epoch": 0.46, + "learning_rate": 1.9195879837581757e-05, + "loss": 1.3065, + "step": 5511 + }, + { + "epoch": 0.46, + "learning_rate": 1.919552267850626e-05, + "loss": 1.3822, + "step": 5512 + }, + { + "epoch": 0.46, + "learning_rate": 1.919516544345441e-05, + "loss": 1.3229, + "step": 5513 + }, + { + "epoch": 0.46, + "learning_rate": 1.919480813242916e-05, + "loss": 1.3309, + "step": 5514 + }, + { + "epoch": 0.46, + "learning_rate": 1.9194450745433463e-05, + "loss": 1.2218, + "step": 5515 + }, + { + "epoch": 0.46, + "learning_rate": 1.919409328247027e-05, + "loss": 1.3038, + "step": 5516 + }, + { + "epoch": 0.46, + "learning_rate": 1.9193735743542538e-05, + "loss": 1.2654, + "step": 5517 + }, + { + "epoch": 0.46, + "learning_rate": 1.9193378128653217e-05, + "loss": 1.3357, + "step": 5518 + }, + { + "epoch": 0.46, + "learning_rate": 1.9193020437805264e-05, + "loss": 1.2757, + "step": 5519 + }, + { + "epoch": 0.46, + "learning_rate": 1.9192662671001635e-05, + "loss": 1.2999, + "step": 5520 + }, + { + "epoch": 0.46, + "learning_rate": 1.9192304828245282e-05, + "loss": 1.349, + "step": 5521 + }, + { + "epoch": 0.46, + "learning_rate": 1.919194690953917e-05, + "loss": 1.2899, + "step": 5522 + }, + { + "epoch": 0.46, + "learning_rate": 1.9191588914886248e-05, + "loss": 1.2925, + "step": 5523 + }, + { + "epoch": 0.47, + "learning_rate": 1.9191230844289474e-05, + "loss": 1.2563, + "step": 5524 + }, + { + "epoch": 0.47, + "learning_rate": 1.9190872697751813e-05, + "loss": 1.2784, + "step": 5525 + }, + { + "epoch": 0.47, + "learning_rate": 1.9190514475276218e-05, + "loss": 1.2367, + "step": 5526 + }, + { + "epoch": 0.47, + "learning_rate": 1.9190156176865654e-05, + "loss": 1.4, + "step": 5527 + }, + { + "epoch": 0.47, + "learning_rate": 1.9189797802523076e-05, + "loss": 1.2675, + "step": 5528 + }, + { + "epoch": 0.47, + "learning_rate": 1.9189439352251448e-05, + "loss": 1.4351, + "step": 5529 + }, + { + "epoch": 0.47, + "learning_rate": 1.918908082605373e-05, + "loss": 1.4091, + "step": 5530 + }, + { + "epoch": 0.47, + "learning_rate": 1.9188722223932887e-05, + "loss": 1.3205, + "step": 5531 + }, + { + "epoch": 0.47, + "learning_rate": 1.918836354589188e-05, + "loss": 1.2991, + "step": 5532 + }, + { + "epoch": 0.47, + "learning_rate": 1.9188004791933675e-05, + "loss": 1.3045, + "step": 5533 + }, + { + "epoch": 0.47, + "learning_rate": 1.918764596206123e-05, + "loss": 1.2927, + "step": 5534 + }, + { + "epoch": 0.47, + "learning_rate": 1.9187287056277516e-05, + "loss": 1.2743, + "step": 5535 + }, + { + "epoch": 0.47, + "learning_rate": 1.9186928074585494e-05, + "loss": 1.3285, + "step": 5536 + }, + { + "epoch": 0.47, + "learning_rate": 1.9186569016988132e-05, + "loss": 1.4098, + "step": 5537 + }, + { + "epoch": 0.47, + "learning_rate": 1.91862098834884e-05, + "loss": 1.3861, + "step": 5538 + }, + { + "epoch": 0.47, + "learning_rate": 1.918585067408926e-05, + "loss": 1.3502, + "step": 5539 + }, + { + "epoch": 0.47, + "learning_rate": 1.918549138879368e-05, + "loss": 1.2307, + "step": 5540 + }, + { + "epoch": 0.47, + "learning_rate": 1.9185132027604632e-05, + "loss": 1.3085, + "step": 5541 + }, + { + "epoch": 0.47, + "learning_rate": 1.9184772590525082e-05, + "loss": 1.3042, + "step": 5542 + }, + { + "epoch": 0.47, + "learning_rate": 1.9184413077558004e-05, + "loss": 1.3046, + "step": 5543 + }, + { + "epoch": 0.47, + "learning_rate": 1.9184053488706365e-05, + "loss": 1.298, + "step": 5544 + }, + { + "epoch": 0.47, + "learning_rate": 1.9183693823973137e-05, + "loss": 1.284, + "step": 5545 + }, + { + "epoch": 0.47, + "learning_rate": 1.9183334083361292e-05, + "loss": 1.3556, + "step": 5546 + }, + { + "epoch": 0.47, + "learning_rate": 1.9182974266873798e-05, + "loss": 1.3585, + "step": 5547 + }, + { + "epoch": 0.47, + "learning_rate": 1.9182614374513635e-05, + "loss": 1.3382, + "step": 5548 + }, + { + "epoch": 0.47, + "learning_rate": 1.9182254406283773e-05, + "loss": 1.3822, + "step": 5549 + }, + { + "epoch": 0.47, + "learning_rate": 1.9181894362187186e-05, + "loss": 1.2314, + "step": 5550 + }, + { + "epoch": 0.47, + "learning_rate": 1.9181534242226848e-05, + "loss": 1.2148, + "step": 5551 + }, + { + "epoch": 0.47, + "learning_rate": 1.9181174046405738e-05, + "loss": 1.2845, + "step": 5552 + }, + { + "epoch": 0.47, + "learning_rate": 1.918081377472683e-05, + "loss": 1.3722, + "step": 5553 + }, + { + "epoch": 0.47, + "learning_rate": 1.9180453427193097e-05, + "loss": 1.3356, + "step": 5554 + }, + { + "epoch": 0.47, + "learning_rate": 1.9180093003807522e-05, + "loss": 1.3228, + "step": 5555 + }, + { + "epoch": 0.47, + "learning_rate": 1.9179732504573078e-05, + "loss": 1.2551, + "step": 5556 + }, + { + "epoch": 0.47, + "learning_rate": 1.9179371929492752e-05, + "loss": 1.4026, + "step": 5557 + }, + { + "epoch": 0.47, + "learning_rate": 1.917901127856951e-05, + "loss": 1.3653, + "step": 5558 + }, + { + "epoch": 0.47, + "learning_rate": 1.9178650551806342e-05, + "loss": 1.3167, + "step": 5559 + }, + { + "epoch": 0.47, + "learning_rate": 1.9178289749206228e-05, + "loss": 1.3268, + "step": 5560 + }, + { + "epoch": 0.47, + "learning_rate": 1.9177928870772145e-05, + "loss": 1.4325, + "step": 5561 + }, + { + "epoch": 0.47, + "learning_rate": 1.9177567916507075e-05, + "loss": 1.3136, + "step": 5562 + }, + { + "epoch": 0.47, + "learning_rate": 1.9177206886414006e-05, + "loss": 1.4256, + "step": 5563 + }, + { + "epoch": 0.47, + "learning_rate": 1.9176845780495913e-05, + "loss": 1.243, + "step": 5564 + }, + { + "epoch": 0.47, + "learning_rate": 1.9176484598755784e-05, + "loss": 1.3753, + "step": 5565 + }, + { + "epoch": 0.47, + "learning_rate": 1.91761233411966e-05, + "loss": 1.3924, + "step": 5566 + }, + { + "epoch": 0.47, + "learning_rate": 1.9175762007821353e-05, + "loss": 1.1837, + "step": 5567 + }, + { + "epoch": 0.47, + "learning_rate": 1.917540059863302e-05, + "loss": 1.2919, + "step": 5568 + }, + { + "epoch": 0.47, + "learning_rate": 1.917503911363459e-05, + "loss": 1.3588, + "step": 5569 + }, + { + "epoch": 0.47, + "learning_rate": 1.9174677552829055e-05, + "loss": 1.4367, + "step": 5570 + }, + { + "epoch": 0.47, + "learning_rate": 1.9174315916219396e-05, + "loss": 1.2881, + "step": 5571 + }, + { + "epoch": 0.47, + "learning_rate": 1.91739542038086e-05, + "loss": 1.3297, + "step": 5572 + }, + { + "epoch": 0.47, + "learning_rate": 1.9173592415599664e-05, + "loss": 1.4045, + "step": 5573 + }, + { + "epoch": 0.47, + "learning_rate": 1.9173230551595567e-05, + "loss": 1.3596, + "step": 5574 + }, + { + "epoch": 0.47, + "learning_rate": 1.9172868611799304e-05, + "loss": 1.3056, + "step": 5575 + }, + { + "epoch": 0.47, + "learning_rate": 1.9172506596213864e-05, + "loss": 1.3307, + "step": 5576 + }, + { + "epoch": 0.47, + "learning_rate": 1.9172144504842237e-05, + "loss": 1.3742, + "step": 5577 + }, + { + "epoch": 0.47, + "learning_rate": 1.917178233768742e-05, + "loss": 1.2799, + "step": 5578 + }, + { + "epoch": 0.47, + "learning_rate": 1.9171420094752404e-05, + "loss": 1.3462, + "step": 5579 + }, + { + "epoch": 0.47, + "learning_rate": 1.9171057776040178e-05, + "loss": 1.3492, + "step": 5580 + }, + { + "epoch": 0.47, + "learning_rate": 1.9170695381553735e-05, + "loss": 1.4078, + "step": 5581 + }, + { + "epoch": 0.47, + "learning_rate": 1.9170332911296076e-05, + "loss": 1.3934, + "step": 5582 + }, + { + "epoch": 0.47, + "learning_rate": 1.9169970365270188e-05, + "loss": 1.3417, + "step": 5583 + }, + { + "epoch": 0.47, + "learning_rate": 1.9169607743479073e-05, + "loss": 1.3342, + "step": 5584 + }, + { + "epoch": 0.47, + "learning_rate": 1.9169245045925722e-05, + "loss": 1.3801, + "step": 5585 + }, + { + "epoch": 0.47, + "learning_rate": 1.9168882272613135e-05, + "loss": 1.2856, + "step": 5586 + }, + { + "epoch": 0.47, + "learning_rate": 1.9168519423544304e-05, + "loss": 1.3037, + "step": 5587 + }, + { + "epoch": 0.47, + "learning_rate": 1.916815649872224e-05, + "loss": 1.293, + "step": 5588 + }, + { + "epoch": 0.47, + "learning_rate": 1.9167793498149926e-05, + "loss": 1.4163, + "step": 5589 + }, + { + "epoch": 0.47, + "learning_rate": 1.9167430421830367e-05, + "loss": 1.2807, + "step": 5590 + }, + { + "epoch": 0.47, + "learning_rate": 1.9167067269766568e-05, + "loss": 1.407, + "step": 5591 + }, + { + "epoch": 0.47, + "learning_rate": 1.9166704041961526e-05, + "loss": 1.3485, + "step": 5592 + }, + { + "epoch": 0.47, + "learning_rate": 1.916634073841824e-05, + "loss": 1.4276, + "step": 5593 + }, + { + "epoch": 0.47, + "learning_rate": 1.9165977359139713e-05, + "loss": 1.3103, + "step": 5594 + }, + { + "epoch": 0.47, + "learning_rate": 1.9165613904128945e-05, + "loss": 1.2569, + "step": 5595 + }, + { + "epoch": 0.47, + "learning_rate": 1.9165250373388943e-05, + "loss": 1.2927, + "step": 5596 + }, + { + "epoch": 0.47, + "learning_rate": 1.916488676692271e-05, + "loss": 1.3235, + "step": 5597 + }, + { + "epoch": 0.47, + "learning_rate": 1.916452308473325e-05, + "loss": 1.4249, + "step": 5598 + }, + { + "epoch": 0.47, + "learning_rate": 1.916415932682357e-05, + "loss": 1.2875, + "step": 5599 + }, + { + "epoch": 0.47, + "learning_rate": 1.9163795493196667e-05, + "loss": 1.326, + "step": 5600 + }, + { + "epoch": 0.47, + "learning_rate": 1.9163431583855557e-05, + "loss": 1.3256, + "step": 5601 + }, + { + "epoch": 0.47, + "learning_rate": 1.916306759880324e-05, + "loss": 1.3448, + "step": 5602 + }, + { + "epoch": 0.47, + "learning_rate": 1.9162703538042725e-05, + "loss": 1.3953, + "step": 5603 + }, + { + "epoch": 0.47, + "learning_rate": 1.9162339401577027e-05, + "loss": 1.3694, + "step": 5604 + }, + { + "epoch": 0.47, + "learning_rate": 1.9161975189409143e-05, + "loss": 1.3933, + "step": 5605 + }, + { + "epoch": 0.47, + "learning_rate": 1.916161090154209e-05, + "loss": 1.3366, + "step": 5606 + }, + { + "epoch": 0.47, + "learning_rate": 1.9161246537978873e-05, + "loss": 1.2516, + "step": 5607 + }, + { + "epoch": 0.47, + "learning_rate": 1.9160882098722505e-05, + "loss": 1.2566, + "step": 5608 + }, + { + "epoch": 0.47, + "learning_rate": 1.9160517583776e-05, + "loss": 1.4347, + "step": 5609 + }, + { + "epoch": 0.47, + "learning_rate": 1.916015299314237e-05, + "loss": 1.388, + "step": 5610 + }, + { + "epoch": 0.47, + "learning_rate": 1.9159788326824616e-05, + "loss": 1.2964, + "step": 5611 + }, + { + "epoch": 0.47, + "learning_rate": 1.9159423584825765e-05, + "loss": 1.2276, + "step": 5612 + }, + { + "epoch": 0.47, + "learning_rate": 1.9159058767148822e-05, + "loss": 1.4214, + "step": 5613 + }, + { + "epoch": 0.47, + "learning_rate": 1.9158693873796807e-05, + "loss": 1.4048, + "step": 5614 + }, + { + "epoch": 0.47, + "learning_rate": 1.9158328904772733e-05, + "loss": 1.4306, + "step": 5615 + }, + { + "epoch": 0.47, + "learning_rate": 1.915796386007961e-05, + "loss": 1.2404, + "step": 5616 + }, + { + "epoch": 0.47, + "learning_rate": 1.915759873972046e-05, + "loss": 1.3384, + "step": 5617 + }, + { + "epoch": 0.47, + "learning_rate": 1.9157233543698303e-05, + "loss": 1.4381, + "step": 5618 + }, + { + "epoch": 0.47, + "learning_rate": 1.9156868272016146e-05, + "loss": 1.332, + "step": 5619 + }, + { + "epoch": 0.47, + "learning_rate": 1.9156502924677016e-05, + "loss": 1.2521, + "step": 5620 + }, + { + "epoch": 0.47, + "learning_rate": 1.915613750168393e-05, + "loss": 1.2847, + "step": 5621 + }, + { + "epoch": 0.47, + "learning_rate": 1.91557720030399e-05, + "loss": 1.2647, + "step": 5622 + }, + { + "epoch": 0.47, + "learning_rate": 1.9155406428747955e-05, + "loss": 1.3381, + "step": 5623 + }, + { + "epoch": 0.47, + "learning_rate": 1.9155040778811113e-05, + "loss": 1.3406, + "step": 5624 + }, + { + "epoch": 0.47, + "learning_rate": 1.9154675053232392e-05, + "loss": 1.4162, + "step": 5625 + }, + { + "epoch": 0.47, + "learning_rate": 1.9154309252014814e-05, + "loss": 1.2971, + "step": 5626 + }, + { + "epoch": 0.47, + "learning_rate": 1.9153943375161406e-05, + "loss": 1.3585, + "step": 5627 + }, + { + "epoch": 0.47, + "learning_rate": 1.915357742267519e-05, + "loss": 1.3907, + "step": 5628 + }, + { + "epoch": 0.47, + "learning_rate": 1.9153211394559187e-05, + "loss": 1.296, + "step": 5629 + }, + { + "epoch": 0.47, + "learning_rate": 1.915284529081642e-05, + "loss": 1.3186, + "step": 5630 + }, + { + "epoch": 0.47, + "learning_rate": 1.9152479111449914e-05, + "loss": 1.2582, + "step": 5631 + }, + { + "epoch": 0.47, + "learning_rate": 1.9152112856462703e-05, + "loss": 1.3031, + "step": 5632 + }, + { + "epoch": 0.47, + "learning_rate": 1.91517465258578e-05, + "loss": 1.3205, + "step": 5633 + }, + { + "epoch": 0.47, + "learning_rate": 1.915138011963824e-05, + "loss": 1.2997, + "step": 5634 + }, + { + "epoch": 0.47, + "learning_rate": 1.9151013637807053e-05, + "loss": 1.2619, + "step": 5635 + }, + { + "epoch": 0.47, + "learning_rate": 1.915064708036726e-05, + "loss": 1.2976, + "step": 5636 + }, + { + "epoch": 0.47, + "learning_rate": 1.9150280447321888e-05, + "loss": 1.2386, + "step": 5637 + }, + { + "epoch": 0.47, + "learning_rate": 1.9149913738673974e-05, + "loss": 1.2074, + "step": 5638 + }, + { + "epoch": 0.47, + "learning_rate": 1.9149546954426546e-05, + "loss": 1.3442, + "step": 5639 + }, + { + "epoch": 0.47, + "learning_rate": 1.9149180094582628e-05, + "loss": 1.3659, + "step": 5640 + }, + { + "epoch": 0.47, + "learning_rate": 1.914881315914526e-05, + "loss": 1.2768, + "step": 5641 + }, + { + "epoch": 0.48, + "learning_rate": 1.914844614811747e-05, + "loss": 1.2496, + "step": 5642 + }, + { + "epoch": 0.48, + "learning_rate": 1.914807906150229e-05, + "loss": 1.3274, + "step": 5643 + }, + { + "epoch": 0.48, + "learning_rate": 1.9147711899302754e-05, + "loss": 1.2208, + "step": 5644 + }, + { + "epoch": 0.48, + "learning_rate": 1.914734466152189e-05, + "loss": 1.3319, + "step": 5645 + }, + { + "epoch": 0.48, + "learning_rate": 1.9146977348162737e-05, + "loss": 1.389, + "step": 5646 + }, + { + "epoch": 0.48, + "learning_rate": 1.914660995922833e-05, + "loss": 1.3871, + "step": 5647 + }, + { + "epoch": 0.48, + "learning_rate": 1.914624249472171e-05, + "loss": 1.3335, + "step": 5648 + }, + { + "epoch": 0.48, + "learning_rate": 1.9145874954645903e-05, + "loss": 1.2902, + "step": 5649 + }, + { + "epoch": 0.48, + "learning_rate": 1.9145507339003952e-05, + "loss": 1.3336, + "step": 5650 + }, + { + "epoch": 0.48, + "learning_rate": 1.914513964779889e-05, + "loss": 1.3525, + "step": 5651 + }, + { + "epoch": 0.48, + "learning_rate": 1.914477188103376e-05, + "loss": 1.3613, + "step": 5652 + }, + { + "epoch": 0.48, + "learning_rate": 1.9144404038711598e-05, + "loss": 1.2849, + "step": 5653 + }, + { + "epoch": 0.48, + "learning_rate": 1.914403612083544e-05, + "loss": 1.359, + "step": 5654 + }, + { + "epoch": 0.48, + "learning_rate": 1.914366812740833e-05, + "loss": 1.3915, + "step": 5655 + }, + { + "epoch": 0.48, + "learning_rate": 1.914330005843331e-05, + "loss": 1.3348, + "step": 5656 + }, + { + "epoch": 0.48, + "learning_rate": 1.9142931913913416e-05, + "loss": 1.436, + "step": 5657 + }, + { + "epoch": 0.48, + "learning_rate": 1.914256369385169e-05, + "loss": 1.2984, + "step": 5658 + }, + { + "epoch": 0.48, + "learning_rate": 1.9142195398251186e-05, + "loss": 1.3721, + "step": 5659 + }, + { + "epoch": 0.48, + "learning_rate": 1.9141827027114928e-05, + "loss": 1.4203, + "step": 5660 + }, + { + "epoch": 0.48, + "learning_rate": 1.9141458580445974e-05, + "loss": 1.3392, + "step": 5661 + }, + { + "epoch": 0.48, + "learning_rate": 1.9141090058247364e-05, + "loss": 1.2117, + "step": 5662 + }, + { + "epoch": 0.48, + "learning_rate": 1.9140721460522142e-05, + "loss": 1.3599, + "step": 5663 + }, + { + "epoch": 0.48, + "learning_rate": 1.9140352787273352e-05, + "loss": 1.4113, + "step": 5664 + }, + { + "epoch": 0.48, + "learning_rate": 1.9139984038504044e-05, + "loss": 1.383, + "step": 5665 + }, + { + "epoch": 0.48, + "learning_rate": 1.9139615214217265e-05, + "loss": 1.2989, + "step": 5666 + }, + { + "epoch": 0.48, + "learning_rate": 1.9139246314416056e-05, + "loss": 1.2814, + "step": 5667 + }, + { + "epoch": 0.48, + "learning_rate": 1.913887733910347e-05, + "loss": 1.3064, + "step": 5668 + }, + { + "epoch": 0.48, + "learning_rate": 1.9138508288282553e-05, + "loss": 1.4243, + "step": 5669 + }, + { + "epoch": 0.48, + "learning_rate": 1.9138139161956358e-05, + "loss": 1.3606, + "step": 5670 + }, + { + "epoch": 0.48, + "learning_rate": 1.9137769960127933e-05, + "loss": 1.3441, + "step": 5671 + }, + { + "epoch": 0.48, + "learning_rate": 1.9137400682800327e-05, + "loss": 1.2889, + "step": 5672 + }, + { + "epoch": 0.48, + "learning_rate": 1.913703132997659e-05, + "loss": 1.2787, + "step": 5673 + }, + { + "epoch": 0.48, + "learning_rate": 1.913666190165978e-05, + "loss": 1.2854, + "step": 5674 + }, + { + "epoch": 0.48, + "learning_rate": 1.9136292397852944e-05, + "loss": 1.3244, + "step": 5675 + }, + { + "epoch": 0.48, + "learning_rate": 1.9135922818559136e-05, + "loss": 1.3597, + "step": 5676 + }, + { + "epoch": 0.48, + "learning_rate": 1.9135553163781408e-05, + "loss": 1.3124, + "step": 5677 + }, + { + "epoch": 0.48, + "learning_rate": 1.913518343352282e-05, + "loss": 1.1867, + "step": 5678 + }, + { + "epoch": 0.48, + "learning_rate": 1.9134813627786418e-05, + "loss": 1.3064, + "step": 5679 + }, + { + "epoch": 0.48, + "learning_rate": 1.9134443746575265e-05, + "loss": 1.3605, + "step": 5680 + }, + { + "epoch": 0.48, + "learning_rate": 1.9134073789892416e-05, + "loss": 1.3139, + "step": 5681 + }, + { + "epoch": 0.48, + "learning_rate": 1.9133703757740924e-05, + "loss": 1.2994, + "step": 5682 + }, + { + "epoch": 0.48, + "learning_rate": 1.9133333650123847e-05, + "loss": 1.2886, + "step": 5683 + }, + { + "epoch": 0.48, + "learning_rate": 1.9132963467044247e-05, + "loss": 1.3436, + "step": 5684 + }, + { + "epoch": 0.48, + "learning_rate": 1.9132593208505178e-05, + "loss": 1.2886, + "step": 5685 + }, + { + "epoch": 0.48, + "learning_rate": 1.9132222874509706e-05, + "loss": 1.3427, + "step": 5686 + }, + { + "epoch": 0.48, + "learning_rate": 1.913185246506088e-05, + "loss": 1.2328, + "step": 5687 + }, + { + "epoch": 0.48, + "learning_rate": 1.913148198016177e-05, + "loss": 1.305, + "step": 5688 + }, + { + "epoch": 0.48, + "learning_rate": 1.9131111419815428e-05, + "loss": 1.3745, + "step": 5689 + }, + { + "epoch": 0.48, + "learning_rate": 1.9130740784024927e-05, + "loss": 1.2743, + "step": 5690 + }, + { + "epoch": 0.48, + "learning_rate": 1.913037007279332e-05, + "loss": 1.2101, + "step": 5691 + }, + { + "epoch": 0.48, + "learning_rate": 1.9129999286123676e-05, + "loss": 1.3476, + "step": 5692 + }, + { + "epoch": 0.48, + "learning_rate": 1.9129628424019056e-05, + "loss": 1.2907, + "step": 5693 + }, + { + "epoch": 0.48, + "learning_rate": 1.9129257486482523e-05, + "loss": 1.2714, + "step": 5694 + }, + { + "epoch": 0.48, + "learning_rate": 1.912888647351714e-05, + "loss": 1.356, + "step": 5695 + }, + { + "epoch": 0.48, + "learning_rate": 1.912851538512598e-05, + "loss": 1.3112, + "step": 5696 + }, + { + "epoch": 0.48, + "learning_rate": 1.9128144221312102e-05, + "loss": 1.2927, + "step": 5697 + }, + { + "epoch": 0.48, + "learning_rate": 1.9127772982078575e-05, + "loss": 1.3263, + "step": 5698 + }, + { + "epoch": 0.48, + "learning_rate": 1.9127401667428466e-05, + "loss": 1.2577, + "step": 5699 + }, + { + "epoch": 0.48, + "learning_rate": 1.9127030277364845e-05, + "loss": 1.3689, + "step": 5700 + }, + { + "epoch": 0.48, + "learning_rate": 1.912665881189078e-05, + "loss": 1.3918, + "step": 5701 + }, + { + "epoch": 0.48, + "learning_rate": 1.9126287271009335e-05, + "loss": 1.2769, + "step": 5702 + }, + { + "epoch": 0.48, + "learning_rate": 1.9125915654723583e-05, + "loss": 1.2446, + "step": 5703 + }, + { + "epoch": 0.48, + "learning_rate": 1.91255439630366e-05, + "loss": 1.2882, + "step": 5704 + }, + { + "epoch": 0.48, + "learning_rate": 1.912517219595145e-05, + "loss": 1.3295, + "step": 5705 + }, + { + "epoch": 0.48, + "learning_rate": 1.9124800353471206e-05, + "loss": 1.4008, + "step": 5706 + }, + { + "epoch": 0.48, + "learning_rate": 1.9124428435598944e-05, + "loss": 1.364, + "step": 5707 + }, + { + "epoch": 0.48, + "learning_rate": 1.9124056442337734e-05, + "loss": 1.3116, + "step": 5708 + }, + { + "epoch": 0.48, + "learning_rate": 1.9123684373690646e-05, + "loss": 1.3386, + "step": 5709 + }, + { + "epoch": 0.48, + "learning_rate": 1.912331222966076e-05, + "loss": 1.2245, + "step": 5710 + }, + { + "epoch": 0.48, + "learning_rate": 1.9122940010251144e-05, + "loss": 1.3728, + "step": 5711 + }, + { + "epoch": 0.48, + "learning_rate": 1.9122567715464883e-05, + "loss": 1.3023, + "step": 5712 + }, + { + "epoch": 0.48, + "learning_rate": 1.9122195345305047e-05, + "loss": 1.237, + "step": 5713 + }, + { + "epoch": 0.48, + "learning_rate": 1.9121822899774715e-05, + "loss": 1.3725, + "step": 5714 + }, + { + "epoch": 0.48, + "learning_rate": 1.912145037887696e-05, + "loss": 1.2912, + "step": 5715 + }, + { + "epoch": 0.48, + "learning_rate": 1.9121077782614868e-05, + "loss": 1.3124, + "step": 5716 + }, + { + "epoch": 0.48, + "learning_rate": 1.9120705110991506e-05, + "loss": 1.3479, + "step": 5717 + }, + { + "epoch": 0.48, + "learning_rate": 1.9120332364009964e-05, + "loss": 1.2954, + "step": 5718 + }, + { + "epoch": 0.48, + "learning_rate": 1.9119959541673312e-05, + "loss": 1.2701, + "step": 5719 + }, + { + "epoch": 0.48, + "learning_rate": 1.911958664398464e-05, + "loss": 1.3053, + "step": 5720 + }, + { + "epoch": 0.48, + "learning_rate": 1.9119213670947022e-05, + "loss": 1.2405, + "step": 5721 + }, + { + "epoch": 0.48, + "learning_rate": 1.911884062256354e-05, + "loss": 1.3049, + "step": 5722 + }, + { + "epoch": 0.48, + "learning_rate": 1.9118467498837282e-05, + "loss": 1.2262, + "step": 5723 + }, + { + "epoch": 0.48, + "learning_rate": 1.911809429977133e-05, + "loss": 1.3219, + "step": 5724 + }, + { + "epoch": 0.48, + "learning_rate": 1.911772102536876e-05, + "loss": 1.3323, + "step": 5725 + }, + { + "epoch": 0.48, + "learning_rate": 1.9117347675632662e-05, + "loss": 1.4515, + "step": 5726 + }, + { + "epoch": 0.48, + "learning_rate": 1.911697425056612e-05, + "loss": 1.4605, + "step": 5727 + }, + { + "epoch": 0.48, + "learning_rate": 1.911660075017222e-05, + "loss": 1.3217, + "step": 5728 + }, + { + "epoch": 0.48, + "learning_rate": 1.9116227174454044e-05, + "loss": 1.2743, + "step": 5729 + }, + { + "epoch": 0.48, + "learning_rate": 1.911585352341468e-05, + "loss": 1.3125, + "step": 5730 + }, + { + "epoch": 0.48, + "learning_rate": 1.9115479797057223e-05, + "loss": 1.3104, + "step": 5731 + }, + { + "epoch": 0.48, + "learning_rate": 1.9115105995384754e-05, + "loss": 1.3176, + "step": 5732 + }, + { + "epoch": 0.48, + "learning_rate": 1.911473211840036e-05, + "loss": 1.3081, + "step": 5733 + }, + { + "epoch": 0.48, + "learning_rate": 1.911435816610713e-05, + "loss": 1.2934, + "step": 5734 + }, + { + "epoch": 0.48, + "learning_rate": 1.9113984138508156e-05, + "loss": 1.3111, + "step": 5735 + }, + { + "epoch": 0.48, + "learning_rate": 1.911361003560653e-05, + "loss": 1.338, + "step": 5736 + }, + { + "epoch": 0.48, + "learning_rate": 1.911323585740534e-05, + "loss": 1.2416, + "step": 5737 + }, + { + "epoch": 0.48, + "learning_rate": 1.911286160390768e-05, + "loss": 1.2833, + "step": 5738 + }, + { + "epoch": 0.48, + "learning_rate": 1.911248727511664e-05, + "loss": 1.3437, + "step": 5739 + }, + { + "epoch": 0.48, + "learning_rate": 1.9112112871035315e-05, + "loss": 1.309, + "step": 5740 + }, + { + "epoch": 0.48, + "learning_rate": 1.9111738391666797e-05, + "loss": 1.3272, + "step": 5741 + }, + { + "epoch": 0.48, + "learning_rate": 1.911136383701418e-05, + "loss": 1.3666, + "step": 5742 + }, + { + "epoch": 0.48, + "learning_rate": 1.911098920708056e-05, + "loss": 1.3333, + "step": 5743 + }, + { + "epoch": 0.48, + "learning_rate": 1.9110614501869028e-05, + "loss": 1.2716, + "step": 5744 + }, + { + "epoch": 0.48, + "learning_rate": 1.9110239721382686e-05, + "loss": 1.4091, + "step": 5745 + }, + { + "epoch": 0.48, + "learning_rate": 1.910986486562463e-05, + "loss": 1.3612, + "step": 5746 + }, + { + "epoch": 0.48, + "learning_rate": 1.910948993459795e-05, + "loss": 1.2997, + "step": 5747 + }, + { + "epoch": 0.48, + "learning_rate": 1.910911492830575e-05, + "loss": 1.2739, + "step": 5748 + }, + { + "epoch": 0.48, + "learning_rate": 1.9108739846751128e-05, + "loss": 1.3184, + "step": 5749 + }, + { + "epoch": 0.48, + "learning_rate": 1.9108364689937182e-05, + "loss": 1.3742, + "step": 5750 + }, + { + "epoch": 0.48, + "learning_rate": 1.9107989457867014e-05, + "loss": 1.3285, + "step": 5751 + }, + { + "epoch": 0.48, + "learning_rate": 1.910761415054372e-05, + "loss": 1.253, + "step": 5752 + }, + { + "epoch": 0.48, + "learning_rate": 1.91072387679704e-05, + "loss": 1.3271, + "step": 5753 + }, + { + "epoch": 0.48, + "learning_rate": 1.9106863310150164e-05, + "loss": 1.3038, + "step": 5754 + }, + { + "epoch": 0.48, + "learning_rate": 1.9106487777086107e-05, + "loss": 1.3193, + "step": 5755 + }, + { + "epoch": 0.48, + "learning_rate": 1.9106112168781333e-05, + "loss": 1.2868, + "step": 5756 + }, + { + "epoch": 0.48, + "learning_rate": 1.9105736485238945e-05, + "loss": 1.2595, + "step": 5757 + }, + { + "epoch": 0.48, + "learning_rate": 1.910536072646205e-05, + "loss": 1.318, + "step": 5758 + }, + { + "epoch": 0.48, + "learning_rate": 1.910498489245375e-05, + "loss": 1.2576, + "step": 5759 + }, + { + "epoch": 0.48, + "learning_rate": 1.910460898321715e-05, + "loss": 1.3817, + "step": 5760 + }, + { + "epoch": 0.49, + "learning_rate": 1.910423299875536e-05, + "loss": 1.3086, + "step": 5761 + }, + { + "epoch": 0.49, + "learning_rate": 1.9103856939071478e-05, + "loss": 1.3041, + "step": 5762 + }, + { + "epoch": 0.49, + "learning_rate": 1.9103480804168618e-05, + "loss": 1.2614, + "step": 5763 + }, + { + "epoch": 0.49, + "learning_rate": 1.910310459404989e-05, + "loss": 1.2759, + "step": 5764 + }, + { + "epoch": 0.49, + "learning_rate": 1.9102728308718392e-05, + "loss": 1.2485, + "step": 5765 + }, + { + "epoch": 0.49, + "learning_rate": 1.9102351948177244e-05, + "loss": 1.2257, + "step": 5766 + }, + { + "epoch": 0.49, + "learning_rate": 1.910197551242955e-05, + "loss": 1.3562, + "step": 5767 + }, + { + "epoch": 0.49, + "learning_rate": 1.9101599001478422e-05, + "loss": 1.2599, + "step": 5768 + }, + { + "epoch": 0.49, + "learning_rate": 1.910122241532697e-05, + "loss": 1.298, + "step": 5769 + }, + { + "epoch": 0.49, + "learning_rate": 1.9100845753978302e-05, + "loss": 1.2431, + "step": 5770 + }, + { + "epoch": 0.49, + "learning_rate": 1.9100469017435536e-05, + "loss": 1.3774, + "step": 5771 + }, + { + "epoch": 0.49, + "learning_rate": 1.9100092205701787e-05, + "loss": 1.4953, + "step": 5772 + }, + { + "epoch": 0.49, + "learning_rate": 1.909971531878016e-05, + "loss": 1.3625, + "step": 5773 + }, + { + "epoch": 0.49, + "learning_rate": 1.9099338356673772e-05, + "loss": 1.3612, + "step": 5774 + }, + { + "epoch": 0.49, + "learning_rate": 1.9098961319385738e-05, + "loss": 1.3057, + "step": 5775 + }, + { + "epoch": 0.49, + "learning_rate": 1.9098584206919175e-05, + "loss": 1.3055, + "step": 5776 + }, + { + "epoch": 0.49, + "learning_rate": 1.9098207019277198e-05, + "loss": 1.2734, + "step": 5777 + }, + { + "epoch": 0.49, + "learning_rate": 1.909782975646292e-05, + "loss": 1.3688, + "step": 5778 + }, + { + "epoch": 0.49, + "learning_rate": 1.9097452418479464e-05, + "loss": 1.3571, + "step": 5779 + }, + { + "epoch": 0.49, + "learning_rate": 1.9097075005329943e-05, + "loss": 1.4203, + "step": 5780 + }, + { + "epoch": 0.49, + "learning_rate": 1.9096697517017478e-05, + "loss": 1.3409, + "step": 5781 + }, + { + "epoch": 0.49, + "learning_rate": 1.9096319953545186e-05, + "loss": 1.3382, + "step": 5782 + }, + { + "epoch": 0.49, + "learning_rate": 1.9095942314916187e-05, + "loss": 1.2664, + "step": 5783 + }, + { + "epoch": 0.49, + "learning_rate": 1.9095564601133603e-05, + "loss": 1.2087, + "step": 5784 + }, + { + "epoch": 0.49, + "learning_rate": 1.9095186812200552e-05, + "loss": 1.3819, + "step": 5785 + }, + { + "epoch": 0.49, + "learning_rate": 1.9094808948120157e-05, + "loss": 1.313, + "step": 5786 + }, + { + "epoch": 0.49, + "learning_rate": 1.9094431008895536e-05, + "loss": 1.2384, + "step": 5787 + }, + { + "epoch": 0.49, + "learning_rate": 1.9094052994529823e-05, + "loss": 1.3851, + "step": 5788 + }, + { + "epoch": 0.49, + "learning_rate": 1.909367490502613e-05, + "loss": 1.3679, + "step": 5789 + }, + { + "epoch": 0.49, + "learning_rate": 1.9093296740387584e-05, + "loss": 1.355, + "step": 5790 + }, + { + "epoch": 0.49, + "learning_rate": 1.909291850061731e-05, + "loss": 1.3268, + "step": 5791 + }, + { + "epoch": 0.49, + "learning_rate": 1.909254018571843e-05, + "loss": 1.3548, + "step": 5792 + }, + { + "epoch": 0.49, + "learning_rate": 1.9092161795694077e-05, + "loss": 1.3148, + "step": 5793 + }, + { + "epoch": 0.49, + "learning_rate": 1.9091783330547377e-05, + "loss": 1.2812, + "step": 5794 + }, + { + "epoch": 0.49, + "learning_rate": 1.9091404790281448e-05, + "loss": 1.2774, + "step": 5795 + }, + { + "epoch": 0.49, + "learning_rate": 1.9091026174899424e-05, + "loss": 1.289, + "step": 5796 + }, + { + "epoch": 0.49, + "learning_rate": 1.9090647484404433e-05, + "loss": 1.2847, + "step": 5797 + }, + { + "epoch": 0.49, + "learning_rate": 1.9090268718799604e-05, + "loss": 1.3145, + "step": 5798 + }, + { + "epoch": 0.49, + "learning_rate": 1.9089889878088063e-05, + "loss": 1.3545, + "step": 5799 + }, + { + "epoch": 0.49, + "learning_rate": 1.9089510962272943e-05, + "loss": 1.2921, + "step": 5800 + }, + { + "epoch": 0.49, + "learning_rate": 1.908913197135738e-05, + "loss": 1.3563, + "step": 5801 + }, + { + "epoch": 0.49, + "learning_rate": 1.9088752905344492e-05, + "loss": 1.3765, + "step": 5802 + }, + { + "epoch": 0.49, + "learning_rate": 1.9088373764237423e-05, + "loss": 1.3983, + "step": 5803 + }, + { + "epoch": 0.49, + "learning_rate": 1.90879945480393e-05, + "loss": 1.3278, + "step": 5804 + }, + { + "epoch": 0.49, + "learning_rate": 1.9087615256753258e-05, + "loss": 1.3211, + "step": 5805 + }, + { + "epoch": 0.49, + "learning_rate": 1.9087235890382428e-05, + "loss": 1.2437, + "step": 5806 + }, + { + "epoch": 0.49, + "learning_rate": 1.908685644892995e-05, + "loss": 1.3282, + "step": 5807 + }, + { + "epoch": 0.49, + "learning_rate": 1.9086476932398952e-05, + "loss": 1.3121, + "step": 5808 + }, + { + "epoch": 0.49, + "learning_rate": 1.9086097340792578e-05, + "loss": 1.2964, + "step": 5809 + }, + { + "epoch": 0.49, + "learning_rate": 1.908571767411396e-05, + "loss": 1.4387, + "step": 5810 + }, + { + "epoch": 0.49, + "learning_rate": 1.908533793236623e-05, + "loss": 1.2508, + "step": 5811 + }, + { + "epoch": 0.49, + "learning_rate": 1.9084958115552536e-05, + "loss": 1.3861, + "step": 5812 + }, + { + "epoch": 0.49, + "learning_rate": 1.9084578223676007e-05, + "loss": 1.2297, + "step": 5813 + }, + { + "epoch": 0.49, + "learning_rate": 1.9084198256739786e-05, + "loss": 1.2977, + "step": 5814 + }, + { + "epoch": 0.49, + "learning_rate": 1.908381821474701e-05, + "loss": 1.2505, + "step": 5815 + }, + { + "epoch": 0.49, + "learning_rate": 1.908343809770082e-05, + "loss": 1.4192, + "step": 5816 + }, + { + "epoch": 0.49, + "learning_rate": 1.9083057905604363e-05, + "loss": 1.2851, + "step": 5817 + }, + { + "epoch": 0.49, + "learning_rate": 1.908267763846077e-05, + "loss": 1.2581, + "step": 5818 + }, + { + "epoch": 0.49, + "learning_rate": 1.9082297296273192e-05, + "loss": 1.3461, + "step": 5819 + }, + { + "epoch": 0.49, + "learning_rate": 1.9081916879044763e-05, + "loss": 1.3456, + "step": 5820 + }, + { + "epoch": 0.49, + "learning_rate": 1.908153638677863e-05, + "loss": 1.3574, + "step": 5821 + }, + { + "epoch": 0.49, + "learning_rate": 1.9081155819477935e-05, + "loss": 1.3425, + "step": 5822 + }, + { + "epoch": 0.49, + "learning_rate": 1.9080775177145828e-05, + "loss": 1.2611, + "step": 5823 + }, + { + "epoch": 0.49, + "learning_rate": 1.9080394459785453e-05, + "loss": 1.3783, + "step": 5824 + }, + { + "epoch": 0.49, + "learning_rate": 1.9080013667399952e-05, + "loss": 1.1556, + "step": 5825 + }, + { + "epoch": 0.49, + "learning_rate": 1.9079632799992468e-05, + "loss": 1.357, + "step": 5826 + }, + { + "epoch": 0.49, + "learning_rate": 1.9079251857566152e-05, + "loss": 1.3365, + "step": 5827 + }, + { + "epoch": 0.49, + "learning_rate": 1.9078870840124153e-05, + "loss": 1.4008, + "step": 5828 + }, + { + "epoch": 0.49, + "learning_rate": 1.9078489747669617e-05, + "loss": 1.4056, + "step": 5829 + }, + { + "epoch": 0.49, + "learning_rate": 1.9078108580205695e-05, + "loss": 1.33, + "step": 5830 + }, + { + "epoch": 0.49, + "learning_rate": 1.9077727337735532e-05, + "loss": 1.1979, + "step": 5831 + }, + { + "epoch": 0.49, + "learning_rate": 1.9077346020262282e-05, + "loss": 1.36, + "step": 5832 + }, + { + "epoch": 0.49, + "learning_rate": 1.9076964627789093e-05, + "loss": 1.2816, + "step": 5833 + }, + { + "epoch": 0.49, + "learning_rate": 1.907658316031912e-05, + "loss": 1.2761, + "step": 5834 + }, + { + "epoch": 0.49, + "learning_rate": 1.9076201617855506e-05, + "loss": 1.268, + "step": 5835 + }, + { + "epoch": 0.49, + "learning_rate": 1.9075820000401415e-05, + "loss": 1.385, + "step": 5836 + }, + { + "epoch": 0.49, + "learning_rate": 1.907543830795999e-05, + "loss": 1.2351, + "step": 5837 + }, + { + "epoch": 0.49, + "learning_rate": 1.9075056540534396e-05, + "loss": 1.3827, + "step": 5838 + }, + { + "epoch": 0.49, + "learning_rate": 1.907467469812778e-05, + "loss": 1.3661, + "step": 5839 + }, + { + "epoch": 0.49, + "learning_rate": 1.907429278074329e-05, + "loss": 1.2994, + "step": 5840 + }, + { + "epoch": 0.49, + "learning_rate": 1.9073910788384092e-05, + "loss": 1.3646, + "step": 5841 + }, + { + "epoch": 0.49, + "learning_rate": 1.9073528721053343e-05, + "loss": 1.3611, + "step": 5842 + }, + { + "epoch": 0.49, + "learning_rate": 1.9073146578754192e-05, + "loss": 1.3286, + "step": 5843 + }, + { + "epoch": 0.49, + "learning_rate": 1.9072764361489802e-05, + "loss": 1.2679, + "step": 5844 + }, + { + "epoch": 0.49, + "learning_rate": 1.9072382069263325e-05, + "loss": 1.2455, + "step": 5845 + }, + { + "epoch": 0.49, + "learning_rate": 1.907199970207793e-05, + "loss": 1.3656, + "step": 5846 + }, + { + "epoch": 0.49, + "learning_rate": 1.9071617259936765e-05, + "loss": 1.3163, + "step": 5847 + }, + { + "epoch": 0.49, + "learning_rate": 1.9071234742843e-05, + "loss": 1.4276, + "step": 5848 + }, + { + "epoch": 0.49, + "learning_rate": 1.9070852150799782e-05, + "loss": 1.385, + "step": 5849 + }, + { + "epoch": 0.49, + "learning_rate": 1.9070469483810286e-05, + "loss": 1.4173, + "step": 5850 + }, + { + "epoch": 0.49, + "learning_rate": 1.907008674187767e-05, + "loss": 1.3262, + "step": 5851 + }, + { + "epoch": 0.49, + "learning_rate": 1.9069703925005094e-05, + "loss": 1.3496, + "step": 5852 + }, + { + "epoch": 0.49, + "learning_rate": 1.906932103319572e-05, + "loss": 1.3882, + "step": 5853 + }, + { + "epoch": 0.49, + "learning_rate": 1.9068938066452713e-05, + "loss": 1.3934, + "step": 5854 + }, + { + "epoch": 0.49, + "learning_rate": 1.9068555024779237e-05, + "loss": 1.3588, + "step": 5855 + }, + { + "epoch": 0.49, + "learning_rate": 1.9068171908178455e-05, + "loss": 1.3169, + "step": 5856 + }, + { + "epoch": 0.49, + "learning_rate": 1.9067788716653538e-05, + "loss": 1.4045, + "step": 5857 + }, + { + "epoch": 0.49, + "learning_rate": 1.9067405450207647e-05, + "loss": 1.399, + "step": 5858 + }, + { + "epoch": 0.49, + "learning_rate": 1.906702210884395e-05, + "loss": 1.2546, + "step": 5859 + }, + { + "epoch": 0.49, + "learning_rate": 1.906663869256562e-05, + "loss": 1.3437, + "step": 5860 + }, + { + "epoch": 0.49, + "learning_rate": 1.9066255201375813e-05, + "loss": 1.3406, + "step": 5861 + }, + { + "epoch": 0.49, + "learning_rate": 1.9065871635277706e-05, + "loss": 1.3197, + "step": 5862 + }, + { + "epoch": 0.49, + "learning_rate": 1.9065487994274467e-05, + "loss": 1.4237, + "step": 5863 + }, + { + "epoch": 0.49, + "learning_rate": 1.9065104278369263e-05, + "loss": 1.3095, + "step": 5864 + }, + { + "epoch": 0.49, + "learning_rate": 1.9064720487565265e-05, + "loss": 1.3468, + "step": 5865 + }, + { + "epoch": 0.49, + "learning_rate": 1.9064336621865645e-05, + "loss": 1.3157, + "step": 5866 + }, + { + "epoch": 0.49, + "learning_rate": 1.9063952681273578e-05, + "loss": 1.3628, + "step": 5867 + }, + { + "epoch": 0.49, + "learning_rate": 1.9063568665792234e-05, + "loss": 1.3644, + "step": 5868 + }, + { + "epoch": 0.49, + "learning_rate": 1.9063184575424782e-05, + "loss": 1.2883, + "step": 5869 + }, + { + "epoch": 0.49, + "learning_rate": 1.90628004101744e-05, + "loss": 1.2714, + "step": 5870 + }, + { + "epoch": 0.49, + "learning_rate": 1.906241617004426e-05, + "loss": 1.3605, + "step": 5871 + }, + { + "epoch": 0.49, + "learning_rate": 1.9062031855037537e-05, + "loss": 1.394, + "step": 5872 + }, + { + "epoch": 0.49, + "learning_rate": 1.9061647465157405e-05, + "loss": 1.3706, + "step": 5873 + }, + { + "epoch": 0.49, + "learning_rate": 1.9061263000407045e-05, + "loss": 1.3955, + "step": 5874 + }, + { + "epoch": 0.49, + "learning_rate": 1.906087846078963e-05, + "loss": 1.1931, + "step": 5875 + }, + { + "epoch": 0.49, + "learning_rate": 1.9060493846308333e-05, + "loss": 1.3884, + "step": 5876 + }, + { + "epoch": 0.49, + "learning_rate": 1.906010915696634e-05, + "loss": 1.3004, + "step": 5877 + }, + { + "epoch": 0.49, + "learning_rate": 1.9059724392766825e-05, + "loss": 1.4156, + "step": 5878 + }, + { + "epoch": 0.49, + "learning_rate": 1.905933955371297e-05, + "loss": 1.3457, + "step": 5879 + }, + { + "epoch": 0.5, + "learning_rate": 1.905895463980795e-05, + "loss": 1.3771, + "step": 5880 + }, + { + "epoch": 0.5, + "learning_rate": 1.9058569651054944e-05, + "loss": 1.3174, + "step": 5881 + }, + { + "epoch": 0.5, + "learning_rate": 1.905818458745714e-05, + "loss": 1.3145, + "step": 5882 + }, + { + "epoch": 0.5, + "learning_rate": 1.9057799449017715e-05, + "loss": 1.3032, + "step": 5883 + }, + { + "epoch": 0.5, + "learning_rate": 1.9057414235739854e-05, + "loss": 1.3159, + "step": 5884 + }, + { + "epoch": 0.5, + "learning_rate": 1.9057028947626733e-05, + "loss": 1.3086, + "step": 5885 + }, + { + "epoch": 0.5, + "learning_rate": 1.9056643584681545e-05, + "loss": 1.3113, + "step": 5886 + }, + { + "epoch": 0.5, + "learning_rate": 1.9056258146907467e-05, + "loss": 1.3573, + "step": 5887 + }, + { + "epoch": 0.5, + "learning_rate": 1.9055872634307685e-05, + "loss": 1.2683, + "step": 5888 + }, + { + "epoch": 0.5, + "learning_rate": 1.905548704688539e-05, + "loss": 1.2875, + "step": 5889 + }, + { + "epoch": 0.5, + "learning_rate": 1.9055101384643756e-05, + "loss": 1.2003, + "step": 5890 + }, + { + "epoch": 0.5, + "learning_rate": 1.9054715647585982e-05, + "loss": 1.2778, + "step": 5891 + }, + { + "epoch": 0.5, + "learning_rate": 1.9054329835715246e-05, + "loss": 1.3513, + "step": 5892 + }, + { + "epoch": 0.5, + "learning_rate": 1.905394394903474e-05, + "loss": 1.3313, + "step": 5893 + }, + { + "epoch": 0.5, + "learning_rate": 1.9053557987547652e-05, + "loss": 1.2609, + "step": 5894 + }, + { + "epoch": 0.5, + "learning_rate": 1.905317195125717e-05, + "loss": 1.3518, + "step": 5895 + }, + { + "epoch": 0.5, + "learning_rate": 1.9052785840166485e-05, + "loss": 1.2336, + "step": 5896 + }, + { + "epoch": 0.5, + "learning_rate": 1.9052399654278787e-05, + "loss": 1.3528, + "step": 5897 + }, + { + "epoch": 0.5, + "learning_rate": 1.9052013393597263e-05, + "loss": 1.3508, + "step": 5898 + }, + { + "epoch": 0.5, + "learning_rate": 1.9051627058125105e-05, + "loss": 1.3155, + "step": 5899 + }, + { + "epoch": 0.5, + "learning_rate": 1.905124064786551e-05, + "loss": 1.2922, + "step": 5900 + }, + { + "epoch": 0.5, + "learning_rate": 1.9050854162821672e-05, + "loss": 1.4224, + "step": 5901 + }, + { + "epoch": 0.5, + "learning_rate": 1.9050467602996775e-05, + "loss": 1.3446, + "step": 5902 + }, + { + "epoch": 0.5, + "learning_rate": 1.9050080968394023e-05, + "loss": 1.318, + "step": 5903 + }, + { + "epoch": 0.5, + "learning_rate": 1.9049694259016602e-05, + "loss": 1.3984, + "step": 5904 + }, + { + "epoch": 0.5, + "learning_rate": 1.904930747486771e-05, + "loss": 1.2393, + "step": 5905 + }, + { + "epoch": 0.5, + "learning_rate": 1.904892061595055e-05, + "loss": 1.3153, + "step": 5906 + }, + { + "epoch": 0.5, + "learning_rate": 1.90485336822683e-05, + "loss": 1.1935, + "step": 5907 + }, + { + "epoch": 0.5, + "learning_rate": 1.904814667382418e-05, + "loss": 1.3348, + "step": 5908 + }, + { + "epoch": 0.5, + "learning_rate": 1.904775959062137e-05, + "loss": 1.3418, + "step": 5909 + }, + { + "epoch": 0.5, + "learning_rate": 1.9047372432663078e-05, + "loss": 1.2637, + "step": 5910 + }, + { + "epoch": 0.5, + "learning_rate": 1.90469851999525e-05, + "loss": 1.1963, + "step": 5911 + }, + { + "epoch": 0.5, + "learning_rate": 1.904659789249283e-05, + "loss": 1.2505, + "step": 5912 + }, + { + "epoch": 0.5, + "learning_rate": 1.904621051028728e-05, + "loss": 1.3922, + "step": 5913 + }, + { + "epoch": 0.5, + "learning_rate": 1.9045823053339036e-05, + "loss": 1.3274, + "step": 5914 + }, + { + "epoch": 0.5, + "learning_rate": 1.9045435521651313e-05, + "loss": 1.2615, + "step": 5915 + }, + { + "epoch": 0.5, + "learning_rate": 1.9045047915227302e-05, + "loss": 1.2675, + "step": 5916 + }, + { + "epoch": 0.5, + "learning_rate": 1.9044660234070214e-05, + "loss": 1.2961, + "step": 5917 + }, + { + "epoch": 0.5, + "learning_rate": 1.9044272478183248e-05, + "loss": 1.2313, + "step": 5918 + }, + { + "epoch": 0.5, + "learning_rate": 1.9043884647569606e-05, + "loss": 1.2614, + "step": 5919 + }, + { + "epoch": 0.5, + "learning_rate": 1.9043496742232495e-05, + "loss": 1.3451, + "step": 5920 + }, + { + "epoch": 0.5, + "learning_rate": 1.9043108762175122e-05, + "loss": 1.3383, + "step": 5921 + }, + { + "epoch": 0.5, + "learning_rate": 1.904272070740069e-05, + "loss": 1.2573, + "step": 5922 + }, + { + "epoch": 0.5, + "learning_rate": 1.9042332577912402e-05, + "loss": 1.214, + "step": 5923 + }, + { + "epoch": 0.5, + "learning_rate": 1.9041944373713472e-05, + "loss": 1.3232, + "step": 5924 + }, + { + "epoch": 0.5, + "learning_rate": 1.9041556094807104e-05, + "loss": 1.4073, + "step": 5925 + }, + { + "epoch": 0.5, + "learning_rate": 1.9041167741196505e-05, + "loss": 1.4027, + "step": 5926 + }, + { + "epoch": 0.5, + "learning_rate": 1.9040779312884882e-05, + "loss": 1.4091, + "step": 5927 + }, + { + "epoch": 0.5, + "learning_rate": 1.904039080987545e-05, + "loss": 1.3265, + "step": 5928 + }, + { + "epoch": 0.5, + "learning_rate": 1.9040002232171412e-05, + "loss": 1.3352, + "step": 5929 + }, + { + "epoch": 0.5, + "learning_rate": 1.9039613579775986e-05, + "loss": 1.2787, + "step": 5930 + }, + { + "epoch": 0.5, + "learning_rate": 1.903922485269238e-05, + "loss": 1.3303, + "step": 5931 + }, + { + "epoch": 0.5, + "learning_rate": 1.9038836050923804e-05, + "loss": 1.3341, + "step": 5932 + }, + { + "epoch": 0.5, + "learning_rate": 1.903844717447347e-05, + "loss": 1.3342, + "step": 5933 + }, + { + "epoch": 0.5, + "learning_rate": 1.9038058223344596e-05, + "loss": 1.3531, + "step": 5934 + }, + { + "epoch": 0.5, + "learning_rate": 1.903766919754039e-05, + "loss": 1.3758, + "step": 5935 + }, + { + "epoch": 0.5, + "learning_rate": 1.903728009706407e-05, + "loss": 1.3188, + "step": 5936 + }, + { + "epoch": 0.5, + "learning_rate": 1.903689092191885e-05, + "loss": 1.32, + "step": 5937 + }, + { + "epoch": 0.5, + "learning_rate": 1.9036501672107944e-05, + "loss": 1.3084, + "step": 5938 + }, + { + "epoch": 0.5, + "learning_rate": 1.903611234763457e-05, + "loss": 1.3298, + "step": 5939 + }, + { + "epoch": 0.5, + "learning_rate": 1.9035722948501943e-05, + "loss": 1.3876, + "step": 5940 + }, + { + "epoch": 0.5, + "learning_rate": 1.9035333474713277e-05, + "loss": 1.3485, + "step": 5941 + }, + { + "epoch": 0.5, + "learning_rate": 1.90349439262718e-05, + "loss": 1.3588, + "step": 5942 + }, + { + "epoch": 0.5, + "learning_rate": 1.9034554303180725e-05, + "loss": 1.2993, + "step": 5943 + }, + { + "epoch": 0.5, + "learning_rate": 1.9034164605443267e-05, + "loss": 1.2649, + "step": 5944 + }, + { + "epoch": 0.5, + "learning_rate": 1.9033774833062653e-05, + "loss": 1.2889, + "step": 5945 + }, + { + "epoch": 0.5, + "learning_rate": 1.90333849860421e-05, + "loss": 1.4489, + "step": 5946 + }, + { + "epoch": 0.5, + "learning_rate": 1.9032995064384826e-05, + "loss": 1.4237, + "step": 5947 + }, + { + "epoch": 0.5, + "learning_rate": 1.9032605068094058e-05, + "loss": 1.336, + "step": 5948 + }, + { + "epoch": 0.5, + "learning_rate": 1.903221499717302e-05, + "loss": 1.3205, + "step": 5949 + }, + { + "epoch": 0.5, + "learning_rate": 1.9031824851624925e-05, + "loss": 1.2842, + "step": 5950 + }, + { + "epoch": 0.5, + "learning_rate": 1.9031434631453005e-05, + "loss": 1.3265, + "step": 5951 + }, + { + "epoch": 0.5, + "learning_rate": 1.903104433666048e-05, + "loss": 1.3342, + "step": 5952 + }, + { + "epoch": 0.5, + "learning_rate": 1.9030653967250577e-05, + "loss": 1.3889, + "step": 5953 + }, + { + "epoch": 0.5, + "learning_rate": 1.903026352322652e-05, + "loss": 1.2733, + "step": 5954 + }, + { + "epoch": 0.5, + "learning_rate": 1.902987300459154e-05, + "loss": 1.2852, + "step": 5955 + }, + { + "epoch": 0.5, + "learning_rate": 1.9029482411348858e-05, + "loss": 1.3427, + "step": 5956 + }, + { + "epoch": 0.5, + "learning_rate": 1.90290917435017e-05, + "loss": 1.2794, + "step": 5957 + }, + { + "epoch": 0.5, + "learning_rate": 1.9028701001053295e-05, + "loss": 1.3297, + "step": 5958 + }, + { + "epoch": 0.5, + "learning_rate": 1.902831018400688e-05, + "loss": 1.3686, + "step": 5959 + }, + { + "epoch": 0.5, + "learning_rate": 1.902791929236567e-05, + "loss": 1.3664, + "step": 5960 + }, + { + "epoch": 0.5, + "learning_rate": 1.9027528326132903e-05, + "loss": 1.2375, + "step": 5961 + }, + { + "epoch": 0.5, + "learning_rate": 1.9027137285311807e-05, + "loss": 1.3127, + "step": 5962 + }, + { + "epoch": 0.5, + "learning_rate": 1.9026746169905615e-05, + "loss": 1.3348, + "step": 5963 + }, + { + "epoch": 0.5, + "learning_rate": 1.902635497991756e-05, + "loss": 1.3722, + "step": 5964 + }, + { + "epoch": 0.5, + "learning_rate": 1.9025963715350867e-05, + "loss": 1.259, + "step": 5965 + }, + { + "epoch": 0.5, + "learning_rate": 1.9025572376208775e-05, + "loss": 1.2923, + "step": 5966 + }, + { + "epoch": 0.5, + "learning_rate": 1.9025180962494513e-05, + "loss": 1.3338, + "step": 5967 + }, + { + "epoch": 0.5, + "learning_rate": 1.902478947421132e-05, + "loss": 1.281, + "step": 5968 + }, + { + "epoch": 0.5, + "learning_rate": 1.9024397911362425e-05, + "loss": 1.3075, + "step": 5969 + }, + { + "epoch": 0.5, + "learning_rate": 1.902400627395107e-05, + "loss": 1.27, + "step": 5970 + }, + { + "epoch": 0.5, + "learning_rate": 1.902361456198049e-05, + "loss": 1.3469, + "step": 5971 + }, + { + "epoch": 0.5, + "learning_rate": 1.902322277545391e-05, + "loss": 1.3107, + "step": 5972 + }, + { + "epoch": 0.5, + "learning_rate": 1.9022830914374583e-05, + "loss": 1.3241, + "step": 5973 + }, + { + "epoch": 0.5, + "learning_rate": 1.9022438978745737e-05, + "loss": 1.2206, + "step": 5974 + }, + { + "epoch": 0.5, + "learning_rate": 1.9022046968570612e-05, + "loss": 1.3289, + "step": 5975 + }, + { + "epoch": 0.5, + "learning_rate": 1.9021654883852448e-05, + "loss": 1.28, + "step": 5976 + }, + { + "epoch": 0.5, + "learning_rate": 1.9021262724594483e-05, + "loss": 1.2461, + "step": 5977 + }, + { + "epoch": 0.5, + "learning_rate": 1.9020870490799957e-05, + "loss": 1.3288, + "step": 5978 + }, + { + "epoch": 0.5, + "learning_rate": 1.9020478182472115e-05, + "loss": 1.2861, + "step": 5979 + }, + { + "epoch": 0.5, + "learning_rate": 1.9020085799614193e-05, + "loss": 1.2824, + "step": 5980 + }, + { + "epoch": 0.5, + "learning_rate": 1.901969334222944e-05, + "loss": 1.1898, + "step": 5981 + }, + { + "epoch": 0.5, + "learning_rate": 1.901930081032109e-05, + "loss": 1.245, + "step": 5982 + }, + { + "epoch": 0.5, + "learning_rate": 1.9018908203892388e-05, + "loss": 1.3617, + "step": 5983 + }, + { + "epoch": 0.5, + "learning_rate": 1.9018515522946583e-05, + "loss": 1.3808, + "step": 5984 + }, + { + "epoch": 0.5, + "learning_rate": 1.9018122767486918e-05, + "loss": 1.2524, + "step": 5985 + }, + { + "epoch": 0.5, + "learning_rate": 1.9017729937516634e-05, + "loss": 1.4263, + "step": 5986 + }, + { + "epoch": 0.5, + "learning_rate": 1.901733703303898e-05, + "loss": 1.3324, + "step": 5987 + }, + { + "epoch": 0.5, + "learning_rate": 1.90169440540572e-05, + "loss": 1.3957, + "step": 5988 + }, + { + "epoch": 0.5, + "learning_rate": 1.9016551000574542e-05, + "loss": 1.264, + "step": 5989 + }, + { + "epoch": 0.5, + "learning_rate": 1.901615787259426e-05, + "loss": 1.269, + "step": 5990 + }, + { + "epoch": 0.5, + "learning_rate": 1.901576467011959e-05, + "loss": 1.2958, + "step": 5991 + }, + { + "epoch": 0.5, + "learning_rate": 1.9015371393153786e-05, + "loss": 1.2866, + "step": 5992 + }, + { + "epoch": 0.5, + "learning_rate": 1.9014978041700103e-05, + "loss": 1.3368, + "step": 5993 + }, + { + "epoch": 0.5, + "learning_rate": 1.9014584615761784e-05, + "loss": 1.3113, + "step": 5994 + }, + { + "epoch": 0.5, + "learning_rate": 1.9014191115342083e-05, + "loss": 1.3131, + "step": 5995 + }, + { + "epoch": 0.5, + "learning_rate": 1.901379754044425e-05, + "loss": 1.2041, + "step": 5996 + }, + { + "epoch": 0.5, + "learning_rate": 1.9013403891071537e-05, + "loss": 1.3768, + "step": 5997 + }, + { + "epoch": 0.5, + "learning_rate": 1.9013010167227194e-05, + "loss": 1.3224, + "step": 5998 + }, + { + "epoch": 0.51, + "learning_rate": 1.9012616368914475e-05, + "loss": 1.3355, + "step": 5999 + }, + { + "epoch": 0.51, + "learning_rate": 1.9012222496136638e-05, + "loss": 1.3081, + "step": 6000 + }, + { + "epoch": 0.51, + "learning_rate": 1.9011828548896936e-05, + "loss": 1.2281, + "step": 6001 + }, + { + "epoch": 0.51, + "learning_rate": 1.901143452719862e-05, + "loss": 1.3219, + "step": 6002 + }, + { + "epoch": 0.51, + "learning_rate": 1.901104043104495e-05, + "loss": 1.3226, + "step": 6003 + }, + { + "epoch": 0.51, + "learning_rate": 1.9010646260439182e-05, + "loss": 1.3019, + "step": 6004 + }, + { + "epoch": 0.51, + "learning_rate": 1.9010252015384567e-05, + "loss": 1.2401, + "step": 6005 + }, + { + "epoch": 0.51, + "learning_rate": 1.9009857695884364e-05, + "loss": 1.2546, + "step": 6006 + }, + { + "epoch": 0.51, + "learning_rate": 1.900946330194184e-05, + "loss": 1.4075, + "step": 6007 + }, + { + "epoch": 0.51, + "learning_rate": 1.900906883356024e-05, + "loss": 1.4245, + "step": 6008 + }, + { + "epoch": 0.51, + "learning_rate": 1.900867429074283e-05, + "loss": 1.2992, + "step": 6009 + }, + { + "epoch": 0.51, + "learning_rate": 1.9008279673492876e-05, + "loss": 1.4335, + "step": 6010 + }, + { + "epoch": 0.51, + "learning_rate": 1.900788498181363e-05, + "loss": 1.3383, + "step": 6011 + }, + { + "epoch": 0.51, + "learning_rate": 1.9007490215708354e-05, + "loss": 1.3167, + "step": 6012 + }, + { + "epoch": 0.51, + "learning_rate": 1.900709537518031e-05, + "loss": 1.3662, + "step": 6013 + }, + { + "epoch": 0.51, + "learning_rate": 1.9006700460232762e-05, + "loss": 1.2924, + "step": 6014 + }, + { + "epoch": 0.51, + "learning_rate": 1.9006305470868975e-05, + "loss": 1.243, + "step": 6015 + }, + { + "epoch": 0.51, + "learning_rate": 1.900591040709221e-05, + "loss": 1.2533, + "step": 6016 + }, + { + "epoch": 0.51, + "learning_rate": 1.9005515268905728e-05, + "loss": 1.2364, + "step": 6017 + }, + { + "epoch": 0.51, + "learning_rate": 1.9005120056312797e-05, + "loss": 1.2581, + "step": 6018 + }, + { + "epoch": 0.51, + "learning_rate": 1.9004724769316682e-05, + "loss": 1.2967, + "step": 6019 + }, + { + "epoch": 0.51, + "learning_rate": 1.9004329407920652e-05, + "loss": 1.2617, + "step": 6020 + }, + { + "epoch": 0.51, + "learning_rate": 1.9003933972127967e-05, + "loss": 1.3096, + "step": 6021 + }, + { + "epoch": 0.51, + "learning_rate": 1.90035384619419e-05, + "loss": 1.2758, + "step": 6022 + }, + { + "epoch": 0.51, + "learning_rate": 1.9003142877365715e-05, + "loss": 1.354, + "step": 6023 + }, + { + "epoch": 0.51, + "learning_rate": 1.9002747218402684e-05, + "loss": 1.2206, + "step": 6024 + }, + { + "epoch": 0.51, + "learning_rate": 1.9002351485056076e-05, + "loss": 1.2816, + "step": 6025 + }, + { + "epoch": 0.51, + "learning_rate": 1.9001955677329157e-05, + "loss": 1.2763, + "step": 6026 + }, + { + "epoch": 0.51, + "learning_rate": 1.9001559795225196e-05, + "loss": 1.3083, + "step": 6027 + }, + { + "epoch": 0.51, + "learning_rate": 1.900116383874747e-05, + "loss": 1.2111, + "step": 6028 + }, + { + "epoch": 0.51, + "learning_rate": 1.900076780789925e-05, + "loss": 1.2952, + "step": 6029 + }, + { + "epoch": 0.51, + "learning_rate": 1.90003717026838e-05, + "loss": 1.3163, + "step": 6030 + }, + { + "epoch": 0.51, + "learning_rate": 1.89999755231044e-05, + "loss": 1.3447, + "step": 6031 + }, + { + "epoch": 0.51, + "learning_rate": 1.8999579269164325e-05, + "loss": 1.2908, + "step": 6032 + }, + { + "epoch": 0.51, + "learning_rate": 1.8999182940866844e-05, + "loss": 1.298, + "step": 6033 + }, + { + "epoch": 0.51, + "learning_rate": 1.8998786538215235e-05, + "loss": 1.1943, + "step": 6034 + }, + { + "epoch": 0.51, + "learning_rate": 1.899839006121277e-05, + "loss": 1.3224, + "step": 6035 + }, + { + "epoch": 0.51, + "learning_rate": 1.899799350986273e-05, + "loss": 1.2961, + "step": 6036 + }, + { + "epoch": 0.51, + "learning_rate": 1.8997596884168384e-05, + "loss": 1.2803, + "step": 6037 + }, + { + "epoch": 0.51, + "learning_rate": 1.899720018413301e-05, + "loss": 1.3022, + "step": 6038 + }, + { + "epoch": 0.51, + "learning_rate": 1.8996803409759895e-05, + "loss": 1.2705, + "step": 6039 + }, + { + "epoch": 0.51, + "learning_rate": 1.8996406561052306e-05, + "loss": 1.2403, + "step": 6040 + }, + { + "epoch": 0.51, + "learning_rate": 1.899600963801353e-05, + "loss": 1.3168, + "step": 6041 + }, + { + "epoch": 0.51, + "learning_rate": 1.8995612640646842e-05, + "loss": 1.2646, + "step": 6042 + }, + { + "epoch": 0.51, + "learning_rate": 1.8995215568955526e-05, + "loss": 1.2954, + "step": 6043 + }, + { + "epoch": 0.51, + "learning_rate": 1.899481842294286e-05, + "loss": 1.3654, + "step": 6044 + }, + { + "epoch": 0.51, + "learning_rate": 1.899442120261212e-05, + "loss": 1.3668, + "step": 6045 + }, + { + "epoch": 0.51, + "learning_rate": 1.8994023907966598e-05, + "loss": 1.3041, + "step": 6046 + }, + { + "epoch": 0.51, + "learning_rate": 1.899362653900957e-05, + "loss": 1.3274, + "step": 6047 + }, + { + "epoch": 0.51, + "learning_rate": 1.899322909574432e-05, + "loss": 1.2326, + "step": 6048 + }, + { + "epoch": 0.51, + "learning_rate": 1.8992831578174136e-05, + "loss": 1.3417, + "step": 6049 + }, + { + "epoch": 0.51, + "learning_rate": 1.89924339863023e-05, + "loss": 1.3124, + "step": 6050 + }, + { + "epoch": 0.51, + "learning_rate": 1.8992036320132093e-05, + "loss": 1.3485, + "step": 6051 + }, + { + "epoch": 0.51, + "learning_rate": 1.8991638579666808e-05, + "loss": 1.2734, + "step": 6052 + }, + { + "epoch": 0.51, + "learning_rate": 1.8991240764909723e-05, + "loss": 1.3013, + "step": 6053 + }, + { + "epoch": 0.51, + "learning_rate": 1.8990842875864132e-05, + "loss": 1.3448, + "step": 6054 + }, + { + "epoch": 0.51, + "learning_rate": 1.8990444912533318e-05, + "loss": 1.1824, + "step": 6055 + }, + { + "epoch": 0.51, + "learning_rate": 1.8990046874920574e-05, + "loss": 1.3146, + "step": 6056 + }, + { + "epoch": 0.51, + "learning_rate": 1.8989648763029183e-05, + "loss": 1.2551, + "step": 6057 + }, + { + "epoch": 0.51, + "learning_rate": 1.8989250576862435e-05, + "loss": 1.3326, + "step": 6058 + }, + { + "epoch": 0.51, + "learning_rate": 1.8988852316423627e-05, + "loss": 1.3725, + "step": 6059 + }, + { + "epoch": 0.51, + "learning_rate": 1.8988453981716042e-05, + "loss": 1.3227, + "step": 6060 + }, + { + "epoch": 0.51, + "learning_rate": 1.8988055572742973e-05, + "loss": 1.3163, + "step": 6061 + }, + { + "epoch": 0.51, + "learning_rate": 1.8987657089507713e-05, + "loss": 1.2961, + "step": 6062 + }, + { + "epoch": 0.51, + "learning_rate": 1.8987258532013553e-05, + "loss": 1.2982, + "step": 6063 + }, + { + "epoch": 0.51, + "learning_rate": 1.8986859900263788e-05, + "loss": 1.3117, + "step": 6064 + }, + { + "epoch": 0.51, + "learning_rate": 1.8986461194261706e-05, + "loss": 1.3236, + "step": 6065 + }, + { + "epoch": 0.51, + "learning_rate": 1.898606241401061e-05, + "loss": 1.3003, + "step": 6066 + }, + { + "epoch": 0.51, + "learning_rate": 1.898566355951379e-05, + "loss": 1.2437, + "step": 6067 + }, + { + "epoch": 0.51, + "learning_rate": 1.8985264630774544e-05, + "loss": 1.2609, + "step": 6068 + }, + { + "epoch": 0.51, + "learning_rate": 1.898486562779616e-05, + "loss": 1.3621, + "step": 6069 + }, + { + "epoch": 0.51, + "learning_rate": 1.8984466550581944e-05, + "loss": 1.3363, + "step": 6070 + }, + { + "epoch": 0.51, + "learning_rate": 1.8984067399135192e-05, + "loss": 1.2991, + "step": 6071 + }, + { + "epoch": 0.51, + "learning_rate": 1.8983668173459197e-05, + "loss": 1.2596, + "step": 6072 + }, + { + "epoch": 0.51, + "learning_rate": 1.898326887355726e-05, + "loss": 1.1884, + "step": 6073 + }, + { + "epoch": 0.51, + "learning_rate": 1.8982869499432682e-05, + "loss": 1.2793, + "step": 6074 + }, + { + "epoch": 0.51, + "learning_rate": 1.8982470051088763e-05, + "loss": 1.3437, + "step": 6075 + }, + { + "epoch": 0.51, + "learning_rate": 1.8982070528528796e-05, + "loss": 1.3226, + "step": 6076 + }, + { + "epoch": 0.51, + "learning_rate": 1.8981670931756093e-05, + "loss": 1.2787, + "step": 6077 + }, + { + "epoch": 0.51, + "learning_rate": 1.8981271260773948e-05, + "loss": 1.2938, + "step": 6078 + }, + { + "epoch": 0.51, + "learning_rate": 1.8980871515585666e-05, + "loss": 1.3758, + "step": 6079 + }, + { + "epoch": 0.51, + "learning_rate": 1.898047169619455e-05, + "loss": 1.3003, + "step": 6080 + }, + { + "epoch": 0.51, + "learning_rate": 1.89800718026039e-05, + "loss": 1.3482, + "step": 6081 + }, + { + "epoch": 0.51, + "learning_rate": 1.8979671834817026e-05, + "loss": 1.2994, + "step": 6082 + }, + { + "epoch": 0.51, + "learning_rate": 1.897927179283723e-05, + "loss": 1.3225, + "step": 6083 + }, + { + "epoch": 0.51, + "learning_rate": 1.8978871676667814e-05, + "loss": 1.3545, + "step": 6084 + }, + { + "epoch": 0.51, + "learning_rate": 1.8978471486312087e-05, + "loss": 1.3709, + "step": 6085 + }, + { + "epoch": 0.51, + "learning_rate": 1.8978071221773356e-05, + "loss": 1.2867, + "step": 6086 + }, + { + "epoch": 0.51, + "learning_rate": 1.8977670883054925e-05, + "loss": 1.3149, + "step": 6087 + }, + { + "epoch": 0.51, + "learning_rate": 1.8977270470160108e-05, + "loss": 1.3043, + "step": 6088 + }, + { + "epoch": 0.51, + "learning_rate": 1.8976869983092205e-05, + "loss": 1.3422, + "step": 6089 + }, + { + "epoch": 0.51, + "learning_rate": 1.8976469421854533e-05, + "loss": 1.2706, + "step": 6090 + }, + { + "epoch": 0.51, + "learning_rate": 1.8976068786450398e-05, + "loss": 1.2822, + "step": 6091 + }, + { + "epoch": 0.51, + "learning_rate": 1.8975668076883105e-05, + "loss": 1.3222, + "step": 6092 + }, + { + "epoch": 0.51, + "learning_rate": 1.8975267293155976e-05, + "loss": 1.3424, + "step": 6093 + }, + { + "epoch": 0.51, + "learning_rate": 1.897486643527231e-05, + "loss": 1.349, + "step": 6094 + }, + { + "epoch": 0.51, + "learning_rate": 1.8974465503235427e-05, + "loss": 1.2782, + "step": 6095 + }, + { + "epoch": 0.51, + "learning_rate": 1.8974064497048637e-05, + "loss": 1.294, + "step": 6096 + }, + { + "epoch": 0.51, + "learning_rate": 1.8973663416715256e-05, + "loss": 1.265, + "step": 6097 + }, + { + "epoch": 0.51, + "learning_rate": 1.8973262262238598e-05, + "loss": 1.3095, + "step": 6098 + }, + { + "epoch": 0.51, + "learning_rate": 1.8972861033621976e-05, + "loss": 1.3006, + "step": 6099 + }, + { + "epoch": 0.51, + "learning_rate": 1.89724597308687e-05, + "loss": 1.1673, + "step": 6100 + }, + { + "epoch": 0.51, + "learning_rate": 1.897205835398209e-05, + "loss": 1.2511, + "step": 6101 + }, + { + "epoch": 0.51, + "learning_rate": 1.8971656902965466e-05, + "loss": 1.3283, + "step": 6102 + }, + { + "epoch": 0.51, + "learning_rate": 1.897125537782214e-05, + "loss": 1.3591, + "step": 6103 + }, + { + "epoch": 0.51, + "learning_rate": 1.8970853778555432e-05, + "loss": 1.403, + "step": 6104 + }, + { + "epoch": 0.51, + "learning_rate": 1.8970452105168657e-05, + "loss": 1.3356, + "step": 6105 + }, + { + "epoch": 0.51, + "learning_rate": 1.8970050357665137e-05, + "loss": 1.3234, + "step": 6106 + }, + { + "epoch": 0.51, + "learning_rate": 1.8969648536048192e-05, + "loss": 1.3832, + "step": 6107 + }, + { + "epoch": 0.51, + "learning_rate": 1.8969246640321137e-05, + "loss": 1.2508, + "step": 6108 + }, + { + "epoch": 0.51, + "learning_rate": 1.8968844670487294e-05, + "loss": 1.2612, + "step": 6109 + }, + { + "epoch": 0.51, + "learning_rate": 1.896844262654999e-05, + "loss": 1.3364, + "step": 6110 + }, + { + "epoch": 0.51, + "learning_rate": 1.8968040508512538e-05, + "loss": 1.3478, + "step": 6111 + }, + { + "epoch": 0.51, + "learning_rate": 1.8967638316378273e-05, + "loss": 1.315, + "step": 6112 + }, + { + "epoch": 0.51, + "learning_rate": 1.8967236050150502e-05, + "loss": 1.2717, + "step": 6113 + }, + { + "epoch": 0.51, + "learning_rate": 1.8966833709832563e-05, + "loss": 1.2929, + "step": 6114 + }, + { + "epoch": 0.51, + "learning_rate": 1.8966431295427772e-05, + "loss": 1.2289, + "step": 6115 + }, + { + "epoch": 0.51, + "learning_rate": 1.8966028806939453e-05, + "loss": 1.3412, + "step": 6116 + }, + { + "epoch": 0.52, + "learning_rate": 1.8965626244370938e-05, + "loss": 1.3294, + "step": 6117 + }, + { + "epoch": 0.52, + "learning_rate": 1.896522360772555e-05, + "loss": 1.2995, + "step": 6118 + }, + { + "epoch": 0.52, + "learning_rate": 1.8964820897006614e-05, + "loss": 1.3255, + "step": 6119 + }, + { + "epoch": 0.52, + "learning_rate": 1.8964418112217456e-05, + "loss": 1.2508, + "step": 6120 + }, + { + "epoch": 0.52, + "learning_rate": 1.896401525336141e-05, + "loss": 1.2976, + "step": 6121 + }, + { + "epoch": 0.52, + "learning_rate": 1.8963612320441802e-05, + "loss": 1.3257, + "step": 6122 + }, + { + "epoch": 0.52, + "learning_rate": 1.896320931346196e-05, + "loss": 1.3906, + "step": 6123 + }, + { + "epoch": 0.52, + "learning_rate": 1.896280623242521e-05, + "loss": 1.2575, + "step": 6124 + }, + { + "epoch": 0.52, + "learning_rate": 1.8962403077334892e-05, + "loss": 1.3071, + "step": 6125 + }, + { + "epoch": 0.52, + "learning_rate": 1.8961999848194327e-05, + "loss": 1.3404, + "step": 6126 + }, + { + "epoch": 0.52, + "learning_rate": 1.8961596545006854e-05, + "loss": 1.376, + "step": 6127 + }, + { + "epoch": 0.52, + "learning_rate": 1.89611931677758e-05, + "loss": 1.4443, + "step": 6128 + }, + { + "epoch": 0.52, + "learning_rate": 1.8960789716504502e-05, + "loss": 1.3045, + "step": 6129 + }, + { + "epoch": 0.52, + "learning_rate": 1.896038619119629e-05, + "loss": 1.3215, + "step": 6130 + }, + { + "epoch": 0.52, + "learning_rate": 1.89599825918545e-05, + "loss": 1.2738, + "step": 6131 + }, + { + "epoch": 0.52, + "learning_rate": 1.8959578918482467e-05, + "loss": 1.3129, + "step": 6132 + }, + { + "epoch": 0.52, + "learning_rate": 1.8959175171083523e-05, + "loss": 1.3729, + "step": 6133 + }, + { + "epoch": 0.52, + "learning_rate": 1.895877134966101e-05, + "loss": 1.2852, + "step": 6134 + }, + { + "epoch": 0.52, + "learning_rate": 1.8958367454218258e-05, + "loss": 1.2456, + "step": 6135 + }, + { + "epoch": 0.52, + "learning_rate": 1.8957963484758612e-05, + "loss": 1.3519, + "step": 6136 + }, + { + "epoch": 0.52, + "learning_rate": 1.89575594412854e-05, + "loss": 1.2922, + "step": 6137 + }, + { + "epoch": 0.52, + "learning_rate": 1.8957155323801965e-05, + "loss": 1.2954, + "step": 6138 + }, + { + "epoch": 0.52, + "learning_rate": 1.895675113231165e-05, + "loss": 1.3634, + "step": 6139 + }, + { + "epoch": 0.52, + "learning_rate": 1.8956346866817786e-05, + "loss": 1.4267, + "step": 6140 + }, + { + "epoch": 0.52, + "learning_rate": 1.895594252732372e-05, + "loss": 1.2934, + "step": 6141 + }, + { + "epoch": 0.52, + "learning_rate": 1.895553811383279e-05, + "loss": 1.3283, + "step": 6142 + }, + { + "epoch": 0.52, + "learning_rate": 1.8955133626348336e-05, + "loss": 1.3642, + "step": 6143 + }, + { + "epoch": 0.52, + "learning_rate": 1.8954729064873704e-05, + "loss": 1.3692, + "step": 6144 + }, + { + "epoch": 0.52, + "learning_rate": 1.8954324429412238e-05, + "loss": 1.2637, + "step": 6145 + }, + { + "epoch": 0.52, + "learning_rate": 1.8953919719967272e-05, + "loss": 1.2689, + "step": 6146 + }, + { + "epoch": 0.52, + "learning_rate": 1.895351493654216e-05, + "loss": 1.3082, + "step": 6147 + }, + { + "epoch": 0.52, + "learning_rate": 1.895311007914024e-05, + "loss": 1.264, + "step": 6148 + }, + { + "epoch": 0.52, + "learning_rate": 1.895270514776486e-05, + "loss": 1.3701, + "step": 6149 + }, + { + "epoch": 0.52, + "learning_rate": 1.8952300142419366e-05, + "loss": 1.3183, + "step": 6150 + }, + { + "epoch": 0.52, + "learning_rate": 1.89518950631071e-05, + "loss": 1.3128, + "step": 6151 + }, + { + "epoch": 0.52, + "learning_rate": 1.8951489909831414e-05, + "loss": 1.2774, + "step": 6152 + }, + { + "epoch": 0.52, + "learning_rate": 1.895108468259565e-05, + "loss": 1.2894, + "step": 6153 + }, + { + "epoch": 0.52, + "learning_rate": 1.895067938140316e-05, + "loss": 1.3488, + "step": 6154 + }, + { + "epoch": 0.52, + "learning_rate": 1.89502740062573e-05, + "loss": 1.3379, + "step": 6155 + }, + { + "epoch": 0.52, + "learning_rate": 1.8949868557161405e-05, + "loss": 1.2973, + "step": 6156 + }, + { + "epoch": 0.52, + "learning_rate": 1.8949463034118834e-05, + "loss": 1.2897, + "step": 6157 + }, + { + "epoch": 0.52, + "learning_rate": 1.8949057437132935e-05, + "loss": 1.2173, + "step": 6158 + }, + { + "epoch": 0.52, + "learning_rate": 1.894865176620706e-05, + "loss": 1.3331, + "step": 6159 + }, + { + "epoch": 0.52, + "learning_rate": 1.8948246021344558e-05, + "loss": 1.2675, + "step": 6160 + }, + { + "epoch": 0.52, + "learning_rate": 1.8947840202548784e-05, + "loss": 1.3457, + "step": 6161 + }, + { + "epoch": 0.52, + "learning_rate": 1.894743430982309e-05, + "loss": 1.2297, + "step": 6162 + }, + { + "epoch": 0.52, + "learning_rate": 1.8947028343170835e-05, + "loss": 1.3441, + "step": 6163 + }, + { + "epoch": 0.52, + "learning_rate": 1.8946622302595364e-05, + "loss": 1.2608, + "step": 6164 + }, + { + "epoch": 0.52, + "learning_rate": 1.8946216188100036e-05, + "loss": 1.3515, + "step": 6165 + }, + { + "epoch": 0.52, + "learning_rate": 1.894580999968821e-05, + "loss": 1.3202, + "step": 6166 + }, + { + "epoch": 0.52, + "learning_rate": 1.8945403737363236e-05, + "loss": 1.2274, + "step": 6167 + }, + { + "epoch": 0.52, + "learning_rate": 1.8944997401128472e-05, + "loss": 1.3585, + "step": 6168 + }, + { + "epoch": 0.52, + "learning_rate": 1.8944590990987278e-05, + "loss": 1.279, + "step": 6169 + }, + { + "epoch": 0.52, + "learning_rate": 1.8944184506943015e-05, + "loss": 1.3844, + "step": 6170 + }, + { + "epoch": 0.52, + "learning_rate": 1.8943777948999035e-05, + "loss": 1.2329, + "step": 6171 + }, + { + "epoch": 0.52, + "learning_rate": 1.8943371317158696e-05, + "loss": 1.295, + "step": 6172 + }, + { + "epoch": 0.52, + "learning_rate": 1.8942964611425362e-05, + "loss": 1.2857, + "step": 6173 + }, + { + "epoch": 0.52, + "learning_rate": 1.8942557831802393e-05, + "loss": 1.274, + "step": 6174 + }, + { + "epoch": 0.52, + "learning_rate": 1.894215097829315e-05, + "loss": 1.3062, + "step": 6175 + }, + { + "epoch": 0.52, + "learning_rate": 1.894174405090099e-05, + "loss": 1.3276, + "step": 6176 + }, + { + "epoch": 0.52, + "learning_rate": 1.8941337049629282e-05, + "loss": 1.3011, + "step": 6177 + }, + { + "epoch": 0.52, + "learning_rate": 1.8940929974481386e-05, + "loss": 1.2047, + "step": 6178 + }, + { + "epoch": 0.52, + "learning_rate": 1.8940522825460664e-05, + "loss": 1.2997, + "step": 6179 + }, + { + "epoch": 0.52, + "learning_rate": 1.8940115602570483e-05, + "loss": 1.2792, + "step": 6180 + }, + { + "epoch": 0.52, + "learning_rate": 1.8939708305814205e-05, + "loss": 1.3101, + "step": 6181 + }, + { + "epoch": 0.52, + "learning_rate": 1.8939300935195196e-05, + "loss": 1.3494, + "step": 6182 + }, + { + "epoch": 0.52, + "learning_rate": 1.8938893490716822e-05, + "loss": 1.3214, + "step": 6183 + }, + { + "epoch": 0.52, + "learning_rate": 1.8938485972382447e-05, + "loss": 1.277, + "step": 6184 + }, + { + "epoch": 0.52, + "learning_rate": 1.893807838019544e-05, + "loss": 1.3416, + "step": 6185 + }, + { + "epoch": 0.52, + "learning_rate": 1.893767071415917e-05, + "loss": 1.3294, + "step": 6186 + }, + { + "epoch": 0.52, + "learning_rate": 1.8937262974277003e-05, + "loss": 1.2712, + "step": 6187 + }, + { + "epoch": 0.52, + "learning_rate": 1.8936855160552312e-05, + "loss": 1.3963, + "step": 6188 + }, + { + "epoch": 0.52, + "learning_rate": 1.8936447272988462e-05, + "loss": 1.3245, + "step": 6189 + }, + { + "epoch": 0.52, + "learning_rate": 1.893603931158882e-05, + "loss": 1.1347, + "step": 6190 + }, + { + "epoch": 0.52, + "learning_rate": 1.893563127635677e-05, + "loss": 1.3677, + "step": 6191 + }, + { + "epoch": 0.52, + "learning_rate": 1.8935223167295672e-05, + "loss": 1.3208, + "step": 6192 + }, + { + "epoch": 0.52, + "learning_rate": 1.8934814984408896e-05, + "loss": 1.3438, + "step": 6193 + }, + { + "epoch": 0.52, + "learning_rate": 1.8934406727699824e-05, + "loss": 1.323, + "step": 6194 + }, + { + "epoch": 0.52, + "learning_rate": 1.893399839717182e-05, + "loss": 1.2426, + "step": 6195 + }, + { + "epoch": 0.52, + "learning_rate": 1.8933589992828264e-05, + "loss": 1.3727, + "step": 6196 + }, + { + "epoch": 0.52, + "learning_rate": 1.893318151467253e-05, + "loss": 1.2656, + "step": 6197 + }, + { + "epoch": 0.52, + "learning_rate": 1.893277296270799e-05, + "loss": 1.3246, + "step": 6198 + }, + { + "epoch": 0.52, + "learning_rate": 1.8932364336938024e-05, + "loss": 1.3391, + "step": 6199 + }, + { + "epoch": 0.52, + "learning_rate": 1.8931955637366002e-05, + "loss": 1.2656, + "step": 6200 + }, + { + "epoch": 0.52, + "learning_rate": 1.8931546863995303e-05, + "loss": 1.2651, + "step": 6201 + }, + { + "epoch": 0.52, + "learning_rate": 1.8931138016829304e-05, + "loss": 1.3304, + "step": 6202 + }, + { + "epoch": 0.52, + "learning_rate": 1.8930729095871388e-05, + "loss": 1.2894, + "step": 6203 + }, + { + "epoch": 0.52, + "learning_rate": 1.893032010112493e-05, + "loss": 1.3036, + "step": 6204 + }, + { + "epoch": 0.52, + "learning_rate": 1.892991103259331e-05, + "loss": 1.3712, + "step": 6205 + }, + { + "epoch": 0.52, + "learning_rate": 1.8929501890279903e-05, + "loss": 1.3125, + "step": 6206 + }, + { + "epoch": 0.52, + "learning_rate": 1.8929092674188095e-05, + "loss": 1.2898, + "step": 6207 + }, + { + "epoch": 0.52, + "learning_rate": 1.8928683384321265e-05, + "loss": 1.3568, + "step": 6208 + }, + { + "epoch": 0.52, + "learning_rate": 1.8928274020682798e-05, + "loss": 1.3162, + "step": 6209 + }, + { + "epoch": 0.52, + "learning_rate": 1.8927864583276073e-05, + "loss": 1.211, + "step": 6210 + }, + { + "epoch": 0.52, + "learning_rate": 1.8927455072104474e-05, + "loss": 1.2745, + "step": 6211 + }, + { + "epoch": 0.52, + "learning_rate": 1.892704548717138e-05, + "loss": 1.3218, + "step": 6212 + }, + { + "epoch": 0.52, + "learning_rate": 1.8926635828480183e-05, + "loss": 1.2966, + "step": 6213 + }, + { + "epoch": 0.52, + "learning_rate": 1.8926226096034264e-05, + "loss": 1.3074, + "step": 6214 + }, + { + "epoch": 0.52, + "learning_rate": 1.8925816289837007e-05, + "loss": 1.3026, + "step": 6215 + }, + { + "epoch": 0.52, + "learning_rate": 1.89254064098918e-05, + "loss": 1.2862, + "step": 6216 + }, + { + "epoch": 0.52, + "learning_rate": 1.8924996456202027e-05, + "loss": 1.2084, + "step": 6217 + }, + { + "epoch": 0.52, + "learning_rate": 1.8924586428771076e-05, + "loss": 1.3494, + "step": 6218 + }, + { + "epoch": 0.52, + "learning_rate": 1.892417632760234e-05, + "loss": 1.3442, + "step": 6219 + }, + { + "epoch": 0.52, + "learning_rate": 1.89237661526992e-05, + "loss": 1.3606, + "step": 6220 + }, + { + "epoch": 0.52, + "learning_rate": 1.8923355904065047e-05, + "loss": 1.283, + "step": 6221 + }, + { + "epoch": 0.52, + "learning_rate": 1.8922945581703276e-05, + "loss": 1.3565, + "step": 6222 + }, + { + "epoch": 0.52, + "learning_rate": 1.892253518561727e-05, + "loss": 1.269, + "step": 6223 + }, + { + "epoch": 0.52, + "learning_rate": 1.8922124715810422e-05, + "loss": 1.3079, + "step": 6224 + }, + { + "epoch": 0.52, + "learning_rate": 1.8921714172286126e-05, + "loss": 1.2763, + "step": 6225 + }, + { + "epoch": 0.52, + "learning_rate": 1.892130355504777e-05, + "loss": 1.222, + "step": 6226 + }, + { + "epoch": 0.52, + "learning_rate": 1.892089286409875e-05, + "loss": 1.2714, + "step": 6227 + }, + { + "epoch": 0.52, + "learning_rate": 1.892048209944246e-05, + "loss": 1.3156, + "step": 6228 + }, + { + "epoch": 0.52, + "learning_rate": 1.892007126108229e-05, + "loss": 1.2892, + "step": 6229 + }, + { + "epoch": 0.52, + "learning_rate": 1.891966034902164e-05, + "loss": 1.3713, + "step": 6230 + }, + { + "epoch": 0.52, + "learning_rate": 1.8919249363263897e-05, + "loss": 1.1944, + "step": 6231 + }, + { + "epoch": 0.52, + "learning_rate": 1.8918838303812463e-05, + "loss": 1.2974, + "step": 6232 + }, + { + "epoch": 0.52, + "learning_rate": 1.891842717067073e-05, + "loss": 1.2641, + "step": 6233 + }, + { + "epoch": 0.52, + "learning_rate": 1.89180159638421e-05, + "loss": 1.362, + "step": 6234 + }, + { + "epoch": 0.52, + "learning_rate": 1.8917604683329967e-05, + "loss": 1.2839, + "step": 6235 + }, + { + "epoch": 0.53, + "learning_rate": 1.891719332913773e-05, + "loss": 1.3132, + "step": 6236 + }, + { + "epoch": 0.53, + "learning_rate": 1.891678190126879e-05, + "loss": 1.3114, + "step": 6237 + }, + { + "epoch": 0.53, + "learning_rate": 1.8916370399726546e-05, + "loss": 1.2676, + "step": 6238 + }, + { + "epoch": 0.53, + "learning_rate": 1.8915958824514388e-05, + "loss": 1.3715, + "step": 6239 + }, + { + "epoch": 0.53, + "learning_rate": 1.891554717563573e-05, + "loss": 1.4145, + "step": 6240 + }, + { + "epoch": 0.53, + "learning_rate": 1.891513545309397e-05, + "loss": 1.3674, + "step": 6241 + }, + { + "epoch": 0.53, + "learning_rate": 1.8914723656892506e-05, + "loss": 1.3121, + "step": 6242 + }, + { + "epoch": 0.53, + "learning_rate": 1.8914311787034737e-05, + "loss": 1.3121, + "step": 6243 + }, + { + "epoch": 0.53, + "learning_rate": 1.891389984352408e-05, + "loss": 1.3788, + "step": 6244 + }, + { + "epoch": 0.53, + "learning_rate": 1.8913487826363923e-05, + "loss": 1.3406, + "step": 6245 + }, + { + "epoch": 0.53, + "learning_rate": 1.8913075735557677e-05, + "loss": 1.3539, + "step": 6246 + }, + { + "epoch": 0.53, + "learning_rate": 1.891266357110875e-05, + "loss": 1.2892, + "step": 6247 + }, + { + "epoch": 0.53, + "learning_rate": 1.8912251333020545e-05, + "loss": 1.3066, + "step": 6248 + }, + { + "epoch": 0.53, + "learning_rate": 1.8911839021296462e-05, + "loss": 1.2862, + "step": 6249 + }, + { + "epoch": 0.53, + "learning_rate": 1.8911426635939917e-05, + "loss": 1.3377, + "step": 6250 + }, + { + "epoch": 0.53, + "learning_rate": 1.891101417695431e-05, + "loss": 1.2725, + "step": 6251 + }, + { + "epoch": 0.53, + "learning_rate": 1.8910601644343056e-05, + "loss": 1.2662, + "step": 6252 + }, + { + "epoch": 0.53, + "learning_rate": 1.8910189038109558e-05, + "loss": 1.3029, + "step": 6253 + }, + { + "epoch": 0.53, + "learning_rate": 1.8909776358257224e-05, + "loss": 1.3337, + "step": 6254 + }, + { + "epoch": 0.53, + "learning_rate": 1.8909363604789467e-05, + "loss": 1.2675, + "step": 6255 + }, + { + "epoch": 0.53, + "learning_rate": 1.8908950777709702e-05, + "loss": 1.3829, + "step": 6256 + }, + { + "epoch": 0.53, + "learning_rate": 1.890853787702133e-05, + "loss": 1.2977, + "step": 6257 + }, + { + "epoch": 0.53, + "learning_rate": 1.8908124902727768e-05, + "loss": 1.325, + "step": 6258 + }, + { + "epoch": 0.53, + "learning_rate": 1.8907711854832422e-05, + "loss": 1.251, + "step": 6259 + }, + { + "epoch": 0.53, + "learning_rate": 1.8907298733338716e-05, + "loss": 1.2993, + "step": 6260 + }, + { + "epoch": 0.53, + "learning_rate": 1.8906885538250053e-05, + "loss": 1.3654, + "step": 6261 + }, + { + "epoch": 0.53, + "learning_rate": 1.8906472269569854e-05, + "loss": 1.3797, + "step": 6262 + }, + { + "epoch": 0.53, + "learning_rate": 1.890605892730153e-05, + "loss": 1.3491, + "step": 6263 + }, + { + "epoch": 0.53, + "learning_rate": 1.89056455114485e-05, + "loss": 1.2973, + "step": 6264 + }, + { + "epoch": 0.53, + "learning_rate": 1.890523202201417e-05, + "loss": 1.3642, + "step": 6265 + }, + { + "epoch": 0.53, + "learning_rate": 1.890481845900197e-05, + "loss": 1.3605, + "step": 6266 + }, + { + "epoch": 0.53, + "learning_rate": 1.8904404822415303e-05, + "loss": 1.3247, + "step": 6267 + }, + { + "epoch": 0.53, + "learning_rate": 1.8903991112257598e-05, + "loss": 1.3336, + "step": 6268 + }, + { + "epoch": 0.53, + "learning_rate": 1.890357732853227e-05, + "loss": 1.2554, + "step": 6269 + }, + { + "epoch": 0.53, + "learning_rate": 1.8903163471242732e-05, + "loss": 1.3238, + "step": 6270 + }, + { + "epoch": 0.53, + "learning_rate": 1.8902749540392413e-05, + "loss": 1.3649, + "step": 6271 + }, + { + "epoch": 0.53, + "learning_rate": 1.8902335535984725e-05, + "loss": 1.3945, + "step": 6272 + }, + { + "epoch": 0.53, + "learning_rate": 1.890192145802309e-05, + "loss": 1.3776, + "step": 6273 + }, + { + "epoch": 0.53, + "learning_rate": 1.8901507306510935e-05, + "loss": 1.3339, + "step": 6274 + }, + { + "epoch": 0.53, + "learning_rate": 1.8901093081451676e-05, + "loss": 1.2781, + "step": 6275 + }, + { + "epoch": 0.53, + "learning_rate": 1.8900678782848734e-05, + "loss": 1.2724, + "step": 6276 + }, + { + "epoch": 0.53, + "learning_rate": 1.890026441070554e-05, + "loss": 1.3172, + "step": 6277 + }, + { + "epoch": 0.53, + "learning_rate": 1.8899849965025513e-05, + "loss": 1.3146, + "step": 6278 + }, + { + "epoch": 0.53, + "learning_rate": 1.8899435445812075e-05, + "loss": 1.2511, + "step": 6279 + }, + { + "epoch": 0.53, + "learning_rate": 1.889902085306865e-05, + "loss": 1.3457, + "step": 6280 + }, + { + "epoch": 0.53, + "learning_rate": 1.8898606186798675e-05, + "loss": 1.3225, + "step": 6281 + }, + { + "epoch": 0.53, + "learning_rate": 1.889819144700556e-05, + "loss": 1.3834, + "step": 6282 + }, + { + "epoch": 0.53, + "learning_rate": 1.8897776633692742e-05, + "loss": 1.2785, + "step": 6283 + }, + { + "epoch": 0.53, + "learning_rate": 1.889736174686365e-05, + "loss": 1.3122, + "step": 6284 + }, + { + "epoch": 0.53, + "learning_rate": 1.8896946786521704e-05, + "loss": 1.2678, + "step": 6285 + }, + { + "epoch": 0.53, + "learning_rate": 1.8896531752670335e-05, + "loss": 1.3219, + "step": 6286 + }, + { + "epoch": 0.53, + "learning_rate": 1.8896116645312976e-05, + "loss": 1.3695, + "step": 6287 + }, + { + "epoch": 0.53, + "learning_rate": 1.8895701464453053e-05, + "loss": 1.4685, + "step": 6288 + }, + { + "epoch": 0.53, + "learning_rate": 1.8895286210093997e-05, + "loss": 1.2875, + "step": 6289 + }, + { + "epoch": 0.53, + "learning_rate": 1.8894870882239244e-05, + "loss": 1.2943, + "step": 6290 + }, + { + "epoch": 0.53, + "learning_rate": 1.8894455480892216e-05, + "loss": 1.258, + "step": 6291 + }, + { + "epoch": 0.53, + "learning_rate": 1.889404000605635e-05, + "loss": 1.2593, + "step": 6292 + }, + { + "epoch": 0.53, + "learning_rate": 1.889362445773508e-05, + "loss": 1.2096, + "step": 6293 + }, + { + "epoch": 0.53, + "learning_rate": 1.8893208835931838e-05, + "loss": 1.2847, + "step": 6294 + }, + { + "epoch": 0.53, + "learning_rate": 1.889279314065006e-05, + "loss": 1.384, + "step": 6295 + }, + { + "epoch": 0.53, + "learning_rate": 1.889237737189318e-05, + "loss": 1.32, + "step": 6296 + }, + { + "epoch": 0.53, + "learning_rate": 1.8891961529664628e-05, + "loss": 1.3799, + "step": 6297 + }, + { + "epoch": 0.53, + "learning_rate": 1.8891545613967847e-05, + "loss": 1.3742, + "step": 6298 + }, + { + "epoch": 0.53, + "learning_rate": 1.889112962480627e-05, + "loss": 1.3039, + "step": 6299 + }, + { + "epoch": 0.53, + "learning_rate": 1.8890713562183337e-05, + "loss": 1.2435, + "step": 6300 + }, + { + "epoch": 0.53, + "learning_rate": 1.8890297426102478e-05, + "loss": 1.3347, + "step": 6301 + }, + { + "epoch": 0.53, + "learning_rate": 1.888988121656714e-05, + "loss": 1.3329, + "step": 6302 + }, + { + "epoch": 0.53, + "learning_rate": 1.8889464933580755e-05, + "loss": 1.3276, + "step": 6303 + }, + { + "epoch": 0.53, + "learning_rate": 1.888904857714677e-05, + "loss": 1.2625, + "step": 6304 + }, + { + "epoch": 0.53, + "learning_rate": 1.888863214726862e-05, + "loss": 1.2828, + "step": 6305 + }, + { + "epoch": 0.53, + "learning_rate": 1.8888215643949745e-05, + "loss": 1.3076, + "step": 6306 + }, + { + "epoch": 0.53, + "learning_rate": 1.8887799067193585e-05, + "loss": 1.2816, + "step": 6307 + }, + { + "epoch": 0.53, + "learning_rate": 1.888738241700359e-05, + "loss": 1.3217, + "step": 6308 + }, + { + "epoch": 0.53, + "learning_rate": 1.888696569338319e-05, + "loss": 1.3173, + "step": 6309 + }, + { + "epoch": 0.53, + "learning_rate": 1.8886548896335846e-05, + "loss": 1.3017, + "step": 6310 + }, + { + "epoch": 0.53, + "learning_rate": 1.8886132025864984e-05, + "loss": 1.3359, + "step": 6311 + }, + { + "epoch": 0.53, + "learning_rate": 1.8885715081974055e-05, + "loss": 1.3116, + "step": 6312 + }, + { + "epoch": 0.53, + "learning_rate": 1.8885298064666505e-05, + "loss": 1.3267, + "step": 6313 + }, + { + "epoch": 0.53, + "learning_rate": 1.888488097394578e-05, + "loss": 1.3375, + "step": 6314 + }, + { + "epoch": 0.53, + "learning_rate": 1.8884463809815323e-05, + "loss": 1.2694, + "step": 6315 + }, + { + "epoch": 0.53, + "learning_rate": 1.8884046572278587e-05, + "loss": 1.3174, + "step": 6316 + }, + { + "epoch": 0.53, + "learning_rate": 1.8883629261339008e-05, + "loss": 1.3278, + "step": 6317 + }, + { + "epoch": 0.53, + "learning_rate": 1.8883211877000044e-05, + "loss": 1.2788, + "step": 6318 + }, + { + "epoch": 0.53, + "learning_rate": 1.8882794419265146e-05, + "loss": 1.3604, + "step": 6319 + }, + { + "epoch": 0.53, + "learning_rate": 1.888237688813775e-05, + "loss": 1.2928, + "step": 6320 + }, + { + "epoch": 0.53, + "learning_rate": 1.8881959283621316e-05, + "loss": 1.3163, + "step": 6321 + }, + { + "epoch": 0.53, + "learning_rate": 1.8881541605719292e-05, + "loss": 1.1839, + "step": 6322 + }, + { + "epoch": 0.53, + "learning_rate": 1.888112385443513e-05, + "loss": 1.3256, + "step": 6323 + }, + { + "epoch": 0.53, + "learning_rate": 1.8880706029772276e-05, + "loss": 1.2778, + "step": 6324 + }, + { + "epoch": 0.53, + "learning_rate": 1.8880288131734193e-05, + "loss": 1.254, + "step": 6325 + }, + { + "epoch": 0.53, + "learning_rate": 1.8879870160324323e-05, + "loss": 1.2977, + "step": 6326 + }, + { + "epoch": 0.53, + "learning_rate": 1.8879452115546126e-05, + "loss": 1.3369, + "step": 6327 + }, + { + "epoch": 0.53, + "learning_rate": 1.887903399740305e-05, + "loss": 1.2992, + "step": 6328 + }, + { + "epoch": 0.53, + "learning_rate": 1.8878615805898558e-05, + "loss": 1.2867, + "step": 6329 + }, + { + "epoch": 0.53, + "learning_rate": 1.8878197541036098e-05, + "loss": 1.3507, + "step": 6330 + }, + { + "epoch": 0.53, + "learning_rate": 1.8877779202819132e-05, + "loss": 1.3354, + "step": 6331 + }, + { + "epoch": 0.53, + "learning_rate": 1.887736079125111e-05, + "loss": 1.3547, + "step": 6332 + }, + { + "epoch": 0.53, + "learning_rate": 1.887694230633549e-05, + "loss": 1.2825, + "step": 6333 + }, + { + "epoch": 0.53, + "learning_rate": 1.8876523748075736e-05, + "loss": 1.3477, + "step": 6334 + }, + { + "epoch": 0.53, + "learning_rate": 1.8876105116475298e-05, + "loss": 1.2838, + "step": 6335 + }, + { + "epoch": 0.53, + "learning_rate": 1.8875686411537643e-05, + "loss": 1.331, + "step": 6336 + }, + { + "epoch": 0.53, + "learning_rate": 1.8875267633266224e-05, + "loss": 1.2893, + "step": 6337 + }, + { + "epoch": 0.53, + "learning_rate": 1.8874848781664503e-05, + "loss": 1.3038, + "step": 6338 + }, + { + "epoch": 0.53, + "learning_rate": 1.887442985673594e-05, + "loss": 1.4194, + "step": 6339 + }, + { + "epoch": 0.53, + "learning_rate": 1.8874010858484003e-05, + "loss": 1.3378, + "step": 6340 + }, + { + "epoch": 0.53, + "learning_rate": 1.887359178691214e-05, + "loss": 1.3652, + "step": 6341 + }, + { + "epoch": 0.53, + "learning_rate": 1.8873172642023826e-05, + "loss": 1.2987, + "step": 6342 + }, + { + "epoch": 0.53, + "learning_rate": 1.887275342382252e-05, + "loss": 1.3194, + "step": 6343 + }, + { + "epoch": 0.53, + "learning_rate": 1.8872334132311688e-05, + "loss": 1.389, + "step": 6344 + }, + { + "epoch": 0.53, + "learning_rate": 1.8871914767494788e-05, + "loss": 1.3848, + "step": 6345 + }, + { + "epoch": 0.53, + "learning_rate": 1.887149532937529e-05, + "loss": 1.3575, + "step": 6346 + }, + { + "epoch": 0.53, + "learning_rate": 1.887107581795666e-05, + "loss": 1.3492, + "step": 6347 + }, + { + "epoch": 0.53, + "learning_rate": 1.8870656233242362e-05, + "loss": 1.23, + "step": 6348 + }, + { + "epoch": 0.53, + "learning_rate": 1.887023657523586e-05, + "loss": 1.3155, + "step": 6349 + }, + { + "epoch": 0.53, + "learning_rate": 1.8869816843940627e-05, + "loss": 1.4375, + "step": 6350 + }, + { + "epoch": 0.53, + "learning_rate": 1.8869397039360128e-05, + "loss": 1.361, + "step": 6351 + }, + { + "epoch": 0.53, + "learning_rate": 1.886897716149783e-05, + "loss": 1.21, + "step": 6352 + }, + { + "epoch": 0.53, + "learning_rate": 1.8868557210357207e-05, + "loss": 1.3002, + "step": 6353 + }, + { + "epoch": 0.53, + "learning_rate": 1.8868137185941727e-05, + "loss": 1.3701, + "step": 6354 + }, + { + "epoch": 0.54, + "learning_rate": 1.886771708825486e-05, + "loss": 1.2767, + "step": 6355 + }, + { + "epoch": 0.54, + "learning_rate": 1.8867296917300074e-05, + "loss": 1.2458, + "step": 6356 + }, + { + "epoch": 0.54, + "learning_rate": 1.8866876673080842e-05, + "loss": 1.3206, + "step": 6357 + }, + { + "epoch": 0.54, + "learning_rate": 1.886645635560064e-05, + "loss": 1.3406, + "step": 6358 + }, + { + "epoch": 0.54, + "learning_rate": 1.8866035964862936e-05, + "loss": 1.3539, + "step": 6359 + }, + { + "epoch": 0.54, + "learning_rate": 1.8865615500871204e-05, + "loss": 1.3256, + "step": 6360 + }, + { + "epoch": 0.54, + "learning_rate": 1.886519496362892e-05, + "loss": 1.314, + "step": 6361 + }, + { + "epoch": 0.54, + "learning_rate": 1.886477435313956e-05, + "loss": 1.2443, + "step": 6362 + }, + { + "epoch": 0.54, + "learning_rate": 1.8864353669406598e-05, + "loss": 1.2954, + "step": 6363 + }, + { + "epoch": 0.54, + "learning_rate": 1.8863932912433508e-05, + "loss": 1.2726, + "step": 6364 + }, + { + "epoch": 0.54, + "learning_rate": 1.8863512082223768e-05, + "loss": 1.3014, + "step": 6365 + }, + { + "epoch": 0.54, + "learning_rate": 1.886309117878085e-05, + "loss": 1.224, + "step": 6366 + }, + { + "epoch": 0.54, + "learning_rate": 1.886267020210824e-05, + "loss": 1.333, + "step": 6367 + }, + { + "epoch": 0.54, + "learning_rate": 1.8862249152209413e-05, + "loss": 1.3298, + "step": 6368 + }, + { + "epoch": 0.54, + "learning_rate": 1.8861828029087846e-05, + "loss": 1.3669, + "step": 6369 + }, + { + "epoch": 0.54, + "learning_rate": 1.886140683274702e-05, + "loss": 1.3272, + "step": 6370 + }, + { + "epoch": 0.54, + "learning_rate": 1.886098556319041e-05, + "loss": 1.3633, + "step": 6371 + }, + { + "epoch": 0.54, + "learning_rate": 1.8860564220421512e-05, + "loss": 1.4462, + "step": 6372 + }, + { + "epoch": 0.54, + "learning_rate": 1.886014280444379e-05, + "loss": 1.3321, + "step": 6373 + }, + { + "epoch": 0.54, + "learning_rate": 1.885972131526073e-05, + "loss": 1.2749, + "step": 6374 + }, + { + "epoch": 0.54, + "learning_rate": 1.885929975287582e-05, + "loss": 1.2672, + "step": 6375 + }, + { + "epoch": 0.54, + "learning_rate": 1.885887811729254e-05, + "loss": 1.3211, + "step": 6376 + }, + { + "epoch": 0.54, + "learning_rate": 1.8858456408514377e-05, + "loss": 1.2918, + "step": 6377 + }, + { + "epoch": 0.54, + "learning_rate": 1.8858034626544807e-05, + "loss": 1.2119, + "step": 6378 + }, + { + "epoch": 0.54, + "learning_rate": 1.885761277138732e-05, + "loss": 1.3554, + "step": 6379 + }, + { + "epoch": 0.54, + "learning_rate": 1.8857190843045406e-05, + "loss": 1.4261, + "step": 6380 + }, + { + "epoch": 0.54, + "learning_rate": 1.8856768841522545e-05, + "loss": 1.324, + "step": 6381 + }, + { + "epoch": 0.54, + "learning_rate": 1.8856346766822224e-05, + "loss": 1.1946, + "step": 6382 + }, + { + "epoch": 0.54, + "learning_rate": 1.8855924618947933e-05, + "loss": 1.384, + "step": 6383 + }, + { + "epoch": 0.54, + "learning_rate": 1.8855502397903158e-05, + "loss": 1.3071, + "step": 6384 + }, + { + "epoch": 0.54, + "learning_rate": 1.885508010369139e-05, + "loss": 1.2874, + "step": 6385 + }, + { + "epoch": 0.54, + "learning_rate": 1.885465773631611e-05, + "loss": 1.2887, + "step": 6386 + }, + { + "epoch": 0.54, + "learning_rate": 1.8854235295780817e-05, + "loss": 1.3051, + "step": 6387 + }, + { + "epoch": 0.54, + "learning_rate": 1.8853812782089002e-05, + "loss": 1.2497, + "step": 6388 + }, + { + "epoch": 0.54, + "learning_rate": 1.8853390195244148e-05, + "loss": 1.2862, + "step": 6389 + }, + { + "epoch": 0.54, + "learning_rate": 1.8852967535249752e-05, + "loss": 1.3686, + "step": 6390 + }, + { + "epoch": 0.54, + "learning_rate": 1.8852544802109305e-05, + "loss": 1.2926, + "step": 6391 + }, + { + "epoch": 0.54, + "learning_rate": 1.8852121995826298e-05, + "loss": 1.2825, + "step": 6392 + }, + { + "epoch": 0.54, + "learning_rate": 1.8851699116404226e-05, + "loss": 1.2579, + "step": 6393 + }, + { + "epoch": 0.54, + "learning_rate": 1.885127616384659e-05, + "loss": 1.2497, + "step": 6394 + }, + { + "epoch": 0.54, + "learning_rate": 1.885085313815687e-05, + "loss": 1.3256, + "step": 6395 + }, + { + "epoch": 0.54, + "learning_rate": 1.885043003933857e-05, + "loss": 1.3427, + "step": 6396 + }, + { + "epoch": 0.54, + "learning_rate": 1.8850006867395186e-05, + "loss": 1.33, + "step": 6397 + }, + { + "epoch": 0.54, + "learning_rate": 1.884958362233021e-05, + "loss": 1.1926, + "step": 6398 + }, + { + "epoch": 0.54, + "learning_rate": 1.8849160304147146e-05, + "loss": 1.3768, + "step": 6399 + }, + { + "epoch": 0.54, + "learning_rate": 1.8848736912849487e-05, + "loss": 1.324, + "step": 6400 + }, + { + "epoch": 0.54, + "learning_rate": 1.884831344844073e-05, + "loss": 1.2826, + "step": 6401 + }, + { + "epoch": 0.54, + "learning_rate": 1.8847889910924374e-05, + "loss": 1.3127, + "step": 6402 + }, + { + "epoch": 0.54, + "learning_rate": 1.8847466300303923e-05, + "loss": 1.2126, + "step": 6403 + }, + { + "epoch": 0.54, + "learning_rate": 1.8847042616582874e-05, + "loss": 1.3231, + "step": 6404 + }, + { + "epoch": 0.54, + "learning_rate": 1.8846618859764726e-05, + "loss": 1.2736, + "step": 6405 + }, + { + "epoch": 0.54, + "learning_rate": 1.884619502985298e-05, + "loss": 1.2795, + "step": 6406 + }, + { + "epoch": 0.54, + "learning_rate": 1.884577112685114e-05, + "loss": 1.2754, + "step": 6407 + }, + { + "epoch": 0.54, + "learning_rate": 1.8845347150762712e-05, + "loss": 1.2743, + "step": 6408 + }, + { + "epoch": 0.54, + "learning_rate": 1.8844923101591193e-05, + "loss": 1.31, + "step": 6409 + }, + { + "epoch": 0.54, + "learning_rate": 1.8844498979340087e-05, + "loss": 1.2304, + "step": 6410 + }, + { + "epoch": 0.54, + "learning_rate": 1.8844074784012897e-05, + "loss": 1.2358, + "step": 6411 + }, + { + "epoch": 0.54, + "learning_rate": 1.8843650515613137e-05, + "loss": 1.264, + "step": 6412 + }, + { + "epoch": 0.54, + "learning_rate": 1.8843226174144306e-05, + "loss": 1.2934, + "step": 6413 + }, + { + "epoch": 0.54, + "learning_rate": 1.8842801759609902e-05, + "loss": 1.3428, + "step": 6414 + }, + { + "epoch": 0.54, + "learning_rate": 1.8842377272013448e-05, + "loss": 1.4011, + "step": 6415 + }, + { + "epoch": 0.54, + "learning_rate": 1.884195271135844e-05, + "loss": 1.3456, + "step": 6416 + }, + { + "epoch": 0.54, + "learning_rate": 1.884152807764839e-05, + "loss": 1.2289, + "step": 6417 + }, + { + "epoch": 0.54, + "learning_rate": 1.8841103370886803e-05, + "loss": 1.2617, + "step": 6418 + }, + { + "epoch": 0.54, + "learning_rate": 1.8840678591077192e-05, + "loss": 1.4328, + "step": 6419 + }, + { + "epoch": 0.54, + "learning_rate": 1.8840253738223066e-05, + "loss": 1.3452, + "step": 6420 + }, + { + "epoch": 0.54, + "learning_rate": 1.8839828812327933e-05, + "loss": 1.4033, + "step": 6421 + }, + { + "epoch": 0.54, + "learning_rate": 1.8839403813395307e-05, + "loss": 1.3157, + "step": 6422 + }, + { + "epoch": 0.54, + "learning_rate": 1.8838978741428695e-05, + "loss": 1.3157, + "step": 6423 + }, + { + "epoch": 0.54, + "learning_rate": 1.8838553596431617e-05, + "loss": 1.3662, + "step": 6424 + }, + { + "epoch": 0.54, + "learning_rate": 1.8838128378407577e-05, + "loss": 1.2525, + "step": 6425 + }, + { + "epoch": 0.54, + "learning_rate": 1.883770308736009e-05, + "loss": 1.1864, + "step": 6426 + }, + { + "epoch": 0.54, + "learning_rate": 1.8837277723292673e-05, + "loss": 1.3073, + "step": 6427 + }, + { + "epoch": 0.54, + "learning_rate": 1.883685228620884e-05, + "loss": 1.2382, + "step": 6428 + }, + { + "epoch": 0.54, + "learning_rate": 1.8836426776112104e-05, + "loss": 1.2487, + "step": 6429 + }, + { + "epoch": 0.54, + "learning_rate": 1.8836001193005984e-05, + "loss": 1.3012, + "step": 6430 + }, + { + "epoch": 0.54, + "learning_rate": 1.883557553689399e-05, + "loss": 1.2868, + "step": 6431 + }, + { + "epoch": 0.54, + "learning_rate": 1.8835149807779647e-05, + "loss": 1.2803, + "step": 6432 + }, + { + "epoch": 0.54, + "learning_rate": 1.883472400566647e-05, + "loss": 1.2708, + "step": 6433 + }, + { + "epoch": 0.54, + "learning_rate": 1.883429813055797e-05, + "loss": 1.3123, + "step": 6434 + }, + { + "epoch": 0.54, + "learning_rate": 1.8833872182457676e-05, + "loss": 1.2575, + "step": 6435 + }, + { + "epoch": 0.54, + "learning_rate": 1.88334461613691e-05, + "loss": 1.2409, + "step": 6436 + }, + { + "epoch": 0.54, + "learning_rate": 1.883302006729577e-05, + "loss": 1.2273, + "step": 6437 + }, + { + "epoch": 0.54, + "learning_rate": 1.88325939002412e-05, + "loss": 1.3342, + "step": 6438 + }, + { + "epoch": 0.54, + "learning_rate": 1.883216766020891e-05, + "loss": 1.35, + "step": 6439 + }, + { + "epoch": 0.54, + "learning_rate": 1.883174134720242e-05, + "loss": 1.2996, + "step": 6440 + }, + { + "epoch": 0.54, + "learning_rate": 1.8831314961225266e-05, + "loss": 1.2394, + "step": 6441 + }, + { + "epoch": 0.54, + "learning_rate": 1.8830888502280956e-05, + "loss": 1.3775, + "step": 6442 + }, + { + "epoch": 0.54, + "learning_rate": 1.883046197037302e-05, + "loss": 1.2596, + "step": 6443 + }, + { + "epoch": 0.54, + "learning_rate": 1.8830035365504976e-05, + "loss": 1.3176, + "step": 6444 + }, + { + "epoch": 0.54, + "learning_rate": 1.882960868768036e-05, + "loss": 1.4416, + "step": 6445 + }, + { + "epoch": 0.54, + "learning_rate": 1.882918193690269e-05, + "loss": 1.2538, + "step": 6446 + }, + { + "epoch": 0.54, + "learning_rate": 1.8828755113175493e-05, + "loss": 1.2527, + "step": 6447 + }, + { + "epoch": 0.54, + "learning_rate": 1.8828328216502293e-05, + "loss": 1.309, + "step": 6448 + }, + { + "epoch": 0.54, + "learning_rate": 1.882790124688662e-05, + "loss": 1.2545, + "step": 6449 + }, + { + "epoch": 0.54, + "learning_rate": 1.8827474204332005e-05, + "loss": 1.3382, + "step": 6450 + }, + { + "epoch": 0.54, + "learning_rate": 1.8827047088841972e-05, + "loss": 1.2702, + "step": 6451 + }, + { + "epoch": 0.54, + "learning_rate": 1.882661990042005e-05, + "loss": 1.2188, + "step": 6452 + }, + { + "epoch": 0.54, + "learning_rate": 1.882619263906977e-05, + "loss": 1.3327, + "step": 6453 + }, + { + "epoch": 0.54, + "learning_rate": 1.882576530479466e-05, + "loss": 1.3021, + "step": 6454 + }, + { + "epoch": 0.54, + "learning_rate": 1.8825337897598258e-05, + "loss": 1.2907, + "step": 6455 + }, + { + "epoch": 0.54, + "learning_rate": 1.8824910417484085e-05, + "loss": 1.367, + "step": 6456 + }, + { + "epoch": 0.54, + "learning_rate": 1.882448286445568e-05, + "loss": 1.3433, + "step": 6457 + }, + { + "epoch": 0.54, + "learning_rate": 1.8824055238516572e-05, + "loss": 1.4081, + "step": 6458 + }, + { + "epoch": 0.54, + "learning_rate": 1.8823627539670296e-05, + "loss": 1.3181, + "step": 6459 + }, + { + "epoch": 0.54, + "learning_rate": 1.8823199767920385e-05, + "loss": 1.4397, + "step": 6460 + }, + { + "epoch": 0.54, + "learning_rate": 1.8822771923270374e-05, + "loss": 1.3347, + "step": 6461 + }, + { + "epoch": 0.54, + "learning_rate": 1.88223440057238e-05, + "loss": 1.3252, + "step": 6462 + }, + { + "epoch": 0.54, + "learning_rate": 1.8821916015284192e-05, + "loss": 1.1848, + "step": 6463 + }, + { + "epoch": 0.54, + "learning_rate": 1.882148795195509e-05, + "loss": 1.3192, + "step": 6464 + }, + { + "epoch": 0.54, + "learning_rate": 1.8821059815740036e-05, + "loss": 1.3021, + "step": 6465 + }, + { + "epoch": 0.54, + "learning_rate": 1.882063160664256e-05, + "loss": 1.3372, + "step": 6466 + }, + { + "epoch": 0.54, + "learning_rate": 1.8820203324666205e-05, + "loss": 1.367, + "step": 6467 + }, + { + "epoch": 0.54, + "learning_rate": 1.8819774969814504e-05, + "loss": 1.3199, + "step": 6468 + }, + { + "epoch": 0.54, + "learning_rate": 1.8819346542091e-05, + "loss": 1.2771, + "step": 6469 + }, + { + "epoch": 0.54, + "learning_rate": 1.8818918041499234e-05, + "loss": 1.3165, + "step": 6470 + }, + { + "epoch": 0.54, + "learning_rate": 1.8818489468042745e-05, + "loss": 1.2846, + "step": 6471 + }, + { + "epoch": 0.54, + "learning_rate": 1.881806082172507e-05, + "loss": 1.3334, + "step": 6472 + }, + { + "epoch": 0.54, + "learning_rate": 1.8817632102549756e-05, + "loss": 1.2502, + "step": 6473 + }, + { + "epoch": 0.55, + "learning_rate": 1.8817203310520345e-05, + "loss": 1.3562, + "step": 6474 + }, + { + "epoch": 0.55, + "learning_rate": 1.881677444564038e-05, + "loss": 1.373, + "step": 6475 + }, + { + "epoch": 0.55, + "learning_rate": 1.88163455079134e-05, + "loss": 1.2574, + "step": 6476 + }, + { + "epoch": 0.55, + "learning_rate": 1.881591649734295e-05, + "loss": 1.391, + "step": 6477 + }, + { + "epoch": 0.55, + "learning_rate": 1.881548741393258e-05, + "loss": 1.3564, + "step": 6478 + }, + { + "epoch": 0.55, + "learning_rate": 1.881505825768583e-05, + "loss": 1.3766, + "step": 6479 + }, + { + "epoch": 0.55, + "learning_rate": 1.881462902860625e-05, + "loss": 1.2603, + "step": 6480 + }, + { + "epoch": 0.55, + "learning_rate": 1.8814199726697383e-05, + "loss": 1.2661, + "step": 6481 + }, + { + "epoch": 0.55, + "learning_rate": 1.8813770351962777e-05, + "loss": 1.2307, + "step": 6482 + }, + { + "epoch": 0.55, + "learning_rate": 1.8813340904405976e-05, + "loss": 1.2987, + "step": 6483 + }, + { + "epoch": 0.55, + "learning_rate": 1.8812911384030536e-05, + "loss": 1.2932, + "step": 6484 + }, + { + "epoch": 0.55, + "learning_rate": 1.881248179084e-05, + "loss": 1.259, + "step": 6485 + }, + { + "epoch": 0.55, + "learning_rate": 1.881205212483792e-05, + "loss": 1.2281, + "step": 6486 + }, + { + "epoch": 0.55, + "learning_rate": 1.8811622386027844e-05, + "loss": 1.3166, + "step": 6487 + }, + { + "epoch": 0.55, + "learning_rate": 1.8811192574413325e-05, + "loss": 1.3208, + "step": 6488 + }, + { + "epoch": 0.55, + "learning_rate": 1.8810762689997913e-05, + "loss": 1.3244, + "step": 6489 + }, + { + "epoch": 0.55, + "learning_rate": 1.8810332732785156e-05, + "loss": 1.2378, + "step": 6490 + }, + { + "epoch": 0.55, + "learning_rate": 1.880990270277861e-05, + "loss": 1.2916, + "step": 6491 + }, + { + "epoch": 0.55, + "learning_rate": 1.8809472599981836e-05, + "loss": 1.3944, + "step": 6492 + }, + { + "epoch": 0.55, + "learning_rate": 1.8809042424398374e-05, + "loss": 1.2835, + "step": 6493 + }, + { + "epoch": 0.55, + "learning_rate": 1.8808612176031787e-05, + "loss": 1.2251, + "step": 6494 + }, + { + "epoch": 0.55, + "learning_rate": 1.8808181854885626e-05, + "loss": 1.3583, + "step": 6495 + }, + { + "epoch": 0.55, + "learning_rate": 1.8807751460963447e-05, + "loss": 1.3172, + "step": 6496 + }, + { + "epoch": 0.55, + "learning_rate": 1.880732099426881e-05, + "loss": 1.3117, + "step": 6497 + }, + { + "epoch": 0.55, + "learning_rate": 1.8806890454805265e-05, + "loss": 1.269, + "step": 6498 + }, + { + "epoch": 0.55, + "learning_rate": 1.8806459842576373e-05, + "loss": 1.3218, + "step": 6499 + }, + { + "epoch": 0.55, + "learning_rate": 1.8806029157585688e-05, + "loss": 1.3353, + "step": 6500 + }, + { + "epoch": 0.55, + "learning_rate": 1.8805598399836776e-05, + "loss": 1.2926, + "step": 6501 + }, + { + "epoch": 0.55, + "learning_rate": 1.880516756933319e-05, + "loss": 1.3287, + "step": 6502 + }, + { + "epoch": 0.55, + "learning_rate": 1.8804736666078495e-05, + "loss": 1.3058, + "step": 6503 + }, + { + "epoch": 0.55, + "learning_rate": 1.8804305690076244e-05, + "loss": 1.3075, + "step": 6504 + }, + { + "epoch": 0.55, + "learning_rate": 1.8803874641330003e-05, + "loss": 1.3529, + "step": 6505 + }, + { + "epoch": 0.55, + "learning_rate": 1.8803443519843328e-05, + "loss": 1.3476, + "step": 6506 + }, + { + "epoch": 0.55, + "learning_rate": 1.8803012325619793e-05, + "loss": 1.4445, + "step": 6507 + }, + { + "epoch": 0.55, + "learning_rate": 1.8802581058662947e-05, + "loss": 1.3389, + "step": 6508 + }, + { + "epoch": 0.55, + "learning_rate": 1.880214971897636e-05, + "loss": 1.3369, + "step": 6509 + }, + { + "epoch": 0.55, + "learning_rate": 1.8801718306563598e-05, + "loss": 1.383, + "step": 6510 + }, + { + "epoch": 0.55, + "learning_rate": 1.880128682142822e-05, + "loss": 1.329, + "step": 6511 + }, + { + "epoch": 0.55, + "learning_rate": 1.8800855263573795e-05, + "loss": 1.2791, + "step": 6512 + }, + { + "epoch": 0.55, + "learning_rate": 1.8800423633003886e-05, + "loss": 1.2819, + "step": 6513 + }, + { + "epoch": 0.55, + "learning_rate": 1.879999192972206e-05, + "loss": 1.3643, + "step": 6514 + }, + { + "epoch": 0.55, + "learning_rate": 1.8799560153731885e-05, + "loss": 1.29, + "step": 6515 + }, + { + "epoch": 0.55, + "learning_rate": 1.8799128305036928e-05, + "loss": 1.3245, + "step": 6516 + }, + { + "epoch": 0.55, + "learning_rate": 1.8798696383640757e-05, + "loss": 1.237, + "step": 6517 + }, + { + "epoch": 0.55, + "learning_rate": 1.879826438954694e-05, + "loss": 1.3103, + "step": 6518 + }, + { + "epoch": 0.55, + "learning_rate": 1.8797832322759047e-05, + "loss": 1.328, + "step": 6519 + }, + { + "epoch": 0.55, + "learning_rate": 1.8797400183280644e-05, + "loss": 1.3512, + "step": 6520 + }, + { + "epoch": 0.55, + "learning_rate": 1.879696797111531e-05, + "loss": 1.3454, + "step": 6521 + }, + { + "epoch": 0.55, + "learning_rate": 1.879653568626661e-05, + "loss": 1.2934, + "step": 6522 + }, + { + "epoch": 0.55, + "learning_rate": 1.8796103328738118e-05, + "loss": 1.359, + "step": 6523 + }, + { + "epoch": 0.55, + "learning_rate": 1.8795670898533402e-05, + "loss": 1.3814, + "step": 6524 + }, + { + "epoch": 0.55, + "learning_rate": 1.8795238395656042e-05, + "loss": 1.3685, + "step": 6525 + }, + { + "epoch": 0.55, + "learning_rate": 1.8794805820109604e-05, + "loss": 1.3771, + "step": 6526 + }, + { + "epoch": 0.55, + "learning_rate": 1.879437317189767e-05, + "loss": 1.2212, + "step": 6527 + }, + { + "epoch": 0.55, + "learning_rate": 1.8793940451023807e-05, + "loss": 1.3835, + "step": 6528 + }, + { + "epoch": 0.55, + "learning_rate": 1.8793507657491593e-05, + "loss": 1.3087, + "step": 6529 + }, + { + "epoch": 0.55, + "learning_rate": 1.8793074791304607e-05, + "loss": 1.2819, + "step": 6530 + }, + { + "epoch": 0.55, + "learning_rate": 1.8792641852466422e-05, + "loss": 1.2789, + "step": 6531 + }, + { + "epoch": 0.55, + "learning_rate": 1.8792208840980614e-05, + "loss": 1.2752, + "step": 6532 + }, + { + "epoch": 0.55, + "learning_rate": 1.8791775756850766e-05, + "loss": 1.3214, + "step": 6533 + }, + { + "epoch": 0.55, + "learning_rate": 1.879134260008045e-05, + "loss": 1.329, + "step": 6534 + }, + { + "epoch": 0.55, + "learning_rate": 1.879090937067325e-05, + "loss": 1.2712, + "step": 6535 + }, + { + "epoch": 0.55, + "learning_rate": 1.8790476068632742e-05, + "loss": 1.3099, + "step": 6536 + }, + { + "epoch": 0.55, + "learning_rate": 1.8790042693962506e-05, + "loss": 1.3194, + "step": 6537 + }, + { + "epoch": 0.55, + "learning_rate": 1.8789609246666127e-05, + "loss": 1.2964, + "step": 6538 + }, + { + "epoch": 0.55, + "learning_rate": 1.878917572674718e-05, + "loss": 1.2852, + "step": 6539 + }, + { + "epoch": 0.55, + "learning_rate": 1.8788742134209254e-05, + "loss": 1.3555, + "step": 6540 + }, + { + "epoch": 0.55, + "learning_rate": 1.8788308469055926e-05, + "loss": 1.2942, + "step": 6541 + }, + { + "epoch": 0.55, + "learning_rate": 1.8787874731290783e-05, + "loss": 1.2649, + "step": 6542 + }, + { + "epoch": 0.55, + "learning_rate": 1.8787440920917404e-05, + "loss": 1.2996, + "step": 6543 + }, + { + "epoch": 0.55, + "learning_rate": 1.8787007037939376e-05, + "loss": 1.3652, + "step": 6544 + }, + { + "epoch": 0.55, + "learning_rate": 1.8786573082360283e-05, + "loss": 1.3264, + "step": 6545 + }, + { + "epoch": 0.55, + "learning_rate": 1.878613905418371e-05, + "loss": 1.3433, + "step": 6546 + }, + { + "epoch": 0.55, + "learning_rate": 1.8785704953413248e-05, + "loss": 1.2263, + "step": 6547 + }, + { + "epoch": 0.55, + "learning_rate": 1.8785270780052478e-05, + "loss": 1.3137, + "step": 6548 + }, + { + "epoch": 0.55, + "learning_rate": 1.878483653410499e-05, + "loss": 1.3299, + "step": 6549 + }, + { + "epoch": 0.55, + "learning_rate": 1.878440221557437e-05, + "loss": 1.3188, + "step": 6550 + }, + { + "epoch": 0.55, + "learning_rate": 1.8783967824464206e-05, + "loss": 1.3623, + "step": 6551 + }, + { + "epoch": 0.55, + "learning_rate": 1.8783533360778093e-05, + "loss": 1.26, + "step": 6552 + }, + { + "epoch": 0.55, + "learning_rate": 1.8783098824519612e-05, + "loss": 1.3367, + "step": 6553 + }, + { + "epoch": 0.55, + "learning_rate": 1.878266421569236e-05, + "loss": 1.3203, + "step": 6554 + }, + { + "epoch": 0.55, + "learning_rate": 1.8782229534299923e-05, + "loss": 1.2968, + "step": 6555 + }, + { + "epoch": 0.55, + "learning_rate": 1.8781794780345898e-05, + "loss": 1.3134, + "step": 6556 + }, + { + "epoch": 0.55, + "learning_rate": 1.8781359953833872e-05, + "loss": 1.2646, + "step": 6557 + }, + { + "epoch": 0.55, + "learning_rate": 1.8780925054767443e-05, + "loss": 1.3369, + "step": 6558 + }, + { + "epoch": 0.55, + "learning_rate": 1.87804900831502e-05, + "loss": 1.2805, + "step": 6559 + }, + { + "epoch": 0.55, + "learning_rate": 1.8780055038985732e-05, + "loss": 1.3954, + "step": 6560 + }, + { + "epoch": 0.55, + "learning_rate": 1.8779619922277646e-05, + "loss": 1.2471, + "step": 6561 + }, + { + "epoch": 0.55, + "learning_rate": 1.8779184733029527e-05, + "loss": 1.3268, + "step": 6562 + }, + { + "epoch": 0.55, + "learning_rate": 1.8778749471244975e-05, + "loss": 1.2982, + "step": 6563 + }, + { + "epoch": 0.55, + "learning_rate": 1.8778314136927583e-05, + "loss": 1.3502, + "step": 6564 + }, + { + "epoch": 0.55, + "learning_rate": 1.8777878730080955e-05, + "loss": 1.2985, + "step": 6565 + }, + { + "epoch": 0.55, + "learning_rate": 1.877744325070868e-05, + "loss": 1.4059, + "step": 6566 + }, + { + "epoch": 0.55, + "learning_rate": 1.877700769881436e-05, + "loss": 1.2825, + "step": 6567 + }, + { + "epoch": 0.55, + "learning_rate": 1.8776572074401594e-05, + "loss": 1.2327, + "step": 6568 + }, + { + "epoch": 0.55, + "learning_rate": 1.877613637747398e-05, + "loss": 1.2551, + "step": 6569 + }, + { + "epoch": 0.55, + "learning_rate": 1.877570060803512e-05, + "loss": 1.3311, + "step": 6570 + }, + { + "epoch": 0.55, + "learning_rate": 1.8775264766088613e-05, + "loss": 1.326, + "step": 6571 + }, + { + "epoch": 0.55, + "learning_rate": 1.8774828851638057e-05, + "loss": 1.2643, + "step": 6572 + }, + { + "epoch": 0.55, + "learning_rate": 1.877439286468706e-05, + "loss": 1.3451, + "step": 6573 + }, + { + "epoch": 0.55, + "learning_rate": 1.8773956805239223e-05, + "loss": 1.1953, + "step": 6574 + }, + { + "epoch": 0.55, + "learning_rate": 1.877352067329814e-05, + "loss": 1.3548, + "step": 6575 + }, + { + "epoch": 0.55, + "learning_rate": 1.8773084468867425e-05, + "loss": 1.3006, + "step": 6576 + }, + { + "epoch": 0.55, + "learning_rate": 1.877264819195068e-05, + "loss": 1.2583, + "step": 6577 + }, + { + "epoch": 0.55, + "learning_rate": 1.8772211842551506e-05, + "loss": 1.2979, + "step": 6578 + }, + { + "epoch": 0.55, + "learning_rate": 1.8771775420673513e-05, + "loss": 1.3177, + "step": 6579 + }, + { + "epoch": 0.55, + "learning_rate": 1.8771338926320298e-05, + "loss": 1.256, + "step": 6580 + }, + { + "epoch": 0.55, + "learning_rate": 1.877090235949548e-05, + "loss": 1.3482, + "step": 6581 + }, + { + "epoch": 0.55, + "learning_rate": 1.877046572020266e-05, + "loss": 1.2373, + "step": 6582 + }, + { + "epoch": 0.55, + "learning_rate": 1.877002900844544e-05, + "loss": 1.1624, + "step": 6583 + }, + { + "epoch": 0.55, + "learning_rate": 1.8769592224227435e-05, + "loss": 1.2733, + "step": 6584 + }, + { + "epoch": 0.55, + "learning_rate": 1.8769155367552255e-05, + "loss": 1.3316, + "step": 6585 + }, + { + "epoch": 0.55, + "learning_rate": 1.8768718438423503e-05, + "loss": 1.3041, + "step": 6586 + }, + { + "epoch": 0.55, + "learning_rate": 1.8768281436844797e-05, + "loss": 1.2034, + "step": 6587 + }, + { + "epoch": 0.55, + "learning_rate": 1.876784436281974e-05, + "loss": 1.2702, + "step": 6588 + }, + { + "epoch": 0.55, + "learning_rate": 1.8767407216351952e-05, + "loss": 1.2584, + "step": 6589 + }, + { + "epoch": 0.55, + "learning_rate": 1.8766969997445033e-05, + "loss": 1.3548, + "step": 6590 + }, + { + "epoch": 0.55, + "learning_rate": 1.8766532706102608e-05, + "loss": 1.263, + "step": 6591 + }, + { + "epoch": 0.55, + "learning_rate": 1.876609534232828e-05, + "loss": 1.2141, + "step": 6592 + }, + { + "epoch": 0.56, + "learning_rate": 1.876565790612567e-05, + "loss": 1.2969, + "step": 6593 + }, + { + "epoch": 0.56, + "learning_rate": 1.8765220397498385e-05, + "loss": 1.3231, + "step": 6594 + }, + { + "epoch": 0.56, + "learning_rate": 1.8764782816450048e-05, + "loss": 1.2709, + "step": 6595 + }, + { + "epoch": 0.56, + "learning_rate": 1.8764345162984266e-05, + "loss": 1.28, + "step": 6596 + }, + { + "epoch": 0.56, + "learning_rate": 1.8763907437104665e-05, + "loss": 1.2956, + "step": 6597 + }, + { + "epoch": 0.56, + "learning_rate": 1.876346963881485e-05, + "loss": 1.3971, + "step": 6598 + }, + { + "epoch": 0.56, + "learning_rate": 1.876303176811845e-05, + "loss": 1.3264, + "step": 6599 + }, + { + "epoch": 0.56, + "learning_rate": 1.8762593825019072e-05, + "loss": 1.2891, + "step": 6600 + }, + { + "epoch": 0.56, + "learning_rate": 1.8762155809520343e-05, + "loss": 1.3066, + "step": 6601 + }, + { + "epoch": 0.56, + "learning_rate": 1.8761717721625877e-05, + "loss": 1.2802, + "step": 6602 + }, + { + "epoch": 0.56, + "learning_rate": 1.8761279561339294e-05, + "loss": 1.3367, + "step": 6603 + }, + { + "epoch": 0.56, + "learning_rate": 1.8760841328664214e-05, + "loss": 1.3398, + "step": 6604 + }, + { + "epoch": 0.56, + "learning_rate": 1.8760403023604263e-05, + "loss": 1.3269, + "step": 6605 + }, + { + "epoch": 0.56, + "learning_rate": 1.8759964646163056e-05, + "loss": 1.2938, + "step": 6606 + }, + { + "epoch": 0.56, + "learning_rate": 1.875952619634422e-05, + "loss": 1.378, + "step": 6607 + }, + { + "epoch": 0.56, + "learning_rate": 1.8759087674151374e-05, + "loss": 1.3361, + "step": 6608 + }, + { + "epoch": 0.56, + "learning_rate": 1.875864907958814e-05, + "loss": 1.2469, + "step": 6609 + }, + { + "epoch": 0.56, + "learning_rate": 1.8758210412658145e-05, + "loss": 1.2733, + "step": 6610 + }, + { + "epoch": 0.56, + "learning_rate": 1.8757771673365016e-05, + "loss": 1.3011, + "step": 6611 + }, + { + "epoch": 0.56, + "learning_rate": 1.875733286171237e-05, + "loss": 1.3035, + "step": 6612 + }, + { + "epoch": 0.56, + "learning_rate": 1.875689397770384e-05, + "loss": 1.3149, + "step": 6613 + }, + { + "epoch": 0.56, + "learning_rate": 1.875645502134305e-05, + "loss": 1.2927, + "step": 6614 + }, + { + "epoch": 0.56, + "learning_rate": 1.8756015992633624e-05, + "loss": 1.3383, + "step": 6615 + }, + { + "epoch": 0.56, + "learning_rate": 1.8755576891579194e-05, + "loss": 1.3039, + "step": 6616 + }, + { + "epoch": 0.56, + "learning_rate": 1.875513771818338e-05, + "loss": 1.3458, + "step": 6617 + }, + { + "epoch": 0.56, + "learning_rate": 1.875469847244982e-05, + "loss": 1.3139, + "step": 6618 + }, + { + "epoch": 0.56, + "learning_rate": 1.875425915438214e-05, + "loss": 1.2608, + "step": 6619 + }, + { + "epoch": 0.56, + "learning_rate": 1.875381976398397e-05, + "loss": 1.34, + "step": 6620 + }, + { + "epoch": 0.56, + "learning_rate": 1.8753380301258937e-05, + "loss": 1.2128, + "step": 6621 + }, + { + "epoch": 0.56, + "learning_rate": 1.8752940766210677e-05, + "loss": 1.3121, + "step": 6622 + }, + { + "epoch": 0.56, + "learning_rate": 1.875250115884282e-05, + "loss": 1.3589, + "step": 6623 + }, + { + "epoch": 0.56, + "learning_rate": 1.8752061479158992e-05, + "loss": 1.329, + "step": 6624 + }, + { + "epoch": 0.56, + "learning_rate": 1.8751621727162836e-05, + "loss": 1.2758, + "step": 6625 + }, + { + "epoch": 0.56, + "learning_rate": 1.875118190285798e-05, + "loss": 1.1962, + "step": 6626 + }, + { + "epoch": 0.56, + "learning_rate": 1.8750742006248055e-05, + "loss": 1.2951, + "step": 6627 + }, + { + "epoch": 0.56, + "learning_rate": 1.8750302037336702e-05, + "loss": 1.3073, + "step": 6628 + }, + { + "epoch": 0.56, + "learning_rate": 1.8749861996127553e-05, + "loss": 1.3102, + "step": 6629 + }, + { + "epoch": 0.56, + "learning_rate": 1.8749421882624243e-05, + "loss": 1.2403, + "step": 6630 + }, + { + "epoch": 0.56, + "learning_rate": 1.8748981696830408e-05, + "loss": 1.321, + "step": 6631 + }, + { + "epoch": 0.56, + "learning_rate": 1.8748541438749688e-05, + "loss": 1.3968, + "step": 6632 + }, + { + "epoch": 0.56, + "learning_rate": 1.874810110838572e-05, + "loss": 1.3798, + "step": 6633 + }, + { + "epoch": 0.56, + "learning_rate": 1.8747660705742137e-05, + "loss": 1.3878, + "step": 6634 + }, + { + "epoch": 0.56, + "learning_rate": 1.8747220230822584e-05, + "loss": 1.402, + "step": 6635 + }, + { + "epoch": 0.56, + "learning_rate": 1.8746779683630697e-05, + "loss": 1.2313, + "step": 6636 + }, + { + "epoch": 0.56, + "learning_rate": 1.8746339064170115e-05, + "loss": 1.2944, + "step": 6637 + }, + { + "epoch": 0.56, + "learning_rate": 1.8745898372444484e-05, + "loss": 1.3068, + "step": 6638 + }, + { + "epoch": 0.56, + "learning_rate": 1.8745457608457442e-05, + "loss": 1.3701, + "step": 6639 + }, + { + "epoch": 0.56, + "learning_rate": 1.8745016772212628e-05, + "loss": 1.3329, + "step": 6640 + }, + { + "epoch": 0.56, + "learning_rate": 1.874457586371369e-05, + "loss": 1.2066, + "step": 6641 + }, + { + "epoch": 0.56, + "learning_rate": 1.8744134882964263e-05, + "loss": 1.333, + "step": 6642 + }, + { + "epoch": 0.56, + "learning_rate": 1.8743693829968e-05, + "loss": 1.3235, + "step": 6643 + }, + { + "epoch": 0.56, + "learning_rate": 1.8743252704728535e-05, + "loss": 1.3321, + "step": 6644 + }, + { + "epoch": 0.56, + "learning_rate": 1.8742811507249522e-05, + "loss": 1.3544, + "step": 6645 + }, + { + "epoch": 0.56, + "learning_rate": 1.87423702375346e-05, + "loss": 1.1873, + "step": 6646 + }, + { + "epoch": 0.56, + "learning_rate": 1.8741928895587418e-05, + "loss": 1.2545, + "step": 6647 + }, + { + "epoch": 0.56, + "learning_rate": 1.874148748141162e-05, + "loss": 1.3882, + "step": 6648 + }, + { + "epoch": 0.56, + "learning_rate": 1.874104599501086e-05, + "loss": 1.2781, + "step": 6649 + }, + { + "epoch": 0.56, + "learning_rate": 1.8740604436388775e-05, + "loss": 1.3352, + "step": 6650 + }, + { + "epoch": 0.56, + "learning_rate": 1.8740162805549024e-05, + "loss": 1.3853, + "step": 6651 + }, + { + "epoch": 0.56, + "learning_rate": 1.8739721102495246e-05, + "loss": 1.3101, + "step": 6652 + }, + { + "epoch": 0.56, + "learning_rate": 1.8739279327231096e-05, + "loss": 1.3278, + "step": 6653 + }, + { + "epoch": 0.56, + "learning_rate": 1.8738837479760222e-05, + "loss": 1.3891, + "step": 6654 + }, + { + "epoch": 0.56, + "learning_rate": 1.873839556008628e-05, + "loss": 1.3766, + "step": 6655 + }, + { + "epoch": 0.56, + "learning_rate": 1.8737953568212914e-05, + "loss": 1.3016, + "step": 6656 + }, + { + "epoch": 0.56, + "learning_rate": 1.873751150414378e-05, + "loss": 1.2105, + "step": 6657 + }, + { + "epoch": 0.56, + "learning_rate": 1.873706936788253e-05, + "loss": 1.1968, + "step": 6658 + }, + { + "epoch": 0.56, + "learning_rate": 1.8736627159432817e-05, + "loss": 1.3073, + "step": 6659 + }, + { + "epoch": 0.56, + "learning_rate": 1.873618487879829e-05, + "loss": 1.2936, + "step": 6660 + }, + { + "epoch": 0.56, + "learning_rate": 1.8735742525982614e-05, + "loss": 1.2588, + "step": 6661 + }, + { + "epoch": 0.56, + "learning_rate": 1.8735300100989434e-05, + "loss": 1.3162, + "step": 6662 + }, + { + "epoch": 0.56, + "learning_rate": 1.873485760382241e-05, + "loss": 1.2894, + "step": 6663 + }, + { + "epoch": 0.56, + "learning_rate": 1.8734415034485197e-05, + "loss": 1.283, + "step": 6664 + }, + { + "epoch": 0.56, + "learning_rate": 1.873397239298145e-05, + "loss": 1.3789, + "step": 6665 + }, + { + "epoch": 0.56, + "learning_rate": 1.8733529679314832e-05, + "loss": 1.2555, + "step": 6666 + }, + { + "epoch": 0.56, + "learning_rate": 1.873308689348899e-05, + "loss": 1.3339, + "step": 6667 + }, + { + "epoch": 0.56, + "learning_rate": 1.8732644035507595e-05, + "loss": 1.3027, + "step": 6668 + }, + { + "epoch": 0.56, + "learning_rate": 1.87322011053743e-05, + "loss": 1.3034, + "step": 6669 + }, + { + "epoch": 0.56, + "learning_rate": 1.8731758103092764e-05, + "loss": 1.2656, + "step": 6670 + }, + { + "epoch": 0.56, + "learning_rate": 1.8731315028666647e-05, + "loss": 1.3214, + "step": 6671 + }, + { + "epoch": 0.56, + "learning_rate": 1.873087188209961e-05, + "loss": 1.3677, + "step": 6672 + }, + { + "epoch": 0.56, + "learning_rate": 1.8730428663395316e-05, + "loss": 1.3974, + "step": 6673 + }, + { + "epoch": 0.56, + "learning_rate": 1.8729985372557427e-05, + "loss": 1.2583, + "step": 6674 + }, + { + "epoch": 0.56, + "learning_rate": 1.8729542009589605e-05, + "loss": 1.32, + "step": 6675 + }, + { + "epoch": 0.56, + "learning_rate": 1.872909857449551e-05, + "loss": 1.27, + "step": 6676 + }, + { + "epoch": 0.56, + "learning_rate": 1.872865506727881e-05, + "loss": 1.2883, + "step": 6677 + }, + { + "epoch": 0.56, + "learning_rate": 1.872821148794317e-05, + "loss": 1.322, + "step": 6678 + }, + { + "epoch": 0.56, + "learning_rate": 1.8727767836492256e-05, + "loss": 1.3409, + "step": 6679 + }, + { + "epoch": 0.56, + "learning_rate": 1.8727324112929728e-05, + "loss": 1.3488, + "step": 6680 + }, + { + "epoch": 0.56, + "learning_rate": 1.8726880317259254e-05, + "loss": 1.3286, + "step": 6681 + }, + { + "epoch": 0.56, + "learning_rate": 1.8726436449484504e-05, + "loss": 1.2427, + "step": 6682 + }, + { + "epoch": 0.56, + "learning_rate": 1.872599250960914e-05, + "loss": 1.3457, + "step": 6683 + }, + { + "epoch": 0.56, + "learning_rate": 1.8725548497636837e-05, + "loss": 1.2409, + "step": 6684 + }, + { + "epoch": 0.56, + "learning_rate": 1.872510441357126e-05, + "loss": 1.2885, + "step": 6685 + }, + { + "epoch": 0.56, + "learning_rate": 1.8724660257416075e-05, + "loss": 1.4476, + "step": 6686 + }, + { + "epoch": 0.56, + "learning_rate": 1.8724216029174952e-05, + "loss": 1.2489, + "step": 6687 + }, + { + "epoch": 0.56, + "learning_rate": 1.872377172885157e-05, + "loss": 1.3641, + "step": 6688 + }, + { + "epoch": 0.56, + "learning_rate": 1.872332735644959e-05, + "loss": 1.315, + "step": 6689 + }, + { + "epoch": 0.56, + "learning_rate": 1.8722882911972687e-05, + "loss": 1.2601, + "step": 6690 + }, + { + "epoch": 0.56, + "learning_rate": 1.872243839542454e-05, + "loss": 1.2805, + "step": 6691 + }, + { + "epoch": 0.56, + "learning_rate": 1.8721993806808808e-05, + "loss": 1.1946, + "step": 6692 + }, + { + "epoch": 0.56, + "learning_rate": 1.8721549146129176e-05, + "loss": 1.3154, + "step": 6693 + }, + { + "epoch": 0.56, + "learning_rate": 1.8721104413389314e-05, + "loss": 1.3171, + "step": 6694 + }, + { + "epoch": 0.56, + "learning_rate": 1.8720659608592898e-05, + "loss": 1.2933, + "step": 6695 + }, + { + "epoch": 0.56, + "learning_rate": 1.8720214731743597e-05, + "loss": 1.3493, + "step": 6696 + }, + { + "epoch": 0.56, + "learning_rate": 1.871976978284509e-05, + "loss": 1.3, + "step": 6697 + }, + { + "epoch": 0.56, + "learning_rate": 1.871932476190106e-05, + "loss": 1.2397, + "step": 6698 + }, + { + "epoch": 0.56, + "learning_rate": 1.8718879668915173e-05, + "loss": 1.2875, + "step": 6699 + }, + { + "epoch": 0.56, + "learning_rate": 1.8718434503891115e-05, + "loss": 1.2625, + "step": 6700 + }, + { + "epoch": 0.56, + "learning_rate": 1.8717989266832562e-05, + "loss": 1.3002, + "step": 6701 + }, + { + "epoch": 0.56, + "learning_rate": 1.871754395774319e-05, + "loss": 1.2673, + "step": 6702 + }, + { + "epoch": 0.56, + "learning_rate": 1.8717098576626676e-05, + "loss": 1.2577, + "step": 6703 + }, + { + "epoch": 0.56, + "learning_rate": 1.871665312348671e-05, + "loss": 1.3202, + "step": 6704 + }, + { + "epoch": 0.56, + "learning_rate": 1.8716207598326962e-05, + "loss": 1.2747, + "step": 6705 + }, + { + "epoch": 0.56, + "learning_rate": 1.871576200115112e-05, + "loss": 1.2822, + "step": 6706 + }, + { + "epoch": 0.56, + "learning_rate": 1.8715316331962858e-05, + "loss": 1.1888, + "step": 6707 + }, + { + "epoch": 0.56, + "learning_rate": 1.8714870590765867e-05, + "loss": 1.351, + "step": 6708 + }, + { + "epoch": 0.56, + "learning_rate": 1.8714424777563825e-05, + "loss": 1.3464, + "step": 6709 + }, + { + "epoch": 0.56, + "learning_rate": 1.871397889236042e-05, + "loss": 1.2544, + "step": 6710 + }, + { + "epoch": 0.57, + "learning_rate": 1.8713532935159326e-05, + "loss": 1.3994, + "step": 6711 + }, + { + "epoch": 0.57, + "learning_rate": 1.871308690596424e-05, + "loss": 1.4506, + "step": 6712 + }, + { + "epoch": 0.57, + "learning_rate": 1.871264080477884e-05, + "loss": 1.2774, + "step": 6713 + }, + { + "epoch": 0.57, + "learning_rate": 1.871219463160681e-05, + "loss": 1.2313, + "step": 6714 + }, + { + "epoch": 0.57, + "learning_rate": 1.8711748386451844e-05, + "loss": 1.3056, + "step": 6715 + }, + { + "epoch": 0.57, + "learning_rate": 1.8711302069317626e-05, + "loss": 1.3496, + "step": 6716 + }, + { + "epoch": 0.57, + "learning_rate": 1.8710855680207837e-05, + "loss": 1.3754, + "step": 6717 + }, + { + "epoch": 0.57, + "learning_rate": 1.8710409219126174e-05, + "loss": 1.3227, + "step": 6718 + }, + { + "epoch": 0.57, + "learning_rate": 1.8709962686076323e-05, + "loss": 1.2911, + "step": 6719 + }, + { + "epoch": 0.57, + "learning_rate": 1.870951608106197e-05, + "loss": 1.2504, + "step": 6720 + }, + { + "epoch": 0.57, + "learning_rate": 1.8709069404086807e-05, + "loss": 1.3097, + "step": 6721 + }, + { + "epoch": 0.57, + "learning_rate": 1.8708622655154526e-05, + "loss": 1.3631, + "step": 6722 + }, + { + "epoch": 0.57, + "learning_rate": 1.870817583426882e-05, + "loss": 1.3298, + "step": 6723 + }, + { + "epoch": 0.57, + "learning_rate": 1.8707728941433377e-05, + "loss": 1.3162, + "step": 6724 + }, + { + "epoch": 0.57, + "learning_rate": 1.870728197665189e-05, + "loss": 1.3401, + "step": 6725 + }, + { + "epoch": 0.57, + "learning_rate": 1.8706834939928055e-05, + "loss": 1.2041, + "step": 6726 + }, + { + "epoch": 0.57, + "learning_rate": 1.870638783126556e-05, + "loss": 1.3177, + "step": 6727 + }, + { + "epoch": 0.57, + "learning_rate": 1.8705940650668104e-05, + "loss": 1.3324, + "step": 6728 + }, + { + "epoch": 0.57, + "learning_rate": 1.870549339813938e-05, + "loss": 1.3029, + "step": 6729 + }, + { + "epoch": 0.57, + "learning_rate": 1.870504607368308e-05, + "loss": 1.3023, + "step": 6730 + }, + { + "epoch": 0.57, + "learning_rate": 1.8704598677302908e-05, + "loss": 1.2127, + "step": 6731 + }, + { + "epoch": 0.57, + "learning_rate": 1.8704151209002556e-05, + "loss": 1.363, + "step": 6732 + }, + { + "epoch": 0.57, + "learning_rate": 1.870370366878572e-05, + "loss": 1.3461, + "step": 6733 + }, + { + "epoch": 0.57, + "learning_rate": 1.8703256056656096e-05, + "loss": 1.4062, + "step": 6734 + }, + { + "epoch": 0.57, + "learning_rate": 1.8702808372617388e-05, + "loss": 1.3574, + "step": 6735 + }, + { + "epoch": 0.57, + "learning_rate": 1.870236061667329e-05, + "loss": 1.2222, + "step": 6736 + }, + { + "epoch": 0.57, + "learning_rate": 1.8701912788827507e-05, + "loss": 1.2967, + "step": 6737 + }, + { + "epoch": 0.57, + "learning_rate": 1.870146488908373e-05, + "loss": 1.2739, + "step": 6738 + }, + { + "epoch": 0.57, + "learning_rate": 1.870101691744567e-05, + "loss": 1.3744, + "step": 6739 + }, + { + "epoch": 0.57, + "learning_rate": 1.8700568873917025e-05, + "loss": 1.3088, + "step": 6740 + }, + { + "epoch": 0.57, + "learning_rate": 1.870012075850149e-05, + "loss": 1.2791, + "step": 6741 + }, + { + "epoch": 0.57, + "learning_rate": 1.8699672571202777e-05, + "loss": 1.3133, + "step": 6742 + }, + { + "epoch": 0.57, + "learning_rate": 1.869922431202458e-05, + "loss": 1.3931, + "step": 6743 + }, + { + "epoch": 0.57, + "learning_rate": 1.8698775980970614e-05, + "loss": 1.3039, + "step": 6744 + }, + { + "epoch": 0.57, + "learning_rate": 1.8698327578044574e-05, + "loss": 1.332, + "step": 6745 + }, + { + "epoch": 0.57, + "learning_rate": 1.8697879103250167e-05, + "loss": 1.2446, + "step": 6746 + }, + { + "epoch": 0.57, + "learning_rate": 1.86974305565911e-05, + "loss": 1.2725, + "step": 6747 + }, + { + "epoch": 0.57, + "learning_rate": 1.8696981938071078e-05, + "loss": 1.2371, + "step": 6748 + }, + { + "epoch": 0.57, + "learning_rate": 1.869653324769381e-05, + "loss": 1.2949, + "step": 6749 + }, + { + "epoch": 0.57, + "learning_rate": 1.8696084485462997e-05, + "loss": 1.3298, + "step": 6750 + }, + { + "epoch": 0.57, + "learning_rate": 1.8695635651382353e-05, + "loss": 1.2384, + "step": 6751 + }, + { + "epoch": 0.57, + "learning_rate": 1.8695186745455586e-05, + "loss": 1.288, + "step": 6752 + }, + { + "epoch": 0.57, + "learning_rate": 1.86947377676864e-05, + "loss": 1.3808, + "step": 6753 + }, + { + "epoch": 0.57, + "learning_rate": 1.869428871807851e-05, + "loss": 1.3278, + "step": 6754 + }, + { + "epoch": 0.57, + "learning_rate": 1.8693839596635623e-05, + "loss": 1.2345, + "step": 6755 + }, + { + "epoch": 0.57, + "learning_rate": 1.869339040336145e-05, + "loss": 1.2706, + "step": 6756 + }, + { + "epoch": 0.57, + "learning_rate": 1.8692941138259705e-05, + "loss": 1.2218, + "step": 6757 + }, + { + "epoch": 0.57, + "learning_rate": 1.8692491801334096e-05, + "loss": 1.2175, + "step": 6758 + }, + { + "epoch": 0.57, + "learning_rate": 1.869204239258834e-05, + "loss": 1.3367, + "step": 6759 + }, + { + "epoch": 0.57, + "learning_rate": 1.869159291202615e-05, + "loss": 1.2998, + "step": 6760 + }, + { + "epoch": 0.57, + "learning_rate": 1.8691143359651235e-05, + "loss": 1.228, + "step": 6761 + }, + { + "epoch": 0.57, + "learning_rate": 1.8690693735467312e-05, + "loss": 1.2217, + "step": 6762 + }, + { + "epoch": 0.57, + "learning_rate": 1.8690244039478095e-05, + "loss": 1.3511, + "step": 6763 + }, + { + "epoch": 0.57, + "learning_rate": 1.8689794271687304e-05, + "loss": 1.3057, + "step": 6764 + }, + { + "epoch": 0.57, + "learning_rate": 1.8689344432098646e-05, + "loss": 1.361, + "step": 6765 + }, + { + "epoch": 0.57, + "learning_rate": 1.8688894520715846e-05, + "loss": 1.4097, + "step": 6766 + }, + { + "epoch": 0.57, + "learning_rate": 1.8688444537542622e-05, + "loss": 1.3405, + "step": 6767 + }, + { + "epoch": 0.57, + "learning_rate": 1.8687994482582685e-05, + "loss": 1.2752, + "step": 6768 + }, + { + "epoch": 0.57, + "learning_rate": 1.8687544355839758e-05, + "loss": 1.2887, + "step": 6769 + }, + { + "epoch": 0.57, + "learning_rate": 1.868709415731756e-05, + "loss": 1.378, + "step": 6770 + }, + { + "epoch": 0.57, + "learning_rate": 1.8686643887019807e-05, + "loss": 1.2905, + "step": 6771 + }, + { + "epoch": 0.57, + "learning_rate": 1.8686193544950225e-05, + "loss": 1.3362, + "step": 6772 + }, + { + "epoch": 0.57, + "learning_rate": 1.868574313111253e-05, + "loss": 1.2757, + "step": 6773 + }, + { + "epoch": 0.57, + "learning_rate": 1.8685292645510447e-05, + "loss": 1.308, + "step": 6774 + }, + { + "epoch": 0.57, + "learning_rate": 1.8684842088147696e-05, + "loss": 1.2532, + "step": 6775 + }, + { + "epoch": 0.57, + "learning_rate": 1.8684391459028e-05, + "loss": 1.2984, + "step": 6776 + }, + { + "epoch": 0.57, + "learning_rate": 1.868394075815508e-05, + "loss": 1.2935, + "step": 6777 + }, + { + "epoch": 0.57, + "learning_rate": 1.8683489985532665e-05, + "loss": 1.2815, + "step": 6778 + }, + { + "epoch": 0.57, + "learning_rate": 1.868303914116448e-05, + "loss": 1.3078, + "step": 6779 + }, + { + "epoch": 0.57, + "learning_rate": 1.8682588225054242e-05, + "loss": 1.2746, + "step": 6780 + }, + { + "epoch": 0.57, + "learning_rate": 1.8682137237205685e-05, + "loss": 1.3744, + "step": 6781 + }, + { + "epoch": 0.57, + "learning_rate": 1.8681686177622527e-05, + "loss": 1.3277, + "step": 6782 + }, + { + "epoch": 0.57, + "learning_rate": 1.86812350463085e-05, + "loss": 1.3463, + "step": 6783 + }, + { + "epoch": 0.57, + "learning_rate": 1.8680783843267333e-05, + "loss": 1.3448, + "step": 6784 + }, + { + "epoch": 0.57, + "learning_rate": 1.868033256850275e-05, + "loss": 1.3354, + "step": 6785 + }, + { + "epoch": 0.57, + "learning_rate": 1.8679881222018484e-05, + "loss": 1.2419, + "step": 6786 + }, + { + "epoch": 0.57, + "learning_rate": 1.867942980381826e-05, + "loss": 1.4125, + "step": 6787 + }, + { + "epoch": 0.57, + "learning_rate": 1.8678978313905808e-05, + "loss": 1.3017, + "step": 6788 + }, + { + "epoch": 0.57, + "learning_rate": 1.8678526752284857e-05, + "loss": 1.3185, + "step": 6789 + }, + { + "epoch": 0.57, + "learning_rate": 1.8678075118959145e-05, + "loss": 1.2914, + "step": 6790 + }, + { + "epoch": 0.57, + "learning_rate": 1.8677623413932396e-05, + "loss": 1.2455, + "step": 6791 + }, + { + "epoch": 0.57, + "learning_rate": 1.8677171637208346e-05, + "loss": 1.2553, + "step": 6792 + }, + { + "epoch": 0.57, + "learning_rate": 1.867671978879073e-05, + "loss": 1.2288, + "step": 6793 + }, + { + "epoch": 0.57, + "learning_rate": 1.8676267868683275e-05, + "loss": 1.2917, + "step": 6794 + }, + { + "epoch": 0.57, + "learning_rate": 1.867581587688972e-05, + "loss": 1.2085, + "step": 6795 + }, + { + "epoch": 0.57, + "learning_rate": 1.8675363813413796e-05, + "loss": 1.258, + "step": 6796 + }, + { + "epoch": 0.57, + "learning_rate": 1.8674911678259244e-05, + "loss": 1.2966, + "step": 6797 + }, + { + "epoch": 0.57, + "learning_rate": 1.867445947142979e-05, + "loss": 1.371, + "step": 6798 + }, + { + "epoch": 0.57, + "learning_rate": 1.8674007192929176e-05, + "loss": 1.3212, + "step": 6799 + }, + { + "epoch": 0.57, + "learning_rate": 1.8673554842761138e-05, + "loss": 1.2693, + "step": 6800 + }, + { + "epoch": 0.57, + "learning_rate": 1.8673102420929416e-05, + "loss": 1.2349, + "step": 6801 + }, + { + "epoch": 0.57, + "learning_rate": 1.867264992743775e-05, + "loss": 1.2844, + "step": 6802 + }, + { + "epoch": 0.57, + "learning_rate": 1.867219736228987e-05, + "loss": 1.33, + "step": 6803 + }, + { + "epoch": 0.57, + "learning_rate": 1.8671744725489516e-05, + "loss": 1.3143, + "step": 6804 + }, + { + "epoch": 0.57, + "learning_rate": 1.867129201704044e-05, + "loss": 1.2608, + "step": 6805 + }, + { + "epoch": 0.57, + "learning_rate": 1.867083923694637e-05, + "loss": 1.4304, + "step": 6806 + }, + { + "epoch": 0.57, + "learning_rate": 1.867038638521105e-05, + "loss": 1.3029, + "step": 6807 + }, + { + "epoch": 0.57, + "learning_rate": 1.8669933461838225e-05, + "loss": 1.2984, + "step": 6808 + }, + { + "epoch": 0.57, + "learning_rate": 1.8669480466831635e-05, + "loss": 1.3836, + "step": 6809 + }, + { + "epoch": 0.57, + "learning_rate": 1.8669027400195025e-05, + "loss": 1.3061, + "step": 6810 + }, + { + "epoch": 0.57, + "learning_rate": 1.866857426193213e-05, + "loss": 1.4017, + "step": 6811 + }, + { + "epoch": 0.57, + "learning_rate": 1.8668121052046704e-05, + "loss": 1.338, + "step": 6812 + }, + { + "epoch": 0.57, + "learning_rate": 1.8667667770542488e-05, + "loss": 1.3588, + "step": 6813 + }, + { + "epoch": 0.57, + "learning_rate": 1.8667214417423225e-05, + "loss": 1.3255, + "step": 6814 + }, + { + "epoch": 0.57, + "learning_rate": 1.8666760992692663e-05, + "loss": 1.3087, + "step": 6815 + }, + { + "epoch": 0.57, + "learning_rate": 1.866630749635455e-05, + "loss": 1.2322, + "step": 6816 + }, + { + "epoch": 0.57, + "learning_rate": 1.866585392841263e-05, + "loss": 1.3828, + "step": 6817 + }, + { + "epoch": 0.57, + "learning_rate": 1.866540028887065e-05, + "loss": 1.2721, + "step": 6818 + }, + { + "epoch": 0.57, + "learning_rate": 1.866494657773236e-05, + "loss": 1.2032, + "step": 6819 + }, + { + "epoch": 0.57, + "learning_rate": 1.866449279500151e-05, + "loss": 1.3098, + "step": 6820 + }, + { + "epoch": 0.57, + "learning_rate": 1.8664038940681845e-05, + "loss": 1.3528, + "step": 6821 + }, + { + "epoch": 0.57, + "learning_rate": 1.8663585014777118e-05, + "loss": 1.2851, + "step": 6822 + }, + { + "epoch": 0.57, + "learning_rate": 1.8663131017291082e-05, + "loss": 1.3049, + "step": 6823 + }, + { + "epoch": 0.57, + "learning_rate": 1.8662676948227483e-05, + "loss": 1.3691, + "step": 6824 + }, + { + "epoch": 0.57, + "learning_rate": 1.8662222807590075e-05, + "loss": 1.3339, + "step": 6825 + }, + { + "epoch": 0.57, + "learning_rate": 1.866176859538261e-05, + "loss": 1.1582, + "step": 6826 + }, + { + "epoch": 0.57, + "learning_rate": 1.8661314311608837e-05, + "loss": 1.3617, + "step": 6827 + }, + { + "epoch": 0.57, + "learning_rate": 1.8660859956272517e-05, + "loss": 1.3333, + "step": 6828 + }, + { + "epoch": 0.57, + "learning_rate": 1.86604055293774e-05, + "loss": 1.2951, + "step": 6829 + }, + { + "epoch": 0.58, + "learning_rate": 1.865995103092724e-05, + "loss": 1.3031, + "step": 6830 + }, + { + "epoch": 0.58, + "learning_rate": 1.865949646092579e-05, + "loss": 1.3293, + "step": 6831 + }, + { + "epoch": 0.58, + "learning_rate": 1.8659041819376814e-05, + "loss": 1.3139, + "step": 6832 + }, + { + "epoch": 0.58, + "learning_rate": 1.8658587106284063e-05, + "loss": 1.2881, + "step": 6833 + }, + { + "epoch": 0.58, + "learning_rate": 1.8658132321651293e-05, + "loss": 1.2669, + "step": 6834 + }, + { + "epoch": 0.58, + "learning_rate": 1.865767746548226e-05, + "loss": 1.298, + "step": 6835 + }, + { + "epoch": 0.58, + "learning_rate": 1.865722253778073e-05, + "loss": 1.2579, + "step": 6836 + }, + { + "epoch": 0.58, + "learning_rate": 1.8656767538550453e-05, + "loss": 1.3837, + "step": 6837 + }, + { + "epoch": 0.58, + "learning_rate": 1.865631246779519e-05, + "loss": 1.2691, + "step": 6838 + }, + { + "epoch": 0.58, + "learning_rate": 1.865585732551871e-05, + "loss": 1.3989, + "step": 6839 + }, + { + "epoch": 0.58, + "learning_rate": 1.8655402111724764e-05, + "loss": 1.3614, + "step": 6840 + }, + { + "epoch": 0.58, + "learning_rate": 1.8654946826417113e-05, + "loss": 1.3555, + "step": 6841 + }, + { + "epoch": 0.58, + "learning_rate": 1.8654491469599522e-05, + "loss": 1.2382, + "step": 6842 + }, + { + "epoch": 0.58, + "learning_rate": 1.8654036041275753e-05, + "loss": 1.3337, + "step": 6843 + }, + { + "epoch": 0.58, + "learning_rate": 1.865358054144957e-05, + "loss": 1.269, + "step": 6844 + }, + { + "epoch": 0.58, + "learning_rate": 1.8653124970124736e-05, + "loss": 1.2972, + "step": 6845 + }, + { + "epoch": 0.58, + "learning_rate": 1.865266932730501e-05, + "loss": 1.341, + "step": 6846 + }, + { + "epoch": 0.58, + "learning_rate": 1.8652213612994162e-05, + "loss": 1.2313, + "step": 6847 + }, + { + "epoch": 0.58, + "learning_rate": 1.8651757827195956e-05, + "loss": 1.3507, + "step": 6848 + }, + { + "epoch": 0.58, + "learning_rate": 1.865130196991416e-05, + "loss": 1.3267, + "step": 6849 + }, + { + "epoch": 0.58, + "learning_rate": 1.865084604115254e-05, + "loss": 1.2537, + "step": 6850 + }, + { + "epoch": 0.58, + "learning_rate": 1.865039004091486e-05, + "loss": 1.3541, + "step": 6851 + }, + { + "epoch": 0.58, + "learning_rate": 1.8649933969204883e-05, + "loss": 1.1991, + "step": 6852 + }, + { + "epoch": 0.58, + "learning_rate": 1.8649477826026387e-05, + "loss": 1.3701, + "step": 6853 + }, + { + "epoch": 0.58, + "learning_rate": 1.864902161138314e-05, + "loss": 1.4376, + "step": 6854 + }, + { + "epoch": 0.58, + "learning_rate": 1.8648565325278908e-05, + "loss": 1.2438, + "step": 6855 + }, + { + "epoch": 0.58, + "learning_rate": 1.8648108967717457e-05, + "loss": 1.285, + "step": 6856 + }, + { + "epoch": 0.58, + "learning_rate": 1.8647652538702566e-05, + "loss": 1.3143, + "step": 6857 + }, + { + "epoch": 0.58, + "learning_rate": 1.8647196038238e-05, + "loss": 1.3218, + "step": 6858 + }, + { + "epoch": 0.58, + "learning_rate": 1.8646739466327534e-05, + "loss": 1.3162, + "step": 6859 + }, + { + "epoch": 0.58, + "learning_rate": 1.864628282297494e-05, + "loss": 1.3576, + "step": 6860 + }, + { + "epoch": 0.58, + "learning_rate": 1.8645826108183988e-05, + "loss": 1.3726, + "step": 6861 + }, + { + "epoch": 0.58, + "learning_rate": 1.8645369321958454e-05, + "loss": 1.3377, + "step": 6862 + }, + { + "epoch": 0.58, + "learning_rate": 1.8644912464302114e-05, + "loss": 1.3215, + "step": 6863 + }, + { + "epoch": 0.58, + "learning_rate": 1.8644455535218738e-05, + "loss": 1.3183, + "step": 6864 + }, + { + "epoch": 0.58, + "learning_rate": 1.8643998534712104e-05, + "loss": 1.3117, + "step": 6865 + }, + { + "epoch": 0.58, + "learning_rate": 1.8643541462785993e-05, + "loss": 1.3986, + "step": 6866 + }, + { + "epoch": 0.58, + "learning_rate": 1.864308431944417e-05, + "loss": 1.3255, + "step": 6867 + }, + { + "epoch": 0.58, + "learning_rate": 1.864262710469042e-05, + "loss": 1.339, + "step": 6868 + }, + { + "epoch": 0.58, + "learning_rate": 1.864216981852852e-05, + "loss": 1.3007, + "step": 6869 + }, + { + "epoch": 0.58, + "learning_rate": 1.8641712460962243e-05, + "loss": 1.2419, + "step": 6870 + }, + { + "epoch": 0.58, + "learning_rate": 1.8641255031995373e-05, + "loss": 1.3258, + "step": 6871 + }, + { + "epoch": 0.58, + "learning_rate": 1.8640797531631693e-05, + "loss": 1.267, + "step": 6872 + }, + { + "epoch": 0.58, + "learning_rate": 1.8640339959874976e-05, + "loss": 1.3261, + "step": 6873 + }, + { + "epoch": 0.58, + "learning_rate": 1.8639882316729002e-05, + "loss": 1.358, + "step": 6874 + }, + { + "epoch": 0.58, + "learning_rate": 1.863942460219756e-05, + "loss": 1.3396, + "step": 6875 + }, + { + "epoch": 0.58, + "learning_rate": 1.8638966816284424e-05, + "loss": 1.2563, + "step": 6876 + }, + { + "epoch": 0.58, + "learning_rate": 1.863850895899338e-05, + "loss": 1.3378, + "step": 6877 + }, + { + "epoch": 0.58, + "learning_rate": 1.863805103032821e-05, + "loss": 1.2628, + "step": 6878 + }, + { + "epoch": 0.58, + "learning_rate": 1.8637593030292697e-05, + "loss": 1.2742, + "step": 6879 + }, + { + "epoch": 0.58, + "learning_rate": 1.863713495889063e-05, + "loss": 1.3305, + "step": 6880 + }, + { + "epoch": 0.58, + "learning_rate": 1.8636676816125784e-05, + "loss": 1.4502, + "step": 6881 + }, + { + "epoch": 0.58, + "learning_rate": 1.8636218602001954e-05, + "loss": 1.3758, + "step": 6882 + }, + { + "epoch": 0.58, + "learning_rate": 1.8635760316522923e-05, + "loss": 1.2808, + "step": 6883 + }, + { + "epoch": 0.58, + "learning_rate": 1.863530195969247e-05, + "loss": 1.3842, + "step": 6884 + }, + { + "epoch": 0.58, + "learning_rate": 1.8634843531514395e-05, + "loss": 1.3817, + "step": 6885 + }, + { + "epoch": 0.58, + "learning_rate": 1.8634385031992476e-05, + "loss": 1.3229, + "step": 6886 + }, + { + "epoch": 0.58, + "learning_rate": 1.8633926461130506e-05, + "loss": 1.2471, + "step": 6887 + }, + { + "epoch": 0.58, + "learning_rate": 1.8633467818932274e-05, + "loss": 1.5066, + "step": 6888 + }, + { + "epoch": 0.58, + "learning_rate": 1.8633009105401565e-05, + "loss": 1.3486, + "step": 6889 + }, + { + "epoch": 0.58, + "learning_rate": 1.8632550320542172e-05, + "loss": 1.4092, + "step": 6890 + }, + { + "epoch": 0.58, + "learning_rate": 1.8632091464357888e-05, + "loss": 1.2952, + "step": 6891 + }, + { + "epoch": 0.58, + "learning_rate": 1.86316325368525e-05, + "loss": 1.3761, + "step": 6892 + }, + { + "epoch": 0.58, + "learning_rate": 1.86311735380298e-05, + "loss": 1.2574, + "step": 6893 + }, + { + "epoch": 0.58, + "learning_rate": 1.8630714467893583e-05, + "loss": 1.3122, + "step": 6894 + }, + { + "epoch": 0.58, + "learning_rate": 1.863025532644764e-05, + "loss": 1.3196, + "step": 6895 + }, + { + "epoch": 0.58, + "learning_rate": 1.8629796113695767e-05, + "loss": 1.307, + "step": 6896 + }, + { + "epoch": 0.58, + "learning_rate": 1.862933682964176e-05, + "loss": 1.2565, + "step": 6897 + }, + { + "epoch": 0.58, + "learning_rate": 1.8628877474289405e-05, + "loss": 1.389, + "step": 6898 + }, + { + "epoch": 0.58, + "learning_rate": 1.8628418047642505e-05, + "loss": 1.2745, + "step": 6899 + }, + { + "epoch": 0.58, + "learning_rate": 1.862795854970485e-05, + "loss": 1.3374, + "step": 6900 + }, + { + "epoch": 0.58, + "learning_rate": 1.8627498980480246e-05, + "loss": 1.3057, + "step": 6901 + }, + { + "epoch": 0.58, + "learning_rate": 1.862703933997248e-05, + "loss": 1.2978, + "step": 6902 + }, + { + "epoch": 0.58, + "learning_rate": 1.8626579628185358e-05, + "loss": 1.3397, + "step": 6903 + }, + { + "epoch": 0.58, + "learning_rate": 1.862611984512267e-05, + "loss": 1.2858, + "step": 6904 + }, + { + "epoch": 0.58, + "learning_rate": 1.862565999078822e-05, + "loss": 1.3626, + "step": 6905 + }, + { + "epoch": 0.58, + "learning_rate": 1.8625200065185808e-05, + "loss": 1.3385, + "step": 6906 + }, + { + "epoch": 0.58, + "learning_rate": 1.8624740068319232e-05, + "loss": 1.307, + "step": 6907 + }, + { + "epoch": 0.58, + "learning_rate": 1.8624280000192293e-05, + "loss": 1.3133, + "step": 6908 + }, + { + "epoch": 0.58, + "learning_rate": 1.8623819860808794e-05, + "loss": 1.3009, + "step": 6909 + }, + { + "epoch": 0.58, + "learning_rate": 1.8623359650172532e-05, + "loss": 1.3758, + "step": 6910 + }, + { + "epoch": 0.58, + "learning_rate": 1.8622899368287318e-05, + "loss": 1.2688, + "step": 6911 + }, + { + "epoch": 0.58, + "learning_rate": 1.8622439015156946e-05, + "loss": 1.2661, + "step": 6912 + }, + { + "epoch": 0.58, + "learning_rate": 1.8621978590785225e-05, + "loss": 1.2655, + "step": 6913 + }, + { + "epoch": 0.58, + "learning_rate": 1.8621518095175953e-05, + "loss": 1.2887, + "step": 6914 + }, + { + "epoch": 0.58, + "learning_rate": 1.8621057528332945e-05, + "loss": 1.2079, + "step": 6915 + }, + { + "epoch": 0.58, + "learning_rate": 1.8620596890259998e-05, + "loss": 1.2904, + "step": 6916 + }, + { + "epoch": 0.58, + "learning_rate": 1.862013618096092e-05, + "loss": 1.2815, + "step": 6917 + }, + { + "epoch": 0.58, + "learning_rate": 1.8619675400439517e-05, + "loss": 1.2545, + "step": 6918 + }, + { + "epoch": 0.58, + "learning_rate": 1.8619214548699603e-05, + "loss": 1.3816, + "step": 6919 + }, + { + "epoch": 0.58, + "learning_rate": 1.8618753625744973e-05, + "loss": 1.3922, + "step": 6920 + }, + { + "epoch": 0.58, + "learning_rate": 1.8618292631579447e-05, + "loss": 1.3234, + "step": 6921 + }, + { + "epoch": 0.58, + "learning_rate": 1.861783156620683e-05, + "loss": 1.3805, + "step": 6922 + }, + { + "epoch": 0.58, + "learning_rate": 1.8617370429630927e-05, + "loss": 1.3296, + "step": 6923 + }, + { + "epoch": 0.58, + "learning_rate": 1.861690922185555e-05, + "loss": 1.1764, + "step": 6924 + }, + { + "epoch": 0.58, + "learning_rate": 1.8616447942884512e-05, + "loss": 1.3707, + "step": 6925 + }, + { + "epoch": 0.58, + "learning_rate": 1.8615986592721626e-05, + "loss": 1.353, + "step": 6926 + }, + { + "epoch": 0.58, + "learning_rate": 1.86155251713707e-05, + "loss": 1.2838, + "step": 6927 + }, + { + "epoch": 0.58, + "learning_rate": 1.8615063678835548e-05, + "loss": 1.2743, + "step": 6928 + }, + { + "epoch": 0.58, + "learning_rate": 1.8614602115119985e-05, + "loss": 1.3688, + "step": 6929 + }, + { + "epoch": 0.58, + "learning_rate": 1.8614140480227818e-05, + "loss": 1.1992, + "step": 6930 + }, + { + "epoch": 0.58, + "learning_rate": 1.861367877416287e-05, + "loss": 1.3175, + "step": 6931 + }, + { + "epoch": 0.58, + "learning_rate": 1.8613216996928946e-05, + "loss": 1.317, + "step": 6932 + }, + { + "epoch": 0.58, + "learning_rate": 1.861275514852987e-05, + "loss": 1.2897, + "step": 6933 + }, + { + "epoch": 0.58, + "learning_rate": 1.8612293228969457e-05, + "loss": 1.294, + "step": 6934 + }, + { + "epoch": 0.58, + "learning_rate": 1.8611831238251515e-05, + "loss": 1.2228, + "step": 6935 + }, + { + "epoch": 0.58, + "learning_rate": 1.861136917637987e-05, + "loss": 1.2725, + "step": 6936 + }, + { + "epoch": 0.58, + "learning_rate": 1.861090704335834e-05, + "loss": 1.2888, + "step": 6937 + }, + { + "epoch": 0.58, + "learning_rate": 1.8610444839190734e-05, + "loss": 1.353, + "step": 6938 + }, + { + "epoch": 0.58, + "learning_rate": 1.8609982563880882e-05, + "loss": 1.3623, + "step": 6939 + }, + { + "epoch": 0.58, + "learning_rate": 1.86095202174326e-05, + "loss": 1.3368, + "step": 6940 + }, + { + "epoch": 0.58, + "learning_rate": 1.86090577998497e-05, + "loss": 1.2003, + "step": 6941 + }, + { + "epoch": 0.58, + "learning_rate": 1.8608595311136014e-05, + "loss": 1.3131, + "step": 6942 + }, + { + "epoch": 0.58, + "learning_rate": 1.8608132751295354e-05, + "loss": 1.3633, + "step": 6943 + }, + { + "epoch": 0.58, + "learning_rate": 1.860767012033155e-05, + "loss": 1.3001, + "step": 6944 + }, + { + "epoch": 0.58, + "learning_rate": 1.8607207418248423e-05, + "loss": 1.2778, + "step": 6945 + }, + { + "epoch": 0.58, + "learning_rate": 1.860674464504979e-05, + "loss": 1.4097, + "step": 6946 + }, + { + "epoch": 0.58, + "learning_rate": 1.860628180073948e-05, + "loss": 1.2101, + "step": 6947 + }, + { + "epoch": 0.58, + "learning_rate": 1.8605818885321313e-05, + "loss": 1.3191, + "step": 6948 + }, + { + "epoch": 0.59, + "learning_rate": 1.860535589879912e-05, + "loss": 1.1907, + "step": 6949 + }, + { + "epoch": 0.59, + "learning_rate": 1.8604892841176717e-05, + "loss": 1.3787, + "step": 6950 + }, + { + "epoch": 0.59, + "learning_rate": 1.8604429712457935e-05, + "loss": 1.2987, + "step": 6951 + }, + { + "epoch": 0.59, + "learning_rate": 1.8603966512646607e-05, + "loss": 1.2949, + "step": 6952 + }, + { + "epoch": 0.59, + "learning_rate": 1.860350324174655e-05, + "loss": 1.2964, + "step": 6953 + }, + { + "epoch": 0.59, + "learning_rate": 1.86030398997616e-05, + "loss": 1.3987, + "step": 6954 + }, + { + "epoch": 0.59, + "learning_rate": 1.8602576486695575e-05, + "loss": 1.3295, + "step": 6955 + }, + { + "epoch": 0.59, + "learning_rate": 1.8602113002552315e-05, + "loss": 1.338, + "step": 6956 + }, + { + "epoch": 0.59, + "learning_rate": 1.8601649447335642e-05, + "loss": 1.287, + "step": 6957 + }, + { + "epoch": 0.59, + "learning_rate": 1.8601185821049388e-05, + "loss": 1.3078, + "step": 6958 + }, + { + "epoch": 0.59, + "learning_rate": 1.8600722123697385e-05, + "loss": 1.3336, + "step": 6959 + }, + { + "epoch": 0.59, + "learning_rate": 1.8600258355283464e-05, + "loss": 1.4014, + "step": 6960 + }, + { + "epoch": 0.59, + "learning_rate": 1.8599794515811455e-05, + "loss": 1.3175, + "step": 6961 + }, + { + "epoch": 0.59, + "learning_rate": 1.859933060528519e-05, + "loss": 1.4207, + "step": 6962 + }, + { + "epoch": 0.59, + "learning_rate": 1.8598866623708505e-05, + "loss": 1.2943, + "step": 6963 + }, + { + "epoch": 0.59, + "learning_rate": 1.8598402571085232e-05, + "loss": 1.1928, + "step": 6964 + }, + { + "epoch": 0.59, + "learning_rate": 1.8597938447419202e-05, + "loss": 1.2151, + "step": 6965 + }, + { + "epoch": 0.59, + "learning_rate": 1.8597474252714257e-05, + "loss": 1.276, + "step": 6966 + }, + { + "epoch": 0.59, + "learning_rate": 1.8597009986974228e-05, + "loss": 1.3181, + "step": 6967 + }, + { + "epoch": 0.59, + "learning_rate": 1.859654565020295e-05, + "loss": 1.1676, + "step": 6968 + }, + { + "epoch": 0.59, + "learning_rate": 1.8596081242404263e-05, + "loss": 1.302, + "step": 6969 + }, + { + "epoch": 0.59, + "learning_rate": 1.8595616763581997e-05, + "loss": 1.3244, + "step": 6970 + }, + { + "epoch": 0.59, + "learning_rate": 1.859515221374e-05, + "loss": 1.2814, + "step": 6971 + }, + { + "epoch": 0.59, + "learning_rate": 1.85946875928821e-05, + "loss": 1.2743, + "step": 6972 + }, + { + "epoch": 0.59, + "learning_rate": 1.859422290101214e-05, + "loss": 1.3543, + "step": 6973 + }, + { + "epoch": 0.59, + "learning_rate": 1.859375813813396e-05, + "loss": 1.3852, + "step": 6974 + }, + { + "epoch": 0.59, + "learning_rate": 1.8593293304251403e-05, + "loss": 1.3272, + "step": 6975 + }, + { + "epoch": 0.59, + "learning_rate": 1.8592828399368305e-05, + "loss": 1.3386, + "step": 6976 + }, + { + "epoch": 0.59, + "learning_rate": 1.859236342348851e-05, + "loss": 1.3593, + "step": 6977 + }, + { + "epoch": 0.59, + "learning_rate": 1.8591898376615855e-05, + "loss": 1.2927, + "step": 6978 + }, + { + "epoch": 0.59, + "learning_rate": 1.859143325875419e-05, + "loss": 1.2977, + "step": 6979 + }, + { + "epoch": 0.59, + "learning_rate": 1.8590968069907348e-05, + "loss": 1.2439, + "step": 6980 + }, + { + "epoch": 0.59, + "learning_rate": 1.8590502810079187e-05, + "loss": 1.2935, + "step": 6981 + }, + { + "epoch": 0.59, + "learning_rate": 1.8590037479273537e-05, + "loss": 1.3635, + "step": 6982 + }, + { + "epoch": 0.59, + "learning_rate": 1.8589572077494247e-05, + "loss": 1.3606, + "step": 6983 + }, + { + "epoch": 0.59, + "learning_rate": 1.8589106604745167e-05, + "loss": 1.24, + "step": 6984 + }, + { + "epoch": 0.59, + "learning_rate": 1.8588641061030136e-05, + "loss": 1.2606, + "step": 6985 + }, + { + "epoch": 0.59, + "learning_rate": 1.8588175446353008e-05, + "loss": 1.3169, + "step": 6986 + }, + { + "epoch": 0.59, + "learning_rate": 1.858770976071762e-05, + "loss": 1.2706, + "step": 6987 + }, + { + "epoch": 0.59, + "learning_rate": 1.858724400412783e-05, + "loss": 1.337, + "step": 6988 + }, + { + "epoch": 0.59, + "learning_rate": 1.858677817658748e-05, + "loss": 1.3367, + "step": 6989 + }, + { + "epoch": 0.59, + "learning_rate": 1.8586312278100423e-05, + "loss": 1.3739, + "step": 6990 + }, + { + "epoch": 0.59, + "learning_rate": 1.8585846308670502e-05, + "loss": 1.3056, + "step": 6991 + }, + { + "epoch": 0.59, + "learning_rate": 1.8585380268301578e-05, + "loss": 1.2928, + "step": 6992 + }, + { + "epoch": 0.59, + "learning_rate": 1.8584914156997486e-05, + "loss": 1.3246, + "step": 6993 + }, + { + "epoch": 0.59, + "learning_rate": 1.858444797476209e-05, + "loss": 1.2001, + "step": 6994 + }, + { + "epoch": 0.59, + "learning_rate": 1.8583981721599235e-05, + "loss": 1.2572, + "step": 6995 + }, + { + "epoch": 0.59, + "learning_rate": 1.858351539751278e-05, + "loss": 1.3845, + "step": 6996 + }, + { + "epoch": 0.59, + "learning_rate": 1.8583049002506573e-05, + "loss": 1.2197, + "step": 6997 + }, + { + "epoch": 0.59, + "learning_rate": 1.8582582536584467e-05, + "loss": 1.2963, + "step": 6998 + }, + { + "epoch": 0.59, + "learning_rate": 1.8582115999750315e-05, + "loss": 1.3921, + "step": 6999 + }, + { + "epoch": 0.59, + "learning_rate": 1.8581649392007975e-05, + "loss": 1.2597, + "step": 7000 + }, + { + "epoch": 0.59, + "learning_rate": 1.8581182713361304e-05, + "loss": 1.3227, + "step": 7001 + }, + { + "epoch": 0.59, + "learning_rate": 1.8580715963814153e-05, + "loss": 1.2447, + "step": 7002 + }, + { + "epoch": 0.59, + "learning_rate": 1.8580249143370382e-05, + "loss": 1.4227, + "step": 7003 + }, + { + "epoch": 0.59, + "learning_rate": 1.8579782252033846e-05, + "loss": 1.2453, + "step": 7004 + }, + { + "epoch": 0.59, + "learning_rate": 1.85793152898084e-05, + "loss": 1.3401, + "step": 7005 + }, + { + "epoch": 0.59, + "learning_rate": 1.857884825669791e-05, + "loss": 1.3954, + "step": 7006 + }, + { + "epoch": 0.59, + "learning_rate": 1.8578381152706227e-05, + "loss": 1.3046, + "step": 7007 + }, + { + "epoch": 0.59, + "learning_rate": 1.8577913977837216e-05, + "loss": 1.2775, + "step": 7008 + }, + { + "epoch": 0.59, + "learning_rate": 1.8577446732094732e-05, + "loss": 1.3682, + "step": 7009 + }, + { + "epoch": 0.59, + "learning_rate": 1.8576979415482638e-05, + "loss": 1.3185, + "step": 7010 + }, + { + "epoch": 0.59, + "learning_rate": 1.8576512028004797e-05, + "loss": 1.3738, + "step": 7011 + }, + { + "epoch": 0.59, + "learning_rate": 1.8576044569665066e-05, + "loss": 1.3732, + "step": 7012 + }, + { + "epoch": 0.59, + "learning_rate": 1.8575577040467313e-05, + "loss": 1.2622, + "step": 7013 + }, + { + "epoch": 0.59, + "learning_rate": 1.8575109440415397e-05, + "loss": 1.2891, + "step": 7014 + }, + { + "epoch": 0.59, + "learning_rate": 1.857464176951318e-05, + "loss": 1.3924, + "step": 7015 + }, + { + "epoch": 0.59, + "learning_rate": 1.8574174027764532e-05, + "loss": 1.2951, + "step": 7016 + }, + { + "epoch": 0.59, + "learning_rate": 1.8573706215173312e-05, + "loss": 1.2755, + "step": 7017 + }, + { + "epoch": 0.59, + "learning_rate": 1.857323833174339e-05, + "loss": 1.3986, + "step": 7018 + }, + { + "epoch": 0.59, + "learning_rate": 1.8572770377478628e-05, + "loss": 1.2585, + "step": 7019 + }, + { + "epoch": 0.59, + "learning_rate": 1.8572302352382888e-05, + "loss": 1.3643, + "step": 7020 + }, + { + "epoch": 0.59, + "learning_rate": 1.857183425646005e-05, + "loss": 1.3825, + "step": 7021 + }, + { + "epoch": 0.59, + "learning_rate": 1.8571366089713972e-05, + "loss": 1.336, + "step": 7022 + }, + { + "epoch": 0.59, + "learning_rate": 1.8570897852148522e-05, + "loss": 1.3464, + "step": 7023 + }, + { + "epoch": 0.59, + "learning_rate": 1.8570429543767573e-05, + "loss": 1.4015, + "step": 7024 + }, + { + "epoch": 0.59, + "learning_rate": 1.856996116457499e-05, + "loss": 1.3124, + "step": 7025 + }, + { + "epoch": 0.59, + "learning_rate": 1.8569492714574647e-05, + "loss": 1.3372, + "step": 7026 + }, + { + "epoch": 0.59, + "learning_rate": 1.856902419377041e-05, + "loss": 1.2939, + "step": 7027 + }, + { + "epoch": 0.59, + "learning_rate": 1.8568555602166158e-05, + "loss": 1.361, + "step": 7028 + }, + { + "epoch": 0.59, + "learning_rate": 1.8568086939765756e-05, + "loss": 1.2578, + "step": 7029 + }, + { + "epoch": 0.59, + "learning_rate": 1.856761820657307e-05, + "loss": 1.3914, + "step": 7030 + }, + { + "epoch": 0.59, + "learning_rate": 1.856714940259199e-05, + "loss": 1.298, + "step": 7031 + }, + { + "epoch": 0.59, + "learning_rate": 1.8566680527826374e-05, + "loss": 1.3208, + "step": 7032 + }, + { + "epoch": 0.59, + "learning_rate": 1.8566211582280106e-05, + "loss": 1.3281, + "step": 7033 + }, + { + "epoch": 0.59, + "learning_rate": 1.8565742565957053e-05, + "loss": 1.4345, + "step": 7034 + }, + { + "epoch": 0.59, + "learning_rate": 1.8565273478861098e-05, + "loss": 1.3906, + "step": 7035 + }, + { + "epoch": 0.59, + "learning_rate": 1.856480432099611e-05, + "loss": 1.4021, + "step": 7036 + }, + { + "epoch": 0.59, + "learning_rate": 1.856433509236596e-05, + "loss": 1.4468, + "step": 7037 + }, + { + "epoch": 0.59, + "learning_rate": 1.8563865792974543e-05, + "loss": 1.2882, + "step": 7038 + }, + { + "epoch": 0.59, + "learning_rate": 1.856339642282572e-05, + "loss": 1.2467, + "step": 7039 + }, + { + "epoch": 0.59, + "learning_rate": 1.8562926981923377e-05, + "loss": 1.351, + "step": 7040 + }, + { + "epoch": 0.59, + "learning_rate": 1.8562457470271394e-05, + "loss": 1.2544, + "step": 7041 + }, + { + "epoch": 0.59, + "learning_rate": 1.856198788787364e-05, + "loss": 1.3395, + "step": 7042 + }, + { + "epoch": 0.59, + "learning_rate": 1.856151823473401e-05, + "loss": 1.2932, + "step": 7043 + }, + { + "epoch": 0.59, + "learning_rate": 1.856104851085637e-05, + "loss": 1.2878, + "step": 7044 + }, + { + "epoch": 0.59, + "learning_rate": 1.856057871624461e-05, + "loss": 1.2694, + "step": 7045 + }, + { + "epoch": 0.59, + "learning_rate": 1.8560108850902608e-05, + "loss": 1.3344, + "step": 7046 + }, + { + "epoch": 0.59, + "learning_rate": 1.8559638914834247e-05, + "loss": 1.2915, + "step": 7047 + }, + { + "epoch": 0.59, + "learning_rate": 1.8559168908043412e-05, + "loss": 1.3515, + "step": 7048 + }, + { + "epoch": 0.59, + "learning_rate": 1.8558698830533982e-05, + "loss": 1.2634, + "step": 7049 + }, + { + "epoch": 0.59, + "learning_rate": 1.8558228682309845e-05, + "loss": 1.2779, + "step": 7050 + }, + { + "epoch": 0.59, + "learning_rate": 1.8557758463374882e-05, + "loss": 1.4059, + "step": 7051 + }, + { + "epoch": 0.59, + "learning_rate": 1.8557288173732983e-05, + "loss": 1.2787, + "step": 7052 + }, + { + "epoch": 0.59, + "learning_rate": 1.8556817813388024e-05, + "loss": 1.3816, + "step": 7053 + }, + { + "epoch": 0.59, + "learning_rate": 1.8556347382343902e-05, + "loss": 1.3126, + "step": 7054 + }, + { + "epoch": 0.59, + "learning_rate": 1.85558768806045e-05, + "loss": 1.4387, + "step": 7055 + }, + { + "epoch": 0.59, + "learning_rate": 1.8555406308173702e-05, + "loss": 1.3061, + "step": 7056 + }, + { + "epoch": 0.59, + "learning_rate": 1.85549356650554e-05, + "loss": 1.2521, + "step": 7057 + }, + { + "epoch": 0.59, + "learning_rate": 1.8554464951253485e-05, + "loss": 1.3555, + "step": 7058 + }, + { + "epoch": 0.59, + "learning_rate": 1.855399416677184e-05, + "loss": 1.284, + "step": 7059 + }, + { + "epoch": 0.59, + "learning_rate": 1.8553523311614355e-05, + "loss": 1.4372, + "step": 7060 + }, + { + "epoch": 0.59, + "learning_rate": 1.8553052385784926e-05, + "loss": 1.24, + "step": 7061 + }, + { + "epoch": 0.59, + "learning_rate": 1.8552581389287443e-05, + "loss": 1.2749, + "step": 7062 + }, + { + "epoch": 0.59, + "learning_rate": 1.8552110322125792e-05, + "loss": 1.2793, + "step": 7063 + }, + { + "epoch": 0.59, + "learning_rate": 1.8551639184303867e-05, + "loss": 1.2814, + "step": 7064 + }, + { + "epoch": 0.59, + "learning_rate": 1.8551167975825567e-05, + "loss": 1.3427, + "step": 7065 + }, + { + "epoch": 0.59, + "learning_rate": 1.8550696696694778e-05, + "loss": 1.2906, + "step": 7066 + }, + { + "epoch": 0.59, + "learning_rate": 1.8550225346915397e-05, + "loss": 1.3006, + "step": 7067 + }, + { + "epoch": 0.6, + "learning_rate": 1.8549753926491317e-05, + "loss": 1.3225, + "step": 7068 + }, + { + "epoch": 0.6, + "learning_rate": 1.854928243542643e-05, + "loss": 1.2887, + "step": 7069 + }, + { + "epoch": 0.6, + "learning_rate": 1.8548810873724644e-05, + "loss": 1.3479, + "step": 7070 + }, + { + "epoch": 0.6, + "learning_rate": 1.8548339241389838e-05, + "loss": 1.2257, + "step": 7071 + }, + { + "epoch": 0.6, + "learning_rate": 1.8547867538425924e-05, + "loss": 1.3466, + "step": 7072 + }, + { + "epoch": 0.6, + "learning_rate": 1.854739576483679e-05, + "loss": 1.2706, + "step": 7073 + }, + { + "epoch": 0.6, + "learning_rate": 1.8546923920626337e-05, + "loss": 1.2932, + "step": 7074 + }, + { + "epoch": 0.6, + "learning_rate": 1.8546452005798464e-05, + "loss": 1.3309, + "step": 7075 + }, + { + "epoch": 0.6, + "learning_rate": 1.8545980020357065e-05, + "loss": 1.343, + "step": 7076 + }, + { + "epoch": 0.6, + "learning_rate": 1.8545507964306048e-05, + "loss": 1.2886, + "step": 7077 + }, + { + "epoch": 0.6, + "learning_rate": 1.854503583764931e-05, + "loss": 1.3042, + "step": 7078 + }, + { + "epoch": 0.6, + "learning_rate": 1.854456364039075e-05, + "loss": 1.2593, + "step": 7079 + }, + { + "epoch": 0.6, + "learning_rate": 1.8544091372534268e-05, + "loss": 1.2614, + "step": 7080 + }, + { + "epoch": 0.6, + "learning_rate": 1.854361903408377e-05, + "loss": 1.313, + "step": 7081 + }, + { + "epoch": 0.6, + "learning_rate": 1.854314662504316e-05, + "loss": 1.3937, + "step": 7082 + }, + { + "epoch": 0.6, + "learning_rate": 1.8542674145416336e-05, + "loss": 1.405, + "step": 7083 + }, + { + "epoch": 0.6, + "learning_rate": 1.8542201595207206e-05, + "loss": 1.386, + "step": 7084 + }, + { + "epoch": 0.6, + "learning_rate": 1.854172897441967e-05, + "loss": 1.3042, + "step": 7085 + }, + { + "epoch": 0.6, + "learning_rate": 1.854125628305764e-05, + "loss": 1.2826, + "step": 7086 + }, + { + "epoch": 0.6, + "learning_rate": 1.854078352112501e-05, + "loss": 1.291, + "step": 7087 + }, + { + "epoch": 0.6, + "learning_rate": 1.8540310688625697e-05, + "loss": 1.3336, + "step": 7088 + }, + { + "epoch": 0.6, + "learning_rate": 1.8539837785563603e-05, + "loss": 1.2099, + "step": 7089 + }, + { + "epoch": 0.6, + "learning_rate": 1.8539364811942638e-05, + "loss": 1.3339, + "step": 7090 + }, + { + "epoch": 0.6, + "learning_rate": 1.8538891767766708e-05, + "loss": 1.323, + "step": 7091 + }, + { + "epoch": 0.6, + "learning_rate": 1.8538418653039717e-05, + "loss": 1.4006, + "step": 7092 + }, + { + "epoch": 0.6, + "learning_rate": 1.8537945467765587e-05, + "loss": 1.3297, + "step": 7093 + }, + { + "epoch": 0.6, + "learning_rate": 1.853747221194821e-05, + "loss": 1.4369, + "step": 7094 + }, + { + "epoch": 0.6, + "learning_rate": 1.853699888559151e-05, + "loss": 1.2837, + "step": 7095 + }, + { + "epoch": 0.6, + "learning_rate": 1.8536525488699392e-05, + "loss": 1.4429, + "step": 7096 + }, + { + "epoch": 0.6, + "learning_rate": 1.853605202127577e-05, + "loss": 1.3527, + "step": 7097 + }, + { + "epoch": 0.6, + "learning_rate": 1.853557848332455e-05, + "loss": 1.2871, + "step": 7098 + }, + { + "epoch": 0.6, + "learning_rate": 1.8535104874849654e-05, + "loss": 1.3409, + "step": 7099 + }, + { + "epoch": 0.6, + "learning_rate": 1.8534631195854985e-05, + "loss": 1.29, + "step": 7100 + }, + { + "epoch": 0.6, + "learning_rate": 1.8534157446344467e-05, + "loss": 1.2926, + "step": 7101 + }, + { + "epoch": 0.6, + "learning_rate": 1.8533683626322004e-05, + "loss": 1.2121, + "step": 7102 + }, + { + "epoch": 0.6, + "learning_rate": 1.8533209735791516e-05, + "loss": 1.2993, + "step": 7103 + }, + { + "epoch": 0.6, + "learning_rate": 1.8532735774756922e-05, + "loss": 1.2724, + "step": 7104 + }, + { + "epoch": 0.6, + "learning_rate": 1.8532261743222133e-05, + "loss": 1.2277, + "step": 7105 + }, + { + "epoch": 0.6, + "learning_rate": 1.8531787641191065e-05, + "loss": 1.4173, + "step": 7106 + }, + { + "epoch": 0.6, + "learning_rate": 1.8531313468667638e-05, + "loss": 1.3406, + "step": 7107 + }, + { + "epoch": 0.6, + "learning_rate": 1.853083922565577e-05, + "loss": 1.2397, + "step": 7108 + }, + { + "epoch": 0.6, + "learning_rate": 1.853036491215938e-05, + "loss": 1.2927, + "step": 7109 + }, + { + "epoch": 0.6, + "learning_rate": 1.852989052818238e-05, + "loss": 1.2749, + "step": 7110 + }, + { + "epoch": 0.6, + "learning_rate": 1.8529416073728695e-05, + "loss": 1.3773, + "step": 7111 + }, + { + "epoch": 0.6, + "learning_rate": 1.8528941548802245e-05, + "loss": 1.2857, + "step": 7112 + }, + { + "epoch": 0.6, + "learning_rate": 1.8528466953406954e-05, + "loss": 1.3468, + "step": 7113 + }, + { + "epoch": 0.6, + "learning_rate": 1.8527992287546734e-05, + "loss": 1.3426, + "step": 7114 + }, + { + "epoch": 0.6, + "learning_rate": 1.8527517551225516e-05, + "loss": 1.3241, + "step": 7115 + }, + { + "epoch": 0.6, + "learning_rate": 1.852704274444722e-05, + "loss": 1.3175, + "step": 7116 + }, + { + "epoch": 0.6, + "learning_rate": 1.8526567867215767e-05, + "loss": 1.2216, + "step": 7117 + }, + { + "epoch": 0.6, + "learning_rate": 1.8526092919535078e-05, + "loss": 1.2685, + "step": 7118 + }, + { + "epoch": 0.6, + "learning_rate": 1.8525617901409082e-05, + "loss": 1.2937, + "step": 7119 + }, + { + "epoch": 0.6, + "learning_rate": 1.8525142812841705e-05, + "loss": 1.2631, + "step": 7120 + }, + { + "epoch": 0.6, + "learning_rate": 1.8524667653836867e-05, + "loss": 1.3158, + "step": 7121 + }, + { + "epoch": 0.6, + "learning_rate": 1.85241924243985e-05, + "loss": 1.3819, + "step": 7122 + }, + { + "epoch": 0.6, + "learning_rate": 1.8523717124530523e-05, + "loss": 1.2718, + "step": 7123 + }, + { + "epoch": 0.6, + "learning_rate": 1.852324175423687e-05, + "loss": 1.3161, + "step": 7124 + }, + { + "epoch": 0.6, + "learning_rate": 1.8522766313521465e-05, + "loss": 1.3562, + "step": 7125 + }, + { + "epoch": 0.6, + "learning_rate": 1.8522290802388236e-05, + "loss": 1.2889, + "step": 7126 + }, + { + "epoch": 0.6, + "learning_rate": 1.8521815220841114e-05, + "loss": 1.3563, + "step": 7127 + }, + { + "epoch": 0.6, + "learning_rate": 1.8521339568884026e-05, + "loss": 1.3644, + "step": 7128 + }, + { + "epoch": 0.6, + "learning_rate": 1.8520863846520905e-05, + "loss": 1.2807, + "step": 7129 + }, + { + "epoch": 0.6, + "learning_rate": 1.852038805375568e-05, + "loss": 1.2498, + "step": 7130 + }, + { + "epoch": 0.6, + "learning_rate": 1.851991219059228e-05, + "loss": 1.3868, + "step": 7131 + }, + { + "epoch": 0.6, + "learning_rate": 1.851943625703464e-05, + "loss": 1.3622, + "step": 7132 + }, + { + "epoch": 0.6, + "learning_rate": 1.851896025308669e-05, + "loss": 1.2516, + "step": 7133 + }, + { + "epoch": 0.6, + "learning_rate": 1.8518484178752367e-05, + "loss": 1.3149, + "step": 7134 + }, + { + "epoch": 0.6, + "learning_rate": 1.85180080340356e-05, + "loss": 1.2895, + "step": 7135 + }, + { + "epoch": 0.6, + "learning_rate": 1.8517531818940325e-05, + "loss": 1.3109, + "step": 7136 + }, + { + "epoch": 0.6, + "learning_rate": 1.8517055533470474e-05, + "loss": 1.2921, + "step": 7137 + }, + { + "epoch": 0.6, + "learning_rate": 1.8516579177629986e-05, + "loss": 1.3474, + "step": 7138 + }, + { + "epoch": 0.6, + "learning_rate": 1.8516102751422795e-05, + "loss": 1.3525, + "step": 7139 + }, + { + "epoch": 0.6, + "learning_rate": 1.8515626254852838e-05, + "loss": 1.2975, + "step": 7140 + }, + { + "epoch": 0.6, + "learning_rate": 1.851514968792405e-05, + "loss": 1.2236, + "step": 7141 + }, + { + "epoch": 0.6, + "learning_rate": 1.8514673050640373e-05, + "loss": 1.3539, + "step": 7142 + }, + { + "epoch": 0.6, + "learning_rate": 1.851419634300574e-05, + "loss": 1.3825, + "step": 7143 + }, + { + "epoch": 0.6, + "learning_rate": 1.8513719565024092e-05, + "loss": 1.4109, + "step": 7144 + }, + { + "epoch": 0.6, + "learning_rate": 1.8513242716699367e-05, + "loss": 1.3277, + "step": 7145 + }, + { + "epoch": 0.6, + "learning_rate": 1.8512765798035507e-05, + "loss": 1.2474, + "step": 7146 + }, + { + "epoch": 0.6, + "learning_rate": 1.851228880903645e-05, + "loss": 1.4454, + "step": 7147 + }, + { + "epoch": 0.6, + "learning_rate": 1.8511811749706142e-05, + "loss": 1.3076, + "step": 7148 + }, + { + "epoch": 0.6, + "learning_rate": 1.8511334620048524e-05, + "loss": 1.2145, + "step": 7149 + }, + { + "epoch": 0.6, + "learning_rate": 1.8510857420067527e-05, + "loss": 1.2644, + "step": 7150 + }, + { + "epoch": 0.6, + "learning_rate": 1.8510380149767107e-05, + "loss": 1.3834, + "step": 7151 + }, + { + "epoch": 0.6, + "learning_rate": 1.85099028091512e-05, + "loss": 1.2762, + "step": 7152 + }, + { + "epoch": 0.6, + "learning_rate": 1.8509425398223756e-05, + "loss": 1.3736, + "step": 7153 + }, + { + "epoch": 0.6, + "learning_rate": 1.8508947916988714e-05, + "loss": 1.3018, + "step": 7154 + }, + { + "epoch": 0.6, + "learning_rate": 1.8508470365450023e-05, + "loss": 1.3275, + "step": 7155 + }, + { + "epoch": 0.6, + "learning_rate": 1.8507992743611625e-05, + "loss": 1.2111, + "step": 7156 + }, + { + "epoch": 0.6, + "learning_rate": 1.8507515051477468e-05, + "loss": 1.3471, + "step": 7157 + }, + { + "epoch": 0.6, + "learning_rate": 1.85070372890515e-05, + "loss": 1.288, + "step": 7158 + }, + { + "epoch": 0.6, + "learning_rate": 1.8506559456337663e-05, + "loss": 1.3745, + "step": 7159 + }, + { + "epoch": 0.6, + "learning_rate": 1.8506081553339913e-05, + "loss": 1.3455, + "step": 7160 + }, + { + "epoch": 0.6, + "learning_rate": 1.8505603580062195e-05, + "loss": 1.3128, + "step": 7161 + }, + { + "epoch": 0.6, + "learning_rate": 1.8505125536508458e-05, + "loss": 1.3494, + "step": 7162 + }, + { + "epoch": 0.6, + "learning_rate": 1.8504647422682652e-05, + "loss": 1.2557, + "step": 7163 + }, + { + "epoch": 0.6, + "learning_rate": 1.8504169238588727e-05, + "loss": 1.2602, + "step": 7164 + }, + { + "epoch": 0.6, + "learning_rate": 1.8503690984230634e-05, + "loss": 1.2484, + "step": 7165 + }, + { + "epoch": 0.6, + "learning_rate": 1.8503212659612324e-05, + "loss": 1.4375, + "step": 7166 + }, + { + "epoch": 0.6, + "learning_rate": 1.850273426473775e-05, + "loss": 1.3304, + "step": 7167 + }, + { + "epoch": 0.6, + "learning_rate": 1.8502255799610863e-05, + "loss": 1.3369, + "step": 7168 + }, + { + "epoch": 0.6, + "learning_rate": 1.850177726423562e-05, + "loss": 1.2868, + "step": 7169 + }, + { + "epoch": 0.6, + "learning_rate": 1.850129865861597e-05, + "loss": 1.3796, + "step": 7170 + }, + { + "epoch": 0.6, + "learning_rate": 1.850081998275587e-05, + "loss": 1.2801, + "step": 7171 + }, + { + "epoch": 0.6, + "learning_rate": 1.850034123665928e-05, + "loss": 1.3081, + "step": 7172 + }, + { + "epoch": 0.6, + "learning_rate": 1.8499862420330144e-05, + "loss": 1.348, + "step": 7173 + }, + { + "epoch": 0.6, + "learning_rate": 1.8499383533772426e-05, + "loss": 1.2914, + "step": 7174 + }, + { + "epoch": 0.6, + "learning_rate": 1.849890457699008e-05, + "loss": 1.2284, + "step": 7175 + }, + { + "epoch": 0.6, + "learning_rate": 1.8498425549987065e-05, + "loss": 1.3657, + "step": 7176 + }, + { + "epoch": 0.6, + "learning_rate": 1.849794645276734e-05, + "loss": 1.283, + "step": 7177 + }, + { + "epoch": 0.6, + "learning_rate": 1.8497467285334863e-05, + "loss": 1.2058, + "step": 7178 + }, + { + "epoch": 0.6, + "learning_rate": 1.849698804769359e-05, + "loss": 1.1921, + "step": 7179 + }, + { + "epoch": 0.6, + "learning_rate": 1.8496508739847477e-05, + "loss": 1.3572, + "step": 7180 + }, + { + "epoch": 0.6, + "learning_rate": 1.8496029361800498e-05, + "loss": 1.3249, + "step": 7181 + }, + { + "epoch": 0.6, + "learning_rate": 1.8495549913556598e-05, + "loss": 1.2855, + "step": 7182 + }, + { + "epoch": 0.6, + "learning_rate": 1.849507039511975e-05, + "loss": 1.1732, + "step": 7183 + }, + { + "epoch": 0.6, + "learning_rate": 1.8494590806493913e-05, + "loss": 1.3678, + "step": 7184 + }, + { + "epoch": 0.6, + "learning_rate": 1.8494111147683044e-05, + "loss": 1.2237, + "step": 7185 + }, + { + "epoch": 0.61, + "learning_rate": 1.8493631418691113e-05, + "loss": 1.2733, + "step": 7186 + }, + { + "epoch": 0.61, + "learning_rate": 1.849315161952208e-05, + "loss": 1.3444, + "step": 7187 + }, + { + "epoch": 0.61, + "learning_rate": 1.849267175017991e-05, + "loss": 1.3851, + "step": 7188 + }, + { + "epoch": 0.61, + "learning_rate": 1.849219181066857e-05, + "loss": 1.2299, + "step": 7189 + }, + { + "epoch": 0.61, + "learning_rate": 1.8491711800992022e-05, + "loss": 1.3007, + "step": 7190 + }, + { + "epoch": 0.61, + "learning_rate": 1.8491231721154234e-05, + "loss": 1.3392, + "step": 7191 + }, + { + "epoch": 0.61, + "learning_rate": 1.849075157115917e-05, + "loss": 1.1867, + "step": 7192 + }, + { + "epoch": 0.61, + "learning_rate": 1.84902713510108e-05, + "loss": 1.3566, + "step": 7193 + }, + { + "epoch": 0.61, + "learning_rate": 1.848979106071309e-05, + "loss": 1.3789, + "step": 7194 + }, + { + "epoch": 0.61, + "learning_rate": 1.848931070027001e-05, + "loss": 1.225, + "step": 7195 + }, + { + "epoch": 0.61, + "learning_rate": 1.848883026968553e-05, + "loss": 1.3273, + "step": 7196 + }, + { + "epoch": 0.61, + "learning_rate": 1.8488349768963616e-05, + "loss": 1.3387, + "step": 7197 + }, + { + "epoch": 0.61, + "learning_rate": 1.8487869198108238e-05, + "loss": 1.3052, + "step": 7198 + }, + { + "epoch": 0.61, + "learning_rate": 1.848738855712337e-05, + "loss": 1.28, + "step": 7199 + }, + { + "epoch": 0.61, + "learning_rate": 1.8486907846012984e-05, + "loss": 1.3023, + "step": 7200 + }, + { + "epoch": 0.61, + "learning_rate": 1.8486427064781046e-05, + "loss": 1.3685, + "step": 7201 + }, + { + "epoch": 0.61, + "learning_rate": 1.8485946213431528e-05, + "loss": 1.3121, + "step": 7202 + }, + { + "epoch": 0.61, + "learning_rate": 1.8485465291968413e-05, + "loss": 1.3016, + "step": 7203 + }, + { + "epoch": 0.61, + "learning_rate": 1.8484984300395664e-05, + "loss": 1.3434, + "step": 7204 + }, + { + "epoch": 0.61, + "learning_rate": 1.8484503238717263e-05, + "loss": 1.363, + "step": 7205 + }, + { + "epoch": 0.61, + "learning_rate": 1.848402210693718e-05, + "loss": 1.3219, + "step": 7206 + }, + { + "epoch": 0.61, + "learning_rate": 1.848354090505939e-05, + "loss": 1.3041, + "step": 7207 + }, + { + "epoch": 0.61, + "learning_rate": 1.848305963308787e-05, + "loss": 1.1896, + "step": 7208 + }, + { + "epoch": 0.61, + "learning_rate": 1.8482578291026593e-05, + "loss": 1.3058, + "step": 7209 + }, + { + "epoch": 0.61, + "learning_rate": 1.8482096878879542e-05, + "loss": 1.233, + "step": 7210 + }, + { + "epoch": 0.61, + "learning_rate": 1.8481615396650693e-05, + "loss": 1.2515, + "step": 7211 + }, + { + "epoch": 0.61, + "learning_rate": 1.8481133844344023e-05, + "loss": 1.2996, + "step": 7212 + }, + { + "epoch": 0.61, + "learning_rate": 1.848065222196351e-05, + "loss": 1.2861, + "step": 7213 + }, + { + "epoch": 0.61, + "learning_rate": 1.8480170529513134e-05, + "loss": 1.3614, + "step": 7214 + }, + { + "epoch": 0.61, + "learning_rate": 1.8479688766996878e-05, + "loss": 1.2786, + "step": 7215 + }, + { + "epoch": 0.61, + "learning_rate": 1.8479206934418715e-05, + "loss": 1.2655, + "step": 7216 + }, + { + "epoch": 0.61, + "learning_rate": 1.8478725031782633e-05, + "loss": 1.2546, + "step": 7217 + }, + { + "epoch": 0.61, + "learning_rate": 1.8478243059092613e-05, + "loss": 1.2469, + "step": 7218 + }, + { + "epoch": 0.61, + "learning_rate": 1.847776101635263e-05, + "loss": 1.2968, + "step": 7219 + }, + { + "epoch": 0.61, + "learning_rate": 1.8477278903566677e-05, + "loss": 1.3107, + "step": 7220 + }, + { + "epoch": 0.61, + "learning_rate": 1.8476796720738734e-05, + "loss": 1.3464, + "step": 7221 + }, + { + "epoch": 0.61, + "learning_rate": 1.847631446787278e-05, + "loss": 1.1668, + "step": 7222 + }, + { + "epoch": 0.61, + "learning_rate": 1.8475832144972804e-05, + "loss": 1.2417, + "step": 7223 + }, + { + "epoch": 0.61, + "learning_rate": 1.847534975204279e-05, + "loss": 1.2149, + "step": 7224 + }, + { + "epoch": 0.61, + "learning_rate": 1.8474867289086725e-05, + "loss": 1.4207, + "step": 7225 + }, + { + "epoch": 0.61, + "learning_rate": 1.8474384756108592e-05, + "loss": 1.3181, + "step": 7226 + }, + { + "epoch": 0.61, + "learning_rate": 1.847390215311238e-05, + "loss": 1.27, + "step": 7227 + }, + { + "epoch": 0.61, + "learning_rate": 1.847341948010208e-05, + "loss": 1.1729, + "step": 7228 + }, + { + "epoch": 0.61, + "learning_rate": 1.8472936737081673e-05, + "loss": 1.3093, + "step": 7229 + }, + { + "epoch": 0.61, + "learning_rate": 1.8472453924055155e-05, + "loss": 1.2035, + "step": 7230 + }, + { + "epoch": 0.61, + "learning_rate": 1.8471971041026508e-05, + "loss": 1.2601, + "step": 7231 + }, + { + "epoch": 0.61, + "learning_rate": 1.847148808799972e-05, + "loss": 1.3037, + "step": 7232 + }, + { + "epoch": 0.61, + "learning_rate": 1.8471005064978794e-05, + "loss": 1.2273, + "step": 7233 + }, + { + "epoch": 0.61, + "learning_rate": 1.847052197196771e-05, + "loss": 1.3333, + "step": 7234 + }, + { + "epoch": 0.61, + "learning_rate": 1.8470038808970466e-05, + "loss": 1.313, + "step": 7235 + }, + { + "epoch": 0.61, + "learning_rate": 1.8469555575991048e-05, + "loss": 1.3228, + "step": 7236 + }, + { + "epoch": 0.61, + "learning_rate": 1.846907227303345e-05, + "loss": 1.325, + "step": 7237 + }, + { + "epoch": 0.61, + "learning_rate": 1.8468588900101667e-05, + "loss": 1.3667, + "step": 7238 + }, + { + "epoch": 0.61, + "learning_rate": 1.846810545719969e-05, + "loss": 1.2968, + "step": 7239 + }, + { + "epoch": 0.61, + "learning_rate": 1.846762194433152e-05, + "loss": 1.368, + "step": 7240 + }, + { + "epoch": 0.61, + "learning_rate": 1.8467138361501145e-05, + "loss": 1.2968, + "step": 7241 + }, + { + "epoch": 0.61, + "learning_rate": 1.8466654708712563e-05, + "loss": 1.2584, + "step": 7242 + }, + { + "epoch": 0.61, + "learning_rate": 1.846617098596977e-05, + "loss": 1.2558, + "step": 7243 + }, + { + "epoch": 0.61, + "learning_rate": 1.8465687193276767e-05, + "loss": 1.368, + "step": 7244 + }, + { + "epoch": 0.61, + "learning_rate": 1.846520333063754e-05, + "loss": 1.291, + "step": 7245 + }, + { + "epoch": 0.61, + "learning_rate": 1.8464719398056098e-05, + "loss": 1.3204, + "step": 7246 + }, + { + "epoch": 0.61, + "learning_rate": 1.8464235395536435e-05, + "loss": 1.2536, + "step": 7247 + }, + { + "epoch": 0.61, + "learning_rate": 1.846375132308255e-05, + "loss": 1.3563, + "step": 7248 + }, + { + "epoch": 0.61, + "learning_rate": 1.8463267180698443e-05, + "loss": 1.239, + "step": 7249 + }, + { + "epoch": 0.61, + "learning_rate": 1.8462782968388113e-05, + "loss": 1.2922, + "step": 7250 + }, + { + "epoch": 0.61, + "learning_rate": 1.8462298686155565e-05, + "loss": 1.2772, + "step": 7251 + }, + { + "epoch": 0.61, + "learning_rate": 1.846181433400479e-05, + "loss": 1.2388, + "step": 7252 + }, + { + "epoch": 0.61, + "learning_rate": 1.8461329911939803e-05, + "loss": 1.4206, + "step": 7253 + }, + { + "epoch": 0.61, + "learning_rate": 1.84608454199646e-05, + "loss": 1.3641, + "step": 7254 + }, + { + "epoch": 0.61, + "learning_rate": 1.846036085808318e-05, + "loss": 1.3127, + "step": 7255 + }, + { + "epoch": 0.61, + "learning_rate": 1.8459876226299557e-05, + "loss": 1.2816, + "step": 7256 + }, + { + "epoch": 0.61, + "learning_rate": 1.8459391524617727e-05, + "loss": 1.3771, + "step": 7257 + }, + { + "epoch": 0.61, + "learning_rate": 1.8458906753041696e-05, + "loss": 1.3519, + "step": 7258 + }, + { + "epoch": 0.61, + "learning_rate": 1.845842191157547e-05, + "loss": 1.3284, + "step": 7259 + }, + { + "epoch": 0.61, + "learning_rate": 1.845793700022305e-05, + "loss": 1.1976, + "step": 7260 + }, + { + "epoch": 0.61, + "learning_rate": 1.8457452018988455e-05, + "loss": 1.2133, + "step": 7261 + }, + { + "epoch": 0.61, + "learning_rate": 1.8456966967875685e-05, + "loss": 1.2504, + "step": 7262 + }, + { + "epoch": 0.61, + "learning_rate": 1.845648184688874e-05, + "loss": 1.3603, + "step": 7263 + }, + { + "epoch": 0.61, + "learning_rate": 1.8455996656031642e-05, + "loss": 1.3248, + "step": 7264 + }, + { + "epoch": 0.61, + "learning_rate": 1.8455511395308393e-05, + "loss": 1.3659, + "step": 7265 + }, + { + "epoch": 0.61, + "learning_rate": 1.8455026064723e-05, + "loss": 1.3504, + "step": 7266 + }, + { + "epoch": 0.61, + "learning_rate": 1.845454066427948e-05, + "loss": 1.2797, + "step": 7267 + }, + { + "epoch": 0.61, + "learning_rate": 1.8454055193981834e-05, + "loss": 1.2164, + "step": 7268 + }, + { + "epoch": 0.61, + "learning_rate": 1.8453569653834084e-05, + "loss": 1.2688, + "step": 7269 + }, + { + "epoch": 0.61, + "learning_rate": 1.845308404384023e-05, + "loss": 1.3168, + "step": 7270 + }, + { + "epoch": 0.61, + "learning_rate": 1.8452598364004292e-05, + "loss": 1.2886, + "step": 7271 + }, + { + "epoch": 0.61, + "learning_rate": 1.845211261433028e-05, + "loss": 1.2977, + "step": 7272 + }, + { + "epoch": 0.61, + "learning_rate": 1.8451626794822215e-05, + "loss": 1.3391, + "step": 7273 + }, + { + "epoch": 0.61, + "learning_rate": 1.84511409054841e-05, + "loss": 1.274, + "step": 7274 + }, + { + "epoch": 0.61, + "learning_rate": 1.8450654946319954e-05, + "loss": 1.4048, + "step": 7275 + }, + { + "epoch": 0.61, + "learning_rate": 1.8450168917333793e-05, + "loss": 1.3533, + "step": 7276 + }, + { + "epoch": 0.61, + "learning_rate": 1.844968281852963e-05, + "loss": 1.2604, + "step": 7277 + }, + { + "epoch": 0.61, + "learning_rate": 1.8449196649911486e-05, + "loss": 1.3407, + "step": 7278 + }, + { + "epoch": 0.61, + "learning_rate": 1.8448710411483374e-05, + "loss": 1.3902, + "step": 7279 + }, + { + "epoch": 0.61, + "learning_rate": 1.844822410324931e-05, + "loss": 1.1767, + "step": 7280 + }, + { + "epoch": 0.61, + "learning_rate": 1.844773772521332e-05, + "loss": 1.2318, + "step": 7281 + }, + { + "epoch": 0.61, + "learning_rate": 1.8447251277379415e-05, + "loss": 1.217, + "step": 7282 + }, + { + "epoch": 0.61, + "learning_rate": 1.8446764759751615e-05, + "loss": 1.3497, + "step": 7283 + }, + { + "epoch": 0.61, + "learning_rate": 1.8446278172333945e-05, + "loss": 1.2644, + "step": 7284 + }, + { + "epoch": 0.61, + "learning_rate": 1.8445791515130416e-05, + "loss": 1.2812, + "step": 7285 + }, + { + "epoch": 0.61, + "learning_rate": 1.844530478814506e-05, + "loss": 1.3205, + "step": 7286 + }, + { + "epoch": 0.61, + "learning_rate": 1.8444817991381893e-05, + "loss": 1.2821, + "step": 7287 + }, + { + "epoch": 0.61, + "learning_rate": 1.8444331124844932e-05, + "loss": 1.3526, + "step": 7288 + }, + { + "epoch": 0.61, + "learning_rate": 1.8443844188538206e-05, + "loss": 1.3192, + "step": 7289 + }, + { + "epoch": 0.61, + "learning_rate": 1.844335718246574e-05, + "loss": 1.2535, + "step": 7290 + }, + { + "epoch": 0.61, + "learning_rate": 1.8442870106631552e-05, + "loss": 1.3, + "step": 7291 + }, + { + "epoch": 0.61, + "learning_rate": 1.844238296103967e-05, + "loss": 1.3174, + "step": 7292 + }, + { + "epoch": 0.61, + "learning_rate": 1.844189574569412e-05, + "loss": 1.396, + "step": 7293 + }, + { + "epoch": 0.61, + "learning_rate": 1.8441408460598923e-05, + "loss": 1.3879, + "step": 7294 + }, + { + "epoch": 0.61, + "learning_rate": 1.8440921105758107e-05, + "loss": 1.4336, + "step": 7295 + }, + { + "epoch": 0.61, + "learning_rate": 1.84404336811757e-05, + "loss": 1.3689, + "step": 7296 + }, + { + "epoch": 0.61, + "learning_rate": 1.843994618685573e-05, + "loss": 1.2774, + "step": 7297 + }, + { + "epoch": 0.61, + "learning_rate": 1.8439458622802225e-05, + "loss": 1.2891, + "step": 7298 + }, + { + "epoch": 0.61, + "learning_rate": 1.843897098901921e-05, + "loss": 1.311, + "step": 7299 + }, + { + "epoch": 0.61, + "learning_rate": 1.8438483285510717e-05, + "loss": 1.3501, + "step": 7300 + }, + { + "epoch": 0.61, + "learning_rate": 1.8437995512280775e-05, + "loss": 1.3659, + "step": 7301 + }, + { + "epoch": 0.61, + "learning_rate": 1.8437507669333407e-05, + "loss": 1.273, + "step": 7302 + }, + { + "epoch": 0.61, + "learning_rate": 1.8437019756672657e-05, + "loss": 1.2195, + "step": 7303 + }, + { + "epoch": 0.61, + "learning_rate": 1.8436531774302548e-05, + "loss": 1.256, + "step": 7304 + }, + { + "epoch": 0.62, + "learning_rate": 1.843604372222711e-05, + "loss": 1.4382, + "step": 7305 + }, + { + "epoch": 0.62, + "learning_rate": 1.8435555600450384e-05, + "loss": 1.2338, + "step": 7306 + }, + { + "epoch": 0.62, + "learning_rate": 1.8435067408976396e-05, + "loss": 1.2464, + "step": 7307 + }, + { + "epoch": 0.62, + "learning_rate": 1.843457914780918e-05, + "loss": 1.3106, + "step": 7308 + }, + { + "epoch": 0.62, + "learning_rate": 1.8434090816952774e-05, + "loss": 1.3997, + "step": 7309 + }, + { + "epoch": 0.62, + "learning_rate": 1.843360241641121e-05, + "loss": 1.2707, + "step": 7310 + }, + { + "epoch": 0.62, + "learning_rate": 1.8433113946188522e-05, + "loss": 1.396, + "step": 7311 + }, + { + "epoch": 0.62, + "learning_rate": 1.843262540628875e-05, + "loss": 1.3386, + "step": 7312 + }, + { + "epoch": 0.62, + "learning_rate": 1.843213679671593e-05, + "loss": 1.2715, + "step": 7313 + }, + { + "epoch": 0.62, + "learning_rate": 1.843164811747409e-05, + "loss": 1.3042, + "step": 7314 + }, + { + "epoch": 0.62, + "learning_rate": 1.843115936856728e-05, + "loss": 1.3285, + "step": 7315 + }, + { + "epoch": 0.62, + "learning_rate": 1.843067054999953e-05, + "loss": 1.2595, + "step": 7316 + }, + { + "epoch": 0.62, + "learning_rate": 1.8430181661774885e-05, + "loss": 1.2722, + "step": 7317 + }, + { + "epoch": 0.62, + "learning_rate": 1.8429692703897378e-05, + "loss": 1.4199, + "step": 7318 + }, + { + "epoch": 0.62, + "learning_rate": 1.8429203676371056e-05, + "loss": 1.2279, + "step": 7319 + }, + { + "epoch": 0.62, + "learning_rate": 1.8428714579199954e-05, + "loss": 1.3248, + "step": 7320 + }, + { + "epoch": 0.62, + "learning_rate": 1.8428225412388113e-05, + "loss": 1.3981, + "step": 7321 + }, + { + "epoch": 0.62, + "learning_rate": 1.8427736175939577e-05, + "loss": 1.3378, + "step": 7322 + }, + { + "epoch": 0.62, + "learning_rate": 1.842724686985839e-05, + "loss": 1.2648, + "step": 7323 + }, + { + "epoch": 0.62, + "learning_rate": 1.842675749414859e-05, + "loss": 1.2816, + "step": 7324 + }, + { + "epoch": 0.62, + "learning_rate": 1.8426268048814223e-05, + "loss": 1.3114, + "step": 7325 + }, + { + "epoch": 0.62, + "learning_rate": 1.8425778533859332e-05, + "loss": 1.2273, + "step": 7326 + }, + { + "epoch": 0.62, + "learning_rate": 1.842528894928796e-05, + "loss": 1.3786, + "step": 7327 + }, + { + "epoch": 0.62, + "learning_rate": 1.8424799295104155e-05, + "loss": 1.2788, + "step": 7328 + }, + { + "epoch": 0.62, + "learning_rate": 1.8424309571311963e-05, + "loss": 1.3452, + "step": 7329 + }, + { + "epoch": 0.62, + "learning_rate": 1.842381977791543e-05, + "loss": 1.2557, + "step": 7330 + }, + { + "epoch": 0.62, + "learning_rate": 1.8423329914918606e-05, + "loss": 1.3088, + "step": 7331 + }, + { + "epoch": 0.62, + "learning_rate": 1.8422839982325528e-05, + "loss": 1.3222, + "step": 7332 + }, + { + "epoch": 0.62, + "learning_rate": 1.8422349980140254e-05, + "loss": 1.2803, + "step": 7333 + }, + { + "epoch": 0.62, + "learning_rate": 1.842185990836683e-05, + "loss": 1.2363, + "step": 7334 + }, + { + "epoch": 0.62, + "learning_rate": 1.8421369767009298e-05, + "loss": 1.1888, + "step": 7335 + }, + { + "epoch": 0.62, + "learning_rate": 1.8420879556071717e-05, + "loss": 1.264, + "step": 7336 + }, + { + "epoch": 0.62, + "learning_rate": 1.8420389275558138e-05, + "loss": 1.3164, + "step": 7337 + }, + { + "epoch": 0.62, + "learning_rate": 1.84198989254726e-05, + "loss": 1.2922, + "step": 7338 + }, + { + "epoch": 0.62, + "learning_rate": 1.841940850581917e-05, + "loss": 1.2537, + "step": 7339 + }, + { + "epoch": 0.62, + "learning_rate": 1.841891801660189e-05, + "loss": 1.3587, + "step": 7340 + }, + { + "epoch": 0.62, + "learning_rate": 1.841842745782481e-05, + "loss": 1.3682, + "step": 7341 + }, + { + "epoch": 0.62, + "learning_rate": 1.8417936829491997e-05, + "loss": 1.2504, + "step": 7342 + }, + { + "epoch": 0.62, + "learning_rate": 1.8417446131607487e-05, + "loss": 1.347, + "step": 7343 + }, + { + "epoch": 0.62, + "learning_rate": 1.841695536417535e-05, + "loss": 1.3329, + "step": 7344 + }, + { + "epoch": 0.62, + "learning_rate": 1.8416464527199632e-05, + "loss": 1.235, + "step": 7345 + }, + { + "epoch": 0.62, + "learning_rate": 1.841597362068439e-05, + "loss": 1.3241, + "step": 7346 + }, + { + "epoch": 0.62, + "learning_rate": 1.841548264463368e-05, + "loss": 1.3373, + "step": 7347 + }, + { + "epoch": 0.62, + "learning_rate": 1.8414991599051558e-05, + "loss": 1.3476, + "step": 7348 + }, + { + "epoch": 0.62, + "learning_rate": 1.8414500483942084e-05, + "loss": 1.3172, + "step": 7349 + }, + { + "epoch": 0.62, + "learning_rate": 1.8414009299309315e-05, + "loss": 1.2424, + "step": 7350 + }, + { + "epoch": 0.62, + "learning_rate": 1.8413518045157307e-05, + "loss": 1.2584, + "step": 7351 + }, + { + "epoch": 0.62, + "learning_rate": 1.841302672149012e-05, + "loss": 1.292, + "step": 7352 + }, + { + "epoch": 0.62, + "learning_rate": 1.8412535328311813e-05, + "loss": 1.3736, + "step": 7353 + }, + { + "epoch": 0.62, + "learning_rate": 1.841204386562645e-05, + "loss": 1.3035, + "step": 7354 + }, + { + "epoch": 0.62, + "learning_rate": 1.8411552333438084e-05, + "loss": 1.2667, + "step": 7355 + }, + { + "epoch": 0.62, + "learning_rate": 1.8411060731750782e-05, + "loss": 1.3086, + "step": 7356 + }, + { + "epoch": 0.62, + "learning_rate": 1.8410569060568607e-05, + "loss": 1.3015, + "step": 7357 + }, + { + "epoch": 0.62, + "learning_rate": 1.8410077319895615e-05, + "loss": 1.3774, + "step": 7358 + }, + { + "epoch": 0.62, + "learning_rate": 1.8409585509735876e-05, + "loss": 1.33, + "step": 7359 + }, + { + "epoch": 0.62, + "learning_rate": 1.8409093630093448e-05, + "loss": 1.2881, + "step": 7360 + }, + { + "epoch": 0.62, + "learning_rate": 1.8408601680972397e-05, + "loss": 1.2702, + "step": 7361 + }, + { + "epoch": 0.62, + "learning_rate": 1.840810966237679e-05, + "loss": 1.2599, + "step": 7362 + }, + { + "epoch": 0.62, + "learning_rate": 1.840761757431069e-05, + "loss": 1.2968, + "step": 7363 + }, + { + "epoch": 0.62, + "learning_rate": 1.8407125416778158e-05, + "loss": 1.3467, + "step": 7364 + }, + { + "epoch": 0.62, + "learning_rate": 1.8406633189783266e-05, + "loss": 1.2585, + "step": 7365 + }, + { + "epoch": 0.62, + "learning_rate": 1.8406140893330087e-05, + "loss": 1.3002, + "step": 7366 + }, + { + "epoch": 0.62, + "learning_rate": 1.8405648527422676e-05, + "loss": 1.3242, + "step": 7367 + }, + { + "epoch": 0.62, + "learning_rate": 1.8405156092065106e-05, + "loss": 1.3408, + "step": 7368 + }, + { + "epoch": 0.62, + "learning_rate": 1.840466358726145e-05, + "loss": 1.3798, + "step": 7369 + }, + { + "epoch": 0.62, + "learning_rate": 1.840417101301577e-05, + "loss": 1.2187, + "step": 7370 + }, + { + "epoch": 0.62, + "learning_rate": 1.8403678369332143e-05, + "loss": 1.264, + "step": 7371 + }, + { + "epoch": 0.62, + "learning_rate": 1.8403185656214635e-05, + "loss": 1.327, + "step": 7372 + }, + { + "epoch": 0.62, + "learning_rate": 1.8402692873667316e-05, + "loss": 1.2917, + "step": 7373 + }, + { + "epoch": 0.62, + "learning_rate": 1.840220002169426e-05, + "loss": 1.3299, + "step": 7374 + }, + { + "epoch": 0.62, + "learning_rate": 1.8401707100299536e-05, + "loss": 1.2469, + "step": 7375 + }, + { + "epoch": 0.62, + "learning_rate": 1.8401214109487226e-05, + "loss": 1.2585, + "step": 7376 + }, + { + "epoch": 0.62, + "learning_rate": 1.840072104926139e-05, + "loss": 1.3394, + "step": 7377 + }, + { + "epoch": 0.62, + "learning_rate": 1.840022791962611e-05, + "loss": 1.3823, + "step": 7378 + }, + { + "epoch": 0.62, + "learning_rate": 1.839973472058546e-05, + "loss": 1.3182, + "step": 7379 + }, + { + "epoch": 0.62, + "learning_rate": 1.839924145214351e-05, + "loss": 1.3568, + "step": 7380 + }, + { + "epoch": 0.62, + "learning_rate": 1.839874811430434e-05, + "loss": 1.3355, + "step": 7381 + }, + { + "epoch": 0.62, + "learning_rate": 1.839825470707203e-05, + "loss": 1.3456, + "step": 7382 + }, + { + "epoch": 0.62, + "learning_rate": 1.8397761230450647e-05, + "loss": 1.2381, + "step": 7383 + }, + { + "epoch": 0.62, + "learning_rate": 1.8397267684444272e-05, + "loss": 1.2568, + "step": 7384 + }, + { + "epoch": 0.62, + "learning_rate": 1.8396774069056985e-05, + "loss": 1.2633, + "step": 7385 + }, + { + "epoch": 0.62, + "learning_rate": 1.8396280384292865e-05, + "loss": 1.2731, + "step": 7386 + }, + { + "epoch": 0.62, + "learning_rate": 1.8395786630155988e-05, + "loss": 1.3778, + "step": 7387 + }, + { + "epoch": 0.62, + "learning_rate": 1.8395292806650436e-05, + "loss": 1.2979, + "step": 7388 + }, + { + "epoch": 0.62, + "learning_rate": 1.839479891378029e-05, + "loss": 1.319, + "step": 7389 + }, + { + "epoch": 0.62, + "learning_rate": 1.8394304951549625e-05, + "loss": 1.3073, + "step": 7390 + }, + { + "epoch": 0.62, + "learning_rate": 1.8393810919962525e-05, + "loss": 1.2996, + "step": 7391 + }, + { + "epoch": 0.62, + "learning_rate": 1.8393316819023078e-05, + "loss": 1.2508, + "step": 7392 + }, + { + "epoch": 0.62, + "learning_rate": 1.8392822648735356e-05, + "loss": 1.3786, + "step": 7393 + }, + { + "epoch": 0.62, + "learning_rate": 1.839232840910345e-05, + "loss": 1.29, + "step": 7394 + }, + { + "epoch": 0.62, + "learning_rate": 1.839183410013144e-05, + "loss": 1.3117, + "step": 7395 + }, + { + "epoch": 0.62, + "learning_rate": 1.839133972182341e-05, + "loss": 1.3003, + "step": 7396 + }, + { + "epoch": 0.62, + "learning_rate": 1.839084527418345e-05, + "loss": 1.2471, + "step": 7397 + }, + { + "epoch": 0.62, + "learning_rate": 1.8390350757215633e-05, + "loss": 1.2717, + "step": 7398 + }, + { + "epoch": 0.62, + "learning_rate": 1.8389856170924057e-05, + "loss": 1.3137, + "step": 7399 + }, + { + "epoch": 0.62, + "learning_rate": 1.838936151531281e-05, + "loss": 1.318, + "step": 7400 + }, + { + "epoch": 0.62, + "learning_rate": 1.8388866790385965e-05, + "loss": 1.2829, + "step": 7401 + }, + { + "epoch": 0.62, + "learning_rate": 1.838837199614762e-05, + "loss": 1.3585, + "step": 7402 + }, + { + "epoch": 0.62, + "learning_rate": 1.838787713260186e-05, + "loss": 1.2845, + "step": 7403 + }, + { + "epoch": 0.62, + "learning_rate": 1.838738219975278e-05, + "loss": 1.332, + "step": 7404 + }, + { + "epoch": 0.62, + "learning_rate": 1.8386887197604458e-05, + "loss": 1.2921, + "step": 7405 + }, + { + "epoch": 0.62, + "learning_rate": 1.8386392126160992e-05, + "loss": 1.3249, + "step": 7406 + }, + { + "epoch": 0.62, + "learning_rate": 1.838589698542647e-05, + "loss": 1.2717, + "step": 7407 + }, + { + "epoch": 0.62, + "learning_rate": 1.8385401775404986e-05, + "loss": 1.2647, + "step": 7408 + }, + { + "epoch": 0.62, + "learning_rate": 1.8384906496100625e-05, + "loss": 1.2929, + "step": 7409 + }, + { + "epoch": 0.62, + "learning_rate": 1.8384411147517486e-05, + "loss": 1.3559, + "step": 7410 + }, + { + "epoch": 0.62, + "learning_rate": 1.8383915729659656e-05, + "loss": 1.2883, + "step": 7411 + }, + { + "epoch": 0.62, + "learning_rate": 1.8383420242531234e-05, + "loss": 1.3735, + "step": 7412 + }, + { + "epoch": 0.62, + "learning_rate": 1.8382924686136307e-05, + "loss": 1.2603, + "step": 7413 + }, + { + "epoch": 0.62, + "learning_rate": 1.8382429060478977e-05, + "loss": 1.3497, + "step": 7414 + }, + { + "epoch": 0.62, + "learning_rate": 1.8381933365563334e-05, + "loss": 1.3629, + "step": 7415 + }, + { + "epoch": 0.62, + "learning_rate": 1.8381437601393475e-05, + "loss": 1.2802, + "step": 7416 + }, + { + "epoch": 0.62, + "learning_rate": 1.83809417679735e-05, + "loss": 1.3207, + "step": 7417 + }, + { + "epoch": 0.62, + "learning_rate": 1.8380445865307494e-05, + "loss": 1.2949, + "step": 7418 + }, + { + "epoch": 0.62, + "learning_rate": 1.8379949893399565e-05, + "loss": 1.3845, + "step": 7419 + }, + { + "epoch": 0.62, + "learning_rate": 1.837945385225381e-05, + "loss": 1.2976, + "step": 7420 + }, + { + "epoch": 0.62, + "learning_rate": 1.8378957741874326e-05, + "loss": 1.2393, + "step": 7421 + }, + { + "epoch": 0.62, + "learning_rate": 1.837846156226521e-05, + "loss": 1.2628, + "step": 7422 + }, + { + "epoch": 0.62, + "learning_rate": 1.8377965313430563e-05, + "loss": 1.1962, + "step": 7423 + }, + { + "epoch": 0.63, + "learning_rate": 1.8377468995374485e-05, + "loss": 1.3076, + "step": 7424 + }, + { + "epoch": 0.63, + "learning_rate": 1.8376972608101077e-05, + "loss": 1.3893, + "step": 7425 + }, + { + "epoch": 0.63, + "learning_rate": 1.8376476151614443e-05, + "loss": 1.2725, + "step": 7426 + }, + { + "epoch": 0.63, + "learning_rate": 1.837597962591868e-05, + "loss": 1.3416, + "step": 7427 + }, + { + "epoch": 0.63, + "learning_rate": 1.837548303101789e-05, + "loss": 1.3246, + "step": 7428 + }, + { + "epoch": 0.63, + "learning_rate": 1.8374986366916185e-05, + "loss": 1.3149, + "step": 7429 + }, + { + "epoch": 0.63, + "learning_rate": 1.8374489633617658e-05, + "loss": 1.2734, + "step": 7430 + }, + { + "epoch": 0.63, + "learning_rate": 1.8373992831126416e-05, + "loss": 1.3523, + "step": 7431 + }, + { + "epoch": 0.63, + "learning_rate": 1.837349595944657e-05, + "loss": 1.3773, + "step": 7432 + }, + { + "epoch": 0.63, + "learning_rate": 1.8372999018582218e-05, + "loss": 1.3734, + "step": 7433 + }, + { + "epoch": 0.63, + "learning_rate": 1.8372502008537468e-05, + "loss": 1.3813, + "step": 7434 + }, + { + "epoch": 0.63, + "learning_rate": 1.837200492931643e-05, + "loss": 1.3134, + "step": 7435 + }, + { + "epoch": 0.63, + "learning_rate": 1.8371507780923202e-05, + "loss": 1.1986, + "step": 7436 + }, + { + "epoch": 0.63, + "learning_rate": 1.8371010563361903e-05, + "loss": 1.3241, + "step": 7437 + }, + { + "epoch": 0.63, + "learning_rate": 1.8370513276636633e-05, + "loss": 1.3389, + "step": 7438 + }, + { + "epoch": 0.63, + "learning_rate": 1.83700159207515e-05, + "loss": 1.3149, + "step": 7439 + }, + { + "epoch": 0.63, + "learning_rate": 1.836951849571062e-05, + "loss": 1.2862, + "step": 7440 + }, + { + "epoch": 0.63, + "learning_rate": 1.8369021001518104e-05, + "loss": 1.3751, + "step": 7441 + }, + { + "epoch": 0.63, + "learning_rate": 1.8368523438178053e-05, + "loss": 1.3966, + "step": 7442 + }, + { + "epoch": 0.63, + "learning_rate": 1.8368025805694584e-05, + "loss": 1.2943, + "step": 7443 + }, + { + "epoch": 0.63, + "learning_rate": 1.8367528104071807e-05, + "loss": 1.3065, + "step": 7444 + }, + { + "epoch": 0.63, + "learning_rate": 1.8367030333313835e-05, + "loss": 1.187, + "step": 7445 + }, + { + "epoch": 0.63, + "learning_rate": 1.836653249342478e-05, + "loss": 1.3086, + "step": 7446 + }, + { + "epoch": 0.63, + "learning_rate": 1.836603458440876e-05, + "loss": 1.2489, + "step": 7447 + }, + { + "epoch": 0.63, + "learning_rate": 1.8365536606269884e-05, + "loss": 1.3305, + "step": 7448 + }, + { + "epoch": 0.63, + "learning_rate": 1.8365038559012263e-05, + "loss": 1.2787, + "step": 7449 + }, + { + "epoch": 0.63, + "learning_rate": 1.8364540442640017e-05, + "loss": 1.2673, + "step": 7450 + }, + { + "epoch": 0.63, + "learning_rate": 1.8364042257157265e-05, + "loss": 1.3347, + "step": 7451 + }, + { + "epoch": 0.63, + "learning_rate": 1.8363544002568114e-05, + "loss": 1.3373, + "step": 7452 + }, + { + "epoch": 0.63, + "learning_rate": 1.836304567887669e-05, + "loss": 1.2411, + "step": 7453 + }, + { + "epoch": 0.63, + "learning_rate": 1.8362547286087104e-05, + "loss": 1.3128, + "step": 7454 + }, + { + "epoch": 0.63, + "learning_rate": 1.8362048824203474e-05, + "loss": 1.3671, + "step": 7455 + }, + { + "epoch": 0.63, + "learning_rate": 1.8361550293229925e-05, + "loss": 1.2509, + "step": 7456 + }, + { + "epoch": 0.63, + "learning_rate": 1.836105169317057e-05, + "loss": 1.3732, + "step": 7457 + }, + { + "epoch": 0.63, + "learning_rate": 1.8360553024029528e-05, + "loss": 1.2033, + "step": 7458 + }, + { + "epoch": 0.63, + "learning_rate": 1.836005428581092e-05, + "loss": 1.299, + "step": 7459 + }, + { + "epoch": 0.63, + "learning_rate": 1.8359555478518873e-05, + "loss": 1.3094, + "step": 7460 + }, + { + "epoch": 0.63, + "learning_rate": 1.8359056602157502e-05, + "loss": 1.2675, + "step": 7461 + }, + { + "epoch": 0.63, + "learning_rate": 1.8358557656730925e-05, + "loss": 1.3034, + "step": 7462 + }, + { + "epoch": 0.63, + "learning_rate": 1.8358058642243276e-05, + "loss": 1.2483, + "step": 7463 + }, + { + "epoch": 0.63, + "learning_rate": 1.8357559558698666e-05, + "loss": 1.3848, + "step": 7464 + }, + { + "epoch": 0.63, + "learning_rate": 1.835706040610123e-05, + "loss": 1.2619, + "step": 7465 + }, + { + "epoch": 0.63, + "learning_rate": 1.835656118445508e-05, + "loss": 1.2918, + "step": 7466 + }, + { + "epoch": 0.63, + "learning_rate": 1.8356061893764353e-05, + "loss": 1.1518, + "step": 7467 + }, + { + "epoch": 0.63, + "learning_rate": 1.8355562534033165e-05, + "loss": 1.2519, + "step": 7468 + }, + { + "epoch": 0.63, + "learning_rate": 1.8355063105265648e-05, + "loss": 1.2735, + "step": 7469 + }, + { + "epoch": 0.63, + "learning_rate": 1.8354563607465922e-05, + "loss": 1.3011, + "step": 7470 + }, + { + "epoch": 0.63, + "learning_rate": 1.835406404063812e-05, + "loss": 1.395, + "step": 7471 + }, + { + "epoch": 0.63, + "learning_rate": 1.8353564404786364e-05, + "loss": 1.3319, + "step": 7472 + }, + { + "epoch": 0.63, + "learning_rate": 1.835306469991479e-05, + "loss": 1.3341, + "step": 7473 + }, + { + "epoch": 0.63, + "learning_rate": 1.8352564926027517e-05, + "loss": 1.2892, + "step": 7474 + }, + { + "epoch": 0.63, + "learning_rate": 1.8352065083128685e-05, + "loss": 1.2985, + "step": 7475 + }, + { + "epoch": 0.63, + "learning_rate": 1.8351565171222418e-05, + "loss": 1.3057, + "step": 7476 + }, + { + "epoch": 0.63, + "learning_rate": 1.835106519031284e-05, + "loss": 1.2112, + "step": 7477 + }, + { + "epoch": 0.63, + "learning_rate": 1.8350565140404094e-05, + "loss": 1.2903, + "step": 7478 + }, + { + "epoch": 0.63, + "learning_rate": 1.8350065021500305e-05, + "loss": 1.3147, + "step": 7479 + }, + { + "epoch": 0.63, + "learning_rate": 1.8349564833605608e-05, + "loss": 1.3095, + "step": 7480 + }, + { + "epoch": 0.63, + "learning_rate": 1.8349064576724132e-05, + "loss": 1.2789, + "step": 7481 + }, + { + "epoch": 0.63, + "learning_rate": 1.8348564250860016e-05, + "loss": 1.2828, + "step": 7482 + }, + { + "epoch": 0.63, + "learning_rate": 1.8348063856017385e-05, + "loss": 1.2759, + "step": 7483 + }, + { + "epoch": 0.63, + "learning_rate": 1.8347563392200384e-05, + "loss": 1.3503, + "step": 7484 + }, + { + "epoch": 0.63, + "learning_rate": 1.834706285941314e-05, + "loss": 1.3388, + "step": 7485 + }, + { + "epoch": 0.63, + "learning_rate": 1.834656225765979e-05, + "loss": 1.2325, + "step": 7486 + }, + { + "epoch": 0.63, + "learning_rate": 1.834606158694447e-05, + "loss": 1.3221, + "step": 7487 + }, + { + "epoch": 0.63, + "learning_rate": 1.8345560847271323e-05, + "loss": 1.2742, + "step": 7488 + }, + { + "epoch": 0.63, + "learning_rate": 1.834506003864448e-05, + "loss": 1.3902, + "step": 7489 + }, + { + "epoch": 0.63, + "learning_rate": 1.834455916106808e-05, + "loss": 1.3182, + "step": 7490 + }, + { + "epoch": 0.63, + "learning_rate": 1.834405821454626e-05, + "loss": 1.2917, + "step": 7491 + }, + { + "epoch": 0.63, + "learning_rate": 1.8343557199083162e-05, + "loss": 1.2635, + "step": 7492 + }, + { + "epoch": 0.63, + "learning_rate": 1.834305611468292e-05, + "loss": 1.2781, + "step": 7493 + }, + { + "epoch": 0.63, + "learning_rate": 1.8342554961349686e-05, + "loss": 1.3043, + "step": 7494 + }, + { + "epoch": 0.63, + "learning_rate": 1.8342053739087587e-05, + "loss": 1.1763, + "step": 7495 + }, + { + "epoch": 0.63, + "learning_rate": 1.8341552447900773e-05, + "loss": 1.3346, + "step": 7496 + }, + { + "epoch": 0.63, + "learning_rate": 1.834105108779338e-05, + "loss": 1.2357, + "step": 7497 + }, + { + "epoch": 0.63, + "learning_rate": 1.8340549658769557e-05, + "loss": 1.2947, + "step": 7498 + }, + { + "epoch": 0.63, + "learning_rate": 1.8340048160833442e-05, + "loss": 1.3368, + "step": 7499 + }, + { + "epoch": 0.63, + "learning_rate": 1.833954659398918e-05, + "loss": 1.2681, + "step": 7500 + }, + { + "epoch": 0.63, + "learning_rate": 1.8339044958240918e-05, + "loss": 1.2903, + "step": 7501 + }, + { + "epoch": 0.63, + "learning_rate": 1.8338543253592793e-05, + "loss": 1.4713, + "step": 7502 + }, + { + "epoch": 0.63, + "learning_rate": 1.8338041480048955e-05, + "loss": 1.2902, + "step": 7503 + }, + { + "epoch": 0.63, + "learning_rate": 1.8337539637613553e-05, + "loss": 1.3107, + "step": 7504 + }, + { + "epoch": 0.63, + "learning_rate": 1.8337037726290727e-05, + "loss": 1.3507, + "step": 7505 + }, + { + "epoch": 0.63, + "learning_rate": 1.833653574608463e-05, + "loss": 1.2541, + "step": 7506 + }, + { + "epoch": 0.63, + "learning_rate": 1.8336033696999406e-05, + "loss": 1.2697, + "step": 7507 + }, + { + "epoch": 0.63, + "learning_rate": 1.83355315790392e-05, + "loss": 1.3805, + "step": 7508 + }, + { + "epoch": 0.63, + "learning_rate": 1.833502939220817e-05, + "loss": 1.3603, + "step": 7509 + }, + { + "epoch": 0.63, + "learning_rate": 1.8334527136510458e-05, + "loss": 1.3906, + "step": 7510 + }, + { + "epoch": 0.63, + "learning_rate": 1.833402481195021e-05, + "loss": 1.3754, + "step": 7511 + }, + { + "epoch": 0.63, + "learning_rate": 1.833352241853159e-05, + "loss": 1.3367, + "step": 7512 + }, + { + "epoch": 0.63, + "learning_rate": 1.8333019956258734e-05, + "loss": 1.2475, + "step": 7513 + }, + { + "epoch": 0.63, + "learning_rate": 1.8332517425135805e-05, + "loss": 1.318, + "step": 7514 + }, + { + "epoch": 0.63, + "learning_rate": 1.833201482516695e-05, + "loss": 1.2474, + "step": 7515 + }, + { + "epoch": 0.63, + "learning_rate": 1.833151215635632e-05, + "loss": 1.2827, + "step": 7516 + }, + { + "epoch": 0.63, + "learning_rate": 1.833100941870807e-05, + "loss": 1.2937, + "step": 7517 + }, + { + "epoch": 0.63, + "learning_rate": 1.8330506612226354e-05, + "loss": 1.3206, + "step": 7518 + }, + { + "epoch": 0.63, + "learning_rate": 1.8330003736915327e-05, + "loss": 1.3467, + "step": 7519 + }, + { + "epoch": 0.63, + "learning_rate": 1.8329500792779143e-05, + "loss": 1.259, + "step": 7520 + }, + { + "epoch": 0.63, + "learning_rate": 1.832899777982196e-05, + "loss": 1.2337, + "step": 7521 + }, + { + "epoch": 0.63, + "learning_rate": 1.8328494698047925e-05, + "loss": 1.245, + "step": 7522 + }, + { + "epoch": 0.63, + "learning_rate": 1.8327991547461206e-05, + "loss": 1.2842, + "step": 7523 + }, + { + "epoch": 0.63, + "learning_rate": 1.8327488328065958e-05, + "loss": 1.3692, + "step": 7524 + }, + { + "epoch": 0.63, + "learning_rate": 1.8326985039866332e-05, + "loss": 1.249, + "step": 7525 + }, + { + "epoch": 0.63, + "learning_rate": 1.832648168286649e-05, + "loss": 1.197, + "step": 7526 + }, + { + "epoch": 0.63, + "learning_rate": 1.8325978257070597e-05, + "loss": 1.3121, + "step": 7527 + }, + { + "epoch": 0.63, + "learning_rate": 1.83254747624828e-05, + "loss": 1.3013, + "step": 7528 + }, + { + "epoch": 0.63, + "learning_rate": 1.832497119910727e-05, + "loss": 1.3222, + "step": 7529 + }, + { + "epoch": 0.63, + "learning_rate": 1.8324467566948164e-05, + "loss": 1.3197, + "step": 7530 + }, + { + "epoch": 0.63, + "learning_rate": 1.8323963866009645e-05, + "loss": 1.2311, + "step": 7531 + }, + { + "epoch": 0.63, + "learning_rate": 1.8323460096295868e-05, + "loss": 1.3308, + "step": 7532 + }, + { + "epoch": 0.63, + "learning_rate": 1.8322956257811e-05, + "loss": 1.2818, + "step": 7533 + }, + { + "epoch": 0.63, + "learning_rate": 1.8322452350559207e-05, + "loss": 1.3749, + "step": 7534 + }, + { + "epoch": 0.63, + "learning_rate": 1.8321948374544643e-05, + "loss": 1.2719, + "step": 7535 + }, + { + "epoch": 0.63, + "learning_rate": 1.8321444329771486e-05, + "loss": 1.3174, + "step": 7536 + }, + { + "epoch": 0.63, + "learning_rate": 1.8320940216243888e-05, + "loss": 1.2721, + "step": 7537 + }, + { + "epoch": 0.63, + "learning_rate": 1.832043603396602e-05, + "loss": 1.3265, + "step": 7538 + }, + { + "epoch": 0.63, + "learning_rate": 1.8319931782942045e-05, + "loss": 1.3539, + "step": 7539 + }, + { + "epoch": 0.63, + "learning_rate": 1.8319427463176135e-05, + "loss": 1.3058, + "step": 7540 + }, + { + "epoch": 0.63, + "learning_rate": 1.831892307467245e-05, + "loss": 1.2622, + "step": 7541 + }, + { + "epoch": 0.63, + "learning_rate": 1.831841861743516e-05, + "loss": 1.2706, + "step": 7542 + }, + { + "epoch": 0.64, + "learning_rate": 1.8317914091468433e-05, + "loss": 1.2452, + "step": 7543 + }, + { + "epoch": 0.64, + "learning_rate": 1.8317409496776438e-05, + "loss": 1.3348, + "step": 7544 + }, + { + "epoch": 0.64, + "learning_rate": 1.8316904833363345e-05, + "loss": 1.3481, + "step": 7545 + }, + { + "epoch": 0.64, + "learning_rate": 1.8316400101233322e-05, + "loss": 1.2822, + "step": 7546 + }, + { + "epoch": 0.64, + "learning_rate": 1.831589530039054e-05, + "loss": 1.3104, + "step": 7547 + }, + { + "epoch": 0.64, + "learning_rate": 1.831539043083917e-05, + "loss": 1.3037, + "step": 7548 + }, + { + "epoch": 0.64, + "learning_rate": 1.831488549258338e-05, + "loss": 1.2626, + "step": 7549 + }, + { + "epoch": 0.64, + "learning_rate": 1.8314380485627347e-05, + "loss": 1.2553, + "step": 7550 + }, + { + "epoch": 0.64, + "learning_rate": 1.8313875409975242e-05, + "loss": 1.2875, + "step": 7551 + }, + { + "epoch": 0.64, + "learning_rate": 1.8313370265631238e-05, + "loss": 1.2595, + "step": 7552 + }, + { + "epoch": 0.64, + "learning_rate": 1.8312865052599506e-05, + "loss": 1.3203, + "step": 7553 + }, + { + "epoch": 0.64, + "learning_rate": 1.8312359770884223e-05, + "loss": 1.2164, + "step": 7554 + }, + { + "epoch": 0.64, + "learning_rate": 1.8311854420489565e-05, + "loss": 1.2401, + "step": 7555 + }, + { + "epoch": 0.64, + "learning_rate": 1.8311349001419705e-05, + "loss": 1.2307, + "step": 7556 + }, + { + "epoch": 0.64, + "learning_rate": 1.8310843513678816e-05, + "loss": 1.3169, + "step": 7557 + }, + { + "epoch": 0.64, + "learning_rate": 1.8310337957271085e-05, + "loss": 1.2927, + "step": 7558 + }, + { + "epoch": 0.64, + "learning_rate": 1.8309832332200677e-05, + "loss": 1.3014, + "step": 7559 + }, + { + "epoch": 0.64, + "learning_rate": 1.8309326638471777e-05, + "loss": 1.2949, + "step": 7560 + }, + { + "epoch": 0.64, + "learning_rate": 1.830882087608856e-05, + "loss": 1.3305, + "step": 7561 + }, + { + "epoch": 0.64, + "learning_rate": 1.8308315045055204e-05, + "loss": 1.2627, + "step": 7562 + }, + { + "epoch": 0.64, + "learning_rate": 1.8307809145375894e-05, + "loss": 1.3072, + "step": 7563 + }, + { + "epoch": 0.64, + "learning_rate": 1.8307303177054804e-05, + "loss": 1.325, + "step": 7564 + }, + { + "epoch": 0.64, + "learning_rate": 1.8306797140096116e-05, + "loss": 1.3182, + "step": 7565 + }, + { + "epoch": 0.64, + "learning_rate": 1.8306291034504012e-05, + "loss": 1.2499, + "step": 7566 + }, + { + "epoch": 0.64, + "learning_rate": 1.8305784860282677e-05, + "loss": 1.3351, + "step": 7567 + }, + { + "epoch": 0.64, + "learning_rate": 1.830527861743628e-05, + "loss": 1.2707, + "step": 7568 + }, + { + "epoch": 0.64, + "learning_rate": 1.8304772305969024e-05, + "loss": 1.2866, + "step": 7569 + }, + { + "epoch": 0.64, + "learning_rate": 1.8304265925885075e-05, + "loss": 1.2975, + "step": 7570 + }, + { + "epoch": 0.64, + "learning_rate": 1.8303759477188622e-05, + "loss": 1.2464, + "step": 7571 + }, + { + "epoch": 0.64, + "learning_rate": 1.830325295988385e-05, + "loss": 1.3471, + "step": 7572 + }, + { + "epoch": 0.64, + "learning_rate": 1.8302746373974952e-05, + "loss": 1.2897, + "step": 7573 + }, + { + "epoch": 0.64, + "learning_rate": 1.8302239719466098e-05, + "loss": 1.3341, + "step": 7574 + }, + { + "epoch": 0.64, + "learning_rate": 1.8301732996361487e-05, + "loss": 1.3452, + "step": 7575 + }, + { + "epoch": 0.64, + "learning_rate": 1.83012262046653e-05, + "loss": 1.2655, + "step": 7576 + }, + { + "epoch": 0.64, + "learning_rate": 1.8300719344381727e-05, + "loss": 1.2889, + "step": 7577 + }, + { + "epoch": 0.64, + "learning_rate": 1.8300212415514954e-05, + "loss": 1.3231, + "step": 7578 + }, + { + "epoch": 0.64, + "learning_rate": 1.8299705418069168e-05, + "loss": 1.3464, + "step": 7579 + }, + { + "epoch": 0.64, + "learning_rate": 1.829919835204856e-05, + "loss": 1.3505, + "step": 7580 + }, + { + "epoch": 0.64, + "learning_rate": 1.829869121745732e-05, + "loss": 1.3243, + "step": 7581 + }, + { + "epoch": 0.64, + "learning_rate": 1.8298184014299634e-05, + "loss": 1.3931, + "step": 7582 + }, + { + "epoch": 0.64, + "learning_rate": 1.8297676742579697e-05, + "loss": 1.1948, + "step": 7583 + }, + { + "epoch": 0.64, + "learning_rate": 1.8297169402301703e-05, + "loss": 1.3367, + "step": 7584 + }, + { + "epoch": 0.64, + "learning_rate": 1.8296661993469837e-05, + "loss": 1.2896, + "step": 7585 + }, + { + "epoch": 0.64, + "learning_rate": 1.829615451608829e-05, + "loss": 1.3167, + "step": 7586 + }, + { + "epoch": 0.64, + "learning_rate": 1.8295646970161265e-05, + "loss": 1.2671, + "step": 7587 + }, + { + "epoch": 0.64, + "learning_rate": 1.829513935569295e-05, + "loss": 1.2766, + "step": 7588 + }, + { + "epoch": 0.64, + "learning_rate": 1.8294631672687536e-05, + "loss": 1.2502, + "step": 7589 + }, + { + "epoch": 0.64, + "learning_rate": 1.8294123921149226e-05, + "loss": 1.323, + "step": 7590 + }, + { + "epoch": 0.64, + "learning_rate": 1.8293616101082202e-05, + "loss": 1.2502, + "step": 7591 + }, + { + "epoch": 0.64, + "learning_rate": 1.8293108212490674e-05, + "loss": 1.3899, + "step": 7592 + }, + { + "epoch": 0.64, + "learning_rate": 1.8292600255378826e-05, + "loss": 1.2272, + "step": 7593 + }, + { + "epoch": 0.64, + "learning_rate": 1.8292092229750865e-05, + "loss": 1.2417, + "step": 7594 + }, + { + "epoch": 0.64, + "learning_rate": 1.8291584135610984e-05, + "loss": 1.3083, + "step": 7595 + }, + { + "epoch": 0.64, + "learning_rate": 1.829107597296338e-05, + "loss": 1.2376, + "step": 7596 + }, + { + "epoch": 0.64, + "learning_rate": 1.829056774181225e-05, + "loss": 1.3141, + "step": 7597 + }, + { + "epoch": 0.64, + "learning_rate": 1.82900594421618e-05, + "loss": 1.3406, + "step": 7598 + }, + { + "epoch": 0.64, + "learning_rate": 1.8289551074016224e-05, + "loss": 1.3518, + "step": 7599 + }, + { + "epoch": 0.64, + "learning_rate": 1.828904263737973e-05, + "loss": 1.3056, + "step": 7600 + }, + { + "epoch": 0.64, + "learning_rate": 1.8288534132256503e-05, + "loss": 1.3316, + "step": 7601 + }, + { + "epoch": 0.64, + "learning_rate": 1.8288025558650758e-05, + "loss": 1.4205, + "step": 7602 + }, + { + "epoch": 0.64, + "learning_rate": 1.82875169165667e-05, + "loss": 1.2506, + "step": 7603 + }, + { + "epoch": 0.64, + "learning_rate": 1.8287008206008518e-05, + "loss": 1.3339, + "step": 7604 + }, + { + "epoch": 0.64, + "learning_rate": 1.828649942698042e-05, + "loss": 1.1646, + "step": 7605 + }, + { + "epoch": 0.64, + "learning_rate": 1.8285990579486617e-05, + "loss": 1.251, + "step": 7606 + }, + { + "epoch": 0.64, + "learning_rate": 1.8285481663531307e-05, + "loss": 1.3626, + "step": 7607 + }, + { + "epoch": 0.64, + "learning_rate": 1.8284972679118693e-05, + "loss": 1.2958, + "step": 7608 + }, + { + "epoch": 0.64, + "learning_rate": 1.8284463626252982e-05, + "loss": 1.345, + "step": 7609 + }, + { + "epoch": 0.64, + "learning_rate": 1.8283954504938387e-05, + "loss": 1.2122, + "step": 7610 + }, + { + "epoch": 0.64, + "learning_rate": 1.8283445315179108e-05, + "loss": 1.3324, + "step": 7611 + }, + { + "epoch": 0.64, + "learning_rate": 1.828293605697935e-05, + "loss": 1.2453, + "step": 7612 + }, + { + "epoch": 0.64, + "learning_rate": 1.8282426730343324e-05, + "loss": 1.4051, + "step": 7613 + }, + { + "epoch": 0.64, + "learning_rate": 1.8281917335275236e-05, + "loss": 1.2596, + "step": 7614 + }, + { + "epoch": 0.64, + "learning_rate": 1.82814078717793e-05, + "loss": 1.3455, + "step": 7615 + }, + { + "epoch": 0.64, + "learning_rate": 1.828089833985972e-05, + "loss": 1.2542, + "step": 7616 + }, + { + "epoch": 0.64, + "learning_rate": 1.8280388739520708e-05, + "loss": 1.1563, + "step": 7617 + }, + { + "epoch": 0.64, + "learning_rate": 1.8279879070766472e-05, + "loss": 1.2934, + "step": 7618 + }, + { + "epoch": 0.64, + "learning_rate": 1.8279369333601224e-05, + "loss": 1.2905, + "step": 7619 + }, + { + "epoch": 0.64, + "learning_rate": 1.8278859528029182e-05, + "loss": 1.3858, + "step": 7620 + }, + { + "epoch": 0.64, + "learning_rate": 1.827834965405455e-05, + "loss": 1.268, + "step": 7621 + }, + { + "epoch": 0.64, + "learning_rate": 1.8277839711681544e-05, + "loss": 1.3148, + "step": 7622 + }, + { + "epoch": 0.64, + "learning_rate": 1.8277329700914378e-05, + "loss": 1.3037, + "step": 7623 + }, + { + "epoch": 0.64, + "learning_rate": 1.8276819621757264e-05, + "loss": 1.3773, + "step": 7624 + }, + { + "epoch": 0.64, + "learning_rate": 1.827630947421442e-05, + "loss": 1.2553, + "step": 7625 + }, + { + "epoch": 0.64, + "learning_rate": 1.8275799258290055e-05, + "loss": 1.2599, + "step": 7626 + }, + { + "epoch": 0.64, + "learning_rate": 1.8275288973988388e-05, + "loss": 1.313, + "step": 7627 + }, + { + "epoch": 0.64, + "learning_rate": 1.8274778621313635e-05, + "loss": 1.2677, + "step": 7628 + }, + { + "epoch": 0.64, + "learning_rate": 1.8274268200270015e-05, + "loss": 1.3807, + "step": 7629 + }, + { + "epoch": 0.64, + "learning_rate": 1.8273757710861742e-05, + "loss": 1.4119, + "step": 7630 + }, + { + "epoch": 0.64, + "learning_rate": 1.8273247153093035e-05, + "loss": 1.3118, + "step": 7631 + }, + { + "epoch": 0.64, + "learning_rate": 1.8272736526968115e-05, + "loss": 1.1809, + "step": 7632 + }, + { + "epoch": 0.64, + "learning_rate": 1.8272225832491193e-05, + "loss": 1.3233, + "step": 7633 + }, + { + "epoch": 0.64, + "learning_rate": 1.8271715069666496e-05, + "loss": 1.2712, + "step": 7634 + }, + { + "epoch": 0.64, + "learning_rate": 1.827120423849824e-05, + "loss": 1.3644, + "step": 7635 + }, + { + "epoch": 0.64, + "learning_rate": 1.8270693338990648e-05, + "loss": 1.2566, + "step": 7636 + }, + { + "epoch": 0.64, + "learning_rate": 1.8270182371147944e-05, + "loss": 1.3587, + "step": 7637 + }, + { + "epoch": 0.64, + "learning_rate": 1.8269671334974344e-05, + "loss": 1.2099, + "step": 7638 + }, + { + "epoch": 0.64, + "learning_rate": 1.8269160230474072e-05, + "loss": 1.2166, + "step": 7639 + }, + { + "epoch": 0.64, + "learning_rate": 1.8268649057651353e-05, + "loss": 1.4283, + "step": 7640 + }, + { + "epoch": 0.64, + "learning_rate": 1.826813781651041e-05, + "loss": 1.3086, + "step": 7641 + }, + { + "epoch": 0.64, + "learning_rate": 1.8267626507055466e-05, + "loss": 1.2418, + "step": 7642 + }, + { + "epoch": 0.64, + "learning_rate": 1.8267115129290744e-05, + "loss": 1.3117, + "step": 7643 + }, + { + "epoch": 0.64, + "learning_rate": 1.8266603683220472e-05, + "loss": 1.2818, + "step": 7644 + }, + { + "epoch": 0.64, + "learning_rate": 1.8266092168848875e-05, + "loss": 1.2716, + "step": 7645 + }, + { + "epoch": 0.64, + "learning_rate": 1.8265580586180178e-05, + "loss": 1.3516, + "step": 7646 + }, + { + "epoch": 0.64, + "learning_rate": 1.826506893521861e-05, + "loss": 1.4267, + "step": 7647 + }, + { + "epoch": 0.64, + "learning_rate": 1.8264557215968397e-05, + "loss": 1.3756, + "step": 7648 + }, + { + "epoch": 0.64, + "learning_rate": 1.8264045428433767e-05, + "loss": 1.3271, + "step": 7649 + }, + { + "epoch": 0.64, + "learning_rate": 1.826353357261895e-05, + "loss": 1.2788, + "step": 7650 + }, + { + "epoch": 0.64, + "learning_rate": 1.8263021648528173e-05, + "loss": 1.3341, + "step": 7651 + }, + { + "epoch": 0.64, + "learning_rate": 1.8262509656165665e-05, + "loss": 1.2185, + "step": 7652 + }, + { + "epoch": 0.64, + "learning_rate": 1.826199759553566e-05, + "loss": 1.3739, + "step": 7653 + }, + { + "epoch": 0.64, + "learning_rate": 1.8261485466642387e-05, + "loss": 1.3371, + "step": 7654 + }, + { + "epoch": 0.64, + "learning_rate": 1.8260973269490074e-05, + "loss": 1.1824, + "step": 7655 + }, + { + "epoch": 0.64, + "learning_rate": 1.826046100408296e-05, + "loss": 1.2692, + "step": 7656 + }, + { + "epoch": 0.64, + "learning_rate": 1.8259948670425272e-05, + "loss": 1.2965, + "step": 7657 + }, + { + "epoch": 0.64, + "learning_rate": 1.8259436268521244e-05, + "loss": 1.3115, + "step": 7658 + }, + { + "epoch": 0.64, + "learning_rate": 1.825892379837511e-05, + "loss": 1.2955, + "step": 7659 + }, + { + "epoch": 0.64, + "learning_rate": 1.8258411259991103e-05, + "loss": 1.3285, + "step": 7660 + }, + { + "epoch": 0.64, + "learning_rate": 1.825789865337346e-05, + "loss": 1.3131, + "step": 7661 + }, + { + "epoch": 0.65, + "learning_rate": 1.825738597852642e-05, + "loss": 1.2772, + "step": 7662 + }, + { + "epoch": 0.65, + "learning_rate": 1.825687323545421e-05, + "loss": 1.3011, + "step": 7663 + }, + { + "epoch": 0.65, + "learning_rate": 1.825636042416107e-05, + "loss": 1.3095, + "step": 7664 + }, + { + "epoch": 0.65, + "learning_rate": 1.825584754465124e-05, + "loss": 1.2868, + "step": 7665 + }, + { + "epoch": 0.65, + "learning_rate": 1.8255334596928954e-05, + "loss": 1.3488, + "step": 7666 + }, + { + "epoch": 0.65, + "learning_rate": 1.8254821580998453e-05, + "loss": 1.3754, + "step": 7667 + }, + { + "epoch": 0.65, + "learning_rate": 1.8254308496863975e-05, + "loss": 1.2335, + "step": 7668 + }, + { + "epoch": 0.65, + "learning_rate": 1.8253795344529758e-05, + "loss": 1.2822, + "step": 7669 + }, + { + "epoch": 0.65, + "learning_rate": 1.825328212400004e-05, + "loss": 1.2694, + "step": 7670 + }, + { + "epoch": 0.65, + "learning_rate": 1.8252768835279066e-05, + "loss": 1.2383, + "step": 7671 + }, + { + "epoch": 0.65, + "learning_rate": 1.825225547837107e-05, + "loss": 1.3866, + "step": 7672 + }, + { + "epoch": 0.65, + "learning_rate": 1.8251742053280302e-05, + "loss": 1.3608, + "step": 7673 + }, + { + "epoch": 0.65, + "learning_rate": 1.8251228560011004e-05, + "loss": 1.3636, + "step": 7674 + }, + { + "epoch": 0.65, + "learning_rate": 1.825071499856741e-05, + "loss": 1.2791, + "step": 7675 + }, + { + "epoch": 0.65, + "learning_rate": 1.825020136895377e-05, + "loss": 1.3156, + "step": 7676 + }, + { + "epoch": 0.65, + "learning_rate": 1.8249687671174325e-05, + "loss": 1.3024, + "step": 7677 + }, + { + "epoch": 0.65, + "learning_rate": 1.824917390523332e-05, + "loss": 1.2129, + "step": 7678 + }, + { + "epoch": 0.65, + "learning_rate": 1.8248660071134997e-05, + "loss": 1.3324, + "step": 7679 + }, + { + "epoch": 0.65, + "learning_rate": 1.824814616888361e-05, + "loss": 1.3004, + "step": 7680 + }, + { + "epoch": 0.65, + "learning_rate": 1.82476321984834e-05, + "loss": 1.3404, + "step": 7681 + }, + { + "epoch": 0.65, + "learning_rate": 1.8247118159938607e-05, + "loss": 1.3351, + "step": 7682 + }, + { + "epoch": 0.65, + "learning_rate": 1.824660405325349e-05, + "loss": 1.3985, + "step": 7683 + }, + { + "epoch": 0.65, + "learning_rate": 1.824608987843229e-05, + "loss": 1.1933, + "step": 7684 + }, + { + "epoch": 0.65, + "learning_rate": 1.8245575635479254e-05, + "loss": 1.3947, + "step": 7685 + }, + { + "epoch": 0.65, + "learning_rate": 1.8245061324398636e-05, + "loss": 1.3922, + "step": 7686 + }, + { + "epoch": 0.65, + "learning_rate": 1.824454694519468e-05, + "loss": 1.2665, + "step": 7687 + }, + { + "epoch": 0.65, + "learning_rate": 1.824403249787164e-05, + "loss": 1.3232, + "step": 7688 + }, + { + "epoch": 0.65, + "learning_rate": 1.8243517982433765e-05, + "loss": 1.2438, + "step": 7689 + }, + { + "epoch": 0.65, + "learning_rate": 1.8243003398885307e-05, + "loss": 1.2636, + "step": 7690 + }, + { + "epoch": 0.65, + "learning_rate": 1.8242488747230515e-05, + "loss": 1.323, + "step": 7691 + }, + { + "epoch": 0.65, + "learning_rate": 1.8241974027473644e-05, + "loss": 1.2637, + "step": 7692 + }, + { + "epoch": 0.65, + "learning_rate": 1.8241459239618945e-05, + "loss": 1.298, + "step": 7693 + }, + { + "epoch": 0.65, + "learning_rate": 1.8240944383670672e-05, + "loss": 1.3483, + "step": 7694 + }, + { + "epoch": 0.65, + "learning_rate": 1.8240429459633083e-05, + "loss": 1.4919, + "step": 7695 + }, + { + "epoch": 0.65, + "learning_rate": 1.8239914467510426e-05, + "loss": 1.2854, + "step": 7696 + }, + { + "epoch": 0.65, + "learning_rate": 1.8239399407306954e-05, + "loss": 1.3111, + "step": 7697 + }, + { + "epoch": 0.65, + "learning_rate": 1.8238884279026932e-05, + "loss": 1.2443, + "step": 7698 + }, + { + "epoch": 0.65, + "learning_rate": 1.8238369082674613e-05, + "loss": 1.2884, + "step": 7699 + }, + { + "epoch": 0.65, + "learning_rate": 1.823785381825425e-05, + "loss": 1.2996, + "step": 7700 + }, + { + "epoch": 0.65, + "learning_rate": 1.8237338485770098e-05, + "loss": 1.282, + "step": 7701 + }, + { + "epoch": 0.65, + "learning_rate": 1.8236823085226427e-05, + "loss": 1.3525, + "step": 7702 + }, + { + "epoch": 0.65, + "learning_rate": 1.8236307616627483e-05, + "loss": 1.3341, + "step": 7703 + }, + { + "epoch": 0.65, + "learning_rate": 1.823579207997753e-05, + "loss": 1.2775, + "step": 7704 + }, + { + "epoch": 0.65, + "learning_rate": 1.8235276475280824e-05, + "loss": 1.2661, + "step": 7705 + }, + { + "epoch": 0.65, + "learning_rate": 1.8234760802541635e-05, + "loss": 1.3338, + "step": 7706 + }, + { + "epoch": 0.65, + "learning_rate": 1.823424506176421e-05, + "loss": 1.2321, + "step": 7707 + }, + { + "epoch": 0.65, + "learning_rate": 1.823372925295282e-05, + "loss": 1.2257, + "step": 7708 + }, + { + "epoch": 0.65, + "learning_rate": 1.8233213376111725e-05, + "loss": 1.3475, + "step": 7709 + }, + { + "epoch": 0.65, + "learning_rate": 1.8232697431245183e-05, + "loss": 1.4028, + "step": 7710 + }, + { + "epoch": 0.65, + "learning_rate": 1.8232181418357463e-05, + "loss": 1.2327, + "step": 7711 + }, + { + "epoch": 0.65, + "learning_rate": 1.8231665337452825e-05, + "loss": 1.3562, + "step": 7712 + }, + { + "epoch": 0.65, + "learning_rate": 1.8231149188535534e-05, + "loss": 1.2241, + "step": 7713 + }, + { + "epoch": 0.65, + "learning_rate": 1.8230632971609853e-05, + "loss": 1.3217, + "step": 7714 + }, + { + "epoch": 0.65, + "learning_rate": 1.8230116686680046e-05, + "loss": 1.2418, + "step": 7715 + }, + { + "epoch": 0.65, + "learning_rate": 1.8229600333750384e-05, + "loss": 1.3079, + "step": 7716 + }, + { + "epoch": 0.65, + "learning_rate": 1.822908391282513e-05, + "loss": 1.3469, + "step": 7717 + }, + { + "epoch": 0.65, + "learning_rate": 1.8228567423908548e-05, + "loss": 1.317, + "step": 7718 + }, + { + "epoch": 0.65, + "learning_rate": 1.8228050867004913e-05, + "loss": 1.3165, + "step": 7719 + }, + { + "epoch": 0.65, + "learning_rate": 1.8227534242118484e-05, + "loss": 1.3441, + "step": 7720 + }, + { + "epoch": 0.65, + "learning_rate": 1.8227017549253537e-05, + "loss": 1.2872, + "step": 7721 + }, + { + "epoch": 0.65, + "learning_rate": 1.822650078841434e-05, + "loss": 1.256, + "step": 7722 + }, + { + "epoch": 0.65, + "learning_rate": 1.822598395960516e-05, + "loss": 1.3101, + "step": 7723 + }, + { + "epoch": 0.65, + "learning_rate": 1.822546706283026e-05, + "loss": 1.3695, + "step": 7724 + }, + { + "epoch": 0.65, + "learning_rate": 1.822495009809393e-05, + "loss": 1.4289, + "step": 7725 + }, + { + "epoch": 0.65, + "learning_rate": 1.822443306540042e-05, + "loss": 1.337, + "step": 7726 + }, + { + "epoch": 0.65, + "learning_rate": 1.8223915964754018e-05, + "loss": 1.3261, + "step": 7727 + }, + { + "epoch": 0.65, + "learning_rate": 1.822339879615899e-05, + "loss": 1.3124, + "step": 7728 + }, + { + "epoch": 0.65, + "learning_rate": 1.8222881559619607e-05, + "loss": 1.3126, + "step": 7729 + }, + { + "epoch": 0.65, + "learning_rate": 1.8222364255140146e-05, + "loss": 1.3186, + "step": 7730 + }, + { + "epoch": 0.65, + "learning_rate": 1.822184688272488e-05, + "loss": 1.3566, + "step": 7731 + }, + { + "epoch": 0.65, + "learning_rate": 1.8221329442378085e-05, + "loss": 1.2897, + "step": 7732 + }, + { + "epoch": 0.65, + "learning_rate": 1.822081193410403e-05, + "loss": 1.2464, + "step": 7733 + }, + { + "epoch": 0.65, + "learning_rate": 1.8220294357907e-05, + "loss": 1.423, + "step": 7734 + }, + { + "epoch": 0.65, + "learning_rate": 1.8219776713791265e-05, + "loss": 1.2405, + "step": 7735 + }, + { + "epoch": 0.65, + "learning_rate": 1.8219259001761108e-05, + "loss": 1.2477, + "step": 7736 + }, + { + "epoch": 0.65, + "learning_rate": 1.8218741221820803e-05, + "loss": 1.1992, + "step": 7737 + }, + { + "epoch": 0.65, + "learning_rate": 1.8218223373974622e-05, + "loss": 1.3044, + "step": 7738 + }, + { + "epoch": 0.65, + "learning_rate": 1.8217705458226854e-05, + "loss": 1.2405, + "step": 7739 + }, + { + "epoch": 0.65, + "learning_rate": 1.821718747458177e-05, + "loss": 1.4287, + "step": 7740 + }, + { + "epoch": 0.65, + "learning_rate": 1.8216669423043656e-05, + "loss": 1.2858, + "step": 7741 + }, + { + "epoch": 0.65, + "learning_rate": 1.8216151303616788e-05, + "loss": 1.3244, + "step": 7742 + }, + { + "epoch": 0.65, + "learning_rate": 1.821563311630545e-05, + "loss": 1.2817, + "step": 7743 + }, + { + "epoch": 0.65, + "learning_rate": 1.8215114861113924e-05, + "loss": 1.3176, + "step": 7744 + }, + { + "epoch": 0.65, + "learning_rate": 1.8214596538046484e-05, + "loss": 1.2588, + "step": 7745 + }, + { + "epoch": 0.65, + "learning_rate": 1.8214078147107423e-05, + "loss": 1.2874, + "step": 7746 + }, + { + "epoch": 0.65, + "learning_rate": 1.821355968830102e-05, + "loss": 1.2598, + "step": 7747 + }, + { + "epoch": 0.65, + "learning_rate": 1.8213041161631554e-05, + "loss": 1.2253, + "step": 7748 + }, + { + "epoch": 0.65, + "learning_rate": 1.8212522567103313e-05, + "loss": 1.2764, + "step": 7749 + }, + { + "epoch": 0.65, + "learning_rate": 1.8212003904720586e-05, + "loss": 1.3094, + "step": 7750 + }, + { + "epoch": 0.65, + "learning_rate": 1.821148517448765e-05, + "loss": 1.2785, + "step": 7751 + }, + { + "epoch": 0.65, + "learning_rate": 1.82109663764088e-05, + "loss": 1.3033, + "step": 7752 + }, + { + "epoch": 0.65, + "learning_rate": 1.8210447510488315e-05, + "loss": 1.2494, + "step": 7753 + }, + { + "epoch": 0.65, + "learning_rate": 1.8209928576730487e-05, + "loss": 1.1392, + "step": 7754 + }, + { + "epoch": 0.65, + "learning_rate": 1.82094095751396e-05, + "loss": 1.3364, + "step": 7755 + }, + { + "epoch": 0.65, + "learning_rate": 1.8208890505719948e-05, + "loss": 1.2778, + "step": 7756 + }, + { + "epoch": 0.65, + "learning_rate": 1.820837136847581e-05, + "loss": 1.3316, + "step": 7757 + }, + { + "epoch": 0.65, + "learning_rate": 1.8207852163411484e-05, + "loss": 1.2995, + "step": 7758 + }, + { + "epoch": 0.65, + "learning_rate": 1.8207332890531255e-05, + "loss": 1.3286, + "step": 7759 + }, + { + "epoch": 0.65, + "learning_rate": 1.8206813549839415e-05, + "loss": 1.3864, + "step": 7760 + }, + { + "epoch": 0.65, + "learning_rate": 1.8206294141340257e-05, + "loss": 1.4003, + "step": 7761 + }, + { + "epoch": 0.65, + "learning_rate": 1.820577466503807e-05, + "loss": 1.2769, + "step": 7762 + }, + { + "epoch": 0.65, + "learning_rate": 1.8205255120937145e-05, + "loss": 1.3336, + "step": 7763 + }, + { + "epoch": 0.65, + "learning_rate": 1.8204735509041774e-05, + "loss": 1.2692, + "step": 7764 + }, + { + "epoch": 0.65, + "learning_rate": 1.8204215829356256e-05, + "loss": 1.3008, + "step": 7765 + }, + { + "epoch": 0.65, + "learning_rate": 1.820369608188488e-05, + "loss": 1.2601, + "step": 7766 + }, + { + "epoch": 0.65, + "learning_rate": 1.820317626663194e-05, + "loss": 1.3326, + "step": 7767 + }, + { + "epoch": 0.65, + "learning_rate": 1.8202656383601737e-05, + "loss": 1.2673, + "step": 7768 + }, + { + "epoch": 0.65, + "learning_rate": 1.8202136432798557e-05, + "loss": 1.1995, + "step": 7769 + }, + { + "epoch": 0.65, + "learning_rate": 1.8201616414226703e-05, + "loss": 1.2801, + "step": 7770 + }, + { + "epoch": 0.65, + "learning_rate": 1.8201096327890472e-05, + "loss": 1.4435, + "step": 7771 + }, + { + "epoch": 0.65, + "learning_rate": 1.820057617379415e-05, + "loss": 1.2539, + "step": 7772 + }, + { + "epoch": 0.65, + "learning_rate": 1.8200055951942053e-05, + "loss": 1.3314, + "step": 7773 + }, + { + "epoch": 0.65, + "learning_rate": 1.8199535662338463e-05, + "loss": 1.3032, + "step": 7774 + }, + { + "epoch": 0.65, + "learning_rate": 1.8199015304987688e-05, + "loss": 1.2605, + "step": 7775 + }, + { + "epoch": 0.65, + "learning_rate": 1.8198494879894023e-05, + "loss": 1.2219, + "step": 7776 + }, + { + "epoch": 0.65, + "learning_rate": 1.8197974387061768e-05, + "loss": 1.244, + "step": 7777 + }, + { + "epoch": 0.65, + "learning_rate": 1.819745382649523e-05, + "loss": 1.3193, + "step": 7778 + }, + { + "epoch": 0.65, + "learning_rate": 1.81969331981987e-05, + "loss": 1.2427, + "step": 7779 + }, + { + "epoch": 0.66, + "learning_rate": 1.8196412502176487e-05, + "loss": 1.367, + "step": 7780 + }, + { + "epoch": 0.66, + "learning_rate": 1.819589173843289e-05, + "loss": 1.2316, + "step": 7781 + }, + { + "epoch": 0.66, + "learning_rate": 1.8195370906972214e-05, + "loss": 1.315, + "step": 7782 + }, + { + "epoch": 0.66, + "learning_rate": 1.8194850007798755e-05, + "loss": 1.2365, + "step": 7783 + }, + { + "epoch": 0.66, + "learning_rate": 1.819432904091683e-05, + "loss": 1.316, + "step": 7784 + }, + { + "epoch": 0.66, + "learning_rate": 1.819380800633073e-05, + "loss": 1.3383, + "step": 7785 + }, + { + "epoch": 0.66, + "learning_rate": 1.819328690404477e-05, + "loss": 1.2694, + "step": 7786 + }, + { + "epoch": 0.66, + "learning_rate": 1.8192765734063252e-05, + "loss": 1.2885, + "step": 7787 + }, + { + "epoch": 0.66, + "learning_rate": 1.8192244496390478e-05, + "loss": 1.291, + "step": 7788 + }, + { + "epoch": 0.66, + "learning_rate": 1.8191723191030758e-05, + "loss": 1.3614, + "step": 7789 + }, + { + "epoch": 0.66, + "learning_rate": 1.8191201817988403e-05, + "loss": 1.2716, + "step": 7790 + }, + { + "epoch": 0.66, + "learning_rate": 1.8190680377267713e-05, + "loss": 1.281, + "step": 7791 + }, + { + "epoch": 0.66, + "learning_rate": 1.8190158868873003e-05, + "loss": 1.3762, + "step": 7792 + }, + { + "epoch": 0.66, + "learning_rate": 1.818963729280858e-05, + "loss": 1.3336, + "step": 7793 + }, + { + "epoch": 0.66, + "learning_rate": 1.818911564907875e-05, + "loss": 1.2344, + "step": 7794 + }, + { + "epoch": 0.66, + "learning_rate": 1.818859393768783e-05, + "loss": 1.3794, + "step": 7795 + }, + { + "epoch": 0.66, + "learning_rate": 1.8188072158640123e-05, + "loss": 1.2766, + "step": 7796 + }, + { + "epoch": 0.66, + "learning_rate": 1.8187550311939942e-05, + "loss": 1.2631, + "step": 7797 + }, + { + "epoch": 0.66, + "learning_rate": 1.81870283975916e-05, + "loss": 1.2035, + "step": 7798 + }, + { + "epoch": 0.66, + "learning_rate": 1.8186506415599412e-05, + "loss": 1.3553, + "step": 7799 + }, + { + "epoch": 0.66, + "learning_rate": 1.8185984365967688e-05, + "loss": 1.4464, + "step": 7800 + }, + { + "epoch": 0.66, + "learning_rate": 1.818546224870074e-05, + "loss": 1.2847, + "step": 7801 + }, + { + "epoch": 0.66, + "learning_rate": 1.818494006380288e-05, + "loss": 1.2752, + "step": 7802 + }, + { + "epoch": 0.66, + "learning_rate": 1.818441781127843e-05, + "loss": 1.3051, + "step": 7803 + }, + { + "epoch": 0.66, + "learning_rate": 1.81838954911317e-05, + "loss": 1.4154, + "step": 7804 + }, + { + "epoch": 0.66, + "learning_rate": 1.8183373103367007e-05, + "loss": 1.2645, + "step": 7805 + }, + { + "epoch": 0.66, + "learning_rate": 1.8182850647988663e-05, + "loss": 1.2505, + "step": 7806 + }, + { + "epoch": 0.66, + "learning_rate": 1.818232812500099e-05, + "loss": 1.2906, + "step": 7807 + }, + { + "epoch": 0.66, + "learning_rate": 1.8181805534408302e-05, + "loss": 1.362, + "step": 7808 + }, + { + "epoch": 0.66, + "learning_rate": 1.8181282876214922e-05, + "loss": 1.2603, + "step": 7809 + }, + { + "epoch": 0.66, + "learning_rate": 1.818076015042516e-05, + "loss": 1.3779, + "step": 7810 + }, + { + "epoch": 0.66, + "learning_rate": 1.8180237357043344e-05, + "loss": 1.2876, + "step": 7811 + }, + { + "epoch": 0.66, + "learning_rate": 1.8179714496073787e-05, + "loss": 1.318, + "step": 7812 + }, + { + "epoch": 0.66, + "learning_rate": 1.8179191567520812e-05, + "loss": 1.2512, + "step": 7813 + }, + { + "epoch": 0.66, + "learning_rate": 1.8178668571388736e-05, + "loss": 1.2562, + "step": 7814 + }, + { + "epoch": 0.66, + "learning_rate": 1.8178145507681883e-05, + "loss": 1.257, + "step": 7815 + }, + { + "epoch": 0.66, + "learning_rate": 1.817762237640458e-05, + "loss": 1.2833, + "step": 7816 + }, + { + "epoch": 0.66, + "learning_rate": 1.8177099177561135e-05, + "loss": 1.1863, + "step": 7817 + }, + { + "epoch": 0.66, + "learning_rate": 1.8176575911155886e-05, + "loss": 1.2607, + "step": 7818 + }, + { + "epoch": 0.66, + "learning_rate": 1.8176052577193145e-05, + "loss": 1.2732, + "step": 7819 + }, + { + "epoch": 0.66, + "learning_rate": 1.8175529175677244e-05, + "loss": 1.2441, + "step": 7820 + }, + { + "epoch": 0.66, + "learning_rate": 1.8175005706612505e-05, + "loss": 1.2716, + "step": 7821 + }, + { + "epoch": 0.66, + "learning_rate": 1.817448217000325e-05, + "loss": 1.3422, + "step": 7822 + }, + { + "epoch": 0.66, + "learning_rate": 1.8173958565853812e-05, + "loss": 1.2678, + "step": 7823 + }, + { + "epoch": 0.66, + "learning_rate": 1.817343489416851e-05, + "loss": 1.2245, + "step": 7824 + }, + { + "epoch": 0.66, + "learning_rate": 1.8172911154951667e-05, + "loss": 1.3007, + "step": 7825 + }, + { + "epoch": 0.66, + "learning_rate": 1.8172387348207625e-05, + "loss": 1.2017, + "step": 7826 + }, + { + "epoch": 0.66, + "learning_rate": 1.81718634739407e-05, + "loss": 1.3014, + "step": 7827 + }, + { + "epoch": 0.66, + "learning_rate": 1.8171339532155222e-05, + "loss": 1.4176, + "step": 7828 + }, + { + "epoch": 0.66, + "learning_rate": 1.8170815522855525e-05, + "loss": 1.2752, + "step": 7829 + }, + { + "epoch": 0.66, + "learning_rate": 1.8170291446045934e-05, + "loss": 1.3192, + "step": 7830 + }, + { + "epoch": 0.66, + "learning_rate": 1.8169767301730778e-05, + "loss": 1.297, + "step": 7831 + }, + { + "epoch": 0.66, + "learning_rate": 1.8169243089914394e-05, + "loss": 1.3402, + "step": 7832 + }, + { + "epoch": 0.66, + "learning_rate": 1.81687188106011e-05, + "loss": 1.4027, + "step": 7833 + }, + { + "epoch": 0.66, + "learning_rate": 1.816819446379525e-05, + "loss": 1.2777, + "step": 7834 + }, + { + "epoch": 0.66, + "learning_rate": 1.8167670049501154e-05, + "loss": 1.2747, + "step": 7835 + }, + { + "epoch": 0.66, + "learning_rate": 1.8167145567723156e-05, + "loss": 1.2627, + "step": 7836 + }, + { + "epoch": 0.66, + "learning_rate": 1.816662101846559e-05, + "loss": 1.3723, + "step": 7837 + }, + { + "epoch": 0.66, + "learning_rate": 1.8166096401732783e-05, + "loss": 1.2074, + "step": 7838 + }, + { + "epoch": 0.66, + "learning_rate": 1.816557171752908e-05, + "loss": 1.2706, + "step": 7839 + }, + { + "epoch": 0.66, + "learning_rate": 1.8165046965858804e-05, + "loss": 1.435, + "step": 7840 + }, + { + "epoch": 0.66, + "learning_rate": 1.81645221467263e-05, + "loss": 1.2918, + "step": 7841 + }, + { + "epoch": 0.66, + "learning_rate": 1.8163997260135903e-05, + "loss": 1.3174, + "step": 7842 + }, + { + "epoch": 0.66, + "learning_rate": 1.8163472306091944e-05, + "loss": 1.2479, + "step": 7843 + }, + { + "epoch": 0.66, + "learning_rate": 1.8162947284598768e-05, + "loss": 1.3189, + "step": 7844 + }, + { + "epoch": 0.66, + "learning_rate": 1.8162422195660702e-05, + "loss": 1.2656, + "step": 7845 + }, + { + "epoch": 0.66, + "learning_rate": 1.81618970392821e-05, + "loss": 1.2561, + "step": 7846 + }, + { + "epoch": 0.66, + "learning_rate": 1.8161371815467288e-05, + "loss": 1.3689, + "step": 7847 + }, + { + "epoch": 0.66, + "learning_rate": 1.816084652422061e-05, + "loss": 1.2114, + "step": 7848 + }, + { + "epoch": 0.66, + "learning_rate": 1.8160321165546407e-05, + "loss": 1.3288, + "step": 7849 + }, + { + "epoch": 0.66, + "learning_rate": 1.815979573944902e-05, + "loss": 1.2853, + "step": 7850 + }, + { + "epoch": 0.66, + "learning_rate": 1.8159270245932786e-05, + "loss": 1.3236, + "step": 7851 + }, + { + "epoch": 0.66, + "learning_rate": 1.8158744685002053e-05, + "loss": 1.3518, + "step": 7852 + }, + { + "epoch": 0.66, + "learning_rate": 1.815821905666116e-05, + "loss": 1.3944, + "step": 7853 + }, + { + "epoch": 0.66, + "learning_rate": 1.815769336091445e-05, + "loss": 1.2625, + "step": 7854 + }, + { + "epoch": 0.66, + "learning_rate": 1.8157167597766265e-05, + "loss": 1.3331, + "step": 7855 + }, + { + "epoch": 0.66, + "learning_rate": 1.8156641767220953e-05, + "loss": 1.2873, + "step": 7856 + }, + { + "epoch": 0.66, + "learning_rate": 1.8156115869282854e-05, + "loss": 1.3473, + "step": 7857 + }, + { + "epoch": 0.66, + "learning_rate": 1.8155589903956315e-05, + "loss": 1.2514, + "step": 7858 + }, + { + "epoch": 0.66, + "learning_rate": 1.8155063871245686e-05, + "loss": 1.2295, + "step": 7859 + }, + { + "epoch": 0.66, + "learning_rate": 1.8154537771155304e-05, + "loss": 1.3493, + "step": 7860 + }, + { + "epoch": 0.66, + "learning_rate": 1.8154011603689525e-05, + "loss": 1.2961, + "step": 7861 + }, + { + "epoch": 0.66, + "learning_rate": 1.8153485368852695e-05, + "loss": 1.3465, + "step": 7862 + }, + { + "epoch": 0.66, + "learning_rate": 1.8152959066649155e-05, + "loss": 1.2523, + "step": 7863 + }, + { + "epoch": 0.66, + "learning_rate": 1.8152432697083257e-05, + "loss": 1.3022, + "step": 7864 + }, + { + "epoch": 0.66, + "learning_rate": 1.8151906260159354e-05, + "loss": 1.2928, + "step": 7865 + }, + { + "epoch": 0.66, + "learning_rate": 1.8151379755881793e-05, + "loss": 1.2921, + "step": 7866 + }, + { + "epoch": 0.66, + "learning_rate": 1.815085318425492e-05, + "loss": 1.2003, + "step": 7867 + }, + { + "epoch": 0.66, + "learning_rate": 1.8150326545283093e-05, + "loss": 1.3089, + "step": 7868 + }, + { + "epoch": 0.66, + "learning_rate": 1.8149799838970655e-05, + "loss": 1.2964, + "step": 7869 + }, + { + "epoch": 0.66, + "learning_rate": 1.8149273065321967e-05, + "loss": 1.254, + "step": 7870 + }, + { + "epoch": 0.66, + "learning_rate": 1.8148746224341376e-05, + "loss": 1.3351, + "step": 7871 + }, + { + "epoch": 0.66, + "learning_rate": 1.8148219316033233e-05, + "loss": 1.432, + "step": 7872 + }, + { + "epoch": 0.66, + "learning_rate": 1.8147692340401898e-05, + "loss": 1.3529, + "step": 7873 + }, + { + "epoch": 0.66, + "learning_rate": 1.8147165297451722e-05, + "loss": 1.2105, + "step": 7874 + }, + { + "epoch": 0.66, + "learning_rate": 1.8146638187187052e-05, + "loss": 1.3313, + "step": 7875 + }, + { + "epoch": 0.66, + "learning_rate": 1.8146111009612258e-05, + "loss": 1.3264, + "step": 7876 + }, + { + "epoch": 0.66, + "learning_rate": 1.8145583764731684e-05, + "loss": 1.2966, + "step": 7877 + }, + { + "epoch": 0.66, + "learning_rate": 1.814505645254969e-05, + "loss": 1.4234, + "step": 7878 + }, + { + "epoch": 0.66, + "learning_rate": 1.814452907307063e-05, + "loss": 1.3676, + "step": 7879 + }, + { + "epoch": 0.66, + "learning_rate": 1.8144001626298872e-05, + "loss": 1.2771, + "step": 7880 + }, + { + "epoch": 0.66, + "learning_rate": 1.8143474112238763e-05, + "loss": 1.3277, + "step": 7881 + }, + { + "epoch": 0.66, + "learning_rate": 1.8142946530894662e-05, + "loss": 1.3531, + "step": 7882 + }, + { + "epoch": 0.66, + "learning_rate": 1.8142418882270928e-05, + "loss": 1.3687, + "step": 7883 + }, + { + "epoch": 0.66, + "learning_rate": 1.8141891166371926e-05, + "loss": 1.3585, + "step": 7884 + }, + { + "epoch": 0.66, + "learning_rate": 1.8141363383202016e-05, + "loss": 1.2949, + "step": 7885 + }, + { + "epoch": 0.66, + "learning_rate": 1.8140835532765557e-05, + "loss": 1.342, + "step": 7886 + }, + { + "epoch": 0.66, + "learning_rate": 1.814030761506691e-05, + "loss": 1.2347, + "step": 7887 + }, + { + "epoch": 0.66, + "learning_rate": 1.813977963011043e-05, + "loss": 1.3169, + "step": 7888 + }, + { + "epoch": 0.66, + "learning_rate": 1.813925157790049e-05, + "loss": 1.2681, + "step": 7889 + }, + { + "epoch": 0.66, + "learning_rate": 1.8138723458441447e-05, + "loss": 1.266, + "step": 7890 + }, + { + "epoch": 0.66, + "learning_rate": 1.8138195271737668e-05, + "loss": 1.3366, + "step": 7891 + }, + { + "epoch": 0.66, + "learning_rate": 1.8137667017793513e-05, + "loss": 1.3132, + "step": 7892 + }, + { + "epoch": 0.66, + "learning_rate": 1.813713869661335e-05, + "loss": 1.2101, + "step": 7893 + }, + { + "epoch": 0.66, + "learning_rate": 1.8136610308201543e-05, + "loss": 1.3329, + "step": 7894 + }, + { + "epoch": 0.66, + "learning_rate": 1.8136081852562455e-05, + "loss": 1.2779, + "step": 7895 + }, + { + "epoch": 0.66, + "learning_rate": 1.813555332970046e-05, + "loss": 1.354, + "step": 7896 + }, + { + "epoch": 0.66, + "learning_rate": 1.8135024739619914e-05, + "loss": 1.356, + "step": 7897 + }, + { + "epoch": 0.66, + "learning_rate": 1.8134496082325192e-05, + "loss": 1.2533, + "step": 7898 + }, + { + "epoch": 0.67, + "learning_rate": 1.8133967357820662e-05, + "loss": 1.2779, + "step": 7899 + }, + { + "epoch": 0.67, + "learning_rate": 1.813343856611069e-05, + "loss": 1.2749, + "step": 7900 + }, + { + "epoch": 0.67, + "learning_rate": 1.8132909707199644e-05, + "loss": 1.3253, + "step": 7901 + }, + { + "epoch": 0.67, + "learning_rate": 1.8132380781091894e-05, + "loss": 1.2486, + "step": 7902 + }, + { + "epoch": 0.67, + "learning_rate": 1.8131851787791813e-05, + "loss": 1.3073, + "step": 7903 + }, + { + "epoch": 0.67, + "learning_rate": 1.813132272730377e-05, + "loss": 1.3084, + "step": 7904 + }, + { + "epoch": 0.67, + "learning_rate": 1.8130793599632137e-05, + "loss": 1.2917, + "step": 7905 + }, + { + "epoch": 0.67, + "learning_rate": 1.8130264404781287e-05, + "loss": 1.2309, + "step": 7906 + }, + { + "epoch": 0.67, + "learning_rate": 1.8129735142755588e-05, + "loss": 1.2286, + "step": 7907 + }, + { + "epoch": 0.67, + "learning_rate": 1.8129205813559414e-05, + "loss": 1.3558, + "step": 7908 + }, + { + "epoch": 0.67, + "learning_rate": 1.812867641719714e-05, + "loss": 1.3269, + "step": 7909 + }, + { + "epoch": 0.67, + "learning_rate": 1.812814695367314e-05, + "loss": 1.2446, + "step": 7910 + }, + { + "epoch": 0.67, + "learning_rate": 1.8127617422991793e-05, + "loss": 1.2702, + "step": 7911 + }, + { + "epoch": 0.67, + "learning_rate": 1.8127087825157463e-05, + "loss": 1.1935, + "step": 7912 + }, + { + "epoch": 0.67, + "learning_rate": 1.8126558160174535e-05, + "loss": 1.3687, + "step": 7913 + }, + { + "epoch": 0.67, + "learning_rate": 1.8126028428047387e-05, + "loss": 1.3109, + "step": 7914 + }, + { + "epoch": 0.67, + "learning_rate": 1.8125498628780386e-05, + "loss": 1.3431, + "step": 7915 + }, + { + "epoch": 0.67, + "learning_rate": 1.8124968762377916e-05, + "loss": 1.3016, + "step": 7916 + }, + { + "epoch": 0.67, + "learning_rate": 1.8124438828844354e-05, + "loss": 1.2542, + "step": 7917 + }, + { + "epoch": 0.67, + "learning_rate": 1.8123908828184082e-05, + "loss": 1.3409, + "step": 7918 + }, + { + "epoch": 0.67, + "learning_rate": 1.812337876040147e-05, + "loss": 1.3471, + "step": 7919 + }, + { + "epoch": 0.67, + "learning_rate": 1.8122848625500902e-05, + "loss": 1.1716, + "step": 7920 + }, + { + "epoch": 0.67, + "learning_rate": 1.8122318423486763e-05, + "loss": 1.3112, + "step": 7921 + }, + { + "epoch": 0.67, + "learning_rate": 1.812178815436343e-05, + "loss": 1.2693, + "step": 7922 + }, + { + "epoch": 0.67, + "learning_rate": 1.8121257818135284e-05, + "loss": 1.3264, + "step": 7923 + }, + { + "epoch": 0.67, + "learning_rate": 1.8120727414806706e-05, + "loss": 1.3486, + "step": 7924 + }, + { + "epoch": 0.67, + "learning_rate": 1.8120196944382077e-05, + "loss": 1.2433, + "step": 7925 + }, + { + "epoch": 0.67, + "learning_rate": 1.811966640686578e-05, + "loss": 1.3422, + "step": 7926 + }, + { + "epoch": 0.67, + "learning_rate": 1.8119135802262207e-05, + "loss": 1.2633, + "step": 7927 + }, + { + "epoch": 0.67, + "learning_rate": 1.8118605130575733e-05, + "loss": 1.3492, + "step": 7928 + }, + { + "epoch": 0.67, + "learning_rate": 1.8118074391810744e-05, + "loss": 1.1651, + "step": 7929 + }, + { + "epoch": 0.67, + "learning_rate": 1.8117543585971628e-05, + "loss": 1.2404, + "step": 7930 + }, + { + "epoch": 0.67, + "learning_rate": 1.8117012713062766e-05, + "loss": 1.3193, + "step": 7931 + }, + { + "epoch": 0.67, + "learning_rate": 1.8116481773088546e-05, + "loss": 1.4135, + "step": 7932 + }, + { + "epoch": 0.67, + "learning_rate": 1.811595076605336e-05, + "loss": 1.3208, + "step": 7933 + }, + { + "epoch": 0.67, + "learning_rate": 1.8115419691961592e-05, + "loss": 1.2889, + "step": 7934 + }, + { + "epoch": 0.67, + "learning_rate": 1.8114888550817627e-05, + "loss": 1.2472, + "step": 7935 + }, + { + "epoch": 0.67, + "learning_rate": 1.8114357342625857e-05, + "loss": 1.2716, + "step": 7936 + }, + { + "epoch": 0.67, + "learning_rate": 1.811382606739067e-05, + "loss": 1.3313, + "step": 7937 + }, + { + "epoch": 0.67, + "learning_rate": 1.811329472511645e-05, + "loss": 1.2896, + "step": 7938 + }, + { + "epoch": 0.67, + "learning_rate": 1.81127633158076e-05, + "loss": 1.2791, + "step": 7939 + }, + { + "epoch": 0.67, + "learning_rate": 1.81122318394685e-05, + "loss": 1.2592, + "step": 7940 + }, + { + "epoch": 0.67, + "learning_rate": 1.8111700296103546e-05, + "loss": 1.3276, + "step": 7941 + }, + { + "epoch": 0.67, + "learning_rate": 1.811116868571712e-05, + "loss": 1.2496, + "step": 7942 + }, + { + "epoch": 0.67, + "learning_rate": 1.8110637008313632e-05, + "loss": 1.3151, + "step": 7943 + }, + { + "epoch": 0.67, + "learning_rate": 1.8110105263897463e-05, + "loss": 1.3428, + "step": 7944 + }, + { + "epoch": 0.67, + "learning_rate": 1.8109573452473006e-05, + "loss": 1.3725, + "step": 7945 + }, + { + "epoch": 0.67, + "learning_rate": 1.8109041574044658e-05, + "loss": 1.3232, + "step": 7946 + }, + { + "epoch": 0.67, + "learning_rate": 1.8108509628616814e-05, + "loss": 1.3057, + "step": 7947 + }, + { + "epoch": 0.67, + "learning_rate": 1.8107977616193874e-05, + "loss": 1.2442, + "step": 7948 + }, + { + "epoch": 0.67, + "learning_rate": 1.810744553678022e-05, + "loss": 1.2245, + "step": 7949 + }, + { + "epoch": 0.67, + "learning_rate": 1.810691339038026e-05, + "loss": 1.2399, + "step": 7950 + }, + { + "epoch": 0.67, + "learning_rate": 1.8106381176998384e-05, + "loss": 1.225, + "step": 7951 + }, + { + "epoch": 0.67, + "learning_rate": 1.8105848896638995e-05, + "loss": 1.3086, + "step": 7952 + }, + { + "epoch": 0.67, + "learning_rate": 1.8105316549306492e-05, + "loss": 1.2777, + "step": 7953 + }, + { + "epoch": 0.67, + "learning_rate": 1.8104784135005264e-05, + "loss": 1.2417, + "step": 7954 + }, + { + "epoch": 0.67, + "learning_rate": 1.8104251653739715e-05, + "loss": 1.2515, + "step": 7955 + }, + { + "epoch": 0.67, + "learning_rate": 1.8103719105514247e-05, + "loss": 1.3096, + "step": 7956 + }, + { + "epoch": 0.67, + "learning_rate": 1.810318649033326e-05, + "loss": 1.2517, + "step": 7957 + }, + { + "epoch": 0.67, + "learning_rate": 1.810265380820115e-05, + "loss": 1.3558, + "step": 7958 + }, + { + "epoch": 0.67, + "learning_rate": 1.8102121059122325e-05, + "loss": 1.2178, + "step": 7959 + }, + { + "epoch": 0.67, + "learning_rate": 1.8101588243101177e-05, + "loss": 1.3178, + "step": 7960 + }, + { + "epoch": 0.67, + "learning_rate": 1.8101055360142118e-05, + "loss": 1.3011, + "step": 7961 + }, + { + "epoch": 0.67, + "learning_rate": 1.8100522410249546e-05, + "loss": 1.2855, + "step": 7962 + }, + { + "epoch": 0.67, + "learning_rate": 1.8099989393427867e-05, + "loss": 1.3412, + "step": 7963 + }, + { + "epoch": 0.67, + "learning_rate": 1.809945630968148e-05, + "loss": 1.2553, + "step": 7964 + }, + { + "epoch": 0.67, + "learning_rate": 1.8098923159014796e-05, + "loss": 1.3383, + "step": 7965 + }, + { + "epoch": 0.67, + "learning_rate": 1.8098389941432217e-05, + "loss": 1.2283, + "step": 7966 + }, + { + "epoch": 0.67, + "learning_rate": 1.8097856656938145e-05, + "loss": 1.2456, + "step": 7967 + }, + { + "epoch": 0.67, + "learning_rate": 1.8097323305536995e-05, + "loss": 1.2541, + "step": 7968 + }, + { + "epoch": 0.67, + "learning_rate": 1.809678988723316e-05, + "loss": 1.3585, + "step": 7969 + }, + { + "epoch": 0.67, + "learning_rate": 1.8096256402031064e-05, + "loss": 1.324, + "step": 7970 + }, + { + "epoch": 0.67, + "learning_rate": 1.8095722849935106e-05, + "loss": 1.2832, + "step": 7971 + }, + { + "epoch": 0.67, + "learning_rate": 1.809518923094969e-05, + "loss": 1.3432, + "step": 7972 + }, + { + "epoch": 0.67, + "learning_rate": 1.8094655545079232e-05, + "loss": 1.3402, + "step": 7973 + }, + { + "epoch": 0.67, + "learning_rate": 1.8094121792328135e-05, + "loss": 1.2563, + "step": 7974 + }, + { + "epoch": 0.67, + "learning_rate": 1.8093587972700822e-05, + "loss": 1.3781, + "step": 7975 + }, + { + "epoch": 0.67, + "learning_rate": 1.8093054086201688e-05, + "loss": 1.2515, + "step": 7976 + }, + { + "epoch": 0.67, + "learning_rate": 1.8092520132835154e-05, + "loss": 1.2248, + "step": 7977 + }, + { + "epoch": 0.67, + "learning_rate": 1.8091986112605628e-05, + "loss": 1.2565, + "step": 7978 + }, + { + "epoch": 0.67, + "learning_rate": 1.8091452025517525e-05, + "loss": 1.3025, + "step": 7979 + }, + { + "epoch": 0.67, + "learning_rate": 1.8090917871575252e-05, + "loss": 1.1814, + "step": 7980 + }, + { + "epoch": 0.67, + "learning_rate": 1.809038365078323e-05, + "loss": 1.3195, + "step": 7981 + }, + { + "epoch": 0.67, + "learning_rate": 1.8089849363145865e-05, + "loss": 1.2634, + "step": 7982 + }, + { + "epoch": 0.67, + "learning_rate": 1.8089315008667577e-05, + "loss": 1.3032, + "step": 7983 + }, + { + "epoch": 0.67, + "learning_rate": 1.8088780587352782e-05, + "loss": 1.2743, + "step": 7984 + }, + { + "epoch": 0.67, + "learning_rate": 1.8088246099205888e-05, + "loss": 1.418, + "step": 7985 + }, + { + "epoch": 0.67, + "learning_rate": 1.808771154423132e-05, + "loss": 1.2823, + "step": 7986 + }, + { + "epoch": 0.67, + "learning_rate": 1.808717692243349e-05, + "loss": 1.2984, + "step": 7987 + }, + { + "epoch": 0.67, + "learning_rate": 1.8086642233816817e-05, + "loss": 1.3282, + "step": 7988 + }, + { + "epoch": 0.67, + "learning_rate": 1.8086107478385715e-05, + "loss": 1.3033, + "step": 7989 + }, + { + "epoch": 0.67, + "learning_rate": 1.8085572656144608e-05, + "loss": 1.2999, + "step": 7990 + }, + { + "epoch": 0.67, + "learning_rate": 1.808503776709791e-05, + "loss": 1.35, + "step": 7991 + }, + { + "epoch": 0.67, + "learning_rate": 1.8084502811250045e-05, + "loss": 1.2878, + "step": 7992 + }, + { + "epoch": 0.67, + "learning_rate": 1.808396778860543e-05, + "loss": 1.2608, + "step": 7993 + }, + { + "epoch": 0.67, + "learning_rate": 1.808343269916848e-05, + "loss": 1.3398, + "step": 7994 + }, + { + "epoch": 0.67, + "learning_rate": 1.8082897542943628e-05, + "loss": 1.3004, + "step": 7995 + }, + { + "epoch": 0.67, + "learning_rate": 1.808236231993529e-05, + "loss": 1.3348, + "step": 7996 + }, + { + "epoch": 0.67, + "learning_rate": 1.8081827030147885e-05, + "loss": 1.2923, + "step": 7997 + }, + { + "epoch": 0.67, + "learning_rate": 1.8081291673585842e-05, + "loss": 1.3011, + "step": 7998 + }, + { + "epoch": 0.67, + "learning_rate": 1.8080756250253575e-05, + "loss": 1.307, + "step": 7999 + }, + { + "epoch": 0.67, + "learning_rate": 1.808022076015552e-05, + "loss": 1.2953, + "step": 8000 + }, + { + "epoch": 0.67, + "learning_rate": 1.807968520329609e-05, + "loss": 1.3693, + "step": 8001 + }, + { + "epoch": 0.67, + "learning_rate": 1.807914957967972e-05, + "loss": 1.333, + "step": 8002 + }, + { + "epoch": 0.67, + "learning_rate": 1.8078613889310824e-05, + "loss": 1.3087, + "step": 8003 + }, + { + "epoch": 0.67, + "learning_rate": 1.807807813219384e-05, + "loss": 1.3501, + "step": 8004 + }, + { + "epoch": 0.67, + "learning_rate": 1.807754230833319e-05, + "loss": 1.2917, + "step": 8005 + }, + { + "epoch": 0.67, + "learning_rate": 1.8077006417733295e-05, + "loss": 1.3408, + "step": 8006 + }, + { + "epoch": 0.67, + "learning_rate": 1.8076470460398592e-05, + "loss": 1.2563, + "step": 8007 + }, + { + "epoch": 0.67, + "learning_rate": 1.8075934436333506e-05, + "loss": 1.3445, + "step": 8008 + }, + { + "epoch": 0.67, + "learning_rate": 1.807539834554246e-05, + "loss": 1.322, + "step": 8009 + }, + { + "epoch": 0.67, + "learning_rate": 1.807486218802989e-05, + "loss": 1.2499, + "step": 8010 + }, + { + "epoch": 0.67, + "learning_rate": 1.8074325963800226e-05, + "loss": 1.3463, + "step": 8011 + }, + { + "epoch": 0.67, + "learning_rate": 1.8073789672857897e-05, + "loss": 1.3256, + "step": 8012 + }, + { + "epoch": 0.67, + "learning_rate": 1.8073253315207336e-05, + "loss": 1.2988, + "step": 8013 + }, + { + "epoch": 0.67, + "learning_rate": 1.807271689085297e-05, + "loss": 1.2995, + "step": 8014 + }, + { + "epoch": 0.67, + "learning_rate": 1.8072180399799234e-05, + "loss": 1.2806, + "step": 8015 + }, + { + "epoch": 0.67, + "learning_rate": 1.8071643842050556e-05, + "loss": 1.2853, + "step": 8016 + }, + { + "epoch": 0.67, + "learning_rate": 1.807110721761138e-05, + "loss": 1.2148, + "step": 8017 + }, + { + "epoch": 0.68, + "learning_rate": 1.807057052648613e-05, + "loss": 1.2788, + "step": 8018 + }, + { + "epoch": 0.68, + "learning_rate": 1.807003376867924e-05, + "loss": 1.4045, + "step": 8019 + }, + { + "epoch": 0.68, + "learning_rate": 1.8069496944195154e-05, + "loss": 1.2925, + "step": 8020 + }, + { + "epoch": 0.68, + "learning_rate": 1.80689600530383e-05, + "loss": 1.2547, + "step": 8021 + }, + { + "epoch": 0.68, + "learning_rate": 1.806842309521311e-05, + "loss": 1.3086, + "step": 8022 + }, + { + "epoch": 0.68, + "learning_rate": 1.806788607072403e-05, + "loss": 1.2668, + "step": 8023 + }, + { + "epoch": 0.68, + "learning_rate": 1.8067348979575498e-05, + "loss": 1.2819, + "step": 8024 + }, + { + "epoch": 0.68, + "learning_rate": 1.8066811821771943e-05, + "loss": 1.2638, + "step": 8025 + }, + { + "epoch": 0.68, + "learning_rate": 1.8066274597317804e-05, + "loss": 1.1753, + "step": 8026 + }, + { + "epoch": 0.68, + "learning_rate": 1.806573730621753e-05, + "loss": 1.3569, + "step": 8027 + }, + { + "epoch": 0.68, + "learning_rate": 1.8065199948475546e-05, + "loss": 1.3413, + "step": 8028 + }, + { + "epoch": 0.68, + "learning_rate": 1.8064662524096303e-05, + "loss": 1.3312, + "step": 8029 + }, + { + "epoch": 0.68, + "learning_rate": 1.8064125033084233e-05, + "loss": 1.3359, + "step": 8030 + }, + { + "epoch": 0.68, + "learning_rate": 1.8063587475443783e-05, + "loss": 1.289, + "step": 8031 + }, + { + "epoch": 0.68, + "learning_rate": 1.8063049851179398e-05, + "loss": 1.2514, + "step": 8032 + }, + { + "epoch": 0.68, + "learning_rate": 1.8062512160295508e-05, + "loss": 1.2655, + "step": 8033 + }, + { + "epoch": 0.68, + "learning_rate": 1.8061974402796563e-05, + "loss": 1.3001, + "step": 8034 + }, + { + "epoch": 0.68, + "learning_rate": 1.806143657868701e-05, + "loss": 1.3074, + "step": 8035 + }, + { + "epoch": 0.68, + "learning_rate": 1.8060898687971283e-05, + "loss": 1.3059, + "step": 8036 + }, + { + "epoch": 0.68, + "learning_rate": 1.8060360730653834e-05, + "loss": 1.2491, + "step": 8037 + }, + { + "epoch": 0.68, + "learning_rate": 1.8059822706739103e-05, + "loss": 1.3539, + "step": 8038 + }, + { + "epoch": 0.68, + "learning_rate": 1.8059284616231537e-05, + "loss": 1.2794, + "step": 8039 + }, + { + "epoch": 0.68, + "learning_rate": 1.8058746459135584e-05, + "loss": 1.2506, + "step": 8040 + }, + { + "epoch": 0.68, + "learning_rate": 1.805820823545569e-05, + "loss": 1.2334, + "step": 8041 + }, + { + "epoch": 0.68, + "learning_rate": 1.8057669945196297e-05, + "loss": 1.2818, + "step": 8042 + }, + { + "epoch": 0.68, + "learning_rate": 1.8057131588361857e-05, + "loss": 1.3291, + "step": 8043 + }, + { + "epoch": 0.68, + "learning_rate": 1.8056593164956817e-05, + "loss": 1.2242, + "step": 8044 + }, + { + "epoch": 0.68, + "learning_rate": 1.8056054674985627e-05, + "loss": 1.2952, + "step": 8045 + }, + { + "epoch": 0.68, + "learning_rate": 1.8055516118452737e-05, + "loss": 1.3096, + "step": 8046 + }, + { + "epoch": 0.68, + "learning_rate": 1.805497749536259e-05, + "loss": 1.3225, + "step": 8047 + }, + { + "epoch": 0.68, + "learning_rate": 1.8054438805719645e-05, + "loss": 1.3258, + "step": 8048 + }, + { + "epoch": 0.68, + "learning_rate": 1.8053900049528348e-05, + "loss": 1.3006, + "step": 8049 + }, + { + "epoch": 0.68, + "learning_rate": 1.805336122679315e-05, + "loss": 1.2262, + "step": 8050 + }, + { + "epoch": 0.68, + "learning_rate": 1.8052822337518504e-05, + "loss": 1.304, + "step": 8051 + }, + { + "epoch": 0.68, + "learning_rate": 1.8052283381708865e-05, + "loss": 1.2123, + "step": 8052 + }, + { + "epoch": 0.68, + "learning_rate": 1.8051744359368682e-05, + "loss": 1.2739, + "step": 8053 + }, + { + "epoch": 0.68, + "learning_rate": 1.805120527050241e-05, + "loss": 1.2743, + "step": 8054 + }, + { + "epoch": 0.68, + "learning_rate": 1.8050666115114502e-05, + "loss": 1.2791, + "step": 8055 + }, + { + "epoch": 0.68, + "learning_rate": 1.8050126893209416e-05, + "loss": 1.2795, + "step": 8056 + }, + { + "epoch": 0.68, + "learning_rate": 1.8049587604791603e-05, + "loss": 1.2397, + "step": 8057 + }, + { + "epoch": 0.68, + "learning_rate": 1.8049048249865525e-05, + "loss": 1.294, + "step": 8058 + }, + { + "epoch": 0.68, + "learning_rate": 1.804850882843563e-05, + "loss": 1.3066, + "step": 8059 + }, + { + "epoch": 0.68, + "learning_rate": 1.8047969340506383e-05, + "loss": 1.3746, + "step": 8060 + }, + { + "epoch": 0.68, + "learning_rate": 1.8047429786082237e-05, + "loss": 1.3207, + "step": 8061 + }, + { + "epoch": 0.68, + "learning_rate": 1.8046890165167648e-05, + "loss": 1.2786, + "step": 8062 + }, + { + "epoch": 0.68, + "learning_rate": 1.8046350477767078e-05, + "loss": 1.3174, + "step": 8063 + }, + { + "epoch": 0.68, + "learning_rate": 1.8045810723884985e-05, + "loss": 1.2139, + "step": 8064 + }, + { + "epoch": 0.68, + "learning_rate": 1.804527090352583e-05, + "loss": 1.3741, + "step": 8065 + }, + { + "epoch": 0.68, + "learning_rate": 1.8044731016694072e-05, + "loss": 1.2784, + "step": 8066 + }, + { + "epoch": 0.68, + "learning_rate": 1.8044191063394173e-05, + "loss": 1.28, + "step": 8067 + }, + { + "epoch": 0.68, + "learning_rate": 1.804365104363059e-05, + "loss": 1.276, + "step": 8068 + }, + { + "epoch": 0.68, + "learning_rate": 1.804311095740779e-05, + "loss": 1.2692, + "step": 8069 + }, + { + "epoch": 0.68, + "learning_rate": 1.804257080473023e-05, + "loss": 1.3045, + "step": 8070 + }, + { + "epoch": 0.68, + "learning_rate": 1.804203058560238e-05, + "loss": 1.3007, + "step": 8071 + }, + { + "epoch": 0.68, + "learning_rate": 1.8041490300028698e-05, + "loss": 1.3279, + "step": 8072 + }, + { + "epoch": 0.68, + "learning_rate": 1.804094994801365e-05, + "loss": 1.3133, + "step": 8073 + }, + { + "epoch": 0.68, + "learning_rate": 1.8040409529561697e-05, + "loss": 1.2535, + "step": 8074 + }, + { + "epoch": 0.68, + "learning_rate": 1.803986904467731e-05, + "loss": 1.2997, + "step": 8075 + }, + { + "epoch": 0.68, + "learning_rate": 1.803932849336495e-05, + "loss": 1.2672, + "step": 8076 + }, + { + "epoch": 0.68, + "learning_rate": 1.803878787562909e-05, + "loss": 1.2441, + "step": 8077 + }, + { + "epoch": 0.68, + "learning_rate": 1.8038247191474188e-05, + "loss": 1.3671, + "step": 8078 + }, + { + "epoch": 0.68, + "learning_rate": 1.8037706440904713e-05, + "loss": 1.3383, + "step": 8079 + }, + { + "epoch": 0.68, + "learning_rate": 1.803716562392514e-05, + "loss": 1.3715, + "step": 8080 + }, + { + "epoch": 0.68, + "learning_rate": 1.8036624740539928e-05, + "loss": 1.4019, + "step": 8081 + }, + { + "epoch": 0.68, + "learning_rate": 1.8036083790753553e-05, + "loss": 1.3017, + "step": 8082 + }, + { + "epoch": 0.68, + "learning_rate": 1.8035542774570483e-05, + "loss": 1.2419, + "step": 8083 + }, + { + "epoch": 0.68, + "learning_rate": 1.8035001691995183e-05, + "loss": 1.2292, + "step": 8084 + }, + { + "epoch": 0.68, + "learning_rate": 1.803446054303213e-05, + "loss": 1.2738, + "step": 8085 + }, + { + "epoch": 0.68, + "learning_rate": 1.8033919327685794e-05, + "loss": 1.2715, + "step": 8086 + }, + { + "epoch": 0.68, + "learning_rate": 1.803337804596064e-05, + "loss": 1.2293, + "step": 8087 + }, + { + "epoch": 0.68, + "learning_rate": 1.803283669786115e-05, + "loss": 1.3149, + "step": 8088 + }, + { + "epoch": 0.68, + "learning_rate": 1.8032295283391793e-05, + "loss": 1.3664, + "step": 8089 + }, + { + "epoch": 0.68, + "learning_rate": 1.803175380255704e-05, + "loss": 1.291, + "step": 8090 + }, + { + "epoch": 0.68, + "learning_rate": 1.8031212255361366e-05, + "loss": 1.3758, + "step": 8091 + }, + { + "epoch": 0.68, + "learning_rate": 1.8030670641809246e-05, + "loss": 1.2584, + "step": 8092 + }, + { + "epoch": 0.68, + "learning_rate": 1.8030128961905155e-05, + "loss": 1.3276, + "step": 8093 + }, + { + "epoch": 0.68, + "learning_rate": 1.802958721565357e-05, + "loss": 1.3205, + "step": 8094 + }, + { + "epoch": 0.68, + "learning_rate": 1.8029045403058964e-05, + "loss": 1.3958, + "step": 8095 + }, + { + "epoch": 0.68, + "learning_rate": 1.8028503524125817e-05, + "loss": 1.2474, + "step": 8096 + }, + { + "epoch": 0.68, + "learning_rate": 1.8027961578858604e-05, + "loss": 1.3508, + "step": 8097 + }, + { + "epoch": 0.68, + "learning_rate": 1.8027419567261805e-05, + "loss": 1.184, + "step": 8098 + }, + { + "epoch": 0.68, + "learning_rate": 1.8026877489339894e-05, + "loss": 1.2193, + "step": 8099 + }, + { + "epoch": 0.68, + "learning_rate": 1.802633534509735e-05, + "loss": 1.2996, + "step": 8100 + }, + { + "epoch": 0.68, + "learning_rate": 1.8025793134538658e-05, + "loss": 1.3249, + "step": 8101 + }, + { + "epoch": 0.68, + "learning_rate": 1.8025250857668295e-05, + "loss": 1.3679, + "step": 8102 + }, + { + "epoch": 0.68, + "learning_rate": 1.8024708514490735e-05, + "loss": 1.1873, + "step": 8103 + }, + { + "epoch": 0.68, + "learning_rate": 1.8024166105010472e-05, + "loss": 1.2465, + "step": 8104 + }, + { + "epoch": 0.68, + "learning_rate": 1.802362362923198e-05, + "loss": 1.2165, + "step": 8105 + }, + { + "epoch": 0.68, + "learning_rate": 1.802308108715974e-05, + "loss": 1.2647, + "step": 8106 + }, + { + "epoch": 0.68, + "learning_rate": 1.8022538478798234e-05, + "loss": 1.4159, + "step": 8107 + }, + { + "epoch": 0.68, + "learning_rate": 1.802199580415195e-05, + "loss": 1.3769, + "step": 8108 + }, + { + "epoch": 0.68, + "learning_rate": 1.802145306322537e-05, + "loss": 1.2158, + "step": 8109 + }, + { + "epoch": 0.68, + "learning_rate": 1.8020910256022975e-05, + "loss": 1.3258, + "step": 8110 + }, + { + "epoch": 0.68, + "learning_rate": 1.8020367382549255e-05, + "loss": 1.3212, + "step": 8111 + }, + { + "epoch": 0.68, + "learning_rate": 1.8019824442808695e-05, + "loss": 1.2837, + "step": 8112 + }, + { + "epoch": 0.68, + "learning_rate": 1.8019281436805777e-05, + "loss": 1.3252, + "step": 8113 + }, + { + "epoch": 0.68, + "learning_rate": 1.8018738364544984e-05, + "loss": 1.3663, + "step": 8114 + }, + { + "epoch": 0.68, + "learning_rate": 1.8018195226030812e-05, + "loss": 1.3521, + "step": 8115 + }, + { + "epoch": 0.68, + "learning_rate": 1.8017652021267748e-05, + "loss": 1.3657, + "step": 8116 + }, + { + "epoch": 0.68, + "learning_rate": 1.8017108750260272e-05, + "loss": 1.2708, + "step": 8117 + }, + { + "epoch": 0.68, + "learning_rate": 1.801656541301288e-05, + "loss": 1.3143, + "step": 8118 + }, + { + "epoch": 0.68, + "learning_rate": 1.8016022009530058e-05, + "loss": 1.2812, + "step": 8119 + }, + { + "epoch": 0.68, + "learning_rate": 1.8015478539816297e-05, + "loss": 1.2881, + "step": 8120 + }, + { + "epoch": 0.68, + "learning_rate": 1.8014935003876092e-05, + "loss": 1.3434, + "step": 8121 + }, + { + "epoch": 0.68, + "learning_rate": 1.8014391401713924e-05, + "loss": 1.2803, + "step": 8122 + }, + { + "epoch": 0.68, + "learning_rate": 1.8013847733334287e-05, + "loss": 1.2731, + "step": 8123 + }, + { + "epoch": 0.68, + "learning_rate": 1.801330399874168e-05, + "loss": 1.2573, + "step": 8124 + }, + { + "epoch": 0.68, + "learning_rate": 1.8012760197940587e-05, + "loss": 1.3739, + "step": 8125 + }, + { + "epoch": 0.68, + "learning_rate": 1.8012216330935505e-05, + "loss": 1.2847, + "step": 8126 + }, + { + "epoch": 0.68, + "learning_rate": 1.801167239773093e-05, + "loss": 1.3222, + "step": 8127 + }, + { + "epoch": 0.68, + "learning_rate": 1.8011128398331356e-05, + "loss": 1.229, + "step": 8128 + }, + { + "epoch": 0.68, + "learning_rate": 1.801058433274127e-05, + "loss": 1.3004, + "step": 8129 + }, + { + "epoch": 0.68, + "learning_rate": 1.801004020096517e-05, + "loss": 1.3437, + "step": 8130 + }, + { + "epoch": 0.68, + "learning_rate": 1.8009496003007563e-05, + "loss": 1.3391, + "step": 8131 + }, + { + "epoch": 0.68, + "learning_rate": 1.8008951738872934e-05, + "loss": 1.3085, + "step": 8132 + }, + { + "epoch": 0.68, + "learning_rate": 1.800840740856578e-05, + "loss": 1.3112, + "step": 8133 + }, + { + "epoch": 0.68, + "learning_rate": 1.8007863012090602e-05, + "loss": 1.2422, + "step": 8134 + }, + { + "epoch": 0.68, + "learning_rate": 1.8007318549451898e-05, + "loss": 1.3473, + "step": 8135 + }, + { + "epoch": 0.68, + "learning_rate": 1.8006774020654164e-05, + "loss": 1.3037, + "step": 8136 + }, + { + "epoch": 0.69, + "learning_rate": 1.80062294257019e-05, + "loss": 1.2345, + "step": 8137 + }, + { + "epoch": 0.69, + "learning_rate": 1.800568476459961e-05, + "loss": 1.3694, + "step": 8138 + }, + { + "epoch": 0.69, + "learning_rate": 1.8005140037351786e-05, + "loss": 1.3839, + "step": 8139 + }, + { + "epoch": 0.69, + "learning_rate": 1.800459524396293e-05, + "loss": 1.3027, + "step": 8140 + }, + { + "epoch": 0.69, + "learning_rate": 1.8004050384437554e-05, + "loss": 1.2758, + "step": 8141 + }, + { + "epoch": 0.69, + "learning_rate": 1.8003505458780148e-05, + "loss": 1.317, + "step": 8142 + }, + { + "epoch": 0.69, + "learning_rate": 1.800296046699522e-05, + "loss": 1.2599, + "step": 8143 + }, + { + "epoch": 0.69, + "learning_rate": 1.800241540908727e-05, + "loss": 1.2576, + "step": 8144 + }, + { + "epoch": 0.69, + "learning_rate": 1.8001870285060806e-05, + "loss": 1.3085, + "step": 8145 + }, + { + "epoch": 0.69, + "learning_rate": 1.8001325094920323e-05, + "loss": 1.2173, + "step": 8146 + }, + { + "epoch": 0.69, + "learning_rate": 1.8000779838670338e-05, + "loss": 1.3453, + "step": 8147 + }, + { + "epoch": 0.69, + "learning_rate": 1.8000234516315344e-05, + "loss": 1.361, + "step": 8148 + }, + { + "epoch": 0.69, + "learning_rate": 1.7999689127859853e-05, + "loss": 1.3203, + "step": 8149 + }, + { + "epoch": 0.69, + "learning_rate": 1.7999143673308373e-05, + "loss": 1.3071, + "step": 8150 + }, + { + "epoch": 0.69, + "learning_rate": 1.7998598152665402e-05, + "loss": 1.304, + "step": 8151 + }, + { + "epoch": 0.69, + "learning_rate": 1.799805256593546e-05, + "loss": 1.2737, + "step": 8152 + }, + { + "epoch": 0.69, + "learning_rate": 1.7997506913123043e-05, + "loss": 1.273, + "step": 8153 + }, + { + "epoch": 0.69, + "learning_rate": 1.799696119423267e-05, + "loss": 1.3038, + "step": 8154 + }, + { + "epoch": 0.69, + "learning_rate": 1.7996415409268838e-05, + "loss": 1.3483, + "step": 8155 + }, + { + "epoch": 0.69, + "learning_rate": 1.7995869558236068e-05, + "loss": 1.3012, + "step": 8156 + }, + { + "epoch": 0.69, + "learning_rate": 1.7995323641138862e-05, + "loss": 1.2974, + "step": 8157 + }, + { + "epoch": 0.69, + "learning_rate": 1.7994777657981736e-05, + "loss": 1.3244, + "step": 8158 + }, + { + "epoch": 0.69, + "learning_rate": 1.7994231608769196e-05, + "loss": 1.4308, + "step": 8159 + }, + { + "epoch": 0.69, + "learning_rate": 1.799368549350575e-05, + "loss": 1.2939, + "step": 8160 + }, + { + "epoch": 0.69, + "learning_rate": 1.7993139312195923e-05, + "loss": 1.2754, + "step": 8161 + }, + { + "epoch": 0.69, + "learning_rate": 1.7992593064844223e-05, + "loss": 1.2205, + "step": 8162 + }, + { + "epoch": 0.69, + "learning_rate": 1.7992046751455158e-05, + "loss": 1.233, + "step": 8163 + }, + { + "epoch": 0.69, + "learning_rate": 1.7991500372033244e-05, + "loss": 1.2472, + "step": 8164 + }, + { + "epoch": 0.69, + "learning_rate": 1.7990953926582998e-05, + "loss": 1.3116, + "step": 8165 + }, + { + "epoch": 0.69, + "learning_rate": 1.7990407415108932e-05, + "loss": 1.3433, + "step": 8166 + }, + { + "epoch": 0.69, + "learning_rate": 1.7989860837615563e-05, + "loss": 1.231, + "step": 8167 + }, + { + "epoch": 0.69, + "learning_rate": 1.7989314194107404e-05, + "loss": 1.3468, + "step": 8168 + }, + { + "epoch": 0.69, + "learning_rate": 1.798876748458898e-05, + "loss": 1.317, + "step": 8169 + }, + { + "epoch": 0.69, + "learning_rate": 1.79882207090648e-05, + "loss": 1.2847, + "step": 8170 + }, + { + "epoch": 0.69, + "learning_rate": 1.798767386753938e-05, + "loss": 1.2953, + "step": 8171 + }, + { + "epoch": 0.69, + "learning_rate": 1.7987126960017246e-05, + "loss": 1.3395, + "step": 8172 + }, + { + "epoch": 0.69, + "learning_rate": 1.7986579986502913e-05, + "loss": 1.2757, + "step": 8173 + }, + { + "epoch": 0.69, + "learning_rate": 1.79860329470009e-05, + "loss": 1.2428, + "step": 8174 + }, + { + "epoch": 0.69, + "learning_rate": 1.798548584151573e-05, + "loss": 1.4034, + "step": 8175 + }, + { + "epoch": 0.69, + "learning_rate": 1.7984938670051913e-05, + "loss": 1.3009, + "step": 8176 + }, + { + "epoch": 0.69, + "learning_rate": 1.7984391432613982e-05, + "loss": 1.2542, + "step": 8177 + }, + { + "epoch": 0.69, + "learning_rate": 1.7983844129206452e-05, + "loss": 1.3748, + "step": 8178 + }, + { + "epoch": 0.69, + "learning_rate": 1.798329675983385e-05, + "loss": 1.3146, + "step": 8179 + }, + { + "epoch": 0.69, + "learning_rate": 1.798274932450069e-05, + "loss": 1.2478, + "step": 8180 + }, + { + "epoch": 0.69, + "learning_rate": 1.79822018232115e-05, + "loss": 1.3229, + "step": 8181 + }, + { + "epoch": 0.69, + "learning_rate": 1.7981654255970807e-05, + "loss": 1.3265, + "step": 8182 + }, + { + "epoch": 0.69, + "learning_rate": 1.7981106622783132e-05, + "loss": 1.2671, + "step": 8183 + }, + { + "epoch": 0.69, + "learning_rate": 1.7980558923653e-05, + "loss": 1.3053, + "step": 8184 + }, + { + "epoch": 0.69, + "learning_rate": 1.7980011158584935e-05, + "loss": 1.3176, + "step": 8185 + }, + { + "epoch": 0.69, + "learning_rate": 1.7979463327583465e-05, + "loss": 1.3074, + "step": 8186 + }, + { + "epoch": 0.69, + "learning_rate": 1.7978915430653115e-05, + "loss": 1.2591, + "step": 8187 + }, + { + "epoch": 0.69, + "learning_rate": 1.797836746779841e-05, + "loss": 1.3984, + "step": 8188 + }, + { + "epoch": 0.69, + "learning_rate": 1.7977819439023883e-05, + "loss": 1.2147, + "step": 8189 + }, + { + "epoch": 0.69, + "learning_rate": 1.7977271344334054e-05, + "loss": 1.3626, + "step": 8190 + }, + { + "epoch": 0.69, + "learning_rate": 1.797672318373346e-05, + "loss": 1.2451, + "step": 8191 + }, + { + "epoch": 0.69, + "learning_rate": 1.7976174957226626e-05, + "loss": 1.2918, + "step": 8192 + }, + { + "epoch": 0.69, + "learning_rate": 1.7975626664818084e-05, + "loss": 1.3073, + "step": 8193 + }, + { + "epoch": 0.69, + "learning_rate": 1.7975078306512353e-05, + "loss": 1.2626, + "step": 8194 + }, + { + "epoch": 0.69, + "learning_rate": 1.7974529882313982e-05, + "loss": 1.2792, + "step": 8195 + }, + { + "epoch": 0.69, + "learning_rate": 1.797398139222749e-05, + "loss": 1.3403, + "step": 8196 + }, + { + "epoch": 0.69, + "learning_rate": 1.7973432836257408e-05, + "loss": 1.2801, + "step": 8197 + }, + { + "epoch": 0.69, + "learning_rate": 1.7972884214408276e-05, + "loss": 1.3252, + "step": 8198 + }, + { + "epoch": 0.69, + "learning_rate": 1.7972335526684626e-05, + "loss": 1.311, + "step": 8199 + }, + { + "epoch": 0.69, + "learning_rate": 1.7971786773090984e-05, + "loss": 1.1958, + "step": 8200 + }, + { + "epoch": 0.69, + "learning_rate": 1.7971237953631887e-05, + "loss": 1.236, + "step": 8201 + }, + { + "epoch": 0.69, + "learning_rate": 1.7970689068311876e-05, + "loss": 1.2625, + "step": 8202 + }, + { + "epoch": 0.69, + "learning_rate": 1.7970140117135473e-05, + "loss": 1.2711, + "step": 8203 + }, + { + "epoch": 0.69, + "learning_rate": 1.796959110010723e-05, + "loss": 1.2886, + "step": 8204 + }, + { + "epoch": 0.69, + "learning_rate": 1.796904201723167e-05, + "loss": 1.3736, + "step": 8205 + }, + { + "epoch": 0.69, + "learning_rate": 1.7968492868513337e-05, + "loss": 1.3314, + "step": 8206 + }, + { + "epoch": 0.69, + "learning_rate": 1.7967943653956763e-05, + "loss": 1.266, + "step": 8207 + }, + { + "epoch": 0.69, + "learning_rate": 1.796739437356649e-05, + "loss": 1.2881, + "step": 8208 + }, + { + "epoch": 0.69, + "learning_rate": 1.7966845027347054e-05, + "loss": 1.2763, + "step": 8209 + }, + { + "epoch": 0.69, + "learning_rate": 1.7966295615302996e-05, + "loss": 1.3054, + "step": 8210 + }, + { + "epoch": 0.69, + "learning_rate": 1.7965746137438853e-05, + "loss": 1.279, + "step": 8211 + }, + { + "epoch": 0.69, + "learning_rate": 1.7965196593759167e-05, + "loss": 1.2166, + "step": 8212 + }, + { + "epoch": 0.69, + "learning_rate": 1.7964646984268477e-05, + "loss": 1.3126, + "step": 8213 + }, + { + "epoch": 0.69, + "learning_rate": 1.7964097308971325e-05, + "loss": 1.3061, + "step": 8214 + }, + { + "epoch": 0.69, + "learning_rate": 1.796354756787225e-05, + "loss": 1.3294, + "step": 8215 + }, + { + "epoch": 0.69, + "learning_rate": 1.79629977609758e-05, + "loss": 1.3333, + "step": 8216 + }, + { + "epoch": 0.69, + "learning_rate": 1.7962447888286513e-05, + "loss": 1.3096, + "step": 8217 + }, + { + "epoch": 0.69, + "learning_rate": 1.796189794980893e-05, + "loss": 1.3457, + "step": 8218 + }, + { + "epoch": 0.69, + "learning_rate": 1.79613479455476e-05, + "loss": 1.391, + "step": 8219 + }, + { + "epoch": 0.69, + "learning_rate": 1.796079787550707e-05, + "loss": 1.351, + "step": 8220 + }, + { + "epoch": 0.69, + "learning_rate": 1.796024773969187e-05, + "loss": 1.3371, + "step": 8221 + }, + { + "epoch": 0.69, + "learning_rate": 1.7959697538106566e-05, + "loss": 1.277, + "step": 8222 + }, + { + "epoch": 0.69, + "learning_rate": 1.7959147270755685e-05, + "loss": 1.4708, + "step": 8223 + }, + { + "epoch": 0.69, + "learning_rate": 1.795859693764379e-05, + "loss": 1.3407, + "step": 8224 + }, + { + "epoch": 0.69, + "learning_rate": 1.7958046538775417e-05, + "loss": 1.3041, + "step": 8225 + }, + { + "epoch": 0.69, + "learning_rate": 1.7957496074155113e-05, + "loss": 1.416, + "step": 8226 + }, + { + "epoch": 0.69, + "learning_rate": 1.795694554378743e-05, + "loss": 1.288, + "step": 8227 + }, + { + "epoch": 0.69, + "learning_rate": 1.7956394947676918e-05, + "loss": 1.1691, + "step": 8228 + }, + { + "epoch": 0.69, + "learning_rate": 1.795584428582813e-05, + "loss": 1.2429, + "step": 8229 + }, + { + "epoch": 0.69, + "learning_rate": 1.7955293558245604e-05, + "loss": 1.2967, + "step": 8230 + }, + { + "epoch": 0.69, + "learning_rate": 1.7954742764933897e-05, + "loss": 1.262, + "step": 8231 + }, + { + "epoch": 0.69, + "learning_rate": 1.795419190589756e-05, + "loss": 1.247, + "step": 8232 + }, + { + "epoch": 0.69, + "learning_rate": 1.7953640981141146e-05, + "loss": 1.3155, + "step": 8233 + }, + { + "epoch": 0.69, + "learning_rate": 1.7953089990669202e-05, + "loss": 1.2815, + "step": 8234 + }, + { + "epoch": 0.69, + "learning_rate": 1.7952538934486282e-05, + "loss": 1.3546, + "step": 8235 + }, + { + "epoch": 0.69, + "learning_rate": 1.7951987812596946e-05, + "loss": 1.3627, + "step": 8236 + }, + { + "epoch": 0.69, + "learning_rate": 1.7951436625005734e-05, + "loss": 1.252, + "step": 8237 + }, + { + "epoch": 0.69, + "learning_rate": 1.7950885371717214e-05, + "loss": 1.2203, + "step": 8238 + }, + { + "epoch": 0.69, + "learning_rate": 1.7950334052735935e-05, + "loss": 1.2323, + "step": 8239 + }, + { + "epoch": 0.69, + "learning_rate": 1.7949782668066445e-05, + "loss": 1.1547, + "step": 8240 + }, + { + "epoch": 0.69, + "learning_rate": 1.794923121771331e-05, + "loss": 1.3586, + "step": 8241 + }, + { + "epoch": 0.69, + "learning_rate": 1.7948679701681086e-05, + "loss": 1.3135, + "step": 8242 + }, + { + "epoch": 0.69, + "learning_rate": 1.7948128119974324e-05, + "loss": 1.179, + "step": 8243 + }, + { + "epoch": 0.69, + "learning_rate": 1.794757647259758e-05, + "loss": 1.2757, + "step": 8244 + }, + { + "epoch": 0.69, + "learning_rate": 1.794702475955542e-05, + "loss": 1.3445, + "step": 8245 + }, + { + "epoch": 0.69, + "learning_rate": 1.7946472980852398e-05, + "loss": 1.3644, + "step": 8246 + }, + { + "epoch": 0.69, + "learning_rate": 1.7945921136493072e-05, + "loss": 1.325, + "step": 8247 + }, + { + "epoch": 0.69, + "learning_rate": 1.7945369226482e-05, + "loss": 1.3453, + "step": 8248 + }, + { + "epoch": 0.69, + "learning_rate": 1.794481725082375e-05, + "loss": 1.3617, + "step": 8249 + }, + { + "epoch": 0.69, + "learning_rate": 1.7944265209522874e-05, + "loss": 1.334, + "step": 8250 + }, + { + "epoch": 0.69, + "learning_rate": 1.7943713102583934e-05, + "loss": 1.268, + "step": 8251 + }, + { + "epoch": 0.69, + "learning_rate": 1.79431609300115e-05, + "loss": 1.2736, + "step": 8252 + }, + { + "epoch": 0.69, + "learning_rate": 1.7942608691810123e-05, + "loss": 1.2584, + "step": 8253 + }, + { + "epoch": 0.69, + "learning_rate": 1.7942056387984372e-05, + "loss": 1.3224, + "step": 8254 + }, + { + "epoch": 0.7, + "learning_rate": 1.7941504018538812e-05, + "loss": 1.2694, + "step": 8255 + }, + { + "epoch": 0.7, + "learning_rate": 1.7940951583478004e-05, + "loss": 1.3029, + "step": 8256 + }, + { + "epoch": 0.7, + "learning_rate": 1.794039908280651e-05, + "loss": 1.3334, + "step": 8257 + }, + { + "epoch": 0.7, + "learning_rate": 1.7939846516528897e-05, + "loss": 1.3066, + "step": 8258 + }, + { + "epoch": 0.7, + "learning_rate": 1.793929388464973e-05, + "loss": 1.3309, + "step": 8259 + }, + { + "epoch": 0.7, + "learning_rate": 1.793874118717358e-05, + "loss": 1.2611, + "step": 8260 + }, + { + "epoch": 0.7, + "learning_rate": 1.793818842410501e-05, + "loss": 1.2499, + "step": 8261 + }, + { + "epoch": 0.7, + "learning_rate": 1.7937635595448585e-05, + "loss": 1.2987, + "step": 8262 + }, + { + "epoch": 0.7, + "learning_rate": 1.7937082701208875e-05, + "loss": 1.2674, + "step": 8263 + }, + { + "epoch": 0.7, + "learning_rate": 1.7936529741390445e-05, + "loss": 1.2508, + "step": 8264 + }, + { + "epoch": 0.7, + "learning_rate": 1.793597671599787e-05, + "loss": 1.2206, + "step": 8265 + }, + { + "epoch": 0.7, + "learning_rate": 1.7935423625035712e-05, + "loss": 1.2928, + "step": 8266 + }, + { + "epoch": 0.7, + "learning_rate": 1.7934870468508548e-05, + "loss": 1.3054, + "step": 8267 + }, + { + "epoch": 0.7, + "learning_rate": 1.793431724642094e-05, + "loss": 1.3245, + "step": 8268 + }, + { + "epoch": 0.7, + "learning_rate": 1.7933763958777467e-05, + "loss": 1.3739, + "step": 8269 + }, + { + "epoch": 0.7, + "learning_rate": 1.7933210605582696e-05, + "loss": 1.3118, + "step": 8270 + }, + { + "epoch": 0.7, + "learning_rate": 1.7932657186841202e-05, + "loss": 1.2948, + "step": 8271 + }, + { + "epoch": 0.7, + "learning_rate": 1.7932103702557552e-05, + "loss": 1.2508, + "step": 8272 + }, + { + "epoch": 0.7, + "learning_rate": 1.7931550152736323e-05, + "loss": 1.247, + "step": 8273 + }, + { + "epoch": 0.7, + "learning_rate": 1.7930996537382094e-05, + "loss": 1.298, + "step": 8274 + }, + { + "epoch": 0.7, + "learning_rate": 1.7930442856499427e-05, + "loss": 1.3227, + "step": 8275 + }, + { + "epoch": 0.7, + "learning_rate": 1.7929889110092903e-05, + "loss": 1.2757, + "step": 8276 + }, + { + "epoch": 0.7, + "learning_rate": 1.7929335298167097e-05, + "loss": 1.2536, + "step": 8277 + }, + { + "epoch": 0.7, + "learning_rate": 1.792878142072659e-05, + "loss": 1.3106, + "step": 8278 + }, + { + "epoch": 0.7, + "learning_rate": 1.7928227477775954e-05, + "loss": 1.2596, + "step": 8279 + }, + { + "epoch": 0.7, + "learning_rate": 1.792767346931976e-05, + "loss": 1.3201, + "step": 8280 + }, + { + "epoch": 0.7, + "learning_rate": 1.7927119395362594e-05, + "loss": 1.2138, + "step": 8281 + }, + { + "epoch": 0.7, + "learning_rate": 1.7926565255909026e-05, + "loss": 1.3527, + "step": 8282 + }, + { + "epoch": 0.7, + "learning_rate": 1.7926011050963642e-05, + "loss": 1.2917, + "step": 8283 + }, + { + "epoch": 0.7, + "learning_rate": 1.7925456780531018e-05, + "loss": 1.3763, + "step": 8284 + }, + { + "epoch": 0.7, + "learning_rate": 1.7924902444615736e-05, + "loss": 1.229, + "step": 8285 + }, + { + "epoch": 0.7, + "learning_rate": 1.7924348043222372e-05, + "loss": 1.3126, + "step": 8286 + }, + { + "epoch": 0.7, + "learning_rate": 1.7923793576355505e-05, + "loss": 1.3764, + "step": 8287 + }, + { + "epoch": 0.7, + "learning_rate": 1.7923239044019724e-05, + "loss": 1.2623, + "step": 8288 + }, + { + "epoch": 0.7, + "learning_rate": 1.7922684446219606e-05, + "loss": 1.3107, + "step": 8289 + }, + { + "epoch": 0.7, + "learning_rate": 1.792212978295973e-05, + "loss": 1.3059, + "step": 8290 + }, + { + "epoch": 0.7, + "learning_rate": 1.7921575054244687e-05, + "loss": 1.3035, + "step": 8291 + }, + { + "epoch": 0.7, + "learning_rate": 1.7921020260079052e-05, + "loss": 1.2025, + "step": 8292 + }, + { + "epoch": 0.7, + "learning_rate": 1.7920465400467413e-05, + "loss": 1.3856, + "step": 8293 + }, + { + "epoch": 0.7, + "learning_rate": 1.7919910475414353e-05, + "loss": 1.333, + "step": 8294 + }, + { + "epoch": 0.7, + "learning_rate": 1.7919355484924462e-05, + "loss": 1.2993, + "step": 8295 + }, + { + "epoch": 0.7, + "learning_rate": 1.791880042900232e-05, + "loss": 1.2948, + "step": 8296 + }, + { + "epoch": 0.7, + "learning_rate": 1.7918245307652514e-05, + "loss": 1.438, + "step": 8297 + }, + { + "epoch": 0.7, + "learning_rate": 1.791769012087963e-05, + "loss": 1.2877, + "step": 8298 + }, + { + "epoch": 0.7, + "learning_rate": 1.7917134868688254e-05, + "loss": 1.2244, + "step": 8299 + }, + { + "epoch": 0.7, + "learning_rate": 1.7916579551082983e-05, + "loss": 1.3322, + "step": 8300 + }, + { + "epoch": 0.7, + "learning_rate": 1.791602416806839e-05, + "loss": 1.3611, + "step": 8301 + }, + { + "epoch": 0.7, + "learning_rate": 1.791546871964908e-05, + "loss": 1.2763, + "step": 8302 + }, + { + "epoch": 0.7, + "learning_rate": 1.791491320582963e-05, + "loss": 1.2234, + "step": 8303 + }, + { + "epoch": 0.7, + "learning_rate": 1.7914357626614634e-05, + "loss": 1.2802, + "step": 8304 + }, + { + "epoch": 0.7, + "learning_rate": 1.791380198200868e-05, + "loss": 1.2642, + "step": 8305 + }, + { + "epoch": 0.7, + "learning_rate": 1.7913246272016366e-05, + "loss": 1.2918, + "step": 8306 + }, + { + "epoch": 0.7, + "learning_rate": 1.7912690496642276e-05, + "loss": 1.2824, + "step": 8307 + }, + { + "epoch": 0.7, + "learning_rate": 1.7912134655891e-05, + "loss": 1.3774, + "step": 8308 + }, + { + "epoch": 0.7, + "learning_rate": 1.7911578749767144e-05, + "loss": 1.3683, + "step": 8309 + }, + { + "epoch": 0.7, + "learning_rate": 1.791102277827529e-05, + "loss": 1.402, + "step": 8310 + }, + { + "epoch": 0.7, + "learning_rate": 1.7910466741420035e-05, + "loss": 1.2569, + "step": 8311 + }, + { + "epoch": 0.7, + "learning_rate": 1.7909910639205967e-05, + "loss": 1.2653, + "step": 8312 + }, + { + "epoch": 0.7, + "learning_rate": 1.790935447163769e-05, + "loss": 1.2739, + "step": 8313 + }, + { + "epoch": 0.7, + "learning_rate": 1.7908798238719792e-05, + "loss": 1.3238, + "step": 8314 + }, + { + "epoch": 0.7, + "learning_rate": 1.7908241940456874e-05, + "loss": 1.1832, + "step": 8315 + }, + { + "epoch": 0.7, + "learning_rate": 1.790768557685353e-05, + "loss": 1.2275, + "step": 8316 + }, + { + "epoch": 0.7, + "learning_rate": 1.7907129147914358e-05, + "loss": 1.3445, + "step": 8317 + }, + { + "epoch": 0.7, + "learning_rate": 1.790657265364395e-05, + "loss": 1.3652, + "step": 8318 + }, + { + "epoch": 0.7, + "learning_rate": 1.7906016094046913e-05, + "loss": 1.2949, + "step": 8319 + }, + { + "epoch": 0.7, + "learning_rate": 1.7905459469127837e-05, + "loss": 1.258, + "step": 8320 + }, + { + "epoch": 0.7, + "learning_rate": 1.790490277889133e-05, + "loss": 1.2621, + "step": 8321 + }, + { + "epoch": 0.7, + "learning_rate": 1.790434602334198e-05, + "loss": 1.2805, + "step": 8322 + }, + { + "epoch": 0.7, + "learning_rate": 1.7903789202484398e-05, + "loss": 1.3297, + "step": 8323 + }, + { + "epoch": 0.7, + "learning_rate": 1.790323231632318e-05, + "loss": 1.2699, + "step": 8324 + }, + { + "epoch": 0.7, + "learning_rate": 1.7902675364862924e-05, + "loss": 1.2466, + "step": 8325 + }, + { + "epoch": 0.7, + "learning_rate": 1.7902118348108236e-05, + "loss": 1.3132, + "step": 8326 + }, + { + "epoch": 0.7, + "learning_rate": 1.790156126606372e-05, + "loss": 1.2952, + "step": 8327 + }, + { + "epoch": 0.7, + "learning_rate": 1.7901004118733976e-05, + "loss": 1.323, + "step": 8328 + }, + { + "epoch": 0.7, + "learning_rate": 1.7900446906123604e-05, + "loss": 1.2312, + "step": 8329 + }, + { + "epoch": 0.7, + "learning_rate": 1.7899889628237213e-05, + "loss": 1.2045, + "step": 8330 + }, + { + "epoch": 0.7, + "learning_rate": 1.7899332285079405e-05, + "loss": 1.4721, + "step": 8331 + }, + { + "epoch": 0.7, + "learning_rate": 1.7898774876654785e-05, + "loss": 1.2064, + "step": 8332 + }, + { + "epoch": 0.7, + "learning_rate": 1.789821740296796e-05, + "loss": 1.3962, + "step": 8333 + }, + { + "epoch": 0.7, + "learning_rate": 1.7897659864023535e-05, + "loss": 1.2954, + "step": 8334 + }, + { + "epoch": 0.7, + "learning_rate": 1.7897102259826117e-05, + "loss": 1.2891, + "step": 8335 + }, + { + "epoch": 0.7, + "learning_rate": 1.7896544590380315e-05, + "loss": 1.2882, + "step": 8336 + }, + { + "epoch": 0.7, + "learning_rate": 1.789598685569073e-05, + "loss": 1.2387, + "step": 8337 + }, + { + "epoch": 0.7, + "learning_rate": 1.7895429055761975e-05, + "loss": 1.22, + "step": 8338 + }, + { + "epoch": 0.7, + "learning_rate": 1.789487119059866e-05, + "loss": 1.3076, + "step": 8339 + }, + { + "epoch": 0.7, + "learning_rate": 1.7894313260205392e-05, + "loss": 1.2707, + "step": 8340 + }, + { + "epoch": 0.7, + "learning_rate": 1.7893755264586782e-05, + "loss": 1.2281, + "step": 8341 + }, + { + "epoch": 0.7, + "learning_rate": 1.7893197203747438e-05, + "loss": 1.2333, + "step": 8342 + }, + { + "epoch": 0.7, + "learning_rate": 1.7892639077691975e-05, + "loss": 1.2756, + "step": 8343 + }, + { + "epoch": 0.7, + "learning_rate": 1.7892080886425e-05, + "loss": 1.3343, + "step": 8344 + }, + { + "epoch": 0.7, + "learning_rate": 1.7891522629951125e-05, + "loss": 1.3055, + "step": 8345 + }, + { + "epoch": 0.7, + "learning_rate": 1.7890964308274968e-05, + "loss": 1.2399, + "step": 8346 + }, + { + "epoch": 0.7, + "learning_rate": 1.789040592140114e-05, + "loss": 1.285, + "step": 8347 + }, + { + "epoch": 0.7, + "learning_rate": 1.788984746933425e-05, + "loss": 1.1856, + "step": 8348 + }, + { + "epoch": 0.7, + "learning_rate": 1.7889288952078913e-05, + "loss": 1.1849, + "step": 8349 + }, + { + "epoch": 0.7, + "learning_rate": 1.788873036963975e-05, + "loss": 1.248, + "step": 8350 + }, + { + "epoch": 0.7, + "learning_rate": 1.788817172202137e-05, + "loss": 1.3644, + "step": 8351 + }, + { + "epoch": 0.7, + "learning_rate": 1.788761300922839e-05, + "loss": 1.3575, + "step": 8352 + }, + { + "epoch": 0.7, + "learning_rate": 1.7887054231265427e-05, + "loss": 1.3276, + "step": 8353 + }, + { + "epoch": 0.7, + "learning_rate": 1.7886495388137102e-05, + "loss": 1.2755, + "step": 8354 + }, + { + "epoch": 0.7, + "learning_rate": 1.7885936479848023e-05, + "loss": 1.288, + "step": 8355 + }, + { + "epoch": 0.7, + "learning_rate": 1.7885377506402815e-05, + "loss": 1.2547, + "step": 8356 + }, + { + "epoch": 0.7, + "learning_rate": 1.7884818467806093e-05, + "loss": 1.3099, + "step": 8357 + }, + { + "epoch": 0.7, + "learning_rate": 1.788425936406248e-05, + "loss": 1.3505, + "step": 8358 + }, + { + "epoch": 0.7, + "learning_rate": 1.788370019517659e-05, + "loss": 1.3262, + "step": 8359 + }, + { + "epoch": 0.7, + "learning_rate": 1.788314096115305e-05, + "loss": 1.2596, + "step": 8360 + }, + { + "epoch": 0.7, + "learning_rate": 1.788258166199647e-05, + "loss": 1.2173, + "step": 8361 + }, + { + "epoch": 0.7, + "learning_rate": 1.7882022297711484e-05, + "loss": 1.2628, + "step": 8362 + }, + { + "epoch": 0.7, + "learning_rate": 1.7881462868302704e-05, + "loss": 1.1968, + "step": 8363 + }, + { + "epoch": 0.7, + "learning_rate": 1.7880903373774754e-05, + "loss": 1.2951, + "step": 8364 + }, + { + "epoch": 0.7, + "learning_rate": 1.7880343814132263e-05, + "loss": 1.3321, + "step": 8365 + }, + { + "epoch": 0.7, + "learning_rate": 1.7879784189379844e-05, + "loss": 1.2751, + "step": 8366 + }, + { + "epoch": 0.7, + "learning_rate": 1.787922449952213e-05, + "loss": 1.2655, + "step": 8367 + }, + { + "epoch": 0.7, + "learning_rate": 1.787866474456374e-05, + "loss": 1.3115, + "step": 8368 + }, + { + "epoch": 0.7, + "learning_rate": 1.7878104924509298e-05, + "loss": 1.3696, + "step": 8369 + }, + { + "epoch": 0.7, + "learning_rate": 1.7877545039363436e-05, + "loss": 1.2631, + "step": 8370 + }, + { + "epoch": 0.7, + "learning_rate": 1.7876985089130774e-05, + "loss": 1.3669, + "step": 8371 + }, + { + "epoch": 0.7, + "learning_rate": 1.7876425073815942e-05, + "loss": 1.2642, + "step": 8372 + }, + { + "epoch": 0.7, + "learning_rate": 1.7875864993423562e-05, + "loss": 1.2954, + "step": 8373 + }, + { + "epoch": 0.71, + "learning_rate": 1.7875304847958265e-05, + "loss": 1.3424, + "step": 8374 + }, + { + "epoch": 0.71, + "learning_rate": 1.787474463742468e-05, + "loss": 1.2064, + "step": 8375 + }, + { + "epoch": 0.71, + "learning_rate": 1.7874184361827435e-05, + "loss": 1.3657, + "step": 8376 + }, + { + "epoch": 0.71, + "learning_rate": 1.7873624021171162e-05, + "loss": 1.3123, + "step": 8377 + }, + { + "epoch": 0.71, + "learning_rate": 1.787306361546048e-05, + "loss": 1.2785, + "step": 8378 + }, + { + "epoch": 0.71, + "learning_rate": 1.7872503144700035e-05, + "loss": 1.2584, + "step": 8379 + }, + { + "epoch": 0.71, + "learning_rate": 1.7871942608894444e-05, + "loss": 1.302, + "step": 8380 + }, + { + "epoch": 0.71, + "learning_rate": 1.7871382008048345e-05, + "loss": 1.3309, + "step": 8381 + }, + { + "epoch": 0.71, + "learning_rate": 1.7870821342166368e-05, + "loss": 1.1909, + "step": 8382 + }, + { + "epoch": 0.71, + "learning_rate": 1.7870260611253145e-05, + "loss": 1.2549, + "step": 8383 + }, + { + "epoch": 0.71, + "learning_rate": 1.786969981531331e-05, + "loss": 1.2145, + "step": 8384 + }, + { + "epoch": 0.71, + "learning_rate": 1.78691389543515e-05, + "loss": 1.2117, + "step": 8385 + }, + { + "epoch": 0.71, + "learning_rate": 1.7868578028372344e-05, + "loss": 1.2918, + "step": 8386 + }, + { + "epoch": 0.71, + "learning_rate": 1.7868017037380478e-05, + "loss": 1.2625, + "step": 8387 + }, + { + "epoch": 0.71, + "learning_rate": 1.7867455981380536e-05, + "loss": 1.4091, + "step": 8388 + }, + { + "epoch": 0.71, + "learning_rate": 1.786689486037716e-05, + "loss": 1.2857, + "step": 8389 + }, + { + "epoch": 0.71, + "learning_rate": 1.7866333674374973e-05, + "loss": 1.2771, + "step": 8390 + }, + { + "epoch": 0.71, + "learning_rate": 1.7865772423378625e-05, + "loss": 1.2142, + "step": 8391 + }, + { + "epoch": 0.71, + "learning_rate": 1.7865211107392746e-05, + "loss": 1.2905, + "step": 8392 + }, + { + "epoch": 0.71, + "learning_rate": 1.7864649726421977e-05, + "loss": 1.2499, + "step": 8393 + }, + { + "epoch": 0.71, + "learning_rate": 1.7864088280470953e-05, + "loss": 1.2349, + "step": 8394 + }, + { + "epoch": 0.71, + "learning_rate": 1.7863526769544314e-05, + "loss": 1.272, + "step": 8395 + }, + { + "epoch": 0.71, + "learning_rate": 1.7862965193646703e-05, + "loss": 1.2851, + "step": 8396 + }, + { + "epoch": 0.71, + "learning_rate": 1.7862403552782756e-05, + "loss": 1.3171, + "step": 8397 + }, + { + "epoch": 0.71, + "learning_rate": 1.7861841846957112e-05, + "loss": 1.2449, + "step": 8398 + }, + { + "epoch": 0.71, + "learning_rate": 1.786128007617442e-05, + "loss": 1.2313, + "step": 8399 + }, + { + "epoch": 0.71, + "learning_rate": 1.7860718240439314e-05, + "loss": 1.2001, + "step": 8400 + }, + { + "epoch": 0.71, + "learning_rate": 1.7860156339756437e-05, + "loss": 1.3297, + "step": 8401 + }, + { + "epoch": 0.71, + "learning_rate": 1.7859594374130432e-05, + "loss": 1.3528, + "step": 8402 + }, + { + "epoch": 0.71, + "learning_rate": 1.7859032343565942e-05, + "loss": 1.2358, + "step": 8403 + }, + { + "epoch": 0.71, + "learning_rate": 1.7858470248067615e-05, + "loss": 1.3516, + "step": 8404 + }, + { + "epoch": 0.71, + "learning_rate": 1.785790808764009e-05, + "loss": 1.3335, + "step": 8405 + }, + { + "epoch": 0.71, + "learning_rate": 1.7857345862288013e-05, + "loss": 1.1496, + "step": 8406 + }, + { + "epoch": 0.71, + "learning_rate": 1.785678357201603e-05, + "loss": 1.3198, + "step": 8407 + }, + { + "epoch": 0.71, + "learning_rate": 1.7856221216828787e-05, + "loss": 1.3024, + "step": 8408 + }, + { + "epoch": 0.71, + "learning_rate": 1.785565879673093e-05, + "loss": 1.3689, + "step": 8409 + }, + { + "epoch": 0.71, + "learning_rate": 1.785509631172711e-05, + "loss": 1.3222, + "step": 8410 + }, + { + "epoch": 0.71, + "learning_rate": 1.7854533761821965e-05, + "loss": 1.3811, + "step": 8411 + }, + { + "epoch": 0.71, + "learning_rate": 1.785397114702015e-05, + "loss": 1.2507, + "step": 8412 + }, + { + "epoch": 0.71, + "learning_rate": 1.785340846732631e-05, + "loss": 1.3263, + "step": 8413 + }, + { + "epoch": 0.71, + "learning_rate": 1.7852845722745095e-05, + "loss": 1.3007, + "step": 8414 + }, + { + "epoch": 0.71, + "learning_rate": 1.7852282913281162e-05, + "loss": 1.1979, + "step": 8415 + }, + { + "epoch": 0.71, + "learning_rate": 1.785172003893915e-05, + "loss": 1.2924, + "step": 8416 + }, + { + "epoch": 0.71, + "learning_rate": 1.785115709972371e-05, + "loss": 1.2665, + "step": 8417 + }, + { + "epoch": 0.71, + "learning_rate": 1.7850594095639505e-05, + "loss": 1.3527, + "step": 8418 + }, + { + "epoch": 0.71, + "learning_rate": 1.7850031026691175e-05, + "loss": 1.3036, + "step": 8419 + }, + { + "epoch": 0.71, + "learning_rate": 1.7849467892883375e-05, + "loss": 1.2493, + "step": 8420 + }, + { + "epoch": 0.71, + "learning_rate": 1.7848904694220762e-05, + "loss": 1.2455, + "step": 8421 + }, + { + "epoch": 0.71, + "learning_rate": 1.784834143070799e-05, + "loss": 1.3138, + "step": 8422 + }, + { + "epoch": 0.71, + "learning_rate": 1.7847778102349703e-05, + "loss": 1.3457, + "step": 8423 + }, + { + "epoch": 0.71, + "learning_rate": 1.784721470915056e-05, + "loss": 1.3198, + "step": 8424 + }, + { + "epoch": 0.71, + "learning_rate": 1.7846651251115225e-05, + "loss": 1.2824, + "step": 8425 + }, + { + "epoch": 0.71, + "learning_rate": 1.784608772824834e-05, + "loss": 1.3119, + "step": 8426 + }, + { + "epoch": 0.71, + "learning_rate": 1.784552414055457e-05, + "loss": 1.35, + "step": 8427 + }, + { + "epoch": 0.71, + "learning_rate": 1.7844960488038568e-05, + "loss": 1.2276, + "step": 8428 + }, + { + "epoch": 0.71, + "learning_rate": 1.7844396770704993e-05, + "loss": 1.3269, + "step": 8429 + }, + { + "epoch": 0.71, + "learning_rate": 1.78438329885585e-05, + "loss": 1.271, + "step": 8430 + }, + { + "epoch": 0.71, + "learning_rate": 1.784326914160375e-05, + "loss": 1.3268, + "step": 8431 + }, + { + "epoch": 0.71, + "learning_rate": 1.7842705229845397e-05, + "loss": 1.2025, + "step": 8432 + }, + { + "epoch": 0.71, + "learning_rate": 1.7842141253288105e-05, + "loss": 1.2301, + "step": 8433 + }, + { + "epoch": 0.71, + "learning_rate": 1.7841577211936533e-05, + "loss": 1.2623, + "step": 8434 + }, + { + "epoch": 0.71, + "learning_rate": 1.7841013105795343e-05, + "loss": 1.3237, + "step": 8435 + }, + { + "epoch": 0.71, + "learning_rate": 1.784044893486919e-05, + "loss": 1.3684, + "step": 8436 + }, + { + "epoch": 0.71, + "learning_rate": 1.783988469916274e-05, + "loss": 1.2902, + "step": 8437 + }, + { + "epoch": 0.71, + "learning_rate": 1.783932039868065e-05, + "loss": 1.2679, + "step": 8438 + }, + { + "epoch": 0.71, + "learning_rate": 1.7838756033427585e-05, + "loss": 1.2268, + "step": 8439 + }, + { + "epoch": 0.71, + "learning_rate": 1.7838191603408216e-05, + "loss": 1.2631, + "step": 8440 + }, + { + "epoch": 0.71, + "learning_rate": 1.7837627108627195e-05, + "loss": 1.2663, + "step": 8441 + }, + { + "epoch": 0.71, + "learning_rate": 1.783706254908919e-05, + "loss": 1.3217, + "step": 8442 + }, + { + "epoch": 0.71, + "learning_rate": 1.783649792479887e-05, + "loss": 1.2726, + "step": 8443 + }, + { + "epoch": 0.71, + "learning_rate": 1.7835933235760894e-05, + "loss": 1.3119, + "step": 8444 + }, + { + "epoch": 0.71, + "learning_rate": 1.7835368481979925e-05, + "loss": 1.3431, + "step": 8445 + }, + { + "epoch": 0.71, + "learning_rate": 1.7834803663460642e-05, + "loss": 1.2026, + "step": 8446 + }, + { + "epoch": 0.71, + "learning_rate": 1.78342387802077e-05, + "loss": 1.2708, + "step": 8447 + }, + { + "epoch": 0.71, + "learning_rate": 1.783367383222577e-05, + "loss": 1.2977, + "step": 8448 + }, + { + "epoch": 0.71, + "learning_rate": 1.7833108819519517e-05, + "loss": 1.2776, + "step": 8449 + }, + { + "epoch": 0.71, + "learning_rate": 1.7832543742093615e-05, + "loss": 1.2719, + "step": 8450 + }, + { + "epoch": 0.71, + "learning_rate": 1.783197859995273e-05, + "loss": 1.3251, + "step": 8451 + }, + { + "epoch": 0.71, + "learning_rate": 1.7831413393101532e-05, + "loss": 1.2243, + "step": 8452 + }, + { + "epoch": 0.71, + "learning_rate": 1.7830848121544688e-05, + "loss": 1.3173, + "step": 8453 + }, + { + "epoch": 0.71, + "learning_rate": 1.7830282785286873e-05, + "loss": 1.2911, + "step": 8454 + }, + { + "epoch": 0.71, + "learning_rate": 1.7829717384332753e-05, + "loss": 1.3156, + "step": 8455 + }, + { + "epoch": 0.71, + "learning_rate": 1.7829151918687003e-05, + "loss": 1.3278, + "step": 8456 + }, + { + "epoch": 0.71, + "learning_rate": 1.7828586388354295e-05, + "loss": 1.4165, + "step": 8457 + }, + { + "epoch": 0.71, + "learning_rate": 1.78280207933393e-05, + "loss": 1.3441, + "step": 8458 + }, + { + "epoch": 0.71, + "learning_rate": 1.782745513364669e-05, + "loss": 1.3185, + "step": 8459 + }, + { + "epoch": 0.71, + "learning_rate": 1.7826889409281147e-05, + "loss": 1.2869, + "step": 8460 + }, + { + "epoch": 0.71, + "learning_rate": 1.782632362024733e-05, + "loss": 1.348, + "step": 8461 + }, + { + "epoch": 0.71, + "learning_rate": 1.782575776654993e-05, + "loss": 1.3168, + "step": 8462 + }, + { + "epoch": 0.71, + "learning_rate": 1.782519184819361e-05, + "loss": 1.255, + "step": 8463 + }, + { + "epoch": 0.71, + "learning_rate": 1.782462586518305e-05, + "loss": 1.3743, + "step": 8464 + }, + { + "epoch": 0.71, + "learning_rate": 1.782405981752293e-05, + "loss": 1.3458, + "step": 8465 + }, + { + "epoch": 0.71, + "learning_rate": 1.7823493705217925e-05, + "loss": 1.3734, + "step": 8466 + }, + { + "epoch": 0.71, + "learning_rate": 1.782292752827271e-05, + "loss": 1.3287, + "step": 8467 + }, + { + "epoch": 0.71, + "learning_rate": 1.782236128669196e-05, + "loss": 1.26, + "step": 8468 + }, + { + "epoch": 0.71, + "learning_rate": 1.7821794980480362e-05, + "loss": 1.3256, + "step": 8469 + }, + { + "epoch": 0.71, + "learning_rate": 1.782122860964259e-05, + "loss": 1.2663, + "step": 8470 + }, + { + "epoch": 0.71, + "learning_rate": 1.7820662174183323e-05, + "loss": 1.2129, + "step": 8471 + }, + { + "epoch": 0.71, + "learning_rate": 1.7820095674107243e-05, + "loss": 1.3236, + "step": 8472 + }, + { + "epoch": 0.71, + "learning_rate": 1.781952910941903e-05, + "loss": 1.3012, + "step": 8473 + }, + { + "epoch": 0.71, + "learning_rate": 1.7818962480123363e-05, + "loss": 1.2226, + "step": 8474 + }, + { + "epoch": 0.71, + "learning_rate": 1.7818395786224928e-05, + "loss": 1.3264, + "step": 8475 + }, + { + "epoch": 0.71, + "learning_rate": 1.7817829027728403e-05, + "loss": 1.3658, + "step": 8476 + }, + { + "epoch": 0.71, + "learning_rate": 1.7817262204638473e-05, + "loss": 1.2851, + "step": 8477 + }, + { + "epoch": 0.71, + "learning_rate": 1.7816695316959822e-05, + "loss": 1.2235, + "step": 8478 + }, + { + "epoch": 0.71, + "learning_rate": 1.781612836469713e-05, + "loss": 1.4167, + "step": 8479 + }, + { + "epoch": 0.71, + "learning_rate": 1.781556134785509e-05, + "loss": 1.2488, + "step": 8480 + }, + { + "epoch": 0.71, + "learning_rate": 1.7814994266438376e-05, + "loss": 1.2432, + "step": 8481 + }, + { + "epoch": 0.71, + "learning_rate": 1.7814427120451677e-05, + "loss": 1.3226, + "step": 8482 + }, + { + "epoch": 0.71, + "learning_rate": 1.7813859909899682e-05, + "loss": 1.2031, + "step": 8483 + }, + { + "epoch": 0.71, + "learning_rate": 1.7813292634787076e-05, + "loss": 1.2902, + "step": 8484 + }, + { + "epoch": 0.71, + "learning_rate": 1.7812725295118545e-05, + "loss": 1.368, + "step": 8485 + }, + { + "epoch": 0.71, + "learning_rate": 1.781215789089878e-05, + "loss": 1.2607, + "step": 8486 + }, + { + "epoch": 0.71, + "learning_rate": 1.7811590422132462e-05, + "loss": 1.2636, + "step": 8487 + }, + { + "epoch": 0.71, + "learning_rate": 1.7811022888824285e-05, + "loss": 1.3197, + "step": 8488 + }, + { + "epoch": 0.71, + "learning_rate": 1.781045529097894e-05, + "loss": 1.3445, + "step": 8489 + }, + { + "epoch": 0.71, + "learning_rate": 1.780988762860111e-05, + "loss": 1.2838, + "step": 8490 + }, + { + "epoch": 0.71, + "learning_rate": 1.780931990169549e-05, + "loss": 1.246, + "step": 8491 + }, + { + "epoch": 0.71, + "learning_rate": 1.7808752110266773e-05, + "loss": 1.2345, + "step": 8492 + }, + { + "epoch": 0.72, + "learning_rate": 1.780818425431964e-05, + "loss": 1.255, + "step": 8493 + }, + { + "epoch": 0.72, + "learning_rate": 1.7807616333858794e-05, + "loss": 1.2955, + "step": 8494 + }, + { + "epoch": 0.72, + "learning_rate": 1.7807048348888926e-05, + "loss": 1.3273, + "step": 8495 + }, + { + "epoch": 0.72, + "learning_rate": 1.780648029941472e-05, + "loss": 1.2983, + "step": 8496 + }, + { + "epoch": 0.72, + "learning_rate": 1.780591218544088e-05, + "loss": 1.3069, + "step": 8497 + }, + { + "epoch": 0.72, + "learning_rate": 1.780534400697209e-05, + "loss": 1.338, + "step": 8498 + }, + { + "epoch": 0.72, + "learning_rate": 1.7804775764013057e-05, + "loss": 1.2162, + "step": 8499 + }, + { + "epoch": 0.72, + "learning_rate": 1.7804207456568465e-05, + "loss": 1.2176, + "step": 8500 + }, + { + "epoch": 0.72, + "learning_rate": 1.7803639084643013e-05, + "loss": 1.3128, + "step": 8501 + }, + { + "epoch": 0.72, + "learning_rate": 1.78030706482414e-05, + "loss": 1.2347, + "step": 8502 + }, + { + "epoch": 0.72, + "learning_rate": 1.7802502147368317e-05, + "loss": 1.3304, + "step": 8503 + }, + { + "epoch": 0.72, + "learning_rate": 1.7801933582028464e-05, + "loss": 1.333, + "step": 8504 + }, + { + "epoch": 0.72, + "learning_rate": 1.780136495222654e-05, + "loss": 1.3488, + "step": 8505 + }, + { + "epoch": 0.72, + "learning_rate": 1.780079625796724e-05, + "loss": 1.3181, + "step": 8506 + }, + { + "epoch": 0.72, + "learning_rate": 1.7800227499255266e-05, + "loss": 1.3402, + "step": 8507 + }, + { + "epoch": 0.72, + "learning_rate": 1.779965867609532e-05, + "loss": 1.2555, + "step": 8508 + }, + { + "epoch": 0.72, + "learning_rate": 1.779908978849209e-05, + "loss": 1.3102, + "step": 8509 + }, + { + "epoch": 0.72, + "learning_rate": 1.7798520836450294e-05, + "loss": 1.3385, + "step": 8510 + }, + { + "epoch": 0.72, + "learning_rate": 1.7797951819974616e-05, + "loss": 1.3274, + "step": 8511 + }, + { + "epoch": 0.72, + "learning_rate": 1.7797382739069763e-05, + "loss": 1.2439, + "step": 8512 + }, + { + "epoch": 0.72, + "learning_rate": 1.7796813593740444e-05, + "loss": 1.3998, + "step": 8513 + }, + { + "epoch": 0.72, + "learning_rate": 1.7796244383991353e-05, + "loss": 1.3122, + "step": 8514 + }, + { + "epoch": 0.72, + "learning_rate": 1.7795675109827195e-05, + "loss": 1.1718, + "step": 8515 + }, + { + "epoch": 0.72, + "learning_rate": 1.7795105771252676e-05, + "loss": 1.3235, + "step": 8516 + }, + { + "epoch": 0.72, + "learning_rate": 1.7794536368272496e-05, + "loss": 1.3666, + "step": 8517 + }, + { + "epoch": 0.72, + "learning_rate": 1.7793966900891364e-05, + "loss": 1.1649, + "step": 8518 + }, + { + "epoch": 0.72, + "learning_rate": 1.7793397369113983e-05, + "loss": 1.2579, + "step": 8519 + }, + { + "epoch": 0.72, + "learning_rate": 1.779282777294506e-05, + "loss": 1.2575, + "step": 8520 + }, + { + "epoch": 0.72, + "learning_rate": 1.77922581123893e-05, + "loss": 1.223, + "step": 8521 + }, + { + "epoch": 0.72, + "learning_rate": 1.7791688387451405e-05, + "loss": 1.3325, + "step": 8522 + }, + { + "epoch": 0.72, + "learning_rate": 1.779111859813609e-05, + "loss": 1.1771, + "step": 8523 + }, + { + "epoch": 0.72, + "learning_rate": 1.779054874444806e-05, + "loss": 1.32, + "step": 8524 + }, + { + "epoch": 0.72, + "learning_rate": 1.7789978826392022e-05, + "loss": 1.2947, + "step": 8525 + }, + { + "epoch": 0.72, + "learning_rate": 1.7789408843972683e-05, + "loss": 1.3393, + "step": 8526 + }, + { + "epoch": 0.72, + "learning_rate": 1.778883879719476e-05, + "loss": 1.2605, + "step": 8527 + }, + { + "epoch": 0.72, + "learning_rate": 1.7788268686062958e-05, + "loss": 1.3867, + "step": 8528 + }, + { + "epoch": 0.72, + "learning_rate": 1.7787698510581984e-05, + "loss": 1.2952, + "step": 8529 + }, + { + "epoch": 0.72, + "learning_rate": 1.7787128270756556e-05, + "loss": 1.2985, + "step": 8530 + }, + { + "epoch": 0.72, + "learning_rate": 1.778655796659138e-05, + "loss": 1.2909, + "step": 8531 + }, + { + "epoch": 0.72, + "learning_rate": 1.778598759809117e-05, + "loss": 1.3106, + "step": 8532 + }, + { + "epoch": 0.72, + "learning_rate": 1.778541716526064e-05, + "loss": 1.3259, + "step": 8533 + }, + { + "epoch": 0.72, + "learning_rate": 1.7784846668104503e-05, + "loss": 1.3004, + "step": 8534 + }, + { + "epoch": 0.72, + "learning_rate": 1.7784276106627468e-05, + "loss": 1.3301, + "step": 8535 + }, + { + "epoch": 0.72, + "learning_rate": 1.7783705480834252e-05, + "loss": 1.2146, + "step": 8536 + }, + { + "epoch": 0.72, + "learning_rate": 1.7783134790729572e-05, + "loss": 1.263, + "step": 8537 + }, + { + "epoch": 0.72, + "learning_rate": 1.7782564036318143e-05, + "loss": 1.3188, + "step": 8538 + }, + { + "epoch": 0.72, + "learning_rate": 1.7781993217604676e-05, + "loss": 1.2804, + "step": 8539 + }, + { + "epoch": 0.72, + "learning_rate": 1.7781422334593893e-05, + "loss": 1.2845, + "step": 8540 + }, + { + "epoch": 0.72, + "learning_rate": 1.778085138729051e-05, + "loss": 1.3633, + "step": 8541 + }, + { + "epoch": 0.72, + "learning_rate": 1.778028037569924e-05, + "loss": 1.3131, + "step": 8542 + }, + { + "epoch": 0.72, + "learning_rate": 1.7779709299824804e-05, + "loss": 1.2918, + "step": 8543 + }, + { + "epoch": 0.72, + "learning_rate": 1.777913815967192e-05, + "loss": 1.293, + "step": 8544 + }, + { + "epoch": 0.72, + "learning_rate": 1.7778566955245306e-05, + "loss": 1.2863, + "step": 8545 + }, + { + "epoch": 0.72, + "learning_rate": 1.7777995686549682e-05, + "loss": 1.255, + "step": 8546 + }, + { + "epoch": 0.72, + "learning_rate": 1.777742435358977e-05, + "loss": 1.2615, + "step": 8547 + }, + { + "epoch": 0.72, + "learning_rate": 1.777685295637029e-05, + "loss": 1.1974, + "step": 8548 + }, + { + "epoch": 0.72, + "learning_rate": 1.7776281494895956e-05, + "loss": 1.2772, + "step": 8549 + }, + { + "epoch": 0.72, + "learning_rate": 1.77757099691715e-05, + "loss": 1.2656, + "step": 8550 + }, + { + "epoch": 0.72, + "learning_rate": 1.777513837920164e-05, + "loss": 1.275, + "step": 8551 + }, + { + "epoch": 0.72, + "learning_rate": 1.7774566724991096e-05, + "loss": 1.3184, + "step": 8552 + }, + { + "epoch": 0.72, + "learning_rate": 1.77739950065446e-05, + "loss": 1.3245, + "step": 8553 + }, + { + "epoch": 0.72, + "learning_rate": 1.7773423223866862e-05, + "loss": 1.2519, + "step": 8554 + }, + { + "epoch": 0.72, + "learning_rate": 1.7772851376962618e-05, + "loss": 1.2749, + "step": 8555 + }, + { + "epoch": 0.72, + "learning_rate": 1.7772279465836583e-05, + "loss": 1.3573, + "step": 8556 + }, + { + "epoch": 0.72, + "learning_rate": 1.777170749049349e-05, + "loss": 1.2942, + "step": 8557 + }, + { + "epoch": 0.72, + "learning_rate": 1.7771135450938067e-05, + "loss": 1.2607, + "step": 8558 + }, + { + "epoch": 0.72, + "learning_rate": 1.777056334717503e-05, + "loss": 1.3316, + "step": 8559 + }, + { + "epoch": 0.72, + "learning_rate": 1.7769991179209115e-05, + "loss": 1.3238, + "step": 8560 + }, + { + "epoch": 0.72, + "learning_rate": 1.7769418947045043e-05, + "loss": 1.3005, + "step": 8561 + }, + { + "epoch": 0.72, + "learning_rate": 1.7768846650687546e-05, + "loss": 1.2551, + "step": 8562 + }, + { + "epoch": 0.72, + "learning_rate": 1.7768274290141354e-05, + "loss": 1.2545, + "step": 8563 + }, + { + "epoch": 0.72, + "learning_rate": 1.7767701865411193e-05, + "loss": 1.2941, + "step": 8564 + }, + { + "epoch": 0.72, + "learning_rate": 1.7767129376501792e-05, + "loss": 1.2811, + "step": 8565 + }, + { + "epoch": 0.72, + "learning_rate": 1.7766556823417885e-05, + "loss": 1.2901, + "step": 8566 + }, + { + "epoch": 0.72, + "learning_rate": 1.7765984206164196e-05, + "loss": 1.3143, + "step": 8567 + }, + { + "epoch": 0.72, + "learning_rate": 1.7765411524745464e-05, + "loss": 1.2442, + "step": 8568 + }, + { + "epoch": 0.72, + "learning_rate": 1.7764838779166415e-05, + "loss": 1.31, + "step": 8569 + }, + { + "epoch": 0.72, + "learning_rate": 1.776426596943178e-05, + "loss": 1.1877, + "step": 8570 + }, + { + "epoch": 0.72, + "learning_rate": 1.77636930955463e-05, + "loss": 1.2945, + "step": 8571 + }, + { + "epoch": 0.72, + "learning_rate": 1.7763120157514698e-05, + "loss": 1.2749, + "step": 8572 + }, + { + "epoch": 0.72, + "learning_rate": 1.7762547155341712e-05, + "loss": 1.3494, + "step": 8573 + }, + { + "epoch": 0.72, + "learning_rate": 1.7761974089032083e-05, + "loss": 1.3337, + "step": 8574 + }, + { + "epoch": 0.72, + "learning_rate": 1.7761400958590536e-05, + "loss": 1.1512, + "step": 8575 + }, + { + "epoch": 0.72, + "learning_rate": 1.776082776402181e-05, + "loss": 1.2092, + "step": 8576 + }, + { + "epoch": 0.72, + "learning_rate": 1.7760254505330645e-05, + "loss": 1.232, + "step": 8577 + }, + { + "epoch": 0.72, + "learning_rate": 1.775968118252177e-05, + "loss": 1.2991, + "step": 8578 + }, + { + "epoch": 0.72, + "learning_rate": 1.7759107795599928e-05, + "loss": 1.222, + "step": 8579 + }, + { + "epoch": 0.72, + "learning_rate": 1.7758534344569855e-05, + "loss": 1.2268, + "step": 8580 + }, + { + "epoch": 0.72, + "learning_rate": 1.7757960829436284e-05, + "loss": 1.2816, + "step": 8581 + }, + { + "epoch": 0.72, + "learning_rate": 1.775738725020396e-05, + "loss": 1.3037, + "step": 8582 + }, + { + "epoch": 0.72, + "learning_rate": 1.775681360687762e-05, + "loss": 1.3305, + "step": 8583 + }, + { + "epoch": 0.72, + "learning_rate": 1.7756239899462e-05, + "loss": 1.3282, + "step": 8584 + }, + { + "epoch": 0.72, + "learning_rate": 1.775566612796185e-05, + "loss": 1.3428, + "step": 8585 + }, + { + "epoch": 0.72, + "learning_rate": 1.7755092292381903e-05, + "loss": 1.3529, + "step": 8586 + }, + { + "epoch": 0.72, + "learning_rate": 1.7754518392726897e-05, + "loss": 1.3335, + "step": 8587 + }, + { + "epoch": 0.72, + "learning_rate": 1.7753944429001584e-05, + "loss": 1.2792, + "step": 8588 + }, + { + "epoch": 0.72, + "learning_rate": 1.7753370401210694e-05, + "loss": 1.3388, + "step": 8589 + }, + { + "epoch": 0.72, + "learning_rate": 1.7752796309358983e-05, + "loss": 1.283, + "step": 8590 + }, + { + "epoch": 0.72, + "learning_rate": 1.775222215345118e-05, + "loss": 1.2557, + "step": 8591 + }, + { + "epoch": 0.72, + "learning_rate": 1.7751647933492043e-05, + "loss": 1.2693, + "step": 8592 + }, + { + "epoch": 0.72, + "learning_rate": 1.7751073649486307e-05, + "loss": 1.2031, + "step": 8593 + }, + { + "epoch": 0.72, + "learning_rate": 1.7750499301438724e-05, + "loss": 1.3163, + "step": 8594 + }, + { + "epoch": 0.72, + "learning_rate": 1.774992488935403e-05, + "loss": 1.2401, + "step": 8595 + }, + { + "epoch": 0.72, + "learning_rate": 1.774935041323698e-05, + "loss": 1.3575, + "step": 8596 + }, + { + "epoch": 0.72, + "learning_rate": 1.7748775873092316e-05, + "loss": 1.255, + "step": 8597 + }, + { + "epoch": 0.72, + "learning_rate": 1.7748201268924785e-05, + "loss": 1.356, + "step": 8598 + }, + { + "epoch": 0.72, + "learning_rate": 1.7747626600739135e-05, + "loss": 1.2314, + "step": 8599 + }, + { + "epoch": 0.72, + "learning_rate": 1.7747051868540114e-05, + "loss": 1.3406, + "step": 8600 + }, + { + "epoch": 0.72, + "learning_rate": 1.774647707233247e-05, + "loss": 1.2991, + "step": 8601 + }, + { + "epoch": 0.72, + "learning_rate": 1.7745902212120958e-05, + "loss": 1.3381, + "step": 8602 + }, + { + "epoch": 0.72, + "learning_rate": 1.774532728791032e-05, + "loss": 1.2023, + "step": 8603 + }, + { + "epoch": 0.72, + "learning_rate": 1.774475229970531e-05, + "loss": 1.3122, + "step": 8604 + }, + { + "epoch": 0.72, + "learning_rate": 1.7744177247510677e-05, + "loss": 1.267, + "step": 8605 + }, + { + "epoch": 0.72, + "learning_rate": 1.7743602131331174e-05, + "loss": 1.3091, + "step": 8606 + }, + { + "epoch": 0.72, + "learning_rate": 1.774302695117155e-05, + "loss": 1.3075, + "step": 8607 + }, + { + "epoch": 0.72, + "learning_rate": 1.774245170703656e-05, + "loss": 1.2173, + "step": 8608 + }, + { + "epoch": 0.72, + "learning_rate": 1.7741876398930954e-05, + "loss": 1.3939, + "step": 8609 + }, + { + "epoch": 0.72, + "learning_rate": 1.774130102685949e-05, + "loss": 1.1963, + "step": 8610 + }, + { + "epoch": 0.72, + "learning_rate": 1.774072559082692e-05, + "loss": 1.359, + "step": 8611 + }, + { + "epoch": 0.73, + "learning_rate": 1.7740150090838e-05, + "loss": 1.3022, + "step": 8612 + }, + { + "epoch": 0.73, + "learning_rate": 1.773957452689748e-05, + "loss": 1.3119, + "step": 8613 + }, + { + "epoch": 0.73, + "learning_rate": 1.773899889901012e-05, + "loss": 1.3399, + "step": 8614 + }, + { + "epoch": 0.73, + "learning_rate": 1.773842320718067e-05, + "loss": 1.2639, + "step": 8615 + }, + { + "epoch": 0.73, + "learning_rate": 1.7737847451413893e-05, + "loss": 1.3284, + "step": 8616 + }, + { + "epoch": 0.73, + "learning_rate": 1.7737271631714544e-05, + "loss": 1.3089, + "step": 8617 + }, + { + "epoch": 0.73, + "learning_rate": 1.7736695748087382e-05, + "loss": 1.2856, + "step": 8618 + }, + { + "epoch": 0.73, + "learning_rate": 1.7736119800537164e-05, + "loss": 1.2454, + "step": 8619 + }, + { + "epoch": 0.73, + "learning_rate": 1.7735543789068647e-05, + "loss": 1.2547, + "step": 8620 + }, + { + "epoch": 0.73, + "learning_rate": 1.7734967713686592e-05, + "loss": 1.3134, + "step": 8621 + }, + { + "epoch": 0.73, + "learning_rate": 1.7734391574395755e-05, + "loss": 1.3252, + "step": 8622 + }, + { + "epoch": 0.73, + "learning_rate": 1.77338153712009e-05, + "loss": 1.3923, + "step": 8623 + }, + { + "epoch": 0.73, + "learning_rate": 1.773323910410679e-05, + "loss": 1.2446, + "step": 8624 + }, + { + "epoch": 0.73, + "learning_rate": 1.7732662773118186e-05, + "loss": 1.2084, + "step": 8625 + }, + { + "epoch": 0.73, + "learning_rate": 1.7732086378239843e-05, + "loss": 1.2677, + "step": 8626 + }, + { + "epoch": 0.73, + "learning_rate": 1.7731509919476527e-05, + "loss": 1.2046, + "step": 8627 + }, + { + "epoch": 0.73, + "learning_rate": 1.7730933396833005e-05, + "loss": 1.3489, + "step": 8628 + }, + { + "epoch": 0.73, + "learning_rate": 1.7730356810314037e-05, + "loss": 1.1906, + "step": 8629 + }, + { + "epoch": 0.73, + "learning_rate": 1.7729780159924387e-05, + "loss": 1.3143, + "step": 8630 + }, + { + "epoch": 0.73, + "learning_rate": 1.772920344566882e-05, + "loss": 1.2956, + "step": 8631 + }, + { + "epoch": 0.73, + "learning_rate": 1.7728626667552093e-05, + "loss": 1.3756, + "step": 8632 + }, + { + "epoch": 0.73, + "learning_rate": 1.7728049825578986e-05, + "loss": 1.2209, + "step": 8633 + }, + { + "epoch": 0.73, + "learning_rate": 1.772747291975426e-05, + "loss": 1.2659, + "step": 8634 + }, + { + "epoch": 0.73, + "learning_rate": 1.7726895950082675e-05, + "loss": 1.2004, + "step": 8635 + }, + { + "epoch": 0.73, + "learning_rate": 1.7726318916569005e-05, + "loss": 1.3519, + "step": 8636 + }, + { + "epoch": 0.73, + "learning_rate": 1.7725741819218016e-05, + "loss": 1.2016, + "step": 8637 + }, + { + "epoch": 0.73, + "learning_rate": 1.772516465803447e-05, + "loss": 1.2261, + "step": 8638 + }, + { + "epoch": 0.73, + "learning_rate": 1.7724587433023145e-05, + "loss": 1.251, + "step": 8639 + }, + { + "epoch": 0.73, + "learning_rate": 1.7724010144188808e-05, + "loss": 1.2529, + "step": 8640 + }, + { + "epoch": 0.73, + "learning_rate": 1.7723432791536227e-05, + "loss": 1.3678, + "step": 8641 + }, + { + "epoch": 0.73, + "learning_rate": 1.7722855375070173e-05, + "loss": 1.2286, + "step": 8642 + }, + { + "epoch": 0.73, + "learning_rate": 1.7722277894795412e-05, + "loss": 1.3212, + "step": 8643 + }, + { + "epoch": 0.73, + "learning_rate": 1.7721700350716723e-05, + "loss": 1.3075, + "step": 8644 + }, + { + "epoch": 0.73, + "learning_rate": 1.772112274283887e-05, + "loss": 1.2065, + "step": 8645 + }, + { + "epoch": 0.73, + "learning_rate": 1.7720545071166638e-05, + "loss": 1.3397, + "step": 8646 + }, + { + "epoch": 0.73, + "learning_rate": 1.7719967335704788e-05, + "loss": 1.2891, + "step": 8647 + }, + { + "epoch": 0.73, + "learning_rate": 1.7719389536458097e-05, + "loss": 1.202, + "step": 8648 + }, + { + "epoch": 0.73, + "learning_rate": 1.7718811673431336e-05, + "loss": 1.2849, + "step": 8649 + }, + { + "epoch": 0.73, + "learning_rate": 1.7718233746629286e-05, + "loss": 1.3582, + "step": 8650 + }, + { + "epoch": 0.73, + "learning_rate": 1.7717655756056715e-05, + "loss": 1.2419, + "step": 8651 + }, + { + "epoch": 0.73, + "learning_rate": 1.7717077701718408e-05, + "loss": 1.3027, + "step": 8652 + }, + { + "epoch": 0.73, + "learning_rate": 1.771649958361913e-05, + "loss": 1.1717, + "step": 8653 + }, + { + "epoch": 0.73, + "learning_rate": 1.7715921401763666e-05, + "loss": 1.3242, + "step": 8654 + }, + { + "epoch": 0.73, + "learning_rate": 1.771534315615679e-05, + "loss": 1.1953, + "step": 8655 + }, + { + "epoch": 0.73, + "learning_rate": 1.771476484680328e-05, + "loss": 1.3754, + "step": 8656 + }, + { + "epoch": 0.73, + "learning_rate": 1.771418647370791e-05, + "loss": 1.2725, + "step": 8657 + }, + { + "epoch": 0.73, + "learning_rate": 1.7713608036875465e-05, + "loss": 1.3951, + "step": 8658 + }, + { + "epoch": 0.73, + "learning_rate": 1.7713029536310722e-05, + "loss": 1.3267, + "step": 8659 + }, + { + "epoch": 0.73, + "learning_rate": 1.771245097201846e-05, + "loss": 1.2893, + "step": 8660 + }, + { + "epoch": 0.73, + "learning_rate": 1.771187234400346e-05, + "loss": 1.3006, + "step": 8661 + }, + { + "epoch": 0.73, + "learning_rate": 1.7711293652270503e-05, + "loss": 1.2336, + "step": 8662 + }, + { + "epoch": 0.73, + "learning_rate": 1.771071489682437e-05, + "loss": 1.3567, + "step": 8663 + }, + { + "epoch": 0.73, + "learning_rate": 1.7710136077669838e-05, + "loss": 1.2439, + "step": 8664 + }, + { + "epoch": 0.73, + "learning_rate": 1.7709557194811703e-05, + "loss": 1.2694, + "step": 8665 + }, + { + "epoch": 0.73, + "learning_rate": 1.7708978248254734e-05, + "loss": 1.3059, + "step": 8666 + }, + { + "epoch": 0.73, + "learning_rate": 1.770839923800372e-05, + "loss": 1.411, + "step": 8667 + }, + { + "epoch": 0.73, + "learning_rate": 1.7707820164063446e-05, + "loss": 1.3737, + "step": 8668 + }, + { + "epoch": 0.73, + "learning_rate": 1.770724102643869e-05, + "loss": 1.3561, + "step": 8669 + }, + { + "epoch": 0.73, + "learning_rate": 1.770666182513425e-05, + "loss": 1.267, + "step": 8670 + }, + { + "epoch": 0.73, + "learning_rate": 1.7706082560154896e-05, + "loss": 1.2848, + "step": 8671 + }, + { + "epoch": 0.73, + "learning_rate": 1.7705503231505424e-05, + "loss": 1.2052, + "step": 8672 + }, + { + "epoch": 0.73, + "learning_rate": 1.7704923839190618e-05, + "loss": 1.2427, + "step": 8673 + }, + { + "epoch": 0.73, + "learning_rate": 1.7704344383215266e-05, + "loss": 1.3038, + "step": 8674 + }, + { + "epoch": 0.73, + "learning_rate": 1.7703764863584157e-05, + "loss": 1.3068, + "step": 8675 + }, + { + "epoch": 0.73, + "learning_rate": 1.7703185280302073e-05, + "loss": 1.272, + "step": 8676 + }, + { + "epoch": 0.73, + "learning_rate": 1.7702605633373812e-05, + "loss": 1.3308, + "step": 8677 + }, + { + "epoch": 0.73, + "learning_rate": 1.7702025922804153e-05, + "loss": 1.3095, + "step": 8678 + }, + { + "epoch": 0.73, + "learning_rate": 1.7701446148597896e-05, + "loss": 1.3046, + "step": 8679 + }, + { + "epoch": 0.73, + "learning_rate": 1.7700866310759823e-05, + "loss": 1.3304, + "step": 8680 + }, + { + "epoch": 0.73, + "learning_rate": 1.770028640929473e-05, + "loss": 1.1843, + "step": 8681 + }, + { + "epoch": 0.73, + "learning_rate": 1.7699706444207404e-05, + "loss": 1.343, + "step": 8682 + }, + { + "epoch": 0.73, + "learning_rate": 1.769912641550264e-05, + "loss": 1.3147, + "step": 8683 + }, + { + "epoch": 0.73, + "learning_rate": 1.7698546323185227e-05, + "loss": 1.3004, + "step": 8684 + }, + { + "epoch": 0.73, + "learning_rate": 1.769796616725996e-05, + "loss": 1.2851, + "step": 8685 + }, + { + "epoch": 0.73, + "learning_rate": 1.769738594773164e-05, + "loss": 1.2133, + "step": 8686 + }, + { + "epoch": 0.73, + "learning_rate": 1.769680566460505e-05, + "loss": 1.3387, + "step": 8687 + }, + { + "epoch": 0.73, + "learning_rate": 1.7696225317884987e-05, + "loss": 1.3235, + "step": 8688 + }, + { + "epoch": 0.73, + "learning_rate": 1.7695644907576248e-05, + "loss": 1.318, + "step": 8689 + }, + { + "epoch": 0.73, + "learning_rate": 1.7695064433683628e-05, + "loss": 1.3885, + "step": 8690 + }, + { + "epoch": 0.73, + "learning_rate": 1.769448389621192e-05, + "loss": 1.1794, + "step": 8691 + }, + { + "epoch": 0.73, + "learning_rate": 1.769390329516593e-05, + "loss": 1.3212, + "step": 8692 + }, + { + "epoch": 0.73, + "learning_rate": 1.769332263055044e-05, + "loss": 1.2893, + "step": 8693 + }, + { + "epoch": 0.73, + "learning_rate": 1.7692741902370263e-05, + "loss": 1.2866, + "step": 8694 + }, + { + "epoch": 0.73, + "learning_rate": 1.7692161110630187e-05, + "loss": 1.2371, + "step": 8695 + }, + { + "epoch": 0.73, + "learning_rate": 1.7691580255335016e-05, + "loss": 1.3498, + "step": 8696 + }, + { + "epoch": 0.73, + "learning_rate": 1.769099933648955e-05, + "loss": 1.2891, + "step": 8697 + }, + { + "epoch": 0.73, + "learning_rate": 1.7690418354098578e-05, + "loss": 1.3399, + "step": 8698 + }, + { + "epoch": 0.73, + "learning_rate": 1.7689837308166916e-05, + "loss": 1.2555, + "step": 8699 + }, + { + "epoch": 0.73, + "learning_rate": 1.7689256198699354e-05, + "loss": 1.2853, + "step": 8700 + }, + { + "epoch": 0.73, + "learning_rate": 1.7688675025700698e-05, + "loss": 1.216, + "step": 8701 + }, + { + "epoch": 0.73, + "learning_rate": 1.7688093789175747e-05, + "loss": 1.3488, + "step": 8702 + }, + { + "epoch": 0.73, + "learning_rate": 1.7687512489129305e-05, + "loss": 1.3412, + "step": 8703 + }, + { + "epoch": 0.73, + "learning_rate": 1.7686931125566173e-05, + "loss": 1.3389, + "step": 8704 + }, + { + "epoch": 0.73, + "learning_rate": 1.768634969849116e-05, + "loss": 1.3242, + "step": 8705 + }, + { + "epoch": 0.73, + "learning_rate": 1.768576820790906e-05, + "loss": 1.2299, + "step": 8706 + }, + { + "epoch": 0.73, + "learning_rate": 1.7685186653824688e-05, + "loss": 1.2844, + "step": 8707 + }, + { + "epoch": 0.73, + "learning_rate": 1.7684605036242843e-05, + "loss": 1.3853, + "step": 8708 + }, + { + "epoch": 0.73, + "learning_rate": 1.768402335516833e-05, + "loss": 1.289, + "step": 8709 + }, + { + "epoch": 0.73, + "learning_rate": 1.7683441610605956e-05, + "loss": 1.3061, + "step": 8710 + }, + { + "epoch": 0.73, + "learning_rate": 1.768285980256053e-05, + "loss": 1.2592, + "step": 8711 + }, + { + "epoch": 0.73, + "learning_rate": 1.768227793103686e-05, + "loss": 1.3205, + "step": 8712 + }, + { + "epoch": 0.73, + "learning_rate": 1.7681695996039748e-05, + "loss": 1.3555, + "step": 8713 + }, + { + "epoch": 0.73, + "learning_rate": 1.7681113997574007e-05, + "loss": 1.2802, + "step": 8714 + }, + { + "epoch": 0.73, + "learning_rate": 1.768053193564444e-05, + "loss": 1.3068, + "step": 8715 + }, + { + "epoch": 0.73, + "learning_rate": 1.7679949810255865e-05, + "loss": 1.2911, + "step": 8716 + }, + { + "epoch": 0.73, + "learning_rate": 1.7679367621413086e-05, + "loss": 1.2962, + "step": 8717 + }, + { + "epoch": 0.73, + "learning_rate": 1.7678785369120915e-05, + "loss": 1.2462, + "step": 8718 + }, + { + "epoch": 0.73, + "learning_rate": 1.767820305338416e-05, + "loss": 1.2009, + "step": 8719 + }, + { + "epoch": 0.73, + "learning_rate": 1.767762067420763e-05, + "loss": 1.3048, + "step": 8720 + }, + { + "epoch": 0.73, + "learning_rate": 1.7677038231596146e-05, + "loss": 1.2542, + "step": 8721 + }, + { + "epoch": 0.73, + "learning_rate": 1.7676455725554514e-05, + "loss": 1.3351, + "step": 8722 + }, + { + "epoch": 0.73, + "learning_rate": 1.767587315608755e-05, + "loss": 1.3597, + "step": 8723 + }, + { + "epoch": 0.73, + "learning_rate": 1.7675290523200064e-05, + "loss": 1.2837, + "step": 8724 + }, + { + "epoch": 0.73, + "learning_rate": 1.7674707826896868e-05, + "loss": 1.3249, + "step": 8725 + }, + { + "epoch": 0.73, + "learning_rate": 1.7674125067182783e-05, + "loss": 1.2838, + "step": 8726 + }, + { + "epoch": 0.73, + "learning_rate": 1.767354224406262e-05, + "loss": 1.2563, + "step": 8727 + }, + { + "epoch": 0.73, + "learning_rate": 1.7672959357541195e-05, + "loss": 1.211, + "step": 8728 + }, + { + "epoch": 0.73, + "learning_rate": 1.7672376407623325e-05, + "loss": 1.2541, + "step": 8729 + }, + { + "epoch": 0.73, + "learning_rate": 1.7671793394313823e-05, + "loss": 1.2918, + "step": 8730 + }, + { + "epoch": 0.74, + "learning_rate": 1.767121031761751e-05, + "loss": 1.3173, + "step": 8731 + }, + { + "epoch": 0.74, + "learning_rate": 1.7670627177539204e-05, + "loss": 1.2829, + "step": 8732 + }, + { + "epoch": 0.74, + "learning_rate": 1.7670043974083718e-05, + "loss": 1.3125, + "step": 8733 + }, + { + "epoch": 0.74, + "learning_rate": 1.7669460707255877e-05, + "loss": 1.2327, + "step": 8734 + }, + { + "epoch": 0.74, + "learning_rate": 1.7668877377060498e-05, + "loss": 1.3702, + "step": 8735 + }, + { + "epoch": 0.74, + "learning_rate": 1.7668293983502395e-05, + "loss": 1.3471, + "step": 8736 + }, + { + "epoch": 0.74, + "learning_rate": 1.7667710526586398e-05, + "loss": 1.311, + "step": 8737 + }, + { + "epoch": 0.74, + "learning_rate": 1.766712700631732e-05, + "loss": 1.3308, + "step": 8738 + }, + { + "epoch": 0.74, + "learning_rate": 1.7666543422699987e-05, + "loss": 1.3199, + "step": 8739 + }, + { + "epoch": 0.74, + "learning_rate": 1.7665959775739215e-05, + "loss": 1.315, + "step": 8740 + }, + { + "epoch": 0.74, + "learning_rate": 1.766537606543983e-05, + "loss": 1.3369, + "step": 8741 + }, + { + "epoch": 0.74, + "learning_rate": 1.766479229180666e-05, + "loss": 1.3044, + "step": 8742 + }, + { + "epoch": 0.74, + "learning_rate": 1.7664208454844517e-05, + "loss": 1.4249, + "step": 8743 + }, + { + "epoch": 0.74, + "learning_rate": 1.7663624554558234e-05, + "loss": 1.2297, + "step": 8744 + }, + { + "epoch": 0.74, + "learning_rate": 1.7663040590952634e-05, + "loss": 1.3082, + "step": 8745 + }, + { + "epoch": 0.74, + "learning_rate": 1.7662456564032537e-05, + "loss": 1.2656, + "step": 8746 + }, + { + "epoch": 0.74, + "learning_rate": 1.766187247380277e-05, + "loss": 1.2833, + "step": 8747 + }, + { + "epoch": 0.74, + "learning_rate": 1.7661288320268165e-05, + "loss": 1.2583, + "step": 8748 + }, + { + "epoch": 0.74, + "learning_rate": 1.766070410343354e-05, + "loss": 1.2471, + "step": 8749 + }, + { + "epoch": 0.74, + "learning_rate": 1.766011982330373e-05, + "loss": 1.3835, + "step": 8750 + }, + { + "epoch": 0.74, + "learning_rate": 1.7659535479883557e-05, + "loss": 1.3968, + "step": 8751 + }, + { + "epoch": 0.74, + "learning_rate": 1.765895107317785e-05, + "loss": 1.2887, + "step": 8752 + }, + { + "epoch": 0.74, + "learning_rate": 1.7658366603191437e-05, + "loss": 1.1818, + "step": 8753 + }, + { + "epoch": 0.74, + "learning_rate": 1.765778206992915e-05, + "loss": 1.3353, + "step": 8754 + }, + { + "epoch": 0.74, + "learning_rate": 1.7657197473395814e-05, + "loss": 1.2397, + "step": 8755 + }, + { + "epoch": 0.74, + "learning_rate": 1.7656612813596267e-05, + "loss": 1.2672, + "step": 8756 + }, + { + "epoch": 0.74, + "learning_rate": 1.7656028090535328e-05, + "loss": 1.2827, + "step": 8757 + }, + { + "epoch": 0.74, + "learning_rate": 1.7655443304217837e-05, + "loss": 1.4023, + "step": 8758 + }, + { + "epoch": 0.74, + "learning_rate": 1.7654858454648625e-05, + "loss": 1.2676, + "step": 8759 + }, + { + "epoch": 0.74, + "learning_rate": 1.7654273541832523e-05, + "loss": 1.3363, + "step": 8760 + }, + { + "epoch": 0.74, + "learning_rate": 1.7653688565774363e-05, + "loss": 1.2749, + "step": 8761 + }, + { + "epoch": 0.74, + "learning_rate": 1.7653103526478977e-05, + "loss": 1.2987, + "step": 8762 + }, + { + "epoch": 0.74, + "learning_rate": 1.76525184239512e-05, + "loss": 1.2271, + "step": 8763 + }, + { + "epoch": 0.74, + "learning_rate": 1.765193325819587e-05, + "loss": 1.355, + "step": 8764 + }, + { + "epoch": 0.74, + "learning_rate": 1.7651348029217815e-05, + "loss": 1.2244, + "step": 8765 + }, + { + "epoch": 0.74, + "learning_rate": 1.7650762737021873e-05, + "loss": 1.3349, + "step": 8766 + }, + { + "epoch": 0.74, + "learning_rate": 1.7650177381612886e-05, + "loss": 1.2768, + "step": 8767 + }, + { + "epoch": 0.74, + "learning_rate": 1.764959196299568e-05, + "loss": 1.2856, + "step": 8768 + }, + { + "epoch": 0.74, + "learning_rate": 1.76490064811751e-05, + "loss": 1.3369, + "step": 8769 + }, + { + "epoch": 0.74, + "learning_rate": 1.764842093615598e-05, + "loss": 1.32, + "step": 8770 + }, + { + "epoch": 0.74, + "learning_rate": 1.7647835327943157e-05, + "loss": 1.2764, + "step": 8771 + }, + { + "epoch": 0.74, + "learning_rate": 1.764724965654147e-05, + "loss": 1.2542, + "step": 8772 + }, + { + "epoch": 0.74, + "learning_rate": 1.7646663921955763e-05, + "loss": 1.1845, + "step": 8773 + }, + { + "epoch": 0.74, + "learning_rate": 1.764607812419087e-05, + "loss": 1.321, + "step": 8774 + }, + { + "epoch": 0.74, + "learning_rate": 1.764549226325163e-05, + "loss": 1.197, + "step": 8775 + }, + { + "epoch": 0.74, + "learning_rate": 1.7644906339142886e-05, + "loss": 1.3193, + "step": 8776 + }, + { + "epoch": 0.74, + "learning_rate": 1.7644320351869478e-05, + "loss": 1.3132, + "step": 8777 + }, + { + "epoch": 0.74, + "learning_rate": 1.7643734301436256e-05, + "loss": 1.43, + "step": 8778 + }, + { + "epoch": 0.74, + "learning_rate": 1.764314818784805e-05, + "loss": 1.3161, + "step": 8779 + }, + { + "epoch": 0.74, + "learning_rate": 1.7642562011109707e-05, + "loss": 1.3336, + "step": 8780 + }, + { + "epoch": 0.74, + "learning_rate": 1.7641975771226073e-05, + "loss": 1.1908, + "step": 8781 + }, + { + "epoch": 0.74, + "learning_rate": 1.7641389468201985e-05, + "loss": 1.315, + "step": 8782 + }, + { + "epoch": 0.74, + "learning_rate": 1.7640803102042297e-05, + "loss": 1.31, + "step": 8783 + }, + { + "epoch": 0.74, + "learning_rate": 1.7640216672751847e-05, + "loss": 1.2334, + "step": 8784 + }, + { + "epoch": 0.74, + "learning_rate": 1.7639630180335478e-05, + "loss": 1.3084, + "step": 8785 + }, + { + "epoch": 0.74, + "learning_rate": 1.7639043624798044e-05, + "loss": 1.2522, + "step": 8786 + }, + { + "epoch": 0.74, + "learning_rate": 1.7638457006144383e-05, + "loss": 1.2906, + "step": 8787 + }, + { + "epoch": 0.74, + "learning_rate": 1.763787032437935e-05, + "loss": 1.2243, + "step": 8788 + }, + { + "epoch": 0.74, + "learning_rate": 1.7637283579507784e-05, + "loss": 1.3289, + "step": 8789 + }, + { + "epoch": 0.74, + "learning_rate": 1.7636696771534538e-05, + "loss": 1.2624, + "step": 8790 + }, + { + "epoch": 0.74, + "learning_rate": 1.763610990046446e-05, + "loss": 1.256, + "step": 8791 + }, + { + "epoch": 0.74, + "learning_rate": 1.7635522966302397e-05, + "loss": 1.2826, + "step": 8792 + }, + { + "epoch": 0.74, + "learning_rate": 1.76349359690532e-05, + "loss": 1.35, + "step": 8793 + }, + { + "epoch": 0.74, + "learning_rate": 1.763434890872172e-05, + "loss": 1.2673, + "step": 8794 + }, + { + "epoch": 0.74, + "learning_rate": 1.7633761785312808e-05, + "loss": 1.3524, + "step": 8795 + }, + { + "epoch": 0.74, + "learning_rate": 1.7633174598831306e-05, + "loss": 1.3126, + "step": 8796 + }, + { + "epoch": 0.74, + "learning_rate": 1.7632587349282078e-05, + "loss": 1.3298, + "step": 8797 + }, + { + "epoch": 0.74, + "learning_rate": 1.7632000036669967e-05, + "loss": 1.2756, + "step": 8798 + }, + { + "epoch": 0.74, + "learning_rate": 1.763141266099983e-05, + "loss": 1.2777, + "step": 8799 + }, + { + "epoch": 0.74, + "learning_rate": 1.763082522227652e-05, + "loss": 1.3353, + "step": 8800 + }, + { + "epoch": 0.74, + "learning_rate": 1.7630237720504894e-05, + "loss": 1.2628, + "step": 8801 + }, + { + "epoch": 0.74, + "learning_rate": 1.7629650155689794e-05, + "loss": 1.3721, + "step": 8802 + }, + { + "epoch": 0.74, + "learning_rate": 1.762906252783609e-05, + "loss": 1.4979, + "step": 8803 + }, + { + "epoch": 0.74, + "learning_rate": 1.7628474836948625e-05, + "loss": 1.2361, + "step": 8804 + }, + { + "epoch": 0.74, + "learning_rate": 1.762788708303226e-05, + "loss": 1.3466, + "step": 8805 + }, + { + "epoch": 0.74, + "learning_rate": 1.7627299266091853e-05, + "loss": 1.2718, + "step": 8806 + }, + { + "epoch": 0.74, + "learning_rate": 1.7626711386132256e-05, + "loss": 1.2877, + "step": 8807 + }, + { + "epoch": 0.74, + "learning_rate": 1.762612344315833e-05, + "loss": 1.2684, + "step": 8808 + }, + { + "epoch": 0.74, + "learning_rate": 1.7625535437174932e-05, + "loss": 1.2662, + "step": 8809 + }, + { + "epoch": 0.74, + "learning_rate": 1.762494736818692e-05, + "loss": 1.3309, + "step": 8810 + }, + { + "epoch": 0.74, + "learning_rate": 1.762435923619915e-05, + "loss": 1.3065, + "step": 8811 + }, + { + "epoch": 0.74, + "learning_rate": 1.7623771041216487e-05, + "loss": 1.2654, + "step": 8812 + }, + { + "epoch": 0.74, + "learning_rate": 1.7623182783243785e-05, + "loss": 1.3591, + "step": 8813 + }, + { + "epoch": 0.74, + "learning_rate": 1.762259446228591e-05, + "loss": 1.2004, + "step": 8814 + }, + { + "epoch": 0.74, + "learning_rate": 1.7622006078347717e-05, + "loss": 1.2721, + "step": 8815 + }, + { + "epoch": 0.74, + "learning_rate": 1.762141763143407e-05, + "loss": 1.2813, + "step": 8816 + }, + { + "epoch": 0.74, + "learning_rate": 1.7620829121549833e-05, + "loss": 1.2801, + "step": 8817 + }, + { + "epoch": 0.74, + "learning_rate": 1.7620240548699866e-05, + "loss": 1.3121, + "step": 8818 + }, + { + "epoch": 0.74, + "learning_rate": 1.7619651912889037e-05, + "loss": 1.3149, + "step": 8819 + }, + { + "epoch": 0.74, + "learning_rate": 1.76190632141222e-05, + "loss": 1.319, + "step": 8820 + }, + { + "epoch": 0.74, + "learning_rate": 1.7618474452404227e-05, + "loss": 1.3509, + "step": 8821 + }, + { + "epoch": 0.74, + "learning_rate": 1.7617885627739978e-05, + "loss": 1.3235, + "step": 8822 + }, + { + "epoch": 0.74, + "learning_rate": 1.7617296740134324e-05, + "loss": 1.2873, + "step": 8823 + }, + { + "epoch": 0.74, + "learning_rate": 1.761670778959212e-05, + "loss": 1.2839, + "step": 8824 + }, + { + "epoch": 0.74, + "learning_rate": 1.7616118776118243e-05, + "loss": 1.2883, + "step": 8825 + }, + { + "epoch": 0.74, + "learning_rate": 1.7615529699717557e-05, + "loss": 1.3311, + "step": 8826 + }, + { + "epoch": 0.74, + "learning_rate": 1.7614940560394923e-05, + "loss": 1.3672, + "step": 8827 + }, + { + "epoch": 0.74, + "learning_rate": 1.761435135815521e-05, + "loss": 1.3889, + "step": 8828 + }, + { + "epoch": 0.74, + "learning_rate": 1.7613762093003296e-05, + "loss": 1.2682, + "step": 8829 + }, + { + "epoch": 0.74, + "learning_rate": 1.7613172764944043e-05, + "loss": 1.373, + "step": 8830 + }, + { + "epoch": 0.74, + "learning_rate": 1.7612583373982317e-05, + "loss": 1.3737, + "step": 8831 + }, + { + "epoch": 0.74, + "learning_rate": 1.761199392012299e-05, + "loss": 1.3234, + "step": 8832 + }, + { + "epoch": 0.74, + "learning_rate": 1.7611404403370937e-05, + "loss": 1.3522, + "step": 8833 + }, + { + "epoch": 0.74, + "learning_rate": 1.761081482373102e-05, + "loss": 1.1986, + "step": 8834 + }, + { + "epoch": 0.74, + "learning_rate": 1.7610225181208118e-05, + "loss": 1.2422, + "step": 8835 + }, + { + "epoch": 0.74, + "learning_rate": 1.76096354758071e-05, + "loss": 1.2706, + "step": 8836 + }, + { + "epoch": 0.74, + "learning_rate": 1.7609045707532838e-05, + "loss": 1.2755, + "step": 8837 + }, + { + "epoch": 0.74, + "learning_rate": 1.7608455876390204e-05, + "loss": 1.2564, + "step": 8838 + }, + { + "epoch": 0.74, + "learning_rate": 1.7607865982384074e-05, + "loss": 1.2627, + "step": 8839 + }, + { + "epoch": 0.74, + "learning_rate": 1.760727602551932e-05, + "loss": 1.2524, + "step": 8840 + }, + { + "epoch": 0.74, + "learning_rate": 1.7606686005800817e-05, + "loss": 1.3121, + "step": 8841 + }, + { + "epoch": 0.74, + "learning_rate": 1.760609592323344e-05, + "loss": 1.3474, + "step": 8842 + }, + { + "epoch": 0.74, + "learning_rate": 1.7605505777822064e-05, + "loss": 1.2945, + "step": 8843 + }, + { + "epoch": 0.74, + "learning_rate": 1.7604915569571567e-05, + "loss": 1.2904, + "step": 8844 + }, + { + "epoch": 0.74, + "learning_rate": 1.760432529848682e-05, + "loss": 1.3441, + "step": 8845 + }, + { + "epoch": 0.74, + "learning_rate": 1.7603734964572706e-05, + "loss": 1.3541, + "step": 8846 + }, + { + "epoch": 0.74, + "learning_rate": 1.76031445678341e-05, + "loss": 1.3541, + "step": 8847 + }, + { + "epoch": 0.74, + "learning_rate": 1.760255410827588e-05, + "loss": 1.228, + "step": 8848 + }, + { + "epoch": 0.75, + "learning_rate": 1.7601963585902922e-05, + "loss": 1.2742, + "step": 8849 + }, + { + "epoch": 0.75, + "learning_rate": 1.7601373000720114e-05, + "loss": 1.2145, + "step": 8850 + }, + { + "epoch": 0.75, + "learning_rate": 1.7600782352732326e-05, + "loss": 1.2646, + "step": 8851 + }, + { + "epoch": 0.75, + "learning_rate": 1.760019164194444e-05, + "loss": 1.3366, + "step": 8852 + }, + { + "epoch": 0.75, + "learning_rate": 1.759960086836134e-05, + "loss": 1.3403, + "step": 8853 + }, + { + "epoch": 0.75, + "learning_rate": 1.7599010031987904e-05, + "loss": 1.3847, + "step": 8854 + }, + { + "epoch": 0.75, + "learning_rate": 1.759841913282902e-05, + "loss": 1.0971, + "step": 8855 + }, + { + "epoch": 0.75, + "learning_rate": 1.759782817088956e-05, + "loss": 1.2848, + "step": 8856 + }, + { + "epoch": 0.75, + "learning_rate": 1.7597237146174416e-05, + "loss": 1.3032, + "step": 8857 + }, + { + "epoch": 0.75, + "learning_rate": 1.7596646058688463e-05, + "loss": 1.3294, + "step": 8858 + }, + { + "epoch": 0.75, + "learning_rate": 1.759605490843659e-05, + "loss": 1.2706, + "step": 8859 + }, + { + "epoch": 0.75, + "learning_rate": 1.7595463695423682e-05, + "loss": 1.2872, + "step": 8860 + }, + { + "epoch": 0.75, + "learning_rate": 1.7594872419654623e-05, + "loss": 1.3426, + "step": 8861 + }, + { + "epoch": 0.75, + "learning_rate": 1.7594281081134298e-05, + "loss": 1.2896, + "step": 8862 + }, + { + "epoch": 0.75, + "learning_rate": 1.7593689679867587e-05, + "loss": 1.2467, + "step": 8863 + }, + { + "epoch": 0.75, + "learning_rate": 1.7593098215859385e-05, + "loss": 1.1964, + "step": 8864 + }, + { + "epoch": 0.75, + "learning_rate": 1.7592506689114575e-05, + "loss": 1.2959, + "step": 8865 + }, + { + "epoch": 0.75, + "learning_rate": 1.7591915099638045e-05, + "loss": 1.389, + "step": 8866 + }, + { + "epoch": 0.75, + "learning_rate": 1.7591323447434683e-05, + "loss": 1.3083, + "step": 8867 + }, + { + "epoch": 0.75, + "learning_rate": 1.759073173250938e-05, + "loss": 1.1907, + "step": 8868 + }, + { + "epoch": 0.75, + "learning_rate": 1.7590139954867017e-05, + "loss": 1.2868, + "step": 8869 + }, + { + "epoch": 0.75, + "learning_rate": 1.758954811451249e-05, + "loss": 1.2225, + "step": 8870 + }, + { + "epoch": 0.75, + "learning_rate": 1.7588956211450686e-05, + "loss": 1.2505, + "step": 8871 + }, + { + "epoch": 0.75, + "learning_rate": 1.75883642456865e-05, + "loss": 1.3607, + "step": 8872 + }, + { + "epoch": 0.75, + "learning_rate": 1.7587772217224823e-05, + "loss": 1.2904, + "step": 8873 + }, + { + "epoch": 0.75, + "learning_rate": 1.758718012607054e-05, + "loss": 1.2815, + "step": 8874 + }, + { + "epoch": 0.75, + "learning_rate": 1.7586587972228543e-05, + "loss": 1.3353, + "step": 8875 + }, + { + "epoch": 0.75, + "learning_rate": 1.758599575570373e-05, + "loss": 1.3543, + "step": 8876 + }, + { + "epoch": 0.75, + "learning_rate": 1.7585403476500994e-05, + "loss": 1.3042, + "step": 8877 + }, + { + "epoch": 0.75, + "learning_rate": 1.758481113462523e-05, + "loss": 1.3107, + "step": 8878 + }, + { + "epoch": 0.75, + "learning_rate": 1.7584218730081325e-05, + "loss": 1.1858, + "step": 8879 + }, + { + "epoch": 0.75, + "learning_rate": 1.7583626262874175e-05, + "loss": 1.2344, + "step": 8880 + }, + { + "epoch": 0.75, + "learning_rate": 1.7583033733008683e-05, + "loss": 1.3071, + "step": 8881 + }, + { + "epoch": 0.75, + "learning_rate": 1.7582441140489737e-05, + "loss": 1.2125, + "step": 8882 + }, + { + "epoch": 0.75, + "learning_rate": 1.7581848485322238e-05, + "loss": 1.2753, + "step": 8883 + }, + { + "epoch": 0.75, + "learning_rate": 1.7581255767511076e-05, + "loss": 1.3225, + "step": 8884 + }, + { + "epoch": 0.75, + "learning_rate": 1.7580662987061153e-05, + "loss": 1.2289, + "step": 8885 + }, + { + "epoch": 0.75, + "learning_rate": 1.758007014397737e-05, + "loss": 1.2788, + "step": 8886 + }, + { + "epoch": 0.75, + "learning_rate": 1.757947723826462e-05, + "loss": 1.284, + "step": 8887 + }, + { + "epoch": 0.75, + "learning_rate": 1.7578884269927803e-05, + "loss": 1.2423, + "step": 8888 + }, + { + "epoch": 0.75, + "learning_rate": 1.7578291238971816e-05, + "loss": 1.2736, + "step": 8889 + }, + { + "epoch": 0.75, + "learning_rate": 1.7577698145401565e-05, + "loss": 1.2735, + "step": 8890 + }, + { + "epoch": 0.75, + "learning_rate": 1.7577104989221948e-05, + "loss": 1.3585, + "step": 8891 + }, + { + "epoch": 0.75, + "learning_rate": 1.7576511770437862e-05, + "loss": 1.3066, + "step": 8892 + }, + { + "epoch": 0.75, + "learning_rate": 1.757591848905421e-05, + "loss": 1.3162, + "step": 8893 + }, + { + "epoch": 0.75, + "learning_rate": 1.7575325145075896e-05, + "loss": 1.2675, + "step": 8894 + }, + { + "epoch": 0.75, + "learning_rate": 1.757473173850782e-05, + "loss": 1.2465, + "step": 8895 + }, + { + "epoch": 0.75, + "learning_rate": 1.7574138269354887e-05, + "loss": 1.2597, + "step": 8896 + }, + { + "epoch": 0.75, + "learning_rate": 1.7573544737622e-05, + "loss": 1.3346, + "step": 8897 + }, + { + "epoch": 0.75, + "learning_rate": 1.7572951143314063e-05, + "loss": 1.3251, + "step": 8898 + }, + { + "epoch": 0.75, + "learning_rate": 1.757235748643598e-05, + "loss": 1.2838, + "step": 8899 + }, + { + "epoch": 0.75, + "learning_rate": 1.7571763766992656e-05, + "loss": 1.3439, + "step": 8900 + }, + { + "epoch": 0.75, + "learning_rate": 1.7571169984988994e-05, + "loss": 1.3831, + "step": 8901 + }, + { + "epoch": 0.75, + "learning_rate": 1.7570576140429906e-05, + "loss": 1.381, + "step": 8902 + }, + { + "epoch": 0.75, + "learning_rate": 1.7569982233320294e-05, + "loss": 1.2575, + "step": 8903 + }, + { + "epoch": 0.75, + "learning_rate": 1.7569388263665065e-05, + "loss": 1.3135, + "step": 8904 + }, + { + "epoch": 0.75, + "learning_rate": 1.7568794231469128e-05, + "loss": 1.2228, + "step": 8905 + }, + { + "epoch": 0.75, + "learning_rate": 1.756820013673739e-05, + "loss": 1.29, + "step": 8906 + }, + { + "epoch": 0.75, + "learning_rate": 1.756760597947476e-05, + "loss": 1.2788, + "step": 8907 + }, + { + "epoch": 0.75, + "learning_rate": 1.756701175968615e-05, + "loss": 1.2786, + "step": 8908 + }, + { + "epoch": 0.75, + "learning_rate": 1.7566417477376464e-05, + "loss": 1.3245, + "step": 8909 + }, + { + "epoch": 0.75, + "learning_rate": 1.756582313255062e-05, + "loss": 1.3868, + "step": 8910 + }, + { + "epoch": 0.75, + "learning_rate": 1.7565228725213518e-05, + "loss": 1.2913, + "step": 8911 + }, + { + "epoch": 0.75, + "learning_rate": 1.756463425537008e-05, + "loss": 1.2511, + "step": 8912 + }, + { + "epoch": 0.75, + "learning_rate": 1.756403972302521e-05, + "loss": 1.3245, + "step": 8913 + }, + { + "epoch": 0.75, + "learning_rate": 1.7563445128183825e-05, + "loss": 1.3793, + "step": 8914 + }, + { + "epoch": 0.75, + "learning_rate": 1.7562850470850834e-05, + "loss": 1.3345, + "step": 8915 + }, + { + "epoch": 0.75, + "learning_rate": 1.756225575103115e-05, + "loss": 1.2871, + "step": 8916 + }, + { + "epoch": 0.75, + "learning_rate": 1.7561660968729693e-05, + "loss": 1.3178, + "step": 8917 + }, + { + "epoch": 0.75, + "learning_rate": 1.7561066123951373e-05, + "loss": 1.3282, + "step": 8918 + }, + { + "epoch": 0.75, + "learning_rate": 1.7560471216701104e-05, + "loss": 1.3357, + "step": 8919 + }, + { + "epoch": 0.75, + "learning_rate": 1.75598762469838e-05, + "loss": 1.284, + "step": 8920 + }, + { + "epoch": 0.75, + "learning_rate": 1.755928121480438e-05, + "loss": 1.3397, + "step": 8921 + }, + { + "epoch": 0.75, + "learning_rate": 1.755868612016776e-05, + "loss": 1.3145, + "step": 8922 + }, + { + "epoch": 0.75, + "learning_rate": 1.7558090963078856e-05, + "loss": 1.2001, + "step": 8923 + }, + { + "epoch": 0.75, + "learning_rate": 1.7557495743542586e-05, + "loss": 1.3025, + "step": 8924 + }, + { + "epoch": 0.75, + "learning_rate": 1.7556900461563863e-05, + "loss": 1.3059, + "step": 8925 + }, + { + "epoch": 0.75, + "learning_rate": 1.7556305117147615e-05, + "loss": 1.3149, + "step": 8926 + }, + { + "epoch": 0.75, + "learning_rate": 1.7555709710298753e-05, + "loss": 1.421, + "step": 8927 + }, + { + "epoch": 0.75, + "learning_rate": 1.7555114241022204e-05, + "loss": 1.2617, + "step": 8928 + }, + { + "epoch": 0.75, + "learning_rate": 1.7554518709322878e-05, + "loss": 1.2675, + "step": 8929 + }, + { + "epoch": 0.75, + "learning_rate": 1.7553923115205704e-05, + "loss": 1.3164, + "step": 8930 + }, + { + "epoch": 0.75, + "learning_rate": 1.75533274586756e-05, + "loss": 1.2395, + "step": 8931 + }, + { + "epoch": 0.75, + "learning_rate": 1.755273173973748e-05, + "loss": 1.2835, + "step": 8932 + }, + { + "epoch": 0.75, + "learning_rate": 1.7552135958396282e-05, + "loss": 1.3748, + "step": 8933 + }, + { + "epoch": 0.75, + "learning_rate": 1.7551540114656916e-05, + "loss": 1.2564, + "step": 8934 + }, + { + "epoch": 0.75, + "learning_rate": 1.755094420852431e-05, + "loss": 1.3493, + "step": 8935 + }, + { + "epoch": 0.75, + "learning_rate": 1.7550348240003386e-05, + "loss": 1.3921, + "step": 8936 + }, + { + "epoch": 0.75, + "learning_rate": 1.7549752209099066e-05, + "loss": 1.2966, + "step": 8937 + }, + { + "epoch": 0.75, + "learning_rate": 1.754915611581628e-05, + "loss": 1.2346, + "step": 8938 + }, + { + "epoch": 0.75, + "learning_rate": 1.7548559960159948e-05, + "loss": 1.2919, + "step": 8939 + }, + { + "epoch": 0.75, + "learning_rate": 1.7547963742135e-05, + "loss": 1.3514, + "step": 8940 + }, + { + "epoch": 0.75, + "learning_rate": 1.7547367461746358e-05, + "loss": 1.2609, + "step": 8941 + }, + { + "epoch": 0.75, + "learning_rate": 1.754677111899895e-05, + "loss": 1.3545, + "step": 8942 + }, + { + "epoch": 0.75, + "learning_rate": 1.7546174713897705e-05, + "loss": 1.3601, + "step": 8943 + }, + { + "epoch": 0.75, + "learning_rate": 1.754557824644755e-05, + "loss": 1.2448, + "step": 8944 + }, + { + "epoch": 0.75, + "learning_rate": 1.754498171665341e-05, + "loss": 1.2844, + "step": 8945 + }, + { + "epoch": 0.75, + "learning_rate": 1.754438512452022e-05, + "loss": 1.4185, + "step": 8946 + }, + { + "epoch": 0.75, + "learning_rate": 1.7543788470052904e-05, + "loss": 1.2151, + "step": 8947 + }, + { + "epoch": 0.75, + "learning_rate": 1.7543191753256392e-05, + "loss": 1.3134, + "step": 8948 + }, + { + "epoch": 0.75, + "learning_rate": 1.754259497413562e-05, + "loss": 1.243, + "step": 8949 + }, + { + "epoch": 0.75, + "learning_rate": 1.754199813269551e-05, + "loss": 1.2496, + "step": 8950 + }, + { + "epoch": 0.75, + "learning_rate": 1.7541401228941e-05, + "loss": 1.2561, + "step": 8951 + }, + { + "epoch": 0.75, + "learning_rate": 1.7540804262877018e-05, + "loss": 1.3712, + "step": 8952 + }, + { + "epoch": 0.75, + "learning_rate": 1.7540207234508496e-05, + "loss": 1.26, + "step": 8953 + }, + { + "epoch": 0.75, + "learning_rate": 1.753961014384037e-05, + "loss": 1.3471, + "step": 8954 + }, + { + "epoch": 0.75, + "learning_rate": 1.7539012990877575e-05, + "loss": 1.2035, + "step": 8955 + }, + { + "epoch": 0.75, + "learning_rate": 1.7538415775625043e-05, + "loss": 1.3524, + "step": 8956 + }, + { + "epoch": 0.75, + "learning_rate": 1.7537818498087702e-05, + "loss": 1.2457, + "step": 8957 + }, + { + "epoch": 0.75, + "learning_rate": 1.753722115827049e-05, + "loss": 1.2947, + "step": 8958 + }, + { + "epoch": 0.75, + "learning_rate": 1.7536623756178354e-05, + "loss": 1.3223, + "step": 8959 + }, + { + "epoch": 0.75, + "learning_rate": 1.7536026291816214e-05, + "loss": 1.2199, + "step": 8960 + }, + { + "epoch": 0.75, + "learning_rate": 1.7535428765189017e-05, + "loss": 1.2814, + "step": 8961 + }, + { + "epoch": 0.75, + "learning_rate": 1.753483117630169e-05, + "loss": 1.2889, + "step": 8962 + }, + { + "epoch": 0.75, + "learning_rate": 1.753423352515918e-05, + "loss": 1.3814, + "step": 8963 + }, + { + "epoch": 0.75, + "learning_rate": 1.753363581176642e-05, + "loss": 1.3703, + "step": 8964 + }, + { + "epoch": 0.75, + "learning_rate": 1.753303803612835e-05, + "loss": 1.2348, + "step": 8965 + }, + { + "epoch": 0.75, + "learning_rate": 1.7532440198249912e-05, + "loss": 1.2259, + "step": 8966 + }, + { + "epoch": 0.75, + "learning_rate": 1.753184229813604e-05, + "loss": 1.3031, + "step": 8967 + }, + { + "epoch": 0.76, + "learning_rate": 1.7531244335791676e-05, + "loss": 1.1888, + "step": 8968 + }, + { + "epoch": 0.76, + "learning_rate": 1.7530646311221762e-05, + "loss": 1.3374, + "step": 8969 + }, + { + "epoch": 0.76, + "learning_rate": 1.7530048224431237e-05, + "loss": 1.2472, + "step": 8970 + }, + { + "epoch": 0.76, + "learning_rate": 1.7529450075425044e-05, + "loss": 1.1438, + "step": 8971 + }, + { + "epoch": 0.76, + "learning_rate": 1.7528851864208123e-05, + "loss": 1.2998, + "step": 8972 + }, + { + "epoch": 0.76, + "learning_rate": 1.7528253590785424e-05, + "loss": 1.2532, + "step": 8973 + }, + { + "epoch": 0.76, + "learning_rate": 1.7527655255161877e-05, + "loss": 1.3316, + "step": 8974 + }, + { + "epoch": 0.76, + "learning_rate": 1.7527056857342436e-05, + "loss": 1.3051, + "step": 8975 + }, + { + "epoch": 0.76, + "learning_rate": 1.7526458397332043e-05, + "loss": 1.233, + "step": 8976 + }, + { + "epoch": 0.76, + "learning_rate": 1.7525859875135643e-05, + "loss": 1.2916, + "step": 8977 + }, + { + "epoch": 0.76, + "learning_rate": 1.752526129075818e-05, + "loss": 1.313, + "step": 8978 + }, + { + "epoch": 0.76, + "learning_rate": 1.75246626442046e-05, + "loss": 1.3459, + "step": 8979 + }, + { + "epoch": 0.76, + "learning_rate": 1.7524063935479846e-05, + "loss": 1.2521, + "step": 8980 + }, + { + "epoch": 0.76, + "learning_rate": 1.752346516458887e-05, + "loss": 1.3116, + "step": 8981 + }, + { + "epoch": 0.76, + "learning_rate": 1.7522866331536615e-05, + "loss": 1.1694, + "step": 8982 + }, + { + "epoch": 0.76, + "learning_rate": 1.7522267436328033e-05, + "loss": 1.2224, + "step": 8983 + }, + { + "epoch": 0.76, + "learning_rate": 1.752166847896807e-05, + "loss": 1.274, + "step": 8984 + }, + { + "epoch": 0.76, + "learning_rate": 1.7521069459461675e-05, + "loss": 1.3719, + "step": 8985 + }, + { + "epoch": 0.76, + "learning_rate": 1.7520470377813795e-05, + "loss": 1.3448, + "step": 8986 + }, + { + "epoch": 0.76, + "learning_rate": 1.7519871234029385e-05, + "loss": 1.3147, + "step": 8987 + }, + { + "epoch": 0.76, + "learning_rate": 1.751927202811339e-05, + "loss": 1.276, + "step": 8988 + }, + { + "epoch": 0.76, + "learning_rate": 1.7518672760070764e-05, + "loss": 1.329, + "step": 8989 + }, + { + "epoch": 0.76, + "learning_rate": 1.7518073429906457e-05, + "loss": 1.3173, + "step": 8990 + }, + { + "epoch": 0.76, + "learning_rate": 1.7517474037625422e-05, + "loss": 1.2101, + "step": 8991 + }, + { + "epoch": 0.76, + "learning_rate": 1.7516874583232607e-05, + "loss": 1.2667, + "step": 8992 + }, + { + "epoch": 0.76, + "learning_rate": 1.7516275066732974e-05, + "loss": 1.2827, + "step": 8993 + }, + { + "epoch": 0.76, + "learning_rate": 1.7515675488131468e-05, + "loss": 1.3338, + "step": 8994 + }, + { + "epoch": 0.76, + "learning_rate": 1.7515075847433044e-05, + "loss": 1.2383, + "step": 8995 + }, + { + "epoch": 0.76, + "learning_rate": 1.751447614464266e-05, + "loss": 1.3375, + "step": 8996 + }, + { + "epoch": 0.76, + "learning_rate": 1.751387637976527e-05, + "loss": 1.3073, + "step": 8997 + }, + { + "epoch": 0.76, + "learning_rate": 1.7513276552805826e-05, + "loss": 1.2679, + "step": 8998 + }, + { + "epoch": 0.76, + "learning_rate": 1.751267666376929e-05, + "loss": 1.2701, + "step": 8999 + }, + { + "epoch": 0.76, + "learning_rate": 1.7512076712660612e-05, + "loss": 1.312, + "step": 9000 + }, + { + "epoch": 0.76, + "learning_rate": 1.7511476699484756e-05, + "loss": 1.2097, + "step": 9001 + }, + { + "epoch": 0.76, + "learning_rate": 1.751087662424667e-05, + "loss": 1.2599, + "step": 9002 + }, + { + "epoch": 0.76, + "learning_rate": 1.751027648695132e-05, + "loss": 1.2678, + "step": 9003 + }, + { + "epoch": 0.76, + "learning_rate": 1.7509676287603665e-05, + "loss": 1.2224, + "step": 9004 + }, + { + "epoch": 0.76, + "learning_rate": 1.750907602620866e-05, + "loss": 1.2753, + "step": 9005 + }, + { + "epoch": 0.76, + "learning_rate": 1.7508475702771264e-05, + "loss": 1.2313, + "step": 9006 + }, + { + "epoch": 0.76, + "learning_rate": 1.7507875317296435e-05, + "loss": 1.2894, + "step": 9007 + }, + { + "epoch": 0.76, + "learning_rate": 1.7507274869789144e-05, + "loss": 1.2362, + "step": 9008 + }, + { + "epoch": 0.76, + "learning_rate": 1.7506674360254343e-05, + "loss": 1.2708, + "step": 9009 + }, + { + "epoch": 0.76, + "learning_rate": 1.7506073788696996e-05, + "loss": 1.2611, + "step": 9010 + }, + { + "epoch": 0.76, + "learning_rate": 1.7505473155122063e-05, + "loss": 1.2929, + "step": 9011 + }, + { + "epoch": 0.76, + "learning_rate": 1.7504872459534513e-05, + "loss": 1.2878, + "step": 9012 + }, + { + "epoch": 0.76, + "learning_rate": 1.75042717019393e-05, + "loss": 1.3086, + "step": 9013 + }, + { + "epoch": 0.76, + "learning_rate": 1.7503670882341395e-05, + "loss": 1.2691, + "step": 9014 + }, + { + "epoch": 0.76, + "learning_rate": 1.750307000074576e-05, + "loss": 1.2841, + "step": 9015 + }, + { + "epoch": 0.76, + "learning_rate": 1.7502469057157362e-05, + "loss": 1.3951, + "step": 9016 + }, + { + "epoch": 0.76, + "learning_rate": 1.750186805158116e-05, + "loss": 1.2717, + "step": 9017 + }, + { + "epoch": 0.76, + "learning_rate": 1.750126698402212e-05, + "loss": 1.3373, + "step": 9018 + }, + { + "epoch": 0.76, + "learning_rate": 1.750066585448522e-05, + "loss": 1.3057, + "step": 9019 + }, + { + "epoch": 0.76, + "learning_rate": 1.7500064662975412e-05, + "loss": 1.2523, + "step": 9020 + }, + { + "epoch": 0.76, + "learning_rate": 1.749946340949767e-05, + "loss": 1.305, + "step": 9021 + }, + { + "epoch": 0.76, + "learning_rate": 1.749886209405696e-05, + "loss": 1.3084, + "step": 9022 + }, + { + "epoch": 0.76, + "learning_rate": 1.7498260716658255e-05, + "loss": 1.3013, + "step": 9023 + }, + { + "epoch": 0.76, + "learning_rate": 1.749765927730652e-05, + "loss": 1.3262, + "step": 9024 + }, + { + "epoch": 0.76, + "learning_rate": 1.7497057776006722e-05, + "loss": 1.261, + "step": 9025 + }, + { + "epoch": 0.76, + "learning_rate": 1.7496456212763835e-05, + "loss": 1.3191, + "step": 9026 + }, + { + "epoch": 0.76, + "learning_rate": 1.749585458758283e-05, + "loss": 1.2791, + "step": 9027 + }, + { + "epoch": 0.76, + "learning_rate": 1.7495252900468668e-05, + "loss": 1.2785, + "step": 9028 + }, + { + "epoch": 0.76, + "learning_rate": 1.7494651151426333e-05, + "loss": 1.3238, + "step": 9029 + }, + { + "epoch": 0.76, + "learning_rate": 1.749404934046079e-05, + "loss": 1.3126, + "step": 9030 + }, + { + "epoch": 0.76, + "learning_rate": 1.7493447467577013e-05, + "loss": 1.2959, + "step": 9031 + }, + { + "epoch": 0.76, + "learning_rate": 1.7492845532779976e-05, + "loss": 1.3369, + "step": 9032 + }, + { + "epoch": 0.76, + "learning_rate": 1.7492243536074653e-05, + "loss": 1.2201, + "step": 9033 + }, + { + "epoch": 0.76, + "learning_rate": 1.7491641477466014e-05, + "loss": 1.2872, + "step": 9034 + }, + { + "epoch": 0.76, + "learning_rate": 1.7491039356959033e-05, + "loss": 1.2057, + "step": 9035 + }, + { + "epoch": 0.76, + "learning_rate": 1.7490437174558693e-05, + "loss": 1.1785, + "step": 9036 + }, + { + "epoch": 0.76, + "learning_rate": 1.748983493026996e-05, + "loss": 1.3174, + "step": 9037 + }, + { + "epoch": 0.76, + "learning_rate": 1.7489232624097815e-05, + "loss": 1.2323, + "step": 9038 + }, + { + "epoch": 0.76, + "learning_rate": 1.748863025604723e-05, + "loss": 1.3186, + "step": 9039 + }, + { + "epoch": 0.76, + "learning_rate": 1.748802782612319e-05, + "loss": 1.2689, + "step": 9040 + }, + { + "epoch": 0.76, + "learning_rate": 1.7487425334330663e-05, + "loss": 1.3124, + "step": 9041 + }, + { + "epoch": 0.76, + "learning_rate": 1.7486822780674635e-05, + "loss": 1.3704, + "step": 9042 + }, + { + "epoch": 0.76, + "learning_rate": 1.7486220165160078e-05, + "loss": 1.3316, + "step": 9043 + }, + { + "epoch": 0.76, + "learning_rate": 1.748561748779198e-05, + "loss": 1.3972, + "step": 9044 + }, + { + "epoch": 0.76, + "learning_rate": 1.748501474857531e-05, + "loss": 1.292, + "step": 9045 + }, + { + "epoch": 0.76, + "learning_rate": 1.748441194751505e-05, + "loss": 1.4077, + "step": 9046 + }, + { + "epoch": 0.76, + "learning_rate": 1.7483809084616187e-05, + "loss": 1.3258, + "step": 9047 + }, + { + "epoch": 0.76, + "learning_rate": 1.7483206159883698e-05, + "loss": 1.4332, + "step": 9048 + }, + { + "epoch": 0.76, + "learning_rate": 1.7482603173322564e-05, + "loss": 1.335, + "step": 9049 + }, + { + "epoch": 0.76, + "learning_rate": 1.748200012493777e-05, + "loss": 1.3212, + "step": 9050 + }, + { + "epoch": 0.76, + "learning_rate": 1.7481397014734294e-05, + "loss": 1.2033, + "step": 9051 + }, + { + "epoch": 0.76, + "learning_rate": 1.7480793842717123e-05, + "loss": 1.3595, + "step": 9052 + }, + { + "epoch": 0.76, + "learning_rate": 1.748019060889124e-05, + "loss": 1.294, + "step": 9053 + }, + { + "epoch": 0.76, + "learning_rate": 1.7479587313261626e-05, + "loss": 1.2829, + "step": 9054 + }, + { + "epoch": 0.76, + "learning_rate": 1.747898395583327e-05, + "loss": 1.3355, + "step": 9055 + }, + { + "epoch": 0.76, + "learning_rate": 1.747838053661115e-05, + "loss": 1.4016, + "step": 9056 + }, + { + "epoch": 0.76, + "learning_rate": 1.7477777055600264e-05, + "loss": 1.1367, + "step": 9057 + }, + { + "epoch": 0.76, + "learning_rate": 1.7477173512805583e-05, + "loss": 1.3703, + "step": 9058 + }, + { + "epoch": 0.76, + "learning_rate": 1.7476569908232108e-05, + "loss": 1.3696, + "step": 9059 + }, + { + "epoch": 0.76, + "learning_rate": 1.7475966241884817e-05, + "loss": 1.2642, + "step": 9060 + }, + { + "epoch": 0.76, + "learning_rate": 1.74753625137687e-05, + "loss": 1.2917, + "step": 9061 + }, + { + "epoch": 0.76, + "learning_rate": 1.7474758723888745e-05, + "loss": 1.2514, + "step": 9062 + }, + { + "epoch": 0.76, + "learning_rate": 1.747415487224994e-05, + "loss": 1.3393, + "step": 9063 + }, + { + "epoch": 0.76, + "learning_rate": 1.7473550958857276e-05, + "loss": 1.2826, + "step": 9064 + }, + { + "epoch": 0.76, + "learning_rate": 1.7472946983715746e-05, + "loss": 1.4159, + "step": 9065 + }, + { + "epoch": 0.76, + "learning_rate": 1.7472342946830336e-05, + "loss": 1.3195, + "step": 9066 + }, + { + "epoch": 0.76, + "learning_rate": 1.7471738848206032e-05, + "loss": 1.3258, + "step": 9067 + }, + { + "epoch": 0.76, + "learning_rate": 1.7471134687847832e-05, + "loss": 1.3125, + "step": 9068 + }, + { + "epoch": 0.76, + "learning_rate": 1.747053046576073e-05, + "loss": 1.2807, + "step": 9069 + }, + { + "epoch": 0.76, + "learning_rate": 1.746992618194971e-05, + "loss": 1.314, + "step": 9070 + }, + { + "epoch": 0.76, + "learning_rate": 1.746932183641977e-05, + "loss": 1.362, + "step": 9071 + }, + { + "epoch": 0.76, + "learning_rate": 1.7468717429175905e-05, + "loss": 1.2687, + "step": 9072 + }, + { + "epoch": 0.76, + "learning_rate": 1.7468112960223104e-05, + "loss": 1.3354, + "step": 9073 + }, + { + "epoch": 0.76, + "learning_rate": 1.7467508429566366e-05, + "loss": 1.3353, + "step": 9074 + }, + { + "epoch": 0.76, + "learning_rate": 1.746690383721068e-05, + "loss": 1.3317, + "step": 9075 + }, + { + "epoch": 0.76, + "learning_rate": 1.746629918316105e-05, + "loss": 1.2985, + "step": 9076 + }, + { + "epoch": 0.76, + "learning_rate": 1.7465694467422462e-05, + "loss": 1.2229, + "step": 9077 + }, + { + "epoch": 0.76, + "learning_rate": 1.746508968999992e-05, + "loss": 1.2758, + "step": 9078 + }, + { + "epoch": 0.76, + "learning_rate": 1.7464484850898414e-05, + "loss": 1.3339, + "step": 9079 + }, + { + "epoch": 0.76, + "learning_rate": 1.746387995012295e-05, + "loss": 1.2502, + "step": 9080 + }, + { + "epoch": 0.76, + "learning_rate": 1.7463274987678514e-05, + "loss": 1.314, + "step": 9081 + }, + { + "epoch": 0.76, + "learning_rate": 1.7462669963570115e-05, + "loss": 1.3352, + "step": 9082 + }, + { + "epoch": 0.76, + "learning_rate": 1.7462064877802752e-05, + "loss": 1.2557, + "step": 9083 + }, + { + "epoch": 0.76, + "learning_rate": 1.746145973038142e-05, + "loss": 1.305, + "step": 9084 + }, + { + "epoch": 0.76, + "learning_rate": 1.7460854521311118e-05, + "loss": 1.2764, + "step": 9085 + }, + { + "epoch": 0.76, + "learning_rate": 1.7460249250596846e-05, + "loss": 1.2766, + "step": 9086 + }, + { + "epoch": 0.77, + "learning_rate": 1.745964391824361e-05, + "loss": 1.276, + "step": 9087 + }, + { + "epoch": 0.77, + "learning_rate": 1.7459038524256407e-05, + "loss": 1.2655, + "step": 9088 + }, + { + "epoch": 0.77, + "learning_rate": 1.7458433068640242e-05, + "loss": 1.3371, + "step": 9089 + }, + { + "epoch": 0.77, + "learning_rate": 1.7457827551400113e-05, + "loss": 1.3932, + "step": 9090 + }, + { + "epoch": 0.77, + "learning_rate": 1.745722197254103e-05, + "loss": 1.2545, + "step": 9091 + }, + { + "epoch": 0.77, + "learning_rate": 1.745661633206799e-05, + "loss": 1.2468, + "step": 9092 + }, + { + "epoch": 0.77, + "learning_rate": 1.7456010629986e-05, + "loss": 1.4162, + "step": 9093 + }, + { + "epoch": 0.77, + "learning_rate": 1.7455404866300067e-05, + "loss": 1.2236, + "step": 9094 + }, + { + "epoch": 0.77, + "learning_rate": 1.7454799041015187e-05, + "loss": 1.2282, + "step": 9095 + }, + { + "epoch": 0.77, + "learning_rate": 1.7454193154136375e-05, + "loss": 1.2882, + "step": 9096 + }, + { + "epoch": 0.77, + "learning_rate": 1.7453587205668632e-05, + "loss": 1.3609, + "step": 9097 + }, + { + "epoch": 0.77, + "learning_rate": 1.7452981195616965e-05, + "loss": 1.3629, + "step": 9098 + }, + { + "epoch": 0.77, + "learning_rate": 1.7452375123986386e-05, + "loss": 1.2987, + "step": 9099 + }, + { + "epoch": 0.77, + "learning_rate": 1.7451768990781896e-05, + "loss": 1.2514, + "step": 9100 + }, + { + "epoch": 0.77, + "learning_rate": 1.745116279600851e-05, + "loss": 1.298, + "step": 9101 + }, + { + "epoch": 0.77, + "learning_rate": 1.7450556539671225e-05, + "loss": 1.3782, + "step": 9102 + }, + { + "epoch": 0.77, + "learning_rate": 1.7449950221775063e-05, + "loss": 1.2641, + "step": 9103 + }, + { + "epoch": 0.77, + "learning_rate": 1.7449343842325026e-05, + "loss": 1.3319, + "step": 9104 + }, + { + "epoch": 0.77, + "learning_rate": 1.7448737401326127e-05, + "loss": 1.3334, + "step": 9105 + }, + { + "epoch": 0.77, + "learning_rate": 1.7448130898783376e-05, + "loss": 1.3077, + "step": 9106 + }, + { + "epoch": 0.77, + "learning_rate": 1.7447524334701784e-05, + "loss": 1.3347, + "step": 9107 + }, + { + "epoch": 0.77, + "learning_rate": 1.7446917709086358e-05, + "loss": 1.2101, + "step": 9108 + }, + { + "epoch": 0.77, + "learning_rate": 1.744631102194212e-05, + "loss": 1.3061, + "step": 9109 + }, + { + "epoch": 0.77, + "learning_rate": 1.7445704273274075e-05, + "loss": 1.3178, + "step": 9110 + }, + { + "epoch": 0.77, + "learning_rate": 1.7445097463087236e-05, + "loss": 1.2737, + "step": 9111 + }, + { + "epoch": 0.77, + "learning_rate": 1.744449059138662e-05, + "loss": 1.2163, + "step": 9112 + }, + { + "epoch": 0.77, + "learning_rate": 1.7443883658177242e-05, + "loss": 1.217, + "step": 9113 + }, + { + "epoch": 0.77, + "learning_rate": 1.7443276663464117e-05, + "loss": 1.2928, + "step": 9114 + }, + { + "epoch": 0.77, + "learning_rate": 1.7442669607252255e-05, + "loss": 1.2183, + "step": 9115 + }, + { + "epoch": 0.77, + "learning_rate": 1.7442062489546673e-05, + "loss": 1.309, + "step": 9116 + }, + { + "epoch": 0.77, + "learning_rate": 1.7441455310352393e-05, + "loss": 1.2815, + "step": 9117 + }, + { + "epoch": 0.77, + "learning_rate": 1.7440848069674425e-05, + "loss": 1.3004, + "step": 9118 + }, + { + "epoch": 0.77, + "learning_rate": 1.7440240767517793e-05, + "loss": 1.2168, + "step": 9119 + }, + { + "epoch": 0.77, + "learning_rate": 1.7439633403887503e-05, + "loss": 1.3925, + "step": 9120 + }, + { + "epoch": 0.77, + "learning_rate": 1.7439025978788585e-05, + "loss": 1.2085, + "step": 9121 + }, + { + "epoch": 0.77, + "learning_rate": 1.7438418492226053e-05, + "loss": 1.279, + "step": 9122 + }, + { + "epoch": 0.77, + "learning_rate": 1.7437810944204928e-05, + "loss": 1.2718, + "step": 9123 + }, + { + "epoch": 0.77, + "learning_rate": 1.7437203334730227e-05, + "loss": 1.31, + "step": 9124 + }, + { + "epoch": 0.77, + "learning_rate": 1.7436595663806976e-05, + "loss": 1.1576, + "step": 9125 + }, + { + "epoch": 0.77, + "learning_rate": 1.743598793144019e-05, + "loss": 1.3111, + "step": 9126 + }, + { + "epoch": 0.77, + "learning_rate": 1.7435380137634886e-05, + "loss": 1.2524, + "step": 9127 + }, + { + "epoch": 0.77, + "learning_rate": 1.74347722823961e-05, + "loss": 1.2083, + "step": 9128 + }, + { + "epoch": 0.77, + "learning_rate": 1.743416436572884e-05, + "loss": 1.2132, + "step": 9129 + }, + { + "epoch": 0.77, + "learning_rate": 1.7433556387638138e-05, + "loss": 1.2434, + "step": 9130 + }, + { + "epoch": 0.77, + "learning_rate": 1.743294834812901e-05, + "loss": 1.336, + "step": 9131 + }, + { + "epoch": 0.77, + "learning_rate": 1.743234024720649e-05, + "loss": 1.3274, + "step": 9132 + }, + { + "epoch": 0.77, + "learning_rate": 1.7431732084875593e-05, + "loss": 1.3218, + "step": 9133 + }, + { + "epoch": 0.77, + "learning_rate": 1.7431123861141344e-05, + "loss": 1.326, + "step": 9134 + }, + { + "epoch": 0.77, + "learning_rate": 1.7430515576008777e-05, + "loss": 1.295, + "step": 9135 + }, + { + "epoch": 0.77, + "learning_rate": 1.742990722948291e-05, + "loss": 1.2492, + "step": 9136 + }, + { + "epoch": 0.77, + "learning_rate": 1.7429298821568772e-05, + "loss": 1.3144, + "step": 9137 + }, + { + "epoch": 0.77, + "learning_rate": 1.7428690352271388e-05, + "loss": 1.2944, + "step": 9138 + }, + { + "epoch": 0.77, + "learning_rate": 1.7428081821595787e-05, + "loss": 1.2377, + "step": 9139 + }, + { + "epoch": 0.77, + "learning_rate": 1.7427473229546998e-05, + "loss": 1.2672, + "step": 9140 + }, + { + "epoch": 0.77, + "learning_rate": 1.7426864576130048e-05, + "loss": 1.2271, + "step": 9141 + }, + { + "epoch": 0.77, + "learning_rate": 1.7426255861349965e-05, + "loss": 1.3049, + "step": 9142 + }, + { + "epoch": 0.77, + "learning_rate": 1.742564708521178e-05, + "loss": 1.3246, + "step": 9143 + }, + { + "epoch": 0.77, + "learning_rate": 1.7425038247720522e-05, + "loss": 1.2564, + "step": 9144 + }, + { + "epoch": 0.77, + "learning_rate": 1.742442934888122e-05, + "loss": 1.3157, + "step": 9145 + }, + { + "epoch": 0.77, + "learning_rate": 1.742382038869891e-05, + "loss": 1.2727, + "step": 9146 + }, + { + "epoch": 0.77, + "learning_rate": 1.7423211367178618e-05, + "loss": 1.311, + "step": 9147 + }, + { + "epoch": 0.77, + "learning_rate": 1.742260228432538e-05, + "loss": 1.2293, + "step": 9148 + }, + { + "epoch": 0.77, + "learning_rate": 1.742199314014422e-05, + "loss": 1.368, + "step": 9149 + }, + { + "epoch": 0.77, + "learning_rate": 1.7421383934640185e-05, + "loss": 1.2799, + "step": 9150 + }, + { + "epoch": 0.77, + "learning_rate": 1.7420774667818295e-05, + "loss": 1.2998, + "step": 9151 + }, + { + "epoch": 0.77, + "learning_rate": 1.7420165339683594e-05, + "loss": 1.247, + "step": 9152 + }, + { + "epoch": 0.77, + "learning_rate": 1.7419555950241112e-05, + "loss": 1.3366, + "step": 9153 + }, + { + "epoch": 0.77, + "learning_rate": 1.741894649949588e-05, + "loss": 1.2449, + "step": 9154 + }, + { + "epoch": 0.77, + "learning_rate": 1.7418336987452945e-05, + "loss": 1.3215, + "step": 9155 + }, + { + "epoch": 0.77, + "learning_rate": 1.7417727414117327e-05, + "loss": 1.3289, + "step": 9156 + }, + { + "epoch": 0.77, + "learning_rate": 1.7417117779494077e-05, + "loss": 1.328, + "step": 9157 + }, + { + "epoch": 0.77, + "learning_rate": 1.7416508083588224e-05, + "loss": 1.3277, + "step": 9158 + }, + { + "epoch": 0.77, + "learning_rate": 1.741589832640481e-05, + "loss": 1.2386, + "step": 9159 + }, + { + "epoch": 0.77, + "learning_rate": 1.741528850794887e-05, + "loss": 1.3566, + "step": 9160 + }, + { + "epoch": 0.77, + "learning_rate": 1.741467862822544e-05, + "loss": 1.3468, + "step": 9161 + }, + { + "epoch": 0.77, + "learning_rate": 1.7414068687239564e-05, + "loss": 1.4018, + "step": 9162 + }, + { + "epoch": 0.77, + "learning_rate": 1.741345868499628e-05, + "loss": 1.2764, + "step": 9163 + }, + { + "epoch": 0.77, + "learning_rate": 1.7412848621500625e-05, + "loss": 1.2519, + "step": 9164 + }, + { + "epoch": 0.77, + "learning_rate": 1.7412238496757644e-05, + "loss": 1.184, + "step": 9165 + }, + { + "epoch": 0.77, + "learning_rate": 1.7411628310772378e-05, + "loss": 1.209, + "step": 9166 + }, + { + "epoch": 0.77, + "learning_rate": 1.7411018063549866e-05, + "loss": 1.3305, + "step": 9167 + }, + { + "epoch": 0.77, + "learning_rate": 1.7410407755095152e-05, + "loss": 1.2307, + "step": 9168 + }, + { + "epoch": 0.77, + "learning_rate": 1.7409797385413275e-05, + "loss": 1.2551, + "step": 9169 + }, + { + "epoch": 0.77, + "learning_rate": 1.7409186954509283e-05, + "loss": 1.252, + "step": 9170 + }, + { + "epoch": 0.77, + "learning_rate": 1.7408576462388217e-05, + "loss": 1.2633, + "step": 9171 + }, + { + "epoch": 0.77, + "learning_rate": 1.7407965909055116e-05, + "loss": 1.2999, + "step": 9172 + }, + { + "epoch": 0.77, + "learning_rate": 1.7407355294515035e-05, + "loss": 1.2944, + "step": 9173 + }, + { + "epoch": 0.77, + "learning_rate": 1.7406744618773015e-05, + "loss": 1.2573, + "step": 9174 + }, + { + "epoch": 0.77, + "learning_rate": 1.74061338818341e-05, + "loss": 1.2541, + "step": 9175 + }, + { + "epoch": 0.77, + "learning_rate": 1.7405523083703333e-05, + "loss": 1.3363, + "step": 9176 + }, + { + "epoch": 0.77, + "learning_rate": 1.7404912224385766e-05, + "loss": 1.3694, + "step": 9177 + }, + { + "epoch": 0.77, + "learning_rate": 1.7404301303886444e-05, + "loss": 1.2841, + "step": 9178 + }, + { + "epoch": 0.77, + "learning_rate": 1.740369032221042e-05, + "loss": 1.224, + "step": 9179 + }, + { + "epoch": 0.77, + "learning_rate": 1.740307927936273e-05, + "loss": 1.2583, + "step": 9180 + }, + { + "epoch": 0.77, + "learning_rate": 1.740246817534843e-05, + "loss": 1.3461, + "step": 9181 + }, + { + "epoch": 0.77, + "learning_rate": 1.740185701017257e-05, + "loss": 1.2196, + "step": 9182 + }, + { + "epoch": 0.77, + "learning_rate": 1.7401245783840202e-05, + "loss": 1.286, + "step": 9183 + }, + { + "epoch": 0.77, + "learning_rate": 1.740063449635637e-05, + "loss": 1.3319, + "step": 9184 + }, + { + "epoch": 0.77, + "learning_rate": 1.740002314772613e-05, + "loss": 1.2263, + "step": 9185 + }, + { + "epoch": 0.77, + "learning_rate": 1.7399411737954526e-05, + "loss": 1.2906, + "step": 9186 + }, + { + "epoch": 0.77, + "learning_rate": 1.7398800267046615e-05, + "loss": 1.3743, + "step": 9187 + }, + { + "epoch": 0.77, + "learning_rate": 1.739818873500745e-05, + "loss": 1.4161, + "step": 9188 + }, + { + "epoch": 0.77, + "learning_rate": 1.7397577141842082e-05, + "loss": 1.2934, + "step": 9189 + }, + { + "epoch": 0.77, + "learning_rate": 1.739696548755556e-05, + "loss": 1.4375, + "step": 9190 + }, + { + "epoch": 0.77, + "learning_rate": 1.739635377215295e-05, + "loss": 1.292, + "step": 9191 + }, + { + "epoch": 0.77, + "learning_rate": 1.7395741995639293e-05, + "loss": 1.2823, + "step": 9192 + }, + { + "epoch": 0.77, + "learning_rate": 1.739513015801965e-05, + "loss": 1.2674, + "step": 9193 + }, + { + "epoch": 0.77, + "learning_rate": 1.7394518259299072e-05, + "loss": 1.3233, + "step": 9194 + }, + { + "epoch": 0.77, + "learning_rate": 1.739390629948262e-05, + "loss": 1.2809, + "step": 9195 + }, + { + "epoch": 0.77, + "learning_rate": 1.7393294278575345e-05, + "loss": 1.2933, + "step": 9196 + }, + { + "epoch": 0.77, + "learning_rate": 1.7392682196582308e-05, + "loss": 1.309, + "step": 9197 + }, + { + "epoch": 0.77, + "learning_rate": 1.7392070053508566e-05, + "loss": 1.4073, + "step": 9198 + }, + { + "epoch": 0.77, + "learning_rate": 1.7391457849359178e-05, + "loss": 1.315, + "step": 9199 + }, + { + "epoch": 0.77, + "learning_rate": 1.7390845584139195e-05, + "loss": 1.2171, + "step": 9200 + }, + { + "epoch": 0.77, + "learning_rate": 1.739023325785368e-05, + "loss": 1.2817, + "step": 9201 + }, + { + "epoch": 0.77, + "learning_rate": 1.7389620870507698e-05, + "loss": 1.3367, + "step": 9202 + }, + { + "epoch": 0.77, + "learning_rate": 1.73890084221063e-05, + "loss": 1.2743, + "step": 9203 + }, + { + "epoch": 0.77, + "learning_rate": 1.738839591265455e-05, + "loss": 1.2844, + "step": 9204 + }, + { + "epoch": 0.77, + "learning_rate": 1.7387783342157505e-05, + "loss": 1.2633, + "step": 9205 + }, + { + "epoch": 0.78, + "learning_rate": 1.7387170710620232e-05, + "loss": 1.2635, + "step": 9206 + }, + { + "epoch": 0.78, + "learning_rate": 1.738655801804779e-05, + "loss": 1.2601, + "step": 9207 + }, + { + "epoch": 0.78, + "learning_rate": 1.7385945264445247e-05, + "loss": 1.3415, + "step": 9208 + }, + { + "epoch": 0.78, + "learning_rate": 1.7385332449817655e-05, + "loss": 1.2414, + "step": 9209 + }, + { + "epoch": 0.78, + "learning_rate": 1.7384719574170087e-05, + "loss": 1.4203, + "step": 9210 + }, + { + "epoch": 0.78, + "learning_rate": 1.7384106637507596e-05, + "loss": 1.266, + "step": 9211 + }, + { + "epoch": 0.78, + "learning_rate": 1.7383493639835256e-05, + "loss": 1.2664, + "step": 9212 + }, + { + "epoch": 0.78, + "learning_rate": 1.738288058115813e-05, + "loss": 1.2746, + "step": 9213 + }, + { + "epoch": 0.78, + "learning_rate": 1.738226746148128e-05, + "loss": 1.201, + "step": 9214 + }, + { + "epoch": 0.78, + "learning_rate": 1.7381654280809773e-05, + "loss": 1.362, + "step": 9215 + }, + { + "epoch": 0.78, + "learning_rate": 1.7381041039148678e-05, + "loss": 1.3267, + "step": 9216 + }, + { + "epoch": 0.78, + "learning_rate": 1.7380427736503057e-05, + "loss": 1.2631, + "step": 9217 + }, + { + "epoch": 0.78, + "learning_rate": 1.7379814372877982e-05, + "loss": 1.2385, + "step": 9218 + }, + { + "epoch": 0.78, + "learning_rate": 1.737920094827852e-05, + "loss": 1.3721, + "step": 9219 + }, + { + "epoch": 0.78, + "learning_rate": 1.7378587462709736e-05, + "loss": 1.2579, + "step": 9220 + }, + { + "epoch": 0.78, + "learning_rate": 1.73779739161767e-05, + "loss": 1.2469, + "step": 9221 + }, + { + "epoch": 0.78, + "learning_rate": 1.7377360308684483e-05, + "loss": 1.1996, + "step": 9222 + }, + { + "epoch": 0.78, + "learning_rate": 1.737674664023816e-05, + "loss": 1.2553, + "step": 9223 + }, + { + "epoch": 0.78, + "learning_rate": 1.7376132910842787e-05, + "loss": 1.2968, + "step": 9224 + }, + { + "epoch": 0.78, + "learning_rate": 1.7375519120503446e-05, + "loss": 1.2554, + "step": 9225 + }, + { + "epoch": 0.78, + "learning_rate": 1.7374905269225206e-05, + "loss": 1.1556, + "step": 9226 + }, + { + "epoch": 0.78, + "learning_rate": 1.737429135701314e-05, + "loss": 1.2747, + "step": 9227 + }, + { + "epoch": 0.78, + "learning_rate": 1.7373677383872316e-05, + "loss": 1.2921, + "step": 9228 + }, + { + "epoch": 0.78, + "learning_rate": 1.737306334980781e-05, + "loss": 1.2714, + "step": 9229 + }, + { + "epoch": 0.78, + "learning_rate": 1.7372449254824696e-05, + "loss": 1.2327, + "step": 9230 + }, + { + "epoch": 0.78, + "learning_rate": 1.7371835098928045e-05, + "loss": 1.3132, + "step": 9231 + }, + { + "epoch": 0.78, + "learning_rate": 1.7371220882122936e-05, + "loss": 1.2208, + "step": 9232 + }, + { + "epoch": 0.78, + "learning_rate": 1.737060660441444e-05, + "loss": 1.3477, + "step": 9233 + }, + { + "epoch": 0.78, + "learning_rate": 1.7369992265807632e-05, + "loss": 1.2493, + "step": 9234 + }, + { + "epoch": 0.78, + "learning_rate": 1.736937786630759e-05, + "loss": 1.2601, + "step": 9235 + }, + { + "epoch": 0.78, + "learning_rate": 1.736876340591939e-05, + "loss": 1.2077, + "step": 9236 + }, + { + "epoch": 0.78, + "learning_rate": 1.7368148884648107e-05, + "loss": 1.1889, + "step": 9237 + }, + { + "epoch": 0.78, + "learning_rate": 1.7367534302498822e-05, + "loss": 1.2704, + "step": 9238 + }, + { + "epoch": 0.78, + "learning_rate": 1.736691965947661e-05, + "loss": 1.3475, + "step": 9239 + }, + { + "epoch": 0.78, + "learning_rate": 1.736630495558655e-05, + "loss": 1.2788, + "step": 9240 + }, + { + "epoch": 0.78, + "learning_rate": 1.736569019083372e-05, + "loss": 1.3099, + "step": 9241 + }, + { + "epoch": 0.78, + "learning_rate": 1.73650753652232e-05, + "loss": 1.2499, + "step": 9242 + }, + { + "epoch": 0.78, + "learning_rate": 1.736446047876007e-05, + "loss": 1.4007, + "step": 9243 + }, + { + "epoch": 0.78, + "learning_rate": 1.736384553144941e-05, + "loss": 1.2781, + "step": 9244 + }, + { + "epoch": 0.78, + "learning_rate": 1.7363230523296306e-05, + "loss": 1.2696, + "step": 9245 + }, + { + "epoch": 0.78, + "learning_rate": 1.736261545430583e-05, + "loss": 1.2922, + "step": 9246 + }, + { + "epoch": 0.78, + "learning_rate": 1.7362000324483073e-05, + "loss": 1.2315, + "step": 9247 + }, + { + "epoch": 0.78, + "learning_rate": 1.736138513383311e-05, + "loss": 1.3005, + "step": 9248 + }, + { + "epoch": 0.78, + "learning_rate": 1.7360769882361028e-05, + "loss": 1.2218, + "step": 9249 + }, + { + "epoch": 0.78, + "learning_rate": 1.736015457007191e-05, + "loss": 1.2012, + "step": 9250 + }, + { + "epoch": 0.78, + "learning_rate": 1.735953919697084e-05, + "loss": 1.3734, + "step": 9251 + }, + { + "epoch": 0.78, + "learning_rate": 1.73589237630629e-05, + "loss": 1.2189, + "step": 9252 + }, + { + "epoch": 0.78, + "learning_rate": 1.7358308268353178e-05, + "loss": 1.3015, + "step": 9253 + }, + { + "epoch": 0.78, + "learning_rate": 1.7357692712846756e-05, + "loss": 1.1964, + "step": 9254 + }, + { + "epoch": 0.78, + "learning_rate": 1.7357077096548726e-05, + "loss": 1.3677, + "step": 9255 + }, + { + "epoch": 0.78, + "learning_rate": 1.7356461419464165e-05, + "loss": 1.2611, + "step": 9256 + }, + { + "epoch": 0.78, + "learning_rate": 1.735584568159817e-05, + "loss": 1.3023, + "step": 9257 + }, + { + "epoch": 0.78, + "learning_rate": 1.7355229882955823e-05, + "loss": 1.2628, + "step": 9258 + }, + { + "epoch": 0.78, + "learning_rate": 1.735461402354221e-05, + "loss": 1.3049, + "step": 9259 + }, + { + "epoch": 0.78, + "learning_rate": 1.7353998103362422e-05, + "loss": 1.2308, + "step": 9260 + }, + { + "epoch": 0.78, + "learning_rate": 1.7353382122421553e-05, + "loss": 1.2881, + "step": 9261 + }, + { + "epoch": 0.78, + "learning_rate": 1.7352766080724684e-05, + "loss": 1.3038, + "step": 9262 + }, + { + "epoch": 0.78, + "learning_rate": 1.7352149978276908e-05, + "loss": 1.332, + "step": 9263 + }, + { + "epoch": 0.78, + "learning_rate": 1.735153381508332e-05, + "loss": 1.2437, + "step": 9264 + }, + { + "epoch": 0.78, + "learning_rate": 1.7350917591149e-05, + "loss": 1.3193, + "step": 9265 + }, + { + "epoch": 0.78, + "learning_rate": 1.7350301306479053e-05, + "loss": 1.3161, + "step": 9266 + }, + { + "epoch": 0.78, + "learning_rate": 1.734968496107856e-05, + "loss": 1.2439, + "step": 9267 + }, + { + "epoch": 0.78, + "learning_rate": 1.7349068554952618e-05, + "loss": 1.2696, + "step": 9268 + }, + { + "epoch": 0.78, + "learning_rate": 1.7348452088106325e-05, + "loss": 1.2376, + "step": 9269 + }, + { + "epoch": 0.78, + "learning_rate": 1.734783556054476e-05, + "loss": 1.3482, + "step": 9270 + }, + { + "epoch": 0.78, + "learning_rate": 1.734721897227303e-05, + "loss": 1.3068, + "step": 9271 + }, + { + "epoch": 0.78, + "learning_rate": 1.734660232329623e-05, + "loss": 1.3503, + "step": 9272 + }, + { + "epoch": 0.78, + "learning_rate": 1.7345985613619448e-05, + "loss": 1.2351, + "step": 9273 + }, + { + "epoch": 0.78, + "learning_rate": 1.7345368843247778e-05, + "loss": 1.2905, + "step": 9274 + }, + { + "epoch": 0.78, + "learning_rate": 1.7344752012186323e-05, + "loss": 1.2902, + "step": 9275 + }, + { + "epoch": 0.78, + "learning_rate": 1.7344135120440175e-05, + "loss": 1.2931, + "step": 9276 + }, + { + "epoch": 0.78, + "learning_rate": 1.7343518168014433e-05, + "loss": 1.2692, + "step": 9277 + }, + { + "epoch": 0.78, + "learning_rate": 1.7342901154914194e-05, + "loss": 1.2833, + "step": 9278 + }, + { + "epoch": 0.78, + "learning_rate": 1.734228408114455e-05, + "loss": 1.2497, + "step": 9279 + }, + { + "epoch": 0.78, + "learning_rate": 1.7341666946710613e-05, + "loss": 1.2322, + "step": 9280 + }, + { + "epoch": 0.78, + "learning_rate": 1.734104975161747e-05, + "loss": 1.3128, + "step": 9281 + }, + { + "epoch": 0.78, + "learning_rate": 1.7340432495870226e-05, + "loss": 1.3292, + "step": 9282 + }, + { + "epoch": 0.78, + "learning_rate": 1.733981517947398e-05, + "loss": 1.3233, + "step": 9283 + }, + { + "epoch": 0.78, + "learning_rate": 1.733919780243383e-05, + "loss": 1.2888, + "step": 9284 + }, + { + "epoch": 0.78, + "learning_rate": 1.7338580364754883e-05, + "loss": 1.2989, + "step": 9285 + }, + { + "epoch": 0.78, + "learning_rate": 1.733796286644223e-05, + "loss": 1.2991, + "step": 9286 + }, + { + "epoch": 0.78, + "learning_rate": 1.7337345307500984e-05, + "loss": 1.2213, + "step": 9287 + }, + { + "epoch": 0.78, + "learning_rate": 1.733672768793624e-05, + "loss": 1.2762, + "step": 9288 + }, + { + "epoch": 0.78, + "learning_rate": 1.7336110007753105e-05, + "loss": 1.147, + "step": 9289 + }, + { + "epoch": 0.78, + "learning_rate": 1.7335492266956682e-05, + "loss": 1.2284, + "step": 9290 + }, + { + "epoch": 0.78, + "learning_rate": 1.7334874465552074e-05, + "loss": 1.34, + "step": 9291 + }, + { + "epoch": 0.78, + "learning_rate": 1.7334256603544383e-05, + "loss": 1.2531, + "step": 9292 + }, + { + "epoch": 0.78, + "learning_rate": 1.733363868093872e-05, + "loss": 1.3111, + "step": 9293 + }, + { + "epoch": 0.78, + "learning_rate": 1.7333020697740188e-05, + "loss": 1.2216, + "step": 9294 + }, + { + "epoch": 0.78, + "learning_rate": 1.733240265395389e-05, + "loss": 1.3291, + "step": 9295 + }, + { + "epoch": 0.78, + "learning_rate": 1.7331784549584936e-05, + "loss": 1.2843, + "step": 9296 + }, + { + "epoch": 0.78, + "learning_rate": 1.733116638463843e-05, + "loss": 1.2127, + "step": 9297 + }, + { + "epoch": 0.78, + "learning_rate": 1.733054815911948e-05, + "loss": 1.2214, + "step": 9298 + }, + { + "epoch": 0.78, + "learning_rate": 1.7329929873033197e-05, + "loss": 1.3872, + "step": 9299 + }, + { + "epoch": 0.78, + "learning_rate": 1.732931152638469e-05, + "loss": 1.2753, + "step": 9300 + }, + { + "epoch": 0.78, + "learning_rate": 1.732869311917906e-05, + "loss": 1.3225, + "step": 9301 + }, + { + "epoch": 0.78, + "learning_rate": 1.7328074651421425e-05, + "loss": 1.3061, + "step": 9302 + }, + { + "epoch": 0.78, + "learning_rate": 1.7327456123116892e-05, + "loss": 1.2604, + "step": 9303 + }, + { + "epoch": 0.78, + "learning_rate": 1.732683753427057e-05, + "loss": 1.3185, + "step": 9304 + }, + { + "epoch": 0.78, + "learning_rate": 1.7326218884887574e-05, + "loss": 1.3542, + "step": 9305 + }, + { + "epoch": 0.78, + "learning_rate": 1.732560017497301e-05, + "loss": 1.202, + "step": 9306 + }, + { + "epoch": 0.78, + "learning_rate": 1.7324981404531994e-05, + "loss": 1.264, + "step": 9307 + }, + { + "epoch": 0.78, + "learning_rate": 1.7324362573569637e-05, + "loss": 1.3147, + "step": 9308 + }, + { + "epoch": 0.78, + "learning_rate": 1.7323743682091054e-05, + "loss": 1.1528, + "step": 9309 + }, + { + "epoch": 0.78, + "learning_rate": 1.7323124730101356e-05, + "loss": 1.3105, + "step": 9310 + }, + { + "epoch": 0.78, + "learning_rate": 1.7322505717605658e-05, + "loss": 1.3062, + "step": 9311 + }, + { + "epoch": 0.78, + "learning_rate": 1.732188664460907e-05, + "loss": 1.3197, + "step": 9312 + }, + { + "epoch": 0.78, + "learning_rate": 1.732126751111672e-05, + "loss": 1.2314, + "step": 9313 + }, + { + "epoch": 0.78, + "learning_rate": 1.7320648317133705e-05, + "loss": 1.3256, + "step": 9314 + }, + { + "epoch": 0.78, + "learning_rate": 1.7320029062665157e-05, + "loss": 1.4157, + "step": 9315 + }, + { + "epoch": 0.78, + "learning_rate": 1.7319409747716186e-05, + "loss": 1.3146, + "step": 9316 + }, + { + "epoch": 0.78, + "learning_rate": 1.7318790372291906e-05, + "loss": 1.2551, + "step": 9317 + }, + { + "epoch": 0.78, + "learning_rate": 1.731817093639744e-05, + "loss": 1.3166, + "step": 9318 + }, + { + "epoch": 0.78, + "learning_rate": 1.73175514400379e-05, + "loss": 1.2777, + "step": 9319 + }, + { + "epoch": 0.78, + "learning_rate": 1.7316931883218413e-05, + "loss": 1.2943, + "step": 9320 + }, + { + "epoch": 0.78, + "learning_rate": 1.731631226594409e-05, + "loss": 1.3138, + "step": 9321 + }, + { + "epoch": 0.78, + "learning_rate": 1.7315692588220057e-05, + "loss": 1.3126, + "step": 9322 + }, + { + "epoch": 0.78, + "learning_rate": 1.7315072850051427e-05, + "loss": 1.2505, + "step": 9323 + }, + { + "epoch": 0.79, + "learning_rate": 1.7314453051443326e-05, + "loss": 1.2082, + "step": 9324 + }, + { + "epoch": 0.79, + "learning_rate": 1.731383319240087e-05, + "loss": 1.2964, + "step": 9325 + }, + { + "epoch": 0.79, + "learning_rate": 1.7313213272929185e-05, + "loss": 1.2168, + "step": 9326 + }, + { + "epoch": 0.79, + "learning_rate": 1.7312593293033394e-05, + "loss": 1.2889, + "step": 9327 + }, + { + "epoch": 0.79, + "learning_rate": 1.7311973252718615e-05, + "loss": 1.2753, + "step": 9328 + }, + { + "epoch": 0.79, + "learning_rate": 1.7311353151989972e-05, + "loss": 1.2705, + "step": 9329 + }, + { + "epoch": 0.79, + "learning_rate": 1.731073299085259e-05, + "loss": 1.3606, + "step": 9330 + }, + { + "epoch": 0.79, + "learning_rate": 1.731011276931159e-05, + "loss": 1.3022, + "step": 9331 + }, + { + "epoch": 0.79, + "learning_rate": 1.7309492487372105e-05, + "loss": 1.341, + "step": 9332 + }, + { + "epoch": 0.79, + "learning_rate": 1.7308872145039247e-05, + "loss": 1.2874, + "step": 9333 + }, + { + "epoch": 0.79, + "learning_rate": 1.730825174231815e-05, + "loss": 1.1912, + "step": 9334 + }, + { + "epoch": 0.79, + "learning_rate": 1.730763127921394e-05, + "loss": 1.2672, + "step": 9335 + }, + { + "epoch": 0.79, + "learning_rate": 1.730701075573174e-05, + "loss": 1.3016, + "step": 9336 + }, + { + "epoch": 0.79, + "learning_rate": 1.730639017187668e-05, + "loss": 1.2972, + "step": 9337 + }, + { + "epoch": 0.79, + "learning_rate": 1.7305769527653887e-05, + "loss": 1.2687, + "step": 9338 + }, + { + "epoch": 0.79, + "learning_rate": 1.7305148823068484e-05, + "loss": 1.403, + "step": 9339 + }, + { + "epoch": 0.79, + "learning_rate": 1.7304528058125603e-05, + "loss": 1.3048, + "step": 9340 + }, + { + "epoch": 0.79, + "learning_rate": 1.7303907232830378e-05, + "loss": 1.3381, + "step": 9341 + }, + { + "epoch": 0.79, + "learning_rate": 1.7303286347187932e-05, + "loss": 1.2476, + "step": 9342 + }, + { + "epoch": 0.79, + "learning_rate": 1.73026654012034e-05, + "loss": 1.253, + "step": 9343 + }, + { + "epoch": 0.79, + "learning_rate": 1.7302044394881907e-05, + "loss": 1.2555, + "step": 9344 + }, + { + "epoch": 0.79, + "learning_rate": 1.7301423328228584e-05, + "loss": 1.2796, + "step": 9345 + }, + { + "epoch": 0.79, + "learning_rate": 1.7300802201248568e-05, + "loss": 1.2721, + "step": 9346 + }, + { + "epoch": 0.79, + "learning_rate": 1.7300181013946982e-05, + "loss": 1.2279, + "step": 9347 + }, + { + "epoch": 0.79, + "learning_rate": 1.729955976632897e-05, + "loss": 1.3105, + "step": 9348 + }, + { + "epoch": 0.79, + "learning_rate": 1.7298938458399657e-05, + "loss": 1.2588, + "step": 9349 + }, + { + "epoch": 0.79, + "learning_rate": 1.729831709016418e-05, + "loss": 1.3792, + "step": 9350 + }, + { + "epoch": 0.79, + "learning_rate": 1.729769566162767e-05, + "loss": 1.2851, + "step": 9351 + }, + { + "epoch": 0.79, + "learning_rate": 1.729707417279526e-05, + "loss": 1.2198, + "step": 9352 + }, + { + "epoch": 0.79, + "learning_rate": 1.7296452623672092e-05, + "loss": 1.3189, + "step": 9353 + }, + { + "epoch": 0.79, + "learning_rate": 1.7295831014263296e-05, + "loss": 1.2097, + "step": 9354 + }, + { + "epoch": 0.79, + "learning_rate": 1.7295209344574007e-05, + "loss": 1.2455, + "step": 9355 + }, + { + "epoch": 0.79, + "learning_rate": 1.729458761460937e-05, + "loss": 1.2758, + "step": 9356 + }, + { + "epoch": 0.79, + "learning_rate": 1.729396582437451e-05, + "loss": 1.2962, + "step": 9357 + }, + { + "epoch": 0.79, + "learning_rate": 1.729334397387457e-05, + "loss": 1.2835, + "step": 9358 + }, + { + "epoch": 0.79, + "learning_rate": 1.729272206311469e-05, + "loss": 1.3277, + "step": 9359 + }, + { + "epoch": 0.79, + "learning_rate": 1.7292100092100007e-05, + "loss": 1.2932, + "step": 9360 + }, + { + "epoch": 0.79, + "learning_rate": 1.7291478060835656e-05, + "loss": 1.2563, + "step": 9361 + }, + { + "epoch": 0.79, + "learning_rate": 1.7290855969326783e-05, + "loss": 1.1571, + "step": 9362 + }, + { + "epoch": 0.79, + "learning_rate": 1.7290233817578525e-05, + "loss": 1.1681, + "step": 9363 + }, + { + "epoch": 0.79, + "learning_rate": 1.7289611605596016e-05, + "loss": 1.2829, + "step": 9364 + }, + { + "epoch": 0.79, + "learning_rate": 1.728898933338441e-05, + "loss": 1.269, + "step": 9365 + }, + { + "epoch": 0.79, + "learning_rate": 1.7288367000948838e-05, + "loss": 1.3086, + "step": 9366 + }, + { + "epoch": 0.79, + "learning_rate": 1.7287744608294445e-05, + "loss": 1.384, + "step": 9367 + }, + { + "epoch": 0.79, + "learning_rate": 1.7287122155426374e-05, + "loss": 1.3102, + "step": 9368 + }, + { + "epoch": 0.79, + "learning_rate": 1.728649964234977e-05, + "loss": 1.3529, + "step": 9369 + }, + { + "epoch": 0.79, + "learning_rate": 1.728587706906977e-05, + "loss": 1.3857, + "step": 9370 + }, + { + "epoch": 0.79, + "learning_rate": 1.7285254435591524e-05, + "loss": 1.3145, + "step": 9371 + }, + { + "epoch": 0.79, + "learning_rate": 1.7284631741920175e-05, + "loss": 1.3151, + "step": 9372 + }, + { + "epoch": 0.79, + "learning_rate": 1.7284008988060868e-05, + "loss": 1.2858, + "step": 9373 + }, + { + "epoch": 0.79, + "learning_rate": 1.7283386174018744e-05, + "loss": 1.2821, + "step": 9374 + }, + { + "epoch": 0.79, + "learning_rate": 1.7282763299798955e-05, + "loss": 1.2386, + "step": 9375 + }, + { + "epoch": 0.79, + "learning_rate": 1.7282140365406643e-05, + "loss": 1.2777, + "step": 9376 + }, + { + "epoch": 0.79, + "learning_rate": 1.728151737084696e-05, + "loss": 1.2863, + "step": 9377 + }, + { + "epoch": 0.79, + "learning_rate": 1.728089431612505e-05, + "loss": 1.3238, + "step": 9378 + }, + { + "epoch": 0.79, + "learning_rate": 1.7280271201246055e-05, + "loss": 1.3277, + "step": 9379 + }, + { + "epoch": 0.79, + "learning_rate": 1.7279648026215135e-05, + "loss": 1.2985, + "step": 9380 + }, + { + "epoch": 0.79, + "learning_rate": 1.727902479103743e-05, + "loss": 1.2168, + "step": 9381 + }, + { + "epoch": 0.79, + "learning_rate": 1.7278401495718094e-05, + "loss": 1.3597, + "step": 9382 + }, + { + "epoch": 0.79, + "learning_rate": 1.7277778140262275e-05, + "loss": 1.3058, + "step": 9383 + }, + { + "epoch": 0.79, + "learning_rate": 1.7277154724675127e-05, + "loss": 1.2557, + "step": 9384 + }, + { + "epoch": 0.79, + "learning_rate": 1.7276531248961795e-05, + "loss": 1.1798, + "step": 9385 + }, + { + "epoch": 0.79, + "learning_rate": 1.7275907713127434e-05, + "loss": 1.2827, + "step": 9386 + }, + { + "epoch": 0.79, + "learning_rate": 1.7275284117177193e-05, + "loss": 1.3363, + "step": 9387 + }, + { + "epoch": 0.79, + "learning_rate": 1.7274660461116225e-05, + "loss": 1.2352, + "step": 9388 + }, + { + "epoch": 0.79, + "learning_rate": 1.7274036744949688e-05, + "loss": 1.389, + "step": 9389 + }, + { + "epoch": 0.79, + "learning_rate": 1.7273412968682727e-05, + "loss": 1.3554, + "step": 9390 + }, + { + "epoch": 0.79, + "learning_rate": 1.7272789132320507e-05, + "loss": 1.2895, + "step": 9391 + }, + { + "epoch": 0.79, + "learning_rate": 1.727216523586817e-05, + "loss": 1.2671, + "step": 9392 + }, + { + "epoch": 0.79, + "learning_rate": 1.7271541279330877e-05, + "loss": 1.2817, + "step": 9393 + }, + { + "epoch": 0.79, + "learning_rate": 1.7270917262713785e-05, + "loss": 1.3079, + "step": 9394 + }, + { + "epoch": 0.79, + "learning_rate": 1.7270293186022046e-05, + "loss": 1.2666, + "step": 9395 + }, + { + "epoch": 0.79, + "learning_rate": 1.7269669049260817e-05, + "loss": 1.1782, + "step": 9396 + }, + { + "epoch": 0.79, + "learning_rate": 1.7269044852435256e-05, + "loss": 1.3166, + "step": 9397 + }, + { + "epoch": 0.79, + "learning_rate": 1.7268420595550522e-05, + "loss": 1.2462, + "step": 9398 + }, + { + "epoch": 0.79, + "learning_rate": 1.726779627861177e-05, + "loss": 1.3453, + "step": 9399 + }, + { + "epoch": 0.79, + "learning_rate": 1.726717190162416e-05, + "loss": 1.3729, + "step": 9400 + }, + { + "epoch": 0.79, + "learning_rate": 1.726654746459285e-05, + "loss": 1.3268, + "step": 9401 + }, + { + "epoch": 0.79, + "learning_rate": 1.7265922967522998e-05, + "loss": 1.4105, + "step": 9402 + }, + { + "epoch": 0.79, + "learning_rate": 1.7265298410419765e-05, + "loss": 1.3067, + "step": 9403 + }, + { + "epoch": 0.79, + "learning_rate": 1.726467379328831e-05, + "loss": 1.2971, + "step": 9404 + }, + { + "epoch": 0.79, + "learning_rate": 1.7264049116133797e-05, + "loss": 1.3406, + "step": 9405 + }, + { + "epoch": 0.79, + "learning_rate": 1.7263424378961386e-05, + "loss": 1.3591, + "step": 9406 + }, + { + "epoch": 0.79, + "learning_rate": 1.7262799581776236e-05, + "loss": 1.2596, + "step": 9407 + }, + { + "epoch": 0.79, + "learning_rate": 1.7262174724583514e-05, + "loss": 1.2727, + "step": 9408 + }, + { + "epoch": 0.79, + "learning_rate": 1.726154980738838e-05, + "loss": 1.2765, + "step": 9409 + }, + { + "epoch": 0.79, + "learning_rate": 1.7260924830195994e-05, + "loss": 1.2688, + "step": 9410 + }, + { + "epoch": 0.79, + "learning_rate": 1.726029979301153e-05, + "loss": 1.3155, + "step": 9411 + }, + { + "epoch": 0.79, + "learning_rate": 1.725967469584014e-05, + "loss": 1.3272, + "step": 9412 + }, + { + "epoch": 0.79, + "learning_rate": 1.7259049538686994e-05, + "loss": 1.2537, + "step": 9413 + }, + { + "epoch": 0.79, + "learning_rate": 1.7258424321557257e-05, + "loss": 1.3499, + "step": 9414 + }, + { + "epoch": 0.79, + "learning_rate": 1.7257799044456097e-05, + "loss": 1.2486, + "step": 9415 + }, + { + "epoch": 0.79, + "learning_rate": 1.725717370738868e-05, + "loss": 1.2631, + "step": 9416 + }, + { + "epoch": 0.79, + "learning_rate": 1.725654831036017e-05, + "loss": 1.2236, + "step": 9417 + }, + { + "epoch": 0.79, + "learning_rate": 1.725592285337574e-05, + "loss": 1.3547, + "step": 9418 + }, + { + "epoch": 0.79, + "learning_rate": 1.7255297336440544e-05, + "loss": 1.353, + "step": 9419 + }, + { + "epoch": 0.79, + "learning_rate": 1.7254671759559767e-05, + "loss": 1.2513, + "step": 9420 + }, + { + "epoch": 0.79, + "learning_rate": 1.7254046122738564e-05, + "loss": 1.2726, + "step": 9421 + }, + { + "epoch": 0.79, + "learning_rate": 1.7253420425982117e-05, + "loss": 1.2704, + "step": 9422 + }, + { + "epoch": 0.79, + "learning_rate": 1.7252794669295584e-05, + "loss": 1.4033, + "step": 9423 + }, + { + "epoch": 0.79, + "learning_rate": 1.7252168852684144e-05, + "loss": 1.4281, + "step": 9424 + }, + { + "epoch": 0.79, + "learning_rate": 1.725154297615296e-05, + "loss": 1.3222, + "step": 9425 + }, + { + "epoch": 0.79, + "learning_rate": 1.7250917039707208e-05, + "loss": 1.2188, + "step": 9426 + }, + { + "epoch": 0.79, + "learning_rate": 1.725029104335206e-05, + "loss": 1.2218, + "step": 9427 + }, + { + "epoch": 0.79, + "learning_rate": 1.7249664987092688e-05, + "loss": 1.391, + "step": 9428 + }, + { + "epoch": 0.79, + "learning_rate": 1.7249038870934263e-05, + "loss": 1.3076, + "step": 9429 + }, + { + "epoch": 0.79, + "learning_rate": 1.724841269488196e-05, + "loss": 1.2335, + "step": 9430 + }, + { + "epoch": 0.79, + "learning_rate": 1.724778645894095e-05, + "loss": 1.3256, + "step": 9431 + }, + { + "epoch": 0.79, + "learning_rate": 1.7247160163116405e-05, + "loss": 1.2934, + "step": 9432 + }, + { + "epoch": 0.79, + "learning_rate": 1.7246533807413508e-05, + "loss": 1.2387, + "step": 9433 + }, + { + "epoch": 0.79, + "learning_rate": 1.7245907391837427e-05, + "loss": 1.409, + "step": 9434 + }, + { + "epoch": 0.79, + "learning_rate": 1.7245280916393342e-05, + "loss": 1.3351, + "step": 9435 + }, + { + "epoch": 0.79, + "learning_rate": 1.7244654381086427e-05, + "loss": 1.3601, + "step": 9436 + }, + { + "epoch": 0.79, + "learning_rate": 1.7244027785921858e-05, + "loss": 1.233, + "step": 9437 + }, + { + "epoch": 0.79, + "learning_rate": 1.7243401130904816e-05, + "loss": 1.2331, + "step": 9438 + }, + { + "epoch": 0.79, + "learning_rate": 1.7242774416040472e-05, + "loss": 1.331, + "step": 9439 + }, + { + "epoch": 0.79, + "learning_rate": 1.7242147641334012e-05, + "loss": 1.3415, + "step": 9440 + }, + { + "epoch": 0.79, + "learning_rate": 1.7241520806790607e-05, + "loss": 1.2776, + "step": 9441 + }, + { + "epoch": 0.79, + "learning_rate": 1.724089391241544e-05, + "loss": 1.2904, + "step": 9442 + }, + { + "epoch": 0.8, + "learning_rate": 1.724026695821369e-05, + "loss": 1.3492, + "step": 9443 + }, + { + "epoch": 0.8, + "learning_rate": 1.7239639944190537e-05, + "loss": 1.2401, + "step": 9444 + }, + { + "epoch": 0.8, + "learning_rate": 1.7239012870351163e-05, + "loss": 1.3159, + "step": 9445 + }, + { + "epoch": 0.8, + "learning_rate": 1.723838573670075e-05, + "loss": 1.2929, + "step": 9446 + }, + { + "epoch": 0.8, + "learning_rate": 1.723775854324447e-05, + "loss": 1.3094, + "step": 9447 + }, + { + "epoch": 0.8, + "learning_rate": 1.723713128998752e-05, + "loss": 1.2427, + "step": 9448 + }, + { + "epoch": 0.8, + "learning_rate": 1.7236503976935074e-05, + "loss": 1.2473, + "step": 9449 + }, + { + "epoch": 0.8, + "learning_rate": 1.723587660409231e-05, + "loss": 1.2827, + "step": 9450 + }, + { + "epoch": 0.8, + "learning_rate": 1.7235249171464423e-05, + "loss": 1.3408, + "step": 9451 + }, + { + "epoch": 0.8, + "learning_rate": 1.7234621679056593e-05, + "loss": 1.2381, + "step": 9452 + }, + { + "epoch": 0.8, + "learning_rate": 1.7233994126874e-05, + "loss": 1.3392, + "step": 9453 + }, + { + "epoch": 0.8, + "learning_rate": 1.7233366514921836e-05, + "loss": 1.3468, + "step": 9454 + }, + { + "epoch": 0.8, + "learning_rate": 1.723273884320528e-05, + "loss": 1.3099, + "step": 9455 + }, + { + "epoch": 0.8, + "learning_rate": 1.7232111111729518e-05, + "loss": 1.3251, + "step": 9456 + }, + { + "epoch": 0.8, + "learning_rate": 1.7231483320499743e-05, + "loss": 1.2658, + "step": 9457 + }, + { + "epoch": 0.8, + "learning_rate": 1.723085546952114e-05, + "loss": 1.3687, + "step": 9458 + }, + { + "epoch": 0.8, + "learning_rate": 1.7230227558798894e-05, + "loss": 1.279, + "step": 9459 + }, + { + "epoch": 0.8, + "learning_rate": 1.722959958833819e-05, + "loss": 1.2006, + "step": 9460 + }, + { + "epoch": 0.8, + "learning_rate": 1.7228971558144224e-05, + "loss": 1.1676, + "step": 9461 + }, + { + "epoch": 0.8, + "learning_rate": 1.7228343468222185e-05, + "loss": 1.2692, + "step": 9462 + }, + { + "epoch": 0.8, + "learning_rate": 1.7227715318577252e-05, + "loss": 1.3196, + "step": 9463 + }, + { + "epoch": 0.8, + "learning_rate": 1.7227087109214627e-05, + "loss": 1.2639, + "step": 9464 + }, + { + "epoch": 0.8, + "learning_rate": 1.7226458840139492e-05, + "loss": 1.2617, + "step": 9465 + }, + { + "epoch": 0.8, + "learning_rate": 1.7225830511357042e-05, + "loss": 1.2517, + "step": 9466 + }, + { + "epoch": 0.8, + "learning_rate": 1.7225202122872467e-05, + "loss": 1.3012, + "step": 9467 + }, + { + "epoch": 0.8, + "learning_rate": 1.7224573674690962e-05, + "loss": 1.2607, + "step": 9468 + }, + { + "epoch": 0.8, + "learning_rate": 1.7223945166817717e-05, + "loss": 1.183, + "step": 9469 + }, + { + "epoch": 0.8, + "learning_rate": 1.7223316599257923e-05, + "loss": 1.2282, + "step": 9470 + }, + { + "epoch": 0.8, + "learning_rate": 1.722268797201678e-05, + "loss": 1.3052, + "step": 9471 + }, + { + "epoch": 0.8, + "learning_rate": 1.722205928509947e-05, + "loss": 1.231, + "step": 9472 + }, + { + "epoch": 0.8, + "learning_rate": 1.72214305385112e-05, + "loss": 1.3092, + "step": 9473 + }, + { + "epoch": 0.8, + "learning_rate": 1.7220801732257156e-05, + "loss": 1.3367, + "step": 9474 + }, + { + "epoch": 0.8, + "learning_rate": 1.7220172866342542e-05, + "loss": 1.3748, + "step": 9475 + }, + { + "epoch": 0.8, + "learning_rate": 1.7219543940772546e-05, + "loss": 1.3334, + "step": 9476 + }, + { + "epoch": 0.8, + "learning_rate": 1.721891495555237e-05, + "loss": 1.2243, + "step": 9477 + }, + { + "epoch": 0.8, + "learning_rate": 1.7218285910687205e-05, + "loss": 1.3204, + "step": 9478 + }, + { + "epoch": 0.8, + "learning_rate": 1.7217656806182255e-05, + "loss": 1.2752, + "step": 9479 + }, + { + "epoch": 0.8, + "learning_rate": 1.7217027642042712e-05, + "loss": 1.3262, + "step": 9480 + }, + { + "epoch": 0.8, + "learning_rate": 1.7216398418273778e-05, + "loss": 1.312, + "step": 9481 + }, + { + "epoch": 0.8, + "learning_rate": 1.721576913488065e-05, + "loss": 1.2195, + "step": 9482 + }, + { + "epoch": 0.8, + "learning_rate": 1.7215139791868528e-05, + "loss": 1.2645, + "step": 9483 + }, + { + "epoch": 0.8, + "learning_rate": 1.721451038924261e-05, + "loss": 1.3915, + "step": 9484 + }, + { + "epoch": 0.8, + "learning_rate": 1.72138809270081e-05, + "loss": 1.272, + "step": 9485 + }, + { + "epoch": 0.8, + "learning_rate": 1.72132514051702e-05, + "loss": 1.2878, + "step": 9486 + }, + { + "epoch": 0.8, + "learning_rate": 1.7212621823734106e-05, + "loss": 1.2595, + "step": 9487 + }, + { + "epoch": 0.8, + "learning_rate": 1.7211992182705022e-05, + "loss": 1.3072, + "step": 9488 + }, + { + "epoch": 0.8, + "learning_rate": 1.721136248208815e-05, + "loss": 1.3054, + "step": 9489 + }, + { + "epoch": 0.8, + "learning_rate": 1.7210732721888693e-05, + "loss": 1.1825, + "step": 9490 + }, + { + "epoch": 0.8, + "learning_rate": 1.7210102902111856e-05, + "loss": 1.3458, + "step": 9491 + }, + { + "epoch": 0.8, + "learning_rate": 1.720947302276284e-05, + "loss": 1.3329, + "step": 9492 + }, + { + "epoch": 0.8, + "learning_rate": 1.720884308384685e-05, + "loss": 1.2942, + "step": 9493 + }, + { + "epoch": 0.8, + "learning_rate": 1.7208213085369094e-05, + "loss": 1.3198, + "step": 9494 + }, + { + "epoch": 0.8, + "learning_rate": 1.720758302733477e-05, + "loss": 1.22, + "step": 9495 + }, + { + "epoch": 0.8, + "learning_rate": 1.720695290974909e-05, + "loss": 1.3154, + "step": 9496 + }, + { + "epoch": 0.8, + "learning_rate": 1.7206322732617263e-05, + "loss": 1.3396, + "step": 9497 + }, + { + "epoch": 0.8, + "learning_rate": 1.7205692495944487e-05, + "loss": 1.3928, + "step": 9498 + }, + { + "epoch": 0.8, + "learning_rate": 1.7205062199735975e-05, + "loss": 1.2743, + "step": 9499 + }, + { + "epoch": 0.8, + "learning_rate": 1.7204431843996933e-05, + "loss": 1.2607, + "step": 9500 + }, + { + "epoch": 0.8, + "learning_rate": 1.720380142873257e-05, + "loss": 1.3402, + "step": 9501 + }, + { + "epoch": 0.8, + "learning_rate": 1.720317095394809e-05, + "loss": 1.3604, + "step": 9502 + }, + { + "epoch": 0.8, + "learning_rate": 1.720254041964871e-05, + "loss": 1.3362, + "step": 9503 + }, + { + "epoch": 0.8, + "learning_rate": 1.7201909825839637e-05, + "loss": 1.2819, + "step": 9504 + }, + { + "epoch": 0.8, + "learning_rate": 1.720127917252608e-05, + "loss": 1.2666, + "step": 9505 + }, + { + "epoch": 0.8, + "learning_rate": 1.7200648459713248e-05, + "loss": 1.2907, + "step": 9506 + }, + { + "epoch": 0.8, + "learning_rate": 1.7200017687406354e-05, + "loss": 1.2804, + "step": 9507 + }, + { + "epoch": 0.8, + "learning_rate": 1.7199386855610608e-05, + "loss": 1.3618, + "step": 9508 + }, + { + "epoch": 0.8, + "learning_rate": 1.7198755964331227e-05, + "loss": 1.3743, + "step": 9509 + }, + { + "epoch": 0.8, + "learning_rate": 1.7198125013573417e-05, + "loss": 1.2755, + "step": 9510 + }, + { + "epoch": 0.8, + "learning_rate": 1.7197494003342398e-05, + "loss": 1.3085, + "step": 9511 + }, + { + "epoch": 0.8, + "learning_rate": 1.7196862933643378e-05, + "loss": 1.3362, + "step": 9512 + }, + { + "epoch": 0.8, + "learning_rate": 1.7196231804481574e-05, + "loss": 1.2876, + "step": 9513 + }, + { + "epoch": 0.8, + "learning_rate": 1.7195600615862198e-05, + "loss": 1.2365, + "step": 9514 + }, + { + "epoch": 0.8, + "learning_rate": 1.719496936779047e-05, + "loss": 1.3624, + "step": 9515 + }, + { + "epoch": 0.8, + "learning_rate": 1.7194338060271598e-05, + "loss": 1.225, + "step": 9516 + }, + { + "epoch": 0.8, + "learning_rate": 1.7193706693310807e-05, + "loss": 1.3057, + "step": 9517 + }, + { + "epoch": 0.8, + "learning_rate": 1.7193075266913307e-05, + "loss": 1.2822, + "step": 9518 + }, + { + "epoch": 0.8, + "learning_rate": 1.7192443781084316e-05, + "loss": 1.3151, + "step": 9519 + }, + { + "epoch": 0.8, + "learning_rate": 1.7191812235829053e-05, + "loss": 1.3106, + "step": 9520 + }, + { + "epoch": 0.8, + "learning_rate": 1.7191180631152734e-05, + "loss": 1.2283, + "step": 9521 + }, + { + "epoch": 0.8, + "learning_rate": 1.7190548967060584e-05, + "loss": 1.2439, + "step": 9522 + }, + { + "epoch": 0.8, + "learning_rate": 1.7189917243557815e-05, + "loss": 1.3577, + "step": 9523 + }, + { + "epoch": 0.8, + "learning_rate": 1.7189285460649646e-05, + "loss": 1.1808, + "step": 9524 + }, + { + "epoch": 0.8, + "learning_rate": 1.7188653618341304e-05, + "loss": 1.3241, + "step": 9525 + }, + { + "epoch": 0.8, + "learning_rate": 1.7188021716638e-05, + "loss": 1.375, + "step": 9526 + }, + { + "epoch": 0.8, + "learning_rate": 1.7187389755544967e-05, + "loss": 1.3225, + "step": 9527 + }, + { + "epoch": 0.8, + "learning_rate": 1.7186757735067415e-05, + "loss": 1.3079, + "step": 9528 + }, + { + "epoch": 0.8, + "learning_rate": 1.718612565521057e-05, + "loss": 1.1947, + "step": 9529 + }, + { + "epoch": 0.8, + "learning_rate": 1.7185493515979655e-05, + "loss": 1.2902, + "step": 9530 + }, + { + "epoch": 0.8, + "learning_rate": 1.7184861317379897e-05, + "loss": 1.208, + "step": 9531 + }, + { + "epoch": 0.8, + "learning_rate": 1.7184229059416512e-05, + "loss": 1.2647, + "step": 9532 + }, + { + "epoch": 0.8, + "learning_rate": 1.7183596742094725e-05, + "loss": 1.3068, + "step": 9533 + }, + { + "epoch": 0.8, + "learning_rate": 1.7182964365419767e-05, + "loss": 1.2707, + "step": 9534 + }, + { + "epoch": 0.8, + "learning_rate": 1.7182331929396857e-05, + "loss": 1.3628, + "step": 9535 + }, + { + "epoch": 0.8, + "learning_rate": 1.718169943403122e-05, + "loss": 1.2451, + "step": 9536 + }, + { + "epoch": 0.8, + "learning_rate": 1.718106687932809e-05, + "loss": 1.4322, + "step": 9537 + }, + { + "epoch": 0.8, + "learning_rate": 1.718043426529268e-05, + "loss": 1.2879, + "step": 9538 + }, + { + "epoch": 0.8, + "learning_rate": 1.717980159193023e-05, + "loss": 1.2527, + "step": 9539 + }, + { + "epoch": 0.8, + "learning_rate": 1.7179168859245955e-05, + "loss": 1.3593, + "step": 9540 + }, + { + "epoch": 0.8, + "learning_rate": 1.7178536067245092e-05, + "loss": 1.3823, + "step": 9541 + }, + { + "epoch": 0.8, + "learning_rate": 1.7177903215932868e-05, + "loss": 1.1689, + "step": 9542 + }, + { + "epoch": 0.8, + "learning_rate": 1.717727030531451e-05, + "loss": 1.3846, + "step": 9543 + }, + { + "epoch": 0.8, + "learning_rate": 1.7176637335395247e-05, + "loss": 1.2429, + "step": 9544 + }, + { + "epoch": 0.8, + "learning_rate": 1.717600430618031e-05, + "loss": 1.3376, + "step": 9545 + }, + { + "epoch": 0.8, + "learning_rate": 1.717537121767493e-05, + "loss": 1.318, + "step": 9546 + }, + { + "epoch": 0.8, + "learning_rate": 1.7174738069884335e-05, + "loss": 1.3156, + "step": 9547 + }, + { + "epoch": 0.8, + "learning_rate": 1.7174104862813757e-05, + "loss": 1.27, + "step": 9548 + }, + { + "epoch": 0.8, + "learning_rate": 1.7173471596468433e-05, + "loss": 1.2969, + "step": 9549 + }, + { + "epoch": 0.8, + "learning_rate": 1.7172838270853588e-05, + "loss": 1.2551, + "step": 9550 + }, + { + "epoch": 0.8, + "learning_rate": 1.7172204885974456e-05, + "loss": 1.2783, + "step": 9551 + }, + { + "epoch": 0.8, + "learning_rate": 1.7171571441836272e-05, + "loss": 1.343, + "step": 9552 + }, + { + "epoch": 0.8, + "learning_rate": 1.717093793844427e-05, + "loss": 1.2692, + "step": 9553 + }, + { + "epoch": 0.8, + "learning_rate": 1.7170304375803685e-05, + "loss": 1.3747, + "step": 9554 + }, + { + "epoch": 0.8, + "learning_rate": 1.716967075391975e-05, + "loss": 1.2833, + "step": 9555 + }, + { + "epoch": 0.8, + "learning_rate": 1.71690370727977e-05, + "loss": 1.3322, + "step": 9556 + }, + { + "epoch": 0.8, + "learning_rate": 1.716840333244277e-05, + "loss": 1.2087, + "step": 9557 + }, + { + "epoch": 0.8, + "learning_rate": 1.71677695328602e-05, + "loss": 1.3356, + "step": 9558 + }, + { + "epoch": 0.8, + "learning_rate": 1.7167135674055226e-05, + "loss": 1.2811, + "step": 9559 + }, + { + "epoch": 0.8, + "learning_rate": 1.716650175603308e-05, + "loss": 1.4077, + "step": 9560 + }, + { + "epoch": 0.8, + "learning_rate": 1.7165867778799003e-05, + "loss": 1.2521, + "step": 9561 + }, + { + "epoch": 0.81, + "learning_rate": 1.7165233742358235e-05, + "loss": 1.3818, + "step": 9562 + }, + { + "epoch": 0.81, + "learning_rate": 1.716459964671601e-05, + "loss": 1.3042, + "step": 9563 + }, + { + "epoch": 0.81, + "learning_rate": 1.7163965491877573e-05, + "loss": 1.3585, + "step": 9564 + }, + { + "epoch": 0.81, + "learning_rate": 1.7163331277848156e-05, + "loss": 1.2823, + "step": 9565 + }, + { + "epoch": 0.81, + "learning_rate": 1.7162697004633005e-05, + "loss": 1.3097, + "step": 9566 + }, + { + "epoch": 0.81, + "learning_rate": 1.7162062672237364e-05, + "loss": 1.3873, + "step": 9567 + }, + { + "epoch": 0.81, + "learning_rate": 1.7161428280666462e-05, + "loss": 1.294, + "step": 9568 + }, + { + "epoch": 0.81, + "learning_rate": 1.716079382992555e-05, + "loss": 1.2649, + "step": 9569 + }, + { + "epoch": 0.81, + "learning_rate": 1.716015932001987e-05, + "loss": 1.3018, + "step": 9570 + }, + { + "epoch": 0.81, + "learning_rate": 1.715952475095466e-05, + "loss": 1.2472, + "step": 9571 + }, + { + "epoch": 0.81, + "learning_rate": 1.7158890122735164e-05, + "loss": 1.3191, + "step": 9572 + }, + { + "epoch": 0.81, + "learning_rate": 1.715825543536663e-05, + "loss": 1.1884, + "step": 9573 + }, + { + "epoch": 0.81, + "learning_rate": 1.7157620688854296e-05, + "loss": 1.2254, + "step": 9574 + }, + { + "epoch": 0.81, + "learning_rate": 1.715698588320341e-05, + "loss": 1.2479, + "step": 9575 + }, + { + "epoch": 0.81, + "learning_rate": 1.7156351018419216e-05, + "loss": 1.1946, + "step": 9576 + }, + { + "epoch": 0.81, + "learning_rate": 1.715571609450696e-05, + "loss": 1.4335, + "step": 9577 + }, + { + "epoch": 0.81, + "learning_rate": 1.7155081111471886e-05, + "loss": 1.2333, + "step": 9578 + }, + { + "epoch": 0.81, + "learning_rate": 1.715444606931924e-05, + "loss": 1.3219, + "step": 9579 + }, + { + "epoch": 0.81, + "learning_rate": 1.7153810968054276e-05, + "loss": 1.2704, + "step": 9580 + }, + { + "epoch": 0.81, + "learning_rate": 1.7153175807682238e-05, + "loss": 1.2743, + "step": 9581 + }, + { + "epoch": 0.81, + "learning_rate": 1.7152540588208368e-05, + "loss": 1.3045, + "step": 9582 + }, + { + "epoch": 0.81, + "learning_rate": 1.7151905309637916e-05, + "loss": 1.2297, + "step": 9583 + }, + { + "epoch": 0.81, + "learning_rate": 1.7151269971976136e-05, + "loss": 1.274, + "step": 9584 + }, + { + "epoch": 0.81, + "learning_rate": 1.7150634575228275e-05, + "loss": 1.2191, + "step": 9585 + }, + { + "epoch": 0.81, + "learning_rate": 1.7149999119399584e-05, + "loss": 1.2985, + "step": 9586 + }, + { + "epoch": 0.81, + "learning_rate": 1.714936360449531e-05, + "loss": 1.2597, + "step": 9587 + }, + { + "epoch": 0.81, + "learning_rate": 1.7148728030520705e-05, + "loss": 1.3386, + "step": 9588 + }, + { + "epoch": 0.81, + "learning_rate": 1.7148092397481023e-05, + "loss": 1.3127, + "step": 9589 + }, + { + "epoch": 0.81, + "learning_rate": 1.7147456705381514e-05, + "loss": 1.2898, + "step": 9590 + }, + { + "epoch": 0.81, + "learning_rate": 1.714682095422743e-05, + "loss": 1.3319, + "step": 9591 + }, + { + "epoch": 0.81, + "learning_rate": 1.7146185144024025e-05, + "loss": 1.2508, + "step": 9592 + }, + { + "epoch": 0.81, + "learning_rate": 1.714554927477655e-05, + "loss": 1.2793, + "step": 9593 + }, + { + "epoch": 0.81, + "learning_rate": 1.714491334649026e-05, + "loss": 1.3635, + "step": 9594 + }, + { + "epoch": 0.81, + "learning_rate": 1.7144277359170408e-05, + "loss": 1.3287, + "step": 9595 + }, + { + "epoch": 0.81, + "learning_rate": 1.7143641312822253e-05, + "loss": 1.2194, + "step": 9596 + }, + { + "epoch": 0.81, + "learning_rate": 1.7143005207451046e-05, + "loss": 1.1753, + "step": 9597 + }, + { + "epoch": 0.81, + "learning_rate": 1.7142369043062043e-05, + "loss": 1.291, + "step": 9598 + }, + { + "epoch": 0.81, + "learning_rate": 1.7141732819660498e-05, + "loss": 1.338, + "step": 9599 + }, + { + "epoch": 0.81, + "learning_rate": 1.7141096537251677e-05, + "loss": 1.3285, + "step": 9600 + }, + { + "epoch": 0.81, + "learning_rate": 1.7140460195840828e-05, + "loss": 1.2339, + "step": 9601 + }, + { + "epoch": 0.81, + "learning_rate": 1.713982379543321e-05, + "loss": 1.2141, + "step": 9602 + }, + { + "epoch": 0.81, + "learning_rate": 1.7139187336034085e-05, + "loss": 1.2369, + "step": 9603 + }, + { + "epoch": 0.81, + "learning_rate": 1.7138550817648712e-05, + "loss": 1.3194, + "step": 9604 + }, + { + "epoch": 0.81, + "learning_rate": 1.713791424028234e-05, + "loss": 1.33, + "step": 9605 + }, + { + "epoch": 0.81, + "learning_rate": 1.7137277603940244e-05, + "loss": 1.3503, + "step": 9606 + }, + { + "epoch": 0.81, + "learning_rate": 1.7136640908627673e-05, + "loss": 1.2659, + "step": 9607 + }, + { + "epoch": 0.81, + "learning_rate": 1.7136004154349892e-05, + "loss": 1.261, + "step": 9608 + }, + { + "epoch": 0.81, + "learning_rate": 1.713536734111216e-05, + "loss": 1.3324, + "step": 9609 + }, + { + "epoch": 0.81, + "learning_rate": 1.713473046891974e-05, + "loss": 1.2781, + "step": 9610 + }, + { + "epoch": 0.81, + "learning_rate": 1.713409353777789e-05, + "loss": 1.307, + "step": 9611 + }, + { + "epoch": 0.81, + "learning_rate": 1.7133456547691877e-05, + "loss": 1.3562, + "step": 9612 + }, + { + "epoch": 0.81, + "learning_rate": 1.7132819498666966e-05, + "loss": 1.3473, + "step": 9613 + }, + { + "epoch": 0.81, + "learning_rate": 1.7132182390708417e-05, + "loss": 1.2613, + "step": 9614 + }, + { + "epoch": 0.81, + "learning_rate": 1.7131545223821494e-05, + "loss": 1.3523, + "step": 9615 + }, + { + "epoch": 0.81, + "learning_rate": 1.7130907998011462e-05, + "loss": 1.3617, + "step": 9616 + }, + { + "epoch": 0.81, + "learning_rate": 1.7130270713283584e-05, + "loss": 1.232, + "step": 9617 + }, + { + "epoch": 0.81, + "learning_rate": 1.7129633369643128e-05, + "loss": 1.2153, + "step": 9618 + }, + { + "epoch": 0.81, + "learning_rate": 1.712899596709536e-05, + "loss": 1.2483, + "step": 9619 + }, + { + "epoch": 0.81, + "learning_rate": 1.7128358505645546e-05, + "loss": 1.3216, + "step": 9620 + }, + { + "epoch": 0.81, + "learning_rate": 1.7127720985298953e-05, + "loss": 1.2149, + "step": 9621 + }, + { + "epoch": 0.81, + "learning_rate": 1.712708340606085e-05, + "loss": 1.1486, + "step": 9622 + }, + { + "epoch": 0.81, + "learning_rate": 1.7126445767936498e-05, + "loss": 1.2975, + "step": 9623 + }, + { + "epoch": 0.81, + "learning_rate": 1.7125808070931172e-05, + "loss": 1.2965, + "step": 9624 + }, + { + "epoch": 0.81, + "learning_rate": 1.7125170315050142e-05, + "loss": 1.2319, + "step": 9625 + }, + { + "epoch": 0.81, + "learning_rate": 1.712453250029867e-05, + "loss": 1.3826, + "step": 9626 + }, + { + "epoch": 0.81, + "learning_rate": 1.7123894626682033e-05, + "loss": 1.2626, + "step": 9627 + }, + { + "epoch": 0.81, + "learning_rate": 1.7123256694205502e-05, + "loss": 1.2924, + "step": 9628 + }, + { + "epoch": 0.81, + "learning_rate": 1.712261870287434e-05, + "loss": 1.364, + "step": 9629 + }, + { + "epoch": 0.81, + "learning_rate": 1.7121980652693823e-05, + "loss": 1.2768, + "step": 9630 + }, + { + "epoch": 0.81, + "learning_rate": 1.7121342543669225e-05, + "loss": 1.2479, + "step": 9631 + }, + { + "epoch": 0.81, + "learning_rate": 1.7120704375805813e-05, + "loss": 1.3093, + "step": 9632 + }, + { + "epoch": 0.81, + "learning_rate": 1.7120066149108864e-05, + "loss": 1.2386, + "step": 9633 + }, + { + "epoch": 0.81, + "learning_rate": 1.7119427863583647e-05, + "loss": 1.2708, + "step": 9634 + }, + { + "epoch": 0.81, + "learning_rate": 1.711878951923544e-05, + "loss": 1.3002, + "step": 9635 + }, + { + "epoch": 0.81, + "learning_rate": 1.711815111606952e-05, + "loss": 1.3256, + "step": 9636 + }, + { + "epoch": 0.81, + "learning_rate": 1.7117512654091154e-05, + "loss": 1.2727, + "step": 9637 + }, + { + "epoch": 0.81, + "learning_rate": 1.711687413330562e-05, + "loss": 1.3579, + "step": 9638 + }, + { + "epoch": 0.81, + "learning_rate": 1.7116235553718194e-05, + "loss": 1.3827, + "step": 9639 + }, + { + "epoch": 0.81, + "learning_rate": 1.7115596915334152e-05, + "loss": 1.3533, + "step": 9640 + }, + { + "epoch": 0.81, + "learning_rate": 1.7114958218158773e-05, + "loss": 1.2744, + "step": 9641 + }, + { + "epoch": 0.81, + "learning_rate": 1.711431946219733e-05, + "loss": 1.3003, + "step": 9642 + }, + { + "epoch": 0.81, + "learning_rate": 1.7113680647455106e-05, + "loss": 1.3591, + "step": 9643 + }, + { + "epoch": 0.81, + "learning_rate": 1.711304177393737e-05, + "loss": 1.3385, + "step": 9644 + }, + { + "epoch": 0.81, + "learning_rate": 1.711240284164941e-05, + "loss": 1.3189, + "step": 9645 + }, + { + "epoch": 0.81, + "learning_rate": 1.7111763850596506e-05, + "loss": 1.3069, + "step": 9646 + }, + { + "epoch": 0.81, + "learning_rate": 1.7111124800783927e-05, + "loss": 1.3252, + "step": 9647 + }, + { + "epoch": 0.81, + "learning_rate": 1.711048569221696e-05, + "loss": 1.3146, + "step": 9648 + }, + { + "epoch": 0.81, + "learning_rate": 1.7109846524900885e-05, + "loss": 1.3559, + "step": 9649 + }, + { + "epoch": 0.81, + "learning_rate": 1.7109207298840984e-05, + "loss": 1.313, + "step": 9650 + }, + { + "epoch": 0.81, + "learning_rate": 1.7108568014042536e-05, + "loss": 1.1865, + "step": 9651 + }, + { + "epoch": 0.81, + "learning_rate": 1.7107928670510823e-05, + "loss": 1.2726, + "step": 9652 + }, + { + "epoch": 0.81, + "learning_rate": 1.7107289268251128e-05, + "loss": 1.1649, + "step": 9653 + }, + { + "epoch": 0.81, + "learning_rate": 1.7106649807268735e-05, + "loss": 1.2105, + "step": 9654 + }, + { + "epoch": 0.81, + "learning_rate": 1.710601028756893e-05, + "loss": 1.2967, + "step": 9655 + }, + { + "epoch": 0.81, + "learning_rate": 1.7105370709156992e-05, + "loss": 1.2479, + "step": 9656 + }, + { + "epoch": 0.81, + "learning_rate": 1.7104731072038205e-05, + "loss": 1.2843, + "step": 9657 + }, + { + "epoch": 0.81, + "learning_rate": 1.710409137621786e-05, + "loss": 1.2422, + "step": 9658 + }, + { + "epoch": 0.81, + "learning_rate": 1.7103451621701234e-05, + "loss": 1.3325, + "step": 9659 + }, + { + "epoch": 0.81, + "learning_rate": 1.710281180849362e-05, + "loss": 1.2498, + "step": 9660 + }, + { + "epoch": 0.81, + "learning_rate": 1.71021719366003e-05, + "loss": 1.244, + "step": 9661 + }, + { + "epoch": 0.81, + "learning_rate": 1.7101532006026565e-05, + "loss": 1.3082, + "step": 9662 + }, + { + "epoch": 0.81, + "learning_rate": 1.7100892016777696e-05, + "loss": 1.3137, + "step": 9663 + }, + { + "epoch": 0.81, + "learning_rate": 1.7100251968858988e-05, + "loss": 1.3019, + "step": 9664 + }, + { + "epoch": 0.81, + "learning_rate": 1.7099611862275723e-05, + "loss": 1.3003, + "step": 9665 + }, + { + "epoch": 0.81, + "learning_rate": 1.7098971697033198e-05, + "loss": 1.2343, + "step": 9666 + }, + { + "epoch": 0.81, + "learning_rate": 1.709833147313669e-05, + "loss": 1.3765, + "step": 9667 + }, + { + "epoch": 0.81, + "learning_rate": 1.70976911905915e-05, + "loss": 1.2869, + "step": 9668 + }, + { + "epoch": 0.81, + "learning_rate": 1.7097050849402914e-05, + "loss": 1.2477, + "step": 9669 + }, + { + "epoch": 0.81, + "learning_rate": 1.7096410449576223e-05, + "loss": 1.2578, + "step": 9670 + }, + { + "epoch": 0.81, + "learning_rate": 1.7095769991116716e-05, + "loss": 1.2952, + "step": 9671 + }, + { + "epoch": 0.81, + "learning_rate": 1.7095129474029684e-05, + "loss": 1.2679, + "step": 9672 + }, + { + "epoch": 0.81, + "learning_rate": 1.7094488898320424e-05, + "loss": 1.2392, + "step": 9673 + }, + { + "epoch": 0.81, + "learning_rate": 1.7093848263994227e-05, + "loss": 1.2012, + "step": 9674 + }, + { + "epoch": 0.81, + "learning_rate": 1.7093207571056386e-05, + "loss": 1.1946, + "step": 9675 + }, + { + "epoch": 0.81, + "learning_rate": 1.709256681951219e-05, + "loss": 1.2857, + "step": 9676 + }, + { + "epoch": 0.81, + "learning_rate": 1.709192600936694e-05, + "loss": 1.2831, + "step": 9677 + }, + { + "epoch": 0.81, + "learning_rate": 1.7091285140625927e-05, + "loss": 1.2754, + "step": 9678 + }, + { + "epoch": 0.81, + "learning_rate": 1.709064421329445e-05, + "loss": 1.2738, + "step": 9679 + }, + { + "epoch": 0.81, + "learning_rate": 1.7090003227377794e-05, + "loss": 1.2849, + "step": 9680 + }, + { + "epoch": 0.82, + "learning_rate": 1.708936218288127e-05, + "loss": 1.2698, + "step": 9681 + }, + { + "epoch": 0.82, + "learning_rate": 1.708872107981016e-05, + "loss": 1.271, + "step": 9682 + }, + { + "epoch": 0.82, + "learning_rate": 1.708807991816977e-05, + "loss": 1.3616, + "step": 9683 + }, + { + "epoch": 0.82, + "learning_rate": 1.7087438697965397e-05, + "loss": 1.3423, + "step": 9684 + }, + { + "epoch": 0.82, + "learning_rate": 1.7086797419202334e-05, + "loss": 1.391, + "step": 9685 + }, + { + "epoch": 0.82, + "learning_rate": 1.7086156081885886e-05, + "loss": 1.2793, + "step": 9686 + }, + { + "epoch": 0.82, + "learning_rate": 1.7085514686021348e-05, + "loss": 1.2926, + "step": 9687 + }, + { + "epoch": 0.82, + "learning_rate": 1.708487323161402e-05, + "loss": 1.2619, + "step": 9688 + }, + { + "epoch": 0.82, + "learning_rate": 1.7084231718669203e-05, + "loss": 1.2995, + "step": 9689 + }, + { + "epoch": 0.82, + "learning_rate": 1.708359014719219e-05, + "loss": 1.2559, + "step": 9690 + }, + { + "epoch": 0.82, + "learning_rate": 1.7082948517188292e-05, + "loss": 1.19, + "step": 9691 + }, + { + "epoch": 0.82, + "learning_rate": 1.708230682866281e-05, + "loss": 1.4031, + "step": 9692 + }, + { + "epoch": 0.82, + "learning_rate": 1.708166508162104e-05, + "loss": 1.3437, + "step": 9693 + }, + { + "epoch": 0.82, + "learning_rate": 1.7081023276068286e-05, + "loss": 1.2266, + "step": 9694 + }, + { + "epoch": 0.82, + "learning_rate": 1.708038141200985e-05, + "loss": 1.2308, + "step": 9695 + }, + { + "epoch": 0.82, + "learning_rate": 1.7079739489451038e-05, + "loss": 1.2878, + "step": 9696 + }, + { + "epoch": 0.82, + "learning_rate": 1.7079097508397156e-05, + "loss": 1.284, + "step": 9697 + }, + { + "epoch": 0.82, + "learning_rate": 1.70784554688535e-05, + "loss": 1.3293, + "step": 9698 + }, + { + "epoch": 0.82, + "learning_rate": 1.7077813370825383e-05, + "loss": 1.3056, + "step": 9699 + }, + { + "epoch": 0.82, + "learning_rate": 1.7077171214318106e-05, + "loss": 1.2259, + "step": 9700 + }, + { + "epoch": 0.82, + "learning_rate": 1.7076528999336974e-05, + "loss": 1.3041, + "step": 9701 + }, + { + "epoch": 0.82, + "learning_rate": 1.7075886725887295e-05, + "loss": 1.2757, + "step": 9702 + }, + { + "epoch": 0.82, + "learning_rate": 1.7075244393974376e-05, + "loss": 1.3352, + "step": 9703 + }, + { + "epoch": 0.82, + "learning_rate": 1.7074602003603522e-05, + "loss": 1.2272, + "step": 9704 + }, + { + "epoch": 0.82, + "learning_rate": 1.7073959554780045e-05, + "loss": 1.417, + "step": 9705 + }, + { + "epoch": 0.82, + "learning_rate": 1.7073317047509246e-05, + "loss": 1.3347, + "step": 9706 + }, + { + "epoch": 0.82, + "learning_rate": 1.7072674481796442e-05, + "loss": 1.313, + "step": 9707 + }, + { + "epoch": 0.82, + "learning_rate": 1.7072031857646934e-05, + "loss": 1.3232, + "step": 9708 + }, + { + "epoch": 0.82, + "learning_rate": 1.7071389175066037e-05, + "loss": 1.2754, + "step": 9709 + }, + { + "epoch": 0.82, + "learning_rate": 1.707074643405906e-05, + "loss": 1.2634, + "step": 9710 + }, + { + "epoch": 0.82, + "learning_rate": 1.7070103634631314e-05, + "loss": 1.297, + "step": 9711 + }, + { + "epoch": 0.82, + "learning_rate": 1.7069460776788106e-05, + "loss": 1.2836, + "step": 9712 + }, + { + "epoch": 0.82, + "learning_rate": 1.7068817860534756e-05, + "loss": 1.2226, + "step": 9713 + }, + { + "epoch": 0.82, + "learning_rate": 1.7068174885876566e-05, + "loss": 1.3701, + "step": 9714 + }, + { + "epoch": 0.82, + "learning_rate": 1.7067531852818853e-05, + "loss": 1.3483, + "step": 9715 + }, + { + "epoch": 0.82, + "learning_rate": 1.706688876136693e-05, + "loss": 1.266, + "step": 9716 + }, + { + "epoch": 0.82, + "learning_rate": 1.7066245611526115e-05, + "loss": 1.3203, + "step": 9717 + }, + { + "epoch": 0.82, + "learning_rate": 1.706560240330171e-05, + "loss": 1.2732, + "step": 9718 + }, + { + "epoch": 0.82, + "learning_rate": 1.706495913669904e-05, + "loss": 1.3126, + "step": 9719 + }, + { + "epoch": 0.82, + "learning_rate": 1.7064315811723416e-05, + "loss": 1.186, + "step": 9720 + }, + { + "epoch": 0.82, + "learning_rate": 1.7063672428380154e-05, + "loss": 1.3171, + "step": 9721 + }, + { + "epoch": 0.82, + "learning_rate": 1.706302898667457e-05, + "loss": 1.2827, + "step": 9722 + }, + { + "epoch": 0.82, + "learning_rate": 1.7062385486611982e-05, + "loss": 1.3328, + "step": 9723 + }, + { + "epoch": 0.82, + "learning_rate": 1.7061741928197703e-05, + "loss": 1.2284, + "step": 9724 + }, + { + "epoch": 0.82, + "learning_rate": 1.7061098311437053e-05, + "loss": 1.274, + "step": 9725 + }, + { + "epoch": 0.82, + "learning_rate": 1.706045463633535e-05, + "loss": 1.2975, + "step": 9726 + }, + { + "epoch": 0.82, + "learning_rate": 1.7059810902897907e-05, + "loss": 1.3345, + "step": 9727 + }, + { + "epoch": 0.82, + "learning_rate": 1.705916711113005e-05, + "loss": 1.1832, + "step": 9728 + }, + { + "epoch": 0.82, + "learning_rate": 1.7058523261037098e-05, + "loss": 1.2419, + "step": 9729 + }, + { + "epoch": 0.82, + "learning_rate": 1.705787935262436e-05, + "loss": 1.3534, + "step": 9730 + }, + { + "epoch": 0.82, + "learning_rate": 1.7057235385897168e-05, + "loss": 1.3215, + "step": 9731 + }, + { + "epoch": 0.82, + "learning_rate": 1.705659136086084e-05, + "loss": 1.2549, + "step": 9732 + }, + { + "epoch": 0.82, + "learning_rate": 1.7055947277520695e-05, + "loss": 1.2991, + "step": 9733 + }, + { + "epoch": 0.82, + "learning_rate": 1.7055303135882056e-05, + "loss": 1.349, + "step": 9734 + }, + { + "epoch": 0.82, + "learning_rate": 1.705465893595024e-05, + "loss": 1.2662, + "step": 9735 + }, + { + "epoch": 0.82, + "learning_rate": 1.7054014677730578e-05, + "loss": 1.3203, + "step": 9736 + }, + { + "epoch": 0.82, + "learning_rate": 1.705337036122839e-05, + "loss": 1.41, + "step": 9737 + }, + { + "epoch": 0.82, + "learning_rate": 1.7052725986449e-05, + "loss": 1.2859, + "step": 9738 + }, + { + "epoch": 0.82, + "learning_rate": 1.7052081553397724e-05, + "loss": 1.3215, + "step": 9739 + }, + { + "epoch": 0.82, + "learning_rate": 1.70514370620799e-05, + "loss": 1.3508, + "step": 9740 + }, + { + "epoch": 0.82, + "learning_rate": 1.7050792512500843e-05, + "loss": 1.2274, + "step": 9741 + }, + { + "epoch": 0.82, + "learning_rate": 1.7050147904665882e-05, + "loss": 1.3196, + "step": 9742 + }, + { + "epoch": 0.82, + "learning_rate": 1.7049503238580346e-05, + "loss": 1.3278, + "step": 9743 + }, + { + "epoch": 0.82, + "learning_rate": 1.7048858514249556e-05, + "loss": 1.2905, + "step": 9744 + }, + { + "epoch": 0.82, + "learning_rate": 1.704821373167884e-05, + "loss": 1.3284, + "step": 9745 + }, + { + "epoch": 0.82, + "learning_rate": 1.7047568890873524e-05, + "loss": 1.2666, + "step": 9746 + }, + { + "epoch": 0.82, + "learning_rate": 1.7046923991838943e-05, + "loss": 1.2993, + "step": 9747 + }, + { + "epoch": 0.82, + "learning_rate": 1.7046279034580417e-05, + "loss": 1.2427, + "step": 9748 + }, + { + "epoch": 0.82, + "learning_rate": 1.704563401910328e-05, + "loss": 1.3759, + "step": 9749 + }, + { + "epoch": 0.82, + "learning_rate": 1.704498894541286e-05, + "loss": 1.291, + "step": 9750 + }, + { + "epoch": 0.82, + "learning_rate": 1.7044343813514488e-05, + "loss": 1.2781, + "step": 9751 + }, + { + "epoch": 0.82, + "learning_rate": 1.704369862341349e-05, + "loss": 1.2242, + "step": 9752 + }, + { + "epoch": 0.82, + "learning_rate": 1.7043053375115203e-05, + "loss": 1.1851, + "step": 9753 + }, + { + "epoch": 0.82, + "learning_rate": 1.7042408068624952e-05, + "loss": 1.318, + "step": 9754 + }, + { + "epoch": 0.82, + "learning_rate": 1.7041762703948075e-05, + "loss": 1.317, + "step": 9755 + }, + { + "epoch": 0.82, + "learning_rate": 1.7041117281089898e-05, + "loss": 1.1791, + "step": 9756 + }, + { + "epoch": 0.82, + "learning_rate": 1.704047180005576e-05, + "loss": 1.2964, + "step": 9757 + }, + { + "epoch": 0.82, + "learning_rate": 1.7039826260850986e-05, + "loss": 1.2938, + "step": 9758 + }, + { + "epoch": 0.82, + "learning_rate": 1.7039180663480917e-05, + "loss": 1.3526, + "step": 9759 + }, + { + "epoch": 0.82, + "learning_rate": 1.7038535007950883e-05, + "loss": 1.3622, + "step": 9760 + }, + { + "epoch": 0.82, + "learning_rate": 1.7037889294266224e-05, + "loss": 1.2956, + "step": 9761 + }, + { + "epoch": 0.82, + "learning_rate": 1.7037243522432264e-05, + "loss": 1.2667, + "step": 9762 + }, + { + "epoch": 0.82, + "learning_rate": 1.7036597692454355e-05, + "loss": 1.2885, + "step": 9763 + }, + { + "epoch": 0.82, + "learning_rate": 1.7035951804337818e-05, + "loss": 1.356, + "step": 9764 + }, + { + "epoch": 0.82, + "learning_rate": 1.7035305858087997e-05, + "loss": 1.1648, + "step": 9765 + }, + { + "epoch": 0.82, + "learning_rate": 1.7034659853710226e-05, + "loss": 1.2717, + "step": 9766 + }, + { + "epoch": 0.82, + "learning_rate": 1.7034013791209846e-05, + "loss": 1.1998, + "step": 9767 + }, + { + "epoch": 0.82, + "learning_rate": 1.703336767059219e-05, + "loss": 1.3917, + "step": 9768 + }, + { + "epoch": 0.82, + "learning_rate": 1.70327214918626e-05, + "loss": 1.3001, + "step": 9769 + }, + { + "epoch": 0.82, + "learning_rate": 1.703207525502642e-05, + "loss": 1.374, + "step": 9770 + }, + { + "epoch": 0.82, + "learning_rate": 1.7031428960088977e-05, + "loss": 1.2002, + "step": 9771 + }, + { + "epoch": 0.82, + "learning_rate": 1.703078260705562e-05, + "loss": 1.2305, + "step": 9772 + }, + { + "epoch": 0.82, + "learning_rate": 1.7030136195931687e-05, + "loss": 1.3053, + "step": 9773 + }, + { + "epoch": 0.82, + "learning_rate": 1.7029489726722515e-05, + "loss": 1.3686, + "step": 9774 + }, + { + "epoch": 0.82, + "learning_rate": 1.7028843199433453e-05, + "loss": 1.295, + "step": 9775 + }, + { + "epoch": 0.82, + "learning_rate": 1.702819661406984e-05, + "loss": 1.2423, + "step": 9776 + }, + { + "epoch": 0.82, + "learning_rate": 1.7027549970637016e-05, + "loss": 1.1901, + "step": 9777 + }, + { + "epoch": 0.82, + "learning_rate": 1.7026903269140324e-05, + "loss": 1.3021, + "step": 9778 + }, + { + "epoch": 0.82, + "learning_rate": 1.702625650958511e-05, + "loss": 1.2531, + "step": 9779 + }, + { + "epoch": 0.82, + "learning_rate": 1.7025609691976715e-05, + "loss": 1.2504, + "step": 9780 + }, + { + "epoch": 0.82, + "learning_rate": 1.7024962816320485e-05, + "loss": 1.3501, + "step": 9781 + }, + { + "epoch": 0.82, + "learning_rate": 1.702431588262176e-05, + "loss": 1.2459, + "step": 9782 + }, + { + "epoch": 0.82, + "learning_rate": 1.7023668890885896e-05, + "loss": 1.2021, + "step": 9783 + }, + { + "epoch": 0.82, + "learning_rate": 1.7023021841118228e-05, + "loss": 1.262, + "step": 9784 + }, + { + "epoch": 0.82, + "learning_rate": 1.7022374733324104e-05, + "loss": 1.3713, + "step": 9785 + }, + { + "epoch": 0.82, + "learning_rate": 1.7021727567508877e-05, + "loss": 1.259, + "step": 9786 + }, + { + "epoch": 0.82, + "learning_rate": 1.7021080343677886e-05, + "loss": 1.2729, + "step": 9787 + }, + { + "epoch": 0.82, + "learning_rate": 1.7020433061836486e-05, + "loss": 1.2406, + "step": 9788 + }, + { + "epoch": 0.82, + "learning_rate": 1.7019785721990015e-05, + "loss": 1.3272, + "step": 9789 + }, + { + "epoch": 0.82, + "learning_rate": 1.7019138324143837e-05, + "loss": 1.2127, + "step": 9790 + }, + { + "epoch": 0.82, + "learning_rate": 1.7018490868303282e-05, + "loss": 1.3902, + "step": 9791 + }, + { + "epoch": 0.82, + "learning_rate": 1.7017843354473713e-05, + "loss": 1.2619, + "step": 9792 + }, + { + "epoch": 0.82, + "learning_rate": 1.701719578266048e-05, + "loss": 1.2797, + "step": 9793 + }, + { + "epoch": 0.82, + "learning_rate": 1.7016548152868925e-05, + "loss": 1.3033, + "step": 9794 + }, + { + "epoch": 0.82, + "learning_rate": 1.7015900465104403e-05, + "loss": 1.2362, + "step": 9795 + }, + { + "epoch": 0.82, + "learning_rate": 1.701525271937227e-05, + "loss": 1.3213, + "step": 9796 + }, + { + "epoch": 0.82, + "learning_rate": 1.7014604915677872e-05, + "loss": 1.3408, + "step": 9797 + }, + { + "epoch": 0.82, + "learning_rate": 1.7013957054026562e-05, + "loss": 1.3151, + "step": 9798 + }, + { + "epoch": 0.82, + "learning_rate": 1.7013309134423695e-05, + "loss": 1.2688, + "step": 9799 + }, + { + "epoch": 0.83, + "learning_rate": 1.7012661156874622e-05, + "loss": 1.3348, + "step": 9800 + }, + { + "epoch": 0.83, + "learning_rate": 1.70120131213847e-05, + "loss": 1.2561, + "step": 9801 + }, + { + "epoch": 0.83, + "learning_rate": 1.7011365027959282e-05, + "loss": 1.2353, + "step": 9802 + }, + { + "epoch": 0.83, + "learning_rate": 1.701071687660372e-05, + "loss": 1.3137, + "step": 9803 + }, + { + "epoch": 0.83, + "learning_rate": 1.7010068667323375e-05, + "loss": 1.3171, + "step": 9804 + }, + { + "epoch": 0.83, + "learning_rate": 1.7009420400123594e-05, + "loss": 1.3174, + "step": 9805 + }, + { + "epoch": 0.83, + "learning_rate": 1.7008772075009744e-05, + "loss": 1.3009, + "step": 9806 + }, + { + "epoch": 0.83, + "learning_rate": 1.700812369198717e-05, + "loss": 1.329, + "step": 9807 + }, + { + "epoch": 0.83, + "learning_rate": 1.700747525106124e-05, + "loss": 1.2388, + "step": 9808 + }, + { + "epoch": 0.83, + "learning_rate": 1.7006826752237305e-05, + "loss": 1.2619, + "step": 9809 + }, + { + "epoch": 0.83, + "learning_rate": 1.7006178195520723e-05, + "loss": 1.2947, + "step": 9810 + }, + { + "epoch": 0.83, + "learning_rate": 1.7005529580916857e-05, + "loss": 1.2759, + "step": 9811 + }, + { + "epoch": 0.83, + "learning_rate": 1.7004880908431065e-05, + "loss": 1.2288, + "step": 9812 + }, + { + "epoch": 0.83, + "learning_rate": 1.7004232178068704e-05, + "loss": 1.255, + "step": 9813 + }, + { + "epoch": 0.83, + "learning_rate": 1.7003583389835135e-05, + "loss": 1.2131, + "step": 9814 + }, + { + "epoch": 0.83, + "learning_rate": 1.7002934543735717e-05, + "loss": 1.2561, + "step": 9815 + }, + { + "epoch": 0.83, + "learning_rate": 1.7002285639775814e-05, + "loss": 1.258, + "step": 9816 + }, + { + "epoch": 0.83, + "learning_rate": 1.7001636677960785e-05, + "loss": 1.2924, + "step": 9817 + }, + { + "epoch": 0.83, + "learning_rate": 1.7000987658295992e-05, + "loss": 1.2859, + "step": 9818 + }, + { + "epoch": 0.83, + "learning_rate": 1.70003385807868e-05, + "loss": 1.2698, + "step": 9819 + }, + { + "epoch": 0.83, + "learning_rate": 1.6999689445438573e-05, + "loss": 1.3222, + "step": 9820 + }, + { + "epoch": 0.83, + "learning_rate": 1.6999040252256668e-05, + "loss": 1.2527, + "step": 9821 + }, + { + "epoch": 0.83, + "learning_rate": 1.699839100124645e-05, + "loss": 1.2245, + "step": 9822 + }, + { + "epoch": 0.83, + "learning_rate": 1.699774169241329e-05, + "loss": 1.1339, + "step": 9823 + }, + { + "epoch": 0.83, + "learning_rate": 1.6997092325762548e-05, + "loss": 1.3287, + "step": 9824 + }, + { + "epoch": 0.83, + "learning_rate": 1.699644290129959e-05, + "loss": 1.3392, + "step": 9825 + }, + { + "epoch": 0.83, + "learning_rate": 1.699579341902978e-05, + "loss": 1.2529, + "step": 9826 + }, + { + "epoch": 0.83, + "learning_rate": 1.6995143878958487e-05, + "loss": 1.2754, + "step": 9827 + }, + { + "epoch": 0.83, + "learning_rate": 1.6994494281091077e-05, + "loss": 1.3887, + "step": 9828 + }, + { + "epoch": 0.83, + "learning_rate": 1.6993844625432916e-05, + "loss": 1.2471, + "step": 9829 + }, + { + "epoch": 0.83, + "learning_rate": 1.6993194911989372e-05, + "loss": 1.357, + "step": 9830 + }, + { + "epoch": 0.83, + "learning_rate": 1.6992545140765814e-05, + "loss": 1.2959, + "step": 9831 + }, + { + "epoch": 0.83, + "learning_rate": 1.699189531176761e-05, + "loss": 1.2321, + "step": 9832 + }, + { + "epoch": 0.83, + "learning_rate": 1.699124542500013e-05, + "loss": 1.217, + "step": 9833 + }, + { + "epoch": 0.83, + "learning_rate": 1.6990595480468742e-05, + "loss": 1.3042, + "step": 9834 + }, + { + "epoch": 0.83, + "learning_rate": 1.6989945478178814e-05, + "loss": 1.3039, + "step": 9835 + }, + { + "epoch": 0.83, + "learning_rate": 1.6989295418135724e-05, + "loss": 1.2898, + "step": 9836 + }, + { + "epoch": 0.83, + "learning_rate": 1.6988645300344837e-05, + "loss": 1.3007, + "step": 9837 + }, + { + "epoch": 0.83, + "learning_rate": 1.6987995124811525e-05, + "loss": 1.3106, + "step": 9838 + }, + { + "epoch": 0.83, + "learning_rate": 1.6987344891541164e-05, + "loss": 1.3083, + "step": 9839 + }, + { + "epoch": 0.83, + "learning_rate": 1.6986694600539117e-05, + "loss": 1.3284, + "step": 9840 + }, + { + "epoch": 0.83, + "learning_rate": 1.698604425181077e-05, + "loss": 1.2554, + "step": 9841 + }, + { + "epoch": 0.83, + "learning_rate": 1.6985393845361486e-05, + "loss": 1.296, + "step": 9842 + }, + { + "epoch": 0.83, + "learning_rate": 1.6984743381196645e-05, + "loss": 1.3672, + "step": 9843 + }, + { + "epoch": 0.83, + "learning_rate": 1.6984092859321614e-05, + "loss": 1.1979, + "step": 9844 + }, + { + "epoch": 0.83, + "learning_rate": 1.6983442279741774e-05, + "loss": 1.2976, + "step": 9845 + }, + { + "epoch": 0.83, + "learning_rate": 1.6982791642462505e-05, + "loss": 1.3169, + "step": 9846 + }, + { + "epoch": 0.83, + "learning_rate": 1.6982140947489168e-05, + "loss": 1.1719, + "step": 9847 + }, + { + "epoch": 0.83, + "learning_rate": 1.6981490194827152e-05, + "loss": 1.2624, + "step": 9848 + }, + { + "epoch": 0.83, + "learning_rate": 1.698083938448183e-05, + "loss": 1.2411, + "step": 9849 + }, + { + "epoch": 0.83, + "learning_rate": 1.6980188516458578e-05, + "loss": 1.3115, + "step": 9850 + }, + { + "epoch": 0.83, + "learning_rate": 1.6979537590762777e-05, + "loss": 1.2509, + "step": 9851 + }, + { + "epoch": 0.83, + "learning_rate": 1.69788866073998e-05, + "loss": 1.3111, + "step": 9852 + }, + { + "epoch": 0.83, + "learning_rate": 1.6978235566375026e-05, + "loss": 1.2604, + "step": 9853 + }, + { + "epoch": 0.83, + "learning_rate": 1.697758446769384e-05, + "loss": 1.2816, + "step": 9854 + }, + { + "epoch": 0.83, + "learning_rate": 1.6976933311361617e-05, + "loss": 1.2692, + "step": 9855 + }, + { + "epoch": 0.83, + "learning_rate": 1.697628209738374e-05, + "loss": 1.3619, + "step": 9856 + }, + { + "epoch": 0.83, + "learning_rate": 1.6975630825765586e-05, + "loss": 1.3997, + "step": 9857 + }, + { + "epoch": 0.83, + "learning_rate": 1.6974979496512538e-05, + "loss": 1.3945, + "step": 9858 + }, + { + "epoch": 0.83, + "learning_rate": 1.6974328109629974e-05, + "loss": 1.2837, + "step": 9859 + }, + { + "epoch": 0.83, + "learning_rate": 1.6973676665123285e-05, + "loss": 1.3015, + "step": 9860 + }, + { + "epoch": 0.83, + "learning_rate": 1.6973025162997843e-05, + "loss": 1.3506, + "step": 9861 + }, + { + "epoch": 0.83, + "learning_rate": 1.6972373603259037e-05, + "loss": 1.362, + "step": 9862 + }, + { + "epoch": 0.83, + "learning_rate": 1.697172198591225e-05, + "loss": 1.2962, + "step": 9863 + }, + { + "epoch": 0.83, + "learning_rate": 1.697107031096286e-05, + "loss": 1.2749, + "step": 9864 + }, + { + "epoch": 0.83, + "learning_rate": 1.697041857841626e-05, + "loss": 1.1753, + "step": 9865 + }, + { + "epoch": 0.83, + "learning_rate": 1.6969766788277834e-05, + "loss": 1.2613, + "step": 9866 + }, + { + "epoch": 0.83, + "learning_rate": 1.696911494055296e-05, + "loss": 1.2874, + "step": 9867 + }, + { + "epoch": 0.83, + "learning_rate": 1.6968463035247027e-05, + "loss": 1.3018, + "step": 9868 + }, + { + "epoch": 0.83, + "learning_rate": 1.6967811072365423e-05, + "loss": 1.2846, + "step": 9869 + }, + { + "epoch": 0.83, + "learning_rate": 1.6967159051913535e-05, + "loss": 1.2619, + "step": 9870 + }, + { + "epoch": 0.83, + "learning_rate": 1.6966506973896752e-05, + "loss": 1.3018, + "step": 9871 + }, + { + "epoch": 0.83, + "learning_rate": 1.696585483832045e-05, + "loss": 1.2369, + "step": 9872 + }, + { + "epoch": 0.83, + "learning_rate": 1.6965202645190034e-05, + "loss": 1.2409, + "step": 9873 + }, + { + "epoch": 0.83, + "learning_rate": 1.696455039451088e-05, + "loss": 1.1704, + "step": 9874 + }, + { + "epoch": 0.83, + "learning_rate": 1.6963898086288385e-05, + "loss": 1.3515, + "step": 9875 + }, + { + "epoch": 0.83, + "learning_rate": 1.6963245720527933e-05, + "loss": 1.2774, + "step": 9876 + }, + { + "epoch": 0.83, + "learning_rate": 1.696259329723492e-05, + "loss": 1.2631, + "step": 9877 + }, + { + "epoch": 0.83, + "learning_rate": 1.696194081641473e-05, + "loss": 1.27, + "step": 9878 + }, + { + "epoch": 0.83, + "learning_rate": 1.6961288278072755e-05, + "loss": 1.2103, + "step": 9879 + }, + { + "epoch": 0.83, + "learning_rate": 1.6960635682214394e-05, + "loss": 1.245, + "step": 9880 + }, + { + "epoch": 0.83, + "learning_rate": 1.695998302884503e-05, + "loss": 1.3069, + "step": 9881 + }, + { + "epoch": 0.83, + "learning_rate": 1.6959330317970057e-05, + "loss": 1.2983, + "step": 9882 + }, + { + "epoch": 0.83, + "learning_rate": 1.695867754959487e-05, + "loss": 1.3031, + "step": 9883 + }, + { + "epoch": 0.83, + "learning_rate": 1.6958024723724864e-05, + "loss": 1.3411, + "step": 9884 + }, + { + "epoch": 0.83, + "learning_rate": 1.6957371840365428e-05, + "loss": 1.267, + "step": 9885 + }, + { + "epoch": 0.83, + "learning_rate": 1.6956718899521962e-05, + "loss": 1.3538, + "step": 9886 + }, + { + "epoch": 0.83, + "learning_rate": 1.6956065901199857e-05, + "loss": 1.2839, + "step": 9887 + }, + { + "epoch": 0.83, + "learning_rate": 1.695541284540451e-05, + "loss": 1.2687, + "step": 9888 + }, + { + "epoch": 0.83, + "learning_rate": 1.6954759732141317e-05, + "loss": 1.3611, + "step": 9889 + }, + { + "epoch": 0.83, + "learning_rate": 1.6954106561415674e-05, + "loss": 1.2549, + "step": 9890 + }, + { + "epoch": 0.83, + "learning_rate": 1.695345333323297e-05, + "loss": 1.2938, + "step": 9891 + }, + { + "epoch": 0.83, + "learning_rate": 1.6952800047598618e-05, + "loss": 1.2676, + "step": 9892 + }, + { + "epoch": 0.83, + "learning_rate": 1.6952146704518003e-05, + "loss": 1.389, + "step": 9893 + }, + { + "epoch": 0.83, + "learning_rate": 1.6951493303996526e-05, + "loss": 1.1589, + "step": 9894 + }, + { + "epoch": 0.83, + "learning_rate": 1.6950839846039588e-05, + "loss": 1.4136, + "step": 9895 + }, + { + "epoch": 0.83, + "learning_rate": 1.6950186330652586e-05, + "loss": 1.2459, + "step": 9896 + }, + { + "epoch": 0.83, + "learning_rate": 1.694953275784092e-05, + "loss": 1.2648, + "step": 9897 + }, + { + "epoch": 0.83, + "learning_rate": 1.6948879127609988e-05, + "loss": 1.2752, + "step": 9898 + }, + { + "epoch": 0.83, + "learning_rate": 1.69482254399652e-05, + "loss": 1.2477, + "step": 9899 + }, + { + "epoch": 0.83, + "learning_rate": 1.6947571694911943e-05, + "loss": 1.2684, + "step": 9900 + }, + { + "epoch": 0.83, + "learning_rate": 1.6946917892455626e-05, + "loss": 1.2997, + "step": 9901 + }, + { + "epoch": 0.83, + "learning_rate": 1.694626403260165e-05, + "loss": 1.2832, + "step": 9902 + }, + { + "epoch": 0.83, + "learning_rate": 1.694561011535542e-05, + "loss": 1.3565, + "step": 9903 + }, + { + "epoch": 0.83, + "learning_rate": 1.6944956140722335e-05, + "loss": 1.3407, + "step": 9904 + }, + { + "epoch": 0.83, + "learning_rate": 1.69443021087078e-05, + "loss": 1.2035, + "step": 9905 + }, + { + "epoch": 0.83, + "learning_rate": 1.694364801931722e-05, + "loss": 1.3003, + "step": 9906 + }, + { + "epoch": 0.83, + "learning_rate": 1.6942993872555992e-05, + "loss": 1.2856, + "step": 9907 + }, + { + "epoch": 0.83, + "learning_rate": 1.6942339668429534e-05, + "loss": 1.2132, + "step": 9908 + }, + { + "epoch": 0.83, + "learning_rate": 1.694168540694324e-05, + "loss": 1.3024, + "step": 9909 + }, + { + "epoch": 0.83, + "learning_rate": 1.6941031088102517e-05, + "loss": 1.1969, + "step": 9910 + }, + { + "epoch": 0.83, + "learning_rate": 1.6940376711912777e-05, + "loss": 1.1931, + "step": 9911 + }, + { + "epoch": 0.83, + "learning_rate": 1.6939722278379425e-05, + "loss": 1.2148, + "step": 9912 + }, + { + "epoch": 0.83, + "learning_rate": 1.6939067787507864e-05, + "loss": 1.3231, + "step": 9913 + }, + { + "epoch": 0.83, + "learning_rate": 1.6938413239303504e-05, + "loss": 1.1771, + "step": 9914 + }, + { + "epoch": 0.83, + "learning_rate": 1.6937758633771756e-05, + "loss": 1.323, + "step": 9915 + }, + { + "epoch": 0.83, + "learning_rate": 1.693710397091802e-05, + "loss": 1.1769, + "step": 9916 + }, + { + "epoch": 0.83, + "learning_rate": 1.693644925074772e-05, + "loss": 1.3663, + "step": 9917 + }, + { + "epoch": 0.84, + "learning_rate": 1.693579447326625e-05, + "loss": 1.1974, + "step": 9918 + }, + { + "epoch": 0.84, + "learning_rate": 1.6935139638479026e-05, + "loss": 1.2777, + "step": 9919 + }, + { + "epoch": 0.84, + "learning_rate": 1.693448474639146e-05, + "loss": 1.3087, + "step": 9920 + }, + { + "epoch": 0.84, + "learning_rate": 1.6933829797008963e-05, + "loss": 1.158, + "step": 9921 + }, + { + "epoch": 0.84, + "learning_rate": 1.693317479033694e-05, + "loss": 1.2573, + "step": 9922 + }, + { + "epoch": 0.84, + "learning_rate": 1.6932519726380815e-05, + "loss": 1.2464, + "step": 9923 + }, + { + "epoch": 0.84, + "learning_rate": 1.693186460514599e-05, + "loss": 1.309, + "step": 9924 + }, + { + "epoch": 0.84, + "learning_rate": 1.693120942663788e-05, + "loss": 1.1945, + "step": 9925 + }, + { + "epoch": 0.84, + "learning_rate": 1.69305541908619e-05, + "loss": 1.2351, + "step": 9926 + }, + { + "epoch": 0.84, + "learning_rate": 1.6929898897823465e-05, + "loss": 1.2989, + "step": 9927 + }, + { + "epoch": 0.84, + "learning_rate": 1.6929243547527987e-05, + "loss": 1.3413, + "step": 9928 + }, + { + "epoch": 0.84, + "learning_rate": 1.692858813998088e-05, + "loss": 1.1975, + "step": 9929 + }, + { + "epoch": 0.84, + "learning_rate": 1.6927932675187562e-05, + "loss": 1.2794, + "step": 9930 + }, + { + "epoch": 0.84, + "learning_rate": 1.6927277153153444e-05, + "loss": 1.2984, + "step": 9931 + }, + { + "epoch": 0.84, + "learning_rate": 1.6926621573883947e-05, + "loss": 1.2213, + "step": 9932 + }, + { + "epoch": 0.84, + "learning_rate": 1.692596593738449e-05, + "loss": 1.284, + "step": 9933 + }, + { + "epoch": 0.84, + "learning_rate": 1.692531024366048e-05, + "loss": 1.268, + "step": 9934 + }, + { + "epoch": 0.84, + "learning_rate": 1.692465449271734e-05, + "loss": 1.2568, + "step": 9935 + }, + { + "epoch": 0.84, + "learning_rate": 1.6923998684560492e-05, + "loss": 1.265, + "step": 9936 + }, + { + "epoch": 0.84, + "learning_rate": 1.6923342819195348e-05, + "loss": 1.271, + "step": 9937 + }, + { + "epoch": 0.84, + "learning_rate": 1.692268689662733e-05, + "loss": 1.2661, + "step": 9938 + }, + { + "epoch": 0.84, + "learning_rate": 1.692203091686186e-05, + "loss": 1.3052, + "step": 9939 + }, + { + "epoch": 0.84, + "learning_rate": 1.6921374879904354e-05, + "loss": 1.2625, + "step": 9940 + }, + { + "epoch": 0.84, + "learning_rate": 1.6920718785760235e-05, + "loss": 1.2543, + "step": 9941 + }, + { + "epoch": 0.84, + "learning_rate": 1.692006263443492e-05, + "loss": 1.3202, + "step": 9942 + }, + { + "epoch": 0.84, + "learning_rate": 1.6919406425933837e-05, + "loss": 1.317, + "step": 9943 + }, + { + "epoch": 0.84, + "learning_rate": 1.69187501602624e-05, + "loss": 1.2223, + "step": 9944 + }, + { + "epoch": 0.84, + "learning_rate": 1.6918093837426032e-05, + "loss": 1.2594, + "step": 9945 + }, + { + "epoch": 0.84, + "learning_rate": 1.6917437457430165e-05, + "loss": 1.1891, + "step": 9946 + }, + { + "epoch": 0.84, + "learning_rate": 1.691678102028021e-05, + "loss": 1.2574, + "step": 9947 + }, + { + "epoch": 0.84, + "learning_rate": 1.69161245259816e-05, + "loss": 1.3048, + "step": 9948 + }, + { + "epoch": 0.84, + "learning_rate": 1.6915467974539755e-05, + "loss": 1.2348, + "step": 9949 + }, + { + "epoch": 0.84, + "learning_rate": 1.6914811365960103e-05, + "loss": 1.259, + "step": 9950 + }, + { + "epoch": 0.84, + "learning_rate": 1.6914154700248062e-05, + "loss": 1.3083, + "step": 9951 + }, + { + "epoch": 0.84, + "learning_rate": 1.6913497977409067e-05, + "loss": 1.1876, + "step": 9952 + }, + { + "epoch": 0.84, + "learning_rate": 1.6912841197448534e-05, + "loss": 1.3083, + "step": 9953 + }, + { + "epoch": 0.84, + "learning_rate": 1.69121843603719e-05, + "loss": 1.2789, + "step": 9954 + }, + { + "epoch": 0.84, + "learning_rate": 1.691152746618458e-05, + "loss": 1.2901, + "step": 9955 + }, + { + "epoch": 0.84, + "learning_rate": 1.6910870514892012e-05, + "loss": 1.201, + "step": 9956 + }, + { + "epoch": 0.84, + "learning_rate": 1.691021350649962e-05, + "loss": 1.237, + "step": 9957 + }, + { + "epoch": 0.84, + "learning_rate": 1.690955644101283e-05, + "loss": 1.2779, + "step": 9958 + }, + { + "epoch": 0.84, + "learning_rate": 1.6908899318437077e-05, + "loss": 1.1808, + "step": 9959 + }, + { + "epoch": 0.84, + "learning_rate": 1.6908242138777783e-05, + "loss": 1.2434, + "step": 9960 + }, + { + "epoch": 0.84, + "learning_rate": 1.6907584902040385e-05, + "loss": 1.2952, + "step": 9961 + }, + { + "epoch": 0.84, + "learning_rate": 1.6906927608230307e-05, + "loss": 1.2373, + "step": 9962 + }, + { + "epoch": 0.84, + "learning_rate": 1.6906270257352983e-05, + "loss": 1.3322, + "step": 9963 + }, + { + "epoch": 0.84, + "learning_rate": 1.6905612849413847e-05, + "loss": 1.2778, + "step": 9964 + }, + { + "epoch": 0.84, + "learning_rate": 1.6904955384418326e-05, + "loss": 1.2123, + "step": 9965 + }, + { + "epoch": 0.84, + "learning_rate": 1.6904297862371852e-05, + "loss": 1.3362, + "step": 9966 + }, + { + "epoch": 0.84, + "learning_rate": 1.690364028327986e-05, + "loss": 1.3333, + "step": 9967 + }, + { + "epoch": 0.84, + "learning_rate": 1.6902982647147783e-05, + "loss": 1.2133, + "step": 9968 + }, + { + "epoch": 0.84, + "learning_rate": 1.6902324953981053e-05, + "loss": 1.3138, + "step": 9969 + }, + { + "epoch": 0.84, + "learning_rate": 1.690166720378511e-05, + "loss": 1.3229, + "step": 9970 + }, + { + "epoch": 0.84, + "learning_rate": 1.6901009396565377e-05, + "loss": 1.2675, + "step": 9971 + }, + { + "epoch": 0.84, + "learning_rate": 1.6900351532327296e-05, + "loss": 1.2897, + "step": 9972 + }, + { + "epoch": 0.84, + "learning_rate": 1.6899693611076305e-05, + "loss": 1.2914, + "step": 9973 + }, + { + "epoch": 0.84, + "learning_rate": 1.6899035632817838e-05, + "loss": 1.3181, + "step": 9974 + }, + { + "epoch": 0.84, + "learning_rate": 1.6898377597557334e-05, + "loss": 1.2552, + "step": 9975 + }, + { + "epoch": 0.84, + "learning_rate": 1.689771950530022e-05, + "loss": 1.4349, + "step": 9976 + }, + { + "epoch": 0.84, + "learning_rate": 1.6897061356051944e-05, + "loss": 1.2826, + "step": 9977 + }, + { + "epoch": 0.84, + "learning_rate": 1.6896403149817937e-05, + "loss": 1.3004, + "step": 9978 + }, + { + "epoch": 0.84, + "learning_rate": 1.689574488660364e-05, + "loss": 1.3162, + "step": 9979 + }, + { + "epoch": 0.84, + "learning_rate": 1.6895086566414495e-05, + "loss": 1.3415, + "step": 9980 + }, + { + "epoch": 0.84, + "learning_rate": 1.6894428189255934e-05, + "loss": 1.265, + "step": 9981 + }, + { + "epoch": 0.84, + "learning_rate": 1.6893769755133406e-05, + "loss": 1.2444, + "step": 9982 + }, + { + "epoch": 0.84, + "learning_rate": 1.6893111264052343e-05, + "loss": 1.2688, + "step": 9983 + }, + { + "epoch": 0.84, + "learning_rate": 1.6892452716018187e-05, + "loss": 1.3853, + "step": 9984 + }, + { + "epoch": 0.84, + "learning_rate": 1.689179411103638e-05, + "loss": 1.2403, + "step": 9985 + }, + { + "epoch": 0.84, + "learning_rate": 1.6891135449112367e-05, + "loss": 1.2987, + "step": 9986 + }, + { + "epoch": 0.84, + "learning_rate": 1.6890476730251588e-05, + "loss": 1.2865, + "step": 9987 + }, + { + "epoch": 0.84, + "learning_rate": 1.6889817954459485e-05, + "loss": 1.2594, + "step": 9988 + }, + { + "epoch": 0.84, + "learning_rate": 1.68891591217415e-05, + "loss": 1.2672, + "step": 9989 + }, + { + "epoch": 0.84, + "learning_rate": 1.6888500232103076e-05, + "loss": 1.3749, + "step": 9990 + }, + { + "epoch": 0.84, + "learning_rate": 1.6887841285549663e-05, + "loss": 1.2388, + "step": 9991 + }, + { + "epoch": 0.84, + "learning_rate": 1.6887182282086695e-05, + "loss": 1.3328, + "step": 9992 + }, + { + "epoch": 0.84, + "learning_rate": 1.6886523221719627e-05, + "loss": 1.2123, + "step": 9993 + }, + { + "epoch": 0.84, + "learning_rate": 1.6885864104453903e-05, + "loss": 1.2627, + "step": 9994 + }, + { + "epoch": 0.84, + "learning_rate": 1.688520493029496e-05, + "loss": 1.4147, + "step": 9995 + }, + { + "epoch": 0.84, + "learning_rate": 1.6884545699248253e-05, + "loss": 1.2902, + "step": 9996 + }, + { + "epoch": 0.84, + "learning_rate": 1.6883886411319224e-05, + "loss": 1.2829, + "step": 9997 + }, + { + "epoch": 0.84, + "learning_rate": 1.6883227066513325e-05, + "loss": 1.3396, + "step": 9998 + }, + { + "epoch": 0.84, + "learning_rate": 1.6882567664835997e-05, + "loss": 1.2331, + "step": 9999 + }, + { + "epoch": 0.84, + "learning_rate": 1.6881908206292696e-05, + "loss": 1.3218, + "step": 10000 + }, + { + "epoch": 0.84, + "learning_rate": 1.6881248690888866e-05, + "loss": 1.2212, + "step": 10001 + }, + { + "epoch": 0.84, + "learning_rate": 1.6880589118629958e-05, + "loss": 1.2462, + "step": 10002 + }, + { + "epoch": 0.84, + "learning_rate": 1.6879929489521418e-05, + "loss": 1.3365, + "step": 10003 + }, + { + "epoch": 0.84, + "learning_rate": 1.6879269803568703e-05, + "loss": 1.2203, + "step": 10004 + }, + { + "epoch": 0.84, + "learning_rate": 1.6878610060777255e-05, + "loss": 1.3138, + "step": 10005 + }, + { + "epoch": 0.84, + "learning_rate": 1.6877950261152533e-05, + "loss": 1.2257, + "step": 10006 + }, + { + "epoch": 0.84, + "learning_rate": 1.6877290404699983e-05, + "loss": 1.1727, + "step": 10007 + }, + { + "epoch": 0.84, + "learning_rate": 1.687663049142506e-05, + "loss": 1.3403, + "step": 10008 + }, + { + "epoch": 0.84, + "learning_rate": 1.687597052133321e-05, + "loss": 1.3142, + "step": 10009 + }, + { + "epoch": 0.84, + "learning_rate": 1.6875310494429894e-05, + "loss": 1.2216, + "step": 10010 + }, + { + "epoch": 0.84, + "learning_rate": 1.6874650410720563e-05, + "loss": 1.2577, + "step": 10011 + }, + { + "epoch": 0.84, + "learning_rate": 1.687399027021067e-05, + "loss": 1.2748, + "step": 10012 + }, + { + "epoch": 0.84, + "learning_rate": 1.687333007290567e-05, + "loss": 1.3623, + "step": 10013 + }, + { + "epoch": 0.84, + "learning_rate": 1.6872669818811016e-05, + "loss": 1.251, + "step": 10014 + }, + { + "epoch": 0.84, + "learning_rate": 1.6872009507932165e-05, + "loss": 1.2616, + "step": 10015 + }, + { + "epoch": 0.84, + "learning_rate": 1.687134914027457e-05, + "loss": 1.1927, + "step": 10016 + }, + { + "epoch": 0.84, + "learning_rate": 1.6870688715843693e-05, + "loss": 1.2666, + "step": 10017 + }, + { + "epoch": 0.84, + "learning_rate": 1.6870028234644985e-05, + "loss": 1.2148, + "step": 10018 + }, + { + "epoch": 0.84, + "learning_rate": 1.6869367696683906e-05, + "loss": 1.24, + "step": 10019 + }, + { + "epoch": 0.84, + "learning_rate": 1.6868707101965913e-05, + "loss": 1.318, + "step": 10020 + }, + { + "epoch": 0.84, + "learning_rate": 1.6868046450496463e-05, + "loss": 1.3493, + "step": 10021 + }, + { + "epoch": 0.84, + "learning_rate": 1.6867385742281012e-05, + "loss": 1.3334, + "step": 10022 + }, + { + "epoch": 0.84, + "learning_rate": 1.6866724977325027e-05, + "loss": 1.3019, + "step": 10023 + }, + { + "epoch": 0.84, + "learning_rate": 1.6866064155633958e-05, + "loss": 1.2047, + "step": 10024 + }, + { + "epoch": 0.84, + "learning_rate": 1.686540327721327e-05, + "loss": 1.2721, + "step": 10025 + }, + { + "epoch": 0.84, + "learning_rate": 1.6864742342068424e-05, + "loss": 1.3179, + "step": 10026 + }, + { + "epoch": 0.84, + "learning_rate": 1.6864081350204882e-05, + "loss": 1.3204, + "step": 10027 + }, + { + "epoch": 0.84, + "learning_rate": 1.6863420301628104e-05, + "loss": 1.2592, + "step": 10028 + }, + { + "epoch": 0.84, + "learning_rate": 1.6862759196343547e-05, + "loss": 1.2301, + "step": 10029 + }, + { + "epoch": 0.84, + "learning_rate": 1.686209803435668e-05, + "loss": 1.2459, + "step": 10030 + }, + { + "epoch": 0.84, + "learning_rate": 1.686143681567296e-05, + "loss": 1.2332, + "step": 10031 + }, + { + "epoch": 0.84, + "learning_rate": 1.6860775540297856e-05, + "loss": 1.353, + "step": 10032 + }, + { + "epoch": 0.84, + "learning_rate": 1.686011420823683e-05, + "loss": 1.2527, + "step": 10033 + }, + { + "epoch": 0.84, + "learning_rate": 1.685945281949534e-05, + "loss": 1.2536, + "step": 10034 + }, + { + "epoch": 0.84, + "learning_rate": 1.6858791374078854e-05, + "loss": 1.2947, + "step": 10035 + }, + { + "epoch": 0.84, + "learning_rate": 1.6858129871992846e-05, + "loss": 1.238, + "step": 10036 + }, + { + "epoch": 0.85, + "learning_rate": 1.685746831324277e-05, + "loss": 1.3072, + "step": 10037 + }, + { + "epoch": 0.85, + "learning_rate": 1.6856806697834098e-05, + "loss": 1.374, + "step": 10038 + }, + { + "epoch": 0.85, + "learning_rate": 1.685614502577229e-05, + "loss": 1.2632, + "step": 10039 + }, + { + "epoch": 0.85, + "learning_rate": 1.685548329706282e-05, + "loss": 1.2588, + "step": 10040 + }, + { + "epoch": 0.85, + "learning_rate": 1.6854821511711158e-05, + "loss": 1.2992, + "step": 10041 + }, + { + "epoch": 0.85, + "learning_rate": 1.685415966972276e-05, + "loss": 1.2818, + "step": 10042 + }, + { + "epoch": 0.85, + "learning_rate": 1.6853497771103104e-05, + "loss": 1.3085, + "step": 10043 + }, + { + "epoch": 0.85, + "learning_rate": 1.6852835815857657e-05, + "loss": 1.2667, + "step": 10044 + }, + { + "epoch": 0.85, + "learning_rate": 1.6852173803991887e-05, + "loss": 1.3203, + "step": 10045 + }, + { + "epoch": 0.85, + "learning_rate": 1.6851511735511264e-05, + "loss": 1.2298, + "step": 10046 + }, + { + "epoch": 0.85, + "learning_rate": 1.6850849610421255e-05, + "loss": 1.2697, + "step": 10047 + }, + { + "epoch": 0.85, + "learning_rate": 1.6850187428727337e-05, + "loss": 1.2334, + "step": 10048 + }, + { + "epoch": 0.85, + "learning_rate": 1.6849525190434975e-05, + "loss": 1.2285, + "step": 10049 + }, + { + "epoch": 0.85, + "learning_rate": 1.684886289554965e-05, + "loss": 1.3008, + "step": 10050 + }, + { + "epoch": 0.85, + "learning_rate": 1.6848200544076824e-05, + "loss": 1.2792, + "step": 10051 + }, + { + "epoch": 0.85, + "learning_rate": 1.6847538136021973e-05, + "loss": 1.2902, + "step": 10052 + }, + { + "epoch": 0.85, + "learning_rate": 1.684687567139057e-05, + "loss": 1.3483, + "step": 10053 + }, + { + "epoch": 0.85, + "learning_rate": 1.6846213150188093e-05, + "loss": 1.1356, + "step": 10054 + }, + { + "epoch": 0.85, + "learning_rate": 1.6845550572420007e-05, + "loss": 1.2874, + "step": 10055 + }, + { + "epoch": 0.85, + "learning_rate": 1.6844887938091794e-05, + "loss": 1.2716, + "step": 10056 + }, + { + "epoch": 0.85, + "learning_rate": 1.6844225247208926e-05, + "loss": 1.3925, + "step": 10057 + }, + { + "epoch": 0.85, + "learning_rate": 1.6843562499776878e-05, + "loss": 1.2126, + "step": 10058 + }, + { + "epoch": 0.85, + "learning_rate": 1.6842899695801125e-05, + "loss": 1.343, + "step": 10059 + }, + { + "epoch": 0.85, + "learning_rate": 1.6842236835287146e-05, + "loss": 1.3131, + "step": 10060 + }, + { + "epoch": 0.85, + "learning_rate": 1.6841573918240416e-05, + "loss": 1.1474, + "step": 10061 + }, + { + "epoch": 0.85, + "learning_rate": 1.6840910944666414e-05, + "loss": 1.3745, + "step": 10062 + }, + { + "epoch": 0.85, + "learning_rate": 1.6840247914570615e-05, + "loss": 1.2366, + "step": 10063 + }, + { + "epoch": 0.85, + "learning_rate": 1.6839584827958497e-05, + "loss": 1.2422, + "step": 10064 + }, + { + "epoch": 0.85, + "learning_rate": 1.6838921684835546e-05, + "loss": 1.2956, + "step": 10065 + }, + { + "epoch": 0.85, + "learning_rate": 1.6838258485207228e-05, + "loss": 1.3175, + "step": 10066 + }, + { + "epoch": 0.85, + "learning_rate": 1.6837595229079037e-05, + "loss": 1.2425, + "step": 10067 + }, + { + "epoch": 0.85, + "learning_rate": 1.683693191645644e-05, + "loss": 1.2035, + "step": 10068 + }, + { + "epoch": 0.85, + "learning_rate": 1.6836268547344927e-05, + "loss": 1.3061, + "step": 10069 + }, + { + "epoch": 0.85, + "learning_rate": 1.6835605121749974e-05, + "loss": 1.3739, + "step": 10070 + }, + { + "epoch": 0.85, + "learning_rate": 1.6834941639677065e-05, + "loss": 1.3183, + "step": 10071 + }, + { + "epoch": 0.85, + "learning_rate": 1.6834278101131677e-05, + "loss": 1.2832, + "step": 10072 + }, + { + "epoch": 0.85, + "learning_rate": 1.68336145061193e-05, + "loss": 1.2995, + "step": 10073 + }, + { + "epoch": 0.85, + "learning_rate": 1.683295085464541e-05, + "loss": 1.2101, + "step": 10074 + }, + { + "epoch": 0.85, + "learning_rate": 1.6832287146715498e-05, + "loss": 1.2332, + "step": 10075 + }, + { + "epoch": 0.85, + "learning_rate": 1.683162338233504e-05, + "loss": 1.2061, + "step": 10076 + }, + { + "epoch": 0.85, + "learning_rate": 1.683095956150952e-05, + "loss": 1.3422, + "step": 10077 + }, + { + "epoch": 0.85, + "learning_rate": 1.683029568424443e-05, + "loss": 1.2737, + "step": 10078 + }, + { + "epoch": 0.85, + "learning_rate": 1.6829631750545253e-05, + "loss": 1.3297, + "step": 10079 + }, + { + "epoch": 0.85, + "learning_rate": 1.682896776041747e-05, + "loss": 1.3173, + "step": 10080 + }, + { + "epoch": 0.85, + "learning_rate": 1.682830371386657e-05, + "loss": 1.3207, + "step": 10081 + }, + { + "epoch": 0.85, + "learning_rate": 1.682763961089804e-05, + "loss": 1.2685, + "step": 10082 + }, + { + "epoch": 0.85, + "learning_rate": 1.6826975451517365e-05, + "loss": 1.2134, + "step": 10083 + }, + { + "epoch": 0.85, + "learning_rate": 1.6826311235730033e-05, + "loss": 1.3222, + "step": 10084 + }, + { + "epoch": 0.85, + "learning_rate": 1.6825646963541532e-05, + "loss": 1.2895, + "step": 10085 + }, + { + "epoch": 0.85, + "learning_rate": 1.6824982634957354e-05, + "loss": 1.3737, + "step": 10086 + }, + { + "epoch": 0.85, + "learning_rate": 1.6824318249982984e-05, + "loss": 1.2546, + "step": 10087 + }, + { + "epoch": 0.85, + "learning_rate": 1.6823653808623914e-05, + "loss": 1.2578, + "step": 10088 + }, + { + "epoch": 0.85, + "learning_rate": 1.682298931088563e-05, + "loss": 1.2042, + "step": 10089 + }, + { + "epoch": 0.85, + "learning_rate": 1.682232475677363e-05, + "loss": 1.2756, + "step": 10090 + }, + { + "epoch": 0.85, + "learning_rate": 1.6821660146293394e-05, + "loss": 1.2259, + "step": 10091 + }, + { + "epoch": 0.85, + "learning_rate": 1.6820995479450423e-05, + "loss": 1.279, + "step": 10092 + }, + { + "epoch": 0.85, + "learning_rate": 1.68203307562502e-05, + "loss": 1.3187, + "step": 10093 + }, + { + "epoch": 0.85, + "learning_rate": 1.6819665976698223e-05, + "loss": 1.277, + "step": 10094 + }, + { + "epoch": 0.85, + "learning_rate": 1.6819001140799985e-05, + "loss": 1.2857, + "step": 10095 + }, + { + "epoch": 0.85, + "learning_rate": 1.6818336248560974e-05, + "loss": 1.1864, + "step": 10096 + }, + { + "epoch": 0.85, + "learning_rate": 1.681767129998669e-05, + "loss": 1.2569, + "step": 10097 + }, + { + "epoch": 0.85, + "learning_rate": 1.6817006295082618e-05, + "loss": 1.2687, + "step": 10098 + }, + { + "epoch": 0.85, + "learning_rate": 1.6816341233854266e-05, + "loss": 1.3147, + "step": 10099 + }, + { + "epoch": 0.85, + "learning_rate": 1.6815676116307117e-05, + "loss": 1.3086, + "step": 10100 + }, + { + "epoch": 0.85, + "learning_rate": 1.6815010942446674e-05, + "loss": 1.2408, + "step": 10101 + }, + { + "epoch": 0.85, + "learning_rate": 1.6814345712278424e-05, + "loss": 1.2832, + "step": 10102 + }, + { + "epoch": 0.85, + "learning_rate": 1.6813680425807874e-05, + "loss": 1.3749, + "step": 10103 + }, + { + "epoch": 0.85, + "learning_rate": 1.6813015083040512e-05, + "loss": 1.3131, + "step": 10104 + }, + { + "epoch": 0.85, + "learning_rate": 1.6812349683981838e-05, + "loss": 1.2112, + "step": 10105 + }, + { + "epoch": 0.85, + "learning_rate": 1.6811684228637354e-05, + "loss": 1.3371, + "step": 10106 + }, + { + "epoch": 0.85, + "learning_rate": 1.6811018717012554e-05, + "loss": 1.1953, + "step": 10107 + }, + { + "epoch": 0.85, + "learning_rate": 1.681035314911294e-05, + "loss": 1.2407, + "step": 10108 + }, + { + "epoch": 0.85, + "learning_rate": 1.6809687524944004e-05, + "loss": 1.3461, + "step": 10109 + }, + { + "epoch": 0.85, + "learning_rate": 1.6809021844511254e-05, + "loss": 1.2838, + "step": 10110 + }, + { + "epoch": 0.85, + "learning_rate": 1.6808356107820182e-05, + "loss": 1.3274, + "step": 10111 + }, + { + "epoch": 0.85, + "learning_rate": 1.6807690314876296e-05, + "loss": 1.2599, + "step": 10112 + }, + { + "epoch": 0.85, + "learning_rate": 1.6807024465685097e-05, + "loss": 1.2713, + "step": 10113 + }, + { + "epoch": 0.85, + "learning_rate": 1.6806358560252075e-05, + "loss": 1.3121, + "step": 10114 + }, + { + "epoch": 0.85, + "learning_rate": 1.6805692598582746e-05, + "loss": 1.2849, + "step": 10115 + }, + { + "epoch": 0.85, + "learning_rate": 1.6805026580682607e-05, + "loss": 1.246, + "step": 10116 + }, + { + "epoch": 0.85, + "learning_rate": 1.680436050655716e-05, + "loss": 1.3018, + "step": 10117 + }, + { + "epoch": 0.85, + "learning_rate": 1.6803694376211906e-05, + "loss": 1.2815, + "step": 10118 + }, + { + "epoch": 0.85, + "learning_rate": 1.6803028189652356e-05, + "loss": 1.2151, + "step": 10119 + }, + { + "epoch": 0.85, + "learning_rate": 1.6802361946884007e-05, + "loss": 1.3361, + "step": 10120 + }, + { + "epoch": 0.85, + "learning_rate": 1.680169564791237e-05, + "loss": 1.322, + "step": 10121 + }, + { + "epoch": 0.85, + "learning_rate": 1.6801029292742938e-05, + "loss": 1.399, + "step": 10122 + }, + { + "epoch": 0.85, + "learning_rate": 1.6800362881381233e-05, + "loss": 1.3098, + "step": 10123 + }, + { + "epoch": 0.85, + "learning_rate": 1.6799696413832753e-05, + "loss": 1.3649, + "step": 10124 + }, + { + "epoch": 0.85, + "learning_rate": 1.6799029890103003e-05, + "loss": 1.2356, + "step": 10125 + }, + { + "epoch": 0.85, + "learning_rate": 1.6798363310197493e-05, + "loss": 1.2588, + "step": 10126 + }, + { + "epoch": 0.85, + "learning_rate": 1.679769667412173e-05, + "loss": 1.277, + "step": 10127 + }, + { + "epoch": 0.85, + "learning_rate": 1.6797029981881217e-05, + "loss": 1.2618, + "step": 10128 + }, + { + "epoch": 0.85, + "learning_rate": 1.6796363233481474e-05, + "loss": 1.2796, + "step": 10129 + }, + { + "epoch": 0.85, + "learning_rate": 1.6795696428928e-05, + "loss": 1.304, + "step": 10130 + }, + { + "epoch": 0.85, + "learning_rate": 1.679502956822631e-05, + "loss": 1.3013, + "step": 10131 + }, + { + "epoch": 0.85, + "learning_rate": 1.6794362651381905e-05, + "loss": 1.2936, + "step": 10132 + }, + { + "epoch": 0.85, + "learning_rate": 1.6793695678400307e-05, + "loss": 1.2677, + "step": 10133 + }, + { + "epoch": 0.85, + "learning_rate": 1.6793028649287014e-05, + "loss": 1.2756, + "step": 10134 + }, + { + "epoch": 0.85, + "learning_rate": 1.679236156404755e-05, + "loss": 1.3404, + "step": 10135 + }, + { + "epoch": 0.85, + "learning_rate": 1.679169442268742e-05, + "loss": 1.2397, + "step": 10136 + }, + { + "epoch": 0.85, + "learning_rate": 1.6791027225212135e-05, + "loss": 1.286, + "step": 10137 + }, + { + "epoch": 0.85, + "learning_rate": 1.6790359971627213e-05, + "loss": 1.253, + "step": 10138 + }, + { + "epoch": 0.85, + "learning_rate": 1.678969266193816e-05, + "loss": 1.1628, + "step": 10139 + }, + { + "epoch": 0.85, + "learning_rate": 1.6789025296150497e-05, + "loss": 1.2098, + "step": 10140 + }, + { + "epoch": 0.85, + "learning_rate": 1.678835787426973e-05, + "loss": 1.3035, + "step": 10141 + }, + { + "epoch": 0.85, + "learning_rate": 1.678769039630138e-05, + "loss": 1.3363, + "step": 10142 + }, + { + "epoch": 0.85, + "learning_rate": 1.6787022862250962e-05, + "loss": 1.3554, + "step": 10143 + }, + { + "epoch": 0.85, + "learning_rate": 1.6786355272123987e-05, + "loss": 1.2312, + "step": 10144 + }, + { + "epoch": 0.85, + "learning_rate": 1.678568762592597e-05, + "loss": 1.2495, + "step": 10145 + }, + { + "epoch": 0.85, + "learning_rate": 1.6785019923662434e-05, + "loss": 1.3889, + "step": 10146 + }, + { + "epoch": 0.85, + "learning_rate": 1.678435216533889e-05, + "loss": 1.2215, + "step": 10147 + }, + { + "epoch": 0.85, + "learning_rate": 1.6783684350960857e-05, + "loss": 1.2913, + "step": 10148 + }, + { + "epoch": 0.85, + "learning_rate": 1.6783016480533855e-05, + "loss": 1.2503, + "step": 10149 + }, + { + "epoch": 0.85, + "learning_rate": 1.67823485540634e-05, + "loss": 1.3402, + "step": 10150 + }, + { + "epoch": 0.85, + "learning_rate": 1.6781680571555008e-05, + "loss": 1.2461, + "step": 10151 + }, + { + "epoch": 0.85, + "learning_rate": 1.6781012533014202e-05, + "loss": 1.1883, + "step": 10152 + }, + { + "epoch": 0.85, + "learning_rate": 1.67803444384465e-05, + "loss": 1.2537, + "step": 10153 + }, + { + "epoch": 0.85, + "learning_rate": 1.6779676287857424e-05, + "loss": 1.3435, + "step": 10154 + }, + { + "epoch": 0.85, + "learning_rate": 1.6779008081252488e-05, + "loss": 1.3543, + "step": 10155 + }, + { + "epoch": 0.86, + "learning_rate": 1.6778339818637223e-05, + "loss": 1.3559, + "step": 10156 + }, + { + "epoch": 0.86, + "learning_rate": 1.677767150001714e-05, + "loss": 1.2245, + "step": 10157 + }, + { + "epoch": 0.86, + "learning_rate": 1.677700312539777e-05, + "loss": 1.2219, + "step": 10158 + }, + { + "epoch": 0.86, + "learning_rate": 1.677633469478463e-05, + "loss": 1.2378, + "step": 10159 + }, + { + "epoch": 0.86, + "learning_rate": 1.6775666208183242e-05, + "loss": 1.3132, + "step": 10160 + }, + { + "epoch": 0.86, + "learning_rate": 1.6774997665599137e-05, + "loss": 1.2041, + "step": 10161 + }, + { + "epoch": 0.86, + "learning_rate": 1.677432906703783e-05, + "loss": 1.2226, + "step": 10162 + }, + { + "epoch": 0.86, + "learning_rate": 1.6773660412504848e-05, + "loss": 1.2952, + "step": 10163 + }, + { + "epoch": 0.86, + "learning_rate": 1.6772991702005713e-05, + "loss": 1.3283, + "step": 10164 + }, + { + "epoch": 0.86, + "learning_rate": 1.6772322935545954e-05, + "loss": 1.2745, + "step": 10165 + }, + { + "epoch": 0.86, + "learning_rate": 1.67716541131311e-05, + "loss": 1.349, + "step": 10166 + }, + { + "epoch": 0.86, + "learning_rate": 1.677098523476667e-05, + "loss": 1.2239, + "step": 10167 + }, + { + "epoch": 0.86, + "learning_rate": 1.677031630045819e-05, + "loss": 1.3483, + "step": 10168 + }, + { + "epoch": 0.86, + "learning_rate": 1.6769647310211192e-05, + "loss": 1.2569, + "step": 10169 + }, + { + "epoch": 0.86, + "learning_rate": 1.6768978264031203e-05, + "loss": 1.275, + "step": 10170 + }, + { + "epoch": 0.86, + "learning_rate": 1.676830916192375e-05, + "loss": 1.1326, + "step": 10171 + }, + { + "epoch": 0.86, + "learning_rate": 1.6767640003894354e-05, + "loss": 1.1321, + "step": 10172 + }, + { + "epoch": 0.86, + "learning_rate": 1.6766970789948556e-05, + "loss": 1.193, + "step": 10173 + }, + { + "epoch": 0.86, + "learning_rate": 1.676630152009188e-05, + "loss": 1.3376, + "step": 10174 + }, + { + "epoch": 0.86, + "learning_rate": 1.676563219432985e-05, + "loss": 1.3202, + "step": 10175 + }, + { + "epoch": 0.86, + "learning_rate": 1.6764962812668006e-05, + "loss": 1.3148, + "step": 10176 + }, + { + "epoch": 0.86, + "learning_rate": 1.6764293375111874e-05, + "loss": 1.2281, + "step": 10177 + }, + { + "epoch": 0.86, + "learning_rate": 1.6763623881666986e-05, + "loss": 1.2125, + "step": 10178 + }, + { + "epoch": 0.86, + "learning_rate": 1.6762954332338872e-05, + "loss": 1.2468, + "step": 10179 + }, + { + "epoch": 0.86, + "learning_rate": 1.6762284727133065e-05, + "loss": 1.2786, + "step": 10180 + }, + { + "epoch": 0.86, + "learning_rate": 1.6761615066055095e-05, + "loss": 1.3995, + "step": 10181 + }, + { + "epoch": 0.86, + "learning_rate": 1.6760945349110498e-05, + "loss": 1.2613, + "step": 10182 + }, + { + "epoch": 0.86, + "learning_rate": 1.676027557630481e-05, + "loss": 1.3926, + "step": 10183 + }, + { + "epoch": 0.86, + "learning_rate": 1.675960574764356e-05, + "loss": 1.212, + "step": 10184 + }, + { + "epoch": 0.86, + "learning_rate": 1.675893586313228e-05, + "loss": 1.2263, + "step": 10185 + }, + { + "epoch": 0.86, + "learning_rate": 1.6758265922776513e-05, + "loss": 1.2777, + "step": 10186 + }, + { + "epoch": 0.86, + "learning_rate": 1.675759592658179e-05, + "loss": 1.2048, + "step": 10187 + }, + { + "epoch": 0.86, + "learning_rate": 1.6756925874553648e-05, + "loss": 1.3148, + "step": 10188 + }, + { + "epoch": 0.86, + "learning_rate": 1.675625576669762e-05, + "loss": 1.2458, + "step": 10189 + }, + { + "epoch": 0.86, + "learning_rate": 1.6755585603019243e-05, + "loss": 1.2312, + "step": 10190 + }, + { + "epoch": 0.86, + "learning_rate": 1.6754915383524056e-05, + "loss": 1.1965, + "step": 10191 + }, + { + "epoch": 0.86, + "learning_rate": 1.67542451082176e-05, + "loss": 1.2847, + "step": 10192 + }, + { + "epoch": 0.86, + "learning_rate": 1.6753574777105404e-05, + "loss": 1.2029, + "step": 10193 + }, + { + "epoch": 0.86, + "learning_rate": 1.6752904390193016e-05, + "loss": 1.3385, + "step": 10194 + }, + { + "epoch": 0.86, + "learning_rate": 1.6752233947485968e-05, + "loss": 1.1836, + "step": 10195 + }, + { + "epoch": 0.86, + "learning_rate": 1.6751563448989807e-05, + "loss": 1.2884, + "step": 10196 + }, + { + "epoch": 0.86, + "learning_rate": 1.675089289471006e-05, + "loss": 1.2478, + "step": 10197 + }, + { + "epoch": 0.86, + "learning_rate": 1.675022228465228e-05, + "loss": 1.3674, + "step": 10198 + }, + { + "epoch": 0.86, + "learning_rate": 1.6749551618822006e-05, + "loss": 1.3292, + "step": 10199 + }, + { + "epoch": 0.86, + "learning_rate": 1.6748880897224772e-05, + "loss": 1.2875, + "step": 10200 + }, + { + "epoch": 0.86, + "learning_rate": 1.6748210119866123e-05, + "loss": 1.2886, + "step": 10201 + }, + { + "epoch": 0.86, + "learning_rate": 1.674753928675161e-05, + "loss": 1.2963, + "step": 10202 + }, + { + "epoch": 0.86, + "learning_rate": 1.674686839788676e-05, + "loss": 1.3394, + "step": 10203 + }, + { + "epoch": 0.86, + "learning_rate": 1.6746197453277126e-05, + "loss": 1.2586, + "step": 10204 + }, + { + "epoch": 0.86, + "learning_rate": 1.674552645292825e-05, + "loss": 1.2445, + "step": 10205 + }, + { + "epoch": 0.86, + "learning_rate": 1.6744855396845676e-05, + "loss": 1.2285, + "step": 10206 + }, + { + "epoch": 0.86, + "learning_rate": 1.6744184285034946e-05, + "loss": 1.2983, + "step": 10207 + }, + { + "epoch": 0.86, + "learning_rate": 1.6743513117501612e-05, + "loss": 1.3197, + "step": 10208 + }, + { + "epoch": 0.86, + "learning_rate": 1.6742841894251213e-05, + "loss": 1.2636, + "step": 10209 + }, + { + "epoch": 0.86, + "learning_rate": 1.674217061528929e-05, + "loss": 1.26, + "step": 10210 + }, + { + "epoch": 0.86, + "learning_rate": 1.67414992806214e-05, + "loss": 1.2816, + "step": 10211 + }, + { + "epoch": 0.86, + "learning_rate": 1.6740827890253086e-05, + "loss": 1.2207, + "step": 10212 + }, + { + "epoch": 0.86, + "learning_rate": 1.6740156444189897e-05, + "loss": 1.2796, + "step": 10213 + }, + { + "epoch": 0.86, + "learning_rate": 1.6739484942437372e-05, + "loss": 1.2808, + "step": 10214 + }, + { + "epoch": 0.86, + "learning_rate": 1.673881338500107e-05, + "loss": 1.3299, + "step": 10215 + }, + { + "epoch": 0.86, + "learning_rate": 1.6738141771886536e-05, + "loss": 1.2032, + "step": 10216 + }, + { + "epoch": 0.86, + "learning_rate": 1.6737470103099317e-05, + "loss": 1.2576, + "step": 10217 + }, + { + "epoch": 0.86, + "learning_rate": 1.6736798378644964e-05, + "loss": 1.2083, + "step": 10218 + }, + { + "epoch": 0.86, + "learning_rate": 1.6736126598529027e-05, + "loss": 1.35, + "step": 10219 + }, + { + "epoch": 0.86, + "learning_rate": 1.6735454762757052e-05, + "loss": 1.3318, + "step": 10220 + }, + { + "epoch": 0.86, + "learning_rate": 1.67347828713346e-05, + "loss": 1.226, + "step": 10221 + }, + { + "epoch": 0.86, + "learning_rate": 1.673411092426721e-05, + "loss": 1.2293, + "step": 10222 + }, + { + "epoch": 0.86, + "learning_rate": 1.6733438921560444e-05, + "loss": 1.2323, + "step": 10223 + }, + { + "epoch": 0.86, + "learning_rate": 1.6732766863219852e-05, + "loss": 1.2366, + "step": 10224 + }, + { + "epoch": 0.86, + "learning_rate": 1.6732094749250985e-05, + "loss": 1.2435, + "step": 10225 + }, + { + "epoch": 0.86, + "learning_rate": 1.6731422579659392e-05, + "loss": 1.2212, + "step": 10226 + }, + { + "epoch": 0.86, + "learning_rate": 1.673075035445064e-05, + "loss": 1.3071, + "step": 10227 + }, + { + "epoch": 0.86, + "learning_rate": 1.6730078073630268e-05, + "loss": 1.2501, + "step": 10228 + }, + { + "epoch": 0.86, + "learning_rate": 1.672940573720384e-05, + "loss": 1.2939, + "step": 10229 + }, + { + "epoch": 0.86, + "learning_rate": 1.6728733345176905e-05, + "loss": 1.2636, + "step": 10230 + }, + { + "epoch": 0.86, + "learning_rate": 1.6728060897555024e-05, + "loss": 1.2092, + "step": 10231 + }, + { + "epoch": 0.86, + "learning_rate": 1.672738839434375e-05, + "loss": 1.2343, + "step": 10232 + }, + { + "epoch": 0.86, + "learning_rate": 1.672671583554864e-05, + "loss": 1.2715, + "step": 10233 + }, + { + "epoch": 0.86, + "learning_rate": 1.672604322117525e-05, + "loss": 1.3379, + "step": 10234 + }, + { + "epoch": 0.86, + "learning_rate": 1.672537055122914e-05, + "loss": 1.2955, + "step": 10235 + }, + { + "epoch": 0.86, + "learning_rate": 1.6724697825715863e-05, + "loss": 1.184, + "step": 10236 + }, + { + "epoch": 0.86, + "learning_rate": 1.6724025044640984e-05, + "loss": 1.3692, + "step": 10237 + }, + { + "epoch": 0.86, + "learning_rate": 1.6723352208010054e-05, + "loss": 1.3386, + "step": 10238 + }, + { + "epoch": 0.86, + "learning_rate": 1.672267931582864e-05, + "loss": 1.1454, + "step": 10239 + }, + { + "epoch": 0.86, + "learning_rate": 1.67220063681023e-05, + "loss": 1.2447, + "step": 10240 + }, + { + "epoch": 0.86, + "learning_rate": 1.6721333364836588e-05, + "loss": 1.2487, + "step": 10241 + }, + { + "epoch": 0.86, + "learning_rate": 1.6720660306037067e-05, + "loss": 1.3371, + "step": 10242 + }, + { + "epoch": 0.86, + "learning_rate": 1.6719987191709306e-05, + "loss": 1.326, + "step": 10243 + }, + { + "epoch": 0.86, + "learning_rate": 1.6719314021858857e-05, + "loss": 1.2755, + "step": 10244 + }, + { + "epoch": 0.86, + "learning_rate": 1.6718640796491286e-05, + "loss": 1.2614, + "step": 10245 + }, + { + "epoch": 0.86, + "learning_rate": 1.671796751561215e-05, + "loss": 1.2575, + "step": 10246 + }, + { + "epoch": 0.86, + "learning_rate": 1.671729417922702e-05, + "loss": 1.2798, + "step": 10247 + }, + { + "epoch": 0.86, + "learning_rate": 1.6716620787341455e-05, + "loss": 1.2842, + "step": 10248 + }, + { + "epoch": 0.86, + "learning_rate": 1.671594733996102e-05, + "loss": 1.2596, + "step": 10249 + }, + { + "epoch": 0.86, + "learning_rate": 1.671527383709128e-05, + "loss": 1.2325, + "step": 10250 + }, + { + "epoch": 0.86, + "learning_rate": 1.6714600278737795e-05, + "loss": 1.2916, + "step": 10251 + }, + { + "epoch": 0.86, + "learning_rate": 1.6713926664906136e-05, + "loss": 1.1436, + "step": 10252 + }, + { + "epoch": 0.86, + "learning_rate": 1.6713252995601866e-05, + "loss": 1.3006, + "step": 10253 + }, + { + "epoch": 0.86, + "learning_rate": 1.671257927083055e-05, + "loss": 1.3446, + "step": 10254 + }, + { + "epoch": 0.86, + "learning_rate": 1.6711905490597758e-05, + "loss": 1.2357, + "step": 10255 + }, + { + "epoch": 0.86, + "learning_rate": 1.671123165490905e-05, + "loss": 1.1662, + "step": 10256 + }, + { + "epoch": 0.86, + "learning_rate": 1.6710557763770006e-05, + "loss": 1.2488, + "step": 10257 + }, + { + "epoch": 0.86, + "learning_rate": 1.670988381718618e-05, + "loss": 1.2644, + "step": 10258 + }, + { + "epoch": 0.86, + "learning_rate": 1.6709209815163147e-05, + "loss": 1.2125, + "step": 10259 + }, + { + "epoch": 0.86, + "learning_rate": 1.6708535757706476e-05, + "loss": 1.1937, + "step": 10260 + }, + { + "epoch": 0.86, + "learning_rate": 1.6707861644821737e-05, + "loss": 1.2795, + "step": 10261 + }, + { + "epoch": 0.86, + "learning_rate": 1.6707187476514498e-05, + "loss": 1.3147, + "step": 10262 + }, + { + "epoch": 0.86, + "learning_rate": 1.6706513252790328e-05, + "loss": 1.267, + "step": 10263 + }, + { + "epoch": 0.86, + "learning_rate": 1.67058389736548e-05, + "loss": 1.1876, + "step": 10264 + }, + { + "epoch": 0.86, + "learning_rate": 1.6705164639113483e-05, + "loss": 1.318, + "step": 10265 + }, + { + "epoch": 0.86, + "learning_rate": 1.670449024917195e-05, + "loss": 1.2124, + "step": 10266 + }, + { + "epoch": 0.86, + "learning_rate": 1.6703815803835777e-05, + "loss": 1.242, + "step": 10267 + }, + { + "epoch": 0.86, + "learning_rate": 1.6703141303110525e-05, + "loss": 1.2697, + "step": 10268 + }, + { + "epoch": 0.86, + "learning_rate": 1.6702466747001783e-05, + "loss": 1.3167, + "step": 10269 + }, + { + "epoch": 0.86, + "learning_rate": 1.670179213551511e-05, + "loss": 1.361, + "step": 10270 + }, + { + "epoch": 0.86, + "learning_rate": 1.6701117468656084e-05, + "loss": 1.2002, + "step": 10271 + }, + { + "epoch": 0.86, + "learning_rate": 1.6700442746430283e-05, + "loss": 1.2372, + "step": 10272 + }, + { + "epoch": 0.86, + "learning_rate": 1.669976796884328e-05, + "loss": 1.277, + "step": 10273 + }, + { + "epoch": 0.86, + "learning_rate": 1.669909313590065e-05, + "loss": 1.3229, + "step": 10274 + }, + { + "epoch": 0.87, + "learning_rate": 1.669841824760797e-05, + "loss": 1.3579, + "step": 10275 + }, + { + "epoch": 0.87, + "learning_rate": 1.6697743303970813e-05, + "loss": 1.3245, + "step": 10276 + }, + { + "epoch": 0.87, + "learning_rate": 1.6697068304994757e-05, + "loss": 1.2918, + "step": 10277 + }, + { + "epoch": 0.87, + "learning_rate": 1.669639325068538e-05, + "loss": 1.2313, + "step": 10278 + }, + { + "epoch": 0.87, + "learning_rate": 1.6695718141048258e-05, + "loss": 1.356, + "step": 10279 + }, + { + "epoch": 0.87, + "learning_rate": 1.669504297608897e-05, + "loss": 1.2811, + "step": 10280 + }, + { + "epoch": 0.87, + "learning_rate": 1.6694367755813096e-05, + "loss": 1.2125, + "step": 10281 + }, + { + "epoch": 0.87, + "learning_rate": 1.669369248022621e-05, + "loss": 1.3229, + "step": 10282 + }, + { + "epoch": 0.87, + "learning_rate": 1.6693017149333898e-05, + "loss": 1.3287, + "step": 10283 + }, + { + "epoch": 0.87, + "learning_rate": 1.6692341763141735e-05, + "loss": 1.3232, + "step": 10284 + }, + { + "epoch": 0.87, + "learning_rate": 1.66916663216553e-05, + "loss": 1.3274, + "step": 10285 + }, + { + "epoch": 0.87, + "learning_rate": 1.669099082488018e-05, + "loss": 1.2016, + "step": 10286 + }, + { + "epoch": 0.87, + "learning_rate": 1.669031527282195e-05, + "loss": 1.2045, + "step": 10287 + }, + { + "epoch": 0.87, + "learning_rate": 1.6689639665486194e-05, + "loss": 1.3287, + "step": 10288 + }, + { + "epoch": 0.87, + "learning_rate": 1.6688964002878493e-05, + "loss": 1.1633, + "step": 10289 + }, + { + "epoch": 0.87, + "learning_rate": 1.6688288285004433e-05, + "loss": 1.3037, + "step": 10290 + }, + { + "epoch": 0.87, + "learning_rate": 1.6687612511869595e-05, + "loss": 1.2545, + "step": 10291 + }, + { + "epoch": 0.87, + "learning_rate": 1.6686936683479562e-05, + "loss": 1.2297, + "step": 10292 + }, + { + "epoch": 0.87, + "learning_rate": 1.6686260799839917e-05, + "loss": 1.3569, + "step": 10293 + }, + { + "epoch": 0.87, + "learning_rate": 1.6685584860956246e-05, + "loss": 1.247, + "step": 10294 + }, + { + "epoch": 0.87, + "learning_rate": 1.6684908866834133e-05, + "loss": 1.2479, + "step": 10295 + }, + { + "epoch": 0.87, + "learning_rate": 1.6684232817479164e-05, + "loss": 1.254, + "step": 10296 + }, + { + "epoch": 0.87, + "learning_rate": 1.668355671289692e-05, + "loss": 1.2932, + "step": 10297 + }, + { + "epoch": 0.87, + "learning_rate": 1.6682880553092993e-05, + "loss": 1.248, + "step": 10298 + }, + { + "epoch": 0.87, + "learning_rate": 1.6682204338072972e-05, + "loss": 1.2203, + "step": 10299 + }, + { + "epoch": 0.87, + "learning_rate": 1.668152806784244e-05, + "loss": 1.2204, + "step": 10300 + }, + { + "epoch": 0.87, + "learning_rate": 1.668085174240698e-05, + "loss": 1.3329, + "step": 10301 + }, + { + "epoch": 0.87, + "learning_rate": 1.668017536177219e-05, + "loss": 1.2167, + "step": 10302 + }, + { + "epoch": 0.87, + "learning_rate": 1.6679498925943654e-05, + "loss": 1.2919, + "step": 10303 + }, + { + "epoch": 0.87, + "learning_rate": 1.6678822434926957e-05, + "loss": 1.251, + "step": 10304 + }, + { + "epoch": 0.87, + "learning_rate": 1.6678145888727692e-05, + "loss": 1.3256, + "step": 10305 + }, + { + "epoch": 0.87, + "learning_rate": 1.6677469287351448e-05, + "loss": 1.1817, + "step": 10306 + }, + { + "epoch": 0.87, + "learning_rate": 1.6676792630803814e-05, + "loss": 1.3185, + "step": 10307 + }, + { + "epoch": 0.87, + "learning_rate": 1.6676115919090384e-05, + "loss": 1.2358, + "step": 10308 + }, + { + "epoch": 0.87, + "learning_rate": 1.6675439152216747e-05, + "loss": 1.1601, + "step": 10309 + }, + { + "epoch": 0.87, + "learning_rate": 1.6674762330188497e-05, + "loss": 1.3061, + "step": 10310 + }, + { + "epoch": 0.87, + "learning_rate": 1.6674085453011224e-05, + "loss": 1.2223, + "step": 10311 + }, + { + "epoch": 0.87, + "learning_rate": 1.667340852069052e-05, + "loss": 1.3034, + "step": 10312 + }, + { + "epoch": 0.87, + "learning_rate": 1.667273153323198e-05, + "loss": 1.2385, + "step": 10313 + }, + { + "epoch": 0.87, + "learning_rate": 1.6672054490641195e-05, + "loss": 1.4385, + "step": 10314 + }, + { + "epoch": 0.87, + "learning_rate": 1.6671377392923763e-05, + "loss": 1.3606, + "step": 10315 + }, + { + "epoch": 0.87, + "learning_rate": 1.6670700240085275e-05, + "loss": 1.3446, + "step": 10316 + }, + { + "epoch": 0.87, + "learning_rate": 1.6670023032131325e-05, + "loss": 1.3579, + "step": 10317 + }, + { + "epoch": 0.87, + "learning_rate": 1.6669345769067515e-05, + "loss": 1.3078, + "step": 10318 + }, + { + "epoch": 0.87, + "learning_rate": 1.6668668450899427e-05, + "loss": 1.2692, + "step": 10319 + }, + { + "epoch": 0.87, + "learning_rate": 1.6667991077632673e-05, + "loss": 1.2771, + "step": 10320 + }, + { + "epoch": 0.87, + "learning_rate": 1.666731364927284e-05, + "loss": 1.2334, + "step": 10321 + }, + { + "epoch": 0.87, + "learning_rate": 1.666663616582553e-05, + "loss": 1.2935, + "step": 10322 + }, + { + "epoch": 0.87, + "learning_rate": 1.6665958627296335e-05, + "loss": 1.3549, + "step": 10323 + }, + { + "epoch": 0.87, + "learning_rate": 1.666528103369086e-05, + "loss": 1.2811, + "step": 10324 + }, + { + "epoch": 0.87, + "learning_rate": 1.6664603385014697e-05, + "loss": 1.3299, + "step": 10325 + }, + { + "epoch": 0.87, + "learning_rate": 1.666392568127345e-05, + "loss": 1.3019, + "step": 10326 + }, + { + "epoch": 0.87, + "learning_rate": 1.6663247922472715e-05, + "loss": 1.2482, + "step": 10327 + }, + { + "epoch": 0.87, + "learning_rate": 1.6662570108618094e-05, + "loss": 1.2008, + "step": 10328 + }, + { + "epoch": 0.87, + "learning_rate": 1.6661892239715185e-05, + "loss": 1.2436, + "step": 10329 + }, + { + "epoch": 0.87, + "learning_rate": 1.666121431576959e-05, + "loss": 1.3112, + "step": 10330 + }, + { + "epoch": 0.87, + "learning_rate": 1.6660536336786916e-05, + "loss": 1.282, + "step": 10331 + }, + { + "epoch": 0.87, + "learning_rate": 1.6659858302772755e-05, + "loss": 1.2927, + "step": 10332 + }, + { + "epoch": 0.87, + "learning_rate": 1.6659180213732715e-05, + "loss": 1.123, + "step": 10333 + }, + { + "epoch": 0.87, + "learning_rate": 1.6658502069672392e-05, + "loss": 1.2315, + "step": 10334 + }, + { + "epoch": 0.87, + "learning_rate": 1.6657823870597395e-05, + "loss": 1.3921, + "step": 10335 + }, + { + "epoch": 0.87, + "learning_rate": 1.665714561651333e-05, + "loss": 1.1795, + "step": 10336 + }, + { + "epoch": 0.87, + "learning_rate": 1.6656467307425796e-05, + "loss": 1.2128, + "step": 10337 + }, + { + "epoch": 0.87, + "learning_rate": 1.6655788943340398e-05, + "loss": 1.2049, + "step": 10338 + }, + { + "epoch": 0.87, + "learning_rate": 1.665511052426274e-05, + "loss": 1.3151, + "step": 10339 + }, + { + "epoch": 0.87, + "learning_rate": 1.6654432050198434e-05, + "loss": 1.2178, + "step": 10340 + }, + { + "epoch": 0.87, + "learning_rate": 1.6653753521153078e-05, + "loss": 1.2024, + "step": 10341 + }, + { + "epoch": 0.87, + "learning_rate": 1.665307493713228e-05, + "loss": 1.2276, + "step": 10342 + }, + { + "epoch": 0.87, + "learning_rate": 1.6652396298141648e-05, + "loss": 1.3217, + "step": 10343 + }, + { + "epoch": 0.87, + "learning_rate": 1.665171760418679e-05, + "loss": 1.2296, + "step": 10344 + }, + { + "epoch": 0.87, + "learning_rate": 1.665103885527331e-05, + "loss": 1.2456, + "step": 10345 + }, + { + "epoch": 0.87, + "learning_rate": 1.6650360051406822e-05, + "loss": 1.2985, + "step": 10346 + }, + { + "epoch": 0.87, + "learning_rate": 1.664968119259293e-05, + "loss": 1.2002, + "step": 10347 + }, + { + "epoch": 0.87, + "learning_rate": 1.6649002278837237e-05, + "loss": 1.4514, + "step": 10348 + }, + { + "epoch": 0.87, + "learning_rate": 1.664832331014537e-05, + "loss": 1.3201, + "step": 10349 + }, + { + "epoch": 0.87, + "learning_rate": 1.6647644286522918e-05, + "loss": 1.1875, + "step": 10350 + }, + { + "epoch": 0.87, + "learning_rate": 1.664696520797551e-05, + "loss": 1.4449, + "step": 10351 + }, + { + "epoch": 0.87, + "learning_rate": 1.664628607450874e-05, + "loss": 1.2296, + "step": 10352 + }, + { + "epoch": 0.87, + "learning_rate": 1.6645606886128236e-05, + "loss": 1.2558, + "step": 10353 + }, + { + "epoch": 0.87, + "learning_rate": 1.6644927642839593e-05, + "loss": 1.2836, + "step": 10354 + }, + { + "epoch": 0.87, + "learning_rate": 1.664424834464844e-05, + "loss": 1.3102, + "step": 10355 + }, + { + "epoch": 0.87, + "learning_rate": 1.6643568991560373e-05, + "loss": 1.3786, + "step": 10356 + }, + { + "epoch": 0.87, + "learning_rate": 1.6642889583581014e-05, + "loss": 1.3102, + "step": 10357 + }, + { + "epoch": 0.87, + "learning_rate": 1.664221012071598e-05, + "loss": 1.3082, + "step": 10358 + }, + { + "epoch": 0.87, + "learning_rate": 1.6641530602970878e-05, + "loss": 1.4031, + "step": 10359 + }, + { + "epoch": 0.87, + "learning_rate": 1.6640851030351323e-05, + "loss": 1.2553, + "step": 10360 + }, + { + "epoch": 0.87, + "learning_rate": 1.6640171402862928e-05, + "loss": 1.3539, + "step": 10361 + }, + { + "epoch": 0.87, + "learning_rate": 1.663949172051132e-05, + "loss": 1.2614, + "step": 10362 + }, + { + "epoch": 0.87, + "learning_rate": 1.6638811983302097e-05, + "loss": 1.2353, + "step": 10363 + }, + { + "epoch": 0.87, + "learning_rate": 1.6638132191240893e-05, + "loss": 1.3854, + "step": 10364 + }, + { + "epoch": 0.87, + "learning_rate": 1.6637452344333313e-05, + "loss": 1.2349, + "step": 10365 + }, + { + "epoch": 0.87, + "learning_rate": 1.6636772442584977e-05, + "loss": 1.3005, + "step": 10366 + }, + { + "epoch": 0.87, + "learning_rate": 1.6636092486001504e-05, + "loss": 1.279, + "step": 10367 + }, + { + "epoch": 0.87, + "learning_rate": 1.663541247458851e-05, + "loss": 1.2692, + "step": 10368 + }, + { + "epoch": 0.87, + "learning_rate": 1.663473240835162e-05, + "loss": 1.3752, + "step": 10369 + }, + { + "epoch": 0.87, + "learning_rate": 1.6634052287296438e-05, + "loss": 1.3265, + "step": 10370 + }, + { + "epoch": 0.87, + "learning_rate": 1.66333721114286e-05, + "loss": 1.3025, + "step": 10371 + }, + { + "epoch": 0.87, + "learning_rate": 1.6632691880753715e-05, + "loss": 1.2904, + "step": 10372 + }, + { + "epoch": 0.87, + "learning_rate": 1.6632011595277408e-05, + "loss": 1.3403, + "step": 10373 + }, + { + "epoch": 0.87, + "learning_rate": 1.66313312550053e-05, + "loss": 1.3582, + "step": 10374 + }, + { + "epoch": 0.87, + "learning_rate": 1.6630650859943006e-05, + "loss": 1.3183, + "step": 10375 + }, + { + "epoch": 0.87, + "learning_rate": 1.6629970410096158e-05, + "loss": 1.2989, + "step": 10376 + }, + { + "epoch": 0.87, + "learning_rate": 1.662928990547037e-05, + "loss": 1.2803, + "step": 10377 + }, + { + "epoch": 0.87, + "learning_rate": 1.662860934607127e-05, + "loss": 1.276, + "step": 10378 + }, + { + "epoch": 0.87, + "learning_rate": 1.6627928731904477e-05, + "loss": 1.281, + "step": 10379 + }, + { + "epoch": 0.87, + "learning_rate": 1.6627248062975612e-05, + "loss": 1.2721, + "step": 10380 + }, + { + "epoch": 0.87, + "learning_rate": 1.6626567339290307e-05, + "loss": 1.2998, + "step": 10381 + }, + { + "epoch": 0.87, + "learning_rate": 1.662588656085418e-05, + "loss": 1.2722, + "step": 10382 + }, + { + "epoch": 0.87, + "learning_rate": 1.662520572767286e-05, + "loss": 1.292, + "step": 10383 + }, + { + "epoch": 0.87, + "learning_rate": 1.6624524839751966e-05, + "loss": 1.3426, + "step": 10384 + }, + { + "epoch": 0.87, + "learning_rate": 1.662384389709713e-05, + "loss": 1.3735, + "step": 10385 + }, + { + "epoch": 0.87, + "learning_rate": 1.6623162899713975e-05, + "loss": 1.192, + "step": 10386 + }, + { + "epoch": 0.87, + "learning_rate": 1.662248184760813e-05, + "loss": 1.3248, + "step": 10387 + }, + { + "epoch": 0.87, + "learning_rate": 1.6621800740785218e-05, + "loss": 1.3495, + "step": 10388 + }, + { + "epoch": 0.87, + "learning_rate": 1.662111957925087e-05, + "loss": 1.286, + "step": 10389 + }, + { + "epoch": 0.87, + "learning_rate": 1.6620438363010713e-05, + "loss": 1.3281, + "step": 10390 + }, + { + "epoch": 0.87, + "learning_rate": 1.6619757092070376e-05, + "loss": 1.2556, + "step": 10391 + }, + { + "epoch": 0.87, + "learning_rate": 1.661907576643549e-05, + "loss": 1.2658, + "step": 10392 + }, + { + "epoch": 0.88, + "learning_rate": 1.6618394386111675e-05, + "loss": 1.3985, + "step": 10393 + }, + { + "epoch": 0.88, + "learning_rate": 1.6617712951104573e-05, + "loss": 1.2543, + "step": 10394 + }, + { + "epoch": 0.88, + "learning_rate": 1.6617031461419807e-05, + "loss": 1.1801, + "step": 10395 + }, + { + "epoch": 0.88, + "learning_rate": 1.6616349917063012e-05, + "loss": 1.2642, + "step": 10396 + }, + { + "epoch": 0.88, + "learning_rate": 1.661566831803981e-05, + "loss": 1.2974, + "step": 10397 + }, + { + "epoch": 0.88, + "learning_rate": 1.661498666435584e-05, + "loss": 1.2573, + "step": 10398 + }, + { + "epoch": 0.88, + "learning_rate": 1.661430495601674e-05, + "loss": 1.2838, + "step": 10399 + }, + { + "epoch": 0.88, + "learning_rate": 1.661362319302813e-05, + "loss": 1.2685, + "step": 10400 + }, + { + "epoch": 0.88, + "learning_rate": 1.661294137539565e-05, + "loss": 1.2933, + "step": 10401 + }, + { + "epoch": 0.88, + "learning_rate": 1.661225950312493e-05, + "loss": 1.3067, + "step": 10402 + }, + { + "epoch": 0.88, + "learning_rate": 1.661157757622161e-05, + "loss": 1.2648, + "step": 10403 + }, + { + "epoch": 0.88, + "learning_rate": 1.6610895594691316e-05, + "loss": 1.3533, + "step": 10404 + }, + { + "epoch": 0.88, + "learning_rate": 1.661021355853969e-05, + "loss": 1.3468, + "step": 10405 + }, + { + "epoch": 0.88, + "learning_rate": 1.6609531467772362e-05, + "loss": 1.3166, + "step": 10406 + }, + { + "epoch": 0.88, + "learning_rate": 1.660884932239497e-05, + "loss": 1.2109, + "step": 10407 + }, + { + "epoch": 0.88, + "learning_rate": 1.6608167122413152e-05, + "loss": 1.2131, + "step": 10408 + }, + { + "epoch": 0.88, + "learning_rate": 1.660748486783254e-05, + "loss": 1.2251, + "step": 10409 + }, + { + "epoch": 0.88, + "learning_rate": 1.660680255865877e-05, + "loss": 1.2414, + "step": 10410 + }, + { + "epoch": 0.88, + "learning_rate": 1.6606120194897492e-05, + "loss": 1.2085, + "step": 10411 + }, + { + "epoch": 0.88, + "learning_rate": 1.6605437776554327e-05, + "loss": 1.2235, + "step": 10412 + }, + { + "epoch": 0.88, + "learning_rate": 1.6604755303634922e-05, + "loss": 1.2589, + "step": 10413 + }, + { + "epoch": 0.88, + "learning_rate": 1.660407277614492e-05, + "loss": 1.3408, + "step": 10414 + }, + { + "epoch": 0.88, + "learning_rate": 1.660339019408995e-05, + "loss": 1.3546, + "step": 10415 + }, + { + "epoch": 0.88, + "learning_rate": 1.6602707557475658e-05, + "loss": 1.2092, + "step": 10416 + }, + { + "epoch": 0.88, + "learning_rate": 1.6602024866307684e-05, + "loss": 1.3162, + "step": 10417 + }, + { + "epoch": 0.88, + "learning_rate": 1.6601342120591665e-05, + "loss": 1.3219, + "step": 10418 + }, + { + "epoch": 0.88, + "learning_rate": 1.6600659320333247e-05, + "loss": 1.2552, + "step": 10419 + }, + { + "epoch": 0.88, + "learning_rate": 1.6599976465538067e-05, + "loss": 1.3041, + "step": 10420 + }, + { + "epoch": 0.88, + "learning_rate": 1.6599293556211774e-05, + "loss": 1.3345, + "step": 10421 + }, + { + "epoch": 0.88, + "learning_rate": 1.659861059236e-05, + "loss": 1.2655, + "step": 10422 + }, + { + "epoch": 0.88, + "learning_rate": 1.6597927573988397e-05, + "loss": 1.249, + "step": 10423 + }, + { + "epoch": 0.88, + "learning_rate": 1.6597244501102606e-05, + "loss": 1.2827, + "step": 10424 + }, + { + "epoch": 0.88, + "learning_rate": 1.6596561373708266e-05, + "loss": 1.3563, + "step": 10425 + }, + { + "epoch": 0.88, + "learning_rate": 1.659587819181103e-05, + "loss": 1.2731, + "step": 10426 + }, + { + "epoch": 0.88, + "learning_rate": 1.6595194955416532e-05, + "loss": 1.2393, + "step": 10427 + }, + { + "epoch": 0.88, + "learning_rate": 1.6594511664530424e-05, + "loss": 1.2787, + "step": 10428 + }, + { + "epoch": 0.88, + "learning_rate": 1.659382831915835e-05, + "loss": 1.3332, + "step": 10429 + }, + { + "epoch": 0.88, + "learning_rate": 1.6593144919305958e-05, + "loss": 1.3241, + "step": 10430 + }, + { + "epoch": 0.88, + "learning_rate": 1.659246146497889e-05, + "loss": 1.342, + "step": 10431 + }, + { + "epoch": 0.88, + "learning_rate": 1.6591777956182795e-05, + "loss": 1.3136, + "step": 10432 + }, + { + "epoch": 0.88, + "learning_rate": 1.6591094392923322e-05, + "loss": 1.2224, + "step": 10433 + }, + { + "epoch": 0.88, + "learning_rate": 1.659041077520612e-05, + "loss": 1.2398, + "step": 10434 + }, + { + "epoch": 0.88, + "learning_rate": 1.658972710303683e-05, + "loss": 1.3358, + "step": 10435 + }, + { + "epoch": 0.88, + "learning_rate": 1.658904337642111e-05, + "loss": 1.3443, + "step": 10436 + }, + { + "epoch": 0.88, + "learning_rate": 1.6588359595364605e-05, + "loss": 1.2494, + "step": 10437 + }, + { + "epoch": 0.88, + "learning_rate": 1.6587675759872964e-05, + "loss": 1.2827, + "step": 10438 + }, + { + "epoch": 0.88, + "learning_rate": 1.6586991869951834e-05, + "loss": 1.3396, + "step": 10439 + }, + { + "epoch": 0.88, + "learning_rate": 1.6586307925606874e-05, + "loss": 1.3466, + "step": 10440 + }, + { + "epoch": 0.88, + "learning_rate": 1.6585623926843724e-05, + "loss": 1.2596, + "step": 10441 + }, + { + "epoch": 0.88, + "learning_rate": 1.658493987366805e-05, + "loss": 1.1896, + "step": 10442 + }, + { + "epoch": 0.88, + "learning_rate": 1.6584255766085488e-05, + "loss": 1.2759, + "step": 10443 + }, + { + "epoch": 0.88, + "learning_rate": 1.65835716041017e-05, + "loss": 1.2607, + "step": 10444 + }, + { + "epoch": 0.88, + "learning_rate": 1.6582887387722336e-05, + "loss": 1.2403, + "step": 10445 + }, + { + "epoch": 0.88, + "learning_rate": 1.6582203116953054e-05, + "loss": 1.3242, + "step": 10446 + }, + { + "epoch": 0.88, + "learning_rate": 1.6581518791799497e-05, + "loss": 1.2706, + "step": 10447 + }, + { + "epoch": 0.88, + "learning_rate": 1.6580834412267326e-05, + "loss": 1.2824, + "step": 10448 + }, + { + "epoch": 0.88, + "learning_rate": 1.6580149978362195e-05, + "loss": 1.1913, + "step": 10449 + }, + { + "epoch": 0.88, + "learning_rate": 1.6579465490089764e-05, + "loss": 1.3414, + "step": 10450 + }, + { + "epoch": 0.88, + "learning_rate": 1.6578780947455683e-05, + "loss": 1.434, + "step": 10451 + }, + { + "epoch": 0.88, + "learning_rate": 1.6578096350465603e-05, + "loss": 1.3254, + "step": 10452 + }, + { + "epoch": 0.88, + "learning_rate": 1.657741169912519e-05, + "loss": 1.2919, + "step": 10453 + }, + { + "epoch": 0.88, + "learning_rate": 1.6576726993440093e-05, + "loss": 1.2651, + "step": 10454 + }, + { + "epoch": 0.88, + "learning_rate": 1.6576042233415975e-05, + "loss": 1.2857, + "step": 10455 + }, + { + "epoch": 0.88, + "learning_rate": 1.657535741905849e-05, + "loss": 1.2137, + "step": 10456 + }, + { + "epoch": 0.88, + "learning_rate": 1.65746725503733e-05, + "loss": 1.3394, + "step": 10457 + }, + { + "epoch": 0.88, + "learning_rate": 1.6573987627366057e-05, + "loss": 1.2885, + "step": 10458 + }, + { + "epoch": 0.88, + "learning_rate": 1.657330265004243e-05, + "loss": 1.3416, + "step": 10459 + }, + { + "epoch": 0.88, + "learning_rate": 1.6572617618408066e-05, + "loss": 1.2569, + "step": 10460 + }, + { + "epoch": 0.88, + "learning_rate": 1.6571932532468638e-05, + "loss": 1.2203, + "step": 10461 + }, + { + "epoch": 0.88, + "learning_rate": 1.6571247392229797e-05, + "loss": 1.313, + "step": 10462 + }, + { + "epoch": 0.88, + "learning_rate": 1.6570562197697208e-05, + "loss": 1.225, + "step": 10463 + }, + { + "epoch": 0.88, + "learning_rate": 1.656987694887653e-05, + "loss": 1.2522, + "step": 10464 + }, + { + "epoch": 0.88, + "learning_rate": 1.656919164577343e-05, + "loss": 1.2769, + "step": 10465 + }, + { + "epoch": 0.88, + "learning_rate": 1.6568506288393562e-05, + "loss": 1.2579, + "step": 10466 + }, + { + "epoch": 0.88, + "learning_rate": 1.6567820876742593e-05, + "loss": 1.3266, + "step": 10467 + }, + { + "epoch": 0.88, + "learning_rate": 1.6567135410826187e-05, + "loss": 1.2992, + "step": 10468 + }, + { + "epoch": 0.88, + "learning_rate": 1.6566449890650006e-05, + "loss": 1.2215, + "step": 10469 + }, + { + "epoch": 0.88, + "learning_rate": 1.6565764316219715e-05, + "loss": 1.286, + "step": 10470 + }, + { + "epoch": 0.88, + "learning_rate": 1.656507868754098e-05, + "loss": 1.2675, + "step": 10471 + }, + { + "epoch": 0.88, + "learning_rate": 1.6564393004619458e-05, + "loss": 1.3512, + "step": 10472 + }, + { + "epoch": 0.88, + "learning_rate": 1.6563707267460825e-05, + "loss": 1.2055, + "step": 10473 + }, + { + "epoch": 0.88, + "learning_rate": 1.656302147607074e-05, + "loss": 1.3666, + "step": 10474 + }, + { + "epoch": 0.88, + "learning_rate": 1.656233563045487e-05, + "loss": 1.3065, + "step": 10475 + }, + { + "epoch": 0.88, + "learning_rate": 1.656164973061889e-05, + "loss": 1.2303, + "step": 10476 + }, + { + "epoch": 0.88, + "learning_rate": 1.6560963776568454e-05, + "loss": 1.35, + "step": 10477 + }, + { + "epoch": 0.88, + "learning_rate": 1.6560277768309233e-05, + "loss": 1.3342, + "step": 10478 + }, + { + "epoch": 0.88, + "learning_rate": 1.6559591705846902e-05, + "loss": 1.3028, + "step": 10479 + }, + { + "epoch": 0.88, + "learning_rate": 1.655890558918712e-05, + "loss": 1.2693, + "step": 10480 + }, + { + "epoch": 0.88, + "learning_rate": 1.6558219418335565e-05, + "loss": 1.2281, + "step": 10481 + }, + { + "epoch": 0.88, + "learning_rate": 1.6557533193297905e-05, + "loss": 1.2599, + "step": 10482 + }, + { + "epoch": 0.88, + "learning_rate": 1.6556846914079803e-05, + "loss": 1.2642, + "step": 10483 + }, + { + "epoch": 0.88, + "learning_rate": 1.6556160580686937e-05, + "loss": 1.2558, + "step": 10484 + }, + { + "epoch": 0.88, + "learning_rate": 1.655547419312497e-05, + "loss": 1.2178, + "step": 10485 + }, + { + "epoch": 0.88, + "learning_rate": 1.6554787751399577e-05, + "loss": 1.3185, + "step": 10486 + }, + { + "epoch": 0.88, + "learning_rate": 1.655410125551643e-05, + "loss": 1.3079, + "step": 10487 + }, + { + "epoch": 0.88, + "learning_rate": 1.6553414705481204e-05, + "loss": 1.3115, + "step": 10488 + }, + { + "epoch": 0.88, + "learning_rate": 1.6552728101299565e-05, + "loss": 1.306, + "step": 10489 + }, + { + "epoch": 0.88, + "learning_rate": 1.655204144297719e-05, + "loss": 1.3181, + "step": 10490 + }, + { + "epoch": 0.88, + "learning_rate": 1.6551354730519753e-05, + "loss": 1.2546, + "step": 10491 + }, + { + "epoch": 0.88, + "learning_rate": 1.6550667963932923e-05, + "loss": 1.3143, + "step": 10492 + }, + { + "epoch": 0.88, + "learning_rate": 1.6549981143222386e-05, + "loss": 1.2284, + "step": 10493 + }, + { + "epoch": 0.88, + "learning_rate": 1.6549294268393802e-05, + "loss": 1.2674, + "step": 10494 + }, + { + "epoch": 0.88, + "learning_rate": 1.6548607339452853e-05, + "loss": 1.2798, + "step": 10495 + }, + { + "epoch": 0.88, + "learning_rate": 1.6547920356405212e-05, + "loss": 1.3214, + "step": 10496 + }, + { + "epoch": 0.88, + "learning_rate": 1.654723331925656e-05, + "loss": 1.2231, + "step": 10497 + }, + { + "epoch": 0.88, + "learning_rate": 1.6546546228012573e-05, + "loss": 1.3408, + "step": 10498 + }, + { + "epoch": 0.88, + "learning_rate": 1.6545859082678925e-05, + "loss": 1.3173, + "step": 10499 + }, + { + "epoch": 0.88, + "learning_rate": 1.654517188326129e-05, + "loss": 1.2008, + "step": 10500 + }, + { + "epoch": 0.88, + "learning_rate": 1.6544484629765358e-05, + "loss": 1.3369, + "step": 10501 + }, + { + "epoch": 0.88, + "learning_rate": 1.6543797322196796e-05, + "loss": 1.3194, + "step": 10502 + }, + { + "epoch": 0.88, + "learning_rate": 1.6543109960561287e-05, + "loss": 1.247, + "step": 10503 + }, + { + "epoch": 0.88, + "learning_rate": 1.6542422544864507e-05, + "loss": 1.3124, + "step": 10504 + }, + { + "epoch": 0.88, + "learning_rate": 1.6541735075112138e-05, + "loss": 1.2306, + "step": 10505 + }, + { + "epoch": 0.88, + "learning_rate": 1.654104755130986e-05, + "loss": 1.3286, + "step": 10506 + }, + { + "epoch": 0.88, + "learning_rate": 1.6540359973463358e-05, + "loss": 1.2775, + "step": 10507 + }, + { + "epoch": 0.88, + "learning_rate": 1.6539672341578305e-05, + "loss": 1.253, + "step": 10508 + }, + { + "epoch": 0.88, + "learning_rate": 1.6538984655660386e-05, + "loss": 1.3371, + "step": 10509 + }, + { + "epoch": 0.88, + "learning_rate": 1.6538296915715284e-05, + "loss": 1.1917, + "step": 10510 + }, + { + "epoch": 0.88, + "learning_rate": 1.6537609121748677e-05, + "loss": 1.2867, + "step": 10511 + }, + { + "epoch": 0.89, + "learning_rate": 1.6536921273766258e-05, + "loss": 1.2167, + "step": 10512 + }, + { + "epoch": 0.89, + "learning_rate": 1.6536233371773704e-05, + "loss": 1.1514, + "step": 10513 + }, + { + "epoch": 0.89, + "learning_rate": 1.653554541577669e-05, + "loss": 1.2699, + "step": 10514 + }, + { + "epoch": 0.89, + "learning_rate": 1.6534857405780914e-05, + "loss": 1.2445, + "step": 10515 + }, + { + "epoch": 0.89, + "learning_rate": 1.6534169341792052e-05, + "loss": 1.284, + "step": 10516 + }, + { + "epoch": 0.89, + "learning_rate": 1.653348122381579e-05, + "loss": 1.2839, + "step": 10517 + }, + { + "epoch": 0.89, + "learning_rate": 1.653279305185782e-05, + "loss": 1.3143, + "step": 10518 + }, + { + "epoch": 0.89, + "learning_rate": 1.653210482592382e-05, + "loss": 1.3918, + "step": 10519 + }, + { + "epoch": 0.89, + "learning_rate": 1.653141654601948e-05, + "loss": 1.2429, + "step": 10520 + }, + { + "epoch": 0.89, + "learning_rate": 1.6530728212150485e-05, + "loss": 1.3235, + "step": 10521 + }, + { + "epoch": 0.89, + "learning_rate": 1.6530039824322526e-05, + "loss": 1.2576, + "step": 10522 + }, + { + "epoch": 0.89, + "learning_rate": 1.6529351382541286e-05, + "loss": 1.2701, + "step": 10523 + }, + { + "epoch": 0.89, + "learning_rate": 1.6528662886812456e-05, + "loss": 1.3144, + "step": 10524 + }, + { + "epoch": 0.89, + "learning_rate": 1.6527974337141723e-05, + "loss": 1.2641, + "step": 10525 + }, + { + "epoch": 0.89, + "learning_rate": 1.652728573353478e-05, + "loss": 1.3446, + "step": 10526 + }, + { + "epoch": 0.89, + "learning_rate": 1.6526597075997308e-05, + "loss": 1.2603, + "step": 10527 + }, + { + "epoch": 0.89, + "learning_rate": 1.6525908364535004e-05, + "loss": 1.2393, + "step": 10528 + }, + { + "epoch": 0.89, + "learning_rate": 1.652521959915356e-05, + "loss": 1.2921, + "step": 10529 + }, + { + "epoch": 0.89, + "learning_rate": 1.652453077985866e-05, + "loss": 1.1915, + "step": 10530 + }, + { + "epoch": 0.89, + "learning_rate": 1.6523841906656e-05, + "loss": 1.2916, + "step": 10531 + }, + { + "epoch": 0.89, + "learning_rate": 1.652315297955127e-05, + "loss": 1.2519, + "step": 10532 + }, + { + "epoch": 0.89, + "learning_rate": 1.652246399855016e-05, + "loss": 1.2492, + "step": 10533 + }, + { + "epoch": 0.89, + "learning_rate": 1.6521774963658366e-05, + "loss": 1.2472, + "step": 10534 + }, + { + "epoch": 0.89, + "learning_rate": 1.652108587488158e-05, + "loss": 1.3199, + "step": 10535 + }, + { + "epoch": 0.89, + "learning_rate": 1.6520396732225496e-05, + "loss": 1.2632, + "step": 10536 + }, + { + "epoch": 0.89, + "learning_rate": 1.651970753569581e-05, + "loss": 1.2425, + "step": 10537 + }, + { + "epoch": 0.89, + "learning_rate": 1.651901828529821e-05, + "loss": 1.2606, + "step": 10538 + }, + { + "epoch": 0.89, + "learning_rate": 1.65183289810384e-05, + "loss": 1.3261, + "step": 10539 + }, + { + "epoch": 0.89, + "learning_rate": 1.6517639622922062e-05, + "loss": 1.2127, + "step": 10540 + }, + { + "epoch": 0.89, + "learning_rate": 1.6516950210954905e-05, + "loss": 1.295, + "step": 10541 + }, + { + "epoch": 0.89, + "learning_rate": 1.651626074514262e-05, + "loss": 1.16, + "step": 10542 + }, + { + "epoch": 0.89, + "learning_rate": 1.65155712254909e-05, + "loss": 1.2754, + "step": 10543 + }, + { + "epoch": 0.89, + "learning_rate": 1.651488165200545e-05, + "loss": 1.295, + "step": 10544 + }, + { + "epoch": 0.89, + "learning_rate": 1.6514192024691956e-05, + "loss": 1.2967, + "step": 10545 + }, + { + "epoch": 0.89, + "learning_rate": 1.6513502343556125e-05, + "loss": 1.3507, + "step": 10546 + }, + { + "epoch": 0.89, + "learning_rate": 1.6512812608603657e-05, + "loss": 1.3019, + "step": 10547 + }, + { + "epoch": 0.89, + "learning_rate": 1.6512122819840243e-05, + "loss": 1.3304, + "step": 10548 + }, + { + "epoch": 0.89, + "learning_rate": 1.6511432977271588e-05, + "loss": 1.2532, + "step": 10549 + }, + { + "epoch": 0.89, + "learning_rate": 1.651074308090339e-05, + "loss": 1.2975, + "step": 10550 + }, + { + "epoch": 0.89, + "learning_rate": 1.651005313074135e-05, + "loss": 1.2498, + "step": 10551 + }, + { + "epoch": 0.89, + "learning_rate": 1.6509363126791163e-05, + "loss": 1.2356, + "step": 10552 + }, + { + "epoch": 0.89, + "learning_rate": 1.6508673069058535e-05, + "loss": 1.2436, + "step": 10553 + }, + { + "epoch": 0.89, + "learning_rate": 1.650798295754917e-05, + "loss": 1.19, + "step": 10554 + }, + { + "epoch": 0.89, + "learning_rate": 1.650729279226877e-05, + "loss": 1.2323, + "step": 10555 + }, + { + "epoch": 0.89, + "learning_rate": 1.650660257322303e-05, + "loss": 1.3323, + "step": 10556 + }, + { + "epoch": 0.89, + "learning_rate": 1.6505912300417657e-05, + "loss": 1.334, + "step": 10557 + }, + { + "epoch": 0.89, + "learning_rate": 1.6505221973858353e-05, + "loss": 1.3293, + "step": 10558 + }, + { + "epoch": 0.89, + "learning_rate": 1.650453159355083e-05, + "loss": 1.2246, + "step": 10559 + }, + { + "epoch": 0.89, + "learning_rate": 1.6503841159500775e-05, + "loss": 1.3192, + "step": 10560 + }, + { + "epoch": 0.89, + "learning_rate": 1.6503150671713907e-05, + "loss": 1.3036, + "step": 10561 + }, + { + "epoch": 0.89, + "learning_rate": 1.650246013019593e-05, + "loss": 1.3198, + "step": 10562 + }, + { + "epoch": 0.89, + "learning_rate": 1.6501769534952545e-05, + "loss": 1.1784, + "step": 10563 + }, + { + "epoch": 0.89, + "learning_rate": 1.650107888598946e-05, + "loss": 1.3548, + "step": 10564 + }, + { + "epoch": 0.89, + "learning_rate": 1.650038818331238e-05, + "loss": 1.3122, + "step": 10565 + }, + { + "epoch": 0.89, + "learning_rate": 1.6499697426927014e-05, + "loss": 1.1881, + "step": 10566 + }, + { + "epoch": 0.89, + "learning_rate": 1.6499006616839064e-05, + "loss": 1.3179, + "step": 10567 + }, + { + "epoch": 0.89, + "learning_rate": 1.649831575305424e-05, + "loss": 1.3014, + "step": 10568 + }, + { + "epoch": 0.89, + "learning_rate": 1.6497624835578258e-05, + "loss": 1.3092, + "step": 10569 + }, + { + "epoch": 0.89, + "learning_rate": 1.6496933864416816e-05, + "loss": 1.2351, + "step": 10570 + }, + { + "epoch": 0.89, + "learning_rate": 1.6496242839575626e-05, + "loss": 1.375, + "step": 10571 + }, + { + "epoch": 0.89, + "learning_rate": 1.64955517610604e-05, + "loss": 1.2461, + "step": 10572 + }, + { + "epoch": 0.89, + "learning_rate": 1.6494860628876848e-05, + "loss": 1.3223, + "step": 10573 + }, + { + "epoch": 0.89, + "learning_rate": 1.649416944303068e-05, + "loss": 1.2485, + "step": 10574 + }, + { + "epoch": 0.89, + "learning_rate": 1.64934782035276e-05, + "loss": 1.3332, + "step": 10575 + }, + { + "epoch": 0.89, + "learning_rate": 1.649278691037333e-05, + "loss": 1.3329, + "step": 10576 + }, + { + "epoch": 0.89, + "learning_rate": 1.6492095563573572e-05, + "loss": 1.2452, + "step": 10577 + }, + { + "epoch": 0.89, + "learning_rate": 1.649140416313405e-05, + "loss": 1.2422, + "step": 10578 + }, + { + "epoch": 0.89, + "learning_rate": 1.6490712709060463e-05, + "loss": 1.3349, + "step": 10579 + }, + { + "epoch": 0.89, + "learning_rate": 1.6490021201358533e-05, + "loss": 1.2198, + "step": 10580 + }, + { + "epoch": 0.89, + "learning_rate": 1.648932964003397e-05, + "loss": 1.2123, + "step": 10581 + }, + { + "epoch": 0.89, + "learning_rate": 1.6488638025092488e-05, + "loss": 1.2354, + "step": 10582 + }, + { + "epoch": 0.89, + "learning_rate": 1.6487946356539804e-05, + "loss": 1.2297, + "step": 10583 + }, + { + "epoch": 0.89, + "learning_rate": 1.6487254634381627e-05, + "loss": 1.3529, + "step": 10584 + }, + { + "epoch": 0.89, + "learning_rate": 1.6486562858623677e-05, + "loss": 1.175, + "step": 10585 + }, + { + "epoch": 0.89, + "learning_rate": 1.648587102927167e-05, + "loss": 1.284, + "step": 10586 + }, + { + "epoch": 0.89, + "learning_rate": 1.6485179146331323e-05, + "loss": 1.223, + "step": 10587 + }, + { + "epoch": 0.89, + "learning_rate": 1.648448720980835e-05, + "loss": 1.242, + "step": 10588 + }, + { + "epoch": 0.89, + "learning_rate": 1.6483795219708464e-05, + "loss": 1.2265, + "step": 10589 + }, + { + "epoch": 0.89, + "learning_rate": 1.648310317603739e-05, + "loss": 1.2354, + "step": 10590 + }, + { + "epoch": 0.89, + "learning_rate": 1.648241107880084e-05, + "loss": 1.1604, + "step": 10591 + }, + { + "epoch": 0.89, + "learning_rate": 1.648171892800454e-05, + "loss": 1.2638, + "step": 10592 + }, + { + "epoch": 0.89, + "learning_rate": 1.64810267236542e-05, + "loss": 1.3514, + "step": 10593 + }, + { + "epoch": 0.89, + "learning_rate": 1.6480334465755545e-05, + "loss": 1.3443, + "step": 10594 + }, + { + "epoch": 0.89, + "learning_rate": 1.6479642154314294e-05, + "loss": 1.229, + "step": 10595 + }, + { + "epoch": 0.89, + "learning_rate": 1.647894978933616e-05, + "loss": 1.3148, + "step": 10596 + }, + { + "epoch": 0.89, + "learning_rate": 1.6478257370826874e-05, + "loss": 1.2065, + "step": 10597 + }, + { + "epoch": 0.89, + "learning_rate": 1.647756489879215e-05, + "loss": 1.3243, + "step": 10598 + }, + { + "epoch": 0.89, + "learning_rate": 1.6476872373237714e-05, + "loss": 1.1569, + "step": 10599 + }, + { + "epoch": 0.89, + "learning_rate": 1.6476179794169285e-05, + "loss": 1.2693, + "step": 10600 + }, + { + "epoch": 0.89, + "learning_rate": 1.6475487161592585e-05, + "loss": 1.1603, + "step": 10601 + }, + { + "epoch": 0.89, + "learning_rate": 1.6474794475513336e-05, + "loss": 1.276, + "step": 10602 + }, + { + "epoch": 0.89, + "learning_rate": 1.6474101735937266e-05, + "loss": 1.2389, + "step": 10603 + }, + { + "epoch": 0.89, + "learning_rate": 1.6473408942870093e-05, + "loss": 1.2945, + "step": 10604 + }, + { + "epoch": 0.89, + "learning_rate": 1.6472716096317546e-05, + "loss": 1.3347, + "step": 10605 + }, + { + "epoch": 0.89, + "learning_rate": 1.6472023196285344e-05, + "loss": 1.2742, + "step": 10606 + }, + { + "epoch": 0.89, + "learning_rate": 1.647133024277922e-05, + "loss": 1.3393, + "step": 10607 + }, + { + "epoch": 0.89, + "learning_rate": 1.647063723580489e-05, + "loss": 1.2571, + "step": 10608 + }, + { + "epoch": 0.89, + "learning_rate": 1.6469944175368085e-05, + "loss": 1.212, + "step": 10609 + }, + { + "epoch": 0.89, + "learning_rate": 1.6469251061474532e-05, + "loss": 1.314, + "step": 10610 + }, + { + "epoch": 0.89, + "learning_rate": 1.6468557894129954e-05, + "loss": 1.2049, + "step": 10611 + }, + { + "epoch": 0.89, + "learning_rate": 1.6467864673340083e-05, + "loss": 1.3868, + "step": 10612 + }, + { + "epoch": 0.89, + "learning_rate": 1.6467171399110643e-05, + "loss": 1.2542, + "step": 10613 + }, + { + "epoch": 0.89, + "learning_rate": 1.6466478071447363e-05, + "loss": 1.2864, + "step": 10614 + }, + { + "epoch": 0.89, + "learning_rate": 1.6465784690355972e-05, + "loss": 1.271, + "step": 10615 + }, + { + "epoch": 0.89, + "learning_rate": 1.64650912558422e-05, + "loss": 1.2293, + "step": 10616 + }, + { + "epoch": 0.89, + "learning_rate": 1.6464397767911772e-05, + "loss": 1.2669, + "step": 10617 + }, + { + "epoch": 0.89, + "learning_rate": 1.6463704226570424e-05, + "loss": 1.2708, + "step": 10618 + }, + { + "epoch": 0.89, + "learning_rate": 1.646301063182388e-05, + "loss": 1.2295, + "step": 10619 + }, + { + "epoch": 0.89, + "learning_rate": 1.6462316983677875e-05, + "loss": 1.3038, + "step": 10620 + }, + { + "epoch": 0.89, + "learning_rate": 1.6461623282138138e-05, + "loss": 1.1961, + "step": 10621 + }, + { + "epoch": 0.89, + "learning_rate": 1.6460929527210402e-05, + "loss": 1.2273, + "step": 10622 + }, + { + "epoch": 0.89, + "learning_rate": 1.64602357189004e-05, + "loss": 1.3072, + "step": 10623 + }, + { + "epoch": 0.89, + "learning_rate": 1.6459541857213863e-05, + "loss": 1.2201, + "step": 10624 + }, + { + "epoch": 0.89, + "learning_rate": 1.645884794215652e-05, + "loss": 1.2376, + "step": 10625 + }, + { + "epoch": 0.89, + "learning_rate": 1.6458153973734113e-05, + "loss": 1.2341, + "step": 10626 + }, + { + "epoch": 0.89, + "learning_rate": 1.6457459951952368e-05, + "loss": 1.3119, + "step": 10627 + }, + { + "epoch": 0.89, + "learning_rate": 1.6456765876817023e-05, + "loss": 1.2892, + "step": 10628 + }, + { + "epoch": 0.89, + "learning_rate": 1.6456071748333813e-05, + "loss": 1.2794, + "step": 10629 + }, + { + "epoch": 0.89, + "learning_rate": 1.645537756650847e-05, + "loss": 1.3059, + "step": 10630 + }, + { + "epoch": 0.9, + "learning_rate": 1.6454683331346733e-05, + "loss": 1.2268, + "step": 10631 + }, + { + "epoch": 0.9, + "learning_rate": 1.6453989042854336e-05, + "loss": 1.2847, + "step": 10632 + }, + { + "epoch": 0.9, + "learning_rate": 1.6453294701037014e-05, + "loss": 1.3001, + "step": 10633 + }, + { + "epoch": 0.9, + "learning_rate": 1.645260030590051e-05, + "loss": 1.2489, + "step": 10634 + }, + { + "epoch": 0.9, + "learning_rate": 1.6451905857450552e-05, + "loss": 1.2973, + "step": 10635 + }, + { + "epoch": 0.9, + "learning_rate": 1.6451211355692888e-05, + "loss": 1.4072, + "step": 10636 + }, + { + "epoch": 0.9, + "learning_rate": 1.645051680063325e-05, + "loss": 1.2705, + "step": 10637 + }, + { + "epoch": 0.9, + "learning_rate": 1.6449822192277377e-05, + "loss": 1.2199, + "step": 10638 + }, + { + "epoch": 0.9, + "learning_rate": 1.6449127530631005e-05, + "loss": 1.2565, + "step": 10639 + }, + { + "epoch": 0.9, + "learning_rate": 1.644843281569988e-05, + "loss": 1.3218, + "step": 10640 + }, + { + "epoch": 0.9, + "learning_rate": 1.6447738047489745e-05, + "loss": 1.2342, + "step": 10641 + }, + { + "epoch": 0.9, + "learning_rate": 1.644704322600633e-05, + "loss": 1.3248, + "step": 10642 + }, + { + "epoch": 0.9, + "learning_rate": 1.644634835125538e-05, + "loss": 1.2825, + "step": 10643 + }, + { + "epoch": 0.9, + "learning_rate": 1.6445653423242637e-05, + "loss": 1.1687, + "step": 10644 + }, + { + "epoch": 0.9, + "learning_rate": 1.6444958441973844e-05, + "loss": 1.2243, + "step": 10645 + }, + { + "epoch": 0.9, + "learning_rate": 1.644426340745474e-05, + "loss": 1.3226, + "step": 10646 + }, + { + "epoch": 0.9, + "learning_rate": 1.6443568319691067e-05, + "loss": 1.3393, + "step": 10647 + }, + { + "epoch": 0.9, + "learning_rate": 1.6442873178688574e-05, + "loss": 1.2247, + "step": 10648 + }, + { + "epoch": 0.9, + "learning_rate": 1.6442177984453e-05, + "loss": 1.3253, + "step": 10649 + }, + { + "epoch": 0.9, + "learning_rate": 1.6441482736990092e-05, + "loss": 1.2405, + "step": 10650 + }, + { + "epoch": 0.9, + "learning_rate": 1.6440787436305588e-05, + "loss": 1.3157, + "step": 10651 + }, + { + "epoch": 0.9, + "learning_rate": 1.644009208240524e-05, + "loss": 1.2941, + "step": 10652 + }, + { + "epoch": 0.9, + "learning_rate": 1.643939667529479e-05, + "loss": 1.2634, + "step": 10653 + }, + { + "epoch": 0.9, + "learning_rate": 1.6438701214979983e-05, + "loss": 1.2555, + "step": 10654 + }, + { + "epoch": 0.9, + "learning_rate": 1.6438005701466565e-05, + "loss": 1.3113, + "step": 10655 + }, + { + "epoch": 0.9, + "learning_rate": 1.6437310134760286e-05, + "loss": 1.2017, + "step": 10656 + }, + { + "epoch": 0.9, + "learning_rate": 1.643661451486689e-05, + "loss": 1.3109, + "step": 10657 + }, + { + "epoch": 0.9, + "learning_rate": 1.643591884179212e-05, + "loss": 1.302, + "step": 10658 + }, + { + "epoch": 0.9, + "learning_rate": 1.6435223115541733e-05, + "loss": 1.2437, + "step": 10659 + }, + { + "epoch": 0.9, + "learning_rate": 1.643452733612147e-05, + "loss": 1.2926, + "step": 10660 + }, + { + "epoch": 0.9, + "learning_rate": 1.643383150353709e-05, + "loss": 1.3844, + "step": 10661 + }, + { + "epoch": 0.9, + "learning_rate": 1.6433135617794328e-05, + "loss": 1.3123, + "step": 10662 + }, + { + "epoch": 0.9, + "learning_rate": 1.6432439678898942e-05, + "loss": 1.2396, + "step": 10663 + }, + { + "epoch": 0.9, + "learning_rate": 1.643174368685668e-05, + "loss": 1.3014, + "step": 10664 + }, + { + "epoch": 0.9, + "learning_rate": 1.6431047641673297e-05, + "loss": 1.3354, + "step": 10665 + }, + { + "epoch": 0.9, + "learning_rate": 1.6430351543354536e-05, + "loss": 1.318, + "step": 10666 + }, + { + "epoch": 0.9, + "learning_rate": 1.6429655391906156e-05, + "loss": 1.3343, + "step": 10667 + }, + { + "epoch": 0.9, + "learning_rate": 1.6428959187333906e-05, + "loss": 1.2829, + "step": 10668 + }, + { + "epoch": 0.9, + "learning_rate": 1.6428262929643535e-05, + "loss": 1.2028, + "step": 10669 + }, + { + "epoch": 0.9, + "learning_rate": 1.64275666188408e-05, + "loss": 1.2547, + "step": 10670 + }, + { + "epoch": 0.9, + "learning_rate": 1.642687025493145e-05, + "loss": 1.1998, + "step": 10671 + }, + { + "epoch": 0.9, + "learning_rate": 1.6426173837921243e-05, + "loss": 1.3314, + "step": 10672 + }, + { + "epoch": 0.9, + "learning_rate": 1.642547736781593e-05, + "loss": 1.2576, + "step": 10673 + }, + { + "epoch": 0.9, + "learning_rate": 1.642478084462127e-05, + "loss": 1.3254, + "step": 10674 + }, + { + "epoch": 0.9, + "learning_rate": 1.642408426834301e-05, + "loss": 1.3016, + "step": 10675 + }, + { + "epoch": 0.9, + "learning_rate": 1.642338763898691e-05, + "loss": 1.259, + "step": 10676 + }, + { + "epoch": 0.9, + "learning_rate": 1.642269095655873e-05, + "loss": 1.2966, + "step": 10677 + }, + { + "epoch": 0.9, + "learning_rate": 1.642199422106422e-05, + "loss": 1.3183, + "step": 10678 + }, + { + "epoch": 0.9, + "learning_rate": 1.6421297432509134e-05, + "loss": 1.3143, + "step": 10679 + }, + { + "epoch": 0.9, + "learning_rate": 1.6420600590899237e-05, + "loss": 1.2452, + "step": 10680 + }, + { + "epoch": 0.9, + "learning_rate": 1.6419903696240283e-05, + "loss": 1.349, + "step": 10681 + }, + { + "epoch": 0.9, + "learning_rate": 1.6419206748538032e-05, + "loss": 1.3046, + "step": 10682 + }, + { + "epoch": 0.9, + "learning_rate": 1.6418509747798237e-05, + "loss": 1.2608, + "step": 10683 + }, + { + "epoch": 0.9, + "learning_rate": 1.6417812694026664e-05, + "loss": 1.3014, + "step": 10684 + }, + { + "epoch": 0.9, + "learning_rate": 1.6417115587229064e-05, + "loss": 1.246, + "step": 10685 + }, + { + "epoch": 0.9, + "learning_rate": 1.6416418427411202e-05, + "loss": 1.3208, + "step": 10686 + }, + { + "epoch": 0.9, + "learning_rate": 1.641572121457884e-05, + "loss": 1.2585, + "step": 10687 + }, + { + "epoch": 0.9, + "learning_rate": 1.6415023948737732e-05, + "loss": 1.255, + "step": 10688 + }, + { + "epoch": 0.9, + "learning_rate": 1.641432662989365e-05, + "loss": 1.1459, + "step": 10689 + }, + { + "epoch": 0.9, + "learning_rate": 1.6413629258052342e-05, + "loss": 1.268, + "step": 10690 + }, + { + "epoch": 0.9, + "learning_rate": 1.6412931833219583e-05, + "loss": 1.3692, + "step": 10691 + }, + { + "epoch": 0.9, + "learning_rate": 1.641223435540112e-05, + "loss": 1.2932, + "step": 10692 + }, + { + "epoch": 0.9, + "learning_rate": 1.6411536824602734e-05, + "loss": 1.2071, + "step": 10693 + }, + { + "epoch": 0.9, + "learning_rate": 1.6410839240830175e-05, + "loss": 1.257, + "step": 10694 + }, + { + "epoch": 0.9, + "learning_rate": 1.6410141604089206e-05, + "loss": 1.3337, + "step": 10695 + }, + { + "epoch": 0.9, + "learning_rate": 1.6409443914385602e-05, + "loss": 1.3214, + "step": 10696 + }, + { + "epoch": 0.9, + "learning_rate": 1.6408746171725115e-05, + "loss": 1.2451, + "step": 10697 + }, + { + "epoch": 0.9, + "learning_rate": 1.6408048376113522e-05, + "loss": 1.2927, + "step": 10698 + }, + { + "epoch": 0.9, + "learning_rate": 1.640735052755658e-05, + "loss": 1.2455, + "step": 10699 + }, + { + "epoch": 0.9, + "learning_rate": 1.6406652626060058e-05, + "loss": 1.3117, + "step": 10700 + }, + { + "epoch": 0.9, + "learning_rate": 1.640595467162972e-05, + "loss": 1.2613, + "step": 10701 + }, + { + "epoch": 0.9, + "learning_rate": 1.6405256664271335e-05, + "loss": 1.2034, + "step": 10702 + }, + { + "epoch": 0.9, + "learning_rate": 1.640455860399067e-05, + "loss": 1.21, + "step": 10703 + }, + { + "epoch": 0.9, + "learning_rate": 1.640386049079349e-05, + "loss": 1.3408, + "step": 10704 + }, + { + "epoch": 0.9, + "learning_rate": 1.640316232468557e-05, + "loss": 1.2158, + "step": 10705 + }, + { + "epoch": 0.9, + "learning_rate": 1.640246410567267e-05, + "loss": 1.2963, + "step": 10706 + }, + { + "epoch": 0.9, + "learning_rate": 1.6401765833760565e-05, + "loss": 1.3929, + "step": 10707 + }, + { + "epoch": 0.9, + "learning_rate": 1.6401067508955018e-05, + "loss": 1.2139, + "step": 10708 + }, + { + "epoch": 0.9, + "learning_rate": 1.6400369131261804e-05, + "loss": 1.3629, + "step": 10709 + }, + { + "epoch": 0.9, + "learning_rate": 1.639967070068669e-05, + "loss": 1.3117, + "step": 10710 + }, + { + "epoch": 0.9, + "learning_rate": 1.6398972217235453e-05, + "loss": 1.2835, + "step": 10711 + }, + { + "epoch": 0.9, + "learning_rate": 1.639827368091386e-05, + "loss": 1.3159, + "step": 10712 + }, + { + "epoch": 0.9, + "learning_rate": 1.6397575091727677e-05, + "loss": 1.22, + "step": 10713 + }, + { + "epoch": 0.9, + "learning_rate": 1.6396876449682683e-05, + "loss": 1.2726, + "step": 10714 + }, + { + "epoch": 0.9, + "learning_rate": 1.639617775478465e-05, + "loss": 1.3068, + "step": 10715 + }, + { + "epoch": 0.9, + "learning_rate": 1.639547900703935e-05, + "loss": 1.2713, + "step": 10716 + }, + { + "epoch": 0.9, + "learning_rate": 1.6394780206452554e-05, + "loss": 1.3581, + "step": 10717 + }, + { + "epoch": 0.9, + "learning_rate": 1.6394081353030035e-05, + "loss": 1.2732, + "step": 10718 + }, + { + "epoch": 0.9, + "learning_rate": 1.6393382446777576e-05, + "loss": 1.4192, + "step": 10719 + }, + { + "epoch": 0.9, + "learning_rate": 1.6392683487700937e-05, + "loss": 1.3337, + "step": 10720 + }, + { + "epoch": 0.9, + "learning_rate": 1.639198447580591e-05, + "loss": 1.2715, + "step": 10721 + }, + { + "epoch": 0.9, + "learning_rate": 1.6391285411098255e-05, + "loss": 1.3962, + "step": 10722 + }, + { + "epoch": 0.9, + "learning_rate": 1.6390586293583754e-05, + "loss": 1.2955, + "step": 10723 + }, + { + "epoch": 0.9, + "learning_rate": 1.6389887123268188e-05, + "loss": 1.3956, + "step": 10724 + }, + { + "epoch": 0.9, + "learning_rate": 1.6389187900157326e-05, + "loss": 1.2561, + "step": 10725 + }, + { + "epoch": 0.9, + "learning_rate": 1.638848862425695e-05, + "loss": 1.2259, + "step": 10726 + }, + { + "epoch": 0.9, + "learning_rate": 1.6387789295572837e-05, + "loss": 1.2861, + "step": 10727 + }, + { + "epoch": 0.9, + "learning_rate": 1.6387089914110765e-05, + "loss": 1.2977, + "step": 10728 + }, + { + "epoch": 0.9, + "learning_rate": 1.6386390479876513e-05, + "loss": 1.3166, + "step": 10729 + }, + { + "epoch": 0.9, + "learning_rate": 1.6385690992875857e-05, + "loss": 1.2501, + "step": 10730 + }, + { + "epoch": 0.9, + "learning_rate": 1.6384991453114575e-05, + "loss": 1.2698, + "step": 10731 + }, + { + "epoch": 0.9, + "learning_rate": 1.6384291860598458e-05, + "loss": 1.1748, + "step": 10732 + }, + { + "epoch": 0.9, + "learning_rate": 1.6383592215333273e-05, + "loss": 1.2874, + "step": 10733 + }, + { + "epoch": 0.9, + "learning_rate": 1.6382892517324807e-05, + "loss": 1.2649, + "step": 10734 + }, + { + "epoch": 0.9, + "learning_rate": 1.638219276657884e-05, + "loss": 1.3308, + "step": 10735 + }, + { + "epoch": 0.9, + "learning_rate": 1.6381492963101154e-05, + "loss": 1.2313, + "step": 10736 + }, + { + "epoch": 0.9, + "learning_rate": 1.638079310689753e-05, + "loss": 1.1601, + "step": 10737 + }, + { + "epoch": 0.9, + "learning_rate": 1.638009319797375e-05, + "loss": 1.2673, + "step": 10738 + }, + { + "epoch": 0.9, + "learning_rate": 1.6379393236335604e-05, + "loss": 1.1841, + "step": 10739 + }, + { + "epoch": 0.9, + "learning_rate": 1.6378693221988867e-05, + "loss": 1.306, + "step": 10740 + }, + { + "epoch": 0.9, + "learning_rate": 1.6377993154939323e-05, + "loss": 1.2285, + "step": 10741 + }, + { + "epoch": 0.9, + "learning_rate": 1.637729303519276e-05, + "loss": 1.2243, + "step": 10742 + }, + { + "epoch": 0.9, + "learning_rate": 1.637659286275496e-05, + "loss": 1.2154, + "step": 10743 + }, + { + "epoch": 0.9, + "learning_rate": 1.637589263763171e-05, + "loss": 1.2838, + "step": 10744 + }, + { + "epoch": 0.9, + "learning_rate": 1.637519235982879e-05, + "loss": 1.3882, + "step": 10745 + }, + { + "epoch": 0.9, + "learning_rate": 1.6374492029351998e-05, + "loss": 1.1929, + "step": 10746 + }, + { + "epoch": 0.9, + "learning_rate": 1.6373791646207108e-05, + "loss": 1.2537, + "step": 10747 + }, + { + "epoch": 0.9, + "learning_rate": 1.637309121039991e-05, + "loss": 1.2476, + "step": 10748 + }, + { + "epoch": 0.9, + "learning_rate": 1.6372390721936198e-05, + "loss": 1.3015, + "step": 10749 + }, + { + "epoch": 0.91, + "learning_rate": 1.637169018082175e-05, + "loss": 1.2917, + "step": 10750 + }, + { + "epoch": 0.91, + "learning_rate": 1.6370989587062364e-05, + "loss": 1.1936, + "step": 10751 + }, + { + "epoch": 0.91, + "learning_rate": 1.6370288940663815e-05, + "loss": 1.2542, + "step": 10752 + }, + { + "epoch": 0.91, + "learning_rate": 1.6369588241631907e-05, + "loss": 1.2398, + "step": 10753 + }, + { + "epoch": 0.91, + "learning_rate": 1.636888748997242e-05, + "loss": 1.3074, + "step": 10754 + }, + { + "epoch": 0.91, + "learning_rate": 1.6368186685691143e-05, + "loss": 1.2694, + "step": 10755 + }, + { + "epoch": 0.91, + "learning_rate": 1.6367485828793876e-05, + "loss": 1.3779, + "step": 10756 + }, + { + "epoch": 0.91, + "learning_rate": 1.63667849192864e-05, + "loss": 1.2656, + "step": 10757 + }, + { + "epoch": 0.91, + "learning_rate": 1.6366083957174507e-05, + "loss": 1.2763, + "step": 10758 + }, + { + "epoch": 0.91, + "learning_rate": 1.636538294246399e-05, + "loss": 1.3264, + "step": 10759 + }, + { + "epoch": 0.91, + "learning_rate": 1.6364681875160646e-05, + "loss": 1.2382, + "step": 10760 + }, + { + "epoch": 0.91, + "learning_rate": 1.6363980755270263e-05, + "loss": 1.2463, + "step": 10761 + }, + { + "epoch": 0.91, + "learning_rate": 1.6363279582798633e-05, + "loss": 1.3438, + "step": 10762 + }, + { + "epoch": 0.91, + "learning_rate": 1.636257835775155e-05, + "loss": 1.3081, + "step": 10763 + }, + { + "epoch": 0.91, + "learning_rate": 1.6361877080134813e-05, + "loss": 1.284, + "step": 10764 + }, + { + "epoch": 0.91, + "learning_rate": 1.6361175749954207e-05, + "loss": 1.2634, + "step": 10765 + }, + { + "epoch": 0.91, + "learning_rate": 1.6360474367215534e-05, + "loss": 1.2935, + "step": 10766 + }, + { + "epoch": 0.91, + "learning_rate": 1.635977293192458e-05, + "loss": 1.3209, + "step": 10767 + }, + { + "epoch": 0.91, + "learning_rate": 1.6359071444087155e-05, + "loss": 1.3157, + "step": 10768 + }, + { + "epoch": 0.91, + "learning_rate": 1.635836990370904e-05, + "loss": 1.2783, + "step": 10769 + }, + { + "epoch": 0.91, + "learning_rate": 1.6357668310796043e-05, + "loss": 1.2819, + "step": 10770 + }, + { + "epoch": 0.91, + "learning_rate": 1.635696666535395e-05, + "loss": 1.2792, + "step": 10771 + }, + { + "epoch": 0.91, + "learning_rate": 1.6356264967388566e-05, + "loss": 1.2159, + "step": 10772 + }, + { + "epoch": 0.91, + "learning_rate": 1.635556321690569e-05, + "loss": 1.2749, + "step": 10773 + }, + { + "epoch": 0.91, + "learning_rate": 1.6354861413911115e-05, + "loss": 1.3088, + "step": 10774 + }, + { + "epoch": 0.91, + "learning_rate": 1.635415955841064e-05, + "loss": 1.317, + "step": 10775 + }, + { + "epoch": 0.91, + "learning_rate": 1.6353457650410065e-05, + "loss": 1.2956, + "step": 10776 + }, + { + "epoch": 0.91, + "learning_rate": 1.635275568991519e-05, + "loss": 1.2964, + "step": 10777 + }, + { + "epoch": 0.91, + "learning_rate": 1.6352053676931815e-05, + "loss": 1.1894, + "step": 10778 + }, + { + "epoch": 0.91, + "learning_rate": 1.635135161146574e-05, + "loss": 1.3003, + "step": 10779 + }, + { + "epoch": 0.91, + "learning_rate": 1.6350649493522764e-05, + "loss": 1.2494, + "step": 10780 + }, + { + "epoch": 0.91, + "learning_rate": 1.634994732310869e-05, + "loss": 1.3556, + "step": 10781 + }, + { + "epoch": 0.91, + "learning_rate": 1.634924510022932e-05, + "loss": 1.1746, + "step": 10782 + }, + { + "epoch": 0.91, + "learning_rate": 1.6348542824890448e-05, + "loss": 1.2697, + "step": 10783 + }, + { + "epoch": 0.91, + "learning_rate": 1.634784049709789e-05, + "loss": 1.307, + "step": 10784 + }, + { + "epoch": 0.91, + "learning_rate": 1.6347138116857443e-05, + "loss": 1.2462, + "step": 10785 + }, + { + "epoch": 0.91, + "learning_rate": 1.6346435684174907e-05, + "loss": 1.3228, + "step": 10786 + }, + { + "epoch": 0.91, + "learning_rate": 1.634573319905609e-05, + "loss": 1.2086, + "step": 10787 + }, + { + "epoch": 0.91, + "learning_rate": 1.6345030661506795e-05, + "loss": 1.2334, + "step": 10788 + }, + { + "epoch": 0.91, + "learning_rate": 1.6344328071532822e-05, + "loss": 1.246, + "step": 10789 + }, + { + "epoch": 0.91, + "learning_rate": 1.6343625429139982e-05, + "loss": 1.2903, + "step": 10790 + }, + { + "epoch": 0.91, + "learning_rate": 1.6342922734334078e-05, + "loss": 1.2767, + "step": 10791 + }, + { + "epoch": 0.91, + "learning_rate": 1.634221998712092e-05, + "loss": 1.3641, + "step": 10792 + }, + { + "epoch": 0.91, + "learning_rate": 1.6341517187506307e-05, + "loss": 1.3402, + "step": 10793 + }, + { + "epoch": 0.91, + "learning_rate": 1.634081433549605e-05, + "loss": 1.3504, + "step": 10794 + }, + { + "epoch": 0.91, + "learning_rate": 1.634011143109596e-05, + "loss": 1.2049, + "step": 10795 + }, + { + "epoch": 0.91, + "learning_rate": 1.6339408474311837e-05, + "loss": 1.2148, + "step": 10796 + }, + { + "epoch": 0.91, + "learning_rate": 1.633870546514949e-05, + "loss": 1.2867, + "step": 10797 + }, + { + "epoch": 0.91, + "learning_rate": 1.633800240361473e-05, + "loss": 1.3114, + "step": 10798 + }, + { + "epoch": 0.91, + "learning_rate": 1.6337299289713367e-05, + "loss": 1.2981, + "step": 10799 + }, + { + "epoch": 0.91, + "learning_rate": 1.6336596123451208e-05, + "loss": 1.2428, + "step": 10800 + }, + { + "epoch": 0.91, + "learning_rate": 1.6335892904834068e-05, + "loss": 1.2985, + "step": 10801 + }, + { + "epoch": 0.91, + "learning_rate": 1.633518963386775e-05, + "loss": 1.2764, + "step": 10802 + }, + { + "epoch": 0.91, + "learning_rate": 1.633448631055807e-05, + "loss": 1.2526, + "step": 10803 + }, + { + "epoch": 0.91, + "learning_rate": 1.633378293491083e-05, + "loss": 1.2892, + "step": 10804 + }, + { + "epoch": 0.91, + "learning_rate": 1.6333079506931855e-05, + "loss": 1.3776, + "step": 10805 + }, + { + "epoch": 0.91, + "learning_rate": 1.633237602662695e-05, + "loss": 1.2158, + "step": 10806 + }, + { + "epoch": 0.91, + "learning_rate": 1.6331672494001926e-05, + "loss": 1.2385, + "step": 10807 + }, + { + "epoch": 0.91, + "learning_rate": 1.6330968909062596e-05, + "loss": 1.2749, + "step": 10808 + }, + { + "epoch": 0.91, + "learning_rate": 1.633026527181478e-05, + "loss": 1.3932, + "step": 10809 + }, + { + "epoch": 0.91, + "learning_rate": 1.632956158226428e-05, + "loss": 1.2622, + "step": 10810 + }, + { + "epoch": 0.91, + "learning_rate": 1.6328857840416916e-05, + "loss": 1.2565, + "step": 10811 + }, + { + "epoch": 0.91, + "learning_rate": 1.6328154046278507e-05, + "loss": 1.375, + "step": 10812 + }, + { + "epoch": 0.91, + "learning_rate": 1.6327450199854865e-05, + "loss": 1.1858, + "step": 10813 + }, + { + "epoch": 0.91, + "learning_rate": 1.6326746301151804e-05, + "loss": 1.1734, + "step": 10814 + }, + { + "epoch": 0.91, + "learning_rate": 1.6326042350175137e-05, + "loss": 1.2001, + "step": 10815 + }, + { + "epoch": 0.91, + "learning_rate": 1.6325338346930687e-05, + "loss": 1.2972, + "step": 10816 + }, + { + "epoch": 0.91, + "learning_rate": 1.6324634291424263e-05, + "loss": 1.3109, + "step": 10817 + }, + { + "epoch": 0.91, + "learning_rate": 1.632393018366169e-05, + "loss": 1.1781, + "step": 10818 + }, + { + "epoch": 0.91, + "learning_rate": 1.6323226023648778e-05, + "loss": 1.2805, + "step": 10819 + }, + { + "epoch": 0.91, + "learning_rate": 1.6322521811391355e-05, + "loss": 1.2451, + "step": 10820 + }, + { + "epoch": 0.91, + "learning_rate": 1.6321817546895232e-05, + "loss": 1.3321, + "step": 10821 + }, + { + "epoch": 0.91, + "learning_rate": 1.6321113230166222e-05, + "loss": 1.2147, + "step": 10822 + }, + { + "epoch": 0.91, + "learning_rate": 1.6320408861210157e-05, + "loss": 1.3621, + "step": 10823 + }, + { + "epoch": 0.91, + "learning_rate": 1.6319704440032854e-05, + "loss": 1.3094, + "step": 10824 + }, + { + "epoch": 0.91, + "learning_rate": 1.6318999966640127e-05, + "loss": 1.3306, + "step": 10825 + }, + { + "epoch": 0.91, + "learning_rate": 1.63182954410378e-05, + "loss": 1.2531, + "step": 10826 + }, + { + "epoch": 0.91, + "learning_rate": 1.6317590863231694e-05, + "loss": 1.3238, + "step": 10827 + }, + { + "epoch": 0.91, + "learning_rate": 1.631688623322763e-05, + "loss": 1.2265, + "step": 10828 + }, + { + "epoch": 0.91, + "learning_rate": 1.631618155103143e-05, + "loss": 1.3701, + "step": 10829 + }, + { + "epoch": 0.91, + "learning_rate": 1.6315476816648918e-05, + "loss": 1.2544, + "step": 10830 + }, + { + "epoch": 0.91, + "learning_rate": 1.6314772030085914e-05, + "loss": 1.2433, + "step": 10831 + }, + { + "epoch": 0.91, + "learning_rate": 1.6314067191348243e-05, + "loss": 1.3203, + "step": 10832 + }, + { + "epoch": 0.91, + "learning_rate": 1.6313362300441727e-05, + "loss": 1.2537, + "step": 10833 + }, + { + "epoch": 0.91, + "learning_rate": 1.6312657357372195e-05, + "loss": 1.2714, + "step": 10834 + }, + { + "epoch": 0.91, + "learning_rate": 1.631195236214546e-05, + "loss": 1.292, + "step": 10835 + }, + { + "epoch": 0.91, + "learning_rate": 1.6311247314767363e-05, + "loss": 1.2748, + "step": 10836 + }, + { + "epoch": 0.91, + "learning_rate": 1.6310542215243717e-05, + "loss": 1.1823, + "step": 10837 + }, + { + "epoch": 0.91, + "learning_rate": 1.6309837063580355e-05, + "loss": 1.2458, + "step": 10838 + }, + { + "epoch": 0.91, + "learning_rate": 1.6309131859783095e-05, + "loss": 1.2636, + "step": 10839 + }, + { + "epoch": 0.91, + "learning_rate": 1.630842660385777e-05, + "loss": 1.3507, + "step": 10840 + }, + { + "epoch": 0.91, + "learning_rate": 1.6307721295810204e-05, + "loss": 1.275, + "step": 10841 + }, + { + "epoch": 0.91, + "learning_rate": 1.630701593564623e-05, + "loss": 1.3103, + "step": 10842 + }, + { + "epoch": 0.91, + "learning_rate": 1.630631052337167e-05, + "loss": 1.2578, + "step": 10843 + }, + { + "epoch": 0.91, + "learning_rate": 1.630560505899235e-05, + "loss": 1.2739, + "step": 10844 + }, + { + "epoch": 0.91, + "learning_rate": 1.6304899542514107e-05, + "loss": 1.3135, + "step": 10845 + }, + { + "epoch": 0.91, + "learning_rate": 1.6304193973942765e-05, + "loss": 1.2586, + "step": 10846 + }, + { + "epoch": 0.91, + "learning_rate": 1.6303488353284155e-05, + "loss": 1.2647, + "step": 10847 + }, + { + "epoch": 0.91, + "learning_rate": 1.630278268054411e-05, + "loss": 1.2608, + "step": 10848 + }, + { + "epoch": 0.91, + "learning_rate": 1.6302076955728456e-05, + "loss": 1.3349, + "step": 10849 + }, + { + "epoch": 0.91, + "learning_rate": 1.630137117884302e-05, + "loss": 1.2618, + "step": 10850 + }, + { + "epoch": 0.91, + "learning_rate": 1.6300665349893643e-05, + "loss": 1.3281, + "step": 10851 + }, + { + "epoch": 0.91, + "learning_rate": 1.629995946888615e-05, + "loss": 1.3085, + "step": 10852 + }, + { + "epoch": 0.91, + "learning_rate": 1.629925353582638e-05, + "loss": 1.2592, + "step": 10853 + }, + { + "epoch": 0.91, + "learning_rate": 1.629854755072016e-05, + "loss": 1.3183, + "step": 10854 + }, + { + "epoch": 0.91, + "learning_rate": 1.6297841513573322e-05, + "loss": 1.2797, + "step": 10855 + }, + { + "epoch": 0.91, + "learning_rate": 1.6297135424391702e-05, + "loss": 1.1557, + "step": 10856 + }, + { + "epoch": 0.91, + "learning_rate": 1.6296429283181137e-05, + "loss": 1.2935, + "step": 10857 + }, + { + "epoch": 0.91, + "learning_rate": 1.6295723089947455e-05, + "loss": 1.3074, + "step": 10858 + }, + { + "epoch": 0.91, + "learning_rate": 1.6295016844696494e-05, + "loss": 1.2398, + "step": 10859 + }, + { + "epoch": 0.91, + "learning_rate": 1.629431054743409e-05, + "loss": 1.2843, + "step": 10860 + }, + { + "epoch": 0.91, + "learning_rate": 1.629360419816608e-05, + "loss": 1.4042, + "step": 10861 + }, + { + "epoch": 0.91, + "learning_rate": 1.6292897796898293e-05, + "loss": 1.2272, + "step": 10862 + }, + { + "epoch": 0.91, + "learning_rate": 1.6292191343636576e-05, + "loss": 1.2946, + "step": 10863 + }, + { + "epoch": 0.91, + "learning_rate": 1.6291484838386754e-05, + "loss": 1.263, + "step": 10864 + }, + { + "epoch": 0.91, + "learning_rate": 1.6290778281154677e-05, + "loss": 1.3419, + "step": 10865 + }, + { + "epoch": 0.91, + "learning_rate": 1.6290071671946174e-05, + "loss": 1.4412, + "step": 10866 + }, + { + "epoch": 0.91, + "learning_rate": 1.6289365010767084e-05, + "loss": 1.2873, + "step": 10867 + }, + { + "epoch": 0.91, + "learning_rate": 1.628865829762325e-05, + "loss": 1.2475, + "step": 10868 + }, + { + "epoch": 0.92, + "learning_rate": 1.6287951532520507e-05, + "loss": 1.295, + "step": 10869 + }, + { + "epoch": 0.92, + "learning_rate": 1.6287244715464698e-05, + "loss": 1.2087, + "step": 10870 + }, + { + "epoch": 0.92, + "learning_rate": 1.628653784646166e-05, + "loss": 1.209, + "step": 10871 + }, + { + "epoch": 0.92, + "learning_rate": 1.628583092551723e-05, + "loss": 1.2555, + "step": 10872 + }, + { + "epoch": 0.92, + "learning_rate": 1.6285123952637255e-05, + "loss": 1.2759, + "step": 10873 + }, + { + "epoch": 0.92, + "learning_rate": 1.628441692782758e-05, + "loss": 1.2082, + "step": 10874 + }, + { + "epoch": 0.92, + "learning_rate": 1.628370985109404e-05, + "loss": 1.1416, + "step": 10875 + }, + { + "epoch": 0.92, + "learning_rate": 1.6283002722442473e-05, + "loss": 1.2981, + "step": 10876 + }, + { + "epoch": 0.92, + "learning_rate": 1.6282295541878728e-05, + "loss": 1.3265, + "step": 10877 + }, + { + "epoch": 0.92, + "learning_rate": 1.6281588309408648e-05, + "loss": 1.2237, + "step": 10878 + }, + { + "epoch": 0.92, + "learning_rate": 1.6280881025038077e-05, + "loss": 1.2344, + "step": 10879 + }, + { + "epoch": 0.92, + "learning_rate": 1.6280173688772857e-05, + "loss": 1.2383, + "step": 10880 + }, + { + "epoch": 0.92, + "learning_rate": 1.6279466300618826e-05, + "loss": 1.3101, + "step": 10881 + }, + { + "epoch": 0.92, + "learning_rate": 1.627875886058184e-05, + "loss": 1.2935, + "step": 10882 + }, + { + "epoch": 0.92, + "learning_rate": 1.6278051368667738e-05, + "loss": 1.331, + "step": 10883 + }, + { + "epoch": 0.92, + "learning_rate": 1.6277343824882367e-05, + "loss": 1.2034, + "step": 10884 + }, + { + "epoch": 0.92, + "learning_rate": 1.627663622923157e-05, + "loss": 1.2014, + "step": 10885 + }, + { + "epoch": 0.92, + "learning_rate": 1.6275928581721196e-05, + "loss": 1.2669, + "step": 10886 + }, + { + "epoch": 0.92, + "learning_rate": 1.6275220882357092e-05, + "loss": 1.2049, + "step": 10887 + }, + { + "epoch": 0.92, + "learning_rate": 1.6274513131145105e-05, + "loss": 1.2211, + "step": 10888 + }, + { + "epoch": 0.92, + "learning_rate": 1.627380532809108e-05, + "loss": 1.3149, + "step": 10889 + }, + { + "epoch": 0.92, + "learning_rate": 1.6273097473200872e-05, + "loss": 1.2409, + "step": 10890 + }, + { + "epoch": 0.92, + "learning_rate": 1.6272389566480324e-05, + "loss": 1.2452, + "step": 10891 + }, + { + "epoch": 0.92, + "learning_rate": 1.6271681607935286e-05, + "loss": 1.2488, + "step": 10892 + }, + { + "epoch": 0.92, + "learning_rate": 1.6270973597571607e-05, + "loss": 1.3255, + "step": 10893 + }, + { + "epoch": 0.92, + "learning_rate": 1.6270265535395133e-05, + "loss": 1.2467, + "step": 10894 + }, + { + "epoch": 0.92, + "learning_rate": 1.6269557421411722e-05, + "loss": 1.2713, + "step": 10895 + }, + { + "epoch": 0.92, + "learning_rate": 1.626884925562722e-05, + "loss": 1.2901, + "step": 10896 + }, + { + "epoch": 0.92, + "learning_rate": 1.626814103804748e-05, + "loss": 1.2712, + "step": 10897 + }, + { + "epoch": 0.92, + "learning_rate": 1.6267432768678355e-05, + "loss": 1.2705, + "step": 10898 + }, + { + "epoch": 0.92, + "learning_rate": 1.6266724447525694e-05, + "loss": 1.2789, + "step": 10899 + }, + { + "epoch": 0.92, + "learning_rate": 1.626601607459535e-05, + "loss": 1.2822, + "step": 10900 + }, + { + "epoch": 0.92, + "learning_rate": 1.6265307649893172e-05, + "loss": 1.246, + "step": 10901 + }, + { + "epoch": 0.92, + "learning_rate": 1.626459917342502e-05, + "loss": 1.2869, + "step": 10902 + }, + { + "epoch": 0.92, + "learning_rate": 1.6263890645196744e-05, + "loss": 1.2878, + "step": 10903 + }, + { + "epoch": 0.92, + "learning_rate": 1.62631820652142e-05, + "loss": 1.2399, + "step": 10904 + }, + { + "epoch": 0.92, + "learning_rate": 1.6262473433483238e-05, + "loss": 1.3333, + "step": 10905 + }, + { + "epoch": 0.92, + "learning_rate": 1.626176475000972e-05, + "loss": 1.2619, + "step": 10906 + }, + { + "epoch": 0.92, + "learning_rate": 1.6261056014799494e-05, + "loss": 1.3339, + "step": 10907 + }, + { + "epoch": 0.92, + "learning_rate": 1.626034722785842e-05, + "loss": 1.1102, + "step": 10908 + }, + { + "epoch": 0.92, + "learning_rate": 1.6259638389192357e-05, + "loss": 1.2928, + "step": 10909 + }, + { + "epoch": 0.92, + "learning_rate": 1.6258929498807156e-05, + "loss": 1.2215, + "step": 10910 + }, + { + "epoch": 0.92, + "learning_rate": 1.6258220556708675e-05, + "loss": 1.2043, + "step": 10911 + }, + { + "epoch": 0.92, + "learning_rate": 1.6257511562902777e-05, + "loss": 1.2775, + "step": 10912 + }, + { + "epoch": 0.92, + "learning_rate": 1.6256802517395312e-05, + "loss": 1.29, + "step": 10913 + }, + { + "epoch": 0.92, + "learning_rate": 1.6256093420192146e-05, + "loss": 1.2032, + "step": 10914 + }, + { + "epoch": 0.92, + "learning_rate": 1.625538427129913e-05, + "loss": 1.3008, + "step": 10915 + }, + { + "epoch": 0.92, + "learning_rate": 1.625467507072213e-05, + "loss": 1.2864, + "step": 10916 + }, + { + "epoch": 0.92, + "learning_rate": 1.6253965818467e-05, + "loss": 1.2974, + "step": 10917 + }, + { + "epoch": 0.92, + "learning_rate": 1.6253256514539605e-05, + "loss": 1.2217, + "step": 10918 + }, + { + "epoch": 0.92, + "learning_rate": 1.6252547158945803e-05, + "loss": 1.3372, + "step": 10919 + }, + { + "epoch": 0.92, + "learning_rate": 1.6251837751691454e-05, + "loss": 1.1753, + "step": 10920 + }, + { + "epoch": 0.92, + "learning_rate": 1.6251128292782425e-05, + "loss": 1.3183, + "step": 10921 + }, + { + "epoch": 0.92, + "learning_rate": 1.6250418782224568e-05, + "loss": 1.3059, + "step": 10922 + }, + { + "epoch": 0.92, + "learning_rate": 1.6249709220023753e-05, + "loss": 1.2454, + "step": 10923 + }, + { + "epoch": 0.92, + "learning_rate": 1.6248999606185838e-05, + "loss": 1.2378, + "step": 10924 + }, + { + "epoch": 0.92, + "learning_rate": 1.624828994071669e-05, + "loss": 1.2572, + "step": 10925 + }, + { + "epoch": 0.92, + "learning_rate": 1.6247580223622172e-05, + "loss": 1.2427, + "step": 10926 + }, + { + "epoch": 0.92, + "learning_rate": 1.6246870454908145e-05, + "loss": 1.1962, + "step": 10927 + }, + { + "epoch": 0.92, + "learning_rate": 1.624616063458048e-05, + "loss": 1.2732, + "step": 10928 + }, + { + "epoch": 0.92, + "learning_rate": 1.624545076264503e-05, + "loss": 1.2961, + "step": 10929 + }, + { + "epoch": 0.92, + "learning_rate": 1.624474083910767e-05, + "loss": 1.2995, + "step": 10930 + }, + { + "epoch": 0.92, + "learning_rate": 1.624403086397426e-05, + "loss": 1.2815, + "step": 10931 + }, + { + "epoch": 0.92, + "learning_rate": 1.6243320837250674e-05, + "loss": 1.3337, + "step": 10932 + }, + { + "epoch": 0.92, + "learning_rate": 1.624261075894277e-05, + "loss": 1.3612, + "step": 10933 + }, + { + "epoch": 0.92, + "learning_rate": 1.6241900629056416e-05, + "loss": 1.2028, + "step": 10934 + }, + { + "epoch": 0.92, + "learning_rate": 1.6241190447597484e-05, + "loss": 1.2404, + "step": 10935 + }, + { + "epoch": 0.92, + "learning_rate": 1.624048021457184e-05, + "loss": 1.3024, + "step": 10936 + }, + { + "epoch": 0.92, + "learning_rate": 1.623976992998535e-05, + "loss": 1.2459, + "step": 10937 + }, + { + "epoch": 0.92, + "learning_rate": 1.6239059593843882e-05, + "loss": 1.324, + "step": 10938 + }, + { + "epoch": 0.92, + "learning_rate": 1.623834920615331e-05, + "loss": 1.233, + "step": 10939 + }, + { + "epoch": 0.92, + "learning_rate": 1.6237638766919496e-05, + "loss": 1.2752, + "step": 10940 + }, + { + "epoch": 0.92, + "learning_rate": 1.623692827614832e-05, + "loss": 1.336, + "step": 10941 + }, + { + "epoch": 0.92, + "learning_rate": 1.623621773384564e-05, + "loss": 1.3877, + "step": 10942 + }, + { + "epoch": 0.92, + "learning_rate": 1.6235507140017338e-05, + "loss": 1.1629, + "step": 10943 + }, + { + "epoch": 0.92, + "learning_rate": 1.623479649466928e-05, + "loss": 1.3062, + "step": 10944 + }, + { + "epoch": 0.92, + "learning_rate": 1.6234085797807336e-05, + "loss": 1.224, + "step": 10945 + }, + { + "epoch": 0.92, + "learning_rate": 1.6233375049437384e-05, + "loss": 1.3827, + "step": 10946 + }, + { + "epoch": 0.92, + "learning_rate": 1.6232664249565288e-05, + "loss": 1.3069, + "step": 10947 + }, + { + "epoch": 0.92, + "learning_rate": 1.6231953398196925e-05, + "loss": 1.2861, + "step": 10948 + }, + { + "epoch": 0.92, + "learning_rate": 1.6231242495338172e-05, + "loss": 1.2962, + "step": 10949 + }, + { + "epoch": 0.92, + "learning_rate": 1.62305315409949e-05, + "loss": 1.2573, + "step": 10950 + }, + { + "epoch": 0.92, + "learning_rate": 1.6229820535172976e-05, + "loss": 1.296, + "step": 10951 + }, + { + "epoch": 0.92, + "learning_rate": 1.6229109477878287e-05, + "loss": 1.2197, + "step": 10952 + }, + { + "epoch": 0.92, + "learning_rate": 1.62283983691167e-05, + "loss": 1.2745, + "step": 10953 + }, + { + "epoch": 0.92, + "learning_rate": 1.622768720889409e-05, + "loss": 1.3837, + "step": 10954 + }, + { + "epoch": 0.92, + "learning_rate": 1.6226975997216338e-05, + "loss": 1.1717, + "step": 10955 + }, + { + "epoch": 0.92, + "learning_rate": 1.622626473408932e-05, + "loss": 1.3181, + "step": 10956 + }, + { + "epoch": 0.92, + "learning_rate": 1.6225553419518906e-05, + "loss": 1.3613, + "step": 10957 + }, + { + "epoch": 0.92, + "learning_rate": 1.6224842053510977e-05, + "loss": 1.3319, + "step": 10958 + }, + { + "epoch": 0.92, + "learning_rate": 1.6224130636071414e-05, + "loss": 1.2925, + "step": 10959 + }, + { + "epoch": 0.92, + "learning_rate": 1.6223419167206086e-05, + "loss": 1.2908, + "step": 10960 + }, + { + "epoch": 0.92, + "learning_rate": 1.622270764692088e-05, + "loss": 1.3204, + "step": 10961 + }, + { + "epoch": 0.92, + "learning_rate": 1.6221996075221676e-05, + "loss": 1.415, + "step": 10962 + }, + { + "epoch": 0.92, + "learning_rate": 1.6221284452114345e-05, + "loss": 1.2635, + "step": 10963 + }, + { + "epoch": 0.92, + "learning_rate": 1.6220572777604772e-05, + "loss": 1.3157, + "step": 10964 + }, + { + "epoch": 0.92, + "learning_rate": 1.6219861051698832e-05, + "loss": 1.3277, + "step": 10965 + }, + { + "epoch": 0.92, + "learning_rate": 1.6219149274402414e-05, + "loss": 1.2198, + "step": 10966 + }, + { + "epoch": 0.92, + "learning_rate": 1.6218437445721393e-05, + "loss": 1.1746, + "step": 10967 + }, + { + "epoch": 0.92, + "learning_rate": 1.6217725565661654e-05, + "loss": 1.2219, + "step": 10968 + }, + { + "epoch": 0.92, + "learning_rate": 1.621701363422907e-05, + "loss": 1.2884, + "step": 10969 + }, + { + "epoch": 0.92, + "learning_rate": 1.6216301651429538e-05, + "loss": 1.2785, + "step": 10970 + }, + { + "epoch": 0.92, + "learning_rate": 1.621558961726893e-05, + "loss": 1.2666, + "step": 10971 + }, + { + "epoch": 0.92, + "learning_rate": 1.6214877531753125e-05, + "loss": 1.265, + "step": 10972 + }, + { + "epoch": 0.92, + "learning_rate": 1.6214165394888016e-05, + "loss": 1.2993, + "step": 10973 + }, + { + "epoch": 0.92, + "learning_rate": 1.6213453206679486e-05, + "loss": 1.2038, + "step": 10974 + }, + { + "epoch": 0.92, + "learning_rate": 1.6212740967133414e-05, + "loss": 1.3254, + "step": 10975 + }, + { + "epoch": 0.92, + "learning_rate": 1.6212028676255692e-05, + "loss": 1.1875, + "step": 10976 + }, + { + "epoch": 0.92, + "learning_rate": 1.6211316334052202e-05, + "loss": 1.2334, + "step": 10977 + }, + { + "epoch": 0.92, + "learning_rate": 1.6210603940528825e-05, + "loss": 1.3067, + "step": 10978 + }, + { + "epoch": 0.92, + "learning_rate": 1.620989149569145e-05, + "loss": 1.2803, + "step": 10979 + }, + { + "epoch": 0.92, + "learning_rate": 1.6209178999545966e-05, + "loss": 1.2764, + "step": 10980 + }, + { + "epoch": 0.92, + "learning_rate": 1.6208466452098255e-05, + "loss": 1.2299, + "step": 10981 + }, + { + "epoch": 0.92, + "learning_rate": 1.620775385335421e-05, + "loss": 1.2808, + "step": 10982 + }, + { + "epoch": 0.92, + "learning_rate": 1.6207041203319716e-05, + "loss": 1.2882, + "step": 10983 + }, + { + "epoch": 0.92, + "learning_rate": 1.620632850200066e-05, + "loss": 1.2865, + "step": 10984 + }, + { + "epoch": 0.92, + "learning_rate": 1.620561574940293e-05, + "loss": 1.322, + "step": 10985 + }, + { + "epoch": 0.92, + "learning_rate": 1.6204902945532417e-05, + "loss": 1.2354, + "step": 10986 + }, + { + "epoch": 0.93, + "learning_rate": 1.6204190090395013e-05, + "loss": 1.3406, + "step": 10987 + }, + { + "epoch": 0.93, + "learning_rate": 1.6203477183996602e-05, + "loss": 1.3093, + "step": 10988 + }, + { + "epoch": 0.93, + "learning_rate": 1.620276422634308e-05, + "loss": 1.2687, + "step": 10989 + }, + { + "epoch": 0.93, + "learning_rate": 1.620205121744033e-05, + "loss": 1.1595, + "step": 10990 + }, + { + "epoch": 0.93, + "learning_rate": 1.6201338157294256e-05, + "loss": 1.3687, + "step": 10991 + }, + { + "epoch": 0.93, + "learning_rate": 1.6200625045910735e-05, + "loss": 1.2688, + "step": 10992 + }, + { + "epoch": 0.93, + "learning_rate": 1.619991188329567e-05, + "loss": 1.2714, + "step": 10993 + }, + { + "epoch": 0.93, + "learning_rate": 1.6199198669454943e-05, + "loss": 1.2261, + "step": 10994 + }, + { + "epoch": 0.93, + "learning_rate": 1.6198485404394454e-05, + "loss": 1.405, + "step": 10995 + }, + { + "epoch": 0.93, + "learning_rate": 1.6197772088120096e-05, + "loss": 1.2927, + "step": 10996 + }, + { + "epoch": 0.93, + "learning_rate": 1.6197058720637762e-05, + "loss": 1.2997, + "step": 10997 + }, + { + "epoch": 0.93, + "learning_rate": 1.6196345301953343e-05, + "loss": 1.3118, + "step": 10998 + }, + { + "epoch": 0.93, + "learning_rate": 1.6195631832072742e-05, + "loss": 1.302, + "step": 10999 + }, + { + "epoch": 0.93, + "learning_rate": 1.6194918311001844e-05, + "loss": 1.3395, + "step": 11000 + }, + { + "epoch": 0.93, + "learning_rate": 1.6194204738746547e-05, + "loss": 1.3207, + "step": 11001 + }, + { + "epoch": 0.93, + "learning_rate": 1.619349111531275e-05, + "loss": 1.2446, + "step": 11002 + }, + { + "epoch": 0.93, + "learning_rate": 1.619277744070635e-05, + "loss": 1.2826, + "step": 11003 + }, + { + "epoch": 0.93, + "learning_rate": 1.6192063714933235e-05, + "loss": 1.279, + "step": 11004 + }, + { + "epoch": 0.93, + "learning_rate": 1.619134993799931e-05, + "loss": 1.3195, + "step": 11005 + }, + { + "epoch": 0.93, + "learning_rate": 1.6190636109910472e-05, + "loss": 1.3051, + "step": 11006 + }, + { + "epoch": 0.93, + "learning_rate": 1.6189922230672617e-05, + "loss": 1.2318, + "step": 11007 + }, + { + "epoch": 0.93, + "learning_rate": 1.6189208300291644e-05, + "loss": 1.1661, + "step": 11008 + }, + { + "epoch": 0.93, + "learning_rate": 1.618849431877345e-05, + "loss": 1.2582, + "step": 11009 + }, + { + "epoch": 0.93, + "learning_rate": 1.6187780286123935e-05, + "loss": 1.3866, + "step": 11010 + }, + { + "epoch": 0.93, + "learning_rate": 1.6187066202348995e-05, + "loss": 1.2717, + "step": 11011 + }, + { + "epoch": 0.93, + "learning_rate": 1.618635206745454e-05, + "loss": 1.2145, + "step": 11012 + }, + { + "epoch": 0.93, + "learning_rate": 1.618563788144646e-05, + "loss": 1.2832, + "step": 11013 + }, + { + "epoch": 0.93, + "learning_rate": 1.6184923644330663e-05, + "loss": 1.2443, + "step": 11014 + }, + { + "epoch": 0.93, + "learning_rate": 1.618420935611305e-05, + "loss": 1.2618, + "step": 11015 + }, + { + "epoch": 0.93, + "learning_rate": 1.6183495016799518e-05, + "loss": 1.2485, + "step": 11016 + }, + { + "epoch": 0.93, + "learning_rate": 1.6182780626395968e-05, + "loss": 1.2947, + "step": 11017 + }, + { + "epoch": 0.93, + "learning_rate": 1.6182066184908308e-05, + "loss": 1.3058, + "step": 11018 + }, + { + "epoch": 0.93, + "learning_rate": 1.6181351692342437e-05, + "loss": 1.2678, + "step": 11019 + }, + { + "epoch": 0.93, + "learning_rate": 1.6180637148704264e-05, + "loss": 1.2493, + "step": 11020 + }, + { + "epoch": 0.93, + "learning_rate": 1.6179922553999684e-05, + "loss": 1.2395, + "step": 11021 + }, + { + "epoch": 0.93, + "learning_rate": 1.617920790823461e-05, + "loss": 1.3802, + "step": 11022 + }, + { + "epoch": 0.93, + "learning_rate": 1.617849321141494e-05, + "loss": 1.405, + "step": 11023 + }, + { + "epoch": 0.93, + "learning_rate": 1.6177778463546583e-05, + "loss": 1.3519, + "step": 11024 + }, + { + "epoch": 0.93, + "learning_rate": 1.6177063664635444e-05, + "loss": 1.484, + "step": 11025 + }, + { + "epoch": 0.93, + "learning_rate": 1.617634881468743e-05, + "loss": 1.2051, + "step": 11026 + }, + { + "epoch": 0.93, + "learning_rate": 1.617563391370844e-05, + "loss": 1.2373, + "step": 11027 + }, + { + "epoch": 0.93, + "learning_rate": 1.6174918961704385e-05, + "loss": 1.3232, + "step": 11028 + }, + { + "epoch": 0.93, + "learning_rate": 1.617420395868118e-05, + "loss": 1.355, + "step": 11029 + }, + { + "epoch": 0.93, + "learning_rate": 1.6173488904644724e-05, + "loss": 1.2594, + "step": 11030 + }, + { + "epoch": 0.93, + "learning_rate": 1.6172773799600923e-05, + "loss": 1.1871, + "step": 11031 + }, + { + "epoch": 0.93, + "learning_rate": 1.6172058643555697e-05, + "loss": 1.2701, + "step": 11032 + }, + { + "epoch": 0.93, + "learning_rate": 1.6171343436514943e-05, + "loss": 1.2719, + "step": 11033 + }, + { + "epoch": 0.93, + "learning_rate": 1.6170628178484574e-05, + "loss": 1.1937, + "step": 11034 + }, + { + "epoch": 0.93, + "learning_rate": 1.6169912869470496e-05, + "loss": 1.2256, + "step": 11035 + }, + { + "epoch": 0.93, + "learning_rate": 1.6169197509478628e-05, + "loss": 1.2737, + "step": 11036 + }, + { + "epoch": 0.93, + "learning_rate": 1.6168482098514875e-05, + "loss": 1.2648, + "step": 11037 + }, + { + "epoch": 0.93, + "learning_rate": 1.6167766636585147e-05, + "loss": 1.286, + "step": 11038 + }, + { + "epoch": 0.93, + "learning_rate": 1.616705112369536e-05, + "loss": 1.2959, + "step": 11039 + }, + { + "epoch": 0.93, + "learning_rate": 1.616633555985142e-05, + "loss": 1.2896, + "step": 11040 + }, + { + "epoch": 0.93, + "learning_rate": 1.6165619945059245e-05, + "loss": 1.2288, + "step": 11041 + }, + { + "epoch": 0.93, + "learning_rate": 1.6164904279324743e-05, + "loss": 1.2649, + "step": 11042 + }, + { + "epoch": 0.93, + "learning_rate": 1.616418856265383e-05, + "loss": 1.2697, + "step": 11043 + }, + { + "epoch": 0.93, + "learning_rate": 1.616347279505242e-05, + "loss": 1.2754, + "step": 11044 + }, + { + "epoch": 0.93, + "learning_rate": 1.6162756976526423e-05, + "loss": 1.3732, + "step": 11045 + }, + { + "epoch": 0.93, + "learning_rate": 1.6162041107081754e-05, + "loss": 1.324, + "step": 11046 + }, + { + "epoch": 0.93, + "learning_rate": 1.616132518672433e-05, + "loss": 1.2723, + "step": 11047 + }, + { + "epoch": 0.93, + "learning_rate": 1.6160609215460066e-05, + "loss": 1.3481, + "step": 11048 + }, + { + "epoch": 0.93, + "learning_rate": 1.615989319329488e-05, + "loss": 1.2699, + "step": 11049 + }, + { + "epoch": 0.93, + "learning_rate": 1.615917712023468e-05, + "loss": 1.3302, + "step": 11050 + }, + { + "epoch": 0.93, + "learning_rate": 1.615846099628539e-05, + "loss": 1.2833, + "step": 11051 + }, + { + "epoch": 0.93, + "learning_rate": 1.6157744821452925e-05, + "loss": 1.3213, + "step": 11052 + }, + { + "epoch": 0.93, + "learning_rate": 1.61570285957432e-05, + "loss": 1.3838, + "step": 11053 + }, + { + "epoch": 0.93, + "learning_rate": 1.6156312319162136e-05, + "loss": 1.3232, + "step": 11054 + }, + { + "epoch": 0.93, + "learning_rate": 1.6155595991715646e-05, + "loss": 1.3851, + "step": 11055 + }, + { + "epoch": 0.93, + "learning_rate": 1.6154879613409655e-05, + "loss": 1.3192, + "step": 11056 + }, + { + "epoch": 0.93, + "learning_rate": 1.6154163184250077e-05, + "loss": 1.3425, + "step": 11057 + }, + { + "epoch": 0.93, + "learning_rate": 1.6153446704242833e-05, + "loss": 1.2771, + "step": 11058 + }, + { + "epoch": 0.93, + "learning_rate": 1.6152730173393845e-05, + "loss": 1.2728, + "step": 11059 + }, + { + "epoch": 0.93, + "learning_rate": 1.6152013591709027e-05, + "loss": 1.2852, + "step": 11060 + }, + { + "epoch": 0.93, + "learning_rate": 1.6151296959194307e-05, + "loss": 1.3425, + "step": 11061 + }, + { + "epoch": 0.93, + "learning_rate": 1.6150580275855605e-05, + "loss": 1.3099, + "step": 11062 + }, + { + "epoch": 0.93, + "learning_rate": 1.614986354169884e-05, + "loss": 1.3269, + "step": 11063 + }, + { + "epoch": 0.93, + "learning_rate": 1.614914675672993e-05, + "loss": 1.2826, + "step": 11064 + }, + { + "epoch": 0.93, + "learning_rate": 1.6148429920954806e-05, + "loss": 1.3147, + "step": 11065 + }, + { + "epoch": 0.93, + "learning_rate": 1.6147713034379384e-05, + "loss": 1.3969, + "step": 11066 + }, + { + "epoch": 0.93, + "learning_rate": 1.6146996097009588e-05, + "loss": 1.2465, + "step": 11067 + }, + { + "epoch": 0.93, + "learning_rate": 1.6146279108851344e-05, + "loss": 1.3363, + "step": 11068 + }, + { + "epoch": 0.93, + "learning_rate": 1.6145562069910575e-05, + "loss": 1.2862, + "step": 11069 + }, + { + "epoch": 0.93, + "learning_rate": 1.6144844980193207e-05, + "loss": 1.1577, + "step": 11070 + }, + { + "epoch": 0.93, + "learning_rate": 1.614412783970516e-05, + "loss": 1.3439, + "step": 11071 + }, + { + "epoch": 0.93, + "learning_rate": 1.6143410648452364e-05, + "loss": 1.2827, + "step": 11072 + }, + { + "epoch": 0.93, + "learning_rate": 1.6142693406440745e-05, + "loss": 1.2034, + "step": 11073 + }, + { + "epoch": 0.93, + "learning_rate": 1.6141976113676223e-05, + "loss": 1.3622, + "step": 11074 + }, + { + "epoch": 0.93, + "learning_rate": 1.6141258770164734e-05, + "loss": 1.1887, + "step": 11075 + }, + { + "epoch": 0.93, + "learning_rate": 1.6140541375912193e-05, + "loss": 1.3744, + "step": 11076 + }, + { + "epoch": 0.93, + "learning_rate": 1.6139823930924538e-05, + "loss": 1.2746, + "step": 11077 + }, + { + "epoch": 0.93, + "learning_rate": 1.6139106435207696e-05, + "loss": 1.3126, + "step": 11078 + }, + { + "epoch": 0.93, + "learning_rate": 1.6138388888767587e-05, + "loss": 1.2817, + "step": 11079 + }, + { + "epoch": 0.93, + "learning_rate": 1.6137671291610147e-05, + "loss": 1.2213, + "step": 11080 + }, + { + "epoch": 0.93, + "learning_rate": 1.61369536437413e-05, + "loss": 1.1943, + "step": 11081 + }, + { + "epoch": 0.93, + "learning_rate": 1.6136235945166978e-05, + "loss": 1.3203, + "step": 11082 + }, + { + "epoch": 0.93, + "learning_rate": 1.6135518195893114e-05, + "loss": 1.2919, + "step": 11083 + }, + { + "epoch": 0.93, + "learning_rate": 1.6134800395925632e-05, + "loss": 1.2915, + "step": 11084 + }, + { + "epoch": 0.93, + "learning_rate": 1.6134082545270463e-05, + "loss": 1.2893, + "step": 11085 + }, + { + "epoch": 0.93, + "learning_rate": 1.6133364643933547e-05, + "loss": 1.4005, + "step": 11086 + }, + { + "epoch": 0.93, + "learning_rate": 1.6132646691920806e-05, + "loss": 1.3519, + "step": 11087 + }, + { + "epoch": 0.93, + "learning_rate": 1.6131928689238177e-05, + "loss": 1.2827, + "step": 11088 + }, + { + "epoch": 0.93, + "learning_rate": 1.613121063589159e-05, + "loss": 1.2701, + "step": 11089 + }, + { + "epoch": 0.93, + "learning_rate": 1.613049253188698e-05, + "loss": 1.3576, + "step": 11090 + }, + { + "epoch": 0.93, + "learning_rate": 1.6129774377230275e-05, + "loss": 1.3091, + "step": 11091 + }, + { + "epoch": 0.93, + "learning_rate": 1.6129056171927414e-05, + "loss": 1.2694, + "step": 11092 + }, + { + "epoch": 0.93, + "learning_rate": 1.612833791598433e-05, + "loss": 1.2665, + "step": 11093 + }, + { + "epoch": 0.93, + "learning_rate": 1.6127619609406958e-05, + "loss": 1.1753, + "step": 11094 + }, + { + "epoch": 0.93, + "learning_rate": 1.612690125220123e-05, + "loss": 1.2246, + "step": 11095 + }, + { + "epoch": 0.93, + "learning_rate": 1.6126182844373083e-05, + "loss": 1.3561, + "step": 11096 + }, + { + "epoch": 0.93, + "learning_rate": 1.6125464385928453e-05, + "loss": 1.17, + "step": 11097 + }, + { + "epoch": 0.93, + "learning_rate": 1.6124745876873276e-05, + "loss": 1.2479, + "step": 11098 + }, + { + "epoch": 0.93, + "learning_rate": 1.6124027317213487e-05, + "loss": 1.3935, + "step": 11099 + }, + { + "epoch": 0.93, + "learning_rate": 1.6123308706955027e-05, + "loss": 1.269, + "step": 11100 + }, + { + "epoch": 0.93, + "learning_rate": 1.6122590046103827e-05, + "loss": 1.2873, + "step": 11101 + }, + { + "epoch": 0.93, + "learning_rate": 1.6121871334665834e-05, + "loss": 1.322, + "step": 11102 + }, + { + "epoch": 0.93, + "learning_rate": 1.6121152572646978e-05, + "loss": 1.2464, + "step": 11103 + }, + { + "epoch": 0.93, + "learning_rate": 1.6120433760053197e-05, + "loss": 1.2756, + "step": 11104 + }, + { + "epoch": 0.93, + "learning_rate": 1.6119714896890437e-05, + "loss": 1.3066, + "step": 11105 + }, + { + "epoch": 0.94, + "learning_rate": 1.6118995983164636e-05, + "loss": 1.4306, + "step": 11106 + }, + { + "epoch": 0.94, + "learning_rate": 1.6118277018881728e-05, + "loss": 1.3482, + "step": 11107 + }, + { + "epoch": 0.94, + "learning_rate": 1.611755800404766e-05, + "loss": 1.3513, + "step": 11108 + }, + { + "epoch": 0.94, + "learning_rate": 1.6116838938668365e-05, + "loss": 1.2854, + "step": 11109 + }, + { + "epoch": 0.94, + "learning_rate": 1.6116119822749795e-05, + "loss": 1.2913, + "step": 11110 + }, + { + "epoch": 0.94, + "learning_rate": 1.611540065629788e-05, + "loss": 1.127, + "step": 11111 + }, + { + "epoch": 0.94, + "learning_rate": 1.6114681439318576e-05, + "loss": 1.2969, + "step": 11112 + }, + { + "epoch": 0.94, + "learning_rate": 1.611396217181781e-05, + "loss": 1.2504, + "step": 11113 + }, + { + "epoch": 0.94, + "learning_rate": 1.6113242853801535e-05, + "loss": 1.2192, + "step": 11114 + }, + { + "epoch": 0.94, + "learning_rate": 1.6112523485275687e-05, + "loss": 1.2644, + "step": 11115 + }, + { + "epoch": 0.94, + "learning_rate": 1.6111804066246218e-05, + "loss": 1.2719, + "step": 11116 + }, + { + "epoch": 0.94, + "learning_rate": 1.6111084596719067e-05, + "loss": 1.3378, + "step": 11117 + }, + { + "epoch": 0.94, + "learning_rate": 1.6110365076700177e-05, + "loss": 1.3708, + "step": 11118 + }, + { + "epoch": 0.94, + "learning_rate": 1.6109645506195497e-05, + "loss": 1.338, + "step": 11119 + }, + { + "epoch": 0.94, + "learning_rate": 1.610892588521097e-05, + "loss": 1.3261, + "step": 11120 + }, + { + "epoch": 0.94, + "learning_rate": 1.610820621375254e-05, + "loss": 1.2722, + "step": 11121 + }, + { + "epoch": 0.94, + "learning_rate": 1.610748649182616e-05, + "loss": 1.0903, + "step": 11122 + }, + { + "epoch": 0.94, + "learning_rate": 1.610676671943777e-05, + "loss": 1.3632, + "step": 11123 + }, + { + "epoch": 0.94, + "learning_rate": 1.6106046896593316e-05, + "loss": 1.2344, + "step": 11124 + }, + { + "epoch": 0.94, + "learning_rate": 1.6105327023298753e-05, + "loss": 1.318, + "step": 11125 + }, + { + "epoch": 0.94, + "learning_rate": 1.610460709956002e-05, + "loss": 1.3509, + "step": 11126 + }, + { + "epoch": 0.94, + "learning_rate": 1.6103887125383073e-05, + "loss": 1.3516, + "step": 11127 + }, + { + "epoch": 0.94, + "learning_rate": 1.6103167100773856e-05, + "loss": 1.2616, + "step": 11128 + }, + { + "epoch": 0.94, + "learning_rate": 1.6102447025738318e-05, + "loss": 1.3307, + "step": 11129 + }, + { + "epoch": 0.94, + "learning_rate": 1.6101726900282412e-05, + "loss": 1.2952, + "step": 11130 + }, + { + "epoch": 0.94, + "learning_rate": 1.6101006724412084e-05, + "loss": 1.1906, + "step": 11131 + }, + { + "epoch": 0.94, + "learning_rate": 1.6100286498133284e-05, + "loss": 1.2113, + "step": 11132 + }, + { + "epoch": 0.94, + "learning_rate": 1.6099566221451968e-05, + "loss": 1.2896, + "step": 11133 + }, + { + "epoch": 0.94, + "learning_rate": 1.6098845894374078e-05, + "loss": 1.2985, + "step": 11134 + }, + { + "epoch": 0.94, + "learning_rate": 1.609812551690558e-05, + "loss": 1.1749, + "step": 11135 + }, + { + "epoch": 0.94, + "learning_rate": 1.6097405089052412e-05, + "loss": 1.2685, + "step": 11136 + }, + { + "epoch": 0.94, + "learning_rate": 1.609668461082053e-05, + "loss": 1.3857, + "step": 11137 + }, + { + "epoch": 0.94, + "learning_rate": 1.6095964082215892e-05, + "loss": 1.2803, + "step": 11138 + }, + { + "epoch": 0.94, + "learning_rate": 1.6095243503244448e-05, + "loss": 1.3563, + "step": 11139 + }, + { + "epoch": 0.94, + "learning_rate": 1.6094522873912152e-05, + "loss": 1.2037, + "step": 11140 + }, + { + "epoch": 0.94, + "learning_rate": 1.6093802194224953e-05, + "loss": 1.3091, + "step": 11141 + }, + { + "epoch": 0.94, + "learning_rate": 1.6093081464188815e-05, + "loss": 1.2405, + "step": 11142 + }, + { + "epoch": 0.94, + "learning_rate": 1.6092360683809687e-05, + "loss": 1.2575, + "step": 11143 + }, + { + "epoch": 0.94, + "learning_rate": 1.6091639853093522e-05, + "loss": 1.1832, + "step": 11144 + }, + { + "epoch": 0.94, + "learning_rate": 1.6090918972046284e-05, + "loss": 1.2475, + "step": 11145 + }, + { + "epoch": 0.94, + "learning_rate": 1.6090198040673917e-05, + "loss": 1.2711, + "step": 11146 + }, + { + "epoch": 0.94, + "learning_rate": 1.6089477058982387e-05, + "loss": 1.2142, + "step": 11147 + }, + { + "epoch": 0.94, + "learning_rate": 1.6088756026977654e-05, + "loss": 1.2466, + "step": 11148 + }, + { + "epoch": 0.94, + "learning_rate": 1.6088034944665663e-05, + "loss": 1.2324, + "step": 11149 + }, + { + "epoch": 0.94, + "learning_rate": 1.6087313812052384e-05, + "loss": 1.3577, + "step": 11150 + }, + { + "epoch": 0.94, + "learning_rate": 1.6086592629143768e-05, + "loss": 1.3518, + "step": 11151 + }, + { + "epoch": 0.94, + "learning_rate": 1.6085871395945777e-05, + "loss": 1.2502, + "step": 11152 + }, + { + "epoch": 0.94, + "learning_rate": 1.608515011246437e-05, + "loss": 1.2892, + "step": 11153 + }, + { + "epoch": 0.94, + "learning_rate": 1.6084428778705504e-05, + "loss": 1.1811, + "step": 11154 + }, + { + "epoch": 0.94, + "learning_rate": 1.608370739467514e-05, + "loss": 1.2774, + "step": 11155 + }, + { + "epoch": 0.94, + "learning_rate": 1.6082985960379234e-05, + "loss": 1.3746, + "step": 11156 + }, + { + "epoch": 0.94, + "learning_rate": 1.608226447582376e-05, + "loss": 1.2792, + "step": 11157 + }, + { + "epoch": 0.94, + "learning_rate": 1.608154294101466e-05, + "loss": 1.2258, + "step": 11158 + }, + { + "epoch": 0.94, + "learning_rate": 1.6080821355957913e-05, + "loss": 1.3366, + "step": 11159 + }, + { + "epoch": 0.94, + "learning_rate": 1.6080099720659473e-05, + "loss": 1.3907, + "step": 11160 + }, + { + "epoch": 0.94, + "learning_rate": 1.60793780351253e-05, + "loss": 1.3012, + "step": 11161 + }, + { + "epoch": 0.94, + "learning_rate": 1.6078656299361364e-05, + "loss": 1.186, + "step": 11162 + }, + { + "epoch": 0.94, + "learning_rate": 1.6077934513373622e-05, + "loss": 1.2573, + "step": 11163 + }, + { + "epoch": 0.94, + "learning_rate": 1.607721267716804e-05, + "loss": 1.2938, + "step": 11164 + }, + { + "epoch": 0.94, + "learning_rate": 1.6076490790750586e-05, + "loss": 1.2645, + "step": 11165 + }, + { + "epoch": 0.94, + "learning_rate": 1.6075768854127213e-05, + "loss": 1.293, + "step": 11166 + }, + { + "epoch": 0.94, + "learning_rate": 1.60750468673039e-05, + "loss": 1.2068, + "step": 11167 + }, + { + "epoch": 0.94, + "learning_rate": 1.60743248302866e-05, + "loss": 1.328, + "step": 11168 + }, + { + "epoch": 0.94, + "learning_rate": 1.6073602743081286e-05, + "loss": 1.3139, + "step": 11169 + }, + { + "epoch": 0.94, + "learning_rate": 1.6072880605693926e-05, + "loss": 1.2129, + "step": 11170 + }, + { + "epoch": 0.94, + "learning_rate": 1.607215841813048e-05, + "loss": 1.2673, + "step": 11171 + }, + { + "epoch": 0.94, + "learning_rate": 1.6071436180396915e-05, + "loss": 1.2786, + "step": 11172 + }, + { + "epoch": 0.94, + "learning_rate": 1.6070713892499203e-05, + "loss": 1.2589, + "step": 11173 + }, + { + "epoch": 0.94, + "learning_rate": 1.6069991554443314e-05, + "loss": 1.2899, + "step": 11174 + }, + { + "epoch": 0.94, + "learning_rate": 1.6069269166235207e-05, + "loss": 1.2885, + "step": 11175 + }, + { + "epoch": 0.94, + "learning_rate": 1.606854672788086e-05, + "loss": 1.3231, + "step": 11176 + }, + { + "epoch": 0.94, + "learning_rate": 1.6067824239386234e-05, + "loss": 1.3094, + "step": 11177 + }, + { + "epoch": 0.94, + "learning_rate": 1.6067101700757303e-05, + "loss": 1.2223, + "step": 11178 + }, + { + "epoch": 0.94, + "learning_rate": 1.6066379112000038e-05, + "loss": 1.1799, + "step": 11179 + }, + { + "epoch": 0.94, + "learning_rate": 1.6065656473120404e-05, + "loss": 1.3419, + "step": 11180 + }, + { + "epoch": 0.94, + "learning_rate": 1.606493378412438e-05, + "loss": 1.2824, + "step": 11181 + }, + { + "epoch": 0.94, + "learning_rate": 1.606421104501793e-05, + "loss": 1.3371, + "step": 11182 + }, + { + "epoch": 0.94, + "learning_rate": 1.6063488255807023e-05, + "loss": 1.2669, + "step": 11183 + }, + { + "epoch": 0.94, + "learning_rate": 1.606276541649764e-05, + "loss": 1.1878, + "step": 11184 + }, + { + "epoch": 0.94, + "learning_rate": 1.6062042527095746e-05, + "loss": 1.3071, + "step": 11185 + }, + { + "epoch": 0.94, + "learning_rate": 1.6061319587607318e-05, + "loss": 1.2704, + "step": 11186 + }, + { + "epoch": 0.94, + "learning_rate": 1.606059659803833e-05, + "loss": 1.2293, + "step": 11187 + }, + { + "epoch": 0.94, + "learning_rate": 1.6059873558394752e-05, + "loss": 1.328, + "step": 11188 + }, + { + "epoch": 0.94, + "learning_rate": 1.6059150468682558e-05, + "loss": 1.174, + "step": 11189 + }, + { + "epoch": 0.94, + "learning_rate": 1.6058427328907724e-05, + "loss": 1.1866, + "step": 11190 + }, + { + "epoch": 0.94, + "learning_rate": 1.6057704139076225e-05, + "loss": 1.2473, + "step": 11191 + }, + { + "epoch": 0.94, + "learning_rate": 1.6056980899194033e-05, + "loss": 1.2817, + "step": 11192 + }, + { + "epoch": 0.94, + "learning_rate": 1.6056257609267133e-05, + "loss": 1.2296, + "step": 11193 + }, + { + "epoch": 0.94, + "learning_rate": 1.6055534269301487e-05, + "loss": 1.2824, + "step": 11194 + }, + { + "epoch": 0.94, + "learning_rate": 1.605481087930308e-05, + "loss": 1.2561, + "step": 11195 + }, + { + "epoch": 0.94, + "learning_rate": 1.6054087439277894e-05, + "loss": 1.1972, + "step": 11196 + }, + { + "epoch": 0.94, + "learning_rate": 1.6053363949231894e-05, + "loss": 1.3055, + "step": 11197 + }, + { + "epoch": 0.94, + "learning_rate": 1.6052640409171065e-05, + "loss": 1.4013, + "step": 11198 + }, + { + "epoch": 0.94, + "learning_rate": 1.6051916819101384e-05, + "loss": 1.2173, + "step": 11199 + }, + { + "epoch": 0.94, + "learning_rate": 1.605119317902883e-05, + "loss": 1.2635, + "step": 11200 + }, + { + "epoch": 0.94, + "learning_rate": 1.605046948895938e-05, + "loss": 1.3299, + "step": 11201 + }, + { + "epoch": 0.94, + "learning_rate": 1.6049745748899016e-05, + "loss": 1.2641, + "step": 11202 + }, + { + "epoch": 0.94, + "learning_rate": 1.6049021958853716e-05, + "loss": 1.191, + "step": 11203 + }, + { + "epoch": 0.94, + "learning_rate": 1.604829811882946e-05, + "loss": 1.2603, + "step": 11204 + }, + { + "epoch": 0.94, + "learning_rate": 1.604757422883223e-05, + "loss": 1.2539, + "step": 11205 + }, + { + "epoch": 0.94, + "learning_rate": 1.6046850288868007e-05, + "loss": 1.2959, + "step": 11206 + }, + { + "epoch": 0.94, + "learning_rate": 1.6046126298942772e-05, + "loss": 1.263, + "step": 11207 + }, + { + "epoch": 0.94, + "learning_rate": 1.6045402259062502e-05, + "loss": 1.3048, + "step": 11208 + }, + { + "epoch": 0.94, + "learning_rate": 1.604467816923319e-05, + "loss": 1.2744, + "step": 11209 + }, + { + "epoch": 0.94, + "learning_rate": 1.6043954029460807e-05, + "loss": 1.2092, + "step": 11210 + }, + { + "epoch": 0.94, + "learning_rate": 1.6043229839751346e-05, + "loss": 1.2248, + "step": 11211 + }, + { + "epoch": 0.94, + "learning_rate": 1.6042505600110786e-05, + "loss": 1.2647, + "step": 11212 + }, + { + "epoch": 0.94, + "learning_rate": 1.6041781310545105e-05, + "loss": 1.2171, + "step": 11213 + }, + { + "epoch": 0.94, + "learning_rate": 1.60410569710603e-05, + "loss": 1.2696, + "step": 11214 + }, + { + "epoch": 0.94, + "learning_rate": 1.6040332581662347e-05, + "loss": 1.2605, + "step": 11215 + }, + { + "epoch": 0.94, + "learning_rate": 1.603960814235723e-05, + "loss": 1.3227, + "step": 11216 + }, + { + "epoch": 0.94, + "learning_rate": 1.6038883653150938e-05, + "loss": 1.3207, + "step": 11217 + }, + { + "epoch": 0.94, + "learning_rate": 1.6038159114049462e-05, + "loss": 1.349, + "step": 11218 + }, + { + "epoch": 0.94, + "learning_rate": 1.6037434525058776e-05, + "loss": 1.2779, + "step": 11219 + }, + { + "epoch": 0.94, + "learning_rate": 1.6036709886184877e-05, + "loss": 1.2147, + "step": 11220 + }, + { + "epoch": 0.94, + "learning_rate": 1.603598519743375e-05, + "loss": 1.3102, + "step": 11221 + }, + { + "epoch": 0.94, + "learning_rate": 1.603526045881138e-05, + "loss": 1.2396, + "step": 11222 + }, + { + "epoch": 0.94, + "learning_rate": 1.6034535670323758e-05, + "loss": 1.2338, + "step": 11223 + }, + { + "epoch": 0.94, + "learning_rate": 1.603381083197687e-05, + "loss": 1.2522, + "step": 11224 + }, + { + "epoch": 0.95, + "learning_rate": 1.6033085943776707e-05, + "loss": 1.1742, + "step": 11225 + }, + { + "epoch": 0.95, + "learning_rate": 1.6032361005729252e-05, + "loss": 1.208, + "step": 11226 + }, + { + "epoch": 0.95, + "learning_rate": 1.6031636017840505e-05, + "loss": 1.1878, + "step": 11227 + }, + { + "epoch": 0.95, + "learning_rate": 1.603091098011645e-05, + "loss": 1.2697, + "step": 11228 + }, + { + "epoch": 0.95, + "learning_rate": 1.603018589256308e-05, + "loss": 1.2864, + "step": 11229 + }, + { + "epoch": 0.95, + "learning_rate": 1.602946075518638e-05, + "loss": 1.3076, + "step": 11230 + }, + { + "epoch": 0.95, + "learning_rate": 1.602873556799235e-05, + "loss": 1.3027, + "step": 11231 + }, + { + "epoch": 0.95, + "learning_rate": 1.602801033098697e-05, + "loss": 1.2613, + "step": 11232 + }, + { + "epoch": 0.95, + "learning_rate": 1.6027285044176245e-05, + "loss": 1.28, + "step": 11233 + }, + { + "epoch": 0.95, + "learning_rate": 1.602655970756616e-05, + "loss": 1.1676, + "step": 11234 + }, + { + "epoch": 0.95, + "learning_rate": 1.6025834321162714e-05, + "loss": 1.2401, + "step": 11235 + }, + { + "epoch": 0.95, + "learning_rate": 1.6025108884971894e-05, + "loss": 1.3201, + "step": 11236 + }, + { + "epoch": 0.95, + "learning_rate": 1.6024383398999693e-05, + "loss": 1.2995, + "step": 11237 + }, + { + "epoch": 0.95, + "learning_rate": 1.6023657863252113e-05, + "loss": 1.2196, + "step": 11238 + }, + { + "epoch": 0.95, + "learning_rate": 1.6022932277735144e-05, + "loss": 1.2701, + "step": 11239 + }, + { + "epoch": 0.95, + "learning_rate": 1.602220664245478e-05, + "loss": 1.2853, + "step": 11240 + }, + { + "epoch": 0.95, + "learning_rate": 1.6021480957417015e-05, + "loss": 1.312, + "step": 11241 + }, + { + "epoch": 0.95, + "learning_rate": 1.602075522262785e-05, + "loss": 1.2022, + "step": 11242 + }, + { + "epoch": 0.95, + "learning_rate": 1.6020029438093275e-05, + "loss": 1.2779, + "step": 11243 + }, + { + "epoch": 0.95, + "learning_rate": 1.601930360381929e-05, + "loss": 1.2206, + "step": 11244 + }, + { + "epoch": 0.95, + "learning_rate": 1.6018577719811895e-05, + "loss": 1.2069, + "step": 11245 + }, + { + "epoch": 0.95, + "learning_rate": 1.6017851786077086e-05, + "loss": 1.3032, + "step": 11246 + }, + { + "epoch": 0.95, + "learning_rate": 1.6017125802620856e-05, + "loss": 1.1868, + "step": 11247 + }, + { + "epoch": 0.95, + "learning_rate": 1.601639976944921e-05, + "loss": 1.2961, + "step": 11248 + }, + { + "epoch": 0.95, + "learning_rate": 1.6015673686568142e-05, + "loss": 1.2947, + "step": 11249 + }, + { + "epoch": 0.95, + "learning_rate": 1.601494755398365e-05, + "loss": 1.2151, + "step": 11250 + }, + { + "epoch": 0.95, + "learning_rate": 1.6014221371701738e-05, + "loss": 1.2408, + "step": 11251 + }, + { + "epoch": 0.95, + "learning_rate": 1.6013495139728406e-05, + "loss": 1.2351, + "step": 11252 + }, + { + "epoch": 0.95, + "learning_rate": 1.6012768858069654e-05, + "loss": 1.2899, + "step": 11253 + }, + { + "epoch": 0.95, + "learning_rate": 1.601204252673148e-05, + "loss": 1.2967, + "step": 11254 + }, + { + "epoch": 0.95, + "learning_rate": 1.6011316145719886e-05, + "loss": 1.2781, + "step": 11255 + }, + { + "epoch": 0.95, + "learning_rate": 1.6010589715040872e-05, + "loss": 1.2346, + "step": 11256 + }, + { + "epoch": 0.95, + "learning_rate": 1.6009863234700445e-05, + "loss": 1.3457, + "step": 11257 + }, + { + "epoch": 0.95, + "learning_rate": 1.6009136704704603e-05, + "loss": 1.3586, + "step": 11258 + }, + { + "epoch": 0.95, + "learning_rate": 1.600841012505935e-05, + "loss": 1.2834, + "step": 11259 + }, + { + "epoch": 0.95, + "learning_rate": 1.600768349577069e-05, + "loss": 1.144, + "step": 11260 + }, + { + "epoch": 0.95, + "learning_rate": 1.600695681684463e-05, + "loss": 1.3007, + "step": 11261 + }, + { + "epoch": 0.95, + "learning_rate": 1.6006230088287167e-05, + "loss": 1.2206, + "step": 11262 + }, + { + "epoch": 0.95, + "learning_rate": 1.6005503310104308e-05, + "loss": 1.2846, + "step": 11263 + }, + { + "epoch": 0.95, + "learning_rate": 1.6004776482302062e-05, + "loss": 1.2864, + "step": 11264 + }, + { + "epoch": 0.95, + "learning_rate": 1.6004049604886426e-05, + "loss": 1.324, + "step": 11265 + }, + { + "epoch": 0.95, + "learning_rate": 1.6003322677863412e-05, + "loss": 1.2352, + "step": 11266 + }, + { + "epoch": 0.95, + "learning_rate": 1.600259570123903e-05, + "loss": 1.2564, + "step": 11267 + }, + { + "epoch": 0.95, + "learning_rate": 1.6001868675019274e-05, + "loss": 1.2349, + "step": 11268 + }, + { + "epoch": 0.95, + "learning_rate": 1.6001141599210162e-05, + "loss": 1.1613, + "step": 11269 + }, + { + "epoch": 0.95, + "learning_rate": 1.6000414473817696e-05, + "loss": 1.3179, + "step": 11270 + }, + { + "epoch": 0.95, + "learning_rate": 1.5999687298847887e-05, + "loss": 1.3318, + "step": 11271 + }, + { + "epoch": 0.95, + "learning_rate": 1.599896007430674e-05, + "loss": 1.2261, + "step": 11272 + }, + { + "epoch": 0.95, + "learning_rate": 1.5998232800200265e-05, + "loss": 1.2118, + "step": 11273 + }, + { + "epoch": 0.95, + "learning_rate": 1.599750547653447e-05, + "loss": 1.2096, + "step": 11274 + }, + { + "epoch": 0.95, + "learning_rate": 1.599677810331537e-05, + "loss": 1.3075, + "step": 11275 + }, + { + "epoch": 0.95, + "learning_rate": 1.5996050680548964e-05, + "loss": 1.3234, + "step": 11276 + }, + { + "epoch": 0.95, + "learning_rate": 1.599532320824127e-05, + "loss": 1.2674, + "step": 11277 + }, + { + "epoch": 0.95, + "learning_rate": 1.59945956863983e-05, + "loss": 1.3237, + "step": 11278 + }, + { + "epoch": 0.95, + "learning_rate": 1.599386811502606e-05, + "loss": 1.3837, + "step": 11279 + }, + { + "epoch": 0.95, + "learning_rate": 1.599314049413056e-05, + "loss": 1.2954, + "step": 11280 + }, + { + "epoch": 0.95, + "learning_rate": 1.5992412823717822e-05, + "loss": 1.3455, + "step": 11281 + }, + { + "epoch": 0.95, + "learning_rate": 1.5991685103793846e-05, + "loss": 1.3375, + "step": 11282 + }, + { + "epoch": 0.95, + "learning_rate": 1.5990957334364652e-05, + "loss": 1.3473, + "step": 11283 + }, + { + "epoch": 0.95, + "learning_rate": 1.599022951543625e-05, + "loss": 1.2697, + "step": 11284 + }, + { + "epoch": 0.95, + "learning_rate": 1.5989501647014654e-05, + "loss": 1.3575, + "step": 11285 + }, + { + "epoch": 0.95, + "learning_rate": 1.598877372910588e-05, + "loss": 1.2186, + "step": 11286 + }, + { + "epoch": 0.95, + "learning_rate": 1.5988045761715943e-05, + "loss": 1.3034, + "step": 11287 + }, + { + "epoch": 0.95, + "learning_rate": 1.5987317744850855e-05, + "loss": 1.18, + "step": 11288 + }, + { + "epoch": 0.95, + "learning_rate": 1.5986589678516627e-05, + "loss": 1.3277, + "step": 11289 + }, + { + "epoch": 0.95, + "learning_rate": 1.5985861562719283e-05, + "loss": 1.2356, + "step": 11290 + }, + { + "epoch": 0.95, + "learning_rate": 1.5985133397464834e-05, + "loss": 1.2553, + "step": 11291 + }, + { + "epoch": 0.95, + "learning_rate": 1.59844051827593e-05, + "loss": 1.2265, + "step": 11292 + }, + { + "epoch": 0.95, + "learning_rate": 1.5983676918608693e-05, + "loss": 1.3574, + "step": 11293 + }, + { + "epoch": 0.95, + "learning_rate": 1.598294860501903e-05, + "loss": 1.2789, + "step": 11294 + }, + { + "epoch": 0.95, + "learning_rate": 1.5982220241996333e-05, + "loss": 1.4166, + "step": 11295 + }, + { + "epoch": 0.95, + "learning_rate": 1.5981491829546618e-05, + "loss": 1.2684, + "step": 11296 + }, + { + "epoch": 0.95, + "learning_rate": 1.59807633676759e-05, + "loss": 1.29, + "step": 11297 + }, + { + "epoch": 0.95, + "learning_rate": 1.5980034856390204e-05, + "loss": 1.2959, + "step": 11298 + }, + { + "epoch": 0.95, + "learning_rate": 1.5979306295695547e-05, + "loss": 1.2748, + "step": 11299 + }, + { + "epoch": 0.95, + "learning_rate": 1.5978577685597948e-05, + "loss": 1.2096, + "step": 11300 + }, + { + "epoch": 0.95, + "learning_rate": 1.5977849026103423e-05, + "loss": 1.2407, + "step": 11301 + }, + { + "epoch": 0.95, + "learning_rate": 1.5977120317218e-05, + "loss": 1.2823, + "step": 11302 + }, + { + "epoch": 0.95, + "learning_rate": 1.5976391558947698e-05, + "loss": 1.2279, + "step": 11303 + }, + { + "epoch": 0.95, + "learning_rate": 1.5975662751298532e-05, + "loss": 1.2881, + "step": 11304 + }, + { + "epoch": 0.95, + "learning_rate": 1.5974933894276527e-05, + "loss": 1.3613, + "step": 11305 + }, + { + "epoch": 0.95, + "learning_rate": 1.597420498788771e-05, + "loss": 1.3492, + "step": 11306 + }, + { + "epoch": 0.95, + "learning_rate": 1.5973476032138097e-05, + "loss": 1.1899, + "step": 11307 + }, + { + "epoch": 0.95, + "learning_rate": 1.5972747027033713e-05, + "loss": 1.3007, + "step": 11308 + }, + { + "epoch": 0.95, + "learning_rate": 1.597201797258058e-05, + "loss": 1.325, + "step": 11309 + }, + { + "epoch": 0.95, + "learning_rate": 1.5971288868784724e-05, + "loss": 1.2019, + "step": 11310 + }, + { + "epoch": 0.95, + "learning_rate": 1.597055971565217e-05, + "loss": 1.2183, + "step": 11311 + }, + { + "epoch": 0.95, + "learning_rate": 1.596983051318894e-05, + "loss": 1.3091, + "step": 11312 + }, + { + "epoch": 0.95, + "learning_rate": 1.596910126140106e-05, + "loss": 1.2585, + "step": 11313 + }, + { + "epoch": 0.95, + "learning_rate": 1.5968371960294556e-05, + "loss": 1.2437, + "step": 11314 + }, + { + "epoch": 0.95, + "learning_rate": 1.5967642609875453e-05, + "loss": 1.1705, + "step": 11315 + }, + { + "epoch": 0.95, + "learning_rate": 1.5966913210149776e-05, + "loss": 1.3066, + "step": 11316 + }, + { + "epoch": 0.95, + "learning_rate": 1.5966183761123552e-05, + "loss": 1.2663, + "step": 11317 + }, + { + "epoch": 0.95, + "learning_rate": 1.596545426280281e-05, + "loss": 1.2691, + "step": 11318 + }, + { + "epoch": 0.95, + "learning_rate": 1.5964724715193572e-05, + "loss": 1.2285, + "step": 11319 + }, + { + "epoch": 0.95, + "learning_rate": 1.596399511830187e-05, + "loss": 1.3384, + "step": 11320 + }, + { + "epoch": 0.95, + "learning_rate": 1.5963265472133733e-05, + "loss": 1.3659, + "step": 11321 + }, + { + "epoch": 0.95, + "learning_rate": 1.596253577669519e-05, + "loss": 1.3048, + "step": 11322 + }, + { + "epoch": 0.95, + "learning_rate": 1.5961806031992265e-05, + "loss": 1.2409, + "step": 11323 + }, + { + "epoch": 0.95, + "learning_rate": 1.5961076238030993e-05, + "loss": 1.1862, + "step": 11324 + }, + { + "epoch": 0.95, + "learning_rate": 1.59603463948174e-05, + "loss": 1.333, + "step": 11325 + }, + { + "epoch": 0.95, + "learning_rate": 1.5959616502357517e-05, + "loss": 1.3201, + "step": 11326 + }, + { + "epoch": 0.95, + "learning_rate": 1.5958886560657378e-05, + "loss": 1.3493, + "step": 11327 + }, + { + "epoch": 0.95, + "learning_rate": 1.5958156569723006e-05, + "loss": 1.2481, + "step": 11328 + }, + { + "epoch": 0.95, + "learning_rate": 1.5957426529560443e-05, + "loss": 1.184, + "step": 11329 + }, + { + "epoch": 0.95, + "learning_rate": 1.5956696440175713e-05, + "loss": 1.26, + "step": 11330 + }, + { + "epoch": 0.95, + "learning_rate": 1.595596630157485e-05, + "loss": 1.3225, + "step": 11331 + }, + { + "epoch": 0.95, + "learning_rate": 1.595523611376389e-05, + "loss": 1.3342, + "step": 11332 + }, + { + "epoch": 0.95, + "learning_rate": 1.595450587674886e-05, + "loss": 1.2629, + "step": 11333 + }, + { + "epoch": 0.95, + "learning_rate": 1.5953775590535797e-05, + "loss": 1.3019, + "step": 11334 + }, + { + "epoch": 0.95, + "learning_rate": 1.5953045255130736e-05, + "loss": 1.3076, + "step": 11335 + }, + { + "epoch": 0.95, + "learning_rate": 1.595231487053971e-05, + "loss": 1.2833, + "step": 11336 + }, + { + "epoch": 0.95, + "learning_rate": 1.5951584436768752e-05, + "loss": 1.3081, + "step": 11337 + }, + { + "epoch": 0.95, + "learning_rate": 1.59508539538239e-05, + "loss": 1.2588, + "step": 11338 + }, + { + "epoch": 0.95, + "learning_rate": 1.5950123421711187e-05, + "loss": 1.2189, + "step": 11339 + }, + { + "epoch": 0.95, + "learning_rate": 1.5949392840436653e-05, + "loss": 1.3116, + "step": 11340 + }, + { + "epoch": 0.95, + "learning_rate": 1.594866221000633e-05, + "loss": 1.2858, + "step": 11341 + }, + { + "epoch": 0.95, + "learning_rate": 1.5947931530426253e-05, + "loss": 1.2789, + "step": 11342 + }, + { + "epoch": 0.95, + "learning_rate": 1.5947200801702467e-05, + "loss": 1.1922, + "step": 11343 + }, + { + "epoch": 0.96, + "learning_rate": 1.5946470023841e-05, + "loss": 1.2379, + "step": 11344 + }, + { + "epoch": 0.96, + "learning_rate": 1.5945739196847898e-05, + "loss": 1.2911, + "step": 11345 + }, + { + "epoch": 0.96, + "learning_rate": 1.594500832072919e-05, + "loss": 1.3559, + "step": 11346 + }, + { + "epoch": 0.96, + "learning_rate": 1.5944277395490928e-05, + "loss": 1.303, + "step": 11347 + }, + { + "epoch": 0.96, + "learning_rate": 1.594354642113914e-05, + "loss": 1.1259, + "step": 11348 + }, + { + "epoch": 0.96, + "learning_rate": 1.594281539767987e-05, + "loss": 1.2556, + "step": 11349 + }, + { + "epoch": 0.96, + "learning_rate": 1.5942084325119157e-05, + "loss": 1.2978, + "step": 11350 + }, + { + "epoch": 0.96, + "learning_rate": 1.5941353203463044e-05, + "loss": 1.3475, + "step": 11351 + }, + { + "epoch": 0.96, + "learning_rate": 1.5940622032717564e-05, + "loss": 1.2349, + "step": 11352 + }, + { + "epoch": 0.96, + "learning_rate": 1.5939890812888767e-05, + "loss": 1.3198, + "step": 11353 + }, + { + "epoch": 0.96, + "learning_rate": 1.5939159543982695e-05, + "loss": 1.2639, + "step": 11354 + }, + { + "epoch": 0.96, + "learning_rate": 1.593842822600538e-05, + "loss": 1.2625, + "step": 11355 + }, + { + "epoch": 0.96, + "learning_rate": 1.593769685896287e-05, + "loss": 1.2496, + "step": 11356 + }, + { + "epoch": 0.96, + "learning_rate": 1.5936965442861213e-05, + "loss": 1.3026, + "step": 11357 + }, + { + "epoch": 0.96, + "learning_rate": 1.5936233977706447e-05, + "loss": 1.2301, + "step": 11358 + }, + { + "epoch": 0.96, + "learning_rate": 1.593550246350461e-05, + "loss": 1.2949, + "step": 11359 + }, + { + "epoch": 0.96, + "learning_rate": 1.5934770900261755e-05, + "loss": 1.2238, + "step": 11360 + }, + { + "epoch": 0.96, + "learning_rate": 1.5934039287983925e-05, + "loss": 1.2174, + "step": 11361 + }, + { + "epoch": 0.96, + "learning_rate": 1.5933307626677165e-05, + "loss": 1.1997, + "step": 11362 + }, + { + "epoch": 0.96, + "learning_rate": 1.5932575916347516e-05, + "loss": 1.3236, + "step": 11363 + }, + { + "epoch": 0.96, + "learning_rate": 1.5931844157001025e-05, + "loss": 1.2481, + "step": 11364 + }, + { + "epoch": 0.96, + "learning_rate": 1.5931112348643742e-05, + "loss": 1.2471, + "step": 11365 + }, + { + "epoch": 0.96, + "learning_rate": 1.593038049128171e-05, + "loss": 1.2826, + "step": 11366 + }, + { + "epoch": 0.96, + "learning_rate": 1.5929648584920972e-05, + "loss": 1.233, + "step": 11367 + }, + { + "epoch": 0.96, + "learning_rate": 1.5928916629567585e-05, + "loss": 1.2259, + "step": 11368 + }, + { + "epoch": 0.96, + "learning_rate": 1.5928184625227593e-05, + "loss": 1.248, + "step": 11369 + }, + { + "epoch": 0.96, + "learning_rate": 1.5927452571907035e-05, + "loss": 1.3281, + "step": 11370 + }, + { + "epoch": 0.96, + "learning_rate": 1.5926720469611972e-05, + "loss": 1.2834, + "step": 11371 + }, + { + "epoch": 0.96, + "learning_rate": 1.5925988318348443e-05, + "loss": 1.2298, + "step": 11372 + }, + { + "epoch": 0.96, + "learning_rate": 1.5925256118122506e-05, + "loss": 1.2031, + "step": 11373 + }, + { + "epoch": 0.96, + "learning_rate": 1.5924523868940207e-05, + "loss": 1.329, + "step": 11374 + }, + { + "epoch": 0.96, + "learning_rate": 1.5923791570807596e-05, + "loss": 1.3252, + "step": 11375 + }, + { + "epoch": 0.96, + "learning_rate": 1.592305922373072e-05, + "loss": 1.2361, + "step": 11376 + }, + { + "epoch": 0.96, + "learning_rate": 1.5922326827715635e-05, + "loss": 1.2839, + "step": 11377 + }, + { + "epoch": 0.96, + "learning_rate": 1.5921594382768392e-05, + "loss": 1.2482, + "step": 11378 + }, + { + "epoch": 0.96, + "learning_rate": 1.592086188889504e-05, + "loss": 1.2585, + "step": 11379 + }, + { + "epoch": 0.96, + "learning_rate": 1.5920129346101633e-05, + "loss": 1.2799, + "step": 11380 + }, + { + "epoch": 0.96, + "learning_rate": 1.591939675439422e-05, + "loss": 1.2963, + "step": 11381 + }, + { + "epoch": 0.96, + "learning_rate": 1.5918664113778858e-05, + "loss": 1.2813, + "step": 11382 + }, + { + "epoch": 0.96, + "learning_rate": 1.59179314242616e-05, + "loss": 1.245, + "step": 11383 + }, + { + "epoch": 0.96, + "learning_rate": 1.5917198685848495e-05, + "loss": 1.2927, + "step": 11384 + }, + { + "epoch": 0.96, + "learning_rate": 1.59164658985456e-05, + "loss": 1.3598, + "step": 11385 + }, + { + "epoch": 0.96, + "learning_rate": 1.5915733062358977e-05, + "loss": 1.3005, + "step": 11386 + }, + { + "epoch": 0.96, + "learning_rate": 1.5915000177294668e-05, + "loss": 1.1948, + "step": 11387 + }, + { + "epoch": 0.96, + "learning_rate": 1.591426724335874e-05, + "loss": 1.2456, + "step": 11388 + }, + { + "epoch": 0.96, + "learning_rate": 1.5913534260557234e-05, + "loss": 1.2001, + "step": 11389 + }, + { + "epoch": 0.96, + "learning_rate": 1.5912801228896225e-05, + "loss": 1.1709, + "step": 11390 + }, + { + "epoch": 0.96, + "learning_rate": 1.5912068148381757e-05, + "loss": 1.2659, + "step": 11391 + }, + { + "epoch": 0.96, + "learning_rate": 1.5911335019019888e-05, + "loss": 1.2627, + "step": 11392 + }, + { + "epoch": 0.96, + "learning_rate": 1.591060184081668e-05, + "loss": 1.278, + "step": 11393 + }, + { + "epoch": 0.96, + "learning_rate": 1.5909868613778184e-05, + "loss": 1.2255, + "step": 11394 + }, + { + "epoch": 0.96, + "learning_rate": 1.5909135337910468e-05, + "loss": 1.3154, + "step": 11395 + }, + { + "epoch": 0.96, + "learning_rate": 1.590840201321958e-05, + "loss": 1.337, + "step": 11396 + }, + { + "epoch": 0.96, + "learning_rate": 1.5907668639711584e-05, + "loss": 1.2814, + "step": 11397 + }, + { + "epoch": 0.96, + "learning_rate": 1.5906935217392538e-05, + "loss": 1.2766, + "step": 11398 + }, + { + "epoch": 0.96, + "learning_rate": 1.5906201746268505e-05, + "loss": 1.2388, + "step": 11399 + }, + { + "epoch": 0.96, + "learning_rate": 1.590546822634554e-05, + "loss": 1.3486, + "step": 11400 + }, + { + "epoch": 0.96, + "learning_rate": 1.5904734657629712e-05, + "loss": 1.2777, + "step": 11401 + }, + { + "epoch": 0.96, + "learning_rate": 1.5904001040127073e-05, + "loss": 1.3184, + "step": 11402 + }, + { + "epoch": 0.96, + "learning_rate": 1.590326737384369e-05, + "loss": 1.2706, + "step": 11403 + }, + { + "epoch": 0.96, + "learning_rate": 1.590253365878562e-05, + "loss": 1.3034, + "step": 11404 + }, + { + "epoch": 0.96, + "learning_rate": 1.590179989495893e-05, + "loss": 1.2741, + "step": 11405 + }, + { + "epoch": 0.96, + "learning_rate": 1.5901066082369678e-05, + "loss": 1.2268, + "step": 11406 + }, + { + "epoch": 0.96, + "learning_rate": 1.5900332221023932e-05, + "loss": 1.302, + "step": 11407 + }, + { + "epoch": 0.96, + "learning_rate": 1.589959831092775e-05, + "loss": 1.3546, + "step": 11408 + }, + { + "epoch": 0.96, + "learning_rate": 1.58988643520872e-05, + "loss": 1.2299, + "step": 11409 + }, + { + "epoch": 0.96, + "learning_rate": 1.5898130344508345e-05, + "loss": 1.3761, + "step": 11410 + }, + { + "epoch": 0.96, + "learning_rate": 1.589739628819725e-05, + "loss": 1.3397, + "step": 11411 + }, + { + "epoch": 0.96, + "learning_rate": 1.5896662183159978e-05, + "loss": 1.2613, + "step": 11412 + }, + { + "epoch": 0.96, + "learning_rate": 1.58959280294026e-05, + "loss": 1.2579, + "step": 11413 + }, + { + "epoch": 0.96, + "learning_rate": 1.5895193826931172e-05, + "loss": 1.3221, + "step": 11414 + }, + { + "epoch": 0.96, + "learning_rate": 1.5894459575751767e-05, + "loss": 1.2237, + "step": 11415 + }, + { + "epoch": 0.96, + "learning_rate": 1.5893725275870454e-05, + "loss": 1.204, + "step": 11416 + }, + { + "epoch": 0.96, + "learning_rate": 1.5892990927293294e-05, + "loss": 1.2336, + "step": 11417 + }, + { + "epoch": 0.96, + "learning_rate": 1.589225653002636e-05, + "loss": 1.1618, + "step": 11418 + }, + { + "epoch": 0.96, + "learning_rate": 1.5891522084075713e-05, + "loss": 1.2223, + "step": 11419 + }, + { + "epoch": 0.96, + "learning_rate": 1.5890787589447424e-05, + "loss": 1.1595, + "step": 11420 + }, + { + "epoch": 0.96, + "learning_rate": 1.5890053046147568e-05, + "loss": 1.2816, + "step": 11421 + }, + { + "epoch": 0.96, + "learning_rate": 1.5889318454182203e-05, + "loss": 1.2713, + "step": 11422 + }, + { + "epoch": 0.96, + "learning_rate": 1.5888583813557408e-05, + "loss": 1.2959, + "step": 11423 + }, + { + "epoch": 0.96, + "learning_rate": 1.5887849124279247e-05, + "loss": 1.2858, + "step": 11424 + }, + { + "epoch": 0.96, + "learning_rate": 1.588711438635379e-05, + "loss": 1.2403, + "step": 11425 + }, + { + "epoch": 0.96, + "learning_rate": 1.5886379599787117e-05, + "loss": 1.1934, + "step": 11426 + }, + { + "epoch": 0.96, + "learning_rate": 1.5885644764585287e-05, + "loss": 1.2721, + "step": 11427 + }, + { + "epoch": 0.96, + "learning_rate": 1.5884909880754375e-05, + "loss": 1.3121, + "step": 11428 + }, + { + "epoch": 0.96, + "learning_rate": 1.5884174948300457e-05, + "loss": 1.1718, + "step": 11429 + }, + { + "epoch": 0.96, + "learning_rate": 1.58834399672296e-05, + "loss": 1.2222, + "step": 11430 + }, + { + "epoch": 0.96, + "learning_rate": 1.5882704937547885e-05, + "loss": 1.3012, + "step": 11431 + }, + { + "epoch": 0.96, + "learning_rate": 1.5881969859261372e-05, + "loss": 1.1915, + "step": 11432 + }, + { + "epoch": 0.96, + "learning_rate": 1.5881234732376144e-05, + "loss": 1.2829, + "step": 11433 + }, + { + "epoch": 0.96, + "learning_rate": 1.5880499556898273e-05, + "loss": 1.2318, + "step": 11434 + }, + { + "epoch": 0.96, + "learning_rate": 1.587976433283383e-05, + "loss": 1.2559, + "step": 11435 + }, + { + "epoch": 0.96, + "learning_rate": 1.5879029060188895e-05, + "loss": 1.2951, + "step": 11436 + }, + { + "epoch": 0.96, + "learning_rate": 1.587829373896954e-05, + "loss": 1.2405, + "step": 11437 + }, + { + "epoch": 0.96, + "learning_rate": 1.587755836918184e-05, + "loss": 1.1613, + "step": 11438 + }, + { + "epoch": 0.96, + "learning_rate": 1.5876822950831875e-05, + "loss": 1.3151, + "step": 11439 + }, + { + "epoch": 0.96, + "learning_rate": 1.587608748392571e-05, + "loss": 1.2452, + "step": 11440 + }, + { + "epoch": 0.96, + "learning_rate": 1.5875351968469438e-05, + "loss": 1.3006, + "step": 11441 + }, + { + "epoch": 0.96, + "learning_rate": 1.5874616404469124e-05, + "loss": 1.2295, + "step": 11442 + }, + { + "epoch": 0.96, + "learning_rate": 1.587388079193085e-05, + "loss": 1.3277, + "step": 11443 + }, + { + "epoch": 0.96, + "learning_rate": 1.5873145130860694e-05, + "loss": 1.2985, + "step": 11444 + }, + { + "epoch": 0.96, + "learning_rate": 1.5872409421264733e-05, + "loss": 1.2612, + "step": 11445 + }, + { + "epoch": 0.96, + "learning_rate": 1.5871673663149042e-05, + "loss": 1.2905, + "step": 11446 + }, + { + "epoch": 0.96, + "learning_rate": 1.5870937856519708e-05, + "loss": 1.2537, + "step": 11447 + }, + { + "epoch": 0.96, + "learning_rate": 1.5870202001382806e-05, + "loss": 1.3107, + "step": 11448 + }, + { + "epoch": 0.96, + "learning_rate": 1.5869466097744415e-05, + "loss": 1.3066, + "step": 11449 + }, + { + "epoch": 0.96, + "learning_rate": 1.5868730145610616e-05, + "loss": 1.1802, + "step": 11450 + }, + { + "epoch": 0.96, + "learning_rate": 1.586799414498749e-05, + "loss": 1.2934, + "step": 11451 + }, + { + "epoch": 0.96, + "learning_rate": 1.5867258095881122e-05, + "loss": 1.2653, + "step": 11452 + }, + { + "epoch": 0.96, + "learning_rate": 1.586652199829759e-05, + "loss": 1.2539, + "step": 11453 + }, + { + "epoch": 0.96, + "learning_rate": 1.5865785852242974e-05, + "loss": 1.2831, + "step": 11454 + }, + { + "epoch": 0.96, + "learning_rate": 1.5865049657723356e-05, + "loss": 1.296, + "step": 11455 + }, + { + "epoch": 0.96, + "learning_rate": 1.5864313414744826e-05, + "loss": 1.2789, + "step": 11456 + }, + { + "epoch": 0.96, + "learning_rate": 1.5863577123313457e-05, + "loss": 1.2339, + "step": 11457 + }, + { + "epoch": 0.96, + "learning_rate": 1.5862840783435338e-05, + "loss": 1.2452, + "step": 11458 + }, + { + "epoch": 0.96, + "learning_rate": 1.586210439511655e-05, + "loss": 1.2126, + "step": 11459 + }, + { + "epoch": 0.96, + "learning_rate": 1.5861367958363183e-05, + "loss": 1.3611, + "step": 11460 + }, + { + "epoch": 0.96, + "learning_rate": 1.586063147318132e-05, + "loss": 1.2299, + "step": 11461 + }, + { + "epoch": 0.97, + "learning_rate": 1.5859894939577037e-05, + "loss": 1.2444, + "step": 11462 + }, + { + "epoch": 0.97, + "learning_rate": 1.5859158357556433e-05, + "loss": 1.3388, + "step": 11463 + }, + { + "epoch": 0.97, + "learning_rate": 1.5858421727125586e-05, + "loss": 1.2846, + "step": 11464 + }, + { + "epoch": 0.97, + "learning_rate": 1.585768504829058e-05, + "loss": 1.2718, + "step": 11465 + }, + { + "epoch": 0.97, + "learning_rate": 1.5856948321057507e-05, + "loss": 1.2678, + "step": 11466 + }, + { + "epoch": 0.97, + "learning_rate": 1.5856211545432456e-05, + "loss": 1.1591, + "step": 11467 + }, + { + "epoch": 0.97, + "learning_rate": 1.585547472142151e-05, + "loss": 1.1313, + "step": 11468 + }, + { + "epoch": 0.97, + "learning_rate": 1.5854737849030756e-05, + "loss": 1.318, + "step": 11469 + }, + { + "epoch": 0.97, + "learning_rate": 1.5854000928266288e-05, + "loss": 1.2592, + "step": 11470 + }, + { + "epoch": 0.97, + "learning_rate": 1.5853263959134185e-05, + "loss": 1.2591, + "step": 11471 + }, + { + "epoch": 0.97, + "learning_rate": 1.5852526941640546e-05, + "loss": 1.2988, + "step": 11472 + }, + { + "epoch": 0.97, + "learning_rate": 1.5851789875791453e-05, + "loss": 1.3393, + "step": 11473 + }, + { + "epoch": 0.97, + "learning_rate": 1.5851052761592998e-05, + "loss": 1.2347, + "step": 11474 + }, + { + "epoch": 0.97, + "learning_rate": 1.585031559905128e-05, + "loss": 1.3249, + "step": 11475 + }, + { + "epoch": 0.97, + "learning_rate": 1.584957838817238e-05, + "loss": 1.2693, + "step": 11476 + }, + { + "epoch": 0.97, + "learning_rate": 1.5848841128962384e-05, + "loss": 1.2199, + "step": 11477 + }, + { + "epoch": 0.97, + "learning_rate": 1.58481038214274e-05, + "loss": 1.4223, + "step": 11478 + }, + { + "epoch": 0.97, + "learning_rate": 1.5847366465573503e-05, + "loss": 1.2421, + "step": 11479 + }, + { + "epoch": 0.97, + "learning_rate": 1.5846629061406796e-05, + "loss": 1.3038, + "step": 11480 + }, + { + "epoch": 0.97, + "learning_rate": 1.584589160893337e-05, + "loss": 1.3036, + "step": 11481 + }, + { + "epoch": 0.97, + "learning_rate": 1.5845154108159314e-05, + "loss": 1.1658, + "step": 11482 + }, + { + "epoch": 0.97, + "learning_rate": 1.5844416559090728e-05, + "loss": 1.3019, + "step": 11483 + }, + { + "epoch": 0.97, + "learning_rate": 1.58436789617337e-05, + "loss": 1.2911, + "step": 11484 + }, + { + "epoch": 0.97, + "learning_rate": 1.584294131609432e-05, + "loss": 1.2588, + "step": 11485 + }, + { + "epoch": 0.97, + "learning_rate": 1.5842203622178697e-05, + "loss": 1.2779, + "step": 11486 + }, + { + "epoch": 0.97, + "learning_rate": 1.5841465879992915e-05, + "loss": 1.2681, + "step": 11487 + }, + { + "epoch": 0.97, + "learning_rate": 1.5840728089543074e-05, + "loss": 1.2613, + "step": 11488 + }, + { + "epoch": 0.97, + "learning_rate": 1.5839990250835265e-05, + "loss": 1.2423, + "step": 11489 + }, + { + "epoch": 0.97, + "learning_rate": 1.583925236387559e-05, + "loss": 1.2677, + "step": 11490 + }, + { + "epoch": 0.97, + "learning_rate": 1.5838514428670143e-05, + "loss": 1.3604, + "step": 11491 + }, + { + "epoch": 0.97, + "learning_rate": 1.583777644522502e-05, + "loss": 1.2391, + "step": 11492 + }, + { + "epoch": 0.97, + "learning_rate": 1.583703841354632e-05, + "loss": 1.3019, + "step": 11493 + }, + { + "epoch": 0.97, + "learning_rate": 1.583630033364014e-05, + "loss": 1.1631, + "step": 11494 + }, + { + "epoch": 0.97, + "learning_rate": 1.5835562205512586e-05, + "loss": 1.2774, + "step": 11495 + }, + { + "epoch": 0.97, + "learning_rate": 1.5834824029169743e-05, + "loss": 1.2202, + "step": 11496 + }, + { + "epoch": 0.97, + "learning_rate": 1.5834085804617715e-05, + "loss": 1.2517, + "step": 11497 + }, + { + "epoch": 0.97, + "learning_rate": 1.5833347531862604e-05, + "loss": 1.2538, + "step": 11498 + }, + { + "epoch": 0.97, + "learning_rate": 1.583260921091051e-05, + "loss": 1.2684, + "step": 11499 + }, + { + "epoch": 0.97, + "learning_rate": 1.583187084176753e-05, + "loss": 1.3355, + "step": 11500 + }, + { + "epoch": 0.97, + "learning_rate": 1.583113242443977e-05, + "loss": 1.2249, + "step": 11501 + }, + { + "epoch": 0.97, + "learning_rate": 1.5830393958933326e-05, + "loss": 1.2967, + "step": 11502 + }, + { + "epoch": 0.97, + "learning_rate": 1.58296554452543e-05, + "loss": 1.4088, + "step": 11503 + }, + { + "epoch": 0.97, + "learning_rate": 1.5828916883408795e-05, + "loss": 1.242, + "step": 11504 + }, + { + "epoch": 0.97, + "learning_rate": 1.5828178273402915e-05, + "loss": 1.183, + "step": 11505 + }, + { + "epoch": 0.97, + "learning_rate": 1.582743961524276e-05, + "loss": 1.3341, + "step": 11506 + }, + { + "epoch": 0.97, + "learning_rate": 1.5826700908934432e-05, + "loss": 1.2489, + "step": 11507 + }, + { + "epoch": 0.97, + "learning_rate": 1.5825962154484044e-05, + "loss": 1.2229, + "step": 11508 + }, + { + "epoch": 0.97, + "learning_rate": 1.582522335189768e-05, + "loss": 1.2113, + "step": 11509 + }, + { + "epoch": 0.97, + "learning_rate": 1.5824484501181464e-05, + "loss": 1.2574, + "step": 11510 + }, + { + "epoch": 0.97, + "learning_rate": 1.5823745602341492e-05, + "loss": 1.2262, + "step": 11511 + }, + { + "epoch": 0.97, + "learning_rate": 1.5823006655383873e-05, + "loss": 1.2407, + "step": 11512 + }, + { + "epoch": 0.97, + "learning_rate": 1.5822267660314704e-05, + "loss": 1.263, + "step": 11513 + }, + { + "epoch": 0.97, + "learning_rate": 1.58215286171401e-05, + "loss": 1.358, + "step": 11514 + }, + { + "epoch": 0.97, + "learning_rate": 1.582078952586616e-05, + "loss": 1.2432, + "step": 11515 + }, + { + "epoch": 0.97, + "learning_rate": 1.5820050386498994e-05, + "loss": 1.3417, + "step": 11516 + }, + { + "epoch": 0.97, + "learning_rate": 1.5819311199044712e-05, + "loss": 1.3531, + "step": 11517 + }, + { + "epoch": 0.97, + "learning_rate": 1.581857196350942e-05, + "loss": 1.3671, + "step": 11518 + }, + { + "epoch": 0.97, + "learning_rate": 1.5817832679899222e-05, + "loss": 1.2528, + "step": 11519 + }, + { + "epoch": 0.97, + "learning_rate": 1.5817093348220227e-05, + "loss": 1.2529, + "step": 11520 + }, + { + "epoch": 0.97, + "learning_rate": 1.5816353968478547e-05, + "loss": 1.2934, + "step": 11521 + }, + { + "epoch": 0.97, + "learning_rate": 1.581561454068029e-05, + "loss": 1.2939, + "step": 11522 + }, + { + "epoch": 0.97, + "learning_rate": 1.581487506483156e-05, + "loss": 1.3225, + "step": 11523 + }, + { + "epoch": 0.97, + "learning_rate": 1.5814135540938473e-05, + "loss": 1.1917, + "step": 11524 + }, + { + "epoch": 0.97, + "learning_rate": 1.5813395969007138e-05, + "loss": 1.3174, + "step": 11525 + }, + { + "epoch": 0.97, + "learning_rate": 1.5812656349043668e-05, + "loss": 1.2276, + "step": 11526 + }, + { + "epoch": 0.97, + "learning_rate": 1.5811916681054165e-05, + "loss": 1.3321, + "step": 11527 + }, + { + "epoch": 0.97, + "learning_rate": 1.581117696504475e-05, + "loss": 1.2101, + "step": 11528 + }, + { + "epoch": 0.97, + "learning_rate": 1.5810437201021533e-05, + "loss": 1.2595, + "step": 11529 + }, + { + "epoch": 0.97, + "learning_rate": 1.5809697388990622e-05, + "loss": 1.2628, + "step": 11530 + }, + { + "epoch": 0.97, + "learning_rate": 1.580895752895813e-05, + "loss": 1.2783, + "step": 11531 + }, + { + "epoch": 0.97, + "learning_rate": 1.5808217620930176e-05, + "loss": 1.3235, + "step": 11532 + }, + { + "epoch": 0.97, + "learning_rate": 1.5807477664912865e-05, + "loss": 1.2527, + "step": 11533 + }, + { + "epoch": 0.97, + "learning_rate": 1.5806737660912314e-05, + "loss": 1.2403, + "step": 11534 + }, + { + "epoch": 0.97, + "learning_rate": 1.580599760893464e-05, + "loss": 1.3009, + "step": 11535 + }, + { + "epoch": 0.97, + "learning_rate": 1.580525750898596e-05, + "loss": 1.2574, + "step": 11536 + }, + { + "epoch": 0.97, + "learning_rate": 1.580451736107238e-05, + "loss": 1.224, + "step": 11537 + }, + { + "epoch": 0.97, + "learning_rate": 1.580377716520002e-05, + "loss": 1.2091, + "step": 11538 + }, + { + "epoch": 0.97, + "learning_rate": 1.5803036921374996e-05, + "loss": 1.2866, + "step": 11539 + }, + { + "epoch": 0.97, + "learning_rate": 1.5802296629603424e-05, + "loss": 1.2013, + "step": 11540 + }, + { + "epoch": 0.97, + "learning_rate": 1.5801556289891423e-05, + "loss": 1.3262, + "step": 11541 + }, + { + "epoch": 0.97, + "learning_rate": 1.58008159022451e-05, + "loss": 1.2407, + "step": 11542 + }, + { + "epoch": 0.97, + "learning_rate": 1.5800075466670586e-05, + "loss": 1.2689, + "step": 11543 + }, + { + "epoch": 0.97, + "learning_rate": 1.579933498317399e-05, + "loss": 1.2819, + "step": 11544 + }, + { + "epoch": 0.97, + "learning_rate": 1.5798594451761436e-05, + "loss": 1.246, + "step": 11545 + }, + { + "epoch": 0.97, + "learning_rate": 1.5797853872439034e-05, + "loss": 1.2884, + "step": 11546 + }, + { + "epoch": 0.97, + "learning_rate": 1.5797113245212908e-05, + "loss": 1.2472, + "step": 11547 + }, + { + "epoch": 0.97, + "learning_rate": 1.579637257008918e-05, + "loss": 1.3372, + "step": 11548 + }, + { + "epoch": 0.97, + "learning_rate": 1.5795631847073964e-05, + "loss": 1.2594, + "step": 11549 + }, + { + "epoch": 0.97, + "learning_rate": 1.579489107617338e-05, + "loss": 1.1645, + "step": 11550 + }, + { + "epoch": 0.97, + "learning_rate": 1.579415025739356e-05, + "loss": 1.2095, + "step": 11551 + }, + { + "epoch": 0.97, + "learning_rate": 1.579340939074061e-05, + "loss": 1.2586, + "step": 11552 + }, + { + "epoch": 0.97, + "learning_rate": 1.5792668476220657e-05, + "loss": 1.2256, + "step": 11553 + }, + { + "epoch": 0.97, + "learning_rate": 1.5791927513839826e-05, + "loss": 1.2806, + "step": 11554 + }, + { + "epoch": 0.97, + "learning_rate": 1.5791186503604236e-05, + "loss": 1.2194, + "step": 11555 + }, + { + "epoch": 0.97, + "learning_rate": 1.5790445445520008e-05, + "loss": 1.1498, + "step": 11556 + }, + { + "epoch": 0.97, + "learning_rate": 1.578970433959327e-05, + "loss": 1.253, + "step": 11557 + }, + { + "epoch": 0.97, + "learning_rate": 1.5788963185830138e-05, + "loss": 1.2566, + "step": 11558 + }, + { + "epoch": 0.97, + "learning_rate": 1.5788221984236736e-05, + "loss": 1.2691, + "step": 11559 + }, + { + "epoch": 0.97, + "learning_rate": 1.5787480734819198e-05, + "loss": 1.2931, + "step": 11560 + }, + { + "epoch": 0.97, + "learning_rate": 1.5786739437583644e-05, + "loss": 1.1871, + "step": 11561 + }, + { + "epoch": 0.97, + "learning_rate": 1.578599809253619e-05, + "loss": 1.3022, + "step": 11562 + }, + { + "epoch": 0.97, + "learning_rate": 1.5785256699682973e-05, + "loss": 1.2025, + "step": 11563 + }, + { + "epoch": 0.97, + "learning_rate": 1.5784515259030108e-05, + "loss": 1.2853, + "step": 11564 + }, + { + "epoch": 0.97, + "learning_rate": 1.5783773770583734e-05, + "loss": 1.3916, + "step": 11565 + }, + { + "epoch": 0.97, + "learning_rate": 1.5783032234349963e-05, + "loss": 1.178, + "step": 11566 + }, + { + "epoch": 0.97, + "learning_rate": 1.578229065033493e-05, + "loss": 1.2518, + "step": 11567 + }, + { + "epoch": 0.97, + "learning_rate": 1.5781549018544765e-05, + "loss": 1.3162, + "step": 11568 + }, + { + "epoch": 0.97, + "learning_rate": 1.5780807338985593e-05, + "loss": 1.3644, + "step": 11569 + }, + { + "epoch": 0.97, + "learning_rate": 1.5780065611663537e-05, + "loss": 1.2901, + "step": 11570 + }, + { + "epoch": 0.97, + "learning_rate": 1.5779323836584728e-05, + "loss": 1.3219, + "step": 11571 + }, + { + "epoch": 0.97, + "learning_rate": 1.5778582013755296e-05, + "loss": 1.2429, + "step": 11572 + }, + { + "epoch": 0.97, + "learning_rate": 1.5777840143181375e-05, + "loss": 1.2582, + "step": 11573 + }, + { + "epoch": 0.97, + "learning_rate": 1.5777098224869087e-05, + "loss": 1.3427, + "step": 11574 + }, + { + "epoch": 0.97, + "learning_rate": 1.577635625882456e-05, + "loss": 1.2244, + "step": 11575 + }, + { + "epoch": 0.97, + "learning_rate": 1.5775614245053933e-05, + "loss": 1.2604, + "step": 11576 + }, + { + "epoch": 0.97, + "learning_rate": 1.5774872183563333e-05, + "loss": 1.2016, + "step": 11577 + }, + { + "epoch": 0.97, + "learning_rate": 1.577413007435889e-05, + "loss": 1.2559, + "step": 11578 + }, + { + "epoch": 0.97, + "learning_rate": 1.5773387917446737e-05, + "loss": 1.2041, + "step": 11579 + }, + { + "epoch": 0.97, + "learning_rate": 1.5772645712833003e-05, + "loss": 1.205, + "step": 11580 + }, + { + "epoch": 0.98, + "learning_rate": 1.5771903460523826e-05, + "loss": 1.2846, + "step": 11581 + }, + { + "epoch": 0.98, + "learning_rate": 1.5771161160525333e-05, + "loss": 1.259, + "step": 11582 + }, + { + "epoch": 0.98, + "learning_rate": 1.5770418812843658e-05, + "loss": 1.3177, + "step": 11583 + }, + { + "epoch": 0.98, + "learning_rate": 1.5769676417484942e-05, + "loss": 1.342, + "step": 11584 + }, + { + "epoch": 0.98, + "learning_rate": 1.5768933974455306e-05, + "loss": 1.2844, + "step": 11585 + }, + { + "epoch": 0.98, + "learning_rate": 1.5768191483760894e-05, + "loss": 1.2789, + "step": 11586 + }, + { + "epoch": 0.98, + "learning_rate": 1.576744894540784e-05, + "loss": 1.269, + "step": 11587 + }, + { + "epoch": 0.98, + "learning_rate": 1.5766706359402275e-05, + "loss": 1.1528, + "step": 11588 + }, + { + "epoch": 0.98, + "learning_rate": 1.5765963725750335e-05, + "loss": 1.4045, + "step": 11589 + }, + { + "epoch": 0.98, + "learning_rate": 1.576522104445816e-05, + "loss": 1.2015, + "step": 11590 + }, + { + "epoch": 0.98, + "learning_rate": 1.5764478315531883e-05, + "loss": 1.2944, + "step": 11591 + }, + { + "epoch": 0.98, + "learning_rate": 1.576373553897764e-05, + "loss": 1.3662, + "step": 11592 + }, + { + "epoch": 0.98, + "learning_rate": 1.5762992714801573e-05, + "loss": 1.285, + "step": 11593 + }, + { + "epoch": 0.98, + "learning_rate": 1.5762249843009812e-05, + "loss": 1.3183, + "step": 11594 + }, + { + "epoch": 0.98, + "learning_rate": 1.5761506923608498e-05, + "loss": 1.199, + "step": 11595 + }, + { + "epoch": 0.98, + "learning_rate": 1.5760763956603776e-05, + "loss": 1.2463, + "step": 11596 + }, + { + "epoch": 0.98, + "learning_rate": 1.576002094200177e-05, + "loss": 1.2589, + "step": 11597 + }, + { + "epoch": 0.98, + "learning_rate": 1.5759277879808634e-05, + "loss": 1.1315, + "step": 11598 + }, + { + "epoch": 0.98, + "learning_rate": 1.5758534770030495e-05, + "loss": 1.2059, + "step": 11599 + }, + { + "epoch": 0.98, + "learning_rate": 1.5757791612673505e-05, + "loss": 1.3054, + "step": 11600 + }, + { + "epoch": 0.98, + "learning_rate": 1.5757048407743795e-05, + "loss": 1.2139, + "step": 11601 + }, + { + "epoch": 0.98, + "learning_rate": 1.575630515524751e-05, + "loss": 1.2222, + "step": 11602 + }, + { + "epoch": 0.98, + "learning_rate": 1.5755561855190785e-05, + "loss": 1.3654, + "step": 11603 + }, + { + "epoch": 0.98, + "learning_rate": 1.5754818507579774e-05, + "loss": 1.2834, + "step": 11604 + }, + { + "epoch": 0.98, + "learning_rate": 1.5754075112420605e-05, + "loss": 1.3037, + "step": 11605 + }, + { + "epoch": 0.98, + "learning_rate": 1.5753331669719426e-05, + "loss": 1.2169, + "step": 11606 + }, + { + "epoch": 0.98, + "learning_rate": 1.575258817948238e-05, + "loss": 1.2782, + "step": 11607 + }, + { + "epoch": 0.98, + "learning_rate": 1.575184464171561e-05, + "loss": 1.3041, + "step": 11608 + }, + { + "epoch": 0.98, + "learning_rate": 1.5751101056425257e-05, + "loss": 1.1885, + "step": 11609 + }, + { + "epoch": 0.98, + "learning_rate": 1.5750357423617467e-05, + "loss": 1.3944, + "step": 11610 + }, + { + "epoch": 0.98, + "learning_rate": 1.5749613743298385e-05, + "loss": 1.2365, + "step": 11611 + }, + { + "epoch": 0.98, + "learning_rate": 1.5748870015474156e-05, + "loss": 1.3031, + "step": 11612 + }, + { + "epoch": 0.98, + "learning_rate": 1.574812624015092e-05, + "loss": 1.2573, + "step": 11613 + }, + { + "epoch": 0.98, + "learning_rate": 1.5747382417334826e-05, + "loss": 1.2777, + "step": 11614 + }, + { + "epoch": 0.98, + "learning_rate": 1.574663854703202e-05, + "loss": 1.2405, + "step": 11615 + }, + { + "epoch": 0.98, + "learning_rate": 1.5745894629248645e-05, + "loss": 1.192, + "step": 11616 + }, + { + "epoch": 0.98, + "learning_rate": 1.5745150663990855e-05, + "loss": 1.2657, + "step": 11617 + }, + { + "epoch": 0.98, + "learning_rate": 1.5744406651264786e-05, + "loss": 1.3276, + "step": 11618 + }, + { + "epoch": 0.98, + "learning_rate": 1.5743662591076595e-05, + "loss": 1.3198, + "step": 11619 + }, + { + "epoch": 0.98, + "learning_rate": 1.574291848343242e-05, + "loss": 1.2378, + "step": 11620 + }, + { + "epoch": 0.98, + "learning_rate": 1.574217432833842e-05, + "loss": 1.4466, + "step": 11621 + }, + { + "epoch": 0.98, + "learning_rate": 1.574143012580073e-05, + "loss": 1.1997, + "step": 11622 + }, + { + "epoch": 0.98, + "learning_rate": 1.5740685875825514e-05, + "loss": 1.2928, + "step": 11623 + }, + { + "epoch": 0.98, + "learning_rate": 1.5739941578418912e-05, + "loss": 1.2897, + "step": 11624 + }, + { + "epoch": 0.98, + "learning_rate": 1.5739197233587077e-05, + "loss": 1.302, + "step": 11625 + }, + { + "epoch": 0.98, + "learning_rate": 1.5738452841336157e-05, + "loss": 1.1985, + "step": 11626 + }, + { + "epoch": 0.98, + "learning_rate": 1.5737708401672302e-05, + "loss": 1.2434, + "step": 11627 + }, + { + "epoch": 0.98, + "learning_rate": 1.5736963914601664e-05, + "loss": 1.2314, + "step": 11628 + }, + { + "epoch": 0.98, + "learning_rate": 1.5736219380130395e-05, + "loss": 1.3539, + "step": 11629 + }, + { + "epoch": 0.98, + "learning_rate": 1.5735474798264643e-05, + "loss": 1.225, + "step": 11630 + }, + { + "epoch": 0.98, + "learning_rate": 1.5734730169010567e-05, + "loss": 1.2735, + "step": 11631 + }, + { + "epoch": 0.98, + "learning_rate": 1.5733985492374314e-05, + "loss": 1.29, + "step": 11632 + }, + { + "epoch": 0.98, + "learning_rate": 1.5733240768362038e-05, + "loss": 1.2427, + "step": 11633 + }, + { + "epoch": 0.98, + "learning_rate": 1.573249599697989e-05, + "loss": 1.1986, + "step": 11634 + }, + { + "epoch": 0.98, + "learning_rate": 1.573175117823402e-05, + "loss": 1.3205, + "step": 11635 + }, + { + "epoch": 0.98, + "learning_rate": 1.57310063121306e-05, + "loss": 1.3627, + "step": 11636 + }, + { + "epoch": 0.98, + "learning_rate": 1.5730261398675763e-05, + "loss": 1.3123, + "step": 11637 + }, + { + "epoch": 0.98, + "learning_rate": 1.5729516437875674e-05, + "loss": 1.277, + "step": 11638 + }, + { + "epoch": 0.98, + "learning_rate": 1.5728771429736485e-05, + "loss": 1.3109, + "step": 11639 + }, + { + "epoch": 0.98, + "learning_rate": 1.5728026374264354e-05, + "loss": 1.3843, + "step": 11640 + }, + { + "epoch": 0.98, + "learning_rate": 1.572728127146544e-05, + "loss": 1.2286, + "step": 11641 + }, + { + "epoch": 0.98, + "learning_rate": 1.5726536121345894e-05, + "loss": 1.2368, + "step": 11642 + }, + { + "epoch": 0.98, + "learning_rate": 1.572579092391187e-05, + "loss": 1.2275, + "step": 11643 + }, + { + "epoch": 0.98, + "learning_rate": 1.572504567916953e-05, + "loss": 1.1941, + "step": 11644 + }, + { + "epoch": 0.98, + "learning_rate": 1.5724300387125032e-05, + "loss": 1.2527, + "step": 11645 + }, + { + "epoch": 0.98, + "learning_rate": 1.5723555047784534e-05, + "loss": 1.2392, + "step": 11646 + }, + { + "epoch": 0.98, + "learning_rate": 1.5722809661154185e-05, + "loss": 1.2087, + "step": 11647 + }, + { + "epoch": 0.98, + "learning_rate": 1.5722064227240157e-05, + "loss": 1.3076, + "step": 11648 + }, + { + "epoch": 0.98, + "learning_rate": 1.5721318746048603e-05, + "loss": 1.2494, + "step": 11649 + }, + { + "epoch": 0.98, + "learning_rate": 1.572057321758568e-05, + "loss": 1.1919, + "step": 11650 + }, + { + "epoch": 0.98, + "learning_rate": 1.571982764185755e-05, + "loss": 1.2705, + "step": 11651 + }, + { + "epoch": 0.98, + "learning_rate": 1.5719082018870377e-05, + "loss": 1.2374, + "step": 11652 + }, + { + "epoch": 0.98, + "learning_rate": 1.5718336348630318e-05, + "loss": 1.2584, + "step": 11653 + }, + { + "epoch": 0.98, + "learning_rate": 1.571759063114353e-05, + "loss": 1.2379, + "step": 11654 + }, + { + "epoch": 0.98, + "learning_rate": 1.571684486641618e-05, + "loss": 1.1952, + "step": 11655 + }, + { + "epoch": 0.98, + "learning_rate": 1.571609905445443e-05, + "loss": 1.2025, + "step": 11656 + }, + { + "epoch": 0.98, + "learning_rate": 1.571535319526444e-05, + "loss": 1.3013, + "step": 11657 + }, + { + "epoch": 0.98, + "learning_rate": 1.571460728885237e-05, + "loss": 1.348, + "step": 11658 + }, + { + "epoch": 0.98, + "learning_rate": 1.5713861335224384e-05, + "loss": 1.4037, + "step": 11659 + }, + { + "epoch": 0.98, + "learning_rate": 1.571311533438665e-05, + "loss": 1.2831, + "step": 11660 + }, + { + "epoch": 0.98, + "learning_rate": 1.571236928634533e-05, + "loss": 1.2989, + "step": 11661 + }, + { + "epoch": 0.98, + "learning_rate": 1.5711623191106588e-05, + "loss": 1.303, + "step": 11662 + }, + { + "epoch": 0.98, + "learning_rate": 1.5710877048676584e-05, + "loss": 1.3181, + "step": 11663 + }, + { + "epoch": 0.98, + "learning_rate": 1.5710130859061485e-05, + "loss": 1.2798, + "step": 11664 + }, + { + "epoch": 0.98, + "learning_rate": 1.570938462226746e-05, + "loss": 1.2444, + "step": 11665 + }, + { + "epoch": 0.98, + "learning_rate": 1.570863833830067e-05, + "loss": 1.3598, + "step": 11666 + }, + { + "epoch": 0.98, + "learning_rate": 1.570789200716729e-05, + "loss": 1.2557, + "step": 11667 + }, + { + "epoch": 0.98, + "learning_rate": 1.570714562887347e-05, + "loss": 1.2454, + "step": 11668 + }, + { + "epoch": 0.98, + "learning_rate": 1.570639920342539e-05, + "loss": 1.2587, + "step": 11669 + }, + { + "epoch": 0.98, + "learning_rate": 1.5705652730829215e-05, + "loss": 1.2626, + "step": 11670 + }, + { + "epoch": 0.98, + "learning_rate": 1.570490621109111e-05, + "loss": 1.3891, + "step": 11671 + }, + { + "epoch": 0.98, + "learning_rate": 1.570415964421724e-05, + "loss": 1.311, + "step": 11672 + }, + { + "epoch": 0.98, + "learning_rate": 1.5703413030213782e-05, + "loss": 1.3152, + "step": 11673 + }, + { + "epoch": 0.98, + "learning_rate": 1.57026663690869e-05, + "loss": 1.3084, + "step": 11674 + }, + { + "epoch": 0.98, + "learning_rate": 1.570191966084276e-05, + "loss": 1.167, + "step": 11675 + }, + { + "epoch": 0.98, + "learning_rate": 1.570117290548754e-05, + "loss": 1.3704, + "step": 11676 + }, + { + "epoch": 0.98, + "learning_rate": 1.5700426103027403e-05, + "loss": 1.2401, + "step": 11677 + }, + { + "epoch": 0.98, + "learning_rate": 1.569967925346852e-05, + "loss": 1.1757, + "step": 11678 + }, + { + "epoch": 0.98, + "learning_rate": 1.5698932356817065e-05, + "loss": 1.231, + "step": 11679 + }, + { + "epoch": 0.98, + "learning_rate": 1.5698185413079205e-05, + "loss": 1.2225, + "step": 11680 + }, + { + "epoch": 0.98, + "learning_rate": 1.5697438422261117e-05, + "loss": 1.3244, + "step": 11681 + }, + { + "epoch": 0.98, + "learning_rate": 1.5696691384368966e-05, + "loss": 1.2642, + "step": 11682 + }, + { + "epoch": 0.98, + "learning_rate": 1.569594429940893e-05, + "loss": 1.3881, + "step": 11683 + }, + { + "epoch": 0.98, + "learning_rate": 1.569519716738717e-05, + "loss": 1.223, + "step": 11684 + }, + { + "epoch": 0.98, + "learning_rate": 1.5694449988309878e-05, + "loss": 1.2137, + "step": 11685 + }, + { + "epoch": 0.98, + "learning_rate": 1.5693702762183215e-05, + "loss": 1.2926, + "step": 11686 + }, + { + "epoch": 0.98, + "learning_rate": 1.5692955489013358e-05, + "loss": 1.21, + "step": 11687 + }, + { + "epoch": 0.98, + "learning_rate": 1.569220816880648e-05, + "loss": 1.2383, + "step": 11688 + }, + { + "epoch": 0.98, + "learning_rate": 1.569146080156876e-05, + "loss": 1.2862, + "step": 11689 + }, + { + "epoch": 0.98, + "learning_rate": 1.5690713387306365e-05, + "loss": 1.2155, + "step": 11690 + }, + { + "epoch": 0.98, + "learning_rate": 1.5689965926025473e-05, + "loss": 1.3905, + "step": 11691 + }, + { + "epoch": 0.98, + "learning_rate": 1.5689218417732267e-05, + "loss": 1.2335, + "step": 11692 + }, + { + "epoch": 0.98, + "learning_rate": 1.5688470862432916e-05, + "loss": 1.3553, + "step": 11693 + }, + { + "epoch": 0.98, + "learning_rate": 1.5687723260133595e-05, + "loss": 1.3249, + "step": 11694 + }, + { + "epoch": 0.98, + "learning_rate": 1.568697561084049e-05, + "loss": 1.2501, + "step": 11695 + }, + { + "epoch": 0.98, + "learning_rate": 1.568622791455977e-05, + "loss": 1.2421, + "step": 11696 + }, + { + "epoch": 0.98, + "learning_rate": 1.568548017129761e-05, + "loss": 1.1751, + "step": 11697 + }, + { + "epoch": 0.98, + "learning_rate": 1.5684732381060197e-05, + "loss": 1.2224, + "step": 11698 + }, + { + "epoch": 0.98, + "learning_rate": 1.5683984543853706e-05, + "loss": 1.2315, + "step": 11699 + }, + { + "epoch": 0.99, + "learning_rate": 1.5683236659684317e-05, + "loss": 1.2081, + "step": 11700 + }, + { + "epoch": 0.99, + "learning_rate": 1.5682488728558207e-05, + "loss": 1.2746, + "step": 11701 + }, + { + "epoch": 0.99, + "learning_rate": 1.5681740750481554e-05, + "loss": 1.2981, + "step": 11702 + }, + { + "epoch": 0.99, + "learning_rate": 1.5680992725460544e-05, + "loss": 1.2476, + "step": 11703 + }, + { + "epoch": 0.99, + "learning_rate": 1.568024465350135e-05, + "loss": 1.2265, + "step": 11704 + }, + { + "epoch": 0.99, + "learning_rate": 1.567949653461016e-05, + "loss": 1.2247, + "step": 11705 + }, + { + "epoch": 0.99, + "learning_rate": 1.5678748368793155e-05, + "loss": 1.2299, + "step": 11706 + }, + { + "epoch": 0.99, + "learning_rate": 1.5678000156056508e-05, + "loss": 1.2914, + "step": 11707 + }, + { + "epoch": 0.99, + "learning_rate": 1.567725189640641e-05, + "loss": 1.3269, + "step": 11708 + }, + { + "epoch": 0.99, + "learning_rate": 1.5676503589849038e-05, + "loss": 1.3387, + "step": 11709 + }, + { + "epoch": 0.99, + "learning_rate": 1.5675755236390572e-05, + "loss": 1.3194, + "step": 11710 + }, + { + "epoch": 0.99, + "learning_rate": 1.5675006836037207e-05, + "loss": 1.2531, + "step": 11711 + }, + { + "epoch": 0.99, + "learning_rate": 1.5674258388795118e-05, + "loss": 1.2046, + "step": 11712 + }, + { + "epoch": 0.99, + "learning_rate": 1.567350989467049e-05, + "loss": 1.3045, + "step": 11713 + }, + { + "epoch": 0.99, + "learning_rate": 1.5672761353669503e-05, + "loss": 1.2153, + "step": 11714 + }, + { + "epoch": 0.99, + "learning_rate": 1.5672012765798352e-05, + "loss": 1.2821, + "step": 11715 + }, + { + "epoch": 0.99, + "learning_rate": 1.5671264131063212e-05, + "loss": 1.3272, + "step": 11716 + }, + { + "epoch": 0.99, + "learning_rate": 1.5670515449470278e-05, + "loss": 1.2775, + "step": 11717 + }, + { + "epoch": 0.99, + "learning_rate": 1.5669766721025725e-05, + "loss": 1.3026, + "step": 11718 + }, + { + "epoch": 0.99, + "learning_rate": 1.5669017945735747e-05, + "loss": 1.2363, + "step": 11719 + }, + { + "epoch": 0.99, + "learning_rate": 1.5668269123606528e-05, + "loss": 1.2481, + "step": 11720 + }, + { + "epoch": 0.99, + "learning_rate": 1.5667520254644257e-05, + "loss": 1.2469, + "step": 11721 + }, + { + "epoch": 0.99, + "learning_rate": 1.566677133885512e-05, + "loss": 1.3186, + "step": 11722 + }, + { + "epoch": 0.99, + "learning_rate": 1.56660223762453e-05, + "loss": 1.3686, + "step": 11723 + }, + { + "epoch": 0.99, + "learning_rate": 1.5665273366820994e-05, + "loss": 1.2311, + "step": 11724 + }, + { + "epoch": 0.99, + "learning_rate": 1.5664524310588386e-05, + "loss": 1.2574, + "step": 11725 + }, + { + "epoch": 0.99, + "learning_rate": 1.5663775207553666e-05, + "loss": 1.2938, + "step": 11726 + }, + { + "epoch": 0.99, + "learning_rate": 1.5663026057723022e-05, + "loss": 1.2402, + "step": 11727 + }, + { + "epoch": 0.99, + "learning_rate": 1.566227686110264e-05, + "loss": 1.2049, + "step": 11728 + }, + { + "epoch": 0.99, + "learning_rate": 1.566152761769872e-05, + "loss": 1.2322, + "step": 11729 + }, + { + "epoch": 0.99, + "learning_rate": 1.5660778327517446e-05, + "loss": 1.2314, + "step": 11730 + }, + { + "epoch": 0.99, + "learning_rate": 1.5660028990565007e-05, + "loss": 1.2438, + "step": 11731 + }, + { + "epoch": 0.99, + "learning_rate": 1.5659279606847597e-05, + "loss": 1.2851, + "step": 11732 + }, + { + "epoch": 0.99, + "learning_rate": 1.565853017637141e-05, + "loss": 1.255, + "step": 11733 + }, + { + "epoch": 0.99, + "learning_rate": 1.5657780699142634e-05, + "loss": 1.2395, + "step": 11734 + }, + { + "epoch": 0.99, + "learning_rate": 1.5657031175167465e-05, + "loss": 1.253, + "step": 11735 + }, + { + "epoch": 0.99, + "learning_rate": 1.5656281604452095e-05, + "loss": 1.3382, + "step": 11736 + }, + { + "epoch": 0.99, + "learning_rate": 1.5655531987002713e-05, + "loss": 1.2074, + "step": 11737 + }, + { + "epoch": 0.99, + "learning_rate": 1.5654782322825515e-05, + "loss": 1.2734, + "step": 11738 + }, + { + "epoch": 0.99, + "learning_rate": 1.5654032611926698e-05, + "loss": 1.2526, + "step": 11739 + }, + { + "epoch": 0.99, + "learning_rate": 1.5653282854312456e-05, + "loss": 1.3109, + "step": 11740 + }, + { + "epoch": 0.99, + "learning_rate": 1.565253304998898e-05, + "loss": 1.2744, + "step": 11741 + }, + { + "epoch": 0.99, + "learning_rate": 1.5651783198962464e-05, + "loss": 1.1879, + "step": 11742 + }, + { + "epoch": 0.99, + "learning_rate": 1.5651033301239112e-05, + "loss": 1.3061, + "step": 11743 + }, + { + "epoch": 0.99, + "learning_rate": 1.5650283356825113e-05, + "loss": 1.3337, + "step": 11744 + }, + { + "epoch": 0.99, + "learning_rate": 1.5649533365726658e-05, + "loss": 1.2524, + "step": 11745 + }, + { + "epoch": 0.99, + "learning_rate": 1.5648783327949956e-05, + "loss": 1.2058, + "step": 11746 + }, + { + "epoch": 0.99, + "learning_rate": 1.5648033243501197e-05, + "loss": 1.243, + "step": 11747 + }, + { + "epoch": 0.99, + "learning_rate": 1.564728311238658e-05, + "loss": 1.3376, + "step": 11748 + }, + { + "epoch": 0.99, + "learning_rate": 1.5646532934612303e-05, + "loss": 1.2323, + "step": 11749 + }, + { + "epoch": 0.99, + "learning_rate": 1.5645782710184562e-05, + "loss": 1.1992, + "step": 11750 + }, + { + "epoch": 0.99, + "learning_rate": 1.564503243910956e-05, + "loss": 1.36, + "step": 11751 + }, + { + "epoch": 0.99, + "learning_rate": 1.5644282121393492e-05, + "loss": 1.302, + "step": 11752 + }, + { + "epoch": 0.99, + "learning_rate": 1.564353175704256e-05, + "loss": 1.2594, + "step": 11753 + }, + { + "epoch": 0.99, + "learning_rate": 1.564278134606296e-05, + "loss": 1.2387, + "step": 11754 + }, + { + "epoch": 0.99, + "learning_rate": 1.5642030888460897e-05, + "loss": 1.2676, + "step": 11755 + }, + { + "epoch": 0.99, + "learning_rate": 1.564128038424257e-05, + "loss": 1.3615, + "step": 11756 + }, + { + "epoch": 0.99, + "learning_rate": 1.5640529833414175e-05, + "loss": 1.2619, + "step": 11757 + }, + { + "epoch": 0.99, + "learning_rate": 1.5639779235981917e-05, + "loss": 1.2869, + "step": 11758 + }, + { + "epoch": 0.99, + "learning_rate": 1.5639028591952e-05, + "loss": 1.2368, + "step": 11759 + }, + { + "epoch": 0.99, + "learning_rate": 1.5638277901330625e-05, + "loss": 1.2963, + "step": 11760 + }, + { + "epoch": 0.99, + "learning_rate": 1.5637527164123995e-05, + "loss": 1.3767, + "step": 11761 + }, + { + "epoch": 0.99, + "learning_rate": 1.563677638033831e-05, + "loss": 1.2896, + "step": 11762 + }, + { + "epoch": 0.99, + "learning_rate": 1.5636025549979772e-05, + "loss": 1.2651, + "step": 11763 + }, + { + "epoch": 0.99, + "learning_rate": 1.5635274673054587e-05, + "loss": 1.2971, + "step": 11764 + }, + { + "epoch": 0.99, + "learning_rate": 1.5634523749568963e-05, + "loss": 1.2971, + "step": 11765 + }, + { + "epoch": 0.99, + "learning_rate": 1.5633772779529096e-05, + "loss": 1.202, + "step": 11766 + }, + { + "epoch": 0.99, + "learning_rate": 1.5633021762941202e-05, + "loss": 1.2416, + "step": 11767 + }, + { + "epoch": 0.99, + "learning_rate": 1.5632270699811473e-05, + "loss": 1.3713, + "step": 11768 + }, + { + "epoch": 0.99, + "learning_rate": 1.5631519590146123e-05, + "loss": 1.3105, + "step": 11769 + }, + { + "epoch": 0.99, + "learning_rate": 1.5630768433951356e-05, + "loss": 1.1854, + "step": 11770 + }, + { + "epoch": 0.99, + "learning_rate": 1.563001723123338e-05, + "loss": 1.3319, + "step": 11771 + }, + { + "epoch": 0.99, + "learning_rate": 1.5629265981998395e-05, + "loss": 1.2081, + "step": 11772 + }, + { + "epoch": 0.99, + "learning_rate": 1.5628514686252615e-05, + "loss": 1.3271, + "step": 11773 + }, + { + "epoch": 0.99, + "learning_rate": 1.5627763344002247e-05, + "loss": 1.2389, + "step": 11774 + }, + { + "epoch": 0.99, + "learning_rate": 1.5627011955253497e-05, + "loss": 1.3398, + "step": 11775 + }, + { + "epoch": 0.99, + "learning_rate": 1.562626052001257e-05, + "loss": 1.2314, + "step": 11776 + }, + { + "epoch": 0.99, + "learning_rate": 1.562550903828568e-05, + "loss": 1.2377, + "step": 11777 + }, + { + "epoch": 0.99, + "learning_rate": 1.5624757510079035e-05, + "loss": 1.212, + "step": 11778 + }, + { + "epoch": 0.99, + "learning_rate": 1.5624005935398845e-05, + "loss": 1.2937, + "step": 11779 + }, + { + "epoch": 0.99, + "learning_rate": 1.5623254314251315e-05, + "loss": 1.27, + "step": 11780 + }, + { + "epoch": 0.99, + "learning_rate": 1.562250264664266e-05, + "loss": 1.2935, + "step": 11781 + }, + { + "epoch": 0.99, + "learning_rate": 1.5621750932579084e-05, + "loss": 1.2356, + "step": 11782 + }, + { + "epoch": 0.99, + "learning_rate": 1.5620999172066807e-05, + "loss": 1.2945, + "step": 11783 + }, + { + "epoch": 0.99, + "learning_rate": 1.5620247365112036e-05, + "loss": 1.2763, + "step": 11784 + }, + { + "epoch": 0.99, + "learning_rate": 1.561949551172098e-05, + "loss": 1.1805, + "step": 11785 + }, + { + "epoch": 0.99, + "learning_rate": 1.5618743611899855e-05, + "loss": 1.27, + "step": 11786 + }, + { + "epoch": 0.99, + "learning_rate": 1.5617991665654876e-05, + "loss": 1.2808, + "step": 11787 + }, + { + "epoch": 0.99, + "learning_rate": 1.5617239672992248e-05, + "loss": 1.3227, + "step": 11788 + }, + { + "epoch": 0.99, + "learning_rate": 1.5616487633918187e-05, + "loss": 1.256, + "step": 11789 + }, + { + "epoch": 0.99, + "learning_rate": 1.561573554843891e-05, + "loss": 1.186, + "step": 11790 + }, + { + "epoch": 0.99, + "learning_rate": 1.5614983416560624e-05, + "loss": 1.2603, + "step": 11791 + }, + { + "epoch": 0.99, + "learning_rate": 1.5614231238289554e-05, + "loss": 1.3067, + "step": 11792 + }, + { + "epoch": 0.99, + "learning_rate": 1.5613479013631906e-05, + "loss": 1.2741, + "step": 11793 + }, + { + "epoch": 0.99, + "learning_rate": 1.56127267425939e-05, + "loss": 1.2565, + "step": 11794 + }, + { + "epoch": 0.99, + "learning_rate": 1.5611974425181745e-05, + "loss": 1.1834, + "step": 11795 + }, + { + "epoch": 0.99, + "learning_rate": 1.5611222061401666e-05, + "loss": 1.2353, + "step": 11796 + }, + { + "epoch": 0.99, + "learning_rate": 1.561046965125987e-05, + "loss": 1.3495, + "step": 11797 + }, + { + "epoch": 0.99, + "learning_rate": 1.5609717194762582e-05, + "loss": 1.2364, + "step": 11798 + }, + { + "epoch": 0.99, + "learning_rate": 1.5608964691916014e-05, + "loss": 1.2146, + "step": 11799 + }, + { + "epoch": 0.99, + "learning_rate": 1.5608212142726383e-05, + "loss": 1.3204, + "step": 11800 + }, + { + "epoch": 0.99, + "learning_rate": 1.560745954719991e-05, + "loss": 1.2769, + "step": 11801 + }, + { + "epoch": 0.99, + "learning_rate": 1.560670690534281e-05, + "loss": 1.3667, + "step": 11802 + }, + { + "epoch": 0.99, + "learning_rate": 1.5605954217161304e-05, + "loss": 1.2006, + "step": 11803 + }, + { + "epoch": 0.99, + "learning_rate": 1.560520148266161e-05, + "loss": 1.2443, + "step": 11804 + }, + { + "epoch": 0.99, + "learning_rate": 1.560444870184995e-05, + "loss": 1.3133, + "step": 11805 + }, + { + "epoch": 0.99, + "learning_rate": 1.5603695874732538e-05, + "loss": 1.2164, + "step": 11806 + }, + { + "epoch": 0.99, + "learning_rate": 1.56029430013156e-05, + "loss": 1.2353, + "step": 11807 + }, + { + "epoch": 0.99, + "learning_rate": 1.560219008160535e-05, + "loss": 1.2694, + "step": 11808 + }, + { + "epoch": 0.99, + "learning_rate": 1.5601437115608014e-05, + "loss": 1.213, + "step": 11809 + }, + { + "epoch": 0.99, + "learning_rate": 1.5600684103329814e-05, + "loss": 1.2054, + "step": 11810 + }, + { + "epoch": 0.99, + "learning_rate": 1.5599931044776968e-05, + "loss": 1.3408, + "step": 11811 + }, + { + "epoch": 0.99, + "learning_rate": 1.55991779399557e-05, + "loss": 1.2859, + "step": 11812 + }, + { + "epoch": 0.99, + "learning_rate": 1.559842478887223e-05, + "loss": 1.2633, + "step": 11813 + }, + { + "epoch": 0.99, + "learning_rate": 1.5597671591532786e-05, + "loss": 1.3398, + "step": 11814 + }, + { + "epoch": 0.99, + "learning_rate": 1.5596918347943587e-05, + "loss": 1.3074, + "step": 11815 + }, + { + "epoch": 0.99, + "learning_rate": 1.5596165058110854e-05, + "loss": 1.2931, + "step": 11816 + }, + { + "epoch": 0.99, + "learning_rate": 1.5595411722040815e-05, + "loss": 1.3234, + "step": 11817 + }, + { + "epoch": 0.99, + "learning_rate": 1.5594658339739697e-05, + "loss": 1.2864, + "step": 11818 + }, + { + "epoch": 1.0, + "learning_rate": 1.5593904911213722e-05, + "loss": 1.2766, + "step": 11819 + }, + { + "epoch": 1.0, + "learning_rate": 1.5593151436469113e-05, + "loss": 1.2424, + "step": 11820 + }, + { + "epoch": 1.0, + "learning_rate": 1.5592397915512093e-05, + "loss": 1.3201, + "step": 11821 + }, + { + "epoch": 1.0, + "learning_rate": 1.5591644348348895e-05, + "loss": 1.2396, + "step": 11822 + }, + { + "epoch": 1.0, + "learning_rate": 1.5590890734985738e-05, + "loss": 1.2933, + "step": 11823 + }, + { + "epoch": 1.0, + "learning_rate": 1.5590137075428856e-05, + "loss": 1.2565, + "step": 11824 + }, + { + "epoch": 1.0, + "learning_rate": 1.558938336968447e-05, + "loss": 1.3399, + "step": 11825 + }, + { + "epoch": 1.0, + "learning_rate": 1.5588629617758816e-05, + "loss": 1.2193, + "step": 11826 + }, + { + "epoch": 1.0, + "learning_rate": 1.5587875819658108e-05, + "loss": 1.2608, + "step": 11827 + }, + { + "epoch": 1.0, + "learning_rate": 1.5587121975388586e-05, + "loss": 1.3033, + "step": 11828 + }, + { + "epoch": 1.0, + "learning_rate": 1.5586368084956472e-05, + "loss": 1.2445, + "step": 11829 + }, + { + "epoch": 1.0, + "learning_rate": 1.5585614148368e-05, + "loss": 1.2199, + "step": 11830 + }, + { + "epoch": 1.0, + "learning_rate": 1.5584860165629393e-05, + "loss": 1.3288, + "step": 11831 + }, + { + "epoch": 1.0, + "learning_rate": 1.5584106136746888e-05, + "loss": 1.2864, + "step": 11832 + }, + { + "epoch": 1.0, + "learning_rate": 1.5583352061726705e-05, + "loss": 1.2067, + "step": 11833 + }, + { + "epoch": 1.0, + "learning_rate": 1.5582597940575082e-05, + "loss": 1.3429, + "step": 11834 + }, + { + "epoch": 1.0, + "learning_rate": 1.558184377329825e-05, + "loss": 1.3158, + "step": 11835 + }, + { + "epoch": 1.0, + "learning_rate": 1.558108955990244e-05, + "loss": 1.2703, + "step": 11836 + }, + { + "epoch": 1.0, + "learning_rate": 1.558033530039388e-05, + "loss": 1.278, + "step": 11837 + }, + { + "epoch": 1.0, + "learning_rate": 1.5579580994778805e-05, + "loss": 1.2912, + "step": 11838 + }, + { + "epoch": 1.0, + "learning_rate": 1.5578826643063444e-05, + "loss": 1.264, + "step": 11839 + }, + { + "epoch": 1.0, + "learning_rate": 1.5578072245254032e-05, + "loss": 1.3062, + "step": 11840 + }, + { + "epoch": 1.0, + "learning_rate": 1.5577317801356805e-05, + "loss": 1.2672, + "step": 11841 + }, + { + "epoch": 1.0, + "learning_rate": 1.557656331137799e-05, + "loss": 1.3261, + "step": 11842 + }, + { + "epoch": 1.0, + "learning_rate": 1.5575808775323825e-05, + "loss": 1.3274, + "step": 11843 + }, + { + "epoch": 1.0, + "learning_rate": 1.557505419320054e-05, + "loss": 1.2899, + "step": 11844 + }, + { + "epoch": 1.0, + "learning_rate": 1.5574299565014383e-05, + "loss": 1.2874, + "step": 11845 + }, + { + "epoch": 1.0, + "learning_rate": 1.557354489077157e-05, + "loss": 1.2208, + "step": 11846 + }, + { + "epoch": 1.0, + "learning_rate": 1.5572790170478347e-05, + "loss": 1.3052, + "step": 11847 + }, + { + "epoch": 1.0, + "learning_rate": 1.557203540414095e-05, + "loss": 1.2155, + "step": 11848 + }, + { + "epoch": 1.0, + "learning_rate": 1.5571280591765613e-05, + "loss": 1.3089, + "step": 11849 + }, + { + "epoch": 1.0, + "learning_rate": 1.557052573335857e-05, + "loss": 1.2901, + "step": 11850 + }, + { + "epoch": 1.0, + "learning_rate": 1.5569770828926062e-05, + "loss": 1.4083, + "step": 11851 + }, + { + "epoch": 1.0, + "learning_rate": 1.5569015878474326e-05, + "loss": 1.3506, + "step": 11852 + }, + { + "epoch": 1.0, + "learning_rate": 1.5568260882009596e-05, + "loss": 1.2921, + "step": 11853 + }, + { + "epoch": 1.0, + "learning_rate": 1.5567505839538115e-05, + "loss": 1.3871, + "step": 11854 + }, + { + "epoch": 1.0, + "learning_rate": 1.556675075106612e-05, + "loss": 1.2944, + "step": 11855 + }, + { + "epoch": 1.0, + "learning_rate": 1.5565995616599846e-05, + "loss": 1.2114, + "step": 11856 + }, + { + "epoch": 1.0, + "learning_rate": 1.5565240436145536e-05, + "loss": 1.2322, + "step": 11857 + }, + { + "epoch": 1.0, + "learning_rate": 1.556448520970943e-05, + "loss": 1.2708, + "step": 11858 + }, + { + "epoch": 1.0, + "learning_rate": 1.5563729937297763e-05, + "loss": 1.2695, + "step": 11859 + }, + { + "epoch": 1.0, + "learning_rate": 1.556297461891678e-05, + "loss": 1.2897, + "step": 11860 + }, + { + "epoch": 1.0, + "learning_rate": 1.556221925457272e-05, + "loss": 1.1785, + "step": 11861 + }, + { + "epoch": 1.0, + "learning_rate": 1.5561463844271825e-05, + "loss": 1.3257, + "step": 11862 + }, + { + "epoch": 1.0, + "learning_rate": 1.5560708388020335e-05, + "loss": 1.2758, + "step": 11863 + }, + { + "epoch": 1.0, + "learning_rate": 1.5559952885824494e-05, + "loss": 1.2617, + "step": 11864 + }, + { + "epoch": 1.0, + "learning_rate": 1.5559197337690545e-05, + "loss": 1.184, + "step": 11865 + }, + { + "epoch": 1.0, + "learning_rate": 1.5558441743624724e-05, + "loss": 1.2839, + "step": 11866 + }, + { + "epoch": 1.0, + "learning_rate": 1.5557686103633282e-05, + "loss": 1.3313, + "step": 11867 + }, + { + "epoch": 1.0, + "learning_rate": 1.5556930417722455e-05, + "loss": 1.2008, + "step": 11868 + }, + { + "epoch": 1.0, + "learning_rate": 1.555617468589849e-05, + "loss": 1.3058, + "step": 11869 + }, + { + "epoch": 1.0, + "learning_rate": 1.5555418908167634e-05, + "loss": 1.3253, + "step": 11870 + }, + { + "epoch": 1.0, + "learning_rate": 1.555466308453613e-05, + "loss": 1.2983, + "step": 11871 + }, + { + "epoch": 1.0, + "learning_rate": 1.555390721501022e-05, + "loss": 1.2632, + "step": 11872 + }, + { + "epoch": 1.0, + "learning_rate": 1.555315129959615e-05, + "loss": 1.258, + "step": 11873 + }, + { + "epoch": 1.0, + "learning_rate": 1.5552395338300172e-05, + "loss": 1.2811, + "step": 11874 + }, + { + "epoch": 1.0, + "learning_rate": 1.5551639331128522e-05, + "loss": 1.2397, + "step": 11875 + }, + { + "epoch": 1.0, + "learning_rate": 1.5550883278087453e-05, + "loss": 1.2836, + "step": 11876 + }, + { + "epoch": 1.0, + "learning_rate": 1.555012717918321e-05, + "loss": 1.2186, + "step": 11877 + }, + { + "epoch": 1.0, + "learning_rate": 1.5549371034422034e-05, + "loss": 1.3733, + "step": 11878 + }, + { + "epoch": 1.0, + "learning_rate": 1.5548614843810187e-05, + "loss": 1.2219, + "step": 11879 + }, + { + "epoch": 1.0, + "learning_rate": 1.5547858607353905e-05, + "loss": 1.1362, + "step": 11880 + }, + { + "epoch": 1.0, + "learning_rate": 1.554710232505944e-05, + "loss": 1.2462, + "step": 11881 + }, + { + "epoch": 1.0, + "learning_rate": 1.5546345996933042e-05, + "loss": 1.1392, + "step": 11882 + }, + { + "epoch": 1.0, + "learning_rate": 1.5545589622980954e-05, + "loss": 1.2174, + "step": 11883 + }, + { + "epoch": 1.0, + "learning_rate": 1.5544833203209432e-05, + "loss": 1.2448, + "step": 11884 + }, + { + "epoch": 1.0, + "learning_rate": 1.5544076737624724e-05, + "loss": 1.1085, + "step": 11885 + }, + { + "epoch": 1.0, + "learning_rate": 1.554332022623308e-05, + "loss": 1.2921, + "step": 11886 + }, + { + "epoch": 1.0, + "learning_rate": 1.5542563669040745e-05, + "loss": 1.1911, + "step": 11887 + }, + { + "epoch": 1.0, + "learning_rate": 1.554180706605398e-05, + "loss": 1.2635, + "step": 11888 + }, + { + "epoch": 1.0, + "learning_rate": 1.5541050417279034e-05, + "loss": 1.1308, + "step": 11889 + }, + { + "epoch": 1.0, + "learning_rate": 1.5540293722722153e-05, + "loss": 1.245, + "step": 11890 + }, + { + "epoch": 1.0, + "learning_rate": 1.5539536982389593e-05, + "loss": 1.1301, + "step": 11891 + }, + { + "epoch": 1.0, + "learning_rate": 1.5538780196287603e-05, + "loss": 1.1408, + "step": 11892 + }, + { + "epoch": 1.0, + "learning_rate": 1.553802336442244e-05, + "loss": 1.1008, + "step": 11893 + }, + { + "epoch": 1.0, + "learning_rate": 1.5537266486800356e-05, + "loss": 1.1698, + "step": 11894 + }, + { + "epoch": 1.0, + "learning_rate": 1.5536509563427605e-05, + "loss": 1.2226, + "step": 11895 + }, + { + "epoch": 1.0, + "learning_rate": 1.553575259431044e-05, + "loss": 1.258, + "step": 11896 + }, + { + "epoch": 1.0, + "learning_rate": 1.553499557945511e-05, + "loss": 1.2646, + "step": 11897 + }, + { + "epoch": 1.0, + "learning_rate": 1.553423851886788e-05, + "loss": 1.2108, + "step": 11898 + }, + { + "epoch": 1.0, + "learning_rate": 1.5533481412555e-05, + "loss": 1.2655, + "step": 11899 + }, + { + "epoch": 1.0, + "learning_rate": 1.5532724260522723e-05, + "loss": 1.1724, + "step": 11900 + }, + { + "epoch": 1.0, + "learning_rate": 1.553196706277731e-05, + "loss": 1.1713, + "step": 11901 + }, + { + "epoch": 1.0, + "learning_rate": 1.5531209819325015e-05, + "loss": 1.2094, + "step": 11902 + }, + { + "epoch": 1.0, + "learning_rate": 1.5530452530172096e-05, + "loss": 1.2456, + "step": 11903 + }, + { + "epoch": 1.0, + "learning_rate": 1.5529695195324803e-05, + "loss": 1.2649, + "step": 11904 + }, + { + "epoch": 1.0, + "learning_rate": 1.5528937814789402e-05, + "loss": 1.2208, + "step": 11905 + }, + { + "epoch": 1.0, + "learning_rate": 1.5528180388572146e-05, + "loss": 1.1955, + "step": 11906 + }, + { + "epoch": 1.0, + "learning_rate": 1.5527422916679296e-05, + "loss": 1.2042, + "step": 11907 + }, + { + "epoch": 1.0, + "learning_rate": 1.552666539911711e-05, + "loss": 1.2365, + "step": 11908 + }, + { + "epoch": 1.0, + "learning_rate": 1.552590783589184e-05, + "loss": 1.1884, + "step": 11909 + }, + { + "epoch": 1.0, + "learning_rate": 1.5525150227009754e-05, + "loss": 1.2431, + "step": 11910 + }, + { + "epoch": 1.0, + "learning_rate": 1.552439257247711e-05, + "loss": 1.1428, + "step": 11911 + }, + { + "epoch": 1.0, + "learning_rate": 1.5523634872300165e-05, + "loss": 1.1651, + "step": 11912 + }, + { + "epoch": 1.0, + "learning_rate": 1.5522877126485184e-05, + "loss": 1.2214, + "step": 11913 + }, + { + "epoch": 1.0, + "learning_rate": 1.552211933503842e-05, + "loss": 1.1393, + "step": 11914 + }, + { + "epoch": 1.0, + "learning_rate": 1.5521361497966143e-05, + "loss": 1.2715, + "step": 11915 + }, + { + "epoch": 1.0, + "learning_rate": 1.5520603615274606e-05, + "loss": 1.1715, + "step": 11916 + }, + { + "epoch": 1.0, + "learning_rate": 1.551984568697008e-05, + "loss": 1.1863, + "step": 11917 + }, + { + "epoch": 1.0, + "learning_rate": 1.551908771305882e-05, + "loss": 1.2394, + "step": 11918 + }, + { + "epoch": 1.0, + "learning_rate": 1.5518329693547087e-05, + "loss": 1.184, + "step": 11919 + }, + { + "epoch": 1.0, + "learning_rate": 1.5517571628441153e-05, + "loss": 1.1972, + "step": 11920 + }, + { + "epoch": 1.0, + "learning_rate": 1.5516813517747275e-05, + "loss": 1.3135, + "step": 11921 + }, + { + "epoch": 1.0, + "learning_rate": 1.5516055361471716e-05, + "loss": 1.1807, + "step": 11922 + }, + { + "epoch": 1.0, + "learning_rate": 1.5515297159620745e-05, + "loss": 1.1432, + "step": 11923 + }, + { + "epoch": 1.0, + "learning_rate": 1.5514538912200622e-05, + "loss": 1.3024, + "step": 11924 + }, + { + "epoch": 1.0, + "learning_rate": 1.5513780619217614e-05, + "loss": 1.1815, + "step": 11925 + }, + { + "epoch": 1.0, + "learning_rate": 1.5513022280677986e-05, + "loss": 1.2236, + "step": 11926 + }, + { + "epoch": 1.0, + "learning_rate": 1.5512263896588006e-05, + "loss": 1.1438, + "step": 11927 + }, + { + "epoch": 1.0, + "learning_rate": 1.551150546695394e-05, + "loss": 1.2485, + "step": 11928 + }, + { + "epoch": 1.0, + "learning_rate": 1.551074699178204e-05, + "loss": 1.2938, + "step": 11929 + }, + { + "epoch": 1.0, + "learning_rate": 1.5509988471078596e-05, + "loss": 1.2133, + "step": 11930 + }, + { + "epoch": 1.0, + "learning_rate": 1.550922990484986e-05, + "loss": 1.1723, + "step": 11931 + }, + { + "epoch": 1.0, + "learning_rate": 1.5508471293102104e-05, + "loss": 1.2221, + "step": 11932 + }, + { + "epoch": 1.0, + "learning_rate": 1.5507712635841596e-05, + "loss": 1.1461, + "step": 11933 + }, + { + "epoch": 1.0, + "learning_rate": 1.55069539330746e-05, + "loss": 1.3389, + "step": 11934 + }, + { + "epoch": 1.0, + "learning_rate": 1.550619518480739e-05, + "loss": 1.127, + "step": 11935 + }, + { + "epoch": 1.0, + "learning_rate": 1.5505436391046232e-05, + "loss": 1.206, + "step": 11936 + }, + { + "epoch": 1.0, + "learning_rate": 1.55046775517974e-05, + "loss": 1.1751, + "step": 11937 + }, + { + "epoch": 1.01, + "learning_rate": 1.550391866706716e-05, + "loss": 1.1639, + "step": 11938 + }, + { + "epoch": 1.01, + "learning_rate": 1.550315973686178e-05, + "loss": 1.1823, + "step": 11939 + }, + { + "epoch": 1.01, + "learning_rate": 1.5502400761187533e-05, + "loss": 1.1297, + "step": 11940 + }, + { + "epoch": 1.01, + "learning_rate": 1.550164174005069e-05, + "loss": 1.1952, + "step": 11941 + }, + { + "epoch": 1.01, + "learning_rate": 1.5500882673457526e-05, + "loss": 1.1973, + "step": 11942 + }, + { + "epoch": 1.01, + "learning_rate": 1.5500123561414305e-05, + "loss": 1.1855, + "step": 11943 + }, + { + "epoch": 1.01, + "learning_rate": 1.5499364403927303e-05, + "loss": 1.1144, + "step": 11944 + }, + { + "epoch": 1.01, + "learning_rate": 1.5498605201002793e-05, + "loss": 1.1958, + "step": 11945 + }, + { + "epoch": 1.01, + "learning_rate": 1.5497845952647045e-05, + "loss": 1.2541, + "step": 11946 + }, + { + "epoch": 1.01, + "learning_rate": 1.5497086658866336e-05, + "loss": 1.1854, + "step": 11947 + }, + { + "epoch": 1.01, + "learning_rate": 1.5496327319666938e-05, + "loss": 1.2938, + "step": 11948 + }, + { + "epoch": 1.01, + "learning_rate": 1.5495567935055125e-05, + "loss": 1.2323, + "step": 11949 + }, + { + "epoch": 1.01, + "learning_rate": 1.549480850503717e-05, + "loss": 1.2437, + "step": 11950 + }, + { + "epoch": 1.01, + "learning_rate": 1.5494049029619348e-05, + "loss": 1.1857, + "step": 11951 + }, + { + "epoch": 1.01, + "learning_rate": 1.5493289508807937e-05, + "loss": 1.2077, + "step": 11952 + }, + { + "epoch": 1.01, + "learning_rate": 1.5492529942609208e-05, + "loss": 1.1747, + "step": 11953 + }, + { + "epoch": 1.01, + "learning_rate": 1.549177033102944e-05, + "loss": 1.2572, + "step": 11954 + }, + { + "epoch": 1.01, + "learning_rate": 1.5491010674074906e-05, + "loss": 1.2308, + "step": 11955 + }, + { + "epoch": 1.01, + "learning_rate": 1.5490250971751887e-05, + "loss": 1.2265, + "step": 11956 + }, + { + "epoch": 1.01, + "learning_rate": 1.5489491224066656e-05, + "loss": 1.3027, + "step": 11957 + }, + { + "epoch": 1.01, + "learning_rate": 1.5488731431025493e-05, + "loss": 1.1553, + "step": 11958 + }, + { + "epoch": 1.01, + "learning_rate": 1.548797159263467e-05, + "loss": 1.2444, + "step": 11959 + }, + { + "epoch": 1.01, + "learning_rate": 1.5487211708900474e-05, + "loss": 1.1814, + "step": 11960 + }, + { + "epoch": 1.01, + "learning_rate": 1.5486451779829178e-05, + "loss": 1.2731, + "step": 11961 + }, + { + "epoch": 1.01, + "learning_rate": 1.548569180542706e-05, + "loss": 1.1933, + "step": 11962 + }, + { + "epoch": 1.01, + "learning_rate": 1.5484931785700402e-05, + "loss": 1.286, + "step": 11963 + }, + { + "epoch": 1.01, + "learning_rate": 1.548417172065548e-05, + "loss": 1.1296, + "step": 11964 + }, + { + "epoch": 1.01, + "learning_rate": 1.548341161029858e-05, + "loss": 1.317, + "step": 11965 + }, + { + "epoch": 1.01, + "learning_rate": 1.5482651454635976e-05, + "loss": 1.1978, + "step": 11966 + }, + { + "epoch": 1.01, + "learning_rate": 1.548189125367395e-05, + "loss": 1.2145, + "step": 11967 + }, + { + "epoch": 1.01, + "learning_rate": 1.5481131007418783e-05, + "loss": 1.2557, + "step": 11968 + }, + { + "epoch": 1.01, + "learning_rate": 1.5480370715876763e-05, + "loss": 1.2495, + "step": 11969 + }, + { + "epoch": 1.01, + "learning_rate": 1.5479610379054163e-05, + "loss": 1.2762, + "step": 11970 + }, + { + "epoch": 1.01, + "learning_rate": 1.5478849996957266e-05, + "loss": 1.3135, + "step": 11971 + }, + { + "epoch": 1.01, + "learning_rate": 1.5478089569592357e-05, + "loss": 1.1694, + "step": 11972 + }, + { + "epoch": 1.01, + "learning_rate": 1.5477329096965724e-05, + "loss": 1.246, + "step": 11973 + }, + { + "epoch": 1.01, + "learning_rate": 1.547656857908364e-05, + "loss": 1.1117, + "step": 11974 + }, + { + "epoch": 1.01, + "learning_rate": 1.5475808015952395e-05, + "loss": 1.2096, + "step": 11975 + }, + { + "epoch": 1.01, + "learning_rate": 1.5475047407578273e-05, + "loss": 1.1265, + "step": 11976 + }, + { + "epoch": 1.01, + "learning_rate": 1.5474286753967557e-05, + "loss": 1.1974, + "step": 11977 + }, + { + "epoch": 1.01, + "learning_rate": 1.5473526055126532e-05, + "loss": 1.1843, + "step": 11978 + }, + { + "epoch": 1.01, + "learning_rate": 1.547276531106148e-05, + "loss": 1.2071, + "step": 11979 + }, + { + "epoch": 1.01, + "learning_rate": 1.5472004521778694e-05, + "loss": 1.1913, + "step": 11980 + }, + { + "epoch": 1.01, + "learning_rate": 1.547124368728445e-05, + "loss": 1.1228, + "step": 11981 + }, + { + "epoch": 1.01, + "learning_rate": 1.5470482807585042e-05, + "loss": 1.2024, + "step": 11982 + }, + { + "epoch": 1.01, + "learning_rate": 1.5469721882686756e-05, + "loss": 1.097, + "step": 11983 + }, + { + "epoch": 1.01, + "learning_rate": 1.5468960912595873e-05, + "loss": 1.2219, + "step": 11984 + }, + { + "epoch": 1.01, + "learning_rate": 1.5468199897318687e-05, + "loss": 1.2659, + "step": 11985 + }, + { + "epoch": 1.01, + "learning_rate": 1.5467438836861485e-05, + "loss": 1.244, + "step": 11986 + }, + { + "epoch": 1.01, + "learning_rate": 1.5466677731230552e-05, + "loss": 1.2084, + "step": 11987 + }, + { + "epoch": 1.01, + "learning_rate": 1.546591658043218e-05, + "loss": 1.2785, + "step": 11988 + }, + { + "epoch": 1.01, + "learning_rate": 1.5465155384472655e-05, + "loss": 1.1009, + "step": 11989 + }, + { + "epoch": 1.01, + "learning_rate": 1.5464394143358267e-05, + "loss": 1.1611, + "step": 11990 + }, + { + "epoch": 1.01, + "learning_rate": 1.5463632857095302e-05, + "loss": 1.2274, + "step": 11991 + }, + { + "epoch": 1.01, + "learning_rate": 1.5462871525690058e-05, + "loss": 1.2195, + "step": 11992 + }, + { + "epoch": 1.01, + "learning_rate": 1.546211014914882e-05, + "loss": 1.2645, + "step": 11993 + }, + { + "epoch": 1.01, + "learning_rate": 1.5461348727477884e-05, + "loss": 1.2252, + "step": 11994 + }, + { + "epoch": 1.01, + "learning_rate": 1.546058726068353e-05, + "loss": 1.2111, + "step": 11995 + }, + { + "epoch": 1.01, + "learning_rate": 1.545982574877206e-05, + "loss": 1.1645, + "step": 11996 + }, + { + "epoch": 1.01, + "learning_rate": 1.545906419174976e-05, + "loss": 1.2202, + "step": 11997 + }, + { + "epoch": 1.01, + "learning_rate": 1.5458302589622927e-05, + "loss": 1.2832, + "step": 11998 + }, + { + "epoch": 1.01, + "learning_rate": 1.545754094239785e-05, + "loss": 1.2328, + "step": 11999 + }, + { + "epoch": 1.01, + "learning_rate": 1.5456779250080823e-05, + "loss": 1.2272, + "step": 12000 + }, + { + "epoch": 1.01, + "learning_rate": 1.5456017512678142e-05, + "loss": 1.1928, + "step": 12001 + }, + { + "epoch": 1.01, + "learning_rate": 1.5455255730196097e-05, + "loss": 1.2187, + "step": 12002 + }, + { + "epoch": 1.01, + "learning_rate": 1.5454493902640982e-05, + "loss": 1.1916, + "step": 12003 + }, + { + "epoch": 1.01, + "learning_rate": 1.5453732030019093e-05, + "loss": 1.234, + "step": 12004 + }, + { + "epoch": 1.01, + "learning_rate": 1.5452970112336725e-05, + "loss": 1.3693, + "step": 12005 + }, + { + "epoch": 1.01, + "learning_rate": 1.545220814960017e-05, + "loss": 1.1787, + "step": 12006 + }, + { + "epoch": 1.01, + "learning_rate": 1.545144614181573e-05, + "loss": 1.1627, + "step": 12007 + }, + { + "epoch": 1.01, + "learning_rate": 1.5450684088989695e-05, + "loss": 1.2844, + "step": 12008 + }, + { + "epoch": 1.01, + "learning_rate": 1.5449921991128366e-05, + "loss": 1.1955, + "step": 12009 + }, + { + "epoch": 1.01, + "learning_rate": 1.5449159848238037e-05, + "loss": 1.1617, + "step": 12010 + }, + { + "epoch": 1.01, + "learning_rate": 1.5448397660325003e-05, + "loss": 1.2286, + "step": 12011 + }, + { + "epoch": 1.01, + "learning_rate": 1.5447635427395563e-05, + "loss": 1.2196, + "step": 12012 + }, + { + "epoch": 1.01, + "learning_rate": 1.5446873149456018e-05, + "loss": 1.1764, + "step": 12013 + }, + { + "epoch": 1.01, + "learning_rate": 1.5446110826512664e-05, + "loss": 1.1659, + "step": 12014 + }, + { + "epoch": 1.01, + "learning_rate": 1.54453484585718e-05, + "loss": 1.1467, + "step": 12015 + }, + { + "epoch": 1.01, + "learning_rate": 1.5444586045639724e-05, + "loss": 1.1547, + "step": 12016 + }, + { + "epoch": 1.01, + "learning_rate": 1.5443823587722734e-05, + "loss": 1.0975, + "step": 12017 + }, + { + "epoch": 1.01, + "learning_rate": 1.5443061084827132e-05, + "loss": 1.1916, + "step": 12018 + }, + { + "epoch": 1.01, + "learning_rate": 1.5442298536959218e-05, + "loss": 1.2061, + "step": 12019 + }, + { + "epoch": 1.01, + "learning_rate": 1.544153594412529e-05, + "loss": 1.1352, + "step": 12020 + }, + { + "epoch": 1.01, + "learning_rate": 1.5440773306331653e-05, + "loss": 1.249, + "step": 12021 + }, + { + "epoch": 1.01, + "learning_rate": 1.5440010623584604e-05, + "loss": 1.2675, + "step": 12022 + }, + { + "epoch": 1.01, + "learning_rate": 1.5439247895890446e-05, + "loss": 1.2195, + "step": 12023 + }, + { + "epoch": 1.01, + "learning_rate": 1.543848512325548e-05, + "loss": 1.1521, + "step": 12024 + }, + { + "epoch": 1.01, + "learning_rate": 1.5437722305686015e-05, + "loss": 1.1796, + "step": 12025 + }, + { + "epoch": 1.01, + "learning_rate": 1.5436959443188342e-05, + "loss": 1.1969, + "step": 12026 + }, + { + "epoch": 1.01, + "learning_rate": 1.5436196535768772e-05, + "loss": 1.2303, + "step": 12027 + }, + { + "epoch": 1.01, + "learning_rate": 1.5435433583433605e-05, + "loss": 1.2988, + "step": 12028 + }, + { + "epoch": 1.01, + "learning_rate": 1.543467058618915e-05, + "loss": 1.1953, + "step": 12029 + }, + { + "epoch": 1.01, + "learning_rate": 1.5433907544041704e-05, + "loss": 1.1205, + "step": 12030 + }, + { + "epoch": 1.01, + "learning_rate": 1.5433144456997575e-05, + "loss": 1.2349, + "step": 12031 + }, + { + "epoch": 1.01, + "learning_rate": 1.543238132506307e-05, + "loss": 1.2305, + "step": 12032 + }, + { + "epoch": 1.01, + "learning_rate": 1.543161814824449e-05, + "loss": 1.1902, + "step": 12033 + }, + { + "epoch": 1.01, + "learning_rate": 1.543085492654814e-05, + "loss": 1.2114, + "step": 12034 + }, + { + "epoch": 1.01, + "learning_rate": 1.5430091659980332e-05, + "loss": 1.0986, + "step": 12035 + }, + { + "epoch": 1.01, + "learning_rate": 1.542932834854737e-05, + "loss": 1.2084, + "step": 12036 + }, + { + "epoch": 1.01, + "learning_rate": 1.542856499225556e-05, + "loss": 1.2034, + "step": 12037 + }, + { + "epoch": 1.01, + "learning_rate": 1.5427801591111203e-05, + "loss": 1.1856, + "step": 12038 + }, + { + "epoch": 1.01, + "learning_rate": 1.5427038145120617e-05, + "loss": 1.2326, + "step": 12039 + }, + { + "epoch": 1.01, + "learning_rate": 1.5426274654290103e-05, + "loss": 1.1551, + "step": 12040 + }, + { + "epoch": 1.01, + "learning_rate": 1.542551111862597e-05, + "loss": 1.2877, + "step": 12041 + }, + { + "epoch": 1.01, + "learning_rate": 1.542474753813453e-05, + "loss": 1.1936, + "step": 12042 + }, + { + "epoch": 1.01, + "learning_rate": 1.5423983912822088e-05, + "loss": 1.1501, + "step": 12043 + }, + { + "epoch": 1.01, + "learning_rate": 1.542322024269496e-05, + "loss": 1.2572, + "step": 12044 + }, + { + "epoch": 1.01, + "learning_rate": 1.5422456527759446e-05, + "loss": 1.1781, + "step": 12045 + }, + { + "epoch": 1.01, + "learning_rate": 1.542169276802186e-05, + "loss": 1.2284, + "step": 12046 + }, + { + "epoch": 1.01, + "learning_rate": 1.5420928963488515e-05, + "loss": 1.2714, + "step": 12047 + }, + { + "epoch": 1.01, + "learning_rate": 1.5420165114165722e-05, + "loss": 1.1695, + "step": 12048 + }, + { + "epoch": 1.01, + "learning_rate": 1.5419401220059786e-05, + "loss": 1.2809, + "step": 12049 + }, + { + "epoch": 1.01, + "learning_rate": 1.5418637281177025e-05, + "loss": 1.2533, + "step": 12050 + }, + { + "epoch": 1.01, + "learning_rate": 1.5417873297523748e-05, + "loss": 1.2277, + "step": 12051 + }, + { + "epoch": 1.01, + "learning_rate": 1.5417109269106267e-05, + "loss": 1.2191, + "step": 12052 + }, + { + "epoch": 1.01, + "learning_rate": 1.5416345195930898e-05, + "loss": 1.1789, + "step": 12053 + }, + { + "epoch": 1.01, + "learning_rate": 1.5415581078003952e-05, + "loss": 1.22, + "step": 12054 + }, + { + "epoch": 1.01, + "learning_rate": 1.5414816915331742e-05, + "loss": 1.2155, + "step": 12055 + }, + { + "epoch": 1.02, + "learning_rate": 1.5414052707920573e-05, + "loss": 1.1673, + "step": 12056 + }, + { + "epoch": 1.02, + "learning_rate": 1.5413288455776778e-05, + "loss": 1.2245, + "step": 12057 + }, + { + "epoch": 1.02, + "learning_rate": 1.5412524158906658e-05, + "loss": 1.2566, + "step": 12058 + }, + { + "epoch": 1.02, + "learning_rate": 1.541175981731653e-05, + "loss": 1.2781, + "step": 12059 + }, + { + "epoch": 1.02, + "learning_rate": 1.5410995431012714e-05, + "loss": 1.2106, + "step": 12060 + }, + { + "epoch": 1.02, + "learning_rate": 1.5410231000001516e-05, + "loss": 1.1971, + "step": 12061 + }, + { + "epoch": 1.02, + "learning_rate": 1.5409466524289263e-05, + "loss": 1.2067, + "step": 12062 + }, + { + "epoch": 1.02, + "learning_rate": 1.5408702003882266e-05, + "loss": 1.2068, + "step": 12063 + }, + { + "epoch": 1.02, + "learning_rate": 1.540793743878684e-05, + "loss": 1.3145, + "step": 12064 + }, + { + "epoch": 1.02, + "learning_rate": 1.5407172829009306e-05, + "loss": 1.1196, + "step": 12065 + }, + { + "epoch": 1.02, + "learning_rate": 1.5406408174555978e-05, + "loss": 1.307, + "step": 12066 + }, + { + "epoch": 1.02, + "learning_rate": 1.5405643475433178e-05, + "loss": 1.2161, + "step": 12067 + }, + { + "epoch": 1.02, + "learning_rate": 1.5404878731647218e-05, + "loss": 1.1706, + "step": 12068 + }, + { + "epoch": 1.02, + "learning_rate": 1.540411394320442e-05, + "loss": 1.2014, + "step": 12069 + }, + { + "epoch": 1.02, + "learning_rate": 1.5403349110111104e-05, + "loss": 1.2234, + "step": 12070 + }, + { + "epoch": 1.02, + "learning_rate": 1.5402584232373592e-05, + "loss": 1.1865, + "step": 12071 + }, + { + "epoch": 1.02, + "learning_rate": 1.5401819309998193e-05, + "loss": 1.2085, + "step": 12072 + }, + { + "epoch": 1.02, + "learning_rate": 1.5401054342991238e-05, + "loss": 1.1662, + "step": 12073 + }, + { + "epoch": 1.02, + "learning_rate": 1.5400289331359047e-05, + "loss": 1.1411, + "step": 12074 + }, + { + "epoch": 1.02, + "learning_rate": 1.5399524275107932e-05, + "loss": 1.1542, + "step": 12075 + }, + { + "epoch": 1.02, + "learning_rate": 1.5398759174244222e-05, + "loss": 1.225, + "step": 12076 + }, + { + "epoch": 1.02, + "learning_rate": 1.5397994028774235e-05, + "loss": 1.1593, + "step": 12077 + }, + { + "epoch": 1.02, + "learning_rate": 1.5397228838704295e-05, + "loss": 1.1718, + "step": 12078 + }, + { + "epoch": 1.02, + "learning_rate": 1.539646360404072e-05, + "loss": 1.1771, + "step": 12079 + }, + { + "epoch": 1.02, + "learning_rate": 1.5395698324789837e-05, + "loss": 1.239, + "step": 12080 + }, + { + "epoch": 1.02, + "learning_rate": 1.539493300095797e-05, + "loss": 1.138, + "step": 12081 + }, + { + "epoch": 1.02, + "learning_rate": 1.5394167632551437e-05, + "loss": 1.1424, + "step": 12082 + }, + { + "epoch": 1.02, + "learning_rate": 1.5393402219576564e-05, + "loss": 1.1956, + "step": 12083 + }, + { + "epoch": 1.02, + "learning_rate": 1.539263676203968e-05, + "loss": 1.2444, + "step": 12084 + }, + { + "epoch": 1.02, + "learning_rate": 1.53918712599471e-05, + "loss": 1.1592, + "step": 12085 + }, + { + "epoch": 1.02, + "learning_rate": 1.539110571330516e-05, + "loss": 1.3002, + "step": 12086 + }, + { + "epoch": 1.02, + "learning_rate": 1.5390340122120172e-05, + "loss": 1.2245, + "step": 12087 + }, + { + "epoch": 1.02, + "learning_rate": 1.5389574486398476e-05, + "loss": 1.2554, + "step": 12088 + }, + { + "epoch": 1.02, + "learning_rate": 1.5388808806146385e-05, + "loss": 1.2873, + "step": 12089 + }, + { + "epoch": 1.02, + "learning_rate": 1.5388043081370234e-05, + "loss": 1.2371, + "step": 12090 + }, + { + "epoch": 1.02, + "learning_rate": 1.5387277312076348e-05, + "loss": 1.2039, + "step": 12091 + }, + { + "epoch": 1.02, + "learning_rate": 1.5386511498271047e-05, + "loss": 1.141, + "step": 12092 + }, + { + "epoch": 1.02, + "learning_rate": 1.538574563996067e-05, + "loss": 1.1834, + "step": 12093 + }, + { + "epoch": 1.02, + "learning_rate": 1.5384979737151536e-05, + "loss": 1.223, + "step": 12094 + }, + { + "epoch": 1.02, + "learning_rate": 1.5384213789849976e-05, + "loss": 1.1706, + "step": 12095 + }, + { + "epoch": 1.02, + "learning_rate": 1.538344779806232e-05, + "loss": 1.1468, + "step": 12096 + }, + { + "epoch": 1.02, + "learning_rate": 1.5382681761794892e-05, + "loss": 1.2464, + "step": 12097 + }, + { + "epoch": 1.02, + "learning_rate": 1.5381915681054033e-05, + "loss": 1.0588, + "step": 12098 + }, + { + "epoch": 1.02, + "learning_rate": 1.5381149555846054e-05, + "loss": 1.2544, + "step": 12099 + }, + { + "epoch": 1.02, + "learning_rate": 1.5380383386177302e-05, + "loss": 1.159, + "step": 12100 + }, + { + "epoch": 1.02, + "learning_rate": 1.5379617172054097e-05, + "loss": 1.3008, + "step": 12101 + }, + { + "epoch": 1.02, + "learning_rate": 1.5378850913482776e-05, + "loss": 1.199, + "step": 12102 + }, + { + "epoch": 1.02, + "learning_rate": 1.5378084610469666e-05, + "loss": 1.1801, + "step": 12103 + }, + { + "epoch": 1.02, + "learning_rate": 1.5377318263021096e-05, + "loss": 1.221, + "step": 12104 + }, + { + "epoch": 1.02, + "learning_rate": 1.5376551871143405e-05, + "loss": 1.1905, + "step": 12105 + }, + { + "epoch": 1.02, + "learning_rate": 1.5375785434842922e-05, + "loss": 1.1907, + "step": 12106 + }, + { + "epoch": 1.02, + "learning_rate": 1.537501895412598e-05, + "loss": 1.2456, + "step": 12107 + }, + { + "epoch": 1.02, + "learning_rate": 1.537425242899891e-05, + "loss": 1.1697, + "step": 12108 + }, + { + "epoch": 1.02, + "learning_rate": 1.5373485859468048e-05, + "loss": 1.1546, + "step": 12109 + }, + { + "epoch": 1.02, + "learning_rate": 1.537271924553972e-05, + "loss": 1.1718, + "step": 12110 + }, + { + "epoch": 1.02, + "learning_rate": 1.537195258722027e-05, + "loss": 1.1698, + "step": 12111 + }, + { + "epoch": 1.02, + "learning_rate": 1.537118588451603e-05, + "loss": 1.1841, + "step": 12112 + }, + { + "epoch": 1.02, + "learning_rate": 1.5370419137433335e-05, + "loss": 1.2449, + "step": 12113 + }, + { + "epoch": 1.02, + "learning_rate": 1.5369652345978513e-05, + "loss": 1.0974, + "step": 12114 + }, + { + "epoch": 1.02, + "learning_rate": 1.536888551015791e-05, + "loss": 1.2035, + "step": 12115 + }, + { + "epoch": 1.02, + "learning_rate": 1.5368118629977856e-05, + "loss": 1.2297, + "step": 12116 + }, + { + "epoch": 1.02, + "learning_rate": 1.5367351705444686e-05, + "loss": 1.1428, + "step": 12117 + }, + { + "epoch": 1.02, + "learning_rate": 1.536658473656474e-05, + "loss": 1.1888, + "step": 12118 + }, + { + "epoch": 1.02, + "learning_rate": 1.536581772334435e-05, + "loss": 1.2513, + "step": 12119 + }, + { + "epoch": 1.02, + "learning_rate": 1.536505066578986e-05, + "loss": 1.1851, + "step": 12120 + }, + { + "epoch": 1.02, + "learning_rate": 1.5364283563907602e-05, + "loss": 1.2175, + "step": 12121 + }, + { + "epoch": 1.02, + "learning_rate": 1.5363516417703913e-05, + "loss": 1.2068, + "step": 12122 + }, + { + "epoch": 1.02, + "learning_rate": 1.5362749227185145e-05, + "loss": 1.2453, + "step": 12123 + }, + { + "epoch": 1.02, + "learning_rate": 1.536198199235762e-05, + "loss": 1.1877, + "step": 12124 + }, + { + "epoch": 1.02, + "learning_rate": 1.536121471322768e-05, + "loss": 1.1925, + "step": 12125 + }, + { + "epoch": 1.02, + "learning_rate": 1.5360447389801674e-05, + "loss": 1.1587, + "step": 12126 + }, + { + "epoch": 1.02, + "learning_rate": 1.5359680022085933e-05, + "loss": 1.191, + "step": 12127 + }, + { + "epoch": 1.02, + "learning_rate": 1.5358912610086806e-05, + "loss": 1.2392, + "step": 12128 + }, + { + "epoch": 1.02, + "learning_rate": 1.5358145153810623e-05, + "loss": 1.1574, + "step": 12129 + }, + { + "epoch": 1.02, + "learning_rate": 1.535737765326373e-05, + "loss": 1.2181, + "step": 12130 + }, + { + "epoch": 1.02, + "learning_rate": 1.535661010845247e-05, + "loss": 1.1835, + "step": 12131 + }, + { + "epoch": 1.02, + "learning_rate": 1.5355842519383183e-05, + "loss": 1.1401, + "step": 12132 + }, + { + "epoch": 1.02, + "learning_rate": 1.535507488606221e-05, + "loss": 1.1674, + "step": 12133 + }, + { + "epoch": 1.02, + "learning_rate": 1.535430720849589e-05, + "loss": 1.1321, + "step": 12134 + }, + { + "epoch": 1.02, + "learning_rate": 1.5353539486690577e-05, + "loss": 1.2041, + "step": 12135 + }, + { + "epoch": 1.02, + "learning_rate": 1.5352771720652606e-05, + "loss": 1.2567, + "step": 12136 + }, + { + "epoch": 1.02, + "learning_rate": 1.535200391038832e-05, + "loss": 1.1448, + "step": 12137 + }, + { + "epoch": 1.02, + "learning_rate": 1.5351236055904068e-05, + "loss": 1.1758, + "step": 12138 + }, + { + "epoch": 1.02, + "learning_rate": 1.535046815720619e-05, + "loss": 1.2151, + "step": 12139 + }, + { + "epoch": 1.02, + "learning_rate": 1.534970021430103e-05, + "loss": 1.2403, + "step": 12140 + }, + { + "epoch": 1.02, + "learning_rate": 1.5348932227194933e-05, + "loss": 1.1736, + "step": 12141 + }, + { + "epoch": 1.02, + "learning_rate": 1.5348164195894253e-05, + "loss": 1.2035, + "step": 12142 + }, + { + "epoch": 1.02, + "learning_rate": 1.5347396120405324e-05, + "loss": 1.2121, + "step": 12143 + }, + { + "epoch": 1.02, + "learning_rate": 1.5346628000734496e-05, + "loss": 1.1655, + "step": 12144 + }, + { + "epoch": 1.02, + "learning_rate": 1.5345859836888118e-05, + "loss": 1.2284, + "step": 12145 + }, + { + "epoch": 1.02, + "learning_rate": 1.5345091628872536e-05, + "loss": 1.185, + "step": 12146 + }, + { + "epoch": 1.02, + "learning_rate": 1.5344323376694093e-05, + "loss": 1.2074, + "step": 12147 + }, + { + "epoch": 1.02, + "learning_rate": 1.534355508035914e-05, + "loss": 1.2817, + "step": 12148 + }, + { + "epoch": 1.02, + "learning_rate": 1.534278673987403e-05, + "loss": 1.1329, + "step": 12149 + }, + { + "epoch": 1.02, + "learning_rate": 1.5342018355245102e-05, + "loss": 1.2697, + "step": 12150 + }, + { + "epoch": 1.02, + "learning_rate": 1.5341249926478707e-05, + "loss": 1.0869, + "step": 12151 + }, + { + "epoch": 1.02, + "learning_rate": 1.53404814535812e-05, + "loss": 1.2531, + "step": 12152 + }, + { + "epoch": 1.02, + "learning_rate": 1.5339712936558923e-05, + "loss": 1.1694, + "step": 12153 + }, + { + "epoch": 1.02, + "learning_rate": 1.533894437541823e-05, + "loss": 1.2125, + "step": 12154 + }, + { + "epoch": 1.02, + "learning_rate": 1.533817577016547e-05, + "loss": 1.2118, + "step": 12155 + }, + { + "epoch": 1.02, + "learning_rate": 1.5337407120806992e-05, + "loss": 1.2902, + "step": 12156 + }, + { + "epoch": 1.02, + "learning_rate": 1.533663842734915e-05, + "loss": 1.2822, + "step": 12157 + }, + { + "epoch": 1.02, + "learning_rate": 1.5335869689798294e-05, + "loss": 1.2382, + "step": 12158 + }, + { + "epoch": 1.02, + "learning_rate": 1.5335100908160772e-05, + "loss": 1.1597, + "step": 12159 + }, + { + "epoch": 1.02, + "learning_rate": 1.533433208244294e-05, + "loss": 1.1574, + "step": 12160 + }, + { + "epoch": 1.02, + "learning_rate": 1.533356321265115e-05, + "loss": 1.2169, + "step": 12161 + }, + { + "epoch": 1.02, + "learning_rate": 1.5332794298791753e-05, + "loss": 1.1586, + "step": 12162 + }, + { + "epoch": 1.02, + "learning_rate": 1.5332025340871103e-05, + "loss": 1.1829, + "step": 12163 + }, + { + "epoch": 1.02, + "learning_rate": 1.5331256338895554e-05, + "loss": 1.2043, + "step": 12164 + }, + { + "epoch": 1.02, + "learning_rate": 1.533048729287146e-05, + "loss": 1.2124, + "step": 12165 + }, + { + "epoch": 1.02, + "learning_rate": 1.532971820280517e-05, + "loss": 1.1404, + "step": 12166 + }, + { + "epoch": 1.02, + "learning_rate": 1.5328949068703048e-05, + "loss": 1.2647, + "step": 12167 + }, + { + "epoch": 1.02, + "learning_rate": 1.532817989057144e-05, + "loss": 1.1854, + "step": 12168 + }, + { + "epoch": 1.02, + "learning_rate": 1.5327410668416703e-05, + "loss": 1.2113, + "step": 12169 + }, + { + "epoch": 1.02, + "learning_rate": 1.5326641402245198e-05, + "loss": 1.2394, + "step": 12170 + }, + { + "epoch": 1.02, + "learning_rate": 1.5325872092063274e-05, + "loss": 1.1843, + "step": 12171 + }, + { + "epoch": 1.02, + "learning_rate": 1.5325102737877287e-05, + "loss": 1.1451, + "step": 12172 + }, + { + "epoch": 1.02, + "learning_rate": 1.5324333339693604e-05, + "loss": 1.1759, + "step": 12173 + }, + { + "epoch": 1.02, + "learning_rate": 1.5323563897518573e-05, + "loss": 1.2421, + "step": 12174 + }, + { + "epoch": 1.03, + "learning_rate": 1.532279441135855e-05, + "loss": 1.179, + "step": 12175 + }, + { + "epoch": 1.03, + "learning_rate": 1.5322024881219897e-05, + "loss": 1.2111, + "step": 12176 + }, + { + "epoch": 1.03, + "learning_rate": 1.532125530710897e-05, + "loss": 1.2383, + "step": 12177 + }, + { + "epoch": 1.03, + "learning_rate": 1.5320485689032137e-05, + "loss": 1.1388, + "step": 12178 + }, + { + "epoch": 1.03, + "learning_rate": 1.531971602699574e-05, + "loss": 1.2197, + "step": 12179 + }, + { + "epoch": 1.03, + "learning_rate": 1.531894632100615e-05, + "loss": 1.1392, + "step": 12180 + }, + { + "epoch": 1.03, + "learning_rate": 1.5318176571069724e-05, + "loss": 1.1629, + "step": 12181 + }, + { + "epoch": 1.03, + "learning_rate": 1.5317406777192817e-05, + "loss": 1.2573, + "step": 12182 + }, + { + "epoch": 1.03, + "learning_rate": 1.5316636939381795e-05, + "loss": 1.27, + "step": 12183 + }, + { + "epoch": 1.03, + "learning_rate": 1.5315867057643018e-05, + "loss": 1.2416, + "step": 12184 + }, + { + "epoch": 1.03, + "learning_rate": 1.5315097131982845e-05, + "loss": 1.2647, + "step": 12185 + }, + { + "epoch": 1.03, + "learning_rate": 1.531432716240764e-05, + "loss": 1.2019, + "step": 12186 + }, + { + "epoch": 1.03, + "learning_rate": 1.531355714892376e-05, + "loss": 1.2484, + "step": 12187 + }, + { + "epoch": 1.03, + "learning_rate": 1.5312787091537575e-05, + "loss": 1.1334, + "step": 12188 + }, + { + "epoch": 1.03, + "learning_rate": 1.531201699025544e-05, + "loss": 1.1299, + "step": 12189 + }, + { + "epoch": 1.03, + "learning_rate": 1.5311246845083724e-05, + "loss": 1.1832, + "step": 12190 + }, + { + "epoch": 1.03, + "learning_rate": 1.5310476656028782e-05, + "loss": 1.1986, + "step": 12191 + }, + { + "epoch": 1.03, + "learning_rate": 1.5309706423096986e-05, + "loss": 1.1557, + "step": 12192 + }, + { + "epoch": 1.03, + "learning_rate": 1.5308936146294696e-05, + "loss": 1.2436, + "step": 12193 + }, + { + "epoch": 1.03, + "learning_rate": 1.530816582562827e-05, + "loss": 1.1434, + "step": 12194 + }, + { + "epoch": 1.03, + "learning_rate": 1.5307395461104087e-05, + "loss": 1.2041, + "step": 12195 + }, + { + "epoch": 1.03, + "learning_rate": 1.5306625052728504e-05, + "loss": 1.1672, + "step": 12196 + }, + { + "epoch": 1.03, + "learning_rate": 1.530585460050788e-05, + "loss": 1.1252, + "step": 12197 + }, + { + "epoch": 1.03, + "learning_rate": 1.5305084104448595e-05, + "loss": 1.2132, + "step": 12198 + }, + { + "epoch": 1.03, + "learning_rate": 1.5304313564556998e-05, + "loss": 1.2005, + "step": 12199 + }, + { + "epoch": 1.03, + "learning_rate": 1.530354298083947e-05, + "loss": 1.1783, + "step": 12200 + }, + { + "epoch": 1.03, + "learning_rate": 1.5302772353302374e-05, + "loss": 1.2206, + "step": 12201 + }, + { + "epoch": 1.03, + "learning_rate": 1.5302001681952073e-05, + "loss": 1.1827, + "step": 12202 + }, + { + "epoch": 1.03, + "learning_rate": 1.530123096679494e-05, + "loss": 1.2184, + "step": 12203 + }, + { + "epoch": 1.03, + "learning_rate": 1.530046020783734e-05, + "loss": 1.2155, + "step": 12204 + }, + { + "epoch": 1.03, + "learning_rate": 1.5299689405085638e-05, + "loss": 1.1615, + "step": 12205 + }, + { + "epoch": 1.03, + "learning_rate": 1.5298918558546207e-05, + "loss": 1.2163, + "step": 12206 + }, + { + "epoch": 1.03, + "learning_rate": 1.5298147668225417e-05, + "loss": 1.1449, + "step": 12207 + }, + { + "epoch": 1.03, + "learning_rate": 1.5297376734129633e-05, + "loss": 1.2705, + "step": 12208 + }, + { + "epoch": 1.03, + "learning_rate": 1.5296605756265228e-05, + "loss": 1.1531, + "step": 12209 + }, + { + "epoch": 1.03, + "learning_rate": 1.5295834734638573e-05, + "loss": 1.2358, + "step": 12210 + }, + { + "epoch": 1.03, + "learning_rate": 1.5295063669256033e-05, + "loss": 1.2067, + "step": 12211 + }, + { + "epoch": 1.03, + "learning_rate": 1.5294292560123984e-05, + "loss": 1.1354, + "step": 12212 + }, + { + "epoch": 1.03, + "learning_rate": 1.52935214072488e-05, + "loss": 1.2473, + "step": 12213 + }, + { + "epoch": 1.03, + "learning_rate": 1.529275021063684e-05, + "loss": 1.287, + "step": 12214 + }, + { + "epoch": 1.03, + "learning_rate": 1.529197897029449e-05, + "loss": 1.1678, + "step": 12215 + }, + { + "epoch": 1.03, + "learning_rate": 1.5291207686228112e-05, + "loss": 1.1991, + "step": 12216 + }, + { + "epoch": 1.03, + "learning_rate": 1.5290436358444086e-05, + "loss": 1.1845, + "step": 12217 + }, + { + "epoch": 1.03, + "learning_rate": 1.528966498694878e-05, + "loss": 1.2667, + "step": 12218 + }, + { + "epoch": 1.03, + "learning_rate": 1.528889357174857e-05, + "loss": 1.2059, + "step": 12219 + }, + { + "epoch": 1.03, + "learning_rate": 1.5288122112849827e-05, + "loss": 1.0964, + "step": 12220 + }, + { + "epoch": 1.03, + "learning_rate": 1.5287350610258926e-05, + "loss": 1.1916, + "step": 12221 + }, + { + "epoch": 1.03, + "learning_rate": 1.5286579063982243e-05, + "loss": 1.2889, + "step": 12222 + }, + { + "epoch": 1.03, + "learning_rate": 1.5285807474026154e-05, + "loss": 1.1382, + "step": 12223 + }, + { + "epoch": 1.03, + "learning_rate": 1.528503584039703e-05, + "loss": 1.2828, + "step": 12224 + }, + { + "epoch": 1.03, + "learning_rate": 1.528426416310125e-05, + "loss": 1.2269, + "step": 12225 + }, + { + "epoch": 1.03, + "learning_rate": 1.5283492442145186e-05, + "loss": 1.1993, + "step": 12226 + }, + { + "epoch": 1.03, + "learning_rate": 1.5282720677535217e-05, + "loss": 1.237, + "step": 12227 + }, + { + "epoch": 1.03, + "learning_rate": 1.5281948869277718e-05, + "loss": 1.252, + "step": 12228 + }, + { + "epoch": 1.03, + "learning_rate": 1.528117701737907e-05, + "loss": 1.1174, + "step": 12229 + }, + { + "epoch": 1.03, + "learning_rate": 1.5280405121845646e-05, + "loss": 1.1309, + "step": 12230 + }, + { + "epoch": 1.03, + "learning_rate": 1.5279633182683826e-05, + "loss": 1.3607, + "step": 12231 + }, + { + "epoch": 1.03, + "learning_rate": 1.5278861199899986e-05, + "loss": 1.2352, + "step": 12232 + }, + { + "epoch": 1.03, + "learning_rate": 1.5278089173500505e-05, + "loss": 1.286, + "step": 12233 + }, + { + "epoch": 1.03, + "learning_rate": 1.527731710349176e-05, + "loss": 1.2315, + "step": 12234 + }, + { + "epoch": 1.03, + "learning_rate": 1.5276544989880133e-05, + "loss": 1.1944, + "step": 12235 + }, + { + "epoch": 1.03, + "learning_rate": 1.5275772832672005e-05, + "loss": 1.2855, + "step": 12236 + }, + { + "epoch": 1.03, + "learning_rate": 1.5275000631873753e-05, + "loss": 1.2146, + "step": 12237 + }, + { + "epoch": 1.03, + "learning_rate": 1.5274228387491756e-05, + "loss": 1.2339, + "step": 12238 + }, + { + "epoch": 1.03, + "learning_rate": 1.5273456099532397e-05, + "loss": 1.2852, + "step": 12239 + }, + { + "epoch": 1.03, + "learning_rate": 1.5272683768002056e-05, + "loss": 1.1213, + "step": 12240 + }, + { + "epoch": 1.03, + "learning_rate": 1.5271911392907114e-05, + "loss": 1.2078, + "step": 12241 + }, + { + "epoch": 1.03, + "learning_rate": 1.5271138974253953e-05, + "loss": 1.2202, + "step": 12242 + }, + { + "epoch": 1.03, + "learning_rate": 1.5270366512048956e-05, + "loss": 1.2364, + "step": 12243 + }, + { + "epoch": 1.03, + "learning_rate": 1.52695940062985e-05, + "loss": 1.1883, + "step": 12244 + }, + { + "epoch": 1.03, + "learning_rate": 1.526882145700898e-05, + "loss": 1.1976, + "step": 12245 + }, + { + "epoch": 1.03, + "learning_rate": 1.5268048864186765e-05, + "loss": 1.102, + "step": 12246 + }, + { + "epoch": 1.03, + "learning_rate": 1.5267276227838242e-05, + "loss": 1.1748, + "step": 12247 + }, + { + "epoch": 1.03, + "learning_rate": 1.52665035479698e-05, + "loss": 1.2414, + "step": 12248 + }, + { + "epoch": 1.03, + "learning_rate": 1.526573082458782e-05, + "loss": 1.1941, + "step": 12249 + }, + { + "epoch": 1.03, + "learning_rate": 1.5264958057698687e-05, + "loss": 1.2652, + "step": 12250 + }, + { + "epoch": 1.03, + "learning_rate": 1.5264185247308783e-05, + "loss": 1.2536, + "step": 12251 + }, + { + "epoch": 1.03, + "learning_rate": 1.5263412393424503e-05, + "loss": 1.211, + "step": 12252 + }, + { + "epoch": 1.03, + "learning_rate": 1.526263949605222e-05, + "loss": 1.2308, + "step": 12253 + }, + { + "epoch": 1.03, + "learning_rate": 1.5261866555198324e-05, + "loss": 1.1361, + "step": 12254 + }, + { + "epoch": 1.03, + "learning_rate": 1.5261093570869205e-05, + "loss": 1.2659, + "step": 12255 + }, + { + "epoch": 1.03, + "learning_rate": 1.5260320543071246e-05, + "loss": 1.3102, + "step": 12256 + }, + { + "epoch": 1.03, + "learning_rate": 1.5259547471810837e-05, + "loss": 1.1866, + "step": 12257 + }, + { + "epoch": 1.03, + "learning_rate": 1.5258774357094362e-05, + "loss": 1.1488, + "step": 12258 + }, + { + "epoch": 1.03, + "learning_rate": 1.525800119892821e-05, + "loss": 1.0721, + "step": 12259 + }, + { + "epoch": 1.03, + "learning_rate": 1.5257227997318768e-05, + "loss": 1.1857, + "step": 12260 + }, + { + "epoch": 1.03, + "learning_rate": 1.5256454752272425e-05, + "loss": 1.1761, + "step": 12261 + }, + { + "epoch": 1.03, + "learning_rate": 1.5255681463795574e-05, + "loss": 1.1812, + "step": 12262 + }, + { + "epoch": 1.03, + "learning_rate": 1.5254908131894599e-05, + "loss": 1.2434, + "step": 12263 + }, + { + "epoch": 1.03, + "learning_rate": 1.5254134756575893e-05, + "loss": 1.2134, + "step": 12264 + }, + { + "epoch": 1.03, + "learning_rate": 1.5253361337845842e-05, + "loss": 1.1925, + "step": 12265 + }, + { + "epoch": 1.03, + "learning_rate": 1.525258787571084e-05, + "loss": 1.1592, + "step": 12266 + }, + { + "epoch": 1.03, + "learning_rate": 1.5251814370177276e-05, + "loss": 1.1986, + "step": 12267 + }, + { + "epoch": 1.03, + "learning_rate": 1.5251040821251539e-05, + "loss": 1.2045, + "step": 12268 + }, + { + "epoch": 1.03, + "learning_rate": 1.5250267228940024e-05, + "loss": 1.1707, + "step": 12269 + }, + { + "epoch": 1.03, + "learning_rate": 1.5249493593249121e-05, + "loss": 1.0981, + "step": 12270 + }, + { + "epoch": 1.03, + "learning_rate": 1.5248719914185222e-05, + "loss": 1.2163, + "step": 12271 + }, + { + "epoch": 1.03, + "learning_rate": 1.5247946191754715e-05, + "loss": 1.2305, + "step": 12272 + }, + { + "epoch": 1.03, + "learning_rate": 1.5247172425964003e-05, + "loss": 1.0923, + "step": 12273 + }, + { + "epoch": 1.03, + "learning_rate": 1.5246398616819468e-05, + "loss": 1.2075, + "step": 12274 + }, + { + "epoch": 1.03, + "learning_rate": 1.5245624764327511e-05, + "loss": 1.1168, + "step": 12275 + }, + { + "epoch": 1.03, + "learning_rate": 1.5244850868494523e-05, + "loss": 1.2119, + "step": 12276 + }, + { + "epoch": 1.03, + "learning_rate": 1.52440769293269e-05, + "loss": 1.2128, + "step": 12277 + }, + { + "epoch": 1.03, + "learning_rate": 1.5243302946831034e-05, + "loss": 1.1672, + "step": 12278 + }, + { + "epoch": 1.03, + "learning_rate": 1.524252892101332e-05, + "loss": 1.1539, + "step": 12279 + }, + { + "epoch": 1.03, + "learning_rate": 1.5241754851880155e-05, + "loss": 1.2157, + "step": 12280 + }, + { + "epoch": 1.03, + "learning_rate": 1.5240980739437934e-05, + "loss": 1.1324, + "step": 12281 + }, + { + "epoch": 1.03, + "learning_rate": 1.5240206583693053e-05, + "loss": 1.2675, + "step": 12282 + }, + { + "epoch": 1.03, + "learning_rate": 1.523943238465191e-05, + "loss": 1.2365, + "step": 12283 + }, + { + "epoch": 1.03, + "learning_rate": 1.5238658142320897e-05, + "loss": 1.1988, + "step": 12284 + }, + { + "epoch": 1.03, + "learning_rate": 1.5237883856706413e-05, + "loss": 1.1644, + "step": 12285 + }, + { + "epoch": 1.03, + "learning_rate": 1.5237109527814857e-05, + "loss": 1.1638, + "step": 12286 + }, + { + "epoch": 1.03, + "learning_rate": 1.5236335155652627e-05, + "loss": 1.0737, + "step": 12287 + }, + { + "epoch": 1.03, + "learning_rate": 1.523556074022612e-05, + "loss": 1.1467, + "step": 12288 + }, + { + "epoch": 1.03, + "learning_rate": 1.5234786281541736e-05, + "loss": 1.1423, + "step": 12289 + }, + { + "epoch": 1.03, + "learning_rate": 1.523401177960587e-05, + "loss": 1.2583, + "step": 12290 + }, + { + "epoch": 1.03, + "learning_rate": 1.5233237234424923e-05, + "loss": 1.2775, + "step": 12291 + }, + { + "epoch": 1.03, + "learning_rate": 1.52324626460053e-05, + "loss": 1.1928, + "step": 12292 + }, + { + "epoch": 1.03, + "learning_rate": 1.5231688014353392e-05, + "loss": 1.1892, + "step": 12293 + }, + { + "epoch": 1.04, + "learning_rate": 1.5230913339475603e-05, + "loss": 1.1463, + "step": 12294 + }, + { + "epoch": 1.04, + "learning_rate": 1.5230138621378334e-05, + "loss": 1.1665, + "step": 12295 + }, + { + "epoch": 1.04, + "learning_rate": 1.5229363860067987e-05, + "loss": 1.2471, + "step": 12296 + }, + { + "epoch": 1.04, + "learning_rate": 1.522858905555096e-05, + "loss": 1.1935, + "step": 12297 + }, + { + "epoch": 1.04, + "learning_rate": 1.522781420783366e-05, + "loss": 1.2141, + "step": 12298 + }, + { + "epoch": 1.04, + "learning_rate": 1.5227039316922483e-05, + "loss": 1.2017, + "step": 12299 + }, + { + "epoch": 1.04, + "learning_rate": 1.5226264382823837e-05, + "loss": 1.1457, + "step": 12300 + }, + { + "epoch": 1.04, + "learning_rate": 1.522548940554412e-05, + "loss": 1.217, + "step": 12301 + }, + { + "epoch": 1.04, + "learning_rate": 1.522471438508974e-05, + "loss": 1.2371, + "step": 12302 + }, + { + "epoch": 1.04, + "learning_rate": 1.5223939321467094e-05, + "loss": 1.1345, + "step": 12303 + }, + { + "epoch": 1.04, + "learning_rate": 1.5223164214682591e-05, + "loss": 1.1694, + "step": 12304 + }, + { + "epoch": 1.04, + "learning_rate": 1.5222389064742635e-05, + "loss": 1.2673, + "step": 12305 + }, + { + "epoch": 1.04, + "learning_rate": 1.522161387165363e-05, + "loss": 1.1389, + "step": 12306 + }, + { + "epoch": 1.04, + "learning_rate": 1.5220838635421978e-05, + "loss": 1.1531, + "step": 12307 + }, + { + "epoch": 1.04, + "learning_rate": 1.5220063356054083e-05, + "loss": 1.2329, + "step": 12308 + }, + { + "epoch": 1.04, + "learning_rate": 1.521928803355636e-05, + "loss": 1.2059, + "step": 12309 + }, + { + "epoch": 1.04, + "learning_rate": 1.5218512667935204e-05, + "loss": 1.1454, + "step": 12310 + }, + { + "epoch": 1.04, + "learning_rate": 1.5217737259197028e-05, + "loss": 1.3009, + "step": 12311 + }, + { + "epoch": 1.04, + "learning_rate": 1.5216961807348236e-05, + "loss": 1.2633, + "step": 12312 + }, + { + "epoch": 1.04, + "learning_rate": 1.5216186312395238e-05, + "loss": 1.1839, + "step": 12313 + }, + { + "epoch": 1.04, + "learning_rate": 1.521541077434444e-05, + "loss": 1.2368, + "step": 12314 + }, + { + "epoch": 1.04, + "learning_rate": 1.5214635193202244e-05, + "loss": 1.2547, + "step": 12315 + }, + { + "epoch": 1.04, + "learning_rate": 1.5213859568975066e-05, + "loss": 1.1496, + "step": 12316 + }, + { + "epoch": 1.04, + "learning_rate": 1.5213083901669317e-05, + "loss": 1.1866, + "step": 12317 + }, + { + "epoch": 1.04, + "learning_rate": 1.5212308191291394e-05, + "loss": 1.1537, + "step": 12318 + }, + { + "epoch": 1.04, + "learning_rate": 1.5211532437847711e-05, + "loss": 1.1474, + "step": 12319 + }, + { + "epoch": 1.04, + "learning_rate": 1.5210756641344684e-05, + "loss": 1.2717, + "step": 12320 + }, + { + "epoch": 1.04, + "learning_rate": 1.5209980801788717e-05, + "loss": 1.2414, + "step": 12321 + }, + { + "epoch": 1.04, + "learning_rate": 1.5209204919186218e-05, + "loss": 1.1626, + "step": 12322 + }, + { + "epoch": 1.04, + "learning_rate": 1.5208428993543604e-05, + "loss": 1.1896, + "step": 12323 + }, + { + "epoch": 1.04, + "learning_rate": 1.5207653024867281e-05, + "loss": 1.2057, + "step": 12324 + }, + { + "epoch": 1.04, + "learning_rate": 1.5206877013163665e-05, + "loss": 1.0987, + "step": 12325 + }, + { + "epoch": 1.04, + "learning_rate": 1.520610095843916e-05, + "loss": 1.1176, + "step": 12326 + }, + { + "epoch": 1.04, + "learning_rate": 1.5205324860700189e-05, + "loss": 1.289, + "step": 12327 + }, + { + "epoch": 1.04, + "learning_rate": 1.5204548719953154e-05, + "loss": 1.1452, + "step": 12328 + }, + { + "epoch": 1.04, + "learning_rate": 1.5203772536204472e-05, + "loss": 1.1533, + "step": 12329 + }, + { + "epoch": 1.04, + "learning_rate": 1.5202996309460556e-05, + "loss": 1.2716, + "step": 12330 + }, + { + "epoch": 1.04, + "learning_rate": 1.520222003972782e-05, + "loss": 1.2634, + "step": 12331 + }, + { + "epoch": 1.04, + "learning_rate": 1.5201443727012674e-05, + "loss": 1.2155, + "step": 12332 + }, + { + "epoch": 1.04, + "learning_rate": 1.520066737132154e-05, + "loss": 1.1577, + "step": 12333 + }, + { + "epoch": 1.04, + "learning_rate": 1.5199890972660824e-05, + "loss": 1.1662, + "step": 12334 + }, + { + "epoch": 1.04, + "learning_rate": 1.5199114531036947e-05, + "loss": 1.1501, + "step": 12335 + }, + { + "epoch": 1.04, + "learning_rate": 1.519833804645632e-05, + "loss": 1.1801, + "step": 12336 + }, + { + "epoch": 1.04, + "learning_rate": 1.5197561518925362e-05, + "loss": 1.1885, + "step": 12337 + }, + { + "epoch": 1.04, + "learning_rate": 1.5196784948450487e-05, + "loss": 1.1143, + "step": 12338 + }, + { + "epoch": 1.04, + "learning_rate": 1.519600833503811e-05, + "loss": 1.242, + "step": 12339 + }, + { + "epoch": 1.04, + "learning_rate": 1.519523167869465e-05, + "loss": 1.2021, + "step": 12340 + }, + { + "epoch": 1.04, + "learning_rate": 1.5194454979426522e-05, + "loss": 1.1622, + "step": 12341 + }, + { + "epoch": 1.04, + "learning_rate": 1.519367823724015e-05, + "loss": 1.1836, + "step": 12342 + }, + { + "epoch": 1.04, + "learning_rate": 1.5192901452141941e-05, + "loss": 1.3135, + "step": 12343 + }, + { + "epoch": 1.04, + "learning_rate": 1.519212462413832e-05, + "loss": 1.1772, + "step": 12344 + }, + { + "epoch": 1.04, + "learning_rate": 1.5191347753235703e-05, + "loss": 1.2027, + "step": 12345 + }, + { + "epoch": 1.04, + "learning_rate": 1.5190570839440511e-05, + "loss": 1.1848, + "step": 12346 + }, + { + "epoch": 1.04, + "learning_rate": 1.5189793882759158e-05, + "loss": 1.2684, + "step": 12347 + }, + { + "epoch": 1.04, + "learning_rate": 1.518901688319807e-05, + "loss": 1.2155, + "step": 12348 + }, + { + "epoch": 1.04, + "learning_rate": 1.5188239840763663e-05, + "loss": 1.2439, + "step": 12349 + }, + { + "epoch": 1.04, + "learning_rate": 1.518746275546236e-05, + "loss": 1.1526, + "step": 12350 + }, + { + "epoch": 1.04, + "learning_rate": 1.5186685627300577e-05, + "loss": 1.2112, + "step": 12351 + }, + { + "epoch": 1.04, + "learning_rate": 1.5185908456284738e-05, + "loss": 1.1204, + "step": 12352 + }, + { + "epoch": 1.04, + "learning_rate": 1.5185131242421263e-05, + "loss": 1.1615, + "step": 12353 + }, + { + "epoch": 1.04, + "learning_rate": 1.5184353985716574e-05, + "loss": 1.1633, + "step": 12354 + }, + { + "epoch": 1.04, + "learning_rate": 1.5183576686177094e-05, + "loss": 1.1915, + "step": 12355 + }, + { + "epoch": 1.04, + "learning_rate": 1.5182799343809246e-05, + "loss": 1.1944, + "step": 12356 + }, + { + "epoch": 1.04, + "learning_rate": 1.518202195861945e-05, + "loss": 1.2646, + "step": 12357 + }, + { + "epoch": 1.04, + "learning_rate": 1.5181244530614127e-05, + "loss": 1.2153, + "step": 12358 + }, + { + "epoch": 1.04, + "learning_rate": 1.5180467059799706e-05, + "loss": 1.2518, + "step": 12359 + }, + { + "epoch": 1.04, + "learning_rate": 1.5179689546182607e-05, + "loss": 1.2053, + "step": 12360 + }, + { + "epoch": 1.04, + "learning_rate": 1.5178911989769259e-05, + "loss": 1.1706, + "step": 12361 + }, + { + "epoch": 1.04, + "learning_rate": 1.517813439056608e-05, + "loss": 1.1877, + "step": 12362 + }, + { + "epoch": 1.04, + "learning_rate": 1.5177356748579497e-05, + "loss": 1.0884, + "step": 12363 + }, + { + "epoch": 1.04, + "learning_rate": 1.5176579063815937e-05, + "loss": 1.1897, + "step": 12364 + }, + { + "epoch": 1.04, + "learning_rate": 1.5175801336281822e-05, + "loss": 1.2333, + "step": 12365 + }, + { + "epoch": 1.04, + "learning_rate": 1.5175023565983583e-05, + "loss": 1.1969, + "step": 12366 + }, + { + "epoch": 1.04, + "learning_rate": 1.517424575292764e-05, + "loss": 1.1337, + "step": 12367 + }, + { + "epoch": 1.04, + "learning_rate": 1.5173467897120424e-05, + "loss": 1.2061, + "step": 12368 + }, + { + "epoch": 1.04, + "learning_rate": 1.5172689998568362e-05, + "loss": 1.2819, + "step": 12369 + }, + { + "epoch": 1.04, + "learning_rate": 1.5171912057277878e-05, + "loss": 1.2296, + "step": 12370 + }, + { + "epoch": 1.04, + "learning_rate": 1.51711340732554e-05, + "loss": 1.1527, + "step": 12371 + }, + { + "epoch": 1.04, + "learning_rate": 1.5170356046507358e-05, + "loss": 1.1711, + "step": 12372 + }, + { + "epoch": 1.04, + "learning_rate": 1.5169577977040182e-05, + "loss": 1.1801, + "step": 12373 + }, + { + "epoch": 1.04, + "learning_rate": 1.5168799864860297e-05, + "loss": 1.2272, + "step": 12374 + }, + { + "epoch": 1.04, + "learning_rate": 1.5168021709974133e-05, + "loss": 1.2101, + "step": 12375 + }, + { + "epoch": 1.04, + "learning_rate": 1.516724351238812e-05, + "loss": 1.1933, + "step": 12376 + }, + { + "epoch": 1.04, + "learning_rate": 1.5166465272108686e-05, + "loss": 1.3418, + "step": 12377 + }, + { + "epoch": 1.04, + "learning_rate": 1.5165686989142265e-05, + "loss": 1.2406, + "step": 12378 + }, + { + "epoch": 1.04, + "learning_rate": 1.5164908663495283e-05, + "loss": 1.1887, + "step": 12379 + }, + { + "epoch": 1.04, + "learning_rate": 1.5164130295174172e-05, + "loss": 1.1992, + "step": 12380 + }, + { + "epoch": 1.04, + "learning_rate": 1.5163351884185366e-05, + "loss": 1.1742, + "step": 12381 + }, + { + "epoch": 1.04, + "learning_rate": 1.5162573430535292e-05, + "loss": 1.1966, + "step": 12382 + }, + { + "epoch": 1.04, + "learning_rate": 1.5161794934230388e-05, + "loss": 1.2171, + "step": 12383 + }, + { + "epoch": 1.04, + "learning_rate": 1.5161016395277078e-05, + "loss": 1.2027, + "step": 12384 + }, + { + "epoch": 1.04, + "learning_rate": 1.5160237813681798e-05, + "loss": 1.1949, + "step": 12385 + }, + { + "epoch": 1.04, + "learning_rate": 1.5159459189450983e-05, + "loss": 1.2101, + "step": 12386 + }, + { + "epoch": 1.04, + "learning_rate": 1.5158680522591068e-05, + "loss": 1.2044, + "step": 12387 + }, + { + "epoch": 1.04, + "learning_rate": 1.5157901813108482e-05, + "loss": 1.2029, + "step": 12388 + }, + { + "epoch": 1.04, + "learning_rate": 1.5157123061009658e-05, + "loss": 1.1727, + "step": 12389 + }, + { + "epoch": 1.04, + "learning_rate": 1.5156344266301036e-05, + "loss": 1.2338, + "step": 12390 + }, + { + "epoch": 1.04, + "learning_rate": 1.5155565428989044e-05, + "loss": 1.2761, + "step": 12391 + }, + { + "epoch": 1.04, + "learning_rate": 1.5154786549080125e-05, + "loss": 1.2339, + "step": 12392 + }, + { + "epoch": 1.04, + "learning_rate": 1.5154007626580707e-05, + "loss": 1.1891, + "step": 12393 + }, + { + "epoch": 1.04, + "learning_rate": 1.5153228661497228e-05, + "loss": 1.2367, + "step": 12394 + }, + { + "epoch": 1.04, + "learning_rate": 1.5152449653836123e-05, + "loss": 1.1366, + "step": 12395 + }, + { + "epoch": 1.04, + "learning_rate": 1.5151670603603834e-05, + "loss": 1.13, + "step": 12396 + }, + { + "epoch": 1.04, + "learning_rate": 1.5150891510806791e-05, + "loss": 1.1101, + "step": 12397 + }, + { + "epoch": 1.04, + "learning_rate": 1.5150112375451435e-05, + "loss": 1.1603, + "step": 12398 + }, + { + "epoch": 1.04, + "learning_rate": 1.5149333197544203e-05, + "loss": 1.2359, + "step": 12399 + }, + { + "epoch": 1.04, + "learning_rate": 1.5148553977091532e-05, + "loss": 1.2363, + "step": 12400 + }, + { + "epoch": 1.04, + "learning_rate": 1.5147774714099859e-05, + "loss": 1.2025, + "step": 12401 + }, + { + "epoch": 1.04, + "learning_rate": 1.5146995408575626e-05, + "loss": 1.1838, + "step": 12402 + }, + { + "epoch": 1.04, + "learning_rate": 1.5146216060525267e-05, + "loss": 1.1538, + "step": 12403 + }, + { + "epoch": 1.04, + "learning_rate": 1.5145436669955225e-05, + "loss": 1.1752, + "step": 12404 + }, + { + "epoch": 1.04, + "learning_rate": 1.5144657236871939e-05, + "loss": 1.125, + "step": 12405 + }, + { + "epoch": 1.04, + "learning_rate": 1.5143877761281847e-05, + "loss": 1.2123, + "step": 12406 + }, + { + "epoch": 1.04, + "learning_rate": 1.5143098243191393e-05, + "loss": 1.2211, + "step": 12407 + }, + { + "epoch": 1.04, + "learning_rate": 1.5142318682607013e-05, + "loss": 1.2274, + "step": 12408 + }, + { + "epoch": 1.04, + "learning_rate": 1.5141539079535151e-05, + "loss": 1.2391, + "step": 12409 + }, + { + "epoch": 1.04, + "learning_rate": 1.5140759433982245e-05, + "loss": 1.2139, + "step": 12410 + }, + { + "epoch": 1.04, + "learning_rate": 1.5139979745954742e-05, + "loss": 1.215, + "step": 12411 + }, + { + "epoch": 1.04, + "learning_rate": 1.513920001545908e-05, + "loss": 1.2389, + "step": 12412 + }, + { + "epoch": 1.05, + "learning_rate": 1.5138420242501708e-05, + "loss": 1.1648, + "step": 12413 + }, + { + "epoch": 1.05, + "learning_rate": 1.513764042708906e-05, + "loss": 1.1275, + "step": 12414 + }, + { + "epoch": 1.05, + "learning_rate": 1.513686056922758e-05, + "loss": 1.1661, + "step": 12415 + }, + { + "epoch": 1.05, + "learning_rate": 1.5136080668923715e-05, + "loss": 1.1754, + "step": 12416 + }, + { + "epoch": 1.05, + "learning_rate": 1.5135300726183913e-05, + "loss": 1.1302, + "step": 12417 + }, + { + "epoch": 1.05, + "learning_rate": 1.5134520741014607e-05, + "loss": 1.1636, + "step": 12418 + }, + { + "epoch": 1.05, + "learning_rate": 1.513374071342225e-05, + "loss": 1.224, + "step": 12419 + }, + { + "epoch": 1.05, + "learning_rate": 1.5132960643413282e-05, + "loss": 1.2438, + "step": 12420 + }, + { + "epoch": 1.05, + "learning_rate": 1.5132180530994149e-05, + "loss": 1.1487, + "step": 12421 + }, + { + "epoch": 1.05, + "learning_rate": 1.5131400376171303e-05, + "loss": 1.2645, + "step": 12422 + }, + { + "epoch": 1.05, + "learning_rate": 1.513062017895118e-05, + "loss": 1.1597, + "step": 12423 + }, + { + "epoch": 1.05, + "learning_rate": 1.5129839939340237e-05, + "loss": 1.1641, + "step": 12424 + }, + { + "epoch": 1.05, + "learning_rate": 1.5129059657344911e-05, + "loss": 1.1639, + "step": 12425 + }, + { + "epoch": 1.05, + "learning_rate": 1.5128279332971651e-05, + "loss": 1.2788, + "step": 12426 + }, + { + "epoch": 1.05, + "learning_rate": 1.512749896622691e-05, + "loss": 1.257, + "step": 12427 + }, + { + "epoch": 1.05, + "learning_rate": 1.5126718557117128e-05, + "loss": 1.18, + "step": 12428 + }, + { + "epoch": 1.05, + "learning_rate": 1.5125938105648759e-05, + "loss": 1.218, + "step": 12429 + }, + { + "epoch": 1.05, + "learning_rate": 1.5125157611828245e-05, + "loss": 1.2385, + "step": 12430 + }, + { + "epoch": 1.05, + "learning_rate": 1.5124377075662044e-05, + "loss": 1.1849, + "step": 12431 + }, + { + "epoch": 1.05, + "learning_rate": 1.5123596497156596e-05, + "loss": 1.1225, + "step": 12432 + }, + { + "epoch": 1.05, + "learning_rate": 1.5122815876318355e-05, + "loss": 1.2361, + "step": 12433 + }, + { + "epoch": 1.05, + "learning_rate": 1.512203521315377e-05, + "loss": 1.2612, + "step": 12434 + }, + { + "epoch": 1.05, + "learning_rate": 1.512125450766929e-05, + "loss": 1.1825, + "step": 12435 + }, + { + "epoch": 1.05, + "learning_rate": 1.5120473759871366e-05, + "loss": 1.1429, + "step": 12436 + }, + { + "epoch": 1.05, + "learning_rate": 1.511969296976645e-05, + "loss": 1.2512, + "step": 12437 + }, + { + "epoch": 1.05, + "learning_rate": 1.511891213736099e-05, + "loss": 1.2157, + "step": 12438 + }, + { + "epoch": 1.05, + "learning_rate": 1.5118131262661444e-05, + "loss": 1.177, + "step": 12439 + }, + { + "epoch": 1.05, + "learning_rate": 1.5117350345674256e-05, + "loss": 1.1499, + "step": 12440 + }, + { + "epoch": 1.05, + "learning_rate": 1.5116569386405882e-05, + "loss": 1.2526, + "step": 12441 + }, + { + "epoch": 1.05, + "learning_rate": 1.5115788384862775e-05, + "loss": 1.2005, + "step": 12442 + }, + { + "epoch": 1.05, + "learning_rate": 1.5115007341051388e-05, + "loss": 1.1455, + "step": 12443 + }, + { + "epoch": 1.05, + "learning_rate": 1.5114226254978172e-05, + "loss": 1.2507, + "step": 12444 + }, + { + "epoch": 1.05, + "learning_rate": 1.5113445126649579e-05, + "loss": 1.2202, + "step": 12445 + }, + { + "epoch": 1.05, + "learning_rate": 1.5112663956072072e-05, + "loss": 1.1023, + "step": 12446 + }, + { + "epoch": 1.05, + "learning_rate": 1.5111882743252095e-05, + "loss": 1.2906, + "step": 12447 + }, + { + "epoch": 1.05, + "learning_rate": 1.5111101488196107e-05, + "loss": 1.1848, + "step": 12448 + }, + { + "epoch": 1.05, + "learning_rate": 1.5110320190910565e-05, + "loss": 1.2136, + "step": 12449 + }, + { + "epoch": 1.05, + "learning_rate": 1.5109538851401922e-05, + "loss": 1.1899, + "step": 12450 + }, + { + "epoch": 1.05, + "learning_rate": 1.510875746967663e-05, + "loss": 1.1598, + "step": 12451 + }, + { + "epoch": 1.05, + "learning_rate": 1.5107976045741152e-05, + "loss": 1.1789, + "step": 12452 + }, + { + "epoch": 1.05, + "learning_rate": 1.5107194579601942e-05, + "loss": 1.2245, + "step": 12453 + }, + { + "epoch": 1.05, + "learning_rate": 1.5106413071265453e-05, + "loss": 1.204, + "step": 12454 + }, + { + "epoch": 1.05, + "learning_rate": 1.5105631520738144e-05, + "loss": 1.1199, + "step": 12455 + }, + { + "epoch": 1.05, + "learning_rate": 1.5104849928026479e-05, + "loss": 1.2369, + "step": 12456 + }, + { + "epoch": 1.05, + "learning_rate": 1.5104068293136906e-05, + "loss": 1.2186, + "step": 12457 + }, + { + "epoch": 1.05, + "learning_rate": 1.5103286616075887e-05, + "loss": 1.1589, + "step": 12458 + }, + { + "epoch": 1.05, + "learning_rate": 1.5102504896849883e-05, + "loss": 1.237, + "step": 12459 + }, + { + "epoch": 1.05, + "learning_rate": 1.510172313546535e-05, + "loss": 1.1378, + "step": 12460 + }, + { + "epoch": 1.05, + "learning_rate": 1.5100941331928744e-05, + "loss": 1.2796, + "step": 12461 + }, + { + "epoch": 1.05, + "learning_rate": 1.5100159486246534e-05, + "loss": 1.1813, + "step": 12462 + }, + { + "epoch": 1.05, + "learning_rate": 1.5099377598425169e-05, + "loss": 1.2054, + "step": 12463 + }, + { + "epoch": 1.05, + "learning_rate": 1.5098595668471117e-05, + "loss": 1.2175, + "step": 12464 + }, + { + "epoch": 1.05, + "learning_rate": 1.5097813696390835e-05, + "loss": 1.1882, + "step": 12465 + }, + { + "epoch": 1.05, + "learning_rate": 1.5097031682190784e-05, + "loss": 1.1655, + "step": 12466 + }, + { + "epoch": 1.05, + "learning_rate": 1.509624962587743e-05, + "loss": 1.2161, + "step": 12467 + }, + { + "epoch": 1.05, + "learning_rate": 1.5095467527457226e-05, + "loss": 1.2057, + "step": 12468 + }, + { + "epoch": 1.05, + "learning_rate": 1.5094685386936641e-05, + "loss": 1.2101, + "step": 12469 + }, + { + "epoch": 1.05, + "learning_rate": 1.509390320432213e-05, + "loss": 1.2026, + "step": 12470 + }, + { + "epoch": 1.05, + "learning_rate": 1.5093120979620166e-05, + "loss": 1.3193, + "step": 12471 + }, + { + "epoch": 1.05, + "learning_rate": 1.5092338712837204e-05, + "loss": 1.1272, + "step": 12472 + }, + { + "epoch": 1.05, + "learning_rate": 1.509155640397971e-05, + "loss": 1.1813, + "step": 12473 + }, + { + "epoch": 1.05, + "learning_rate": 1.5090774053054151e-05, + "loss": 1.308, + "step": 12474 + }, + { + "epoch": 1.05, + "learning_rate": 1.5089991660066985e-05, + "loss": 1.1814, + "step": 12475 + }, + { + "epoch": 1.05, + "learning_rate": 1.5089209225024678e-05, + "loss": 1.2623, + "step": 12476 + }, + { + "epoch": 1.05, + "learning_rate": 1.5088426747933696e-05, + "loss": 1.1359, + "step": 12477 + }, + { + "epoch": 1.05, + "learning_rate": 1.5087644228800504e-05, + "loss": 1.2609, + "step": 12478 + }, + { + "epoch": 1.05, + "learning_rate": 1.5086861667631567e-05, + "loss": 1.1912, + "step": 12479 + }, + { + "epoch": 1.05, + "learning_rate": 1.508607906443335e-05, + "loss": 1.1927, + "step": 12480 + }, + { + "epoch": 1.05, + "learning_rate": 1.5085296419212323e-05, + "loss": 1.203, + "step": 12481 + }, + { + "epoch": 1.05, + "learning_rate": 1.5084513731974948e-05, + "loss": 1.1267, + "step": 12482 + }, + { + "epoch": 1.05, + "learning_rate": 1.508373100272769e-05, + "loss": 1.1673, + "step": 12483 + }, + { + "epoch": 1.05, + "learning_rate": 1.5082948231477024e-05, + "loss": 1.3527, + "step": 12484 + }, + { + "epoch": 1.05, + "learning_rate": 1.5082165418229408e-05, + "loss": 1.1896, + "step": 12485 + }, + { + "epoch": 1.05, + "learning_rate": 1.5081382562991321e-05, + "loss": 1.1361, + "step": 12486 + }, + { + "epoch": 1.05, + "learning_rate": 1.508059966576922e-05, + "loss": 1.1286, + "step": 12487 + }, + { + "epoch": 1.05, + "learning_rate": 1.5079816726569583e-05, + "loss": 1.1871, + "step": 12488 + }, + { + "epoch": 1.05, + "learning_rate": 1.507903374539887e-05, + "loss": 1.2737, + "step": 12489 + }, + { + "epoch": 1.05, + "learning_rate": 1.5078250722263557e-05, + "loss": 1.1731, + "step": 12490 + }, + { + "epoch": 1.05, + "learning_rate": 1.5077467657170108e-05, + "loss": 1.1181, + "step": 12491 + }, + { + "epoch": 1.05, + "learning_rate": 1.5076684550125003e-05, + "loss": 1.2779, + "step": 12492 + }, + { + "epoch": 1.05, + "learning_rate": 1.50759014011347e-05, + "loss": 1.2253, + "step": 12493 + }, + { + "epoch": 1.05, + "learning_rate": 1.5075118210205676e-05, + "loss": 1.1568, + "step": 12494 + }, + { + "epoch": 1.05, + "learning_rate": 1.5074334977344403e-05, + "loss": 1.2716, + "step": 12495 + }, + { + "epoch": 1.05, + "learning_rate": 1.507355170255735e-05, + "loss": 1.2475, + "step": 12496 + }, + { + "epoch": 1.05, + "learning_rate": 1.5072768385850987e-05, + "loss": 1.1755, + "step": 12497 + }, + { + "epoch": 1.05, + "learning_rate": 1.507198502723179e-05, + "loss": 1.1645, + "step": 12498 + }, + { + "epoch": 1.05, + "learning_rate": 1.5071201626706229e-05, + "loss": 1.2084, + "step": 12499 + }, + { + "epoch": 1.05, + "learning_rate": 1.5070418184280778e-05, + "loss": 1.2284, + "step": 12500 + }, + { + "epoch": 1.05, + "learning_rate": 1.5069634699961909e-05, + "loss": 1.1634, + "step": 12501 + }, + { + "epoch": 1.05, + "learning_rate": 1.5068851173756093e-05, + "loss": 1.1202, + "step": 12502 + }, + { + "epoch": 1.05, + "learning_rate": 1.506806760566981e-05, + "loss": 1.2168, + "step": 12503 + }, + { + "epoch": 1.05, + "learning_rate": 1.5067283995709528e-05, + "loss": 1.2591, + "step": 12504 + }, + { + "epoch": 1.05, + "learning_rate": 1.5066500343881725e-05, + "loss": 1.1676, + "step": 12505 + }, + { + "epoch": 1.05, + "learning_rate": 1.5065716650192874e-05, + "loss": 1.169, + "step": 12506 + }, + { + "epoch": 1.05, + "learning_rate": 1.506493291464945e-05, + "loss": 1.1937, + "step": 12507 + }, + { + "epoch": 1.05, + "learning_rate": 1.5064149137257933e-05, + "loss": 1.2208, + "step": 12508 + }, + { + "epoch": 1.05, + "learning_rate": 1.5063365318024791e-05, + "loss": 1.1364, + "step": 12509 + }, + { + "epoch": 1.05, + "learning_rate": 1.5062581456956506e-05, + "loss": 1.1555, + "step": 12510 + }, + { + "epoch": 1.05, + "learning_rate": 1.5061797554059552e-05, + "loss": 1.1453, + "step": 12511 + }, + { + "epoch": 1.05, + "learning_rate": 1.5061013609340407e-05, + "loss": 1.2229, + "step": 12512 + }, + { + "epoch": 1.05, + "learning_rate": 1.5060229622805548e-05, + "loss": 1.1793, + "step": 12513 + }, + { + "epoch": 1.05, + "learning_rate": 1.5059445594461453e-05, + "loss": 1.1758, + "step": 12514 + }, + { + "epoch": 1.05, + "learning_rate": 1.5058661524314597e-05, + "loss": 1.2088, + "step": 12515 + }, + { + "epoch": 1.05, + "learning_rate": 1.5057877412371462e-05, + "loss": 1.163, + "step": 12516 + }, + { + "epoch": 1.05, + "learning_rate": 1.5057093258638527e-05, + "loss": 1.1643, + "step": 12517 + }, + { + "epoch": 1.05, + "learning_rate": 1.5056309063122267e-05, + "loss": 1.1934, + "step": 12518 + }, + { + "epoch": 1.05, + "learning_rate": 1.5055524825829162e-05, + "loss": 1.1818, + "step": 12519 + }, + { + "epoch": 1.05, + "learning_rate": 1.5054740546765696e-05, + "loss": 1.1977, + "step": 12520 + }, + { + "epoch": 1.05, + "learning_rate": 1.5053956225938344e-05, + "loss": 1.1353, + "step": 12521 + }, + { + "epoch": 1.05, + "learning_rate": 1.5053171863353586e-05, + "loss": 1.1611, + "step": 12522 + }, + { + "epoch": 1.05, + "learning_rate": 1.505238745901791e-05, + "loss": 1.2334, + "step": 12523 + }, + { + "epoch": 1.05, + "learning_rate": 1.505160301293779e-05, + "loss": 1.2269, + "step": 12524 + }, + { + "epoch": 1.05, + "learning_rate": 1.5050818525119706e-05, + "loss": 1.2815, + "step": 12525 + }, + { + "epoch": 1.05, + "learning_rate": 1.5050033995570146e-05, + "loss": 1.1126, + "step": 12526 + }, + { + "epoch": 1.05, + "learning_rate": 1.5049249424295589e-05, + "loss": 1.1697, + "step": 12527 + }, + { + "epoch": 1.05, + "learning_rate": 1.5048464811302517e-05, + "loss": 1.1783, + "step": 12528 + }, + { + "epoch": 1.05, + "learning_rate": 1.504768015659741e-05, + "loss": 1.2015, + "step": 12529 + }, + { + "epoch": 1.05, + "learning_rate": 1.5046895460186757e-05, + "loss": 1.2598, + "step": 12530 + }, + { + "epoch": 1.06, + "learning_rate": 1.5046110722077041e-05, + "loss": 1.1338, + "step": 12531 + }, + { + "epoch": 1.06, + "learning_rate": 1.5045325942274738e-05, + "loss": 1.1362, + "step": 12532 + }, + { + "epoch": 1.06, + "learning_rate": 1.504454112078634e-05, + "loss": 1.1888, + "step": 12533 + }, + { + "epoch": 1.06, + "learning_rate": 1.5043756257618326e-05, + "loss": 1.2798, + "step": 12534 + }, + { + "epoch": 1.06, + "learning_rate": 1.5042971352777187e-05, + "loss": 1.1576, + "step": 12535 + }, + { + "epoch": 1.06, + "learning_rate": 1.5042186406269404e-05, + "loss": 1.2176, + "step": 12536 + }, + { + "epoch": 1.06, + "learning_rate": 1.5041401418101464e-05, + "loss": 1.2822, + "step": 12537 + }, + { + "epoch": 1.06, + "learning_rate": 1.5040616388279852e-05, + "loss": 1.1602, + "step": 12538 + }, + { + "epoch": 1.06, + "learning_rate": 1.503983131681105e-05, + "loss": 1.1742, + "step": 12539 + }, + { + "epoch": 1.06, + "learning_rate": 1.5039046203701553e-05, + "loss": 1.1987, + "step": 12540 + }, + { + "epoch": 1.06, + "learning_rate": 1.5038261048957843e-05, + "loss": 1.1821, + "step": 12541 + }, + { + "epoch": 1.06, + "learning_rate": 1.503747585258641e-05, + "loss": 1.1921, + "step": 12542 + }, + { + "epoch": 1.06, + "learning_rate": 1.5036690614593736e-05, + "loss": 1.2085, + "step": 12543 + }, + { + "epoch": 1.06, + "learning_rate": 1.5035905334986311e-05, + "loss": 1.228, + "step": 12544 + }, + { + "epoch": 1.06, + "learning_rate": 1.5035120013770627e-05, + "loss": 1.2727, + "step": 12545 + }, + { + "epoch": 1.06, + "learning_rate": 1.5034334650953169e-05, + "loss": 1.1681, + "step": 12546 + }, + { + "epoch": 1.06, + "learning_rate": 1.5033549246540427e-05, + "loss": 1.1943, + "step": 12547 + }, + { + "epoch": 1.06, + "learning_rate": 1.503276380053889e-05, + "loss": 1.2353, + "step": 12548 + }, + { + "epoch": 1.06, + "learning_rate": 1.5031978312955046e-05, + "loss": 1.2355, + "step": 12549 + }, + { + "epoch": 1.06, + "learning_rate": 1.503119278379539e-05, + "loss": 1.1375, + "step": 12550 + }, + { + "epoch": 1.06, + "learning_rate": 1.5030407213066406e-05, + "loss": 1.1919, + "step": 12551 + }, + { + "epoch": 1.06, + "learning_rate": 1.5029621600774591e-05, + "loss": 1.2525, + "step": 12552 + }, + { + "epoch": 1.06, + "learning_rate": 1.5028835946926433e-05, + "loss": 1.1382, + "step": 12553 + }, + { + "epoch": 1.06, + "learning_rate": 1.5028050251528422e-05, + "loss": 1.2572, + "step": 12554 + }, + { + "epoch": 1.06, + "learning_rate": 1.5027264514587048e-05, + "loss": 1.1837, + "step": 12555 + }, + { + "epoch": 1.06, + "learning_rate": 1.5026478736108806e-05, + "loss": 1.1677, + "step": 12556 + }, + { + "epoch": 1.06, + "learning_rate": 1.5025692916100192e-05, + "loss": 1.3118, + "step": 12557 + }, + { + "epoch": 1.06, + "learning_rate": 1.502490705456769e-05, + "loss": 1.1694, + "step": 12558 + }, + { + "epoch": 1.06, + "learning_rate": 1.50241211515178e-05, + "loss": 1.1723, + "step": 12559 + }, + { + "epoch": 1.06, + "learning_rate": 1.5023335206957015e-05, + "loss": 1.1858, + "step": 12560 + }, + { + "epoch": 1.06, + "learning_rate": 1.5022549220891821e-05, + "loss": 1.3187, + "step": 12561 + }, + { + "epoch": 1.06, + "learning_rate": 1.5021763193328722e-05, + "loss": 1.1911, + "step": 12562 + }, + { + "epoch": 1.06, + "learning_rate": 1.5020977124274212e-05, + "loss": 1.1443, + "step": 12563 + }, + { + "epoch": 1.06, + "learning_rate": 1.5020191013734776e-05, + "loss": 1.2362, + "step": 12564 + }, + { + "epoch": 1.06, + "learning_rate": 1.5019404861716917e-05, + "loss": 1.1592, + "step": 12565 + }, + { + "epoch": 1.06, + "learning_rate": 1.501861866822713e-05, + "loss": 1.2024, + "step": 12566 + }, + { + "epoch": 1.06, + "learning_rate": 1.5017832433271908e-05, + "loss": 1.1557, + "step": 12567 + }, + { + "epoch": 1.06, + "learning_rate": 1.501704615685775e-05, + "loss": 1.2094, + "step": 12568 + }, + { + "epoch": 1.06, + "learning_rate": 1.5016259838991149e-05, + "loss": 1.1261, + "step": 12569 + }, + { + "epoch": 1.06, + "learning_rate": 1.5015473479678604e-05, + "loss": 1.2196, + "step": 12570 + }, + { + "epoch": 1.06, + "learning_rate": 1.5014687078926611e-05, + "loss": 1.1865, + "step": 12571 + }, + { + "epoch": 1.06, + "learning_rate": 1.5013900636741672e-05, + "loss": 1.1402, + "step": 12572 + }, + { + "epoch": 1.06, + "learning_rate": 1.5013114153130279e-05, + "loss": 1.1768, + "step": 12573 + }, + { + "epoch": 1.06, + "learning_rate": 1.5012327628098933e-05, + "loss": 1.2061, + "step": 12574 + }, + { + "epoch": 1.06, + "learning_rate": 1.5011541061654131e-05, + "loss": 1.1944, + "step": 12575 + }, + { + "epoch": 1.06, + "learning_rate": 1.5010754453802375e-05, + "loss": 1.1927, + "step": 12576 + }, + { + "epoch": 1.06, + "learning_rate": 1.500996780455016e-05, + "loss": 1.1839, + "step": 12577 + }, + { + "epoch": 1.06, + "learning_rate": 1.5009181113903988e-05, + "loss": 1.195, + "step": 12578 + }, + { + "epoch": 1.06, + "learning_rate": 1.500839438187036e-05, + "loss": 1.1935, + "step": 12579 + }, + { + "epoch": 1.06, + "learning_rate": 1.5007607608455771e-05, + "loss": 1.3061, + "step": 12580 + }, + { + "epoch": 1.06, + "learning_rate": 1.5006820793666732e-05, + "loss": 1.2605, + "step": 12581 + }, + { + "epoch": 1.06, + "learning_rate": 1.5006033937509732e-05, + "loss": 1.1099, + "step": 12582 + }, + { + "epoch": 1.06, + "learning_rate": 1.5005247039991277e-05, + "loss": 1.2035, + "step": 12583 + }, + { + "epoch": 1.06, + "learning_rate": 1.500446010111787e-05, + "loss": 1.1978, + "step": 12584 + }, + { + "epoch": 1.06, + "learning_rate": 1.5003673120896013e-05, + "loss": 1.1831, + "step": 12585 + }, + { + "epoch": 1.06, + "learning_rate": 1.5002886099332208e-05, + "loss": 1.1842, + "step": 12586 + }, + { + "epoch": 1.06, + "learning_rate": 1.5002099036432953e-05, + "loss": 1.152, + "step": 12587 + }, + { + "epoch": 1.06, + "learning_rate": 1.5001311932204758e-05, + "loss": 1.2622, + "step": 12588 + }, + { + "epoch": 1.06, + "learning_rate": 1.5000524786654125e-05, + "loss": 1.1533, + "step": 12589 + }, + { + "epoch": 1.06, + "learning_rate": 1.4999737599787555e-05, + "loss": 1.2208, + "step": 12590 + }, + { + "epoch": 1.06, + "learning_rate": 1.499895037161155e-05, + "loss": 1.2361, + "step": 12591 + }, + { + "epoch": 1.06, + "learning_rate": 1.4998163102132619e-05, + "loss": 1.1692, + "step": 12592 + }, + { + "epoch": 1.06, + "learning_rate": 1.4997375791357265e-05, + "loss": 1.2589, + "step": 12593 + }, + { + "epoch": 1.06, + "learning_rate": 1.4996588439291993e-05, + "loss": 1.1783, + "step": 12594 + }, + { + "epoch": 1.06, + "learning_rate": 1.4995801045943305e-05, + "loss": 1.2161, + "step": 12595 + }, + { + "epoch": 1.06, + "learning_rate": 1.4995013611317712e-05, + "loss": 1.1052, + "step": 12596 + }, + { + "epoch": 1.06, + "learning_rate": 1.4994226135421719e-05, + "loss": 1.22, + "step": 12597 + }, + { + "epoch": 1.06, + "learning_rate": 1.499343861826183e-05, + "loss": 1.1676, + "step": 12598 + }, + { + "epoch": 1.06, + "learning_rate": 1.4992651059844554e-05, + "loss": 1.2932, + "step": 12599 + }, + { + "epoch": 1.06, + "learning_rate": 1.4991863460176398e-05, + "loss": 1.1606, + "step": 12600 + }, + { + "epoch": 1.06, + "learning_rate": 1.4991075819263868e-05, + "loss": 1.1949, + "step": 12601 + }, + { + "epoch": 1.06, + "learning_rate": 1.4990288137113468e-05, + "loss": 1.2231, + "step": 12602 + }, + { + "epoch": 1.06, + "learning_rate": 1.4989500413731716e-05, + "loss": 1.1726, + "step": 12603 + }, + { + "epoch": 1.06, + "learning_rate": 1.4988712649125114e-05, + "loss": 1.1246, + "step": 12604 + }, + { + "epoch": 1.06, + "learning_rate": 1.498792484330017e-05, + "loss": 1.277, + "step": 12605 + }, + { + "epoch": 1.06, + "learning_rate": 1.4987136996263396e-05, + "loss": 1.2679, + "step": 12606 + }, + { + "epoch": 1.06, + "learning_rate": 1.4986349108021298e-05, + "loss": 1.1842, + "step": 12607 + }, + { + "epoch": 1.06, + "learning_rate": 1.4985561178580388e-05, + "loss": 1.2036, + "step": 12608 + }, + { + "epoch": 1.06, + "learning_rate": 1.4984773207947177e-05, + "loss": 1.2127, + "step": 12609 + }, + { + "epoch": 1.06, + "learning_rate": 1.4983985196128173e-05, + "loss": 1.1984, + "step": 12610 + }, + { + "epoch": 1.06, + "learning_rate": 1.4983197143129889e-05, + "loss": 1.0942, + "step": 12611 + }, + { + "epoch": 1.06, + "learning_rate": 1.4982409048958835e-05, + "loss": 1.1991, + "step": 12612 + }, + { + "epoch": 1.06, + "learning_rate": 1.4981620913621525e-05, + "loss": 1.2119, + "step": 12613 + }, + { + "epoch": 1.06, + "learning_rate": 1.4980832737124466e-05, + "loss": 1.1556, + "step": 12614 + }, + { + "epoch": 1.06, + "learning_rate": 1.4980044519474176e-05, + "loss": 1.1797, + "step": 12615 + }, + { + "epoch": 1.06, + "learning_rate": 1.497925626067716e-05, + "loss": 1.1785, + "step": 12616 + }, + { + "epoch": 1.06, + "learning_rate": 1.497846796073994e-05, + "loss": 1.1428, + "step": 12617 + }, + { + "epoch": 1.06, + "learning_rate": 1.4977679619669021e-05, + "loss": 1.2555, + "step": 12618 + }, + { + "epoch": 1.06, + "learning_rate": 1.497689123747092e-05, + "loss": 1.2304, + "step": 12619 + }, + { + "epoch": 1.06, + "learning_rate": 1.497610281415215e-05, + "loss": 1.287, + "step": 12620 + }, + { + "epoch": 1.06, + "learning_rate": 1.497531434971923e-05, + "loss": 1.2683, + "step": 12621 + }, + { + "epoch": 1.06, + "learning_rate": 1.4974525844178667e-05, + "loss": 1.2622, + "step": 12622 + }, + { + "epoch": 1.06, + "learning_rate": 1.497373729753698e-05, + "loss": 1.1521, + "step": 12623 + }, + { + "epoch": 1.06, + "learning_rate": 1.4972948709800686e-05, + "loss": 1.2513, + "step": 12624 + }, + { + "epoch": 1.06, + "learning_rate": 1.4972160080976297e-05, + "loss": 1.2428, + "step": 12625 + }, + { + "epoch": 1.06, + "learning_rate": 1.4971371411070328e-05, + "loss": 1.1556, + "step": 12626 + }, + { + "epoch": 1.06, + "learning_rate": 1.49705827000893e-05, + "loss": 1.1782, + "step": 12627 + }, + { + "epoch": 1.06, + "learning_rate": 1.4969793948039727e-05, + "loss": 1.2116, + "step": 12628 + }, + { + "epoch": 1.06, + "learning_rate": 1.4969005154928125e-05, + "loss": 1.2529, + "step": 12629 + }, + { + "epoch": 1.06, + "learning_rate": 1.4968216320761012e-05, + "loss": 1.1417, + "step": 12630 + }, + { + "epoch": 1.06, + "learning_rate": 1.4967427445544904e-05, + "loss": 1.2587, + "step": 12631 + }, + { + "epoch": 1.06, + "learning_rate": 1.4966638529286324e-05, + "loss": 1.2275, + "step": 12632 + }, + { + "epoch": 1.06, + "learning_rate": 1.4965849571991784e-05, + "loss": 1.2023, + "step": 12633 + }, + { + "epoch": 1.06, + "learning_rate": 1.4965060573667804e-05, + "loss": 1.2221, + "step": 12634 + }, + { + "epoch": 1.06, + "learning_rate": 1.4964271534320908e-05, + "loss": 1.2235, + "step": 12635 + }, + { + "epoch": 1.06, + "learning_rate": 1.4963482453957612e-05, + "loss": 1.2497, + "step": 12636 + }, + { + "epoch": 1.06, + "learning_rate": 1.4962693332584435e-05, + "loss": 1.1539, + "step": 12637 + }, + { + "epoch": 1.06, + "learning_rate": 1.4961904170207899e-05, + "loss": 1.2944, + "step": 12638 + }, + { + "epoch": 1.06, + "learning_rate": 1.4961114966834517e-05, + "loss": 1.2526, + "step": 12639 + }, + { + "epoch": 1.06, + "learning_rate": 1.4960325722470821e-05, + "loss": 1.2199, + "step": 12640 + }, + { + "epoch": 1.06, + "learning_rate": 1.4959536437123326e-05, + "loss": 1.257, + "step": 12641 + }, + { + "epoch": 1.06, + "learning_rate": 1.4958747110798553e-05, + "loss": 1.2072, + "step": 12642 + }, + { + "epoch": 1.06, + "learning_rate": 1.4957957743503023e-05, + "loss": 1.2219, + "step": 12643 + }, + { + "epoch": 1.06, + "learning_rate": 1.4957168335243259e-05, + "loss": 1.1927, + "step": 12644 + }, + { + "epoch": 1.06, + "learning_rate": 1.4956378886025785e-05, + "loss": 1.098, + "step": 12645 + }, + { + "epoch": 1.06, + "learning_rate": 1.495558939585712e-05, + "loss": 1.2683, + "step": 12646 + }, + { + "epoch": 1.06, + "learning_rate": 1.4954799864743791e-05, + "loss": 1.2148, + "step": 12647 + }, + { + "epoch": 1.06, + "learning_rate": 1.495401029269232e-05, + "loss": 1.2423, + "step": 12648 + }, + { + "epoch": 1.06, + "learning_rate": 1.4953220679709233e-05, + "loss": 1.0751, + "step": 12649 + }, + { + "epoch": 1.07, + "learning_rate": 1.4952431025801048e-05, + "loss": 1.1643, + "step": 12650 + }, + { + "epoch": 1.07, + "learning_rate": 1.4951641330974294e-05, + "loss": 1.2441, + "step": 12651 + }, + { + "epoch": 1.07, + "learning_rate": 1.4950851595235494e-05, + "loss": 1.2642, + "step": 12652 + }, + { + "epoch": 1.07, + "learning_rate": 1.4950061818591178e-05, + "loss": 1.1144, + "step": 12653 + }, + { + "epoch": 1.07, + "learning_rate": 1.494927200104786e-05, + "loss": 1.26, + "step": 12654 + }, + { + "epoch": 1.07, + "learning_rate": 1.4948482142612078e-05, + "loss": 1.1881, + "step": 12655 + }, + { + "epoch": 1.07, + "learning_rate": 1.494769224329035e-05, + "loss": 1.1696, + "step": 12656 + }, + { + "epoch": 1.07, + "learning_rate": 1.4946902303089208e-05, + "loss": 1.3117, + "step": 12657 + }, + { + "epoch": 1.07, + "learning_rate": 1.4946112322015174e-05, + "loss": 1.2865, + "step": 12658 + }, + { + "epoch": 1.07, + "learning_rate": 1.4945322300074773e-05, + "loss": 1.2547, + "step": 12659 + }, + { + "epoch": 1.07, + "learning_rate": 1.4944532237274543e-05, + "loss": 1.1235, + "step": 12660 + }, + { + "epoch": 1.07, + "learning_rate": 1.4943742133621001e-05, + "loss": 1.2052, + "step": 12661 + }, + { + "epoch": 1.07, + "learning_rate": 1.4942951989120683e-05, + "loss": 1.1543, + "step": 12662 + }, + { + "epoch": 1.07, + "learning_rate": 1.4942161803780113e-05, + "loss": 1.1126, + "step": 12663 + }, + { + "epoch": 1.07, + "learning_rate": 1.4941371577605819e-05, + "loss": 1.2166, + "step": 12664 + }, + { + "epoch": 1.07, + "learning_rate": 1.4940581310604333e-05, + "loss": 1.1302, + "step": 12665 + }, + { + "epoch": 1.07, + "learning_rate": 1.493979100278218e-05, + "loss": 1.1845, + "step": 12666 + }, + { + "epoch": 1.07, + "learning_rate": 1.4939000654145897e-05, + "loss": 1.192, + "step": 12667 + }, + { + "epoch": 1.07, + "learning_rate": 1.4938210264702007e-05, + "loss": 1.2312, + "step": 12668 + }, + { + "epoch": 1.07, + "learning_rate": 1.4937419834457043e-05, + "loss": 1.1634, + "step": 12669 + }, + { + "epoch": 1.07, + "learning_rate": 1.4936629363417536e-05, + "loss": 1.2689, + "step": 12670 + }, + { + "epoch": 1.07, + "learning_rate": 1.493583885159002e-05, + "loss": 1.2125, + "step": 12671 + }, + { + "epoch": 1.07, + "learning_rate": 1.493504829898102e-05, + "loss": 1.1587, + "step": 12672 + }, + { + "epoch": 1.07, + "learning_rate": 1.4934257705597076e-05, + "loss": 1.1659, + "step": 12673 + }, + { + "epoch": 1.07, + "learning_rate": 1.4933467071444713e-05, + "loss": 1.1772, + "step": 12674 + }, + { + "epoch": 1.07, + "learning_rate": 1.4932676396530466e-05, + "loss": 1.2111, + "step": 12675 + }, + { + "epoch": 1.07, + "learning_rate": 1.4931885680860867e-05, + "loss": 1.1633, + "step": 12676 + }, + { + "epoch": 1.07, + "learning_rate": 1.493109492444245e-05, + "loss": 1.1369, + "step": 12677 + }, + { + "epoch": 1.07, + "learning_rate": 1.4930304127281755e-05, + "loss": 1.1892, + "step": 12678 + }, + { + "epoch": 1.07, + "learning_rate": 1.4929513289385302e-05, + "loss": 1.1979, + "step": 12679 + }, + { + "epoch": 1.07, + "learning_rate": 1.4928722410759636e-05, + "loss": 1.1468, + "step": 12680 + }, + { + "epoch": 1.07, + "learning_rate": 1.4927931491411287e-05, + "loss": 1.2438, + "step": 12681 + }, + { + "epoch": 1.07, + "learning_rate": 1.4927140531346793e-05, + "loss": 1.2818, + "step": 12682 + }, + { + "epoch": 1.07, + "learning_rate": 1.4926349530572684e-05, + "loss": 1.1011, + "step": 12683 + }, + { + "epoch": 1.07, + "learning_rate": 1.4925558489095497e-05, + "loss": 1.2358, + "step": 12684 + }, + { + "epoch": 1.07, + "learning_rate": 1.4924767406921775e-05, + "loss": 1.2146, + "step": 12685 + }, + { + "epoch": 1.07, + "learning_rate": 1.4923976284058044e-05, + "loss": 1.2279, + "step": 12686 + }, + { + "epoch": 1.07, + "learning_rate": 1.4923185120510845e-05, + "loss": 1.2474, + "step": 12687 + }, + { + "epoch": 1.07, + "learning_rate": 1.4922393916286719e-05, + "loss": 1.0982, + "step": 12688 + }, + { + "epoch": 1.07, + "learning_rate": 1.4921602671392197e-05, + "loss": 1.2316, + "step": 12689 + }, + { + "epoch": 1.07, + "learning_rate": 1.4920811385833817e-05, + "loss": 1.2262, + "step": 12690 + }, + { + "epoch": 1.07, + "learning_rate": 1.4920020059618118e-05, + "loss": 1.108, + "step": 12691 + }, + { + "epoch": 1.07, + "learning_rate": 1.4919228692751643e-05, + "loss": 1.145, + "step": 12692 + }, + { + "epoch": 1.07, + "learning_rate": 1.4918437285240924e-05, + "loss": 1.264, + "step": 12693 + }, + { + "epoch": 1.07, + "learning_rate": 1.49176458370925e-05, + "loss": 1.1841, + "step": 12694 + }, + { + "epoch": 1.07, + "learning_rate": 1.4916854348312915e-05, + "loss": 1.1028, + "step": 12695 + }, + { + "epoch": 1.07, + "learning_rate": 1.4916062818908703e-05, + "loss": 1.166, + "step": 12696 + }, + { + "epoch": 1.07, + "learning_rate": 1.491527124888641e-05, + "loss": 1.169, + "step": 12697 + }, + { + "epoch": 1.07, + "learning_rate": 1.491447963825257e-05, + "loss": 1.2863, + "step": 12698 + }, + { + "epoch": 1.07, + "learning_rate": 1.4913687987013728e-05, + "loss": 1.2178, + "step": 12699 + }, + { + "epoch": 1.07, + "learning_rate": 1.4912896295176421e-05, + "loss": 1.233, + "step": 12700 + }, + { + "epoch": 1.07, + "learning_rate": 1.4912104562747196e-05, + "loss": 1.282, + "step": 12701 + }, + { + "epoch": 1.07, + "learning_rate": 1.4911312789732588e-05, + "loss": 1.1995, + "step": 12702 + }, + { + "epoch": 1.07, + "learning_rate": 1.4910520976139144e-05, + "loss": 1.1118, + "step": 12703 + }, + { + "epoch": 1.07, + "learning_rate": 1.4909729121973405e-05, + "loss": 1.1698, + "step": 12704 + }, + { + "epoch": 1.07, + "learning_rate": 1.4908937227241911e-05, + "loss": 1.1706, + "step": 12705 + }, + { + "epoch": 1.07, + "learning_rate": 1.4908145291951207e-05, + "loss": 1.1636, + "step": 12706 + }, + { + "epoch": 1.07, + "learning_rate": 1.4907353316107836e-05, + "loss": 1.2155, + "step": 12707 + }, + { + "epoch": 1.07, + "learning_rate": 1.4906561299718338e-05, + "loss": 1.1173, + "step": 12708 + }, + { + "epoch": 1.07, + "learning_rate": 1.4905769242789262e-05, + "loss": 1.1621, + "step": 12709 + }, + { + "epoch": 1.07, + "learning_rate": 1.4904977145327153e-05, + "loss": 1.1572, + "step": 12710 + }, + { + "epoch": 1.07, + "learning_rate": 1.490418500733855e-05, + "loss": 1.1974, + "step": 12711 + }, + { + "epoch": 1.07, + "learning_rate": 1.4903392828830003e-05, + "loss": 1.2362, + "step": 12712 + }, + { + "epoch": 1.07, + "learning_rate": 1.4902600609808055e-05, + "loss": 1.1792, + "step": 12713 + }, + { + "epoch": 1.07, + "learning_rate": 1.4901808350279253e-05, + "loss": 1.2316, + "step": 12714 + }, + { + "epoch": 1.07, + "learning_rate": 1.4901016050250139e-05, + "loss": 1.2274, + "step": 12715 + }, + { + "epoch": 1.07, + "learning_rate": 1.4900223709727261e-05, + "loss": 1.2614, + "step": 12716 + }, + { + "epoch": 1.07, + "learning_rate": 1.4899431328717171e-05, + "loss": 1.3478, + "step": 12717 + }, + { + "epoch": 1.07, + "learning_rate": 1.4898638907226407e-05, + "loss": 1.1677, + "step": 12718 + }, + { + "epoch": 1.07, + "learning_rate": 1.4897846445261523e-05, + "loss": 1.1453, + "step": 12719 + }, + { + "epoch": 1.07, + "learning_rate": 1.4897053942829063e-05, + "loss": 1.2835, + "step": 12720 + }, + { + "epoch": 1.07, + "learning_rate": 1.4896261399935576e-05, + "loss": 1.2885, + "step": 12721 + }, + { + "epoch": 1.07, + "learning_rate": 1.489546881658761e-05, + "loss": 1.2555, + "step": 12722 + }, + { + "epoch": 1.07, + "learning_rate": 1.4894676192791715e-05, + "loss": 1.2703, + "step": 12723 + }, + { + "epoch": 1.07, + "learning_rate": 1.4893883528554441e-05, + "loss": 1.1694, + "step": 12724 + }, + { + "epoch": 1.07, + "learning_rate": 1.489309082388233e-05, + "loss": 1.2116, + "step": 12725 + }, + { + "epoch": 1.07, + "learning_rate": 1.4892298078781939e-05, + "loss": 1.277, + "step": 12726 + }, + { + "epoch": 1.07, + "learning_rate": 1.4891505293259813e-05, + "loss": 1.1557, + "step": 12727 + }, + { + "epoch": 1.07, + "learning_rate": 1.489071246732251e-05, + "loss": 1.1425, + "step": 12728 + }, + { + "epoch": 1.07, + "learning_rate": 1.4889919600976573e-05, + "loss": 1.2535, + "step": 12729 + }, + { + "epoch": 1.07, + "learning_rate": 1.4889126694228555e-05, + "loss": 1.2032, + "step": 12730 + }, + { + "epoch": 1.07, + "learning_rate": 1.4888333747085006e-05, + "loss": 1.2148, + "step": 12731 + }, + { + "epoch": 1.07, + "learning_rate": 1.4887540759552483e-05, + "loss": 1.2008, + "step": 12732 + }, + { + "epoch": 1.07, + "learning_rate": 1.4886747731637532e-05, + "loss": 1.1237, + "step": 12733 + }, + { + "epoch": 1.07, + "learning_rate": 1.4885954663346704e-05, + "loss": 1.2144, + "step": 12734 + }, + { + "epoch": 1.07, + "learning_rate": 1.488516155468656e-05, + "loss": 1.2158, + "step": 12735 + }, + { + "epoch": 1.07, + "learning_rate": 1.4884368405663645e-05, + "loss": 1.1967, + "step": 12736 + }, + { + "epoch": 1.07, + "learning_rate": 1.4883575216284514e-05, + "loss": 1.2074, + "step": 12737 + }, + { + "epoch": 1.07, + "learning_rate": 1.4882781986555725e-05, + "loss": 1.1497, + "step": 12738 + }, + { + "epoch": 1.07, + "learning_rate": 1.4881988716483826e-05, + "loss": 1.1756, + "step": 12739 + }, + { + "epoch": 1.07, + "learning_rate": 1.4881195406075374e-05, + "loss": 1.2106, + "step": 12740 + }, + { + "epoch": 1.07, + "learning_rate": 1.4880402055336925e-05, + "loss": 1.2263, + "step": 12741 + }, + { + "epoch": 1.07, + "learning_rate": 1.4879608664275034e-05, + "loss": 1.1582, + "step": 12742 + }, + { + "epoch": 1.07, + "learning_rate": 1.4878815232896249e-05, + "loss": 1.1365, + "step": 12743 + }, + { + "epoch": 1.07, + "learning_rate": 1.4878021761207135e-05, + "loss": 1.1791, + "step": 12744 + }, + { + "epoch": 1.07, + "learning_rate": 1.4877228249214243e-05, + "loss": 1.1165, + "step": 12745 + }, + { + "epoch": 1.07, + "learning_rate": 1.487643469692413e-05, + "loss": 1.1903, + "step": 12746 + }, + { + "epoch": 1.07, + "learning_rate": 1.4875641104343352e-05, + "loss": 1.3121, + "step": 12747 + }, + { + "epoch": 1.07, + "learning_rate": 1.4874847471478466e-05, + "loss": 1.2816, + "step": 12748 + }, + { + "epoch": 1.07, + "learning_rate": 1.4874053798336034e-05, + "loss": 1.1936, + "step": 12749 + }, + { + "epoch": 1.07, + "learning_rate": 1.4873260084922608e-05, + "loss": 1.2045, + "step": 12750 + }, + { + "epoch": 1.07, + "learning_rate": 1.4872466331244745e-05, + "loss": 1.1905, + "step": 12751 + }, + { + "epoch": 1.07, + "learning_rate": 1.487167253730901e-05, + "loss": 1.1887, + "step": 12752 + }, + { + "epoch": 1.07, + "learning_rate": 1.4870878703121956e-05, + "loss": 1.2247, + "step": 12753 + }, + { + "epoch": 1.07, + "learning_rate": 1.4870084828690142e-05, + "loss": 1.1767, + "step": 12754 + }, + { + "epoch": 1.07, + "learning_rate": 1.4869290914020129e-05, + "loss": 1.2169, + "step": 12755 + }, + { + "epoch": 1.07, + "learning_rate": 1.4868496959118477e-05, + "loss": 1.1656, + "step": 12756 + }, + { + "epoch": 1.07, + "learning_rate": 1.4867702963991743e-05, + "loss": 1.1434, + "step": 12757 + }, + { + "epoch": 1.07, + "learning_rate": 1.4866908928646491e-05, + "loss": 1.2167, + "step": 12758 + }, + { + "epoch": 1.07, + "learning_rate": 1.4866114853089277e-05, + "loss": 1.2254, + "step": 12759 + }, + { + "epoch": 1.07, + "learning_rate": 1.486532073732667e-05, + "loss": 1.2157, + "step": 12760 + }, + { + "epoch": 1.07, + "learning_rate": 1.4864526581365221e-05, + "loss": 1.1229, + "step": 12761 + }, + { + "epoch": 1.07, + "learning_rate": 1.4863732385211498e-05, + "loss": 1.1981, + "step": 12762 + }, + { + "epoch": 1.07, + "learning_rate": 1.4862938148872063e-05, + "loss": 1.2212, + "step": 12763 + }, + { + "epoch": 1.07, + "learning_rate": 1.4862143872353476e-05, + "loss": 1.2596, + "step": 12764 + }, + { + "epoch": 1.07, + "learning_rate": 1.4861349555662299e-05, + "loss": 1.1208, + "step": 12765 + }, + { + "epoch": 1.07, + "learning_rate": 1.4860555198805097e-05, + "loss": 1.3228, + "step": 12766 + }, + { + "epoch": 1.07, + "learning_rate": 1.4859760801788434e-05, + "loss": 1.2691, + "step": 12767 + }, + { + "epoch": 1.07, + "learning_rate": 1.485896636461887e-05, + "loss": 1.152, + "step": 12768 + }, + { + "epoch": 1.08, + "learning_rate": 1.485817188730297e-05, + "loss": 1.174, + "step": 12769 + }, + { + "epoch": 1.08, + "learning_rate": 1.48573773698473e-05, + "loss": 1.205, + "step": 12770 + }, + { + "epoch": 1.08, + "learning_rate": 1.4856582812258425e-05, + "loss": 1.2371, + "step": 12771 + }, + { + "epoch": 1.08, + "learning_rate": 1.4855788214542905e-05, + "loss": 1.2112, + "step": 12772 + }, + { + "epoch": 1.08, + "learning_rate": 1.485499357670731e-05, + "loss": 1.2783, + "step": 12773 + }, + { + "epoch": 1.08, + "learning_rate": 1.4854198898758207e-05, + "loss": 1.1067, + "step": 12774 + }, + { + "epoch": 1.08, + "learning_rate": 1.4853404180702158e-05, + "loss": 1.2348, + "step": 12775 + }, + { + "epoch": 1.08, + "learning_rate": 1.485260942254573e-05, + "loss": 1.2354, + "step": 12776 + }, + { + "epoch": 1.08, + "learning_rate": 1.4851814624295489e-05, + "loss": 1.2257, + "step": 12777 + }, + { + "epoch": 1.08, + "learning_rate": 1.4851019785958005e-05, + "loss": 1.2154, + "step": 12778 + }, + { + "epoch": 1.08, + "learning_rate": 1.485022490753984e-05, + "loss": 1.2466, + "step": 12779 + }, + { + "epoch": 1.08, + "learning_rate": 1.4849429989047565e-05, + "loss": 1.2146, + "step": 12780 + }, + { + "epoch": 1.08, + "learning_rate": 1.4848635030487751e-05, + "loss": 1.1771, + "step": 12781 + }, + { + "epoch": 1.08, + "learning_rate": 1.484784003186696e-05, + "loss": 1.208, + "step": 12782 + }, + { + "epoch": 1.08, + "learning_rate": 1.4847044993191761e-05, + "loss": 1.2712, + "step": 12783 + }, + { + "epoch": 1.08, + "learning_rate": 1.4846249914468726e-05, + "loss": 1.1894, + "step": 12784 + }, + { + "epoch": 1.08, + "learning_rate": 1.4845454795704426e-05, + "loss": 1.2822, + "step": 12785 + }, + { + "epoch": 1.08, + "learning_rate": 1.4844659636905427e-05, + "loss": 1.2269, + "step": 12786 + }, + { + "epoch": 1.08, + "learning_rate": 1.4843864438078297e-05, + "loss": 1.2258, + "step": 12787 + }, + { + "epoch": 1.08, + "learning_rate": 1.4843069199229612e-05, + "loss": 1.261, + "step": 12788 + }, + { + "epoch": 1.08, + "learning_rate": 1.4842273920365937e-05, + "loss": 1.1552, + "step": 12789 + }, + { + "epoch": 1.08, + "learning_rate": 1.4841478601493846e-05, + "loss": 1.2966, + "step": 12790 + }, + { + "epoch": 1.08, + "learning_rate": 1.484068324261991e-05, + "loss": 1.2046, + "step": 12791 + }, + { + "epoch": 1.08, + "learning_rate": 1.4839887843750699e-05, + "loss": 1.197, + "step": 12792 + }, + { + "epoch": 1.08, + "learning_rate": 1.4839092404892785e-05, + "loss": 1.3076, + "step": 12793 + }, + { + "epoch": 1.08, + "learning_rate": 1.4838296926052741e-05, + "loss": 1.2961, + "step": 12794 + }, + { + "epoch": 1.08, + "learning_rate": 1.4837501407237143e-05, + "loss": 1.203, + "step": 12795 + }, + { + "epoch": 1.08, + "learning_rate": 1.4836705848452554e-05, + "loss": 1.1802, + "step": 12796 + }, + { + "epoch": 1.08, + "learning_rate": 1.4835910249705555e-05, + "loss": 1.1641, + "step": 12797 + }, + { + "epoch": 1.08, + "learning_rate": 1.4835114611002722e-05, + "loss": 1.1432, + "step": 12798 + }, + { + "epoch": 1.08, + "learning_rate": 1.4834318932350623e-05, + "loss": 1.2271, + "step": 12799 + }, + { + "epoch": 1.08, + "learning_rate": 1.4833523213755831e-05, + "loss": 1.2367, + "step": 12800 + }, + { + "epoch": 1.08, + "learning_rate": 1.4832727455224927e-05, + "loss": 1.1958, + "step": 12801 + }, + { + "epoch": 1.08, + "learning_rate": 1.4831931656764478e-05, + "loss": 1.2104, + "step": 12802 + }, + { + "epoch": 1.08, + "learning_rate": 1.4831135818381067e-05, + "loss": 1.1553, + "step": 12803 + }, + { + "epoch": 1.08, + "learning_rate": 1.4830339940081264e-05, + "loss": 1.212, + "step": 12804 + }, + { + "epoch": 1.08, + "learning_rate": 1.4829544021871647e-05, + "loss": 1.2039, + "step": 12805 + }, + { + "epoch": 1.08, + "learning_rate": 1.482874806375879e-05, + "loss": 1.1921, + "step": 12806 + }, + { + "epoch": 1.08, + "learning_rate": 1.4827952065749274e-05, + "loss": 1.1646, + "step": 12807 + }, + { + "epoch": 1.08, + "learning_rate": 1.4827156027849671e-05, + "loss": 1.1922, + "step": 12808 + }, + { + "epoch": 1.08, + "learning_rate": 1.4826359950066557e-05, + "loss": 1.0928, + "step": 12809 + }, + { + "epoch": 1.08, + "learning_rate": 1.4825563832406517e-05, + "loss": 1.2408, + "step": 12810 + }, + { + "epoch": 1.08, + "learning_rate": 1.4824767674876124e-05, + "loss": 1.2408, + "step": 12811 + }, + { + "epoch": 1.08, + "learning_rate": 1.4823971477481954e-05, + "loss": 1.1413, + "step": 12812 + }, + { + "epoch": 1.08, + "learning_rate": 1.482317524023059e-05, + "loss": 1.1978, + "step": 12813 + }, + { + "epoch": 1.08, + "learning_rate": 1.4822378963128608e-05, + "loss": 1.2509, + "step": 12814 + }, + { + "epoch": 1.08, + "learning_rate": 1.4821582646182587e-05, + "loss": 1.1492, + "step": 12815 + }, + { + "epoch": 1.08, + "learning_rate": 1.4820786289399107e-05, + "loss": 1.173, + "step": 12816 + }, + { + "epoch": 1.08, + "learning_rate": 1.481998989278475e-05, + "loss": 1.2714, + "step": 12817 + }, + { + "epoch": 1.08, + "learning_rate": 1.4819193456346091e-05, + "loss": 1.1957, + "step": 12818 + }, + { + "epoch": 1.08, + "learning_rate": 1.4818396980089715e-05, + "loss": 1.2104, + "step": 12819 + }, + { + "epoch": 1.08, + "learning_rate": 1.48176004640222e-05, + "loss": 1.1908, + "step": 12820 + }, + { + "epoch": 1.08, + "learning_rate": 1.4816803908150128e-05, + "loss": 1.1767, + "step": 12821 + }, + { + "epoch": 1.08, + "learning_rate": 1.4816007312480081e-05, + "loss": 1.2, + "step": 12822 + }, + { + "epoch": 1.08, + "learning_rate": 1.481521067701864e-05, + "loss": 1.1576, + "step": 12823 + }, + { + "epoch": 1.08, + "learning_rate": 1.4814414001772388e-05, + "loss": 1.1226, + "step": 12824 + }, + { + "epoch": 1.08, + "learning_rate": 1.4813617286747905e-05, + "loss": 1.2401, + "step": 12825 + }, + { + "epoch": 1.08, + "learning_rate": 1.481282053195178e-05, + "loss": 1.2493, + "step": 12826 + }, + { + "epoch": 1.08, + "learning_rate": 1.4812023737390586e-05, + "loss": 1.1642, + "step": 12827 + }, + { + "epoch": 1.08, + "learning_rate": 1.4811226903070917e-05, + "loss": 1.1807, + "step": 12828 + }, + { + "epoch": 1.08, + "learning_rate": 1.4810430028999347e-05, + "loss": 1.2546, + "step": 12829 + }, + { + "epoch": 1.08, + "learning_rate": 1.4809633115182467e-05, + "loss": 1.1937, + "step": 12830 + }, + { + "epoch": 1.08, + "learning_rate": 1.4808836161626856e-05, + "loss": 1.2103, + "step": 12831 + }, + { + "epoch": 1.08, + "learning_rate": 1.4808039168339106e-05, + "loss": 1.2097, + "step": 12832 + }, + { + "epoch": 1.08, + "learning_rate": 1.4807242135325795e-05, + "loss": 1.3108, + "step": 12833 + }, + { + "epoch": 1.08, + "learning_rate": 1.4806445062593508e-05, + "loss": 1.1222, + "step": 12834 + }, + { + "epoch": 1.08, + "learning_rate": 1.480564795014884e-05, + "loss": 1.1091, + "step": 12835 + }, + { + "epoch": 1.08, + "learning_rate": 1.4804850797998367e-05, + "loss": 1.2553, + "step": 12836 + }, + { + "epoch": 1.08, + "learning_rate": 1.480405360614868e-05, + "loss": 1.1918, + "step": 12837 + }, + { + "epoch": 1.08, + "learning_rate": 1.4803256374606363e-05, + "loss": 1.1912, + "step": 12838 + }, + { + "epoch": 1.08, + "learning_rate": 1.4802459103378008e-05, + "loss": 1.1554, + "step": 12839 + }, + { + "epoch": 1.08, + "learning_rate": 1.4801661792470196e-05, + "loss": 1.1647, + "step": 12840 + }, + { + "epoch": 1.08, + "learning_rate": 1.4800864441889517e-05, + "loss": 1.1883, + "step": 12841 + }, + { + "epoch": 1.08, + "learning_rate": 1.4800067051642564e-05, + "loss": 1.2539, + "step": 12842 + }, + { + "epoch": 1.08, + "learning_rate": 1.4799269621735916e-05, + "loss": 1.2563, + "step": 12843 + }, + { + "epoch": 1.08, + "learning_rate": 1.4798472152176166e-05, + "loss": 1.1485, + "step": 12844 + }, + { + "epoch": 1.08, + "learning_rate": 1.4797674642969906e-05, + "loss": 1.2095, + "step": 12845 + }, + { + "epoch": 1.08, + "learning_rate": 1.4796877094123723e-05, + "loss": 1.2432, + "step": 12846 + }, + { + "epoch": 1.08, + "learning_rate": 1.4796079505644203e-05, + "loss": 1.1831, + "step": 12847 + }, + { + "epoch": 1.08, + "learning_rate": 1.4795281877537941e-05, + "loss": 1.1464, + "step": 12848 + }, + { + "epoch": 1.08, + "learning_rate": 1.4794484209811528e-05, + "loss": 1.2203, + "step": 12849 + }, + { + "epoch": 1.08, + "learning_rate": 1.4793686502471549e-05, + "loss": 1.1317, + "step": 12850 + }, + { + "epoch": 1.08, + "learning_rate": 1.4792888755524599e-05, + "loss": 1.2862, + "step": 12851 + }, + { + "epoch": 1.08, + "learning_rate": 1.4792090968977268e-05, + "loss": 1.2296, + "step": 12852 + }, + { + "epoch": 1.08, + "learning_rate": 1.479129314283615e-05, + "loss": 1.191, + "step": 12853 + }, + { + "epoch": 1.08, + "learning_rate": 1.479049527710783e-05, + "loss": 1.2649, + "step": 12854 + }, + { + "epoch": 1.08, + "learning_rate": 1.4789697371798909e-05, + "loss": 1.176, + "step": 12855 + }, + { + "epoch": 1.08, + "learning_rate": 1.4788899426915973e-05, + "loss": 1.1575, + "step": 12856 + }, + { + "epoch": 1.08, + "learning_rate": 1.4788101442465618e-05, + "loss": 1.1765, + "step": 12857 + }, + { + "epoch": 1.08, + "learning_rate": 1.4787303418454437e-05, + "loss": 1.1886, + "step": 12858 + }, + { + "epoch": 1.08, + "learning_rate": 1.478650535488902e-05, + "loss": 1.2076, + "step": 12859 + }, + { + "epoch": 1.08, + "learning_rate": 1.4785707251775967e-05, + "loss": 1.2525, + "step": 12860 + }, + { + "epoch": 1.08, + "learning_rate": 1.478490910912187e-05, + "loss": 1.1834, + "step": 12861 + }, + { + "epoch": 1.08, + "learning_rate": 1.478411092693332e-05, + "loss": 1.1402, + "step": 12862 + }, + { + "epoch": 1.08, + "learning_rate": 1.4783312705216915e-05, + "loss": 1.2034, + "step": 12863 + }, + { + "epoch": 1.08, + "learning_rate": 1.4782514443979253e-05, + "loss": 1.27, + "step": 12864 + }, + { + "epoch": 1.08, + "learning_rate": 1.478171614322692e-05, + "loss": 1.2125, + "step": 12865 + }, + { + "epoch": 1.08, + "learning_rate": 1.4780917802966522e-05, + "loss": 1.2306, + "step": 12866 + }, + { + "epoch": 1.08, + "learning_rate": 1.478011942320465e-05, + "loss": 1.2392, + "step": 12867 + }, + { + "epoch": 1.08, + "learning_rate": 1.4779321003947902e-05, + "loss": 1.2078, + "step": 12868 + }, + { + "epoch": 1.08, + "learning_rate": 1.4778522545202873e-05, + "loss": 1.2546, + "step": 12869 + }, + { + "epoch": 1.08, + "learning_rate": 1.4777724046976162e-05, + "loss": 1.1365, + "step": 12870 + }, + { + "epoch": 1.08, + "learning_rate": 1.4776925509274364e-05, + "loss": 1.1101, + "step": 12871 + }, + { + "epoch": 1.08, + "learning_rate": 1.4776126932104083e-05, + "loss": 1.2997, + "step": 12872 + }, + { + "epoch": 1.08, + "learning_rate": 1.477532831547191e-05, + "loss": 1.3178, + "step": 12873 + }, + { + "epoch": 1.08, + "learning_rate": 1.4774529659384446e-05, + "loss": 1.1749, + "step": 12874 + }, + { + "epoch": 1.08, + "learning_rate": 1.4773730963848292e-05, + "loss": 1.1934, + "step": 12875 + }, + { + "epoch": 1.08, + "learning_rate": 1.4772932228870044e-05, + "loss": 1.1674, + "step": 12876 + }, + { + "epoch": 1.08, + "learning_rate": 1.47721334544563e-05, + "loss": 1.204, + "step": 12877 + }, + { + "epoch": 1.08, + "learning_rate": 1.4771334640613666e-05, + "loss": 1.1616, + "step": 12878 + }, + { + "epoch": 1.08, + "learning_rate": 1.4770535787348735e-05, + "loss": 1.1045, + "step": 12879 + }, + { + "epoch": 1.08, + "learning_rate": 1.4769736894668112e-05, + "loss": 1.2195, + "step": 12880 + }, + { + "epoch": 1.08, + "learning_rate": 1.4768937962578394e-05, + "loss": 1.1605, + "step": 12881 + }, + { + "epoch": 1.08, + "learning_rate": 1.4768138991086188e-05, + "loss": 1.1905, + "step": 12882 + }, + { + "epoch": 1.08, + "learning_rate": 1.476733998019809e-05, + "loss": 1.1961, + "step": 12883 + }, + { + "epoch": 1.08, + "learning_rate": 1.47665409299207e-05, + "loss": 1.2643, + "step": 12884 + }, + { + "epoch": 1.08, + "learning_rate": 1.4765741840260628e-05, + "loss": 1.1757, + "step": 12885 + }, + { + "epoch": 1.08, + "learning_rate": 1.476494271122447e-05, + "loss": 1.2268, + "step": 12886 + }, + { + "epoch": 1.08, + "learning_rate": 1.4764143542818832e-05, + "loss": 1.1922, + "step": 12887 + }, + { + "epoch": 1.09, + "learning_rate": 1.4763344335050311e-05, + "loss": 1.2311, + "step": 12888 + }, + { + "epoch": 1.09, + "learning_rate": 1.4762545087925521e-05, + "loss": 1.2454, + "step": 12889 + }, + { + "epoch": 1.09, + "learning_rate": 1.4761745801451054e-05, + "loss": 1.1145, + "step": 12890 + }, + { + "epoch": 1.09, + "learning_rate": 1.4760946475633522e-05, + "loss": 1.1876, + "step": 12891 + }, + { + "epoch": 1.09, + "learning_rate": 1.4760147110479529e-05, + "loss": 1.2393, + "step": 12892 + }, + { + "epoch": 1.09, + "learning_rate": 1.475934770599567e-05, + "loss": 1.2015, + "step": 12893 + }, + { + "epoch": 1.09, + "learning_rate": 1.4758548262188562e-05, + "loss": 1.1446, + "step": 12894 + }, + { + "epoch": 1.09, + "learning_rate": 1.4757748779064805e-05, + "loss": 1.1806, + "step": 12895 + }, + { + "epoch": 1.09, + "learning_rate": 1.4756949256631001e-05, + "loss": 1.1906, + "step": 12896 + }, + { + "epoch": 1.09, + "learning_rate": 1.4756149694893761e-05, + "loss": 1.2082, + "step": 12897 + }, + { + "epoch": 1.09, + "learning_rate": 1.4755350093859693e-05, + "loss": 1.1482, + "step": 12898 + }, + { + "epoch": 1.09, + "learning_rate": 1.4754550453535403e-05, + "loss": 1.23, + "step": 12899 + }, + { + "epoch": 1.09, + "learning_rate": 1.4753750773927491e-05, + "loss": 1.2054, + "step": 12900 + }, + { + "epoch": 1.09, + "learning_rate": 1.4752951055042571e-05, + "loss": 1.2763, + "step": 12901 + }, + { + "epoch": 1.09, + "learning_rate": 1.4752151296887245e-05, + "loss": 1.2177, + "step": 12902 + }, + { + "epoch": 1.09, + "learning_rate": 1.4751351499468127e-05, + "loss": 1.218, + "step": 12903 + }, + { + "epoch": 1.09, + "learning_rate": 1.4750551662791821e-05, + "loss": 1.1693, + "step": 12904 + }, + { + "epoch": 1.09, + "learning_rate": 1.4749751786864936e-05, + "loss": 1.3277, + "step": 12905 + }, + { + "epoch": 1.09, + "learning_rate": 1.4748951871694084e-05, + "loss": 1.1705, + "step": 12906 + }, + { + "epoch": 1.09, + "learning_rate": 1.4748151917285872e-05, + "loss": 1.2371, + "step": 12907 + }, + { + "epoch": 1.09, + "learning_rate": 1.4747351923646906e-05, + "loss": 1.3109, + "step": 12908 + }, + { + "epoch": 1.09, + "learning_rate": 1.47465518907838e-05, + "loss": 1.1216, + "step": 12909 + }, + { + "epoch": 1.09, + "learning_rate": 1.4745751818703164e-05, + "loss": 1.2455, + "step": 12910 + }, + { + "epoch": 1.09, + "learning_rate": 1.4744951707411609e-05, + "loss": 1.2155, + "step": 12911 + }, + { + "epoch": 1.09, + "learning_rate": 1.4744151556915743e-05, + "loss": 1.2314, + "step": 12912 + }, + { + "epoch": 1.09, + "learning_rate": 1.4743351367222176e-05, + "loss": 1.2013, + "step": 12913 + }, + { + "epoch": 1.09, + "learning_rate": 1.4742551138337527e-05, + "loss": 1.0985, + "step": 12914 + }, + { + "epoch": 1.09, + "learning_rate": 1.4741750870268401e-05, + "loss": 1.1586, + "step": 12915 + }, + { + "epoch": 1.09, + "learning_rate": 1.474095056302141e-05, + "loss": 1.1964, + "step": 12916 + }, + { + "epoch": 1.09, + "learning_rate": 1.474015021660317e-05, + "loss": 1.2569, + "step": 12917 + }, + { + "epoch": 1.09, + "learning_rate": 1.4739349831020291e-05, + "loss": 1.2177, + "step": 12918 + }, + { + "epoch": 1.09, + "learning_rate": 1.4738549406279387e-05, + "loss": 1.2683, + "step": 12919 + }, + { + "epoch": 1.09, + "learning_rate": 1.473774894238707e-05, + "loss": 1.3389, + "step": 12920 + }, + { + "epoch": 1.09, + "learning_rate": 1.4736948439349956e-05, + "loss": 1.2102, + "step": 12921 + }, + { + "epoch": 1.09, + "learning_rate": 1.4736147897174658e-05, + "loss": 1.1478, + "step": 12922 + }, + { + "epoch": 1.09, + "learning_rate": 1.473534731586779e-05, + "loss": 1.1336, + "step": 12923 + }, + { + "epoch": 1.09, + "learning_rate": 1.473454669543597e-05, + "loss": 1.1437, + "step": 12924 + }, + { + "epoch": 1.09, + "learning_rate": 1.4733746035885808e-05, + "loss": 1.1777, + "step": 12925 + }, + { + "epoch": 1.09, + "learning_rate": 1.4732945337223919e-05, + "loss": 1.2307, + "step": 12926 + }, + { + "epoch": 1.09, + "learning_rate": 1.473214459945692e-05, + "loss": 1.2636, + "step": 12927 + }, + { + "epoch": 1.09, + "learning_rate": 1.4731343822591432e-05, + "loss": 1.1976, + "step": 12928 + }, + { + "epoch": 1.09, + "learning_rate": 1.4730543006634065e-05, + "loss": 1.1837, + "step": 12929 + }, + { + "epoch": 1.09, + "learning_rate": 1.4729742151591439e-05, + "loss": 1.1968, + "step": 12930 + }, + { + "epoch": 1.09, + "learning_rate": 1.4728941257470168e-05, + "loss": 1.1752, + "step": 12931 + }, + { + "epoch": 1.09, + "learning_rate": 1.4728140324276874e-05, + "loss": 1.1737, + "step": 12932 + }, + { + "epoch": 1.09, + "learning_rate": 1.4727339352018169e-05, + "loss": 1.1336, + "step": 12933 + }, + { + "epoch": 1.09, + "learning_rate": 1.4726538340700671e-05, + "loss": 1.2502, + "step": 12934 + }, + { + "epoch": 1.09, + "learning_rate": 1.4725737290331003e-05, + "loss": 1.2058, + "step": 12935 + }, + { + "epoch": 1.09, + "learning_rate": 1.4724936200915782e-05, + "loss": 1.1504, + "step": 12936 + }, + { + "epoch": 1.09, + "learning_rate": 1.4724135072461625e-05, + "loss": 1.15, + "step": 12937 + }, + { + "epoch": 1.09, + "learning_rate": 1.4723333904975153e-05, + "loss": 1.2324, + "step": 12938 + }, + { + "epoch": 1.09, + "learning_rate": 1.4722532698462983e-05, + "loss": 1.215, + "step": 12939 + }, + { + "epoch": 1.09, + "learning_rate": 1.4721731452931737e-05, + "loss": 1.1885, + "step": 12940 + }, + { + "epoch": 1.09, + "learning_rate": 1.4720930168388034e-05, + "loss": 1.177, + "step": 12941 + }, + { + "epoch": 1.09, + "learning_rate": 1.4720128844838496e-05, + "loss": 1.1575, + "step": 12942 + }, + { + "epoch": 1.09, + "learning_rate": 1.4719327482289744e-05, + "loss": 1.1222, + "step": 12943 + }, + { + "epoch": 1.09, + "learning_rate": 1.4718526080748397e-05, + "loss": 1.1988, + "step": 12944 + }, + { + "epoch": 1.09, + "learning_rate": 1.4717724640221077e-05, + "loss": 1.1421, + "step": 12945 + }, + { + "epoch": 1.09, + "learning_rate": 1.4716923160714408e-05, + "loss": 1.18, + "step": 12946 + }, + { + "epoch": 1.09, + "learning_rate": 1.4716121642235006e-05, + "loss": 1.2171, + "step": 12947 + }, + { + "epoch": 1.09, + "learning_rate": 1.47153200847895e-05, + "loss": 1.1612, + "step": 12948 + }, + { + "epoch": 1.09, + "learning_rate": 1.4714518488384512e-05, + "loss": 1.2964, + "step": 12949 + }, + { + "epoch": 1.09, + "learning_rate": 1.471371685302666e-05, + "loss": 1.2282, + "step": 12950 + }, + { + "epoch": 1.09, + "learning_rate": 1.4712915178722573e-05, + "loss": 1.1703, + "step": 12951 + }, + { + "epoch": 1.09, + "learning_rate": 1.4712113465478872e-05, + "loss": 1.1525, + "step": 12952 + }, + { + "epoch": 1.09, + "learning_rate": 1.4711311713302184e-05, + "loss": 1.1565, + "step": 12953 + }, + { + "epoch": 1.09, + "learning_rate": 1.4710509922199128e-05, + "loss": 1.2246, + "step": 12954 + }, + { + "epoch": 1.09, + "learning_rate": 1.4709708092176332e-05, + "loss": 1.165, + "step": 12955 + }, + { + "epoch": 1.09, + "learning_rate": 1.4708906223240418e-05, + "loss": 1.1677, + "step": 12956 + }, + { + "epoch": 1.09, + "learning_rate": 1.4708104315398017e-05, + "loss": 1.1784, + "step": 12957 + }, + { + "epoch": 1.09, + "learning_rate": 1.4707302368655752e-05, + "loss": 1.2294, + "step": 12958 + }, + { + "epoch": 1.09, + "learning_rate": 1.4706500383020244e-05, + "loss": 1.1908, + "step": 12959 + }, + { + "epoch": 1.09, + "learning_rate": 1.4705698358498128e-05, + "loss": 1.1885, + "step": 12960 + }, + { + "epoch": 1.09, + "learning_rate": 1.4704896295096022e-05, + "loss": 1.1919, + "step": 12961 + }, + { + "epoch": 1.09, + "learning_rate": 1.470409419282056e-05, + "loss": 1.2566, + "step": 12962 + }, + { + "epoch": 1.09, + "learning_rate": 1.4703292051678366e-05, + "loss": 1.2144, + "step": 12963 + }, + { + "epoch": 1.09, + "learning_rate": 1.4702489871676068e-05, + "loss": 1.0535, + "step": 12964 + }, + { + "epoch": 1.09, + "learning_rate": 1.4701687652820293e-05, + "loss": 1.212, + "step": 12965 + }, + { + "epoch": 1.09, + "learning_rate": 1.470088539511767e-05, + "loss": 1.3075, + "step": 12966 + }, + { + "epoch": 1.09, + "learning_rate": 1.4700083098574828e-05, + "loss": 1.1476, + "step": 12967 + }, + { + "epoch": 1.09, + "learning_rate": 1.4699280763198395e-05, + "loss": 1.218, + "step": 12968 + }, + { + "epoch": 1.09, + "learning_rate": 1.4698478388995002e-05, + "loss": 1.3128, + "step": 12969 + }, + { + "epoch": 1.09, + "learning_rate": 1.4697675975971272e-05, + "loss": 1.1597, + "step": 12970 + }, + { + "epoch": 1.09, + "learning_rate": 1.4696873524133843e-05, + "loss": 1.1656, + "step": 12971 + }, + { + "epoch": 1.09, + "learning_rate": 1.469607103348934e-05, + "loss": 1.2183, + "step": 12972 + }, + { + "epoch": 1.09, + "learning_rate": 1.4695268504044398e-05, + "loss": 1.1013, + "step": 12973 + }, + { + "epoch": 1.09, + "learning_rate": 1.4694465935805645e-05, + "loss": 1.1657, + "step": 12974 + }, + { + "epoch": 1.09, + "learning_rate": 1.4693663328779711e-05, + "loss": 1.2432, + "step": 12975 + }, + { + "epoch": 1.09, + "learning_rate": 1.4692860682973226e-05, + "loss": 1.1794, + "step": 12976 + }, + { + "epoch": 1.09, + "learning_rate": 1.4692057998392826e-05, + "loss": 1.2154, + "step": 12977 + }, + { + "epoch": 1.09, + "learning_rate": 1.4691255275045142e-05, + "loss": 1.0912, + "step": 12978 + }, + { + "epoch": 1.09, + "learning_rate": 1.4690452512936806e-05, + "loss": 1.2159, + "step": 12979 + }, + { + "epoch": 1.09, + "learning_rate": 1.4689649712074447e-05, + "loss": 1.1708, + "step": 12980 + }, + { + "epoch": 1.09, + "learning_rate": 1.4688846872464703e-05, + "loss": 1.226, + "step": 12981 + }, + { + "epoch": 1.09, + "learning_rate": 1.4688043994114208e-05, + "loss": 1.2014, + "step": 12982 + }, + { + "epoch": 1.09, + "learning_rate": 1.468724107702959e-05, + "loss": 1.1742, + "step": 12983 + }, + { + "epoch": 1.09, + "learning_rate": 1.4686438121217484e-05, + "loss": 1.1936, + "step": 12984 + }, + { + "epoch": 1.09, + "learning_rate": 1.4685635126684532e-05, + "loss": 1.2724, + "step": 12985 + }, + { + "epoch": 1.09, + "learning_rate": 1.4684832093437358e-05, + "loss": 1.2216, + "step": 12986 + }, + { + "epoch": 1.09, + "learning_rate": 1.4684029021482603e-05, + "loss": 1.1605, + "step": 12987 + }, + { + "epoch": 1.09, + "learning_rate": 1.4683225910826904e-05, + "loss": 1.2289, + "step": 12988 + }, + { + "epoch": 1.09, + "learning_rate": 1.4682422761476894e-05, + "loss": 1.2761, + "step": 12989 + }, + { + "epoch": 1.09, + "learning_rate": 1.4681619573439205e-05, + "loss": 1.2958, + "step": 12990 + }, + { + "epoch": 1.09, + "learning_rate": 1.468081634672048e-05, + "loss": 1.1112, + "step": 12991 + }, + { + "epoch": 1.09, + "learning_rate": 1.4680013081327349e-05, + "loss": 1.1467, + "step": 12992 + }, + { + "epoch": 1.09, + "learning_rate": 1.4679209777266455e-05, + "loss": 1.2367, + "step": 12993 + }, + { + "epoch": 1.09, + "learning_rate": 1.467840643454443e-05, + "loss": 1.2428, + "step": 12994 + }, + { + "epoch": 1.09, + "learning_rate": 1.4677603053167914e-05, + "loss": 1.3142, + "step": 12995 + }, + { + "epoch": 1.09, + "learning_rate": 1.4676799633143545e-05, + "loss": 1.2814, + "step": 12996 + }, + { + "epoch": 1.09, + "learning_rate": 1.4675996174477963e-05, + "loss": 1.228, + "step": 12997 + }, + { + "epoch": 1.09, + "learning_rate": 1.46751926771778e-05, + "loss": 1.2218, + "step": 12998 + }, + { + "epoch": 1.09, + "learning_rate": 1.4674389141249703e-05, + "loss": 1.1916, + "step": 12999 + }, + { + "epoch": 1.09, + "learning_rate": 1.4673585566700306e-05, + "loss": 1.1301, + "step": 13000 + }, + { + "epoch": 1.09, + "learning_rate": 1.4672781953536249e-05, + "loss": 1.2411, + "step": 13001 + }, + { + "epoch": 1.09, + "learning_rate": 1.4671978301764169e-05, + "loss": 1.199, + "step": 13002 + }, + { + "epoch": 1.09, + "learning_rate": 1.4671174611390712e-05, + "loss": 1.1898, + "step": 13003 + }, + { + "epoch": 1.09, + "learning_rate": 1.4670370882422515e-05, + "loss": 1.2566, + "step": 13004 + }, + { + "epoch": 1.09, + "learning_rate": 1.4669567114866218e-05, + "loss": 1.1196, + "step": 13005 + }, + { + "epoch": 1.09, + "learning_rate": 1.4668763308728464e-05, + "loss": 1.2217, + "step": 13006 + }, + { + "epoch": 1.1, + "learning_rate": 1.4667959464015896e-05, + "loss": 1.2559, + "step": 13007 + }, + { + "epoch": 1.1, + "learning_rate": 1.4667155580735148e-05, + "loss": 1.1973, + "step": 13008 + }, + { + "epoch": 1.1, + "learning_rate": 1.4666351658892866e-05, + "loss": 1.1484, + "step": 13009 + }, + { + "epoch": 1.1, + "learning_rate": 1.4665547698495699e-05, + "loss": 1.2228, + "step": 13010 + }, + { + "epoch": 1.1, + "learning_rate": 1.4664743699550279e-05, + "loss": 1.2394, + "step": 13011 + }, + { + "epoch": 1.1, + "learning_rate": 1.4663939662063256e-05, + "loss": 1.2397, + "step": 13012 + }, + { + "epoch": 1.1, + "learning_rate": 1.466313558604127e-05, + "loss": 1.2054, + "step": 13013 + }, + { + "epoch": 1.1, + "learning_rate": 1.4662331471490966e-05, + "loss": 1.1692, + "step": 13014 + }, + { + "epoch": 1.1, + "learning_rate": 1.466152731841899e-05, + "loss": 1.1398, + "step": 13015 + }, + { + "epoch": 1.1, + "learning_rate": 1.4660723126831977e-05, + "loss": 1.1568, + "step": 13016 + }, + { + "epoch": 1.1, + "learning_rate": 1.4659918896736582e-05, + "loss": 1.1609, + "step": 13017 + }, + { + "epoch": 1.1, + "learning_rate": 1.4659114628139449e-05, + "loss": 1.2392, + "step": 13018 + }, + { + "epoch": 1.1, + "learning_rate": 1.4658310321047214e-05, + "loss": 1.1995, + "step": 13019 + }, + { + "epoch": 1.1, + "learning_rate": 1.465750597546653e-05, + "loss": 1.2305, + "step": 13020 + }, + { + "epoch": 1.1, + "learning_rate": 1.465670159140404e-05, + "loss": 1.236, + "step": 13021 + }, + { + "epoch": 1.1, + "learning_rate": 1.4655897168866392e-05, + "loss": 1.1035, + "step": 13022 + }, + { + "epoch": 1.1, + "learning_rate": 1.4655092707860233e-05, + "loss": 1.2423, + "step": 13023 + }, + { + "epoch": 1.1, + "learning_rate": 1.4654288208392206e-05, + "loss": 1.2628, + "step": 13024 + }, + { + "epoch": 1.1, + "learning_rate": 1.4653483670468961e-05, + "loss": 1.135, + "step": 13025 + }, + { + "epoch": 1.1, + "learning_rate": 1.4652679094097146e-05, + "loss": 1.1488, + "step": 13026 + }, + { + "epoch": 1.1, + "learning_rate": 1.4651874479283405e-05, + "loss": 1.1795, + "step": 13027 + }, + { + "epoch": 1.1, + "learning_rate": 1.4651069826034392e-05, + "loss": 1.2588, + "step": 13028 + }, + { + "epoch": 1.1, + "learning_rate": 1.465026513435675e-05, + "loss": 1.2519, + "step": 13029 + }, + { + "epoch": 1.1, + "learning_rate": 1.4649460404257125e-05, + "loss": 1.2001, + "step": 13030 + }, + { + "epoch": 1.1, + "learning_rate": 1.4648655635742175e-05, + "loss": 1.1387, + "step": 13031 + }, + { + "epoch": 1.1, + "learning_rate": 1.4647850828818545e-05, + "loss": 1.1961, + "step": 13032 + }, + { + "epoch": 1.1, + "learning_rate": 1.4647045983492881e-05, + "loss": 1.1871, + "step": 13033 + }, + { + "epoch": 1.1, + "learning_rate": 1.4646241099771834e-05, + "loss": 1.1855, + "step": 13034 + }, + { + "epoch": 1.1, + "learning_rate": 1.464543617766206e-05, + "loss": 1.2079, + "step": 13035 + }, + { + "epoch": 1.1, + "learning_rate": 1.4644631217170205e-05, + "loss": 1.2446, + "step": 13036 + }, + { + "epoch": 1.1, + "learning_rate": 1.464382621830292e-05, + "loss": 1.23, + "step": 13037 + }, + { + "epoch": 1.1, + "learning_rate": 1.4643021181066855e-05, + "loss": 1.2243, + "step": 13038 + }, + { + "epoch": 1.1, + "learning_rate": 1.4642216105468662e-05, + "loss": 1.2205, + "step": 13039 + }, + { + "epoch": 1.1, + "learning_rate": 1.4641410991514997e-05, + "loss": 1.1302, + "step": 13040 + }, + { + "epoch": 1.1, + "learning_rate": 1.464060583921251e-05, + "loss": 1.1367, + "step": 13041 + }, + { + "epoch": 1.1, + "learning_rate": 1.4639800648567847e-05, + "loss": 1.1327, + "step": 13042 + }, + { + "epoch": 1.1, + "learning_rate": 1.4638995419587672e-05, + "loss": 1.1424, + "step": 13043 + }, + { + "epoch": 1.1, + "learning_rate": 1.463819015227863e-05, + "loss": 1.1207, + "step": 13044 + }, + { + "epoch": 1.1, + "learning_rate": 1.4637384846647376e-05, + "loss": 1.2226, + "step": 13045 + }, + { + "epoch": 1.1, + "learning_rate": 1.4636579502700563e-05, + "loss": 1.1757, + "step": 13046 + }, + { + "epoch": 1.1, + "learning_rate": 1.4635774120444848e-05, + "loss": 1.1923, + "step": 13047 + }, + { + "epoch": 1.1, + "learning_rate": 1.4634968699886882e-05, + "loss": 1.1433, + "step": 13048 + }, + { + "epoch": 1.1, + "learning_rate": 1.4634163241033324e-05, + "loss": 1.2502, + "step": 13049 + }, + { + "epoch": 1.1, + "learning_rate": 1.4633357743890826e-05, + "loss": 1.1464, + "step": 13050 + }, + { + "epoch": 1.1, + "learning_rate": 1.463255220846604e-05, + "loss": 1.2211, + "step": 13051 + }, + { + "epoch": 1.1, + "learning_rate": 1.4631746634765627e-05, + "loss": 1.2255, + "step": 13052 + }, + { + "epoch": 1.1, + "learning_rate": 1.4630941022796243e-05, + "loss": 1.2562, + "step": 13053 + }, + { + "epoch": 1.1, + "learning_rate": 1.463013537256454e-05, + "loss": 1.3055, + "step": 13054 + }, + { + "epoch": 1.1, + "learning_rate": 1.4629329684077176e-05, + "loss": 1.2167, + "step": 13055 + }, + { + "epoch": 1.1, + "learning_rate": 1.462852395734081e-05, + "loss": 1.196, + "step": 13056 + }, + { + "epoch": 1.1, + "learning_rate": 1.4627718192362099e-05, + "loss": 1.3023, + "step": 13057 + }, + { + "epoch": 1.1, + "learning_rate": 1.4626912389147698e-05, + "loss": 1.1359, + "step": 13058 + }, + { + "epoch": 1.1, + "learning_rate": 1.4626106547704262e-05, + "loss": 1.0822, + "step": 13059 + }, + { + "epoch": 1.1, + "learning_rate": 1.4625300668038459e-05, + "loss": 1.2103, + "step": 13060 + }, + { + "epoch": 1.1, + "learning_rate": 1.462449475015694e-05, + "loss": 1.2409, + "step": 13061 + }, + { + "epoch": 1.1, + "learning_rate": 1.4623688794066362e-05, + "loss": 1.176, + "step": 13062 + }, + { + "epoch": 1.1, + "learning_rate": 1.462288279977339e-05, + "loss": 1.1972, + "step": 13063 + }, + { + "epoch": 1.1, + "learning_rate": 1.4622076767284683e-05, + "loss": 1.1851, + "step": 13064 + }, + { + "epoch": 1.1, + "learning_rate": 1.4621270696606895e-05, + "loss": 1.2334, + "step": 13065 + }, + { + "epoch": 1.1, + "learning_rate": 1.4620464587746692e-05, + "loss": 1.204, + "step": 13066 + }, + { + "epoch": 1.1, + "learning_rate": 1.461965844071073e-05, + "loss": 1.3812, + "step": 13067 + }, + { + "epoch": 1.1, + "learning_rate": 1.4618852255505673e-05, + "loss": 1.1057, + "step": 13068 + }, + { + "epoch": 1.1, + "learning_rate": 1.461804603213818e-05, + "loss": 1.2787, + "step": 13069 + }, + { + "epoch": 1.1, + "learning_rate": 1.4617239770614913e-05, + "loss": 1.1996, + "step": 13070 + }, + { + "epoch": 1.1, + "learning_rate": 1.461643347094253e-05, + "loss": 1.1769, + "step": 13071 + }, + { + "epoch": 1.1, + "learning_rate": 1.46156271331277e-05, + "loss": 1.1692, + "step": 13072 + }, + { + "epoch": 1.1, + "learning_rate": 1.4614820757177079e-05, + "loss": 1.1841, + "step": 13073 + }, + { + "epoch": 1.1, + "learning_rate": 1.4614014343097335e-05, + "loss": 1.1501, + "step": 13074 + }, + { + "epoch": 1.1, + "learning_rate": 1.4613207890895127e-05, + "loss": 1.1724, + "step": 13075 + }, + { + "epoch": 1.1, + "learning_rate": 1.461240140057712e-05, + "loss": 1.2493, + "step": 13076 + }, + { + "epoch": 1.1, + "learning_rate": 1.4611594872149972e-05, + "loss": 1.2453, + "step": 13077 + }, + { + "epoch": 1.1, + "learning_rate": 1.4610788305620356e-05, + "loss": 1.1725, + "step": 13078 + }, + { + "epoch": 1.1, + "learning_rate": 1.4609981700994928e-05, + "loss": 1.2433, + "step": 13079 + }, + { + "epoch": 1.1, + "learning_rate": 1.4609175058280358e-05, + "loss": 1.1875, + "step": 13080 + }, + { + "epoch": 1.1, + "learning_rate": 1.4608368377483306e-05, + "loss": 1.1431, + "step": 13081 + }, + { + "epoch": 1.1, + "learning_rate": 1.4607561658610443e-05, + "loss": 1.2211, + "step": 13082 + }, + { + "epoch": 1.1, + "learning_rate": 1.460675490166843e-05, + "loss": 1.213, + "step": 13083 + }, + { + "epoch": 1.1, + "learning_rate": 1.460594810666393e-05, + "loss": 1.1623, + "step": 13084 + }, + { + "epoch": 1.1, + "learning_rate": 1.4605141273603616e-05, + "loss": 1.2273, + "step": 13085 + }, + { + "epoch": 1.1, + "learning_rate": 1.4604334402494149e-05, + "loss": 1.2056, + "step": 13086 + }, + { + "epoch": 1.1, + "learning_rate": 1.46035274933422e-05, + "loss": 1.241, + "step": 13087 + }, + { + "epoch": 1.1, + "learning_rate": 1.4602720546154432e-05, + "loss": 1.2278, + "step": 13088 + }, + { + "epoch": 1.1, + "learning_rate": 1.4601913560937516e-05, + "loss": 1.2351, + "step": 13089 + }, + { + "epoch": 1.1, + "learning_rate": 1.4601106537698115e-05, + "loss": 1.2063, + "step": 13090 + }, + { + "epoch": 1.1, + "learning_rate": 1.46002994764429e-05, + "loss": 1.129, + "step": 13091 + }, + { + "epoch": 1.1, + "learning_rate": 1.459949237717854e-05, + "loss": 1.1322, + "step": 13092 + }, + { + "epoch": 1.1, + "learning_rate": 1.4598685239911701e-05, + "loss": 1.2956, + "step": 13093 + }, + { + "epoch": 1.1, + "learning_rate": 1.4597878064649052e-05, + "loss": 1.2603, + "step": 13094 + }, + { + "epoch": 1.1, + "learning_rate": 1.4597070851397262e-05, + "loss": 1.22, + "step": 13095 + }, + { + "epoch": 1.1, + "learning_rate": 1.4596263600163001e-05, + "loss": 1.1783, + "step": 13096 + }, + { + "epoch": 1.1, + "learning_rate": 1.459545631095294e-05, + "loss": 1.2582, + "step": 13097 + }, + { + "epoch": 1.1, + "learning_rate": 1.4594648983773749e-05, + "loss": 1.2685, + "step": 13098 + }, + { + "epoch": 1.1, + "learning_rate": 1.4593841618632097e-05, + "loss": 1.2369, + "step": 13099 + }, + { + "epoch": 1.1, + "learning_rate": 1.4593034215534659e-05, + "loss": 1.2546, + "step": 13100 + }, + { + "epoch": 1.1, + "learning_rate": 1.4592226774488096e-05, + "loss": 1.1484, + "step": 13101 + }, + { + "epoch": 1.1, + "learning_rate": 1.459141929549909e-05, + "loss": 1.2101, + "step": 13102 + }, + { + "epoch": 1.1, + "learning_rate": 1.4590611778574309e-05, + "loss": 1.1799, + "step": 13103 + }, + { + "epoch": 1.1, + "learning_rate": 1.458980422372042e-05, + "loss": 1.1573, + "step": 13104 + }, + { + "epoch": 1.1, + "learning_rate": 1.4588996630944103e-05, + "loss": 1.1839, + "step": 13105 + }, + { + "epoch": 1.1, + "learning_rate": 1.4588189000252026e-05, + "loss": 1.2329, + "step": 13106 + }, + { + "epoch": 1.1, + "learning_rate": 1.4587381331650865e-05, + "loss": 1.2055, + "step": 13107 + }, + { + "epoch": 1.1, + "learning_rate": 1.458657362514729e-05, + "loss": 1.2153, + "step": 13108 + }, + { + "epoch": 1.1, + "learning_rate": 1.4585765880747973e-05, + "loss": 1.1638, + "step": 13109 + }, + { + "epoch": 1.1, + "learning_rate": 1.4584958098459595e-05, + "loss": 1.1415, + "step": 13110 + }, + { + "epoch": 1.1, + "learning_rate": 1.4584150278288822e-05, + "loss": 1.2125, + "step": 13111 + }, + { + "epoch": 1.1, + "learning_rate": 1.4583342420242335e-05, + "loss": 1.2075, + "step": 13112 + }, + { + "epoch": 1.1, + "learning_rate": 1.4582534524326804e-05, + "loss": 1.2706, + "step": 13113 + }, + { + "epoch": 1.1, + "learning_rate": 1.4581726590548908e-05, + "loss": 1.1925, + "step": 13114 + }, + { + "epoch": 1.1, + "learning_rate": 1.458091861891532e-05, + "loss": 1.1818, + "step": 13115 + }, + { + "epoch": 1.1, + "learning_rate": 1.4580110609432716e-05, + "loss": 1.1521, + "step": 13116 + }, + { + "epoch": 1.1, + "learning_rate": 1.457930256210777e-05, + "loss": 1.2019, + "step": 13117 + }, + { + "epoch": 1.1, + "learning_rate": 1.4578494476947168e-05, + "loss": 1.1205, + "step": 13118 + }, + { + "epoch": 1.1, + "learning_rate": 1.4577686353957572e-05, + "loss": 1.2006, + "step": 13119 + }, + { + "epoch": 1.1, + "learning_rate": 1.457687819314567e-05, + "loss": 1.1615, + "step": 13120 + }, + { + "epoch": 1.1, + "learning_rate": 1.4576069994518134e-05, + "loss": 1.2733, + "step": 13121 + }, + { + "epoch": 1.1, + "learning_rate": 1.4575261758081643e-05, + "loss": 1.2403, + "step": 13122 + }, + { + "epoch": 1.1, + "learning_rate": 1.4574453483842874e-05, + "loss": 1.186, + "step": 13123 + }, + { + "epoch": 1.1, + "learning_rate": 1.4573645171808508e-05, + "loss": 1.1418, + "step": 13124 + }, + { + "epoch": 1.11, + "learning_rate": 1.4572836821985223e-05, + "loss": 1.1838, + "step": 13125 + }, + { + "epoch": 1.11, + "learning_rate": 1.4572028434379695e-05, + "loss": 1.1641, + "step": 13126 + }, + { + "epoch": 1.11, + "learning_rate": 1.4571220008998605e-05, + "loss": 1.2655, + "step": 13127 + }, + { + "epoch": 1.11, + "learning_rate": 1.4570411545848635e-05, + "loss": 1.1226, + "step": 13128 + }, + { + "epoch": 1.11, + "learning_rate": 1.456960304493646e-05, + "loss": 1.1559, + "step": 13129 + }, + { + "epoch": 1.11, + "learning_rate": 1.4568794506268762e-05, + "loss": 1.1543, + "step": 13130 + }, + { + "epoch": 1.11, + "learning_rate": 1.456798592985222e-05, + "loss": 1.225, + "step": 13131 + }, + { + "epoch": 1.11, + "learning_rate": 1.4567177315693518e-05, + "loss": 1.3117, + "step": 13132 + }, + { + "epoch": 1.11, + "learning_rate": 1.4566368663799335e-05, + "loss": 1.2415, + "step": 13133 + }, + { + "epoch": 1.11, + "learning_rate": 1.4565559974176351e-05, + "loss": 1.2391, + "step": 13134 + }, + { + "epoch": 1.11, + "learning_rate": 1.4564751246831253e-05, + "loss": 1.1713, + "step": 13135 + }, + { + "epoch": 1.11, + "learning_rate": 1.4563942481770717e-05, + "loss": 1.2236, + "step": 13136 + }, + { + "epoch": 1.11, + "learning_rate": 1.4563133679001426e-05, + "loss": 1.2902, + "step": 13137 + }, + { + "epoch": 1.11, + "learning_rate": 1.4562324838530066e-05, + "loss": 1.1639, + "step": 13138 + }, + { + "epoch": 1.11, + "learning_rate": 1.4561515960363322e-05, + "loss": 1.1827, + "step": 13139 + }, + { + "epoch": 1.11, + "learning_rate": 1.4560707044507866e-05, + "loss": 1.2637, + "step": 13140 + }, + { + "epoch": 1.11, + "learning_rate": 1.4559898090970395e-05, + "loss": 1.2145, + "step": 13141 + }, + { + "epoch": 1.11, + "learning_rate": 1.4559089099757582e-05, + "loss": 1.1975, + "step": 13142 + }, + { + "epoch": 1.11, + "learning_rate": 1.4558280070876118e-05, + "loss": 1.1578, + "step": 13143 + }, + { + "epoch": 1.11, + "learning_rate": 1.4557471004332687e-05, + "loss": 1.336, + "step": 13144 + }, + { + "epoch": 1.11, + "learning_rate": 1.4556661900133969e-05, + "loss": 1.2664, + "step": 13145 + }, + { + "epoch": 1.11, + "learning_rate": 1.4555852758286654e-05, + "loss": 1.2603, + "step": 13146 + }, + { + "epoch": 1.11, + "learning_rate": 1.4555043578797423e-05, + "loss": 1.1642, + "step": 13147 + }, + { + "epoch": 1.11, + "learning_rate": 1.4554234361672963e-05, + "loss": 1.1517, + "step": 13148 + }, + { + "epoch": 1.11, + "learning_rate": 1.4553425106919965e-05, + "loss": 1.1917, + "step": 13149 + }, + { + "epoch": 1.11, + "learning_rate": 1.455261581454511e-05, + "loss": 1.1589, + "step": 13150 + }, + { + "epoch": 1.11, + "learning_rate": 1.4551806484555087e-05, + "loss": 1.2385, + "step": 13151 + }, + { + "epoch": 1.11, + "learning_rate": 1.455099711695658e-05, + "loss": 1.2684, + "step": 13152 + }, + { + "epoch": 1.11, + "learning_rate": 1.4550187711756283e-05, + "loss": 1.1571, + "step": 13153 + }, + { + "epoch": 1.11, + "learning_rate": 1.4549378268960874e-05, + "loss": 1.1633, + "step": 13154 + }, + { + "epoch": 1.11, + "learning_rate": 1.4548568788577046e-05, + "loss": 1.2103, + "step": 13155 + }, + { + "epoch": 1.11, + "learning_rate": 1.4547759270611488e-05, + "loss": 1.2066, + "step": 13156 + }, + { + "epoch": 1.11, + "learning_rate": 1.4546949715070891e-05, + "loss": 1.1884, + "step": 13157 + }, + { + "epoch": 1.11, + "learning_rate": 1.4546140121961935e-05, + "loss": 1.2682, + "step": 13158 + }, + { + "epoch": 1.11, + "learning_rate": 1.4545330491291313e-05, + "loss": 1.1749, + "step": 13159 + }, + { + "epoch": 1.11, + "learning_rate": 1.4544520823065722e-05, + "loss": 1.1887, + "step": 13160 + }, + { + "epoch": 1.11, + "learning_rate": 1.4543711117291843e-05, + "loss": 1.222, + "step": 13161 + }, + { + "epoch": 1.11, + "learning_rate": 1.4542901373976368e-05, + "loss": 1.2012, + "step": 13162 + }, + { + "epoch": 1.11, + "learning_rate": 1.4542091593125986e-05, + "loss": 1.1577, + "step": 13163 + }, + { + "epoch": 1.11, + "learning_rate": 1.4541281774747393e-05, + "loss": 1.2334, + "step": 13164 + }, + { + "epoch": 1.11, + "learning_rate": 1.4540471918847273e-05, + "loss": 1.2248, + "step": 13165 + }, + { + "epoch": 1.11, + "learning_rate": 1.4539662025432324e-05, + "loss": 1.1118, + "step": 13166 + }, + { + "epoch": 1.11, + "learning_rate": 1.4538852094509235e-05, + "loss": 1.1477, + "step": 13167 + }, + { + "epoch": 1.11, + "learning_rate": 1.4538042126084697e-05, + "loss": 1.2031, + "step": 13168 + }, + { + "epoch": 1.11, + "learning_rate": 1.4537232120165401e-05, + "loss": 1.1973, + "step": 13169 + }, + { + "epoch": 1.11, + "learning_rate": 1.4536422076758041e-05, + "loss": 1.1713, + "step": 13170 + }, + { + "epoch": 1.11, + "learning_rate": 1.453561199586931e-05, + "loss": 1.1812, + "step": 13171 + }, + { + "epoch": 1.11, + "learning_rate": 1.4534801877505902e-05, + "loss": 1.2044, + "step": 13172 + }, + { + "epoch": 1.11, + "learning_rate": 1.4533991721674508e-05, + "loss": 1.2269, + "step": 13173 + }, + { + "epoch": 1.11, + "learning_rate": 1.4533181528381825e-05, + "loss": 1.1596, + "step": 13174 + }, + { + "epoch": 1.11, + "learning_rate": 1.4532371297634546e-05, + "loss": 1.1927, + "step": 13175 + }, + { + "epoch": 1.11, + "learning_rate": 1.4531561029439365e-05, + "loss": 1.1787, + "step": 13176 + }, + { + "epoch": 1.11, + "learning_rate": 1.4530750723802975e-05, + "loss": 1.1843, + "step": 13177 + }, + { + "epoch": 1.11, + "learning_rate": 1.4529940380732073e-05, + "loss": 1.1843, + "step": 13178 + }, + { + "epoch": 1.11, + "learning_rate": 1.4529130000233354e-05, + "loss": 1.1189, + "step": 13179 + }, + { + "epoch": 1.11, + "learning_rate": 1.4528319582313513e-05, + "loss": 1.2265, + "step": 13180 + }, + { + "epoch": 1.11, + "learning_rate": 1.4527509126979245e-05, + "loss": 1.2074, + "step": 13181 + }, + { + "epoch": 1.11, + "learning_rate": 1.4526698634237251e-05, + "loss": 1.2307, + "step": 13182 + }, + { + "epoch": 1.11, + "learning_rate": 1.4525888104094222e-05, + "loss": 1.2791, + "step": 13183 + }, + { + "epoch": 1.11, + "learning_rate": 1.4525077536556855e-05, + "loss": 1.2728, + "step": 13184 + }, + { + "epoch": 1.11, + "learning_rate": 1.4524266931631851e-05, + "loss": 1.2068, + "step": 13185 + }, + { + "epoch": 1.11, + "learning_rate": 1.4523456289325908e-05, + "loss": 1.1457, + "step": 13186 + }, + { + "epoch": 1.11, + "learning_rate": 1.4522645609645718e-05, + "loss": 1.166, + "step": 13187 + }, + { + "epoch": 1.11, + "learning_rate": 1.4521834892597984e-05, + "loss": 1.1353, + "step": 13188 + }, + { + "epoch": 1.11, + "learning_rate": 1.4521024138189406e-05, + "loss": 1.1888, + "step": 13189 + }, + { + "epoch": 1.11, + "learning_rate": 1.4520213346426674e-05, + "loss": 1.1048, + "step": 13190 + }, + { + "epoch": 1.11, + "learning_rate": 1.4519402517316496e-05, + "loss": 1.1727, + "step": 13191 + }, + { + "epoch": 1.11, + "learning_rate": 1.4518591650865567e-05, + "loss": 1.259, + "step": 13192 + }, + { + "epoch": 1.11, + "learning_rate": 1.451778074708059e-05, + "loss": 1.2866, + "step": 13193 + }, + { + "epoch": 1.11, + "learning_rate": 1.451696980596826e-05, + "loss": 1.2232, + "step": 13194 + }, + { + "epoch": 1.11, + "learning_rate": 1.451615882753528e-05, + "loss": 1.2192, + "step": 13195 + }, + { + "epoch": 1.11, + "learning_rate": 1.451534781178835e-05, + "loss": 1.27, + "step": 13196 + }, + { + "epoch": 1.11, + "learning_rate": 1.4514536758734172e-05, + "loss": 1.2313, + "step": 13197 + }, + { + "epoch": 1.11, + "learning_rate": 1.4513725668379446e-05, + "loss": 1.1792, + "step": 13198 + }, + { + "epoch": 1.11, + "learning_rate": 1.4512914540730873e-05, + "loss": 1.1253, + "step": 13199 + }, + { + "epoch": 1.11, + "learning_rate": 1.4512103375795159e-05, + "loss": 1.2008, + "step": 13200 + }, + { + "epoch": 1.11, + "learning_rate": 1.4511292173579002e-05, + "loss": 1.2427, + "step": 13201 + }, + { + "epoch": 1.11, + "learning_rate": 1.4510480934089104e-05, + "loss": 1.14, + "step": 13202 + }, + { + "epoch": 1.11, + "learning_rate": 1.4509669657332171e-05, + "loss": 1.2055, + "step": 13203 + }, + { + "epoch": 1.11, + "learning_rate": 1.4508858343314903e-05, + "loss": 1.2221, + "step": 13204 + }, + { + "epoch": 1.11, + "learning_rate": 1.4508046992044e-05, + "loss": 1.2407, + "step": 13205 + }, + { + "epoch": 1.11, + "learning_rate": 1.4507235603526175e-05, + "loss": 1.1647, + "step": 13206 + }, + { + "epoch": 1.11, + "learning_rate": 1.4506424177768126e-05, + "loss": 1.1526, + "step": 13207 + }, + { + "epoch": 1.11, + "learning_rate": 1.4505612714776559e-05, + "loss": 1.1762, + "step": 13208 + }, + { + "epoch": 1.11, + "learning_rate": 1.4504801214558174e-05, + "loss": 1.2197, + "step": 13209 + }, + { + "epoch": 1.11, + "learning_rate": 1.4503989677119683e-05, + "loss": 1.1807, + "step": 13210 + }, + { + "epoch": 1.11, + "learning_rate": 1.4503178102467788e-05, + "loss": 1.2064, + "step": 13211 + }, + { + "epoch": 1.11, + "learning_rate": 1.4502366490609192e-05, + "loss": 1.1865, + "step": 13212 + }, + { + "epoch": 1.11, + "learning_rate": 1.4501554841550605e-05, + "loss": 1.2442, + "step": 13213 + }, + { + "epoch": 1.11, + "learning_rate": 1.4500743155298733e-05, + "loss": 1.2178, + "step": 13214 + }, + { + "epoch": 1.11, + "learning_rate": 1.4499931431860276e-05, + "loss": 1.1761, + "step": 13215 + }, + { + "epoch": 1.11, + "learning_rate": 1.4499119671241949e-05, + "loss": 1.2101, + "step": 13216 + }, + { + "epoch": 1.11, + "learning_rate": 1.4498307873450452e-05, + "loss": 1.1458, + "step": 13217 + }, + { + "epoch": 1.11, + "learning_rate": 1.44974960384925e-05, + "loss": 1.1907, + "step": 13218 + }, + { + "epoch": 1.11, + "learning_rate": 1.4496684166374795e-05, + "loss": 1.1428, + "step": 13219 + }, + { + "epoch": 1.11, + "learning_rate": 1.4495872257104044e-05, + "loss": 1.1693, + "step": 13220 + }, + { + "epoch": 1.11, + "learning_rate": 1.4495060310686959e-05, + "loss": 1.2488, + "step": 13221 + }, + { + "epoch": 1.11, + "learning_rate": 1.4494248327130247e-05, + "loss": 1.237, + "step": 13222 + }, + { + "epoch": 1.11, + "learning_rate": 1.4493436306440616e-05, + "loss": 1.272, + "step": 13223 + }, + { + "epoch": 1.11, + "learning_rate": 1.4492624248624778e-05, + "loss": 1.2776, + "step": 13224 + }, + { + "epoch": 1.11, + "learning_rate": 1.449181215368944e-05, + "loss": 1.1746, + "step": 13225 + }, + { + "epoch": 1.11, + "learning_rate": 1.4491000021641313e-05, + "loss": 1.1615, + "step": 13226 + }, + { + "epoch": 1.11, + "learning_rate": 1.4490187852487106e-05, + "loss": 1.1841, + "step": 13227 + }, + { + "epoch": 1.11, + "learning_rate": 1.448937564623353e-05, + "loss": 1.1953, + "step": 13228 + }, + { + "epoch": 1.11, + "learning_rate": 1.4488563402887294e-05, + "loss": 1.1691, + "step": 13229 + }, + { + "epoch": 1.11, + "learning_rate": 1.448775112245511e-05, + "loss": 1.2518, + "step": 13230 + }, + { + "epoch": 1.11, + "learning_rate": 1.4486938804943693e-05, + "loss": 1.1681, + "step": 13231 + }, + { + "epoch": 1.11, + "learning_rate": 1.4486126450359752e-05, + "loss": 1.2081, + "step": 13232 + }, + { + "epoch": 1.11, + "learning_rate": 1.4485314058709998e-05, + "loss": 1.1528, + "step": 13233 + }, + { + "epoch": 1.11, + "learning_rate": 1.448450163000114e-05, + "loss": 1.2348, + "step": 13234 + }, + { + "epoch": 1.11, + "learning_rate": 1.44836891642399e-05, + "loss": 1.2152, + "step": 13235 + }, + { + "epoch": 1.11, + "learning_rate": 1.4482876661432981e-05, + "loss": 1.1765, + "step": 13236 + }, + { + "epoch": 1.11, + "learning_rate": 1.44820641215871e-05, + "loss": 1.297, + "step": 13237 + }, + { + "epoch": 1.11, + "learning_rate": 1.4481251544708973e-05, + "loss": 1.2037, + "step": 13238 + }, + { + "epoch": 1.11, + "learning_rate": 1.4480438930805312e-05, + "loss": 1.0997, + "step": 13239 + }, + { + "epoch": 1.11, + "learning_rate": 1.447962627988283e-05, + "loss": 1.2399, + "step": 13240 + }, + { + "epoch": 1.11, + "learning_rate": 1.447881359194824e-05, + "loss": 1.1318, + "step": 13241 + }, + { + "epoch": 1.11, + "learning_rate": 1.4478000867008262e-05, + "loss": 1.0901, + "step": 13242 + }, + { + "epoch": 1.11, + "learning_rate": 1.4477188105069605e-05, + "loss": 1.2992, + "step": 13243 + }, + { + "epoch": 1.12, + "learning_rate": 1.447637530613899e-05, + "loss": 1.2421, + "step": 13244 + }, + { + "epoch": 1.12, + "learning_rate": 1.4475562470223126e-05, + "loss": 1.1242, + "step": 13245 + }, + { + "epoch": 1.12, + "learning_rate": 1.4474749597328735e-05, + "loss": 1.18, + "step": 13246 + }, + { + "epoch": 1.12, + "learning_rate": 1.4473936687462529e-05, + "loss": 1.1249, + "step": 13247 + }, + { + "epoch": 1.12, + "learning_rate": 1.4473123740631225e-05, + "loss": 1.2578, + "step": 13248 + }, + { + "epoch": 1.12, + "learning_rate": 1.4472310756841544e-05, + "loss": 1.1362, + "step": 13249 + }, + { + "epoch": 1.12, + "learning_rate": 1.4471497736100202e-05, + "loss": 1.1873, + "step": 13250 + }, + { + "epoch": 1.12, + "learning_rate": 1.4470684678413908e-05, + "loss": 1.2454, + "step": 13251 + }, + { + "epoch": 1.12, + "learning_rate": 1.4469871583789392e-05, + "loss": 1.1784, + "step": 13252 + }, + { + "epoch": 1.12, + "learning_rate": 1.4469058452233364e-05, + "loss": 1.125, + "step": 13253 + }, + { + "epoch": 1.12, + "learning_rate": 1.4468245283752547e-05, + "loss": 1.1767, + "step": 13254 + }, + { + "epoch": 1.12, + "learning_rate": 1.4467432078353655e-05, + "loss": 1.1614, + "step": 13255 + }, + { + "epoch": 1.12, + "learning_rate": 1.4466618836043412e-05, + "loss": 1.243, + "step": 13256 + }, + { + "epoch": 1.12, + "learning_rate": 1.4465805556828534e-05, + "loss": 1.137, + "step": 13257 + }, + { + "epoch": 1.12, + "learning_rate": 1.4464992240715741e-05, + "loss": 1.1127, + "step": 13258 + }, + { + "epoch": 1.12, + "learning_rate": 1.4464178887711748e-05, + "loss": 1.132, + "step": 13259 + }, + { + "epoch": 1.12, + "learning_rate": 1.4463365497823288e-05, + "loss": 1.1713, + "step": 13260 + }, + { + "epoch": 1.12, + "learning_rate": 1.446255207105707e-05, + "loss": 1.2735, + "step": 13261 + }, + { + "epoch": 1.12, + "learning_rate": 1.4461738607419817e-05, + "loss": 1.1865, + "step": 13262 + }, + { + "epoch": 1.12, + "learning_rate": 1.4460925106918256e-05, + "loss": 1.1793, + "step": 13263 + }, + { + "epoch": 1.12, + "learning_rate": 1.4460111569559102e-05, + "loss": 1.2531, + "step": 13264 + }, + { + "epoch": 1.12, + "learning_rate": 1.4459297995349078e-05, + "loss": 1.2158, + "step": 13265 + }, + { + "epoch": 1.12, + "learning_rate": 1.4458484384294908e-05, + "loss": 1.0965, + "step": 13266 + }, + { + "epoch": 1.12, + "learning_rate": 1.4457670736403312e-05, + "loss": 1.1806, + "step": 13267 + }, + { + "epoch": 1.12, + "learning_rate": 1.4456857051681015e-05, + "loss": 1.2729, + "step": 13268 + }, + { + "epoch": 1.12, + "learning_rate": 1.445604333013474e-05, + "loss": 1.1496, + "step": 13269 + }, + { + "epoch": 1.12, + "learning_rate": 1.4455229571771205e-05, + "loss": 1.1253, + "step": 13270 + }, + { + "epoch": 1.12, + "learning_rate": 1.445441577659714e-05, + "loss": 1.0833, + "step": 13271 + }, + { + "epoch": 1.12, + "learning_rate": 1.4453601944619265e-05, + "loss": 1.1881, + "step": 13272 + }, + { + "epoch": 1.12, + "learning_rate": 1.4452788075844306e-05, + "loss": 1.2653, + "step": 13273 + }, + { + "epoch": 1.12, + "learning_rate": 1.4451974170278986e-05, + "loss": 1.1675, + "step": 13274 + }, + { + "epoch": 1.12, + "learning_rate": 1.4451160227930032e-05, + "loss": 1.2884, + "step": 13275 + }, + { + "epoch": 1.12, + "learning_rate": 1.4450346248804168e-05, + "loss": 1.1613, + "step": 13276 + }, + { + "epoch": 1.12, + "learning_rate": 1.4449532232908116e-05, + "loss": 1.3398, + "step": 13277 + }, + { + "epoch": 1.12, + "learning_rate": 1.4448718180248606e-05, + "loss": 1.2251, + "step": 13278 + }, + { + "epoch": 1.12, + "learning_rate": 1.4447904090832366e-05, + "loss": 1.1657, + "step": 13279 + }, + { + "epoch": 1.12, + "learning_rate": 1.4447089964666117e-05, + "loss": 1.1605, + "step": 13280 + }, + { + "epoch": 1.12, + "learning_rate": 1.4446275801756584e-05, + "loss": 1.2477, + "step": 13281 + }, + { + "epoch": 1.12, + "learning_rate": 1.4445461602110502e-05, + "loss": 1.2054, + "step": 13282 + }, + { + "epoch": 1.12, + "learning_rate": 1.4444647365734592e-05, + "loss": 1.1475, + "step": 13283 + }, + { + "epoch": 1.12, + "learning_rate": 1.444383309263558e-05, + "loss": 1.1305, + "step": 13284 + }, + { + "epoch": 1.12, + "learning_rate": 1.4443018782820203e-05, + "loss": 1.1901, + "step": 13285 + }, + { + "epoch": 1.12, + "learning_rate": 1.4442204436295178e-05, + "loss": 1.1321, + "step": 13286 + }, + { + "epoch": 1.12, + "learning_rate": 1.4441390053067239e-05, + "loss": 1.2701, + "step": 13287 + }, + { + "epoch": 1.12, + "learning_rate": 1.4440575633143112e-05, + "loss": 1.2784, + "step": 13288 + }, + { + "epoch": 1.12, + "learning_rate": 1.4439761176529534e-05, + "loss": 1.1782, + "step": 13289 + }, + { + "epoch": 1.12, + "learning_rate": 1.4438946683233224e-05, + "loss": 1.1466, + "step": 13290 + }, + { + "epoch": 1.12, + "learning_rate": 1.4438132153260916e-05, + "loss": 1.2403, + "step": 13291 + }, + { + "epoch": 1.12, + "learning_rate": 1.4437317586619339e-05, + "loss": 1.1079, + "step": 13292 + }, + { + "epoch": 1.12, + "learning_rate": 1.4436502983315228e-05, + "loss": 1.2027, + "step": 13293 + }, + { + "epoch": 1.12, + "learning_rate": 1.4435688343355305e-05, + "loss": 1.1451, + "step": 13294 + }, + { + "epoch": 1.12, + "learning_rate": 1.4434873666746307e-05, + "loss": 1.0981, + "step": 13295 + }, + { + "epoch": 1.12, + "learning_rate": 1.4434058953494963e-05, + "loss": 1.2243, + "step": 13296 + }, + { + "epoch": 1.12, + "learning_rate": 1.4433244203608006e-05, + "loss": 1.1881, + "step": 13297 + }, + { + "epoch": 1.12, + "learning_rate": 1.4432429417092164e-05, + "loss": 1.1877, + "step": 13298 + }, + { + "epoch": 1.12, + "learning_rate": 1.4431614593954172e-05, + "loss": 1.2258, + "step": 13299 + }, + { + "epoch": 1.12, + "learning_rate": 1.4430799734200767e-05, + "loss": 1.2394, + "step": 13300 + }, + { + "epoch": 1.12, + "learning_rate": 1.4429984837838673e-05, + "loss": 1.2252, + "step": 13301 + }, + { + "epoch": 1.12, + "learning_rate": 1.4429169904874624e-05, + "loss": 1.3077, + "step": 13302 + }, + { + "epoch": 1.12, + "learning_rate": 1.442835493531536e-05, + "loss": 1.2371, + "step": 13303 + }, + { + "epoch": 1.12, + "learning_rate": 1.4427539929167608e-05, + "loss": 1.175, + "step": 13304 + }, + { + "epoch": 1.12, + "learning_rate": 1.4426724886438107e-05, + "loss": 1.2101, + "step": 13305 + }, + { + "epoch": 1.12, + "learning_rate": 1.4425909807133584e-05, + "loss": 1.2487, + "step": 13306 + }, + { + "epoch": 1.12, + "learning_rate": 1.4425094691260782e-05, + "loss": 1.2056, + "step": 13307 + }, + { + "epoch": 1.12, + "learning_rate": 1.4424279538826427e-05, + "loss": 1.0985, + "step": 13308 + }, + { + "epoch": 1.12, + "learning_rate": 1.4423464349837257e-05, + "loss": 1.2364, + "step": 13309 + }, + { + "epoch": 1.12, + "learning_rate": 1.4422649124300017e-05, + "loss": 1.2692, + "step": 13310 + }, + { + "epoch": 1.12, + "learning_rate": 1.4421833862221429e-05, + "loss": 1.2267, + "step": 13311 + }, + { + "epoch": 1.12, + "learning_rate": 1.4421018563608234e-05, + "loss": 1.1806, + "step": 13312 + }, + { + "epoch": 1.12, + "learning_rate": 1.4420203228467169e-05, + "loss": 1.2059, + "step": 13313 + }, + { + "epoch": 1.12, + "learning_rate": 1.4419387856804974e-05, + "loss": 1.1834, + "step": 13314 + }, + { + "epoch": 1.12, + "learning_rate": 1.4418572448628378e-05, + "loss": 1.1912, + "step": 13315 + }, + { + "epoch": 1.12, + "learning_rate": 1.4417757003944124e-05, + "loss": 1.303, + "step": 13316 + }, + { + "epoch": 1.12, + "learning_rate": 1.4416941522758948e-05, + "loss": 1.2071, + "step": 13317 + }, + { + "epoch": 1.12, + "learning_rate": 1.4416126005079587e-05, + "loss": 1.2055, + "step": 13318 + }, + { + "epoch": 1.12, + "learning_rate": 1.4415310450912778e-05, + "loss": 1.1603, + "step": 13319 + }, + { + "epoch": 1.12, + "learning_rate": 1.4414494860265265e-05, + "loss": 1.243, + "step": 13320 + }, + { + "epoch": 1.12, + "learning_rate": 1.4413679233143779e-05, + "loss": 1.2111, + "step": 13321 + }, + { + "epoch": 1.12, + "learning_rate": 1.4412863569555065e-05, + "loss": 1.2508, + "step": 13322 + }, + { + "epoch": 1.12, + "learning_rate": 1.4412047869505857e-05, + "loss": 1.2184, + "step": 13323 + }, + { + "epoch": 1.12, + "learning_rate": 1.44112321330029e-05, + "loss": 1.1797, + "step": 13324 + }, + { + "epoch": 1.12, + "learning_rate": 1.441041636005293e-05, + "loss": 1.2633, + "step": 13325 + }, + { + "epoch": 1.12, + "learning_rate": 1.440960055066269e-05, + "loss": 1.2995, + "step": 13326 + }, + { + "epoch": 1.12, + "learning_rate": 1.4408784704838915e-05, + "loss": 1.2044, + "step": 13327 + }, + { + "epoch": 1.12, + "learning_rate": 1.4407968822588355e-05, + "loss": 1.2041, + "step": 13328 + }, + { + "epoch": 1.12, + "learning_rate": 1.4407152903917746e-05, + "loss": 1.235, + "step": 13329 + }, + { + "epoch": 1.12, + "learning_rate": 1.4406336948833825e-05, + "loss": 1.2075, + "step": 13330 + }, + { + "epoch": 1.12, + "learning_rate": 1.440552095734334e-05, + "loss": 1.1312, + "step": 13331 + }, + { + "epoch": 1.12, + "learning_rate": 1.4404704929453033e-05, + "loss": 1.2159, + "step": 13332 + }, + { + "epoch": 1.12, + "learning_rate": 1.440388886516964e-05, + "loss": 1.2044, + "step": 13333 + }, + { + "epoch": 1.12, + "learning_rate": 1.4403072764499908e-05, + "loss": 1.1869, + "step": 13334 + }, + { + "epoch": 1.12, + "learning_rate": 1.4402256627450583e-05, + "loss": 1.1846, + "step": 13335 + }, + { + "epoch": 1.12, + "learning_rate": 1.4401440454028403e-05, + "loss": 1.2367, + "step": 13336 + }, + { + "epoch": 1.12, + "learning_rate": 1.4400624244240114e-05, + "loss": 1.2324, + "step": 13337 + }, + { + "epoch": 1.12, + "learning_rate": 1.4399807998092459e-05, + "loss": 1.1664, + "step": 13338 + }, + { + "epoch": 1.12, + "learning_rate": 1.4398991715592181e-05, + "loss": 1.2036, + "step": 13339 + }, + { + "epoch": 1.12, + "learning_rate": 1.4398175396746027e-05, + "loss": 1.1485, + "step": 13340 + }, + { + "epoch": 1.12, + "learning_rate": 1.4397359041560739e-05, + "loss": 1.2405, + "step": 13341 + }, + { + "epoch": 1.12, + "learning_rate": 1.4396542650043064e-05, + "loss": 1.2406, + "step": 13342 + }, + { + "epoch": 1.12, + "learning_rate": 1.4395726222199749e-05, + "loss": 1.2449, + "step": 13343 + }, + { + "epoch": 1.12, + "learning_rate": 1.4394909758037535e-05, + "loss": 1.2133, + "step": 13344 + }, + { + "epoch": 1.12, + "learning_rate": 1.4394093257563169e-05, + "loss": 1.1944, + "step": 13345 + }, + { + "epoch": 1.12, + "learning_rate": 1.4393276720783397e-05, + "loss": 1.2022, + "step": 13346 + }, + { + "epoch": 1.12, + "learning_rate": 1.4392460147704967e-05, + "loss": 1.2127, + "step": 13347 + }, + { + "epoch": 1.12, + "learning_rate": 1.439164353833463e-05, + "loss": 1.1539, + "step": 13348 + }, + { + "epoch": 1.12, + "learning_rate": 1.4390826892679123e-05, + "loss": 1.1813, + "step": 13349 + }, + { + "epoch": 1.12, + "learning_rate": 1.4390010210745204e-05, + "loss": 1.1783, + "step": 13350 + }, + { + "epoch": 1.12, + "learning_rate": 1.4389193492539613e-05, + "loss": 1.205, + "step": 13351 + }, + { + "epoch": 1.12, + "learning_rate": 1.4388376738069101e-05, + "loss": 1.2047, + "step": 13352 + }, + { + "epoch": 1.12, + "learning_rate": 1.4387559947340416e-05, + "loss": 1.1145, + "step": 13353 + }, + { + "epoch": 1.12, + "learning_rate": 1.4386743120360307e-05, + "loss": 1.2067, + "step": 13354 + }, + { + "epoch": 1.12, + "learning_rate": 1.4385926257135522e-05, + "loss": 1.2087, + "step": 13355 + }, + { + "epoch": 1.12, + "learning_rate": 1.4385109357672811e-05, + "loss": 1.1959, + "step": 13356 + }, + { + "epoch": 1.12, + "learning_rate": 1.4384292421978923e-05, + "loss": 1.1305, + "step": 13357 + }, + { + "epoch": 1.12, + "learning_rate": 1.4383475450060606e-05, + "loss": 1.2001, + "step": 13358 + }, + { + "epoch": 1.12, + "learning_rate": 1.4382658441924613e-05, + "loss": 1.1471, + "step": 13359 + }, + { + "epoch": 1.12, + "learning_rate": 1.4381841397577694e-05, + "loss": 1.2563, + "step": 13360 + }, + { + "epoch": 1.12, + "learning_rate": 1.4381024317026598e-05, + "loss": 1.2633, + "step": 13361 + }, + { + "epoch": 1.12, + "learning_rate": 1.4380207200278078e-05, + "loss": 1.3716, + "step": 13362 + }, + { + "epoch": 1.13, + "learning_rate": 1.4379390047338882e-05, + "loss": 1.2095, + "step": 13363 + }, + { + "epoch": 1.13, + "learning_rate": 1.437857285821577e-05, + "loss": 1.0782, + "step": 13364 + }, + { + "epoch": 1.13, + "learning_rate": 1.437775563291548e-05, + "loss": 1.1871, + "step": 13365 + }, + { + "epoch": 1.13, + "learning_rate": 1.4376938371444778e-05, + "loss": 1.203, + "step": 13366 + }, + { + "epoch": 1.13, + "learning_rate": 1.4376121073810406e-05, + "loss": 1.2251, + "step": 13367 + }, + { + "epoch": 1.13, + "learning_rate": 1.4375303740019124e-05, + "loss": 1.171, + "step": 13368 + }, + { + "epoch": 1.13, + "learning_rate": 1.4374486370077681e-05, + "loss": 1.2153, + "step": 13369 + }, + { + "epoch": 1.13, + "learning_rate": 1.437366896399283e-05, + "loss": 1.1863, + "step": 13370 + }, + { + "epoch": 1.13, + "learning_rate": 1.437285152177133e-05, + "loss": 1.201, + "step": 13371 + }, + { + "epoch": 1.13, + "learning_rate": 1.4372034043419927e-05, + "loss": 1.2634, + "step": 13372 + }, + { + "epoch": 1.13, + "learning_rate": 1.4371216528945381e-05, + "loss": 1.1445, + "step": 13373 + }, + { + "epoch": 1.13, + "learning_rate": 1.4370398978354445e-05, + "loss": 1.1886, + "step": 13374 + }, + { + "epoch": 1.13, + "learning_rate": 1.4369581391653878e-05, + "loss": 1.2273, + "step": 13375 + }, + { + "epoch": 1.13, + "learning_rate": 1.4368763768850426e-05, + "loss": 1.2683, + "step": 13376 + }, + { + "epoch": 1.13, + "learning_rate": 1.4367946109950852e-05, + "loss": 1.1784, + "step": 13377 + }, + { + "epoch": 1.13, + "learning_rate": 1.4367128414961908e-05, + "loss": 1.1975, + "step": 13378 + }, + { + "epoch": 1.13, + "learning_rate": 1.4366310683890352e-05, + "loss": 1.1975, + "step": 13379 + }, + { + "epoch": 1.13, + "learning_rate": 1.436549291674294e-05, + "loss": 1.2218, + "step": 13380 + }, + { + "epoch": 1.13, + "learning_rate": 1.436467511352643e-05, + "loss": 1.1994, + "step": 13381 + }, + { + "epoch": 1.13, + "learning_rate": 1.4363857274247575e-05, + "loss": 1.1442, + "step": 13382 + }, + { + "epoch": 1.13, + "learning_rate": 1.4363039398913134e-05, + "loss": 1.3031, + "step": 13383 + }, + { + "epoch": 1.13, + "learning_rate": 1.4362221487529864e-05, + "loss": 1.1764, + "step": 13384 + }, + { + "epoch": 1.13, + "learning_rate": 1.4361403540104529e-05, + "loss": 1.2181, + "step": 13385 + }, + { + "epoch": 1.13, + "learning_rate": 1.436058555664388e-05, + "loss": 1.2041, + "step": 13386 + }, + { + "epoch": 1.13, + "learning_rate": 1.4359767537154677e-05, + "loss": 1.1849, + "step": 13387 + }, + { + "epoch": 1.13, + "learning_rate": 1.4358949481643677e-05, + "loss": 1.1456, + "step": 13388 + }, + { + "epoch": 1.13, + "learning_rate": 1.4358131390117645e-05, + "loss": 1.2886, + "step": 13389 + }, + { + "epoch": 1.13, + "learning_rate": 1.4357313262583335e-05, + "loss": 1.2718, + "step": 13390 + }, + { + "epoch": 1.13, + "learning_rate": 1.4356495099047506e-05, + "loss": 1.1301, + "step": 13391 + }, + { + "epoch": 1.13, + "learning_rate": 1.4355676899516925e-05, + "loss": 1.1368, + "step": 13392 + }, + { + "epoch": 1.13, + "learning_rate": 1.4354858663998345e-05, + "loss": 1.1238, + "step": 13393 + }, + { + "epoch": 1.13, + "learning_rate": 1.435404039249853e-05, + "loss": 1.2765, + "step": 13394 + }, + { + "epoch": 1.13, + "learning_rate": 1.4353222085024239e-05, + "loss": 1.1528, + "step": 13395 + }, + { + "epoch": 1.13, + "learning_rate": 1.4352403741582233e-05, + "loss": 1.2219, + "step": 13396 + }, + { + "epoch": 1.13, + "learning_rate": 1.4351585362179275e-05, + "loss": 1.2515, + "step": 13397 + }, + { + "epoch": 1.13, + "learning_rate": 1.4350766946822126e-05, + "loss": 1.1482, + "step": 13398 + }, + { + "epoch": 1.13, + "learning_rate": 1.4349948495517548e-05, + "loss": 1.2228, + "step": 13399 + }, + { + "epoch": 1.13, + "learning_rate": 1.4349130008272305e-05, + "loss": 1.2589, + "step": 13400 + }, + { + "epoch": 1.13, + "learning_rate": 1.4348311485093156e-05, + "loss": 1.1836, + "step": 13401 + }, + { + "epoch": 1.13, + "learning_rate": 1.4347492925986868e-05, + "loss": 1.2439, + "step": 13402 + }, + { + "epoch": 1.13, + "learning_rate": 1.4346674330960199e-05, + "loss": 1.1722, + "step": 13403 + }, + { + "epoch": 1.13, + "learning_rate": 1.434585570001992e-05, + "loss": 1.221, + "step": 13404 + }, + { + "epoch": 1.13, + "learning_rate": 1.4345037033172787e-05, + "loss": 1.1785, + "step": 13405 + }, + { + "epoch": 1.13, + "learning_rate": 1.4344218330425567e-05, + "loss": 1.0886, + "step": 13406 + }, + { + "epoch": 1.13, + "learning_rate": 1.4343399591785025e-05, + "loss": 1.2445, + "step": 13407 + }, + { + "epoch": 1.13, + "learning_rate": 1.4342580817257927e-05, + "loss": 1.2021, + "step": 13408 + }, + { + "epoch": 1.13, + "learning_rate": 1.4341762006851034e-05, + "loss": 1.224, + "step": 13409 + }, + { + "epoch": 1.13, + "learning_rate": 1.4340943160571117e-05, + "loss": 1.2679, + "step": 13410 + }, + { + "epoch": 1.13, + "learning_rate": 1.4340124278424936e-05, + "loss": 1.2211, + "step": 13411 + }, + { + "epoch": 1.13, + "learning_rate": 1.433930536041926e-05, + "loss": 1.224, + "step": 13412 + }, + { + "epoch": 1.13, + "learning_rate": 1.4338486406560854e-05, + "loss": 1.1554, + "step": 13413 + }, + { + "epoch": 1.13, + "learning_rate": 1.4337667416856489e-05, + "loss": 1.1599, + "step": 13414 + }, + { + "epoch": 1.13, + "learning_rate": 1.4336848391312922e-05, + "loss": 1.3084, + "step": 13415 + }, + { + "epoch": 1.13, + "learning_rate": 1.4336029329936926e-05, + "loss": 1.2266, + "step": 13416 + }, + { + "epoch": 1.13, + "learning_rate": 1.433521023273527e-05, + "loss": 1.2536, + "step": 13417 + }, + { + "epoch": 1.13, + "learning_rate": 1.4334391099714723e-05, + "loss": 1.2337, + "step": 13418 + }, + { + "epoch": 1.13, + "learning_rate": 1.4333571930882042e-05, + "loss": 1.1342, + "step": 13419 + }, + { + "epoch": 1.13, + "learning_rate": 1.4332752726244007e-05, + "loss": 1.1311, + "step": 13420 + }, + { + "epoch": 1.13, + "learning_rate": 1.4331933485807381e-05, + "loss": 1.1663, + "step": 13421 + }, + { + "epoch": 1.13, + "learning_rate": 1.4331114209578935e-05, + "loss": 1.1298, + "step": 13422 + }, + { + "epoch": 1.13, + "learning_rate": 1.4330294897565438e-05, + "loss": 1.1457, + "step": 13423 + }, + { + "epoch": 1.13, + "learning_rate": 1.4329475549773655e-05, + "loss": 1.2219, + "step": 13424 + }, + { + "epoch": 1.13, + "learning_rate": 1.4328656166210367e-05, + "loss": 1.214, + "step": 13425 + }, + { + "epoch": 1.13, + "learning_rate": 1.4327836746882328e-05, + "loss": 1.1495, + "step": 13426 + }, + { + "epoch": 1.13, + "learning_rate": 1.4327017291796321e-05, + "loss": 1.2022, + "step": 13427 + }, + { + "epoch": 1.13, + "learning_rate": 1.4326197800959112e-05, + "loss": 1.1893, + "step": 13428 + }, + { + "epoch": 1.13, + "learning_rate": 1.4325378274377473e-05, + "loss": 1.1924, + "step": 13429 + }, + { + "epoch": 1.13, + "learning_rate": 1.4324558712058173e-05, + "loss": 1.2368, + "step": 13430 + }, + { + "epoch": 1.13, + "learning_rate": 1.4323739114007983e-05, + "loss": 1.2259, + "step": 13431 + }, + { + "epoch": 1.13, + "learning_rate": 1.432291948023368e-05, + "loss": 1.1955, + "step": 13432 + }, + { + "epoch": 1.13, + "learning_rate": 1.4322099810742031e-05, + "loss": 1.2532, + "step": 13433 + }, + { + "epoch": 1.13, + "learning_rate": 1.432128010553981e-05, + "loss": 1.187, + "step": 13434 + }, + { + "epoch": 1.13, + "learning_rate": 1.4320460364633791e-05, + "loss": 1.194, + "step": 13435 + }, + { + "epoch": 1.13, + "learning_rate": 1.4319640588030745e-05, + "loss": 1.1436, + "step": 13436 + }, + { + "epoch": 1.13, + "learning_rate": 1.4318820775737448e-05, + "loss": 1.1543, + "step": 13437 + }, + { + "epoch": 1.13, + "learning_rate": 1.4318000927760668e-05, + "loss": 1.2151, + "step": 13438 + }, + { + "epoch": 1.13, + "learning_rate": 1.4317181044107185e-05, + "loss": 1.2183, + "step": 13439 + }, + { + "epoch": 1.13, + "learning_rate": 1.4316361124783771e-05, + "loss": 1.2642, + "step": 13440 + }, + { + "epoch": 1.13, + "learning_rate": 1.4315541169797198e-05, + "loss": 1.1776, + "step": 13441 + }, + { + "epoch": 1.13, + "learning_rate": 1.4314721179154242e-05, + "loss": 1.1341, + "step": 13442 + }, + { + "epoch": 1.13, + "learning_rate": 1.4313901152861683e-05, + "loss": 1.2509, + "step": 13443 + }, + { + "epoch": 1.13, + "learning_rate": 1.4313081090926288e-05, + "loss": 1.153, + "step": 13444 + }, + { + "epoch": 1.13, + "learning_rate": 1.4312260993354837e-05, + "loss": 1.1625, + "step": 13445 + }, + { + "epoch": 1.13, + "learning_rate": 1.4311440860154107e-05, + "loss": 1.2519, + "step": 13446 + }, + { + "epoch": 1.13, + "learning_rate": 1.431062069133087e-05, + "loss": 1.1051, + "step": 13447 + }, + { + "epoch": 1.13, + "learning_rate": 1.4309800486891907e-05, + "loss": 1.1867, + "step": 13448 + }, + { + "epoch": 1.13, + "learning_rate": 1.4308980246843993e-05, + "loss": 1.2508, + "step": 13449 + }, + { + "epoch": 1.13, + "learning_rate": 1.4308159971193907e-05, + "loss": 1.1594, + "step": 13450 + }, + { + "epoch": 1.13, + "learning_rate": 1.4307339659948426e-05, + "loss": 1.2163, + "step": 13451 + }, + { + "epoch": 1.13, + "learning_rate": 1.430651931311432e-05, + "loss": 1.1906, + "step": 13452 + }, + { + "epoch": 1.13, + "learning_rate": 1.4305698930698378e-05, + "loss": 1.2603, + "step": 13453 + }, + { + "epoch": 1.13, + "learning_rate": 1.4304878512707372e-05, + "loss": 1.1463, + "step": 13454 + }, + { + "epoch": 1.13, + "learning_rate": 1.4304058059148083e-05, + "loss": 1.1734, + "step": 13455 + }, + { + "epoch": 1.13, + "learning_rate": 1.4303237570027287e-05, + "loss": 1.2606, + "step": 13456 + }, + { + "epoch": 1.13, + "learning_rate": 1.4302417045351764e-05, + "loss": 1.2153, + "step": 13457 + }, + { + "epoch": 1.13, + "learning_rate": 1.43015964851283e-05, + "loss": 1.1819, + "step": 13458 + }, + { + "epoch": 1.13, + "learning_rate": 1.4300775889363661e-05, + "loss": 1.153, + "step": 13459 + }, + { + "epoch": 1.13, + "learning_rate": 1.429995525806464e-05, + "loss": 1.1656, + "step": 13460 + }, + { + "epoch": 1.13, + "learning_rate": 1.4299134591238013e-05, + "loss": 1.1257, + "step": 13461 + }, + { + "epoch": 1.13, + "learning_rate": 1.4298313888890558e-05, + "loss": 1.1139, + "step": 13462 + }, + { + "epoch": 1.13, + "learning_rate": 1.4297493151029058e-05, + "loss": 1.2915, + "step": 13463 + }, + { + "epoch": 1.13, + "learning_rate": 1.4296672377660299e-05, + "loss": 1.1202, + "step": 13464 + }, + { + "epoch": 1.13, + "learning_rate": 1.4295851568791053e-05, + "loss": 1.0916, + "step": 13465 + }, + { + "epoch": 1.13, + "learning_rate": 1.4295030724428108e-05, + "loss": 1.2498, + "step": 13466 + }, + { + "epoch": 1.13, + "learning_rate": 1.4294209844578241e-05, + "loss": 1.2595, + "step": 13467 + }, + { + "epoch": 1.13, + "learning_rate": 1.4293388929248244e-05, + "loss": 1.2303, + "step": 13468 + }, + { + "epoch": 1.13, + "learning_rate": 1.429256797844489e-05, + "loss": 1.2399, + "step": 13469 + }, + { + "epoch": 1.13, + "learning_rate": 1.4291746992174967e-05, + "loss": 1.2755, + "step": 13470 + }, + { + "epoch": 1.13, + "learning_rate": 1.4290925970445255e-05, + "loss": 1.113, + "step": 13471 + }, + { + "epoch": 1.13, + "learning_rate": 1.429010491326254e-05, + "loss": 1.2224, + "step": 13472 + }, + { + "epoch": 1.13, + "learning_rate": 1.4289283820633606e-05, + "loss": 1.2516, + "step": 13473 + }, + { + "epoch": 1.13, + "learning_rate": 1.4288462692565236e-05, + "loss": 1.2439, + "step": 13474 + }, + { + "epoch": 1.13, + "learning_rate": 1.4287641529064215e-05, + "loss": 1.2153, + "step": 13475 + }, + { + "epoch": 1.13, + "learning_rate": 1.4286820330137327e-05, + "loss": 1.3005, + "step": 13476 + }, + { + "epoch": 1.13, + "learning_rate": 1.4285999095791358e-05, + "loss": 1.1562, + "step": 13477 + }, + { + "epoch": 1.13, + "learning_rate": 1.4285177826033092e-05, + "loss": 1.118, + "step": 13478 + }, + { + "epoch": 1.13, + "learning_rate": 1.4284356520869316e-05, + "loss": 1.2376, + "step": 13479 + }, + { + "epoch": 1.13, + "learning_rate": 1.4283535180306816e-05, + "loss": 1.2169, + "step": 13480 + }, + { + "epoch": 1.13, + "learning_rate": 1.4282713804352375e-05, + "loss": 1.1858, + "step": 13481 + }, + { + "epoch": 1.14, + "learning_rate": 1.4281892393012784e-05, + "loss": 1.156, + "step": 13482 + }, + { + "epoch": 1.14, + "learning_rate": 1.4281070946294828e-05, + "loss": 1.187, + "step": 13483 + }, + { + "epoch": 1.14, + "learning_rate": 1.4280249464205289e-05, + "loss": 1.0788, + "step": 13484 + }, + { + "epoch": 1.14, + "learning_rate": 1.4279427946750963e-05, + "loss": 1.2158, + "step": 13485 + }, + { + "epoch": 1.14, + "learning_rate": 1.4278606393938636e-05, + "loss": 1.3102, + "step": 13486 + }, + { + "epoch": 1.14, + "learning_rate": 1.427778480577509e-05, + "loss": 1.2248, + "step": 13487 + }, + { + "epoch": 1.14, + "learning_rate": 1.4276963182267119e-05, + "loss": 1.2807, + "step": 13488 + }, + { + "epoch": 1.14, + "learning_rate": 1.427614152342151e-05, + "loss": 1.264, + "step": 13489 + }, + { + "epoch": 1.14, + "learning_rate": 1.427531982924505e-05, + "loss": 1.2751, + "step": 13490 + }, + { + "epoch": 1.14, + "learning_rate": 1.4274498099744528e-05, + "loss": 1.1666, + "step": 13491 + }, + { + "epoch": 1.14, + "learning_rate": 1.4273676334926734e-05, + "loss": 1.166, + "step": 13492 + }, + { + "epoch": 1.14, + "learning_rate": 1.4272854534798462e-05, + "loss": 1.2835, + "step": 13493 + }, + { + "epoch": 1.14, + "learning_rate": 1.4272032699366496e-05, + "loss": 1.1875, + "step": 13494 + }, + { + "epoch": 1.14, + "learning_rate": 1.4271210828637627e-05, + "loss": 1.1859, + "step": 13495 + }, + { + "epoch": 1.14, + "learning_rate": 1.427038892261865e-05, + "loss": 1.2408, + "step": 13496 + }, + { + "epoch": 1.14, + "learning_rate": 1.4269566981316352e-05, + "loss": 1.1561, + "step": 13497 + }, + { + "epoch": 1.14, + "learning_rate": 1.4268745004737526e-05, + "loss": 1.2261, + "step": 13498 + }, + { + "epoch": 1.14, + "learning_rate": 1.4267922992888961e-05, + "loss": 1.1366, + "step": 13499 + }, + { + "epoch": 1.14, + "learning_rate": 1.4267100945777452e-05, + "loss": 1.2276, + "step": 13500 + }, + { + "epoch": 1.14, + "learning_rate": 1.4266278863409788e-05, + "loss": 1.218, + "step": 13501 + }, + { + "epoch": 1.14, + "learning_rate": 1.4265456745792763e-05, + "loss": 1.2702, + "step": 13502 + }, + { + "epoch": 1.14, + "learning_rate": 1.4264634592933168e-05, + "loss": 1.2985, + "step": 13503 + }, + { + "epoch": 1.14, + "learning_rate": 1.4263812404837803e-05, + "loss": 1.1576, + "step": 13504 + }, + { + "epoch": 1.14, + "learning_rate": 1.4262990181513448e-05, + "loss": 1.1257, + "step": 13505 + }, + { + "epoch": 1.14, + "learning_rate": 1.4262167922966907e-05, + "loss": 1.2091, + "step": 13506 + }, + { + "epoch": 1.14, + "learning_rate": 1.426134562920497e-05, + "loss": 1.1357, + "step": 13507 + }, + { + "epoch": 1.14, + "learning_rate": 1.4260523300234433e-05, + "loss": 1.1597, + "step": 13508 + }, + { + "epoch": 1.14, + "learning_rate": 1.4259700936062087e-05, + "loss": 1.2475, + "step": 13509 + }, + { + "epoch": 1.14, + "learning_rate": 1.4258878536694727e-05, + "loss": 1.2426, + "step": 13510 + }, + { + "epoch": 1.14, + "learning_rate": 1.4258056102139156e-05, + "loss": 1.224, + "step": 13511 + }, + { + "epoch": 1.14, + "learning_rate": 1.4257233632402157e-05, + "loss": 1.2147, + "step": 13512 + }, + { + "epoch": 1.14, + "learning_rate": 1.425641112749053e-05, + "loss": 1.2321, + "step": 13513 + }, + { + "epoch": 1.14, + "learning_rate": 1.4255588587411078e-05, + "loss": 1.2707, + "step": 13514 + }, + { + "epoch": 1.14, + "learning_rate": 1.4254766012170586e-05, + "loss": 1.2685, + "step": 13515 + }, + { + "epoch": 1.14, + "learning_rate": 1.4253943401775859e-05, + "loss": 1.1345, + "step": 13516 + }, + { + "epoch": 1.14, + "learning_rate": 1.4253120756233689e-05, + "loss": 1.2589, + "step": 13517 + }, + { + "epoch": 1.14, + "learning_rate": 1.4252298075550874e-05, + "loss": 1.1597, + "step": 13518 + }, + { + "epoch": 1.14, + "learning_rate": 1.4251475359734212e-05, + "loss": 1.2534, + "step": 13519 + }, + { + "epoch": 1.14, + "learning_rate": 1.4250652608790498e-05, + "loss": 1.1123, + "step": 13520 + }, + { + "epoch": 1.14, + "learning_rate": 1.4249829822726534e-05, + "loss": 1.2924, + "step": 13521 + }, + { + "epoch": 1.14, + "learning_rate": 1.4249007001549113e-05, + "loss": 1.2134, + "step": 13522 + }, + { + "epoch": 1.14, + "learning_rate": 1.4248184145265039e-05, + "loss": 1.1222, + "step": 13523 + }, + { + "epoch": 1.14, + "learning_rate": 1.4247361253881104e-05, + "loss": 1.1603, + "step": 13524 + }, + { + "epoch": 1.14, + "learning_rate": 1.4246538327404118e-05, + "loss": 1.1928, + "step": 13525 + }, + { + "epoch": 1.14, + "learning_rate": 1.424571536584087e-05, + "loss": 1.1449, + "step": 13526 + }, + { + "epoch": 1.14, + "learning_rate": 1.424489236919816e-05, + "loss": 1.1912, + "step": 13527 + }, + { + "epoch": 1.14, + "learning_rate": 1.4244069337482793e-05, + "loss": 1.228, + "step": 13528 + }, + { + "epoch": 1.14, + "learning_rate": 1.4243246270701568e-05, + "loss": 1.2177, + "step": 13529 + }, + { + "epoch": 1.14, + "learning_rate": 1.4242423168861282e-05, + "loss": 1.1736, + "step": 13530 + }, + { + "epoch": 1.14, + "learning_rate": 1.4241600031968739e-05, + "loss": 1.1829, + "step": 13531 + }, + { + "epoch": 1.14, + "learning_rate": 1.4240776860030738e-05, + "loss": 1.1994, + "step": 13532 + }, + { + "epoch": 1.14, + "learning_rate": 1.4239953653054087e-05, + "loss": 1.3066, + "step": 13533 + }, + { + "epoch": 1.14, + "learning_rate": 1.4239130411045575e-05, + "loss": 1.273, + "step": 13534 + }, + { + "epoch": 1.14, + "learning_rate": 1.4238307134012011e-05, + "loss": 1.2754, + "step": 13535 + }, + { + "epoch": 1.14, + "learning_rate": 1.42374838219602e-05, + "loss": 1.255, + "step": 13536 + }, + { + "epoch": 1.14, + "learning_rate": 1.4236660474896941e-05, + "loss": 1.2259, + "step": 13537 + }, + { + "epoch": 1.14, + "learning_rate": 1.4235837092829037e-05, + "loss": 1.168, + "step": 13538 + }, + { + "epoch": 1.14, + "learning_rate": 1.4235013675763294e-05, + "loss": 1.1791, + "step": 13539 + }, + { + "epoch": 1.14, + "learning_rate": 1.4234190223706508e-05, + "loss": 1.1889, + "step": 13540 + }, + { + "epoch": 1.14, + "learning_rate": 1.423336673666549e-05, + "loss": 1.1832, + "step": 13541 + }, + { + "epoch": 1.14, + "learning_rate": 1.4232543214647038e-05, + "loss": 1.2278, + "step": 13542 + }, + { + "epoch": 1.14, + "learning_rate": 1.4231719657657966e-05, + "loss": 1.1303, + "step": 13543 + }, + { + "epoch": 1.14, + "learning_rate": 1.4230896065705066e-05, + "loss": 1.2493, + "step": 13544 + }, + { + "epoch": 1.14, + "learning_rate": 1.423007243879515e-05, + "loss": 1.1888, + "step": 13545 + }, + { + "epoch": 1.14, + "learning_rate": 1.4229248776935024e-05, + "loss": 1.238, + "step": 13546 + }, + { + "epoch": 1.14, + "learning_rate": 1.4228425080131488e-05, + "loss": 1.1415, + "step": 13547 + }, + { + "epoch": 1.14, + "learning_rate": 1.4227601348391351e-05, + "loss": 1.2499, + "step": 13548 + }, + { + "epoch": 1.14, + "learning_rate": 1.4226777581721418e-05, + "loss": 1.2254, + "step": 13549 + }, + { + "epoch": 1.14, + "learning_rate": 1.4225953780128499e-05, + "loss": 1.1415, + "step": 13550 + }, + { + "epoch": 1.14, + "learning_rate": 1.4225129943619396e-05, + "loss": 1.1794, + "step": 13551 + }, + { + "epoch": 1.14, + "learning_rate": 1.4224306072200916e-05, + "loss": 1.2139, + "step": 13552 + }, + { + "epoch": 1.14, + "learning_rate": 1.4223482165879867e-05, + "loss": 1.1407, + "step": 13553 + }, + { + "epoch": 1.14, + "learning_rate": 1.4222658224663062e-05, + "loss": 1.2075, + "step": 13554 + }, + { + "epoch": 1.14, + "learning_rate": 1.4221834248557298e-05, + "loss": 1.2835, + "step": 13555 + }, + { + "epoch": 1.14, + "learning_rate": 1.4221010237569388e-05, + "loss": 1.2054, + "step": 13556 + }, + { + "epoch": 1.14, + "learning_rate": 1.4220186191706143e-05, + "loss": 1.1888, + "step": 13557 + }, + { + "epoch": 1.14, + "learning_rate": 1.4219362110974368e-05, + "loss": 1.1769, + "step": 13558 + }, + { + "epoch": 1.14, + "learning_rate": 1.4218537995380872e-05, + "loss": 1.2627, + "step": 13559 + }, + { + "epoch": 1.14, + "learning_rate": 1.4217713844932466e-05, + "loss": 1.1911, + "step": 13560 + }, + { + "epoch": 1.14, + "learning_rate": 1.421688965963596e-05, + "loss": 1.2226, + "step": 13561 + }, + { + "epoch": 1.14, + "learning_rate": 1.421606543949816e-05, + "loss": 1.2815, + "step": 13562 + }, + { + "epoch": 1.14, + "learning_rate": 1.4215241184525876e-05, + "loss": 1.1997, + "step": 13563 + }, + { + "epoch": 1.14, + "learning_rate": 1.4214416894725923e-05, + "loss": 1.2116, + "step": 13564 + }, + { + "epoch": 1.14, + "learning_rate": 1.421359257010511e-05, + "loss": 1.2584, + "step": 13565 + }, + { + "epoch": 1.14, + "learning_rate": 1.4212768210670245e-05, + "loss": 1.1844, + "step": 13566 + }, + { + "epoch": 1.14, + "learning_rate": 1.421194381642814e-05, + "loss": 1.1982, + "step": 13567 + }, + { + "epoch": 1.14, + "learning_rate": 1.421111938738561e-05, + "loss": 1.2075, + "step": 13568 + }, + { + "epoch": 1.14, + "learning_rate": 1.421029492354946e-05, + "loss": 1.2934, + "step": 13569 + }, + { + "epoch": 1.14, + "learning_rate": 1.4209470424926509e-05, + "loss": 1.2356, + "step": 13570 + }, + { + "epoch": 1.14, + "learning_rate": 1.4208645891523564e-05, + "loss": 1.1751, + "step": 13571 + }, + { + "epoch": 1.14, + "learning_rate": 1.4207821323347442e-05, + "loss": 1.2203, + "step": 13572 + }, + { + "epoch": 1.14, + "learning_rate": 1.420699672040495e-05, + "loss": 1.2796, + "step": 13573 + }, + { + "epoch": 1.14, + "learning_rate": 1.4206172082702908e-05, + "loss": 1.2435, + "step": 13574 + }, + { + "epoch": 1.14, + "learning_rate": 1.4205347410248127e-05, + "loss": 1.1946, + "step": 13575 + }, + { + "epoch": 1.14, + "learning_rate": 1.4204522703047417e-05, + "loss": 1.2281, + "step": 13576 + }, + { + "epoch": 1.14, + "learning_rate": 1.4203697961107597e-05, + "loss": 1.1994, + "step": 13577 + }, + { + "epoch": 1.14, + "learning_rate": 1.4202873184435475e-05, + "loss": 1.1909, + "step": 13578 + }, + { + "epoch": 1.14, + "learning_rate": 1.4202048373037876e-05, + "loss": 1.2216, + "step": 13579 + }, + { + "epoch": 1.14, + "learning_rate": 1.4201223526921604e-05, + "loss": 1.2503, + "step": 13580 + }, + { + "epoch": 1.14, + "learning_rate": 1.4200398646093478e-05, + "loss": 1.3398, + "step": 13581 + }, + { + "epoch": 1.14, + "learning_rate": 1.4199573730560316e-05, + "loss": 1.1198, + "step": 13582 + }, + { + "epoch": 1.14, + "learning_rate": 1.4198748780328933e-05, + "loss": 1.3054, + "step": 13583 + }, + { + "epoch": 1.14, + "learning_rate": 1.419792379540614e-05, + "loss": 1.197, + "step": 13584 + }, + { + "epoch": 1.14, + "learning_rate": 1.4197098775798758e-05, + "loss": 1.2076, + "step": 13585 + }, + { + "epoch": 1.14, + "learning_rate": 1.4196273721513608e-05, + "loss": 1.1855, + "step": 13586 + }, + { + "epoch": 1.14, + "learning_rate": 1.4195448632557496e-05, + "loss": 1.1984, + "step": 13587 + }, + { + "epoch": 1.14, + "learning_rate": 1.4194623508937247e-05, + "loss": 1.254, + "step": 13588 + }, + { + "epoch": 1.14, + "learning_rate": 1.4193798350659674e-05, + "loss": 1.24, + "step": 13589 + }, + { + "epoch": 1.14, + "learning_rate": 1.41929731577316e-05, + "loss": 1.1754, + "step": 13590 + }, + { + "epoch": 1.14, + "learning_rate": 1.4192147930159839e-05, + "loss": 1.1801, + "step": 13591 + }, + { + "epoch": 1.14, + "learning_rate": 1.4191322667951206e-05, + "loss": 1.1569, + "step": 13592 + }, + { + "epoch": 1.14, + "learning_rate": 1.419049737111253e-05, + "loss": 1.1817, + "step": 13593 + }, + { + "epoch": 1.14, + "learning_rate": 1.418967203965062e-05, + "loss": 1.1913, + "step": 13594 + }, + { + "epoch": 1.14, + "learning_rate": 1.4188846673572298e-05, + "loss": 1.1453, + "step": 13595 + }, + { + "epoch": 1.14, + "learning_rate": 1.4188021272884384e-05, + "loss": 1.2128, + "step": 13596 + }, + { + "epoch": 1.14, + "learning_rate": 1.4187195837593702e-05, + "loss": 1.141, + "step": 13597 + }, + { + "epoch": 1.14, + "learning_rate": 1.4186370367707063e-05, + "loss": 1.2421, + "step": 13598 + }, + { + "epoch": 1.14, + "learning_rate": 1.4185544863231296e-05, + "loss": 1.199, + "step": 13599 + }, + { + "epoch": 1.15, + "learning_rate": 1.4184719324173218e-05, + "loss": 1.2364, + "step": 13600 + }, + { + "epoch": 1.15, + "learning_rate": 1.4183893750539646e-05, + "loss": 1.2128, + "step": 13601 + }, + { + "epoch": 1.15, + "learning_rate": 1.4183068142337406e-05, + "loss": 1.2397, + "step": 13602 + }, + { + "epoch": 1.15, + "learning_rate": 1.418224249957332e-05, + "loss": 1.1186, + "step": 13603 + }, + { + "epoch": 1.15, + "learning_rate": 1.4181416822254208e-05, + "loss": 1.1535, + "step": 13604 + }, + { + "epoch": 1.15, + "learning_rate": 1.418059111038689e-05, + "loss": 1.2597, + "step": 13605 + }, + { + "epoch": 1.15, + "learning_rate": 1.4179765363978191e-05, + "loss": 1.1844, + "step": 13606 + }, + { + "epoch": 1.15, + "learning_rate": 1.4178939583034933e-05, + "loss": 1.1206, + "step": 13607 + }, + { + "epoch": 1.15, + "learning_rate": 1.4178113767563941e-05, + "loss": 1.2683, + "step": 13608 + }, + { + "epoch": 1.15, + "learning_rate": 1.4177287917572032e-05, + "loss": 1.2531, + "step": 13609 + }, + { + "epoch": 1.15, + "learning_rate": 1.4176462033066033e-05, + "loss": 1.2039, + "step": 13610 + }, + { + "epoch": 1.15, + "learning_rate": 1.4175636114052773e-05, + "loss": 1.1533, + "step": 13611 + }, + { + "epoch": 1.15, + "learning_rate": 1.4174810160539067e-05, + "loss": 1.2182, + "step": 13612 + }, + { + "epoch": 1.15, + "learning_rate": 1.4173984172531744e-05, + "loss": 1.2989, + "step": 13613 + }, + { + "epoch": 1.15, + "learning_rate": 1.417315815003763e-05, + "loss": 1.1537, + "step": 13614 + }, + { + "epoch": 1.15, + "learning_rate": 1.4172332093063546e-05, + "loss": 1.1399, + "step": 13615 + }, + { + "epoch": 1.15, + "learning_rate": 1.417150600161632e-05, + "loss": 1.2129, + "step": 13616 + }, + { + "epoch": 1.15, + "learning_rate": 1.4170679875702775e-05, + "loss": 1.2792, + "step": 13617 + }, + { + "epoch": 1.15, + "learning_rate": 1.416985371532974e-05, + "loss": 1.1488, + "step": 13618 + }, + { + "epoch": 1.15, + "learning_rate": 1.4169027520504038e-05, + "loss": 1.1843, + "step": 13619 + }, + { + "epoch": 1.15, + "learning_rate": 1.4168201291232494e-05, + "loss": 1.1676, + "step": 13620 + }, + { + "epoch": 1.15, + "learning_rate": 1.416737502752194e-05, + "loss": 1.1255, + "step": 13621 + }, + { + "epoch": 1.15, + "learning_rate": 1.4166548729379197e-05, + "loss": 1.3325, + "step": 13622 + }, + { + "epoch": 1.15, + "learning_rate": 1.4165722396811098e-05, + "loss": 1.2648, + "step": 13623 + }, + { + "epoch": 1.15, + "learning_rate": 1.4164896029824465e-05, + "loss": 1.2312, + "step": 13624 + }, + { + "epoch": 1.15, + "learning_rate": 1.416406962842613e-05, + "loss": 1.2541, + "step": 13625 + }, + { + "epoch": 1.15, + "learning_rate": 1.4163243192622917e-05, + "loss": 1.1618, + "step": 13626 + }, + { + "epoch": 1.15, + "learning_rate": 1.4162416722421659e-05, + "loss": 1.1762, + "step": 13627 + }, + { + "epoch": 1.15, + "learning_rate": 1.416159021782918e-05, + "loss": 1.1149, + "step": 13628 + }, + { + "epoch": 1.15, + "learning_rate": 1.4160763678852311e-05, + "loss": 1.1447, + "step": 13629 + }, + { + "epoch": 1.15, + "learning_rate": 1.415993710549788e-05, + "loss": 1.2539, + "step": 13630 + }, + { + "epoch": 1.15, + "learning_rate": 1.4159110497772718e-05, + "loss": 1.1706, + "step": 13631 + }, + { + "epoch": 1.15, + "learning_rate": 1.4158283855683654e-05, + "loss": 1.2706, + "step": 13632 + }, + { + "epoch": 1.15, + "learning_rate": 1.4157457179237522e-05, + "loss": 1.1713, + "step": 13633 + }, + { + "epoch": 1.15, + "learning_rate": 1.415663046844114e-05, + "loss": 1.2428, + "step": 13634 + }, + { + "epoch": 1.15, + "learning_rate": 1.4155803723301352e-05, + "loss": 1.1995, + "step": 13635 + }, + { + "epoch": 1.15, + "learning_rate": 1.4154976943824985e-05, + "loss": 1.2376, + "step": 13636 + }, + { + "epoch": 1.15, + "learning_rate": 1.4154150130018867e-05, + "loss": 1.1724, + "step": 13637 + }, + { + "epoch": 1.15, + "learning_rate": 1.415332328188983e-05, + "loss": 1.2476, + "step": 13638 + }, + { + "epoch": 1.15, + "learning_rate": 1.4152496399444706e-05, + "loss": 1.0637, + "step": 13639 + }, + { + "epoch": 1.15, + "learning_rate": 1.4151669482690331e-05, + "loss": 1.2233, + "step": 13640 + }, + { + "epoch": 1.15, + "learning_rate": 1.4150842531633533e-05, + "loss": 1.1639, + "step": 13641 + }, + { + "epoch": 1.15, + "learning_rate": 1.4150015546281144e-05, + "loss": 1.2829, + "step": 13642 + }, + { + "epoch": 1.15, + "learning_rate": 1.4149188526640001e-05, + "loss": 1.1623, + "step": 13643 + }, + { + "epoch": 1.15, + "learning_rate": 1.4148361472716931e-05, + "loss": 1.1561, + "step": 13644 + }, + { + "epoch": 1.15, + "learning_rate": 1.4147534384518773e-05, + "loss": 1.3105, + "step": 13645 + }, + { + "epoch": 1.15, + "learning_rate": 1.4146707262052357e-05, + "loss": 1.1452, + "step": 13646 + }, + { + "epoch": 1.15, + "learning_rate": 1.4145880105324522e-05, + "loss": 1.146, + "step": 13647 + }, + { + "epoch": 1.15, + "learning_rate": 1.4145052914342095e-05, + "loss": 1.216, + "step": 13648 + }, + { + "epoch": 1.15, + "learning_rate": 1.4144225689111914e-05, + "loss": 1.235, + "step": 13649 + }, + { + "epoch": 1.15, + "learning_rate": 1.4143398429640817e-05, + "loss": 1.2621, + "step": 13650 + }, + { + "epoch": 1.15, + "learning_rate": 1.4142571135935635e-05, + "loss": 1.2387, + "step": 13651 + }, + { + "epoch": 1.15, + "learning_rate": 1.4141743808003204e-05, + "loss": 1.1853, + "step": 13652 + }, + { + "epoch": 1.15, + "learning_rate": 1.4140916445850358e-05, + "loss": 1.1797, + "step": 13653 + }, + { + "epoch": 1.15, + "learning_rate": 1.4140089049483941e-05, + "loss": 1.2024, + "step": 13654 + }, + { + "epoch": 1.15, + "learning_rate": 1.413926161891078e-05, + "loss": 1.1745, + "step": 13655 + }, + { + "epoch": 1.15, + "learning_rate": 1.4138434154137712e-05, + "loss": 1.1916, + "step": 13656 + }, + { + "epoch": 1.15, + "learning_rate": 1.4137606655171578e-05, + "loss": 1.2429, + "step": 13657 + }, + { + "epoch": 1.15, + "learning_rate": 1.4136779122019217e-05, + "loss": 1.1607, + "step": 13658 + }, + { + "epoch": 1.15, + "learning_rate": 1.413595155468746e-05, + "loss": 1.286, + "step": 13659 + }, + { + "epoch": 1.15, + "learning_rate": 1.4135123953183148e-05, + "loss": 1.1612, + "step": 13660 + }, + { + "epoch": 1.15, + "learning_rate": 1.413429631751312e-05, + "loss": 1.1986, + "step": 13661 + }, + { + "epoch": 1.15, + "learning_rate": 1.413346864768421e-05, + "loss": 1.1742, + "step": 13662 + }, + { + "epoch": 1.15, + "learning_rate": 1.413264094370326e-05, + "loss": 1.1745, + "step": 13663 + }, + { + "epoch": 1.15, + "learning_rate": 1.413181320557711e-05, + "loss": 1.1897, + "step": 13664 + }, + { + "epoch": 1.15, + "learning_rate": 1.4130985433312594e-05, + "loss": 1.1805, + "step": 13665 + }, + { + "epoch": 1.15, + "learning_rate": 1.4130157626916557e-05, + "loss": 1.1124, + "step": 13666 + }, + { + "epoch": 1.15, + "learning_rate": 1.4129329786395834e-05, + "loss": 1.1378, + "step": 13667 + }, + { + "epoch": 1.15, + "learning_rate": 1.4128501911757269e-05, + "loss": 1.2298, + "step": 13668 + }, + { + "epoch": 1.15, + "learning_rate": 1.4127674003007696e-05, + "loss": 1.1468, + "step": 13669 + }, + { + "epoch": 1.15, + "learning_rate": 1.4126846060153963e-05, + "loss": 1.0947, + "step": 13670 + }, + { + "epoch": 1.15, + "learning_rate": 1.4126018083202904e-05, + "loss": 1.2444, + "step": 13671 + }, + { + "epoch": 1.15, + "learning_rate": 1.4125190072161365e-05, + "loss": 1.198, + "step": 13672 + }, + { + "epoch": 1.15, + "learning_rate": 1.4124362027036185e-05, + "loss": 1.3018, + "step": 13673 + }, + { + "epoch": 1.15, + "learning_rate": 1.4123533947834205e-05, + "loss": 1.2544, + "step": 13674 + }, + { + "epoch": 1.15, + "learning_rate": 1.412270583456227e-05, + "loss": 1.1112, + "step": 13675 + }, + { + "epoch": 1.15, + "learning_rate": 1.4121877687227219e-05, + "loss": 1.1899, + "step": 13676 + }, + { + "epoch": 1.15, + "learning_rate": 1.4121049505835896e-05, + "loss": 1.1994, + "step": 13677 + }, + { + "epoch": 1.15, + "learning_rate": 1.4120221290395142e-05, + "loss": 1.2332, + "step": 13678 + }, + { + "epoch": 1.15, + "learning_rate": 1.4119393040911801e-05, + "loss": 1.1453, + "step": 13679 + }, + { + "epoch": 1.15, + "learning_rate": 1.4118564757392716e-05, + "loss": 1.2692, + "step": 13680 + }, + { + "epoch": 1.15, + "learning_rate": 1.4117736439844732e-05, + "loss": 1.148, + "step": 13681 + }, + { + "epoch": 1.15, + "learning_rate": 1.4116908088274693e-05, + "loss": 1.199, + "step": 13682 + }, + { + "epoch": 1.15, + "learning_rate": 1.411607970268944e-05, + "loss": 1.1454, + "step": 13683 + }, + { + "epoch": 1.15, + "learning_rate": 1.4115251283095817e-05, + "loss": 1.2334, + "step": 13684 + }, + { + "epoch": 1.15, + "learning_rate": 1.4114422829500671e-05, + "loss": 1.2406, + "step": 13685 + }, + { + "epoch": 1.15, + "learning_rate": 1.4113594341910853e-05, + "loss": 1.2109, + "step": 13686 + }, + { + "epoch": 1.15, + "learning_rate": 1.4112765820333196e-05, + "loss": 1.2755, + "step": 13687 + }, + { + "epoch": 1.15, + "learning_rate": 1.4111937264774553e-05, + "loss": 1.2227, + "step": 13688 + }, + { + "epoch": 1.15, + "learning_rate": 1.411110867524177e-05, + "loss": 1.1419, + "step": 13689 + }, + { + "epoch": 1.15, + "learning_rate": 1.4110280051741692e-05, + "loss": 1.1028, + "step": 13690 + }, + { + "epoch": 1.15, + "learning_rate": 1.4109451394281162e-05, + "loss": 1.2008, + "step": 13691 + }, + { + "epoch": 1.15, + "learning_rate": 1.410862270286703e-05, + "loss": 1.2739, + "step": 13692 + }, + { + "epoch": 1.15, + "learning_rate": 1.4107793977506143e-05, + "loss": 1.1207, + "step": 13693 + }, + { + "epoch": 1.15, + "learning_rate": 1.4106965218205347e-05, + "loss": 1.2045, + "step": 13694 + }, + { + "epoch": 1.15, + "learning_rate": 1.410613642497149e-05, + "loss": 1.1663, + "step": 13695 + }, + { + "epoch": 1.15, + "learning_rate": 1.410530759781142e-05, + "loss": 1.1123, + "step": 13696 + }, + { + "epoch": 1.15, + "learning_rate": 1.4104478736731983e-05, + "loss": 1.1272, + "step": 13697 + }, + { + "epoch": 1.15, + "learning_rate": 1.410364984174003e-05, + "loss": 1.2449, + "step": 13698 + }, + { + "epoch": 1.15, + "learning_rate": 1.410282091284241e-05, + "loss": 1.2136, + "step": 13699 + }, + { + "epoch": 1.15, + "learning_rate": 1.410199195004597e-05, + "loss": 1.2184, + "step": 13700 + }, + { + "epoch": 1.15, + "learning_rate": 1.4101162953357558e-05, + "loss": 1.1563, + "step": 13701 + }, + { + "epoch": 1.15, + "learning_rate": 1.4100333922784028e-05, + "loss": 1.1108, + "step": 13702 + }, + { + "epoch": 1.15, + "learning_rate": 1.4099504858332225e-05, + "loss": 1.1411, + "step": 13703 + }, + { + "epoch": 1.15, + "learning_rate": 1.4098675760009003e-05, + "loss": 1.1749, + "step": 13704 + }, + { + "epoch": 1.15, + "learning_rate": 1.4097846627821208e-05, + "loss": 1.2263, + "step": 13705 + }, + { + "epoch": 1.15, + "learning_rate": 1.4097017461775692e-05, + "loss": 1.1855, + "step": 13706 + }, + { + "epoch": 1.15, + "learning_rate": 1.4096188261879309e-05, + "loss": 1.2284, + "step": 13707 + }, + { + "epoch": 1.15, + "learning_rate": 1.4095359028138906e-05, + "loss": 1.1973, + "step": 13708 + }, + { + "epoch": 1.15, + "learning_rate": 1.4094529760561335e-05, + "loss": 1.1425, + "step": 13709 + }, + { + "epoch": 1.15, + "learning_rate": 1.409370045915345e-05, + "loss": 1.1817, + "step": 13710 + }, + { + "epoch": 1.15, + "learning_rate": 1.4092871123922104e-05, + "loss": 1.1914, + "step": 13711 + }, + { + "epoch": 1.15, + "learning_rate": 1.4092041754874143e-05, + "loss": 1.1788, + "step": 13712 + }, + { + "epoch": 1.15, + "learning_rate": 1.4091212352016426e-05, + "loss": 1.2568, + "step": 13713 + }, + { + "epoch": 1.15, + "learning_rate": 1.4090382915355801e-05, + "loss": 1.2124, + "step": 13714 + }, + { + "epoch": 1.15, + "learning_rate": 1.4089553444899127e-05, + "loss": 1.2009, + "step": 13715 + }, + { + "epoch": 1.15, + "learning_rate": 1.408872394065325e-05, + "loss": 1.2025, + "step": 13716 + }, + { + "epoch": 1.15, + "learning_rate": 1.4087894402625029e-05, + "loss": 1.2307, + "step": 13717 + }, + { + "epoch": 1.15, + "learning_rate": 1.4087064830821314e-05, + "loss": 1.1639, + "step": 13718 + }, + { + "epoch": 1.16, + "learning_rate": 1.4086235225248966e-05, + "loss": 1.1513, + "step": 13719 + }, + { + "epoch": 1.16, + "learning_rate": 1.4085405585914829e-05, + "loss": 1.254, + "step": 13720 + }, + { + "epoch": 1.16, + "learning_rate": 1.4084575912825766e-05, + "loss": 1.1666, + "step": 13721 + }, + { + "epoch": 1.16, + "learning_rate": 1.408374620598863e-05, + "loss": 1.2467, + "step": 13722 + }, + { + "epoch": 1.16, + "learning_rate": 1.4082916465410277e-05, + "loss": 1.2223, + "step": 13723 + }, + { + "epoch": 1.16, + "learning_rate": 1.4082086691097558e-05, + "loss": 1.1479, + "step": 13724 + }, + { + "epoch": 1.16, + "learning_rate": 1.4081256883057335e-05, + "loss": 1.2155, + "step": 13725 + }, + { + "epoch": 1.16, + "learning_rate": 1.4080427041296463e-05, + "loss": 1.2014, + "step": 13726 + }, + { + "epoch": 1.16, + "learning_rate": 1.4079597165821792e-05, + "loss": 1.1656, + "step": 13727 + }, + { + "epoch": 1.16, + "learning_rate": 1.4078767256640186e-05, + "loss": 1.0637, + "step": 13728 + }, + { + "epoch": 1.16, + "learning_rate": 1.4077937313758502e-05, + "loss": 1.2491, + "step": 13729 + }, + { + "epoch": 1.16, + "learning_rate": 1.4077107337183593e-05, + "loss": 1.2428, + "step": 13730 + }, + { + "epoch": 1.16, + "learning_rate": 1.4076277326922317e-05, + "loss": 1.1089, + "step": 13731 + }, + { + "epoch": 1.16, + "learning_rate": 1.4075447282981533e-05, + "loss": 1.3005, + "step": 13732 + }, + { + "epoch": 1.16, + "learning_rate": 1.4074617205368102e-05, + "loss": 1.1939, + "step": 13733 + }, + { + "epoch": 1.16, + "learning_rate": 1.4073787094088877e-05, + "loss": 1.1821, + "step": 13734 + }, + { + "epoch": 1.16, + "learning_rate": 1.407295694915072e-05, + "loss": 1.2191, + "step": 13735 + }, + { + "epoch": 1.16, + "learning_rate": 1.407212677056049e-05, + "loss": 1.2233, + "step": 13736 + }, + { + "epoch": 1.16, + "learning_rate": 1.4071296558325045e-05, + "loss": 1.2349, + "step": 13737 + }, + { + "epoch": 1.16, + "learning_rate": 1.4070466312451246e-05, + "loss": 1.1909, + "step": 13738 + }, + { + "epoch": 1.16, + "learning_rate": 1.4069636032945948e-05, + "loss": 1.1764, + "step": 13739 + }, + { + "epoch": 1.16, + "learning_rate": 1.4068805719816019e-05, + "loss": 1.1856, + "step": 13740 + }, + { + "epoch": 1.16, + "learning_rate": 1.4067975373068313e-05, + "loss": 1.1814, + "step": 13741 + }, + { + "epoch": 1.16, + "learning_rate": 1.4067144992709692e-05, + "loss": 1.2471, + "step": 13742 + }, + { + "epoch": 1.16, + "learning_rate": 1.4066314578747017e-05, + "loss": 1.2225, + "step": 13743 + }, + { + "epoch": 1.16, + "learning_rate": 1.4065484131187152e-05, + "loss": 1.1409, + "step": 13744 + }, + { + "epoch": 1.16, + "learning_rate": 1.4064653650036954e-05, + "loss": 1.2447, + "step": 13745 + }, + { + "epoch": 1.16, + "learning_rate": 1.4063823135303285e-05, + "loss": 1.189, + "step": 13746 + }, + { + "epoch": 1.16, + "learning_rate": 1.4062992586993011e-05, + "loss": 1.1791, + "step": 13747 + }, + { + "epoch": 1.16, + "learning_rate": 1.4062162005112994e-05, + "loss": 1.1518, + "step": 13748 + }, + { + "epoch": 1.16, + "learning_rate": 1.406133138967009e-05, + "loss": 1.2221, + "step": 13749 + }, + { + "epoch": 1.16, + "learning_rate": 1.406050074067117e-05, + "loss": 1.2103, + "step": 13750 + }, + { + "epoch": 1.16, + "learning_rate": 1.4059670058123092e-05, + "loss": 1.1648, + "step": 13751 + }, + { + "epoch": 1.16, + "learning_rate": 1.4058839342032719e-05, + "loss": 1.2001, + "step": 13752 + }, + { + "epoch": 1.16, + "learning_rate": 1.4058008592406917e-05, + "loss": 1.2003, + "step": 13753 + }, + { + "epoch": 1.16, + "learning_rate": 1.4057177809252554e-05, + "loss": 1.1568, + "step": 13754 + }, + { + "epoch": 1.16, + "learning_rate": 1.4056346992576484e-05, + "loss": 1.2163, + "step": 13755 + }, + { + "epoch": 1.16, + "learning_rate": 1.405551614238558e-05, + "loss": 1.2047, + "step": 13756 + }, + { + "epoch": 1.16, + "learning_rate": 1.40546852586867e-05, + "loss": 1.2141, + "step": 13757 + }, + { + "epoch": 1.16, + "learning_rate": 1.4053854341486715e-05, + "loss": 1.0846, + "step": 13758 + }, + { + "epoch": 1.16, + "learning_rate": 1.4053023390792488e-05, + "loss": 1.1804, + "step": 13759 + }, + { + "epoch": 1.16, + "learning_rate": 1.4052192406610882e-05, + "loss": 1.1201, + "step": 13760 + }, + { + "epoch": 1.16, + "learning_rate": 1.4051361388948772e-05, + "loss": 1.282, + "step": 13761 + }, + { + "epoch": 1.16, + "learning_rate": 1.4050530337813013e-05, + "loss": 1.2025, + "step": 13762 + }, + { + "epoch": 1.16, + "learning_rate": 1.4049699253210475e-05, + "loss": 1.1526, + "step": 13763 + }, + { + "epoch": 1.16, + "learning_rate": 1.4048868135148028e-05, + "loss": 1.212, + "step": 13764 + }, + { + "epoch": 1.16, + "learning_rate": 1.4048036983632538e-05, + "loss": 1.2286, + "step": 13765 + }, + { + "epoch": 1.16, + "learning_rate": 1.404720579867087e-05, + "loss": 1.2017, + "step": 13766 + }, + { + "epoch": 1.16, + "learning_rate": 1.404637458026989e-05, + "loss": 1.1012, + "step": 13767 + }, + { + "epoch": 1.16, + "learning_rate": 1.4045543328436469e-05, + "loss": 1.2803, + "step": 13768 + }, + { + "epoch": 1.16, + "learning_rate": 1.4044712043177478e-05, + "loss": 1.1411, + "step": 13769 + }, + { + "epoch": 1.16, + "learning_rate": 1.4043880724499777e-05, + "loss": 1.0996, + "step": 13770 + }, + { + "epoch": 1.16, + "learning_rate": 1.4043049372410238e-05, + "loss": 1.2505, + "step": 13771 + }, + { + "epoch": 1.16, + "learning_rate": 1.4042217986915736e-05, + "loss": 1.2515, + "step": 13772 + }, + { + "epoch": 1.16, + "learning_rate": 1.4041386568023133e-05, + "loss": 1.1912, + "step": 13773 + }, + { + "epoch": 1.16, + "learning_rate": 1.4040555115739301e-05, + "loss": 1.1581, + "step": 13774 + }, + { + "epoch": 1.16, + "learning_rate": 1.403972363007111e-05, + "loss": 1.2301, + "step": 13775 + }, + { + "epoch": 1.16, + "learning_rate": 1.4038892111025428e-05, + "loss": 1.1917, + "step": 13776 + }, + { + "epoch": 1.16, + "learning_rate": 1.4038060558609127e-05, + "loss": 1.1098, + "step": 13777 + }, + { + "epoch": 1.16, + "learning_rate": 1.4037228972829078e-05, + "loss": 1.2705, + "step": 13778 + }, + { + "epoch": 1.16, + "learning_rate": 1.4036397353692153e-05, + "loss": 1.1701, + "step": 13779 + }, + { + "epoch": 1.16, + "learning_rate": 1.403556570120522e-05, + "loss": 1.2636, + "step": 13780 + }, + { + "epoch": 1.16, + "learning_rate": 1.4034734015375148e-05, + "loss": 1.1178, + "step": 13781 + }, + { + "epoch": 1.16, + "learning_rate": 1.4033902296208816e-05, + "loss": 1.1487, + "step": 13782 + }, + { + "epoch": 1.16, + "learning_rate": 1.4033070543713092e-05, + "loss": 1.2435, + "step": 13783 + }, + { + "epoch": 1.16, + "learning_rate": 1.4032238757894847e-05, + "loss": 1.182, + "step": 13784 + }, + { + "epoch": 1.16, + "learning_rate": 1.4031406938760956e-05, + "loss": 1.1287, + "step": 13785 + }, + { + "epoch": 1.16, + "learning_rate": 1.4030575086318293e-05, + "loss": 1.1771, + "step": 13786 + }, + { + "epoch": 1.16, + "learning_rate": 1.4029743200573727e-05, + "loss": 1.1962, + "step": 13787 + }, + { + "epoch": 1.16, + "learning_rate": 1.402891128153413e-05, + "loss": 1.2501, + "step": 13788 + }, + { + "epoch": 1.16, + "learning_rate": 1.4028079329206382e-05, + "loss": 1.2572, + "step": 13789 + }, + { + "epoch": 1.16, + "learning_rate": 1.4027247343597352e-05, + "loss": 1.1765, + "step": 13790 + }, + { + "epoch": 1.16, + "learning_rate": 1.4026415324713915e-05, + "loss": 1.176, + "step": 13791 + }, + { + "epoch": 1.16, + "learning_rate": 1.4025583272562948e-05, + "loss": 1.1671, + "step": 13792 + }, + { + "epoch": 1.16, + "learning_rate": 1.4024751187151322e-05, + "loss": 1.2785, + "step": 13793 + }, + { + "epoch": 1.16, + "learning_rate": 1.4023919068485916e-05, + "loss": 1.1918, + "step": 13794 + }, + { + "epoch": 1.16, + "learning_rate": 1.40230869165736e-05, + "loss": 1.2157, + "step": 13795 + }, + { + "epoch": 1.16, + "learning_rate": 1.402225473142125e-05, + "loss": 1.2027, + "step": 13796 + }, + { + "epoch": 1.16, + "learning_rate": 1.402142251303575e-05, + "loss": 1.1601, + "step": 13797 + }, + { + "epoch": 1.16, + "learning_rate": 1.4020590261423968e-05, + "loss": 1.1658, + "step": 13798 + }, + { + "epoch": 1.16, + "learning_rate": 1.401975797659278e-05, + "loss": 1.1845, + "step": 13799 + }, + { + "epoch": 1.16, + "learning_rate": 1.4018925658549069e-05, + "loss": 1.1656, + "step": 13800 + }, + { + "epoch": 1.16, + "learning_rate": 1.4018093307299706e-05, + "loss": 1.2017, + "step": 13801 + }, + { + "epoch": 1.16, + "learning_rate": 1.401726092285157e-05, + "loss": 1.1646, + "step": 13802 + }, + { + "epoch": 1.16, + "learning_rate": 1.401642850521154e-05, + "loss": 1.2667, + "step": 13803 + }, + { + "epoch": 1.16, + "learning_rate": 1.401559605438649e-05, + "loss": 1.239, + "step": 13804 + }, + { + "epoch": 1.16, + "learning_rate": 1.4014763570383303e-05, + "loss": 1.2633, + "step": 13805 + }, + { + "epoch": 1.16, + "learning_rate": 1.4013931053208851e-05, + "loss": 1.2279, + "step": 13806 + }, + { + "epoch": 1.16, + "learning_rate": 1.4013098502870018e-05, + "loss": 1.1866, + "step": 13807 + }, + { + "epoch": 1.16, + "learning_rate": 1.401226591937368e-05, + "loss": 1.1483, + "step": 13808 + }, + { + "epoch": 1.16, + "learning_rate": 1.4011433302726719e-05, + "loss": 1.2659, + "step": 13809 + }, + { + "epoch": 1.16, + "learning_rate": 1.4010600652936009e-05, + "loss": 1.2268, + "step": 13810 + }, + { + "epoch": 1.16, + "learning_rate": 1.4009767970008437e-05, + "loss": 1.19, + "step": 13811 + }, + { + "epoch": 1.16, + "learning_rate": 1.4008935253950875e-05, + "loss": 1.1942, + "step": 13812 + }, + { + "epoch": 1.16, + "learning_rate": 1.4008102504770207e-05, + "loss": 1.1771, + "step": 13813 + }, + { + "epoch": 1.16, + "learning_rate": 1.4007269722473314e-05, + "loss": 1.1694, + "step": 13814 + }, + { + "epoch": 1.16, + "learning_rate": 1.400643690706708e-05, + "loss": 1.1408, + "step": 13815 + }, + { + "epoch": 1.16, + "learning_rate": 1.4005604058558377e-05, + "loss": 1.1921, + "step": 13816 + }, + { + "epoch": 1.16, + "learning_rate": 1.4004771176954093e-05, + "loss": 1.117, + "step": 13817 + }, + { + "epoch": 1.16, + "learning_rate": 1.4003938262261108e-05, + "loss": 1.1399, + "step": 13818 + }, + { + "epoch": 1.16, + "learning_rate": 1.4003105314486305e-05, + "loss": 1.2767, + "step": 13819 + }, + { + "epoch": 1.16, + "learning_rate": 1.4002272333636562e-05, + "loss": 1.2399, + "step": 13820 + }, + { + "epoch": 1.16, + "learning_rate": 1.4001439319718764e-05, + "loss": 1.2115, + "step": 13821 + }, + { + "epoch": 1.16, + "learning_rate": 1.4000606272739797e-05, + "loss": 1.1368, + "step": 13822 + }, + { + "epoch": 1.16, + "learning_rate": 1.3999773192706538e-05, + "loss": 1.1622, + "step": 13823 + }, + { + "epoch": 1.16, + "learning_rate": 1.3998940079625873e-05, + "loss": 1.1494, + "step": 13824 + }, + { + "epoch": 1.16, + "learning_rate": 1.3998106933504686e-05, + "loss": 1.2407, + "step": 13825 + }, + { + "epoch": 1.16, + "learning_rate": 1.399727375434986e-05, + "loss": 1.1774, + "step": 13826 + }, + { + "epoch": 1.16, + "learning_rate": 1.399644054216828e-05, + "loss": 1.3062, + "step": 13827 + }, + { + "epoch": 1.16, + "learning_rate": 1.3995607296966827e-05, + "loss": 1.1628, + "step": 13828 + }, + { + "epoch": 1.16, + "learning_rate": 1.3994774018752389e-05, + "loss": 1.2857, + "step": 13829 + }, + { + "epoch": 1.16, + "learning_rate": 1.3993940707531848e-05, + "loss": 1.2509, + "step": 13830 + }, + { + "epoch": 1.16, + "learning_rate": 1.3993107363312093e-05, + "loss": 1.2995, + "step": 13831 + }, + { + "epoch": 1.16, + "learning_rate": 1.3992273986100005e-05, + "loss": 1.2489, + "step": 13832 + }, + { + "epoch": 1.16, + "learning_rate": 1.399144057590247e-05, + "loss": 1.1433, + "step": 13833 + }, + { + "epoch": 1.16, + "learning_rate": 1.3990607132726379e-05, + "loss": 1.1885, + "step": 13834 + }, + { + "epoch": 1.16, + "learning_rate": 1.3989773656578613e-05, + "loss": 1.2408, + "step": 13835 + }, + { + "epoch": 1.16, + "learning_rate": 1.3988940147466063e-05, + "loss": 1.2031, + "step": 13836 + }, + { + "epoch": 1.16, + "learning_rate": 1.398810660539561e-05, + "loss": 1.2614, + "step": 13837 + }, + { + "epoch": 1.17, + "learning_rate": 1.3987273030374144e-05, + "loss": 1.2362, + "step": 13838 + }, + { + "epoch": 1.17, + "learning_rate": 1.3986439422408553e-05, + "loss": 1.2912, + "step": 13839 + }, + { + "epoch": 1.17, + "learning_rate": 1.3985605781505727e-05, + "loss": 1.2222, + "step": 13840 + }, + { + "epoch": 1.17, + "learning_rate": 1.3984772107672543e-05, + "loss": 1.2164, + "step": 13841 + }, + { + "epoch": 1.17, + "learning_rate": 1.3983938400915902e-05, + "loss": 1.2908, + "step": 13842 + }, + { + "epoch": 1.17, + "learning_rate": 1.3983104661242685e-05, + "loss": 1.2105, + "step": 13843 + }, + { + "epoch": 1.17, + "learning_rate": 1.3982270888659783e-05, + "loss": 1.1962, + "step": 13844 + }, + { + "epoch": 1.17, + "learning_rate": 1.3981437083174086e-05, + "loss": 1.2598, + "step": 13845 + }, + { + "epoch": 1.17, + "learning_rate": 1.3980603244792477e-05, + "loss": 1.1282, + "step": 13846 + }, + { + "epoch": 1.17, + "learning_rate": 1.3979769373521855e-05, + "loss": 1.1849, + "step": 13847 + }, + { + "epoch": 1.17, + "learning_rate": 1.3978935469369101e-05, + "loss": 1.1528, + "step": 13848 + }, + { + "epoch": 1.17, + "learning_rate": 1.397810153234111e-05, + "loss": 1.1648, + "step": 13849 + }, + { + "epoch": 1.17, + "learning_rate": 1.3977267562444773e-05, + "loss": 1.2429, + "step": 13850 + }, + { + "epoch": 1.17, + "learning_rate": 1.3976433559686977e-05, + "loss": 1.1019, + "step": 13851 + }, + { + "epoch": 1.17, + "learning_rate": 1.3975599524074615e-05, + "loss": 1.199, + "step": 13852 + }, + { + "epoch": 1.17, + "learning_rate": 1.3974765455614577e-05, + "loss": 1.1874, + "step": 13853 + }, + { + "epoch": 1.17, + "learning_rate": 1.3973931354313755e-05, + "loss": 1.2579, + "step": 13854 + }, + { + "epoch": 1.17, + "learning_rate": 1.3973097220179037e-05, + "loss": 1.2295, + "step": 13855 + }, + { + "epoch": 1.17, + "learning_rate": 1.3972263053217324e-05, + "loss": 1.2662, + "step": 13856 + }, + { + "epoch": 1.17, + "learning_rate": 1.3971428853435498e-05, + "loss": 1.2642, + "step": 13857 + }, + { + "epoch": 1.17, + "learning_rate": 1.3970594620840454e-05, + "loss": 1.2004, + "step": 13858 + }, + { + "epoch": 1.17, + "learning_rate": 1.396976035543909e-05, + "loss": 1.1248, + "step": 13859 + }, + { + "epoch": 1.17, + "learning_rate": 1.3968926057238294e-05, + "loss": 1.1681, + "step": 13860 + }, + { + "epoch": 1.17, + "learning_rate": 1.3968091726244963e-05, + "loss": 1.2773, + "step": 13861 + }, + { + "epoch": 1.17, + "learning_rate": 1.396725736246599e-05, + "loss": 1.18, + "step": 13862 + }, + { + "epoch": 1.17, + "learning_rate": 1.3966422965908262e-05, + "loss": 1.2799, + "step": 13863 + }, + { + "epoch": 1.17, + "learning_rate": 1.3965588536578678e-05, + "loss": 1.1148, + "step": 13864 + }, + { + "epoch": 1.17, + "learning_rate": 1.3964754074484137e-05, + "loss": 1.1488, + "step": 13865 + }, + { + "epoch": 1.17, + "learning_rate": 1.3963919579631526e-05, + "loss": 1.1126, + "step": 13866 + }, + { + "epoch": 1.17, + "learning_rate": 1.3963085052027743e-05, + "loss": 1.25, + "step": 13867 + }, + { + "epoch": 1.17, + "learning_rate": 1.3962250491679683e-05, + "loss": 1.2037, + "step": 13868 + }, + { + "epoch": 1.17, + "learning_rate": 1.3961415898594242e-05, + "loss": 1.2235, + "step": 13869 + }, + { + "epoch": 1.17, + "learning_rate": 1.3960581272778314e-05, + "loss": 1.1177, + "step": 13870 + }, + { + "epoch": 1.17, + "learning_rate": 1.3959746614238793e-05, + "loss": 1.1081, + "step": 13871 + }, + { + "epoch": 1.17, + "learning_rate": 1.3958911922982583e-05, + "loss": 1.2573, + "step": 13872 + }, + { + "epoch": 1.17, + "learning_rate": 1.3958077199016576e-05, + "loss": 1.097, + "step": 13873 + }, + { + "epoch": 1.17, + "learning_rate": 1.3957242442347665e-05, + "loss": 1.173, + "step": 13874 + }, + { + "epoch": 1.17, + "learning_rate": 1.395640765298275e-05, + "loss": 1.1775, + "step": 13875 + }, + { + "epoch": 1.17, + "learning_rate": 1.3955572830928734e-05, + "loss": 1.2597, + "step": 13876 + }, + { + "epoch": 1.17, + "learning_rate": 1.3954737976192507e-05, + "loss": 1.1882, + "step": 13877 + }, + { + "epoch": 1.17, + "learning_rate": 1.3953903088780967e-05, + "loss": 1.2068, + "step": 13878 + }, + { + "epoch": 1.17, + "learning_rate": 1.3953068168701015e-05, + "loss": 1.1518, + "step": 13879 + }, + { + "epoch": 1.17, + "learning_rate": 1.395223321595955e-05, + "loss": 1.2118, + "step": 13880 + }, + { + "epoch": 1.17, + "learning_rate": 1.3951398230563469e-05, + "loss": 1.307, + "step": 13881 + }, + { + "epoch": 1.17, + "learning_rate": 1.395056321251967e-05, + "loss": 1.2286, + "step": 13882 + }, + { + "epoch": 1.17, + "learning_rate": 1.3949728161835055e-05, + "loss": 1.2102, + "step": 13883 + }, + { + "epoch": 1.17, + "learning_rate": 1.3948893078516519e-05, + "loss": 1.209, + "step": 13884 + }, + { + "epoch": 1.17, + "learning_rate": 1.3948057962570967e-05, + "loss": 1.2269, + "step": 13885 + }, + { + "epoch": 1.17, + "learning_rate": 1.3947222814005299e-05, + "loss": 1.2398, + "step": 13886 + }, + { + "epoch": 1.17, + "learning_rate": 1.3946387632826408e-05, + "loss": 1.1989, + "step": 13887 + }, + { + "epoch": 1.17, + "learning_rate": 1.3945552419041202e-05, + "loss": 1.1457, + "step": 13888 + }, + { + "epoch": 1.17, + "learning_rate": 1.3944717172656579e-05, + "loss": 1.2023, + "step": 13889 + }, + { + "epoch": 1.17, + "learning_rate": 1.3943881893679441e-05, + "loss": 1.2077, + "step": 13890 + }, + { + "epoch": 1.17, + "learning_rate": 1.3943046582116686e-05, + "loss": 1.1143, + "step": 13891 + }, + { + "epoch": 1.17, + "learning_rate": 1.3942211237975221e-05, + "loss": 1.217, + "step": 13892 + }, + { + "epoch": 1.17, + "learning_rate": 1.394137586126194e-05, + "loss": 1.2345, + "step": 13893 + }, + { + "epoch": 1.17, + "learning_rate": 1.3940540451983758e-05, + "loss": 1.1779, + "step": 13894 + }, + { + "epoch": 1.17, + "learning_rate": 1.3939705010147563e-05, + "loss": 1.207, + "step": 13895 + }, + { + "epoch": 1.17, + "learning_rate": 1.3938869535760264e-05, + "loss": 1.2554, + "step": 13896 + }, + { + "epoch": 1.17, + "learning_rate": 1.393803402882877e-05, + "loss": 1.1907, + "step": 13897 + }, + { + "epoch": 1.17, + "learning_rate": 1.3937198489359976e-05, + "loss": 1.1834, + "step": 13898 + }, + { + "epoch": 1.17, + "learning_rate": 1.3936362917360789e-05, + "loss": 1.234, + "step": 13899 + }, + { + "epoch": 1.17, + "learning_rate": 1.3935527312838108e-05, + "loss": 1.2427, + "step": 13900 + }, + { + "epoch": 1.17, + "learning_rate": 1.3934691675798846e-05, + "loss": 1.192, + "step": 13901 + }, + { + "epoch": 1.17, + "learning_rate": 1.3933856006249897e-05, + "loss": 1.2753, + "step": 13902 + }, + { + "epoch": 1.17, + "learning_rate": 1.3933020304198175e-05, + "loss": 1.2146, + "step": 13903 + }, + { + "epoch": 1.17, + "learning_rate": 1.393218456965058e-05, + "loss": 1.2563, + "step": 13904 + }, + { + "epoch": 1.17, + "learning_rate": 1.3931348802614016e-05, + "loss": 1.1702, + "step": 13905 + }, + { + "epoch": 1.17, + "learning_rate": 1.393051300309539e-05, + "loss": 1.2845, + "step": 13906 + }, + { + "epoch": 1.17, + "learning_rate": 1.3929677171101608e-05, + "loss": 1.2638, + "step": 13907 + }, + { + "epoch": 1.17, + "learning_rate": 1.3928841306639573e-05, + "loss": 1.2984, + "step": 13908 + }, + { + "epoch": 1.17, + "learning_rate": 1.3928005409716196e-05, + "loss": 1.1782, + "step": 13909 + }, + { + "epoch": 1.17, + "learning_rate": 1.392716948033838e-05, + "loss": 1.2601, + "step": 13910 + }, + { + "epoch": 1.17, + "learning_rate": 1.3926333518513037e-05, + "loss": 1.197, + "step": 13911 + }, + { + "epoch": 1.17, + "learning_rate": 1.3925497524247067e-05, + "loss": 1.243, + "step": 13912 + }, + { + "epoch": 1.17, + "learning_rate": 1.3924661497547381e-05, + "loss": 1.1852, + "step": 13913 + }, + { + "epoch": 1.17, + "learning_rate": 1.3923825438420884e-05, + "loss": 1.1762, + "step": 13914 + }, + { + "epoch": 1.17, + "learning_rate": 1.3922989346874487e-05, + "loss": 1.2238, + "step": 13915 + }, + { + "epoch": 1.17, + "learning_rate": 1.3922153222915096e-05, + "loss": 1.2215, + "step": 13916 + }, + { + "epoch": 1.17, + "learning_rate": 1.392131706654962e-05, + "loss": 1.2876, + "step": 13917 + }, + { + "epoch": 1.17, + "learning_rate": 1.3920480877784965e-05, + "loss": 1.1424, + "step": 13918 + }, + { + "epoch": 1.17, + "learning_rate": 1.3919644656628047e-05, + "loss": 1.2953, + "step": 13919 + }, + { + "epoch": 1.17, + "learning_rate": 1.3918808403085766e-05, + "loss": 1.1926, + "step": 13920 + }, + { + "epoch": 1.17, + "learning_rate": 1.3917972117165038e-05, + "loss": 1.1927, + "step": 13921 + }, + { + "epoch": 1.17, + "learning_rate": 1.391713579887277e-05, + "loss": 1.2524, + "step": 13922 + }, + { + "epoch": 1.17, + "learning_rate": 1.3916299448215873e-05, + "loss": 1.1769, + "step": 13923 + }, + { + "epoch": 1.17, + "learning_rate": 1.3915463065201255e-05, + "loss": 1.2224, + "step": 13924 + }, + { + "epoch": 1.17, + "learning_rate": 1.3914626649835827e-05, + "loss": 1.2552, + "step": 13925 + }, + { + "epoch": 1.17, + "learning_rate": 1.3913790202126505e-05, + "loss": 1.2084, + "step": 13926 + }, + { + "epoch": 1.17, + "learning_rate": 1.3912953722080194e-05, + "loss": 1.1764, + "step": 13927 + }, + { + "epoch": 1.17, + "learning_rate": 1.3912117209703806e-05, + "loss": 1.1191, + "step": 13928 + }, + { + "epoch": 1.17, + "learning_rate": 1.3911280665004256e-05, + "loss": 1.2365, + "step": 13929 + }, + { + "epoch": 1.17, + "learning_rate": 1.3910444087988451e-05, + "loss": 1.1855, + "step": 13930 + }, + { + "epoch": 1.17, + "learning_rate": 1.3909607478663305e-05, + "loss": 1.1692, + "step": 13931 + }, + { + "epoch": 1.17, + "learning_rate": 1.3908770837035732e-05, + "loss": 1.2381, + "step": 13932 + }, + { + "epoch": 1.17, + "learning_rate": 1.3907934163112644e-05, + "loss": 1.2338, + "step": 13933 + }, + { + "epoch": 1.17, + "learning_rate": 1.390709745690095e-05, + "loss": 1.1453, + "step": 13934 + }, + { + "epoch": 1.17, + "learning_rate": 1.3906260718407569e-05, + "loss": 1.1552, + "step": 13935 + }, + { + "epoch": 1.17, + "learning_rate": 1.3905423947639413e-05, + "loss": 1.1356, + "step": 13936 + }, + { + "epoch": 1.17, + "learning_rate": 1.3904587144603395e-05, + "loss": 1.1518, + "step": 13937 + }, + { + "epoch": 1.17, + "learning_rate": 1.3903750309306426e-05, + "loss": 1.2363, + "step": 13938 + }, + { + "epoch": 1.17, + "learning_rate": 1.3902913441755422e-05, + "loss": 1.1784, + "step": 13939 + }, + { + "epoch": 1.17, + "learning_rate": 1.3902076541957299e-05, + "loss": 1.118, + "step": 13940 + }, + { + "epoch": 1.17, + "learning_rate": 1.3901239609918972e-05, + "loss": 1.2333, + "step": 13941 + }, + { + "epoch": 1.17, + "learning_rate": 1.3900402645647353e-05, + "loss": 1.2549, + "step": 13942 + }, + { + "epoch": 1.17, + "learning_rate": 1.389956564914936e-05, + "loss": 1.1783, + "step": 13943 + }, + { + "epoch": 1.17, + "learning_rate": 1.3898728620431908e-05, + "loss": 1.2413, + "step": 13944 + }, + { + "epoch": 1.17, + "learning_rate": 1.3897891559501911e-05, + "loss": 1.1179, + "step": 13945 + }, + { + "epoch": 1.17, + "learning_rate": 1.3897054466366285e-05, + "loss": 1.234, + "step": 13946 + }, + { + "epoch": 1.17, + "learning_rate": 1.3896217341031952e-05, + "loss": 1.2012, + "step": 13947 + }, + { + "epoch": 1.17, + "learning_rate": 1.3895380183505822e-05, + "loss": 1.2659, + "step": 13948 + }, + { + "epoch": 1.17, + "learning_rate": 1.3894542993794814e-05, + "loss": 1.2376, + "step": 13949 + }, + { + "epoch": 1.17, + "learning_rate": 1.3893705771905845e-05, + "loss": 1.2114, + "step": 13950 + }, + { + "epoch": 1.17, + "learning_rate": 1.3892868517845837e-05, + "loss": 1.2643, + "step": 13951 + }, + { + "epoch": 1.17, + "learning_rate": 1.3892031231621698e-05, + "loss": 1.27, + "step": 13952 + }, + { + "epoch": 1.17, + "learning_rate": 1.3891193913240354e-05, + "loss": 1.2659, + "step": 13953 + }, + { + "epoch": 1.17, + "learning_rate": 1.3890356562708722e-05, + "loss": 1.2394, + "step": 13954 + }, + { + "epoch": 1.17, + "learning_rate": 1.3889519180033716e-05, + "loss": 1.1629, + "step": 13955 + }, + { + "epoch": 1.17, + "learning_rate": 1.3888681765222258e-05, + "loss": 1.1504, + "step": 13956 + }, + { + "epoch": 1.18, + "learning_rate": 1.3887844318281268e-05, + "loss": 1.1808, + "step": 13957 + }, + { + "epoch": 1.18, + "learning_rate": 1.3887006839217663e-05, + "loss": 1.18, + "step": 13958 + }, + { + "epoch": 1.18, + "learning_rate": 1.3886169328038364e-05, + "loss": 1.1571, + "step": 13959 + }, + { + "epoch": 1.18, + "learning_rate": 1.3885331784750288e-05, + "loss": 1.2094, + "step": 13960 + }, + { + "epoch": 1.18, + "learning_rate": 1.3884494209360362e-05, + "loss": 1.2157, + "step": 13961 + }, + { + "epoch": 1.18, + "learning_rate": 1.3883656601875499e-05, + "loss": 1.2368, + "step": 13962 + }, + { + "epoch": 1.18, + "learning_rate": 1.3882818962302621e-05, + "loss": 1.2255, + "step": 13963 + }, + { + "epoch": 1.18, + "learning_rate": 1.3881981290648651e-05, + "loss": 1.2335, + "step": 13964 + }, + { + "epoch": 1.18, + "learning_rate": 1.388114358692051e-05, + "loss": 1.2477, + "step": 13965 + }, + { + "epoch": 1.18, + "learning_rate": 1.388030585112512e-05, + "loss": 1.1442, + "step": 13966 + }, + { + "epoch": 1.18, + "learning_rate": 1.3879468083269399e-05, + "loss": 1.2408, + "step": 13967 + }, + { + "epoch": 1.18, + "learning_rate": 1.387863028336027e-05, + "loss": 1.1925, + "step": 13968 + }, + { + "epoch": 1.18, + "learning_rate": 1.3877792451404658e-05, + "loss": 1.1413, + "step": 13969 + }, + { + "epoch": 1.18, + "learning_rate": 1.3876954587409485e-05, + "loss": 1.2397, + "step": 13970 + }, + { + "epoch": 1.18, + "learning_rate": 1.387611669138167e-05, + "loss": 1.1142, + "step": 13971 + }, + { + "epoch": 1.18, + "learning_rate": 1.3875278763328139e-05, + "loss": 1.2782, + "step": 13972 + }, + { + "epoch": 1.18, + "learning_rate": 1.3874440803255815e-05, + "loss": 1.2141, + "step": 13973 + }, + { + "epoch": 1.18, + "learning_rate": 1.387360281117162e-05, + "loss": 1.1166, + "step": 13974 + }, + { + "epoch": 1.18, + "learning_rate": 1.387276478708248e-05, + "loss": 1.1483, + "step": 13975 + }, + { + "epoch": 1.18, + "learning_rate": 1.387192673099532e-05, + "loss": 1.2362, + "step": 13976 + }, + { + "epoch": 1.18, + "learning_rate": 1.3871088642917059e-05, + "loss": 1.193, + "step": 13977 + }, + { + "epoch": 1.18, + "learning_rate": 1.3870250522854624e-05, + "loss": 1.2887, + "step": 13978 + }, + { + "epoch": 1.18, + "learning_rate": 1.3869412370814945e-05, + "loss": 1.2482, + "step": 13979 + }, + { + "epoch": 1.18, + "learning_rate": 1.386857418680494e-05, + "loss": 1.2519, + "step": 13980 + }, + { + "epoch": 1.18, + "learning_rate": 1.3867735970831536e-05, + "loss": 1.2071, + "step": 13981 + }, + { + "epoch": 1.18, + "learning_rate": 1.3866897722901659e-05, + "loss": 1.1645, + "step": 13982 + }, + { + "epoch": 1.18, + "learning_rate": 1.3866059443022237e-05, + "loss": 1.1703, + "step": 13983 + }, + { + "epoch": 1.18, + "learning_rate": 1.3865221131200196e-05, + "loss": 1.2444, + "step": 13984 + }, + { + "epoch": 1.18, + "learning_rate": 1.3864382787442456e-05, + "loss": 1.1952, + "step": 13985 + }, + { + "epoch": 1.18, + "learning_rate": 1.3863544411755955e-05, + "loss": 1.1575, + "step": 13986 + }, + { + "epoch": 1.18, + "learning_rate": 1.386270600414761e-05, + "loss": 1.2614, + "step": 13987 + }, + { + "epoch": 1.18, + "learning_rate": 1.3861867564624351e-05, + "loss": 1.226, + "step": 13988 + }, + { + "epoch": 1.18, + "learning_rate": 1.3861029093193109e-05, + "loss": 1.209, + "step": 13989 + }, + { + "epoch": 1.18, + "learning_rate": 1.386019058986081e-05, + "loss": 1.1862, + "step": 13990 + }, + { + "epoch": 1.18, + "learning_rate": 1.3859352054634376e-05, + "loss": 1.1258, + "step": 13991 + }, + { + "epoch": 1.18, + "learning_rate": 1.3858513487520741e-05, + "loss": 1.1331, + "step": 13992 + }, + { + "epoch": 1.18, + "learning_rate": 1.3857674888526835e-05, + "loss": 1.2282, + "step": 13993 + }, + { + "epoch": 1.18, + "learning_rate": 1.3856836257659587e-05, + "loss": 1.2852, + "step": 13994 + }, + { + "epoch": 1.18, + "learning_rate": 1.385599759492592e-05, + "loss": 1.1846, + "step": 13995 + }, + { + "epoch": 1.18, + "learning_rate": 1.3855158900332764e-05, + "loss": 1.1946, + "step": 13996 + }, + { + "epoch": 1.18, + "learning_rate": 1.3854320173887055e-05, + "loss": 1.2784, + "step": 13997 + }, + { + "epoch": 1.18, + "learning_rate": 1.385348141559572e-05, + "loss": 1.2216, + "step": 13998 + }, + { + "epoch": 1.18, + "learning_rate": 1.3852642625465684e-05, + "loss": 1.1484, + "step": 13999 + }, + { + "epoch": 1.18, + "learning_rate": 1.3851803803503887e-05, + "loss": 1.1532, + "step": 14000 + }, + { + "epoch": 1.18, + "learning_rate": 1.3850964949717252e-05, + "loss": 1.2527, + "step": 14001 + }, + { + "epoch": 1.18, + "learning_rate": 1.3850126064112712e-05, + "loss": 1.2134, + "step": 14002 + }, + { + "epoch": 1.18, + "learning_rate": 1.3849287146697196e-05, + "loss": 1.2332, + "step": 14003 + }, + { + "epoch": 1.18, + "learning_rate": 1.3848448197477642e-05, + "loss": 1.2039, + "step": 14004 + }, + { + "epoch": 1.18, + "learning_rate": 1.3847609216460976e-05, + "loss": 1.1769, + "step": 14005 + }, + { + "epoch": 1.18, + "learning_rate": 1.384677020365413e-05, + "loss": 1.3336, + "step": 14006 + }, + { + "epoch": 1.18, + "learning_rate": 1.3845931159064036e-05, + "loss": 1.2094, + "step": 14007 + }, + { + "epoch": 1.18, + "learning_rate": 1.3845092082697631e-05, + "loss": 1.1396, + "step": 14008 + }, + { + "epoch": 1.18, + "learning_rate": 1.3844252974561843e-05, + "loss": 1.1173, + "step": 14009 + }, + { + "epoch": 1.18, + "learning_rate": 1.3843413834663606e-05, + "loss": 1.17, + "step": 14010 + }, + { + "epoch": 1.18, + "learning_rate": 1.3842574663009856e-05, + "loss": 1.1559, + "step": 14011 + }, + { + "epoch": 1.18, + "learning_rate": 1.3841735459607523e-05, + "loss": 1.1327, + "step": 14012 + }, + { + "epoch": 1.18, + "learning_rate": 1.3840896224463542e-05, + "loss": 1.2403, + "step": 14013 + }, + { + "epoch": 1.18, + "learning_rate": 1.3840056957584846e-05, + "loss": 1.2432, + "step": 14014 + }, + { + "epoch": 1.18, + "learning_rate": 1.3839217658978375e-05, + "loss": 1.276, + "step": 14015 + }, + { + "epoch": 1.18, + "learning_rate": 1.3838378328651056e-05, + "loss": 1.2521, + "step": 14016 + }, + { + "epoch": 1.18, + "learning_rate": 1.3837538966609825e-05, + "loss": 1.2115, + "step": 14017 + }, + { + "epoch": 1.18, + "learning_rate": 1.3836699572861621e-05, + "loss": 1.2457, + "step": 14018 + }, + { + "epoch": 1.18, + "learning_rate": 1.3835860147413379e-05, + "loss": 1.1972, + "step": 14019 + }, + { + "epoch": 1.18, + "learning_rate": 1.383502069027203e-05, + "loss": 1.0568, + "step": 14020 + }, + { + "epoch": 1.18, + "learning_rate": 1.383418120144451e-05, + "loss": 1.2114, + "step": 14021 + }, + { + "epoch": 1.18, + "learning_rate": 1.3833341680937764e-05, + "loss": 1.1642, + "step": 14022 + }, + { + "epoch": 1.18, + "learning_rate": 1.3832502128758717e-05, + "loss": 1.2043, + "step": 14023 + }, + { + "epoch": 1.18, + "learning_rate": 1.3831662544914312e-05, + "loss": 1.1554, + "step": 14024 + }, + { + "epoch": 1.18, + "learning_rate": 1.3830822929411487e-05, + "loss": 1.2009, + "step": 14025 + }, + { + "epoch": 1.18, + "learning_rate": 1.3829983282257179e-05, + "loss": 1.2062, + "step": 14026 + }, + { + "epoch": 1.18, + "learning_rate": 1.3829143603458319e-05, + "loss": 1.2118, + "step": 14027 + }, + { + "epoch": 1.18, + "learning_rate": 1.382830389302185e-05, + "loss": 1.2906, + "step": 14028 + }, + { + "epoch": 1.18, + "learning_rate": 1.3827464150954708e-05, + "loss": 1.2438, + "step": 14029 + }, + { + "epoch": 1.18, + "learning_rate": 1.3826624377263836e-05, + "loss": 1.1633, + "step": 14030 + }, + { + "epoch": 1.18, + "learning_rate": 1.3825784571956167e-05, + "loss": 1.1572, + "step": 14031 + }, + { + "epoch": 1.18, + "learning_rate": 1.382494473503864e-05, + "loss": 1.1478, + "step": 14032 + }, + { + "epoch": 1.18, + "learning_rate": 1.3824104866518196e-05, + "loss": 1.2476, + "step": 14033 + }, + { + "epoch": 1.18, + "learning_rate": 1.3823264966401772e-05, + "loss": 1.2506, + "step": 14034 + }, + { + "epoch": 1.18, + "learning_rate": 1.3822425034696311e-05, + "loss": 1.1899, + "step": 14035 + }, + { + "epoch": 1.18, + "learning_rate": 1.3821585071408754e-05, + "loss": 1.3561, + "step": 14036 + }, + { + "epoch": 1.18, + "learning_rate": 1.3820745076546033e-05, + "loss": 1.2256, + "step": 14037 + }, + { + "epoch": 1.18, + "learning_rate": 1.3819905050115097e-05, + "loss": 1.1456, + "step": 14038 + }, + { + "epoch": 1.18, + "learning_rate": 1.3819064992122882e-05, + "loss": 1.2144, + "step": 14039 + }, + { + "epoch": 1.18, + "learning_rate": 1.381822490257633e-05, + "loss": 1.1624, + "step": 14040 + }, + { + "epoch": 1.18, + "learning_rate": 1.381738478148238e-05, + "loss": 1.1676, + "step": 14041 + }, + { + "epoch": 1.18, + "learning_rate": 1.3816544628847976e-05, + "loss": 1.205, + "step": 14042 + }, + { + "epoch": 1.18, + "learning_rate": 1.3815704444680061e-05, + "loss": 1.2168, + "step": 14043 + }, + { + "epoch": 1.18, + "learning_rate": 1.3814864228985574e-05, + "loss": 1.3153, + "step": 14044 + }, + { + "epoch": 1.18, + "learning_rate": 1.3814023981771455e-05, + "loss": 1.1297, + "step": 14045 + }, + { + "epoch": 1.18, + "learning_rate": 1.381318370304465e-05, + "loss": 1.1534, + "step": 14046 + }, + { + "epoch": 1.18, + "learning_rate": 1.3812343392812104e-05, + "loss": 1.1588, + "step": 14047 + }, + { + "epoch": 1.18, + "learning_rate": 1.3811503051080752e-05, + "loss": 1.21, + "step": 14048 + }, + { + "epoch": 1.18, + "learning_rate": 1.3810662677857548e-05, + "loss": 1.1399, + "step": 14049 + }, + { + "epoch": 1.18, + "learning_rate": 1.3809822273149424e-05, + "loss": 1.1453, + "step": 14050 + }, + { + "epoch": 1.18, + "learning_rate": 1.3808981836963334e-05, + "loss": 1.1991, + "step": 14051 + }, + { + "epoch": 1.18, + "learning_rate": 1.3808141369306212e-05, + "loss": 1.2127, + "step": 14052 + }, + { + "epoch": 1.18, + "learning_rate": 1.380730087018501e-05, + "loss": 1.1496, + "step": 14053 + }, + { + "epoch": 1.18, + "learning_rate": 1.3806460339606669e-05, + "loss": 1.2243, + "step": 14054 + }, + { + "epoch": 1.18, + "learning_rate": 1.3805619777578137e-05, + "loss": 1.2148, + "step": 14055 + }, + { + "epoch": 1.18, + "learning_rate": 1.3804779184106351e-05, + "loss": 1.1392, + "step": 14056 + }, + { + "epoch": 1.18, + "learning_rate": 1.3803938559198264e-05, + "loss": 1.2457, + "step": 14057 + }, + { + "epoch": 1.18, + "learning_rate": 1.380309790286082e-05, + "loss": 1.1666, + "step": 14058 + }, + { + "epoch": 1.18, + "learning_rate": 1.3802257215100965e-05, + "loss": 1.1969, + "step": 14059 + }, + { + "epoch": 1.18, + "learning_rate": 1.3801416495925641e-05, + "loss": 1.2078, + "step": 14060 + }, + { + "epoch": 1.18, + "learning_rate": 1.3800575745341802e-05, + "loss": 1.182, + "step": 14061 + }, + { + "epoch": 1.18, + "learning_rate": 1.3799734963356386e-05, + "loss": 1.1446, + "step": 14062 + }, + { + "epoch": 1.18, + "learning_rate": 1.3798894149976344e-05, + "loss": 1.1454, + "step": 14063 + }, + { + "epoch": 1.18, + "learning_rate": 1.3798053305208625e-05, + "loss": 1.0685, + "step": 14064 + }, + { + "epoch": 1.18, + "learning_rate": 1.3797212429060174e-05, + "loss": 1.1559, + "step": 14065 + }, + { + "epoch": 1.18, + "learning_rate": 1.3796371521537936e-05, + "loss": 1.202, + "step": 14066 + }, + { + "epoch": 1.18, + "learning_rate": 1.3795530582648862e-05, + "loss": 1.1814, + "step": 14067 + }, + { + "epoch": 1.18, + "learning_rate": 1.3794689612399901e-05, + "loss": 1.1914, + "step": 14068 + }, + { + "epoch": 1.18, + "learning_rate": 1.3793848610797999e-05, + "loss": 1.2949, + "step": 14069 + }, + { + "epoch": 1.18, + "learning_rate": 1.3793007577850106e-05, + "loss": 1.1708, + "step": 14070 + }, + { + "epoch": 1.18, + "learning_rate": 1.3792166513563167e-05, + "loss": 1.1888, + "step": 14071 + }, + { + "epoch": 1.18, + "learning_rate": 1.3791325417944142e-05, + "loss": 1.2602, + "step": 14072 + }, + { + "epoch": 1.18, + "learning_rate": 1.3790484290999969e-05, + "loss": 1.2042, + "step": 14073 + }, + { + "epoch": 1.18, + "learning_rate": 1.37896431327376e-05, + "loss": 1.2431, + "step": 14074 + }, + { + "epoch": 1.18, + "learning_rate": 1.3788801943163987e-05, + "loss": 1.1949, + "step": 14075 + }, + { + "epoch": 1.19, + "learning_rate": 1.3787960722286085e-05, + "loss": 1.1651, + "step": 14076 + }, + { + "epoch": 1.19, + "learning_rate": 1.3787119470110832e-05, + "loss": 1.1728, + "step": 14077 + }, + { + "epoch": 1.19, + "learning_rate": 1.378627818664519e-05, + "loss": 1.2041, + "step": 14078 + }, + { + "epoch": 1.19, + "learning_rate": 1.3785436871896106e-05, + "loss": 1.1635, + "step": 14079 + }, + { + "epoch": 1.19, + "learning_rate": 1.378459552587053e-05, + "loss": 1.1862, + "step": 14080 + }, + { + "epoch": 1.19, + "learning_rate": 1.3783754148575414e-05, + "loss": 1.2062, + "step": 14081 + }, + { + "epoch": 1.19, + "learning_rate": 1.3782912740017711e-05, + "loss": 1.1498, + "step": 14082 + }, + { + "epoch": 1.19, + "learning_rate": 1.3782071300204372e-05, + "loss": 1.1951, + "step": 14083 + }, + { + "epoch": 1.19, + "learning_rate": 1.3781229829142348e-05, + "loss": 1.1452, + "step": 14084 + }, + { + "epoch": 1.19, + "learning_rate": 1.3780388326838594e-05, + "loss": 1.1756, + "step": 14085 + }, + { + "epoch": 1.19, + "learning_rate": 1.3779546793300065e-05, + "loss": 1.3252, + "step": 14086 + }, + { + "epoch": 1.19, + "learning_rate": 1.3778705228533707e-05, + "loss": 1.2171, + "step": 14087 + }, + { + "epoch": 1.19, + "learning_rate": 1.3777863632546478e-05, + "loss": 1.2162, + "step": 14088 + }, + { + "epoch": 1.19, + "learning_rate": 1.377702200534533e-05, + "loss": 1.2127, + "step": 14089 + }, + { + "epoch": 1.19, + "learning_rate": 1.377618034693722e-05, + "loss": 1.1396, + "step": 14090 + }, + { + "epoch": 1.19, + "learning_rate": 1.3775338657329098e-05, + "loss": 1.1653, + "step": 14091 + }, + { + "epoch": 1.19, + "learning_rate": 1.3774496936527918e-05, + "loss": 1.2065, + "step": 14092 + }, + { + "epoch": 1.19, + "learning_rate": 1.3773655184540637e-05, + "loss": 1.1449, + "step": 14093 + }, + { + "epoch": 1.19, + "learning_rate": 1.3772813401374213e-05, + "loss": 1.1543, + "step": 14094 + }, + { + "epoch": 1.19, + "learning_rate": 1.3771971587035593e-05, + "loss": 1.1887, + "step": 14095 + }, + { + "epoch": 1.19, + "learning_rate": 1.3771129741531736e-05, + "loss": 1.1489, + "step": 14096 + }, + { + "epoch": 1.19, + "learning_rate": 1.37702878648696e-05, + "loss": 1.1733, + "step": 14097 + }, + { + "epoch": 1.19, + "learning_rate": 1.3769445957056141e-05, + "loss": 1.1716, + "step": 14098 + }, + { + "epoch": 1.19, + "learning_rate": 1.3768604018098309e-05, + "loss": 1.0736, + "step": 14099 + }, + { + "epoch": 1.19, + "learning_rate": 1.3767762048003066e-05, + "loss": 1.319, + "step": 14100 + }, + { + "epoch": 1.19, + "learning_rate": 1.3766920046777367e-05, + "loss": 1.1823, + "step": 14101 + }, + { + "epoch": 1.19, + "learning_rate": 1.376607801442817e-05, + "loss": 1.2065, + "step": 14102 + }, + { + "epoch": 1.19, + "learning_rate": 1.3765235950962432e-05, + "loss": 1.1934, + "step": 14103 + }, + { + "epoch": 1.19, + "learning_rate": 1.3764393856387109e-05, + "loss": 1.2242, + "step": 14104 + }, + { + "epoch": 1.19, + "learning_rate": 1.376355173070916e-05, + "loss": 1.1868, + "step": 14105 + }, + { + "epoch": 1.19, + "learning_rate": 1.3762709573935538e-05, + "loss": 1.1955, + "step": 14106 + }, + { + "epoch": 1.19, + "learning_rate": 1.3761867386073209e-05, + "loss": 1.2447, + "step": 14107 + }, + { + "epoch": 1.19, + "learning_rate": 1.3761025167129125e-05, + "loss": 1.154, + "step": 14108 + }, + { + "epoch": 1.19, + "learning_rate": 1.376018291711025e-05, + "loss": 1.2475, + "step": 14109 + }, + { + "epoch": 1.19, + "learning_rate": 1.3759340636023537e-05, + "loss": 1.2202, + "step": 14110 + }, + { + "epoch": 1.19, + "learning_rate": 1.3758498323875953e-05, + "loss": 1.2088, + "step": 14111 + }, + { + "epoch": 1.19, + "learning_rate": 1.3757655980674449e-05, + "loss": 1.2275, + "step": 14112 + }, + { + "epoch": 1.19, + "learning_rate": 1.3756813606425992e-05, + "loss": 1.164, + "step": 14113 + }, + { + "epoch": 1.19, + "learning_rate": 1.3755971201137537e-05, + "loss": 1.2296, + "step": 14114 + }, + { + "epoch": 1.19, + "learning_rate": 1.3755128764816048e-05, + "loss": 1.1738, + "step": 14115 + }, + { + "epoch": 1.19, + "learning_rate": 1.375428629746848e-05, + "loss": 1.3114, + "step": 14116 + }, + { + "epoch": 1.19, + "learning_rate": 1.3753443799101798e-05, + "loss": 1.1536, + "step": 14117 + }, + { + "epoch": 1.19, + "learning_rate": 1.3752601269722963e-05, + "loss": 1.1933, + "step": 14118 + }, + { + "epoch": 1.19, + "learning_rate": 1.3751758709338935e-05, + "loss": 1.1764, + "step": 14119 + }, + { + "epoch": 1.19, + "learning_rate": 1.3750916117956674e-05, + "loss": 1.2179, + "step": 14120 + }, + { + "epoch": 1.19, + "learning_rate": 1.3750073495583145e-05, + "loss": 1.0682, + "step": 14121 + }, + { + "epoch": 1.19, + "learning_rate": 1.374923084222531e-05, + "loss": 1.1537, + "step": 14122 + }, + { + "epoch": 1.19, + "learning_rate": 1.3748388157890129e-05, + "loss": 1.2171, + "step": 14123 + }, + { + "epoch": 1.19, + "learning_rate": 1.3747545442584564e-05, + "loss": 1.2593, + "step": 14124 + }, + { + "epoch": 1.19, + "learning_rate": 1.374670269631558e-05, + "loss": 1.2127, + "step": 14125 + }, + { + "epoch": 1.19, + "learning_rate": 1.3745859919090141e-05, + "loss": 1.2979, + "step": 14126 + }, + { + "epoch": 1.19, + "learning_rate": 1.3745017110915207e-05, + "loss": 1.1772, + "step": 14127 + }, + { + "epoch": 1.19, + "learning_rate": 1.3744174271797742e-05, + "loss": 1.247, + "step": 14128 + }, + { + "epoch": 1.19, + "learning_rate": 1.374333140174471e-05, + "loss": 1.1997, + "step": 14129 + }, + { + "epoch": 1.19, + "learning_rate": 1.374248850076308e-05, + "loss": 1.2028, + "step": 14130 + }, + { + "epoch": 1.19, + "learning_rate": 1.374164556885981e-05, + "loss": 1.2727, + "step": 14131 + }, + { + "epoch": 1.19, + "learning_rate": 1.3740802606041868e-05, + "loss": 1.1766, + "step": 14132 + }, + { + "epoch": 1.19, + "learning_rate": 1.3739959612316216e-05, + "loss": 1.2136, + "step": 14133 + }, + { + "epoch": 1.19, + "learning_rate": 1.373911658768982e-05, + "loss": 1.0815, + "step": 14134 + }, + { + "epoch": 1.19, + "learning_rate": 1.3738273532169648e-05, + "loss": 1.2324, + "step": 14135 + }, + { + "epoch": 1.19, + "learning_rate": 1.3737430445762664e-05, + "loss": 1.1667, + "step": 14136 + }, + { + "epoch": 1.19, + "learning_rate": 1.3736587328475832e-05, + "loss": 1.1313, + "step": 14137 + }, + { + "epoch": 1.19, + "learning_rate": 1.3735744180316121e-05, + "loss": 1.1232, + "step": 14138 + }, + { + "epoch": 1.19, + "learning_rate": 1.3734901001290494e-05, + "loss": 1.2097, + "step": 14139 + }, + { + "epoch": 1.19, + "learning_rate": 1.3734057791405924e-05, + "loss": 1.31, + "step": 14140 + }, + { + "epoch": 1.19, + "learning_rate": 1.373321455066937e-05, + "loss": 1.238, + "step": 14141 + }, + { + "epoch": 1.19, + "learning_rate": 1.3732371279087801e-05, + "loss": 1.2476, + "step": 14142 + }, + { + "epoch": 1.19, + "learning_rate": 1.3731527976668188e-05, + "loss": 1.1649, + "step": 14143 + }, + { + "epoch": 1.19, + "learning_rate": 1.3730684643417497e-05, + "loss": 1.1967, + "step": 14144 + }, + { + "epoch": 1.19, + "learning_rate": 1.3729841279342697e-05, + "loss": 1.1912, + "step": 14145 + }, + { + "epoch": 1.19, + "learning_rate": 1.372899788445075e-05, + "loss": 1.1092, + "step": 14146 + }, + { + "epoch": 1.19, + "learning_rate": 1.3728154458748632e-05, + "loss": 1.2086, + "step": 14147 + }, + { + "epoch": 1.19, + "learning_rate": 1.3727311002243309e-05, + "loss": 1.1376, + "step": 14148 + }, + { + "epoch": 1.19, + "learning_rate": 1.3726467514941748e-05, + "loss": 1.1993, + "step": 14149 + }, + { + "epoch": 1.19, + "learning_rate": 1.3725623996850921e-05, + "loss": 1.2479, + "step": 14150 + }, + { + "epoch": 1.19, + "learning_rate": 1.3724780447977797e-05, + "loss": 1.1733, + "step": 14151 + }, + { + "epoch": 1.19, + "learning_rate": 1.3723936868329344e-05, + "loss": 1.2432, + "step": 14152 + }, + { + "epoch": 1.19, + "learning_rate": 1.3723093257912532e-05, + "loss": 1.1749, + "step": 14153 + }, + { + "epoch": 1.19, + "learning_rate": 1.3722249616734331e-05, + "loss": 1.1355, + "step": 14154 + }, + { + "epoch": 1.19, + "learning_rate": 1.3721405944801718e-05, + "loss": 1.2216, + "step": 14155 + }, + { + "epoch": 1.19, + "learning_rate": 1.3720562242121654e-05, + "loss": 1.2009, + "step": 14156 + }, + { + "epoch": 1.19, + "learning_rate": 1.3719718508701113e-05, + "loss": 1.1837, + "step": 14157 + }, + { + "epoch": 1.19, + "learning_rate": 1.3718874744547069e-05, + "loss": 1.1457, + "step": 14158 + }, + { + "epoch": 1.19, + "learning_rate": 1.3718030949666488e-05, + "loss": 1.1704, + "step": 14159 + }, + { + "epoch": 1.19, + "learning_rate": 1.371718712406635e-05, + "loss": 1.3074, + "step": 14160 + }, + { + "epoch": 1.19, + "learning_rate": 1.3716343267753622e-05, + "loss": 1.126, + "step": 14161 + }, + { + "epoch": 1.19, + "learning_rate": 1.3715499380735274e-05, + "loss": 1.1616, + "step": 14162 + }, + { + "epoch": 1.19, + "learning_rate": 1.3714655463018283e-05, + "loss": 1.2376, + "step": 14163 + }, + { + "epoch": 1.19, + "learning_rate": 1.3713811514609616e-05, + "loss": 1.2333, + "step": 14164 + }, + { + "epoch": 1.19, + "learning_rate": 1.3712967535516258e-05, + "loss": 1.1716, + "step": 14165 + }, + { + "epoch": 1.19, + "learning_rate": 1.3712123525745166e-05, + "loss": 1.1363, + "step": 14166 + }, + { + "epoch": 1.19, + "learning_rate": 1.3711279485303324e-05, + "loss": 1.2677, + "step": 14167 + }, + { + "epoch": 1.19, + "learning_rate": 1.3710435414197703e-05, + "loss": 1.1865, + "step": 14168 + }, + { + "epoch": 1.19, + "learning_rate": 1.3709591312435277e-05, + "loss": 1.176, + "step": 14169 + }, + { + "epoch": 1.19, + "learning_rate": 1.370874718002302e-05, + "loss": 1.2733, + "step": 14170 + }, + { + "epoch": 1.19, + "learning_rate": 1.3707903016967908e-05, + "loss": 1.2223, + "step": 14171 + }, + { + "epoch": 1.19, + "learning_rate": 1.3707058823276915e-05, + "loss": 1.1657, + "step": 14172 + }, + { + "epoch": 1.19, + "learning_rate": 1.3706214598957013e-05, + "loss": 1.088, + "step": 14173 + }, + { + "epoch": 1.19, + "learning_rate": 1.3705370344015181e-05, + "loss": 1.25, + "step": 14174 + }, + { + "epoch": 1.19, + "learning_rate": 1.3704526058458394e-05, + "loss": 1.3067, + "step": 14175 + }, + { + "epoch": 1.19, + "learning_rate": 1.3703681742293629e-05, + "loss": 1.2887, + "step": 14176 + }, + { + "epoch": 1.19, + "learning_rate": 1.3702837395527857e-05, + "loss": 1.2629, + "step": 14177 + }, + { + "epoch": 1.19, + "learning_rate": 1.3701993018168058e-05, + "loss": 1.154, + "step": 14178 + }, + { + "epoch": 1.19, + "learning_rate": 1.3701148610221209e-05, + "loss": 1.2343, + "step": 14179 + }, + { + "epoch": 1.19, + "learning_rate": 1.3700304171694286e-05, + "loss": 1.2787, + "step": 14180 + }, + { + "epoch": 1.19, + "learning_rate": 1.3699459702594264e-05, + "loss": 1.1738, + "step": 14181 + }, + { + "epoch": 1.19, + "learning_rate": 1.3698615202928122e-05, + "loss": 1.1732, + "step": 14182 + }, + { + "epoch": 1.19, + "learning_rate": 1.3697770672702837e-05, + "loss": 1.1436, + "step": 14183 + }, + { + "epoch": 1.19, + "learning_rate": 1.3696926111925389e-05, + "loss": 1.1431, + "step": 14184 + }, + { + "epoch": 1.19, + "learning_rate": 1.3696081520602754e-05, + "loss": 1.1998, + "step": 14185 + }, + { + "epoch": 1.19, + "learning_rate": 1.369523689874191e-05, + "loss": 1.1843, + "step": 14186 + }, + { + "epoch": 1.19, + "learning_rate": 1.3694392246349836e-05, + "loss": 1.2364, + "step": 14187 + }, + { + "epoch": 1.19, + "learning_rate": 1.3693547563433512e-05, + "loss": 1.1548, + "step": 14188 + }, + { + "epoch": 1.19, + "learning_rate": 1.3692702849999915e-05, + "loss": 1.1845, + "step": 14189 + }, + { + "epoch": 1.19, + "learning_rate": 1.3691858106056027e-05, + "loss": 1.212, + "step": 14190 + }, + { + "epoch": 1.19, + "learning_rate": 1.3691013331608822e-05, + "loss": 1.1724, + "step": 14191 + }, + { + "epoch": 1.19, + "learning_rate": 1.3690168526665287e-05, + "loss": 1.188, + "step": 14192 + }, + { + "epoch": 1.19, + "learning_rate": 1.3689323691232396e-05, + "loss": 1.2074, + "step": 14193 + }, + { + "epoch": 1.2, + "learning_rate": 1.3688478825317136e-05, + "loss": 1.1957, + "step": 14194 + }, + { + "epoch": 1.2, + "learning_rate": 1.368763392892648e-05, + "loss": 1.2166, + "step": 14195 + }, + { + "epoch": 1.2, + "learning_rate": 1.3686789002067412e-05, + "loss": 1.2282, + "step": 14196 + }, + { + "epoch": 1.2, + "learning_rate": 1.3685944044746915e-05, + "loss": 1.0719, + "step": 14197 + }, + { + "epoch": 1.2, + "learning_rate": 1.3685099056971969e-05, + "loss": 1.2665, + "step": 14198 + }, + { + "epoch": 1.2, + "learning_rate": 1.3684254038749551e-05, + "loss": 1.1182, + "step": 14199 + }, + { + "epoch": 1.2, + "learning_rate": 1.3683408990086653e-05, + "loss": 1.134, + "step": 14200 + }, + { + "epoch": 1.2, + "learning_rate": 1.3682563910990249e-05, + "loss": 1.1804, + "step": 14201 + }, + { + "epoch": 1.2, + "learning_rate": 1.3681718801467322e-05, + "loss": 1.1216, + "step": 14202 + }, + { + "epoch": 1.2, + "learning_rate": 1.3680873661524855e-05, + "loss": 1.267, + "step": 14203 + }, + { + "epoch": 1.2, + "learning_rate": 1.3680028491169835e-05, + "loss": 1.2489, + "step": 14204 + }, + { + "epoch": 1.2, + "learning_rate": 1.3679183290409242e-05, + "loss": 1.1618, + "step": 14205 + }, + { + "epoch": 1.2, + "learning_rate": 1.3678338059250056e-05, + "loss": 1.2298, + "step": 14206 + }, + { + "epoch": 1.2, + "learning_rate": 1.3677492797699265e-05, + "loss": 1.1987, + "step": 14207 + }, + { + "epoch": 1.2, + "learning_rate": 1.3676647505763851e-05, + "loss": 1.1892, + "step": 14208 + }, + { + "epoch": 1.2, + "learning_rate": 1.3675802183450801e-05, + "loss": 1.2259, + "step": 14209 + }, + { + "epoch": 1.2, + "learning_rate": 1.3674956830767094e-05, + "loss": 1.2176, + "step": 14210 + }, + { + "epoch": 1.2, + "learning_rate": 1.367411144771972e-05, + "loss": 1.2327, + "step": 14211 + }, + { + "epoch": 1.2, + "learning_rate": 1.367326603431566e-05, + "loss": 1.229, + "step": 14212 + }, + { + "epoch": 1.2, + "learning_rate": 1.36724205905619e-05, + "loss": 1.2258, + "step": 14213 + }, + { + "epoch": 1.2, + "learning_rate": 1.3671575116465427e-05, + "loss": 1.2187, + "step": 14214 + }, + { + "epoch": 1.2, + "learning_rate": 1.3670729612033226e-05, + "loss": 1.1708, + "step": 14215 + }, + { + "epoch": 1.2, + "learning_rate": 1.3669884077272282e-05, + "loss": 1.1987, + "step": 14216 + }, + { + "epoch": 1.2, + "learning_rate": 1.366903851218958e-05, + "loss": 1.2248, + "step": 14217 + }, + { + "epoch": 1.2, + "learning_rate": 1.3668192916792106e-05, + "loss": 1.1676, + "step": 14218 + }, + { + "epoch": 1.2, + "learning_rate": 1.3667347291086852e-05, + "loss": 1.2207, + "step": 14219 + }, + { + "epoch": 1.2, + "learning_rate": 1.36665016350808e-05, + "loss": 1.1075, + "step": 14220 + }, + { + "epoch": 1.2, + "learning_rate": 1.3665655948780934e-05, + "loss": 1.2758, + "step": 14221 + }, + { + "epoch": 1.2, + "learning_rate": 1.366481023219425e-05, + "loss": 1.1483, + "step": 14222 + }, + { + "epoch": 1.2, + "learning_rate": 1.3663964485327727e-05, + "loss": 1.2351, + "step": 14223 + }, + { + "epoch": 1.2, + "learning_rate": 1.366311870818836e-05, + "loss": 1.217, + "step": 14224 + }, + { + "epoch": 1.2, + "learning_rate": 1.3662272900783134e-05, + "loss": 1.1869, + "step": 14225 + }, + { + "epoch": 1.2, + "learning_rate": 1.3661427063119037e-05, + "loss": 1.2683, + "step": 14226 + }, + { + "epoch": 1.2, + "learning_rate": 1.3660581195203055e-05, + "loss": 1.201, + "step": 14227 + }, + { + "epoch": 1.2, + "learning_rate": 1.3659735297042183e-05, + "loss": 1.1692, + "step": 14228 + }, + { + "epoch": 1.2, + "learning_rate": 1.3658889368643405e-05, + "loss": 1.223, + "step": 14229 + }, + { + "epoch": 1.2, + "learning_rate": 1.3658043410013714e-05, + "loss": 1.1772, + "step": 14230 + }, + { + "epoch": 1.2, + "learning_rate": 1.3657197421160096e-05, + "loss": 1.2299, + "step": 14231 + }, + { + "epoch": 1.2, + "learning_rate": 1.365635140208954e-05, + "loss": 1.3495, + "step": 14232 + }, + { + "epoch": 1.2, + "learning_rate": 1.3655505352809042e-05, + "loss": 1.2058, + "step": 14233 + }, + { + "epoch": 1.2, + "learning_rate": 1.3654659273325588e-05, + "loss": 1.1928, + "step": 14234 + }, + { + "epoch": 1.2, + "learning_rate": 1.3653813163646168e-05, + "loss": 1.1689, + "step": 14235 + }, + { + "epoch": 1.2, + "learning_rate": 1.3652967023777776e-05, + "loss": 1.2027, + "step": 14236 + }, + { + "epoch": 1.2, + "learning_rate": 1.3652120853727402e-05, + "loss": 1.2127, + "step": 14237 + }, + { + "epoch": 1.2, + "learning_rate": 1.3651274653502036e-05, + "loss": 1.2814, + "step": 14238 + }, + { + "epoch": 1.2, + "learning_rate": 1.3650428423108668e-05, + "loss": 1.1673, + "step": 14239 + }, + { + "epoch": 1.2, + "learning_rate": 1.3649582162554298e-05, + "loss": 1.262, + "step": 14240 + }, + { + "epoch": 1.2, + "learning_rate": 1.3648735871845906e-05, + "loss": 1.2763, + "step": 14241 + }, + { + "epoch": 1.2, + "learning_rate": 1.3647889550990492e-05, + "loss": 1.1676, + "step": 14242 + }, + { + "epoch": 1.2, + "learning_rate": 1.3647043199995049e-05, + "loss": 1.2097, + "step": 14243 + }, + { + "epoch": 1.2, + "learning_rate": 1.3646196818866568e-05, + "loss": 1.14, + "step": 14244 + }, + { + "epoch": 1.2, + "learning_rate": 1.3645350407612039e-05, + "loss": 1.2394, + "step": 14245 + }, + { + "epoch": 1.2, + "learning_rate": 1.3644503966238458e-05, + "loss": 1.2223, + "step": 14246 + }, + { + "epoch": 1.2, + "learning_rate": 1.3643657494752822e-05, + "loss": 1.1723, + "step": 14247 + }, + { + "epoch": 1.2, + "learning_rate": 1.364281099316212e-05, + "loss": 1.2526, + "step": 14248 + }, + { + "epoch": 1.2, + "learning_rate": 1.3641964461473347e-05, + "loss": 1.2162, + "step": 14249 + }, + { + "epoch": 1.2, + "learning_rate": 1.36411178996935e-05, + "loss": 1.2513, + "step": 14250 + }, + { + "epoch": 1.2, + "learning_rate": 1.364027130782957e-05, + "loss": 1.2014, + "step": 14251 + }, + { + "epoch": 1.2, + "learning_rate": 1.3639424685888555e-05, + "loss": 1.1276, + "step": 14252 + }, + { + "epoch": 1.2, + "learning_rate": 1.3638578033877447e-05, + "loss": 1.26, + "step": 14253 + }, + { + "epoch": 1.2, + "learning_rate": 1.3637731351803242e-05, + "loss": 1.1389, + "step": 14254 + }, + { + "epoch": 1.2, + "learning_rate": 1.363688463967294e-05, + "loss": 1.1229, + "step": 14255 + }, + { + "epoch": 1.2, + "learning_rate": 1.363603789749353e-05, + "loss": 1.1485, + "step": 14256 + }, + { + "epoch": 1.2, + "learning_rate": 1.3635191125272012e-05, + "loss": 1.163, + "step": 14257 + }, + { + "epoch": 1.2, + "learning_rate": 1.3634344323015382e-05, + "loss": 1.1247, + "step": 14258 + }, + { + "epoch": 1.2, + "learning_rate": 1.3633497490730636e-05, + "loss": 1.1285, + "step": 14259 + }, + { + "epoch": 1.2, + "learning_rate": 1.363265062842477e-05, + "loss": 1.1674, + "step": 14260 + }, + { + "epoch": 1.2, + "learning_rate": 1.3631803736104781e-05, + "loss": 1.1248, + "step": 14261 + }, + { + "epoch": 1.2, + "learning_rate": 1.3630956813777671e-05, + "loss": 1.122, + "step": 14262 + }, + { + "epoch": 1.2, + "learning_rate": 1.3630109861450432e-05, + "loss": 1.2147, + "step": 14263 + }, + { + "epoch": 1.2, + "learning_rate": 1.3629262879130061e-05, + "loss": 1.1067, + "step": 14264 + }, + { + "epoch": 1.2, + "learning_rate": 1.3628415866823561e-05, + "loss": 1.2122, + "step": 14265 + }, + { + "epoch": 1.2, + "learning_rate": 1.362756882453793e-05, + "loss": 1.1598, + "step": 14266 + }, + { + "epoch": 1.2, + "learning_rate": 1.3626721752280159e-05, + "loss": 1.2122, + "step": 14267 + }, + { + "epoch": 1.2, + "learning_rate": 1.3625874650057256e-05, + "loss": 1.1826, + "step": 14268 + }, + { + "epoch": 1.2, + "learning_rate": 1.3625027517876216e-05, + "loss": 1.1579, + "step": 14269 + }, + { + "epoch": 1.2, + "learning_rate": 1.3624180355744037e-05, + "loss": 1.2062, + "step": 14270 + }, + { + "epoch": 1.2, + "learning_rate": 1.3623333163667718e-05, + "loss": 1.1885, + "step": 14271 + }, + { + "epoch": 1.2, + "learning_rate": 1.3622485941654264e-05, + "loss": 1.1189, + "step": 14272 + }, + { + "epoch": 1.2, + "learning_rate": 1.3621638689710673e-05, + "loss": 1.2725, + "step": 14273 + }, + { + "epoch": 1.2, + "learning_rate": 1.362079140784394e-05, + "loss": 1.2022, + "step": 14274 + }, + { + "epoch": 1.2, + "learning_rate": 1.3619944096061073e-05, + "loss": 1.152, + "step": 14275 + }, + { + "epoch": 1.2, + "learning_rate": 1.361909675436907e-05, + "loss": 1.197, + "step": 14276 + }, + { + "epoch": 1.2, + "learning_rate": 1.3618249382774928e-05, + "loss": 1.129, + "step": 14277 + }, + { + "epoch": 1.2, + "learning_rate": 1.3617401981285654e-05, + "loss": 1.1617, + "step": 14278 + }, + { + "epoch": 1.2, + "learning_rate": 1.3616554549908245e-05, + "loss": 1.2219, + "step": 14279 + }, + { + "epoch": 1.2, + "learning_rate": 1.3615707088649708e-05, + "loss": 1.165, + "step": 14280 + }, + { + "epoch": 1.2, + "learning_rate": 1.3614859597517041e-05, + "loss": 1.1162, + "step": 14281 + }, + { + "epoch": 1.2, + "learning_rate": 1.3614012076517245e-05, + "loss": 1.1472, + "step": 14282 + }, + { + "epoch": 1.2, + "learning_rate": 1.3613164525657326e-05, + "loss": 1.2476, + "step": 14283 + }, + { + "epoch": 1.2, + "learning_rate": 1.3612316944944283e-05, + "loss": 1.1465, + "step": 14284 + }, + { + "epoch": 1.2, + "learning_rate": 1.3611469334385123e-05, + "loss": 1.2105, + "step": 14285 + }, + { + "epoch": 1.2, + "learning_rate": 1.3610621693986849e-05, + "loss": 1.1417, + "step": 14286 + }, + { + "epoch": 1.2, + "learning_rate": 1.3609774023756464e-05, + "loss": 1.1956, + "step": 14287 + }, + { + "epoch": 1.2, + "learning_rate": 1.3608926323700966e-05, + "loss": 1.2264, + "step": 14288 + }, + { + "epoch": 1.2, + "learning_rate": 1.3608078593827368e-05, + "loss": 1.1143, + "step": 14289 + }, + { + "epoch": 1.2, + "learning_rate": 1.3607230834142669e-05, + "loss": 1.1732, + "step": 14290 + }, + { + "epoch": 1.2, + "learning_rate": 1.3606383044653875e-05, + "loss": 1.2218, + "step": 14291 + }, + { + "epoch": 1.2, + "learning_rate": 1.3605535225367987e-05, + "loss": 1.2409, + "step": 14292 + }, + { + "epoch": 1.2, + "learning_rate": 1.3604687376292013e-05, + "loss": 1.2556, + "step": 14293 + }, + { + "epoch": 1.2, + "learning_rate": 1.3603839497432963e-05, + "loss": 1.2089, + "step": 14294 + }, + { + "epoch": 1.2, + "learning_rate": 1.3602991588797835e-05, + "loss": 1.2844, + "step": 14295 + }, + { + "epoch": 1.2, + "learning_rate": 1.3602143650393633e-05, + "loss": 1.1291, + "step": 14296 + }, + { + "epoch": 1.2, + "learning_rate": 1.3601295682227376e-05, + "loss": 1.142, + "step": 14297 + }, + { + "epoch": 1.2, + "learning_rate": 1.3600447684306056e-05, + "loss": 1.1917, + "step": 14298 + }, + { + "epoch": 1.2, + "learning_rate": 1.3599599656636683e-05, + "loss": 1.1042, + "step": 14299 + }, + { + "epoch": 1.2, + "learning_rate": 1.359875159922627e-05, + "loss": 1.2138, + "step": 14300 + }, + { + "epoch": 1.2, + "learning_rate": 1.359790351208182e-05, + "loss": 1.2783, + "step": 14301 + }, + { + "epoch": 1.2, + "learning_rate": 1.3597055395210336e-05, + "loss": 1.1359, + "step": 14302 + }, + { + "epoch": 1.2, + "learning_rate": 1.3596207248618832e-05, + "loss": 1.1894, + "step": 14303 + }, + { + "epoch": 1.2, + "learning_rate": 1.3595359072314311e-05, + "loss": 1.2774, + "step": 14304 + }, + { + "epoch": 1.2, + "learning_rate": 1.3594510866303784e-05, + "loss": 1.2061, + "step": 14305 + }, + { + "epoch": 1.2, + "learning_rate": 1.3593662630594258e-05, + "loss": 1.1401, + "step": 14306 + }, + { + "epoch": 1.2, + "learning_rate": 1.3592814365192739e-05, + "loss": 1.2147, + "step": 14307 + }, + { + "epoch": 1.2, + "learning_rate": 1.3591966070106239e-05, + "loss": 1.2862, + "step": 14308 + }, + { + "epoch": 1.2, + "learning_rate": 1.3591117745341766e-05, + "loss": 1.2659, + "step": 14309 + }, + { + "epoch": 1.2, + "learning_rate": 1.359026939090633e-05, + "loss": 1.15, + "step": 14310 + }, + { + "epoch": 1.2, + "learning_rate": 1.3589421006806938e-05, + "loss": 1.1396, + "step": 14311 + }, + { + "epoch": 1.2, + "learning_rate": 1.3588572593050602e-05, + "loss": 1.2841, + "step": 14312 + }, + { + "epoch": 1.21, + "learning_rate": 1.3587724149644328e-05, + "loss": 1.2057, + "step": 14313 + }, + { + "epoch": 1.21, + "learning_rate": 1.3586875676595131e-05, + "loss": 1.2112, + "step": 14314 + }, + { + "epoch": 1.21, + "learning_rate": 1.3586027173910022e-05, + "loss": 1.1808, + "step": 14315 + }, + { + "epoch": 1.21, + "learning_rate": 1.3585178641596004e-05, + "loss": 1.194, + "step": 14316 + }, + { + "epoch": 1.21, + "learning_rate": 1.3584330079660095e-05, + "loss": 1.2335, + "step": 14317 + }, + { + "epoch": 1.21, + "learning_rate": 1.3583481488109303e-05, + "loss": 1.2508, + "step": 14318 + }, + { + "epoch": 1.21, + "learning_rate": 1.3582632866950641e-05, + "loss": 1.1354, + "step": 14319 + }, + { + "epoch": 1.21, + "learning_rate": 1.3581784216191118e-05, + "loss": 1.2428, + "step": 14320 + }, + { + "epoch": 1.21, + "learning_rate": 1.3580935535837747e-05, + "loss": 1.1869, + "step": 14321 + }, + { + "epoch": 1.21, + "learning_rate": 1.3580086825897543e-05, + "loss": 1.2022, + "step": 14322 + }, + { + "epoch": 1.21, + "learning_rate": 1.3579238086377516e-05, + "loss": 1.1743, + "step": 14323 + }, + { + "epoch": 1.21, + "learning_rate": 1.3578389317284675e-05, + "loss": 1.2528, + "step": 14324 + }, + { + "epoch": 1.21, + "learning_rate": 1.3577540518626038e-05, + "loss": 1.1427, + "step": 14325 + }, + { + "epoch": 1.21, + "learning_rate": 1.3576691690408619e-05, + "loss": 1.1541, + "step": 14326 + }, + { + "epoch": 1.21, + "learning_rate": 1.3575842832639426e-05, + "loss": 1.129, + "step": 14327 + }, + { + "epoch": 1.21, + "learning_rate": 1.3574993945325475e-05, + "loss": 1.2088, + "step": 14328 + }, + { + "epoch": 1.21, + "learning_rate": 1.3574145028473779e-05, + "loss": 1.2113, + "step": 14329 + }, + { + "epoch": 1.21, + "learning_rate": 1.3573296082091357e-05, + "loss": 1.1987, + "step": 14330 + }, + { + "epoch": 1.21, + "learning_rate": 1.3572447106185217e-05, + "loss": 1.1978, + "step": 14331 + }, + { + "epoch": 1.21, + "learning_rate": 1.3571598100762374e-05, + "loss": 1.1727, + "step": 14332 + }, + { + "epoch": 1.21, + "learning_rate": 1.3570749065829845e-05, + "loss": 1.139, + "step": 14333 + }, + { + "epoch": 1.21, + "learning_rate": 1.3569900001394642e-05, + "loss": 1.1021, + "step": 14334 + }, + { + "epoch": 1.21, + "learning_rate": 1.3569050907463788e-05, + "loss": 1.2082, + "step": 14335 + }, + { + "epoch": 1.21, + "learning_rate": 1.356820178404429e-05, + "loss": 1.0645, + "step": 14336 + }, + { + "epoch": 1.21, + "learning_rate": 1.3567352631143167e-05, + "loss": 1.2211, + "step": 14337 + }, + { + "epoch": 1.21, + "learning_rate": 1.3566503448767437e-05, + "loss": 1.2529, + "step": 14338 + }, + { + "epoch": 1.21, + "learning_rate": 1.3565654236924113e-05, + "loss": 1.161, + "step": 14339 + }, + { + "epoch": 1.21, + "learning_rate": 1.356480499562021e-05, + "loss": 1.2442, + "step": 14340 + }, + { + "epoch": 1.21, + "learning_rate": 1.3563955724862752e-05, + "loss": 1.1259, + "step": 14341 + }, + { + "epoch": 1.21, + "learning_rate": 1.3563106424658747e-05, + "loss": 1.2534, + "step": 14342 + }, + { + "epoch": 1.21, + "learning_rate": 1.3562257095015219e-05, + "loss": 1.1674, + "step": 14343 + }, + { + "epoch": 1.21, + "learning_rate": 1.3561407735939183e-05, + "loss": 1.2008, + "step": 14344 + }, + { + "epoch": 1.21, + "learning_rate": 1.3560558347437653e-05, + "loss": 1.1481, + "step": 14345 + }, + { + "epoch": 1.21, + "learning_rate": 1.3559708929517653e-05, + "loss": 1.1964, + "step": 14346 + }, + { + "epoch": 1.21, + "learning_rate": 1.3558859482186197e-05, + "loss": 1.1579, + "step": 14347 + }, + { + "epoch": 1.21, + "learning_rate": 1.3558010005450307e-05, + "loss": 1.1492, + "step": 14348 + }, + { + "epoch": 1.21, + "learning_rate": 1.3557160499317e-05, + "loss": 1.1749, + "step": 14349 + }, + { + "epoch": 1.21, + "learning_rate": 1.3556310963793291e-05, + "loss": 1.2909, + "step": 14350 + }, + { + "epoch": 1.21, + "learning_rate": 1.3555461398886207e-05, + "loss": 1.2402, + "step": 14351 + }, + { + "epoch": 1.21, + "learning_rate": 1.355461180460276e-05, + "loss": 1.1836, + "step": 14352 + }, + { + "epoch": 1.21, + "learning_rate": 1.3553762180949975e-05, + "loss": 1.1381, + "step": 14353 + }, + { + "epoch": 1.21, + "learning_rate": 1.3552912527934867e-05, + "loss": 1.201, + "step": 14354 + }, + { + "epoch": 1.21, + "learning_rate": 1.355206284556446e-05, + "loss": 1.2226, + "step": 14355 + }, + { + "epoch": 1.21, + "learning_rate": 1.3551213133845773e-05, + "loss": 1.2149, + "step": 14356 + }, + { + "epoch": 1.21, + "learning_rate": 1.3550363392785826e-05, + "loss": 1.099, + "step": 14357 + }, + { + "epoch": 1.21, + "learning_rate": 1.354951362239164e-05, + "loss": 1.2058, + "step": 14358 + }, + { + "epoch": 1.21, + "learning_rate": 1.3548663822670238e-05, + "loss": 1.1973, + "step": 14359 + }, + { + "epoch": 1.21, + "learning_rate": 1.3547813993628638e-05, + "loss": 1.231, + "step": 14360 + }, + { + "epoch": 1.21, + "learning_rate": 1.3546964135273863e-05, + "loss": 1.2418, + "step": 14361 + }, + { + "epoch": 1.21, + "learning_rate": 1.3546114247612939e-05, + "loss": 1.2506, + "step": 14362 + }, + { + "epoch": 1.21, + "learning_rate": 1.3545264330652881e-05, + "loss": 1.216, + "step": 14363 + }, + { + "epoch": 1.21, + "learning_rate": 1.3544414384400714e-05, + "loss": 1.3073, + "step": 14364 + }, + { + "epoch": 1.21, + "learning_rate": 1.3543564408863463e-05, + "loss": 1.1465, + "step": 14365 + }, + { + "epoch": 1.21, + "learning_rate": 1.3542714404048148e-05, + "loss": 1.1687, + "step": 14366 + }, + { + "epoch": 1.21, + "learning_rate": 1.3541864369961792e-05, + "loss": 1.2061, + "step": 14367 + }, + { + "epoch": 1.21, + "learning_rate": 1.354101430661142e-05, + "loss": 1.1904, + "step": 14368 + }, + { + "epoch": 1.21, + "learning_rate": 1.3540164214004054e-05, + "loss": 1.129, + "step": 14369 + }, + { + "epoch": 1.21, + "learning_rate": 1.353931409214672e-05, + "loss": 1.1879, + "step": 14370 + }, + { + "epoch": 1.21, + "learning_rate": 1.3538463941046436e-05, + "loss": 1.1469, + "step": 14371 + }, + { + "epoch": 1.21, + "learning_rate": 1.3537613760710235e-05, + "loss": 1.2962, + "step": 14372 + }, + { + "epoch": 1.21, + "learning_rate": 1.3536763551145136e-05, + "loss": 1.2199, + "step": 14373 + }, + { + "epoch": 1.21, + "learning_rate": 1.3535913312358162e-05, + "loss": 1.1445, + "step": 14374 + }, + { + "epoch": 1.21, + "learning_rate": 1.3535063044356343e-05, + "loss": 1.1735, + "step": 14375 + }, + { + "epoch": 1.21, + "learning_rate": 1.3534212747146702e-05, + "loss": 1.234, + "step": 14376 + }, + { + "epoch": 1.21, + "learning_rate": 1.3533362420736262e-05, + "loss": 1.1331, + "step": 14377 + }, + { + "epoch": 1.21, + "learning_rate": 1.353251206513205e-05, + "loss": 1.1907, + "step": 14378 + }, + { + "epoch": 1.21, + "learning_rate": 1.3531661680341098e-05, + "loss": 1.2894, + "step": 14379 + }, + { + "epoch": 1.21, + "learning_rate": 1.3530811266370423e-05, + "loss": 1.2378, + "step": 14380 + }, + { + "epoch": 1.21, + "learning_rate": 1.3529960823227057e-05, + "loss": 1.1824, + "step": 14381 + }, + { + "epoch": 1.21, + "learning_rate": 1.3529110350918021e-05, + "loss": 1.2093, + "step": 14382 + }, + { + "epoch": 1.21, + "learning_rate": 1.3528259849450352e-05, + "loss": 1.243, + "step": 14383 + }, + { + "epoch": 1.21, + "learning_rate": 1.3527409318831067e-05, + "loss": 1.2673, + "step": 14384 + }, + { + "epoch": 1.21, + "learning_rate": 1.3526558759067197e-05, + "loss": 1.1562, + "step": 14385 + }, + { + "epoch": 1.21, + "learning_rate": 1.352570817016577e-05, + "loss": 1.2311, + "step": 14386 + }, + { + "epoch": 1.21, + "learning_rate": 1.3524857552133816e-05, + "loss": 1.1847, + "step": 14387 + }, + { + "epoch": 1.21, + "learning_rate": 1.3524006904978359e-05, + "loss": 1.1714, + "step": 14388 + }, + { + "epoch": 1.21, + "learning_rate": 1.3523156228706432e-05, + "loss": 1.2031, + "step": 14389 + }, + { + "epoch": 1.21, + "learning_rate": 1.3522305523325057e-05, + "loss": 1.1232, + "step": 14390 + }, + { + "epoch": 1.21, + "learning_rate": 1.3521454788841269e-05, + "loss": 1.177, + "step": 14391 + }, + { + "epoch": 1.21, + "learning_rate": 1.3520604025262092e-05, + "loss": 1.239, + "step": 14392 + }, + { + "epoch": 1.21, + "learning_rate": 1.351975323259456e-05, + "loss": 1.1332, + "step": 14393 + }, + { + "epoch": 1.21, + "learning_rate": 1.35189024108457e-05, + "loss": 1.2329, + "step": 14394 + }, + { + "epoch": 1.21, + "learning_rate": 1.3518051560022541e-05, + "loss": 1.17, + "step": 14395 + }, + { + "epoch": 1.21, + "learning_rate": 1.3517200680132112e-05, + "loss": 1.2303, + "step": 14396 + }, + { + "epoch": 1.21, + "learning_rate": 1.3516349771181449e-05, + "loss": 1.1603, + "step": 14397 + }, + { + "epoch": 1.21, + "learning_rate": 1.3515498833177578e-05, + "loss": 1.1914, + "step": 14398 + }, + { + "epoch": 1.21, + "learning_rate": 1.3514647866127531e-05, + "loss": 1.221, + "step": 14399 + }, + { + "epoch": 1.21, + "learning_rate": 1.3513796870038337e-05, + "loss": 1.2085, + "step": 14400 + }, + { + "epoch": 1.21, + "learning_rate": 1.351294584491703e-05, + "loss": 1.2839, + "step": 14401 + }, + { + "epoch": 1.21, + "learning_rate": 1.351209479077064e-05, + "loss": 1.2427, + "step": 14402 + }, + { + "epoch": 1.21, + "learning_rate": 1.35112437076062e-05, + "loss": 1.1517, + "step": 14403 + }, + { + "epoch": 1.21, + "learning_rate": 1.3510392595430736e-05, + "loss": 1.2208, + "step": 14404 + }, + { + "epoch": 1.21, + "learning_rate": 1.3509541454251292e-05, + "loss": 1.1791, + "step": 14405 + }, + { + "epoch": 1.21, + "learning_rate": 1.3508690284074889e-05, + "loss": 1.1609, + "step": 14406 + }, + { + "epoch": 1.21, + "learning_rate": 1.3507839084908563e-05, + "loss": 1.1697, + "step": 14407 + }, + { + "epoch": 1.21, + "learning_rate": 1.3506987856759348e-05, + "loss": 1.1545, + "step": 14408 + }, + { + "epoch": 1.21, + "learning_rate": 1.3506136599634279e-05, + "loss": 1.1408, + "step": 14409 + }, + { + "epoch": 1.21, + "learning_rate": 1.3505285313540386e-05, + "loss": 1.2037, + "step": 14410 + }, + { + "epoch": 1.21, + "learning_rate": 1.3504433998484702e-05, + "loss": 1.1812, + "step": 14411 + }, + { + "epoch": 1.21, + "learning_rate": 1.3503582654474267e-05, + "loss": 1.1323, + "step": 14412 + }, + { + "epoch": 1.21, + "learning_rate": 1.350273128151611e-05, + "loss": 1.1894, + "step": 14413 + }, + { + "epoch": 1.21, + "learning_rate": 1.3501879879617264e-05, + "loss": 1.1928, + "step": 14414 + }, + { + "epoch": 1.21, + "learning_rate": 1.3501028448784767e-05, + "loss": 1.1222, + "step": 14415 + }, + { + "epoch": 1.21, + "learning_rate": 1.3500176989025655e-05, + "loss": 1.1663, + "step": 14416 + }, + { + "epoch": 1.21, + "learning_rate": 1.3499325500346956e-05, + "loss": 1.1143, + "step": 14417 + }, + { + "epoch": 1.21, + "learning_rate": 1.349847398275571e-05, + "loss": 1.1831, + "step": 14418 + }, + { + "epoch": 1.21, + "learning_rate": 1.3497622436258957e-05, + "loss": 1.1891, + "step": 14419 + }, + { + "epoch": 1.21, + "learning_rate": 1.3496770860863724e-05, + "loss": 1.2132, + "step": 14420 + }, + { + "epoch": 1.21, + "learning_rate": 1.349591925657705e-05, + "loss": 1.1919, + "step": 14421 + }, + { + "epoch": 1.21, + "learning_rate": 1.3495067623405975e-05, + "loss": 1.2642, + "step": 14422 + }, + { + "epoch": 1.21, + "learning_rate": 1.3494215961357533e-05, + "loss": 1.2117, + "step": 14423 + }, + { + "epoch": 1.21, + "learning_rate": 1.3493364270438758e-05, + "loss": 1.187, + "step": 14424 + }, + { + "epoch": 1.21, + "learning_rate": 1.3492512550656691e-05, + "loss": 1.1554, + "step": 14425 + }, + { + "epoch": 1.21, + "learning_rate": 1.349166080201837e-05, + "loss": 1.2072, + "step": 14426 + }, + { + "epoch": 1.21, + "learning_rate": 1.3490809024530824e-05, + "loss": 1.1242, + "step": 14427 + }, + { + "epoch": 1.21, + "learning_rate": 1.34899572182011e-05, + "loss": 1.2422, + "step": 14428 + }, + { + "epoch": 1.21, + "learning_rate": 1.3489105383036232e-05, + "loss": 1.1874, + "step": 14429 + }, + { + "epoch": 1.21, + "learning_rate": 1.3488253519043261e-05, + "loss": 1.2067, + "step": 14430 + }, + { + "epoch": 1.21, + "learning_rate": 1.3487401626229222e-05, + "loss": 1.2282, + "step": 14431 + }, + { + "epoch": 1.22, + "learning_rate": 1.3486549704601153e-05, + "loss": 1.1327, + "step": 14432 + }, + { + "epoch": 1.22, + "learning_rate": 1.3485697754166097e-05, + "loss": 1.2146, + "step": 14433 + }, + { + "epoch": 1.22, + "learning_rate": 1.3484845774931085e-05, + "loss": 1.24, + "step": 14434 + }, + { + "epoch": 1.22, + "learning_rate": 1.3483993766903167e-05, + "loss": 1.1925, + "step": 14435 + }, + { + "epoch": 1.22, + "learning_rate": 1.3483141730089375e-05, + "loss": 1.1687, + "step": 14436 + }, + { + "epoch": 1.22, + "learning_rate": 1.3482289664496755e-05, + "loss": 1.2152, + "step": 14437 + }, + { + "epoch": 1.22, + "learning_rate": 1.3481437570132341e-05, + "loss": 1.1515, + "step": 14438 + }, + { + "epoch": 1.22, + "learning_rate": 1.3480585447003174e-05, + "loss": 1.1779, + "step": 14439 + }, + { + "epoch": 1.22, + "learning_rate": 1.3479733295116297e-05, + "loss": 1.1458, + "step": 14440 + }, + { + "epoch": 1.22, + "learning_rate": 1.3478881114478753e-05, + "loss": 1.2307, + "step": 14441 + }, + { + "epoch": 1.22, + "learning_rate": 1.3478028905097578e-05, + "loss": 1.1904, + "step": 14442 + }, + { + "epoch": 1.22, + "learning_rate": 1.3477176666979812e-05, + "loss": 1.2319, + "step": 14443 + }, + { + "epoch": 1.22, + "learning_rate": 1.3476324400132504e-05, + "loss": 1.156, + "step": 14444 + }, + { + "epoch": 1.22, + "learning_rate": 1.347547210456269e-05, + "loss": 1.2018, + "step": 14445 + }, + { + "epoch": 1.22, + "learning_rate": 1.347461978027741e-05, + "loss": 1.2749, + "step": 14446 + }, + { + "epoch": 1.22, + "learning_rate": 1.3473767427283713e-05, + "loss": 1.2054, + "step": 14447 + }, + { + "epoch": 1.22, + "learning_rate": 1.3472915045588635e-05, + "loss": 1.1445, + "step": 14448 + }, + { + "epoch": 1.22, + "learning_rate": 1.3472062635199223e-05, + "loss": 1.2327, + "step": 14449 + }, + { + "epoch": 1.22, + "learning_rate": 1.3471210196122517e-05, + "loss": 1.2017, + "step": 14450 + }, + { + "epoch": 1.22, + "learning_rate": 1.3470357728365566e-05, + "loss": 1.2186, + "step": 14451 + }, + { + "epoch": 1.22, + "learning_rate": 1.3469505231935403e-05, + "loss": 1.2222, + "step": 14452 + }, + { + "epoch": 1.22, + "learning_rate": 1.346865270683908e-05, + "loss": 1.2335, + "step": 14453 + }, + { + "epoch": 1.22, + "learning_rate": 1.346780015308364e-05, + "loss": 1.2312, + "step": 14454 + }, + { + "epoch": 1.22, + "learning_rate": 1.3466947570676124e-05, + "loss": 1.1686, + "step": 14455 + }, + { + "epoch": 1.22, + "learning_rate": 1.3466094959623575e-05, + "loss": 1.266, + "step": 14456 + }, + { + "epoch": 1.22, + "learning_rate": 1.3465242319933043e-05, + "loss": 1.175, + "step": 14457 + }, + { + "epoch": 1.22, + "learning_rate": 1.3464389651611566e-05, + "loss": 1.1663, + "step": 14458 + }, + { + "epoch": 1.22, + "learning_rate": 1.3463536954666197e-05, + "loss": 1.2165, + "step": 14459 + }, + { + "epoch": 1.22, + "learning_rate": 1.3462684229103978e-05, + "loss": 1.1746, + "step": 14460 + }, + { + "epoch": 1.22, + "learning_rate": 1.346183147493195e-05, + "loss": 1.2398, + "step": 14461 + }, + { + "epoch": 1.22, + "learning_rate": 1.3460978692157166e-05, + "loss": 1.3322, + "step": 14462 + }, + { + "epoch": 1.22, + "learning_rate": 1.3460125880786665e-05, + "loss": 1.2174, + "step": 14463 + }, + { + "epoch": 1.22, + "learning_rate": 1.34592730408275e-05, + "loss": 1.1423, + "step": 14464 + }, + { + "epoch": 1.22, + "learning_rate": 1.3458420172286711e-05, + "loss": 1.1133, + "step": 14465 + }, + { + "epoch": 1.22, + "learning_rate": 1.3457567275171351e-05, + "loss": 1.1234, + "step": 14466 + }, + { + "epoch": 1.22, + "learning_rate": 1.345671434948846e-05, + "loss": 1.2398, + "step": 14467 + }, + { + "epoch": 1.22, + "learning_rate": 1.3455861395245088e-05, + "loss": 1.2153, + "step": 14468 + }, + { + "epoch": 1.22, + "learning_rate": 1.345500841244829e-05, + "loss": 1.1809, + "step": 14469 + }, + { + "epoch": 1.22, + "learning_rate": 1.3454155401105103e-05, + "loss": 1.3256, + "step": 14470 + }, + { + "epoch": 1.22, + "learning_rate": 1.3453302361222573e-05, + "loss": 1.1952, + "step": 14471 + }, + { + "epoch": 1.22, + "learning_rate": 1.3452449292807761e-05, + "loss": 1.1292, + "step": 14472 + }, + { + "epoch": 1.22, + "learning_rate": 1.3451596195867705e-05, + "loss": 1.1948, + "step": 14473 + }, + { + "epoch": 1.22, + "learning_rate": 1.3450743070409455e-05, + "loss": 1.1762, + "step": 14474 + }, + { + "epoch": 1.22, + "learning_rate": 1.3449889916440065e-05, + "loss": 1.1419, + "step": 14475 + }, + { + "epoch": 1.22, + "learning_rate": 1.3449036733966581e-05, + "loss": 1.1946, + "step": 14476 + }, + { + "epoch": 1.22, + "learning_rate": 1.3448183522996048e-05, + "loss": 1.1401, + "step": 14477 + }, + { + "epoch": 1.22, + "learning_rate": 1.344733028353552e-05, + "loss": 1.1889, + "step": 14478 + }, + { + "epoch": 1.22, + "learning_rate": 1.3446477015592048e-05, + "loss": 1.1592, + "step": 14479 + }, + { + "epoch": 1.22, + "learning_rate": 1.344562371917268e-05, + "loss": 1.3351, + "step": 14480 + }, + { + "epoch": 1.22, + "learning_rate": 1.3444770394284464e-05, + "loss": 1.1401, + "step": 14481 + }, + { + "epoch": 1.22, + "learning_rate": 1.3443917040934453e-05, + "loss": 1.1768, + "step": 14482 + }, + { + "epoch": 1.22, + "learning_rate": 1.3443063659129696e-05, + "loss": 1.2349, + "step": 14483 + }, + { + "epoch": 1.22, + "learning_rate": 1.3442210248877245e-05, + "loss": 1.1187, + "step": 14484 + }, + { + "epoch": 1.22, + "learning_rate": 1.3441356810184154e-05, + "loss": 1.1734, + "step": 14485 + }, + { + "epoch": 1.22, + "learning_rate": 1.3440503343057469e-05, + "loss": 1.2319, + "step": 14486 + }, + { + "epoch": 1.22, + "learning_rate": 1.3439649847504248e-05, + "loss": 1.2026, + "step": 14487 + }, + { + "epoch": 1.22, + "learning_rate": 1.3438796323531537e-05, + "loss": 1.2037, + "step": 14488 + }, + { + "epoch": 1.22, + "learning_rate": 1.3437942771146388e-05, + "loss": 1.2891, + "step": 14489 + }, + { + "epoch": 1.22, + "learning_rate": 1.343708919035586e-05, + "loss": 1.3165, + "step": 14490 + }, + { + "epoch": 1.22, + "learning_rate": 1.3436235581166999e-05, + "loss": 1.2333, + "step": 14491 + }, + { + "epoch": 1.22, + "learning_rate": 1.3435381943586858e-05, + "loss": 1.1179, + "step": 14492 + }, + { + "epoch": 1.22, + "learning_rate": 1.3434528277622494e-05, + "loss": 1.2191, + "step": 14493 + }, + { + "epoch": 1.22, + "learning_rate": 1.343367458328096e-05, + "loss": 1.1662, + "step": 14494 + }, + { + "epoch": 1.22, + "learning_rate": 1.3432820860569304e-05, + "loss": 1.2553, + "step": 14495 + }, + { + "epoch": 1.22, + "learning_rate": 1.3431967109494583e-05, + "loss": 1.2269, + "step": 14496 + }, + { + "epoch": 1.22, + "learning_rate": 1.3431113330063853e-05, + "loss": 1.1803, + "step": 14497 + }, + { + "epoch": 1.22, + "learning_rate": 1.3430259522284168e-05, + "loss": 1.1389, + "step": 14498 + }, + { + "epoch": 1.22, + "learning_rate": 1.342940568616258e-05, + "loss": 1.2534, + "step": 14499 + }, + { + "epoch": 1.22, + "learning_rate": 1.3428551821706144e-05, + "loss": 1.1683, + "step": 14500 + }, + { + "epoch": 1.22, + "learning_rate": 1.3427697928921916e-05, + "loss": 1.2606, + "step": 14501 + }, + { + "epoch": 1.22, + "learning_rate": 1.3426844007816952e-05, + "loss": 1.1735, + "step": 14502 + }, + { + "epoch": 1.22, + "learning_rate": 1.3425990058398305e-05, + "loss": 1.2066, + "step": 14503 + }, + { + "epoch": 1.22, + "learning_rate": 1.3425136080673032e-05, + "loss": 1.1323, + "step": 14504 + }, + { + "epoch": 1.22, + "learning_rate": 1.3424282074648187e-05, + "loss": 1.1663, + "step": 14505 + }, + { + "epoch": 1.22, + "learning_rate": 1.3423428040330828e-05, + "loss": 1.2051, + "step": 14506 + }, + { + "epoch": 1.22, + "learning_rate": 1.342257397772801e-05, + "loss": 1.216, + "step": 14507 + }, + { + "epoch": 1.22, + "learning_rate": 1.3421719886846793e-05, + "loss": 1.2125, + "step": 14508 + }, + { + "epoch": 1.22, + "learning_rate": 1.342086576769423e-05, + "loss": 1.1645, + "step": 14509 + }, + { + "epoch": 1.22, + "learning_rate": 1.3420011620277378e-05, + "loss": 1.1453, + "step": 14510 + }, + { + "epoch": 1.22, + "learning_rate": 1.3419157444603297e-05, + "loss": 1.1949, + "step": 14511 + }, + { + "epoch": 1.22, + "learning_rate": 1.3418303240679045e-05, + "loss": 1.1432, + "step": 14512 + }, + { + "epoch": 1.22, + "learning_rate": 1.3417449008511674e-05, + "loss": 1.125, + "step": 14513 + }, + { + "epoch": 1.22, + "learning_rate": 1.3416594748108247e-05, + "loss": 1.1744, + "step": 14514 + }, + { + "epoch": 1.22, + "learning_rate": 1.341574045947582e-05, + "loss": 1.1852, + "step": 14515 + }, + { + "epoch": 1.22, + "learning_rate": 1.3414886142621455e-05, + "loss": 1.2055, + "step": 14516 + }, + { + "epoch": 1.22, + "learning_rate": 1.3414031797552205e-05, + "loss": 1.2627, + "step": 14517 + }, + { + "epoch": 1.22, + "learning_rate": 1.341317742427513e-05, + "loss": 1.2669, + "step": 14518 + }, + { + "epoch": 1.22, + "learning_rate": 1.3412323022797294e-05, + "loss": 1.2345, + "step": 14519 + }, + { + "epoch": 1.22, + "learning_rate": 1.3411468593125754e-05, + "loss": 1.1484, + "step": 14520 + }, + { + "epoch": 1.22, + "learning_rate": 1.3410614135267563e-05, + "loss": 1.1425, + "step": 14521 + }, + { + "epoch": 1.22, + "learning_rate": 1.340975964922979e-05, + "loss": 1.2058, + "step": 14522 + }, + { + "epoch": 1.22, + "learning_rate": 1.3408905135019495e-05, + "loss": 1.2659, + "step": 14523 + }, + { + "epoch": 1.22, + "learning_rate": 1.3408050592643731e-05, + "loss": 1.2683, + "step": 14524 + }, + { + "epoch": 1.22, + "learning_rate": 1.340719602210956e-05, + "loss": 1.1557, + "step": 14525 + }, + { + "epoch": 1.22, + "learning_rate": 1.3406341423424052e-05, + "loss": 1.2124, + "step": 14526 + }, + { + "epoch": 1.22, + "learning_rate": 1.3405486796594255e-05, + "loss": 1.1702, + "step": 14527 + }, + { + "epoch": 1.22, + "learning_rate": 1.3404632141627239e-05, + "loss": 1.259, + "step": 14528 + }, + { + "epoch": 1.22, + "learning_rate": 1.340377745853006e-05, + "loss": 1.2302, + "step": 14529 + }, + { + "epoch": 1.22, + "learning_rate": 1.3402922747309785e-05, + "loss": 1.2838, + "step": 14530 + }, + { + "epoch": 1.22, + "learning_rate": 1.3402068007973473e-05, + "loss": 1.1576, + "step": 14531 + }, + { + "epoch": 1.22, + "learning_rate": 1.3401213240528184e-05, + "loss": 1.1737, + "step": 14532 + }, + { + "epoch": 1.22, + "learning_rate": 1.3400358444980986e-05, + "loss": 1.1432, + "step": 14533 + }, + { + "epoch": 1.22, + "learning_rate": 1.3399503621338934e-05, + "loss": 1.1508, + "step": 14534 + }, + { + "epoch": 1.22, + "learning_rate": 1.3398648769609096e-05, + "loss": 1.2331, + "step": 14535 + }, + { + "epoch": 1.22, + "learning_rate": 1.3397793889798536e-05, + "loss": 1.1688, + "step": 14536 + }, + { + "epoch": 1.22, + "learning_rate": 1.3396938981914317e-05, + "loss": 1.1666, + "step": 14537 + }, + { + "epoch": 1.22, + "learning_rate": 1.33960840459635e-05, + "loss": 1.211, + "step": 14538 + }, + { + "epoch": 1.22, + "learning_rate": 1.3395229081953148e-05, + "loss": 1.2364, + "step": 14539 + }, + { + "epoch": 1.22, + "learning_rate": 1.3394374089890327e-05, + "loss": 1.1985, + "step": 14540 + }, + { + "epoch": 1.22, + "learning_rate": 1.3393519069782101e-05, + "loss": 1.2006, + "step": 14541 + }, + { + "epoch": 1.22, + "learning_rate": 1.3392664021635535e-05, + "loss": 1.1686, + "step": 14542 + }, + { + "epoch": 1.22, + "learning_rate": 1.3391808945457694e-05, + "loss": 1.2292, + "step": 14543 + }, + { + "epoch": 1.22, + "learning_rate": 1.3390953841255641e-05, + "loss": 1.1184, + "step": 14544 + }, + { + "epoch": 1.22, + "learning_rate": 1.3390098709036444e-05, + "loss": 1.2446, + "step": 14545 + }, + { + "epoch": 1.22, + "learning_rate": 1.3389243548807161e-05, + "loss": 1.2219, + "step": 14546 + }, + { + "epoch": 1.22, + "learning_rate": 1.3388388360574864e-05, + "loss": 1.1885, + "step": 14547 + }, + { + "epoch": 1.22, + "learning_rate": 1.3387533144346625e-05, + "loss": 1.1866, + "step": 14548 + }, + { + "epoch": 1.22, + "learning_rate": 1.3386677900129496e-05, + "loss": 1.202, + "step": 14549 + }, + { + "epoch": 1.22, + "learning_rate": 1.3385822627930554e-05, + "loss": 1.2515, + "step": 14550 + }, + { + "epoch": 1.23, + "learning_rate": 1.338496732775686e-05, + "loss": 1.2169, + "step": 14551 + }, + { + "epoch": 1.23, + "learning_rate": 1.3384111999615485e-05, + "loss": 1.2387, + "step": 14552 + }, + { + "epoch": 1.23, + "learning_rate": 1.338325664351349e-05, + "loss": 1.1691, + "step": 14553 + }, + { + "epoch": 1.23, + "learning_rate": 1.3382401259457949e-05, + "loss": 1.1777, + "step": 14554 + }, + { + "epoch": 1.23, + "learning_rate": 1.3381545847455925e-05, + "loss": 1.1996, + "step": 14555 + }, + { + "epoch": 1.23, + "learning_rate": 1.3380690407514488e-05, + "loss": 1.2029, + "step": 14556 + }, + { + "epoch": 1.23, + "learning_rate": 1.3379834939640704e-05, + "loss": 1.2673, + "step": 14557 + }, + { + "epoch": 1.23, + "learning_rate": 1.337897944384164e-05, + "loss": 1.1952, + "step": 14558 + }, + { + "epoch": 1.23, + "learning_rate": 1.3378123920124368e-05, + "loss": 1.1696, + "step": 14559 + }, + { + "epoch": 1.23, + "learning_rate": 1.3377268368495955e-05, + "loss": 1.1439, + "step": 14560 + }, + { + "epoch": 1.23, + "learning_rate": 1.3376412788963471e-05, + "loss": 1.2073, + "step": 14561 + }, + { + "epoch": 1.23, + "learning_rate": 1.3375557181533984e-05, + "loss": 1.204, + "step": 14562 + }, + { + "epoch": 1.23, + "learning_rate": 1.3374701546214563e-05, + "loss": 1.1275, + "step": 14563 + }, + { + "epoch": 1.23, + "learning_rate": 1.3373845883012275e-05, + "loss": 1.1088, + "step": 14564 + }, + { + "epoch": 1.23, + "learning_rate": 1.3372990191934196e-05, + "loss": 1.2191, + "step": 14565 + }, + { + "epoch": 1.23, + "learning_rate": 1.3372134472987392e-05, + "loss": 1.1755, + "step": 14566 + }, + { + "epoch": 1.23, + "learning_rate": 1.3371278726178933e-05, + "loss": 1.1754, + "step": 14567 + }, + { + "epoch": 1.23, + "learning_rate": 1.337042295151589e-05, + "loss": 1.2282, + "step": 14568 + }, + { + "epoch": 1.23, + "learning_rate": 1.3369567149005333e-05, + "loss": 1.1796, + "step": 14569 + }, + { + "epoch": 1.23, + "learning_rate": 1.3368711318654337e-05, + "loss": 1.1846, + "step": 14570 + }, + { + "epoch": 1.23, + "learning_rate": 1.3367855460469965e-05, + "loss": 1.1841, + "step": 14571 + }, + { + "epoch": 1.23, + "learning_rate": 1.3366999574459296e-05, + "loss": 1.2226, + "step": 14572 + }, + { + "epoch": 1.23, + "learning_rate": 1.33661436606294e-05, + "loss": 1.2088, + "step": 14573 + }, + { + "epoch": 1.23, + "learning_rate": 1.3365287718987348e-05, + "loss": 1.2092, + "step": 14574 + }, + { + "epoch": 1.23, + "learning_rate": 1.3364431749540209e-05, + "loss": 1.2124, + "step": 14575 + }, + { + "epoch": 1.23, + "learning_rate": 1.3363575752295063e-05, + "loss": 1.1358, + "step": 14576 + }, + { + "epoch": 1.23, + "learning_rate": 1.3362719727258975e-05, + "loss": 1.2599, + "step": 14577 + }, + { + "epoch": 1.23, + "learning_rate": 1.3361863674439019e-05, + "loss": 1.1668, + "step": 14578 + }, + { + "epoch": 1.23, + "learning_rate": 1.336100759384227e-05, + "loss": 1.265, + "step": 14579 + }, + { + "epoch": 1.23, + "learning_rate": 1.3360151485475803e-05, + "loss": 1.256, + "step": 14580 + }, + { + "epoch": 1.23, + "learning_rate": 1.3359295349346686e-05, + "loss": 1.1877, + "step": 14581 + }, + { + "epoch": 1.23, + "learning_rate": 1.3358439185461998e-05, + "loss": 1.1784, + "step": 14582 + }, + { + "epoch": 1.23, + "learning_rate": 1.3357582993828808e-05, + "loss": 1.1972, + "step": 14583 + }, + { + "epoch": 1.23, + "learning_rate": 1.3356726774454194e-05, + "loss": 1.2104, + "step": 14584 + }, + { + "epoch": 1.23, + "learning_rate": 1.335587052734523e-05, + "loss": 1.1343, + "step": 14585 + }, + { + "epoch": 1.23, + "learning_rate": 1.3355014252508987e-05, + "loss": 1.2632, + "step": 14586 + }, + { + "epoch": 1.23, + "learning_rate": 1.3354157949952547e-05, + "loss": 1.1013, + "step": 14587 + }, + { + "epoch": 1.23, + "learning_rate": 1.3353301619682976e-05, + "loss": 1.2583, + "step": 14588 + }, + { + "epoch": 1.23, + "learning_rate": 1.3352445261707356e-05, + "loss": 1.206, + "step": 14589 + }, + { + "epoch": 1.23, + "learning_rate": 1.335158887603276e-05, + "loss": 1.3033, + "step": 14590 + }, + { + "epoch": 1.23, + "learning_rate": 1.3350732462666265e-05, + "loss": 1.1488, + "step": 14591 + }, + { + "epoch": 1.23, + "learning_rate": 1.3349876021614944e-05, + "loss": 1.2162, + "step": 14592 + }, + { + "epoch": 1.23, + "learning_rate": 1.3349019552885878e-05, + "loss": 1.1335, + "step": 14593 + }, + { + "epoch": 1.23, + "learning_rate": 1.3348163056486138e-05, + "loss": 1.2069, + "step": 14594 + }, + { + "epoch": 1.23, + "learning_rate": 1.3347306532422806e-05, + "loss": 1.1336, + "step": 14595 + }, + { + "epoch": 1.23, + "learning_rate": 1.3346449980702951e-05, + "loss": 1.2019, + "step": 14596 + }, + { + "epoch": 1.23, + "learning_rate": 1.3345593401333659e-05, + "loss": 1.2014, + "step": 14597 + }, + { + "epoch": 1.23, + "learning_rate": 1.3344736794322004e-05, + "loss": 1.2449, + "step": 14598 + }, + { + "epoch": 1.23, + "learning_rate": 1.3343880159675062e-05, + "loss": 1.2125, + "step": 14599 + }, + { + "epoch": 1.23, + "learning_rate": 1.3343023497399914e-05, + "loss": 1.0993, + "step": 14600 + }, + { + "epoch": 1.23, + "learning_rate": 1.3342166807503635e-05, + "loss": 1.1364, + "step": 14601 + }, + { + "epoch": 1.23, + "learning_rate": 1.3341310089993305e-05, + "loss": 1.1725, + "step": 14602 + }, + { + "epoch": 1.23, + "learning_rate": 1.3340453344875999e-05, + "loss": 1.1463, + "step": 14603 + }, + { + "epoch": 1.23, + "learning_rate": 1.3339596572158797e-05, + "loss": 1.1654, + "step": 14604 + }, + { + "epoch": 1.23, + "learning_rate": 1.3338739771848785e-05, + "loss": 1.165, + "step": 14605 + }, + { + "epoch": 1.23, + "learning_rate": 1.3337882943953033e-05, + "loss": 1.2765, + "step": 14606 + }, + { + "epoch": 1.23, + "learning_rate": 1.3337026088478625e-05, + "loss": 1.236, + "step": 14607 + }, + { + "epoch": 1.23, + "learning_rate": 1.3336169205432637e-05, + "loss": 1.2469, + "step": 14608 + }, + { + "epoch": 1.23, + "learning_rate": 1.3335312294822154e-05, + "loss": 1.2331, + "step": 14609 + }, + { + "epoch": 1.23, + "learning_rate": 1.3334455356654252e-05, + "loss": 1.1352, + "step": 14610 + }, + { + "epoch": 1.23, + "learning_rate": 1.333359839093601e-05, + "loss": 1.2259, + "step": 14611 + }, + { + "epoch": 1.23, + "learning_rate": 1.3332741397674517e-05, + "loss": 1.1504, + "step": 14612 + }, + { + "epoch": 1.23, + "learning_rate": 1.3331884376876847e-05, + "loss": 1.2366, + "step": 14613 + }, + { + "epoch": 1.23, + "learning_rate": 1.333102732855008e-05, + "loss": 1.1504, + "step": 14614 + }, + { + "epoch": 1.23, + "learning_rate": 1.3330170252701298e-05, + "loss": 1.2448, + "step": 14615 + }, + { + "epoch": 1.23, + "learning_rate": 1.3329313149337586e-05, + "loss": 1.1608, + "step": 14616 + }, + { + "epoch": 1.23, + "learning_rate": 1.3328456018466022e-05, + "loss": 1.2531, + "step": 14617 + }, + { + "epoch": 1.23, + "learning_rate": 1.332759886009369e-05, + "loss": 1.2325, + "step": 14618 + }, + { + "epoch": 1.23, + "learning_rate": 1.3326741674227668e-05, + "loss": 1.197, + "step": 14619 + }, + { + "epoch": 1.23, + "learning_rate": 1.3325884460875047e-05, + "loss": 1.1538, + "step": 14620 + }, + { + "epoch": 1.23, + "learning_rate": 1.33250272200429e-05, + "loss": 1.324, + "step": 14621 + }, + { + "epoch": 1.23, + "learning_rate": 1.3324169951738312e-05, + "loss": 1.2624, + "step": 14622 + }, + { + "epoch": 1.23, + "learning_rate": 1.3323312655968375e-05, + "loss": 1.1758, + "step": 14623 + }, + { + "epoch": 1.23, + "learning_rate": 1.3322455332740161e-05, + "loss": 1.1439, + "step": 14624 + }, + { + "epoch": 1.23, + "learning_rate": 1.332159798206076e-05, + "loss": 1.1679, + "step": 14625 + }, + { + "epoch": 1.23, + "learning_rate": 1.3320740603937251e-05, + "loss": 1.1908, + "step": 14626 + }, + { + "epoch": 1.23, + "learning_rate": 1.331988319837672e-05, + "loss": 1.3491, + "step": 14627 + }, + { + "epoch": 1.23, + "learning_rate": 1.3319025765386253e-05, + "loss": 1.2838, + "step": 14628 + }, + { + "epoch": 1.23, + "learning_rate": 1.3318168304972932e-05, + "loss": 1.229, + "step": 14629 + }, + { + "epoch": 1.23, + "learning_rate": 1.3317310817143844e-05, + "loss": 1.2411, + "step": 14630 + }, + { + "epoch": 1.23, + "learning_rate": 1.3316453301906071e-05, + "loss": 1.1895, + "step": 14631 + }, + { + "epoch": 1.23, + "learning_rate": 1.33155957592667e-05, + "loss": 1.2013, + "step": 14632 + }, + { + "epoch": 1.23, + "learning_rate": 1.3314738189232814e-05, + "loss": 1.1306, + "step": 14633 + }, + { + "epoch": 1.23, + "learning_rate": 1.3313880591811503e-05, + "loss": 1.29, + "step": 14634 + }, + { + "epoch": 1.23, + "learning_rate": 1.3313022967009848e-05, + "loss": 1.2252, + "step": 14635 + }, + { + "epoch": 1.23, + "learning_rate": 1.3312165314834937e-05, + "loss": 1.1804, + "step": 14636 + }, + { + "epoch": 1.23, + "learning_rate": 1.3311307635293857e-05, + "loss": 1.1728, + "step": 14637 + }, + { + "epoch": 1.23, + "learning_rate": 1.3310449928393694e-05, + "loss": 1.1726, + "step": 14638 + }, + { + "epoch": 1.23, + "learning_rate": 1.3309592194141532e-05, + "loss": 1.2074, + "step": 14639 + }, + { + "epoch": 1.23, + "learning_rate": 1.330873443254446e-05, + "loss": 1.2592, + "step": 14640 + }, + { + "epoch": 1.23, + "learning_rate": 1.330787664360957e-05, + "loss": 1.2174, + "step": 14641 + }, + { + "epoch": 1.23, + "learning_rate": 1.330701882734394e-05, + "loss": 1.1668, + "step": 14642 + }, + { + "epoch": 1.23, + "learning_rate": 1.3306160983754665e-05, + "loss": 1.1228, + "step": 14643 + }, + { + "epoch": 1.23, + "learning_rate": 1.3305303112848828e-05, + "loss": 1.2511, + "step": 14644 + }, + { + "epoch": 1.23, + "learning_rate": 1.3304445214633521e-05, + "loss": 1.1994, + "step": 14645 + }, + { + "epoch": 1.23, + "learning_rate": 1.3303587289115825e-05, + "loss": 1.2449, + "step": 14646 + }, + { + "epoch": 1.23, + "learning_rate": 1.3302729336302836e-05, + "loss": 1.0922, + "step": 14647 + }, + { + "epoch": 1.23, + "learning_rate": 1.3301871356201643e-05, + "loss": 1.1119, + "step": 14648 + }, + { + "epoch": 1.23, + "learning_rate": 1.3301013348819329e-05, + "loss": 1.1877, + "step": 14649 + }, + { + "epoch": 1.23, + "learning_rate": 1.3300155314162986e-05, + "loss": 1.1363, + "step": 14650 + }, + { + "epoch": 1.23, + "learning_rate": 1.3299297252239707e-05, + "loss": 1.1965, + "step": 14651 + }, + { + "epoch": 1.23, + "learning_rate": 1.3298439163056576e-05, + "loss": 1.1437, + "step": 14652 + }, + { + "epoch": 1.23, + "learning_rate": 1.3297581046620684e-05, + "loss": 1.2532, + "step": 14653 + }, + { + "epoch": 1.23, + "learning_rate": 1.3296722902939124e-05, + "loss": 1.2218, + "step": 14654 + }, + { + "epoch": 1.23, + "learning_rate": 1.3295864732018986e-05, + "loss": 1.258, + "step": 14655 + }, + { + "epoch": 1.23, + "learning_rate": 1.3295006533867354e-05, + "loss": 1.2151, + "step": 14656 + }, + { + "epoch": 1.23, + "learning_rate": 1.3294148308491326e-05, + "loss": 1.2089, + "step": 14657 + }, + { + "epoch": 1.23, + "learning_rate": 1.3293290055897989e-05, + "loss": 1.2868, + "step": 14658 + }, + { + "epoch": 1.23, + "learning_rate": 1.3292431776094436e-05, + "loss": 1.1979, + "step": 14659 + }, + { + "epoch": 1.23, + "learning_rate": 1.329157346908776e-05, + "loss": 1.1805, + "step": 14660 + }, + { + "epoch": 1.23, + "learning_rate": 1.3290715134885046e-05, + "loss": 1.2186, + "step": 14661 + }, + { + "epoch": 1.23, + "learning_rate": 1.3289856773493397e-05, + "loss": 1.1957, + "step": 14662 + }, + { + "epoch": 1.23, + "learning_rate": 1.3288998384919893e-05, + "loss": 1.24, + "step": 14663 + }, + { + "epoch": 1.23, + "learning_rate": 1.3288139969171635e-05, + "loss": 1.2435, + "step": 14664 + }, + { + "epoch": 1.23, + "learning_rate": 1.3287281526255709e-05, + "loss": 1.1155, + "step": 14665 + }, + { + "epoch": 1.23, + "learning_rate": 1.3286423056179216e-05, + "loss": 1.1622, + "step": 14666 + }, + { + "epoch": 1.23, + "learning_rate": 1.328556455894924e-05, + "loss": 1.1461, + "step": 14667 + }, + { + "epoch": 1.23, + "learning_rate": 1.3284706034572878e-05, + "loss": 1.2566, + "step": 14668 + }, + { + "epoch": 1.24, + "learning_rate": 1.3283847483057223e-05, + "loss": 1.1399, + "step": 14669 + }, + { + "epoch": 1.24, + "learning_rate": 1.3282988904409374e-05, + "loss": 1.2291, + "step": 14670 + }, + { + "epoch": 1.24, + "learning_rate": 1.3282130298636414e-05, + "loss": 1.1513, + "step": 14671 + }, + { + "epoch": 1.24, + "learning_rate": 1.3281271665745445e-05, + "loss": 1.2727, + "step": 14672 + }, + { + "epoch": 1.24, + "learning_rate": 1.3280413005743562e-05, + "loss": 1.2128, + "step": 14673 + }, + { + "epoch": 1.24, + "learning_rate": 1.3279554318637855e-05, + "loss": 1.1631, + "step": 14674 + }, + { + "epoch": 1.24, + "learning_rate": 1.3278695604435418e-05, + "loss": 1.2254, + "step": 14675 + }, + { + "epoch": 1.24, + "learning_rate": 1.3277836863143352e-05, + "loss": 1.2305, + "step": 14676 + }, + { + "epoch": 1.24, + "learning_rate": 1.3276978094768748e-05, + "loss": 1.1962, + "step": 14677 + }, + { + "epoch": 1.24, + "learning_rate": 1.3276119299318704e-05, + "loss": 1.1476, + "step": 14678 + }, + { + "epoch": 1.24, + "learning_rate": 1.327526047680031e-05, + "loss": 1.1205, + "step": 14679 + }, + { + "epoch": 1.24, + "learning_rate": 1.327440162722067e-05, + "loss": 1.2828, + "step": 14680 + }, + { + "epoch": 1.24, + "learning_rate": 1.3273542750586871e-05, + "loss": 1.1043, + "step": 14681 + }, + { + "epoch": 1.24, + "learning_rate": 1.3272683846906019e-05, + "loss": 1.185, + "step": 14682 + }, + { + "epoch": 1.24, + "learning_rate": 1.3271824916185203e-05, + "loss": 1.2541, + "step": 14683 + }, + { + "epoch": 1.24, + "learning_rate": 1.327096595843152e-05, + "loss": 1.2968, + "step": 14684 + }, + { + "epoch": 1.24, + "learning_rate": 1.3270106973652072e-05, + "loss": 1.2094, + "step": 14685 + }, + { + "epoch": 1.24, + "learning_rate": 1.3269247961853955e-05, + "loss": 1.1801, + "step": 14686 + }, + { + "epoch": 1.24, + "learning_rate": 1.3268388923044265e-05, + "loss": 1.1594, + "step": 14687 + }, + { + "epoch": 1.24, + "learning_rate": 1.3267529857230096e-05, + "loss": 1.1489, + "step": 14688 + }, + { + "epoch": 1.24, + "learning_rate": 1.3266670764418552e-05, + "loss": 1.1927, + "step": 14689 + }, + { + "epoch": 1.24, + "learning_rate": 1.3265811644616728e-05, + "loss": 1.2511, + "step": 14690 + }, + { + "epoch": 1.24, + "learning_rate": 1.3264952497831727e-05, + "loss": 1.229, + "step": 14691 + }, + { + "epoch": 1.24, + "learning_rate": 1.326409332407064e-05, + "loss": 1.2142, + "step": 14692 + }, + { + "epoch": 1.24, + "learning_rate": 1.3263234123340569e-05, + "loss": 1.1849, + "step": 14693 + }, + { + "epoch": 1.24, + "learning_rate": 1.3262374895648612e-05, + "loss": 1.221, + "step": 14694 + }, + { + "epoch": 1.24, + "learning_rate": 1.3261515641001874e-05, + "loss": 1.1535, + "step": 14695 + }, + { + "epoch": 1.24, + "learning_rate": 1.3260656359407445e-05, + "loss": 1.2356, + "step": 14696 + }, + { + "epoch": 1.24, + "learning_rate": 1.3259797050872431e-05, + "loss": 1.2596, + "step": 14697 + }, + { + "epoch": 1.24, + "learning_rate": 1.3258937715403933e-05, + "loss": 1.2324, + "step": 14698 + }, + { + "epoch": 1.24, + "learning_rate": 1.3258078353009047e-05, + "loss": 1.2406, + "step": 14699 + }, + { + "epoch": 1.24, + "learning_rate": 1.3257218963694876e-05, + "loss": 1.2373, + "step": 14700 + }, + { + "epoch": 1.24, + "learning_rate": 1.3256359547468516e-05, + "loss": 1.3014, + "step": 14701 + }, + { + "epoch": 1.24, + "learning_rate": 1.3255500104337075e-05, + "loss": 1.2586, + "step": 14702 + }, + { + "epoch": 1.24, + "learning_rate": 1.3254640634307649e-05, + "loss": 1.135, + "step": 14703 + }, + { + "epoch": 1.24, + "learning_rate": 1.3253781137387341e-05, + "loss": 1.2042, + "step": 14704 + }, + { + "epoch": 1.24, + "learning_rate": 1.3252921613583251e-05, + "loss": 1.2814, + "step": 14705 + }, + { + "epoch": 1.24, + "learning_rate": 1.3252062062902482e-05, + "loss": 1.2315, + "step": 14706 + }, + { + "epoch": 1.24, + "learning_rate": 1.3251202485352135e-05, + "loss": 1.1741, + "step": 14707 + }, + { + "epoch": 1.24, + "learning_rate": 1.3250342880939313e-05, + "loss": 1.2115, + "step": 14708 + }, + { + "epoch": 1.24, + "learning_rate": 1.3249483249671117e-05, + "loss": 1.1402, + "step": 14709 + }, + { + "epoch": 1.24, + "learning_rate": 1.3248623591554653e-05, + "loss": 1.1817, + "step": 14710 + }, + { + "epoch": 1.24, + "learning_rate": 1.3247763906597018e-05, + "loss": 1.1419, + "step": 14711 + }, + { + "epoch": 1.24, + "learning_rate": 1.324690419480532e-05, + "loss": 1.1637, + "step": 14712 + }, + { + "epoch": 1.24, + "learning_rate": 1.324604445618666e-05, + "loss": 1.2776, + "step": 14713 + }, + { + "epoch": 1.24, + "learning_rate": 1.3245184690748141e-05, + "loss": 1.2405, + "step": 14714 + }, + { + "epoch": 1.24, + "learning_rate": 1.324432489849687e-05, + "loss": 1.2138, + "step": 14715 + }, + { + "epoch": 1.24, + "learning_rate": 1.3243465079439948e-05, + "loss": 1.0546, + "step": 14716 + }, + { + "epoch": 1.24, + "learning_rate": 1.3242605233584476e-05, + "loss": 1.1649, + "step": 14717 + }, + { + "epoch": 1.24, + "learning_rate": 1.3241745360937563e-05, + "loss": 1.2236, + "step": 14718 + }, + { + "epoch": 1.24, + "learning_rate": 1.3240885461506314e-05, + "loss": 1.1544, + "step": 14719 + }, + { + "epoch": 1.24, + "learning_rate": 1.3240025535297833e-05, + "loss": 1.2676, + "step": 14720 + }, + { + "epoch": 1.24, + "learning_rate": 1.323916558231922e-05, + "loss": 1.2302, + "step": 14721 + }, + { + "epoch": 1.24, + "learning_rate": 1.3238305602577586e-05, + "loss": 1.1148, + "step": 14722 + }, + { + "epoch": 1.24, + "learning_rate": 1.3237445596080036e-05, + "loss": 1.1185, + "step": 14723 + }, + { + "epoch": 1.24, + "learning_rate": 1.3236585562833675e-05, + "loss": 1.1504, + "step": 14724 + }, + { + "epoch": 1.24, + "learning_rate": 1.3235725502845606e-05, + "loss": 1.2193, + "step": 14725 + }, + { + "epoch": 1.24, + "learning_rate": 1.3234865416122937e-05, + "loss": 1.1999, + "step": 14726 + }, + { + "epoch": 1.24, + "learning_rate": 1.323400530267278e-05, + "loss": 1.179, + "step": 14727 + }, + { + "epoch": 1.24, + "learning_rate": 1.323314516250223e-05, + "loss": 1.2161, + "step": 14728 + }, + { + "epoch": 1.24, + "learning_rate": 1.3232284995618403e-05, + "loss": 1.1776, + "step": 14729 + }, + { + "epoch": 1.24, + "learning_rate": 1.3231424802028404e-05, + "loss": 1.2699, + "step": 14730 + }, + { + "epoch": 1.24, + "learning_rate": 1.3230564581739336e-05, + "loss": 1.2034, + "step": 14731 + }, + { + "epoch": 1.24, + "learning_rate": 1.3229704334758311e-05, + "loss": 1.16, + "step": 14732 + }, + { + "epoch": 1.24, + "learning_rate": 1.3228844061092434e-05, + "loss": 1.1387, + "step": 14733 + }, + { + "epoch": 1.24, + "learning_rate": 1.3227983760748812e-05, + "loss": 1.2033, + "step": 14734 + }, + { + "epoch": 1.24, + "learning_rate": 1.322712343373456e-05, + "loss": 1.139, + "step": 14735 + }, + { + "epoch": 1.24, + "learning_rate": 1.3226263080056778e-05, + "loss": 1.2433, + "step": 14736 + }, + { + "epoch": 1.24, + "learning_rate": 1.322540269972258e-05, + "loss": 1.2672, + "step": 14737 + }, + { + "epoch": 1.24, + "learning_rate": 1.322454229273907e-05, + "loss": 1.1498, + "step": 14738 + }, + { + "epoch": 1.24, + "learning_rate": 1.3223681859113361e-05, + "loss": 1.1323, + "step": 14739 + }, + { + "epoch": 1.24, + "learning_rate": 1.3222821398852564e-05, + "loss": 1.1777, + "step": 14740 + }, + { + "epoch": 1.24, + "learning_rate": 1.3221960911963783e-05, + "loss": 1.1803, + "step": 14741 + }, + { + "epoch": 1.24, + "learning_rate": 1.3221100398454127e-05, + "loss": 1.209, + "step": 14742 + }, + { + "epoch": 1.24, + "learning_rate": 1.3220239858330713e-05, + "loss": 1.2393, + "step": 14743 + }, + { + "epoch": 1.24, + "learning_rate": 1.3219379291600643e-05, + "loss": 1.2689, + "step": 14744 + }, + { + "epoch": 1.24, + "learning_rate": 1.3218518698271034e-05, + "loss": 1.2344, + "step": 14745 + }, + { + "epoch": 1.24, + "learning_rate": 1.321765807834899e-05, + "loss": 1.209, + "step": 14746 + }, + { + "epoch": 1.24, + "learning_rate": 1.3216797431841625e-05, + "loss": 1.1839, + "step": 14747 + }, + { + "epoch": 1.24, + "learning_rate": 1.3215936758756055e-05, + "loss": 1.1324, + "step": 14748 + }, + { + "epoch": 1.24, + "learning_rate": 1.3215076059099384e-05, + "loss": 1.2553, + "step": 14749 + }, + { + "epoch": 1.24, + "learning_rate": 1.3214215332878724e-05, + "loss": 1.1753, + "step": 14750 + }, + { + "epoch": 1.24, + "learning_rate": 1.3213354580101187e-05, + "loss": 1.1988, + "step": 14751 + }, + { + "epoch": 1.24, + "learning_rate": 1.3212493800773891e-05, + "loss": 1.2405, + "step": 14752 + }, + { + "epoch": 1.24, + "learning_rate": 1.321163299490394e-05, + "loss": 1.2498, + "step": 14753 + }, + { + "epoch": 1.24, + "learning_rate": 1.3210772162498449e-05, + "loss": 1.1876, + "step": 14754 + }, + { + "epoch": 1.24, + "learning_rate": 1.3209911303564533e-05, + "loss": 1.1569, + "step": 14755 + }, + { + "epoch": 1.24, + "learning_rate": 1.3209050418109302e-05, + "loss": 1.1642, + "step": 14756 + }, + { + "epoch": 1.24, + "learning_rate": 1.3208189506139869e-05, + "loss": 1.2232, + "step": 14757 + }, + { + "epoch": 1.24, + "learning_rate": 1.3207328567663344e-05, + "loss": 1.2553, + "step": 14758 + }, + { + "epoch": 1.24, + "learning_rate": 1.3206467602686849e-05, + "loss": 1.2202, + "step": 14759 + }, + { + "epoch": 1.24, + "learning_rate": 1.3205606611217488e-05, + "loss": 1.2294, + "step": 14760 + }, + { + "epoch": 1.24, + "learning_rate": 1.3204745593262382e-05, + "loss": 1.211, + "step": 14761 + }, + { + "epoch": 1.24, + "learning_rate": 1.3203884548828642e-05, + "loss": 1.187, + "step": 14762 + }, + { + "epoch": 1.24, + "learning_rate": 1.3203023477923382e-05, + "loss": 1.2567, + "step": 14763 + }, + { + "epoch": 1.24, + "learning_rate": 1.3202162380553717e-05, + "loss": 1.1894, + "step": 14764 + }, + { + "epoch": 1.24, + "learning_rate": 1.3201301256726762e-05, + "loss": 1.1327, + "step": 14765 + }, + { + "epoch": 1.24, + "learning_rate": 1.3200440106449635e-05, + "loss": 1.2298, + "step": 14766 + }, + { + "epoch": 1.24, + "learning_rate": 1.3199578929729443e-05, + "loss": 1.1573, + "step": 14767 + }, + { + "epoch": 1.24, + "learning_rate": 1.3198717726573305e-05, + "loss": 1.205, + "step": 14768 + }, + { + "epoch": 1.24, + "learning_rate": 1.319785649698834e-05, + "loss": 1.1852, + "step": 14769 + }, + { + "epoch": 1.24, + "learning_rate": 1.3196995240981663e-05, + "loss": 1.0888, + "step": 14770 + }, + { + "epoch": 1.24, + "learning_rate": 1.3196133958560382e-05, + "loss": 1.1708, + "step": 14771 + }, + { + "epoch": 1.24, + "learning_rate": 1.319527264973162e-05, + "loss": 1.2015, + "step": 14772 + }, + { + "epoch": 1.24, + "learning_rate": 1.31944113145025e-05, + "loss": 1.1943, + "step": 14773 + }, + { + "epoch": 1.24, + "learning_rate": 1.3193549952880126e-05, + "loss": 1.0857, + "step": 14774 + }, + { + "epoch": 1.24, + "learning_rate": 1.3192688564871621e-05, + "loss": 1.2487, + "step": 14775 + }, + { + "epoch": 1.24, + "learning_rate": 1.31918271504841e-05, + "loss": 1.2331, + "step": 14776 + }, + { + "epoch": 1.24, + "learning_rate": 1.3190965709724685e-05, + "loss": 1.1594, + "step": 14777 + }, + { + "epoch": 1.24, + "learning_rate": 1.3190104242600487e-05, + "loss": 1.1037, + "step": 14778 + }, + { + "epoch": 1.24, + "learning_rate": 1.318924274911863e-05, + "loss": 1.1017, + "step": 14779 + }, + { + "epoch": 1.24, + "learning_rate": 1.3188381229286227e-05, + "loss": 1.2356, + "step": 14780 + }, + { + "epoch": 1.24, + "learning_rate": 1.3187519683110397e-05, + "loss": 1.2286, + "step": 14781 + }, + { + "epoch": 1.24, + "learning_rate": 1.3186658110598261e-05, + "loss": 1.1772, + "step": 14782 + }, + { + "epoch": 1.24, + "learning_rate": 1.3185796511756931e-05, + "loss": 1.2287, + "step": 14783 + }, + { + "epoch": 1.24, + "learning_rate": 1.3184934886593537e-05, + "loss": 1.1719, + "step": 14784 + }, + { + "epoch": 1.24, + "learning_rate": 1.3184073235115186e-05, + "loss": 1.2958, + "step": 14785 + }, + { + "epoch": 1.24, + "learning_rate": 1.3183211557329005e-05, + "loss": 1.2286, + "step": 14786 + }, + { + "epoch": 1.24, + "learning_rate": 1.3182349853242115e-05, + "loss": 1.2145, + "step": 14787 + }, + { + "epoch": 1.25, + "learning_rate": 1.318148812286163e-05, + "loss": 1.2413, + "step": 14788 + }, + { + "epoch": 1.25, + "learning_rate": 1.3180626366194669e-05, + "loss": 1.305, + "step": 14789 + }, + { + "epoch": 1.25, + "learning_rate": 1.3179764583248358e-05, + "loss": 1.1939, + "step": 14790 + }, + { + "epoch": 1.25, + "learning_rate": 1.3178902774029814e-05, + "loss": 1.2029, + "step": 14791 + }, + { + "epoch": 1.25, + "learning_rate": 1.3178040938546157e-05, + "loss": 1.1175, + "step": 14792 + }, + { + "epoch": 1.25, + "learning_rate": 1.317717907680451e-05, + "loss": 1.2139, + "step": 14793 + }, + { + "epoch": 1.25, + "learning_rate": 1.3176317188811992e-05, + "loss": 1.2024, + "step": 14794 + }, + { + "epoch": 1.25, + "learning_rate": 1.3175455274575726e-05, + "loss": 1.2157, + "step": 14795 + }, + { + "epoch": 1.25, + "learning_rate": 1.3174593334102827e-05, + "loss": 1.141, + "step": 14796 + }, + { + "epoch": 1.25, + "learning_rate": 1.3173731367400426e-05, + "loss": 1.221, + "step": 14797 + }, + { + "epoch": 1.25, + "learning_rate": 1.317286937447564e-05, + "loss": 1.187, + "step": 14798 + }, + { + "epoch": 1.25, + "learning_rate": 1.3172007355335593e-05, + "loss": 1.2254, + "step": 14799 + }, + { + "epoch": 1.25, + "learning_rate": 1.3171145309987406e-05, + "loss": 1.1691, + "step": 14800 + }, + { + "epoch": 1.25, + "learning_rate": 1.31702832384382e-05, + "loss": 1.1275, + "step": 14801 + }, + { + "epoch": 1.25, + "learning_rate": 1.31694211406951e-05, + "loss": 1.2449, + "step": 14802 + }, + { + "epoch": 1.25, + "learning_rate": 1.3168559016765227e-05, + "loss": 1.1837, + "step": 14803 + }, + { + "epoch": 1.25, + "learning_rate": 1.3167696866655709e-05, + "loss": 1.1455, + "step": 14804 + }, + { + "epoch": 1.25, + "learning_rate": 1.3166834690373664e-05, + "loss": 1.2552, + "step": 14805 + }, + { + "epoch": 1.25, + "learning_rate": 1.3165972487926215e-05, + "loss": 1.2165, + "step": 14806 + }, + { + "epoch": 1.25, + "learning_rate": 1.316511025932049e-05, + "loss": 1.2814, + "step": 14807 + }, + { + "epoch": 1.25, + "learning_rate": 1.3164248004563612e-05, + "loss": 1.2545, + "step": 14808 + }, + { + "epoch": 1.25, + "learning_rate": 1.31633857236627e-05, + "loss": 1.1365, + "step": 14809 + }, + { + "epoch": 1.25, + "learning_rate": 1.3162523416624886e-05, + "loss": 1.2622, + "step": 14810 + }, + { + "epoch": 1.25, + "learning_rate": 1.316166108345729e-05, + "loss": 1.196, + "step": 14811 + }, + { + "epoch": 1.25, + "learning_rate": 1.3160798724167041e-05, + "loss": 1.1932, + "step": 14812 + }, + { + "epoch": 1.25, + "learning_rate": 1.3159936338761262e-05, + "loss": 1.2518, + "step": 14813 + }, + { + "epoch": 1.25, + "learning_rate": 1.3159073927247073e-05, + "loss": 1.1093, + "step": 14814 + }, + { + "epoch": 1.25, + "learning_rate": 1.3158211489631607e-05, + "loss": 1.1797, + "step": 14815 + }, + { + "epoch": 1.25, + "learning_rate": 1.3157349025921989e-05, + "loss": 1.143, + "step": 14816 + }, + { + "epoch": 1.25, + "learning_rate": 1.3156486536125338e-05, + "loss": 1.1661, + "step": 14817 + }, + { + "epoch": 1.25, + "learning_rate": 1.3155624020248787e-05, + "loss": 1.2563, + "step": 14818 + }, + { + "epoch": 1.25, + "learning_rate": 1.3154761478299461e-05, + "loss": 1.2696, + "step": 14819 + }, + { + "epoch": 1.25, + "learning_rate": 1.3153898910284488e-05, + "loss": 1.2101, + "step": 14820 + }, + { + "epoch": 1.25, + "learning_rate": 1.3153036316210987e-05, + "loss": 1.2336, + "step": 14821 + }, + { + "epoch": 1.25, + "learning_rate": 1.3152173696086094e-05, + "loss": 1.1281, + "step": 14822 + }, + { + "epoch": 1.25, + "learning_rate": 1.3151311049916936e-05, + "loss": 1.2207, + "step": 14823 + }, + { + "epoch": 1.25, + "learning_rate": 1.3150448377710634e-05, + "loss": 1.2025, + "step": 14824 + }, + { + "epoch": 1.25, + "learning_rate": 1.3149585679474322e-05, + "loss": 1.2372, + "step": 14825 + }, + { + "epoch": 1.25, + "learning_rate": 1.3148722955215121e-05, + "loss": 1.1967, + "step": 14826 + }, + { + "epoch": 1.25, + "learning_rate": 1.3147860204940168e-05, + "loss": 1.0616, + "step": 14827 + }, + { + "epoch": 1.25, + "learning_rate": 1.3146997428656584e-05, + "loss": 1.2298, + "step": 14828 + }, + { + "epoch": 1.25, + "learning_rate": 1.31461346263715e-05, + "loss": 1.1742, + "step": 14829 + }, + { + "epoch": 1.25, + "learning_rate": 1.3145271798092046e-05, + "loss": 1.2883, + "step": 14830 + }, + { + "epoch": 1.25, + "learning_rate": 1.3144408943825348e-05, + "loss": 1.1899, + "step": 14831 + }, + { + "epoch": 1.25, + "learning_rate": 1.3143546063578538e-05, + "loss": 1.2084, + "step": 14832 + }, + { + "epoch": 1.25, + "learning_rate": 1.3142683157358742e-05, + "loss": 1.0777, + "step": 14833 + }, + { + "epoch": 1.25, + "learning_rate": 1.3141820225173097e-05, + "loss": 1.2113, + "step": 14834 + }, + { + "epoch": 1.25, + "learning_rate": 1.3140957267028725e-05, + "loss": 1.1593, + "step": 14835 + }, + { + "epoch": 1.25, + "learning_rate": 1.3140094282932759e-05, + "loss": 1.1924, + "step": 14836 + }, + { + "epoch": 1.25, + "learning_rate": 1.313923127289233e-05, + "loss": 1.1681, + "step": 14837 + }, + { + "epoch": 1.25, + "learning_rate": 1.3138368236914565e-05, + "loss": 1.2169, + "step": 14838 + }, + { + "epoch": 1.25, + "learning_rate": 1.31375051750066e-05, + "loss": 1.2289, + "step": 14839 + }, + { + "epoch": 1.25, + "learning_rate": 1.3136642087175559e-05, + "loss": 1.2079, + "step": 14840 + }, + { + "epoch": 1.25, + "learning_rate": 1.3135778973428583e-05, + "loss": 1.1006, + "step": 14841 + }, + { + "epoch": 1.25, + "learning_rate": 1.3134915833772795e-05, + "loss": 1.2223, + "step": 14842 + }, + { + "epoch": 1.25, + "learning_rate": 1.3134052668215328e-05, + "loss": 1.1723, + "step": 14843 + }, + { + "epoch": 1.25, + "learning_rate": 1.3133189476763316e-05, + "loss": 1.1976, + "step": 14844 + }, + { + "epoch": 1.25, + "learning_rate": 1.3132326259423888e-05, + "loss": 1.1947, + "step": 14845 + }, + { + "epoch": 1.25, + "learning_rate": 1.3131463016204177e-05, + "loss": 1.1021, + "step": 14846 + }, + { + "epoch": 1.25, + "learning_rate": 1.3130599747111318e-05, + "loss": 1.1741, + "step": 14847 + }, + { + "epoch": 1.25, + "learning_rate": 1.3129736452152444e-05, + "loss": 1.2047, + "step": 14848 + }, + { + "epoch": 1.25, + "learning_rate": 1.3128873131334685e-05, + "loss": 1.2707, + "step": 14849 + }, + { + "epoch": 1.25, + "learning_rate": 1.3128009784665174e-05, + "loss": 1.2496, + "step": 14850 + }, + { + "epoch": 1.25, + "learning_rate": 1.3127146412151044e-05, + "loss": 1.2631, + "step": 14851 + }, + { + "epoch": 1.25, + "learning_rate": 1.3126283013799432e-05, + "loss": 1.2878, + "step": 14852 + }, + { + "epoch": 1.25, + "learning_rate": 1.312541958961747e-05, + "loss": 1.1368, + "step": 14853 + }, + { + "epoch": 1.25, + "learning_rate": 1.3124556139612288e-05, + "loss": 1.1995, + "step": 14854 + }, + { + "epoch": 1.25, + "learning_rate": 1.3123692663791024e-05, + "loss": 1.215, + "step": 14855 + }, + { + "epoch": 1.25, + "learning_rate": 1.3122829162160812e-05, + "loss": 1.1742, + "step": 14856 + }, + { + "epoch": 1.25, + "learning_rate": 1.3121965634728786e-05, + "loss": 1.1605, + "step": 14857 + }, + { + "epoch": 1.25, + "learning_rate": 1.3121102081502077e-05, + "loss": 1.1672, + "step": 14858 + }, + { + "epoch": 1.25, + "learning_rate": 1.312023850248783e-05, + "loss": 1.2998, + "step": 14859 + }, + { + "epoch": 1.25, + "learning_rate": 1.3119374897693172e-05, + "loss": 1.1373, + "step": 14860 + }, + { + "epoch": 1.25, + "learning_rate": 1.3118511267125238e-05, + "loss": 1.2003, + "step": 14861 + }, + { + "epoch": 1.25, + "learning_rate": 1.3117647610791169e-05, + "loss": 1.2204, + "step": 14862 + }, + { + "epoch": 1.25, + "learning_rate": 1.3116783928698095e-05, + "loss": 1.1772, + "step": 14863 + }, + { + "epoch": 1.25, + "learning_rate": 1.3115920220853155e-05, + "loss": 1.264, + "step": 14864 + }, + { + "epoch": 1.25, + "learning_rate": 1.3115056487263486e-05, + "loss": 1.2649, + "step": 14865 + }, + { + "epoch": 1.25, + "learning_rate": 1.3114192727936224e-05, + "loss": 1.2176, + "step": 14866 + }, + { + "epoch": 1.25, + "learning_rate": 1.3113328942878503e-05, + "loss": 1.2703, + "step": 14867 + }, + { + "epoch": 1.25, + "learning_rate": 1.3112465132097463e-05, + "loss": 1.2651, + "step": 14868 + }, + { + "epoch": 1.25, + "learning_rate": 1.3111601295600237e-05, + "loss": 1.1754, + "step": 14869 + }, + { + "epoch": 1.25, + "learning_rate": 1.3110737433393971e-05, + "loss": 1.2556, + "step": 14870 + }, + { + "epoch": 1.25, + "learning_rate": 1.3109873545485792e-05, + "loss": 1.1784, + "step": 14871 + }, + { + "epoch": 1.25, + "learning_rate": 1.3109009631882846e-05, + "loss": 1.09, + "step": 14872 + }, + { + "epoch": 1.25, + "learning_rate": 1.3108145692592267e-05, + "loss": 1.1493, + "step": 14873 + }, + { + "epoch": 1.25, + "learning_rate": 1.310728172762119e-05, + "loss": 1.1361, + "step": 14874 + }, + { + "epoch": 1.25, + "learning_rate": 1.3106417736976759e-05, + "loss": 1.1786, + "step": 14875 + }, + { + "epoch": 1.25, + "learning_rate": 1.3105553720666111e-05, + "loss": 1.1734, + "step": 14876 + }, + { + "epoch": 1.25, + "learning_rate": 1.3104689678696386e-05, + "loss": 1.1899, + "step": 14877 + }, + { + "epoch": 1.25, + "learning_rate": 1.3103825611074718e-05, + "loss": 1.1791, + "step": 14878 + }, + { + "epoch": 1.25, + "learning_rate": 1.310296151780825e-05, + "loss": 1.2206, + "step": 14879 + }, + { + "epoch": 1.25, + "learning_rate": 1.3102097398904122e-05, + "loss": 1.149, + "step": 14880 + }, + { + "epoch": 1.25, + "learning_rate": 1.3101233254369472e-05, + "loss": 1.1779, + "step": 14881 + }, + { + "epoch": 1.25, + "learning_rate": 1.3100369084211439e-05, + "loss": 1.2128, + "step": 14882 + }, + { + "epoch": 1.25, + "learning_rate": 1.3099504888437164e-05, + "loss": 1.1563, + "step": 14883 + }, + { + "epoch": 1.25, + "learning_rate": 1.3098640667053792e-05, + "loss": 1.2177, + "step": 14884 + }, + { + "epoch": 1.25, + "learning_rate": 1.3097776420068456e-05, + "loss": 1.1418, + "step": 14885 + }, + { + "epoch": 1.25, + "learning_rate": 1.3096912147488298e-05, + "loss": 1.2157, + "step": 14886 + }, + { + "epoch": 1.25, + "learning_rate": 1.3096047849320466e-05, + "loss": 1.1269, + "step": 14887 + }, + { + "epoch": 1.25, + "learning_rate": 1.309518352557209e-05, + "loss": 1.1936, + "step": 14888 + }, + { + "epoch": 1.25, + "learning_rate": 1.309431917625032e-05, + "loss": 1.1363, + "step": 14889 + }, + { + "epoch": 1.25, + "learning_rate": 1.3093454801362295e-05, + "loss": 1.0545, + "step": 14890 + }, + { + "epoch": 1.25, + "learning_rate": 1.3092590400915158e-05, + "loss": 1.1779, + "step": 14891 + }, + { + "epoch": 1.25, + "learning_rate": 1.3091725974916046e-05, + "loss": 1.2481, + "step": 14892 + }, + { + "epoch": 1.25, + "learning_rate": 1.3090861523372107e-05, + "loss": 1.1703, + "step": 14893 + }, + { + "epoch": 1.25, + "learning_rate": 1.3089997046290476e-05, + "loss": 1.2591, + "step": 14894 + }, + { + "epoch": 1.25, + "learning_rate": 1.308913254367831e-05, + "loss": 1.204, + "step": 14895 + }, + { + "epoch": 1.25, + "learning_rate": 1.3088268015542732e-05, + "loss": 1.2334, + "step": 14896 + }, + { + "epoch": 1.25, + "learning_rate": 1.3087403461890899e-05, + "loss": 1.1779, + "step": 14897 + }, + { + "epoch": 1.25, + "learning_rate": 1.3086538882729953e-05, + "loss": 1.2441, + "step": 14898 + }, + { + "epoch": 1.25, + "learning_rate": 1.3085674278067033e-05, + "loss": 1.2106, + "step": 14899 + }, + { + "epoch": 1.25, + "learning_rate": 1.3084809647909284e-05, + "loss": 1.1218, + "step": 14900 + }, + { + "epoch": 1.25, + "learning_rate": 1.308394499226385e-05, + "loss": 1.1926, + "step": 14901 + }, + { + "epoch": 1.25, + "learning_rate": 1.3083080311137879e-05, + "loss": 1.1834, + "step": 14902 + }, + { + "epoch": 1.25, + "learning_rate": 1.3082215604538508e-05, + "loss": 1.2178, + "step": 14903 + }, + { + "epoch": 1.25, + "learning_rate": 1.3081350872472887e-05, + "loss": 1.1743, + "step": 14904 + }, + { + "epoch": 1.25, + "learning_rate": 1.3080486114948159e-05, + "loss": 1.2252, + "step": 14905 + }, + { + "epoch": 1.25, + "learning_rate": 1.307962133197147e-05, + "loss": 1.1773, + "step": 14906 + }, + { + "epoch": 1.26, + "learning_rate": 1.307875652354996e-05, + "loss": 1.1523, + "step": 14907 + }, + { + "epoch": 1.26, + "learning_rate": 1.3077891689690781e-05, + "loss": 1.2303, + "step": 14908 + }, + { + "epoch": 1.26, + "learning_rate": 1.3077026830401076e-05, + "loss": 1.2559, + "step": 14909 + }, + { + "epoch": 1.26, + "learning_rate": 1.3076161945687988e-05, + "loss": 1.2756, + "step": 14910 + }, + { + "epoch": 1.26, + "learning_rate": 1.3075297035558667e-05, + "loss": 1.1416, + "step": 14911 + }, + { + "epoch": 1.26, + "learning_rate": 1.307443210002026e-05, + "loss": 1.1674, + "step": 14912 + }, + { + "epoch": 1.26, + "learning_rate": 1.307356713907991e-05, + "loss": 1.2123, + "step": 14913 + }, + { + "epoch": 1.26, + "learning_rate": 1.3072702152744763e-05, + "loss": 1.2548, + "step": 14914 + }, + { + "epoch": 1.26, + "learning_rate": 1.3071837141021969e-05, + "loss": 1.2775, + "step": 14915 + }, + { + "epoch": 1.26, + "learning_rate": 1.3070972103918673e-05, + "loss": 1.2078, + "step": 14916 + }, + { + "epoch": 1.26, + "learning_rate": 1.3070107041442022e-05, + "loss": 1.2074, + "step": 14917 + }, + { + "epoch": 1.26, + "learning_rate": 1.3069241953599163e-05, + "loss": 1.2043, + "step": 14918 + }, + { + "epoch": 1.26, + "learning_rate": 1.3068376840397245e-05, + "loss": 1.1113, + "step": 14919 + }, + { + "epoch": 1.26, + "learning_rate": 1.3067511701843421e-05, + "loss": 1.2295, + "step": 14920 + }, + { + "epoch": 1.26, + "learning_rate": 1.3066646537944825e-05, + "loss": 1.2663, + "step": 14921 + }, + { + "epoch": 1.26, + "learning_rate": 1.3065781348708617e-05, + "loss": 1.2247, + "step": 14922 + }, + { + "epoch": 1.26, + "learning_rate": 1.3064916134141944e-05, + "loss": 1.2974, + "step": 14923 + }, + { + "epoch": 1.26, + "learning_rate": 1.3064050894251953e-05, + "loss": 1.2688, + "step": 14924 + }, + { + "epoch": 1.26, + "learning_rate": 1.3063185629045792e-05, + "loss": 1.1504, + "step": 14925 + }, + { + "epoch": 1.26, + "learning_rate": 1.3062320338530612e-05, + "loss": 1.21, + "step": 14926 + }, + { + "epoch": 1.26, + "learning_rate": 1.306145502271356e-05, + "loss": 1.2662, + "step": 14927 + }, + { + "epoch": 1.26, + "learning_rate": 1.306058968160179e-05, + "loss": 1.192, + "step": 14928 + }, + { + "epoch": 1.26, + "learning_rate": 1.3059724315202444e-05, + "loss": 1.2282, + "step": 14929 + }, + { + "epoch": 1.26, + "learning_rate": 1.3058858923522677e-05, + "loss": 1.1868, + "step": 14930 + }, + { + "epoch": 1.26, + "learning_rate": 1.3057993506569642e-05, + "loss": 1.1542, + "step": 14931 + }, + { + "epoch": 1.26, + "learning_rate": 1.3057128064350484e-05, + "loss": 1.1744, + "step": 14932 + }, + { + "epoch": 1.26, + "learning_rate": 1.3056262596872353e-05, + "loss": 1.2024, + "step": 14933 + }, + { + "epoch": 1.26, + "learning_rate": 1.3055397104142407e-05, + "loss": 1.1553, + "step": 14934 + }, + { + "epoch": 1.26, + "learning_rate": 1.3054531586167789e-05, + "loss": 1.1626, + "step": 14935 + }, + { + "epoch": 1.26, + "learning_rate": 1.3053666042955655e-05, + "loss": 1.1856, + "step": 14936 + }, + { + "epoch": 1.26, + "learning_rate": 1.3052800474513154e-05, + "loss": 1.202, + "step": 14937 + }, + { + "epoch": 1.26, + "learning_rate": 1.3051934880847438e-05, + "loss": 1.0969, + "step": 14938 + }, + { + "epoch": 1.26, + "learning_rate": 1.305106926196566e-05, + "loss": 1.1747, + "step": 14939 + }, + { + "epoch": 1.26, + "learning_rate": 1.3050203617874972e-05, + "loss": 1.151, + "step": 14940 + }, + { + "epoch": 1.26, + "learning_rate": 1.3049337948582526e-05, + "loss": 1.275, + "step": 14941 + }, + { + "epoch": 1.26, + "learning_rate": 1.3048472254095472e-05, + "loss": 1.1572, + "step": 14942 + }, + { + "epoch": 1.26, + "learning_rate": 1.3047606534420964e-05, + "loss": 1.1462, + "step": 14943 + }, + { + "epoch": 1.26, + "learning_rate": 1.3046740789566159e-05, + "loss": 1.191, + "step": 14944 + }, + { + "epoch": 1.26, + "learning_rate": 1.3045875019538206e-05, + "loss": 1.1852, + "step": 14945 + }, + { + "epoch": 1.26, + "learning_rate": 1.3045009224344255e-05, + "loss": 1.3036, + "step": 14946 + }, + { + "epoch": 1.26, + "learning_rate": 1.3044143403991463e-05, + "loss": 1.1903, + "step": 14947 + }, + { + "epoch": 1.26, + "learning_rate": 1.3043277558486989e-05, + "loss": 1.1323, + "step": 14948 + }, + { + "epoch": 1.26, + "learning_rate": 1.304241168783798e-05, + "loss": 1.1965, + "step": 14949 + }, + { + "epoch": 1.26, + "learning_rate": 1.3041545792051592e-05, + "loss": 1.1467, + "step": 14950 + }, + { + "epoch": 1.26, + "learning_rate": 1.3040679871134976e-05, + "loss": 1.2493, + "step": 14951 + }, + { + "epoch": 1.26, + "learning_rate": 1.3039813925095293e-05, + "loss": 1.1954, + "step": 14952 + }, + { + "epoch": 1.26, + "learning_rate": 1.3038947953939693e-05, + "loss": 1.1745, + "step": 14953 + }, + { + "epoch": 1.26, + "learning_rate": 1.3038081957675333e-05, + "loss": 1.2846, + "step": 14954 + }, + { + "epoch": 1.26, + "learning_rate": 1.3037215936309368e-05, + "loss": 1.2342, + "step": 14955 + }, + { + "epoch": 1.26, + "learning_rate": 1.3036349889848953e-05, + "loss": 1.2199, + "step": 14956 + }, + { + "epoch": 1.26, + "learning_rate": 1.3035483818301242e-05, + "loss": 1.2232, + "step": 14957 + }, + { + "epoch": 1.26, + "learning_rate": 1.3034617721673391e-05, + "loss": 1.2277, + "step": 14958 + }, + { + "epoch": 1.26, + "learning_rate": 1.3033751599972562e-05, + "loss": 1.1438, + "step": 14959 + }, + { + "epoch": 1.26, + "learning_rate": 1.3032885453205904e-05, + "loss": 1.1428, + "step": 14960 + }, + { + "epoch": 1.26, + "learning_rate": 1.3032019281380574e-05, + "loss": 1.3251, + "step": 14961 + }, + { + "epoch": 1.26, + "learning_rate": 1.3031153084503733e-05, + "loss": 1.1814, + "step": 14962 + }, + { + "epoch": 1.26, + "learning_rate": 1.3030286862582534e-05, + "loss": 1.1766, + "step": 14963 + }, + { + "epoch": 1.26, + "learning_rate": 1.3029420615624135e-05, + "loss": 1.1847, + "step": 14964 + }, + { + "epoch": 1.26, + "learning_rate": 1.3028554343635693e-05, + "loss": 1.2322, + "step": 14965 + }, + { + "epoch": 1.26, + "learning_rate": 1.3027688046624367e-05, + "loss": 1.2027, + "step": 14966 + }, + { + "epoch": 1.26, + "learning_rate": 1.3026821724597314e-05, + "loss": 1.1809, + "step": 14967 + }, + { + "epoch": 1.26, + "learning_rate": 1.3025955377561688e-05, + "loss": 1.2044, + "step": 14968 + }, + { + "epoch": 1.26, + "learning_rate": 1.3025089005524652e-05, + "loss": 1.1878, + "step": 14969 + }, + { + "epoch": 1.26, + "learning_rate": 1.3024222608493365e-05, + "loss": 1.2381, + "step": 14970 + }, + { + "epoch": 1.26, + "learning_rate": 1.3023356186474975e-05, + "loss": 1.1172, + "step": 14971 + }, + { + "epoch": 1.26, + "learning_rate": 1.3022489739476654e-05, + "loss": 1.1545, + "step": 14972 + }, + { + "epoch": 1.26, + "learning_rate": 1.3021623267505558e-05, + "loss": 1.2134, + "step": 14973 + }, + { + "epoch": 1.26, + "learning_rate": 1.302075677056884e-05, + "loss": 1.2171, + "step": 14974 + }, + { + "epoch": 1.26, + "learning_rate": 1.3019890248673664e-05, + "loss": 1.3083, + "step": 14975 + }, + { + "epoch": 1.26, + "learning_rate": 1.3019023701827187e-05, + "loss": 1.1542, + "step": 14976 + }, + { + "epoch": 1.26, + "learning_rate": 1.3018157130036573e-05, + "loss": 1.1603, + "step": 14977 + }, + { + "epoch": 1.26, + "learning_rate": 1.3017290533308977e-05, + "loss": 1.1056, + "step": 14978 + }, + { + "epoch": 1.26, + "learning_rate": 1.3016423911651562e-05, + "loss": 1.2281, + "step": 14979 + }, + { + "epoch": 1.26, + "learning_rate": 1.3015557265071483e-05, + "loss": 1.2628, + "step": 14980 + }, + { + "epoch": 1.26, + "learning_rate": 1.3014690593575909e-05, + "loss": 1.1355, + "step": 14981 + }, + { + "epoch": 1.26, + "learning_rate": 1.3013823897171996e-05, + "loss": 1.2368, + "step": 14982 + }, + { + "epoch": 1.26, + "learning_rate": 1.3012957175866903e-05, + "loss": 1.1506, + "step": 14983 + }, + { + "epoch": 1.26, + "learning_rate": 1.3012090429667796e-05, + "loss": 1.1839, + "step": 14984 + }, + { + "epoch": 1.26, + "learning_rate": 1.3011223658581833e-05, + "loss": 1.1771, + "step": 14985 + }, + { + "epoch": 1.26, + "learning_rate": 1.3010356862616176e-05, + "loss": 1.257, + "step": 14986 + }, + { + "epoch": 1.26, + "learning_rate": 1.3009490041777988e-05, + "loss": 1.1669, + "step": 14987 + }, + { + "epoch": 1.26, + "learning_rate": 1.300862319607443e-05, + "loss": 1.2657, + "step": 14988 + }, + { + "epoch": 1.26, + "learning_rate": 1.3007756325512663e-05, + "loss": 1.0839, + "step": 14989 + }, + { + "epoch": 1.26, + "learning_rate": 1.3006889430099853e-05, + "loss": 1.2045, + "step": 14990 + }, + { + "epoch": 1.26, + "learning_rate": 1.3006022509843162e-05, + "loss": 1.2095, + "step": 14991 + }, + { + "epoch": 1.26, + "learning_rate": 1.3005155564749745e-05, + "loss": 1.2283, + "step": 14992 + }, + { + "epoch": 1.26, + "learning_rate": 1.3004288594826775e-05, + "loss": 1.1784, + "step": 14993 + }, + { + "epoch": 1.26, + "learning_rate": 1.300342160008141e-05, + "loss": 1.1212, + "step": 14994 + }, + { + "epoch": 1.26, + "learning_rate": 1.3002554580520816e-05, + "loss": 1.2473, + "step": 14995 + }, + { + "epoch": 1.26, + "learning_rate": 1.300168753615215e-05, + "loss": 1.1432, + "step": 14996 + }, + { + "epoch": 1.26, + "learning_rate": 1.3000820466982585e-05, + "loss": 1.1513, + "step": 14997 + }, + { + "epoch": 1.26, + "learning_rate": 1.299995337301928e-05, + "loss": 1.2343, + "step": 14998 + }, + { + "epoch": 1.26, + "learning_rate": 1.2999086254269401e-05, + "loss": 1.2438, + "step": 14999 + }, + { + "epoch": 1.26, + "learning_rate": 1.299821911074011e-05, + "loss": 1.1589, + "step": 15000 + }, + { + "epoch": 1.26, + "learning_rate": 1.2997351942438576e-05, + "loss": 1.2388, + "step": 15001 + }, + { + "epoch": 1.26, + "learning_rate": 1.299648474937196e-05, + "loss": 1.1831, + "step": 15002 + }, + { + "epoch": 1.26, + "learning_rate": 1.2995617531547426e-05, + "loss": 1.1605, + "step": 15003 + }, + { + "epoch": 1.26, + "learning_rate": 1.2994750288972142e-05, + "loss": 1.2075, + "step": 15004 + }, + { + "epoch": 1.26, + "learning_rate": 1.2993883021653274e-05, + "loss": 1.2424, + "step": 15005 + }, + { + "epoch": 1.26, + "learning_rate": 1.2993015729597987e-05, + "loss": 1.1606, + "step": 15006 + }, + { + "epoch": 1.26, + "learning_rate": 1.2992148412813443e-05, + "loss": 1.2454, + "step": 15007 + }, + { + "epoch": 1.26, + "learning_rate": 1.2991281071306812e-05, + "loss": 1.1437, + "step": 15008 + }, + { + "epoch": 1.26, + "learning_rate": 1.2990413705085261e-05, + "loss": 1.1605, + "step": 15009 + }, + { + "epoch": 1.26, + "learning_rate": 1.2989546314155954e-05, + "loss": 1.1127, + "step": 15010 + }, + { + "epoch": 1.26, + "learning_rate": 1.2988678898526058e-05, + "loss": 1.2145, + "step": 15011 + }, + { + "epoch": 1.26, + "learning_rate": 1.2987811458202742e-05, + "loss": 1.1928, + "step": 15012 + }, + { + "epoch": 1.26, + "learning_rate": 1.298694399319317e-05, + "loss": 1.0992, + "step": 15013 + }, + { + "epoch": 1.26, + "learning_rate": 1.2986076503504512e-05, + "loss": 1.2755, + "step": 15014 + }, + { + "epoch": 1.26, + "learning_rate": 1.2985208989143933e-05, + "loss": 1.2577, + "step": 15015 + }, + { + "epoch": 1.26, + "learning_rate": 1.2984341450118605e-05, + "loss": 1.203, + "step": 15016 + }, + { + "epoch": 1.26, + "learning_rate": 1.2983473886435692e-05, + "loss": 1.2436, + "step": 15017 + }, + { + "epoch": 1.26, + "learning_rate": 1.2982606298102361e-05, + "loss": 1.1223, + "step": 15018 + }, + { + "epoch": 1.26, + "learning_rate": 1.2981738685125783e-05, + "loss": 1.1915, + "step": 15019 + }, + { + "epoch": 1.26, + "learning_rate": 1.2980871047513124e-05, + "loss": 1.2278, + "step": 15020 + }, + { + "epoch": 1.26, + "learning_rate": 1.2980003385271553e-05, + "loss": 1.2093, + "step": 15021 + }, + { + "epoch": 1.26, + "learning_rate": 1.2979135698408243e-05, + "loss": 1.1018, + "step": 15022 + }, + { + "epoch": 1.26, + "learning_rate": 1.2978267986930364e-05, + "loss": 1.1718, + "step": 15023 + }, + { + "epoch": 1.26, + "learning_rate": 1.2977400250845076e-05, + "loss": 1.1665, + "step": 15024 + }, + { + "epoch": 1.26, + "learning_rate": 1.2976532490159557e-05, + "loss": 1.1884, + "step": 15025 + }, + { + "epoch": 1.27, + "learning_rate": 1.2975664704880973e-05, + "loss": 1.2403, + "step": 15026 + }, + { + "epoch": 1.27, + "learning_rate": 1.2974796895016497e-05, + "loss": 1.1299, + "step": 15027 + }, + { + "epoch": 1.27, + "learning_rate": 1.2973929060573295e-05, + "loss": 1.2432, + "step": 15028 + }, + { + "epoch": 1.27, + "learning_rate": 1.2973061201558538e-05, + "loss": 1.1003, + "step": 15029 + }, + { + "epoch": 1.27, + "learning_rate": 1.29721933179794e-05, + "loss": 1.1788, + "step": 15030 + }, + { + "epoch": 1.27, + "learning_rate": 1.297132540984305e-05, + "loss": 1.1879, + "step": 15031 + }, + { + "epoch": 1.27, + "learning_rate": 1.2970457477156657e-05, + "loss": 1.2261, + "step": 15032 + }, + { + "epoch": 1.27, + "learning_rate": 1.2969589519927392e-05, + "loss": 1.0954, + "step": 15033 + }, + { + "epoch": 1.27, + "learning_rate": 1.2968721538162431e-05, + "loss": 1.18, + "step": 15034 + }, + { + "epoch": 1.27, + "learning_rate": 1.2967853531868943e-05, + "loss": 1.2364, + "step": 15035 + }, + { + "epoch": 1.27, + "learning_rate": 1.2966985501054097e-05, + "loss": 1.1401, + "step": 15036 + }, + { + "epoch": 1.27, + "learning_rate": 1.2966117445725069e-05, + "loss": 1.1386, + "step": 15037 + }, + { + "epoch": 1.27, + "learning_rate": 1.296524936588903e-05, + "loss": 1.2008, + "step": 15038 + }, + { + "epoch": 1.27, + "learning_rate": 1.296438126155315e-05, + "loss": 1.358, + "step": 15039 + }, + { + "epoch": 1.27, + "learning_rate": 1.2963513132724604e-05, + "loss": 1.1869, + "step": 15040 + }, + { + "epoch": 1.27, + "learning_rate": 1.2962644979410564e-05, + "loss": 1.206, + "step": 15041 + }, + { + "epoch": 1.27, + "learning_rate": 1.2961776801618203e-05, + "loss": 1.231, + "step": 15042 + }, + { + "epoch": 1.27, + "learning_rate": 1.2960908599354693e-05, + "loss": 1.1646, + "step": 15043 + }, + { + "epoch": 1.27, + "learning_rate": 1.296004037262721e-05, + "loss": 1.1854, + "step": 15044 + }, + { + "epoch": 1.27, + "learning_rate": 1.2959172121442925e-05, + "loss": 1.256, + "step": 15045 + }, + { + "epoch": 1.27, + "learning_rate": 1.2958303845809013e-05, + "loss": 1.2518, + "step": 15046 + }, + { + "epoch": 1.27, + "learning_rate": 1.2957435545732647e-05, + "loss": 1.118, + "step": 15047 + }, + { + "epoch": 1.27, + "learning_rate": 1.2956567221221006e-05, + "loss": 1.1921, + "step": 15048 + }, + { + "epoch": 1.27, + "learning_rate": 1.2955698872281258e-05, + "loss": 1.1296, + "step": 15049 + }, + { + "epoch": 1.27, + "learning_rate": 1.2954830498920577e-05, + "loss": 1.2086, + "step": 15050 + }, + { + "epoch": 1.27, + "learning_rate": 1.2953962101146145e-05, + "loss": 1.1638, + "step": 15051 + }, + { + "epoch": 1.27, + "learning_rate": 1.2953093678965131e-05, + "loss": 1.2256, + "step": 15052 + }, + { + "epoch": 1.27, + "learning_rate": 1.2952225232384712e-05, + "loss": 1.1822, + "step": 15053 + }, + { + "epoch": 1.27, + "learning_rate": 1.2951356761412064e-05, + "loss": 1.2027, + "step": 15054 + }, + { + "epoch": 1.27, + "learning_rate": 1.295048826605436e-05, + "loss": 1.2138, + "step": 15055 + }, + { + "epoch": 1.27, + "learning_rate": 1.294961974631878e-05, + "loss": 1.1913, + "step": 15056 + }, + { + "epoch": 1.27, + "learning_rate": 1.2948751202212497e-05, + "loss": 1.0693, + "step": 15057 + }, + { + "epoch": 1.27, + "learning_rate": 1.2947882633742683e-05, + "loss": 1.1824, + "step": 15058 + }, + { + "epoch": 1.27, + "learning_rate": 1.2947014040916526e-05, + "loss": 1.2513, + "step": 15059 + }, + { + "epoch": 1.27, + "learning_rate": 1.2946145423741194e-05, + "loss": 1.2308, + "step": 15060 + }, + { + "epoch": 1.27, + "learning_rate": 1.2945276782223863e-05, + "loss": 1.1039, + "step": 15061 + }, + { + "epoch": 1.27, + "learning_rate": 1.2944408116371716e-05, + "loss": 1.1768, + "step": 15062 + }, + { + "epoch": 1.27, + "learning_rate": 1.2943539426191926e-05, + "loss": 1.2187, + "step": 15063 + }, + { + "epoch": 1.27, + "learning_rate": 1.2942670711691672e-05, + "loss": 1.3322, + "step": 15064 + }, + { + "epoch": 1.27, + "learning_rate": 1.2941801972878128e-05, + "loss": 1.1752, + "step": 15065 + }, + { + "epoch": 1.27, + "learning_rate": 1.2940933209758479e-05, + "loss": 1.1861, + "step": 15066 + }, + { + "epoch": 1.27, + "learning_rate": 1.2940064422339896e-05, + "loss": 1.1904, + "step": 15067 + }, + { + "epoch": 1.27, + "learning_rate": 1.2939195610629558e-05, + "loss": 1.1937, + "step": 15068 + }, + { + "epoch": 1.27, + "learning_rate": 1.2938326774634646e-05, + "loss": 1.1161, + "step": 15069 + }, + { + "epoch": 1.27, + "learning_rate": 1.293745791436234e-05, + "loss": 1.2242, + "step": 15070 + }, + { + "epoch": 1.27, + "learning_rate": 1.2936589029819813e-05, + "loss": 1.2124, + "step": 15071 + }, + { + "epoch": 1.27, + "learning_rate": 1.293572012101425e-05, + "loss": 1.1181, + "step": 15072 + }, + { + "epoch": 1.27, + "learning_rate": 1.2934851187952828e-05, + "loss": 1.1738, + "step": 15073 + }, + { + "epoch": 1.27, + "learning_rate": 1.2933982230642726e-05, + "loss": 1.2089, + "step": 15074 + }, + { + "epoch": 1.27, + "learning_rate": 1.2933113249091123e-05, + "loss": 1.205, + "step": 15075 + }, + { + "epoch": 1.27, + "learning_rate": 1.2932244243305201e-05, + "loss": 1.109, + "step": 15076 + }, + { + "epoch": 1.27, + "learning_rate": 1.2931375213292138e-05, + "loss": 1.2492, + "step": 15077 + }, + { + "epoch": 1.27, + "learning_rate": 1.2930506159059115e-05, + "loss": 1.3067, + "step": 15078 + }, + { + "epoch": 1.27, + "learning_rate": 1.2929637080613312e-05, + "loss": 1.2911, + "step": 15079 + }, + { + "epoch": 1.27, + "learning_rate": 1.2928767977961908e-05, + "loss": 1.2712, + "step": 15080 + }, + { + "epoch": 1.27, + "learning_rate": 1.292789885111209e-05, + "loss": 1.1387, + "step": 15081 + }, + { + "epoch": 1.27, + "learning_rate": 1.2927029700071033e-05, + "loss": 1.254, + "step": 15082 + }, + { + "epoch": 1.27, + "learning_rate": 1.2926160524845914e-05, + "loss": 1.1842, + "step": 15083 + }, + { + "epoch": 1.27, + "learning_rate": 1.2925291325443928e-05, + "loss": 1.2118, + "step": 15084 + }, + { + "epoch": 1.27, + "learning_rate": 1.2924422101872246e-05, + "loss": 1.1675, + "step": 15085 + }, + { + "epoch": 1.27, + "learning_rate": 1.2923552854138052e-05, + "loss": 1.2191, + "step": 15086 + }, + { + "epoch": 1.27, + "learning_rate": 1.292268358224853e-05, + "loss": 1.2392, + "step": 15087 + }, + { + "epoch": 1.27, + "learning_rate": 1.2921814286210861e-05, + "loss": 1.2633, + "step": 15088 + }, + { + "epoch": 1.27, + "learning_rate": 1.2920944966032225e-05, + "loss": 1.1592, + "step": 15089 + }, + { + "epoch": 1.27, + "learning_rate": 1.2920075621719808e-05, + "loss": 1.2147, + "step": 15090 + }, + { + "epoch": 1.27, + "learning_rate": 1.2919206253280794e-05, + "loss": 1.1409, + "step": 15091 + }, + { + "epoch": 1.27, + "learning_rate": 1.291833686072236e-05, + "loss": 1.2357, + "step": 15092 + }, + { + "epoch": 1.27, + "learning_rate": 1.2917467444051694e-05, + "loss": 1.2416, + "step": 15093 + }, + { + "epoch": 1.27, + "learning_rate": 1.291659800327598e-05, + "loss": 1.215, + "step": 15094 + }, + { + "epoch": 1.27, + "learning_rate": 1.2915728538402399e-05, + "loss": 1.2134, + "step": 15095 + }, + { + "epoch": 1.27, + "learning_rate": 1.2914859049438134e-05, + "loss": 1.2091, + "step": 15096 + }, + { + "epoch": 1.27, + "learning_rate": 1.2913989536390372e-05, + "loss": 1.1036, + "step": 15097 + }, + { + "epoch": 1.27, + "learning_rate": 1.2913119999266296e-05, + "loss": 1.1731, + "step": 15098 + }, + { + "epoch": 1.27, + "learning_rate": 1.291225043807309e-05, + "loss": 1.0704, + "step": 15099 + }, + { + "epoch": 1.27, + "learning_rate": 1.2911380852817938e-05, + "loss": 1.1324, + "step": 15100 + }, + { + "epoch": 1.27, + "learning_rate": 1.2910511243508026e-05, + "loss": 1.1917, + "step": 15101 + }, + { + "epoch": 1.27, + "learning_rate": 1.290964161015054e-05, + "loss": 1.1923, + "step": 15102 + }, + { + "epoch": 1.27, + "learning_rate": 1.2908771952752664e-05, + "loss": 1.199, + "step": 15103 + }, + { + "epoch": 1.27, + "learning_rate": 1.2907902271321581e-05, + "loss": 1.2445, + "step": 15104 + }, + { + "epoch": 1.27, + "learning_rate": 1.2907032565864482e-05, + "loss": 1.2248, + "step": 15105 + }, + { + "epoch": 1.27, + "learning_rate": 1.2906162836388546e-05, + "loss": 1.1266, + "step": 15106 + }, + { + "epoch": 1.27, + "learning_rate": 1.2905293082900965e-05, + "loss": 1.1778, + "step": 15107 + }, + { + "epoch": 1.27, + "learning_rate": 1.290442330540892e-05, + "loss": 1.193, + "step": 15108 + }, + { + "epoch": 1.27, + "learning_rate": 1.2903553503919603e-05, + "loss": 1.2477, + "step": 15109 + }, + { + "epoch": 1.27, + "learning_rate": 1.2902683678440198e-05, + "loss": 1.264, + "step": 15110 + }, + { + "epoch": 1.27, + "learning_rate": 1.2901813828977889e-05, + "loss": 1.1208, + "step": 15111 + }, + { + "epoch": 1.27, + "learning_rate": 1.2900943955539867e-05, + "loss": 1.1883, + "step": 15112 + }, + { + "epoch": 1.27, + "learning_rate": 1.2900074058133319e-05, + "loss": 1.1052, + "step": 15113 + }, + { + "epoch": 1.27, + "learning_rate": 1.2899204136765429e-05, + "loss": 1.14, + "step": 15114 + }, + { + "epoch": 1.27, + "learning_rate": 1.2898334191443384e-05, + "loss": 1.0718, + "step": 15115 + }, + { + "epoch": 1.27, + "learning_rate": 1.289746422217438e-05, + "loss": 1.1908, + "step": 15116 + }, + { + "epoch": 1.27, + "learning_rate": 1.2896594228965596e-05, + "loss": 1.3233, + "step": 15117 + }, + { + "epoch": 1.27, + "learning_rate": 1.2895724211824224e-05, + "loss": 1.1736, + "step": 15118 + }, + { + "epoch": 1.27, + "learning_rate": 1.289485417075745e-05, + "loss": 1.1199, + "step": 15119 + }, + { + "epoch": 1.27, + "learning_rate": 1.2893984105772465e-05, + "loss": 1.1816, + "step": 15120 + }, + { + "epoch": 1.27, + "learning_rate": 1.2893114016876457e-05, + "loss": 1.158, + "step": 15121 + }, + { + "epoch": 1.27, + "learning_rate": 1.2892243904076618e-05, + "loss": 1.1548, + "step": 15122 + }, + { + "epoch": 1.27, + "learning_rate": 1.2891373767380133e-05, + "loss": 1.2752, + "step": 15123 + }, + { + "epoch": 1.27, + "learning_rate": 1.2890503606794191e-05, + "loss": 1.2692, + "step": 15124 + }, + { + "epoch": 1.27, + "learning_rate": 1.2889633422325984e-05, + "loss": 1.2536, + "step": 15125 + }, + { + "epoch": 1.27, + "learning_rate": 1.2888763213982701e-05, + "loss": 1.1689, + "step": 15126 + }, + { + "epoch": 1.27, + "learning_rate": 1.2887892981771532e-05, + "loss": 1.1352, + "step": 15127 + }, + { + "epoch": 1.27, + "learning_rate": 1.2887022725699668e-05, + "loss": 1.24, + "step": 15128 + }, + { + "epoch": 1.27, + "learning_rate": 1.2886152445774296e-05, + "loss": 1.1859, + "step": 15129 + }, + { + "epoch": 1.27, + "learning_rate": 1.2885282142002608e-05, + "loss": 1.2231, + "step": 15130 + }, + { + "epoch": 1.27, + "learning_rate": 1.2884411814391799e-05, + "loss": 1.1989, + "step": 15131 + }, + { + "epoch": 1.27, + "learning_rate": 1.2883541462949056e-05, + "loss": 1.1888, + "step": 15132 + }, + { + "epoch": 1.27, + "learning_rate": 1.2882671087681566e-05, + "loss": 1.1225, + "step": 15133 + }, + { + "epoch": 1.27, + "learning_rate": 1.2881800688596531e-05, + "loss": 1.1804, + "step": 15134 + }, + { + "epoch": 1.27, + "learning_rate": 1.2880930265701133e-05, + "loss": 1.198, + "step": 15135 + }, + { + "epoch": 1.27, + "learning_rate": 1.2880059819002569e-05, + "loss": 1.2056, + "step": 15136 + }, + { + "epoch": 1.27, + "learning_rate": 1.2879189348508025e-05, + "loss": 1.1887, + "step": 15137 + }, + { + "epoch": 1.27, + "learning_rate": 1.2878318854224703e-05, + "loss": 1.2386, + "step": 15138 + }, + { + "epoch": 1.27, + "learning_rate": 1.2877448336159787e-05, + "loss": 1.2464, + "step": 15139 + }, + { + "epoch": 1.27, + "learning_rate": 1.287657779432047e-05, + "loss": 1.1422, + "step": 15140 + }, + { + "epoch": 1.27, + "learning_rate": 1.2875707228713948e-05, + "loss": 1.2231, + "step": 15141 + }, + { + "epoch": 1.27, + "learning_rate": 1.2874836639347415e-05, + "loss": 1.2265, + "step": 15142 + }, + { + "epoch": 1.27, + "learning_rate": 1.2873966026228058e-05, + "loss": 1.2671, + "step": 15143 + }, + { + "epoch": 1.27, + "learning_rate": 1.2873095389363074e-05, + "loss": 1.2037, + "step": 15144 + }, + { + "epoch": 1.28, + "learning_rate": 1.2872224728759657e-05, + "loss": 1.1624, + "step": 15145 + }, + { + "epoch": 1.28, + "learning_rate": 1.2871354044425e-05, + "loss": 1.2245, + "step": 15146 + }, + { + "epoch": 1.28, + "learning_rate": 1.2870483336366297e-05, + "loss": 1.239, + "step": 15147 + }, + { + "epoch": 1.28, + "learning_rate": 1.2869612604590744e-05, + "loss": 1.2171, + "step": 15148 + }, + { + "epoch": 1.28, + "learning_rate": 1.286874184910553e-05, + "loss": 1.1933, + "step": 15149 + }, + { + "epoch": 1.28, + "learning_rate": 1.2867871069917857e-05, + "loss": 1.16, + "step": 15150 + }, + { + "epoch": 1.28, + "learning_rate": 1.286700026703491e-05, + "loss": 1.2289, + "step": 15151 + }, + { + "epoch": 1.28, + "learning_rate": 1.2866129440463895e-05, + "loss": 1.1819, + "step": 15152 + }, + { + "epoch": 1.28, + "learning_rate": 1.2865258590211997e-05, + "loss": 1.1346, + "step": 15153 + }, + { + "epoch": 1.28, + "learning_rate": 1.2864387716286417e-05, + "loss": 1.2275, + "step": 15154 + }, + { + "epoch": 1.28, + "learning_rate": 1.2863516818694349e-05, + "loss": 1.2039, + "step": 15155 + }, + { + "epoch": 1.28, + "learning_rate": 1.286264589744299e-05, + "loss": 1.2401, + "step": 15156 + }, + { + "epoch": 1.28, + "learning_rate": 1.2861774952539532e-05, + "loss": 1.1215, + "step": 15157 + }, + { + "epoch": 1.28, + "learning_rate": 1.286090398399117e-05, + "loss": 1.1624, + "step": 15158 + }, + { + "epoch": 1.28, + "learning_rate": 1.2860032991805111e-05, + "loss": 1.247, + "step": 15159 + }, + { + "epoch": 1.28, + "learning_rate": 1.285916197598854e-05, + "loss": 1.1771, + "step": 15160 + }, + { + "epoch": 1.28, + "learning_rate": 1.285829093654866e-05, + "loss": 1.1973, + "step": 15161 + }, + { + "epoch": 1.28, + "learning_rate": 1.2857419873492663e-05, + "loss": 1.1693, + "step": 15162 + }, + { + "epoch": 1.28, + "learning_rate": 1.2856548786827752e-05, + "loss": 1.1703, + "step": 15163 + }, + { + "epoch": 1.28, + "learning_rate": 1.2855677676561117e-05, + "loss": 1.334, + "step": 15164 + }, + { + "epoch": 1.28, + "learning_rate": 1.2854806542699963e-05, + "loss": 1.2495, + "step": 15165 + }, + { + "epoch": 1.28, + "learning_rate": 1.2853935385251479e-05, + "loss": 1.122, + "step": 15166 + }, + { + "epoch": 1.28, + "learning_rate": 1.2853064204222871e-05, + "loss": 1.2807, + "step": 15167 + }, + { + "epoch": 1.28, + "learning_rate": 1.2852192999621333e-05, + "loss": 1.204, + "step": 15168 + }, + { + "epoch": 1.28, + "learning_rate": 1.2851321771454062e-05, + "loss": 1.143, + "step": 15169 + }, + { + "epoch": 1.28, + "learning_rate": 1.285045051972826e-05, + "loss": 1.3262, + "step": 15170 + }, + { + "epoch": 1.28, + "learning_rate": 1.284957924445112e-05, + "loss": 1.15, + "step": 15171 + }, + { + "epoch": 1.28, + "learning_rate": 1.2848707945629846e-05, + "loss": 1.1436, + "step": 15172 + }, + { + "epoch": 1.28, + "learning_rate": 1.284783662327164e-05, + "loss": 1.2132, + "step": 15173 + }, + { + "epoch": 1.28, + "learning_rate": 1.284696527738369e-05, + "loss": 1.2045, + "step": 15174 + }, + { + "epoch": 1.28, + "learning_rate": 1.2846093907973205e-05, + "loss": 1.2415, + "step": 15175 + }, + { + "epoch": 1.28, + "learning_rate": 1.284522251504738e-05, + "loss": 1.1566, + "step": 15176 + }, + { + "epoch": 1.28, + "learning_rate": 1.2844351098613417e-05, + "loss": 1.276, + "step": 15177 + }, + { + "epoch": 1.28, + "learning_rate": 1.2843479658678515e-05, + "loss": 1.167, + "step": 15178 + }, + { + "epoch": 1.28, + "learning_rate": 1.2842608195249873e-05, + "loss": 1.2125, + "step": 15179 + }, + { + "epoch": 1.28, + "learning_rate": 1.2841736708334691e-05, + "loss": 1.1236, + "step": 15180 + }, + { + "epoch": 1.28, + "learning_rate": 1.2840865197940175e-05, + "loss": 1.141, + "step": 15181 + }, + { + "epoch": 1.28, + "learning_rate": 1.2839993664073517e-05, + "loss": 1.1668, + "step": 15182 + }, + { + "epoch": 1.28, + "learning_rate": 1.2839122106741924e-05, + "loss": 1.1673, + "step": 15183 + }, + { + "epoch": 1.28, + "learning_rate": 1.2838250525952597e-05, + "loss": 1.2163, + "step": 15184 + }, + { + "epoch": 1.28, + "learning_rate": 1.2837378921712734e-05, + "loss": 1.2058, + "step": 15185 + }, + { + "epoch": 1.28, + "learning_rate": 1.2836507294029538e-05, + "loss": 1.2153, + "step": 15186 + }, + { + "epoch": 1.28, + "learning_rate": 1.2835635642910211e-05, + "loss": 1.2808, + "step": 15187 + }, + { + "epoch": 1.28, + "learning_rate": 1.2834763968361958e-05, + "loss": 1.1607, + "step": 15188 + }, + { + "epoch": 1.28, + "learning_rate": 1.2833892270391974e-05, + "loss": 1.2367, + "step": 15189 + }, + { + "epoch": 1.28, + "learning_rate": 1.2833020549007466e-05, + "loss": 1.233, + "step": 15190 + }, + { + "epoch": 1.28, + "learning_rate": 1.2832148804215638e-05, + "loss": 1.243, + "step": 15191 + }, + { + "epoch": 1.28, + "learning_rate": 1.283127703602369e-05, + "loss": 1.298, + "step": 15192 + }, + { + "epoch": 1.28, + "learning_rate": 1.2830405244438823e-05, + "loss": 1.2041, + "step": 15193 + }, + { + "epoch": 1.28, + "learning_rate": 1.2829533429468242e-05, + "loss": 1.1407, + "step": 15194 + }, + { + "epoch": 1.28, + "learning_rate": 1.2828661591119152e-05, + "loss": 1.248, + "step": 15195 + }, + { + "epoch": 1.28, + "learning_rate": 1.2827789729398752e-05, + "loss": 1.1508, + "step": 15196 + }, + { + "epoch": 1.28, + "learning_rate": 1.2826917844314248e-05, + "loss": 1.1363, + "step": 15197 + }, + { + "epoch": 1.28, + "learning_rate": 1.2826045935872848e-05, + "loss": 1.253, + "step": 15198 + }, + { + "epoch": 1.28, + "learning_rate": 1.2825174004081748e-05, + "loss": 1.1971, + "step": 15199 + }, + { + "epoch": 1.28, + "learning_rate": 1.2824302048948158e-05, + "loss": 1.1475, + "step": 15200 + }, + { + "epoch": 1.28, + "learning_rate": 1.282343007047928e-05, + "loss": 1.2124, + "step": 15201 + }, + { + "epoch": 1.28, + "learning_rate": 1.2822558068682322e-05, + "loss": 1.1248, + "step": 15202 + }, + { + "epoch": 1.28, + "learning_rate": 1.2821686043564483e-05, + "loss": 1.1812, + "step": 15203 + }, + { + "epoch": 1.28, + "learning_rate": 1.282081399513297e-05, + "loss": 1.223, + "step": 15204 + }, + { + "epoch": 1.28, + "learning_rate": 1.2819941923394991e-05, + "loss": 1.2483, + "step": 15205 + }, + { + "epoch": 1.28, + "learning_rate": 1.281906982835775e-05, + "loss": 1.2068, + "step": 15206 + }, + { + "epoch": 1.28, + "learning_rate": 1.281819771002845e-05, + "loss": 1.2681, + "step": 15207 + }, + { + "epoch": 1.28, + "learning_rate": 1.2817325568414299e-05, + "loss": 1.2026, + "step": 15208 + }, + { + "epoch": 1.28, + "learning_rate": 1.2816453403522501e-05, + "loss": 1.2195, + "step": 15209 + }, + { + "epoch": 1.28, + "learning_rate": 1.2815581215360266e-05, + "loss": 1.106, + "step": 15210 + }, + { + "epoch": 1.28, + "learning_rate": 1.2814709003934796e-05, + "loss": 1.2451, + "step": 15211 + }, + { + "epoch": 1.28, + "learning_rate": 1.2813836769253301e-05, + "loss": 1.1046, + "step": 15212 + }, + { + "epoch": 1.28, + "learning_rate": 1.2812964511322988e-05, + "loss": 1.2656, + "step": 15213 + }, + { + "epoch": 1.28, + "learning_rate": 1.2812092230151058e-05, + "loss": 1.2508, + "step": 15214 + }, + { + "epoch": 1.28, + "learning_rate": 1.2811219925744723e-05, + "loss": 1.2265, + "step": 15215 + }, + { + "epoch": 1.28, + "learning_rate": 1.2810347598111191e-05, + "loss": 1.1848, + "step": 15216 + }, + { + "epoch": 1.28, + "learning_rate": 1.2809475247257667e-05, + "loss": 1.1701, + "step": 15217 + }, + { + "epoch": 1.28, + "learning_rate": 1.2808602873191358e-05, + "loss": 1.1488, + "step": 15218 + }, + { + "epoch": 1.28, + "learning_rate": 1.2807730475919473e-05, + "loss": 1.1078, + "step": 15219 + }, + { + "epoch": 1.28, + "learning_rate": 1.280685805544922e-05, + "loss": 1.2692, + "step": 15220 + }, + { + "epoch": 1.28, + "learning_rate": 1.280598561178781e-05, + "loss": 1.2032, + "step": 15221 + }, + { + "epoch": 1.28, + "learning_rate": 1.2805113144942448e-05, + "loss": 1.2009, + "step": 15222 + }, + { + "epoch": 1.28, + "learning_rate": 1.2804240654920343e-05, + "loss": 1.1982, + "step": 15223 + }, + { + "epoch": 1.28, + "learning_rate": 1.2803368141728705e-05, + "loss": 1.2126, + "step": 15224 + }, + { + "epoch": 1.28, + "learning_rate": 1.280249560537474e-05, + "loss": 1.2561, + "step": 15225 + }, + { + "epoch": 1.28, + "learning_rate": 1.2801623045865661e-05, + "loss": 1.2741, + "step": 15226 + }, + { + "epoch": 1.28, + "learning_rate": 1.2800750463208677e-05, + "loss": 1.2039, + "step": 15227 + }, + { + "epoch": 1.28, + "learning_rate": 1.2799877857410996e-05, + "loss": 1.1713, + "step": 15228 + }, + { + "epoch": 1.28, + "learning_rate": 1.2799005228479827e-05, + "loss": 1.2099, + "step": 15229 + }, + { + "epoch": 1.28, + "learning_rate": 1.279813257642238e-05, + "loss": 1.2604, + "step": 15230 + }, + { + "epoch": 1.28, + "learning_rate": 1.279725990124587e-05, + "loss": 1.2577, + "step": 15231 + }, + { + "epoch": 1.28, + "learning_rate": 1.27963872029575e-05, + "loss": 1.2375, + "step": 15232 + }, + { + "epoch": 1.28, + "learning_rate": 1.2795514481564481e-05, + "loss": 1.2187, + "step": 15233 + }, + { + "epoch": 1.28, + "learning_rate": 1.2794641737074032e-05, + "loss": 1.0838, + "step": 15234 + }, + { + "epoch": 1.28, + "learning_rate": 1.2793768969493356e-05, + "loss": 1.1542, + "step": 15235 + }, + { + "epoch": 1.28, + "learning_rate": 1.2792896178829667e-05, + "loss": 1.2207, + "step": 15236 + }, + { + "epoch": 1.28, + "learning_rate": 1.2792023365090176e-05, + "loss": 1.167, + "step": 15237 + }, + { + "epoch": 1.28, + "learning_rate": 1.2791150528282096e-05, + "loss": 1.2522, + "step": 15238 + }, + { + "epoch": 1.28, + "learning_rate": 1.2790277668412636e-05, + "loss": 1.1725, + "step": 15239 + }, + { + "epoch": 1.28, + "learning_rate": 1.2789404785489008e-05, + "loss": 1.2421, + "step": 15240 + }, + { + "epoch": 1.28, + "learning_rate": 1.2788531879518424e-05, + "loss": 1.2327, + "step": 15241 + }, + { + "epoch": 1.28, + "learning_rate": 1.2787658950508101e-05, + "loss": 1.2647, + "step": 15242 + }, + { + "epoch": 1.28, + "learning_rate": 1.2786785998465245e-05, + "loss": 1.1609, + "step": 15243 + }, + { + "epoch": 1.28, + "learning_rate": 1.2785913023397072e-05, + "loss": 1.147, + "step": 15244 + }, + { + "epoch": 1.28, + "learning_rate": 1.2785040025310793e-05, + "loss": 1.1166, + "step": 15245 + }, + { + "epoch": 1.28, + "learning_rate": 1.2784167004213623e-05, + "loss": 1.2281, + "step": 15246 + }, + { + "epoch": 1.28, + "learning_rate": 1.2783293960112773e-05, + "loss": 1.1115, + "step": 15247 + }, + { + "epoch": 1.28, + "learning_rate": 1.2782420893015458e-05, + "loss": 1.2106, + "step": 15248 + }, + { + "epoch": 1.28, + "learning_rate": 1.2781547802928894e-05, + "loss": 1.1967, + "step": 15249 + }, + { + "epoch": 1.28, + "learning_rate": 1.2780674689860288e-05, + "loss": 1.2624, + "step": 15250 + }, + { + "epoch": 1.28, + "learning_rate": 1.2779801553816857e-05, + "loss": 1.258, + "step": 15251 + }, + { + "epoch": 1.28, + "learning_rate": 1.2778928394805821e-05, + "loss": 1.1465, + "step": 15252 + }, + { + "epoch": 1.28, + "learning_rate": 1.2778055212834386e-05, + "loss": 1.2749, + "step": 15253 + }, + { + "epoch": 1.28, + "learning_rate": 1.277718200790977e-05, + "loss": 1.1267, + "step": 15254 + }, + { + "epoch": 1.28, + "learning_rate": 1.2776308780039185e-05, + "loss": 1.2416, + "step": 15255 + }, + { + "epoch": 1.28, + "learning_rate": 1.2775435529229854e-05, + "loss": 1.2068, + "step": 15256 + }, + { + "epoch": 1.28, + "learning_rate": 1.2774562255488983e-05, + "loss": 1.2128, + "step": 15257 + }, + { + "epoch": 1.28, + "learning_rate": 1.2773688958823786e-05, + "loss": 1.1083, + "step": 15258 + }, + { + "epoch": 1.28, + "learning_rate": 1.2772815639241491e-05, + "loss": 1.1927, + "step": 15259 + }, + { + "epoch": 1.28, + "learning_rate": 1.2771942296749303e-05, + "loss": 1.225, + "step": 15260 + }, + { + "epoch": 1.28, + "learning_rate": 1.277106893135444e-05, + "loss": 1.1502, + "step": 15261 + }, + { + "epoch": 1.28, + "learning_rate": 1.2770195543064117e-05, + "loss": 1.0879, + "step": 15262 + }, + { + "epoch": 1.29, + "learning_rate": 1.2769322131885557e-05, + "loss": 1.2389, + "step": 15263 + }, + { + "epoch": 1.29, + "learning_rate": 1.2768448697825968e-05, + "loss": 1.2006, + "step": 15264 + }, + { + "epoch": 1.29, + "learning_rate": 1.276757524089257e-05, + "loss": 1.2138, + "step": 15265 + }, + { + "epoch": 1.29, + "learning_rate": 1.2766701761092576e-05, + "loss": 1.124, + "step": 15266 + }, + { + "epoch": 1.29, + "learning_rate": 1.2765828258433212e-05, + "loss": 1.2147, + "step": 15267 + }, + { + "epoch": 1.29, + "learning_rate": 1.2764954732921688e-05, + "loss": 1.2222, + "step": 15268 + }, + { + "epoch": 1.29, + "learning_rate": 1.2764081184565223e-05, + "loss": 1.2633, + "step": 15269 + }, + { + "epoch": 1.29, + "learning_rate": 1.2763207613371034e-05, + "loss": 1.1223, + "step": 15270 + }, + { + "epoch": 1.29, + "learning_rate": 1.276233401934634e-05, + "loss": 1.1425, + "step": 15271 + }, + { + "epoch": 1.29, + "learning_rate": 1.2761460402498358e-05, + "loss": 1.093, + "step": 15272 + }, + { + "epoch": 1.29, + "learning_rate": 1.2760586762834309e-05, + "loss": 1.181, + "step": 15273 + }, + { + "epoch": 1.29, + "learning_rate": 1.2759713100361403e-05, + "loss": 1.2096, + "step": 15274 + }, + { + "epoch": 1.29, + "learning_rate": 1.2758839415086868e-05, + "loss": 1.2391, + "step": 15275 + }, + { + "epoch": 1.29, + "learning_rate": 1.2757965707017918e-05, + "loss": 1.181, + "step": 15276 + }, + { + "epoch": 1.29, + "learning_rate": 1.2757091976161776e-05, + "loss": 1.2669, + "step": 15277 + }, + { + "epoch": 1.29, + "learning_rate": 1.2756218222525654e-05, + "loss": 1.1525, + "step": 15278 + }, + { + "epoch": 1.29, + "learning_rate": 1.2755344446116773e-05, + "loss": 1.1664, + "step": 15279 + }, + { + "epoch": 1.29, + "learning_rate": 1.2754470646942356e-05, + "loss": 1.254, + "step": 15280 + }, + { + "epoch": 1.29, + "learning_rate": 1.2753596825009625e-05, + "loss": 1.2101, + "step": 15281 + }, + { + "epoch": 1.29, + "learning_rate": 1.2752722980325792e-05, + "loss": 1.111, + "step": 15282 + }, + { + "epoch": 1.29, + "learning_rate": 1.2751849112898078e-05, + "loss": 1.111, + "step": 15283 + }, + { + "epoch": 1.29, + "learning_rate": 1.275097522273371e-05, + "loss": 1.2209, + "step": 15284 + }, + { + "epoch": 1.29, + "learning_rate": 1.2750101309839905e-05, + "loss": 1.2268, + "step": 15285 + }, + { + "epoch": 1.29, + "learning_rate": 1.274922737422388e-05, + "loss": 1.2004, + "step": 15286 + }, + { + "epoch": 1.29, + "learning_rate": 1.274835341589286e-05, + "loss": 1.131, + "step": 15287 + }, + { + "epoch": 1.29, + "learning_rate": 1.2747479434854066e-05, + "loss": 1.2731, + "step": 15288 + }, + { + "epoch": 1.29, + "learning_rate": 1.2746605431114716e-05, + "loss": 1.2484, + "step": 15289 + }, + { + "epoch": 1.29, + "learning_rate": 1.274573140468203e-05, + "loss": 1.2497, + "step": 15290 + }, + { + "epoch": 1.29, + "learning_rate": 1.2744857355563238e-05, + "loss": 1.2318, + "step": 15291 + }, + { + "epoch": 1.29, + "learning_rate": 1.2743983283765555e-05, + "loss": 1.1402, + "step": 15292 + }, + { + "epoch": 1.29, + "learning_rate": 1.2743109189296202e-05, + "loss": 1.1745, + "step": 15293 + }, + { + "epoch": 1.29, + "learning_rate": 1.2742235072162403e-05, + "loss": 1.1743, + "step": 15294 + }, + { + "epoch": 1.29, + "learning_rate": 1.274136093237138e-05, + "loss": 1.1782, + "step": 15295 + }, + { + "epoch": 1.29, + "learning_rate": 1.2740486769930356e-05, + "loss": 1.2166, + "step": 15296 + }, + { + "epoch": 1.29, + "learning_rate": 1.2739612584846552e-05, + "loss": 1.2975, + "step": 15297 + }, + { + "epoch": 1.29, + "learning_rate": 1.2738738377127192e-05, + "loss": 1.1649, + "step": 15298 + }, + { + "epoch": 1.29, + "learning_rate": 1.2737864146779504e-05, + "loss": 1.1256, + "step": 15299 + }, + { + "epoch": 1.29, + "learning_rate": 1.2736989893810701e-05, + "loss": 1.2198, + "step": 15300 + }, + { + "epoch": 1.29, + "learning_rate": 1.2736115618228012e-05, + "loss": 1.1592, + "step": 15301 + }, + { + "epoch": 1.29, + "learning_rate": 1.2735241320038663e-05, + "loss": 1.1144, + "step": 15302 + }, + { + "epoch": 1.29, + "learning_rate": 1.2734366999249874e-05, + "loss": 1.1738, + "step": 15303 + }, + { + "epoch": 1.29, + "learning_rate": 1.2733492655868869e-05, + "loss": 1.2246, + "step": 15304 + }, + { + "epoch": 1.29, + "learning_rate": 1.2732618289902872e-05, + "loss": 1.2453, + "step": 15305 + }, + { + "epoch": 1.29, + "learning_rate": 1.2731743901359111e-05, + "loss": 1.208, + "step": 15306 + }, + { + "epoch": 1.29, + "learning_rate": 1.2730869490244803e-05, + "loss": 1.1748, + "step": 15307 + }, + { + "epoch": 1.29, + "learning_rate": 1.2729995056567177e-05, + "loss": 1.1998, + "step": 15308 + }, + { + "epoch": 1.29, + "learning_rate": 1.2729120600333462e-05, + "loss": 1.1551, + "step": 15309 + }, + { + "epoch": 1.29, + "learning_rate": 1.272824612155088e-05, + "loss": 1.1787, + "step": 15310 + }, + { + "epoch": 1.29, + "learning_rate": 1.2727371620226652e-05, + "loss": 1.1944, + "step": 15311 + }, + { + "epoch": 1.29, + "learning_rate": 1.2726497096368008e-05, + "loss": 1.1748, + "step": 15312 + }, + { + "epoch": 1.29, + "learning_rate": 1.2725622549982175e-05, + "loss": 1.2322, + "step": 15313 + }, + { + "epoch": 1.29, + "learning_rate": 1.2724747981076373e-05, + "loss": 1.1517, + "step": 15314 + }, + { + "epoch": 1.29, + "learning_rate": 1.272387338965783e-05, + "loss": 1.1844, + "step": 15315 + }, + { + "epoch": 1.29, + "learning_rate": 1.2722998775733774e-05, + "loss": 1.2371, + "step": 15316 + }, + { + "epoch": 1.29, + "learning_rate": 1.2722124139311436e-05, + "loss": 1.2419, + "step": 15317 + }, + { + "epoch": 1.29, + "learning_rate": 1.2721249480398033e-05, + "loss": 1.1383, + "step": 15318 + }, + { + "epoch": 1.29, + "learning_rate": 1.2720374799000794e-05, + "loss": 1.2564, + "step": 15319 + }, + { + "epoch": 1.29, + "learning_rate": 1.271950009512695e-05, + "loss": 1.18, + "step": 15320 + }, + { + "epoch": 1.29, + "learning_rate": 1.2718625368783726e-05, + "loss": 1.2316, + "step": 15321 + }, + { + "epoch": 1.29, + "learning_rate": 1.2717750619978348e-05, + "loss": 1.276, + "step": 15322 + }, + { + "epoch": 1.29, + "learning_rate": 1.2716875848718045e-05, + "loss": 1.2094, + "step": 15323 + }, + { + "epoch": 1.29, + "learning_rate": 1.2716001055010046e-05, + "loss": 1.2218, + "step": 15324 + }, + { + "epoch": 1.29, + "learning_rate": 1.2715126238861574e-05, + "loss": 1.14, + "step": 15325 + }, + { + "epoch": 1.29, + "learning_rate": 1.2714251400279862e-05, + "loss": 1.1578, + "step": 15326 + }, + { + "epoch": 1.29, + "learning_rate": 1.2713376539272137e-05, + "loss": 1.2131, + "step": 15327 + }, + { + "epoch": 1.29, + "learning_rate": 1.2712501655845626e-05, + "loss": 1.1733, + "step": 15328 + }, + { + "epoch": 1.29, + "learning_rate": 1.2711626750007557e-05, + "loss": 1.2011, + "step": 15329 + }, + { + "epoch": 1.29, + "learning_rate": 1.271075182176516e-05, + "loss": 1.2774, + "step": 15330 + }, + { + "epoch": 1.29, + "learning_rate": 1.2709876871125666e-05, + "loss": 1.3036, + "step": 15331 + }, + { + "epoch": 1.29, + "learning_rate": 1.27090018980963e-05, + "loss": 1.3389, + "step": 15332 + }, + { + "epoch": 1.29, + "learning_rate": 1.2708126902684292e-05, + "loss": 1.1999, + "step": 15333 + }, + { + "epoch": 1.29, + "learning_rate": 1.2707251884896876e-05, + "loss": 1.1812, + "step": 15334 + }, + { + "epoch": 1.29, + "learning_rate": 1.2706376844741278e-05, + "loss": 1.1567, + "step": 15335 + }, + { + "epoch": 1.29, + "learning_rate": 1.2705501782224725e-05, + "loss": 1.1667, + "step": 15336 + }, + { + "epoch": 1.29, + "learning_rate": 1.2704626697354455e-05, + "loss": 1.1794, + "step": 15337 + }, + { + "epoch": 1.29, + "learning_rate": 1.2703751590137691e-05, + "loss": 1.2957, + "step": 15338 + }, + { + "epoch": 1.29, + "learning_rate": 1.2702876460581667e-05, + "loss": 1.2604, + "step": 15339 + }, + { + "epoch": 1.29, + "learning_rate": 1.2702001308693613e-05, + "loss": 1.196, + "step": 15340 + }, + { + "epoch": 1.29, + "learning_rate": 1.2701126134480757e-05, + "loss": 1.3032, + "step": 15341 + }, + { + "epoch": 1.29, + "learning_rate": 1.2700250937950336e-05, + "loss": 1.1655, + "step": 15342 + }, + { + "epoch": 1.29, + "learning_rate": 1.2699375719109575e-05, + "loss": 1.2593, + "step": 15343 + }, + { + "epoch": 1.29, + "learning_rate": 1.269850047796571e-05, + "loss": 1.3026, + "step": 15344 + }, + { + "epoch": 1.29, + "learning_rate": 1.2697625214525968e-05, + "loss": 1.171, + "step": 15345 + }, + { + "epoch": 1.29, + "learning_rate": 1.2696749928797582e-05, + "loss": 1.1889, + "step": 15346 + }, + { + "epoch": 1.29, + "learning_rate": 1.269587462078779e-05, + "loss": 1.1565, + "step": 15347 + }, + { + "epoch": 1.29, + "learning_rate": 1.2694999290503814e-05, + "loss": 1.2725, + "step": 15348 + }, + { + "epoch": 1.29, + "learning_rate": 1.2694123937952895e-05, + "loss": 1.2149, + "step": 15349 + }, + { + "epoch": 1.29, + "learning_rate": 1.2693248563142259e-05, + "loss": 1.1813, + "step": 15350 + }, + { + "epoch": 1.29, + "learning_rate": 1.2692373166079144e-05, + "loss": 1.2275, + "step": 15351 + }, + { + "epoch": 1.29, + "learning_rate": 1.269149774677078e-05, + "loss": 1.1982, + "step": 15352 + }, + { + "epoch": 1.29, + "learning_rate": 1.2690622305224401e-05, + "loss": 1.1965, + "step": 15353 + }, + { + "epoch": 1.29, + "learning_rate": 1.2689746841447237e-05, + "loss": 1.1906, + "step": 15354 + }, + { + "epoch": 1.29, + "learning_rate": 1.2688871355446523e-05, + "loss": 1.1921, + "step": 15355 + }, + { + "epoch": 1.29, + "learning_rate": 1.2687995847229497e-05, + "loss": 1.118, + "step": 15356 + }, + { + "epoch": 1.29, + "learning_rate": 1.2687120316803388e-05, + "loss": 1.2789, + "step": 15357 + }, + { + "epoch": 1.29, + "learning_rate": 1.2686244764175426e-05, + "loss": 1.1877, + "step": 15358 + }, + { + "epoch": 1.29, + "learning_rate": 1.2685369189352854e-05, + "loss": 1.1708, + "step": 15359 + }, + { + "epoch": 1.29, + "learning_rate": 1.2684493592342904e-05, + "loss": 1.2433, + "step": 15360 + }, + { + "epoch": 1.29, + "learning_rate": 1.2683617973152807e-05, + "loss": 1.139, + "step": 15361 + }, + { + "epoch": 1.29, + "learning_rate": 1.2682742331789796e-05, + "loss": 1.2389, + "step": 15362 + }, + { + "epoch": 1.29, + "learning_rate": 1.2681866668261116e-05, + "loss": 1.17, + "step": 15363 + }, + { + "epoch": 1.29, + "learning_rate": 1.2680990982573991e-05, + "loss": 1.1669, + "step": 15364 + }, + { + "epoch": 1.29, + "learning_rate": 1.268011527473566e-05, + "loss": 1.139, + "step": 15365 + }, + { + "epoch": 1.29, + "learning_rate": 1.267923954475336e-05, + "loss": 1.2491, + "step": 15366 + }, + { + "epoch": 1.29, + "learning_rate": 1.2678363792634327e-05, + "loss": 1.1567, + "step": 15367 + }, + { + "epoch": 1.29, + "learning_rate": 1.2677488018385794e-05, + "loss": 1.1815, + "step": 15368 + }, + { + "epoch": 1.29, + "learning_rate": 1.2676612222014999e-05, + "loss": 1.1847, + "step": 15369 + }, + { + "epoch": 1.29, + "learning_rate": 1.2675736403529176e-05, + "loss": 1.2435, + "step": 15370 + }, + { + "epoch": 1.29, + "learning_rate": 1.2674860562935561e-05, + "loss": 1.2376, + "step": 15371 + }, + { + "epoch": 1.29, + "learning_rate": 1.2673984700241393e-05, + "loss": 1.1993, + "step": 15372 + }, + { + "epoch": 1.29, + "learning_rate": 1.2673108815453906e-05, + "loss": 1.1998, + "step": 15373 + }, + { + "epoch": 1.29, + "learning_rate": 1.2672232908580345e-05, + "loss": 1.1618, + "step": 15374 + }, + { + "epoch": 1.29, + "learning_rate": 1.2671356979627935e-05, + "loss": 1.1574, + "step": 15375 + }, + { + "epoch": 1.29, + "learning_rate": 1.2670481028603921e-05, + "loss": 1.1193, + "step": 15376 + }, + { + "epoch": 1.29, + "learning_rate": 1.2669605055515537e-05, + "loss": 1.2226, + "step": 15377 + }, + { + "epoch": 1.29, + "learning_rate": 1.2668729060370022e-05, + "loss": 1.1946, + "step": 15378 + }, + { + "epoch": 1.29, + "learning_rate": 1.2667853043174612e-05, + "loss": 1.2219, + "step": 15379 + }, + { + "epoch": 1.29, + "learning_rate": 1.2666977003936546e-05, + "loss": 1.2103, + "step": 15380 + }, + { + "epoch": 1.29, + "learning_rate": 1.2666100942663066e-05, + "loss": 1.2437, + "step": 15381 + }, + { + "epoch": 1.3, + "learning_rate": 1.2665224859361404e-05, + "loss": 1.1497, + "step": 15382 + }, + { + "epoch": 1.3, + "learning_rate": 1.26643487540388e-05, + "loss": 1.232, + "step": 15383 + }, + { + "epoch": 1.3, + "learning_rate": 1.2663472626702498e-05, + "loss": 1.1831, + "step": 15384 + }, + { + "epoch": 1.3, + "learning_rate": 1.266259647735973e-05, + "loss": 1.1465, + "step": 15385 + }, + { + "epoch": 1.3, + "learning_rate": 1.2661720306017738e-05, + "loss": 1.1694, + "step": 15386 + }, + { + "epoch": 1.3, + "learning_rate": 1.266084411268376e-05, + "loss": 1.1084, + "step": 15387 + }, + { + "epoch": 1.3, + "learning_rate": 1.265996789736504e-05, + "loss": 1.2018, + "step": 15388 + }, + { + "epoch": 1.3, + "learning_rate": 1.2659091660068809e-05, + "loss": 1.1881, + "step": 15389 + }, + { + "epoch": 1.3, + "learning_rate": 1.2658215400802314e-05, + "loss": 1.159, + "step": 15390 + }, + { + "epoch": 1.3, + "learning_rate": 1.2657339119572793e-05, + "loss": 1.224, + "step": 15391 + }, + { + "epoch": 1.3, + "learning_rate": 1.2656462816387488e-05, + "loss": 1.2171, + "step": 15392 + }, + { + "epoch": 1.3, + "learning_rate": 1.2655586491253634e-05, + "loss": 1.114, + "step": 15393 + }, + { + "epoch": 1.3, + "learning_rate": 1.2654710144178472e-05, + "loss": 1.1738, + "step": 15394 + }, + { + "epoch": 1.3, + "learning_rate": 1.2653833775169248e-05, + "loss": 1.1172, + "step": 15395 + }, + { + "epoch": 1.3, + "learning_rate": 1.2652957384233203e-05, + "loss": 1.2541, + "step": 15396 + }, + { + "epoch": 1.3, + "learning_rate": 1.265208097137757e-05, + "loss": 1.1572, + "step": 15397 + }, + { + "epoch": 1.3, + "learning_rate": 1.2651204536609597e-05, + "loss": 1.1911, + "step": 15398 + }, + { + "epoch": 1.3, + "learning_rate": 1.2650328079936527e-05, + "loss": 1.2881, + "step": 15399 + }, + { + "epoch": 1.3, + "learning_rate": 1.2649451601365594e-05, + "loss": 1.1684, + "step": 15400 + }, + { + "epoch": 1.3, + "learning_rate": 1.2648575100904044e-05, + "loss": 1.1597, + "step": 15401 + }, + { + "epoch": 1.3, + "learning_rate": 1.2647698578559123e-05, + "loss": 1.2418, + "step": 15402 + }, + { + "epoch": 1.3, + "learning_rate": 1.2646822034338064e-05, + "loss": 1.2496, + "step": 15403 + }, + { + "epoch": 1.3, + "learning_rate": 1.2645945468248118e-05, + "loss": 1.2869, + "step": 15404 + }, + { + "epoch": 1.3, + "learning_rate": 1.264506888029652e-05, + "loss": 1.2576, + "step": 15405 + }, + { + "epoch": 1.3, + "learning_rate": 1.264419227049052e-05, + "loss": 1.2278, + "step": 15406 + }, + { + "epoch": 1.3, + "learning_rate": 1.2643315638837357e-05, + "loss": 1.1547, + "step": 15407 + }, + { + "epoch": 1.3, + "learning_rate": 1.264243898534427e-05, + "loss": 1.1658, + "step": 15408 + }, + { + "epoch": 1.3, + "learning_rate": 1.264156231001851e-05, + "loss": 1.1694, + "step": 15409 + }, + { + "epoch": 1.3, + "learning_rate": 1.2640685612867316e-05, + "loss": 1.1539, + "step": 15410 + }, + { + "epoch": 1.3, + "learning_rate": 1.263980889389793e-05, + "loss": 1.2071, + "step": 15411 + }, + { + "epoch": 1.3, + "learning_rate": 1.2638932153117603e-05, + "loss": 1.2866, + "step": 15412 + }, + { + "epoch": 1.3, + "learning_rate": 1.2638055390533572e-05, + "loss": 1.1088, + "step": 15413 + }, + { + "epoch": 1.3, + "learning_rate": 1.2637178606153082e-05, + "loss": 1.2746, + "step": 15414 + }, + { + "epoch": 1.3, + "learning_rate": 1.263630179998338e-05, + "loss": 1.2286, + "step": 15415 + }, + { + "epoch": 1.3, + "learning_rate": 1.2635424972031705e-05, + "loss": 1.1351, + "step": 15416 + }, + { + "epoch": 1.3, + "learning_rate": 1.263454812230531e-05, + "loss": 1.1669, + "step": 15417 + }, + { + "epoch": 1.3, + "learning_rate": 1.2633671250811434e-05, + "loss": 1.1105, + "step": 15418 + }, + { + "epoch": 1.3, + "learning_rate": 1.2632794357557322e-05, + "loss": 1.2801, + "step": 15419 + }, + { + "epoch": 1.3, + "learning_rate": 1.263191744255022e-05, + "loss": 1.1218, + "step": 15420 + }, + { + "epoch": 1.3, + "learning_rate": 1.2631040505797375e-05, + "loss": 1.2426, + "step": 15421 + }, + { + "epoch": 1.3, + "learning_rate": 1.2630163547306031e-05, + "loss": 1.2082, + "step": 15422 + }, + { + "epoch": 1.3, + "learning_rate": 1.2629286567083435e-05, + "loss": 1.205, + "step": 15423 + }, + { + "epoch": 1.3, + "learning_rate": 1.2628409565136832e-05, + "loss": 1.2299, + "step": 15424 + }, + { + "epoch": 1.3, + "learning_rate": 1.2627532541473467e-05, + "loss": 1.3258, + "step": 15425 + }, + { + "epoch": 1.3, + "learning_rate": 1.2626655496100586e-05, + "loss": 1.1984, + "step": 15426 + }, + { + "epoch": 1.3, + "learning_rate": 1.2625778429025439e-05, + "loss": 1.2053, + "step": 15427 + }, + { + "epoch": 1.3, + "learning_rate": 1.2624901340255269e-05, + "loss": 1.1772, + "step": 15428 + }, + { + "epoch": 1.3, + "learning_rate": 1.2624024229797325e-05, + "loss": 1.2106, + "step": 15429 + }, + { + "epoch": 1.3, + "learning_rate": 1.262314709765885e-05, + "loss": 1.1813, + "step": 15430 + }, + { + "epoch": 1.3, + "learning_rate": 1.2622269943847098e-05, + "loss": 1.31, + "step": 15431 + }, + { + "epoch": 1.3, + "learning_rate": 1.2621392768369311e-05, + "loss": 1.2233, + "step": 15432 + }, + { + "epoch": 1.3, + "learning_rate": 1.2620515571232735e-05, + "loss": 1.2684, + "step": 15433 + }, + { + "epoch": 1.3, + "learning_rate": 1.2619638352444625e-05, + "loss": 1.1775, + "step": 15434 + }, + { + "epoch": 1.3, + "learning_rate": 1.2618761112012223e-05, + "loss": 1.1803, + "step": 15435 + }, + { + "epoch": 1.3, + "learning_rate": 1.2617883849942778e-05, + "loss": 1.2501, + "step": 15436 + }, + { + "epoch": 1.3, + "learning_rate": 1.2617006566243539e-05, + "loss": 1.1867, + "step": 15437 + }, + { + "epoch": 1.3, + "learning_rate": 1.2616129260921757e-05, + "loss": 1.2598, + "step": 15438 + }, + { + "epoch": 1.3, + "learning_rate": 1.2615251933984674e-05, + "loss": 1.1396, + "step": 15439 + }, + { + "epoch": 1.3, + "learning_rate": 1.2614374585439544e-05, + "loss": 1.1201, + "step": 15440 + }, + { + "epoch": 1.3, + "learning_rate": 1.2613497215293612e-05, + "loss": 1.1371, + "step": 15441 + }, + { + "epoch": 1.3, + "learning_rate": 1.2612619823554135e-05, + "loss": 1.1362, + "step": 15442 + }, + { + "epoch": 1.3, + "learning_rate": 1.2611742410228354e-05, + "loss": 1.1861, + "step": 15443 + }, + { + "epoch": 1.3, + "learning_rate": 1.261086497532352e-05, + "loss": 1.1942, + "step": 15444 + }, + { + "epoch": 1.3, + "learning_rate": 1.2609987518846887e-05, + "loss": 1.281, + "step": 15445 + }, + { + "epoch": 1.3, + "learning_rate": 1.2609110040805699e-05, + "loss": 1.2351, + "step": 15446 + }, + { + "epoch": 1.3, + "learning_rate": 1.2608232541207209e-05, + "loss": 1.1702, + "step": 15447 + }, + { + "epoch": 1.3, + "learning_rate": 1.2607355020058668e-05, + "loss": 1.2069, + "step": 15448 + }, + { + "epoch": 1.3, + "learning_rate": 1.2606477477367328e-05, + "loss": 1.25, + "step": 15449 + }, + { + "epoch": 1.3, + "learning_rate": 1.2605599913140432e-05, + "loss": 1.1898, + "step": 15450 + }, + { + "epoch": 1.3, + "learning_rate": 1.2604722327385237e-05, + "loss": 1.1882, + "step": 15451 + }, + { + "epoch": 1.3, + "learning_rate": 1.2603844720108994e-05, + "loss": 1.1997, + "step": 15452 + }, + { + "epoch": 1.3, + "learning_rate": 1.2602967091318953e-05, + "loss": 1.1541, + "step": 15453 + }, + { + "epoch": 1.3, + "learning_rate": 1.2602089441022362e-05, + "loss": 1.2181, + "step": 15454 + }, + { + "epoch": 1.3, + "learning_rate": 1.2601211769226478e-05, + "loss": 1.2209, + "step": 15455 + }, + { + "epoch": 1.3, + "learning_rate": 1.260033407593855e-05, + "loss": 1.1913, + "step": 15456 + }, + { + "epoch": 1.3, + "learning_rate": 1.2599456361165828e-05, + "loss": 1.0787, + "step": 15457 + }, + { + "epoch": 1.3, + "learning_rate": 1.2598578624915563e-05, + "loss": 1.1492, + "step": 15458 + }, + { + "epoch": 1.3, + "learning_rate": 1.2597700867195014e-05, + "loss": 1.2737, + "step": 15459 + }, + { + "epoch": 1.3, + "learning_rate": 1.2596823088011427e-05, + "loss": 1.1495, + "step": 15460 + }, + { + "epoch": 1.3, + "learning_rate": 1.259594528737206e-05, + "loss": 1.1306, + "step": 15461 + }, + { + "epoch": 1.3, + "learning_rate": 1.2595067465284158e-05, + "loss": 1.1661, + "step": 15462 + }, + { + "epoch": 1.3, + "learning_rate": 1.2594189621754982e-05, + "loss": 1.217, + "step": 15463 + }, + { + "epoch": 1.3, + "learning_rate": 1.2593311756791777e-05, + "loss": 1.1873, + "step": 15464 + }, + { + "epoch": 1.3, + "learning_rate": 1.25924338704018e-05, + "loss": 1.3141, + "step": 15465 + }, + { + "epoch": 1.3, + "learning_rate": 1.2591555962592307e-05, + "loss": 1.2592, + "step": 15466 + }, + { + "epoch": 1.3, + "learning_rate": 1.2590678033370551e-05, + "loss": 1.1916, + "step": 15467 + }, + { + "epoch": 1.3, + "learning_rate": 1.2589800082743783e-05, + "loss": 1.1466, + "step": 15468 + }, + { + "epoch": 1.3, + "learning_rate": 1.2588922110719257e-05, + "loss": 1.2135, + "step": 15469 + }, + { + "epoch": 1.3, + "learning_rate": 1.2588044117304228e-05, + "loss": 1.089, + "step": 15470 + }, + { + "epoch": 1.3, + "learning_rate": 1.2587166102505948e-05, + "loss": 1.2105, + "step": 15471 + }, + { + "epoch": 1.3, + "learning_rate": 1.2586288066331676e-05, + "loss": 1.1766, + "step": 15472 + }, + { + "epoch": 1.3, + "learning_rate": 1.2585410008788664e-05, + "loss": 1.1952, + "step": 15473 + }, + { + "epoch": 1.3, + "learning_rate": 1.2584531929884167e-05, + "loss": 1.1883, + "step": 15474 + }, + { + "epoch": 1.3, + "learning_rate": 1.2583653829625442e-05, + "loss": 1.2209, + "step": 15475 + }, + { + "epoch": 1.3, + "learning_rate": 1.2582775708019739e-05, + "loss": 1.1946, + "step": 15476 + }, + { + "epoch": 1.3, + "learning_rate": 1.2581897565074316e-05, + "loss": 1.2424, + "step": 15477 + }, + { + "epoch": 1.3, + "learning_rate": 1.2581019400796435e-05, + "loss": 1.2253, + "step": 15478 + }, + { + "epoch": 1.3, + "learning_rate": 1.258014121519334e-05, + "loss": 1.2163, + "step": 15479 + }, + { + "epoch": 1.3, + "learning_rate": 1.2579263008272294e-05, + "loss": 1.2182, + "step": 15480 + }, + { + "epoch": 1.3, + "learning_rate": 1.2578384780040553e-05, + "loss": 1.2554, + "step": 15481 + }, + { + "epoch": 1.3, + "learning_rate": 1.257750653050537e-05, + "loss": 1.218, + "step": 15482 + }, + { + "epoch": 1.3, + "learning_rate": 1.2576628259674002e-05, + "loss": 1.2026, + "step": 15483 + }, + { + "epoch": 1.3, + "learning_rate": 1.2575749967553707e-05, + "loss": 1.1249, + "step": 15484 + }, + { + "epoch": 1.3, + "learning_rate": 1.2574871654151744e-05, + "loss": 1.2301, + "step": 15485 + }, + { + "epoch": 1.3, + "learning_rate": 1.2573993319475365e-05, + "loss": 1.1006, + "step": 15486 + }, + { + "epoch": 1.3, + "learning_rate": 1.257311496353183e-05, + "loss": 1.2199, + "step": 15487 + }, + { + "epoch": 1.3, + "learning_rate": 1.2572236586328396e-05, + "loss": 1.1549, + "step": 15488 + }, + { + "epoch": 1.3, + "learning_rate": 1.2571358187872322e-05, + "loss": 1.13, + "step": 15489 + }, + { + "epoch": 1.3, + "learning_rate": 1.257047976817086e-05, + "loss": 1.162, + "step": 15490 + }, + { + "epoch": 1.3, + "learning_rate": 1.2569601327231274e-05, + "loss": 1.2366, + "step": 15491 + }, + { + "epoch": 1.3, + "learning_rate": 1.2568722865060821e-05, + "loss": 1.1737, + "step": 15492 + }, + { + "epoch": 1.3, + "learning_rate": 1.2567844381666753e-05, + "loss": 1.2274, + "step": 15493 + }, + { + "epoch": 1.3, + "learning_rate": 1.2566965877056336e-05, + "loss": 1.1935, + "step": 15494 + }, + { + "epoch": 1.3, + "learning_rate": 1.2566087351236824e-05, + "loss": 1.1605, + "step": 15495 + }, + { + "epoch": 1.3, + "learning_rate": 1.2565208804215479e-05, + "loss": 1.2188, + "step": 15496 + }, + { + "epoch": 1.3, + "learning_rate": 1.2564330235999555e-05, + "loss": 1.1824, + "step": 15497 + }, + { + "epoch": 1.3, + "learning_rate": 1.2563451646596314e-05, + "loss": 1.1903, + "step": 15498 + }, + { + "epoch": 1.3, + "learning_rate": 1.2562573036013019e-05, + "loss": 1.3253, + "step": 15499 + }, + { + "epoch": 1.3, + "learning_rate": 1.2561694404256924e-05, + "loss": 1.2447, + "step": 15500 + }, + { + "epoch": 1.31, + "learning_rate": 1.2560815751335286e-05, + "loss": 1.1971, + "step": 15501 + }, + { + "epoch": 1.31, + "learning_rate": 1.2559937077255372e-05, + "loss": 1.238, + "step": 15502 + }, + { + "epoch": 1.31, + "learning_rate": 1.2559058382024441e-05, + "loss": 1.1635, + "step": 15503 + }, + { + "epoch": 1.31, + "learning_rate": 1.2558179665649748e-05, + "loss": 1.2869, + "step": 15504 + }, + { + "epoch": 1.31, + "learning_rate": 1.2557300928138553e-05, + "loss": 1.2101, + "step": 15505 + }, + { + "epoch": 1.31, + "learning_rate": 1.2556422169498126e-05, + "loss": 1.2354, + "step": 15506 + }, + { + "epoch": 1.31, + "learning_rate": 1.2555543389735717e-05, + "loss": 1.2382, + "step": 15507 + }, + { + "epoch": 1.31, + "learning_rate": 1.2554664588858587e-05, + "loss": 1.2802, + "step": 15508 + }, + { + "epoch": 1.31, + "learning_rate": 1.2553785766874005e-05, + "loss": 1.151, + "step": 15509 + }, + { + "epoch": 1.31, + "learning_rate": 1.2552906923789228e-05, + "loss": 1.2147, + "step": 15510 + }, + { + "epoch": 1.31, + "learning_rate": 1.2552028059611513e-05, + "loss": 1.1623, + "step": 15511 + }, + { + "epoch": 1.31, + "learning_rate": 1.2551149174348129e-05, + "loss": 1.1652, + "step": 15512 + }, + { + "epoch": 1.31, + "learning_rate": 1.2550270268006334e-05, + "loss": 1.2245, + "step": 15513 + }, + { + "epoch": 1.31, + "learning_rate": 1.254939134059339e-05, + "loss": 1.1789, + "step": 15514 + }, + { + "epoch": 1.31, + "learning_rate": 1.2548512392116559e-05, + "loss": 1.124, + "step": 15515 + }, + { + "epoch": 1.31, + "learning_rate": 1.2547633422583101e-05, + "loss": 1.1762, + "step": 15516 + }, + { + "epoch": 1.31, + "learning_rate": 1.2546754432000282e-05, + "loss": 1.2903, + "step": 15517 + }, + { + "epoch": 1.31, + "learning_rate": 1.2545875420375362e-05, + "loss": 1.2128, + "step": 15518 + }, + { + "epoch": 1.31, + "learning_rate": 1.2544996387715605e-05, + "loss": 1.1363, + "step": 15519 + }, + { + "epoch": 1.31, + "learning_rate": 1.2544117334028272e-05, + "loss": 1.242, + "step": 15520 + }, + { + "epoch": 1.31, + "learning_rate": 1.2543238259320627e-05, + "loss": 1.2357, + "step": 15521 + }, + { + "epoch": 1.31, + "learning_rate": 1.2542359163599935e-05, + "loss": 1.146, + "step": 15522 + }, + { + "epoch": 1.31, + "learning_rate": 1.2541480046873457e-05, + "loss": 1.1445, + "step": 15523 + }, + { + "epoch": 1.31, + "learning_rate": 1.254060090914846e-05, + "loss": 1.2189, + "step": 15524 + }, + { + "epoch": 1.31, + "learning_rate": 1.2539721750432203e-05, + "loss": 1.2185, + "step": 15525 + }, + { + "epoch": 1.31, + "learning_rate": 1.2538842570731953e-05, + "loss": 1.1331, + "step": 15526 + }, + { + "epoch": 1.31, + "learning_rate": 1.2537963370054972e-05, + "loss": 1.2006, + "step": 15527 + }, + { + "epoch": 1.31, + "learning_rate": 1.2537084148408528e-05, + "loss": 1.1752, + "step": 15528 + }, + { + "epoch": 1.31, + "learning_rate": 1.2536204905799881e-05, + "loss": 1.1848, + "step": 15529 + }, + { + "epoch": 1.31, + "learning_rate": 1.2535325642236296e-05, + "loss": 1.2194, + "step": 15530 + }, + { + "epoch": 1.31, + "learning_rate": 1.253444635772504e-05, + "loss": 1.0497, + "step": 15531 + }, + { + "epoch": 1.31, + "learning_rate": 1.2533567052273379e-05, + "loss": 1.0176, + "step": 15532 + }, + { + "epoch": 1.31, + "learning_rate": 1.2532687725888573e-05, + "loss": 1.1883, + "step": 15533 + }, + { + "epoch": 1.31, + "learning_rate": 1.2531808378577892e-05, + "loss": 1.2241, + "step": 15534 + }, + { + "epoch": 1.31, + "learning_rate": 1.25309290103486e-05, + "loss": 1.2478, + "step": 15535 + }, + { + "epoch": 1.31, + "learning_rate": 1.2530049621207962e-05, + "loss": 1.2854, + "step": 15536 + }, + { + "epoch": 1.31, + "learning_rate": 1.2529170211163243e-05, + "loss": 1.1775, + "step": 15537 + }, + { + "epoch": 1.31, + "learning_rate": 1.2528290780221715e-05, + "loss": 1.2162, + "step": 15538 + }, + { + "epoch": 1.31, + "learning_rate": 1.2527411328390638e-05, + "loss": 1.2097, + "step": 15539 + }, + { + "epoch": 1.31, + "learning_rate": 1.2526531855677275e-05, + "loss": 1.2446, + "step": 15540 + }, + { + "epoch": 1.31, + "learning_rate": 1.25256523620889e-05, + "loss": 1.1108, + "step": 15541 + }, + { + "epoch": 1.31, + "learning_rate": 1.252477284763278e-05, + "loss": 1.2541, + "step": 15542 + }, + { + "epoch": 1.31, + "learning_rate": 1.2523893312316175e-05, + "loss": 1.1922, + "step": 15543 + }, + { + "epoch": 1.31, + "learning_rate": 1.2523013756146357e-05, + "loss": 1.2148, + "step": 15544 + }, + { + "epoch": 1.31, + "learning_rate": 1.2522134179130592e-05, + "loss": 1.1837, + "step": 15545 + }, + { + "epoch": 1.31, + "learning_rate": 1.2521254581276146e-05, + "loss": 1.1684, + "step": 15546 + }, + { + "epoch": 1.31, + "learning_rate": 1.2520374962590286e-05, + "loss": 1.1636, + "step": 15547 + }, + { + "epoch": 1.31, + "learning_rate": 1.2519495323080286e-05, + "loss": 1.1939, + "step": 15548 + }, + { + "epoch": 1.31, + "learning_rate": 1.2518615662753408e-05, + "loss": 1.2217, + "step": 15549 + }, + { + "epoch": 1.31, + "learning_rate": 1.251773598161692e-05, + "loss": 1.3069, + "step": 15550 + }, + { + "epoch": 1.31, + "learning_rate": 1.2516856279678092e-05, + "loss": 1.2686, + "step": 15551 + }, + { + "epoch": 1.31, + "learning_rate": 1.2515976556944191e-05, + "loss": 1.1755, + "step": 15552 + }, + { + "epoch": 1.31, + "learning_rate": 1.2515096813422489e-05, + "loss": 1.2183, + "step": 15553 + }, + { + "epoch": 1.31, + "learning_rate": 1.2514217049120251e-05, + "loss": 1.1463, + "step": 15554 + }, + { + "epoch": 1.31, + "learning_rate": 1.2513337264044748e-05, + "loss": 1.0903, + "step": 15555 + }, + { + "epoch": 1.31, + "learning_rate": 1.2512457458203247e-05, + "loss": 1.2728, + "step": 15556 + }, + { + "epoch": 1.31, + "learning_rate": 1.2511577631603018e-05, + "loss": 1.1989, + "step": 15557 + }, + { + "epoch": 1.31, + "learning_rate": 1.2510697784251328e-05, + "loss": 1.2632, + "step": 15558 + }, + { + "epoch": 1.31, + "learning_rate": 1.2509817916155455e-05, + "loss": 1.2166, + "step": 15559 + }, + { + "epoch": 1.31, + "learning_rate": 1.2508938027322661e-05, + "loss": 1.2343, + "step": 15560 + }, + { + "epoch": 1.31, + "learning_rate": 1.2508058117760217e-05, + "loss": 1.1592, + "step": 15561 + }, + { + "epoch": 1.31, + "learning_rate": 1.2507178187475396e-05, + "loss": 1.2031, + "step": 15562 + }, + { + "epoch": 1.31, + "learning_rate": 1.2506298236475465e-05, + "loss": 1.0541, + "step": 15563 + }, + { + "epoch": 1.31, + "learning_rate": 1.2505418264767696e-05, + "loss": 1.2921, + "step": 15564 + }, + { + "epoch": 1.31, + "learning_rate": 1.2504538272359359e-05, + "loss": 1.2215, + "step": 15565 + }, + { + "epoch": 1.31, + "learning_rate": 1.2503658259257726e-05, + "loss": 1.2469, + "step": 15566 + }, + { + "epoch": 1.31, + "learning_rate": 1.2502778225470069e-05, + "loss": 1.1947, + "step": 15567 + }, + { + "epoch": 1.31, + "learning_rate": 1.2501898171003654e-05, + "loss": 1.2128, + "step": 15568 + }, + { + "epoch": 1.31, + "learning_rate": 1.2501018095865756e-05, + "loss": 1.2366, + "step": 15569 + }, + { + "epoch": 1.31, + "learning_rate": 1.250013800006365e-05, + "loss": 1.235, + "step": 15570 + }, + { + "epoch": 1.31, + "learning_rate": 1.2499257883604599e-05, + "loss": 1.1939, + "step": 15571 + }, + { + "epoch": 1.31, + "learning_rate": 1.249837774649588e-05, + "loss": 1.1934, + "step": 15572 + }, + { + "epoch": 1.31, + "learning_rate": 1.2497497588744765e-05, + "loss": 1.2586, + "step": 15573 + }, + { + "epoch": 1.31, + "learning_rate": 1.249661741035853e-05, + "loss": 1.1703, + "step": 15574 + }, + { + "epoch": 1.31, + "learning_rate": 1.2495737211344437e-05, + "loss": 1.326, + "step": 15575 + }, + { + "epoch": 1.31, + "learning_rate": 1.2494856991709765e-05, + "loss": 1.1845, + "step": 15576 + }, + { + "epoch": 1.31, + "learning_rate": 1.2493976751461786e-05, + "loss": 1.1397, + "step": 15577 + }, + { + "epoch": 1.31, + "learning_rate": 1.249309649060778e-05, + "loss": 1.1477, + "step": 15578 + }, + { + "epoch": 1.31, + "learning_rate": 1.2492216209155006e-05, + "loss": 1.2623, + "step": 15579 + }, + { + "epoch": 1.31, + "learning_rate": 1.2491335907110744e-05, + "loss": 1.1291, + "step": 15580 + }, + { + "epoch": 1.31, + "learning_rate": 1.249045558448227e-05, + "loss": 1.1846, + "step": 15581 + }, + { + "epoch": 1.31, + "learning_rate": 1.2489575241276854e-05, + "loss": 1.2101, + "step": 15582 + }, + { + "epoch": 1.31, + "learning_rate": 1.2488694877501767e-05, + "loss": 1.2908, + "step": 15583 + }, + { + "epoch": 1.31, + "learning_rate": 1.2487814493164294e-05, + "loss": 1.1608, + "step": 15584 + }, + { + "epoch": 1.31, + "learning_rate": 1.2486934088271696e-05, + "loss": 1.1183, + "step": 15585 + }, + { + "epoch": 1.31, + "learning_rate": 1.2486053662831254e-05, + "loss": 1.1659, + "step": 15586 + }, + { + "epoch": 1.31, + "learning_rate": 1.2485173216850241e-05, + "loss": 1.3057, + "step": 15587 + }, + { + "epoch": 1.31, + "learning_rate": 1.2484292750335933e-05, + "loss": 1.2969, + "step": 15588 + }, + { + "epoch": 1.31, + "learning_rate": 1.2483412263295602e-05, + "loss": 1.204, + "step": 15589 + }, + { + "epoch": 1.31, + "learning_rate": 1.2482531755736522e-05, + "loss": 1.1058, + "step": 15590 + }, + { + "epoch": 1.31, + "learning_rate": 1.2481651227665972e-05, + "loss": 1.2198, + "step": 15591 + }, + { + "epoch": 1.31, + "learning_rate": 1.2480770679091227e-05, + "loss": 1.2245, + "step": 15592 + }, + { + "epoch": 1.31, + "learning_rate": 1.247989011001956e-05, + "loss": 1.2878, + "step": 15593 + }, + { + "epoch": 1.31, + "learning_rate": 1.2479009520458246e-05, + "loss": 1.284, + "step": 15594 + }, + { + "epoch": 1.31, + "learning_rate": 1.2478128910414563e-05, + "loss": 1.2353, + "step": 15595 + }, + { + "epoch": 1.31, + "learning_rate": 1.2477248279895782e-05, + "loss": 1.2742, + "step": 15596 + }, + { + "epoch": 1.31, + "learning_rate": 1.2476367628909188e-05, + "loss": 1.1975, + "step": 15597 + }, + { + "epoch": 1.31, + "learning_rate": 1.2475486957462048e-05, + "loss": 1.142, + "step": 15598 + }, + { + "epoch": 1.31, + "learning_rate": 1.2474606265561647e-05, + "loss": 1.1522, + "step": 15599 + }, + { + "epoch": 1.31, + "learning_rate": 1.2473725553215255e-05, + "loss": 1.152, + "step": 15600 + }, + { + "epoch": 1.31, + "learning_rate": 1.2472844820430149e-05, + "loss": 1.1967, + "step": 15601 + }, + { + "epoch": 1.31, + "learning_rate": 1.247196406721361e-05, + "loss": 1.2366, + "step": 15602 + }, + { + "epoch": 1.31, + "learning_rate": 1.2471083293572912e-05, + "loss": 1.1847, + "step": 15603 + }, + { + "epoch": 1.31, + "learning_rate": 1.2470202499515331e-05, + "loss": 1.3091, + "step": 15604 + }, + { + "epoch": 1.31, + "learning_rate": 1.2469321685048149e-05, + "loss": 1.1043, + "step": 15605 + }, + { + "epoch": 1.31, + "learning_rate": 1.246844085017864e-05, + "loss": 1.1439, + "step": 15606 + }, + { + "epoch": 1.31, + "learning_rate": 1.2467559994914085e-05, + "loss": 1.2064, + "step": 15607 + }, + { + "epoch": 1.31, + "learning_rate": 1.2466679119261756e-05, + "loss": 1.1268, + "step": 15608 + }, + { + "epoch": 1.31, + "learning_rate": 1.2465798223228933e-05, + "loss": 1.2548, + "step": 15609 + }, + { + "epoch": 1.31, + "learning_rate": 1.24649173068229e-05, + "loss": 1.2556, + "step": 15610 + }, + { + "epoch": 1.31, + "learning_rate": 1.2464036370050932e-05, + "loss": 1.1801, + "step": 15611 + }, + { + "epoch": 1.31, + "learning_rate": 1.2463155412920305e-05, + "loss": 1.2248, + "step": 15612 + }, + { + "epoch": 1.31, + "learning_rate": 1.2462274435438301e-05, + "loss": 1.2008, + "step": 15613 + }, + { + "epoch": 1.31, + "learning_rate": 1.2461393437612198e-05, + "loss": 1.172, + "step": 15614 + }, + { + "epoch": 1.31, + "learning_rate": 1.246051241944927e-05, + "loss": 1.26, + "step": 15615 + }, + { + "epoch": 1.31, + "learning_rate": 1.2459631380956805e-05, + "loss": 1.0409, + "step": 15616 + }, + { + "epoch": 1.31, + "learning_rate": 1.245875032214208e-05, + "loss": 1.148, + "step": 15617 + }, + { + "epoch": 1.31, + "learning_rate": 1.245786924301237e-05, + "loss": 1.0905, + "step": 15618 + }, + { + "epoch": 1.31, + "learning_rate": 1.2456988143574958e-05, + "loss": 1.1564, + "step": 15619 + }, + { + "epoch": 1.32, + "learning_rate": 1.2456107023837123e-05, + "loss": 1.1836, + "step": 15620 + }, + { + "epoch": 1.32, + "learning_rate": 1.2455225883806146e-05, + "loss": 1.2401, + "step": 15621 + }, + { + "epoch": 1.32, + "learning_rate": 1.2454344723489306e-05, + "loss": 1.2572, + "step": 15622 + }, + { + "epoch": 1.32, + "learning_rate": 1.2453463542893886e-05, + "loss": 1.2417, + "step": 15623 + }, + { + "epoch": 1.32, + "learning_rate": 1.2452582342027167e-05, + "loss": 1.1502, + "step": 15624 + }, + { + "epoch": 1.32, + "learning_rate": 1.2451701120896424e-05, + "loss": 1.1839, + "step": 15625 + }, + { + "epoch": 1.32, + "learning_rate": 1.2450819879508944e-05, + "loss": 1.2136, + "step": 15626 + }, + { + "epoch": 1.32, + "learning_rate": 1.2449938617872003e-05, + "loss": 1.2445, + "step": 15627 + }, + { + "epoch": 1.32, + "learning_rate": 1.2449057335992891e-05, + "loss": 1.2947, + "step": 15628 + }, + { + "epoch": 1.32, + "learning_rate": 1.2448176033878878e-05, + "loss": 1.2256, + "step": 15629 + }, + { + "epoch": 1.32, + "learning_rate": 1.2447294711537253e-05, + "loss": 1.2572, + "step": 15630 + }, + { + "epoch": 1.32, + "learning_rate": 1.2446413368975296e-05, + "loss": 1.1155, + "step": 15631 + }, + { + "epoch": 1.32, + "learning_rate": 1.2445532006200289e-05, + "loss": 1.2589, + "step": 15632 + }, + { + "epoch": 1.32, + "learning_rate": 1.2444650623219512e-05, + "loss": 1.1793, + "step": 15633 + }, + { + "epoch": 1.32, + "learning_rate": 1.244376922004025e-05, + "loss": 1.1444, + "step": 15634 + }, + { + "epoch": 1.32, + "learning_rate": 1.2442887796669786e-05, + "loss": 1.2225, + "step": 15635 + }, + { + "epoch": 1.32, + "learning_rate": 1.2442006353115398e-05, + "loss": 1.1509, + "step": 15636 + }, + { + "epoch": 1.32, + "learning_rate": 1.2441124889384375e-05, + "loss": 1.1968, + "step": 15637 + }, + { + "epoch": 1.32, + "learning_rate": 1.2440243405483999e-05, + "loss": 1.1627, + "step": 15638 + }, + { + "epoch": 1.32, + "learning_rate": 1.2439361901421548e-05, + "loss": 1.2494, + "step": 15639 + }, + { + "epoch": 1.32, + "learning_rate": 1.2438480377204307e-05, + "loss": 1.2818, + "step": 15640 + }, + { + "epoch": 1.32, + "learning_rate": 1.243759883283956e-05, + "loss": 1.2393, + "step": 15641 + }, + { + "epoch": 1.32, + "learning_rate": 1.2436717268334597e-05, + "loss": 1.1512, + "step": 15642 + }, + { + "epoch": 1.32, + "learning_rate": 1.2435835683696693e-05, + "loss": 1.2302, + "step": 15643 + }, + { + "epoch": 1.32, + "learning_rate": 1.2434954078933133e-05, + "loss": 1.0953, + "step": 15644 + }, + { + "epoch": 1.32, + "learning_rate": 1.2434072454051205e-05, + "loss": 1.1496, + "step": 15645 + }, + { + "epoch": 1.32, + "learning_rate": 1.2433190809058191e-05, + "loss": 1.1344, + "step": 15646 + }, + { + "epoch": 1.32, + "learning_rate": 1.2432309143961373e-05, + "loss": 1.2335, + "step": 15647 + }, + { + "epoch": 1.32, + "learning_rate": 1.2431427458768042e-05, + "loss": 1.2111, + "step": 15648 + }, + { + "epoch": 1.32, + "learning_rate": 1.243054575348548e-05, + "loss": 1.1637, + "step": 15649 + }, + { + "epoch": 1.32, + "learning_rate": 1.242966402812097e-05, + "loss": 1.0689, + "step": 15650 + }, + { + "epoch": 1.32, + "learning_rate": 1.2428782282681797e-05, + "loss": 1.2466, + "step": 15651 + }, + { + "epoch": 1.32, + "learning_rate": 1.2427900517175248e-05, + "loss": 1.1473, + "step": 15652 + }, + { + "epoch": 1.32, + "learning_rate": 1.2427018731608608e-05, + "loss": 1.2709, + "step": 15653 + }, + { + "epoch": 1.32, + "learning_rate": 1.2426136925989165e-05, + "loss": 1.2579, + "step": 15654 + }, + { + "epoch": 1.32, + "learning_rate": 1.2425255100324198e-05, + "loss": 1.1487, + "step": 15655 + }, + { + "epoch": 1.32, + "learning_rate": 1.2424373254620998e-05, + "loss": 1.2141, + "step": 15656 + }, + { + "epoch": 1.32, + "learning_rate": 1.2423491388886854e-05, + "loss": 1.2462, + "step": 15657 + }, + { + "epoch": 1.32, + "learning_rate": 1.2422609503129044e-05, + "loss": 1.1254, + "step": 15658 + }, + { + "epoch": 1.32, + "learning_rate": 1.242172759735486e-05, + "loss": 1.1726, + "step": 15659 + }, + { + "epoch": 1.32, + "learning_rate": 1.2420845671571588e-05, + "loss": 1.1586, + "step": 15660 + }, + { + "epoch": 1.32, + "learning_rate": 1.2419963725786517e-05, + "loss": 1.1705, + "step": 15661 + }, + { + "epoch": 1.32, + "learning_rate": 1.2419081760006928e-05, + "loss": 1.1748, + "step": 15662 + }, + { + "epoch": 1.32, + "learning_rate": 1.2418199774240115e-05, + "loss": 1.2677, + "step": 15663 + }, + { + "epoch": 1.32, + "learning_rate": 1.2417317768493357e-05, + "loss": 1.0955, + "step": 15664 + }, + { + "epoch": 1.32, + "learning_rate": 1.2416435742773946e-05, + "loss": 1.2356, + "step": 15665 + }, + { + "epoch": 1.32, + "learning_rate": 1.241555369708917e-05, + "loss": 1.2227, + "step": 15666 + }, + { + "epoch": 1.32, + "learning_rate": 1.241467163144632e-05, + "loss": 1.2792, + "step": 15667 + }, + { + "epoch": 1.32, + "learning_rate": 1.2413789545852678e-05, + "loss": 1.2352, + "step": 15668 + }, + { + "epoch": 1.32, + "learning_rate": 1.2412907440315534e-05, + "loss": 1.1841, + "step": 15669 + }, + { + "epoch": 1.32, + "learning_rate": 1.2412025314842176e-05, + "loss": 1.2359, + "step": 15670 + }, + { + "epoch": 1.32, + "learning_rate": 1.2411143169439893e-05, + "loss": 1.1548, + "step": 15671 + }, + { + "epoch": 1.32, + "learning_rate": 1.2410261004115974e-05, + "loss": 1.1549, + "step": 15672 + }, + { + "epoch": 1.32, + "learning_rate": 1.2409378818877704e-05, + "loss": 1.2157, + "step": 15673 + }, + { + "epoch": 1.32, + "learning_rate": 1.2408496613732381e-05, + "loss": 1.2553, + "step": 15674 + }, + { + "epoch": 1.32, + "learning_rate": 1.2407614388687286e-05, + "loss": 1.2463, + "step": 15675 + }, + { + "epoch": 1.32, + "learning_rate": 1.240673214374971e-05, + "loss": 1.1115, + "step": 15676 + }, + { + "epoch": 1.32, + "learning_rate": 1.2405849878926941e-05, + "loss": 1.3144, + "step": 15677 + }, + { + "epoch": 1.32, + "learning_rate": 1.2404967594226271e-05, + "loss": 1.1849, + "step": 15678 + }, + { + "epoch": 1.32, + "learning_rate": 1.2404085289654991e-05, + "loss": 1.2397, + "step": 15679 + }, + { + "epoch": 1.32, + "learning_rate": 1.2403202965220387e-05, + "loss": 1.1257, + "step": 15680 + }, + { + "epoch": 1.32, + "learning_rate": 1.2402320620929751e-05, + "loss": 1.1243, + "step": 15681 + }, + { + "epoch": 1.32, + "learning_rate": 1.2401438256790374e-05, + "loss": 1.1345, + "step": 15682 + }, + { + "epoch": 1.32, + "learning_rate": 1.2400555872809545e-05, + "loss": 1.0861, + "step": 15683 + }, + { + "epoch": 1.32, + "learning_rate": 1.2399673468994556e-05, + "loss": 1.2502, + "step": 15684 + }, + { + "epoch": 1.32, + "learning_rate": 1.2398791045352696e-05, + "loss": 1.2208, + "step": 15685 + }, + { + "epoch": 1.32, + "learning_rate": 1.2397908601891257e-05, + "loss": 1.2286, + "step": 15686 + }, + { + "epoch": 1.32, + "learning_rate": 1.239702613861753e-05, + "loss": 1.2, + "step": 15687 + }, + { + "epoch": 1.32, + "learning_rate": 1.2396143655538806e-05, + "loss": 1.1609, + "step": 15688 + }, + { + "epoch": 1.32, + "learning_rate": 1.2395261152662377e-05, + "loss": 1.15, + "step": 15689 + }, + { + "epoch": 1.32, + "learning_rate": 1.2394378629995531e-05, + "loss": 1.2603, + "step": 15690 + }, + { + "epoch": 1.32, + "learning_rate": 1.2393496087545565e-05, + "loss": 1.2582, + "step": 15691 + }, + { + "epoch": 1.32, + "learning_rate": 1.2392613525319767e-05, + "loss": 1.2177, + "step": 15692 + }, + { + "epoch": 1.32, + "learning_rate": 1.2391730943325432e-05, + "loss": 1.1298, + "step": 15693 + }, + { + "epoch": 1.32, + "learning_rate": 1.2390848341569847e-05, + "loss": 1.2136, + "step": 15694 + }, + { + "epoch": 1.32, + "learning_rate": 1.238996572006031e-05, + "loss": 1.2269, + "step": 15695 + }, + { + "epoch": 1.32, + "learning_rate": 1.2389083078804112e-05, + "loss": 1.2092, + "step": 15696 + }, + { + "epoch": 1.32, + "learning_rate": 1.2388200417808545e-05, + "loss": 1.2152, + "step": 15697 + }, + { + "epoch": 1.32, + "learning_rate": 1.2387317737080899e-05, + "loss": 1.2234, + "step": 15698 + }, + { + "epoch": 1.32, + "learning_rate": 1.2386435036628474e-05, + "loss": 1.1611, + "step": 15699 + }, + { + "epoch": 1.32, + "learning_rate": 1.2385552316458557e-05, + "loss": 1.1933, + "step": 15700 + }, + { + "epoch": 1.32, + "learning_rate": 1.2384669576578443e-05, + "loss": 1.2295, + "step": 15701 + }, + { + "epoch": 1.32, + "learning_rate": 1.2383786816995428e-05, + "loss": 1.1141, + "step": 15702 + }, + { + "epoch": 1.32, + "learning_rate": 1.23829040377168e-05, + "loss": 1.1785, + "step": 15703 + }, + { + "epoch": 1.32, + "learning_rate": 1.2382021238749857e-05, + "loss": 1.2228, + "step": 15704 + }, + { + "epoch": 1.32, + "learning_rate": 1.2381138420101895e-05, + "loss": 1.0972, + "step": 15705 + }, + { + "epoch": 1.32, + "learning_rate": 1.23802555817802e-05, + "loss": 1.3415, + "step": 15706 + }, + { + "epoch": 1.32, + "learning_rate": 1.2379372723792078e-05, + "loss": 1.186, + "step": 15707 + }, + { + "epoch": 1.32, + "learning_rate": 1.2378489846144813e-05, + "loss": 1.1842, + "step": 15708 + }, + { + "epoch": 1.32, + "learning_rate": 1.2377606948845702e-05, + "loss": 1.1494, + "step": 15709 + }, + { + "epoch": 1.32, + "learning_rate": 1.2376724031902046e-05, + "loss": 1.2171, + "step": 15710 + }, + { + "epoch": 1.32, + "learning_rate": 1.2375841095321132e-05, + "loss": 1.2563, + "step": 15711 + }, + { + "epoch": 1.32, + "learning_rate": 1.237495813911026e-05, + "loss": 1.1659, + "step": 15712 + }, + { + "epoch": 1.32, + "learning_rate": 1.2374075163276726e-05, + "loss": 1.2201, + "step": 15713 + }, + { + "epoch": 1.32, + "learning_rate": 1.2373192167827818e-05, + "loss": 1.2551, + "step": 15714 + }, + { + "epoch": 1.32, + "learning_rate": 1.237230915277084e-05, + "loss": 1.1587, + "step": 15715 + }, + { + "epoch": 1.32, + "learning_rate": 1.2371426118113086e-05, + "loss": 1.1892, + "step": 15716 + }, + { + "epoch": 1.32, + "learning_rate": 1.2370543063861848e-05, + "loss": 1.2282, + "step": 15717 + }, + { + "epoch": 1.32, + "learning_rate": 1.2369659990024425e-05, + "loss": 1.1894, + "step": 15718 + }, + { + "epoch": 1.32, + "learning_rate": 1.2368776896608114e-05, + "loss": 1.249, + "step": 15719 + }, + { + "epoch": 1.32, + "learning_rate": 1.2367893783620208e-05, + "loss": 1.1626, + "step": 15720 + }, + { + "epoch": 1.32, + "learning_rate": 1.2367010651068005e-05, + "loss": 1.2272, + "step": 15721 + }, + { + "epoch": 1.32, + "learning_rate": 1.2366127498958807e-05, + "loss": 1.2294, + "step": 15722 + }, + { + "epoch": 1.32, + "learning_rate": 1.23652443272999e-05, + "loss": 1.1374, + "step": 15723 + }, + { + "epoch": 1.32, + "learning_rate": 1.2364361136098596e-05, + "loss": 1.1859, + "step": 15724 + }, + { + "epoch": 1.32, + "learning_rate": 1.2363477925362177e-05, + "loss": 1.224, + "step": 15725 + }, + { + "epoch": 1.32, + "learning_rate": 1.2362594695097947e-05, + "loss": 1.1706, + "step": 15726 + }, + { + "epoch": 1.32, + "learning_rate": 1.2361711445313205e-05, + "loss": 1.1728, + "step": 15727 + }, + { + "epoch": 1.32, + "learning_rate": 1.236082817601525e-05, + "loss": 1.2121, + "step": 15728 + }, + { + "epoch": 1.32, + "learning_rate": 1.2359944887211373e-05, + "loss": 1.0779, + "step": 15729 + }, + { + "epoch": 1.32, + "learning_rate": 1.2359061578908878e-05, + "loss": 1.143, + "step": 15730 + }, + { + "epoch": 1.32, + "learning_rate": 1.235817825111506e-05, + "loss": 1.0887, + "step": 15731 + }, + { + "epoch": 1.32, + "learning_rate": 1.2357294903837221e-05, + "loss": 1.1934, + "step": 15732 + }, + { + "epoch": 1.32, + "learning_rate": 1.2356411537082652e-05, + "loss": 1.1609, + "step": 15733 + }, + { + "epoch": 1.32, + "learning_rate": 1.235552815085866e-05, + "loss": 1.1711, + "step": 15734 + }, + { + "epoch": 1.32, + "learning_rate": 1.2354644745172542e-05, + "loss": 1.211, + "step": 15735 + }, + { + "epoch": 1.32, + "learning_rate": 1.2353761320031594e-05, + "loss": 1.1534, + "step": 15736 + }, + { + "epoch": 1.32, + "learning_rate": 1.2352877875443117e-05, + "loss": 1.1768, + "step": 15737 + }, + { + "epoch": 1.33, + "learning_rate": 1.2351994411414409e-05, + "loss": 1.2518, + "step": 15738 + }, + { + "epoch": 1.33, + "learning_rate": 1.2351110927952771e-05, + "loss": 1.1838, + "step": 15739 + }, + { + "epoch": 1.33, + "learning_rate": 1.2350227425065503e-05, + "loss": 1.2092, + "step": 15740 + }, + { + "epoch": 1.33, + "learning_rate": 1.2349343902759902e-05, + "loss": 1.1068, + "step": 15741 + }, + { + "epoch": 1.33, + "learning_rate": 1.2348460361043273e-05, + "loss": 1.0967, + "step": 15742 + }, + { + "epoch": 1.33, + "learning_rate": 1.234757679992291e-05, + "loss": 1.2957, + "step": 15743 + }, + { + "epoch": 1.33, + "learning_rate": 1.2346693219406118e-05, + "loss": 1.1413, + "step": 15744 + }, + { + "epoch": 1.33, + "learning_rate": 1.2345809619500193e-05, + "loss": 1.2281, + "step": 15745 + }, + { + "epoch": 1.33, + "learning_rate": 1.234492600021244e-05, + "loss": 1.1405, + "step": 15746 + }, + { + "epoch": 1.33, + "learning_rate": 1.2344042361550158e-05, + "loss": 1.1318, + "step": 15747 + }, + { + "epoch": 1.33, + "learning_rate": 1.2343158703520645e-05, + "loss": 1.1671, + "step": 15748 + }, + { + "epoch": 1.33, + "learning_rate": 1.2342275026131209e-05, + "loss": 1.2761, + "step": 15749 + }, + { + "epoch": 1.33, + "learning_rate": 1.2341391329389147e-05, + "loss": 1.138, + "step": 15750 + }, + { + "epoch": 1.33, + "learning_rate": 1.2340507613301758e-05, + "loss": 1.191, + "step": 15751 + }, + { + "epoch": 1.33, + "learning_rate": 1.2339623877876345e-05, + "loss": 1.1927, + "step": 15752 + }, + { + "epoch": 1.33, + "learning_rate": 1.2338740123120217e-05, + "loss": 1.2104, + "step": 15753 + }, + { + "epoch": 1.33, + "learning_rate": 1.2337856349040665e-05, + "loss": 1.1005, + "step": 15754 + }, + { + "epoch": 1.33, + "learning_rate": 1.2336972555644995e-05, + "loss": 1.1272, + "step": 15755 + }, + { + "epoch": 1.33, + "learning_rate": 1.233608874294051e-05, + "loss": 1.1802, + "step": 15756 + }, + { + "epoch": 1.33, + "learning_rate": 1.2335204910934515e-05, + "loss": 1.2152, + "step": 15757 + }, + { + "epoch": 1.33, + "learning_rate": 1.2334321059634305e-05, + "loss": 1.2672, + "step": 15758 + }, + { + "epoch": 1.33, + "learning_rate": 1.2333437189047189e-05, + "loss": 1.1975, + "step": 15759 + }, + { + "epoch": 1.33, + "learning_rate": 1.233255329918047e-05, + "loss": 1.2724, + "step": 15760 + }, + { + "epoch": 1.33, + "learning_rate": 1.233166939004145e-05, + "loss": 1.1228, + "step": 15761 + }, + { + "epoch": 1.33, + "learning_rate": 1.2330785461637426e-05, + "loss": 1.0927, + "step": 15762 + }, + { + "epoch": 1.33, + "learning_rate": 1.2329901513975711e-05, + "loss": 1.1671, + "step": 15763 + }, + { + "epoch": 1.33, + "learning_rate": 1.2329017547063607e-05, + "loss": 1.2825, + "step": 15764 + }, + { + "epoch": 1.33, + "learning_rate": 1.232813356090841e-05, + "loss": 1.1746, + "step": 15765 + }, + { + "epoch": 1.33, + "learning_rate": 1.2327249555517426e-05, + "loss": 1.1184, + "step": 15766 + }, + { + "epoch": 1.33, + "learning_rate": 1.2326365530897968e-05, + "loss": 1.2074, + "step": 15767 + }, + { + "epoch": 1.33, + "learning_rate": 1.232548148705733e-05, + "loss": 1.1914, + "step": 15768 + }, + { + "epoch": 1.33, + "learning_rate": 1.2324597424002817e-05, + "loss": 1.1866, + "step": 15769 + }, + { + "epoch": 1.33, + "learning_rate": 1.2323713341741739e-05, + "loss": 1.2308, + "step": 15770 + }, + { + "epoch": 1.33, + "learning_rate": 1.2322829240281398e-05, + "loss": 1.1167, + "step": 15771 + }, + { + "epoch": 1.33, + "learning_rate": 1.2321945119629097e-05, + "loss": 1.1543, + "step": 15772 + }, + { + "epoch": 1.33, + "learning_rate": 1.2321060979792143e-05, + "loss": 1.1806, + "step": 15773 + }, + { + "epoch": 1.33, + "learning_rate": 1.2320176820777843e-05, + "loss": 1.2077, + "step": 15774 + }, + { + "epoch": 1.33, + "learning_rate": 1.2319292642593497e-05, + "loss": 1.155, + "step": 15775 + }, + { + "epoch": 1.33, + "learning_rate": 1.231840844524641e-05, + "loss": 1.2086, + "step": 15776 + }, + { + "epoch": 1.33, + "learning_rate": 1.2317524228743893e-05, + "loss": 1.0963, + "step": 15777 + }, + { + "epoch": 1.33, + "learning_rate": 1.2316639993093254e-05, + "loss": 1.2109, + "step": 15778 + }, + { + "epoch": 1.33, + "learning_rate": 1.231575573830179e-05, + "loss": 1.2204, + "step": 15779 + }, + { + "epoch": 1.33, + "learning_rate": 1.2314871464376808e-05, + "loss": 1.19, + "step": 15780 + }, + { + "epoch": 1.33, + "learning_rate": 1.231398717132562e-05, + "loss": 1.2377, + "step": 15781 + }, + { + "epoch": 1.33, + "learning_rate": 1.231310285915553e-05, + "loss": 1.2243, + "step": 15782 + }, + { + "epoch": 1.33, + "learning_rate": 1.2312218527873841e-05, + "loss": 1.1732, + "step": 15783 + }, + { + "epoch": 1.33, + "learning_rate": 1.2311334177487863e-05, + "loss": 1.1879, + "step": 15784 + }, + { + "epoch": 1.33, + "learning_rate": 1.2310449808004905e-05, + "loss": 1.0088, + "step": 15785 + }, + { + "epoch": 1.33, + "learning_rate": 1.230956541943227e-05, + "loss": 1.2291, + "step": 15786 + }, + { + "epoch": 1.33, + "learning_rate": 1.2308681011777266e-05, + "loss": 1.1195, + "step": 15787 + }, + { + "epoch": 1.33, + "learning_rate": 1.23077965850472e-05, + "loss": 1.1263, + "step": 15788 + }, + { + "epoch": 1.33, + "learning_rate": 1.2306912139249383e-05, + "loss": 1.0878, + "step": 15789 + }, + { + "epoch": 1.33, + "learning_rate": 1.2306027674391116e-05, + "loss": 1.1519, + "step": 15790 + }, + { + "epoch": 1.33, + "learning_rate": 1.2305143190479712e-05, + "loss": 1.237, + "step": 15791 + }, + { + "epoch": 1.33, + "learning_rate": 1.230425868752248e-05, + "loss": 1.1492, + "step": 15792 + }, + { + "epoch": 1.33, + "learning_rate": 1.230337416552672e-05, + "loss": 1.1549, + "step": 15793 + }, + { + "epoch": 1.33, + "learning_rate": 1.230248962449975e-05, + "loss": 1.1881, + "step": 15794 + }, + { + "epoch": 1.33, + "learning_rate": 1.230160506444887e-05, + "loss": 1.13, + "step": 15795 + }, + { + "epoch": 1.33, + "learning_rate": 1.2300720485381394e-05, + "loss": 1.135, + "step": 15796 + }, + { + "epoch": 1.33, + "learning_rate": 1.2299835887304627e-05, + "loss": 1.2218, + "step": 15797 + }, + { + "epoch": 1.33, + "learning_rate": 1.2298951270225883e-05, + "loss": 1.2358, + "step": 15798 + }, + { + "epoch": 1.33, + "learning_rate": 1.2298066634152468e-05, + "loss": 1.1697, + "step": 15799 + }, + { + "epoch": 1.33, + "learning_rate": 1.229718197909169e-05, + "loss": 1.2522, + "step": 15800 + }, + { + "epoch": 1.33, + "learning_rate": 1.2296297305050857e-05, + "loss": 1.1848, + "step": 15801 + }, + { + "epoch": 1.33, + "learning_rate": 1.2295412612037283e-05, + "loss": 1.1071, + "step": 15802 + }, + { + "epoch": 1.33, + "learning_rate": 1.2294527900058278e-05, + "loss": 1.1893, + "step": 15803 + }, + { + "epoch": 1.33, + "learning_rate": 1.2293643169121146e-05, + "loss": 1.17, + "step": 15804 + }, + { + "epoch": 1.33, + "learning_rate": 1.22927584192332e-05, + "loss": 1.1783, + "step": 15805 + }, + { + "epoch": 1.33, + "learning_rate": 1.229187365040175e-05, + "loss": 1.2382, + "step": 15806 + }, + { + "epoch": 1.33, + "learning_rate": 1.2290988862634109e-05, + "loss": 1.109, + "step": 15807 + }, + { + "epoch": 1.33, + "learning_rate": 1.229010405593758e-05, + "loss": 1.2274, + "step": 15808 + }, + { + "epoch": 1.33, + "learning_rate": 1.228921923031948e-05, + "loss": 1.2339, + "step": 15809 + }, + { + "epoch": 1.33, + "learning_rate": 1.2288334385787123e-05, + "loss": 1.1554, + "step": 15810 + }, + { + "epoch": 1.33, + "learning_rate": 1.2287449522347808e-05, + "loss": 1.25, + "step": 15811 + }, + { + "epoch": 1.33, + "learning_rate": 1.2286564640008856e-05, + "loss": 1.1444, + "step": 15812 + }, + { + "epoch": 1.33, + "learning_rate": 1.2285679738777575e-05, + "loss": 1.1049, + "step": 15813 + }, + { + "epoch": 1.33, + "learning_rate": 1.2284794818661279e-05, + "loss": 1.2104, + "step": 15814 + }, + { + "epoch": 1.33, + "learning_rate": 1.2283909879667272e-05, + "loss": 1.2399, + "step": 15815 + }, + { + "epoch": 1.33, + "learning_rate": 1.2283024921802875e-05, + "loss": 1.2938, + "step": 15816 + }, + { + "epoch": 1.33, + "learning_rate": 1.2282139945075395e-05, + "loss": 1.085, + "step": 15817 + }, + { + "epoch": 1.33, + "learning_rate": 1.2281254949492143e-05, + "loss": 1.1461, + "step": 15818 + }, + { + "epoch": 1.33, + "learning_rate": 1.2280369935060432e-05, + "loss": 1.2217, + "step": 15819 + }, + { + "epoch": 1.33, + "learning_rate": 1.2279484901787576e-05, + "loss": 1.2516, + "step": 15820 + }, + { + "epoch": 1.33, + "learning_rate": 1.2278599849680886e-05, + "loss": 1.1825, + "step": 15821 + }, + { + "epoch": 1.33, + "learning_rate": 1.2277714778747673e-05, + "loss": 1.1621, + "step": 15822 + }, + { + "epoch": 1.33, + "learning_rate": 1.2276829688995253e-05, + "loss": 1.156, + "step": 15823 + }, + { + "epoch": 1.33, + "learning_rate": 1.2275944580430939e-05, + "loss": 1.1674, + "step": 15824 + }, + { + "epoch": 1.33, + "learning_rate": 1.227505945306204e-05, + "loss": 1.1431, + "step": 15825 + }, + { + "epoch": 1.33, + "learning_rate": 1.2274174306895872e-05, + "loss": 1.1885, + "step": 15826 + }, + { + "epoch": 1.33, + "learning_rate": 1.2273289141939749e-05, + "loss": 1.1821, + "step": 15827 + }, + { + "epoch": 1.33, + "learning_rate": 1.2272403958200985e-05, + "loss": 1.1614, + "step": 15828 + }, + { + "epoch": 1.33, + "learning_rate": 1.2271518755686888e-05, + "loss": 1.2235, + "step": 15829 + }, + { + "epoch": 1.33, + "learning_rate": 1.2270633534404778e-05, + "loss": 1.1827, + "step": 15830 + }, + { + "epoch": 1.33, + "learning_rate": 1.2269748294361969e-05, + "loss": 1.1655, + "step": 15831 + }, + { + "epoch": 1.33, + "learning_rate": 1.2268863035565773e-05, + "loss": 1.1516, + "step": 15832 + }, + { + "epoch": 1.33, + "learning_rate": 1.2267977758023502e-05, + "loss": 1.1849, + "step": 15833 + }, + { + "epoch": 1.33, + "learning_rate": 1.2267092461742474e-05, + "loss": 1.0829, + "step": 15834 + }, + { + "epoch": 1.33, + "learning_rate": 1.2266207146730003e-05, + "loss": 1.2401, + "step": 15835 + }, + { + "epoch": 1.33, + "learning_rate": 1.2265321812993402e-05, + "loss": 1.2006, + "step": 15836 + }, + { + "epoch": 1.33, + "learning_rate": 1.2264436460539988e-05, + "loss": 1.1929, + "step": 15837 + }, + { + "epoch": 1.33, + "learning_rate": 1.2263551089377075e-05, + "loss": 1.2754, + "step": 15838 + }, + { + "epoch": 1.33, + "learning_rate": 1.2262665699511981e-05, + "loss": 1.319, + "step": 15839 + }, + { + "epoch": 1.33, + "learning_rate": 1.2261780290952016e-05, + "loss": 1.1796, + "step": 15840 + }, + { + "epoch": 1.33, + "learning_rate": 1.2260894863704497e-05, + "loss": 1.1901, + "step": 15841 + }, + { + "epoch": 1.33, + "learning_rate": 1.2260009417776745e-05, + "loss": 1.1839, + "step": 15842 + }, + { + "epoch": 1.33, + "learning_rate": 1.2259123953176071e-05, + "loss": 1.1721, + "step": 15843 + }, + { + "epoch": 1.33, + "learning_rate": 1.225823846990979e-05, + "loss": 1.1734, + "step": 15844 + }, + { + "epoch": 1.33, + "learning_rate": 1.2257352967985221e-05, + "loss": 1.1682, + "step": 15845 + }, + { + "epoch": 1.33, + "learning_rate": 1.2256467447409679e-05, + "loss": 1.2107, + "step": 15846 + }, + { + "epoch": 1.33, + "learning_rate": 1.225558190819048e-05, + "loss": 1.2349, + "step": 15847 + }, + { + "epoch": 1.33, + "learning_rate": 1.2254696350334943e-05, + "loss": 1.274, + "step": 15848 + }, + { + "epoch": 1.33, + "learning_rate": 1.2253810773850383e-05, + "loss": 1.2816, + "step": 15849 + }, + { + "epoch": 1.33, + "learning_rate": 1.2252925178744117e-05, + "loss": 1.1701, + "step": 15850 + }, + { + "epoch": 1.33, + "learning_rate": 1.2252039565023462e-05, + "loss": 1.2331, + "step": 15851 + }, + { + "epoch": 1.33, + "learning_rate": 1.2251153932695734e-05, + "loss": 1.1546, + "step": 15852 + }, + { + "epoch": 1.33, + "learning_rate": 1.2250268281768255e-05, + "loss": 1.1879, + "step": 15853 + }, + { + "epoch": 1.33, + "learning_rate": 1.2249382612248337e-05, + "loss": 1.1706, + "step": 15854 + }, + { + "epoch": 1.33, + "learning_rate": 1.22484969241433e-05, + "loss": 1.2096, + "step": 15855 + }, + { + "epoch": 1.33, + "learning_rate": 1.2247611217460462e-05, + "loss": 1.188, + "step": 15856 + }, + { + "epoch": 1.34, + "learning_rate": 1.2246725492207142e-05, + "loss": 1.2996, + "step": 15857 + }, + { + "epoch": 1.34, + "learning_rate": 1.2245839748390654e-05, + "loss": 1.3399, + "step": 15858 + }, + { + "epoch": 1.34, + "learning_rate": 1.224495398601832e-05, + "loss": 1.1136, + "step": 15859 + }, + { + "epoch": 1.34, + "learning_rate": 1.2244068205097459e-05, + "loss": 1.2123, + "step": 15860 + }, + { + "epoch": 1.34, + "learning_rate": 1.2243182405635387e-05, + "loss": 1.1935, + "step": 15861 + }, + { + "epoch": 1.34, + "learning_rate": 1.2242296587639426e-05, + "loss": 1.2211, + "step": 15862 + }, + { + "epoch": 1.34, + "learning_rate": 1.2241410751116891e-05, + "loss": 1.2301, + "step": 15863 + }, + { + "epoch": 1.34, + "learning_rate": 1.2240524896075106e-05, + "loss": 1.2566, + "step": 15864 + }, + { + "epoch": 1.34, + "learning_rate": 1.2239639022521385e-05, + "loss": 1.1632, + "step": 15865 + }, + { + "epoch": 1.34, + "learning_rate": 1.223875313046305e-05, + "loss": 1.1571, + "step": 15866 + }, + { + "epoch": 1.34, + "learning_rate": 1.223786721990742e-05, + "loss": 1.1883, + "step": 15867 + }, + { + "epoch": 1.34, + "learning_rate": 1.2236981290861815e-05, + "loss": 1.1928, + "step": 15868 + }, + { + "epoch": 1.34, + "learning_rate": 1.2236095343333554e-05, + "loss": 1.2166, + "step": 15869 + }, + { + "epoch": 1.34, + "learning_rate": 1.2235209377329958e-05, + "loss": 1.2351, + "step": 15870 + }, + { + "epoch": 1.34, + "learning_rate": 1.2234323392858347e-05, + "loss": 1.043, + "step": 15871 + }, + { + "epoch": 1.34, + "learning_rate": 1.2233437389926039e-05, + "loss": 1.2289, + "step": 15872 + }, + { + "epoch": 1.34, + "learning_rate": 1.2232551368540359e-05, + "loss": 1.1252, + "step": 15873 + }, + { + "epoch": 1.34, + "learning_rate": 1.2231665328708626e-05, + "loss": 1.2747, + "step": 15874 + }, + { + "epoch": 1.34, + "learning_rate": 1.2230779270438158e-05, + "loss": 1.1507, + "step": 15875 + }, + { + "epoch": 1.34, + "learning_rate": 1.2229893193736278e-05, + "loss": 1.1882, + "step": 15876 + }, + { + "epoch": 1.34, + "learning_rate": 1.2229007098610305e-05, + "loss": 1.27, + "step": 15877 + }, + { + "epoch": 1.34, + "learning_rate": 1.2228120985067566e-05, + "loss": 1.2267, + "step": 15878 + }, + { + "epoch": 1.34, + "learning_rate": 1.2227234853115377e-05, + "loss": 1.1996, + "step": 15879 + }, + { + "epoch": 1.34, + "learning_rate": 1.2226348702761059e-05, + "loss": 1.2243, + "step": 15880 + }, + { + "epoch": 1.34, + "learning_rate": 1.2225462534011936e-05, + "loss": 1.2084, + "step": 15881 + }, + { + "epoch": 1.34, + "learning_rate": 1.2224576346875333e-05, + "loss": 1.2111, + "step": 15882 + }, + { + "epoch": 1.34, + "learning_rate": 1.2223690141358561e-05, + "loss": 1.3575, + "step": 15883 + }, + { + "epoch": 1.34, + "learning_rate": 1.2222803917468952e-05, + "loss": 1.1984, + "step": 15884 + }, + { + "epoch": 1.34, + "learning_rate": 1.222191767521383e-05, + "loss": 1.1956, + "step": 15885 + }, + { + "epoch": 1.34, + "learning_rate": 1.222103141460051e-05, + "loss": 1.1965, + "step": 15886 + }, + { + "epoch": 1.34, + "learning_rate": 1.2220145135636316e-05, + "loss": 1.2351, + "step": 15887 + }, + { + "epoch": 1.34, + "learning_rate": 1.2219258838328574e-05, + "loss": 1.1764, + "step": 15888 + }, + { + "epoch": 1.34, + "learning_rate": 1.2218372522684607e-05, + "loss": 1.2798, + "step": 15889 + }, + { + "epoch": 1.34, + "learning_rate": 1.2217486188711735e-05, + "loss": 1.218, + "step": 15890 + }, + { + "epoch": 1.34, + "learning_rate": 1.2216599836417281e-05, + "loss": 1.2123, + "step": 15891 + }, + { + "epoch": 1.34, + "learning_rate": 1.2215713465808574e-05, + "loss": 1.2305, + "step": 15892 + }, + { + "epoch": 1.34, + "learning_rate": 1.221482707689293e-05, + "loss": 1.1658, + "step": 15893 + }, + { + "epoch": 1.34, + "learning_rate": 1.2213940669677675e-05, + "loss": 1.2011, + "step": 15894 + }, + { + "epoch": 1.34, + "learning_rate": 1.2213054244170133e-05, + "loss": 1.2686, + "step": 15895 + }, + { + "epoch": 1.34, + "learning_rate": 1.221216780037763e-05, + "loss": 1.1654, + "step": 15896 + }, + { + "epoch": 1.34, + "learning_rate": 1.2211281338307488e-05, + "loss": 1.126, + "step": 15897 + }, + { + "epoch": 1.34, + "learning_rate": 1.2210394857967032e-05, + "loss": 1.2288, + "step": 15898 + }, + { + "epoch": 1.34, + "learning_rate": 1.2209508359363588e-05, + "loss": 1.2124, + "step": 15899 + }, + { + "epoch": 1.34, + "learning_rate": 1.2208621842504478e-05, + "loss": 1.141, + "step": 15900 + }, + { + "epoch": 1.34, + "learning_rate": 1.2207735307397026e-05, + "loss": 1.2577, + "step": 15901 + }, + { + "epoch": 1.34, + "learning_rate": 1.220684875404856e-05, + "loss": 1.2037, + "step": 15902 + }, + { + "epoch": 1.34, + "learning_rate": 1.2205962182466405e-05, + "loss": 1.2449, + "step": 15903 + }, + { + "epoch": 1.34, + "learning_rate": 1.220507559265788e-05, + "loss": 1.2465, + "step": 15904 + }, + { + "epoch": 1.34, + "learning_rate": 1.2204188984630319e-05, + "loss": 1.2521, + "step": 15905 + }, + { + "epoch": 1.34, + "learning_rate": 1.220330235839104e-05, + "loss": 1.3742, + "step": 15906 + }, + { + "epoch": 1.34, + "learning_rate": 1.2202415713947372e-05, + "loss": 1.1984, + "step": 15907 + }, + { + "epoch": 1.34, + "learning_rate": 1.220152905130664e-05, + "loss": 1.2048, + "step": 15908 + }, + { + "epoch": 1.34, + "learning_rate": 1.2200642370476173e-05, + "loss": 1.231, + "step": 15909 + }, + { + "epoch": 1.34, + "learning_rate": 1.2199755671463293e-05, + "loss": 1.1978, + "step": 15910 + }, + { + "epoch": 1.34, + "learning_rate": 1.219886895427533e-05, + "loss": 1.1383, + "step": 15911 + }, + { + "epoch": 1.34, + "learning_rate": 1.2197982218919604e-05, + "loss": 1.2081, + "step": 15912 + }, + { + "epoch": 1.34, + "learning_rate": 1.2197095465403447e-05, + "loss": 1.1873, + "step": 15913 + }, + { + "epoch": 1.34, + "learning_rate": 1.2196208693734186e-05, + "loss": 1.2114, + "step": 15914 + }, + { + "epoch": 1.34, + "learning_rate": 1.2195321903919144e-05, + "loss": 1.2243, + "step": 15915 + }, + { + "epoch": 1.34, + "learning_rate": 1.219443509596565e-05, + "loss": 1.198, + "step": 15916 + }, + { + "epoch": 1.34, + "learning_rate": 1.219354826988103e-05, + "loss": 1.159, + "step": 15917 + }, + { + "epoch": 1.34, + "learning_rate": 1.2192661425672616e-05, + "loss": 1.11, + "step": 15918 + }, + { + "epoch": 1.34, + "learning_rate": 1.2191774563347727e-05, + "loss": 1.1799, + "step": 15919 + }, + { + "epoch": 1.34, + "learning_rate": 1.2190887682913697e-05, + "loss": 1.1609, + "step": 15920 + }, + { + "epoch": 1.34, + "learning_rate": 1.2190000784377851e-05, + "loss": 1.1806, + "step": 15921 + }, + { + "epoch": 1.34, + "learning_rate": 1.2189113867747518e-05, + "loss": 1.2057, + "step": 15922 + }, + { + "epoch": 1.34, + "learning_rate": 1.2188226933030025e-05, + "loss": 1.2215, + "step": 15923 + }, + { + "epoch": 1.34, + "learning_rate": 1.2187339980232704e-05, + "loss": 1.2713, + "step": 15924 + }, + { + "epoch": 1.34, + "learning_rate": 1.2186453009362878e-05, + "loss": 1.1727, + "step": 15925 + }, + { + "epoch": 1.34, + "learning_rate": 1.2185566020427878e-05, + "loss": 1.2044, + "step": 15926 + }, + { + "epoch": 1.34, + "learning_rate": 1.218467901343503e-05, + "loss": 1.1462, + "step": 15927 + }, + { + "epoch": 1.34, + "learning_rate": 1.2183791988391667e-05, + "loss": 1.165, + "step": 15928 + }, + { + "epoch": 1.34, + "learning_rate": 1.2182904945305114e-05, + "loss": 1.1612, + "step": 15929 + }, + { + "epoch": 1.34, + "learning_rate": 1.2182017884182702e-05, + "loss": 1.1347, + "step": 15930 + }, + { + "epoch": 1.34, + "learning_rate": 1.218113080503176e-05, + "loss": 1.2196, + "step": 15931 + }, + { + "epoch": 1.34, + "learning_rate": 1.218024370785962e-05, + "loss": 1.1703, + "step": 15932 + }, + { + "epoch": 1.34, + "learning_rate": 1.2179356592673604e-05, + "loss": 1.2338, + "step": 15933 + }, + { + "epoch": 1.34, + "learning_rate": 1.2178469459481048e-05, + "loss": 1.255, + "step": 15934 + }, + { + "epoch": 1.34, + "learning_rate": 1.2177582308289282e-05, + "loss": 1.2104, + "step": 15935 + }, + { + "epoch": 1.34, + "learning_rate": 1.2176695139105634e-05, + "loss": 1.2741, + "step": 15936 + }, + { + "epoch": 1.34, + "learning_rate": 1.2175807951937433e-05, + "loss": 1.1341, + "step": 15937 + }, + { + "epoch": 1.34, + "learning_rate": 1.217492074679201e-05, + "loss": 1.2707, + "step": 15938 + }, + { + "epoch": 1.34, + "learning_rate": 1.2174033523676696e-05, + "loss": 1.2241, + "step": 15939 + }, + { + "epoch": 1.34, + "learning_rate": 1.2173146282598822e-05, + "loss": 1.2578, + "step": 15940 + }, + { + "epoch": 1.34, + "learning_rate": 1.2172259023565715e-05, + "loss": 1.1846, + "step": 15941 + }, + { + "epoch": 1.34, + "learning_rate": 1.2171371746584712e-05, + "loss": 1.1095, + "step": 15942 + }, + { + "epoch": 1.34, + "learning_rate": 1.2170484451663142e-05, + "loss": 1.0922, + "step": 15943 + }, + { + "epoch": 1.34, + "learning_rate": 1.2169597138808332e-05, + "loss": 1.2132, + "step": 15944 + }, + { + "epoch": 1.34, + "learning_rate": 1.2168709808027614e-05, + "loss": 1.1443, + "step": 15945 + }, + { + "epoch": 1.34, + "learning_rate": 1.2167822459328324e-05, + "loss": 1.2372, + "step": 15946 + }, + { + "epoch": 1.34, + "learning_rate": 1.2166935092717793e-05, + "loss": 1.2065, + "step": 15947 + }, + { + "epoch": 1.34, + "learning_rate": 1.2166047708203348e-05, + "loss": 1.2657, + "step": 15948 + }, + { + "epoch": 1.34, + "learning_rate": 1.2165160305792325e-05, + "loss": 1.2174, + "step": 15949 + }, + { + "epoch": 1.34, + "learning_rate": 1.2164272885492055e-05, + "loss": 1.2674, + "step": 15950 + }, + { + "epoch": 1.34, + "learning_rate": 1.2163385447309868e-05, + "loss": 1.1847, + "step": 15951 + }, + { + "epoch": 1.34, + "learning_rate": 1.21624979912531e-05, + "loss": 1.1418, + "step": 15952 + }, + { + "epoch": 1.34, + "learning_rate": 1.216161051732908e-05, + "loss": 1.2137, + "step": 15953 + }, + { + "epoch": 1.34, + "learning_rate": 1.2160723025545143e-05, + "loss": 1.1852, + "step": 15954 + }, + { + "epoch": 1.34, + "learning_rate": 1.2159835515908621e-05, + "loss": 1.2463, + "step": 15955 + }, + { + "epoch": 1.34, + "learning_rate": 1.2158947988426846e-05, + "loss": 1.2279, + "step": 15956 + }, + { + "epoch": 1.34, + "learning_rate": 1.2158060443107152e-05, + "loss": 1.1335, + "step": 15957 + }, + { + "epoch": 1.34, + "learning_rate": 1.2157172879956868e-05, + "loss": 1.322, + "step": 15958 + }, + { + "epoch": 1.34, + "learning_rate": 1.2156285298983336e-05, + "loss": 1.187, + "step": 15959 + }, + { + "epoch": 1.34, + "learning_rate": 1.2155397700193884e-05, + "loss": 1.1959, + "step": 15960 + }, + { + "epoch": 1.34, + "learning_rate": 1.2154510083595848e-05, + "loss": 1.146, + "step": 15961 + }, + { + "epoch": 1.34, + "learning_rate": 1.2153622449196558e-05, + "loss": 1.2064, + "step": 15962 + }, + { + "epoch": 1.34, + "learning_rate": 1.2152734797003349e-05, + "loss": 1.2015, + "step": 15963 + }, + { + "epoch": 1.34, + "learning_rate": 1.215184712702356e-05, + "loss": 1.157, + "step": 15964 + }, + { + "epoch": 1.34, + "learning_rate": 1.2150959439264518e-05, + "loss": 1.1825, + "step": 15965 + }, + { + "epoch": 1.34, + "learning_rate": 1.215007173373356e-05, + "loss": 1.193, + "step": 15966 + }, + { + "epoch": 1.34, + "learning_rate": 1.2149184010438024e-05, + "loss": 1.1663, + "step": 15967 + }, + { + "epoch": 1.34, + "learning_rate": 1.2148296269385242e-05, + "loss": 1.1506, + "step": 15968 + }, + { + "epoch": 1.34, + "learning_rate": 1.2147408510582548e-05, + "loss": 1.1127, + "step": 15969 + }, + { + "epoch": 1.34, + "learning_rate": 1.2146520734037274e-05, + "loss": 1.1081, + "step": 15970 + }, + { + "epoch": 1.34, + "learning_rate": 1.2145632939756763e-05, + "loss": 1.3001, + "step": 15971 + }, + { + "epoch": 1.34, + "learning_rate": 1.2144745127748346e-05, + "loss": 1.207, + "step": 15972 + }, + { + "epoch": 1.34, + "learning_rate": 1.2143857298019357e-05, + "loss": 1.2843, + "step": 15973 + }, + { + "epoch": 1.34, + "learning_rate": 1.2142969450577136e-05, + "loss": 1.2434, + "step": 15974 + }, + { + "epoch": 1.34, + "learning_rate": 1.214208158542901e-05, + "loss": 1.2031, + "step": 15975 + }, + { + "epoch": 1.35, + "learning_rate": 1.2141193702582324e-05, + "loss": 1.1728, + "step": 15976 + }, + { + "epoch": 1.35, + "learning_rate": 1.2140305802044408e-05, + "loss": 1.1577, + "step": 15977 + }, + { + "epoch": 1.35, + "learning_rate": 1.2139417883822604e-05, + "loss": 1.2006, + "step": 15978 + }, + { + "epoch": 1.35, + "learning_rate": 1.2138529947924243e-05, + "loss": 1.1733, + "step": 15979 + }, + { + "epoch": 1.35, + "learning_rate": 1.2137641994356662e-05, + "loss": 1.1934, + "step": 15980 + }, + { + "epoch": 1.35, + "learning_rate": 1.2136754023127197e-05, + "loss": 1.1427, + "step": 15981 + }, + { + "epoch": 1.35, + "learning_rate": 1.2135866034243191e-05, + "loss": 1.0977, + "step": 15982 + }, + { + "epoch": 1.35, + "learning_rate": 1.2134978027711972e-05, + "loss": 1.2328, + "step": 15983 + }, + { + "epoch": 1.35, + "learning_rate": 1.2134090003540883e-05, + "loss": 1.2429, + "step": 15984 + }, + { + "epoch": 1.35, + "learning_rate": 1.2133201961737262e-05, + "loss": 1.1764, + "step": 15985 + }, + { + "epoch": 1.35, + "learning_rate": 1.2132313902308442e-05, + "loss": 1.1942, + "step": 15986 + }, + { + "epoch": 1.35, + "learning_rate": 1.213142582526176e-05, + "loss": 1.2232, + "step": 15987 + }, + { + "epoch": 1.35, + "learning_rate": 1.2130537730604557e-05, + "loss": 1.2377, + "step": 15988 + }, + { + "epoch": 1.35, + "learning_rate": 1.212964961834417e-05, + "loss": 1.0188, + "step": 15989 + }, + { + "epoch": 1.35, + "learning_rate": 1.2128761488487937e-05, + "loss": 1.1566, + "step": 15990 + }, + { + "epoch": 1.35, + "learning_rate": 1.2127873341043195e-05, + "loss": 1.1551, + "step": 15991 + }, + { + "epoch": 1.35, + "learning_rate": 1.212698517601728e-05, + "loss": 1.1133, + "step": 15992 + }, + { + "epoch": 1.35, + "learning_rate": 1.2126096993417535e-05, + "loss": 1.1342, + "step": 15993 + }, + { + "epoch": 1.35, + "learning_rate": 1.2125208793251297e-05, + "loss": 1.2212, + "step": 15994 + }, + { + "epoch": 1.35, + "learning_rate": 1.2124320575525899e-05, + "loss": 1.2539, + "step": 15995 + }, + { + "epoch": 1.35, + "learning_rate": 1.212343234024869e-05, + "loss": 1.2012, + "step": 15996 + }, + { + "epoch": 1.35, + "learning_rate": 1.2122544087427002e-05, + "loss": 1.2237, + "step": 15997 + }, + { + "epoch": 1.35, + "learning_rate": 1.2121655817068172e-05, + "loss": 1.1965, + "step": 15998 + }, + { + "epoch": 1.35, + "learning_rate": 1.2120767529179548e-05, + "loss": 1.1002, + "step": 15999 + }, + { + "epoch": 1.35, + "learning_rate": 1.2119879223768463e-05, + "loss": 1.1649, + "step": 16000 + }, + { + "epoch": 1.35, + "learning_rate": 1.2118990900842253e-05, + "loss": 1.1823, + "step": 16001 + }, + { + "epoch": 1.35, + "learning_rate": 1.2118102560408264e-05, + "loss": 1.2161, + "step": 16002 + }, + { + "epoch": 1.35, + "learning_rate": 1.2117214202473837e-05, + "loss": 1.2194, + "step": 16003 + }, + { + "epoch": 1.35, + "learning_rate": 1.2116325827046307e-05, + "loss": 1.2938, + "step": 16004 + }, + { + "epoch": 1.35, + "learning_rate": 1.2115437434133016e-05, + "loss": 1.2502, + "step": 16005 + }, + { + "epoch": 1.35, + "learning_rate": 1.21145490237413e-05, + "loss": 1.2462, + "step": 16006 + }, + { + "epoch": 1.35, + "learning_rate": 1.2113660595878506e-05, + "loss": 1.2487, + "step": 16007 + }, + { + "epoch": 1.35, + "learning_rate": 1.2112772150551972e-05, + "loss": 1.1332, + "step": 16008 + }, + { + "epoch": 1.35, + "learning_rate": 1.2111883687769037e-05, + "loss": 1.2053, + "step": 16009 + }, + { + "epoch": 1.35, + "learning_rate": 1.2110995207537046e-05, + "loss": 1.1719, + "step": 16010 + }, + { + "epoch": 1.35, + "learning_rate": 1.2110106709863333e-05, + "loss": 1.2105, + "step": 16011 + }, + { + "epoch": 1.35, + "learning_rate": 1.2109218194755245e-05, + "loss": 1.1793, + "step": 16012 + }, + { + "epoch": 1.35, + "learning_rate": 1.2108329662220123e-05, + "loss": 1.2342, + "step": 16013 + }, + { + "epoch": 1.35, + "learning_rate": 1.2107441112265306e-05, + "loss": 1.1845, + "step": 16014 + }, + { + "epoch": 1.35, + "learning_rate": 1.2106552544898134e-05, + "loss": 1.1398, + "step": 16015 + }, + { + "epoch": 1.35, + "learning_rate": 1.2105663960125953e-05, + "loss": 1.2291, + "step": 16016 + }, + { + "epoch": 1.35, + "learning_rate": 1.21047753579561e-05, + "loss": 1.1966, + "step": 16017 + }, + { + "epoch": 1.35, + "learning_rate": 1.2103886738395923e-05, + "loss": 1.1959, + "step": 16018 + }, + { + "epoch": 1.35, + "learning_rate": 1.2102998101452757e-05, + "loss": 1.2014, + "step": 16019 + }, + { + "epoch": 1.35, + "learning_rate": 1.2102109447133944e-05, + "loss": 1.1205, + "step": 16020 + }, + { + "epoch": 1.35, + "learning_rate": 1.2101220775446839e-05, + "loss": 1.0809, + "step": 16021 + }, + { + "epoch": 1.35, + "learning_rate": 1.210033208639877e-05, + "loss": 1.2263, + "step": 16022 + }, + { + "epoch": 1.35, + "learning_rate": 1.2099443379997087e-05, + "loss": 1.1651, + "step": 16023 + }, + { + "epoch": 1.35, + "learning_rate": 1.209855465624913e-05, + "loss": 1.2045, + "step": 16024 + }, + { + "epoch": 1.35, + "learning_rate": 1.2097665915162242e-05, + "loss": 1.2338, + "step": 16025 + }, + { + "epoch": 1.35, + "learning_rate": 1.2096777156743768e-05, + "loss": 1.1588, + "step": 16026 + }, + { + "epoch": 1.35, + "learning_rate": 1.209588838100105e-05, + "loss": 1.1606, + "step": 16027 + }, + { + "epoch": 1.35, + "learning_rate": 1.2094999587941434e-05, + "loss": 1.2393, + "step": 16028 + }, + { + "epoch": 1.35, + "learning_rate": 1.2094110777572257e-05, + "loss": 1.2323, + "step": 16029 + }, + { + "epoch": 1.35, + "learning_rate": 1.2093221949900866e-05, + "loss": 1.1557, + "step": 16030 + }, + { + "epoch": 1.35, + "learning_rate": 1.2092333104934609e-05, + "loss": 1.1638, + "step": 16031 + }, + { + "epoch": 1.35, + "learning_rate": 1.2091444242680826e-05, + "loss": 1.1709, + "step": 16032 + }, + { + "epoch": 1.35, + "learning_rate": 1.2090555363146856e-05, + "loss": 1.1808, + "step": 16033 + }, + { + "epoch": 1.35, + "learning_rate": 1.2089666466340053e-05, + "loss": 1.1563, + "step": 16034 + }, + { + "epoch": 1.35, + "learning_rate": 1.2088777552267756e-05, + "loss": 1.2116, + "step": 16035 + }, + { + "epoch": 1.35, + "learning_rate": 1.2087888620937311e-05, + "loss": 1.2133, + "step": 16036 + }, + { + "epoch": 1.35, + "learning_rate": 1.2086999672356061e-05, + "loss": 1.2544, + "step": 16037 + }, + { + "epoch": 1.35, + "learning_rate": 1.2086110706531354e-05, + "loss": 1.2526, + "step": 16038 + }, + { + "epoch": 1.35, + "learning_rate": 1.208522172347053e-05, + "loss": 1.1334, + "step": 16039 + }, + { + "epoch": 1.35, + "learning_rate": 1.2084332723180938e-05, + "loss": 1.1508, + "step": 16040 + }, + { + "epoch": 1.35, + "learning_rate": 1.2083443705669922e-05, + "loss": 1.1791, + "step": 16041 + }, + { + "epoch": 1.35, + "learning_rate": 1.2082554670944826e-05, + "loss": 1.112, + "step": 16042 + }, + { + "epoch": 1.35, + "learning_rate": 1.2081665619012998e-05, + "loss": 1.2365, + "step": 16043 + }, + { + "epoch": 1.35, + "learning_rate": 1.2080776549881782e-05, + "loss": 1.2579, + "step": 16044 + }, + { + "epoch": 1.35, + "learning_rate": 1.2079887463558523e-05, + "loss": 1.1481, + "step": 16045 + }, + { + "epoch": 1.35, + "learning_rate": 1.2078998360050572e-05, + "loss": 1.1471, + "step": 16046 + }, + { + "epoch": 1.35, + "learning_rate": 1.2078109239365268e-05, + "loss": 1.197, + "step": 16047 + }, + { + "epoch": 1.35, + "learning_rate": 1.2077220101509961e-05, + "loss": 1.1829, + "step": 16048 + }, + { + "epoch": 1.35, + "learning_rate": 1.2076330946491998e-05, + "loss": 1.25, + "step": 16049 + }, + { + "epoch": 1.35, + "learning_rate": 1.2075441774318723e-05, + "loss": 1.1581, + "step": 16050 + }, + { + "epoch": 1.35, + "learning_rate": 1.2074552584997482e-05, + "loss": 1.1257, + "step": 16051 + }, + { + "epoch": 1.35, + "learning_rate": 1.2073663378535627e-05, + "loss": 1.1952, + "step": 16052 + }, + { + "epoch": 1.35, + "learning_rate": 1.2072774154940501e-05, + "loss": 1.2069, + "step": 16053 + }, + { + "epoch": 1.35, + "learning_rate": 1.207188491421945e-05, + "loss": 1.1574, + "step": 16054 + }, + { + "epoch": 1.35, + "learning_rate": 1.2070995656379822e-05, + "loss": 1.1601, + "step": 16055 + }, + { + "epoch": 1.35, + "learning_rate": 1.2070106381428968e-05, + "loss": 1.2432, + "step": 16056 + }, + { + "epoch": 1.35, + "learning_rate": 1.2069217089374232e-05, + "loss": 1.2447, + "step": 16057 + }, + { + "epoch": 1.35, + "learning_rate": 1.2068327780222959e-05, + "loss": 1.1827, + "step": 16058 + }, + { + "epoch": 1.35, + "learning_rate": 1.20674384539825e-05, + "loss": 1.2283, + "step": 16059 + }, + { + "epoch": 1.35, + "learning_rate": 1.2066549110660209e-05, + "loss": 1.1518, + "step": 16060 + }, + { + "epoch": 1.35, + "learning_rate": 1.2065659750263422e-05, + "loss": 1.055, + "step": 16061 + }, + { + "epoch": 1.35, + "learning_rate": 1.2064770372799495e-05, + "loss": 1.2428, + "step": 16062 + }, + { + "epoch": 1.35, + "learning_rate": 1.2063880978275773e-05, + "loss": 1.1636, + "step": 16063 + }, + { + "epoch": 1.35, + "learning_rate": 1.2062991566699609e-05, + "loss": 1.2119, + "step": 16064 + }, + { + "epoch": 1.35, + "learning_rate": 1.2062102138078347e-05, + "loss": 1.2629, + "step": 16065 + }, + { + "epoch": 1.35, + "learning_rate": 1.2061212692419334e-05, + "loss": 1.1296, + "step": 16066 + }, + { + "epoch": 1.35, + "learning_rate": 1.2060323229729926e-05, + "loss": 1.1919, + "step": 16067 + }, + { + "epoch": 1.35, + "learning_rate": 1.2059433750017468e-05, + "loss": 1.1878, + "step": 16068 + }, + { + "epoch": 1.35, + "learning_rate": 1.2058544253289308e-05, + "loss": 1.1642, + "step": 16069 + }, + { + "epoch": 1.35, + "learning_rate": 1.2057654739552795e-05, + "loss": 1.2705, + "step": 16070 + }, + { + "epoch": 1.35, + "learning_rate": 1.2056765208815282e-05, + "loss": 1.184, + "step": 16071 + }, + { + "epoch": 1.35, + "learning_rate": 1.2055875661084117e-05, + "loss": 1.2071, + "step": 16072 + }, + { + "epoch": 1.35, + "learning_rate": 1.2054986096366648e-05, + "loss": 1.2303, + "step": 16073 + }, + { + "epoch": 1.35, + "learning_rate": 1.2054096514670225e-05, + "loss": 1.1969, + "step": 16074 + }, + { + "epoch": 1.35, + "learning_rate": 1.2053206916002203e-05, + "loss": 1.2735, + "step": 16075 + }, + { + "epoch": 1.35, + "learning_rate": 1.2052317300369928e-05, + "loss": 1.2946, + "step": 16076 + }, + { + "epoch": 1.35, + "learning_rate": 1.2051427667780747e-05, + "loss": 1.0998, + "step": 16077 + }, + { + "epoch": 1.35, + "learning_rate": 1.2050538018242017e-05, + "loss": 1.1625, + "step": 16078 + }, + { + "epoch": 1.35, + "learning_rate": 1.2049648351761086e-05, + "loss": 1.094, + "step": 16079 + }, + { + "epoch": 1.35, + "learning_rate": 1.2048758668345302e-05, + "loss": 1.2011, + "step": 16080 + }, + { + "epoch": 1.35, + "learning_rate": 1.204786896800202e-05, + "loss": 1.1818, + "step": 16081 + }, + { + "epoch": 1.35, + "learning_rate": 1.2046979250738589e-05, + "loss": 1.122, + "step": 16082 + }, + { + "epoch": 1.35, + "learning_rate": 1.2046089516562356e-05, + "loss": 1.0538, + "step": 16083 + }, + { + "epoch": 1.35, + "learning_rate": 1.204519976548068e-05, + "loss": 1.1799, + "step": 16084 + }, + { + "epoch": 1.35, + "learning_rate": 1.2044309997500912e-05, + "loss": 1.1158, + "step": 16085 + }, + { + "epoch": 1.35, + "learning_rate": 1.2043420212630396e-05, + "loss": 1.2252, + "step": 16086 + }, + { + "epoch": 1.35, + "learning_rate": 1.204253041087649e-05, + "loss": 1.2047, + "step": 16087 + }, + { + "epoch": 1.35, + "learning_rate": 1.2041640592246544e-05, + "loss": 1.1588, + "step": 16088 + }, + { + "epoch": 1.35, + "learning_rate": 1.2040750756747912e-05, + "loss": 1.2551, + "step": 16089 + }, + { + "epoch": 1.35, + "learning_rate": 1.203986090438794e-05, + "loss": 1.198, + "step": 16090 + }, + { + "epoch": 1.35, + "learning_rate": 1.2038971035173987e-05, + "loss": 1.1525, + "step": 16091 + }, + { + "epoch": 1.35, + "learning_rate": 1.2038081149113403e-05, + "loss": 1.2291, + "step": 16092 + }, + { + "epoch": 1.35, + "learning_rate": 1.2037191246213539e-05, + "loss": 1.2593, + "step": 16093 + }, + { + "epoch": 1.35, + "learning_rate": 1.2036301326481751e-05, + "loss": 1.1838, + "step": 16094 + }, + { + "epoch": 1.36, + "learning_rate": 1.2035411389925385e-05, + "loss": 1.2219, + "step": 16095 + }, + { + "epoch": 1.36, + "learning_rate": 1.2034521436551805e-05, + "loss": 1.2133, + "step": 16096 + }, + { + "epoch": 1.36, + "learning_rate": 1.2033631466368354e-05, + "loss": 1.2245, + "step": 16097 + }, + { + "epoch": 1.36, + "learning_rate": 1.2032741479382389e-05, + "loss": 1.1897, + "step": 16098 + }, + { + "epoch": 1.36, + "learning_rate": 1.2031851475601263e-05, + "loss": 1.0351, + "step": 16099 + }, + { + "epoch": 1.36, + "learning_rate": 1.2030961455032332e-05, + "loss": 1.1161, + "step": 16100 + }, + { + "epoch": 1.36, + "learning_rate": 1.2030071417682947e-05, + "loss": 1.1428, + "step": 16101 + }, + { + "epoch": 1.36, + "learning_rate": 1.202918136356046e-05, + "loss": 1.1601, + "step": 16102 + }, + { + "epoch": 1.36, + "learning_rate": 1.2028291292672231e-05, + "loss": 1.1486, + "step": 16103 + }, + { + "epoch": 1.36, + "learning_rate": 1.2027401205025608e-05, + "loss": 1.0811, + "step": 16104 + }, + { + "epoch": 1.36, + "learning_rate": 1.2026511100627947e-05, + "loss": 1.0644, + "step": 16105 + }, + { + "epoch": 1.36, + "learning_rate": 1.2025620979486603e-05, + "loss": 1.2496, + "step": 16106 + }, + { + "epoch": 1.36, + "learning_rate": 1.2024730841608932e-05, + "loss": 1.2326, + "step": 16107 + }, + { + "epoch": 1.36, + "learning_rate": 1.2023840687002282e-05, + "loss": 1.2478, + "step": 16108 + }, + { + "epoch": 1.36, + "learning_rate": 1.2022950515674014e-05, + "loss": 1.2135, + "step": 16109 + }, + { + "epoch": 1.36, + "learning_rate": 1.2022060327631487e-05, + "loss": 1.213, + "step": 16110 + }, + { + "epoch": 1.36, + "learning_rate": 1.2021170122882045e-05, + "loss": 1.1607, + "step": 16111 + }, + { + "epoch": 1.36, + "learning_rate": 1.2020279901433049e-05, + "loss": 1.1544, + "step": 16112 + }, + { + "epoch": 1.36, + "learning_rate": 1.2019389663291854e-05, + "loss": 1.2615, + "step": 16113 + }, + { + "epoch": 1.36, + "learning_rate": 1.2018499408465819e-05, + "loss": 1.2202, + "step": 16114 + }, + { + "epoch": 1.36, + "learning_rate": 1.2017609136962291e-05, + "loss": 1.1846, + "step": 16115 + }, + { + "epoch": 1.36, + "learning_rate": 1.2016718848788634e-05, + "loss": 1.0689, + "step": 16116 + }, + { + "epoch": 1.36, + "learning_rate": 1.2015828543952196e-05, + "loss": 1.198, + "step": 16117 + }, + { + "epoch": 1.36, + "learning_rate": 1.2014938222460344e-05, + "loss": 1.1934, + "step": 16118 + }, + { + "epoch": 1.36, + "learning_rate": 1.2014047884320423e-05, + "loss": 1.1725, + "step": 16119 + }, + { + "epoch": 1.36, + "learning_rate": 1.2013157529539794e-05, + "loss": 1.2131, + "step": 16120 + }, + { + "epoch": 1.36, + "learning_rate": 1.2012267158125815e-05, + "loss": 1.2604, + "step": 16121 + }, + { + "epoch": 1.36, + "learning_rate": 1.2011376770085841e-05, + "loss": 1.209, + "step": 16122 + }, + { + "epoch": 1.36, + "learning_rate": 1.2010486365427225e-05, + "loss": 1.1569, + "step": 16123 + }, + { + "epoch": 1.36, + "learning_rate": 1.2009595944157329e-05, + "loss": 1.1879, + "step": 16124 + }, + { + "epoch": 1.36, + "learning_rate": 1.2008705506283511e-05, + "loss": 1.1753, + "step": 16125 + }, + { + "epoch": 1.36, + "learning_rate": 1.200781505181312e-05, + "loss": 1.2192, + "step": 16126 + }, + { + "epoch": 1.36, + "learning_rate": 1.2006924580753523e-05, + "loss": 1.2409, + "step": 16127 + }, + { + "epoch": 1.36, + "learning_rate": 1.2006034093112073e-05, + "loss": 1.2401, + "step": 16128 + }, + { + "epoch": 1.36, + "learning_rate": 1.2005143588896125e-05, + "loss": 1.2285, + "step": 16129 + }, + { + "epoch": 1.36, + "learning_rate": 1.2004253068113039e-05, + "loss": 1.1325, + "step": 16130 + }, + { + "epoch": 1.36, + "learning_rate": 1.2003362530770173e-05, + "loss": 1.2735, + "step": 16131 + }, + { + "epoch": 1.36, + "learning_rate": 1.2002471976874885e-05, + "loss": 1.1701, + "step": 16132 + }, + { + "epoch": 1.36, + "learning_rate": 1.200158140643453e-05, + "loss": 1.2493, + "step": 16133 + }, + { + "epoch": 1.36, + "learning_rate": 1.2000690819456471e-05, + "loss": 1.1912, + "step": 16134 + }, + { + "epoch": 1.36, + "learning_rate": 1.1999800215948065e-05, + "loss": 1.2453, + "step": 16135 + }, + { + "epoch": 1.36, + "learning_rate": 1.199890959591667e-05, + "loss": 1.1467, + "step": 16136 + }, + { + "epoch": 1.36, + "learning_rate": 1.1998018959369642e-05, + "loss": 1.1862, + "step": 16137 + }, + { + "epoch": 1.36, + "learning_rate": 1.1997128306314343e-05, + "loss": 1.2035, + "step": 16138 + }, + { + "epoch": 1.36, + "learning_rate": 1.1996237636758133e-05, + "loss": 1.2803, + "step": 16139 + }, + { + "epoch": 1.36, + "learning_rate": 1.1995346950708365e-05, + "loss": 1.1865, + "step": 16140 + }, + { + "epoch": 1.36, + "learning_rate": 1.1994456248172403e-05, + "loss": 1.273, + "step": 16141 + }, + { + "epoch": 1.36, + "learning_rate": 1.1993565529157605e-05, + "loss": 1.218, + "step": 16142 + }, + { + "epoch": 1.36, + "learning_rate": 1.1992674793671332e-05, + "loss": 1.2166, + "step": 16143 + }, + { + "epoch": 1.36, + "learning_rate": 1.1991784041720942e-05, + "loss": 1.1907, + "step": 16144 + }, + { + "epoch": 1.36, + "learning_rate": 1.1990893273313793e-05, + "loss": 1.1433, + "step": 16145 + }, + { + "epoch": 1.36, + "learning_rate": 1.1990002488457249e-05, + "loss": 1.1096, + "step": 16146 + }, + { + "epoch": 1.36, + "learning_rate": 1.1989111687158665e-05, + "loss": 1.21, + "step": 16147 + }, + { + "epoch": 1.36, + "learning_rate": 1.1988220869425405e-05, + "loss": 1.1689, + "step": 16148 + }, + { + "epoch": 1.36, + "learning_rate": 1.1987330035264826e-05, + "loss": 1.2073, + "step": 16149 + }, + { + "epoch": 1.36, + "learning_rate": 1.1986439184684295e-05, + "loss": 1.2403, + "step": 16150 + }, + { + "epoch": 1.36, + "learning_rate": 1.1985548317691164e-05, + "loss": 1.2041, + "step": 16151 + }, + { + "epoch": 1.36, + "learning_rate": 1.1984657434292797e-05, + "loss": 1.2772, + "step": 16152 + }, + { + "epoch": 1.36, + "learning_rate": 1.1983766534496559e-05, + "loss": 1.206, + "step": 16153 + }, + { + "epoch": 1.36, + "learning_rate": 1.1982875618309801e-05, + "loss": 1.1984, + "step": 16154 + }, + { + "epoch": 1.36, + "learning_rate": 1.1981984685739894e-05, + "loss": 1.2586, + "step": 16155 + }, + { + "epoch": 1.36, + "learning_rate": 1.1981093736794193e-05, + "loss": 1.2634, + "step": 16156 + }, + { + "epoch": 1.36, + "learning_rate": 1.1980202771480064e-05, + "loss": 1.2318, + "step": 16157 + }, + { + "epoch": 1.36, + "learning_rate": 1.1979311789804864e-05, + "loss": 1.2361, + "step": 16158 + }, + { + "epoch": 1.36, + "learning_rate": 1.1978420791775957e-05, + "loss": 1.1672, + "step": 16159 + }, + { + "epoch": 1.36, + "learning_rate": 1.1977529777400708e-05, + "loss": 1.2425, + "step": 16160 + }, + { + "epoch": 1.36, + "learning_rate": 1.197663874668647e-05, + "loss": 1.1396, + "step": 16161 + }, + { + "epoch": 1.36, + "learning_rate": 1.1975747699640612e-05, + "loss": 1.1592, + "step": 16162 + }, + { + "epoch": 1.36, + "learning_rate": 1.1974856636270493e-05, + "loss": 1.134, + "step": 16163 + }, + { + "epoch": 1.36, + "learning_rate": 1.1973965556583478e-05, + "loss": 1.248, + "step": 16164 + }, + { + "epoch": 1.36, + "learning_rate": 1.197307446058693e-05, + "loss": 1.1989, + "step": 16165 + }, + { + "epoch": 1.36, + "learning_rate": 1.1972183348288205e-05, + "loss": 1.1789, + "step": 16166 + }, + { + "epoch": 1.36, + "learning_rate": 1.197129221969467e-05, + "loss": 1.1692, + "step": 16167 + }, + { + "epoch": 1.36, + "learning_rate": 1.1970401074813691e-05, + "loss": 1.1982, + "step": 16168 + }, + { + "epoch": 1.36, + "learning_rate": 1.1969509913652626e-05, + "loss": 1.2852, + "step": 16169 + }, + { + "epoch": 1.36, + "learning_rate": 1.1968618736218837e-05, + "loss": 1.1098, + "step": 16170 + }, + { + "epoch": 1.36, + "learning_rate": 1.1967727542519694e-05, + "loss": 1.0995, + "step": 16171 + }, + { + "epoch": 1.36, + "learning_rate": 1.1966836332562554e-05, + "loss": 1.142, + "step": 16172 + }, + { + "epoch": 1.36, + "learning_rate": 1.1965945106354783e-05, + "loss": 1.1728, + "step": 16173 + }, + { + "epoch": 1.36, + "learning_rate": 1.1965053863903746e-05, + "loss": 1.1707, + "step": 16174 + }, + { + "epoch": 1.36, + "learning_rate": 1.1964162605216807e-05, + "loss": 1.2067, + "step": 16175 + }, + { + "epoch": 1.36, + "learning_rate": 1.1963271330301325e-05, + "loss": 1.2146, + "step": 16176 + }, + { + "epoch": 1.36, + "learning_rate": 1.1962380039164665e-05, + "loss": 1.152, + "step": 16177 + }, + { + "epoch": 1.36, + "learning_rate": 1.1961488731814197e-05, + "loss": 1.2222, + "step": 16178 + }, + { + "epoch": 1.36, + "learning_rate": 1.1960597408257279e-05, + "loss": 1.2029, + "step": 16179 + }, + { + "epoch": 1.36, + "learning_rate": 1.195970606850128e-05, + "loss": 1.1562, + "step": 16180 + }, + { + "epoch": 1.36, + "learning_rate": 1.195881471255356e-05, + "loss": 1.1789, + "step": 16181 + }, + { + "epoch": 1.36, + "learning_rate": 1.1957923340421487e-05, + "loss": 1.2362, + "step": 16182 + }, + { + "epoch": 1.36, + "learning_rate": 1.1957031952112422e-05, + "loss": 1.2592, + "step": 16183 + }, + { + "epoch": 1.36, + "learning_rate": 1.1956140547633734e-05, + "loss": 1.1198, + "step": 16184 + }, + { + "epoch": 1.36, + "learning_rate": 1.195524912699279e-05, + "loss": 1.1949, + "step": 16185 + }, + { + "epoch": 1.36, + "learning_rate": 1.195435769019695e-05, + "loss": 1.239, + "step": 16186 + }, + { + "epoch": 1.36, + "learning_rate": 1.195346623725358e-05, + "loss": 1.1536, + "step": 16187 + }, + { + "epoch": 1.36, + "learning_rate": 1.1952574768170047e-05, + "loss": 1.1809, + "step": 16188 + }, + { + "epoch": 1.36, + "learning_rate": 1.1951683282953719e-05, + "loss": 1.15, + "step": 16189 + }, + { + "epoch": 1.36, + "learning_rate": 1.1950791781611957e-05, + "loss": 1.0821, + "step": 16190 + }, + { + "epoch": 1.36, + "learning_rate": 1.1949900264152128e-05, + "loss": 1.142, + "step": 16191 + }, + { + "epoch": 1.36, + "learning_rate": 1.19490087305816e-05, + "loss": 1.0988, + "step": 16192 + }, + { + "epoch": 1.36, + "learning_rate": 1.1948117180907742e-05, + "loss": 1.2015, + "step": 16193 + }, + { + "epoch": 1.36, + "learning_rate": 1.194722561513791e-05, + "loss": 1.1706, + "step": 16194 + }, + { + "epoch": 1.36, + "learning_rate": 1.1946334033279476e-05, + "loss": 1.1365, + "step": 16195 + }, + { + "epoch": 1.36, + "learning_rate": 1.1945442435339814e-05, + "loss": 1.2454, + "step": 16196 + }, + { + "epoch": 1.36, + "learning_rate": 1.1944550821326279e-05, + "loss": 1.1291, + "step": 16197 + }, + { + "epoch": 1.36, + "learning_rate": 1.1943659191246245e-05, + "loss": 1.2117, + "step": 16198 + }, + { + "epoch": 1.36, + "learning_rate": 1.1942767545107074e-05, + "loss": 1.1937, + "step": 16199 + }, + { + "epoch": 1.36, + "learning_rate": 1.1941875882916141e-05, + "loss": 1.1831, + "step": 16200 + }, + { + "epoch": 1.36, + "learning_rate": 1.1940984204680803e-05, + "loss": 1.1792, + "step": 16201 + }, + { + "epoch": 1.36, + "learning_rate": 1.1940092510408435e-05, + "loss": 1.2604, + "step": 16202 + }, + { + "epoch": 1.36, + "learning_rate": 1.1939200800106403e-05, + "loss": 1.1457, + "step": 16203 + }, + { + "epoch": 1.36, + "learning_rate": 1.193830907378207e-05, + "loss": 1.1574, + "step": 16204 + }, + { + "epoch": 1.36, + "learning_rate": 1.1937417331442808e-05, + "loss": 1.1963, + "step": 16205 + }, + { + "epoch": 1.36, + "learning_rate": 1.1936525573095985e-05, + "loss": 1.2264, + "step": 16206 + }, + { + "epoch": 1.36, + "learning_rate": 1.1935633798748969e-05, + "loss": 1.2236, + "step": 16207 + }, + { + "epoch": 1.36, + "learning_rate": 1.1934742008409122e-05, + "loss": 1.1386, + "step": 16208 + }, + { + "epoch": 1.36, + "learning_rate": 1.193385020208382e-05, + "loss": 1.2276, + "step": 16209 + }, + { + "epoch": 1.36, + "learning_rate": 1.1932958379780433e-05, + "loss": 1.1774, + "step": 16210 + }, + { + "epoch": 1.36, + "learning_rate": 1.193206654150632e-05, + "loss": 1.2755, + "step": 16211 + }, + { + "epoch": 1.36, + "learning_rate": 1.1931174687268856e-05, + "loss": 1.2576, + "step": 16212 + }, + { + "epoch": 1.36, + "learning_rate": 1.1930282817075409e-05, + "loss": 1.1454, + "step": 16213 + }, + { + "epoch": 1.37, + "learning_rate": 1.1929390930933349e-05, + "loss": 1.2714, + "step": 16214 + }, + { + "epoch": 1.37, + "learning_rate": 1.1928499028850042e-05, + "loss": 1.1885, + "step": 16215 + }, + { + "epoch": 1.37, + "learning_rate": 1.192760711083286e-05, + "loss": 1.1738, + "step": 16216 + }, + { + "epoch": 1.37, + "learning_rate": 1.192671517688917e-05, + "loss": 1.1541, + "step": 16217 + }, + { + "epoch": 1.37, + "learning_rate": 1.1925823227026346e-05, + "loss": 1.1085, + "step": 16218 + }, + { + "epoch": 1.37, + "learning_rate": 1.1924931261251749e-05, + "loss": 1.1308, + "step": 16219 + }, + { + "epoch": 1.37, + "learning_rate": 1.1924039279572755e-05, + "loss": 1.199, + "step": 16220 + }, + { + "epoch": 1.37, + "learning_rate": 1.1923147281996734e-05, + "loss": 1.1824, + "step": 16221 + }, + { + "epoch": 1.37, + "learning_rate": 1.1922255268531053e-05, + "loss": 1.1836, + "step": 16222 + }, + { + "epoch": 1.37, + "learning_rate": 1.1921363239183086e-05, + "loss": 1.1834, + "step": 16223 + }, + { + "epoch": 1.37, + "learning_rate": 1.1920471193960199e-05, + "loss": 1.2222, + "step": 16224 + }, + { + "epoch": 1.37, + "learning_rate": 1.1919579132869767e-05, + "loss": 1.2154, + "step": 16225 + }, + { + "epoch": 1.37, + "learning_rate": 1.1918687055919158e-05, + "loss": 1.1251, + "step": 16226 + }, + { + "epoch": 1.37, + "learning_rate": 1.1917794963115741e-05, + "loss": 1.22, + "step": 16227 + }, + { + "epoch": 1.37, + "learning_rate": 1.1916902854466888e-05, + "loss": 1.2839, + "step": 16228 + }, + { + "epoch": 1.37, + "learning_rate": 1.191601072997997e-05, + "loss": 1.2255, + "step": 16229 + }, + { + "epoch": 1.37, + "learning_rate": 1.1915118589662357e-05, + "loss": 1.1809, + "step": 16230 + }, + { + "epoch": 1.37, + "learning_rate": 1.1914226433521423e-05, + "loss": 1.1636, + "step": 16231 + }, + { + "epoch": 1.37, + "learning_rate": 1.191333426156454e-05, + "loss": 1.2248, + "step": 16232 + }, + { + "epoch": 1.37, + "learning_rate": 1.1912442073799072e-05, + "loss": 1.2397, + "step": 16233 + }, + { + "epoch": 1.37, + "learning_rate": 1.1911549870232397e-05, + "loss": 1.3071, + "step": 16234 + }, + { + "epoch": 1.37, + "learning_rate": 1.1910657650871889e-05, + "loss": 1.3356, + "step": 16235 + }, + { + "epoch": 1.37, + "learning_rate": 1.1909765415724912e-05, + "loss": 1.1883, + "step": 16236 + }, + { + "epoch": 1.37, + "learning_rate": 1.1908873164798842e-05, + "loss": 1.2693, + "step": 16237 + }, + { + "epoch": 1.37, + "learning_rate": 1.1907980898101052e-05, + "loss": 1.2176, + "step": 16238 + }, + { + "epoch": 1.37, + "learning_rate": 1.1907088615638913e-05, + "loss": 1.0514, + "step": 16239 + }, + { + "epoch": 1.37, + "learning_rate": 1.1906196317419799e-05, + "loss": 1.1511, + "step": 16240 + }, + { + "epoch": 1.37, + "learning_rate": 1.1905304003451078e-05, + "loss": 1.2093, + "step": 16241 + }, + { + "epoch": 1.37, + "learning_rate": 1.1904411673740129e-05, + "loss": 1.1391, + "step": 16242 + }, + { + "epoch": 1.37, + "learning_rate": 1.190351932829432e-05, + "loss": 1.1651, + "step": 16243 + }, + { + "epoch": 1.37, + "learning_rate": 1.1902626967121023e-05, + "loss": 1.2233, + "step": 16244 + }, + { + "epoch": 1.37, + "learning_rate": 1.1901734590227614e-05, + "loss": 1.1393, + "step": 16245 + }, + { + "epoch": 1.37, + "learning_rate": 1.1900842197621465e-05, + "loss": 1.1403, + "step": 16246 + }, + { + "epoch": 1.37, + "learning_rate": 1.189994978930995e-05, + "loss": 1.1571, + "step": 16247 + }, + { + "epoch": 1.37, + "learning_rate": 1.1899057365300441e-05, + "loss": 1.1971, + "step": 16248 + }, + { + "epoch": 1.37, + "learning_rate": 1.1898164925600316e-05, + "loss": 1.2437, + "step": 16249 + }, + { + "epoch": 1.37, + "learning_rate": 1.1897272470216943e-05, + "loss": 1.1976, + "step": 16250 + }, + { + "epoch": 1.37, + "learning_rate": 1.1896379999157697e-05, + "loss": 1.2751, + "step": 16251 + }, + { + "epoch": 1.37, + "learning_rate": 1.189548751242995e-05, + "loss": 1.1829, + "step": 16252 + }, + { + "epoch": 1.37, + "learning_rate": 1.1894595010041082e-05, + "loss": 1.2476, + "step": 16253 + }, + { + "epoch": 1.37, + "learning_rate": 1.1893702491998462e-05, + "loss": 1.2296, + "step": 16254 + }, + { + "epoch": 1.37, + "learning_rate": 1.1892809958309468e-05, + "loss": 1.0752, + "step": 16255 + }, + { + "epoch": 1.37, + "learning_rate": 1.189191740898147e-05, + "loss": 1.182, + "step": 16256 + }, + { + "epoch": 1.37, + "learning_rate": 1.1891024844021848e-05, + "loss": 1.227, + "step": 16257 + }, + { + "epoch": 1.37, + "learning_rate": 1.1890132263437967e-05, + "loss": 1.273, + "step": 16258 + }, + { + "epoch": 1.37, + "learning_rate": 1.1889239667237214e-05, + "loss": 1.1768, + "step": 16259 + }, + { + "epoch": 1.37, + "learning_rate": 1.1888347055426956e-05, + "loss": 1.0548, + "step": 16260 + }, + { + "epoch": 1.37, + "learning_rate": 1.1887454428014573e-05, + "loss": 1.2949, + "step": 16261 + }, + { + "epoch": 1.37, + "learning_rate": 1.1886561785007434e-05, + "loss": 1.2363, + "step": 16262 + }, + { + "epoch": 1.37, + "learning_rate": 1.188566912641292e-05, + "loss": 1.1445, + "step": 16263 + }, + { + "epoch": 1.37, + "learning_rate": 1.1884776452238405e-05, + "loss": 1.2279, + "step": 16264 + }, + { + "epoch": 1.37, + "learning_rate": 1.1883883762491261e-05, + "loss": 1.138, + "step": 16265 + }, + { + "epoch": 1.37, + "learning_rate": 1.1882991057178868e-05, + "loss": 1.1791, + "step": 16266 + }, + { + "epoch": 1.37, + "learning_rate": 1.1882098336308598e-05, + "loss": 1.2391, + "step": 16267 + }, + { + "epoch": 1.37, + "learning_rate": 1.1881205599887833e-05, + "loss": 1.1633, + "step": 16268 + }, + { + "epoch": 1.37, + "learning_rate": 1.1880312847923943e-05, + "loss": 1.2217, + "step": 16269 + }, + { + "epoch": 1.37, + "learning_rate": 1.1879420080424304e-05, + "loss": 1.2641, + "step": 16270 + }, + { + "epoch": 1.37, + "learning_rate": 1.18785272973963e-05, + "loss": 1.1599, + "step": 16271 + }, + { + "epoch": 1.37, + "learning_rate": 1.1877634498847299e-05, + "loss": 1.2446, + "step": 16272 + }, + { + "epoch": 1.37, + "learning_rate": 1.187674168478468e-05, + "loss": 1.1743, + "step": 16273 + }, + { + "epoch": 1.37, + "learning_rate": 1.1875848855215823e-05, + "loss": 1.2623, + "step": 16274 + }, + { + "epoch": 1.37, + "learning_rate": 1.1874956010148104e-05, + "loss": 1.185, + "step": 16275 + }, + { + "epoch": 1.37, + "learning_rate": 1.1874063149588894e-05, + "loss": 1.2489, + "step": 16276 + }, + { + "epoch": 1.37, + "learning_rate": 1.1873170273545578e-05, + "loss": 1.2343, + "step": 16277 + }, + { + "epoch": 1.37, + "learning_rate": 1.1872277382025527e-05, + "loss": 1.1101, + "step": 16278 + }, + { + "epoch": 1.37, + "learning_rate": 1.1871384475036125e-05, + "loss": 1.1424, + "step": 16279 + }, + { + "epoch": 1.37, + "learning_rate": 1.1870491552584742e-05, + "loss": 1.1057, + "step": 16280 + }, + { + "epoch": 1.37, + "learning_rate": 1.186959861467876e-05, + "loss": 1.1899, + "step": 16281 + }, + { + "epoch": 1.37, + "learning_rate": 1.1868705661325557e-05, + "loss": 1.2479, + "step": 16282 + }, + { + "epoch": 1.37, + "learning_rate": 1.1867812692532508e-05, + "loss": 1.1952, + "step": 16283 + }, + { + "epoch": 1.37, + "learning_rate": 1.1866919708306992e-05, + "loss": 1.1494, + "step": 16284 + }, + { + "epoch": 1.37, + "learning_rate": 1.186602670865639e-05, + "loss": 1.2053, + "step": 16285 + }, + { + "epoch": 1.37, + "learning_rate": 1.1865133693588078e-05, + "loss": 1.1861, + "step": 16286 + }, + { + "epoch": 1.37, + "learning_rate": 1.1864240663109435e-05, + "loss": 1.2362, + "step": 16287 + }, + { + "epoch": 1.37, + "learning_rate": 1.1863347617227836e-05, + "loss": 1.2002, + "step": 16288 + }, + { + "epoch": 1.37, + "learning_rate": 1.1862454555950667e-05, + "loss": 1.2265, + "step": 16289 + }, + { + "epoch": 1.37, + "learning_rate": 1.1861561479285299e-05, + "loss": 1.2482, + "step": 16290 + }, + { + "epoch": 1.37, + "learning_rate": 1.1860668387239115e-05, + "loss": 1.195, + "step": 16291 + }, + { + "epoch": 1.37, + "learning_rate": 1.1859775279819494e-05, + "loss": 1.1927, + "step": 16292 + }, + { + "epoch": 1.37, + "learning_rate": 1.1858882157033815e-05, + "loss": 1.1396, + "step": 16293 + }, + { + "epoch": 1.37, + "learning_rate": 1.1857989018889455e-05, + "loss": 1.2712, + "step": 16294 + }, + { + "epoch": 1.37, + "learning_rate": 1.1857095865393796e-05, + "loss": 1.2031, + "step": 16295 + }, + { + "epoch": 1.37, + "learning_rate": 1.1856202696554216e-05, + "loss": 1.1758, + "step": 16296 + }, + { + "epoch": 1.37, + "learning_rate": 1.1855309512378095e-05, + "loss": 1.2687, + "step": 16297 + }, + { + "epoch": 1.37, + "learning_rate": 1.1854416312872813e-05, + "loss": 1.2252, + "step": 16298 + }, + { + "epoch": 1.37, + "learning_rate": 1.1853523098045751e-05, + "loss": 1.1385, + "step": 16299 + }, + { + "epoch": 1.37, + "learning_rate": 1.1852629867904288e-05, + "loss": 1.2461, + "step": 16300 + }, + { + "epoch": 1.37, + "learning_rate": 1.1851736622455806e-05, + "loss": 1.2014, + "step": 16301 + }, + { + "epoch": 1.37, + "learning_rate": 1.1850843361707681e-05, + "loss": 1.1974, + "step": 16302 + }, + { + "epoch": 1.37, + "learning_rate": 1.1849950085667294e-05, + "loss": 1.2602, + "step": 16303 + }, + { + "epoch": 1.37, + "learning_rate": 1.184905679434203e-05, + "loss": 1.1181, + "step": 16304 + }, + { + "epoch": 1.37, + "learning_rate": 1.1848163487739265e-05, + "loss": 1.2401, + "step": 16305 + }, + { + "epoch": 1.37, + "learning_rate": 1.1847270165866385e-05, + "loss": 1.2071, + "step": 16306 + }, + { + "epoch": 1.37, + "learning_rate": 1.1846376828730768e-05, + "loss": 1.16, + "step": 16307 + }, + { + "epoch": 1.37, + "learning_rate": 1.184548347633979e-05, + "loss": 1.2249, + "step": 16308 + }, + { + "epoch": 1.37, + "learning_rate": 1.1844590108700837e-05, + "loss": 1.2021, + "step": 16309 + }, + { + "epoch": 1.37, + "learning_rate": 1.1843696725821295e-05, + "loss": 1.2651, + "step": 16310 + }, + { + "epoch": 1.37, + "learning_rate": 1.1842803327708539e-05, + "loss": 1.1758, + "step": 16311 + }, + { + "epoch": 1.37, + "learning_rate": 1.184190991436995e-05, + "loss": 1.2985, + "step": 16312 + }, + { + "epoch": 1.37, + "learning_rate": 1.1841016485812912e-05, + "loss": 1.2726, + "step": 16313 + }, + { + "epoch": 1.37, + "learning_rate": 1.184012304204481e-05, + "loss": 1.165, + "step": 16314 + }, + { + "epoch": 1.37, + "learning_rate": 1.183922958307302e-05, + "loss": 1.2523, + "step": 16315 + }, + { + "epoch": 1.37, + "learning_rate": 1.1838336108904927e-05, + "loss": 1.209, + "step": 16316 + }, + { + "epoch": 1.37, + "learning_rate": 1.1837442619547912e-05, + "loss": 1.2165, + "step": 16317 + }, + { + "epoch": 1.37, + "learning_rate": 1.183654911500936e-05, + "loss": 1.1431, + "step": 16318 + }, + { + "epoch": 1.37, + "learning_rate": 1.1835655595296649e-05, + "loss": 1.1473, + "step": 16319 + }, + { + "epoch": 1.37, + "learning_rate": 1.1834762060417163e-05, + "loss": 1.3005, + "step": 16320 + }, + { + "epoch": 1.37, + "learning_rate": 1.1833868510378287e-05, + "loss": 1.1706, + "step": 16321 + }, + { + "epoch": 1.37, + "learning_rate": 1.1832974945187404e-05, + "loss": 1.1604, + "step": 16322 + }, + { + "epoch": 1.37, + "learning_rate": 1.1832081364851893e-05, + "loss": 1.2058, + "step": 16323 + }, + { + "epoch": 1.37, + "learning_rate": 1.183118776937914e-05, + "loss": 1.1907, + "step": 16324 + }, + { + "epoch": 1.37, + "learning_rate": 1.1830294158776532e-05, + "loss": 1.1983, + "step": 16325 + }, + { + "epoch": 1.37, + "learning_rate": 1.1829400533051443e-05, + "loss": 1.1343, + "step": 16326 + }, + { + "epoch": 1.37, + "learning_rate": 1.1828506892211262e-05, + "loss": 1.1614, + "step": 16327 + }, + { + "epoch": 1.37, + "learning_rate": 1.1827613236263372e-05, + "loss": 1.2421, + "step": 16328 + }, + { + "epoch": 1.37, + "learning_rate": 1.1826719565215159e-05, + "loss": 1.2081, + "step": 16329 + }, + { + "epoch": 1.37, + "learning_rate": 1.1825825879074003e-05, + "loss": 1.2728, + "step": 16330 + }, + { + "epoch": 1.37, + "learning_rate": 1.1824932177847286e-05, + "loss": 1.1665, + "step": 16331 + }, + { + "epoch": 1.38, + "learning_rate": 1.18240384615424e-05, + "loss": 1.1606, + "step": 16332 + }, + { + "epoch": 1.38, + "learning_rate": 1.182314473016672e-05, + "loss": 1.1483, + "step": 16333 + }, + { + "epoch": 1.38, + "learning_rate": 1.1822250983727636e-05, + "loss": 1.2215, + "step": 16334 + }, + { + "epoch": 1.38, + "learning_rate": 1.1821357222232535e-05, + "loss": 1.1966, + "step": 16335 + }, + { + "epoch": 1.38, + "learning_rate": 1.1820463445688794e-05, + "loss": 1.1982, + "step": 16336 + }, + { + "epoch": 1.38, + "learning_rate": 1.1819569654103802e-05, + "loss": 1.2039, + "step": 16337 + }, + { + "epoch": 1.38, + "learning_rate": 1.1818675847484943e-05, + "loss": 1.2131, + "step": 16338 + }, + { + "epoch": 1.38, + "learning_rate": 1.1817782025839605e-05, + "loss": 1.221, + "step": 16339 + }, + { + "epoch": 1.38, + "learning_rate": 1.1816888189175169e-05, + "loss": 1.3007, + "step": 16340 + }, + { + "epoch": 1.38, + "learning_rate": 1.1815994337499018e-05, + "loss": 1.1541, + "step": 16341 + }, + { + "epoch": 1.38, + "learning_rate": 1.1815100470818544e-05, + "loss": 1.1928, + "step": 16342 + }, + { + "epoch": 1.38, + "learning_rate": 1.1814206589141128e-05, + "loss": 1.2031, + "step": 16343 + }, + { + "epoch": 1.38, + "learning_rate": 1.1813312692474155e-05, + "loss": 1.1719, + "step": 16344 + }, + { + "epoch": 1.38, + "learning_rate": 1.181241878082501e-05, + "loss": 1.2195, + "step": 16345 + }, + { + "epoch": 1.38, + "learning_rate": 1.1811524854201087e-05, + "loss": 1.1878, + "step": 16346 + }, + { + "epoch": 1.38, + "learning_rate": 1.1810630912609764e-05, + "loss": 1.1001, + "step": 16347 + }, + { + "epoch": 1.38, + "learning_rate": 1.1809736956058427e-05, + "loss": 1.167, + "step": 16348 + }, + { + "epoch": 1.38, + "learning_rate": 1.1808842984554465e-05, + "loss": 1.14, + "step": 16349 + }, + { + "epoch": 1.38, + "learning_rate": 1.1807948998105265e-05, + "loss": 1.1988, + "step": 16350 + }, + { + "epoch": 1.38, + "learning_rate": 1.1807054996718211e-05, + "loss": 1.2758, + "step": 16351 + }, + { + "epoch": 1.38, + "learning_rate": 1.180616098040069e-05, + "loss": 1.1865, + "step": 16352 + }, + { + "epoch": 1.38, + "learning_rate": 1.1805266949160088e-05, + "loss": 1.1738, + "step": 16353 + }, + { + "epoch": 1.38, + "learning_rate": 1.1804372903003794e-05, + "loss": 1.1242, + "step": 16354 + }, + { + "epoch": 1.38, + "learning_rate": 1.1803478841939194e-05, + "loss": 1.2142, + "step": 16355 + }, + { + "epoch": 1.38, + "learning_rate": 1.1802584765973672e-05, + "loss": 1.1775, + "step": 16356 + }, + { + "epoch": 1.38, + "learning_rate": 1.1801690675114619e-05, + "loss": 1.1242, + "step": 16357 + }, + { + "epoch": 1.38, + "learning_rate": 1.1800796569369421e-05, + "loss": 1.1334, + "step": 16358 + }, + { + "epoch": 1.38, + "learning_rate": 1.1799902448745466e-05, + "loss": 1.2902, + "step": 16359 + }, + { + "epoch": 1.38, + "learning_rate": 1.1799008313250142e-05, + "loss": 1.245, + "step": 16360 + }, + { + "epoch": 1.38, + "learning_rate": 1.1798114162890832e-05, + "loss": 1.2581, + "step": 16361 + }, + { + "epoch": 1.38, + "learning_rate": 1.1797219997674931e-05, + "loss": 1.1355, + "step": 16362 + }, + { + "epoch": 1.38, + "learning_rate": 1.1796325817609822e-05, + "loss": 1.1607, + "step": 16363 + }, + { + "epoch": 1.38, + "learning_rate": 1.1795431622702895e-05, + "loss": 1.1713, + "step": 16364 + }, + { + "epoch": 1.38, + "learning_rate": 1.1794537412961535e-05, + "loss": 1.1591, + "step": 16365 + }, + { + "epoch": 1.38, + "learning_rate": 1.1793643188393134e-05, + "loss": 1.1587, + "step": 16366 + }, + { + "epoch": 1.38, + "learning_rate": 1.1792748949005076e-05, + "loss": 1.1704, + "step": 16367 + }, + { + "epoch": 1.38, + "learning_rate": 1.1791854694804759e-05, + "loss": 1.2341, + "step": 16368 + }, + { + "epoch": 1.38, + "learning_rate": 1.179096042579956e-05, + "loss": 1.1547, + "step": 16369 + }, + { + "epoch": 1.38, + "learning_rate": 1.1790066141996872e-05, + "loss": 1.1897, + "step": 16370 + }, + { + "epoch": 1.38, + "learning_rate": 1.1789171843404088e-05, + "loss": 1.1843, + "step": 16371 + }, + { + "epoch": 1.38, + "learning_rate": 1.1788277530028594e-05, + "loss": 1.1684, + "step": 16372 + }, + { + "epoch": 1.38, + "learning_rate": 1.1787383201877776e-05, + "loss": 1.1436, + "step": 16373 + }, + { + "epoch": 1.38, + "learning_rate": 1.1786488858959026e-05, + "loss": 1.1593, + "step": 16374 + }, + { + "epoch": 1.38, + "learning_rate": 1.1785594501279736e-05, + "loss": 1.2654, + "step": 16375 + }, + { + "epoch": 1.38, + "learning_rate": 1.178470012884729e-05, + "loss": 1.2427, + "step": 16376 + }, + { + "epoch": 1.38, + "learning_rate": 1.1783805741669082e-05, + "loss": 1.1521, + "step": 16377 + }, + { + "epoch": 1.38, + "learning_rate": 1.1782911339752498e-05, + "loss": 1.0968, + "step": 16378 + }, + { + "epoch": 1.38, + "learning_rate": 1.1782016923104933e-05, + "loss": 1.1914, + "step": 16379 + }, + { + "epoch": 1.38, + "learning_rate": 1.1781122491733771e-05, + "loss": 1.21, + "step": 16380 + }, + { + "epoch": 1.38, + "learning_rate": 1.1780228045646407e-05, + "loss": 1.17, + "step": 16381 + }, + { + "epoch": 1.38, + "learning_rate": 1.1779333584850227e-05, + "loss": 1.2111, + "step": 16382 + }, + { + "epoch": 1.38, + "learning_rate": 1.1778439109352627e-05, + "loss": 1.2177, + "step": 16383 + }, + { + "epoch": 1.38, + "learning_rate": 1.177754461916099e-05, + "loss": 1.2102, + "step": 16384 + }, + { + "epoch": 1.38, + "learning_rate": 1.1776650114282715e-05, + "loss": 1.1441, + "step": 16385 + }, + { + "epoch": 1.38, + "learning_rate": 1.1775755594725185e-05, + "loss": 1.252, + "step": 16386 + }, + { + "epoch": 1.38, + "learning_rate": 1.1774861060495792e-05, + "loss": 1.1946, + "step": 16387 + }, + { + "epoch": 1.38, + "learning_rate": 1.177396651160193e-05, + "loss": 1.1814, + "step": 16388 + }, + { + "epoch": 1.38, + "learning_rate": 1.1773071948050994e-05, + "loss": 1.2328, + "step": 16389 + }, + { + "epoch": 1.38, + "learning_rate": 1.1772177369850364e-05, + "loss": 1.1851, + "step": 16390 + }, + { + "epoch": 1.38, + "learning_rate": 1.177128277700744e-05, + "loss": 1.1793, + "step": 16391 + }, + { + "epoch": 1.38, + "learning_rate": 1.177038816952961e-05, + "loss": 1.2071, + "step": 16392 + }, + { + "epoch": 1.38, + "learning_rate": 1.1769493547424269e-05, + "loss": 1.2428, + "step": 16393 + }, + { + "epoch": 1.38, + "learning_rate": 1.1768598910698804e-05, + "loss": 1.2458, + "step": 16394 + }, + { + "epoch": 1.38, + "learning_rate": 1.1767704259360606e-05, + "loss": 1.1974, + "step": 16395 + }, + { + "epoch": 1.38, + "learning_rate": 1.1766809593417072e-05, + "loss": 1.1808, + "step": 16396 + }, + { + "epoch": 1.38, + "learning_rate": 1.1765914912875593e-05, + "loss": 1.1774, + "step": 16397 + }, + { + "epoch": 1.38, + "learning_rate": 1.176502021774356e-05, + "loss": 1.2367, + "step": 16398 + }, + { + "epoch": 1.38, + "learning_rate": 1.176412550802836e-05, + "loss": 1.0858, + "step": 16399 + }, + { + "epoch": 1.38, + "learning_rate": 1.1763230783737399e-05, + "loss": 1.2693, + "step": 16400 + }, + { + "epoch": 1.38, + "learning_rate": 1.1762336044878056e-05, + "loss": 1.1963, + "step": 16401 + }, + { + "epoch": 1.38, + "learning_rate": 1.1761441291457726e-05, + "loss": 1.1587, + "step": 16402 + }, + { + "epoch": 1.38, + "learning_rate": 1.1760546523483806e-05, + "loss": 1.2342, + "step": 16403 + }, + { + "epoch": 1.38, + "learning_rate": 1.1759651740963692e-05, + "loss": 1.201, + "step": 16404 + }, + { + "epoch": 1.38, + "learning_rate": 1.1758756943904768e-05, + "loss": 1.1801, + "step": 16405 + }, + { + "epoch": 1.38, + "learning_rate": 1.175786213231443e-05, + "loss": 1.2993, + "step": 16406 + }, + { + "epoch": 1.38, + "learning_rate": 1.1756967306200075e-05, + "loss": 1.1765, + "step": 16407 + }, + { + "epoch": 1.38, + "learning_rate": 1.1756072465569092e-05, + "loss": 1.1858, + "step": 16408 + }, + { + "epoch": 1.38, + "learning_rate": 1.1755177610428877e-05, + "loss": 1.2038, + "step": 16409 + }, + { + "epoch": 1.38, + "learning_rate": 1.1754282740786822e-05, + "loss": 1.2257, + "step": 16410 + }, + { + "epoch": 1.38, + "learning_rate": 1.1753387856650325e-05, + "loss": 1.224, + "step": 16411 + }, + { + "epoch": 1.38, + "learning_rate": 1.1752492958026774e-05, + "loss": 1.2063, + "step": 16412 + }, + { + "epoch": 1.38, + "learning_rate": 1.1751598044923565e-05, + "loss": 1.2046, + "step": 16413 + }, + { + "epoch": 1.38, + "learning_rate": 1.1750703117348094e-05, + "loss": 1.1336, + "step": 16414 + }, + { + "epoch": 1.38, + "learning_rate": 1.1749808175307754e-05, + "loss": 1.1164, + "step": 16415 + }, + { + "epoch": 1.38, + "learning_rate": 1.1748913218809934e-05, + "loss": 1.2335, + "step": 16416 + }, + { + "epoch": 1.38, + "learning_rate": 1.1748018247862036e-05, + "loss": 1.1798, + "step": 16417 + }, + { + "epoch": 1.38, + "learning_rate": 1.1747123262471454e-05, + "loss": 1.2201, + "step": 16418 + }, + { + "epoch": 1.38, + "learning_rate": 1.1746228262645579e-05, + "loss": 1.1382, + "step": 16419 + }, + { + "epoch": 1.38, + "learning_rate": 1.1745333248391803e-05, + "loss": 1.2139, + "step": 16420 + }, + { + "epoch": 1.38, + "learning_rate": 1.174443821971753e-05, + "loss": 1.2813, + "step": 16421 + }, + { + "epoch": 1.38, + "learning_rate": 1.174354317663015e-05, + "loss": 1.2359, + "step": 16422 + }, + { + "epoch": 1.38, + "learning_rate": 1.1742648119137057e-05, + "loss": 1.2018, + "step": 16423 + }, + { + "epoch": 1.38, + "learning_rate": 1.1741753047245648e-05, + "loss": 1.2363, + "step": 16424 + }, + { + "epoch": 1.38, + "learning_rate": 1.174085796096332e-05, + "loss": 1.2439, + "step": 16425 + }, + { + "epoch": 1.38, + "learning_rate": 1.1739962860297465e-05, + "loss": 1.1439, + "step": 16426 + }, + { + "epoch": 1.38, + "learning_rate": 1.1739067745255476e-05, + "loss": 1.1776, + "step": 16427 + }, + { + "epoch": 1.38, + "learning_rate": 1.1738172615844756e-05, + "loss": 1.1618, + "step": 16428 + }, + { + "epoch": 1.38, + "learning_rate": 1.17372774720727e-05, + "loss": 1.1374, + "step": 16429 + }, + { + "epoch": 1.38, + "learning_rate": 1.17363823139467e-05, + "loss": 1.1426, + "step": 16430 + }, + { + "epoch": 1.38, + "learning_rate": 1.173548714147415e-05, + "loss": 1.1471, + "step": 16431 + }, + { + "epoch": 1.38, + "learning_rate": 1.1734591954662454e-05, + "loss": 1.1593, + "step": 16432 + }, + { + "epoch": 1.38, + "learning_rate": 1.1733696753519002e-05, + "loss": 1.2091, + "step": 16433 + }, + { + "epoch": 1.38, + "learning_rate": 1.1732801538051193e-05, + "loss": 1.2097, + "step": 16434 + }, + { + "epoch": 1.38, + "learning_rate": 1.1731906308266421e-05, + "loss": 1.1598, + "step": 16435 + }, + { + "epoch": 1.38, + "learning_rate": 1.1731011064172089e-05, + "loss": 1.2954, + "step": 16436 + }, + { + "epoch": 1.38, + "learning_rate": 1.1730115805775588e-05, + "loss": 1.0883, + "step": 16437 + }, + { + "epoch": 1.38, + "learning_rate": 1.1729220533084316e-05, + "loss": 1.2415, + "step": 16438 + }, + { + "epoch": 1.38, + "learning_rate": 1.1728325246105671e-05, + "loss": 1.162, + "step": 16439 + }, + { + "epoch": 1.38, + "learning_rate": 1.1727429944847048e-05, + "loss": 1.2393, + "step": 16440 + }, + { + "epoch": 1.38, + "learning_rate": 1.1726534629315849e-05, + "loss": 1.1119, + "step": 16441 + }, + { + "epoch": 1.38, + "learning_rate": 1.1725639299519466e-05, + "loss": 1.1192, + "step": 16442 + }, + { + "epoch": 1.38, + "learning_rate": 1.1724743955465302e-05, + "loss": 1.1086, + "step": 16443 + }, + { + "epoch": 1.38, + "learning_rate": 1.1723848597160748e-05, + "loss": 1.1576, + "step": 16444 + }, + { + "epoch": 1.38, + "learning_rate": 1.1722953224613203e-05, + "loss": 1.1604, + "step": 16445 + }, + { + "epoch": 1.38, + "learning_rate": 1.1722057837830071e-05, + "loss": 1.1782, + "step": 16446 + }, + { + "epoch": 1.38, + "learning_rate": 1.1721162436818745e-05, + "loss": 1.2203, + "step": 16447 + }, + { + "epoch": 1.38, + "learning_rate": 1.1720267021586624e-05, + "loss": 1.2271, + "step": 16448 + }, + { + "epoch": 1.38, + "learning_rate": 1.1719371592141107e-05, + "loss": 1.2733, + "step": 16449 + }, + { + "epoch": 1.38, + "learning_rate": 1.1718476148489593e-05, + "loss": 1.3007, + "step": 16450 + }, + { + "epoch": 1.39, + "learning_rate": 1.1717580690639475e-05, + "loss": 1.1641, + "step": 16451 + }, + { + "epoch": 1.39, + "learning_rate": 1.1716685218598158e-05, + "loss": 1.2879, + "step": 16452 + }, + { + "epoch": 1.39, + "learning_rate": 1.171578973237304e-05, + "loss": 1.2247, + "step": 16453 + }, + { + "epoch": 1.39, + "learning_rate": 1.1714894231971517e-05, + "loss": 1.1405, + "step": 16454 + }, + { + "epoch": 1.39, + "learning_rate": 1.1713998717400987e-05, + "loss": 1.1663, + "step": 16455 + }, + { + "epoch": 1.39, + "learning_rate": 1.1713103188668853e-05, + "loss": 1.2081, + "step": 16456 + }, + { + "epoch": 1.39, + "learning_rate": 1.1712207645782512e-05, + "loss": 1.161, + "step": 16457 + }, + { + "epoch": 1.39, + "learning_rate": 1.1711312088749363e-05, + "loss": 1.1874, + "step": 16458 + }, + { + "epoch": 1.39, + "learning_rate": 1.1710416517576803e-05, + "loss": 1.2348, + "step": 16459 + }, + { + "epoch": 1.39, + "learning_rate": 1.1709520932272238e-05, + "loss": 1.1476, + "step": 16460 + }, + { + "epoch": 1.39, + "learning_rate": 1.1708625332843064e-05, + "loss": 1.1245, + "step": 16461 + }, + { + "epoch": 1.39, + "learning_rate": 1.1707729719296681e-05, + "loss": 1.128, + "step": 16462 + }, + { + "epoch": 1.39, + "learning_rate": 1.1706834091640487e-05, + "loss": 1.1737, + "step": 16463 + }, + { + "epoch": 1.39, + "learning_rate": 1.1705938449881887e-05, + "loss": 1.2125, + "step": 16464 + }, + { + "epoch": 1.39, + "learning_rate": 1.1705042794028272e-05, + "loss": 1.208, + "step": 16465 + }, + { + "epoch": 1.39, + "learning_rate": 1.1704147124087051e-05, + "loss": 1.2005, + "step": 16466 + }, + { + "epoch": 1.39, + "learning_rate": 1.170325144006562e-05, + "loss": 1.2568, + "step": 16467 + }, + { + "epoch": 1.39, + "learning_rate": 1.170235574197138e-05, + "loss": 1.2462, + "step": 16468 + }, + { + "epoch": 1.39, + "learning_rate": 1.1701460029811733e-05, + "loss": 1.2536, + "step": 16469 + }, + { + "epoch": 1.39, + "learning_rate": 1.1700564303594075e-05, + "loss": 1.176, + "step": 16470 + }, + { + "epoch": 1.39, + "learning_rate": 1.1699668563325816e-05, + "loss": 1.2337, + "step": 16471 + }, + { + "epoch": 1.39, + "learning_rate": 1.169877280901435e-05, + "loss": 1.1442, + "step": 16472 + }, + { + "epoch": 1.39, + "learning_rate": 1.1697877040667079e-05, + "loss": 1.2092, + "step": 16473 + }, + { + "epoch": 1.39, + "learning_rate": 1.1696981258291403e-05, + "loss": 1.172, + "step": 16474 + }, + { + "epoch": 1.39, + "learning_rate": 1.1696085461894725e-05, + "loss": 1.2228, + "step": 16475 + }, + { + "epoch": 1.39, + "learning_rate": 1.1695189651484449e-05, + "loss": 1.1777, + "step": 16476 + }, + { + "epoch": 1.39, + "learning_rate": 1.169429382706797e-05, + "loss": 1.2644, + "step": 16477 + }, + { + "epoch": 1.39, + "learning_rate": 1.1693397988652694e-05, + "loss": 1.2145, + "step": 16478 + }, + { + "epoch": 1.39, + "learning_rate": 1.1692502136246025e-05, + "loss": 1.2659, + "step": 16479 + }, + { + "epoch": 1.39, + "learning_rate": 1.169160626985536e-05, + "loss": 1.2349, + "step": 16480 + }, + { + "epoch": 1.39, + "learning_rate": 1.16907103894881e-05, + "loss": 1.1541, + "step": 16481 + }, + { + "epoch": 1.39, + "learning_rate": 1.1689814495151652e-05, + "loss": 1.2007, + "step": 16482 + }, + { + "epoch": 1.39, + "learning_rate": 1.1688918586853415e-05, + "loss": 1.1162, + "step": 16483 + }, + { + "epoch": 1.39, + "learning_rate": 1.1688022664600791e-05, + "loss": 1.1102, + "step": 16484 + }, + { + "epoch": 1.39, + "learning_rate": 1.1687126728401185e-05, + "loss": 1.2728, + "step": 16485 + }, + { + "epoch": 1.39, + "learning_rate": 1.1686230778262e-05, + "loss": 1.0745, + "step": 16486 + }, + { + "epoch": 1.39, + "learning_rate": 1.1685334814190635e-05, + "loss": 1.1967, + "step": 16487 + }, + { + "epoch": 1.39, + "learning_rate": 1.1684438836194493e-05, + "loss": 1.1505, + "step": 16488 + }, + { + "epoch": 1.39, + "learning_rate": 1.1683542844280982e-05, + "loss": 1.1538, + "step": 16489 + }, + { + "epoch": 1.39, + "learning_rate": 1.1682646838457496e-05, + "loss": 1.2228, + "step": 16490 + }, + { + "epoch": 1.39, + "learning_rate": 1.1681750818731446e-05, + "loss": 1.207, + "step": 16491 + }, + { + "epoch": 1.39, + "learning_rate": 1.1680854785110232e-05, + "loss": 1.1453, + "step": 16492 + }, + { + "epoch": 1.39, + "learning_rate": 1.1679958737601261e-05, + "loss": 1.2586, + "step": 16493 + }, + { + "epoch": 1.39, + "learning_rate": 1.167906267621193e-05, + "loss": 1.198, + "step": 16494 + }, + { + "epoch": 1.39, + "learning_rate": 1.1678166600949644e-05, + "loss": 1.1846, + "step": 16495 + }, + { + "epoch": 1.39, + "learning_rate": 1.1677270511821812e-05, + "loss": 1.1394, + "step": 16496 + }, + { + "epoch": 1.39, + "learning_rate": 1.1676374408835832e-05, + "loss": 1.2043, + "step": 16497 + }, + { + "epoch": 1.39, + "learning_rate": 1.167547829199911e-05, + "loss": 1.1669, + "step": 16498 + }, + { + "epoch": 1.39, + "learning_rate": 1.167458216131905e-05, + "loss": 1.2123, + "step": 16499 + }, + { + "epoch": 1.39, + "learning_rate": 1.1673686016803062e-05, + "loss": 1.1977, + "step": 16500 + }, + { + "epoch": 1.39, + "learning_rate": 1.1672789858458537e-05, + "loss": 1.217, + "step": 16501 + }, + { + "epoch": 1.39, + "learning_rate": 1.167189368629289e-05, + "loss": 1.158, + "step": 16502 + }, + { + "epoch": 1.39, + "learning_rate": 1.1670997500313521e-05, + "loss": 1.1747, + "step": 16503 + }, + { + "epoch": 1.39, + "learning_rate": 1.1670101300527838e-05, + "loss": 1.1384, + "step": 16504 + }, + { + "epoch": 1.39, + "learning_rate": 1.1669205086943243e-05, + "loss": 1.2429, + "step": 16505 + }, + { + "epoch": 1.39, + "learning_rate": 1.1668308859567141e-05, + "loss": 1.152, + "step": 16506 + }, + { + "epoch": 1.39, + "learning_rate": 1.1667412618406938e-05, + "loss": 1.1869, + "step": 16507 + }, + { + "epoch": 1.39, + "learning_rate": 1.1666516363470035e-05, + "loss": 1.2562, + "step": 16508 + }, + { + "epoch": 1.39, + "learning_rate": 1.1665620094763842e-05, + "loss": 1.1941, + "step": 16509 + }, + { + "epoch": 1.39, + "learning_rate": 1.1664723812295763e-05, + "loss": 1.1535, + "step": 16510 + }, + { + "epoch": 1.39, + "learning_rate": 1.1663827516073204e-05, + "loss": 1.0775, + "step": 16511 + }, + { + "epoch": 1.39, + "learning_rate": 1.166293120610357e-05, + "loss": 1.2484, + "step": 16512 + }, + { + "epoch": 1.39, + "learning_rate": 1.1662034882394266e-05, + "loss": 1.1856, + "step": 16513 + }, + { + "epoch": 1.39, + "learning_rate": 1.1661138544952699e-05, + "loss": 1.1824, + "step": 16514 + }, + { + "epoch": 1.39, + "learning_rate": 1.1660242193786272e-05, + "loss": 1.2153, + "step": 16515 + }, + { + "epoch": 1.39, + "learning_rate": 1.165934582890239e-05, + "loss": 1.1664, + "step": 16516 + }, + { + "epoch": 1.39, + "learning_rate": 1.1658449450308464e-05, + "loss": 1.2568, + "step": 16517 + }, + { + "epoch": 1.39, + "learning_rate": 1.16575530580119e-05, + "loss": 1.2577, + "step": 16518 + }, + { + "epoch": 1.39, + "learning_rate": 1.1656656652020102e-05, + "loss": 1.1007, + "step": 16519 + }, + { + "epoch": 1.39, + "learning_rate": 1.1655760232340472e-05, + "loss": 1.0732, + "step": 16520 + }, + { + "epoch": 1.39, + "learning_rate": 1.1654863798980426e-05, + "loss": 1.2755, + "step": 16521 + }, + { + "epoch": 1.39, + "learning_rate": 1.1653967351947362e-05, + "loss": 1.1582, + "step": 16522 + }, + { + "epoch": 1.39, + "learning_rate": 1.1653070891248692e-05, + "loss": 1.2164, + "step": 16523 + }, + { + "epoch": 1.39, + "learning_rate": 1.165217441689182e-05, + "loss": 1.2531, + "step": 16524 + }, + { + "epoch": 1.39, + "learning_rate": 1.1651277928884157e-05, + "loss": 1.0923, + "step": 16525 + }, + { + "epoch": 1.39, + "learning_rate": 1.1650381427233106e-05, + "loss": 1.2177, + "step": 16526 + }, + { + "epoch": 1.39, + "learning_rate": 1.1649484911946075e-05, + "loss": 1.3041, + "step": 16527 + }, + { + "epoch": 1.39, + "learning_rate": 1.164858838303047e-05, + "loss": 1.2052, + "step": 16528 + }, + { + "epoch": 1.39, + "learning_rate": 1.1647691840493705e-05, + "loss": 1.0999, + "step": 16529 + }, + { + "epoch": 1.39, + "learning_rate": 1.1646795284343178e-05, + "loss": 1.1964, + "step": 16530 + }, + { + "epoch": 1.39, + "learning_rate": 1.1645898714586303e-05, + "loss": 1.287, + "step": 16531 + }, + { + "epoch": 1.39, + "learning_rate": 1.1645002131230486e-05, + "loss": 1.0749, + "step": 16532 + }, + { + "epoch": 1.39, + "learning_rate": 1.1644105534283134e-05, + "loss": 1.1858, + "step": 16533 + }, + { + "epoch": 1.39, + "learning_rate": 1.1643208923751656e-05, + "loss": 1.135, + "step": 16534 + }, + { + "epoch": 1.39, + "learning_rate": 1.1642312299643461e-05, + "loss": 1.3322, + "step": 16535 + }, + { + "epoch": 1.39, + "learning_rate": 1.1641415661965958e-05, + "loss": 1.2072, + "step": 16536 + }, + { + "epoch": 1.39, + "learning_rate": 1.1640519010726553e-05, + "loss": 1.1637, + "step": 16537 + }, + { + "epoch": 1.39, + "learning_rate": 1.1639622345932651e-05, + "loss": 1.1425, + "step": 16538 + }, + { + "epoch": 1.39, + "learning_rate": 1.1638725667591668e-05, + "loss": 1.1378, + "step": 16539 + }, + { + "epoch": 1.39, + "learning_rate": 1.163782897571101e-05, + "loss": 1.2754, + "step": 16540 + }, + { + "epoch": 1.39, + "learning_rate": 1.1636932270298083e-05, + "loss": 1.1713, + "step": 16541 + }, + { + "epoch": 1.39, + "learning_rate": 1.1636035551360296e-05, + "loss": 1.2069, + "step": 16542 + }, + { + "epoch": 1.39, + "learning_rate": 1.1635138818905062e-05, + "loss": 1.1203, + "step": 16543 + }, + { + "epoch": 1.39, + "learning_rate": 1.1634242072939787e-05, + "loss": 1.1902, + "step": 16544 + }, + { + "epoch": 1.39, + "learning_rate": 1.1633345313471878e-05, + "loss": 1.1235, + "step": 16545 + }, + { + "epoch": 1.39, + "learning_rate": 1.1632448540508751e-05, + "loss": 1.2149, + "step": 16546 + }, + { + "epoch": 1.39, + "learning_rate": 1.1631551754057813e-05, + "loss": 1.243, + "step": 16547 + }, + { + "epoch": 1.39, + "learning_rate": 1.1630654954126468e-05, + "loss": 1.2579, + "step": 16548 + }, + { + "epoch": 1.39, + "learning_rate": 1.1629758140722132e-05, + "loss": 1.1466, + "step": 16549 + }, + { + "epoch": 1.39, + "learning_rate": 1.1628861313852215e-05, + "loss": 1.1675, + "step": 16550 + }, + { + "epoch": 1.39, + "learning_rate": 1.162796447352412e-05, + "loss": 1.2497, + "step": 16551 + }, + { + "epoch": 1.39, + "learning_rate": 1.1627067619745265e-05, + "loss": 1.1809, + "step": 16552 + }, + { + "epoch": 1.39, + "learning_rate": 1.1626170752523054e-05, + "loss": 1.1902, + "step": 16553 + }, + { + "epoch": 1.39, + "learning_rate": 1.1625273871864904e-05, + "loss": 1.2635, + "step": 16554 + }, + { + "epoch": 1.39, + "learning_rate": 1.1624376977778218e-05, + "loss": 1.2462, + "step": 16555 + }, + { + "epoch": 1.39, + "learning_rate": 1.1623480070270409e-05, + "loss": 1.1522, + "step": 16556 + }, + { + "epoch": 1.39, + "learning_rate": 1.1622583149348888e-05, + "loss": 1.1369, + "step": 16557 + }, + { + "epoch": 1.39, + "learning_rate": 1.1621686215021066e-05, + "loss": 1.1834, + "step": 16558 + }, + { + "epoch": 1.39, + "learning_rate": 1.1620789267294354e-05, + "loss": 1.1814, + "step": 16559 + }, + { + "epoch": 1.39, + "learning_rate": 1.1619892306176161e-05, + "loss": 1.1802, + "step": 16560 + }, + { + "epoch": 1.39, + "learning_rate": 1.1618995331673904e-05, + "loss": 1.173, + "step": 16561 + }, + { + "epoch": 1.39, + "learning_rate": 1.1618098343794986e-05, + "loss": 1.0525, + "step": 16562 + }, + { + "epoch": 1.39, + "learning_rate": 1.161720134254682e-05, + "loss": 1.1946, + "step": 16563 + }, + { + "epoch": 1.39, + "learning_rate": 1.161630432793682e-05, + "loss": 1.1797, + "step": 16564 + }, + { + "epoch": 1.39, + "learning_rate": 1.1615407299972402e-05, + "loss": 1.1203, + "step": 16565 + }, + { + "epoch": 1.39, + "learning_rate": 1.1614510258660964e-05, + "loss": 1.1829, + "step": 16566 + }, + { + "epoch": 1.39, + "learning_rate": 1.1613613204009927e-05, + "loss": 1.2342, + "step": 16567 + }, + { + "epoch": 1.39, + "learning_rate": 1.1612716136026707e-05, + "loss": 1.2211, + "step": 16568 + }, + { + "epoch": 1.39, + "learning_rate": 1.1611819054718704e-05, + "loss": 1.1684, + "step": 16569 + }, + { + "epoch": 1.4, + "learning_rate": 1.1610921960093335e-05, + "loss": 1.1864, + "step": 16570 + }, + { + "epoch": 1.4, + "learning_rate": 1.161002485215802e-05, + "loss": 1.1106, + "step": 16571 + }, + { + "epoch": 1.4, + "learning_rate": 1.160912773092016e-05, + "loss": 1.2024, + "step": 16572 + }, + { + "epoch": 1.4, + "learning_rate": 1.160823059638717e-05, + "loss": 1.1914, + "step": 16573 + }, + { + "epoch": 1.4, + "learning_rate": 1.1607333448566467e-05, + "loss": 1.1925, + "step": 16574 + }, + { + "epoch": 1.4, + "learning_rate": 1.1606436287465464e-05, + "loss": 1.1868, + "step": 16575 + }, + { + "epoch": 1.4, + "learning_rate": 1.1605539113091564e-05, + "loss": 1.1426, + "step": 16576 + }, + { + "epoch": 1.4, + "learning_rate": 1.1604641925452189e-05, + "loss": 1.2183, + "step": 16577 + }, + { + "epoch": 1.4, + "learning_rate": 1.1603744724554747e-05, + "loss": 1.1629, + "step": 16578 + }, + { + "epoch": 1.4, + "learning_rate": 1.1602847510406657e-05, + "loss": 1.1879, + "step": 16579 + }, + { + "epoch": 1.4, + "learning_rate": 1.1601950283015324e-05, + "loss": 1.222, + "step": 16580 + }, + { + "epoch": 1.4, + "learning_rate": 1.1601053042388166e-05, + "loss": 1.1552, + "step": 16581 + }, + { + "epoch": 1.4, + "learning_rate": 1.1600155788532594e-05, + "loss": 1.3531, + "step": 16582 + }, + { + "epoch": 1.4, + "learning_rate": 1.1599258521456025e-05, + "loss": 1.136, + "step": 16583 + }, + { + "epoch": 1.4, + "learning_rate": 1.1598361241165868e-05, + "loss": 1.2025, + "step": 16584 + }, + { + "epoch": 1.4, + "learning_rate": 1.159746394766954e-05, + "loss": 1.2106, + "step": 16585 + }, + { + "epoch": 1.4, + "learning_rate": 1.1596566640974457e-05, + "loss": 1.168, + "step": 16586 + }, + { + "epoch": 1.4, + "learning_rate": 1.1595669321088025e-05, + "loss": 1.1919, + "step": 16587 + }, + { + "epoch": 1.4, + "learning_rate": 1.1594771988017663e-05, + "loss": 1.1787, + "step": 16588 + }, + { + "epoch": 1.4, + "learning_rate": 1.1593874641770784e-05, + "loss": 1.264, + "step": 16589 + }, + { + "epoch": 1.4, + "learning_rate": 1.1592977282354807e-05, + "loss": 1.2122, + "step": 16590 + }, + { + "epoch": 1.4, + "learning_rate": 1.159207990977714e-05, + "loss": 1.1542, + "step": 16591 + }, + { + "epoch": 1.4, + "learning_rate": 1.1591182524045194e-05, + "loss": 1.2863, + "step": 16592 + }, + { + "epoch": 1.4, + "learning_rate": 1.1590285125166396e-05, + "loss": 1.3463, + "step": 16593 + }, + { + "epoch": 1.4, + "learning_rate": 1.158938771314815e-05, + "loss": 1.2212, + "step": 16594 + }, + { + "epoch": 1.4, + "learning_rate": 1.1588490287997872e-05, + "loss": 1.1616, + "step": 16595 + }, + { + "epoch": 1.4, + "learning_rate": 1.1587592849722982e-05, + "loss": 1.188, + "step": 16596 + }, + { + "epoch": 1.4, + "learning_rate": 1.158669539833089e-05, + "loss": 1.1687, + "step": 16597 + }, + { + "epoch": 1.4, + "learning_rate": 1.1585797933829013e-05, + "loss": 1.1347, + "step": 16598 + }, + { + "epoch": 1.4, + "learning_rate": 1.1584900456224766e-05, + "loss": 1.3259, + "step": 16599 + }, + { + "epoch": 1.4, + "learning_rate": 1.1584002965525567e-05, + "loss": 1.2166, + "step": 16600 + }, + { + "epoch": 1.4, + "learning_rate": 1.1583105461738827e-05, + "loss": 1.2515, + "step": 16601 + }, + { + "epoch": 1.4, + "learning_rate": 1.158220794487196e-05, + "loss": 1.1753, + "step": 16602 + }, + { + "epoch": 1.4, + "learning_rate": 1.1581310414932387e-05, + "loss": 1.2585, + "step": 16603 + }, + { + "epoch": 1.4, + "learning_rate": 1.1580412871927523e-05, + "loss": 1.1225, + "step": 16604 + }, + { + "epoch": 1.4, + "learning_rate": 1.157951531586478e-05, + "loss": 1.1731, + "step": 16605 + }, + { + "epoch": 1.4, + "learning_rate": 1.1578617746751574e-05, + "loss": 1.1774, + "step": 16606 + }, + { + "epoch": 1.4, + "learning_rate": 1.1577720164595324e-05, + "loss": 1.2461, + "step": 16607 + }, + { + "epoch": 1.4, + "learning_rate": 1.1576822569403448e-05, + "loss": 1.2653, + "step": 16608 + }, + { + "epoch": 1.4, + "learning_rate": 1.1575924961183355e-05, + "loss": 1.1778, + "step": 16609 + }, + { + "epoch": 1.4, + "learning_rate": 1.1575027339942468e-05, + "loss": 1.1113, + "step": 16610 + }, + { + "epoch": 1.4, + "learning_rate": 1.1574129705688204e-05, + "loss": 1.1346, + "step": 16611 + }, + { + "epoch": 1.4, + "learning_rate": 1.1573232058427972e-05, + "loss": 1.1101, + "step": 16612 + }, + { + "epoch": 1.4, + "learning_rate": 1.1572334398169196e-05, + "loss": 1.1524, + "step": 16613 + }, + { + "epoch": 1.4, + "learning_rate": 1.1571436724919288e-05, + "loss": 1.0983, + "step": 16614 + }, + { + "epoch": 1.4, + "learning_rate": 1.1570539038685669e-05, + "loss": 1.2172, + "step": 16615 + }, + { + "epoch": 1.4, + "learning_rate": 1.1569641339475752e-05, + "loss": 1.1963, + "step": 16616 + }, + { + "epoch": 1.4, + "learning_rate": 1.1568743627296957e-05, + "loss": 1.2397, + "step": 16617 + }, + { + "epoch": 1.4, + "learning_rate": 1.15678459021567e-05, + "loss": 1.2207, + "step": 16618 + }, + { + "epoch": 1.4, + "learning_rate": 1.15669481640624e-05, + "loss": 1.1851, + "step": 16619 + }, + { + "epoch": 1.4, + "learning_rate": 1.1566050413021465e-05, + "loss": 1.1235, + "step": 16620 + }, + { + "epoch": 1.4, + "learning_rate": 1.156515264904133e-05, + "loss": 1.1863, + "step": 16621 + }, + { + "epoch": 1.4, + "learning_rate": 1.15642548721294e-05, + "loss": 1.136, + "step": 16622 + }, + { + "epoch": 1.4, + "learning_rate": 1.1563357082293092e-05, + "loss": 1.1553, + "step": 16623 + }, + { + "epoch": 1.4, + "learning_rate": 1.1562459279539829e-05, + "loss": 1.1187, + "step": 16624 + }, + { + "epoch": 1.4, + "learning_rate": 1.156156146387703e-05, + "loss": 1.2748, + "step": 16625 + }, + { + "epoch": 1.4, + "learning_rate": 1.1560663635312108e-05, + "loss": 1.2257, + "step": 16626 + }, + { + "epoch": 1.4, + "learning_rate": 1.1559765793852485e-05, + "loss": 1.162, + "step": 16627 + }, + { + "epoch": 1.4, + "learning_rate": 1.1558867939505576e-05, + "loss": 1.1808, + "step": 16628 + }, + { + "epoch": 1.4, + "learning_rate": 1.1557970072278803e-05, + "loss": 1.1663, + "step": 16629 + }, + { + "epoch": 1.4, + "learning_rate": 1.1557072192179583e-05, + "loss": 1.1584, + "step": 16630 + }, + { + "epoch": 1.4, + "learning_rate": 1.1556174299215332e-05, + "loss": 1.2616, + "step": 16631 + }, + { + "epoch": 1.4, + "learning_rate": 1.1555276393393473e-05, + "loss": 1.1382, + "step": 16632 + }, + { + "epoch": 1.4, + "learning_rate": 1.1554378474721422e-05, + "loss": 1.1903, + "step": 16633 + }, + { + "epoch": 1.4, + "learning_rate": 1.1553480543206597e-05, + "loss": 1.2209, + "step": 16634 + }, + { + "epoch": 1.4, + "learning_rate": 1.155258259885642e-05, + "loss": 1.1959, + "step": 16635 + }, + { + "epoch": 1.4, + "learning_rate": 1.1551684641678311e-05, + "loss": 1.227, + "step": 16636 + }, + { + "epoch": 1.4, + "learning_rate": 1.1550786671679683e-05, + "loss": 1.2123, + "step": 16637 + }, + { + "epoch": 1.4, + "learning_rate": 1.154988868886796e-05, + "loss": 1.2706, + "step": 16638 + }, + { + "epoch": 1.4, + "learning_rate": 1.1548990693250562e-05, + "loss": 1.2574, + "step": 16639 + }, + { + "epoch": 1.4, + "learning_rate": 1.1548092684834908e-05, + "loss": 1.156, + "step": 16640 + }, + { + "epoch": 1.4, + "learning_rate": 1.1547194663628414e-05, + "loss": 1.1311, + "step": 16641 + }, + { + "epoch": 1.4, + "learning_rate": 1.1546296629638505e-05, + "loss": 1.2262, + "step": 16642 + }, + { + "epoch": 1.4, + "learning_rate": 1.1545398582872598e-05, + "loss": 1.2227, + "step": 16643 + }, + { + "epoch": 1.4, + "learning_rate": 1.1544500523338111e-05, + "loss": 1.1424, + "step": 16644 + }, + { + "epoch": 1.4, + "learning_rate": 1.1543602451042466e-05, + "loss": 1.1995, + "step": 16645 + }, + { + "epoch": 1.4, + "learning_rate": 1.1542704365993087e-05, + "loss": 1.2912, + "step": 16646 + }, + { + "epoch": 1.4, + "learning_rate": 1.1541806268197389e-05, + "loss": 1.1049, + "step": 16647 + }, + { + "epoch": 1.4, + "learning_rate": 1.1540908157662793e-05, + "loss": 1.2206, + "step": 16648 + }, + { + "epoch": 1.4, + "learning_rate": 1.1540010034396723e-05, + "loss": 1.2054, + "step": 16649 + }, + { + "epoch": 1.4, + "learning_rate": 1.1539111898406598e-05, + "loss": 1.251, + "step": 16650 + }, + { + "epoch": 1.4, + "learning_rate": 1.1538213749699835e-05, + "loss": 1.2934, + "step": 16651 + }, + { + "epoch": 1.4, + "learning_rate": 1.153731558828386e-05, + "loss": 1.2042, + "step": 16652 + }, + { + "epoch": 1.4, + "learning_rate": 1.153641741416609e-05, + "loss": 1.1694, + "step": 16653 + }, + { + "epoch": 1.4, + "learning_rate": 1.1535519227353948e-05, + "loss": 1.1681, + "step": 16654 + }, + { + "epoch": 1.4, + "learning_rate": 1.1534621027854855e-05, + "loss": 1.2141, + "step": 16655 + }, + { + "epoch": 1.4, + "learning_rate": 1.1533722815676231e-05, + "loss": 1.1858, + "step": 16656 + }, + { + "epoch": 1.4, + "learning_rate": 1.15328245908255e-05, + "loss": 1.2216, + "step": 16657 + }, + { + "epoch": 1.4, + "learning_rate": 1.1531926353310078e-05, + "loss": 1.1388, + "step": 16658 + }, + { + "epoch": 1.4, + "learning_rate": 1.1531028103137393e-05, + "loss": 1.1708, + "step": 16659 + }, + { + "epoch": 1.4, + "learning_rate": 1.1530129840314864e-05, + "loss": 1.2485, + "step": 16660 + }, + { + "epoch": 1.4, + "learning_rate": 1.1529231564849914e-05, + "loss": 1.1686, + "step": 16661 + }, + { + "epoch": 1.4, + "learning_rate": 1.1528333276749962e-05, + "loss": 1.2218, + "step": 16662 + }, + { + "epoch": 1.4, + "learning_rate": 1.1527434976022428e-05, + "loss": 1.2274, + "step": 16663 + }, + { + "epoch": 1.4, + "learning_rate": 1.1526536662674741e-05, + "loss": 1.2216, + "step": 16664 + }, + { + "epoch": 1.4, + "learning_rate": 1.1525638336714322e-05, + "loss": 1.1609, + "step": 16665 + }, + { + "epoch": 1.4, + "learning_rate": 1.1524739998148587e-05, + "loss": 1.1634, + "step": 16666 + }, + { + "epoch": 1.4, + "learning_rate": 1.1523841646984961e-05, + "loss": 1.312, + "step": 16667 + }, + { + "epoch": 1.4, + "learning_rate": 1.1522943283230872e-05, + "loss": 1.223, + "step": 16668 + }, + { + "epoch": 1.4, + "learning_rate": 1.1522044906893734e-05, + "loss": 1.3154, + "step": 16669 + }, + { + "epoch": 1.4, + "learning_rate": 1.1521146517980974e-05, + "loss": 1.2014, + "step": 16670 + }, + { + "epoch": 1.4, + "learning_rate": 1.1520248116500019e-05, + "loss": 1.3285, + "step": 16671 + }, + { + "epoch": 1.4, + "learning_rate": 1.1519349702458283e-05, + "loss": 1.1848, + "step": 16672 + }, + { + "epoch": 1.4, + "learning_rate": 1.1518451275863195e-05, + "loss": 1.2425, + "step": 16673 + }, + { + "epoch": 1.4, + "learning_rate": 1.1517552836722176e-05, + "loss": 1.179, + "step": 16674 + }, + { + "epoch": 1.4, + "learning_rate": 1.1516654385042652e-05, + "loss": 1.1955, + "step": 16675 + }, + { + "epoch": 1.4, + "learning_rate": 1.1515755920832044e-05, + "loss": 1.1348, + "step": 16676 + }, + { + "epoch": 1.4, + "learning_rate": 1.1514857444097771e-05, + "loss": 1.191, + "step": 16677 + }, + { + "epoch": 1.4, + "learning_rate": 1.1513958954847264e-05, + "loss": 1.2233, + "step": 16678 + }, + { + "epoch": 1.4, + "learning_rate": 1.1513060453087946e-05, + "loss": 1.1948, + "step": 16679 + }, + { + "epoch": 1.4, + "learning_rate": 1.1512161938827236e-05, + "loss": 1.1689, + "step": 16680 + }, + { + "epoch": 1.4, + "learning_rate": 1.151126341207256e-05, + "loss": 1.2256, + "step": 16681 + }, + { + "epoch": 1.4, + "learning_rate": 1.151036487283134e-05, + "loss": 1.1632, + "step": 16682 + }, + { + "epoch": 1.4, + "learning_rate": 1.1509466321111004e-05, + "loss": 1.1933, + "step": 16683 + }, + { + "epoch": 1.4, + "learning_rate": 1.1508567756918975e-05, + "loss": 1.1554, + "step": 16684 + }, + { + "epoch": 1.4, + "learning_rate": 1.1507669180262675e-05, + "loss": 1.2147, + "step": 16685 + }, + { + "epoch": 1.4, + "learning_rate": 1.1506770591149531e-05, + "loss": 1.2171, + "step": 16686 + }, + { + "epoch": 1.4, + "learning_rate": 1.1505871989586966e-05, + "loss": 1.2377, + "step": 16687 + }, + { + "epoch": 1.4, + "learning_rate": 1.1504973375582405e-05, + "loss": 1.1792, + "step": 16688 + }, + { + "epoch": 1.41, + "learning_rate": 1.150407474914327e-05, + "loss": 1.1231, + "step": 16689 + }, + { + "epoch": 1.41, + "learning_rate": 1.1503176110276993e-05, + "loss": 1.1618, + "step": 16690 + }, + { + "epoch": 1.41, + "learning_rate": 1.150227745899099e-05, + "loss": 1.2794, + "step": 16691 + }, + { + "epoch": 1.41, + "learning_rate": 1.1501378795292688e-05, + "loss": 1.274, + "step": 16692 + }, + { + "epoch": 1.41, + "learning_rate": 1.150048011918952e-05, + "loss": 1.2046, + "step": 16693 + }, + { + "epoch": 1.41, + "learning_rate": 1.14995814306889e-05, + "loss": 1.1689, + "step": 16694 + }, + { + "epoch": 1.41, + "learning_rate": 1.1498682729798256e-05, + "loss": 1.1834, + "step": 16695 + }, + { + "epoch": 1.41, + "learning_rate": 1.1497784016525021e-05, + "loss": 1.2267, + "step": 16696 + }, + { + "epoch": 1.41, + "learning_rate": 1.1496885290876614e-05, + "loss": 1.1892, + "step": 16697 + }, + { + "epoch": 1.41, + "learning_rate": 1.1495986552860461e-05, + "loss": 1.1541, + "step": 16698 + }, + { + "epoch": 1.41, + "learning_rate": 1.1495087802483987e-05, + "loss": 1.1773, + "step": 16699 + }, + { + "epoch": 1.41, + "learning_rate": 1.1494189039754621e-05, + "loss": 1.2318, + "step": 16700 + }, + { + "epoch": 1.41, + "learning_rate": 1.1493290264679786e-05, + "loss": 1.0802, + "step": 16701 + }, + { + "epoch": 1.41, + "learning_rate": 1.1492391477266908e-05, + "loss": 1.1025, + "step": 16702 + }, + { + "epoch": 1.41, + "learning_rate": 1.1491492677523418e-05, + "loss": 1.2431, + "step": 16703 + }, + { + "epoch": 1.41, + "learning_rate": 1.1490593865456736e-05, + "loss": 1.2378, + "step": 16704 + }, + { + "epoch": 1.41, + "learning_rate": 1.1489695041074289e-05, + "loss": 1.1132, + "step": 16705 + }, + { + "epoch": 1.41, + "learning_rate": 1.1488796204383505e-05, + "loss": 1.2287, + "step": 16706 + }, + { + "epoch": 1.41, + "learning_rate": 1.148789735539181e-05, + "loss": 1.217, + "step": 16707 + }, + { + "epoch": 1.41, + "learning_rate": 1.1486998494106632e-05, + "loss": 1.135, + "step": 16708 + }, + { + "epoch": 1.41, + "learning_rate": 1.1486099620535396e-05, + "loss": 1.2307, + "step": 16709 + }, + { + "epoch": 1.41, + "learning_rate": 1.148520073468553e-05, + "loss": 1.1146, + "step": 16710 + }, + { + "epoch": 1.41, + "learning_rate": 1.1484301836564459e-05, + "loss": 1.0874, + "step": 16711 + }, + { + "epoch": 1.41, + "learning_rate": 1.1483402926179612e-05, + "loss": 1.1833, + "step": 16712 + }, + { + "epoch": 1.41, + "learning_rate": 1.1482504003538417e-05, + "loss": 1.2585, + "step": 16713 + }, + { + "epoch": 1.41, + "learning_rate": 1.1481605068648296e-05, + "loss": 1.0361, + "step": 16714 + }, + { + "epoch": 1.41, + "learning_rate": 1.1480706121516683e-05, + "loss": 1.2317, + "step": 16715 + }, + { + "epoch": 1.41, + "learning_rate": 1.1479807162151002e-05, + "loss": 1.1741, + "step": 16716 + }, + { + "epoch": 1.41, + "learning_rate": 1.1478908190558677e-05, + "loss": 1.1838, + "step": 16717 + }, + { + "epoch": 1.41, + "learning_rate": 1.1478009206747143e-05, + "loss": 1.1617, + "step": 16718 + }, + { + "epoch": 1.41, + "learning_rate": 1.1477110210723823e-05, + "loss": 1.2558, + "step": 16719 + }, + { + "epoch": 1.41, + "learning_rate": 1.1476211202496144e-05, + "loss": 1.1333, + "step": 16720 + }, + { + "epoch": 1.41, + "learning_rate": 1.1475312182071534e-05, + "loss": 1.1582, + "step": 16721 + }, + { + "epoch": 1.41, + "learning_rate": 1.1474413149457429e-05, + "loss": 1.1766, + "step": 16722 + }, + { + "epoch": 1.41, + "learning_rate": 1.1473514104661245e-05, + "loss": 1.1829, + "step": 16723 + }, + { + "epoch": 1.41, + "learning_rate": 1.1472615047690418e-05, + "loss": 1.1521, + "step": 16724 + }, + { + "epoch": 1.41, + "learning_rate": 1.1471715978552376e-05, + "loss": 1.2484, + "step": 16725 + }, + { + "epoch": 1.41, + "learning_rate": 1.1470816897254541e-05, + "loss": 1.1324, + "step": 16726 + }, + { + "epoch": 1.41, + "learning_rate": 1.146991780380435e-05, + "loss": 1.1933, + "step": 16727 + }, + { + "epoch": 1.41, + "learning_rate": 1.1469018698209227e-05, + "loss": 1.2998, + "step": 16728 + }, + { + "epoch": 1.41, + "learning_rate": 1.1468119580476602e-05, + "loss": 1.2062, + "step": 16729 + }, + { + "epoch": 1.41, + "learning_rate": 1.14672204506139e-05, + "loss": 1.1037, + "step": 16730 + }, + { + "epoch": 1.41, + "learning_rate": 1.1466321308628556e-05, + "loss": 1.2281, + "step": 16731 + }, + { + "epoch": 1.41, + "learning_rate": 1.1465422154527994e-05, + "loss": 1.1095, + "step": 16732 + }, + { + "epoch": 1.41, + "learning_rate": 1.1464522988319646e-05, + "loss": 1.1656, + "step": 16733 + }, + { + "epoch": 1.41, + "learning_rate": 1.1463623810010939e-05, + "loss": 1.1209, + "step": 16734 + }, + { + "epoch": 1.41, + "learning_rate": 1.1462724619609307e-05, + "loss": 1.1843, + "step": 16735 + }, + { + "epoch": 1.41, + "learning_rate": 1.1461825417122174e-05, + "loss": 1.2934, + "step": 16736 + }, + { + "epoch": 1.41, + "learning_rate": 1.1460926202556972e-05, + "loss": 1.1959, + "step": 16737 + }, + { + "epoch": 1.41, + "learning_rate": 1.1460026975921131e-05, + "loss": 1.1068, + "step": 16738 + }, + { + "epoch": 1.41, + "learning_rate": 1.1459127737222077e-05, + "loss": 1.245, + "step": 16739 + }, + { + "epoch": 1.41, + "learning_rate": 1.1458228486467247e-05, + "loss": 1.0735, + "step": 16740 + }, + { + "epoch": 1.41, + "learning_rate": 1.1457329223664065e-05, + "loss": 1.129, + "step": 16741 + }, + { + "epoch": 1.41, + "learning_rate": 1.145642994881996e-05, + "loss": 1.1668, + "step": 16742 + }, + { + "epoch": 1.41, + "learning_rate": 1.1455530661942365e-05, + "loss": 1.0964, + "step": 16743 + }, + { + "epoch": 1.41, + "learning_rate": 1.1454631363038713e-05, + "loss": 1.1838, + "step": 16744 + }, + { + "epoch": 1.41, + "learning_rate": 1.1453732052116427e-05, + "loss": 1.2287, + "step": 16745 + }, + { + "epoch": 1.41, + "learning_rate": 1.1452832729182943e-05, + "loss": 1.1151, + "step": 16746 + }, + { + "epoch": 1.41, + "learning_rate": 1.1451933394245693e-05, + "loss": 1.2708, + "step": 16747 + }, + { + "epoch": 1.41, + "learning_rate": 1.1451034047312102e-05, + "loss": 1.1962, + "step": 16748 + }, + { + "epoch": 1.41, + "learning_rate": 1.1450134688389602e-05, + "loss": 1.0812, + "step": 16749 + }, + { + "epoch": 1.41, + "learning_rate": 1.1449235317485629e-05, + "loss": 1.3135, + "step": 16750 + }, + { + "epoch": 1.41, + "learning_rate": 1.1448335934607608e-05, + "loss": 1.1733, + "step": 16751 + }, + { + "epoch": 1.41, + "learning_rate": 1.1447436539762972e-05, + "loss": 1.1229, + "step": 16752 + }, + { + "epoch": 1.41, + "learning_rate": 1.144653713295915e-05, + "loss": 1.1935, + "step": 16753 + }, + { + "epoch": 1.41, + "learning_rate": 1.1445637714203579e-05, + "loss": 1.2441, + "step": 16754 + }, + { + "epoch": 1.41, + "learning_rate": 1.1444738283503684e-05, + "loss": 1.226, + "step": 16755 + }, + { + "epoch": 1.41, + "learning_rate": 1.14438388408669e-05, + "loss": 1.1388, + "step": 16756 + }, + { + "epoch": 1.41, + "learning_rate": 1.1442939386300658e-05, + "loss": 1.1339, + "step": 16757 + }, + { + "epoch": 1.41, + "learning_rate": 1.1442039919812386e-05, + "loss": 1.1961, + "step": 16758 + }, + { + "epoch": 1.41, + "learning_rate": 1.144114044140952e-05, + "loss": 1.243, + "step": 16759 + }, + { + "epoch": 1.41, + "learning_rate": 1.144024095109949e-05, + "loss": 1.1732, + "step": 16760 + }, + { + "epoch": 1.41, + "learning_rate": 1.1439341448889731e-05, + "loss": 1.1913, + "step": 16761 + }, + { + "epoch": 1.41, + "learning_rate": 1.1438441934787672e-05, + "loss": 1.2327, + "step": 16762 + }, + { + "epoch": 1.41, + "learning_rate": 1.1437542408800743e-05, + "loss": 1.2641, + "step": 16763 + }, + { + "epoch": 1.41, + "learning_rate": 1.1436642870936381e-05, + "loss": 1.1891, + "step": 16764 + }, + { + "epoch": 1.41, + "learning_rate": 1.1435743321202016e-05, + "loss": 1.168, + "step": 16765 + }, + { + "epoch": 1.41, + "learning_rate": 1.1434843759605077e-05, + "loss": 1.1233, + "step": 16766 + }, + { + "epoch": 1.41, + "learning_rate": 1.1433944186153001e-05, + "loss": 1.2784, + "step": 16767 + }, + { + "epoch": 1.41, + "learning_rate": 1.143304460085322e-05, + "loss": 1.2336, + "step": 16768 + }, + { + "epoch": 1.41, + "learning_rate": 1.1432145003713167e-05, + "loss": 1.2114, + "step": 16769 + }, + { + "epoch": 1.41, + "learning_rate": 1.1431245394740269e-05, + "loss": 1.2588, + "step": 16770 + }, + { + "epoch": 1.41, + "learning_rate": 1.1430345773941967e-05, + "loss": 1.1175, + "step": 16771 + }, + { + "epoch": 1.41, + "learning_rate": 1.142944614132569e-05, + "loss": 1.1492, + "step": 16772 + }, + { + "epoch": 1.41, + "learning_rate": 1.1428546496898871e-05, + "loss": 1.2251, + "step": 16773 + }, + { + "epoch": 1.41, + "learning_rate": 1.1427646840668944e-05, + "loss": 1.2525, + "step": 16774 + }, + { + "epoch": 1.41, + "learning_rate": 1.1426747172643345e-05, + "loss": 1.2548, + "step": 16775 + }, + { + "epoch": 1.41, + "learning_rate": 1.14258474928295e-05, + "loss": 1.1783, + "step": 16776 + }, + { + "epoch": 1.41, + "learning_rate": 1.1424947801234846e-05, + "loss": 1.0913, + "step": 16777 + }, + { + "epoch": 1.41, + "learning_rate": 1.142404809786682e-05, + "loss": 1.2361, + "step": 16778 + }, + { + "epoch": 1.41, + "learning_rate": 1.1423148382732854e-05, + "loss": 1.1529, + "step": 16779 + }, + { + "epoch": 1.41, + "learning_rate": 1.1422248655840377e-05, + "loss": 1.2185, + "step": 16780 + }, + { + "epoch": 1.41, + "learning_rate": 1.1421348917196828e-05, + "loss": 1.1173, + "step": 16781 + }, + { + "epoch": 1.41, + "learning_rate": 1.142044916680964e-05, + "loss": 1.1242, + "step": 16782 + }, + { + "epoch": 1.41, + "learning_rate": 1.1419549404686246e-05, + "loss": 1.1536, + "step": 16783 + }, + { + "epoch": 1.41, + "learning_rate": 1.141864963083408e-05, + "loss": 1.1967, + "step": 16784 + }, + { + "epoch": 1.41, + "learning_rate": 1.1417749845260575e-05, + "loss": 1.1553, + "step": 16785 + }, + { + "epoch": 1.41, + "learning_rate": 1.1416850047973172e-05, + "loss": 1.2602, + "step": 16786 + }, + { + "epoch": 1.41, + "learning_rate": 1.1415950238979296e-05, + "loss": 1.1427, + "step": 16787 + }, + { + "epoch": 1.41, + "learning_rate": 1.141505041828639e-05, + "loss": 1.2242, + "step": 16788 + }, + { + "epoch": 1.41, + "learning_rate": 1.141415058590188e-05, + "loss": 1.1985, + "step": 16789 + }, + { + "epoch": 1.41, + "learning_rate": 1.1413250741833208e-05, + "loss": 1.2199, + "step": 16790 + }, + { + "epoch": 1.41, + "learning_rate": 1.1412350886087806e-05, + "loss": 1.3039, + "step": 16791 + }, + { + "epoch": 1.41, + "learning_rate": 1.1411451018673107e-05, + "loss": 1.1641, + "step": 16792 + }, + { + "epoch": 1.41, + "learning_rate": 1.1410551139596549e-05, + "loss": 1.088, + "step": 16793 + }, + { + "epoch": 1.41, + "learning_rate": 1.1409651248865567e-05, + "loss": 1.214, + "step": 16794 + }, + { + "epoch": 1.41, + "learning_rate": 1.1408751346487595e-05, + "loss": 1.2376, + "step": 16795 + }, + { + "epoch": 1.41, + "learning_rate": 1.1407851432470065e-05, + "loss": 1.2194, + "step": 16796 + }, + { + "epoch": 1.41, + "learning_rate": 1.1406951506820423e-05, + "loss": 1.3, + "step": 16797 + }, + { + "epoch": 1.41, + "learning_rate": 1.1406051569546092e-05, + "loss": 1.1561, + "step": 16798 + }, + { + "epoch": 1.41, + "learning_rate": 1.1405151620654513e-05, + "loss": 1.1841, + "step": 16799 + }, + { + "epoch": 1.41, + "learning_rate": 1.1404251660153125e-05, + "loss": 1.2421, + "step": 16800 + }, + { + "epoch": 1.41, + "learning_rate": 1.1403351688049358e-05, + "loss": 1.2142, + "step": 16801 + }, + { + "epoch": 1.41, + "learning_rate": 1.140245170435065e-05, + "loss": 1.2425, + "step": 16802 + }, + { + "epoch": 1.41, + "learning_rate": 1.1401551709064438e-05, + "loss": 1.1645, + "step": 16803 + }, + { + "epoch": 1.41, + "learning_rate": 1.1400651702198159e-05, + "loss": 1.2147, + "step": 16804 + }, + { + "epoch": 1.41, + "learning_rate": 1.1399751683759244e-05, + "loss": 1.1498, + "step": 16805 + }, + { + "epoch": 1.41, + "learning_rate": 1.1398851653755134e-05, + "loss": 1.2772, + "step": 16806 + }, + { + "epoch": 1.42, + "learning_rate": 1.1397951612193263e-05, + "loss": 1.2266, + "step": 16807 + }, + { + "epoch": 1.42, + "learning_rate": 1.1397051559081069e-05, + "loss": 1.1159, + "step": 16808 + }, + { + "epoch": 1.42, + "learning_rate": 1.1396151494425988e-05, + "loss": 1.2214, + "step": 16809 + }, + { + "epoch": 1.42, + "learning_rate": 1.1395251418235455e-05, + "loss": 1.1641, + "step": 16810 + }, + { + "epoch": 1.42, + "learning_rate": 1.1394351330516913e-05, + "loss": 1.1499, + "step": 16811 + }, + { + "epoch": 1.42, + "learning_rate": 1.139345123127779e-05, + "loss": 1.2635, + "step": 16812 + }, + { + "epoch": 1.42, + "learning_rate": 1.1392551120525529e-05, + "loss": 1.1573, + "step": 16813 + }, + { + "epoch": 1.42, + "learning_rate": 1.1391650998267563e-05, + "loss": 1.1416, + "step": 16814 + }, + { + "epoch": 1.42, + "learning_rate": 1.1390750864511333e-05, + "loss": 1.1821, + "step": 16815 + }, + { + "epoch": 1.42, + "learning_rate": 1.1389850719264272e-05, + "loss": 1.2721, + "step": 16816 + }, + { + "epoch": 1.42, + "learning_rate": 1.138895056253382e-05, + "loss": 1.1438, + "step": 16817 + }, + { + "epoch": 1.42, + "learning_rate": 1.1388050394327417e-05, + "loss": 1.1841, + "step": 16818 + }, + { + "epoch": 1.42, + "learning_rate": 1.1387150214652496e-05, + "loss": 1.267, + "step": 16819 + }, + { + "epoch": 1.42, + "learning_rate": 1.1386250023516492e-05, + "loss": 1.2175, + "step": 16820 + }, + { + "epoch": 1.42, + "learning_rate": 1.138534982092685e-05, + "loss": 1.2032, + "step": 16821 + }, + { + "epoch": 1.42, + "learning_rate": 1.1384449606891005e-05, + "loss": 1.2382, + "step": 16822 + }, + { + "epoch": 1.42, + "learning_rate": 1.1383549381416393e-05, + "loss": 1.1638, + "step": 16823 + }, + { + "epoch": 1.42, + "learning_rate": 1.1382649144510454e-05, + "loss": 1.1952, + "step": 16824 + }, + { + "epoch": 1.42, + "learning_rate": 1.1381748896180627e-05, + "loss": 1.136, + "step": 16825 + }, + { + "epoch": 1.42, + "learning_rate": 1.1380848636434345e-05, + "loss": 1.2339, + "step": 16826 + }, + { + "epoch": 1.42, + "learning_rate": 1.137994836527905e-05, + "loss": 1.2142, + "step": 16827 + }, + { + "epoch": 1.42, + "learning_rate": 1.1379048082722182e-05, + "loss": 1.2065, + "step": 16828 + }, + { + "epoch": 1.42, + "learning_rate": 1.1378147788771177e-05, + "loss": 1.1612, + "step": 16829 + }, + { + "epoch": 1.42, + "learning_rate": 1.1377247483433474e-05, + "loss": 1.1155, + "step": 16830 + }, + { + "epoch": 1.42, + "learning_rate": 1.1376347166716508e-05, + "loss": 1.1365, + "step": 16831 + }, + { + "epoch": 1.42, + "learning_rate": 1.1375446838627725e-05, + "loss": 1.3143, + "step": 16832 + }, + { + "epoch": 1.42, + "learning_rate": 1.137454649917456e-05, + "loss": 1.2031, + "step": 16833 + }, + { + "epoch": 1.42, + "learning_rate": 1.1373646148364448e-05, + "loss": 1.1461, + "step": 16834 + }, + { + "epoch": 1.42, + "learning_rate": 1.1372745786204835e-05, + "loss": 1.1748, + "step": 16835 + }, + { + "epoch": 1.42, + "learning_rate": 1.1371845412703158e-05, + "loss": 1.2484, + "step": 16836 + }, + { + "epoch": 1.42, + "learning_rate": 1.1370945027866853e-05, + "loss": 1.2264, + "step": 16837 + }, + { + "epoch": 1.42, + "learning_rate": 1.1370044631703364e-05, + "loss": 1.2396, + "step": 16838 + }, + { + "epoch": 1.42, + "learning_rate": 1.1369144224220126e-05, + "loss": 1.1199, + "step": 16839 + }, + { + "epoch": 1.42, + "learning_rate": 1.1368243805424581e-05, + "loss": 1.0258, + "step": 16840 + }, + { + "epoch": 1.42, + "learning_rate": 1.1367343375324166e-05, + "loss": 1.1433, + "step": 16841 + }, + { + "epoch": 1.42, + "learning_rate": 1.1366442933926322e-05, + "loss": 1.241, + "step": 16842 + }, + { + "epoch": 1.42, + "learning_rate": 1.1365542481238492e-05, + "loss": 1.1981, + "step": 16843 + }, + { + "epoch": 1.42, + "learning_rate": 1.1364642017268113e-05, + "loss": 1.1738, + "step": 16844 + }, + { + "epoch": 1.42, + "learning_rate": 1.1363741542022622e-05, + "loss": 1.1795, + "step": 16845 + }, + { + "epoch": 1.42, + "learning_rate": 1.1362841055509462e-05, + "loss": 1.182, + "step": 16846 + }, + { + "epoch": 1.42, + "learning_rate": 1.1361940557736078e-05, + "loss": 1.1397, + "step": 16847 + }, + { + "epoch": 1.42, + "learning_rate": 1.13610400487099e-05, + "loss": 1.2155, + "step": 16848 + }, + { + "epoch": 1.42, + "learning_rate": 1.1360139528438374e-05, + "loss": 1.152, + "step": 16849 + }, + { + "epoch": 1.42, + "learning_rate": 1.1359238996928944e-05, + "loss": 1.1462, + "step": 16850 + }, + { + "epoch": 1.42, + "learning_rate": 1.1358338454189044e-05, + "loss": 1.2461, + "step": 16851 + }, + { + "epoch": 1.42, + "learning_rate": 1.1357437900226115e-05, + "loss": 1.1831, + "step": 16852 + }, + { + "epoch": 1.42, + "learning_rate": 1.13565373350476e-05, + "loss": 1.118, + "step": 16853 + }, + { + "epoch": 1.42, + "learning_rate": 1.1355636758660945e-05, + "loss": 1.1196, + "step": 16854 + }, + { + "epoch": 1.42, + "learning_rate": 1.1354736171073579e-05, + "loss": 1.1412, + "step": 16855 + }, + { + "epoch": 1.42, + "learning_rate": 1.1353835572292951e-05, + "loss": 1.172, + "step": 16856 + }, + { + "epoch": 1.42, + "learning_rate": 1.13529349623265e-05, + "loss": 1.178, + "step": 16857 + }, + { + "epoch": 1.42, + "learning_rate": 1.1352034341181668e-05, + "loss": 1.201, + "step": 16858 + }, + { + "epoch": 1.42, + "learning_rate": 1.1351133708865894e-05, + "loss": 1.0107, + "step": 16859 + }, + { + "epoch": 1.42, + "learning_rate": 1.1350233065386622e-05, + "loss": 1.1626, + "step": 16860 + }, + { + "epoch": 1.42, + "learning_rate": 1.1349332410751295e-05, + "loss": 1.2564, + "step": 16861 + }, + { + "epoch": 1.42, + "learning_rate": 1.1348431744967347e-05, + "loss": 1.16, + "step": 16862 + }, + { + "epoch": 1.42, + "learning_rate": 1.1347531068042225e-05, + "loss": 1.2645, + "step": 16863 + }, + { + "epoch": 1.42, + "learning_rate": 1.1346630379983372e-05, + "loss": 1.1978, + "step": 16864 + }, + { + "epoch": 1.42, + "learning_rate": 1.134572968079823e-05, + "loss": 1.2713, + "step": 16865 + }, + { + "epoch": 1.42, + "learning_rate": 1.1344828970494235e-05, + "loss": 1.1507, + "step": 16866 + }, + { + "epoch": 1.42, + "learning_rate": 1.1343928249078835e-05, + "loss": 1.1202, + "step": 16867 + }, + { + "epoch": 1.42, + "learning_rate": 1.1343027516559467e-05, + "loss": 1.2363, + "step": 16868 + }, + { + "epoch": 1.42, + "learning_rate": 1.1342126772943577e-05, + "loss": 1.2139, + "step": 16869 + }, + { + "epoch": 1.42, + "learning_rate": 1.1341226018238605e-05, + "loss": 1.1809, + "step": 16870 + }, + { + "epoch": 1.42, + "learning_rate": 1.1340325252451996e-05, + "loss": 1.2173, + "step": 16871 + }, + { + "epoch": 1.42, + "learning_rate": 1.1339424475591193e-05, + "loss": 1.231, + "step": 16872 + }, + { + "epoch": 1.42, + "learning_rate": 1.1338523687663632e-05, + "loss": 1.2692, + "step": 16873 + }, + { + "epoch": 1.42, + "learning_rate": 1.1337622888676763e-05, + "loss": 1.1805, + "step": 16874 + }, + { + "epoch": 1.42, + "learning_rate": 1.1336722078638023e-05, + "loss": 1.1655, + "step": 16875 + }, + { + "epoch": 1.42, + "learning_rate": 1.133582125755486e-05, + "loss": 1.3191, + "step": 16876 + }, + { + "epoch": 1.42, + "learning_rate": 1.1334920425434712e-05, + "loss": 1.218, + "step": 16877 + }, + { + "epoch": 1.42, + "learning_rate": 1.1334019582285024e-05, + "loss": 1.209, + "step": 16878 + }, + { + "epoch": 1.42, + "learning_rate": 1.1333118728113244e-05, + "loss": 1.1465, + "step": 16879 + }, + { + "epoch": 1.42, + "learning_rate": 1.1332217862926807e-05, + "loss": 1.2294, + "step": 16880 + }, + { + "epoch": 1.42, + "learning_rate": 1.1331316986733158e-05, + "loss": 1.2932, + "step": 16881 + }, + { + "epoch": 1.42, + "learning_rate": 1.1330416099539742e-05, + "loss": 1.1356, + "step": 16882 + }, + { + "epoch": 1.42, + "learning_rate": 1.1329515201354004e-05, + "loss": 1.2795, + "step": 16883 + }, + { + "epoch": 1.42, + "learning_rate": 1.1328614292183387e-05, + "loss": 1.1121, + "step": 16884 + }, + { + "epoch": 1.42, + "learning_rate": 1.1327713372035331e-05, + "loss": 1.1854, + "step": 16885 + }, + { + "epoch": 1.42, + "learning_rate": 1.1326812440917286e-05, + "loss": 1.1254, + "step": 16886 + }, + { + "epoch": 1.42, + "learning_rate": 1.1325911498836688e-05, + "loss": 1.1381, + "step": 16887 + }, + { + "epoch": 1.42, + "learning_rate": 1.1325010545800987e-05, + "loss": 1.1535, + "step": 16888 + }, + { + "epoch": 1.42, + "learning_rate": 1.1324109581817624e-05, + "loss": 1.1945, + "step": 16889 + }, + { + "epoch": 1.42, + "learning_rate": 1.1323208606894045e-05, + "loss": 1.2762, + "step": 16890 + }, + { + "epoch": 1.42, + "learning_rate": 1.1322307621037692e-05, + "loss": 1.2862, + "step": 16891 + }, + { + "epoch": 1.42, + "learning_rate": 1.132140662425601e-05, + "loss": 1.1486, + "step": 16892 + }, + { + "epoch": 1.42, + "learning_rate": 1.1320505616556443e-05, + "loss": 1.1248, + "step": 16893 + }, + { + "epoch": 1.42, + "learning_rate": 1.1319604597946438e-05, + "loss": 1.1709, + "step": 16894 + }, + { + "epoch": 1.42, + "learning_rate": 1.1318703568433434e-05, + "loss": 1.2279, + "step": 16895 + }, + { + "epoch": 1.42, + "learning_rate": 1.1317802528024882e-05, + "loss": 1.2071, + "step": 16896 + }, + { + "epoch": 1.42, + "learning_rate": 1.1316901476728223e-05, + "loss": 1.1787, + "step": 16897 + }, + { + "epoch": 1.42, + "learning_rate": 1.1316000414550902e-05, + "loss": 1.2524, + "step": 16898 + }, + { + "epoch": 1.42, + "learning_rate": 1.1315099341500366e-05, + "loss": 1.2568, + "step": 16899 + }, + { + "epoch": 1.42, + "learning_rate": 1.1314198257584056e-05, + "loss": 1.1121, + "step": 16900 + }, + { + "epoch": 1.42, + "learning_rate": 1.1313297162809425e-05, + "loss": 1.2008, + "step": 16901 + }, + { + "epoch": 1.42, + "learning_rate": 1.1312396057183907e-05, + "loss": 1.1789, + "step": 16902 + }, + { + "epoch": 1.42, + "learning_rate": 1.1311494940714952e-05, + "loss": 1.2379, + "step": 16903 + }, + { + "epoch": 1.42, + "learning_rate": 1.131059381341001e-05, + "loss": 1.1683, + "step": 16904 + }, + { + "epoch": 1.42, + "learning_rate": 1.1309692675276522e-05, + "loss": 1.1588, + "step": 16905 + }, + { + "epoch": 1.42, + "learning_rate": 1.1308791526321931e-05, + "loss": 1.1974, + "step": 16906 + }, + { + "epoch": 1.42, + "learning_rate": 1.1307890366553686e-05, + "loss": 1.1762, + "step": 16907 + }, + { + "epoch": 1.42, + "learning_rate": 1.1306989195979235e-05, + "loss": 1.2537, + "step": 16908 + }, + { + "epoch": 1.42, + "learning_rate": 1.1306088014606018e-05, + "loss": 1.1852, + "step": 16909 + }, + { + "epoch": 1.42, + "learning_rate": 1.1305186822441485e-05, + "loss": 1.2492, + "step": 16910 + }, + { + "epoch": 1.42, + "learning_rate": 1.1304285619493082e-05, + "loss": 1.15, + "step": 16911 + }, + { + "epoch": 1.42, + "learning_rate": 1.1303384405768254e-05, + "loss": 1.2329, + "step": 16912 + }, + { + "epoch": 1.42, + "learning_rate": 1.1302483181274443e-05, + "loss": 1.1649, + "step": 16913 + }, + { + "epoch": 1.42, + "learning_rate": 1.1301581946019103e-05, + "loss": 1.1351, + "step": 16914 + }, + { + "epoch": 1.42, + "learning_rate": 1.1300680700009676e-05, + "loss": 1.2371, + "step": 16915 + }, + { + "epoch": 1.42, + "learning_rate": 1.129977944325361e-05, + "loss": 1.195, + "step": 16916 + }, + { + "epoch": 1.42, + "learning_rate": 1.1298878175758345e-05, + "loss": 1.2058, + "step": 16917 + }, + { + "epoch": 1.42, + "learning_rate": 1.1297976897531337e-05, + "loss": 1.172, + "step": 16918 + }, + { + "epoch": 1.42, + "learning_rate": 1.1297075608580029e-05, + "loss": 1.2109, + "step": 16919 + }, + { + "epoch": 1.42, + "learning_rate": 1.1296174308911863e-05, + "loss": 1.1701, + "step": 16920 + }, + { + "epoch": 1.42, + "learning_rate": 1.1295272998534292e-05, + "loss": 1.1855, + "step": 16921 + }, + { + "epoch": 1.42, + "learning_rate": 1.1294371677454763e-05, + "loss": 1.2155, + "step": 16922 + }, + { + "epoch": 1.42, + "learning_rate": 1.129347034568072e-05, + "loss": 1.1611, + "step": 16923 + }, + { + "epoch": 1.42, + "learning_rate": 1.1292569003219609e-05, + "loss": 1.2285, + "step": 16924 + }, + { + "epoch": 1.42, + "learning_rate": 1.1291667650078882e-05, + "loss": 1.1176, + "step": 16925 + }, + { + "epoch": 1.43, + "learning_rate": 1.1290766286265985e-05, + "loss": 1.2516, + "step": 16926 + }, + { + "epoch": 1.43, + "learning_rate": 1.128986491178836e-05, + "loss": 1.2394, + "step": 16927 + }, + { + "epoch": 1.43, + "learning_rate": 1.1288963526653459e-05, + "loss": 1.1736, + "step": 16928 + }, + { + "epoch": 1.43, + "learning_rate": 1.128806213086873e-05, + "loss": 1.1205, + "step": 16929 + }, + { + "epoch": 1.43, + "learning_rate": 1.128716072444162e-05, + "loss": 1.1852, + "step": 16930 + }, + { + "epoch": 1.43, + "learning_rate": 1.1286259307379573e-05, + "loss": 1.1244, + "step": 16931 + }, + { + "epoch": 1.43, + "learning_rate": 1.1285357879690043e-05, + "loss": 1.1734, + "step": 16932 + }, + { + "epoch": 1.43, + "learning_rate": 1.1284456441380471e-05, + "loss": 1.1695, + "step": 16933 + }, + { + "epoch": 1.43, + "learning_rate": 1.1283554992458312e-05, + "loss": 1.231, + "step": 16934 + }, + { + "epoch": 1.43, + "learning_rate": 1.1282653532931009e-05, + "loss": 1.2257, + "step": 16935 + }, + { + "epoch": 1.43, + "learning_rate": 1.1281752062806015e-05, + "loss": 1.2021, + "step": 16936 + }, + { + "epoch": 1.43, + "learning_rate": 1.1280850582090774e-05, + "loss": 1.1461, + "step": 16937 + }, + { + "epoch": 1.43, + "learning_rate": 1.1279949090792734e-05, + "loss": 1.2535, + "step": 16938 + }, + { + "epoch": 1.43, + "learning_rate": 1.1279047588919347e-05, + "loss": 1.2275, + "step": 16939 + }, + { + "epoch": 1.43, + "learning_rate": 1.127814607647806e-05, + "loss": 1.1096, + "step": 16940 + }, + { + "epoch": 1.43, + "learning_rate": 1.1277244553476318e-05, + "loss": 1.1358, + "step": 16941 + }, + { + "epoch": 1.43, + "learning_rate": 1.1276343019921575e-05, + "loss": 1.1665, + "step": 16942 + }, + { + "epoch": 1.43, + "learning_rate": 1.1275441475821276e-05, + "loss": 1.0992, + "step": 16943 + }, + { + "epoch": 1.43, + "learning_rate": 1.1274539921182874e-05, + "loss": 1.172, + "step": 16944 + }, + { + "epoch": 1.43, + "learning_rate": 1.1273638356013811e-05, + "loss": 1.1996, + "step": 16945 + }, + { + "epoch": 1.43, + "learning_rate": 1.1272736780321542e-05, + "loss": 1.2652, + "step": 16946 + }, + { + "epoch": 1.43, + "learning_rate": 1.1271835194113519e-05, + "loss": 1.1436, + "step": 16947 + }, + { + "epoch": 1.43, + "learning_rate": 1.127093359739718e-05, + "loss": 1.2464, + "step": 16948 + }, + { + "epoch": 1.43, + "learning_rate": 1.1270031990179984e-05, + "loss": 1.185, + "step": 16949 + }, + { + "epoch": 1.43, + "learning_rate": 1.1269130372469376e-05, + "loss": 1.2197, + "step": 16950 + }, + { + "epoch": 1.43, + "learning_rate": 1.1268228744272809e-05, + "loss": 1.2258, + "step": 16951 + }, + { + "epoch": 1.43, + "learning_rate": 1.126732710559773e-05, + "loss": 1.2021, + "step": 16952 + }, + { + "epoch": 1.43, + "learning_rate": 1.1266425456451586e-05, + "loss": 1.1876, + "step": 16953 + }, + { + "epoch": 1.43, + "learning_rate": 1.1265523796841832e-05, + "loss": 1.1481, + "step": 16954 + }, + { + "epoch": 1.43, + "learning_rate": 1.1264622126775915e-05, + "loss": 1.108, + "step": 16955 + }, + { + "epoch": 1.43, + "learning_rate": 1.1263720446261284e-05, + "loss": 1.0975, + "step": 16956 + }, + { + "epoch": 1.43, + "learning_rate": 1.1262818755305392e-05, + "loss": 1.1761, + "step": 16957 + }, + { + "epoch": 1.43, + "learning_rate": 1.1261917053915685e-05, + "loss": 1.1423, + "step": 16958 + }, + { + "epoch": 1.43, + "learning_rate": 1.1261015342099616e-05, + "loss": 1.1582, + "step": 16959 + }, + { + "epoch": 1.43, + "learning_rate": 1.1260113619864636e-05, + "loss": 1.2291, + "step": 16960 + }, + { + "epoch": 1.43, + "learning_rate": 1.1259211887218195e-05, + "loss": 1.1861, + "step": 16961 + }, + { + "epoch": 1.43, + "learning_rate": 1.125831014416774e-05, + "loss": 1.2102, + "step": 16962 + }, + { + "epoch": 1.43, + "learning_rate": 1.1257408390720724e-05, + "loss": 1.1732, + "step": 16963 + }, + { + "epoch": 1.43, + "learning_rate": 1.1256506626884598e-05, + "loss": 1.1993, + "step": 16964 + }, + { + "epoch": 1.43, + "learning_rate": 1.1255604852666815e-05, + "loss": 1.0929, + "step": 16965 + }, + { + "epoch": 1.43, + "learning_rate": 1.1254703068074819e-05, + "loss": 1.1581, + "step": 16966 + }, + { + "epoch": 1.43, + "learning_rate": 1.1253801273116065e-05, + "loss": 1.1744, + "step": 16967 + }, + { + "epoch": 1.43, + "learning_rate": 1.1252899467798003e-05, + "loss": 1.1365, + "step": 16968 + }, + { + "epoch": 1.43, + "learning_rate": 1.1251997652128088e-05, + "loss": 1.0758, + "step": 16969 + }, + { + "epoch": 1.43, + "learning_rate": 1.1251095826113765e-05, + "loss": 1.1763, + "step": 16970 + }, + { + "epoch": 1.43, + "learning_rate": 1.1250193989762488e-05, + "loss": 1.2213, + "step": 16971 + }, + { + "epoch": 1.43, + "learning_rate": 1.1249292143081709e-05, + "loss": 1.2588, + "step": 16972 + }, + { + "epoch": 1.43, + "learning_rate": 1.1248390286078878e-05, + "loss": 1.1808, + "step": 16973 + }, + { + "epoch": 1.43, + "learning_rate": 1.1247488418761448e-05, + "loss": 1.1501, + "step": 16974 + }, + { + "epoch": 1.43, + "learning_rate": 1.1246586541136868e-05, + "loss": 1.1893, + "step": 16975 + }, + { + "epoch": 1.43, + "learning_rate": 1.1245684653212594e-05, + "loss": 1.1666, + "step": 16976 + }, + { + "epoch": 1.43, + "learning_rate": 1.124478275499607e-05, + "loss": 1.202, + "step": 16977 + }, + { + "epoch": 1.43, + "learning_rate": 1.1243880846494754e-05, + "loss": 1.1581, + "step": 16978 + }, + { + "epoch": 1.43, + "learning_rate": 1.1242978927716102e-05, + "loss": 1.0994, + "step": 16979 + }, + { + "epoch": 1.43, + "learning_rate": 1.1242076998667555e-05, + "loss": 1.136, + "step": 16980 + }, + { + "epoch": 1.43, + "learning_rate": 1.1241175059356569e-05, + "loss": 1.2125, + "step": 16981 + }, + { + "epoch": 1.43, + "learning_rate": 1.1240273109790599e-05, + "loss": 1.1313, + "step": 16982 + }, + { + "epoch": 1.43, + "learning_rate": 1.1239371149977097e-05, + "loss": 1.189, + "step": 16983 + }, + { + "epoch": 1.43, + "learning_rate": 1.1238469179923511e-05, + "loss": 1.2219, + "step": 16984 + }, + { + "epoch": 1.43, + "learning_rate": 1.1237567199637297e-05, + "loss": 1.1761, + "step": 16985 + }, + { + "epoch": 1.43, + "learning_rate": 1.123666520912591e-05, + "loss": 1.2651, + "step": 16986 + }, + { + "epoch": 1.43, + "learning_rate": 1.1235763208396797e-05, + "loss": 1.084, + "step": 16987 + }, + { + "epoch": 1.43, + "learning_rate": 1.1234861197457412e-05, + "loss": 1.1785, + "step": 16988 + }, + { + "epoch": 1.43, + "learning_rate": 1.123395917631521e-05, + "loss": 1.242, + "step": 16989 + }, + { + "epoch": 1.43, + "learning_rate": 1.1233057144977644e-05, + "loss": 1.1862, + "step": 16990 + }, + { + "epoch": 1.43, + "learning_rate": 1.1232155103452163e-05, + "loss": 1.2148, + "step": 16991 + }, + { + "epoch": 1.43, + "learning_rate": 1.123125305174622e-05, + "loss": 1.1776, + "step": 16992 + }, + { + "epoch": 1.43, + "learning_rate": 1.1230350989867273e-05, + "loss": 1.2426, + "step": 16993 + }, + { + "epoch": 1.43, + "learning_rate": 1.1229448917822775e-05, + "loss": 1.179, + "step": 16994 + }, + { + "epoch": 1.43, + "learning_rate": 1.122854683562017e-05, + "loss": 1.1103, + "step": 16995 + }, + { + "epoch": 1.43, + "learning_rate": 1.1227644743266921e-05, + "loss": 1.2553, + "step": 16996 + }, + { + "epoch": 1.43, + "learning_rate": 1.122674264077048e-05, + "loss": 1.2161, + "step": 16997 + }, + { + "epoch": 1.43, + "learning_rate": 1.1225840528138297e-05, + "loss": 1.1284, + "step": 16998 + }, + { + "epoch": 1.43, + "learning_rate": 1.1224938405377828e-05, + "loss": 1.1705, + "step": 16999 + }, + { + "epoch": 1.43, + "learning_rate": 1.1224036272496523e-05, + "loss": 1.1329, + "step": 17000 + }, + { + "epoch": 1.43, + "learning_rate": 1.1223134129501845e-05, + "loss": 1.2219, + "step": 17001 + }, + { + "epoch": 1.43, + "learning_rate": 1.1222231976401235e-05, + "loss": 1.1471, + "step": 17002 + }, + { + "epoch": 1.43, + "learning_rate": 1.1221329813202155e-05, + "loss": 1.1529, + "step": 17003 + }, + { + "epoch": 1.43, + "learning_rate": 1.1220427639912059e-05, + "loss": 1.2191, + "step": 17004 + }, + { + "epoch": 1.43, + "learning_rate": 1.1219525456538399e-05, + "loss": 1.1931, + "step": 17005 + }, + { + "epoch": 1.43, + "learning_rate": 1.1218623263088627e-05, + "loss": 1.1096, + "step": 17006 + }, + { + "epoch": 1.43, + "learning_rate": 1.1217721059570201e-05, + "loss": 1.1493, + "step": 17007 + }, + { + "epoch": 1.43, + "learning_rate": 1.1216818845990573e-05, + "loss": 1.1736, + "step": 17008 + }, + { + "epoch": 1.43, + "learning_rate": 1.12159166223572e-05, + "loss": 1.2674, + "step": 17009 + }, + { + "epoch": 1.43, + "learning_rate": 1.1215014388677533e-05, + "loss": 1.2671, + "step": 17010 + }, + { + "epoch": 1.43, + "learning_rate": 1.1214112144959031e-05, + "loss": 1.1095, + "step": 17011 + }, + { + "epoch": 1.43, + "learning_rate": 1.1213209891209145e-05, + "loss": 1.3097, + "step": 17012 + }, + { + "epoch": 1.43, + "learning_rate": 1.1212307627435328e-05, + "loss": 1.2149, + "step": 17013 + }, + { + "epoch": 1.43, + "learning_rate": 1.121140535364504e-05, + "loss": 1.1894, + "step": 17014 + }, + { + "epoch": 1.43, + "learning_rate": 1.1210503069845732e-05, + "loss": 1.1578, + "step": 17015 + }, + { + "epoch": 1.43, + "learning_rate": 1.120960077604486e-05, + "loss": 1.2391, + "step": 17016 + }, + { + "epoch": 1.43, + "learning_rate": 1.120869847224988e-05, + "loss": 1.136, + "step": 17017 + }, + { + "epoch": 1.43, + "learning_rate": 1.1207796158468245e-05, + "loss": 1.1762, + "step": 17018 + }, + { + "epoch": 1.43, + "learning_rate": 1.1206893834707415e-05, + "loss": 1.145, + "step": 17019 + }, + { + "epoch": 1.43, + "learning_rate": 1.1205991500974838e-05, + "loss": 1.1866, + "step": 17020 + }, + { + "epoch": 1.43, + "learning_rate": 1.1205089157277973e-05, + "loss": 1.1575, + "step": 17021 + }, + { + "epoch": 1.43, + "learning_rate": 1.120418680362428e-05, + "loss": 1.1581, + "step": 17022 + }, + { + "epoch": 1.43, + "learning_rate": 1.1203284440021207e-05, + "loss": 1.2146, + "step": 17023 + }, + { + "epoch": 1.43, + "learning_rate": 1.1202382066476213e-05, + "loss": 1.1217, + "step": 17024 + }, + { + "epoch": 1.43, + "learning_rate": 1.1201479682996754e-05, + "loss": 1.1902, + "step": 17025 + }, + { + "epoch": 1.43, + "learning_rate": 1.1200577289590288e-05, + "loss": 1.1639, + "step": 17026 + }, + { + "epoch": 1.43, + "learning_rate": 1.1199674886264265e-05, + "loss": 1.2358, + "step": 17027 + }, + { + "epoch": 1.43, + "learning_rate": 1.1198772473026144e-05, + "loss": 1.204, + "step": 17028 + }, + { + "epoch": 1.43, + "learning_rate": 1.1197870049883384e-05, + "loss": 1.1811, + "step": 17029 + }, + { + "epoch": 1.43, + "learning_rate": 1.1196967616843437e-05, + "loss": 1.1273, + "step": 17030 + }, + { + "epoch": 1.43, + "learning_rate": 1.1196065173913759e-05, + "loss": 1.1831, + "step": 17031 + }, + { + "epoch": 1.43, + "learning_rate": 1.1195162721101806e-05, + "loss": 1.1829, + "step": 17032 + }, + { + "epoch": 1.43, + "learning_rate": 1.1194260258415042e-05, + "loss": 1.2659, + "step": 17033 + }, + { + "epoch": 1.43, + "learning_rate": 1.1193357785860911e-05, + "loss": 1.2163, + "step": 17034 + }, + { + "epoch": 1.43, + "learning_rate": 1.119245530344688e-05, + "loss": 1.1181, + "step": 17035 + }, + { + "epoch": 1.43, + "learning_rate": 1.1191552811180404e-05, + "loss": 1.2424, + "step": 17036 + }, + { + "epoch": 1.43, + "learning_rate": 1.1190650309068934e-05, + "loss": 1.2933, + "step": 17037 + }, + { + "epoch": 1.43, + "learning_rate": 1.1189747797119929e-05, + "loss": 1.2238, + "step": 17038 + }, + { + "epoch": 1.43, + "learning_rate": 1.1188845275340846e-05, + "loss": 1.2004, + "step": 17039 + }, + { + "epoch": 1.43, + "learning_rate": 1.1187942743739148e-05, + "loss": 1.2887, + "step": 17040 + }, + { + "epoch": 1.43, + "learning_rate": 1.1187040202322286e-05, + "loss": 1.106, + "step": 17041 + }, + { + "epoch": 1.43, + "learning_rate": 1.1186137651097715e-05, + "loss": 1.1695, + "step": 17042 + }, + { + "epoch": 1.43, + "learning_rate": 1.1185235090072894e-05, + "loss": 1.2236, + "step": 17043 + }, + { + "epoch": 1.43, + "learning_rate": 1.1184332519255286e-05, + "loss": 1.3101, + "step": 17044 + }, + { + "epoch": 1.44, + "learning_rate": 1.1183429938652339e-05, + "loss": 1.255, + "step": 17045 + }, + { + "epoch": 1.44, + "learning_rate": 1.1182527348271517e-05, + "loss": 1.2348, + "step": 17046 + }, + { + "epoch": 1.44, + "learning_rate": 1.1181624748120277e-05, + "loss": 1.1546, + "step": 17047 + }, + { + "epoch": 1.44, + "learning_rate": 1.1180722138206073e-05, + "loss": 1.1977, + "step": 17048 + }, + { + "epoch": 1.44, + "learning_rate": 1.1179819518536365e-05, + "loss": 1.2212, + "step": 17049 + }, + { + "epoch": 1.44, + "learning_rate": 1.1178916889118612e-05, + "loss": 1.2153, + "step": 17050 + }, + { + "epoch": 1.44, + "learning_rate": 1.1178014249960272e-05, + "loss": 1.2943, + "step": 17051 + }, + { + "epoch": 1.44, + "learning_rate": 1.1177111601068798e-05, + "loss": 1.1916, + "step": 17052 + }, + { + "epoch": 1.44, + "learning_rate": 1.1176208942451651e-05, + "loss": 1.1672, + "step": 17053 + }, + { + "epoch": 1.44, + "learning_rate": 1.1175306274116292e-05, + "loss": 1.1237, + "step": 17054 + }, + { + "epoch": 1.44, + "learning_rate": 1.1174403596070176e-05, + "loss": 1.1088, + "step": 17055 + }, + { + "epoch": 1.44, + "learning_rate": 1.117350090832076e-05, + "loss": 1.1587, + "step": 17056 + }, + { + "epoch": 1.44, + "learning_rate": 1.11725982108755e-05, + "loss": 1.2435, + "step": 17057 + }, + { + "epoch": 1.44, + "learning_rate": 1.1171695503741866e-05, + "loss": 1.2556, + "step": 17058 + }, + { + "epoch": 1.44, + "learning_rate": 1.1170792786927304e-05, + "loss": 1.2522, + "step": 17059 + }, + { + "epoch": 1.44, + "learning_rate": 1.116989006043928e-05, + "loss": 1.1791, + "step": 17060 + }, + { + "epoch": 1.44, + "learning_rate": 1.116898732428525e-05, + "loss": 1.2094, + "step": 17061 + }, + { + "epoch": 1.44, + "learning_rate": 1.1168084578472671e-05, + "loss": 1.1847, + "step": 17062 + }, + { + "epoch": 1.44, + "learning_rate": 1.1167181823009004e-05, + "loss": 1.2424, + "step": 17063 + }, + { + "epoch": 1.44, + "learning_rate": 1.1166279057901706e-05, + "loss": 1.1317, + "step": 17064 + }, + { + "epoch": 1.44, + "learning_rate": 1.116537628315824e-05, + "loss": 1.2674, + "step": 17065 + }, + { + "epoch": 1.44, + "learning_rate": 1.116447349878606e-05, + "loss": 1.1856, + "step": 17066 + }, + { + "epoch": 1.44, + "learning_rate": 1.1163570704792626e-05, + "loss": 1.1065, + "step": 17067 + }, + { + "epoch": 1.44, + "learning_rate": 1.11626679011854e-05, + "loss": 1.1649, + "step": 17068 + }, + { + "epoch": 1.44, + "learning_rate": 1.116176508797184e-05, + "loss": 1.1757, + "step": 17069 + }, + { + "epoch": 1.44, + "learning_rate": 1.1160862265159405e-05, + "loss": 1.2267, + "step": 17070 + }, + { + "epoch": 1.44, + "learning_rate": 1.1159959432755551e-05, + "loss": 1.2372, + "step": 17071 + }, + { + "epoch": 1.44, + "learning_rate": 1.1159056590767748e-05, + "loss": 1.1164, + "step": 17072 + }, + { + "epoch": 1.44, + "learning_rate": 1.1158153739203443e-05, + "loss": 1.2438, + "step": 17073 + }, + { + "epoch": 1.44, + "learning_rate": 1.1157250878070102e-05, + "loss": 1.1864, + "step": 17074 + }, + { + "epoch": 1.44, + "learning_rate": 1.1156348007375183e-05, + "loss": 1.2203, + "step": 17075 + }, + { + "epoch": 1.44, + "learning_rate": 1.1155445127126152e-05, + "loss": 1.1992, + "step": 17076 + }, + { + "epoch": 1.44, + "learning_rate": 1.115454223733046e-05, + "loss": 1.2012, + "step": 17077 + }, + { + "epoch": 1.44, + "learning_rate": 1.115363933799557e-05, + "loss": 1.2513, + "step": 17078 + }, + { + "epoch": 1.44, + "learning_rate": 1.115273642912894e-05, + "loss": 1.2687, + "step": 17079 + }, + { + "epoch": 1.44, + "learning_rate": 1.1151833510738039e-05, + "loss": 1.1854, + "step": 17080 + }, + { + "epoch": 1.44, + "learning_rate": 1.1150930582830317e-05, + "loss": 1.2003, + "step": 17081 + }, + { + "epoch": 1.44, + "learning_rate": 1.1150027645413238e-05, + "loss": 1.0871, + "step": 17082 + }, + { + "epoch": 1.44, + "learning_rate": 1.1149124698494265e-05, + "loss": 1.1841, + "step": 17083 + }, + { + "epoch": 1.44, + "learning_rate": 1.114822174208085e-05, + "loss": 1.1412, + "step": 17084 + }, + { + "epoch": 1.44, + "learning_rate": 1.1147318776180467e-05, + "loss": 1.2788, + "step": 17085 + }, + { + "epoch": 1.44, + "learning_rate": 1.1146415800800568e-05, + "loss": 1.2472, + "step": 17086 + }, + { + "epoch": 1.44, + "learning_rate": 1.1145512815948612e-05, + "loss": 1.1108, + "step": 17087 + }, + { + "epoch": 1.44, + "learning_rate": 1.1144609821632062e-05, + "loss": 1.1844, + "step": 17088 + }, + { + "epoch": 1.44, + "learning_rate": 1.1143706817858379e-05, + "loss": 1.1424, + "step": 17089 + }, + { + "epoch": 1.44, + "learning_rate": 1.1142803804635027e-05, + "loss": 1.2132, + "step": 17090 + }, + { + "epoch": 1.44, + "learning_rate": 1.1141900781969464e-05, + "loss": 1.1391, + "step": 17091 + }, + { + "epoch": 1.44, + "learning_rate": 1.114099774986915e-05, + "loss": 1.1708, + "step": 17092 + }, + { + "epoch": 1.44, + "learning_rate": 1.1140094708341547e-05, + "loss": 1.1768, + "step": 17093 + }, + { + "epoch": 1.44, + "learning_rate": 1.1139191657394117e-05, + "loss": 1.143, + "step": 17094 + }, + { + "epoch": 1.44, + "learning_rate": 1.1138288597034319e-05, + "loss": 1.1721, + "step": 17095 + }, + { + "epoch": 1.44, + "learning_rate": 1.1137385527269619e-05, + "loss": 1.2039, + "step": 17096 + }, + { + "epoch": 1.44, + "learning_rate": 1.1136482448107478e-05, + "loss": 1.1759, + "step": 17097 + }, + { + "epoch": 1.44, + "learning_rate": 1.113557935955535e-05, + "loss": 1.1444, + "step": 17098 + }, + { + "epoch": 1.44, + "learning_rate": 1.1134676261620705e-05, + "loss": 1.2614, + "step": 17099 + }, + { + "epoch": 1.44, + "learning_rate": 1.1133773154311e-05, + "loss": 1.2142, + "step": 17100 + }, + { + "epoch": 1.44, + "learning_rate": 1.1132870037633704e-05, + "loss": 1.1763, + "step": 17101 + }, + { + "epoch": 1.44, + "learning_rate": 1.1131966911596267e-05, + "loss": 1.1947, + "step": 17102 + }, + { + "epoch": 1.44, + "learning_rate": 1.1131063776206158e-05, + "loss": 1.1463, + "step": 17103 + }, + { + "epoch": 1.44, + "learning_rate": 1.1130160631470836e-05, + "loss": 1.1404, + "step": 17104 + }, + { + "epoch": 1.44, + "learning_rate": 1.112925747739777e-05, + "loss": 1.1323, + "step": 17105 + }, + { + "epoch": 1.44, + "learning_rate": 1.1128354313994416e-05, + "loss": 1.176, + "step": 17106 + }, + { + "epoch": 1.44, + "learning_rate": 1.1127451141268235e-05, + "loss": 1.1993, + "step": 17107 + }, + { + "epoch": 1.44, + "learning_rate": 1.1126547959226694e-05, + "loss": 1.2135, + "step": 17108 + }, + { + "epoch": 1.44, + "learning_rate": 1.1125644767877254e-05, + "loss": 1.1307, + "step": 17109 + }, + { + "epoch": 1.44, + "learning_rate": 1.1124741567227376e-05, + "loss": 1.2715, + "step": 17110 + }, + { + "epoch": 1.44, + "learning_rate": 1.1123838357284522e-05, + "loss": 1.2092, + "step": 17111 + }, + { + "epoch": 1.44, + "learning_rate": 1.1122935138056157e-05, + "loss": 1.1864, + "step": 17112 + }, + { + "epoch": 1.44, + "learning_rate": 1.1122031909549741e-05, + "loss": 1.2119, + "step": 17113 + }, + { + "epoch": 1.44, + "learning_rate": 1.1121128671772738e-05, + "loss": 1.1974, + "step": 17114 + }, + { + "epoch": 1.44, + "learning_rate": 1.1120225424732615e-05, + "loss": 1.267, + "step": 17115 + }, + { + "epoch": 1.44, + "learning_rate": 1.1119322168436829e-05, + "loss": 1.2219, + "step": 17116 + }, + { + "epoch": 1.44, + "learning_rate": 1.1118418902892842e-05, + "loss": 1.1538, + "step": 17117 + }, + { + "epoch": 1.44, + "learning_rate": 1.1117515628108125e-05, + "loss": 1.2333, + "step": 17118 + }, + { + "epoch": 1.44, + "learning_rate": 1.1116612344090134e-05, + "loss": 1.1482, + "step": 17119 + }, + { + "epoch": 1.44, + "learning_rate": 1.1115709050846331e-05, + "loss": 1.2234, + "step": 17120 + }, + { + "epoch": 1.44, + "learning_rate": 1.1114805748384186e-05, + "loss": 1.1863, + "step": 17121 + }, + { + "epoch": 1.44, + "learning_rate": 1.111390243671116e-05, + "loss": 1.2061, + "step": 17122 + }, + { + "epoch": 1.44, + "learning_rate": 1.1112999115834716e-05, + "loss": 1.253, + "step": 17123 + }, + { + "epoch": 1.44, + "learning_rate": 1.1112095785762315e-05, + "loss": 1.2858, + "step": 17124 + }, + { + "epoch": 1.44, + "learning_rate": 1.1111192446501422e-05, + "loss": 1.2133, + "step": 17125 + }, + { + "epoch": 1.44, + "learning_rate": 1.1110289098059506e-05, + "loss": 1.1037, + "step": 17126 + }, + { + "epoch": 1.44, + "learning_rate": 1.1109385740444023e-05, + "loss": 1.2101, + "step": 17127 + }, + { + "epoch": 1.44, + "learning_rate": 1.110848237366244e-05, + "loss": 1.192, + "step": 17128 + }, + { + "epoch": 1.44, + "learning_rate": 1.110757899772222e-05, + "loss": 1.1198, + "step": 17129 + }, + { + "epoch": 1.44, + "learning_rate": 1.1106675612630828e-05, + "loss": 1.2009, + "step": 17130 + }, + { + "epoch": 1.44, + "learning_rate": 1.1105772218395729e-05, + "loss": 1.2417, + "step": 17131 + }, + { + "epoch": 1.44, + "learning_rate": 1.1104868815024384e-05, + "loss": 1.1785, + "step": 17132 + }, + { + "epoch": 1.44, + "learning_rate": 1.1103965402524261e-05, + "loss": 1.1386, + "step": 17133 + }, + { + "epoch": 1.44, + "learning_rate": 1.1103061980902822e-05, + "loss": 1.1587, + "step": 17134 + }, + { + "epoch": 1.44, + "learning_rate": 1.1102158550167532e-05, + "loss": 1.1985, + "step": 17135 + }, + { + "epoch": 1.44, + "learning_rate": 1.1101255110325859e-05, + "loss": 1.2029, + "step": 17136 + }, + { + "epoch": 1.44, + "learning_rate": 1.110035166138526e-05, + "loss": 1.2891, + "step": 17137 + }, + { + "epoch": 1.44, + "learning_rate": 1.1099448203353202e-05, + "loss": 1.1317, + "step": 17138 + }, + { + "epoch": 1.44, + "learning_rate": 1.1098544736237154e-05, + "loss": 1.1916, + "step": 17139 + }, + { + "epoch": 1.44, + "learning_rate": 1.1097641260044579e-05, + "loss": 1.1708, + "step": 17140 + }, + { + "epoch": 1.44, + "learning_rate": 1.1096737774782937e-05, + "loss": 1.1949, + "step": 17141 + }, + { + "epoch": 1.44, + "learning_rate": 1.1095834280459696e-05, + "loss": 1.2191, + "step": 17142 + }, + { + "epoch": 1.44, + "learning_rate": 1.1094930777082325e-05, + "loss": 1.2954, + "step": 17143 + }, + { + "epoch": 1.44, + "learning_rate": 1.1094027264658285e-05, + "loss": 1.131, + "step": 17144 + }, + { + "epoch": 1.44, + "learning_rate": 1.1093123743195038e-05, + "loss": 1.11, + "step": 17145 + }, + { + "epoch": 1.44, + "learning_rate": 1.1092220212700053e-05, + "loss": 1.2486, + "step": 17146 + }, + { + "epoch": 1.44, + "learning_rate": 1.10913166731808e-05, + "loss": 1.2358, + "step": 17147 + }, + { + "epoch": 1.44, + "learning_rate": 1.1090413124644737e-05, + "loss": 1.1433, + "step": 17148 + }, + { + "epoch": 1.44, + "learning_rate": 1.108950956709933e-05, + "loss": 1.1924, + "step": 17149 + }, + { + "epoch": 1.44, + "learning_rate": 1.1088606000552048e-05, + "loss": 1.1164, + "step": 17150 + }, + { + "epoch": 1.44, + "learning_rate": 1.1087702425010356e-05, + "loss": 1.1653, + "step": 17151 + }, + { + "epoch": 1.44, + "learning_rate": 1.1086798840481716e-05, + "loss": 1.1957, + "step": 17152 + }, + { + "epoch": 1.44, + "learning_rate": 1.1085895246973597e-05, + "loss": 1.2395, + "step": 17153 + }, + { + "epoch": 1.44, + "learning_rate": 1.1084991644493463e-05, + "loss": 1.2556, + "step": 17154 + }, + { + "epoch": 1.44, + "learning_rate": 1.1084088033048786e-05, + "loss": 1.2188, + "step": 17155 + }, + { + "epoch": 1.44, + "learning_rate": 1.108318441264702e-05, + "loss": 1.2439, + "step": 17156 + }, + { + "epoch": 1.44, + "learning_rate": 1.1082280783295638e-05, + "loss": 1.2558, + "step": 17157 + }, + { + "epoch": 1.44, + "learning_rate": 1.108137714500211e-05, + "loss": 1.1708, + "step": 17158 + }, + { + "epoch": 1.44, + "learning_rate": 1.1080473497773897e-05, + "loss": 1.1535, + "step": 17159 + }, + { + "epoch": 1.44, + "learning_rate": 1.1079569841618462e-05, + "loss": 1.218, + "step": 17160 + }, + { + "epoch": 1.44, + "learning_rate": 1.1078666176543282e-05, + "loss": 1.1223, + "step": 17161 + }, + { + "epoch": 1.44, + "learning_rate": 1.1077762502555813e-05, + "loss": 1.1414, + "step": 17162 + }, + { + "epoch": 1.44, + "learning_rate": 1.1076858819663525e-05, + "loss": 1.1915, + "step": 17163 + }, + { + "epoch": 1.45, + "learning_rate": 1.1075955127873885e-05, + "loss": 1.2105, + "step": 17164 + }, + { + "epoch": 1.45, + "learning_rate": 1.1075051427194363e-05, + "loss": 1.1907, + "step": 17165 + }, + { + "epoch": 1.45, + "learning_rate": 1.1074147717632418e-05, + "loss": 1.1193, + "step": 17166 + }, + { + "epoch": 1.45, + "learning_rate": 1.1073243999195522e-05, + "loss": 1.2832, + "step": 17167 + }, + { + "epoch": 1.45, + "learning_rate": 1.107234027189114e-05, + "loss": 1.2223, + "step": 17168 + }, + { + "epoch": 1.45, + "learning_rate": 1.107143653572674e-05, + "loss": 1.1423, + "step": 17169 + }, + { + "epoch": 1.45, + "learning_rate": 1.1070532790709787e-05, + "loss": 1.1373, + "step": 17170 + }, + { + "epoch": 1.45, + "learning_rate": 1.1069629036847748e-05, + "loss": 1.1851, + "step": 17171 + }, + { + "epoch": 1.45, + "learning_rate": 1.1068725274148098e-05, + "loss": 1.1916, + "step": 17172 + }, + { + "epoch": 1.45, + "learning_rate": 1.1067821502618294e-05, + "loss": 1.1188, + "step": 17173 + }, + { + "epoch": 1.45, + "learning_rate": 1.1066917722265806e-05, + "loss": 1.1939, + "step": 17174 + }, + { + "epoch": 1.45, + "learning_rate": 1.1066013933098103e-05, + "loss": 1.1761, + "step": 17175 + }, + { + "epoch": 1.45, + "learning_rate": 1.1065110135122654e-05, + "loss": 1.2495, + "step": 17176 + }, + { + "epoch": 1.45, + "learning_rate": 1.1064206328346923e-05, + "loss": 1.3046, + "step": 17177 + }, + { + "epoch": 1.45, + "learning_rate": 1.1063302512778378e-05, + "loss": 1.2056, + "step": 17178 + }, + { + "epoch": 1.45, + "learning_rate": 1.1062398688424485e-05, + "loss": 1.1647, + "step": 17179 + }, + { + "epoch": 1.45, + "learning_rate": 1.1061494855292719e-05, + "loss": 1.1114, + "step": 17180 + }, + { + "epoch": 1.45, + "learning_rate": 1.106059101339054e-05, + "loss": 1.2457, + "step": 17181 + }, + { + "epoch": 1.45, + "learning_rate": 1.1059687162725417e-05, + "loss": 1.1757, + "step": 17182 + }, + { + "epoch": 1.45, + "learning_rate": 1.1058783303304823e-05, + "loss": 1.2396, + "step": 17183 + }, + { + "epoch": 1.45, + "learning_rate": 1.1057879435136218e-05, + "loss": 1.1198, + "step": 17184 + }, + { + "epoch": 1.45, + "learning_rate": 1.1056975558227078e-05, + "loss": 1.1691, + "step": 17185 + }, + { + "epoch": 1.45, + "learning_rate": 1.1056071672584867e-05, + "loss": 1.1015, + "step": 17186 + }, + { + "epoch": 1.45, + "learning_rate": 1.1055167778217055e-05, + "loss": 1.2227, + "step": 17187 + }, + { + "epoch": 1.45, + "learning_rate": 1.1054263875131108e-05, + "loss": 1.1341, + "step": 17188 + }, + { + "epoch": 1.45, + "learning_rate": 1.1053359963334495e-05, + "loss": 1.1411, + "step": 17189 + }, + { + "epoch": 1.45, + "learning_rate": 1.1052456042834687e-05, + "loss": 1.1569, + "step": 17190 + }, + { + "epoch": 1.45, + "learning_rate": 1.1051552113639148e-05, + "loss": 1.1623, + "step": 17191 + }, + { + "epoch": 1.45, + "learning_rate": 1.1050648175755349e-05, + "loss": 1.1742, + "step": 17192 + }, + { + "epoch": 1.45, + "learning_rate": 1.104974422919076e-05, + "loss": 1.1609, + "step": 17193 + }, + { + "epoch": 1.45, + "learning_rate": 1.1048840273952848e-05, + "loss": 1.2541, + "step": 17194 + }, + { + "epoch": 1.45, + "learning_rate": 1.1047936310049078e-05, + "loss": 1.2591, + "step": 17195 + }, + { + "epoch": 1.45, + "learning_rate": 1.1047032337486927e-05, + "loss": 1.2768, + "step": 17196 + }, + { + "epoch": 1.45, + "learning_rate": 1.104612835627386e-05, + "loss": 1.1355, + "step": 17197 + }, + { + "epoch": 1.45, + "learning_rate": 1.1045224366417344e-05, + "loss": 1.2257, + "step": 17198 + }, + { + "epoch": 1.45, + "learning_rate": 1.1044320367924851e-05, + "loss": 1.1931, + "step": 17199 + }, + { + "epoch": 1.45, + "learning_rate": 1.1043416360803847e-05, + "loss": 1.2641, + "step": 17200 + }, + { + "epoch": 1.45, + "learning_rate": 1.1042512345061806e-05, + "loss": 1.202, + "step": 17201 + }, + { + "epoch": 1.45, + "learning_rate": 1.1041608320706193e-05, + "loss": 1.2134, + "step": 17202 + }, + { + "epoch": 1.45, + "learning_rate": 1.1040704287744477e-05, + "loss": 1.2095, + "step": 17203 + }, + { + "epoch": 1.45, + "learning_rate": 1.1039800246184131e-05, + "loss": 1.1556, + "step": 17204 + }, + { + "epoch": 1.45, + "learning_rate": 1.1038896196032622e-05, + "loss": 1.25, + "step": 17205 + }, + { + "epoch": 1.45, + "learning_rate": 1.1037992137297422e-05, + "loss": 1.2416, + "step": 17206 + }, + { + "epoch": 1.45, + "learning_rate": 1.1037088069985992e-05, + "loss": 1.146, + "step": 17207 + }, + { + "epoch": 1.45, + "learning_rate": 1.1036183994105815e-05, + "loss": 1.1354, + "step": 17208 + }, + { + "epoch": 1.45, + "learning_rate": 1.1035279909664354e-05, + "loss": 1.2708, + "step": 17209 + }, + { + "epoch": 1.45, + "learning_rate": 1.1034375816669077e-05, + "loss": 1.1677, + "step": 17210 + }, + { + "epoch": 1.45, + "learning_rate": 1.1033471715127455e-05, + "loss": 1.1555, + "step": 17211 + }, + { + "epoch": 1.45, + "learning_rate": 1.1032567605046964e-05, + "loss": 1.0945, + "step": 17212 + }, + { + "epoch": 1.45, + "learning_rate": 1.1031663486435065e-05, + "loss": 1.226, + "step": 17213 + }, + { + "epoch": 1.45, + "learning_rate": 1.1030759359299232e-05, + "loss": 1.197, + "step": 17214 + }, + { + "epoch": 1.45, + "learning_rate": 1.1029855223646937e-05, + "loss": 1.3111, + "step": 17215 + }, + { + "epoch": 1.45, + "learning_rate": 1.1028951079485648e-05, + "loss": 1.3033, + "step": 17216 + }, + { + "epoch": 1.45, + "learning_rate": 1.1028046926822835e-05, + "loss": 1.1938, + "step": 17217 + }, + { + "epoch": 1.45, + "learning_rate": 1.1027142765665968e-05, + "loss": 1.1741, + "step": 17218 + }, + { + "epoch": 1.45, + "learning_rate": 1.1026238596022523e-05, + "loss": 1.2973, + "step": 17219 + }, + { + "epoch": 1.45, + "learning_rate": 1.102533441789996e-05, + "loss": 1.0876, + "step": 17220 + }, + { + "epoch": 1.45, + "learning_rate": 1.1024430231305761e-05, + "loss": 1.1726, + "step": 17221 + }, + { + "epoch": 1.45, + "learning_rate": 1.1023526036247392e-05, + "loss": 1.2161, + "step": 17222 + }, + { + "epoch": 1.45, + "learning_rate": 1.1022621832732322e-05, + "loss": 1.2295, + "step": 17223 + }, + { + "epoch": 1.45, + "learning_rate": 1.102171762076802e-05, + "loss": 1.1696, + "step": 17224 + }, + { + "epoch": 1.45, + "learning_rate": 1.1020813400361962e-05, + "loss": 1.3013, + "step": 17225 + }, + { + "epoch": 1.45, + "learning_rate": 1.101990917152162e-05, + "loss": 1.1145, + "step": 17226 + }, + { + "epoch": 1.45, + "learning_rate": 1.1019004934254457e-05, + "loss": 1.218, + "step": 17227 + }, + { + "epoch": 1.45, + "learning_rate": 1.1018100688567954e-05, + "loss": 1.106, + "step": 17228 + }, + { + "epoch": 1.45, + "learning_rate": 1.1017196434469574e-05, + "loss": 1.197, + "step": 17229 + }, + { + "epoch": 1.45, + "learning_rate": 1.1016292171966794e-05, + "loss": 1.1594, + "step": 17230 + }, + { + "epoch": 1.45, + "learning_rate": 1.101538790106708e-05, + "loss": 1.1478, + "step": 17231 + }, + { + "epoch": 1.45, + "learning_rate": 1.1014483621777906e-05, + "loss": 1.1715, + "step": 17232 + }, + { + "epoch": 1.45, + "learning_rate": 1.1013579334106747e-05, + "loss": 1.1863, + "step": 17233 + }, + { + "epoch": 1.45, + "learning_rate": 1.1012675038061068e-05, + "loss": 1.1173, + "step": 17234 + }, + { + "epoch": 1.45, + "learning_rate": 1.1011770733648346e-05, + "loss": 1.171, + "step": 17235 + }, + { + "epoch": 1.45, + "learning_rate": 1.101086642087605e-05, + "loss": 1.2275, + "step": 17236 + }, + { + "epoch": 1.45, + "learning_rate": 1.1009962099751652e-05, + "loss": 1.1736, + "step": 17237 + }, + { + "epoch": 1.45, + "learning_rate": 1.1009057770282623e-05, + "loss": 1.1583, + "step": 17238 + }, + { + "epoch": 1.45, + "learning_rate": 1.1008153432476436e-05, + "loss": 1.1716, + "step": 17239 + }, + { + "epoch": 1.45, + "learning_rate": 1.1007249086340563e-05, + "loss": 1.2874, + "step": 17240 + }, + { + "epoch": 1.45, + "learning_rate": 1.1006344731882478e-05, + "loss": 1.1897, + "step": 17241 + }, + { + "epoch": 1.45, + "learning_rate": 1.1005440369109646e-05, + "loss": 1.1486, + "step": 17242 + }, + { + "epoch": 1.45, + "learning_rate": 1.1004535998029547e-05, + "loss": 1.2084, + "step": 17243 + }, + { + "epoch": 1.45, + "learning_rate": 1.100363161864965e-05, + "loss": 1.2213, + "step": 17244 + }, + { + "epoch": 1.45, + "learning_rate": 1.1002727230977424e-05, + "loss": 1.1334, + "step": 17245 + }, + { + "epoch": 1.45, + "learning_rate": 1.1001822835020348e-05, + "loss": 1.2543, + "step": 17246 + }, + { + "epoch": 1.45, + "learning_rate": 1.1000918430785893e-05, + "loss": 1.1886, + "step": 17247 + }, + { + "epoch": 1.45, + "learning_rate": 1.1000014018281527e-05, + "loss": 1.1345, + "step": 17248 + }, + { + "epoch": 1.45, + "learning_rate": 1.0999109597514723e-05, + "loss": 1.1561, + "step": 17249 + }, + { + "epoch": 1.45, + "learning_rate": 1.0998205168492957e-05, + "loss": 1.0806, + "step": 17250 + }, + { + "epoch": 1.45, + "learning_rate": 1.0997300731223703e-05, + "loss": 1.2535, + "step": 17251 + }, + { + "epoch": 1.45, + "learning_rate": 1.099639628571443e-05, + "loss": 1.1468, + "step": 17252 + }, + { + "epoch": 1.45, + "learning_rate": 1.099549183197261e-05, + "loss": 1.1457, + "step": 17253 + }, + { + "epoch": 1.45, + "learning_rate": 1.099458737000572e-05, + "loss": 1.1641, + "step": 17254 + }, + { + "epoch": 1.45, + "learning_rate": 1.0993682899821232e-05, + "loss": 1.181, + "step": 17255 + }, + { + "epoch": 1.45, + "learning_rate": 1.0992778421426615e-05, + "loss": 1.1888, + "step": 17256 + }, + { + "epoch": 1.45, + "learning_rate": 1.0991873934829344e-05, + "loss": 1.1525, + "step": 17257 + }, + { + "epoch": 1.45, + "learning_rate": 1.0990969440036898e-05, + "loss": 1.1583, + "step": 17258 + }, + { + "epoch": 1.45, + "learning_rate": 1.0990064937056742e-05, + "loss": 1.2297, + "step": 17259 + }, + { + "epoch": 1.45, + "learning_rate": 1.0989160425896352e-05, + "loss": 1.2272, + "step": 17260 + }, + { + "epoch": 1.45, + "learning_rate": 1.0988255906563203e-05, + "loss": 1.1132, + "step": 17261 + }, + { + "epoch": 1.45, + "learning_rate": 1.0987351379064771e-05, + "loss": 1.2939, + "step": 17262 + }, + { + "epoch": 1.45, + "learning_rate": 1.098644684340852e-05, + "loss": 1.2198, + "step": 17263 + }, + { + "epoch": 1.45, + "learning_rate": 1.0985542299601932e-05, + "loss": 1.2109, + "step": 17264 + }, + { + "epoch": 1.45, + "learning_rate": 1.098463774765248e-05, + "loss": 1.1937, + "step": 17265 + }, + { + "epoch": 1.45, + "learning_rate": 1.0983733187567633e-05, + "loss": 1.1967, + "step": 17266 + }, + { + "epoch": 1.45, + "learning_rate": 1.0982828619354868e-05, + "loss": 1.1398, + "step": 17267 + }, + { + "epoch": 1.45, + "learning_rate": 1.0981924043021659e-05, + "loss": 1.2476, + "step": 17268 + }, + { + "epoch": 1.45, + "learning_rate": 1.0981019458575481e-05, + "loss": 1.1627, + "step": 17269 + }, + { + "epoch": 1.45, + "learning_rate": 1.0980114866023804e-05, + "loss": 1.2278, + "step": 17270 + }, + { + "epoch": 1.45, + "learning_rate": 1.0979210265374103e-05, + "loss": 1.126, + "step": 17271 + }, + { + "epoch": 1.45, + "learning_rate": 1.0978305656633856e-05, + "loss": 1.1904, + "step": 17272 + }, + { + "epoch": 1.45, + "learning_rate": 1.0977401039810534e-05, + "loss": 1.163, + "step": 17273 + }, + { + "epoch": 1.45, + "learning_rate": 1.097649641491161e-05, + "loss": 1.192, + "step": 17274 + }, + { + "epoch": 1.45, + "learning_rate": 1.0975591781944563e-05, + "loss": 1.1335, + "step": 17275 + }, + { + "epoch": 1.45, + "learning_rate": 1.0974687140916865e-05, + "loss": 1.1653, + "step": 17276 + }, + { + "epoch": 1.45, + "learning_rate": 1.0973782491835987e-05, + "loss": 1.2471, + "step": 17277 + }, + { + "epoch": 1.45, + "learning_rate": 1.0972877834709407e-05, + "loss": 1.1112, + "step": 17278 + }, + { + "epoch": 1.45, + "learning_rate": 1.0971973169544599e-05, + "loss": 1.1156, + "step": 17279 + }, + { + "epoch": 1.45, + "learning_rate": 1.097106849634904e-05, + "loss": 1.2173, + "step": 17280 + }, + { + "epoch": 1.45, + "learning_rate": 1.09701638151302e-05, + "loss": 1.163, + "step": 17281 + }, + { + "epoch": 1.45, + "learning_rate": 1.0969259125895554e-05, + "loss": 1.1729, + "step": 17282 + }, + { + "epoch": 1.46, + "learning_rate": 1.0968354428652584e-05, + "loss": 1.2376, + "step": 17283 + }, + { + "epoch": 1.46, + "learning_rate": 1.0967449723408755e-05, + "loss": 1.1285, + "step": 17284 + }, + { + "epoch": 1.46, + "learning_rate": 1.0966545010171548e-05, + "loss": 1.1994, + "step": 17285 + }, + { + "epoch": 1.46, + "learning_rate": 1.0965640288948437e-05, + "loss": 1.1676, + "step": 17286 + }, + { + "epoch": 1.46, + "learning_rate": 1.09647355597469e-05, + "loss": 1.154, + "step": 17287 + }, + { + "epoch": 1.46, + "learning_rate": 1.0963830822574405e-05, + "loss": 1.1658, + "step": 17288 + }, + { + "epoch": 1.46, + "learning_rate": 1.0962926077438435e-05, + "loss": 1.1305, + "step": 17289 + }, + { + "epoch": 1.46, + "learning_rate": 1.096202132434646e-05, + "loss": 1.1051, + "step": 17290 + }, + { + "epoch": 1.46, + "learning_rate": 1.0961116563305956e-05, + "loss": 1.1269, + "step": 17291 + }, + { + "epoch": 1.46, + "learning_rate": 1.0960211794324398e-05, + "loss": 1.2582, + "step": 17292 + }, + { + "epoch": 1.46, + "learning_rate": 1.0959307017409264e-05, + "loss": 1.198, + "step": 17293 + }, + { + "epoch": 1.46, + "learning_rate": 1.0958402232568031e-05, + "loss": 1.2009, + "step": 17294 + }, + { + "epoch": 1.46, + "learning_rate": 1.0957497439808166e-05, + "loss": 1.2386, + "step": 17295 + }, + { + "epoch": 1.46, + "learning_rate": 1.0956592639137156e-05, + "loss": 1.1838, + "step": 17296 + }, + { + "epoch": 1.46, + "learning_rate": 1.0955687830562471e-05, + "loss": 1.3022, + "step": 17297 + }, + { + "epoch": 1.46, + "learning_rate": 1.0954783014091586e-05, + "loss": 1.1625, + "step": 17298 + }, + { + "epoch": 1.46, + "learning_rate": 1.0953878189731979e-05, + "loss": 1.1434, + "step": 17299 + }, + { + "epoch": 1.46, + "learning_rate": 1.0952973357491124e-05, + "loss": 1.146, + "step": 17300 + }, + { + "epoch": 1.46, + "learning_rate": 1.0952068517376502e-05, + "loss": 1.13, + "step": 17301 + }, + { + "epoch": 1.46, + "learning_rate": 1.0951163669395581e-05, + "loss": 1.2118, + "step": 17302 + }, + { + "epoch": 1.46, + "learning_rate": 1.0950258813555841e-05, + "loss": 1.1372, + "step": 17303 + }, + { + "epoch": 1.46, + "learning_rate": 1.094935394986476e-05, + "loss": 1.3014, + "step": 17304 + }, + { + "epoch": 1.46, + "learning_rate": 1.0948449078329815e-05, + "loss": 1.1211, + "step": 17305 + }, + { + "epoch": 1.46, + "learning_rate": 1.0947544198958478e-05, + "loss": 1.224, + "step": 17306 + }, + { + "epoch": 1.46, + "learning_rate": 1.0946639311758224e-05, + "loss": 1.1996, + "step": 17307 + }, + { + "epoch": 1.46, + "learning_rate": 1.0945734416736538e-05, + "loss": 1.2011, + "step": 17308 + }, + { + "epoch": 1.46, + "learning_rate": 1.094482951390089e-05, + "loss": 1.219, + "step": 17309 + }, + { + "epoch": 1.46, + "learning_rate": 1.0943924603258758e-05, + "loss": 1.2362, + "step": 17310 + }, + { + "epoch": 1.46, + "learning_rate": 1.0943019684817619e-05, + "loss": 1.2089, + "step": 17311 + }, + { + "epoch": 1.46, + "learning_rate": 1.0942114758584952e-05, + "loss": 1.2272, + "step": 17312 + }, + { + "epoch": 1.46, + "learning_rate": 1.0941209824568228e-05, + "loss": 1.2134, + "step": 17313 + }, + { + "epoch": 1.46, + "learning_rate": 1.0940304882774927e-05, + "loss": 1.1754, + "step": 17314 + }, + { + "epoch": 1.46, + "learning_rate": 1.0939399933212529e-05, + "loss": 1.2601, + "step": 17315 + }, + { + "epoch": 1.46, + "learning_rate": 1.0938494975888508e-05, + "loss": 1.1098, + "step": 17316 + }, + { + "epoch": 1.46, + "learning_rate": 1.0937590010810337e-05, + "loss": 1.183, + "step": 17317 + }, + { + "epoch": 1.46, + "learning_rate": 1.0936685037985501e-05, + "loss": 1.2059, + "step": 17318 + }, + { + "epoch": 1.46, + "learning_rate": 1.0935780057421475e-05, + "loss": 1.1942, + "step": 17319 + }, + { + "epoch": 1.46, + "learning_rate": 1.0934875069125728e-05, + "loss": 1.107, + "step": 17320 + }, + { + "epoch": 1.46, + "learning_rate": 1.0933970073105749e-05, + "loss": 1.2489, + "step": 17321 + }, + { + "epoch": 1.46, + "learning_rate": 1.0933065069369011e-05, + "loss": 1.1045, + "step": 17322 + }, + { + "epoch": 1.46, + "learning_rate": 1.0932160057922987e-05, + "loss": 1.1734, + "step": 17323 + }, + { + "epoch": 1.46, + "learning_rate": 1.0931255038775161e-05, + "loss": 1.2084, + "step": 17324 + }, + { + "epoch": 1.46, + "learning_rate": 1.0930350011933006e-05, + "loss": 1.3446, + "step": 17325 + }, + { + "epoch": 1.46, + "learning_rate": 1.0929444977404004e-05, + "loss": 1.1925, + "step": 17326 + }, + { + "epoch": 1.46, + "learning_rate": 1.092853993519563e-05, + "loss": 1.2114, + "step": 17327 + }, + { + "epoch": 1.46, + "learning_rate": 1.0927634885315358e-05, + "loss": 1.2308, + "step": 17328 + }, + { + "epoch": 1.46, + "learning_rate": 1.0926729827770672e-05, + "loss": 1.1635, + "step": 17329 + }, + { + "epoch": 1.46, + "learning_rate": 1.092582476256905e-05, + "loss": 1.2226, + "step": 17330 + }, + { + "epoch": 1.46, + "learning_rate": 1.0924919689717966e-05, + "loss": 1.2441, + "step": 17331 + }, + { + "epoch": 1.46, + "learning_rate": 1.0924014609224897e-05, + "loss": 1.1878, + "step": 17332 + }, + { + "epoch": 1.46, + "learning_rate": 1.0923109521097329e-05, + "loss": 1.1714, + "step": 17333 + }, + { + "epoch": 1.46, + "learning_rate": 1.0922204425342729e-05, + "loss": 1.2408, + "step": 17334 + }, + { + "epoch": 1.46, + "learning_rate": 1.0921299321968584e-05, + "loss": 1.177, + "step": 17335 + }, + { + "epoch": 1.46, + "learning_rate": 1.0920394210982366e-05, + "loss": 1.1625, + "step": 17336 + }, + { + "epoch": 1.46, + "learning_rate": 1.0919489092391561e-05, + "loss": 1.1253, + "step": 17337 + }, + { + "epoch": 1.46, + "learning_rate": 1.091858396620364e-05, + "loss": 1.2418, + "step": 17338 + }, + { + "epoch": 1.46, + "learning_rate": 1.0917678832426085e-05, + "loss": 1.1164, + "step": 17339 + }, + { + "epoch": 1.46, + "learning_rate": 1.0916773691066373e-05, + "loss": 1.1229, + "step": 17340 + }, + { + "epoch": 1.46, + "learning_rate": 1.0915868542131987e-05, + "loss": 1.2329, + "step": 17341 + }, + { + "epoch": 1.46, + "learning_rate": 1.0914963385630398e-05, + "loss": 1.1913, + "step": 17342 + }, + { + "epoch": 1.46, + "learning_rate": 1.0914058221569089e-05, + "loss": 1.2178, + "step": 17343 + }, + { + "epoch": 1.46, + "learning_rate": 1.091315304995554e-05, + "loss": 1.2064, + "step": 17344 + }, + { + "epoch": 1.46, + "learning_rate": 1.0912247870797226e-05, + "loss": 1.2699, + "step": 17345 + }, + { + "epoch": 1.46, + "learning_rate": 1.091134268410163e-05, + "loss": 1.2457, + "step": 17346 + }, + { + "epoch": 1.46, + "learning_rate": 1.0910437489876229e-05, + "loss": 1.2058, + "step": 17347 + }, + { + "epoch": 1.46, + "learning_rate": 1.0909532288128501e-05, + "loss": 1.1586, + "step": 17348 + }, + { + "epoch": 1.46, + "learning_rate": 1.0908627078865928e-05, + "loss": 1.1861, + "step": 17349 + }, + { + "epoch": 1.46, + "learning_rate": 1.0907721862095986e-05, + "loss": 1.2978, + "step": 17350 + }, + { + "epoch": 1.46, + "learning_rate": 1.0906816637826156e-05, + "loss": 1.2081, + "step": 17351 + }, + { + "epoch": 1.46, + "learning_rate": 1.0905911406063916e-05, + "loss": 1.1823, + "step": 17352 + }, + { + "epoch": 1.46, + "learning_rate": 1.0905006166816745e-05, + "loss": 1.1665, + "step": 17353 + }, + { + "epoch": 1.46, + "learning_rate": 1.0904100920092124e-05, + "loss": 1.1511, + "step": 17354 + }, + { + "epoch": 1.46, + "learning_rate": 1.0903195665897534e-05, + "loss": 1.1819, + "step": 17355 + }, + { + "epoch": 1.46, + "learning_rate": 1.0902290404240452e-05, + "loss": 1.3297, + "step": 17356 + }, + { + "epoch": 1.46, + "learning_rate": 1.0901385135128355e-05, + "loss": 1.1975, + "step": 17357 + }, + { + "epoch": 1.46, + "learning_rate": 1.0900479858568729e-05, + "loss": 1.1903, + "step": 17358 + }, + { + "epoch": 1.46, + "learning_rate": 1.0899574574569047e-05, + "loss": 1.2517, + "step": 17359 + }, + { + "epoch": 1.46, + "learning_rate": 1.0898669283136793e-05, + "loss": 1.1281, + "step": 17360 + }, + { + "epoch": 1.46, + "learning_rate": 1.0897763984279449e-05, + "loss": 1.1774, + "step": 17361 + }, + { + "epoch": 1.46, + "learning_rate": 1.089685867800449e-05, + "loss": 1.2626, + "step": 17362 + }, + { + "epoch": 1.46, + "learning_rate": 1.0895953364319396e-05, + "loss": 1.1824, + "step": 17363 + }, + { + "epoch": 1.46, + "learning_rate": 1.0895048043231648e-05, + "loss": 1.1098, + "step": 17364 + }, + { + "epoch": 1.46, + "learning_rate": 1.0894142714748728e-05, + "loss": 1.2106, + "step": 17365 + }, + { + "epoch": 1.46, + "learning_rate": 1.0893237378878116e-05, + "loss": 1.1422, + "step": 17366 + }, + { + "epoch": 1.46, + "learning_rate": 1.0892332035627288e-05, + "loss": 1.1611, + "step": 17367 + }, + { + "epoch": 1.46, + "learning_rate": 1.0891426685003728e-05, + "loss": 1.2472, + "step": 17368 + }, + { + "epoch": 1.46, + "learning_rate": 1.0890521327014917e-05, + "loss": 1.2603, + "step": 17369 + }, + { + "epoch": 1.46, + "learning_rate": 1.0889615961668332e-05, + "loss": 1.2003, + "step": 17370 + }, + { + "epoch": 1.46, + "learning_rate": 1.0888710588971455e-05, + "loss": 1.156, + "step": 17371 + }, + { + "epoch": 1.46, + "learning_rate": 1.088780520893177e-05, + "loss": 1.1327, + "step": 17372 + }, + { + "epoch": 1.46, + "learning_rate": 1.088689982155675e-05, + "loss": 1.2188, + "step": 17373 + }, + { + "epoch": 1.46, + "learning_rate": 1.0885994426853882e-05, + "loss": 1.2599, + "step": 17374 + }, + { + "epoch": 1.46, + "learning_rate": 1.0885089024830644e-05, + "loss": 1.1995, + "step": 17375 + }, + { + "epoch": 1.46, + "learning_rate": 1.088418361549452e-05, + "loss": 1.2476, + "step": 17376 + }, + { + "epoch": 1.46, + "learning_rate": 1.0883278198852985e-05, + "loss": 1.144, + "step": 17377 + }, + { + "epoch": 1.46, + "learning_rate": 1.088237277491352e-05, + "loss": 1.228, + "step": 17378 + }, + { + "epoch": 1.46, + "learning_rate": 1.0881467343683614e-05, + "loss": 1.2059, + "step": 17379 + }, + { + "epoch": 1.46, + "learning_rate": 1.0880561905170741e-05, + "loss": 1.1908, + "step": 17380 + }, + { + "epoch": 1.46, + "learning_rate": 1.0879656459382384e-05, + "loss": 1.2226, + "step": 17381 + }, + { + "epoch": 1.46, + "learning_rate": 1.087875100632602e-05, + "loss": 1.1502, + "step": 17382 + }, + { + "epoch": 1.46, + "learning_rate": 1.087784554600914e-05, + "loss": 1.1735, + "step": 17383 + }, + { + "epoch": 1.46, + "learning_rate": 1.0876940078439215e-05, + "loss": 1.118, + "step": 17384 + }, + { + "epoch": 1.46, + "learning_rate": 1.0876034603623732e-05, + "loss": 1.1574, + "step": 17385 + }, + { + "epoch": 1.46, + "learning_rate": 1.087512912157017e-05, + "loss": 1.2594, + "step": 17386 + }, + { + "epoch": 1.46, + "learning_rate": 1.0874223632286014e-05, + "loss": 1.1984, + "step": 17387 + }, + { + "epoch": 1.46, + "learning_rate": 1.0873318135778738e-05, + "loss": 1.1703, + "step": 17388 + }, + { + "epoch": 1.46, + "learning_rate": 1.0872412632055832e-05, + "loss": 1.1185, + "step": 17389 + }, + { + "epoch": 1.46, + "learning_rate": 1.087150712112477e-05, + "loss": 1.2376, + "step": 17390 + }, + { + "epoch": 1.46, + "learning_rate": 1.0870601602993041e-05, + "loss": 1.1676, + "step": 17391 + }, + { + "epoch": 1.46, + "learning_rate": 1.0869696077668121e-05, + "loss": 1.0754, + "step": 17392 + }, + { + "epoch": 1.46, + "learning_rate": 1.0868790545157492e-05, + "loss": 1.2041, + "step": 17393 + }, + { + "epoch": 1.46, + "learning_rate": 1.0867885005468642e-05, + "loss": 1.2668, + "step": 17394 + }, + { + "epoch": 1.46, + "learning_rate": 1.0866979458609044e-05, + "loss": 1.2839, + "step": 17395 + }, + { + "epoch": 1.46, + "learning_rate": 1.0866073904586183e-05, + "loss": 1.1642, + "step": 17396 + }, + { + "epoch": 1.46, + "learning_rate": 1.0865168343407548e-05, + "loss": 1.1951, + "step": 17397 + }, + { + "epoch": 1.46, + "learning_rate": 1.0864262775080612e-05, + "loss": 1.2701, + "step": 17398 + }, + { + "epoch": 1.46, + "learning_rate": 1.086335719961286e-05, + "loss": 1.1887, + "step": 17399 + }, + { + "epoch": 1.46, + "learning_rate": 1.0862451617011773e-05, + "loss": 1.2175, + "step": 17400 + }, + { + "epoch": 1.47, + "learning_rate": 1.0861546027284839e-05, + "loss": 1.1506, + "step": 17401 + }, + { + "epoch": 1.47, + "learning_rate": 1.0860640430439535e-05, + "loss": 1.1445, + "step": 17402 + }, + { + "epoch": 1.47, + "learning_rate": 1.0859734826483342e-05, + "loss": 1.1983, + "step": 17403 + }, + { + "epoch": 1.47, + "learning_rate": 1.0858829215423746e-05, + "loss": 1.1826, + "step": 17404 + }, + { + "epoch": 1.47, + "learning_rate": 1.085792359726823e-05, + "loss": 1.2275, + "step": 17405 + }, + { + "epoch": 1.47, + "learning_rate": 1.085701797202427e-05, + "loss": 1.098, + "step": 17406 + }, + { + "epoch": 1.47, + "learning_rate": 1.0856112339699354e-05, + "loss": 1.1999, + "step": 17407 + }, + { + "epoch": 1.47, + "learning_rate": 1.085520670030097e-05, + "loss": 1.1692, + "step": 17408 + }, + { + "epoch": 1.47, + "learning_rate": 1.0854301053836589e-05, + "loss": 1.2584, + "step": 17409 + }, + { + "epoch": 1.47, + "learning_rate": 1.08533954003137e-05, + "loss": 1.163, + "step": 17410 + }, + { + "epoch": 1.47, + "learning_rate": 1.0852489739739785e-05, + "loss": 1.2531, + "step": 17411 + }, + { + "epoch": 1.47, + "learning_rate": 1.085158407212233e-05, + "loss": 1.1882, + "step": 17412 + }, + { + "epoch": 1.47, + "learning_rate": 1.085067839746881e-05, + "loss": 1.2043, + "step": 17413 + }, + { + "epoch": 1.47, + "learning_rate": 1.0849772715786716e-05, + "loss": 1.1736, + "step": 17414 + }, + { + "epoch": 1.47, + "learning_rate": 1.0848867027083528e-05, + "loss": 1.2755, + "step": 17415 + }, + { + "epoch": 1.47, + "learning_rate": 1.0847961331366728e-05, + "loss": 1.1642, + "step": 17416 + }, + { + "epoch": 1.47, + "learning_rate": 1.0847055628643801e-05, + "loss": 1.266, + "step": 17417 + }, + { + "epoch": 1.47, + "learning_rate": 1.0846149918922226e-05, + "loss": 1.174, + "step": 17418 + }, + { + "epoch": 1.47, + "learning_rate": 1.0845244202209495e-05, + "loss": 1.272, + "step": 17419 + }, + { + "epoch": 1.47, + "learning_rate": 1.084433847851308e-05, + "loss": 1.1704, + "step": 17420 + }, + { + "epoch": 1.47, + "learning_rate": 1.0843432747840472e-05, + "loss": 1.1288, + "step": 17421 + }, + { + "epoch": 1.47, + "learning_rate": 1.0842527010199158e-05, + "loss": 1.1891, + "step": 17422 + }, + { + "epoch": 1.47, + "learning_rate": 1.0841621265596612e-05, + "loss": 1.1463, + "step": 17423 + }, + { + "epoch": 1.47, + "learning_rate": 1.084071551404032e-05, + "loss": 1.2379, + "step": 17424 + }, + { + "epoch": 1.47, + "learning_rate": 1.0839809755537768e-05, + "loss": 1.1905, + "step": 17425 + }, + { + "epoch": 1.47, + "learning_rate": 1.0838903990096443e-05, + "loss": 1.1925, + "step": 17426 + }, + { + "epoch": 1.47, + "learning_rate": 1.0837998217723822e-05, + "loss": 1.053, + "step": 17427 + }, + { + "epoch": 1.47, + "learning_rate": 1.083709243842739e-05, + "loss": 1.2354, + "step": 17428 + }, + { + "epoch": 1.47, + "learning_rate": 1.0836186652214634e-05, + "loss": 1.1017, + "step": 17429 + }, + { + "epoch": 1.47, + "learning_rate": 1.0835280859093038e-05, + "loss": 1.1706, + "step": 17430 + }, + { + "epoch": 1.47, + "learning_rate": 1.0834375059070081e-05, + "loss": 1.3614, + "step": 17431 + }, + { + "epoch": 1.47, + "learning_rate": 1.083346925215325e-05, + "loss": 1.2152, + "step": 17432 + }, + { + "epoch": 1.47, + "learning_rate": 1.0832563438350032e-05, + "loss": 1.1535, + "step": 17433 + }, + { + "epoch": 1.47, + "learning_rate": 1.0831657617667908e-05, + "loss": 1.1754, + "step": 17434 + }, + { + "epoch": 1.47, + "learning_rate": 1.083075179011436e-05, + "loss": 1.2225, + "step": 17435 + }, + { + "epoch": 1.47, + "learning_rate": 1.0829845955696875e-05, + "loss": 1.2033, + "step": 17436 + }, + { + "epoch": 1.47, + "learning_rate": 1.0828940114422943e-05, + "loss": 1.3407, + "step": 17437 + }, + { + "epoch": 1.47, + "learning_rate": 1.0828034266300038e-05, + "loss": 1.1576, + "step": 17438 + }, + { + "epoch": 1.47, + "learning_rate": 1.0827128411335649e-05, + "loss": 1.173, + "step": 17439 + }, + { + "epoch": 1.47, + "learning_rate": 1.082622254953726e-05, + "loss": 1.2243, + "step": 17440 + }, + { + "epoch": 1.47, + "learning_rate": 1.082531668091236e-05, + "loss": 1.211, + "step": 17441 + }, + { + "epoch": 1.47, + "learning_rate": 1.0824410805468424e-05, + "loss": 1.1465, + "step": 17442 + }, + { + "epoch": 1.47, + "learning_rate": 1.0823504923212944e-05, + "loss": 1.2551, + "step": 17443 + }, + { + "epoch": 1.47, + "learning_rate": 1.0822599034153403e-05, + "loss": 1.2974, + "step": 17444 + }, + { + "epoch": 1.47, + "learning_rate": 1.0821693138297283e-05, + "loss": 1.1643, + "step": 17445 + }, + { + "epoch": 1.47, + "learning_rate": 1.0820787235652073e-05, + "loss": 1.0896, + "step": 17446 + }, + { + "epoch": 1.47, + "learning_rate": 1.0819881326225259e-05, + "loss": 1.1393, + "step": 17447 + }, + { + "epoch": 1.47, + "learning_rate": 1.0818975410024319e-05, + "loss": 1.1294, + "step": 17448 + }, + { + "epoch": 1.47, + "learning_rate": 1.0818069487056744e-05, + "loss": 1.2289, + "step": 17449 + }, + { + "epoch": 1.47, + "learning_rate": 1.0817163557330018e-05, + "loss": 1.1751, + "step": 17450 + }, + { + "epoch": 1.47, + "learning_rate": 1.0816257620851624e-05, + "loss": 1.1201, + "step": 17451 + }, + { + "epoch": 1.47, + "learning_rate": 1.0815351677629049e-05, + "loss": 1.2262, + "step": 17452 + }, + { + "epoch": 1.47, + "learning_rate": 1.0814445727669778e-05, + "loss": 1.2003, + "step": 17453 + }, + { + "epoch": 1.47, + "learning_rate": 1.0813539770981294e-05, + "loss": 1.1768, + "step": 17454 + }, + { + "epoch": 1.47, + "learning_rate": 1.0812633807571084e-05, + "loss": 1.1668, + "step": 17455 + }, + { + "epoch": 1.47, + "learning_rate": 1.0811727837446636e-05, + "loss": 1.2613, + "step": 17456 + }, + { + "epoch": 1.47, + "learning_rate": 1.0810821860615428e-05, + "loss": 1.1856, + "step": 17457 + }, + { + "epoch": 1.47, + "learning_rate": 1.0809915877084955e-05, + "loss": 1.1461, + "step": 17458 + }, + { + "epoch": 1.47, + "learning_rate": 1.0809009886862695e-05, + "loss": 1.1728, + "step": 17459 + }, + { + "epoch": 1.47, + "learning_rate": 1.0808103889956136e-05, + "loss": 1.155, + "step": 17460 + }, + { + "epoch": 1.47, + "learning_rate": 1.0807197886372766e-05, + "loss": 1.2519, + "step": 17461 + }, + { + "epoch": 1.47, + "learning_rate": 1.080629187612007e-05, + "loss": 1.1845, + "step": 17462 + }, + { + "epoch": 1.47, + "learning_rate": 1.0805385859205531e-05, + "loss": 1.1476, + "step": 17463 + }, + { + "epoch": 1.47, + "learning_rate": 1.0804479835636634e-05, + "loss": 1.2678, + "step": 17464 + }, + { + "epoch": 1.47, + "learning_rate": 1.080357380542087e-05, + "loss": 1.1733, + "step": 17465 + }, + { + "epoch": 1.47, + "learning_rate": 1.0802667768565723e-05, + "loss": 1.167, + "step": 17466 + }, + { + "epoch": 1.47, + "learning_rate": 1.0801761725078677e-05, + "loss": 1.26, + "step": 17467 + }, + { + "epoch": 1.47, + "learning_rate": 1.0800855674967217e-05, + "loss": 1.105, + "step": 17468 + }, + { + "epoch": 1.47, + "learning_rate": 1.0799949618238834e-05, + "loss": 1.2699, + "step": 17469 + }, + { + "epoch": 1.47, + "learning_rate": 1.0799043554901008e-05, + "loss": 1.2483, + "step": 17470 + }, + { + "epoch": 1.47, + "learning_rate": 1.0798137484961228e-05, + "loss": 1.261, + "step": 17471 + }, + { + "epoch": 1.47, + "learning_rate": 1.0797231408426985e-05, + "loss": 1.1975, + "step": 17472 + }, + { + "epoch": 1.47, + "learning_rate": 1.0796325325305758e-05, + "loss": 1.1954, + "step": 17473 + }, + { + "epoch": 1.47, + "learning_rate": 1.0795419235605035e-05, + "loss": 1.2326, + "step": 17474 + }, + { + "epoch": 1.47, + "learning_rate": 1.0794513139332305e-05, + "loss": 1.1914, + "step": 17475 + }, + { + "epoch": 1.47, + "learning_rate": 1.0793607036495057e-05, + "loss": 1.1632, + "step": 17476 + }, + { + "epoch": 1.47, + "learning_rate": 1.079270092710077e-05, + "loss": 1.1017, + "step": 17477 + }, + { + "epoch": 1.47, + "learning_rate": 1.0791794811156935e-05, + "loss": 1.1886, + "step": 17478 + }, + { + "epoch": 1.47, + "learning_rate": 1.0790888688671038e-05, + "loss": 1.2034, + "step": 17479 + }, + { + "epoch": 1.47, + "learning_rate": 1.0789982559650564e-05, + "loss": 1.2261, + "step": 17480 + }, + { + "epoch": 1.47, + "learning_rate": 1.0789076424103002e-05, + "loss": 1.2479, + "step": 17481 + }, + { + "epoch": 1.47, + "learning_rate": 1.0788170282035836e-05, + "loss": 1.1463, + "step": 17482 + }, + { + "epoch": 1.47, + "learning_rate": 1.078726413345656e-05, + "loss": 1.1478, + "step": 17483 + }, + { + "epoch": 1.47, + "learning_rate": 1.0786357978372653e-05, + "loss": 1.1635, + "step": 17484 + }, + { + "epoch": 1.47, + "learning_rate": 1.0785451816791604e-05, + "loss": 1.1827, + "step": 17485 + }, + { + "epoch": 1.47, + "learning_rate": 1.07845456487209e-05, + "loss": 1.236, + "step": 17486 + }, + { + "epoch": 1.47, + "learning_rate": 1.0783639474168032e-05, + "loss": 1.1992, + "step": 17487 + }, + { + "epoch": 1.47, + "learning_rate": 1.0782733293140484e-05, + "loss": 1.1708, + "step": 17488 + }, + { + "epoch": 1.47, + "learning_rate": 1.0781827105645738e-05, + "loss": 1.1459, + "step": 17489 + }, + { + "epoch": 1.47, + "learning_rate": 1.078092091169129e-05, + "loss": 1.2304, + "step": 17490 + }, + { + "epoch": 1.47, + "learning_rate": 1.0780014711284625e-05, + "loss": 1.1897, + "step": 17491 + }, + { + "epoch": 1.47, + "learning_rate": 1.0779108504433226e-05, + "loss": 1.1557, + "step": 17492 + }, + { + "epoch": 1.47, + "learning_rate": 1.0778202291144584e-05, + "loss": 1.259, + "step": 17493 + }, + { + "epoch": 1.47, + "learning_rate": 1.0777296071426184e-05, + "loss": 1.2183, + "step": 17494 + }, + { + "epoch": 1.47, + "learning_rate": 1.0776389845285517e-05, + "loss": 1.2358, + "step": 17495 + }, + { + "epoch": 1.47, + "learning_rate": 1.0775483612730069e-05, + "loss": 1.2305, + "step": 17496 + }, + { + "epoch": 1.47, + "learning_rate": 1.0774577373767326e-05, + "loss": 1.2061, + "step": 17497 + }, + { + "epoch": 1.47, + "learning_rate": 1.0773671128404776e-05, + "loss": 1.0937, + "step": 17498 + }, + { + "epoch": 1.47, + "learning_rate": 1.0772764876649911e-05, + "loss": 1.193, + "step": 17499 + }, + { + "epoch": 1.47, + "learning_rate": 1.0771858618510212e-05, + "loss": 1.2016, + "step": 17500 + }, + { + "epoch": 1.47, + "learning_rate": 1.0770952353993174e-05, + "loss": 1.0995, + "step": 17501 + }, + { + "epoch": 1.47, + "learning_rate": 1.0770046083106276e-05, + "loss": 1.2722, + "step": 17502 + }, + { + "epoch": 1.47, + "learning_rate": 1.0769139805857012e-05, + "loss": 1.221, + "step": 17503 + }, + { + "epoch": 1.47, + "learning_rate": 1.0768233522252868e-05, + "loss": 1.1739, + "step": 17504 + }, + { + "epoch": 1.47, + "learning_rate": 1.0767327232301339e-05, + "loss": 1.1345, + "step": 17505 + }, + { + "epoch": 1.47, + "learning_rate": 1.07664209360099e-05, + "loss": 1.2096, + "step": 17506 + }, + { + "epoch": 1.47, + "learning_rate": 1.0765514633386048e-05, + "loss": 1.1429, + "step": 17507 + }, + { + "epoch": 1.47, + "learning_rate": 1.0764608324437268e-05, + "loss": 1.1845, + "step": 17508 + }, + { + "epoch": 1.47, + "learning_rate": 1.0763702009171051e-05, + "loss": 1.1387, + "step": 17509 + }, + { + "epoch": 1.47, + "learning_rate": 1.0762795687594883e-05, + "loss": 1.1471, + "step": 17510 + }, + { + "epoch": 1.47, + "learning_rate": 1.0761889359716254e-05, + "loss": 1.1621, + "step": 17511 + }, + { + "epoch": 1.47, + "learning_rate": 1.076098302554265e-05, + "loss": 1.2056, + "step": 17512 + }, + { + "epoch": 1.47, + "learning_rate": 1.0760076685081562e-05, + "loss": 1.1585, + "step": 17513 + }, + { + "epoch": 1.47, + "learning_rate": 1.0759170338340477e-05, + "loss": 1.0766, + "step": 17514 + }, + { + "epoch": 1.47, + "learning_rate": 1.0758263985326882e-05, + "loss": 1.0978, + "step": 17515 + }, + { + "epoch": 1.47, + "learning_rate": 1.0757357626048269e-05, + "loss": 1.1364, + "step": 17516 + }, + { + "epoch": 1.47, + "learning_rate": 1.0756451260512124e-05, + "loss": 1.2681, + "step": 17517 + }, + { + "epoch": 1.47, + "learning_rate": 1.0755544888725934e-05, + "loss": 1.0455, + "step": 17518 + }, + { + "epoch": 1.47, + "learning_rate": 1.0754638510697195e-05, + "loss": 1.1752, + "step": 17519 + }, + { + "epoch": 1.48, + "learning_rate": 1.0753732126433389e-05, + "loss": 1.1323, + "step": 17520 + }, + { + "epoch": 1.48, + "learning_rate": 1.0752825735942004e-05, + "loss": 1.1391, + "step": 17521 + }, + { + "epoch": 1.48, + "learning_rate": 1.0751919339230537e-05, + "loss": 1.194, + "step": 17522 + }, + { + "epoch": 1.48, + "learning_rate": 1.0751012936306471e-05, + "loss": 1.1721, + "step": 17523 + }, + { + "epoch": 1.48, + "learning_rate": 1.0750106527177293e-05, + "loss": 1.246, + "step": 17524 + }, + { + "epoch": 1.48, + "learning_rate": 1.0749200111850498e-05, + "loss": 1.157, + "step": 17525 + }, + { + "epoch": 1.48, + "learning_rate": 1.074829369033357e-05, + "loss": 1.206, + "step": 17526 + }, + { + "epoch": 1.48, + "learning_rate": 1.0747387262634001e-05, + "loss": 1.2144, + "step": 17527 + }, + { + "epoch": 1.48, + "learning_rate": 1.0746480828759278e-05, + "loss": 1.1785, + "step": 17528 + }, + { + "epoch": 1.48, + "learning_rate": 1.074557438871689e-05, + "loss": 1.0883, + "step": 17529 + }, + { + "epoch": 1.48, + "learning_rate": 1.0744667942514333e-05, + "loss": 1.0901, + "step": 17530 + }, + { + "epoch": 1.48, + "learning_rate": 1.0743761490159088e-05, + "loss": 1.1314, + "step": 17531 + }, + { + "epoch": 1.48, + "learning_rate": 1.0742855031658643e-05, + "loss": 1.1709, + "step": 17532 + }, + { + "epoch": 1.48, + "learning_rate": 1.07419485670205e-05, + "loss": 1.2504, + "step": 17533 + }, + { + "epoch": 1.48, + "learning_rate": 1.0741042096252134e-05, + "loss": 1.2047, + "step": 17534 + }, + { + "epoch": 1.48, + "learning_rate": 1.0740135619361045e-05, + "loss": 1.2178, + "step": 17535 + }, + { + "epoch": 1.48, + "learning_rate": 1.0739229136354716e-05, + "loss": 1.2534, + "step": 17536 + }, + { + "epoch": 1.48, + "learning_rate": 1.0738322647240644e-05, + "loss": 1.1959, + "step": 17537 + }, + { + "epoch": 1.48, + "learning_rate": 1.073741615202631e-05, + "loss": 1.1285, + "step": 17538 + }, + { + "epoch": 1.48, + "learning_rate": 1.0736509650719209e-05, + "loss": 1.2305, + "step": 17539 + }, + { + "epoch": 1.48, + "learning_rate": 1.0735603143326827e-05, + "loss": 1.1671, + "step": 17540 + }, + { + "epoch": 1.48, + "learning_rate": 1.0734696629856662e-05, + "loss": 1.1498, + "step": 17541 + }, + { + "epoch": 1.48, + "learning_rate": 1.0733790110316194e-05, + "loss": 1.1766, + "step": 17542 + }, + { + "epoch": 1.48, + "learning_rate": 1.0732883584712917e-05, + "loss": 1.1695, + "step": 17543 + }, + { + "epoch": 1.48, + "learning_rate": 1.0731977053054324e-05, + "loss": 1.2388, + "step": 17544 + }, + { + "epoch": 1.48, + "learning_rate": 1.0731070515347899e-05, + "loss": 1.2164, + "step": 17545 + }, + { + "epoch": 1.48, + "learning_rate": 1.0730163971601138e-05, + "loss": 1.1586, + "step": 17546 + }, + { + "epoch": 1.48, + "learning_rate": 1.0729257421821527e-05, + "loss": 1.1915, + "step": 17547 + }, + { + "epoch": 1.48, + "learning_rate": 1.0728350866016559e-05, + "loss": 1.2054, + "step": 17548 + }, + { + "epoch": 1.48, + "learning_rate": 1.0727444304193721e-05, + "loss": 1.1649, + "step": 17549 + }, + { + "epoch": 1.48, + "learning_rate": 1.0726537736360508e-05, + "loss": 1.1527, + "step": 17550 + }, + { + "epoch": 1.48, + "learning_rate": 1.0725631162524409e-05, + "loss": 1.2098, + "step": 17551 + }, + { + "epoch": 1.48, + "learning_rate": 1.0724724582692909e-05, + "loss": 1.0958, + "step": 17552 + }, + { + "epoch": 1.48, + "learning_rate": 1.0723817996873505e-05, + "loss": 1.1538, + "step": 17553 + }, + { + "epoch": 1.48, + "learning_rate": 1.0722911405073683e-05, + "loss": 1.2302, + "step": 17554 + }, + { + "epoch": 1.48, + "learning_rate": 1.0722004807300938e-05, + "loss": 1.2738, + "step": 17555 + }, + { + "epoch": 1.48, + "learning_rate": 1.0721098203562757e-05, + "loss": 1.1483, + "step": 17556 + }, + { + "epoch": 1.48, + "learning_rate": 1.0720191593866629e-05, + "loss": 1.1588, + "step": 17557 + }, + { + "epoch": 1.48, + "learning_rate": 1.0719284978220052e-05, + "loss": 1.1706, + "step": 17558 + }, + { + "epoch": 1.48, + "learning_rate": 1.0718378356630512e-05, + "loss": 1.2537, + "step": 17559 + }, + { + "epoch": 1.48, + "learning_rate": 1.0717471729105498e-05, + "loss": 1.115, + "step": 17560 + }, + { + "epoch": 1.48, + "learning_rate": 1.0716565095652503e-05, + "loss": 1.1116, + "step": 17561 + }, + { + "epoch": 1.48, + "learning_rate": 1.0715658456279022e-05, + "loss": 1.2103, + "step": 17562 + }, + { + "epoch": 1.48, + "learning_rate": 1.0714751810992536e-05, + "loss": 1.1822, + "step": 17563 + }, + { + "epoch": 1.48, + "learning_rate": 1.0713845159800545e-05, + "loss": 1.2182, + "step": 17564 + }, + { + "epoch": 1.48, + "learning_rate": 1.0712938502710535e-05, + "loss": 1.2291, + "step": 17565 + }, + { + "epoch": 1.48, + "learning_rate": 1.0712031839730004e-05, + "loss": 1.1995, + "step": 17566 + }, + { + "epoch": 1.48, + "learning_rate": 1.0711125170866432e-05, + "loss": 1.0996, + "step": 17567 + }, + { + "epoch": 1.48, + "learning_rate": 1.0710218496127318e-05, + "loss": 1.2205, + "step": 17568 + }, + { + "epoch": 1.48, + "learning_rate": 1.0709311815520151e-05, + "loss": 1.149, + "step": 17569 + }, + { + "epoch": 1.48, + "learning_rate": 1.0708405129052423e-05, + "loss": 1.2512, + "step": 17570 + }, + { + "epoch": 1.48, + "learning_rate": 1.0707498436731627e-05, + "loss": 1.2079, + "step": 17571 + }, + { + "epoch": 1.48, + "learning_rate": 1.0706591738565251e-05, + "loss": 1.1795, + "step": 17572 + }, + { + "epoch": 1.48, + "learning_rate": 1.070568503456079e-05, + "loss": 1.1711, + "step": 17573 + }, + { + "epoch": 1.48, + "learning_rate": 1.0704778324725731e-05, + "loss": 1.1218, + "step": 17574 + }, + { + "epoch": 1.48, + "learning_rate": 1.0703871609067567e-05, + "loss": 1.1168, + "step": 17575 + }, + { + "epoch": 1.48, + "learning_rate": 1.0702964887593795e-05, + "loss": 1.264, + "step": 17576 + }, + { + "epoch": 1.48, + "learning_rate": 1.0702058160311899e-05, + "loss": 1.1987, + "step": 17577 + }, + { + "epoch": 1.48, + "learning_rate": 1.0701151427229373e-05, + "loss": 1.134, + "step": 17578 + }, + { + "epoch": 1.48, + "learning_rate": 1.070024468835371e-05, + "loss": 1.196, + "step": 17579 + }, + { + "epoch": 1.48, + "learning_rate": 1.0699337943692402e-05, + "loss": 1.2593, + "step": 17580 + }, + { + "epoch": 1.48, + "learning_rate": 1.0698431193252941e-05, + "loss": 1.2031, + "step": 17581 + }, + { + "epoch": 1.48, + "learning_rate": 1.0697524437042815e-05, + "loss": 1.1896, + "step": 17582 + }, + { + "epoch": 1.48, + "learning_rate": 1.0696617675069521e-05, + "loss": 1.2687, + "step": 17583 + }, + { + "epoch": 1.48, + "learning_rate": 1.0695710907340549e-05, + "loss": 1.1659, + "step": 17584 + }, + { + "epoch": 1.48, + "learning_rate": 1.0694804133863388e-05, + "loss": 1.2169, + "step": 17585 + }, + { + "epoch": 1.48, + "learning_rate": 1.0693897354645535e-05, + "loss": 1.3004, + "step": 17586 + }, + { + "epoch": 1.48, + "learning_rate": 1.0692990569694484e-05, + "loss": 1.1772, + "step": 17587 + }, + { + "epoch": 1.48, + "learning_rate": 1.0692083779017717e-05, + "loss": 1.1678, + "step": 17588 + }, + { + "epoch": 1.48, + "learning_rate": 1.0691176982622735e-05, + "loss": 1.2031, + "step": 17589 + }, + { + "epoch": 1.48, + "learning_rate": 1.0690270180517025e-05, + "loss": 1.2798, + "step": 17590 + }, + { + "epoch": 1.48, + "learning_rate": 1.0689363372708087e-05, + "loss": 1.1432, + "step": 17591 + }, + { + "epoch": 1.48, + "learning_rate": 1.0688456559203404e-05, + "loss": 1.1489, + "step": 17592 + }, + { + "epoch": 1.48, + "learning_rate": 1.0687549740010475e-05, + "loss": 1.2156, + "step": 17593 + }, + { + "epoch": 1.48, + "learning_rate": 1.0686642915136787e-05, + "loss": 1.0972, + "step": 17594 + }, + { + "epoch": 1.48, + "learning_rate": 1.0685736084589837e-05, + "loss": 1.0954, + "step": 17595 + }, + { + "epoch": 1.48, + "learning_rate": 1.0684829248377115e-05, + "loss": 1.1396, + "step": 17596 + }, + { + "epoch": 1.48, + "learning_rate": 1.0683922406506117e-05, + "loss": 1.2596, + "step": 17597 + }, + { + "epoch": 1.48, + "learning_rate": 1.0683015558984334e-05, + "loss": 1.2106, + "step": 17598 + }, + { + "epoch": 1.48, + "learning_rate": 1.0682108705819254e-05, + "loss": 1.1626, + "step": 17599 + }, + { + "epoch": 1.48, + "learning_rate": 1.0681201847018376e-05, + "loss": 1.2824, + "step": 17600 + }, + { + "epoch": 1.48, + "learning_rate": 1.0680294982589192e-05, + "loss": 1.3, + "step": 17601 + }, + { + "epoch": 1.48, + "learning_rate": 1.0679388112539191e-05, + "loss": 1.2368, + "step": 17602 + }, + { + "epoch": 1.48, + "learning_rate": 1.0678481236875866e-05, + "loss": 1.1593, + "step": 17603 + }, + { + "epoch": 1.48, + "learning_rate": 1.0677574355606715e-05, + "loss": 1.1716, + "step": 17604 + }, + { + "epoch": 1.48, + "learning_rate": 1.0676667468739228e-05, + "loss": 1.193, + "step": 17605 + }, + { + "epoch": 1.48, + "learning_rate": 1.0675760576280896e-05, + "loss": 1.1458, + "step": 17606 + }, + { + "epoch": 1.48, + "learning_rate": 1.067485367823921e-05, + "loss": 1.1582, + "step": 17607 + }, + { + "epoch": 1.48, + "learning_rate": 1.0673946774621673e-05, + "loss": 1.2803, + "step": 17608 + }, + { + "epoch": 1.48, + "learning_rate": 1.067303986543577e-05, + "loss": 1.1969, + "step": 17609 + }, + { + "epoch": 1.48, + "learning_rate": 1.0672132950688995e-05, + "loss": 1.2339, + "step": 17610 + }, + { + "epoch": 1.48, + "learning_rate": 1.0671226030388847e-05, + "loss": 1.2514, + "step": 17611 + }, + { + "epoch": 1.48, + "learning_rate": 1.0670319104542813e-05, + "loss": 1.1387, + "step": 17612 + }, + { + "epoch": 1.48, + "learning_rate": 1.0669412173158384e-05, + "loss": 1.1361, + "step": 17613 + }, + { + "epoch": 1.48, + "learning_rate": 1.066850523624306e-05, + "loss": 1.1799, + "step": 17614 + }, + { + "epoch": 1.48, + "learning_rate": 1.066759829380433e-05, + "loss": 1.274, + "step": 17615 + }, + { + "epoch": 1.48, + "learning_rate": 1.0666691345849696e-05, + "loss": 1.177, + "step": 17616 + }, + { + "epoch": 1.48, + "learning_rate": 1.0665784392386637e-05, + "loss": 1.2334, + "step": 17617 + }, + { + "epoch": 1.48, + "learning_rate": 1.0664877433422658e-05, + "loss": 1.156, + "step": 17618 + }, + { + "epoch": 1.48, + "learning_rate": 1.0663970468965247e-05, + "loss": 1.1244, + "step": 17619 + }, + { + "epoch": 1.48, + "learning_rate": 1.06630634990219e-05, + "loss": 1.168, + "step": 17620 + }, + { + "epoch": 1.48, + "learning_rate": 1.0662156523600112e-05, + "loss": 1.13, + "step": 17621 + }, + { + "epoch": 1.48, + "learning_rate": 1.066124954270737e-05, + "loss": 1.1439, + "step": 17622 + }, + { + "epoch": 1.48, + "learning_rate": 1.0660342556351179e-05, + "loss": 1.222, + "step": 17623 + }, + { + "epoch": 1.48, + "learning_rate": 1.0659435564539022e-05, + "loss": 1.2215, + "step": 17624 + }, + { + "epoch": 1.48, + "learning_rate": 1.0658528567278398e-05, + "loss": 1.099, + "step": 17625 + }, + { + "epoch": 1.48, + "learning_rate": 1.0657621564576806e-05, + "loss": 1.1731, + "step": 17626 + }, + { + "epoch": 1.48, + "learning_rate": 1.0656714556441728e-05, + "loss": 1.2022, + "step": 17627 + }, + { + "epoch": 1.48, + "learning_rate": 1.0655807542880665e-05, + "loss": 1.1916, + "step": 17628 + }, + { + "epoch": 1.48, + "learning_rate": 1.065490052390111e-05, + "loss": 1.2203, + "step": 17629 + }, + { + "epoch": 1.48, + "learning_rate": 1.065399349951056e-05, + "loss": 1.192, + "step": 17630 + }, + { + "epoch": 1.48, + "learning_rate": 1.0653086469716503e-05, + "loss": 1.2486, + "step": 17631 + }, + { + "epoch": 1.48, + "learning_rate": 1.0652179434526435e-05, + "loss": 1.1224, + "step": 17632 + }, + { + "epoch": 1.48, + "learning_rate": 1.0651272393947859e-05, + "loss": 1.1819, + "step": 17633 + }, + { + "epoch": 1.48, + "learning_rate": 1.0650365347988257e-05, + "loss": 1.2412, + "step": 17634 + }, + { + "epoch": 1.48, + "learning_rate": 1.0649458296655126e-05, + "loss": 1.1493, + "step": 17635 + }, + { + "epoch": 1.48, + "learning_rate": 1.0648551239955966e-05, + "loss": 1.2261, + "step": 17636 + }, + { + "epoch": 1.48, + "learning_rate": 1.064764417789827e-05, + "loss": 1.1931, + "step": 17637 + }, + { + "epoch": 1.48, + "learning_rate": 1.064673711048953e-05, + "loss": 1.2179, + "step": 17638 + }, + { + "epoch": 1.49, + "learning_rate": 1.0645830037737237e-05, + "loss": 1.2153, + "step": 17639 + }, + { + "epoch": 1.49, + "learning_rate": 1.0644922959648892e-05, + "loss": 1.2055, + "step": 17640 + }, + { + "epoch": 1.49, + "learning_rate": 1.0644015876231988e-05, + "loss": 1.2024, + "step": 17641 + }, + { + "epoch": 1.49, + "learning_rate": 1.0643108787494019e-05, + "loss": 1.1549, + "step": 17642 + }, + { + "epoch": 1.49, + "learning_rate": 1.0642201693442478e-05, + "loss": 1.1707, + "step": 17643 + }, + { + "epoch": 1.49, + "learning_rate": 1.0641294594084858e-05, + "loss": 1.2871, + "step": 17644 + }, + { + "epoch": 1.49, + "learning_rate": 1.0640387489428661e-05, + "loss": 1.0897, + "step": 17645 + }, + { + "epoch": 1.49, + "learning_rate": 1.0639480379481378e-05, + "loss": 1.2495, + "step": 17646 + }, + { + "epoch": 1.49, + "learning_rate": 1.06385732642505e-05, + "loss": 1.1971, + "step": 17647 + }, + { + "epoch": 1.49, + "learning_rate": 1.0637666143743531e-05, + "loss": 1.2364, + "step": 17648 + }, + { + "epoch": 1.49, + "learning_rate": 1.0636759017967955e-05, + "loss": 1.1885, + "step": 17649 + }, + { + "epoch": 1.49, + "learning_rate": 1.0635851886931272e-05, + "loss": 1.1625, + "step": 17650 + }, + { + "epoch": 1.49, + "learning_rate": 1.0634944750640982e-05, + "loss": 1.1654, + "step": 17651 + }, + { + "epoch": 1.49, + "learning_rate": 1.0634037609104574e-05, + "loss": 1.1229, + "step": 17652 + }, + { + "epoch": 1.49, + "learning_rate": 1.063313046232954e-05, + "loss": 1.1121, + "step": 17653 + }, + { + "epoch": 1.49, + "learning_rate": 1.0632223310323382e-05, + "loss": 1.1513, + "step": 17654 + }, + { + "epoch": 1.49, + "learning_rate": 1.0631316153093594e-05, + "loss": 1.1955, + "step": 17655 + }, + { + "epoch": 1.49, + "learning_rate": 1.0630408990647669e-05, + "loss": 1.1801, + "step": 17656 + }, + { + "epoch": 1.49, + "learning_rate": 1.06295018229931e-05, + "loss": 1.2147, + "step": 17657 + }, + { + "epoch": 1.49, + "learning_rate": 1.0628594650137391e-05, + "loss": 1.2203, + "step": 17658 + }, + { + "epoch": 1.49, + "learning_rate": 1.0627687472088029e-05, + "loss": 1.3578, + "step": 17659 + }, + { + "epoch": 1.49, + "learning_rate": 1.0626780288852513e-05, + "loss": 1.1254, + "step": 17660 + }, + { + "epoch": 1.49, + "learning_rate": 1.0625873100438335e-05, + "loss": 1.253, + "step": 17661 + }, + { + "epoch": 1.49, + "learning_rate": 1.0624965906852997e-05, + "loss": 1.1789, + "step": 17662 + }, + { + "epoch": 1.49, + "learning_rate": 1.0624058708103988e-05, + "loss": 1.2443, + "step": 17663 + }, + { + "epoch": 1.49, + "learning_rate": 1.0623151504198808e-05, + "loss": 1.1609, + "step": 17664 + }, + { + "epoch": 1.49, + "learning_rate": 1.0622244295144948e-05, + "loss": 1.2234, + "step": 17665 + }, + { + "epoch": 1.49, + "learning_rate": 1.0621337080949911e-05, + "loss": 1.1683, + "step": 17666 + }, + { + "epoch": 1.49, + "learning_rate": 1.0620429861621186e-05, + "loss": 1.2504, + "step": 17667 + }, + { + "epoch": 1.49, + "learning_rate": 1.0619522637166271e-05, + "loss": 1.1513, + "step": 17668 + }, + { + "epoch": 1.49, + "learning_rate": 1.0618615407592659e-05, + "loss": 1.1707, + "step": 17669 + }, + { + "epoch": 1.49, + "learning_rate": 1.061770817290785e-05, + "loss": 1.207, + "step": 17670 + }, + { + "epoch": 1.49, + "learning_rate": 1.061680093311934e-05, + "loss": 1.2355, + "step": 17671 + }, + { + "epoch": 1.49, + "learning_rate": 1.0615893688234622e-05, + "loss": 1.172, + "step": 17672 + }, + { + "epoch": 1.49, + "learning_rate": 1.0614986438261194e-05, + "loss": 1.1546, + "step": 17673 + }, + { + "epoch": 1.49, + "learning_rate": 1.0614079183206552e-05, + "loss": 1.2742, + "step": 17674 + }, + { + "epoch": 1.49, + "learning_rate": 1.0613171923078189e-05, + "loss": 1.2737, + "step": 17675 + }, + { + "epoch": 1.49, + "learning_rate": 1.0612264657883604e-05, + "loss": 1.1771, + "step": 17676 + }, + { + "epoch": 1.49, + "learning_rate": 1.0611357387630295e-05, + "loss": 1.1895, + "step": 17677 + }, + { + "epoch": 1.49, + "learning_rate": 1.0610450112325753e-05, + "loss": 1.2896, + "step": 17678 + }, + { + "epoch": 1.49, + "learning_rate": 1.0609542831977478e-05, + "loss": 1.1912, + "step": 17679 + }, + { + "epoch": 1.49, + "learning_rate": 1.0608635546592964e-05, + "loss": 1.2586, + "step": 17680 + }, + { + "epoch": 1.49, + "learning_rate": 1.0607728256179711e-05, + "loss": 1.1618, + "step": 17681 + }, + { + "epoch": 1.49, + "learning_rate": 1.0606820960745207e-05, + "loss": 1.2789, + "step": 17682 + }, + { + "epoch": 1.49, + "learning_rate": 1.0605913660296958e-05, + "loss": 1.2375, + "step": 17683 + }, + { + "epoch": 1.49, + "learning_rate": 1.0605006354842455e-05, + "loss": 1.1518, + "step": 17684 + }, + { + "epoch": 1.49, + "learning_rate": 1.0604099044389198e-05, + "loss": 1.2275, + "step": 17685 + }, + { + "epoch": 1.49, + "learning_rate": 1.060319172894468e-05, + "loss": 1.2864, + "step": 17686 + }, + { + "epoch": 1.49, + "learning_rate": 1.0602284408516399e-05, + "loss": 1.1893, + "step": 17687 + }, + { + "epoch": 1.49, + "learning_rate": 1.0601377083111852e-05, + "loss": 1.0981, + "step": 17688 + }, + { + "epoch": 1.49, + "learning_rate": 1.0600469752738534e-05, + "loss": 1.2279, + "step": 17689 + }, + { + "epoch": 1.49, + "learning_rate": 1.0599562417403944e-05, + "loss": 1.2378, + "step": 17690 + }, + { + "epoch": 1.49, + "learning_rate": 1.0598655077115576e-05, + "loss": 1.0942, + "step": 17691 + }, + { + "epoch": 1.49, + "learning_rate": 1.059774773188093e-05, + "loss": 1.2091, + "step": 17692 + }, + { + "epoch": 1.49, + "learning_rate": 1.0596840381707498e-05, + "loss": 1.2364, + "step": 17693 + }, + { + "epoch": 1.49, + "learning_rate": 1.0595933026602782e-05, + "loss": 1.2413, + "step": 17694 + }, + { + "epoch": 1.49, + "learning_rate": 1.0595025666574275e-05, + "loss": 1.186, + "step": 17695 + }, + { + "epoch": 1.49, + "learning_rate": 1.0594118301629476e-05, + "loss": 1.136, + "step": 17696 + }, + { + "epoch": 1.49, + "learning_rate": 1.0593210931775881e-05, + "loss": 1.2217, + "step": 17697 + }, + { + "epoch": 1.49, + "learning_rate": 1.059230355702099e-05, + "loss": 1.2391, + "step": 17698 + }, + { + "epoch": 1.49, + "learning_rate": 1.0591396177372298e-05, + "loss": 1.2378, + "step": 17699 + }, + { + "epoch": 1.49, + "learning_rate": 1.05904887928373e-05, + "loss": 1.1557, + "step": 17700 + }, + { + "epoch": 1.49, + "learning_rate": 1.0589581403423493e-05, + "loss": 1.1208, + "step": 17701 + }, + { + "epoch": 1.49, + "learning_rate": 1.0588674009138377e-05, + "loss": 1.1786, + "step": 17702 + }, + { + "epoch": 1.49, + "learning_rate": 1.0587766609989448e-05, + "loss": 1.2438, + "step": 17703 + }, + { + "epoch": 1.49, + "learning_rate": 1.0586859205984201e-05, + "loss": 1.1044, + "step": 17704 + }, + { + "epoch": 1.49, + "learning_rate": 1.0585951797130141e-05, + "loss": 1.2241, + "step": 17705 + }, + { + "epoch": 1.49, + "learning_rate": 1.0585044383434754e-05, + "loss": 1.1565, + "step": 17706 + }, + { + "epoch": 1.49, + "learning_rate": 1.0584136964905543e-05, + "loss": 1.2228, + "step": 17707 + }, + { + "epoch": 1.49, + "learning_rate": 1.0583229541550008e-05, + "loss": 1.1878, + "step": 17708 + }, + { + "epoch": 1.49, + "learning_rate": 1.0582322113375645e-05, + "loss": 1.2768, + "step": 17709 + }, + { + "epoch": 1.49, + "learning_rate": 1.0581414680389947e-05, + "loss": 1.2784, + "step": 17710 + }, + { + "epoch": 1.49, + "learning_rate": 1.0580507242600418e-05, + "loss": 1.2746, + "step": 17711 + }, + { + "epoch": 1.49, + "learning_rate": 1.0579599800014553e-05, + "loss": 1.2542, + "step": 17712 + }, + { + "epoch": 1.49, + "learning_rate": 1.0578692352639846e-05, + "loss": 1.2409, + "step": 17713 + }, + { + "epoch": 1.49, + "learning_rate": 1.0577784900483798e-05, + "loss": 1.1812, + "step": 17714 + }, + { + "epoch": 1.49, + "learning_rate": 1.0576877443553906e-05, + "loss": 1.1672, + "step": 17715 + }, + { + "epoch": 1.49, + "learning_rate": 1.057596998185767e-05, + "loss": 1.2606, + "step": 17716 + }, + { + "epoch": 1.49, + "learning_rate": 1.0575062515402585e-05, + "loss": 1.2056, + "step": 17717 + }, + { + "epoch": 1.49, + "learning_rate": 1.0574155044196148e-05, + "loss": 1.1892, + "step": 17718 + }, + { + "epoch": 1.49, + "learning_rate": 1.0573247568245859e-05, + "loss": 1.1659, + "step": 17719 + }, + { + "epoch": 1.49, + "learning_rate": 1.0572340087559214e-05, + "loss": 1.1697, + "step": 17720 + }, + { + "epoch": 1.49, + "learning_rate": 1.0571432602143714e-05, + "loss": 1.1501, + "step": 17721 + }, + { + "epoch": 1.49, + "learning_rate": 1.0570525112006853e-05, + "loss": 1.1931, + "step": 17722 + }, + { + "epoch": 1.49, + "learning_rate": 1.0569617617156133e-05, + "loss": 1.1627, + "step": 17723 + }, + { + "epoch": 1.49, + "learning_rate": 1.056871011759905e-05, + "loss": 1.1773, + "step": 17724 + }, + { + "epoch": 1.49, + "learning_rate": 1.05678026133431e-05, + "loss": 1.1584, + "step": 17725 + }, + { + "epoch": 1.49, + "learning_rate": 1.0566895104395784e-05, + "loss": 1.1598, + "step": 17726 + }, + { + "epoch": 1.49, + "learning_rate": 1.05659875907646e-05, + "loss": 1.1826, + "step": 17727 + }, + { + "epoch": 1.49, + "learning_rate": 1.0565080072457045e-05, + "loss": 1.1614, + "step": 17728 + }, + { + "epoch": 1.49, + "learning_rate": 1.0564172549480618e-05, + "loss": 1.2275, + "step": 17729 + }, + { + "epoch": 1.49, + "learning_rate": 1.0563265021842817e-05, + "loss": 1.196, + "step": 17730 + }, + { + "epoch": 1.49, + "learning_rate": 1.0562357489551138e-05, + "loss": 1.2711, + "step": 17731 + }, + { + "epoch": 1.49, + "learning_rate": 1.0561449952613081e-05, + "loss": 1.2595, + "step": 17732 + }, + { + "epoch": 1.49, + "learning_rate": 1.0560542411036148e-05, + "loss": 1.1117, + "step": 17733 + }, + { + "epoch": 1.49, + "learning_rate": 1.0559634864827833e-05, + "loss": 1.2491, + "step": 17734 + }, + { + "epoch": 1.49, + "learning_rate": 1.0558727313995635e-05, + "loss": 1.1299, + "step": 17735 + }, + { + "epoch": 1.49, + "learning_rate": 1.0557819758547052e-05, + "loss": 1.1459, + "step": 17736 + }, + { + "epoch": 1.49, + "learning_rate": 1.0556912198489588e-05, + "loss": 1.1407, + "step": 17737 + }, + { + "epoch": 1.49, + "learning_rate": 1.0556004633830734e-05, + "loss": 1.2559, + "step": 17738 + }, + { + "epoch": 1.49, + "learning_rate": 1.0555097064577992e-05, + "loss": 1.1558, + "step": 17739 + }, + { + "epoch": 1.49, + "learning_rate": 1.055418949073886e-05, + "loss": 1.1678, + "step": 17740 + }, + { + "epoch": 1.49, + "learning_rate": 1.0553281912320838e-05, + "loss": 1.1982, + "step": 17741 + }, + { + "epoch": 1.49, + "learning_rate": 1.0552374329331422e-05, + "loss": 1.1926, + "step": 17742 + }, + { + "epoch": 1.49, + "learning_rate": 1.0551466741778115e-05, + "loss": 1.2586, + "step": 17743 + }, + { + "epoch": 1.49, + "learning_rate": 1.055055914966841e-05, + "loss": 1.1241, + "step": 17744 + }, + { + "epoch": 1.49, + "learning_rate": 1.0549651553009813e-05, + "loss": 1.2026, + "step": 17745 + }, + { + "epoch": 1.49, + "learning_rate": 1.0548743951809817e-05, + "loss": 1.1654, + "step": 17746 + }, + { + "epoch": 1.49, + "learning_rate": 1.0547836346075922e-05, + "loss": 1.2097, + "step": 17747 + }, + { + "epoch": 1.49, + "learning_rate": 1.0546928735815631e-05, + "loss": 1.225, + "step": 17748 + }, + { + "epoch": 1.49, + "learning_rate": 1.0546021121036435e-05, + "loss": 1.1694, + "step": 17749 + }, + { + "epoch": 1.49, + "learning_rate": 1.0545113501745844e-05, + "loss": 1.1752, + "step": 17750 + }, + { + "epoch": 1.49, + "learning_rate": 1.0544205877951346e-05, + "loss": 1.2646, + "step": 17751 + }, + { + "epoch": 1.49, + "learning_rate": 1.0543298249660445e-05, + "loss": 1.1334, + "step": 17752 + }, + { + "epoch": 1.49, + "learning_rate": 1.0542390616880641e-05, + "loss": 1.1903, + "step": 17753 + }, + { + "epoch": 1.49, + "learning_rate": 1.0541482979619433e-05, + "loss": 1.1473, + "step": 17754 + }, + { + "epoch": 1.49, + "learning_rate": 1.054057533788432e-05, + "loss": 1.0897, + "step": 17755 + }, + { + "epoch": 1.49, + "learning_rate": 1.0539667691682798e-05, + "loss": 1.1855, + "step": 17756 + }, + { + "epoch": 1.49, + "learning_rate": 1.0538760041022367e-05, + "loss": 1.2518, + "step": 17757 + }, + { + "epoch": 1.5, + "learning_rate": 1.0537852385910535e-05, + "loss": 1.1326, + "step": 17758 + }, + { + "epoch": 1.5, + "learning_rate": 1.0536944726354787e-05, + "loss": 1.1283, + "step": 17759 + }, + { + "epoch": 1.5, + "learning_rate": 1.0536037062362635e-05, + "loss": 1.1781, + "step": 17760 + }, + { + "epoch": 1.5, + "learning_rate": 1.053512939394157e-05, + "loss": 1.2114, + "step": 17761 + }, + { + "epoch": 1.5, + "learning_rate": 1.0534221721099098e-05, + "loss": 1.1065, + "step": 17762 + }, + { + "epoch": 1.5, + "learning_rate": 1.0533314043842713e-05, + "loss": 1.1704, + "step": 17763 + }, + { + "epoch": 1.5, + "learning_rate": 1.0532406362179917e-05, + "loss": 1.1402, + "step": 17764 + }, + { + "epoch": 1.5, + "learning_rate": 1.0531498676118207e-05, + "loss": 1.2331, + "step": 17765 + }, + { + "epoch": 1.5, + "learning_rate": 1.053059098566509e-05, + "loss": 1.2607, + "step": 17766 + }, + { + "epoch": 1.5, + "learning_rate": 1.0529683290828056e-05, + "loss": 1.126, + "step": 17767 + }, + { + "epoch": 1.5, + "learning_rate": 1.052877559161461e-05, + "loss": 1.233, + "step": 17768 + }, + { + "epoch": 1.5, + "learning_rate": 1.0527867888032253e-05, + "loss": 1.2197, + "step": 17769 + }, + { + "epoch": 1.5, + "learning_rate": 1.052696018008848e-05, + "loss": 1.2416, + "step": 17770 + }, + { + "epoch": 1.5, + "learning_rate": 1.0526052467790792e-05, + "loss": 1.1749, + "step": 17771 + }, + { + "epoch": 1.5, + "learning_rate": 1.0525144751146694e-05, + "loss": 1.1962, + "step": 17772 + }, + { + "epoch": 1.5, + "learning_rate": 1.0524237030163681e-05, + "loss": 1.1644, + "step": 17773 + }, + { + "epoch": 1.5, + "learning_rate": 1.0523329304849254e-05, + "loss": 1.1333, + "step": 17774 + }, + { + "epoch": 1.5, + "learning_rate": 1.0522421575210913e-05, + "loss": 1.1817, + "step": 17775 + }, + { + "epoch": 1.5, + "learning_rate": 1.0521513841256155e-05, + "loss": 1.1983, + "step": 17776 + }, + { + "epoch": 1.5, + "learning_rate": 1.0520606102992488e-05, + "loss": 1.1726, + "step": 17777 + }, + { + "epoch": 1.5, + "learning_rate": 1.0519698360427401e-05, + "loss": 1.155, + "step": 17778 + }, + { + "epoch": 1.5, + "learning_rate": 1.0518790613568403e-05, + "loss": 1.1076, + "step": 17779 + }, + { + "epoch": 1.5, + "learning_rate": 1.0517882862422993e-05, + "loss": 1.1866, + "step": 17780 + }, + { + "epoch": 1.5, + "learning_rate": 1.0516975106998663e-05, + "loss": 1.2243, + "step": 17781 + }, + { + "epoch": 1.5, + "learning_rate": 1.0516067347302921e-05, + "loss": 1.2071, + "step": 17782 + }, + { + "epoch": 1.5, + "learning_rate": 1.0515159583343268e-05, + "loss": 1.1334, + "step": 17783 + }, + { + "epoch": 1.5, + "learning_rate": 1.0514251815127201e-05, + "loss": 1.2126, + "step": 17784 + }, + { + "epoch": 1.5, + "learning_rate": 1.051334404266222e-05, + "loss": 1.2372, + "step": 17785 + }, + { + "epoch": 1.5, + "learning_rate": 1.0512436265955826e-05, + "loss": 1.1387, + "step": 17786 + }, + { + "epoch": 1.5, + "learning_rate": 1.0511528485015524e-05, + "loss": 1.2461, + "step": 17787 + }, + { + "epoch": 1.5, + "learning_rate": 1.0510620699848803e-05, + "loss": 1.2081, + "step": 17788 + }, + { + "epoch": 1.5, + "learning_rate": 1.0509712910463175e-05, + "loss": 1.1415, + "step": 17789 + }, + { + "epoch": 1.5, + "learning_rate": 1.0508805116866134e-05, + "loss": 1.3172, + "step": 17790 + }, + { + "epoch": 1.5, + "learning_rate": 1.0507897319065183e-05, + "loss": 1.1159, + "step": 17791 + }, + { + "epoch": 1.5, + "learning_rate": 1.0506989517067823e-05, + "loss": 1.1927, + "step": 17792 + }, + { + "epoch": 1.5, + "learning_rate": 1.0506081710881551e-05, + "loss": 1.2016, + "step": 17793 + }, + { + "epoch": 1.5, + "learning_rate": 1.0505173900513869e-05, + "loss": 1.2478, + "step": 17794 + }, + { + "epoch": 1.5, + "learning_rate": 1.0504266085972281e-05, + "loss": 1.1532, + "step": 17795 + }, + { + "epoch": 1.5, + "learning_rate": 1.0503358267264282e-05, + "loss": 1.2638, + "step": 17796 + }, + { + "epoch": 1.5, + "learning_rate": 1.0502450444397379e-05, + "loss": 1.1761, + "step": 17797 + }, + { + "epoch": 1.5, + "learning_rate": 1.050154261737907e-05, + "loss": 1.0669, + "step": 17798 + }, + { + "epoch": 1.5, + "learning_rate": 1.0500634786216854e-05, + "loss": 1.2279, + "step": 17799 + }, + { + "epoch": 1.5, + "learning_rate": 1.0499726950918235e-05, + "loss": 1.1775, + "step": 17800 + }, + { + "epoch": 1.5, + "learning_rate": 1.049881911149071e-05, + "loss": 1.1785, + "step": 17801 + }, + { + "epoch": 1.5, + "learning_rate": 1.0497911267941783e-05, + "loss": 1.1781, + "step": 17802 + }, + { + "epoch": 1.5, + "learning_rate": 1.0497003420278952e-05, + "loss": 1.1653, + "step": 17803 + }, + { + "epoch": 1.5, + "learning_rate": 1.0496095568509721e-05, + "loss": 1.2689, + "step": 17804 + }, + { + "epoch": 1.5, + "learning_rate": 1.049518771264159e-05, + "loss": 1.1796, + "step": 17805 + }, + { + "epoch": 1.5, + "learning_rate": 1.049427985268206e-05, + "loss": 1.2438, + "step": 17806 + }, + { + "epoch": 1.5, + "learning_rate": 1.049337198863863e-05, + "loss": 1.3488, + "step": 17807 + }, + { + "epoch": 1.5, + "learning_rate": 1.0492464120518802e-05, + "loss": 1.1418, + "step": 17808 + }, + { + "epoch": 1.5, + "learning_rate": 1.0491556248330081e-05, + "loss": 1.2775, + "step": 17809 + }, + { + "epoch": 1.5, + "learning_rate": 1.0490648372079963e-05, + "loss": 1.1123, + "step": 17810 + }, + { + "epoch": 1.5, + "learning_rate": 1.048974049177595e-05, + "loss": 1.2346, + "step": 17811 + }, + { + "epoch": 1.5, + "learning_rate": 1.0488832607425549e-05, + "loss": 1.1604, + "step": 17812 + }, + { + "epoch": 1.5, + "learning_rate": 1.0487924719036253e-05, + "loss": 1.2647, + "step": 17813 + }, + { + "epoch": 1.5, + "learning_rate": 1.0487016826615565e-05, + "loss": 1.2189, + "step": 17814 + }, + { + "epoch": 1.5, + "learning_rate": 1.048610893017099e-05, + "loss": 1.1793, + "step": 17815 + }, + { + "epoch": 1.5, + "learning_rate": 1.048520102971003e-05, + "loss": 1.2174, + "step": 17816 + }, + { + "epoch": 1.5, + "learning_rate": 1.0484293125240182e-05, + "loss": 1.1835, + "step": 17817 + }, + { + "epoch": 1.5, + "learning_rate": 1.048338521676895e-05, + "loss": 1.1965, + "step": 17818 + }, + { + "epoch": 1.5, + "learning_rate": 1.048247730430383e-05, + "loss": 1.0906, + "step": 17819 + }, + { + "epoch": 1.5, + "learning_rate": 1.0481569387852333e-05, + "loss": 1.1707, + "step": 17820 + }, + { + "epoch": 1.5, + "learning_rate": 1.0480661467421953e-05, + "loss": 1.227, + "step": 17821 + }, + { + "epoch": 1.5, + "learning_rate": 1.0479753543020194e-05, + "loss": 1.231, + "step": 17822 + }, + { + "epoch": 1.5, + "learning_rate": 1.047884561465456e-05, + "loss": 1.2602, + "step": 17823 + }, + { + "epoch": 1.5, + "learning_rate": 1.047793768233255e-05, + "loss": 1.2242, + "step": 17824 + }, + { + "epoch": 1.5, + "learning_rate": 1.0477029746061664e-05, + "loss": 1.2186, + "step": 17825 + }, + { + "epoch": 1.5, + "learning_rate": 1.0476121805849406e-05, + "loss": 1.1231, + "step": 17826 + }, + { + "epoch": 1.5, + "learning_rate": 1.047521386170328e-05, + "loss": 1.1345, + "step": 17827 + }, + { + "epoch": 1.5, + "learning_rate": 1.0474305913630782e-05, + "loss": 1.1917, + "step": 17828 + }, + { + "epoch": 1.5, + "learning_rate": 1.0473397961639418e-05, + "loss": 1.1766, + "step": 17829 + }, + { + "epoch": 1.5, + "learning_rate": 1.047249000573669e-05, + "loss": 1.1655, + "step": 17830 + }, + { + "epoch": 1.5, + "learning_rate": 1.0471582045930096e-05, + "loss": 1.2453, + "step": 17831 + }, + { + "epoch": 1.5, + "learning_rate": 1.0470674082227136e-05, + "loss": 1.1762, + "step": 17832 + }, + { + "epoch": 1.5, + "learning_rate": 1.0469766114635323e-05, + "loss": 1.0741, + "step": 17833 + }, + { + "epoch": 1.5, + "learning_rate": 1.046885814316215e-05, + "loss": 1.2483, + "step": 17834 + }, + { + "epoch": 1.5, + "learning_rate": 1.0467950167815122e-05, + "loss": 1.3136, + "step": 17835 + }, + { + "epoch": 1.5, + "learning_rate": 1.0467042188601736e-05, + "loss": 1.2463, + "step": 17836 + }, + { + "epoch": 1.5, + "learning_rate": 1.0466134205529503e-05, + "loss": 1.2206, + "step": 17837 + }, + { + "epoch": 1.5, + "learning_rate": 1.046522621860592e-05, + "loss": 1.2425, + "step": 17838 + }, + { + "epoch": 1.5, + "learning_rate": 1.0464318227838486e-05, + "loss": 1.1838, + "step": 17839 + }, + { + "epoch": 1.5, + "learning_rate": 1.0463410233234706e-05, + "loss": 1.17, + "step": 17840 + }, + { + "epoch": 1.5, + "learning_rate": 1.0462502234802086e-05, + "loss": 1.2224, + "step": 17841 + }, + { + "epoch": 1.5, + "learning_rate": 1.046159423254812e-05, + "loss": 1.1875, + "step": 17842 + }, + { + "epoch": 1.5, + "learning_rate": 1.0460686226480316e-05, + "loss": 1.1692, + "step": 17843 + }, + { + "epoch": 1.5, + "learning_rate": 1.0459778216606175e-05, + "loss": 1.1043, + "step": 17844 + }, + { + "epoch": 1.5, + "learning_rate": 1.04588702029332e-05, + "loss": 1.2036, + "step": 17845 + }, + { + "epoch": 1.5, + "learning_rate": 1.045796218546889e-05, + "loss": 1.2114, + "step": 17846 + }, + { + "epoch": 1.5, + "learning_rate": 1.0457054164220752e-05, + "loss": 1.2262, + "step": 17847 + }, + { + "epoch": 1.5, + "learning_rate": 1.0456146139196291e-05, + "loss": 1.1588, + "step": 17848 + }, + { + "epoch": 1.5, + "learning_rate": 1.0455238110403e-05, + "loss": 1.1737, + "step": 17849 + }, + { + "epoch": 1.5, + "learning_rate": 1.0454330077848382e-05, + "loss": 1.1586, + "step": 17850 + }, + { + "epoch": 1.5, + "learning_rate": 1.0453422041539948e-05, + "loss": 1.1634, + "step": 17851 + }, + { + "epoch": 1.5, + "learning_rate": 1.04525140014852e-05, + "loss": 1.1352, + "step": 17852 + }, + { + "epoch": 1.5, + "learning_rate": 1.0451605957691629e-05, + "loss": 1.141, + "step": 17853 + }, + { + "epoch": 1.5, + "learning_rate": 1.0450697910166748e-05, + "loss": 1.1524, + "step": 17854 + }, + { + "epoch": 1.5, + "learning_rate": 1.0449789858918056e-05, + "loss": 1.123, + "step": 17855 + }, + { + "epoch": 1.5, + "learning_rate": 1.0448881803953059e-05, + "loss": 1.2213, + "step": 17856 + }, + { + "epoch": 1.5, + "learning_rate": 1.0447973745279252e-05, + "loss": 1.215, + "step": 17857 + }, + { + "epoch": 1.5, + "learning_rate": 1.0447065682904145e-05, + "loss": 1.2187, + "step": 17858 + }, + { + "epoch": 1.5, + "learning_rate": 1.044615761683524e-05, + "loss": 1.1836, + "step": 17859 + }, + { + "epoch": 1.5, + "learning_rate": 1.0445249547080036e-05, + "loss": 1.233, + "step": 17860 + }, + { + "epoch": 1.5, + "learning_rate": 1.0444341473646038e-05, + "loss": 1.2047, + "step": 17861 + }, + { + "epoch": 1.5, + "learning_rate": 1.0443433396540752e-05, + "loss": 1.0936, + "step": 17862 + }, + { + "epoch": 1.5, + "learning_rate": 1.0442525315771672e-05, + "loss": 1.2765, + "step": 17863 + }, + { + "epoch": 1.5, + "learning_rate": 1.0441617231346309e-05, + "loss": 1.1588, + "step": 17864 + }, + { + "epoch": 1.5, + "learning_rate": 1.0440709143272161e-05, + "loss": 1.2341, + "step": 17865 + }, + { + "epoch": 1.5, + "learning_rate": 1.0439801051556738e-05, + "loss": 1.1605, + "step": 17866 + }, + { + "epoch": 1.5, + "learning_rate": 1.0438892956207533e-05, + "loss": 1.1266, + "step": 17867 + }, + { + "epoch": 1.5, + "learning_rate": 1.0437984857232056e-05, + "loss": 1.202, + "step": 17868 + }, + { + "epoch": 1.5, + "learning_rate": 1.0437076754637807e-05, + "loss": 1.2261, + "step": 17869 + }, + { + "epoch": 1.5, + "learning_rate": 1.0436168648432289e-05, + "loss": 1.1828, + "step": 17870 + }, + { + "epoch": 1.5, + "learning_rate": 1.0435260538623007e-05, + "loss": 1.1328, + "step": 17871 + }, + { + "epoch": 1.5, + "learning_rate": 1.0434352425217461e-05, + "loss": 1.1443, + "step": 17872 + }, + { + "epoch": 1.5, + "learning_rate": 1.0433444308223161e-05, + "loss": 1.1956, + "step": 17873 + }, + { + "epoch": 1.5, + "learning_rate": 1.0432536187647604e-05, + "loss": 1.1711, + "step": 17874 + }, + { + "epoch": 1.5, + "learning_rate": 1.0431628063498294e-05, + "loss": 1.1267, + "step": 17875 + }, + { + "epoch": 1.51, + "learning_rate": 1.0430719935782734e-05, + "loss": 1.1113, + "step": 17876 + }, + { + "epoch": 1.51, + "learning_rate": 1.042981180450843e-05, + "loss": 1.1868, + "step": 17877 + }, + { + "epoch": 1.51, + "learning_rate": 1.0428903669682882e-05, + "loss": 1.2672, + "step": 17878 + }, + { + "epoch": 1.51, + "learning_rate": 1.0427995531313594e-05, + "loss": 1.2335, + "step": 17879 + }, + { + "epoch": 1.51, + "learning_rate": 1.0427087389408071e-05, + "loss": 1.1931, + "step": 17880 + }, + { + "epoch": 1.51, + "learning_rate": 1.0426179243973815e-05, + "loss": 1.2881, + "step": 17881 + }, + { + "epoch": 1.51, + "learning_rate": 1.0425271095018327e-05, + "loss": 1.1932, + "step": 17882 + }, + { + "epoch": 1.51, + "learning_rate": 1.0424362942549116e-05, + "loss": 1.2201, + "step": 17883 + }, + { + "epoch": 1.51, + "learning_rate": 1.0423454786573684e-05, + "loss": 1.1961, + "step": 17884 + }, + { + "epoch": 1.51, + "learning_rate": 1.042254662709953e-05, + "loss": 1.2527, + "step": 17885 + }, + { + "epoch": 1.51, + "learning_rate": 1.042163846413416e-05, + "loss": 1.2426, + "step": 17886 + }, + { + "epoch": 1.51, + "learning_rate": 1.0420730297685083e-05, + "loss": 1.2361, + "step": 17887 + }, + { + "epoch": 1.51, + "learning_rate": 1.0419822127759794e-05, + "loss": 1.1705, + "step": 17888 + }, + { + "epoch": 1.51, + "learning_rate": 1.04189139543658e-05, + "loss": 1.2002, + "step": 17889 + }, + { + "epoch": 1.51, + "learning_rate": 1.0418005777510605e-05, + "loss": 1.2444, + "step": 17890 + }, + { + "epoch": 1.51, + "learning_rate": 1.0417097597201714e-05, + "loss": 1.2875, + "step": 17891 + }, + { + "epoch": 1.51, + "learning_rate": 1.0416189413446626e-05, + "loss": 1.1193, + "step": 17892 + }, + { + "epoch": 1.51, + "learning_rate": 1.041528122625285e-05, + "loss": 1.1384, + "step": 17893 + }, + { + "epoch": 1.51, + "learning_rate": 1.0414373035627889e-05, + "loss": 1.1416, + "step": 17894 + }, + { + "epoch": 1.51, + "learning_rate": 1.0413464841579244e-05, + "loss": 1.0758, + "step": 17895 + }, + { + "epoch": 1.51, + "learning_rate": 1.0412556644114418e-05, + "loss": 1.2582, + "step": 17896 + }, + { + "epoch": 1.51, + "learning_rate": 1.0411648443240918e-05, + "loss": 1.2106, + "step": 17897 + }, + { + "epoch": 1.51, + "learning_rate": 1.0410740238966251e-05, + "loss": 1.2456, + "step": 17898 + }, + { + "epoch": 1.51, + "learning_rate": 1.0409832031297912e-05, + "loss": 1.2273, + "step": 17899 + }, + { + "epoch": 1.51, + "learning_rate": 1.040892382024341e-05, + "loss": 1.1927, + "step": 17900 + }, + { + "epoch": 1.51, + "learning_rate": 1.0408015605810251e-05, + "loss": 1.1117, + "step": 17901 + }, + { + "epoch": 1.51, + "learning_rate": 1.0407107388005937e-05, + "loss": 1.1708, + "step": 17902 + }, + { + "epoch": 1.51, + "learning_rate": 1.040619916683797e-05, + "loss": 1.2469, + "step": 17903 + }, + { + "epoch": 1.51, + "learning_rate": 1.0405290942313853e-05, + "loss": 1.2101, + "step": 17904 + }, + { + "epoch": 1.51, + "learning_rate": 1.0404382714441095e-05, + "loss": 1.1731, + "step": 17905 + }, + { + "epoch": 1.51, + "learning_rate": 1.0403474483227199e-05, + "loss": 1.2681, + "step": 17906 + }, + { + "epoch": 1.51, + "learning_rate": 1.0402566248679662e-05, + "loss": 1.1377, + "step": 17907 + }, + { + "epoch": 1.51, + "learning_rate": 1.0401658010805997e-05, + "loss": 1.2467, + "step": 17908 + }, + { + "epoch": 1.51, + "learning_rate": 1.040074976961371e-05, + "loss": 1.0708, + "step": 17909 + }, + { + "epoch": 1.51, + "learning_rate": 1.0399841525110295e-05, + "loss": 1.1479, + "step": 17910 + }, + { + "epoch": 1.51, + "learning_rate": 1.039893327730326e-05, + "loss": 1.1943, + "step": 17911 + }, + { + "epoch": 1.51, + "learning_rate": 1.0398025026200116e-05, + "loss": 1.2279, + "step": 17912 + }, + { + "epoch": 1.51, + "learning_rate": 1.0397116771808357e-05, + "loss": 1.2675, + "step": 17913 + }, + { + "epoch": 1.51, + "learning_rate": 1.0396208514135493e-05, + "loss": 1.1551, + "step": 17914 + }, + { + "epoch": 1.51, + "learning_rate": 1.039530025318903e-05, + "loss": 1.173, + "step": 17915 + }, + { + "epoch": 1.51, + "learning_rate": 1.0394391988976469e-05, + "loss": 1.2128, + "step": 17916 + }, + { + "epoch": 1.51, + "learning_rate": 1.0393483721505314e-05, + "loss": 1.1746, + "step": 17917 + }, + { + "epoch": 1.51, + "learning_rate": 1.0392575450783068e-05, + "loss": 1.2622, + "step": 17918 + }, + { + "epoch": 1.51, + "learning_rate": 1.0391667176817242e-05, + "loss": 1.2701, + "step": 17919 + }, + { + "epoch": 1.51, + "learning_rate": 1.0390758899615334e-05, + "loss": 1.1127, + "step": 17920 + }, + { + "epoch": 1.51, + "learning_rate": 1.0389850619184852e-05, + "loss": 1.1933, + "step": 17921 + }, + { + "epoch": 1.51, + "learning_rate": 1.03889423355333e-05, + "loss": 1.1109, + "step": 17922 + }, + { + "epoch": 1.51, + "learning_rate": 1.0388034048668183e-05, + "loss": 1.1861, + "step": 17923 + }, + { + "epoch": 1.51, + "learning_rate": 1.0387125758597003e-05, + "loss": 1.2263, + "step": 17924 + }, + { + "epoch": 1.51, + "learning_rate": 1.0386217465327264e-05, + "loss": 1.2546, + "step": 17925 + }, + { + "epoch": 1.51, + "learning_rate": 1.0385309168866475e-05, + "loss": 1.2506, + "step": 17926 + }, + { + "epoch": 1.51, + "learning_rate": 1.038440086922214e-05, + "loss": 1.1797, + "step": 17927 + }, + { + "epoch": 1.51, + "learning_rate": 1.038349256640176e-05, + "loss": 1.1534, + "step": 17928 + }, + { + "epoch": 1.51, + "learning_rate": 1.038258426041284e-05, + "loss": 1.1334, + "step": 17929 + }, + { + "epoch": 1.51, + "learning_rate": 1.0381675951262888e-05, + "loss": 1.1857, + "step": 17930 + }, + { + "epoch": 1.51, + "learning_rate": 1.038076763895941e-05, + "loss": 1.1244, + "step": 17931 + }, + { + "epoch": 1.51, + "learning_rate": 1.0379859323509902e-05, + "loss": 1.1163, + "step": 17932 + }, + { + "epoch": 1.51, + "learning_rate": 1.0378951004921877e-05, + "loss": 1.1988, + "step": 17933 + }, + { + "epoch": 1.51, + "learning_rate": 1.0378042683202841e-05, + "loss": 1.1805, + "step": 17934 + }, + { + "epoch": 1.51, + "learning_rate": 1.0377134358360293e-05, + "loss": 1.2251, + "step": 17935 + }, + { + "epoch": 1.51, + "learning_rate": 1.0376226030401739e-05, + "loss": 1.1366, + "step": 17936 + }, + { + "epoch": 1.51, + "learning_rate": 1.0375317699334688e-05, + "loss": 1.2399, + "step": 17937 + }, + { + "epoch": 1.51, + "learning_rate": 1.037440936516664e-05, + "loss": 1.2114, + "step": 17938 + }, + { + "epoch": 1.51, + "learning_rate": 1.0373501027905102e-05, + "loss": 1.1561, + "step": 17939 + }, + { + "epoch": 1.51, + "learning_rate": 1.0372592687557576e-05, + "loss": 1.1511, + "step": 17940 + }, + { + "epoch": 1.51, + "learning_rate": 1.0371684344131576e-05, + "loss": 1.1966, + "step": 17941 + }, + { + "epoch": 1.51, + "learning_rate": 1.03707759976346e-05, + "loss": 1.2668, + "step": 17942 + }, + { + "epoch": 1.51, + "learning_rate": 1.036986764807415e-05, + "loss": 1.0995, + "step": 17943 + }, + { + "epoch": 1.51, + "learning_rate": 1.0368959295457736e-05, + "loss": 1.1898, + "step": 17944 + }, + { + "epoch": 1.51, + "learning_rate": 1.0368050939792863e-05, + "loss": 1.1351, + "step": 17945 + }, + { + "epoch": 1.51, + "learning_rate": 1.0367142581087036e-05, + "loss": 1.2173, + "step": 17946 + }, + { + "epoch": 1.51, + "learning_rate": 1.036623421934776e-05, + "loss": 1.1347, + "step": 17947 + }, + { + "epoch": 1.51, + "learning_rate": 1.0365325854582539e-05, + "loss": 1.2831, + "step": 17948 + }, + { + "epoch": 1.51, + "learning_rate": 1.036441748679888e-05, + "loss": 1.1686, + "step": 17949 + }, + { + "epoch": 1.51, + "learning_rate": 1.0363509116004284e-05, + "loss": 1.2113, + "step": 17950 + }, + { + "epoch": 1.51, + "learning_rate": 1.0362600742206262e-05, + "loss": 1.112, + "step": 17951 + }, + { + "epoch": 1.51, + "learning_rate": 1.0361692365412317e-05, + "loss": 1.2013, + "step": 17952 + }, + { + "epoch": 1.51, + "learning_rate": 1.0360783985629953e-05, + "loss": 1.221, + "step": 17953 + }, + { + "epoch": 1.51, + "learning_rate": 1.0359875602866676e-05, + "loss": 1.1683, + "step": 17954 + }, + { + "epoch": 1.51, + "learning_rate": 1.0358967217129991e-05, + "loss": 1.1956, + "step": 17955 + }, + { + "epoch": 1.51, + "learning_rate": 1.0358058828427406e-05, + "loss": 1.1886, + "step": 17956 + }, + { + "epoch": 1.51, + "learning_rate": 1.0357150436766422e-05, + "loss": 1.1525, + "step": 17957 + }, + { + "epoch": 1.51, + "learning_rate": 1.0356242042154545e-05, + "loss": 1.1524, + "step": 17958 + }, + { + "epoch": 1.51, + "learning_rate": 1.0355333644599289e-05, + "loss": 1.17, + "step": 17959 + }, + { + "epoch": 1.51, + "learning_rate": 1.0354425244108147e-05, + "loss": 1.1802, + "step": 17960 + }, + { + "epoch": 1.51, + "learning_rate": 1.035351684068863e-05, + "loss": 1.1678, + "step": 17961 + }, + { + "epoch": 1.51, + "learning_rate": 1.0352608434348246e-05, + "loss": 1.2371, + "step": 17962 + }, + { + "epoch": 1.51, + "learning_rate": 1.0351700025094499e-05, + "loss": 1.2156, + "step": 17963 + }, + { + "epoch": 1.51, + "learning_rate": 1.0350791612934894e-05, + "loss": 1.239, + "step": 17964 + }, + { + "epoch": 1.51, + "learning_rate": 1.0349883197876935e-05, + "loss": 1.1528, + "step": 17965 + }, + { + "epoch": 1.51, + "learning_rate": 1.0348974779928132e-05, + "loss": 1.2073, + "step": 17966 + }, + { + "epoch": 1.51, + "learning_rate": 1.0348066359095983e-05, + "loss": 1.158, + "step": 17967 + }, + { + "epoch": 1.51, + "learning_rate": 1.0347157935388e-05, + "loss": 1.1658, + "step": 17968 + }, + { + "epoch": 1.51, + "learning_rate": 1.0346249508811687e-05, + "loss": 1.2514, + "step": 17969 + }, + { + "epoch": 1.51, + "learning_rate": 1.0345341079374549e-05, + "loss": 1.2284, + "step": 17970 + }, + { + "epoch": 1.51, + "learning_rate": 1.0344432647084093e-05, + "loss": 1.2042, + "step": 17971 + }, + { + "epoch": 1.51, + "learning_rate": 1.0343524211947824e-05, + "loss": 1.2335, + "step": 17972 + }, + { + "epoch": 1.51, + "learning_rate": 1.0342615773973253e-05, + "loss": 1.1997, + "step": 17973 + }, + { + "epoch": 1.51, + "learning_rate": 1.0341707333167874e-05, + "loss": 1.1949, + "step": 17974 + }, + { + "epoch": 1.51, + "learning_rate": 1.0340798889539202e-05, + "loss": 1.2458, + "step": 17975 + }, + { + "epoch": 1.51, + "learning_rate": 1.033989044309474e-05, + "loss": 1.2133, + "step": 17976 + }, + { + "epoch": 1.51, + "learning_rate": 1.0338981993841997e-05, + "loss": 1.1793, + "step": 17977 + }, + { + "epoch": 1.51, + "learning_rate": 1.0338073541788473e-05, + "loss": 1.1929, + "step": 17978 + }, + { + "epoch": 1.51, + "learning_rate": 1.033716508694168e-05, + "loss": 1.1553, + "step": 17979 + }, + { + "epoch": 1.51, + "learning_rate": 1.0336256629309118e-05, + "loss": 1.2056, + "step": 17980 + }, + { + "epoch": 1.51, + "learning_rate": 1.03353481688983e-05, + "loss": 1.1791, + "step": 17981 + }, + { + "epoch": 1.51, + "learning_rate": 1.0334439705716724e-05, + "loss": 1.1675, + "step": 17982 + }, + { + "epoch": 1.51, + "learning_rate": 1.0333531239771902e-05, + "loss": 1.1856, + "step": 17983 + }, + { + "epoch": 1.51, + "learning_rate": 1.033262277107134e-05, + "loss": 1.2073, + "step": 17984 + }, + { + "epoch": 1.51, + "learning_rate": 1.033171429962254e-05, + "loss": 1.2691, + "step": 17985 + }, + { + "epoch": 1.51, + "learning_rate": 1.033080582543301e-05, + "loss": 1.1673, + "step": 17986 + }, + { + "epoch": 1.51, + "learning_rate": 1.0329897348510255e-05, + "loss": 1.1753, + "step": 17987 + }, + { + "epoch": 1.51, + "learning_rate": 1.0328988868861789e-05, + "loss": 1.1681, + "step": 17988 + }, + { + "epoch": 1.51, + "learning_rate": 1.0328080386495107e-05, + "loss": 1.2384, + "step": 17989 + }, + { + "epoch": 1.51, + "learning_rate": 1.0327171901417718e-05, + "loss": 1.1681, + "step": 17990 + }, + { + "epoch": 1.51, + "learning_rate": 1.0326263413637134e-05, + "loss": 1.1934, + "step": 17991 + }, + { + "epoch": 1.51, + "learning_rate": 1.0325354923160857e-05, + "loss": 1.1583, + "step": 17992 + }, + { + "epoch": 1.51, + "learning_rate": 1.032444642999639e-05, + "loss": 1.2145, + "step": 17993 + }, + { + "epoch": 1.51, + "learning_rate": 1.0323537934151245e-05, + "loss": 1.1718, + "step": 17994 + }, + { + "epoch": 1.52, + "learning_rate": 1.0322629435632925e-05, + "loss": 1.1554, + "step": 17995 + }, + { + "epoch": 1.52, + "learning_rate": 1.0321720934448937e-05, + "loss": 1.1509, + "step": 17996 + }, + { + "epoch": 1.52, + "learning_rate": 1.032081243060679e-05, + "loss": 1.2739, + "step": 17997 + }, + { + "epoch": 1.52, + "learning_rate": 1.0319903924113988e-05, + "loss": 1.1882, + "step": 17998 + }, + { + "epoch": 1.52, + "learning_rate": 1.0318995414978036e-05, + "loss": 1.1722, + "step": 17999 + }, + { + "epoch": 1.52, + "learning_rate": 1.0318086903206441e-05, + "loss": 1.1761, + "step": 18000 + }, + { + "epoch": 1.52, + "learning_rate": 1.0317178388806712e-05, + "loss": 1.1684, + "step": 18001 + }, + { + "epoch": 1.52, + "learning_rate": 1.0316269871786355e-05, + "loss": 1.1716, + "step": 18002 + }, + { + "epoch": 1.52, + "learning_rate": 1.0315361352152873e-05, + "loss": 1.1438, + "step": 18003 + }, + { + "epoch": 1.52, + "learning_rate": 1.0314452829913772e-05, + "loss": 1.1635, + "step": 18004 + }, + { + "epoch": 1.52, + "learning_rate": 1.0313544305076564e-05, + "loss": 1.2104, + "step": 18005 + }, + { + "epoch": 1.52, + "learning_rate": 1.0312635777648754e-05, + "loss": 1.1234, + "step": 18006 + }, + { + "epoch": 1.52, + "learning_rate": 1.0311727247637843e-05, + "loss": 1.2829, + "step": 18007 + }, + { + "epoch": 1.52, + "learning_rate": 1.0310818715051344e-05, + "loss": 1.2084, + "step": 18008 + }, + { + "epoch": 1.52, + "learning_rate": 1.0309910179896762e-05, + "loss": 1.1776, + "step": 18009 + }, + { + "epoch": 1.52, + "learning_rate": 1.0309001642181602e-05, + "loss": 1.2954, + "step": 18010 + }, + { + "epoch": 1.52, + "learning_rate": 1.0308093101913372e-05, + "loss": 1.2095, + "step": 18011 + }, + { + "epoch": 1.52, + "learning_rate": 1.0307184559099579e-05, + "loss": 1.2167, + "step": 18012 + }, + { + "epoch": 1.52, + "learning_rate": 1.0306276013747729e-05, + "loss": 1.2266, + "step": 18013 + }, + { + "epoch": 1.52, + "learning_rate": 1.0305367465865325e-05, + "loss": 1.2345, + "step": 18014 + }, + { + "epoch": 1.52, + "learning_rate": 1.030445891545988e-05, + "loss": 1.2289, + "step": 18015 + }, + { + "epoch": 1.52, + "learning_rate": 1.0303550362538901e-05, + "loss": 1.1632, + "step": 18016 + }, + { + "epoch": 1.52, + "learning_rate": 1.0302641807109888e-05, + "loss": 1.1741, + "step": 18017 + }, + { + "epoch": 1.52, + "learning_rate": 1.0301733249180353e-05, + "loss": 1.2348, + "step": 18018 + }, + { + "epoch": 1.52, + "learning_rate": 1.0300824688757797e-05, + "loss": 1.1303, + "step": 18019 + }, + { + "epoch": 1.52, + "learning_rate": 1.0299916125849736e-05, + "loss": 1.225, + "step": 18020 + }, + { + "epoch": 1.52, + "learning_rate": 1.029900756046367e-05, + "loss": 1.2542, + "step": 18021 + }, + { + "epoch": 1.52, + "learning_rate": 1.0298098992607106e-05, + "loss": 1.1032, + "step": 18022 + }, + { + "epoch": 1.52, + "learning_rate": 1.0297190422287557e-05, + "loss": 1.202, + "step": 18023 + }, + { + "epoch": 1.52, + "learning_rate": 1.0296281849512524e-05, + "loss": 1.2073, + "step": 18024 + }, + { + "epoch": 1.52, + "learning_rate": 1.0295373274289515e-05, + "loss": 1.2006, + "step": 18025 + }, + { + "epoch": 1.52, + "learning_rate": 1.0294464696626036e-05, + "loss": 1.2086, + "step": 18026 + }, + { + "epoch": 1.52, + "learning_rate": 1.02935561165296e-05, + "loss": 1.1453, + "step": 18027 + }, + { + "epoch": 1.52, + "learning_rate": 1.0292647534007707e-05, + "loss": 1.1806, + "step": 18028 + }, + { + "epoch": 1.52, + "learning_rate": 1.0291738949067865e-05, + "loss": 1.266, + "step": 18029 + }, + { + "epoch": 1.52, + "learning_rate": 1.0290830361717584e-05, + "loss": 1.1658, + "step": 18030 + }, + { + "epoch": 1.52, + "learning_rate": 1.0289921771964368e-05, + "loss": 1.2069, + "step": 18031 + }, + { + "epoch": 1.52, + "learning_rate": 1.0289013179815726e-05, + "loss": 1.1772, + "step": 18032 + }, + { + "epoch": 1.52, + "learning_rate": 1.0288104585279163e-05, + "loss": 1.1324, + "step": 18033 + }, + { + "epoch": 1.52, + "learning_rate": 1.0287195988362194e-05, + "loss": 1.1426, + "step": 18034 + }, + { + "epoch": 1.52, + "learning_rate": 1.0286287389072314e-05, + "loss": 1.1264, + "step": 18035 + }, + { + "epoch": 1.52, + "learning_rate": 1.0285378787417037e-05, + "loss": 1.1977, + "step": 18036 + }, + { + "epoch": 1.52, + "learning_rate": 1.0284470183403869e-05, + "loss": 1.1407, + "step": 18037 + }, + { + "epoch": 1.52, + "learning_rate": 1.028356157704032e-05, + "loss": 1.1943, + "step": 18038 + }, + { + "epoch": 1.52, + "learning_rate": 1.028265296833389e-05, + "loss": 1.1724, + "step": 18039 + }, + { + "epoch": 1.52, + "learning_rate": 1.0281744357292094e-05, + "loss": 1.1254, + "step": 18040 + }, + { + "epoch": 1.52, + "learning_rate": 1.0280835743922436e-05, + "loss": 1.1087, + "step": 18041 + }, + { + "epoch": 1.52, + "learning_rate": 1.0279927128232423e-05, + "loss": 1.0661, + "step": 18042 + }, + { + "epoch": 1.52, + "learning_rate": 1.0279018510229561e-05, + "loss": 1.2159, + "step": 18043 + }, + { + "epoch": 1.52, + "learning_rate": 1.027810988992136e-05, + "loss": 1.2065, + "step": 18044 + }, + { + "epoch": 1.52, + "learning_rate": 1.0277201267315324e-05, + "loss": 1.1462, + "step": 18045 + }, + { + "epoch": 1.52, + "learning_rate": 1.0276292642418964e-05, + "loss": 1.1225, + "step": 18046 + }, + { + "epoch": 1.52, + "learning_rate": 1.0275384015239786e-05, + "loss": 1.2289, + "step": 18047 + }, + { + "epoch": 1.52, + "learning_rate": 1.0274475385785299e-05, + "loss": 1.1581, + "step": 18048 + }, + { + "epoch": 1.52, + "learning_rate": 1.0273566754063005e-05, + "loss": 1.1587, + "step": 18049 + }, + { + "epoch": 1.52, + "learning_rate": 1.0272658120080416e-05, + "loss": 1.1685, + "step": 18050 + }, + { + "epoch": 1.52, + "learning_rate": 1.027174948384504e-05, + "loss": 1.2396, + "step": 18051 + }, + { + "epoch": 1.52, + "learning_rate": 1.0270840845364383e-05, + "loss": 1.2538, + "step": 18052 + }, + { + "epoch": 1.52, + "learning_rate": 1.0269932204645949e-05, + "loss": 1.1425, + "step": 18053 + }, + { + "epoch": 1.52, + "learning_rate": 1.0269023561697252e-05, + "loss": 1.305, + "step": 18054 + }, + { + "epoch": 1.52, + "learning_rate": 1.0268114916525793e-05, + "loss": 1.1814, + "step": 18055 + }, + { + "epoch": 1.52, + "learning_rate": 1.0267206269139085e-05, + "loss": 1.2093, + "step": 18056 + }, + { + "epoch": 1.52, + "learning_rate": 1.0266297619544631e-05, + "loss": 1.1849, + "step": 18057 + }, + { + "epoch": 1.52, + "learning_rate": 1.0265388967749944e-05, + "loss": 1.1791, + "step": 18058 + }, + { + "epoch": 1.52, + "learning_rate": 1.0264480313762528e-05, + "loss": 1.1341, + "step": 18059 + }, + { + "epoch": 1.52, + "learning_rate": 1.0263571657589888e-05, + "loss": 1.1446, + "step": 18060 + }, + { + "epoch": 1.52, + "learning_rate": 1.0262662999239537e-05, + "loss": 1.1825, + "step": 18061 + }, + { + "epoch": 1.52, + "learning_rate": 1.0261754338718978e-05, + "loss": 1.1654, + "step": 18062 + }, + { + "epoch": 1.52, + "learning_rate": 1.0260845676035726e-05, + "loss": 1.1403, + "step": 18063 + }, + { + "epoch": 1.52, + "learning_rate": 1.025993701119728e-05, + "loss": 1.1573, + "step": 18064 + }, + { + "epoch": 1.52, + "learning_rate": 1.025902834421115e-05, + "loss": 1.1642, + "step": 18065 + }, + { + "epoch": 1.52, + "learning_rate": 1.025811967508485e-05, + "loss": 1.0733, + "step": 18066 + }, + { + "epoch": 1.52, + "learning_rate": 1.0257211003825877e-05, + "loss": 1.2644, + "step": 18067 + }, + { + "epoch": 1.52, + "learning_rate": 1.0256302330441748e-05, + "loss": 1.1332, + "step": 18068 + }, + { + "epoch": 1.52, + "learning_rate": 1.0255393654939965e-05, + "loss": 1.2014, + "step": 18069 + }, + { + "epoch": 1.52, + "learning_rate": 1.0254484977328037e-05, + "loss": 1.263, + "step": 18070 + }, + { + "epoch": 1.52, + "learning_rate": 1.0253576297613475e-05, + "loss": 1.2142, + "step": 18071 + }, + { + "epoch": 1.52, + "learning_rate": 1.0252667615803782e-05, + "loss": 1.2016, + "step": 18072 + }, + { + "epoch": 1.52, + "learning_rate": 1.0251758931906473e-05, + "loss": 1.1635, + "step": 18073 + }, + { + "epoch": 1.52, + "learning_rate": 1.0250850245929047e-05, + "loss": 1.1002, + "step": 18074 + }, + { + "epoch": 1.52, + "learning_rate": 1.0249941557879017e-05, + "loss": 1.0975, + "step": 18075 + }, + { + "epoch": 1.52, + "learning_rate": 1.0249032867763888e-05, + "loss": 1.2178, + "step": 18076 + }, + { + "epoch": 1.52, + "learning_rate": 1.0248124175591173e-05, + "loss": 1.2353, + "step": 18077 + }, + { + "epoch": 1.52, + "learning_rate": 1.0247215481368375e-05, + "loss": 1.1736, + "step": 18078 + }, + { + "epoch": 1.52, + "learning_rate": 1.0246306785103004e-05, + "loss": 1.1705, + "step": 18079 + }, + { + "epoch": 1.52, + "learning_rate": 1.0245398086802567e-05, + "loss": 1.1649, + "step": 18080 + }, + { + "epoch": 1.52, + "learning_rate": 1.0244489386474574e-05, + "loss": 1.2402, + "step": 18081 + }, + { + "epoch": 1.52, + "learning_rate": 1.024358068412653e-05, + "loss": 1.1966, + "step": 18082 + }, + { + "epoch": 1.52, + "learning_rate": 1.0242671979765944e-05, + "loss": 1.2054, + "step": 18083 + }, + { + "epoch": 1.52, + "learning_rate": 1.0241763273400327e-05, + "loss": 1.1703, + "step": 18084 + }, + { + "epoch": 1.52, + "learning_rate": 1.0240854565037181e-05, + "loss": 1.1426, + "step": 18085 + }, + { + "epoch": 1.52, + "learning_rate": 1.023994585468402e-05, + "loss": 1.2655, + "step": 18086 + }, + { + "epoch": 1.52, + "learning_rate": 1.0239037142348347e-05, + "loss": 1.1902, + "step": 18087 + }, + { + "epoch": 1.52, + "learning_rate": 1.0238128428037677e-05, + "loss": 1.1309, + "step": 18088 + }, + { + "epoch": 1.52, + "learning_rate": 1.023721971175951e-05, + "loss": 1.1855, + "step": 18089 + }, + { + "epoch": 1.52, + "learning_rate": 1.0236310993521359e-05, + "loss": 1.1679, + "step": 18090 + }, + { + "epoch": 1.52, + "learning_rate": 1.0235402273330733e-05, + "loss": 1.1528, + "step": 18091 + }, + { + "epoch": 1.52, + "learning_rate": 1.0234493551195134e-05, + "loss": 1.1959, + "step": 18092 + }, + { + "epoch": 1.52, + "learning_rate": 1.0233584827122077e-05, + "loss": 1.1312, + "step": 18093 + }, + { + "epoch": 1.52, + "learning_rate": 1.0232676101119065e-05, + "loss": 1.1771, + "step": 18094 + }, + { + "epoch": 1.52, + "learning_rate": 1.023176737319361e-05, + "loss": 1.2521, + "step": 18095 + }, + { + "epoch": 1.52, + "learning_rate": 1.0230858643353218e-05, + "loss": 1.2583, + "step": 18096 + }, + { + "epoch": 1.52, + "learning_rate": 1.0229949911605399e-05, + "loss": 1.1647, + "step": 18097 + }, + { + "epoch": 1.52, + "learning_rate": 1.0229041177957662e-05, + "loss": 1.1996, + "step": 18098 + }, + { + "epoch": 1.52, + "learning_rate": 1.0228132442417511e-05, + "loss": 1.2486, + "step": 18099 + }, + { + "epoch": 1.52, + "learning_rate": 1.0227223704992458e-05, + "loss": 1.213, + "step": 18100 + }, + { + "epoch": 1.52, + "learning_rate": 1.0226314965690007e-05, + "loss": 1.1306, + "step": 18101 + }, + { + "epoch": 1.52, + "learning_rate": 1.0225406224517674e-05, + "loss": 1.1891, + "step": 18102 + }, + { + "epoch": 1.52, + "learning_rate": 1.022449748148296e-05, + "loss": 1.1427, + "step": 18103 + }, + { + "epoch": 1.52, + "learning_rate": 1.0223588736593375e-05, + "loss": 1.2006, + "step": 18104 + }, + { + "epoch": 1.52, + "learning_rate": 1.022267998985643e-05, + "loss": 1.1545, + "step": 18105 + }, + { + "epoch": 1.52, + "learning_rate": 1.0221771241279632e-05, + "loss": 1.2079, + "step": 18106 + }, + { + "epoch": 1.52, + "learning_rate": 1.0220862490870487e-05, + "loss": 1.2023, + "step": 18107 + }, + { + "epoch": 1.52, + "learning_rate": 1.0219953738636505e-05, + "loss": 1.259, + "step": 18108 + }, + { + "epoch": 1.52, + "learning_rate": 1.02190449845852e-05, + "loss": 1.2322, + "step": 18109 + }, + { + "epoch": 1.52, + "learning_rate": 1.0218136228724071e-05, + "loss": 1.2048, + "step": 18110 + }, + { + "epoch": 1.52, + "learning_rate": 1.0217227471060631e-05, + "loss": 1.183, + "step": 18111 + }, + { + "epoch": 1.52, + "learning_rate": 1.0216318711602386e-05, + "loss": 1.1799, + "step": 18112 + }, + { + "epoch": 1.52, + "learning_rate": 1.0215409950356852e-05, + "loss": 1.1641, + "step": 18113 + }, + { + "epoch": 1.53, + "learning_rate": 1.0214501187331529e-05, + "loss": 1.3133, + "step": 18114 + }, + { + "epoch": 1.53, + "learning_rate": 1.0213592422533929e-05, + "loss": 1.2603, + "step": 18115 + }, + { + "epoch": 1.53, + "learning_rate": 1.0212683655971559e-05, + "loss": 1.1208, + "step": 18116 + }, + { + "epoch": 1.53, + "learning_rate": 1.0211774887651931e-05, + "loss": 1.2423, + "step": 18117 + }, + { + "epoch": 1.53, + "learning_rate": 1.021086611758255e-05, + "loss": 1.2225, + "step": 18118 + }, + { + "epoch": 1.53, + "learning_rate": 1.0209957345770924e-05, + "loss": 1.2013, + "step": 18119 + }, + { + "epoch": 1.53, + "learning_rate": 1.0209048572224562e-05, + "loss": 1.1907, + "step": 18120 + }, + { + "epoch": 1.53, + "learning_rate": 1.0208139796950977e-05, + "loss": 1.1267, + "step": 18121 + }, + { + "epoch": 1.53, + "learning_rate": 1.0207231019957673e-05, + "loss": 1.1704, + "step": 18122 + }, + { + "epoch": 1.53, + "learning_rate": 1.0206322241252162e-05, + "loss": 1.1881, + "step": 18123 + }, + { + "epoch": 1.53, + "learning_rate": 1.0205413460841949e-05, + "loss": 1.2141, + "step": 18124 + }, + { + "epoch": 1.53, + "learning_rate": 1.0204504678734544e-05, + "loss": 1.1373, + "step": 18125 + }, + { + "epoch": 1.53, + "learning_rate": 1.0203595894937454e-05, + "loss": 1.1539, + "step": 18126 + }, + { + "epoch": 1.53, + "learning_rate": 1.0202687109458196e-05, + "loss": 1.1596, + "step": 18127 + }, + { + "epoch": 1.53, + "learning_rate": 1.0201778322304267e-05, + "loss": 1.1395, + "step": 18128 + }, + { + "epoch": 1.53, + "learning_rate": 1.0200869533483183e-05, + "loss": 1.1599, + "step": 18129 + }, + { + "epoch": 1.53, + "learning_rate": 1.0199960743002446e-05, + "loss": 1.1414, + "step": 18130 + }, + { + "epoch": 1.53, + "learning_rate": 1.0199051950869576e-05, + "loss": 1.2216, + "step": 18131 + }, + { + "epoch": 1.53, + "learning_rate": 1.0198143157092068e-05, + "loss": 1.2375, + "step": 18132 + }, + { + "epoch": 1.53, + "learning_rate": 1.0197234361677442e-05, + "loss": 1.2353, + "step": 18133 + }, + { + "epoch": 1.53, + "learning_rate": 1.0196325564633205e-05, + "loss": 1.1794, + "step": 18134 + }, + { + "epoch": 1.53, + "learning_rate": 1.0195416765966859e-05, + "loss": 1.2053, + "step": 18135 + }, + { + "epoch": 1.53, + "learning_rate": 1.019450796568592e-05, + "loss": 1.2474, + "step": 18136 + }, + { + "epoch": 1.53, + "learning_rate": 1.0193599163797892e-05, + "loss": 1.1208, + "step": 18137 + }, + { + "epoch": 1.53, + "learning_rate": 1.0192690360310288e-05, + "loss": 1.1998, + "step": 18138 + }, + { + "epoch": 1.53, + "learning_rate": 1.0191781555230613e-05, + "loss": 1.2907, + "step": 18139 + }, + { + "epoch": 1.53, + "learning_rate": 1.0190872748566376e-05, + "loss": 1.1549, + "step": 18140 + }, + { + "epoch": 1.53, + "learning_rate": 1.018996394032509e-05, + "loss": 1.1896, + "step": 18141 + }, + { + "epoch": 1.53, + "learning_rate": 1.0189055130514261e-05, + "loss": 1.1843, + "step": 18142 + }, + { + "epoch": 1.53, + "learning_rate": 1.0188146319141397e-05, + "loss": 1.2093, + "step": 18143 + }, + { + "epoch": 1.53, + "learning_rate": 1.0187237506214005e-05, + "loss": 1.2136, + "step": 18144 + }, + { + "epoch": 1.53, + "learning_rate": 1.0186328691739601e-05, + "loss": 1.1888, + "step": 18145 + }, + { + "epoch": 1.53, + "learning_rate": 1.0185419875725688e-05, + "loss": 1.2488, + "step": 18146 + }, + { + "epoch": 1.53, + "learning_rate": 1.0184511058179775e-05, + "loss": 1.2843, + "step": 18147 + }, + { + "epoch": 1.53, + "learning_rate": 1.0183602239109379e-05, + "loss": 1.2012, + "step": 18148 + }, + { + "epoch": 1.53, + "learning_rate": 1.0182693418521995e-05, + "loss": 1.1619, + "step": 18149 + }, + { + "epoch": 1.53, + "learning_rate": 1.0181784596425144e-05, + "loss": 1.1967, + "step": 18150 + }, + { + "epoch": 1.53, + "learning_rate": 1.0180875772826328e-05, + "loss": 1.2555, + "step": 18151 + }, + { + "epoch": 1.53, + "learning_rate": 1.0179966947733062e-05, + "loss": 1.1758, + "step": 18152 + }, + { + "epoch": 1.53, + "learning_rate": 1.017905812115285e-05, + "loss": 1.2482, + "step": 18153 + }, + { + "epoch": 1.53, + "learning_rate": 1.01781492930932e-05, + "loss": 1.2728, + "step": 18154 + }, + { + "epoch": 1.53, + "learning_rate": 1.0177240463561624e-05, + "loss": 1.2639, + "step": 18155 + }, + { + "epoch": 1.53, + "learning_rate": 1.0176331632565632e-05, + "loss": 1.3188, + "step": 18156 + }, + { + "epoch": 1.53, + "learning_rate": 1.017542280011273e-05, + "loss": 1.2438, + "step": 18157 + }, + { + "epoch": 1.53, + "learning_rate": 1.0174513966210428e-05, + "loss": 1.1718, + "step": 18158 + }, + { + "epoch": 1.53, + "learning_rate": 1.0173605130866242e-05, + "loss": 1.257, + "step": 18159 + }, + { + "epoch": 1.53, + "learning_rate": 1.017269629408767e-05, + "loss": 1.1716, + "step": 18160 + }, + { + "epoch": 1.53, + "learning_rate": 1.0171787455882225e-05, + "loss": 1.1403, + "step": 18161 + }, + { + "epoch": 1.53, + "learning_rate": 1.0170878616257417e-05, + "loss": 1.1554, + "step": 18162 + }, + { + "epoch": 1.53, + "learning_rate": 1.016996977522076e-05, + "loss": 1.1953, + "step": 18163 + }, + { + "epoch": 1.53, + "learning_rate": 1.0169060932779756e-05, + "loss": 1.1786, + "step": 18164 + }, + { + "epoch": 1.53, + "learning_rate": 1.0168152088941914e-05, + "loss": 1.1496, + "step": 18165 + }, + { + "epoch": 1.53, + "learning_rate": 1.0167243243714746e-05, + "loss": 1.1865, + "step": 18166 + }, + { + "epoch": 1.53, + "learning_rate": 1.0166334397105764e-05, + "loss": 1.247, + "step": 18167 + }, + { + "epoch": 1.53, + "learning_rate": 1.0165425549122472e-05, + "loss": 1.1448, + "step": 18168 + }, + { + "epoch": 1.53, + "learning_rate": 1.0164516699772378e-05, + "loss": 1.1757, + "step": 18169 + }, + { + "epoch": 1.53, + "learning_rate": 1.0163607849063e-05, + "loss": 1.1234, + "step": 18170 + }, + { + "epoch": 1.53, + "learning_rate": 1.0162698997001838e-05, + "loss": 1.1964, + "step": 18171 + }, + { + "epoch": 1.53, + "learning_rate": 1.0161790143596405e-05, + "loss": 1.1713, + "step": 18172 + }, + { + "epoch": 1.53, + "learning_rate": 1.0160881288854214e-05, + "loss": 1.1753, + "step": 18173 + }, + { + "epoch": 1.53, + "learning_rate": 1.0159972432782766e-05, + "loss": 1.075, + "step": 18174 + }, + { + "epoch": 1.53, + "learning_rate": 1.0159063575389575e-05, + "loss": 1.1971, + "step": 18175 + }, + { + "epoch": 1.53, + "learning_rate": 1.0158154716682152e-05, + "loss": 1.201, + "step": 18176 + }, + { + "epoch": 1.53, + "learning_rate": 1.0157245856668003e-05, + "loss": 1.1051, + "step": 18177 + }, + { + "epoch": 1.53, + "learning_rate": 1.0156336995354638e-05, + "loss": 1.1327, + "step": 18178 + }, + { + "epoch": 1.53, + "learning_rate": 1.0155428132749568e-05, + "loss": 1.2107, + "step": 18179 + }, + { + "epoch": 1.53, + "learning_rate": 1.0154519268860298e-05, + "loss": 1.2237, + "step": 18180 + }, + { + "epoch": 1.53, + "learning_rate": 1.0153610403694345e-05, + "loss": 1.1991, + "step": 18181 + }, + { + "epoch": 1.53, + "learning_rate": 1.0152701537259207e-05, + "loss": 1.1836, + "step": 18182 + }, + { + "epoch": 1.53, + "learning_rate": 1.0151792669562404e-05, + "loss": 1.1761, + "step": 18183 + }, + { + "epoch": 1.53, + "learning_rate": 1.0150883800611442e-05, + "loss": 1.1623, + "step": 18184 + }, + { + "epoch": 1.53, + "learning_rate": 1.014997493041383e-05, + "loss": 1.1876, + "step": 18185 + }, + { + "epoch": 1.53, + "learning_rate": 1.0149066058977076e-05, + "loss": 1.1709, + "step": 18186 + }, + { + "epoch": 1.53, + "learning_rate": 1.014815718630869e-05, + "loss": 1.148, + "step": 18187 + }, + { + "epoch": 1.53, + "learning_rate": 1.0147248312416184e-05, + "loss": 1.2972, + "step": 18188 + }, + { + "epoch": 1.53, + "learning_rate": 1.0146339437307064e-05, + "loss": 1.1792, + "step": 18189 + }, + { + "epoch": 1.53, + "learning_rate": 1.014543056098884e-05, + "loss": 1.1609, + "step": 18190 + }, + { + "epoch": 1.53, + "learning_rate": 1.0144521683469022e-05, + "loss": 1.1195, + "step": 18191 + }, + { + "epoch": 1.53, + "learning_rate": 1.0143612804755122e-05, + "loss": 1.2577, + "step": 18192 + }, + { + "epoch": 1.53, + "learning_rate": 1.0142703924854647e-05, + "loss": 1.2179, + "step": 18193 + }, + { + "epoch": 1.53, + "learning_rate": 1.01417950437751e-05, + "loss": 1.1565, + "step": 18194 + }, + { + "epoch": 1.53, + "learning_rate": 1.0140886161524004e-05, + "loss": 1.255, + "step": 18195 + }, + { + "epoch": 1.53, + "learning_rate": 1.0139977278108858e-05, + "loss": 1.2731, + "step": 18196 + }, + { + "epoch": 1.53, + "learning_rate": 1.0139068393537177e-05, + "loss": 1.2041, + "step": 18197 + }, + { + "epoch": 1.53, + "learning_rate": 1.0138159507816469e-05, + "loss": 1.147, + "step": 18198 + }, + { + "epoch": 1.53, + "learning_rate": 1.0137250620954241e-05, + "loss": 1.2988, + "step": 18199 + }, + { + "epoch": 1.53, + "learning_rate": 1.0136341732958006e-05, + "loss": 1.2773, + "step": 18200 + }, + { + "epoch": 1.53, + "learning_rate": 1.0135432843835271e-05, + "loss": 1.2188, + "step": 18201 + }, + { + "epoch": 1.53, + "learning_rate": 1.013452395359355e-05, + "loss": 1.1323, + "step": 18202 + }, + { + "epoch": 1.53, + "learning_rate": 1.0133615062240344e-05, + "loss": 1.1765, + "step": 18203 + }, + { + "epoch": 1.53, + "learning_rate": 1.0132706169783169e-05, + "loss": 1.1765, + "step": 18204 + }, + { + "epoch": 1.53, + "learning_rate": 1.0131797276229532e-05, + "loss": 1.2083, + "step": 18205 + }, + { + "epoch": 1.53, + "learning_rate": 1.0130888381586947e-05, + "loss": 1.1963, + "step": 18206 + }, + { + "epoch": 1.53, + "learning_rate": 1.0129979485862915e-05, + "loss": 1.1481, + "step": 18207 + }, + { + "epoch": 1.53, + "learning_rate": 1.0129070589064956e-05, + "loss": 1.2494, + "step": 18208 + }, + { + "epoch": 1.53, + "learning_rate": 1.0128161691200573e-05, + "loss": 1.1082, + "step": 18209 + }, + { + "epoch": 1.53, + "learning_rate": 1.0127252792277276e-05, + "loss": 1.1874, + "step": 18210 + }, + { + "epoch": 1.53, + "learning_rate": 1.0126343892302577e-05, + "loss": 1.1786, + "step": 18211 + }, + { + "epoch": 1.53, + "learning_rate": 1.0125434991283984e-05, + "loss": 1.152, + "step": 18212 + }, + { + "epoch": 1.53, + "learning_rate": 1.0124526089229008e-05, + "loss": 1.2055, + "step": 18213 + }, + { + "epoch": 1.53, + "learning_rate": 1.0123617186145155e-05, + "loss": 1.2881, + "step": 18214 + }, + { + "epoch": 1.53, + "learning_rate": 1.0122708282039939e-05, + "loss": 1.1475, + "step": 18215 + }, + { + "epoch": 1.53, + "learning_rate": 1.0121799376920867e-05, + "loss": 1.093, + "step": 18216 + }, + { + "epoch": 1.53, + "learning_rate": 1.0120890470795453e-05, + "loss": 1.1428, + "step": 18217 + }, + { + "epoch": 1.53, + "learning_rate": 1.01199815636712e-05, + "loss": 1.228, + "step": 18218 + }, + { + "epoch": 1.53, + "learning_rate": 1.0119072655555618e-05, + "loss": 1.1331, + "step": 18219 + }, + { + "epoch": 1.53, + "learning_rate": 1.0118163746456224e-05, + "loss": 1.2316, + "step": 18220 + }, + { + "epoch": 1.53, + "learning_rate": 1.0117254836380523e-05, + "loss": 1.16, + "step": 18221 + }, + { + "epoch": 1.53, + "learning_rate": 1.0116345925336024e-05, + "loss": 1.109, + "step": 18222 + }, + { + "epoch": 1.53, + "learning_rate": 1.011543701333024e-05, + "loss": 1.2339, + "step": 18223 + }, + { + "epoch": 1.53, + "learning_rate": 1.0114528100370675e-05, + "loss": 1.2562, + "step": 18224 + }, + { + "epoch": 1.53, + "learning_rate": 1.0113619186464845e-05, + "loss": 1.1153, + "step": 18225 + }, + { + "epoch": 1.53, + "learning_rate": 1.0112710271620254e-05, + "loss": 1.2299, + "step": 18226 + }, + { + "epoch": 1.53, + "learning_rate": 1.0111801355844418e-05, + "loss": 1.2002, + "step": 18227 + }, + { + "epoch": 1.53, + "learning_rate": 1.0110892439144841e-05, + "loss": 1.1689, + "step": 18228 + }, + { + "epoch": 1.53, + "learning_rate": 1.0109983521529036e-05, + "loss": 1.2002, + "step": 18229 + }, + { + "epoch": 1.53, + "learning_rate": 1.0109074603004512e-05, + "loss": 1.1701, + "step": 18230 + }, + { + "epoch": 1.53, + "learning_rate": 1.0108165683578779e-05, + "loss": 1.2188, + "step": 18231 + }, + { + "epoch": 1.53, + "learning_rate": 1.0107256763259343e-05, + "loss": 1.2282, + "step": 18232 + }, + { + "epoch": 1.54, + "learning_rate": 1.010634784205372e-05, + "loss": 1.0409, + "step": 18233 + }, + { + "epoch": 1.54, + "learning_rate": 1.010543891996942e-05, + "loss": 1.1979, + "step": 18234 + }, + { + "epoch": 1.54, + "learning_rate": 1.0104529997013946e-05, + "loss": 1.1781, + "step": 18235 + }, + { + "epoch": 1.54, + "learning_rate": 1.0103621073194814e-05, + "loss": 1.2165, + "step": 18236 + }, + { + "epoch": 1.54, + "learning_rate": 1.010271214851953e-05, + "loss": 1.1552, + "step": 18237 + }, + { + "epoch": 1.54, + "learning_rate": 1.0101803222995608e-05, + "loss": 1.193, + "step": 18238 + }, + { + "epoch": 1.54, + "learning_rate": 1.0100894296630552e-05, + "loss": 1.1971, + "step": 18239 + }, + { + "epoch": 1.54, + "learning_rate": 1.0099985369431875e-05, + "loss": 1.1831, + "step": 18240 + }, + { + "epoch": 1.54, + "learning_rate": 1.0099076441407087e-05, + "loss": 1.1985, + "step": 18241 + }, + { + "epoch": 1.54, + "learning_rate": 1.0098167512563701e-05, + "loss": 1.1832, + "step": 18242 + }, + { + "epoch": 1.54, + "learning_rate": 1.0097258582909222e-05, + "loss": 1.2322, + "step": 18243 + }, + { + "epoch": 1.54, + "learning_rate": 1.0096349652451155e-05, + "loss": 1.1176, + "step": 18244 + }, + { + "epoch": 1.54, + "learning_rate": 1.0095440721197026e-05, + "loss": 1.3004, + "step": 18245 + }, + { + "epoch": 1.54, + "learning_rate": 1.0094531789154332e-05, + "loss": 1.149, + "step": 18246 + }, + { + "epoch": 1.54, + "learning_rate": 1.0093622856330585e-05, + "loss": 1.1778, + "step": 18247 + }, + { + "epoch": 1.54, + "learning_rate": 1.0092713922733298e-05, + "loss": 1.1604, + "step": 18248 + }, + { + "epoch": 1.54, + "learning_rate": 1.0091804988369976e-05, + "loss": 1.208, + "step": 18249 + }, + { + "epoch": 1.54, + "learning_rate": 1.0090896053248134e-05, + "loss": 1.1817, + "step": 18250 + }, + { + "epoch": 1.54, + "learning_rate": 1.0089987117375278e-05, + "loss": 1.1572, + "step": 18251 + }, + { + "epoch": 1.54, + "learning_rate": 1.0089078180758923e-05, + "loss": 1.1953, + "step": 18252 + }, + { + "epoch": 1.54, + "learning_rate": 1.0088169243406572e-05, + "loss": 1.133, + "step": 18253 + }, + { + "epoch": 1.54, + "learning_rate": 1.008726030532574e-05, + "loss": 1.1725, + "step": 18254 + }, + { + "epoch": 1.54, + "learning_rate": 1.0086351366523933e-05, + "loss": 1.165, + "step": 18255 + }, + { + "epoch": 1.54, + "learning_rate": 1.0085442427008667e-05, + "loss": 1.2682, + "step": 18256 + }, + { + "epoch": 1.54, + "learning_rate": 1.0084533486787444e-05, + "loss": 1.2373, + "step": 18257 + }, + { + "epoch": 1.54, + "learning_rate": 1.0083624545867781e-05, + "loss": 1.2353, + "step": 18258 + }, + { + "epoch": 1.54, + "learning_rate": 1.0082715604257187e-05, + "loss": 1.2303, + "step": 18259 + }, + { + "epoch": 1.54, + "learning_rate": 1.0081806661963167e-05, + "loss": 1.1441, + "step": 18260 + }, + { + "epoch": 1.54, + "learning_rate": 1.0080897718993235e-05, + "loss": 1.1928, + "step": 18261 + }, + { + "epoch": 1.54, + "learning_rate": 1.00799887753549e-05, + "loss": 1.2518, + "step": 18262 + }, + { + "epoch": 1.54, + "learning_rate": 1.0079079831055675e-05, + "loss": 1.2047, + "step": 18263 + }, + { + "epoch": 1.54, + "learning_rate": 1.0078170886103064e-05, + "loss": 1.1386, + "step": 18264 + }, + { + "epoch": 1.54, + "learning_rate": 1.007726194050458e-05, + "loss": 1.2383, + "step": 18265 + }, + { + "epoch": 1.54, + "learning_rate": 1.0076352994267735e-05, + "loss": 1.1228, + "step": 18266 + }, + { + "epoch": 1.54, + "learning_rate": 1.0075444047400039e-05, + "loss": 1.3241, + "step": 18267 + }, + { + "epoch": 1.54, + "learning_rate": 1.0074535099908997e-05, + "loss": 1.16, + "step": 18268 + }, + { + "epoch": 1.54, + "learning_rate": 1.0073626151802118e-05, + "loss": 1.2406, + "step": 18269 + }, + { + "epoch": 1.54, + "learning_rate": 1.0072717203086924e-05, + "loss": 1.1773, + "step": 18270 + }, + { + "epoch": 1.54, + "learning_rate": 1.0071808253770914e-05, + "loss": 1.1685, + "step": 18271 + }, + { + "epoch": 1.54, + "learning_rate": 1.00708993038616e-05, + "loss": 1.1627, + "step": 18272 + }, + { + "epoch": 1.54, + "learning_rate": 1.0069990353366497e-05, + "loss": 1.1181, + "step": 18273 + }, + { + "epoch": 1.54, + "learning_rate": 1.0069081402293109e-05, + "loss": 1.134, + "step": 18274 + }, + { + "epoch": 1.54, + "learning_rate": 1.0068172450648948e-05, + "loss": 1.1584, + "step": 18275 + }, + { + "epoch": 1.54, + "learning_rate": 1.0067263498441522e-05, + "loss": 1.192, + "step": 18276 + }, + { + "epoch": 1.54, + "learning_rate": 1.006635454567835e-05, + "loss": 1.2697, + "step": 18277 + }, + { + "epoch": 1.54, + "learning_rate": 1.006544559236693e-05, + "loss": 1.2545, + "step": 18278 + }, + { + "epoch": 1.54, + "learning_rate": 1.006453663851478e-05, + "loss": 1.118, + "step": 18279 + }, + { + "epoch": 1.54, + "learning_rate": 1.0063627684129404e-05, + "loss": 1.1448, + "step": 18280 + }, + { + "epoch": 1.54, + "learning_rate": 1.0062718729218322e-05, + "loss": 1.1276, + "step": 18281 + }, + { + "epoch": 1.54, + "learning_rate": 1.0061809773789031e-05, + "loss": 1.1391, + "step": 18282 + }, + { + "epoch": 1.54, + "learning_rate": 1.0060900817849051e-05, + "loss": 1.1077, + "step": 18283 + }, + { + "epoch": 1.54, + "learning_rate": 1.0059991861405889e-05, + "loss": 1.0792, + "step": 18284 + }, + { + "epoch": 1.54, + "learning_rate": 1.0059082904467055e-05, + "loss": 1.1659, + "step": 18285 + }, + { + "epoch": 1.54, + "learning_rate": 1.005817394704006e-05, + "loss": 1.2383, + "step": 18286 + }, + { + "epoch": 1.54, + "learning_rate": 1.0057264989132412e-05, + "loss": 1.1886, + "step": 18287 + }, + { + "epoch": 1.54, + "learning_rate": 1.0056356030751623e-05, + "loss": 1.2105, + "step": 18288 + }, + { + "epoch": 1.54, + "learning_rate": 1.0055447071905202e-05, + "loss": 1.1961, + "step": 18289 + }, + { + "epoch": 1.54, + "learning_rate": 1.0054538112600661e-05, + "loss": 1.2259, + "step": 18290 + }, + { + "epoch": 1.54, + "learning_rate": 1.0053629152845505e-05, + "loss": 1.2185, + "step": 18291 + }, + { + "epoch": 1.54, + "learning_rate": 1.0052720192647252e-05, + "loss": 1.2064, + "step": 18292 + }, + { + "epoch": 1.54, + "learning_rate": 1.0051811232013406e-05, + "loss": 1.2319, + "step": 18293 + }, + { + "epoch": 1.54, + "learning_rate": 1.0050902270951477e-05, + "loss": 1.121, + "step": 18294 + }, + { + "epoch": 1.54, + "learning_rate": 1.0049993309468982e-05, + "loss": 1.2907, + "step": 18295 + }, + { + "epoch": 1.54, + "learning_rate": 1.0049084347573422e-05, + "loss": 1.2609, + "step": 18296 + }, + { + "epoch": 1.54, + "learning_rate": 1.0048175385272312e-05, + "loss": 1.0902, + "step": 18297 + }, + { + "epoch": 1.54, + "learning_rate": 1.0047266422573163e-05, + "loss": 1.1991, + "step": 18298 + }, + { + "epoch": 1.54, + "learning_rate": 1.0046357459483485e-05, + "loss": 1.1316, + "step": 18299 + }, + { + "epoch": 1.54, + "learning_rate": 1.0045448496010783e-05, + "loss": 1.1657, + "step": 18300 + }, + { + "epoch": 1.54, + "learning_rate": 1.0044539532162574e-05, + "loss": 1.1983, + "step": 18301 + }, + { + "epoch": 1.54, + "learning_rate": 1.0043630567946366e-05, + "loss": 1.1825, + "step": 18302 + }, + { + "epoch": 1.54, + "learning_rate": 1.0042721603369665e-05, + "loss": 1.235, + "step": 18303 + }, + { + "epoch": 1.54, + "learning_rate": 1.0041812638439985e-05, + "loss": 1.1614, + "step": 18304 + }, + { + "epoch": 1.54, + "learning_rate": 1.0040903673164838e-05, + "loss": 1.2864, + "step": 18305 + }, + { + "epoch": 1.54, + "learning_rate": 1.003999470755173e-05, + "loss": 1.2079, + "step": 18306 + }, + { + "epoch": 1.54, + "learning_rate": 1.003908574160817e-05, + "loss": 1.1895, + "step": 18307 + }, + { + "epoch": 1.54, + "learning_rate": 1.0038176775341675e-05, + "loss": 1.2612, + "step": 18308 + }, + { + "epoch": 1.54, + "learning_rate": 1.0037267808759753e-05, + "loss": 1.1703, + "step": 18309 + }, + { + "epoch": 1.54, + "learning_rate": 1.0036358841869908e-05, + "loss": 1.2108, + "step": 18310 + }, + { + "epoch": 1.54, + "learning_rate": 1.0035449874679657e-05, + "loss": 1.176, + "step": 18311 + }, + { + "epoch": 1.54, + "learning_rate": 1.0034540907196506e-05, + "loss": 1.183, + "step": 18312 + }, + { + "epoch": 1.54, + "learning_rate": 1.003363193942797e-05, + "loss": 1.2497, + "step": 18313 + }, + { + "epoch": 1.54, + "learning_rate": 1.0032722971381555e-05, + "loss": 1.145, + "step": 18314 + }, + { + "epoch": 1.54, + "learning_rate": 1.003181400306477e-05, + "loss": 1.222, + "step": 18315 + }, + { + "epoch": 1.54, + "learning_rate": 1.003090503448513e-05, + "loss": 1.2383, + "step": 18316 + }, + { + "epoch": 1.54, + "learning_rate": 1.0029996065650142e-05, + "loss": 1.179, + "step": 18317 + }, + { + "epoch": 1.54, + "learning_rate": 1.0029087096567319e-05, + "loss": 1.1716, + "step": 18318 + }, + { + "epoch": 1.54, + "learning_rate": 1.0028178127244164e-05, + "loss": 1.3028, + "step": 18319 + }, + { + "epoch": 1.54, + "learning_rate": 1.0027269157688195e-05, + "loss": 1.1702, + "step": 18320 + }, + { + "epoch": 1.54, + "learning_rate": 1.002636018790692e-05, + "loss": 1.1845, + "step": 18321 + }, + { + "epoch": 1.54, + "learning_rate": 1.0025451217907849e-05, + "loss": 1.1554, + "step": 18322 + }, + { + "epoch": 1.54, + "learning_rate": 1.002454224769849e-05, + "loss": 1.1337, + "step": 18323 + }, + { + "epoch": 1.54, + "learning_rate": 1.002363327728636e-05, + "loss": 1.1966, + "step": 18324 + }, + { + "epoch": 1.54, + "learning_rate": 1.002272430667896e-05, + "loss": 1.1787, + "step": 18325 + }, + { + "epoch": 1.54, + "learning_rate": 1.0021815335883804e-05, + "loss": 1.1745, + "step": 18326 + }, + { + "epoch": 1.54, + "learning_rate": 1.0020906364908406e-05, + "loss": 1.3619, + "step": 18327 + }, + { + "epoch": 1.54, + "learning_rate": 1.001999739376027e-05, + "loss": 1.0851, + "step": 18328 + }, + { + "epoch": 1.54, + "learning_rate": 1.0019088422446909e-05, + "loss": 1.1627, + "step": 18329 + }, + { + "epoch": 1.54, + "learning_rate": 1.0018179450975836e-05, + "loss": 1.128, + "step": 18330 + }, + { + "epoch": 1.54, + "learning_rate": 1.0017270479354557e-05, + "loss": 1.2086, + "step": 18331 + }, + { + "epoch": 1.54, + "learning_rate": 1.0016361507590582e-05, + "loss": 1.2638, + "step": 18332 + }, + { + "epoch": 1.54, + "learning_rate": 1.0015452535691423e-05, + "loss": 1.1497, + "step": 18333 + }, + { + "epoch": 1.54, + "learning_rate": 1.0014543563664593e-05, + "loss": 1.2227, + "step": 18334 + }, + { + "epoch": 1.54, + "learning_rate": 1.00136345915176e-05, + "loss": 1.2559, + "step": 18335 + }, + { + "epoch": 1.54, + "learning_rate": 1.0012725619257951e-05, + "loss": 1.1827, + "step": 18336 + }, + { + "epoch": 1.54, + "learning_rate": 1.001181664689316e-05, + "loss": 1.1448, + "step": 18337 + }, + { + "epoch": 1.54, + "learning_rate": 1.0010907674430736e-05, + "loss": 1.1499, + "step": 18338 + }, + { + "epoch": 1.54, + "learning_rate": 1.000999870187819e-05, + "loss": 1.1816, + "step": 18339 + }, + { + "epoch": 1.54, + "learning_rate": 1.0009089729243031e-05, + "loss": 1.2951, + "step": 18340 + }, + { + "epoch": 1.54, + "learning_rate": 1.0008180756532768e-05, + "loss": 1.0876, + "step": 18341 + }, + { + "epoch": 1.54, + "learning_rate": 1.0007271783754918e-05, + "loss": 1.2271, + "step": 18342 + }, + { + "epoch": 1.54, + "learning_rate": 1.0006362810916981e-05, + "loss": 1.1877, + "step": 18343 + }, + { + "epoch": 1.54, + "learning_rate": 1.000545383802647e-05, + "loss": 1.1876, + "step": 18344 + }, + { + "epoch": 1.54, + "learning_rate": 1.0004544865090906e-05, + "loss": 1.1744, + "step": 18345 + }, + { + "epoch": 1.54, + "learning_rate": 1.0003635892117785e-05, + "loss": 1.0966, + "step": 18346 + }, + { + "epoch": 1.54, + "learning_rate": 1.0002726919114625e-05, + "loss": 1.163, + "step": 18347 + }, + { + "epoch": 1.54, + "learning_rate": 1.0001817946088934e-05, + "loss": 1.1724, + "step": 18348 + }, + { + "epoch": 1.54, + "learning_rate": 1.0000908973048224e-05, + "loss": 1.259, + "step": 18349 + }, + { + "epoch": 1.54, + "learning_rate": 1e-05, + "loss": 1.1361, + "step": 18350 + }, + { + "epoch": 1.55, + "learning_rate": 9.999091026951778e-06, + "loss": 1.1473, + "step": 18351 + }, + { + "epoch": 1.55, + "learning_rate": 9.99818205391107e-06, + "loss": 1.1341, + "step": 18352 + }, + { + "epoch": 1.55, + "learning_rate": 9.997273080885376e-06, + "loss": 1.2552, + "step": 18353 + }, + { + "epoch": 1.55, + "learning_rate": 9.996364107882217e-06, + "loss": 1.2633, + "step": 18354 + }, + { + "epoch": 1.55, + "learning_rate": 9.995455134909097e-06, + "loss": 1.1581, + "step": 18355 + }, + { + "epoch": 1.55, + "learning_rate": 9.994546161973531e-06, + "loss": 1.154, + "step": 18356 + }, + { + "epoch": 1.55, + "learning_rate": 9.993637189083022e-06, + "loss": 1.1775, + "step": 18357 + }, + { + "epoch": 1.55, + "learning_rate": 9.992728216245087e-06, + "loss": 1.2156, + "step": 18358 + }, + { + "epoch": 1.55, + "learning_rate": 9.991819243467234e-06, + "loss": 1.2073, + "step": 18359 + }, + { + "epoch": 1.55, + "learning_rate": 9.990910270756975e-06, + "loss": 1.2359, + "step": 18360 + }, + { + "epoch": 1.55, + "learning_rate": 9.990001298121813e-06, + "loss": 1.2053, + "step": 18361 + }, + { + "epoch": 1.55, + "learning_rate": 9.989092325569267e-06, + "loss": 1.1907, + "step": 18362 + }, + { + "epoch": 1.55, + "learning_rate": 9.988183353106843e-06, + "loss": 1.223, + "step": 18363 + }, + { + "epoch": 1.55, + "learning_rate": 9.987274380742049e-06, + "loss": 1.2054, + "step": 18364 + }, + { + "epoch": 1.55, + "learning_rate": 9.986365408482404e-06, + "loss": 1.1806, + "step": 18365 + }, + { + "epoch": 1.55, + "learning_rate": 9.985456436335408e-06, + "loss": 1.2117, + "step": 18366 + }, + { + "epoch": 1.55, + "learning_rate": 9.984547464308578e-06, + "loss": 1.1447, + "step": 18367 + }, + { + "epoch": 1.55, + "learning_rate": 9.983638492409418e-06, + "loss": 1.319, + "step": 18368 + }, + { + "epoch": 1.55, + "learning_rate": 9.982729520645448e-06, + "loss": 1.1962, + "step": 18369 + }, + { + "epoch": 1.55, + "learning_rate": 9.981820549024169e-06, + "loss": 1.2923, + "step": 18370 + }, + { + "epoch": 1.55, + "learning_rate": 9.980911577553096e-06, + "loss": 1.1375, + "step": 18371 + }, + { + "epoch": 1.55, + "learning_rate": 9.980002606239732e-06, + "loss": 1.2019, + "step": 18372 + }, + { + "epoch": 1.55, + "learning_rate": 9.979093635091599e-06, + "loss": 1.2478, + "step": 18373 + }, + { + "epoch": 1.55, + "learning_rate": 9.978184664116198e-06, + "loss": 1.2177, + "step": 18374 + }, + { + "epoch": 1.55, + "learning_rate": 9.977275693321042e-06, + "loss": 1.2381, + "step": 18375 + }, + { + "epoch": 1.55, + "learning_rate": 9.976366722713644e-06, + "loss": 1.2315, + "step": 18376 + }, + { + "epoch": 1.55, + "learning_rate": 9.97545775230151e-06, + "loss": 1.1387, + "step": 18377 + }, + { + "epoch": 1.55, + "learning_rate": 9.974548782092153e-06, + "loss": 1.1906, + "step": 18378 + }, + { + "epoch": 1.55, + "learning_rate": 9.97363981209308e-06, + "loss": 1.1173, + "step": 18379 + }, + { + "epoch": 1.55, + "learning_rate": 9.972730842311806e-06, + "loss": 1.1831, + "step": 18380 + }, + { + "epoch": 1.55, + "learning_rate": 9.971821872755838e-06, + "loss": 1.1881, + "step": 18381 + }, + { + "epoch": 1.55, + "learning_rate": 9.970912903432686e-06, + "loss": 1.2441, + "step": 18382 + }, + { + "epoch": 1.55, + "learning_rate": 9.97000393434986e-06, + "loss": 1.1465, + "step": 18383 + }, + { + "epoch": 1.55, + "learning_rate": 9.969094965514873e-06, + "loss": 1.1847, + "step": 18384 + }, + { + "epoch": 1.55, + "learning_rate": 9.968185996935232e-06, + "loss": 1.2449, + "step": 18385 + }, + { + "epoch": 1.55, + "learning_rate": 9.96727702861845e-06, + "loss": 1.2012, + "step": 18386 + }, + { + "epoch": 1.55, + "learning_rate": 9.966368060572033e-06, + "loss": 1.2181, + "step": 18387 + }, + { + "epoch": 1.55, + "learning_rate": 9.965459092803494e-06, + "loss": 1.2693, + "step": 18388 + }, + { + "epoch": 1.55, + "learning_rate": 9.964550125320347e-06, + "loss": 1.2417, + "step": 18389 + }, + { + "epoch": 1.55, + "learning_rate": 9.963641158130092e-06, + "loss": 1.1962, + "step": 18390 + }, + { + "epoch": 1.55, + "learning_rate": 9.96273219124025e-06, + "loss": 1.2578, + "step": 18391 + }, + { + "epoch": 1.55, + "learning_rate": 9.961823224658326e-06, + "loss": 1.2459, + "step": 18392 + }, + { + "epoch": 1.55, + "learning_rate": 9.960914258391832e-06, + "loss": 1.2725, + "step": 18393 + }, + { + "epoch": 1.55, + "learning_rate": 9.960005292448272e-06, + "loss": 1.2036, + "step": 18394 + }, + { + "epoch": 1.55, + "learning_rate": 9.959096326835166e-06, + "loss": 1.146, + "step": 18395 + }, + { + "epoch": 1.55, + "learning_rate": 9.958187361560018e-06, + "loss": 1.1946, + "step": 18396 + }, + { + "epoch": 1.55, + "learning_rate": 9.95727839663034e-06, + "loss": 1.2075, + "step": 18397 + }, + { + "epoch": 1.55, + "learning_rate": 9.956369432053638e-06, + "loss": 1.2208, + "step": 18398 + }, + { + "epoch": 1.55, + "learning_rate": 9.955460467837432e-06, + "loss": 1.2576, + "step": 18399 + }, + { + "epoch": 1.55, + "learning_rate": 9.954551503989219e-06, + "loss": 1.1784, + "step": 18400 + }, + { + "epoch": 1.55, + "learning_rate": 9.953642540516515e-06, + "loss": 1.0854, + "step": 18401 + }, + { + "epoch": 1.55, + "learning_rate": 9.95273357742684e-06, + "loss": 1.2441, + "step": 18402 + }, + { + "epoch": 1.55, + "learning_rate": 9.95182461472769e-06, + "loss": 1.1573, + "step": 18403 + }, + { + "epoch": 1.55, + "learning_rate": 9.950915652426581e-06, + "loss": 1.2419, + "step": 18404 + }, + { + "epoch": 1.55, + "learning_rate": 9.950006690531023e-06, + "loss": 1.1462, + "step": 18405 + }, + { + "epoch": 1.55, + "learning_rate": 9.949097729048526e-06, + "loss": 1.2325, + "step": 18406 + }, + { + "epoch": 1.55, + "learning_rate": 9.948188767986596e-06, + "loss": 1.1713, + "step": 18407 + }, + { + "epoch": 1.55, + "learning_rate": 9.947279807352752e-06, + "loss": 1.2479, + "step": 18408 + }, + { + "epoch": 1.55, + "learning_rate": 9.946370847154498e-06, + "loss": 1.1143, + "step": 18409 + }, + { + "epoch": 1.55, + "learning_rate": 9.945461887399346e-06, + "loss": 1.247, + "step": 18410 + }, + { + "epoch": 1.55, + "learning_rate": 9.9445529280948e-06, + "loss": 1.1666, + "step": 18411 + }, + { + "epoch": 1.55, + "learning_rate": 9.94364396924838e-06, + "loss": 1.2352, + "step": 18412 + }, + { + "epoch": 1.55, + "learning_rate": 9.942735010867593e-06, + "loss": 1.045, + "step": 18413 + }, + { + "epoch": 1.55, + "learning_rate": 9.941826052959942e-06, + "loss": 1.1248, + "step": 18414 + }, + { + "epoch": 1.55, + "learning_rate": 9.940917095532948e-06, + "loss": 1.1113, + "step": 18415 + }, + { + "epoch": 1.55, + "learning_rate": 9.940008138594113e-06, + "loss": 1.1678, + "step": 18416 + }, + { + "epoch": 1.55, + "learning_rate": 9.939099182150952e-06, + "loss": 1.161, + "step": 18417 + }, + { + "epoch": 1.55, + "learning_rate": 9.93819022621097e-06, + "loss": 1.1177, + "step": 18418 + }, + { + "epoch": 1.55, + "learning_rate": 9.937281270781685e-06, + "loss": 1.1431, + "step": 18419 + }, + { + "epoch": 1.55, + "learning_rate": 9.936372315870599e-06, + "loss": 1.2023, + "step": 18420 + }, + { + "epoch": 1.55, + "learning_rate": 9.935463361485227e-06, + "loss": 1.1709, + "step": 18421 + }, + { + "epoch": 1.55, + "learning_rate": 9.934554407633073e-06, + "loss": 1.1992, + "step": 18422 + }, + { + "epoch": 1.55, + "learning_rate": 9.933645454321657e-06, + "loss": 1.1904, + "step": 18423 + }, + { + "epoch": 1.55, + "learning_rate": 9.93273650155848e-06, + "loss": 1.2422, + "step": 18424 + }, + { + "epoch": 1.55, + "learning_rate": 9.931827549351054e-06, + "loss": 1.2753, + "step": 18425 + }, + { + "epoch": 1.55, + "learning_rate": 9.930918597706894e-06, + "loss": 1.1725, + "step": 18426 + }, + { + "epoch": 1.55, + "learning_rate": 9.930009646633506e-06, + "loss": 1.1329, + "step": 18427 + }, + { + "epoch": 1.55, + "learning_rate": 9.929100696138403e-06, + "loss": 1.1883, + "step": 18428 + }, + { + "epoch": 1.55, + "learning_rate": 9.928191746229087e-06, + "loss": 1.1921, + "step": 18429 + }, + { + "epoch": 1.55, + "learning_rate": 9.927282796913079e-06, + "loss": 1.2127, + "step": 18430 + }, + { + "epoch": 1.55, + "learning_rate": 9.926373848197882e-06, + "loss": 1.1869, + "step": 18431 + }, + { + "epoch": 1.55, + "learning_rate": 9.925464900091008e-06, + "loss": 1.1872, + "step": 18432 + }, + { + "epoch": 1.55, + "learning_rate": 9.924555952599965e-06, + "loss": 1.2231, + "step": 18433 + }, + { + "epoch": 1.55, + "learning_rate": 9.923647005732268e-06, + "loss": 1.1593, + "step": 18434 + }, + { + "epoch": 1.55, + "learning_rate": 9.922738059495421e-06, + "loss": 1.1375, + "step": 18435 + }, + { + "epoch": 1.55, + "learning_rate": 9.92182911389694e-06, + "loss": 1.2034, + "step": 18436 + }, + { + "epoch": 1.55, + "learning_rate": 9.920920168944328e-06, + "loss": 1.1835, + "step": 18437 + }, + { + "epoch": 1.55, + "learning_rate": 9.9200112246451e-06, + "loss": 1.201, + "step": 18438 + }, + { + "epoch": 1.55, + "learning_rate": 9.919102281006767e-06, + "loss": 1.1435, + "step": 18439 + }, + { + "epoch": 1.55, + "learning_rate": 9.918193338036833e-06, + "loss": 1.2096, + "step": 18440 + }, + { + "epoch": 1.55, + "learning_rate": 9.917284395742816e-06, + "loss": 1.2058, + "step": 18441 + }, + { + "epoch": 1.55, + "learning_rate": 9.91637545413222e-06, + "loss": 1.1806, + "step": 18442 + }, + { + "epoch": 1.55, + "learning_rate": 9.915466513212559e-06, + "loss": 1.0758, + "step": 18443 + }, + { + "epoch": 1.55, + "learning_rate": 9.914557572991335e-06, + "loss": 1.1652, + "step": 18444 + }, + { + "epoch": 1.55, + "learning_rate": 9.91364863347607e-06, + "loss": 1.1812, + "step": 18445 + }, + { + "epoch": 1.55, + "learning_rate": 9.912739694674266e-06, + "loss": 1.1585, + "step": 18446 + }, + { + "epoch": 1.55, + "learning_rate": 9.911830756593435e-06, + "loss": 1.1318, + "step": 18447 + }, + { + "epoch": 1.55, + "learning_rate": 9.910921819241082e-06, + "loss": 1.1874, + "step": 18448 + }, + { + "epoch": 1.55, + "learning_rate": 9.910012882624726e-06, + "loss": 1.1387, + "step": 18449 + }, + { + "epoch": 1.55, + "learning_rate": 9.90910394675187e-06, + "loss": 1.222, + "step": 18450 + }, + { + "epoch": 1.55, + "learning_rate": 9.908195011630026e-06, + "loss": 1.1063, + "step": 18451 + }, + { + "epoch": 1.55, + "learning_rate": 9.907286077266705e-06, + "loss": 1.1981, + "step": 18452 + }, + { + "epoch": 1.55, + "learning_rate": 9.906377143669416e-06, + "loss": 1.1841, + "step": 18453 + }, + { + "epoch": 1.55, + "learning_rate": 9.905468210845672e-06, + "loss": 1.1992, + "step": 18454 + }, + { + "epoch": 1.55, + "learning_rate": 9.904559278802974e-06, + "loss": 1.2095, + "step": 18455 + }, + { + "epoch": 1.55, + "learning_rate": 9.903650347548846e-06, + "loss": 1.3191, + "step": 18456 + }, + { + "epoch": 1.55, + "learning_rate": 9.902741417090781e-06, + "loss": 1.2152, + "step": 18457 + }, + { + "epoch": 1.55, + "learning_rate": 9.901832487436304e-06, + "loss": 1.1066, + "step": 18458 + }, + { + "epoch": 1.55, + "learning_rate": 9.900923558592916e-06, + "loss": 1.1664, + "step": 18459 + }, + { + "epoch": 1.55, + "learning_rate": 9.90001463056813e-06, + "loss": 1.0977, + "step": 18460 + }, + { + "epoch": 1.55, + "learning_rate": 9.899105703369451e-06, + "loss": 1.2249, + "step": 18461 + }, + { + "epoch": 1.55, + "learning_rate": 9.898196777004399e-06, + "loss": 1.189, + "step": 18462 + }, + { + "epoch": 1.55, + "learning_rate": 9.897287851480474e-06, + "loss": 1.1609, + "step": 18463 + }, + { + "epoch": 1.55, + "learning_rate": 9.896378926805188e-06, + "loss": 1.2585, + "step": 18464 + }, + { + "epoch": 1.55, + "learning_rate": 9.895470002986055e-06, + "loss": 1.1128, + "step": 18465 + }, + { + "epoch": 1.55, + "learning_rate": 9.894561080030583e-06, + "loss": 1.1767, + "step": 18466 + }, + { + "epoch": 1.55, + "learning_rate": 9.893652157946283e-06, + "loss": 1.2161, + "step": 18467 + }, + { + "epoch": 1.55, + "learning_rate": 9.892743236740657e-06, + "loss": 1.1849, + "step": 18468 + }, + { + "epoch": 1.55, + "learning_rate": 9.891834316421226e-06, + "loss": 1.2228, + "step": 18469 + }, + { + "epoch": 1.56, + "learning_rate": 9.890925396995491e-06, + "loss": 1.2235, + "step": 18470 + }, + { + "epoch": 1.56, + "learning_rate": 9.890016478470969e-06, + "loss": 1.1879, + "step": 18471 + }, + { + "epoch": 1.56, + "learning_rate": 9.889107560855162e-06, + "loss": 1.1436, + "step": 18472 + }, + { + "epoch": 1.56, + "learning_rate": 9.888198644155587e-06, + "loss": 1.2469, + "step": 18473 + }, + { + "epoch": 1.56, + "learning_rate": 9.887289728379748e-06, + "loss": 1.2102, + "step": 18474 + }, + { + "epoch": 1.56, + "learning_rate": 9.886380813535157e-06, + "loss": 1.2255, + "step": 18475 + }, + { + "epoch": 1.56, + "learning_rate": 9.885471899629327e-06, + "loss": 1.1855, + "step": 18476 + }, + { + "epoch": 1.56, + "learning_rate": 9.884562986669764e-06, + "loss": 1.2312, + "step": 18477 + }, + { + "epoch": 1.56, + "learning_rate": 9.88365407466398e-06, + "loss": 1.1216, + "step": 18478 + }, + { + "epoch": 1.56, + "learning_rate": 9.882745163619478e-06, + "loss": 1.1509, + "step": 18479 + }, + { + "epoch": 1.56, + "learning_rate": 9.88183625354378e-06, + "loss": 1.1056, + "step": 18480 + }, + { + "epoch": 1.56, + "learning_rate": 9.880927344444382e-06, + "loss": 1.1696, + "step": 18481 + }, + { + "epoch": 1.56, + "learning_rate": 9.880018436328805e-06, + "loss": 1.1319, + "step": 18482 + }, + { + "epoch": 1.56, + "learning_rate": 9.87910952920455e-06, + "loss": 1.1256, + "step": 18483 + }, + { + "epoch": 1.56, + "learning_rate": 9.878200623079135e-06, + "loss": 1.254, + "step": 18484 + }, + { + "epoch": 1.56, + "learning_rate": 9.877291717960064e-06, + "loss": 1.1964, + "step": 18485 + }, + { + "epoch": 1.56, + "learning_rate": 9.87638281385485e-06, + "loss": 1.211, + "step": 18486 + }, + { + "epoch": 1.56, + "learning_rate": 9.875473910770996e-06, + "loss": 1.2967, + "step": 18487 + }, + { + "epoch": 1.56, + "learning_rate": 9.874565008716018e-06, + "loss": 1.2013, + "step": 18488 + }, + { + "epoch": 1.56, + "learning_rate": 9.873656107697425e-06, + "loss": 1.1478, + "step": 18489 + }, + { + "epoch": 1.56, + "learning_rate": 9.872747207722724e-06, + "loss": 1.1876, + "step": 18490 + }, + { + "epoch": 1.56, + "learning_rate": 9.87183830879943e-06, + "loss": 1.2171, + "step": 18491 + }, + { + "epoch": 1.56, + "learning_rate": 9.870929410935046e-06, + "loss": 1.2057, + "step": 18492 + }, + { + "epoch": 1.56, + "learning_rate": 9.870020514137088e-06, + "loss": 1.2403, + "step": 18493 + }, + { + "epoch": 1.56, + "learning_rate": 9.869111618413056e-06, + "loss": 1.1789, + "step": 18494 + }, + { + "epoch": 1.56, + "learning_rate": 9.86820272377047e-06, + "loss": 1.105, + "step": 18495 + }, + { + "epoch": 1.56, + "learning_rate": 9.867293830216835e-06, + "loss": 1.1657, + "step": 18496 + }, + { + "epoch": 1.56, + "learning_rate": 9.866384937759661e-06, + "loss": 1.1964, + "step": 18497 + }, + { + "epoch": 1.56, + "learning_rate": 9.865476046406455e-06, + "loss": 1.1541, + "step": 18498 + }, + { + "epoch": 1.56, + "learning_rate": 9.864567156164732e-06, + "loss": 1.2699, + "step": 18499 + }, + { + "epoch": 1.56, + "learning_rate": 9.863658267041997e-06, + "loss": 1.1155, + "step": 18500 + }, + { + "epoch": 1.56, + "learning_rate": 9.862749379045759e-06, + "loss": 1.2113, + "step": 18501 + }, + { + "epoch": 1.56, + "learning_rate": 9.861840492183533e-06, + "loss": 1.086, + "step": 18502 + }, + { + "epoch": 1.56, + "learning_rate": 9.860931606462824e-06, + "loss": 1.204, + "step": 18503 + }, + { + "epoch": 1.56, + "learning_rate": 9.860022721891143e-06, + "loss": 1.147, + "step": 18504 + }, + { + "epoch": 1.56, + "learning_rate": 9.859113838475996e-06, + "loss": 1.1886, + "step": 18505 + }, + { + "epoch": 1.56, + "learning_rate": 9.858204956224902e-06, + "loss": 1.204, + "step": 18506 + }, + { + "epoch": 1.56, + "learning_rate": 9.857296075145357e-06, + "loss": 1.1956, + "step": 18507 + }, + { + "epoch": 1.56, + "learning_rate": 9.856387195244883e-06, + "loss": 1.1763, + "step": 18508 + }, + { + "epoch": 1.56, + "learning_rate": 9.855478316530982e-06, + "loss": 1.1522, + "step": 18509 + }, + { + "epoch": 1.56, + "learning_rate": 9.854569439011165e-06, + "loss": 1.3064, + "step": 18510 + }, + { + "epoch": 1.56, + "learning_rate": 9.853660562692939e-06, + "loss": 1.1705, + "step": 18511 + }, + { + "epoch": 1.56, + "learning_rate": 9.852751687583822e-06, + "loss": 1.1377, + "step": 18512 + }, + { + "epoch": 1.56, + "learning_rate": 9.851842813691314e-06, + "loss": 1.2634, + "step": 18513 + }, + { + "epoch": 1.56, + "learning_rate": 9.850933941022926e-06, + "loss": 1.2353, + "step": 18514 + }, + { + "epoch": 1.56, + "learning_rate": 9.850025069586173e-06, + "loss": 1.1569, + "step": 18515 + }, + { + "epoch": 1.56, + "learning_rate": 9.84911619938856e-06, + "loss": 1.1199, + "step": 18516 + }, + { + "epoch": 1.56, + "learning_rate": 9.8482073304376e-06, + "loss": 1.1437, + "step": 18517 + }, + { + "epoch": 1.56, + "learning_rate": 9.847298462740793e-06, + "loss": 1.1681, + "step": 18518 + }, + { + "epoch": 1.56, + "learning_rate": 9.846389596305661e-06, + "loss": 1.2839, + "step": 18519 + }, + { + "epoch": 1.56, + "learning_rate": 9.845480731139705e-06, + "loss": 1.234, + "step": 18520 + }, + { + "epoch": 1.56, + "learning_rate": 9.844571867250439e-06, + "loss": 1.1251, + "step": 18521 + }, + { + "epoch": 1.56, + "learning_rate": 9.843663004645365e-06, + "loss": 1.1174, + "step": 18522 + }, + { + "epoch": 1.56, + "learning_rate": 9.842754143332002e-06, + "loss": 1.1964, + "step": 18523 + }, + { + "epoch": 1.56, + "learning_rate": 9.841845283317853e-06, + "loss": 1.2072, + "step": 18524 + }, + { + "epoch": 1.56, + "learning_rate": 9.840936424610426e-06, + "loss": 1.2043, + "step": 18525 + }, + { + "epoch": 1.56, + "learning_rate": 9.840027567217236e-06, + "loss": 1.1695, + "step": 18526 + }, + { + "epoch": 1.56, + "learning_rate": 9.839118711145789e-06, + "loss": 1.2501, + "step": 18527 + }, + { + "epoch": 1.56, + "learning_rate": 9.838209856403598e-06, + "loss": 1.3138, + "step": 18528 + }, + { + "epoch": 1.56, + "learning_rate": 9.837301002998164e-06, + "loss": 1.1791, + "step": 18529 + }, + { + "epoch": 1.56, + "learning_rate": 9.836392150937003e-06, + "loss": 1.0898, + "step": 18530 + }, + { + "epoch": 1.56, + "learning_rate": 9.835483300227622e-06, + "loss": 1.1223, + "step": 18531 + }, + { + "epoch": 1.56, + "learning_rate": 9.834574450877533e-06, + "loss": 1.1703, + "step": 18532 + }, + { + "epoch": 1.56, + "learning_rate": 9.83366560289424e-06, + "loss": 1.1972, + "step": 18533 + }, + { + "epoch": 1.56, + "learning_rate": 9.832756756285257e-06, + "loss": 1.1898, + "step": 18534 + }, + { + "epoch": 1.56, + "learning_rate": 9.831847911058089e-06, + "loss": 1.2529, + "step": 18535 + }, + { + "epoch": 1.56, + "learning_rate": 9.830939067220251e-06, + "loss": 1.1831, + "step": 18536 + }, + { + "epoch": 1.56, + "learning_rate": 9.830030224779242e-06, + "loss": 1.117, + "step": 18537 + }, + { + "epoch": 1.56, + "learning_rate": 9.829121383742583e-06, + "loss": 1.2833, + "step": 18538 + }, + { + "epoch": 1.56, + "learning_rate": 9.828212544117778e-06, + "loss": 1.2394, + "step": 18539 + }, + { + "epoch": 1.56, + "learning_rate": 9.827303705912332e-06, + "loss": 1.1065, + "step": 18540 + }, + { + "epoch": 1.56, + "learning_rate": 9.826394869133761e-06, + "loss": 1.2213, + "step": 18541 + }, + { + "epoch": 1.56, + "learning_rate": 9.82548603378957e-06, + "loss": 1.206, + "step": 18542 + }, + { + "epoch": 1.56, + "learning_rate": 9.824577199887273e-06, + "loss": 1.2432, + "step": 18543 + }, + { + "epoch": 1.56, + "learning_rate": 9.82366836743437e-06, + "loss": 1.0822, + "step": 18544 + }, + { + "epoch": 1.56, + "learning_rate": 9.82275953643838e-06, + "loss": 1.0921, + "step": 18545 + }, + { + "epoch": 1.56, + "learning_rate": 9.821850706906804e-06, + "loss": 1.2029, + "step": 18546 + }, + { + "epoch": 1.56, + "learning_rate": 9.820941878847157e-06, + "loss": 1.1044, + "step": 18547 + }, + { + "epoch": 1.56, + "learning_rate": 9.82003305226694e-06, + "loss": 1.2657, + "step": 18548 + }, + { + "epoch": 1.56, + "learning_rate": 9.819124227173674e-06, + "loss": 1.1792, + "step": 18549 + }, + { + "epoch": 1.56, + "learning_rate": 9.81821540357486e-06, + "loss": 1.2479, + "step": 18550 + }, + { + "epoch": 1.56, + "learning_rate": 9.817306581478004e-06, + "loss": 1.1576, + "step": 18551 + }, + { + "epoch": 1.56, + "learning_rate": 9.816397760890625e-06, + "loss": 1.197, + "step": 18552 + }, + { + "epoch": 1.56, + "learning_rate": 9.815488941820223e-06, + "loss": 1.2425, + "step": 18553 + }, + { + "epoch": 1.56, + "learning_rate": 9.814580124274315e-06, + "loss": 1.2079, + "step": 18554 + }, + { + "epoch": 1.56, + "learning_rate": 9.8136713082604e-06, + "loss": 1.1592, + "step": 18555 + }, + { + "epoch": 1.56, + "learning_rate": 9.812762493785998e-06, + "loss": 1.2523, + "step": 18556 + }, + { + "epoch": 1.56, + "learning_rate": 9.811853680858607e-06, + "loss": 1.1215, + "step": 18557 + }, + { + "epoch": 1.56, + "learning_rate": 9.810944869485744e-06, + "loss": 1.2398, + "step": 18558 + }, + { + "epoch": 1.56, + "learning_rate": 9.810036059674914e-06, + "loss": 1.1337, + "step": 18559 + }, + { + "epoch": 1.56, + "learning_rate": 9.809127251433627e-06, + "loss": 1.2053, + "step": 18560 + }, + { + "epoch": 1.56, + "learning_rate": 9.80821844476939e-06, + "loss": 1.193, + "step": 18561 + }, + { + "epoch": 1.56, + "learning_rate": 9.807309639689717e-06, + "loss": 1.2046, + "step": 18562 + }, + { + "epoch": 1.56, + "learning_rate": 9.806400836202111e-06, + "loss": 1.1246, + "step": 18563 + }, + { + "epoch": 1.56, + "learning_rate": 9.805492034314082e-06, + "loss": 1.2388, + "step": 18564 + }, + { + "epoch": 1.56, + "learning_rate": 9.804583234033143e-06, + "loss": 1.1344, + "step": 18565 + }, + { + "epoch": 1.56, + "learning_rate": 9.803674435366798e-06, + "loss": 1.2113, + "step": 18566 + }, + { + "epoch": 1.56, + "learning_rate": 9.80276563832256e-06, + "loss": 1.2127, + "step": 18567 + }, + { + "epoch": 1.56, + "learning_rate": 9.801856842907932e-06, + "loss": 1.1523, + "step": 18568 + }, + { + "epoch": 1.56, + "learning_rate": 9.80094804913043e-06, + "loss": 1.165, + "step": 18569 + }, + { + "epoch": 1.56, + "learning_rate": 9.800039256997556e-06, + "loss": 1.1517, + "step": 18570 + }, + { + "epoch": 1.56, + "learning_rate": 9.799130466516824e-06, + "loss": 1.2436, + "step": 18571 + }, + { + "epoch": 1.56, + "learning_rate": 9.798221677695736e-06, + "loss": 1.2691, + "step": 18572 + }, + { + "epoch": 1.56, + "learning_rate": 9.79731289054181e-06, + "loss": 1.1285, + "step": 18573 + }, + { + "epoch": 1.56, + "learning_rate": 9.796404105062547e-06, + "loss": 1.1722, + "step": 18574 + }, + { + "epoch": 1.56, + "learning_rate": 9.795495321265458e-06, + "loss": 1.1756, + "step": 18575 + }, + { + "epoch": 1.56, + "learning_rate": 9.794586539158054e-06, + "loss": 1.2604, + "step": 18576 + }, + { + "epoch": 1.56, + "learning_rate": 9.793677758747841e-06, + "loss": 1.1727, + "step": 18577 + }, + { + "epoch": 1.56, + "learning_rate": 9.792768980042329e-06, + "loss": 1.1987, + "step": 18578 + }, + { + "epoch": 1.56, + "learning_rate": 9.791860203049025e-06, + "loss": 1.2432, + "step": 18579 + }, + { + "epoch": 1.56, + "learning_rate": 9.79095142777544e-06, + "loss": 1.167, + "step": 18580 + }, + { + "epoch": 1.56, + "learning_rate": 9.79004265422908e-06, + "loss": 1.175, + "step": 18581 + }, + { + "epoch": 1.56, + "learning_rate": 9.789133882417457e-06, + "loss": 1.2599, + "step": 18582 + }, + { + "epoch": 1.56, + "learning_rate": 9.788225112348072e-06, + "loss": 1.2344, + "step": 18583 + }, + { + "epoch": 1.56, + "learning_rate": 9.787316344028443e-06, + "loss": 1.2697, + "step": 18584 + }, + { + "epoch": 1.56, + "learning_rate": 9.786407577466075e-06, + "loss": 1.2541, + "step": 18585 + }, + { + "epoch": 1.56, + "learning_rate": 9.785498812668476e-06, + "loss": 1.1287, + "step": 18586 + }, + { + "epoch": 1.56, + "learning_rate": 9.78459004964315e-06, + "loss": 1.2227, + "step": 18587 + }, + { + "epoch": 1.56, + "learning_rate": 9.783681288397612e-06, + "loss": 1.2349, + "step": 18588 + }, + { + "epoch": 1.57, + "learning_rate": 9.782772528939372e-06, + "loss": 1.231, + "step": 18589 + }, + { + "epoch": 1.57, + "learning_rate": 9.78186377127593e-06, + "loss": 1.1312, + "step": 18590 + }, + { + "epoch": 1.57, + "learning_rate": 9.780955015414804e-06, + "loss": 1.1527, + "step": 18591 + }, + { + "epoch": 1.57, + "learning_rate": 9.780046261363495e-06, + "loss": 1.1021, + "step": 18592 + }, + { + "epoch": 1.57, + "learning_rate": 9.779137509129516e-06, + "loss": 1.1301, + "step": 18593 + }, + { + "epoch": 1.57, + "learning_rate": 9.77822875872037e-06, + "loss": 1.1662, + "step": 18594 + }, + { + "epoch": 1.57, + "learning_rate": 9.777320010143572e-06, + "loss": 1.1434, + "step": 18595 + }, + { + "epoch": 1.57, + "learning_rate": 9.776411263406627e-06, + "loss": 1.1572, + "step": 18596 + }, + { + "epoch": 1.57, + "learning_rate": 9.775502518517045e-06, + "loss": 1.1629, + "step": 18597 + }, + { + "epoch": 1.57, + "learning_rate": 9.774593775482328e-06, + "loss": 1.1585, + "step": 18598 + }, + { + "epoch": 1.57, + "learning_rate": 9.773685034309996e-06, + "loss": 1.1644, + "step": 18599 + }, + { + "epoch": 1.57, + "learning_rate": 9.772776295007547e-06, + "loss": 1.1982, + "step": 18600 + }, + { + "epoch": 1.57, + "learning_rate": 9.771867557582489e-06, + "loss": 1.1925, + "step": 18601 + }, + { + "epoch": 1.57, + "learning_rate": 9.770958822042342e-06, + "loss": 1.1412, + "step": 18602 + }, + { + "epoch": 1.57, + "learning_rate": 9.770050088394603e-06, + "loss": 1.1083, + "step": 18603 + }, + { + "epoch": 1.57, + "learning_rate": 9.769141356646785e-06, + "loss": 1.1232, + "step": 18604 + }, + { + "epoch": 1.57, + "learning_rate": 9.768232626806392e-06, + "loss": 1.25, + "step": 18605 + }, + { + "epoch": 1.57, + "learning_rate": 9.767323898880939e-06, + "loss": 1.2108, + "step": 18606 + }, + { + "epoch": 1.57, + "learning_rate": 9.766415172877927e-06, + "loss": 1.1845, + "step": 18607 + }, + { + "epoch": 1.57, + "learning_rate": 9.76550644880487e-06, + "loss": 1.1753, + "step": 18608 + }, + { + "epoch": 1.57, + "learning_rate": 9.764597726669272e-06, + "loss": 1.1703, + "step": 18609 + }, + { + "epoch": 1.57, + "learning_rate": 9.763689006478648e-06, + "loss": 1.2806, + "step": 18610 + }, + { + "epoch": 1.57, + "learning_rate": 9.762780288240493e-06, + "loss": 1.2509, + "step": 18611 + }, + { + "epoch": 1.57, + "learning_rate": 9.76187157196233e-06, + "loss": 1.2069, + "step": 18612 + }, + { + "epoch": 1.57, + "learning_rate": 9.760962857651655e-06, + "loss": 1.1746, + "step": 18613 + }, + { + "epoch": 1.57, + "learning_rate": 9.760054145315982e-06, + "loss": 1.1949, + "step": 18614 + }, + { + "epoch": 1.57, + "learning_rate": 9.75914543496282e-06, + "loss": 1.1735, + "step": 18615 + }, + { + "epoch": 1.57, + "learning_rate": 9.758236726599676e-06, + "loss": 1.2501, + "step": 18616 + }, + { + "epoch": 1.57, + "learning_rate": 9.75732802023406e-06, + "loss": 1.2767, + "step": 18617 + }, + { + "epoch": 1.57, + "learning_rate": 9.756419315873472e-06, + "loss": 1.2805, + "step": 18618 + }, + { + "epoch": 1.57, + "learning_rate": 9.75551061352543e-06, + "loss": 1.1786, + "step": 18619 + }, + { + "epoch": 1.57, + "learning_rate": 9.754601913197436e-06, + "loss": 1.1975, + "step": 18620 + }, + { + "epoch": 1.57, + "learning_rate": 9.753693214897e-06, + "loss": 1.1255, + "step": 18621 + }, + { + "epoch": 1.57, + "learning_rate": 9.752784518631627e-06, + "loss": 1.1664, + "step": 18622 + }, + { + "epoch": 1.57, + "learning_rate": 9.75187582440883e-06, + "loss": 1.1831, + "step": 18623 + }, + { + "epoch": 1.57, + "learning_rate": 9.750967132236114e-06, + "loss": 1.2063, + "step": 18624 + }, + { + "epoch": 1.57, + "learning_rate": 9.750058442120985e-06, + "loss": 1.1167, + "step": 18625 + }, + { + "epoch": 1.57, + "learning_rate": 9.749149754070956e-06, + "loss": 1.2083, + "step": 18626 + }, + { + "epoch": 1.57, + "learning_rate": 9.74824106809353e-06, + "loss": 1.195, + "step": 18627 + }, + { + "epoch": 1.57, + "learning_rate": 9.747332384196222e-06, + "loss": 1.1693, + "step": 18628 + }, + { + "epoch": 1.57, + "learning_rate": 9.746423702386527e-06, + "loss": 1.1638, + "step": 18629 + }, + { + "epoch": 1.57, + "learning_rate": 9.745515022671965e-06, + "loss": 1.2645, + "step": 18630 + }, + { + "epoch": 1.57, + "learning_rate": 9.744606345060038e-06, + "loss": 1.216, + "step": 18631 + }, + { + "epoch": 1.57, + "learning_rate": 9.743697669558257e-06, + "loss": 1.239, + "step": 18632 + }, + { + "epoch": 1.57, + "learning_rate": 9.742788996174126e-06, + "loss": 1.1673, + "step": 18633 + }, + { + "epoch": 1.57, + "learning_rate": 9.741880324915156e-06, + "loss": 1.1332, + "step": 18634 + }, + { + "epoch": 1.57, + "learning_rate": 9.740971655788852e-06, + "loss": 1.2192, + "step": 18635 + }, + { + "epoch": 1.57, + "learning_rate": 9.740062988802725e-06, + "loss": 1.1713, + "step": 18636 + }, + { + "epoch": 1.57, + "learning_rate": 9.739154323964277e-06, + "loss": 1.2066, + "step": 18637 + }, + { + "epoch": 1.57, + "learning_rate": 9.73824566128102e-06, + "loss": 1.3342, + "step": 18638 + }, + { + "epoch": 1.57, + "learning_rate": 9.737337000760466e-06, + "loss": 1.1831, + "step": 18639 + }, + { + "epoch": 1.57, + "learning_rate": 9.736428342410112e-06, + "loss": 1.1395, + "step": 18640 + }, + { + "epoch": 1.57, + "learning_rate": 9.735519686237475e-06, + "loss": 1.1172, + "step": 18641 + }, + { + "epoch": 1.57, + "learning_rate": 9.734611032250058e-06, + "loss": 1.1665, + "step": 18642 + }, + { + "epoch": 1.57, + "learning_rate": 9.733702380455372e-06, + "loss": 1.1741, + "step": 18643 + }, + { + "epoch": 1.57, + "learning_rate": 9.732793730860917e-06, + "loss": 1.1334, + "step": 18644 + }, + { + "epoch": 1.57, + "learning_rate": 9.731885083474209e-06, + "loss": 1.1895, + "step": 18645 + }, + { + "epoch": 1.57, + "learning_rate": 9.730976438302753e-06, + "loss": 1.2336, + "step": 18646 + }, + { + "epoch": 1.57, + "learning_rate": 9.730067795354056e-06, + "loss": 1.2536, + "step": 18647 + }, + { + "epoch": 1.57, + "learning_rate": 9.729159154635622e-06, + "loss": 1.1376, + "step": 18648 + }, + { + "epoch": 1.57, + "learning_rate": 9.728250516154964e-06, + "loss": 1.1849, + "step": 18649 + }, + { + "epoch": 1.57, + "learning_rate": 9.727341879919586e-06, + "loss": 1.1922, + "step": 18650 + }, + { + "epoch": 1.57, + "learning_rate": 9.726433245936995e-06, + "loss": 1.0931, + "step": 18651 + }, + { + "epoch": 1.57, + "learning_rate": 9.725524614214704e-06, + "loss": 1.2438, + "step": 18652 + }, + { + "epoch": 1.57, + "learning_rate": 9.724615984760216e-06, + "loss": 1.1823, + "step": 18653 + }, + { + "epoch": 1.57, + "learning_rate": 9.723707357581038e-06, + "loss": 1.2325, + "step": 18654 + }, + { + "epoch": 1.57, + "learning_rate": 9.722798732684676e-06, + "loss": 1.2567, + "step": 18655 + }, + { + "epoch": 1.57, + "learning_rate": 9.721890110078644e-06, + "loss": 1.164, + "step": 18656 + }, + { + "epoch": 1.57, + "learning_rate": 9.72098148977044e-06, + "loss": 1.147, + "step": 18657 + }, + { + "epoch": 1.57, + "learning_rate": 9.720072871767582e-06, + "loss": 1.2234, + "step": 18658 + }, + { + "epoch": 1.57, + "learning_rate": 9.719164256077565e-06, + "loss": 1.1468, + "step": 18659 + }, + { + "epoch": 1.57, + "learning_rate": 9.71825564270791e-06, + "loss": 1.1947, + "step": 18660 + }, + { + "epoch": 1.57, + "learning_rate": 9.717347031666111e-06, + "loss": 1.2652, + "step": 18661 + }, + { + "epoch": 1.57, + "learning_rate": 9.716438422959685e-06, + "loss": 1.1171, + "step": 18662 + }, + { + "epoch": 1.57, + "learning_rate": 9.715529816596134e-06, + "loss": 1.1656, + "step": 18663 + }, + { + "epoch": 1.57, + "learning_rate": 9.714621212582965e-06, + "loss": 1.1801, + "step": 18664 + }, + { + "epoch": 1.57, + "learning_rate": 9.71371261092769e-06, + "loss": 1.2103, + "step": 18665 + }, + { + "epoch": 1.57, + "learning_rate": 9.712804011637811e-06, + "loss": 1.1576, + "step": 18666 + }, + { + "epoch": 1.57, + "learning_rate": 9.711895414720838e-06, + "loss": 1.2677, + "step": 18667 + }, + { + "epoch": 1.57, + "learning_rate": 9.710986820184275e-06, + "loss": 1.2327, + "step": 18668 + }, + { + "epoch": 1.57, + "learning_rate": 9.710078228035635e-06, + "loss": 1.14, + "step": 18669 + }, + { + "epoch": 1.57, + "learning_rate": 9.70916963828242e-06, + "loss": 1.1539, + "step": 18670 + }, + { + "epoch": 1.57, + "learning_rate": 9.70826105093214e-06, + "loss": 1.3471, + "step": 18671 + }, + { + "epoch": 1.57, + "learning_rate": 9.707352465992298e-06, + "loss": 1.2123, + "step": 18672 + }, + { + "epoch": 1.57, + "learning_rate": 9.706443883470404e-06, + "loss": 1.1488, + "step": 18673 + }, + { + "epoch": 1.57, + "learning_rate": 9.705535303373965e-06, + "loss": 1.2903, + "step": 18674 + }, + { + "epoch": 1.57, + "learning_rate": 9.704626725710485e-06, + "loss": 1.0988, + "step": 18675 + }, + { + "epoch": 1.57, + "learning_rate": 9.70371815048748e-06, + "loss": 1.1821, + "step": 18676 + }, + { + "epoch": 1.57, + "learning_rate": 9.702809577712444e-06, + "loss": 1.2446, + "step": 18677 + }, + { + "epoch": 1.57, + "learning_rate": 9.701901007392896e-06, + "loss": 1.2679, + "step": 18678 + }, + { + "epoch": 1.57, + "learning_rate": 9.70099243953633e-06, + "loss": 1.2091, + "step": 18679 + }, + { + "epoch": 1.57, + "learning_rate": 9.700083874150267e-06, + "loss": 1.1534, + "step": 18680 + }, + { + "epoch": 1.57, + "learning_rate": 9.699175311242204e-06, + "loss": 1.1825, + "step": 18681 + }, + { + "epoch": 1.57, + "learning_rate": 9.698266750819654e-06, + "loss": 1.1514, + "step": 18682 + }, + { + "epoch": 1.57, + "learning_rate": 9.697358192890115e-06, + "loss": 1.1285, + "step": 18683 + }, + { + "epoch": 1.57, + "learning_rate": 9.696449637461104e-06, + "loss": 1.196, + "step": 18684 + }, + { + "epoch": 1.57, + "learning_rate": 9.695541084540122e-06, + "loss": 1.0985, + "step": 18685 + }, + { + "epoch": 1.57, + "learning_rate": 9.694632534134678e-06, + "loss": 1.1687, + "step": 18686 + }, + { + "epoch": 1.57, + "learning_rate": 9.693723986252273e-06, + "loss": 1.2751, + "step": 18687 + }, + { + "epoch": 1.57, + "learning_rate": 9.692815440900423e-06, + "loss": 1.1064, + "step": 18688 + }, + { + "epoch": 1.57, + "learning_rate": 9.691906898086631e-06, + "loss": 1.171, + "step": 18689 + }, + { + "epoch": 1.57, + "learning_rate": 9.690998357818398e-06, + "loss": 1.1671, + "step": 18690 + }, + { + "epoch": 1.57, + "learning_rate": 9.69008982010324e-06, + "loss": 1.2364, + "step": 18691 + }, + { + "epoch": 1.57, + "learning_rate": 9.689181284948656e-06, + "loss": 1.0738, + "step": 18692 + }, + { + "epoch": 1.57, + "learning_rate": 9.68827275236216e-06, + "loss": 1.1663, + "step": 18693 + }, + { + "epoch": 1.57, + "learning_rate": 9.687364222351249e-06, + "loss": 1.2606, + "step": 18694 + }, + { + "epoch": 1.57, + "learning_rate": 9.68645569492344e-06, + "loss": 1.2427, + "step": 18695 + }, + { + "epoch": 1.57, + "learning_rate": 9.685547170086231e-06, + "loss": 1.1734, + "step": 18696 + }, + { + "epoch": 1.57, + "learning_rate": 9.684638647847133e-06, + "loss": 1.1464, + "step": 18697 + }, + { + "epoch": 1.57, + "learning_rate": 9.683730128213648e-06, + "loss": 1.0907, + "step": 18698 + }, + { + "epoch": 1.57, + "learning_rate": 9.682821611193292e-06, + "loss": 1.2026, + "step": 18699 + }, + { + "epoch": 1.57, + "learning_rate": 9.68191309679356e-06, + "loss": 1.1275, + "step": 18700 + }, + { + "epoch": 1.57, + "learning_rate": 9.681004585021965e-06, + "loss": 1.2016, + "step": 18701 + }, + { + "epoch": 1.57, + "learning_rate": 9.680096075886014e-06, + "loss": 1.2121, + "step": 18702 + }, + { + "epoch": 1.57, + "learning_rate": 9.679187569393211e-06, + "loss": 1.1693, + "step": 18703 + }, + { + "epoch": 1.57, + "learning_rate": 9.678279065551064e-06, + "loss": 1.1278, + "step": 18704 + }, + { + "epoch": 1.57, + "learning_rate": 9.677370564367075e-06, + "loss": 1.1786, + "step": 18705 + }, + { + "epoch": 1.57, + "learning_rate": 9.676462065848757e-06, + "loss": 1.1119, + "step": 18706 + }, + { + "epoch": 1.57, + "learning_rate": 9.675553570003613e-06, + "loss": 1.1189, + "step": 18707 + }, + { + "epoch": 1.58, + "learning_rate": 9.674645076839148e-06, + "loss": 1.1644, + "step": 18708 + }, + { + "epoch": 1.58, + "learning_rate": 9.673736586362868e-06, + "loss": 1.2074, + "step": 18709 + }, + { + "epoch": 1.58, + "learning_rate": 9.672828098582285e-06, + "loss": 1.1891, + "step": 18710 + }, + { + "epoch": 1.58, + "learning_rate": 9.671919613504896e-06, + "loss": 1.1317, + "step": 18711 + }, + { + "epoch": 1.58, + "learning_rate": 9.671011131138216e-06, + "loss": 1.2208, + "step": 18712 + }, + { + "epoch": 1.58, + "learning_rate": 9.670102651489746e-06, + "loss": 1.1675, + "step": 18713 + }, + { + "epoch": 1.58, + "learning_rate": 9.669194174566991e-06, + "loss": 1.3105, + "step": 18714 + }, + { + "epoch": 1.58, + "learning_rate": 9.668285700377464e-06, + "loss": 1.1725, + "step": 18715 + }, + { + "epoch": 1.58, + "learning_rate": 9.667377228928662e-06, + "loss": 1.1874, + "step": 18716 + }, + { + "epoch": 1.58, + "learning_rate": 9.6664687602281e-06, + "loss": 1.1772, + "step": 18717 + }, + { + "epoch": 1.58, + "learning_rate": 9.665560294283276e-06, + "loss": 1.1856, + "step": 18718 + }, + { + "epoch": 1.58, + "learning_rate": 9.664651831101704e-06, + "loss": 1.0688, + "step": 18719 + }, + { + "epoch": 1.58, + "learning_rate": 9.663743370690884e-06, + "loss": 1.1778, + "step": 18720 + }, + { + "epoch": 1.58, + "learning_rate": 9.662834913058325e-06, + "loss": 1.1981, + "step": 18721 + }, + { + "epoch": 1.58, + "learning_rate": 9.661926458211529e-06, + "loss": 1.1761, + "step": 18722 + }, + { + "epoch": 1.58, + "learning_rate": 9.661018006158008e-06, + "loss": 1.2207, + "step": 18723 + }, + { + "epoch": 1.58, + "learning_rate": 9.660109556905261e-06, + "loss": 1.134, + "step": 18724 + }, + { + "epoch": 1.58, + "learning_rate": 9.6592011104608e-06, + "loss": 1.1842, + "step": 18725 + }, + { + "epoch": 1.58, + "learning_rate": 9.65829266683213e-06, + "loss": 1.2635, + "step": 18726 + }, + { + "epoch": 1.58, + "learning_rate": 9.657384226026752e-06, + "loss": 1.2192, + "step": 18727 + }, + { + "epoch": 1.58, + "learning_rate": 9.656475788052178e-06, + "loss": 1.2811, + "step": 18728 + }, + { + "epoch": 1.58, + "learning_rate": 9.655567352915907e-06, + "loss": 1.2346, + "step": 18729 + }, + { + "epoch": 1.58, + "learning_rate": 9.654658920625454e-06, + "loss": 1.2105, + "step": 18730 + }, + { + "epoch": 1.58, + "learning_rate": 9.653750491188316e-06, + "loss": 1.1492, + "step": 18731 + }, + { + "epoch": 1.58, + "learning_rate": 9.652842064612005e-06, + "loss": 1.1765, + "step": 18732 + }, + { + "epoch": 1.58, + "learning_rate": 9.65193364090402e-06, + "loss": 1.2113, + "step": 18733 + }, + { + "epoch": 1.58, + "learning_rate": 9.651025220071875e-06, + "loss": 1.1678, + "step": 18734 + }, + { + "epoch": 1.58, + "learning_rate": 9.650116802123069e-06, + "loss": 1.1824, + "step": 18735 + }, + { + "epoch": 1.58, + "learning_rate": 9.649208387065111e-06, + "loss": 1.1238, + "step": 18736 + }, + { + "epoch": 1.58, + "learning_rate": 9.648299974905502e-06, + "loss": 1.1327, + "step": 18737 + }, + { + "epoch": 1.58, + "learning_rate": 9.647391565651754e-06, + "loss": 1.1505, + "step": 18738 + }, + { + "epoch": 1.58, + "learning_rate": 9.646483159311371e-06, + "loss": 1.1603, + "step": 18739 + }, + { + "epoch": 1.58, + "learning_rate": 9.645574755891853e-06, + "loss": 1.2209, + "step": 18740 + }, + { + "epoch": 1.58, + "learning_rate": 9.644666355400715e-06, + "loss": 1.1706, + "step": 18741 + }, + { + "epoch": 1.58, + "learning_rate": 9.643757957845455e-06, + "loss": 1.0408, + "step": 18742 + }, + { + "epoch": 1.58, + "learning_rate": 9.642849563233583e-06, + "loss": 1.2147, + "step": 18743 + }, + { + "epoch": 1.58, + "learning_rate": 9.641941171572597e-06, + "loss": 1.2395, + "step": 18744 + }, + { + "epoch": 1.58, + "learning_rate": 9.641032782870012e-06, + "loss": 1.1635, + "step": 18745 + }, + { + "epoch": 1.58, + "learning_rate": 9.640124397133327e-06, + "loss": 1.2962, + "step": 18746 + }, + { + "epoch": 1.58, + "learning_rate": 9.639216014370052e-06, + "loss": 1.2319, + "step": 18747 + }, + { + "epoch": 1.58, + "learning_rate": 9.638307634587685e-06, + "loss": 1.201, + "step": 18748 + }, + { + "epoch": 1.58, + "learning_rate": 9.637399257793741e-06, + "loss": 1.1356, + "step": 18749 + }, + { + "epoch": 1.58, + "learning_rate": 9.636490883995718e-06, + "loss": 1.0856, + "step": 18750 + }, + { + "epoch": 1.58, + "learning_rate": 9.635582513201122e-06, + "loss": 1.0852, + "step": 18751 + }, + { + "epoch": 1.58, + "learning_rate": 9.634674145417463e-06, + "loss": 1.2297, + "step": 18752 + }, + { + "epoch": 1.58, + "learning_rate": 9.63376578065224e-06, + "loss": 1.1564, + "step": 18753 + }, + { + "epoch": 1.58, + "learning_rate": 9.632857418912966e-06, + "loss": 1.1569, + "step": 18754 + }, + { + "epoch": 1.58, + "learning_rate": 9.631949060207138e-06, + "loss": 1.2349, + "step": 18755 + }, + { + "epoch": 1.58, + "learning_rate": 9.631040704542265e-06, + "loss": 1.272, + "step": 18756 + }, + { + "epoch": 1.58, + "learning_rate": 9.630132351925852e-06, + "loss": 1.2087, + "step": 18757 + }, + { + "epoch": 1.58, + "learning_rate": 9.629224002365407e-06, + "loss": 1.1609, + "step": 18758 + }, + { + "epoch": 1.58, + "learning_rate": 9.628315655868426e-06, + "loss": 1.19, + "step": 18759 + }, + { + "epoch": 1.58, + "learning_rate": 9.627407312442425e-06, + "loss": 1.1173, + "step": 18760 + }, + { + "epoch": 1.58, + "learning_rate": 9.626498972094902e-06, + "loss": 1.1868, + "step": 18761 + }, + { + "epoch": 1.58, + "learning_rate": 9.625590634833366e-06, + "loss": 1.12, + "step": 18762 + }, + { + "epoch": 1.58, + "learning_rate": 9.624682300665316e-06, + "loss": 1.2028, + "step": 18763 + }, + { + "epoch": 1.58, + "learning_rate": 9.623773969598263e-06, + "loss": 1.2807, + "step": 18764 + }, + { + "epoch": 1.58, + "learning_rate": 9.622865641639712e-06, + "loss": 1.124, + "step": 18765 + }, + { + "epoch": 1.58, + "learning_rate": 9.62195731679716e-06, + "loss": 1.178, + "step": 18766 + }, + { + "epoch": 1.58, + "learning_rate": 9.621048995078124e-06, + "loss": 1.1187, + "step": 18767 + }, + { + "epoch": 1.58, + "learning_rate": 9.620140676490098e-06, + "loss": 1.2552, + "step": 18768 + }, + { + "epoch": 1.58, + "learning_rate": 9.619232361040596e-06, + "loss": 1.2319, + "step": 18769 + }, + { + "epoch": 1.58, + "learning_rate": 9.618324048737115e-06, + "loss": 1.1735, + "step": 18770 + }, + { + "epoch": 1.58, + "learning_rate": 9.617415739587165e-06, + "loss": 1.0856, + "step": 18771 + }, + { + "epoch": 1.58, + "learning_rate": 9.616507433598244e-06, + "loss": 1.1672, + "step": 18772 + }, + { + "epoch": 1.58, + "learning_rate": 9.615599130777866e-06, + "loss": 1.0852, + "step": 18773 + }, + { + "epoch": 1.58, + "learning_rate": 9.614690831133528e-06, + "loss": 1.328, + "step": 18774 + }, + { + "epoch": 1.58, + "learning_rate": 9.613782534672736e-06, + "loss": 1.1414, + "step": 18775 + }, + { + "epoch": 1.58, + "learning_rate": 9.612874241403e-06, + "loss": 1.2148, + "step": 18776 + }, + { + "epoch": 1.58, + "learning_rate": 9.61196595133182e-06, + "loss": 1.2159, + "step": 18777 + }, + { + "epoch": 1.58, + "learning_rate": 9.611057664466702e-06, + "loss": 1.202, + "step": 18778 + }, + { + "epoch": 1.58, + "learning_rate": 9.610149380815148e-06, + "loss": 1.2896, + "step": 18779 + }, + { + "epoch": 1.58, + "learning_rate": 9.609241100384668e-06, + "loss": 1.1438, + "step": 18780 + }, + { + "epoch": 1.58, + "learning_rate": 9.608332823182761e-06, + "loss": 1.1077, + "step": 18781 + }, + { + "epoch": 1.58, + "learning_rate": 9.607424549216937e-06, + "loss": 1.1566, + "step": 18782 + }, + { + "epoch": 1.58, + "learning_rate": 9.60651627849469e-06, + "loss": 1.2443, + "step": 18783 + }, + { + "epoch": 1.58, + "learning_rate": 9.605608011023536e-06, + "loss": 1.1399, + "step": 18784 + }, + { + "epoch": 1.58, + "learning_rate": 9.604699746810974e-06, + "loss": 1.1433, + "step": 18785 + }, + { + "epoch": 1.58, + "learning_rate": 9.60379148586451e-06, + "loss": 1.202, + "step": 18786 + }, + { + "epoch": 1.58, + "learning_rate": 9.602883228191644e-06, + "loss": 1.021, + "step": 18787 + }, + { + "epoch": 1.58, + "learning_rate": 9.601974973799887e-06, + "loss": 1.056, + "step": 18788 + }, + { + "epoch": 1.58, + "learning_rate": 9.601066722696743e-06, + "loss": 1.099, + "step": 18789 + }, + { + "epoch": 1.58, + "learning_rate": 9.600158474889706e-06, + "loss": 1.2369, + "step": 18790 + }, + { + "epoch": 1.58, + "learning_rate": 9.599250230386295e-06, + "loss": 1.1844, + "step": 18791 + }, + { + "epoch": 1.58, + "learning_rate": 9.598341989194002e-06, + "loss": 1.2332, + "step": 18792 + }, + { + "epoch": 1.58, + "learning_rate": 9.59743375132034e-06, + "loss": 1.1596, + "step": 18793 + }, + { + "epoch": 1.58, + "learning_rate": 9.596525516772804e-06, + "loss": 1.2408, + "step": 18794 + }, + { + "epoch": 1.58, + "learning_rate": 9.595617285558908e-06, + "loss": 1.3065, + "step": 18795 + }, + { + "epoch": 1.58, + "learning_rate": 9.594709057686148e-06, + "loss": 1.1422, + "step": 18796 + }, + { + "epoch": 1.58, + "learning_rate": 9.593800833162035e-06, + "loss": 1.1524, + "step": 18797 + }, + { + "epoch": 1.58, + "learning_rate": 9.592892611994067e-06, + "loss": 1.1725, + "step": 18798 + }, + { + "epoch": 1.58, + "learning_rate": 9.591984394189752e-06, + "loss": 1.1712, + "step": 18799 + }, + { + "epoch": 1.58, + "learning_rate": 9.591076179756591e-06, + "loss": 1.2361, + "step": 18800 + }, + { + "epoch": 1.58, + "learning_rate": 9.590167968702088e-06, + "loss": 1.0857, + "step": 18801 + }, + { + "epoch": 1.58, + "learning_rate": 9.589259761033752e-06, + "loss": 1.1287, + "step": 18802 + }, + { + "epoch": 1.58, + "learning_rate": 9.588351556759082e-06, + "loss": 1.3021, + "step": 18803 + }, + { + "epoch": 1.58, + "learning_rate": 9.587443355885584e-06, + "loss": 1.2478, + "step": 18804 + }, + { + "epoch": 1.58, + "learning_rate": 9.586535158420758e-06, + "loss": 1.2219, + "step": 18805 + }, + { + "epoch": 1.58, + "learning_rate": 9.585626964372114e-06, + "loss": 1.1438, + "step": 18806 + }, + { + "epoch": 1.58, + "learning_rate": 9.584718773747151e-06, + "loss": 1.1739, + "step": 18807 + }, + { + "epoch": 1.58, + "learning_rate": 9.583810586553377e-06, + "loss": 1.2181, + "step": 18808 + }, + { + "epoch": 1.58, + "learning_rate": 9.582902402798288e-06, + "loss": 1.185, + "step": 18809 + }, + { + "epoch": 1.58, + "learning_rate": 9.5819942224894e-06, + "loss": 1.1456, + "step": 18810 + }, + { + "epoch": 1.58, + "learning_rate": 9.581086045634204e-06, + "loss": 1.1116, + "step": 18811 + }, + { + "epoch": 1.58, + "learning_rate": 9.580177872240211e-06, + "loss": 1.2109, + "step": 18812 + }, + { + "epoch": 1.58, + "learning_rate": 9.57926970231492e-06, + "loss": 1.1391, + "step": 18813 + }, + { + "epoch": 1.58, + "learning_rate": 9.57836153586584e-06, + "loss": 1.15, + "step": 18814 + }, + { + "epoch": 1.58, + "learning_rate": 9.577453372900474e-06, + "loss": 1.1121, + "step": 18815 + }, + { + "epoch": 1.58, + "learning_rate": 9.576545213426318e-06, + "loss": 1.3299, + "step": 18816 + }, + { + "epoch": 1.58, + "learning_rate": 9.575637057450889e-06, + "loss": 1.1444, + "step": 18817 + }, + { + "epoch": 1.58, + "learning_rate": 9.574728904981675e-06, + "loss": 1.1278, + "step": 18818 + }, + { + "epoch": 1.58, + "learning_rate": 9.57382075602619e-06, + "loss": 1.2181, + "step": 18819 + }, + { + "epoch": 1.58, + "learning_rate": 9.572912610591934e-06, + "loss": 1.292, + "step": 18820 + }, + { + "epoch": 1.58, + "learning_rate": 9.572004468686411e-06, + "loss": 1.1761, + "step": 18821 + }, + { + "epoch": 1.58, + "learning_rate": 9.571096330317121e-06, + "loss": 1.1895, + "step": 18822 + }, + { + "epoch": 1.58, + "learning_rate": 9.570188195491575e-06, + "loss": 1.1525, + "step": 18823 + }, + { + "epoch": 1.58, + "learning_rate": 9.56928006421727e-06, + "loss": 1.2375, + "step": 18824 + }, + { + "epoch": 1.58, + "learning_rate": 9.568371936501707e-06, + "loss": 1.1661, + "step": 18825 + }, + { + "epoch": 1.58, + "learning_rate": 9.5674638123524e-06, + "loss": 1.1855, + "step": 18826 + }, + { + "epoch": 1.59, + "learning_rate": 9.56655569177684e-06, + "loss": 1.1772, + "step": 18827 + }, + { + "epoch": 1.59, + "learning_rate": 9.56564757478254e-06, + "loss": 1.213, + "step": 18828 + }, + { + "epoch": 1.59, + "learning_rate": 9.564739461376993e-06, + "loss": 1.3114, + "step": 18829 + }, + { + "epoch": 1.59, + "learning_rate": 9.563831351567714e-06, + "loss": 1.1948, + "step": 18830 + }, + { + "epoch": 1.59, + "learning_rate": 9.562923245362196e-06, + "loss": 1.0877, + "step": 18831 + }, + { + "epoch": 1.59, + "learning_rate": 9.562015142767949e-06, + "loss": 1.2021, + "step": 18832 + }, + { + "epoch": 1.59, + "learning_rate": 9.56110704379247e-06, + "loss": 1.2711, + "step": 18833 + }, + { + "epoch": 1.59, + "learning_rate": 9.560198948443267e-06, + "loss": 1.2479, + "step": 18834 + }, + { + "epoch": 1.59, + "learning_rate": 9.55929085672784e-06, + "loss": 1.2228, + "step": 18835 + }, + { + "epoch": 1.59, + "learning_rate": 9.558382768653696e-06, + "loss": 1.1745, + "step": 18836 + }, + { + "epoch": 1.59, + "learning_rate": 9.557474684228331e-06, + "loss": 1.1897, + "step": 18837 + }, + { + "epoch": 1.59, + "learning_rate": 9.556566603459251e-06, + "loss": 1.2106, + "step": 18838 + }, + { + "epoch": 1.59, + "learning_rate": 9.555658526353965e-06, + "loss": 1.2445, + "step": 18839 + }, + { + "epoch": 1.59, + "learning_rate": 9.554750452919964e-06, + "loss": 1.1528, + "step": 18840 + }, + { + "epoch": 1.59, + "learning_rate": 9.553842383164762e-06, + "loss": 1.1908, + "step": 18841 + }, + { + "epoch": 1.59, + "learning_rate": 9.552934317095855e-06, + "loss": 1.2547, + "step": 18842 + }, + { + "epoch": 1.59, + "learning_rate": 9.552026254720751e-06, + "loss": 1.107, + "step": 18843 + }, + { + "epoch": 1.59, + "learning_rate": 9.551118196046943e-06, + "loss": 1.2045, + "step": 18844 + }, + { + "epoch": 1.59, + "learning_rate": 9.550210141081948e-06, + "loss": 1.3093, + "step": 18845 + }, + { + "epoch": 1.59, + "learning_rate": 9.549302089833254e-06, + "loss": 1.1304, + "step": 18846 + }, + { + "epoch": 1.59, + "learning_rate": 9.548394042308376e-06, + "loss": 1.2011, + "step": 18847 + }, + { + "epoch": 1.59, + "learning_rate": 9.547485998514805e-06, + "loss": 1.1491, + "step": 18848 + }, + { + "epoch": 1.59, + "learning_rate": 9.546577958460053e-06, + "loss": 1.2222, + "step": 18849 + }, + { + "epoch": 1.59, + "learning_rate": 9.54566992215162e-06, + "loss": 1.2267, + "step": 18850 + }, + { + "epoch": 1.59, + "learning_rate": 9.544761889597002e-06, + "loss": 1.2655, + "step": 18851 + }, + { + "epoch": 1.59, + "learning_rate": 9.543853860803714e-06, + "loss": 1.1399, + "step": 18852 + }, + { + "epoch": 1.59, + "learning_rate": 9.542945835779247e-06, + "loss": 1.142, + "step": 18853 + }, + { + "epoch": 1.59, + "learning_rate": 9.542037814531111e-06, + "loss": 1.1809, + "step": 18854 + }, + { + "epoch": 1.59, + "learning_rate": 9.5411297970668e-06, + "loss": 1.2386, + "step": 18855 + }, + { + "epoch": 1.59, + "learning_rate": 9.540221783393827e-06, + "loss": 1.2088, + "step": 18856 + }, + { + "epoch": 1.59, + "learning_rate": 9.539313773519687e-06, + "loss": 1.1736, + "step": 18857 + }, + { + "epoch": 1.59, + "learning_rate": 9.538405767451885e-06, + "loss": 1.1831, + "step": 18858 + }, + { + "epoch": 1.59, + "learning_rate": 9.537497765197918e-06, + "loss": 1.1909, + "step": 18859 + }, + { + "epoch": 1.59, + "learning_rate": 9.536589766765297e-06, + "loss": 1.1063, + "step": 18860 + }, + { + "epoch": 1.59, + "learning_rate": 9.535681772161519e-06, + "loss": 1.2631, + "step": 18861 + }, + { + "epoch": 1.59, + "learning_rate": 9.534773781394087e-06, + "loss": 1.1918, + "step": 18862 + }, + { + "epoch": 1.59, + "learning_rate": 9.533865794470499e-06, + "loss": 1.2332, + "step": 18863 + }, + { + "epoch": 1.59, + "learning_rate": 9.532957811398263e-06, + "loss": 1.0418, + "step": 18864 + }, + { + "epoch": 1.59, + "learning_rate": 9.532049832184882e-06, + "loss": 1.1339, + "step": 18865 + }, + { + "epoch": 1.59, + "learning_rate": 9.53114185683785e-06, + "loss": 1.0836, + "step": 18866 + }, + { + "epoch": 1.59, + "learning_rate": 9.53023388536468e-06, + "loss": 1.1594, + "step": 18867 + }, + { + "epoch": 1.59, + "learning_rate": 9.529325917772862e-06, + "loss": 1.1555, + "step": 18868 + }, + { + "epoch": 1.59, + "learning_rate": 9.528417954069909e-06, + "loss": 1.2325, + "step": 18869 + }, + { + "epoch": 1.59, + "learning_rate": 9.527509994263316e-06, + "loss": 1.254, + "step": 18870 + }, + { + "epoch": 1.59, + "learning_rate": 9.526602038360587e-06, + "loss": 1.2151, + "step": 18871 + }, + { + "epoch": 1.59, + "learning_rate": 9.52569408636922e-06, + "loss": 1.2003, + "step": 18872 + }, + { + "epoch": 1.59, + "learning_rate": 9.524786138296725e-06, + "loss": 1.232, + "step": 18873 + }, + { + "epoch": 1.59, + "learning_rate": 9.523878194150597e-06, + "loss": 1.166, + "step": 18874 + }, + { + "epoch": 1.59, + "learning_rate": 9.522970253938336e-06, + "loss": 1.2243, + "step": 18875 + }, + { + "epoch": 1.59, + "learning_rate": 9.522062317667453e-06, + "loss": 1.3018, + "step": 18876 + }, + { + "epoch": 1.59, + "learning_rate": 9.52115438534544e-06, + "loss": 1.1563, + "step": 18877 + }, + { + "epoch": 1.59, + "learning_rate": 9.520246456979808e-06, + "loss": 1.1069, + "step": 18878 + }, + { + "epoch": 1.59, + "learning_rate": 9.519338532578047e-06, + "loss": 1.2044, + "step": 18879 + }, + { + "epoch": 1.59, + "learning_rate": 9.51843061214767e-06, + "loss": 1.1545, + "step": 18880 + }, + { + "epoch": 1.59, + "learning_rate": 9.517522695696173e-06, + "loss": 1.0703, + "step": 18881 + }, + { + "epoch": 1.59, + "learning_rate": 9.516614783231057e-06, + "loss": 1.23, + "step": 18882 + }, + { + "epoch": 1.59, + "learning_rate": 9.515706874759821e-06, + "loss": 1.2152, + "step": 18883 + }, + { + "epoch": 1.59, + "learning_rate": 9.514798970289975e-06, + "loss": 1.1994, + "step": 18884 + }, + { + "epoch": 1.59, + "learning_rate": 9.51389106982901e-06, + "loss": 1.1333, + "step": 18885 + }, + { + "epoch": 1.59, + "learning_rate": 9.512983173384438e-06, + "loss": 1.159, + "step": 18886 + }, + { + "epoch": 1.59, + "learning_rate": 9.512075280963749e-06, + "loss": 1.159, + "step": 18887 + }, + { + "epoch": 1.59, + "learning_rate": 9.511167392574454e-06, + "loss": 1.1752, + "step": 18888 + }, + { + "epoch": 1.59, + "learning_rate": 9.510259508224052e-06, + "loss": 1.1049, + "step": 18889 + }, + { + "epoch": 1.59, + "learning_rate": 9.509351627920037e-06, + "loss": 1.1189, + "step": 18890 + }, + { + "epoch": 1.59, + "learning_rate": 9.508443751669922e-06, + "loss": 1.0972, + "step": 18891 + }, + { + "epoch": 1.59, + "learning_rate": 9.507535879481198e-06, + "loss": 1.1664, + "step": 18892 + }, + { + "epoch": 1.59, + "learning_rate": 9.506628011361375e-06, + "loss": 1.2245, + "step": 18893 + }, + { + "epoch": 1.59, + "learning_rate": 9.505720147317941e-06, + "loss": 1.1461, + "step": 18894 + }, + { + "epoch": 1.59, + "learning_rate": 9.504812287358413e-06, + "loss": 1.2126, + "step": 18895 + }, + { + "epoch": 1.59, + "learning_rate": 9.50390443149028e-06, + "loss": 1.1991, + "step": 18896 + }, + { + "epoch": 1.59, + "learning_rate": 9.502996579721051e-06, + "loss": 1.1887, + "step": 18897 + }, + { + "epoch": 1.59, + "learning_rate": 9.502088732058219e-06, + "loss": 1.1543, + "step": 18898 + }, + { + "epoch": 1.59, + "learning_rate": 9.501180888509295e-06, + "loss": 1.0902, + "step": 18899 + }, + { + "epoch": 1.59, + "learning_rate": 9.500273049081769e-06, + "loss": 1.2132, + "step": 18900 + }, + { + "epoch": 1.59, + "learning_rate": 9.499365213783145e-06, + "loss": 1.1223, + "step": 18901 + }, + { + "epoch": 1.59, + "learning_rate": 9.498457382620931e-06, + "loss": 1.15, + "step": 18902 + }, + { + "epoch": 1.59, + "learning_rate": 9.497549555602621e-06, + "loss": 1.1665, + "step": 18903 + }, + { + "epoch": 1.59, + "learning_rate": 9.49664173273572e-06, + "loss": 1.1865, + "step": 18904 + }, + { + "epoch": 1.59, + "learning_rate": 9.49573391402772e-06, + "loss": 1.2289, + "step": 18905 + }, + { + "epoch": 1.59, + "learning_rate": 9.494826099486133e-06, + "loss": 1.2201, + "step": 18906 + }, + { + "epoch": 1.59, + "learning_rate": 9.493918289118452e-06, + "loss": 1.1364, + "step": 18907 + }, + { + "epoch": 1.59, + "learning_rate": 9.493010482932184e-06, + "loss": 1.1539, + "step": 18908 + }, + { + "epoch": 1.59, + "learning_rate": 9.49210268093482e-06, + "loss": 1.1623, + "step": 18909 + }, + { + "epoch": 1.59, + "learning_rate": 9.49119488313387e-06, + "loss": 1.1457, + "step": 18910 + }, + { + "epoch": 1.59, + "learning_rate": 9.490287089536828e-06, + "loss": 1.2398, + "step": 18911 + }, + { + "epoch": 1.59, + "learning_rate": 9.4893793001512e-06, + "loss": 1.2588, + "step": 18912 + }, + { + "epoch": 1.59, + "learning_rate": 9.488471514984481e-06, + "loss": 1.0698, + "step": 18913 + }, + { + "epoch": 1.59, + "learning_rate": 9.487563734044174e-06, + "loss": 1.1554, + "step": 18914 + }, + { + "epoch": 1.59, + "learning_rate": 9.486655957337783e-06, + "loss": 1.1472, + "step": 18915 + }, + { + "epoch": 1.59, + "learning_rate": 9.485748184872799e-06, + "loss": 1.1333, + "step": 18916 + }, + { + "epoch": 1.59, + "learning_rate": 9.484840416656734e-06, + "loss": 1.1345, + "step": 18917 + }, + { + "epoch": 1.59, + "learning_rate": 9.483932652697079e-06, + "loss": 1.1868, + "step": 18918 + }, + { + "epoch": 1.59, + "learning_rate": 9.48302489300134e-06, + "loss": 1.2417, + "step": 18919 + }, + { + "epoch": 1.59, + "learning_rate": 9.48211713757701e-06, + "loss": 1.1606, + "step": 18920 + }, + { + "epoch": 1.59, + "learning_rate": 9.481209386431602e-06, + "loss": 1.1071, + "step": 18921 + }, + { + "epoch": 1.59, + "learning_rate": 9.4803016395726e-06, + "loss": 1.2086, + "step": 18922 + }, + { + "epoch": 1.59, + "learning_rate": 9.479393897007519e-06, + "loss": 1.1976, + "step": 18923 + }, + { + "epoch": 1.59, + "learning_rate": 9.478486158743846e-06, + "loss": 1.1553, + "step": 18924 + }, + { + "epoch": 1.59, + "learning_rate": 9.477578424789088e-06, + "loss": 1.0906, + "step": 18925 + }, + { + "epoch": 1.59, + "learning_rate": 9.47667069515075e-06, + "loss": 1.2626, + "step": 18926 + }, + { + "epoch": 1.59, + "learning_rate": 9.47576296983632e-06, + "loss": 1.1959, + "step": 18927 + }, + { + "epoch": 1.59, + "learning_rate": 9.474855248853308e-06, + "loss": 1.1406, + "step": 18928 + }, + { + "epoch": 1.59, + "learning_rate": 9.473947532209206e-06, + "loss": 1.1765, + "step": 18929 + }, + { + "epoch": 1.59, + "learning_rate": 9.473039819911522e-06, + "loss": 1.1122, + "step": 18930 + }, + { + "epoch": 1.59, + "learning_rate": 9.472132111967752e-06, + "loss": 1.2064, + "step": 18931 + }, + { + "epoch": 1.59, + "learning_rate": 9.471224408385395e-06, + "loss": 1.1755, + "step": 18932 + }, + { + "epoch": 1.59, + "learning_rate": 9.470316709171945e-06, + "loss": 1.2205, + "step": 18933 + }, + { + "epoch": 1.59, + "learning_rate": 9.469409014334915e-06, + "loss": 1.1678, + "step": 18934 + }, + { + "epoch": 1.59, + "learning_rate": 9.468501323881794e-06, + "loss": 1.185, + "step": 18935 + }, + { + "epoch": 1.59, + "learning_rate": 9.46759363782009e-06, + "loss": 1.1884, + "step": 18936 + }, + { + "epoch": 1.59, + "learning_rate": 9.46668595615729e-06, + "loss": 1.268, + "step": 18937 + }, + { + "epoch": 1.59, + "learning_rate": 9.465778278900905e-06, + "loss": 1.1449, + "step": 18938 + }, + { + "epoch": 1.59, + "learning_rate": 9.464870606058433e-06, + "loss": 1.2088, + "step": 18939 + }, + { + "epoch": 1.59, + "learning_rate": 9.463962937637367e-06, + "loss": 1.1025, + "step": 18940 + }, + { + "epoch": 1.59, + "learning_rate": 9.463055273645215e-06, + "loss": 1.1922, + "step": 18941 + }, + { + "epoch": 1.59, + "learning_rate": 9.46214761408947e-06, + "loss": 1.2133, + "step": 18942 + }, + { + "epoch": 1.59, + "learning_rate": 9.461239958977635e-06, + "loss": 1.193, + "step": 18943 + }, + { + "epoch": 1.59, + "learning_rate": 9.460332308317205e-06, + "loss": 1.133, + "step": 18944 + }, + { + "epoch": 1.6, + "learning_rate": 9.459424662115686e-06, + "loss": 1.1675, + "step": 18945 + }, + { + "epoch": 1.6, + "learning_rate": 9.458517020380569e-06, + "loss": 1.1683, + "step": 18946 + }, + { + "epoch": 1.6, + "learning_rate": 9.457609383119362e-06, + "loss": 1.1271, + "step": 18947 + }, + { + "epoch": 1.6, + "learning_rate": 9.456701750339557e-06, + "loss": 1.2772, + "step": 18948 + }, + { + "epoch": 1.6, + "learning_rate": 9.455794122048658e-06, + "loss": 1.2847, + "step": 18949 + }, + { + "epoch": 1.6, + "learning_rate": 9.454886498254161e-06, + "loss": 1.1886, + "step": 18950 + }, + { + "epoch": 1.6, + "learning_rate": 9.453978878963563e-06, + "loss": 1.1958, + "step": 18951 + }, + { + "epoch": 1.6, + "learning_rate": 9.453071264184372e-06, + "loss": 1.2073, + "step": 18952 + }, + { + "epoch": 1.6, + "learning_rate": 9.452163653924078e-06, + "loss": 1.1887, + "step": 18953 + }, + { + "epoch": 1.6, + "learning_rate": 9.451256048190186e-06, + "loss": 1.0953, + "step": 18954 + }, + { + "epoch": 1.6, + "learning_rate": 9.450348446990187e-06, + "loss": 1.1512, + "step": 18955 + }, + { + "epoch": 1.6, + "learning_rate": 9.44944085033159e-06, + "loss": 1.1464, + "step": 18956 + }, + { + "epoch": 1.6, + "learning_rate": 9.448533258221888e-06, + "loss": 1.1507, + "step": 18957 + }, + { + "epoch": 1.6, + "learning_rate": 9.447625670668581e-06, + "loss": 1.1985, + "step": 18958 + }, + { + "epoch": 1.6, + "learning_rate": 9.446718087679163e-06, + "loss": 1.1998, + "step": 18959 + }, + { + "epoch": 1.6, + "learning_rate": 9.445810509261142e-06, + "loss": 1.1701, + "step": 18960 + }, + { + "epoch": 1.6, + "learning_rate": 9.444902935422011e-06, + "loss": 1.1985, + "step": 18961 + }, + { + "epoch": 1.6, + "learning_rate": 9.443995366169271e-06, + "loss": 1.1684, + "step": 18962 + }, + { + "epoch": 1.6, + "learning_rate": 9.443087801510416e-06, + "loss": 1.1506, + "step": 18963 + }, + { + "epoch": 1.6, + "learning_rate": 9.442180241452948e-06, + "loss": 1.1377, + "step": 18964 + }, + { + "epoch": 1.6, + "learning_rate": 9.441272686004368e-06, + "loss": 1.1917, + "step": 18965 + }, + { + "epoch": 1.6, + "learning_rate": 9.440365135172168e-06, + "loss": 1.2401, + "step": 18966 + }, + { + "epoch": 1.6, + "learning_rate": 9.439457588963854e-06, + "loss": 1.1591, + "step": 18967 + }, + { + "epoch": 1.6, + "learning_rate": 9.438550047386919e-06, + "loss": 1.1393, + "step": 18968 + }, + { + "epoch": 1.6, + "learning_rate": 9.437642510448865e-06, + "loss": 1.1681, + "step": 18969 + }, + { + "epoch": 1.6, + "learning_rate": 9.436734978157187e-06, + "loss": 1.1824, + "step": 18970 + }, + { + "epoch": 1.6, + "learning_rate": 9.435827450519389e-06, + "loss": 1.2153, + "step": 18971 + }, + { + "epoch": 1.6, + "learning_rate": 9.434919927542958e-06, + "loss": 1.2044, + "step": 18972 + }, + { + "epoch": 1.6, + "learning_rate": 9.434012409235404e-06, + "loss": 1.2187, + "step": 18973 + }, + { + "epoch": 1.6, + "learning_rate": 9.433104895604219e-06, + "loss": 1.228, + "step": 18974 + }, + { + "epoch": 1.6, + "learning_rate": 9.432197386656901e-06, + "loss": 1.2181, + "step": 18975 + }, + { + "epoch": 1.6, + "learning_rate": 9.431289882400954e-06, + "loss": 1.233, + "step": 18976 + }, + { + "epoch": 1.6, + "learning_rate": 9.430382382843869e-06, + "loss": 1.2205, + "step": 18977 + }, + { + "epoch": 1.6, + "learning_rate": 9.42947488799315e-06, + "loss": 1.0984, + "step": 18978 + }, + { + "epoch": 1.6, + "learning_rate": 9.428567397856288e-06, + "loss": 1.1371, + "step": 18979 + }, + { + "epoch": 1.6, + "learning_rate": 9.427659912440788e-06, + "loss": 1.1552, + "step": 18980 + }, + { + "epoch": 1.6, + "learning_rate": 9.426752431754145e-06, + "loss": 1.0723, + "step": 18981 + }, + { + "epoch": 1.6, + "learning_rate": 9.425844955803857e-06, + "loss": 1.1483, + "step": 18982 + }, + { + "epoch": 1.6, + "learning_rate": 9.424937484597418e-06, + "loss": 1.093, + "step": 18983 + }, + { + "epoch": 1.6, + "learning_rate": 9.424030018142335e-06, + "loss": 1.2063, + "step": 18984 + }, + { + "epoch": 1.6, + "learning_rate": 9.423122556446097e-06, + "loss": 1.1818, + "step": 18985 + }, + { + "epoch": 1.6, + "learning_rate": 9.422215099516207e-06, + "loss": 1.2518, + "step": 18986 + }, + { + "epoch": 1.6, + "learning_rate": 9.421307647360158e-06, + "loss": 1.1368, + "step": 18987 + }, + { + "epoch": 1.6, + "learning_rate": 9.420400199985452e-06, + "loss": 1.1718, + "step": 18988 + }, + { + "epoch": 1.6, + "learning_rate": 9.419492757399585e-06, + "loss": 1.1778, + "step": 18989 + }, + { + "epoch": 1.6, + "learning_rate": 9.418585319610053e-06, + "loss": 1.2514, + "step": 18990 + }, + { + "epoch": 1.6, + "learning_rate": 9.417677886624358e-06, + "loss": 1.0922, + "step": 18991 + }, + { + "epoch": 1.6, + "learning_rate": 9.416770458449993e-06, + "loss": 1.083, + "step": 18992 + }, + { + "epoch": 1.6, + "learning_rate": 9.415863035094458e-06, + "loss": 1.1598, + "step": 18993 + }, + { + "epoch": 1.6, + "learning_rate": 9.414955616565248e-06, + "loss": 1.222, + "step": 18994 + }, + { + "epoch": 1.6, + "learning_rate": 9.414048202869864e-06, + "loss": 1.1596, + "step": 18995 + }, + { + "epoch": 1.6, + "learning_rate": 9.4131407940158e-06, + "loss": 1.2955, + "step": 18996 + }, + { + "epoch": 1.6, + "learning_rate": 9.412233390010557e-06, + "loss": 1.2176, + "step": 18997 + }, + { + "epoch": 1.6, + "learning_rate": 9.411325990861624e-06, + "loss": 1.1306, + "step": 18998 + }, + { + "epoch": 1.6, + "learning_rate": 9.41041859657651e-06, + "loss": 1.1751, + "step": 18999 + }, + { + "epoch": 1.6, + "learning_rate": 9.409511207162704e-06, + "loss": 1.1434, + "step": 19000 + }, + { + "epoch": 1.6, + "learning_rate": 9.408603822627704e-06, + "loss": 1.15, + "step": 19001 + }, + { + "epoch": 1.6, + "learning_rate": 9.407696442979011e-06, + "loss": 1.2915, + "step": 19002 + }, + { + "epoch": 1.6, + "learning_rate": 9.406789068224117e-06, + "loss": 1.1768, + "step": 19003 + }, + { + "epoch": 1.6, + "learning_rate": 9.405881698370526e-06, + "loss": 1.1614, + "step": 19004 + }, + { + "epoch": 1.6, + "learning_rate": 9.404974333425725e-06, + "loss": 1.2239, + "step": 19005 + }, + { + "epoch": 1.6, + "learning_rate": 9.404066973397221e-06, + "loss": 1.1186, + "step": 19006 + }, + { + "epoch": 1.6, + "learning_rate": 9.403159618292504e-06, + "loss": 1.1358, + "step": 19007 + }, + { + "epoch": 1.6, + "learning_rate": 9.402252268119076e-06, + "loss": 1.2778, + "step": 19008 + }, + { + "epoch": 1.6, + "learning_rate": 9.401344922884425e-06, + "loss": 1.1713, + "step": 19009 + }, + { + "epoch": 1.6, + "learning_rate": 9.40043758259606e-06, + "loss": 1.1335, + "step": 19010 + }, + { + "epoch": 1.6, + "learning_rate": 9.39953024726147e-06, + "loss": 1.1236, + "step": 19011 + }, + { + "epoch": 1.6, + "learning_rate": 9.398622916888155e-06, + "loss": 1.1377, + "step": 19012 + }, + { + "epoch": 1.6, + "learning_rate": 9.397715591483603e-06, + "loss": 1.1219, + "step": 19013 + }, + { + "epoch": 1.6, + "learning_rate": 9.396808271055322e-06, + "loss": 1.1865, + "step": 19014 + }, + { + "epoch": 1.6, + "learning_rate": 9.395900955610805e-06, + "loss": 1.1324, + "step": 19015 + }, + { + "epoch": 1.6, + "learning_rate": 9.394993645157543e-06, + "loss": 1.2247, + "step": 19016 + }, + { + "epoch": 1.6, + "learning_rate": 9.394086339703043e-06, + "loss": 1.1736, + "step": 19017 + }, + { + "epoch": 1.6, + "learning_rate": 9.393179039254794e-06, + "loss": 1.2333, + "step": 19018 + }, + { + "epoch": 1.6, + "learning_rate": 9.392271743820294e-06, + "loss": 1.2657, + "step": 19019 + }, + { + "epoch": 1.6, + "learning_rate": 9.391364453407038e-06, + "loss": 1.1083, + "step": 19020 + }, + { + "epoch": 1.6, + "learning_rate": 9.390457168022527e-06, + "loss": 1.1901, + "step": 19021 + }, + { + "epoch": 1.6, + "learning_rate": 9.38954988767425e-06, + "loss": 1.1818, + "step": 19022 + }, + { + "epoch": 1.6, + "learning_rate": 9.38864261236971e-06, + "loss": 1.1566, + "step": 19023 + }, + { + "epoch": 1.6, + "learning_rate": 9.387735342116398e-06, + "loss": 1.2516, + "step": 19024 + }, + { + "epoch": 1.6, + "learning_rate": 9.386828076921811e-06, + "loss": 1.1576, + "step": 19025 + }, + { + "epoch": 1.6, + "learning_rate": 9.385920816793452e-06, + "loss": 1.1386, + "step": 19026 + }, + { + "epoch": 1.6, + "learning_rate": 9.385013561738808e-06, + "loss": 1.0806, + "step": 19027 + }, + { + "epoch": 1.6, + "learning_rate": 9.384106311765381e-06, + "loss": 1.3261, + "step": 19028 + }, + { + "epoch": 1.6, + "learning_rate": 9.383199066880662e-06, + "loss": 1.1865, + "step": 19029 + }, + { + "epoch": 1.6, + "learning_rate": 9.382291827092152e-06, + "loss": 1.0998, + "step": 19030 + }, + { + "epoch": 1.6, + "learning_rate": 9.381384592407343e-06, + "loss": 1.1673, + "step": 19031 + }, + { + "epoch": 1.6, + "learning_rate": 9.380477362833735e-06, + "loss": 1.2107, + "step": 19032 + }, + { + "epoch": 1.6, + "learning_rate": 9.379570138378818e-06, + "loss": 1.0999, + "step": 19033 + }, + { + "epoch": 1.6, + "learning_rate": 9.378662919050094e-06, + "loss": 1.2016, + "step": 19034 + }, + { + "epoch": 1.6, + "learning_rate": 9.377755704855054e-06, + "loss": 1.2249, + "step": 19035 + }, + { + "epoch": 1.6, + "learning_rate": 9.376848495801197e-06, + "loss": 1.1674, + "step": 19036 + }, + { + "epoch": 1.6, + "learning_rate": 9.375941291896015e-06, + "loss": 1.2251, + "step": 19037 + }, + { + "epoch": 1.6, + "learning_rate": 9.375034093147006e-06, + "loss": 1.2092, + "step": 19038 + }, + { + "epoch": 1.6, + "learning_rate": 9.374126899561667e-06, + "loss": 1.2328, + "step": 19039 + }, + { + "epoch": 1.6, + "learning_rate": 9.373219711147489e-06, + "loss": 1.1183, + "step": 19040 + }, + { + "epoch": 1.6, + "learning_rate": 9.372312527911974e-06, + "loss": 1.1477, + "step": 19041 + }, + { + "epoch": 1.6, + "learning_rate": 9.371405349862612e-06, + "loss": 1.0996, + "step": 19042 + }, + { + "epoch": 1.6, + "learning_rate": 9.370498177006903e-06, + "loss": 1.2334, + "step": 19043 + }, + { + "epoch": 1.6, + "learning_rate": 9.369591009352334e-06, + "loss": 1.2296, + "step": 19044 + }, + { + "epoch": 1.6, + "learning_rate": 9.36868384690641e-06, + "loss": 1.2104, + "step": 19045 + }, + { + "epoch": 1.6, + "learning_rate": 9.367776689676621e-06, + "loss": 1.2504, + "step": 19046 + }, + { + "epoch": 1.6, + "learning_rate": 9.366869537670464e-06, + "loss": 1.1835, + "step": 19047 + }, + { + "epoch": 1.6, + "learning_rate": 9.365962390895431e-06, + "loss": 1.1368, + "step": 19048 + }, + { + "epoch": 1.6, + "learning_rate": 9.365055249359023e-06, + "loss": 1.2169, + "step": 19049 + }, + { + "epoch": 1.6, + "learning_rate": 9.36414811306873e-06, + "loss": 1.2173, + "step": 19050 + }, + { + "epoch": 1.6, + "learning_rate": 9.363240982032045e-06, + "loss": 1.227, + "step": 19051 + }, + { + "epoch": 1.6, + "learning_rate": 9.362333856256474e-06, + "loss": 1.2061, + "step": 19052 + }, + { + "epoch": 1.6, + "learning_rate": 9.3614267357495e-06, + "loss": 1.1434, + "step": 19053 + }, + { + "epoch": 1.6, + "learning_rate": 9.360519620518627e-06, + "loss": 1.1755, + "step": 19054 + }, + { + "epoch": 1.6, + "learning_rate": 9.359612510571339e-06, + "loss": 1.2183, + "step": 19055 + }, + { + "epoch": 1.6, + "learning_rate": 9.358705405915144e-06, + "loss": 1.2177, + "step": 19056 + }, + { + "epoch": 1.6, + "learning_rate": 9.357798306557526e-06, + "loss": 1.2343, + "step": 19057 + }, + { + "epoch": 1.6, + "learning_rate": 9.356891212505988e-06, + "loss": 1.223, + "step": 19058 + }, + { + "epoch": 1.6, + "learning_rate": 9.355984123768013e-06, + "loss": 1.1948, + "step": 19059 + }, + { + "epoch": 1.6, + "learning_rate": 9.35507704035111e-06, + "loss": 1.1489, + "step": 19060 + }, + { + "epoch": 1.6, + "learning_rate": 9.354169962262765e-06, + "loss": 1.1359, + "step": 19061 + }, + { + "epoch": 1.6, + "learning_rate": 9.353262889510477e-06, + "loss": 1.1556, + "step": 19062 + }, + { + "epoch": 1.6, + "learning_rate": 9.352355822101733e-06, + "loss": 1.1679, + "step": 19063 + }, + { + "epoch": 1.61, + "learning_rate": 9.351448760044034e-06, + "loss": 1.1374, + "step": 19064 + }, + { + "epoch": 1.61, + "learning_rate": 9.350541703344875e-06, + "loss": 1.1014, + "step": 19065 + }, + { + "epoch": 1.61, + "learning_rate": 9.349634652011745e-06, + "loss": 1.1605, + "step": 19066 + }, + { + "epoch": 1.61, + "learning_rate": 9.348727606052146e-06, + "loss": 1.2007, + "step": 19067 + }, + { + "epoch": 1.61, + "learning_rate": 9.347820565473565e-06, + "loss": 1.2434, + "step": 19068 + }, + { + "epoch": 1.61, + "learning_rate": 9.346913530283502e-06, + "loss": 1.2059, + "step": 19069 + }, + { + "epoch": 1.61, + "learning_rate": 9.346006500489442e-06, + "loss": 1.1646, + "step": 19070 + }, + { + "epoch": 1.61, + "learning_rate": 9.345099476098894e-06, + "loss": 1.1338, + "step": 19071 + }, + { + "epoch": 1.61, + "learning_rate": 9.344192457119337e-06, + "loss": 1.2093, + "step": 19072 + }, + { + "epoch": 1.61, + "learning_rate": 9.343285443558278e-06, + "loss": 1.2086, + "step": 19073 + }, + { + "epoch": 1.61, + "learning_rate": 9.342378435423197e-06, + "loss": 1.1497, + "step": 19074 + }, + { + "epoch": 1.61, + "learning_rate": 9.3414714327216e-06, + "loss": 1.1525, + "step": 19075 + }, + { + "epoch": 1.61, + "learning_rate": 9.34056443546098e-06, + "loss": 1.1104, + "step": 19076 + }, + { + "epoch": 1.61, + "learning_rate": 9.339657443648823e-06, + "loss": 1.2647, + "step": 19077 + }, + { + "epoch": 1.61, + "learning_rate": 9.338750457292632e-06, + "loss": 1.1397, + "step": 19078 + }, + { + "epoch": 1.61, + "learning_rate": 9.33784347639989e-06, + "loss": 1.1548, + "step": 19079 + }, + { + "epoch": 1.61, + "learning_rate": 9.336936500978101e-06, + "loss": 1.2005, + "step": 19080 + }, + { + "epoch": 1.61, + "learning_rate": 9.336029531034756e-06, + "loss": 1.1912, + "step": 19081 + }, + { + "epoch": 1.61, + "learning_rate": 9.335122566577347e-06, + "loss": 1.2085, + "step": 19082 + }, + { + "epoch": 1.61, + "learning_rate": 9.334215607613364e-06, + "loss": 1.1418, + "step": 19083 + }, + { + "epoch": 1.61, + "learning_rate": 9.33330865415031e-06, + "loss": 1.205, + "step": 19084 + }, + { + "epoch": 1.61, + "learning_rate": 9.332401706195671e-06, + "loss": 1.1356, + "step": 19085 + }, + { + "epoch": 1.61, + "learning_rate": 9.331494763756946e-06, + "loss": 1.108, + "step": 19086 + }, + { + "epoch": 1.61, + "learning_rate": 9.330587826841618e-06, + "loss": 1.2131, + "step": 19087 + }, + { + "epoch": 1.61, + "learning_rate": 9.32968089545719e-06, + "loss": 1.2609, + "step": 19088 + }, + { + "epoch": 1.61, + "learning_rate": 9.328773969611158e-06, + "loss": 1.2341, + "step": 19089 + }, + { + "epoch": 1.61, + "learning_rate": 9.327867049311005e-06, + "loss": 1.2226, + "step": 19090 + }, + { + "epoch": 1.61, + "learning_rate": 9.326960134564233e-06, + "loss": 1.2872, + "step": 19091 + }, + { + "epoch": 1.61, + "learning_rate": 9.326053225378329e-06, + "loss": 1.1646, + "step": 19092 + }, + { + "epoch": 1.61, + "learning_rate": 9.325146321760792e-06, + "loss": 1.1582, + "step": 19093 + }, + { + "epoch": 1.61, + "learning_rate": 9.324239423719107e-06, + "loss": 1.1636, + "step": 19094 + }, + { + "epoch": 1.61, + "learning_rate": 9.323332531260777e-06, + "loss": 1.1983, + "step": 19095 + }, + { + "epoch": 1.61, + "learning_rate": 9.322425644393289e-06, + "loss": 1.1181, + "step": 19096 + }, + { + "epoch": 1.61, + "learning_rate": 9.321518763124139e-06, + "loss": 1.2802, + "step": 19097 + }, + { + "epoch": 1.61, + "learning_rate": 9.320611887460814e-06, + "loss": 1.3359, + "step": 19098 + }, + { + "epoch": 1.61, + "learning_rate": 9.319705017410813e-06, + "loss": 1.2121, + "step": 19099 + }, + { + "epoch": 1.61, + "learning_rate": 9.318798152981627e-06, + "loss": 1.2112, + "step": 19100 + }, + { + "epoch": 1.61, + "learning_rate": 9.317891294180746e-06, + "loss": 1.1857, + "step": 19101 + }, + { + "epoch": 1.61, + "learning_rate": 9.31698444101567e-06, + "loss": 1.1778, + "step": 19102 + }, + { + "epoch": 1.61, + "learning_rate": 9.316077593493885e-06, + "loss": 1.2123, + "step": 19103 + }, + { + "epoch": 1.61, + "learning_rate": 9.315170751622887e-06, + "loss": 1.1831, + "step": 19104 + }, + { + "epoch": 1.61, + "learning_rate": 9.314263915410163e-06, + "loss": 1.1768, + "step": 19105 + }, + { + "epoch": 1.61, + "learning_rate": 9.313357084863215e-06, + "loss": 1.2019, + "step": 19106 + }, + { + "epoch": 1.61, + "learning_rate": 9.312450259989529e-06, + "loss": 1.2097, + "step": 19107 + }, + { + "epoch": 1.61, + "learning_rate": 9.3115434407966e-06, + "loss": 1.1186, + "step": 19108 + }, + { + "epoch": 1.61, + "learning_rate": 9.310636627291915e-06, + "loss": 1.143, + "step": 19109 + }, + { + "epoch": 1.61, + "learning_rate": 9.309729819482976e-06, + "loss": 1.1935, + "step": 19110 + }, + { + "epoch": 1.61, + "learning_rate": 9.308823017377268e-06, + "loss": 1.2205, + "step": 19111 + }, + { + "epoch": 1.61, + "learning_rate": 9.307916220982287e-06, + "loss": 1.1134, + "step": 19112 + }, + { + "epoch": 1.61, + "learning_rate": 9.307009430305521e-06, + "loss": 1.2215, + "step": 19113 + }, + { + "epoch": 1.61, + "learning_rate": 9.306102645354465e-06, + "loss": 1.1854, + "step": 19114 + }, + { + "epoch": 1.61, + "learning_rate": 9.305195866136613e-06, + "loss": 1.188, + "step": 19115 + }, + { + "epoch": 1.61, + "learning_rate": 9.304289092659453e-06, + "loss": 1.1999, + "step": 19116 + }, + { + "epoch": 1.61, + "learning_rate": 9.303382324930482e-06, + "loss": 1.1476, + "step": 19117 + }, + { + "epoch": 1.61, + "learning_rate": 9.302475562957187e-06, + "loss": 1.0732, + "step": 19118 + }, + { + "epoch": 1.61, + "learning_rate": 9.301568806747064e-06, + "loss": 1.1857, + "step": 19119 + }, + { + "epoch": 1.61, + "learning_rate": 9.3006620563076e-06, + "loss": 1.1809, + "step": 19120 + }, + { + "epoch": 1.61, + "learning_rate": 9.299755311646293e-06, + "loss": 1.2398, + "step": 19121 + }, + { + "epoch": 1.61, + "learning_rate": 9.29884857277063e-06, + "loss": 1.1324, + "step": 19122 + }, + { + "epoch": 1.61, + "learning_rate": 9.297941839688108e-06, + "loss": 1.1375, + "step": 19123 + }, + { + "epoch": 1.61, + "learning_rate": 9.297035112406209e-06, + "loss": 1.1987, + "step": 19124 + }, + { + "epoch": 1.61, + "learning_rate": 9.296128390932433e-06, + "loss": 1.2269, + "step": 19125 + }, + { + "epoch": 1.61, + "learning_rate": 9.295221675274272e-06, + "loss": 1.2132, + "step": 19126 + }, + { + "epoch": 1.61, + "learning_rate": 9.29431496543921e-06, + "loss": 1.1009, + "step": 19127 + }, + { + "epoch": 1.61, + "learning_rate": 9.293408261434752e-06, + "loss": 1.2442, + "step": 19128 + }, + { + "epoch": 1.61, + "learning_rate": 9.292501563268374e-06, + "loss": 1.1794, + "step": 19129 + }, + { + "epoch": 1.61, + "learning_rate": 9.291594870947579e-06, + "loss": 1.151, + "step": 19130 + }, + { + "epoch": 1.61, + "learning_rate": 9.290688184479852e-06, + "loss": 1.2571, + "step": 19131 + }, + { + "epoch": 1.61, + "learning_rate": 9.289781503872687e-06, + "loss": 1.1373, + "step": 19132 + }, + { + "epoch": 1.61, + "learning_rate": 9.288874829133571e-06, + "loss": 1.1507, + "step": 19133 + }, + { + "epoch": 1.61, + "learning_rate": 9.287968160270001e-06, + "loss": 1.246, + "step": 19134 + }, + { + "epoch": 1.61, + "learning_rate": 9.287061497289467e-06, + "loss": 1.2561, + "step": 19135 + }, + { + "epoch": 1.61, + "learning_rate": 9.28615484019946e-06, + "loss": 1.143, + "step": 19136 + }, + { + "epoch": 1.61, + "learning_rate": 9.285248189007466e-06, + "loss": 1.1321, + "step": 19137 + }, + { + "epoch": 1.61, + "learning_rate": 9.284341543720982e-06, + "loss": 1.1424, + "step": 19138 + }, + { + "epoch": 1.61, + "learning_rate": 9.283434904347499e-06, + "loss": 1.1881, + "step": 19139 + }, + { + "epoch": 1.61, + "learning_rate": 9.282528270894502e-06, + "loss": 1.166, + "step": 19140 + }, + { + "epoch": 1.61, + "learning_rate": 9.281621643369491e-06, + "loss": 1.2691, + "step": 19141 + }, + { + "epoch": 1.61, + "learning_rate": 9.28071502177995e-06, + "loss": 1.2081, + "step": 19142 + }, + { + "epoch": 1.61, + "learning_rate": 9.279808406133373e-06, + "loss": 1.1294, + "step": 19143 + }, + { + "epoch": 1.61, + "learning_rate": 9.278901796437245e-06, + "loss": 1.2278, + "step": 19144 + }, + { + "epoch": 1.61, + "learning_rate": 9.277995192699067e-06, + "loss": 1.246, + "step": 19145 + }, + { + "epoch": 1.61, + "learning_rate": 9.27708859492632e-06, + "loss": 1.1681, + "step": 19146 + }, + { + "epoch": 1.61, + "learning_rate": 9.276182003126502e-06, + "loss": 1.1762, + "step": 19147 + }, + { + "epoch": 1.61, + "learning_rate": 9.275275417307094e-06, + "loss": 1.1684, + "step": 19148 + }, + { + "epoch": 1.61, + "learning_rate": 9.274368837475596e-06, + "loss": 1.1623, + "step": 19149 + }, + { + "epoch": 1.61, + "learning_rate": 9.273462263639496e-06, + "loss": 1.3177, + "step": 19150 + }, + { + "epoch": 1.61, + "learning_rate": 9.272555695806279e-06, + "loss": 1.1619, + "step": 19151 + }, + { + "epoch": 1.61, + "learning_rate": 9.271649133983443e-06, + "loss": 1.106, + "step": 19152 + }, + { + "epoch": 1.61, + "learning_rate": 9.270742578178475e-06, + "loss": 1.2097, + "step": 19153 + }, + { + "epoch": 1.61, + "learning_rate": 9.269836028398866e-06, + "loss": 1.1259, + "step": 19154 + }, + { + "epoch": 1.61, + "learning_rate": 9.268929484652101e-06, + "loss": 1.173, + "step": 19155 + }, + { + "epoch": 1.61, + "learning_rate": 9.268022946945681e-06, + "loss": 1.165, + "step": 19156 + }, + { + "epoch": 1.61, + "learning_rate": 9.267116415287085e-06, + "loss": 1.1999, + "step": 19157 + }, + { + "epoch": 1.61, + "learning_rate": 9.266209889683811e-06, + "loss": 1.2272, + "step": 19158 + }, + { + "epoch": 1.61, + "learning_rate": 9.265303370143341e-06, + "loss": 1.1338, + "step": 19159 + }, + { + "epoch": 1.61, + "learning_rate": 9.264396856673175e-06, + "loss": 1.2022, + "step": 19160 + }, + { + "epoch": 1.61, + "learning_rate": 9.263490349280795e-06, + "loss": 1.1106, + "step": 19161 + }, + { + "epoch": 1.61, + "learning_rate": 9.262583847973694e-06, + "loss": 1.2346, + "step": 19162 + }, + { + "epoch": 1.61, + "learning_rate": 9.26167735275936e-06, + "loss": 1.2098, + "step": 19163 + }, + { + "epoch": 1.61, + "learning_rate": 9.260770863645283e-06, + "loss": 1.2034, + "step": 19164 + }, + { + "epoch": 1.61, + "learning_rate": 9.259864380638958e-06, + "loss": 1.1463, + "step": 19165 + }, + { + "epoch": 1.61, + "learning_rate": 9.258957903747866e-06, + "loss": 1.235, + "step": 19166 + }, + { + "epoch": 1.61, + "learning_rate": 9.258051432979503e-06, + "loss": 1.2662, + "step": 19167 + }, + { + "epoch": 1.61, + "learning_rate": 9.257144968341355e-06, + "loss": 1.1427, + "step": 19168 + }, + { + "epoch": 1.61, + "learning_rate": 9.256238509840918e-06, + "loss": 1.1308, + "step": 19169 + }, + { + "epoch": 1.61, + "learning_rate": 9.25533205748567e-06, + "loss": 1.2028, + "step": 19170 + }, + { + "epoch": 1.61, + "learning_rate": 9.254425611283113e-06, + "loss": 1.1552, + "step": 19171 + }, + { + "epoch": 1.61, + "learning_rate": 9.253519171240727e-06, + "loss": 1.173, + "step": 19172 + }, + { + "epoch": 1.61, + "learning_rate": 9.252612737366004e-06, + "loss": 1.097, + "step": 19173 + }, + { + "epoch": 1.61, + "learning_rate": 9.251706309666431e-06, + "loss": 1.1221, + "step": 19174 + }, + { + "epoch": 1.61, + "learning_rate": 9.250799888149504e-06, + "loss": 1.2193, + "step": 19175 + }, + { + "epoch": 1.61, + "learning_rate": 9.249893472822709e-06, + "loss": 1.172, + "step": 19176 + }, + { + "epoch": 1.61, + "learning_rate": 9.24898706369353e-06, + "loss": 1.1138, + "step": 19177 + }, + { + "epoch": 1.61, + "learning_rate": 9.248080660769466e-06, + "loss": 1.2446, + "step": 19178 + }, + { + "epoch": 1.61, + "learning_rate": 9.247174264057994e-06, + "loss": 1.1569, + "step": 19179 + }, + { + "epoch": 1.61, + "learning_rate": 9.246267873566615e-06, + "loss": 1.2374, + "step": 19180 + }, + { + "epoch": 1.61, + "learning_rate": 9.245361489302808e-06, + "loss": 1.1731, + "step": 19181 + }, + { + "epoch": 1.61, + "learning_rate": 9.244455111274069e-06, + "loss": 1.2527, + "step": 19182 + }, + { + "epoch": 1.62, + "learning_rate": 9.24354873948788e-06, + "loss": 1.1855, + "step": 19183 + }, + { + "epoch": 1.62, + "learning_rate": 9.242642373951736e-06, + "loss": 1.1591, + "step": 19184 + }, + { + "epoch": 1.62, + "learning_rate": 9.241736014673121e-06, + "loss": 1.226, + "step": 19185 + }, + { + "epoch": 1.62, + "learning_rate": 9.24082966165953e-06, + "loss": 1.1191, + "step": 19186 + }, + { + "epoch": 1.62, + "learning_rate": 9.239923314918441e-06, + "loss": 1.1836, + "step": 19187 + }, + { + "epoch": 1.62, + "learning_rate": 9.239016974457351e-06, + "loss": 1.2212, + "step": 19188 + }, + { + "epoch": 1.62, + "learning_rate": 9.23811064028375e-06, + "loss": 1.2716, + "step": 19189 + }, + { + "epoch": 1.62, + "learning_rate": 9.237204312405117e-06, + "loss": 1.1607, + "step": 19190 + }, + { + "epoch": 1.62, + "learning_rate": 9.23629799082895e-06, + "loss": 1.2217, + "step": 19191 + }, + { + "epoch": 1.62, + "learning_rate": 9.235391675562733e-06, + "loss": 1.1049, + "step": 19192 + }, + { + "epoch": 1.62, + "learning_rate": 9.234485366613957e-06, + "loss": 1.153, + "step": 19193 + }, + { + "epoch": 1.62, + "learning_rate": 9.233579063990102e-06, + "loss": 1.21, + "step": 19194 + }, + { + "epoch": 1.62, + "learning_rate": 9.232672767698666e-06, + "loss": 1.1706, + "step": 19195 + }, + { + "epoch": 1.62, + "learning_rate": 9.231766477747134e-06, + "loss": 1.2596, + "step": 19196 + }, + { + "epoch": 1.62, + "learning_rate": 9.230860194142991e-06, + "loss": 1.195, + "step": 19197 + }, + { + "epoch": 1.62, + "learning_rate": 9.229953916893727e-06, + "loss": 1.0526, + "step": 19198 + }, + { + "epoch": 1.62, + "learning_rate": 9.229047646006832e-06, + "loss": 1.2672, + "step": 19199 + }, + { + "epoch": 1.62, + "learning_rate": 9.228141381489792e-06, + "loss": 1.1748, + "step": 19200 + }, + { + "epoch": 1.62, + "learning_rate": 9.22723512335009e-06, + "loss": 1.1678, + "step": 19201 + }, + { + "epoch": 1.62, + "learning_rate": 9.226328871595226e-06, + "loss": 1.2222, + "step": 19202 + }, + { + "epoch": 1.62, + "learning_rate": 9.225422626232676e-06, + "loss": 1.1354, + "step": 19203 + }, + { + "epoch": 1.62, + "learning_rate": 9.224516387269935e-06, + "loss": 1.1224, + "step": 19204 + }, + { + "epoch": 1.62, + "learning_rate": 9.223610154714484e-06, + "loss": 1.1288, + "step": 19205 + }, + { + "epoch": 1.62, + "learning_rate": 9.222703928573818e-06, + "loss": 1.2006, + "step": 19206 + }, + { + "epoch": 1.62, + "learning_rate": 9.22179770885542e-06, + "loss": 1.1651, + "step": 19207 + }, + { + "epoch": 1.62, + "learning_rate": 9.220891495566779e-06, + "loss": 1.1671, + "step": 19208 + }, + { + "epoch": 1.62, + "learning_rate": 9.219985288715378e-06, + "loss": 1.1113, + "step": 19209 + }, + { + "epoch": 1.62, + "learning_rate": 9.219079088308713e-06, + "loss": 1.1417, + "step": 19210 + }, + { + "epoch": 1.62, + "learning_rate": 9.218172894354263e-06, + "loss": 1.1806, + "step": 19211 + }, + { + "epoch": 1.62, + "learning_rate": 9.217266706859522e-06, + "loss": 1.1629, + "step": 19212 + }, + { + "epoch": 1.62, + "learning_rate": 9.21636052583197e-06, + "loss": 1.206, + "step": 19213 + }, + { + "epoch": 1.62, + "learning_rate": 9.2154543512791e-06, + "loss": 1.1915, + "step": 19214 + }, + { + "epoch": 1.62, + "learning_rate": 9.2145481832084e-06, + "loss": 1.1546, + "step": 19215 + }, + { + "epoch": 1.62, + "learning_rate": 9.213642021627347e-06, + "loss": 1.1557, + "step": 19216 + }, + { + "epoch": 1.62, + "learning_rate": 9.212735866543443e-06, + "loss": 1.241, + "step": 19217 + }, + { + "epoch": 1.62, + "learning_rate": 9.211829717964164e-06, + "loss": 1.1489, + "step": 19218 + }, + { + "epoch": 1.62, + "learning_rate": 9.210923575897001e-06, + "loss": 1.1954, + "step": 19219 + }, + { + "epoch": 1.62, + "learning_rate": 9.210017440349439e-06, + "loss": 1.1302, + "step": 19220 + }, + { + "epoch": 1.62, + "learning_rate": 9.209111311328967e-06, + "loss": 1.2005, + "step": 19221 + }, + { + "epoch": 1.62, + "learning_rate": 9.20820518884307e-06, + "loss": 1.1877, + "step": 19222 + }, + { + "epoch": 1.62, + "learning_rate": 9.207299072899235e-06, + "loss": 1.1549, + "step": 19223 + }, + { + "epoch": 1.62, + "learning_rate": 9.206392963504946e-06, + "loss": 1.125, + "step": 19224 + }, + { + "epoch": 1.62, + "learning_rate": 9.205486860667694e-06, + "loss": 1.1755, + "step": 19225 + }, + { + "epoch": 1.62, + "learning_rate": 9.204580764394966e-06, + "loss": 1.1189, + "step": 19226 + }, + { + "epoch": 1.62, + "learning_rate": 9.203674674694243e-06, + "loss": 1.1689, + "step": 19227 + }, + { + "epoch": 1.62, + "learning_rate": 9.202768591573018e-06, + "loss": 1.1572, + "step": 19228 + }, + { + "epoch": 1.62, + "learning_rate": 9.201862515038773e-06, + "loss": 1.2051, + "step": 19229 + }, + { + "epoch": 1.62, + "learning_rate": 9.200956445098997e-06, + "loss": 1.1765, + "step": 19230 + }, + { + "epoch": 1.62, + "learning_rate": 9.20005038176117e-06, + "loss": 1.2401, + "step": 19231 + }, + { + "epoch": 1.62, + "learning_rate": 9.199144325032788e-06, + "loss": 1.1439, + "step": 19232 + }, + { + "epoch": 1.62, + "learning_rate": 9.198238274921328e-06, + "loss": 1.2264, + "step": 19233 + }, + { + "epoch": 1.62, + "learning_rate": 9.197332231434282e-06, + "loss": 1.1799, + "step": 19234 + }, + { + "epoch": 1.62, + "learning_rate": 9.196426194579133e-06, + "loss": 1.2398, + "step": 19235 + }, + { + "epoch": 1.62, + "learning_rate": 9.195520164363369e-06, + "loss": 1.2338, + "step": 19236 + }, + { + "epoch": 1.62, + "learning_rate": 9.194614140794472e-06, + "loss": 1.1682, + "step": 19237 + }, + { + "epoch": 1.62, + "learning_rate": 9.193708123879932e-06, + "loss": 1.1959, + "step": 19238 + }, + { + "epoch": 1.62, + "learning_rate": 9.192802113627235e-06, + "loss": 1.1837, + "step": 19239 + }, + { + "epoch": 1.62, + "learning_rate": 9.191896110043862e-06, + "loss": 1.2261, + "step": 19240 + }, + { + "epoch": 1.62, + "learning_rate": 9.190990113137308e-06, + "loss": 1.2445, + "step": 19241 + }, + { + "epoch": 1.62, + "learning_rate": 9.190084122915048e-06, + "loss": 1.1982, + "step": 19242 + }, + { + "epoch": 1.62, + "learning_rate": 9.189178139384575e-06, + "loss": 1.1935, + "step": 19243 + }, + { + "epoch": 1.62, + "learning_rate": 9.188272162553368e-06, + "loss": 1.2206, + "step": 19244 + }, + { + "epoch": 1.62, + "learning_rate": 9.18736619242892e-06, + "loss": 1.1744, + "step": 19245 + }, + { + "epoch": 1.62, + "learning_rate": 9.18646022901871e-06, + "loss": 1.2122, + "step": 19246 + }, + { + "epoch": 1.62, + "learning_rate": 9.185554272330228e-06, + "loss": 1.1115, + "step": 19247 + }, + { + "epoch": 1.62, + "learning_rate": 9.184648322370955e-06, + "loss": 1.1125, + "step": 19248 + }, + { + "epoch": 1.62, + "learning_rate": 9.18374237914838e-06, + "loss": 1.2441, + "step": 19249 + }, + { + "epoch": 1.62, + "learning_rate": 9.182836442669986e-06, + "loss": 1.1664, + "step": 19250 + }, + { + "epoch": 1.62, + "learning_rate": 9.181930512943256e-06, + "loss": 1.1855, + "step": 19251 + }, + { + "epoch": 1.62, + "learning_rate": 9.181024589975683e-06, + "loss": 1.1931, + "step": 19252 + }, + { + "epoch": 1.62, + "learning_rate": 9.180118673774744e-06, + "loss": 1.2433, + "step": 19253 + }, + { + "epoch": 1.62, + "learning_rate": 9.179212764347928e-06, + "loss": 1.2134, + "step": 19254 + }, + { + "epoch": 1.62, + "learning_rate": 9.178306861702717e-06, + "loss": 1.2533, + "step": 19255 + }, + { + "epoch": 1.62, + "learning_rate": 9.1774009658466e-06, + "loss": 1.2746, + "step": 19256 + }, + { + "epoch": 1.62, + "learning_rate": 9.17649507678706e-06, + "loss": 1.2364, + "step": 19257 + }, + { + "epoch": 1.62, + "learning_rate": 9.175589194531583e-06, + "loss": 1.1482, + "step": 19258 + }, + { + "epoch": 1.62, + "learning_rate": 9.174683319087646e-06, + "loss": 1.2774, + "step": 19259 + }, + { + "epoch": 1.62, + "learning_rate": 9.173777450462744e-06, + "loss": 1.1043, + "step": 19260 + }, + { + "epoch": 1.62, + "learning_rate": 9.172871588664353e-06, + "loss": 1.1678, + "step": 19261 + }, + { + "epoch": 1.62, + "learning_rate": 9.171965733699968e-06, + "loss": 1.2744, + "step": 19262 + }, + { + "epoch": 1.62, + "learning_rate": 9.17105988557706e-06, + "loss": 1.2259, + "step": 19263 + }, + { + "epoch": 1.62, + "learning_rate": 9.170154044303123e-06, + "loss": 1.3226, + "step": 19264 + }, + { + "epoch": 1.62, + "learning_rate": 9.169248209885642e-06, + "loss": 1.249, + "step": 19265 + }, + { + "epoch": 1.62, + "learning_rate": 9.168342382332093e-06, + "loss": 1.1547, + "step": 19266 + }, + { + "epoch": 1.62, + "learning_rate": 9.167436561649971e-06, + "loss": 1.2023, + "step": 19267 + }, + { + "epoch": 1.62, + "learning_rate": 9.166530747846751e-06, + "loss": 1.2245, + "step": 19268 + }, + { + "epoch": 1.62, + "learning_rate": 9.165624940929924e-06, + "loss": 1.2512, + "step": 19269 + }, + { + "epoch": 1.62, + "learning_rate": 9.164719140906964e-06, + "loss": 1.262, + "step": 19270 + }, + { + "epoch": 1.62, + "learning_rate": 9.163813347785368e-06, + "loss": 1.2144, + "step": 19271 + }, + { + "epoch": 1.62, + "learning_rate": 9.162907561572613e-06, + "loss": 1.2131, + "step": 19272 + }, + { + "epoch": 1.62, + "learning_rate": 9.162001782276185e-06, + "loss": 1.2169, + "step": 19273 + }, + { + "epoch": 1.62, + "learning_rate": 9.16109600990356e-06, + "loss": 1.167, + "step": 19274 + }, + { + "epoch": 1.62, + "learning_rate": 9.160190244462233e-06, + "loss": 1.2378, + "step": 19275 + }, + { + "epoch": 1.62, + "learning_rate": 9.159284485959683e-06, + "loss": 1.1887, + "step": 19276 + }, + { + "epoch": 1.62, + "learning_rate": 9.15837873440339e-06, + "loss": 1.1494, + "step": 19277 + }, + { + "epoch": 1.62, + "learning_rate": 9.157472989800847e-06, + "loss": 1.1209, + "step": 19278 + }, + { + "epoch": 1.62, + "learning_rate": 9.156567252159528e-06, + "loss": 1.2045, + "step": 19279 + }, + { + "epoch": 1.62, + "learning_rate": 9.155661521486923e-06, + "loss": 1.1961, + "step": 19280 + }, + { + "epoch": 1.62, + "learning_rate": 9.154755797790508e-06, + "loss": 1.2091, + "step": 19281 + }, + { + "epoch": 1.62, + "learning_rate": 9.153850081077777e-06, + "loss": 1.1927, + "step": 19282 + }, + { + "epoch": 1.62, + "learning_rate": 9.152944371356202e-06, + "loss": 1.1295, + "step": 19283 + }, + { + "epoch": 1.62, + "learning_rate": 9.152038668633276e-06, + "loss": 1.1174, + "step": 19284 + }, + { + "epoch": 1.62, + "learning_rate": 9.151132972916477e-06, + "loss": 1.2315, + "step": 19285 + }, + { + "epoch": 1.62, + "learning_rate": 9.15022728421329e-06, + "loss": 1.1534, + "step": 19286 + }, + { + "epoch": 1.62, + "learning_rate": 9.149321602531191e-06, + "loss": 1.2697, + "step": 19287 + }, + { + "epoch": 1.62, + "learning_rate": 9.148415927877675e-06, + "loss": 1.2437, + "step": 19288 + }, + { + "epoch": 1.62, + "learning_rate": 9.147510260260218e-06, + "loss": 1.1545, + "step": 19289 + }, + { + "epoch": 1.62, + "learning_rate": 9.146604599686302e-06, + "loss": 1.1115, + "step": 19290 + }, + { + "epoch": 1.62, + "learning_rate": 9.145698946163414e-06, + "loss": 1.1675, + "step": 19291 + }, + { + "epoch": 1.62, + "learning_rate": 9.144793299699036e-06, + "loss": 1.1826, + "step": 19292 + }, + { + "epoch": 1.62, + "learning_rate": 9.143887660300648e-06, + "loss": 1.0983, + "step": 19293 + }, + { + "epoch": 1.62, + "learning_rate": 9.14298202797573e-06, + "loss": 1.176, + "step": 19294 + }, + { + "epoch": 1.62, + "learning_rate": 9.142076402731775e-06, + "loss": 1.2173, + "step": 19295 + }, + { + "epoch": 1.62, + "learning_rate": 9.141170784576257e-06, + "loss": 1.1644, + "step": 19296 + }, + { + "epoch": 1.62, + "learning_rate": 9.140265173516663e-06, + "loss": 1.2226, + "step": 19297 + }, + { + "epoch": 1.62, + "learning_rate": 9.139359569560468e-06, + "loss": 1.1131, + "step": 19298 + }, + { + "epoch": 1.62, + "learning_rate": 9.138453972715166e-06, + "loss": 1.2104, + "step": 19299 + }, + { + "epoch": 1.62, + "learning_rate": 9.137548382988229e-06, + "loss": 1.2403, + "step": 19300 + }, + { + "epoch": 1.62, + "learning_rate": 9.13664280038714e-06, + "loss": 1.207, + "step": 19301 + }, + { + "epoch": 1.63, + "learning_rate": 9.135737224919392e-06, + "loss": 1.2053, + "step": 19302 + }, + { + "epoch": 1.63, + "learning_rate": 9.134831656592455e-06, + "loss": 1.1672, + "step": 19303 + }, + { + "epoch": 1.63, + "learning_rate": 9.133926095413818e-06, + "loss": 1.2428, + "step": 19304 + }, + { + "epoch": 1.63, + "learning_rate": 9.133020541390958e-06, + "loss": 1.11, + "step": 19305 + }, + { + "epoch": 1.63, + "learning_rate": 9.132114994531362e-06, + "loss": 1.163, + "step": 19306 + }, + { + "epoch": 1.63, + "learning_rate": 9.13120945484251e-06, + "loss": 1.216, + "step": 19307 + }, + { + "epoch": 1.63, + "learning_rate": 9.130303922331885e-06, + "loss": 1.1254, + "step": 19308 + }, + { + "epoch": 1.63, + "learning_rate": 9.129398397006962e-06, + "loss": 1.2362, + "step": 19309 + }, + { + "epoch": 1.63, + "learning_rate": 9.128492878875232e-06, + "loss": 1.0855, + "step": 19310 + }, + { + "epoch": 1.63, + "learning_rate": 9.127587367944172e-06, + "loss": 1.1982, + "step": 19311 + }, + { + "epoch": 1.63, + "learning_rate": 9.126681864221267e-06, + "loss": 1.1742, + "step": 19312 + }, + { + "epoch": 1.63, + "learning_rate": 9.12577636771399e-06, + "loss": 1.1391, + "step": 19313 + }, + { + "epoch": 1.63, + "learning_rate": 9.12487087842983e-06, + "loss": 1.222, + "step": 19314 + }, + { + "epoch": 1.63, + "learning_rate": 9.123965396376271e-06, + "loss": 1.1924, + "step": 19315 + }, + { + "epoch": 1.63, + "learning_rate": 9.123059921560785e-06, + "loss": 1.1449, + "step": 19316 + }, + { + "epoch": 1.63, + "learning_rate": 9.122154453990864e-06, + "loss": 1.1555, + "step": 19317 + }, + { + "epoch": 1.63, + "learning_rate": 9.121248993673981e-06, + "loss": 1.2132, + "step": 19318 + }, + { + "epoch": 1.63, + "learning_rate": 9.12034354061762e-06, + "loss": 1.2645, + "step": 19319 + }, + { + "epoch": 1.63, + "learning_rate": 9.11943809482926e-06, + "loss": 1.1716, + "step": 19320 + }, + { + "epoch": 1.63, + "learning_rate": 9.11853265631639e-06, + "loss": 1.2051, + "step": 19321 + }, + { + "epoch": 1.63, + "learning_rate": 9.117627225086481e-06, + "loss": 1.2325, + "step": 19322 + }, + { + "epoch": 1.63, + "learning_rate": 9.11672180114702e-06, + "loss": 1.2493, + "step": 19323 + }, + { + "epoch": 1.63, + "learning_rate": 9.115816384505484e-06, + "loss": 1.3276, + "step": 19324 + }, + { + "epoch": 1.63, + "learning_rate": 9.114910975169356e-06, + "loss": 1.2209, + "step": 19325 + }, + { + "epoch": 1.63, + "learning_rate": 9.114005573146121e-06, + "loss": 1.1901, + "step": 19326 + }, + { + "epoch": 1.63, + "learning_rate": 9.11310017844325e-06, + "loss": 1.1622, + "step": 19327 + }, + { + "epoch": 1.63, + "learning_rate": 9.112194791068234e-06, + "loss": 1.1295, + "step": 19328 + }, + { + "epoch": 1.63, + "learning_rate": 9.111289411028545e-06, + "loss": 1.0825, + "step": 19329 + }, + { + "epoch": 1.63, + "learning_rate": 9.110384038331671e-06, + "loss": 1.2216, + "step": 19330 + }, + { + "epoch": 1.63, + "learning_rate": 9.109478672985084e-06, + "loss": 1.2319, + "step": 19331 + }, + { + "epoch": 1.63, + "learning_rate": 9.108573314996277e-06, + "loss": 1.0873, + "step": 19332 + }, + { + "epoch": 1.63, + "learning_rate": 9.107667964372715e-06, + "loss": 1.1452, + "step": 19333 + }, + { + "epoch": 1.63, + "learning_rate": 9.10676262112189e-06, + "loss": 1.2518, + "step": 19334 + }, + { + "epoch": 1.63, + "learning_rate": 9.105857285251277e-06, + "loss": 1.2056, + "step": 19335 + }, + { + "epoch": 1.63, + "learning_rate": 9.104951956768357e-06, + "loss": 1.2004, + "step": 19336 + }, + { + "epoch": 1.63, + "learning_rate": 9.104046635680607e-06, + "loss": 1.1984, + "step": 19337 + }, + { + "epoch": 1.63, + "learning_rate": 9.103141321995514e-06, + "loss": 1.248, + "step": 19338 + }, + { + "epoch": 1.63, + "learning_rate": 9.102236015720556e-06, + "loss": 1.201, + "step": 19339 + }, + { + "epoch": 1.63, + "learning_rate": 9.101330716863207e-06, + "loss": 1.1522, + "step": 19340 + }, + { + "epoch": 1.63, + "learning_rate": 9.100425425430955e-06, + "loss": 1.2553, + "step": 19341 + }, + { + "epoch": 1.63, + "learning_rate": 9.099520141431274e-06, + "loss": 1.2209, + "step": 19342 + }, + { + "epoch": 1.63, + "learning_rate": 9.098614864871649e-06, + "loss": 1.1413, + "step": 19343 + }, + { + "epoch": 1.63, + "learning_rate": 9.09770959575955e-06, + "loss": 1.1492, + "step": 19344 + }, + { + "epoch": 1.63, + "learning_rate": 9.09680433410247e-06, + "loss": 1.1387, + "step": 19345 + }, + { + "epoch": 1.63, + "learning_rate": 9.095899079907878e-06, + "loss": 1.1896, + "step": 19346 + }, + { + "epoch": 1.63, + "learning_rate": 9.094993833183258e-06, + "loss": 1.2092, + "step": 19347 + }, + { + "epoch": 1.63, + "learning_rate": 9.094088593936086e-06, + "loss": 1.233, + "step": 19348 + }, + { + "epoch": 1.63, + "learning_rate": 9.09318336217385e-06, + "loss": 1.1577, + "step": 19349 + }, + { + "epoch": 1.63, + "learning_rate": 9.092278137904019e-06, + "loss": 1.3018, + "step": 19350 + }, + { + "epoch": 1.63, + "learning_rate": 9.091372921134074e-06, + "loss": 1.123, + "step": 19351 + }, + { + "epoch": 1.63, + "learning_rate": 9.090467711871502e-06, + "loss": 1.2429, + "step": 19352 + }, + { + "epoch": 1.63, + "learning_rate": 9.089562510123774e-06, + "loss": 1.1951, + "step": 19353 + }, + { + "epoch": 1.63, + "learning_rate": 9.088657315898373e-06, + "loss": 1.1997, + "step": 19354 + }, + { + "epoch": 1.63, + "learning_rate": 9.087752129202775e-06, + "loss": 1.24, + "step": 19355 + }, + { + "epoch": 1.63, + "learning_rate": 9.086846950044464e-06, + "loss": 1.1557, + "step": 19356 + }, + { + "epoch": 1.63, + "learning_rate": 9.085941778430913e-06, + "loss": 1.2067, + "step": 19357 + }, + { + "epoch": 1.63, + "learning_rate": 9.085036614369607e-06, + "loss": 1.1217, + "step": 19358 + }, + { + "epoch": 1.63, + "learning_rate": 9.084131457868017e-06, + "loss": 1.1015, + "step": 19359 + }, + { + "epoch": 1.63, + "learning_rate": 9.08322630893363e-06, + "loss": 1.2382, + "step": 19360 + }, + { + "epoch": 1.63, + "learning_rate": 9.082321167573917e-06, + "loss": 1.2077, + "step": 19361 + }, + { + "epoch": 1.63, + "learning_rate": 9.081416033796363e-06, + "loss": 1.1826, + "step": 19362 + }, + { + "epoch": 1.63, + "learning_rate": 9.080510907608442e-06, + "loss": 1.1589, + "step": 19363 + }, + { + "epoch": 1.63, + "learning_rate": 9.079605789017634e-06, + "loss": 1.2017, + "step": 19364 + }, + { + "epoch": 1.63, + "learning_rate": 9.07870067803142e-06, + "loss": 1.2469, + "step": 19365 + }, + { + "epoch": 1.63, + "learning_rate": 9.077795574657271e-06, + "loss": 1.1922, + "step": 19366 + }, + { + "epoch": 1.63, + "learning_rate": 9.076890478902676e-06, + "loss": 1.0895, + "step": 19367 + }, + { + "epoch": 1.63, + "learning_rate": 9.075985390775104e-06, + "loss": 1.1665, + "step": 19368 + }, + { + "epoch": 1.63, + "learning_rate": 9.075080310282039e-06, + "loss": 1.1565, + "step": 19369 + }, + { + "epoch": 1.63, + "learning_rate": 9.074175237430952e-06, + "loss": 1.1743, + "step": 19370 + }, + { + "epoch": 1.63, + "learning_rate": 9.07327017222933e-06, + "loss": 1.1708, + "step": 19371 + }, + { + "epoch": 1.63, + "learning_rate": 9.072365114684643e-06, + "loss": 1.1813, + "step": 19372 + }, + { + "epoch": 1.63, + "learning_rate": 9.071460064804377e-06, + "loss": 1.1664, + "step": 19373 + }, + { + "epoch": 1.63, + "learning_rate": 9.070555022595998e-06, + "loss": 1.1232, + "step": 19374 + }, + { + "epoch": 1.63, + "learning_rate": 9.069649988066994e-06, + "loss": 1.2645, + "step": 19375 + }, + { + "epoch": 1.63, + "learning_rate": 9.068744961224842e-06, + "loss": 1.2271, + "step": 19376 + }, + { + "epoch": 1.63, + "learning_rate": 9.067839942077013e-06, + "loss": 1.1629, + "step": 19377 + }, + { + "epoch": 1.63, + "learning_rate": 9.066934930630994e-06, + "loss": 1.1526, + "step": 19378 + }, + { + "epoch": 1.63, + "learning_rate": 9.066029926894253e-06, + "loss": 1.1347, + "step": 19379 + }, + { + "epoch": 1.63, + "learning_rate": 9.065124930874275e-06, + "loss": 1.1763, + "step": 19380 + }, + { + "epoch": 1.63, + "learning_rate": 9.064219942578529e-06, + "loss": 1.1298, + "step": 19381 + }, + { + "epoch": 1.63, + "learning_rate": 9.063314962014502e-06, + "loss": 1.2441, + "step": 19382 + }, + { + "epoch": 1.63, + "learning_rate": 9.062409989189665e-06, + "loss": 1.2697, + "step": 19383 + }, + { + "epoch": 1.63, + "learning_rate": 9.0615050241115e-06, + "loss": 1.1481, + "step": 19384 + }, + { + "epoch": 1.63, + "learning_rate": 9.060600066787473e-06, + "loss": 1.1997, + "step": 19385 + }, + { + "epoch": 1.63, + "learning_rate": 9.059695117225078e-06, + "loss": 1.087, + "step": 19386 + }, + { + "epoch": 1.63, + "learning_rate": 9.058790175431775e-06, + "loss": 1.2296, + "step": 19387 + }, + { + "epoch": 1.63, + "learning_rate": 9.05788524141505e-06, + "loss": 1.1718, + "step": 19388 + }, + { + "epoch": 1.63, + "learning_rate": 9.056980315182385e-06, + "loss": 1.259, + "step": 19389 + }, + { + "epoch": 1.63, + "learning_rate": 9.056075396741243e-06, + "loss": 1.2499, + "step": 19390 + }, + { + "epoch": 1.63, + "learning_rate": 9.055170486099112e-06, + "loss": 1.1299, + "step": 19391 + }, + { + "epoch": 1.63, + "learning_rate": 9.054265583263464e-06, + "loss": 1.1912, + "step": 19392 + }, + { + "epoch": 1.63, + "learning_rate": 9.053360688241779e-06, + "loss": 1.2606, + "step": 19393 + }, + { + "epoch": 1.63, + "learning_rate": 9.052455801041527e-06, + "loss": 1.159, + "step": 19394 + }, + { + "epoch": 1.63, + "learning_rate": 9.05155092167019e-06, + "loss": 1.1259, + "step": 19395 + }, + { + "epoch": 1.63, + "learning_rate": 9.050646050135243e-06, + "loss": 1.0765, + "step": 19396 + }, + { + "epoch": 1.63, + "learning_rate": 9.049741186444164e-06, + "loss": 1.2201, + "step": 19397 + }, + { + "epoch": 1.63, + "learning_rate": 9.048836330604422e-06, + "loss": 1.1183, + "step": 19398 + }, + { + "epoch": 1.63, + "learning_rate": 9.047931482623505e-06, + "loss": 1.2202, + "step": 19399 + }, + { + "epoch": 1.63, + "learning_rate": 9.047026642508878e-06, + "loss": 1.1718, + "step": 19400 + }, + { + "epoch": 1.63, + "learning_rate": 9.046121810268021e-06, + "loss": 1.1862, + "step": 19401 + }, + { + "epoch": 1.63, + "learning_rate": 9.045216985908415e-06, + "loss": 1.2075, + "step": 19402 + }, + { + "epoch": 1.63, + "learning_rate": 9.044312169437532e-06, + "loss": 1.1665, + "step": 19403 + }, + { + "epoch": 1.63, + "learning_rate": 9.043407360862847e-06, + "loss": 1.2109, + "step": 19404 + }, + { + "epoch": 1.63, + "learning_rate": 9.042502560191832e-06, + "loss": 1.1593, + "step": 19405 + }, + { + "epoch": 1.63, + "learning_rate": 9.041597767431972e-06, + "loss": 1.1143, + "step": 19406 + }, + { + "epoch": 1.63, + "learning_rate": 9.04069298259074e-06, + "loss": 1.2138, + "step": 19407 + }, + { + "epoch": 1.63, + "learning_rate": 9.039788205675607e-06, + "loss": 1.166, + "step": 19408 + }, + { + "epoch": 1.63, + "learning_rate": 9.038883436694047e-06, + "loss": 1.1596, + "step": 19409 + }, + { + "epoch": 1.63, + "learning_rate": 9.037978675653546e-06, + "loss": 1.2234, + "step": 19410 + }, + { + "epoch": 1.63, + "learning_rate": 9.03707392256157e-06, + "loss": 1.1431, + "step": 19411 + }, + { + "epoch": 1.63, + "learning_rate": 9.0361691774256e-06, + "loss": 1.2215, + "step": 19412 + }, + { + "epoch": 1.63, + "learning_rate": 9.035264440253104e-06, + "loss": 1.2682, + "step": 19413 + }, + { + "epoch": 1.63, + "learning_rate": 9.034359711051564e-06, + "loss": 1.1876, + "step": 19414 + }, + { + "epoch": 1.63, + "learning_rate": 9.033454989828453e-06, + "loss": 1.2103, + "step": 19415 + }, + { + "epoch": 1.63, + "learning_rate": 9.032550276591245e-06, + "loss": 1.2696, + "step": 19416 + }, + { + "epoch": 1.63, + "learning_rate": 9.031645571347421e-06, + "loss": 1.2024, + "step": 19417 + }, + { + "epoch": 1.63, + "learning_rate": 9.030740874104448e-06, + "loss": 1.1844, + "step": 19418 + }, + { + "epoch": 1.63, + "learning_rate": 9.029836184869806e-06, + "loss": 1.1722, + "step": 19419 + }, + { + "epoch": 1.64, + "learning_rate": 9.028931503650963e-06, + "loss": 1.1596, + "step": 19420 + }, + { + "epoch": 1.64, + "learning_rate": 9.028026830455405e-06, + "loss": 1.2507, + "step": 19421 + }, + { + "epoch": 1.64, + "learning_rate": 9.027122165290596e-06, + "loss": 1.2247, + "step": 19422 + }, + { + "epoch": 1.64, + "learning_rate": 9.026217508164018e-06, + "loss": 1.2036, + "step": 19423 + }, + { + "epoch": 1.64, + "learning_rate": 9.025312859083139e-06, + "loss": 1.1709, + "step": 19424 + }, + { + "epoch": 1.64, + "learning_rate": 9.024408218055439e-06, + "loss": 1.1719, + "step": 19425 + }, + { + "epoch": 1.64, + "learning_rate": 9.023503585088392e-06, + "loss": 1.1717, + "step": 19426 + }, + { + "epoch": 1.64, + "learning_rate": 9.022598960189466e-06, + "loss": 1.2374, + "step": 19427 + }, + { + "epoch": 1.64, + "learning_rate": 9.021694343366146e-06, + "loss": 1.1867, + "step": 19428 + }, + { + "epoch": 1.64, + "learning_rate": 9.020789734625898e-06, + "loss": 1.1956, + "step": 19429 + }, + { + "epoch": 1.64, + "learning_rate": 9.019885133976201e-06, + "loss": 1.1695, + "step": 19430 + }, + { + "epoch": 1.64, + "learning_rate": 9.018980541424522e-06, + "loss": 1.2576, + "step": 19431 + }, + { + "epoch": 1.64, + "learning_rate": 9.018075956978343e-06, + "loss": 1.2375, + "step": 19432 + }, + { + "epoch": 1.64, + "learning_rate": 9.017171380645134e-06, + "loss": 1.1611, + "step": 19433 + }, + { + "epoch": 1.64, + "learning_rate": 9.01626681243237e-06, + "loss": 1.1741, + "step": 19434 + }, + { + "epoch": 1.64, + "learning_rate": 9.015362252347523e-06, + "loss": 1.1854, + "step": 19435 + }, + { + "epoch": 1.64, + "learning_rate": 9.014457700398073e-06, + "loss": 1.3065, + "step": 19436 + }, + { + "epoch": 1.64, + "learning_rate": 9.013553156591482e-06, + "loss": 1.163, + "step": 19437 + }, + { + "epoch": 1.64, + "learning_rate": 9.01264862093523e-06, + "loss": 1.1604, + "step": 19438 + }, + { + "epoch": 1.64, + "learning_rate": 9.0117440934368e-06, + "loss": 1.2094, + "step": 19439 + }, + { + "epoch": 1.64, + "learning_rate": 9.010839574103648e-06, + "loss": 1.2363, + "step": 19440 + }, + { + "epoch": 1.64, + "learning_rate": 9.00993506294326e-06, + "loss": 1.1681, + "step": 19441 + }, + { + "epoch": 1.64, + "learning_rate": 9.009030559963105e-06, + "loss": 1.0654, + "step": 19442 + }, + { + "epoch": 1.64, + "learning_rate": 9.008126065170658e-06, + "loss": 1.0829, + "step": 19443 + }, + { + "epoch": 1.64, + "learning_rate": 9.007221578573386e-06, + "loss": 1.1849, + "step": 19444 + }, + { + "epoch": 1.64, + "learning_rate": 9.006317100178772e-06, + "loss": 1.2871, + "step": 19445 + }, + { + "epoch": 1.64, + "learning_rate": 9.005412629994283e-06, + "loss": 1.1222, + "step": 19446 + }, + { + "epoch": 1.64, + "learning_rate": 9.004508168027393e-06, + "loss": 1.1936, + "step": 19447 + }, + { + "epoch": 1.64, + "learning_rate": 9.003603714285572e-06, + "loss": 1.1176, + "step": 19448 + }, + { + "epoch": 1.64, + "learning_rate": 9.0026992687763e-06, + "loss": 1.1816, + "step": 19449 + }, + { + "epoch": 1.64, + "learning_rate": 9.001794831507044e-06, + "loss": 1.2064, + "step": 19450 + }, + { + "epoch": 1.64, + "learning_rate": 9.000890402485277e-06, + "loss": 1.2855, + "step": 19451 + }, + { + "epoch": 1.64, + "learning_rate": 8.999985981718478e-06, + "loss": 1.1754, + "step": 19452 + }, + { + "epoch": 1.64, + "learning_rate": 8.99908156921411e-06, + "loss": 1.35, + "step": 19453 + }, + { + "epoch": 1.64, + "learning_rate": 8.998177164979655e-06, + "loss": 1.1695, + "step": 19454 + }, + { + "epoch": 1.64, + "learning_rate": 8.997272769022576e-06, + "loss": 1.1943, + "step": 19455 + }, + { + "epoch": 1.64, + "learning_rate": 8.996368381350355e-06, + "loss": 1.1621, + "step": 19456 + }, + { + "epoch": 1.64, + "learning_rate": 8.995464001970457e-06, + "loss": 1.1408, + "step": 19457 + }, + { + "epoch": 1.64, + "learning_rate": 8.994559630890358e-06, + "loss": 1.1258, + "step": 19458 + }, + { + "epoch": 1.64, + "learning_rate": 8.993655268117526e-06, + "loss": 1.1874, + "step": 19459 + }, + { + "epoch": 1.64, + "learning_rate": 8.99275091365944e-06, + "loss": 1.1801, + "step": 19460 + }, + { + "epoch": 1.64, + "learning_rate": 8.991846567523567e-06, + "loss": 1.2861, + "step": 19461 + }, + { + "epoch": 1.64, + "learning_rate": 8.990942229717382e-06, + "loss": 1.141, + "step": 19462 + }, + { + "epoch": 1.64, + "learning_rate": 8.990037900248351e-06, + "loss": 1.1906, + "step": 19463 + }, + { + "epoch": 1.64, + "learning_rate": 8.989133579123951e-06, + "loss": 1.1008, + "step": 19464 + }, + { + "epoch": 1.64, + "learning_rate": 8.988229266351657e-06, + "loss": 1.2671, + "step": 19465 + }, + { + "epoch": 1.64, + "learning_rate": 8.987324961938932e-06, + "loss": 1.1013, + "step": 19466 + }, + { + "epoch": 1.64, + "learning_rate": 8.986420665893256e-06, + "loss": 1.1949, + "step": 19467 + }, + { + "epoch": 1.64, + "learning_rate": 8.985516378222094e-06, + "loss": 1.1732, + "step": 19468 + }, + { + "epoch": 1.64, + "learning_rate": 8.984612098932925e-06, + "loss": 1.2646, + "step": 19469 + }, + { + "epoch": 1.64, + "learning_rate": 8.98370782803321e-06, + "loss": 1.2216, + "step": 19470 + }, + { + "epoch": 1.64, + "learning_rate": 8.982803565530429e-06, + "loss": 1.1865, + "step": 19471 + }, + { + "epoch": 1.64, + "learning_rate": 8.98189931143205e-06, + "loss": 1.1167, + "step": 19472 + }, + { + "epoch": 1.64, + "learning_rate": 8.980995065745546e-06, + "loss": 1.2034, + "step": 19473 + }, + { + "epoch": 1.64, + "learning_rate": 8.980090828478382e-06, + "loss": 1.12, + "step": 19474 + }, + { + "epoch": 1.64, + "learning_rate": 8.979186599638037e-06, + "loss": 1.1477, + "step": 19475 + }, + { + "epoch": 1.64, + "learning_rate": 8.978282379231983e-06, + "loss": 1.1742, + "step": 19476 + }, + { + "epoch": 1.64, + "learning_rate": 8.97737816726768e-06, + "loss": 1.1934, + "step": 19477 + }, + { + "epoch": 1.64, + "learning_rate": 8.976473963752613e-06, + "loss": 1.1438, + "step": 19478 + }, + { + "epoch": 1.64, + "learning_rate": 8.97556976869424e-06, + "loss": 1.137, + "step": 19479 + }, + { + "epoch": 1.64, + "learning_rate": 8.974665582100042e-06, + "loss": 1.2047, + "step": 19480 + }, + { + "epoch": 1.64, + "learning_rate": 8.97376140397748e-06, + "loss": 1.1422, + "step": 19481 + }, + { + "epoch": 1.64, + "learning_rate": 8.972857234334034e-06, + "loss": 1.1584, + "step": 19482 + }, + { + "epoch": 1.64, + "learning_rate": 8.971953073177168e-06, + "loss": 1.2428, + "step": 19483 + }, + { + "epoch": 1.64, + "learning_rate": 8.971048920514357e-06, + "loss": 1.1808, + "step": 19484 + }, + { + "epoch": 1.64, + "learning_rate": 8.970144776353065e-06, + "loss": 1.1649, + "step": 19485 + }, + { + "epoch": 1.64, + "learning_rate": 8.969240640700773e-06, + "loss": 1.1611, + "step": 19486 + }, + { + "epoch": 1.64, + "learning_rate": 8.968336513564938e-06, + "loss": 1.2496, + "step": 19487 + }, + { + "epoch": 1.64, + "learning_rate": 8.967432394953038e-06, + "loss": 1.0988, + "step": 19488 + }, + { + "epoch": 1.64, + "learning_rate": 8.966528284872547e-06, + "loss": 1.3182, + "step": 19489 + }, + { + "epoch": 1.64, + "learning_rate": 8.965624183330923e-06, + "loss": 1.1654, + "step": 19490 + }, + { + "epoch": 1.64, + "learning_rate": 8.96472009033565e-06, + "loss": 1.2458, + "step": 19491 + }, + { + "epoch": 1.64, + "learning_rate": 8.963816005894186e-06, + "loss": 1.1828, + "step": 19492 + }, + { + "epoch": 1.64, + "learning_rate": 8.962911930014009e-06, + "loss": 1.244, + "step": 19493 + }, + { + "epoch": 1.64, + "learning_rate": 8.962007862702581e-06, + "loss": 1.218, + "step": 19494 + }, + { + "epoch": 1.64, + "learning_rate": 8.961103803967381e-06, + "loss": 1.225, + "step": 19495 + }, + { + "epoch": 1.64, + "learning_rate": 8.960199753815872e-06, + "loss": 1.1723, + "step": 19496 + }, + { + "epoch": 1.64, + "learning_rate": 8.959295712255528e-06, + "loss": 1.1374, + "step": 19497 + }, + { + "epoch": 1.64, + "learning_rate": 8.95839167929381e-06, + "loss": 1.2077, + "step": 19498 + }, + { + "epoch": 1.64, + "learning_rate": 8.957487654938199e-06, + "loss": 1.1625, + "step": 19499 + }, + { + "epoch": 1.64, + "learning_rate": 8.956583639196156e-06, + "loss": 1.2548, + "step": 19500 + }, + { + "epoch": 1.64, + "learning_rate": 8.95567963207515e-06, + "loss": 1.1926, + "step": 19501 + }, + { + "epoch": 1.64, + "learning_rate": 8.954775633582658e-06, + "loss": 1.2565, + "step": 19502 + }, + { + "epoch": 1.64, + "learning_rate": 8.953871643726142e-06, + "loss": 1.146, + "step": 19503 + }, + { + "epoch": 1.64, + "learning_rate": 8.952967662513076e-06, + "loss": 1.1592, + "step": 19504 + }, + { + "epoch": 1.64, + "learning_rate": 8.952063689950922e-06, + "loss": 1.2147, + "step": 19505 + }, + { + "epoch": 1.64, + "learning_rate": 8.951159726047157e-06, + "loss": 1.2591, + "step": 19506 + }, + { + "epoch": 1.64, + "learning_rate": 8.950255770809243e-06, + "loss": 1.2272, + "step": 19507 + }, + { + "epoch": 1.64, + "learning_rate": 8.949351824244656e-06, + "loss": 1.2517, + "step": 19508 + }, + { + "epoch": 1.64, + "learning_rate": 8.948447886360856e-06, + "loss": 1.1522, + "step": 19509 + }, + { + "epoch": 1.64, + "learning_rate": 8.947543957165318e-06, + "loss": 1.1811, + "step": 19510 + }, + { + "epoch": 1.64, + "learning_rate": 8.946640036665508e-06, + "loss": 1.2311, + "step": 19511 + }, + { + "epoch": 1.64, + "learning_rate": 8.945736124868897e-06, + "loss": 1.2033, + "step": 19512 + }, + { + "epoch": 1.64, + "learning_rate": 8.944832221782948e-06, + "loss": 1.0558, + "step": 19513 + }, + { + "epoch": 1.64, + "learning_rate": 8.943928327415133e-06, + "loss": 1.2342, + "step": 19514 + }, + { + "epoch": 1.64, + "learning_rate": 8.943024441772925e-06, + "loss": 1.1957, + "step": 19515 + }, + { + "epoch": 1.64, + "learning_rate": 8.94212056486378e-06, + "loss": 1.2381, + "step": 19516 + }, + { + "epoch": 1.64, + "learning_rate": 8.94121669669518e-06, + "loss": 1.2001, + "step": 19517 + }, + { + "epoch": 1.64, + "learning_rate": 8.940312837274585e-06, + "loss": 1.1845, + "step": 19518 + }, + { + "epoch": 1.64, + "learning_rate": 8.939408986609465e-06, + "loss": 1.2022, + "step": 19519 + }, + { + "epoch": 1.64, + "learning_rate": 8.938505144707283e-06, + "loss": 1.16, + "step": 19520 + }, + { + "epoch": 1.64, + "learning_rate": 8.937601311575517e-06, + "loss": 1.1671, + "step": 19521 + }, + { + "epoch": 1.64, + "learning_rate": 8.936697487221626e-06, + "loss": 1.2472, + "step": 19522 + }, + { + "epoch": 1.64, + "learning_rate": 8.935793671653084e-06, + "loss": 1.1127, + "step": 19523 + }, + { + "epoch": 1.64, + "learning_rate": 8.93488986487735e-06, + "loss": 1.1833, + "step": 19524 + }, + { + "epoch": 1.64, + "learning_rate": 8.933986066901897e-06, + "loss": 1.1937, + "step": 19525 + }, + { + "epoch": 1.64, + "learning_rate": 8.933082277734196e-06, + "loss": 1.155, + "step": 19526 + }, + { + "epoch": 1.64, + "learning_rate": 8.932178497381708e-06, + "loss": 1.2179, + "step": 19527 + }, + { + "epoch": 1.64, + "learning_rate": 8.931274725851905e-06, + "loss": 1.1721, + "step": 19528 + }, + { + "epoch": 1.64, + "learning_rate": 8.93037096315225e-06, + "loss": 1.1192, + "step": 19529 + }, + { + "epoch": 1.64, + "learning_rate": 8.929467209290216e-06, + "loss": 1.2222, + "step": 19530 + }, + { + "epoch": 1.64, + "learning_rate": 8.928563464273263e-06, + "loss": 1.2496, + "step": 19531 + }, + { + "epoch": 1.64, + "learning_rate": 8.927659728108865e-06, + "loss": 1.244, + "step": 19532 + }, + { + "epoch": 1.64, + "learning_rate": 8.926756000804482e-06, + "loss": 1.133, + "step": 19533 + }, + { + "epoch": 1.64, + "learning_rate": 8.925852282367587e-06, + "loss": 1.2274, + "step": 19534 + }, + { + "epoch": 1.64, + "learning_rate": 8.92494857280564e-06, + "loss": 1.2795, + "step": 19535 + }, + { + "epoch": 1.64, + "learning_rate": 8.924044872126117e-06, + "loss": 1.1482, + "step": 19536 + }, + { + "epoch": 1.64, + "learning_rate": 8.923141180336477e-06, + "loss": 1.2462, + "step": 19537 + }, + { + "epoch": 1.64, + "learning_rate": 8.922237497444186e-06, + "loss": 1.2966, + "step": 19538 + }, + { + "epoch": 1.65, + "learning_rate": 8.921333823456721e-06, + "loss": 1.2084, + "step": 19539 + }, + { + "epoch": 1.65, + "learning_rate": 8.920430158381536e-06, + "loss": 1.1949, + "step": 19540 + }, + { + "epoch": 1.65, + "learning_rate": 8.919526502226107e-06, + "loss": 1.2398, + "step": 19541 + }, + { + "epoch": 1.65, + "learning_rate": 8.918622854997891e-06, + "loss": 1.1416, + "step": 19542 + }, + { + "epoch": 1.65, + "learning_rate": 8.917719216704365e-06, + "loss": 1.1302, + "step": 19543 + }, + { + "epoch": 1.65, + "learning_rate": 8.916815587352983e-06, + "loss": 1.1627, + "step": 19544 + }, + { + "epoch": 1.65, + "learning_rate": 8.91591196695122e-06, + "loss": 1.1752, + "step": 19545 + }, + { + "epoch": 1.65, + "learning_rate": 8.915008355506539e-06, + "loss": 1.1432, + "step": 19546 + }, + { + "epoch": 1.65, + "learning_rate": 8.914104753026408e-06, + "loss": 1.1559, + "step": 19547 + }, + { + "epoch": 1.65, + "learning_rate": 8.913201159518287e-06, + "loss": 1.1672, + "step": 19548 + }, + { + "epoch": 1.65, + "learning_rate": 8.912297574989649e-06, + "loss": 1.2258, + "step": 19549 + }, + { + "epoch": 1.65, + "learning_rate": 8.911393999447954e-06, + "loss": 1.2205, + "step": 19550 + }, + { + "epoch": 1.65, + "learning_rate": 8.91049043290067e-06, + "loss": 1.2943, + "step": 19551 + }, + { + "epoch": 1.65, + "learning_rate": 8.909586875355266e-06, + "loss": 1.1238, + "step": 19552 + }, + { + "epoch": 1.65, + "learning_rate": 8.908683326819203e-06, + "loss": 1.1634, + "step": 19553 + }, + { + "epoch": 1.65, + "learning_rate": 8.907779787299949e-06, + "loss": 1.1536, + "step": 19554 + }, + { + "epoch": 1.65, + "learning_rate": 8.906876256804964e-06, + "loss": 1.1774, + "step": 19555 + }, + { + "epoch": 1.65, + "learning_rate": 8.90597273534172e-06, + "loss": 1.0685, + "step": 19556 + }, + { + "epoch": 1.65, + "learning_rate": 8.90506922291768e-06, + "loss": 1.2342, + "step": 19557 + }, + { + "epoch": 1.65, + "learning_rate": 8.904165719540307e-06, + "loss": 1.066, + "step": 19558 + }, + { + "epoch": 1.65, + "learning_rate": 8.903262225217066e-06, + "loss": 1.1648, + "step": 19559 + }, + { + "epoch": 1.65, + "learning_rate": 8.902358739955428e-06, + "loss": 1.1478, + "step": 19560 + }, + { + "epoch": 1.65, + "learning_rate": 8.901455263762849e-06, + "loss": 1.2109, + "step": 19561 + }, + { + "epoch": 1.65, + "learning_rate": 8.900551796646803e-06, + "loss": 1.2151, + "step": 19562 + }, + { + "epoch": 1.65, + "learning_rate": 8.899648338614744e-06, + "loss": 1.145, + "step": 19563 + }, + { + "epoch": 1.65, + "learning_rate": 8.898744889674145e-06, + "loss": 1.1683, + "step": 19564 + }, + { + "epoch": 1.65, + "learning_rate": 8.89784144983247e-06, + "loss": 1.1753, + "step": 19565 + }, + { + "epoch": 1.65, + "learning_rate": 8.896938019097178e-06, + "loss": 1.2373, + "step": 19566 + }, + { + "epoch": 1.65, + "learning_rate": 8.89603459747574e-06, + "loss": 1.206, + "step": 19567 + }, + { + "epoch": 1.65, + "learning_rate": 8.895131184975618e-06, + "loss": 1.1841, + "step": 19568 + }, + { + "epoch": 1.65, + "learning_rate": 8.894227781604276e-06, + "loss": 1.1676, + "step": 19569 + }, + { + "epoch": 1.65, + "learning_rate": 8.893324387369174e-06, + "loss": 1.1844, + "step": 19570 + }, + { + "epoch": 1.65, + "learning_rate": 8.892421002277784e-06, + "loss": 1.0806, + "step": 19571 + }, + { + "epoch": 1.65, + "learning_rate": 8.891517626337565e-06, + "loss": 1.1709, + "step": 19572 + }, + { + "epoch": 1.65, + "learning_rate": 8.890614259555983e-06, + "loss": 1.1879, + "step": 19573 + }, + { + "epoch": 1.65, + "learning_rate": 8.889710901940497e-06, + "loss": 1.2522, + "step": 19574 + }, + { + "epoch": 1.65, + "learning_rate": 8.888807553498578e-06, + "loss": 1.1718, + "step": 19575 + }, + { + "epoch": 1.65, + "learning_rate": 8.887904214237688e-06, + "loss": 1.2281, + "step": 19576 + }, + { + "epoch": 1.65, + "learning_rate": 8.887000884165286e-06, + "loss": 1.0621, + "step": 19577 + }, + { + "epoch": 1.65, + "learning_rate": 8.886097563288841e-06, + "loss": 1.1691, + "step": 19578 + }, + { + "epoch": 1.65, + "learning_rate": 8.885194251615814e-06, + "loss": 1.2605, + "step": 19579 + }, + { + "epoch": 1.65, + "learning_rate": 8.88429094915367e-06, + "loss": 1.2013, + "step": 19580 + }, + { + "epoch": 1.65, + "learning_rate": 8.88338765590987e-06, + "loss": 1.1621, + "step": 19581 + }, + { + "epoch": 1.65, + "learning_rate": 8.88248437189188e-06, + "loss": 1.123, + "step": 19582 + }, + { + "epoch": 1.65, + "learning_rate": 8.88158109710716e-06, + "loss": 1.0594, + "step": 19583 + }, + { + "epoch": 1.65, + "learning_rate": 8.880677831563176e-06, + "loss": 1.0449, + "step": 19584 + }, + { + "epoch": 1.65, + "learning_rate": 8.879774575267388e-06, + "loss": 1.2364, + "step": 19585 + }, + { + "epoch": 1.65, + "learning_rate": 8.878871328227264e-06, + "loss": 1.0734, + "step": 19586 + }, + { + "epoch": 1.65, + "learning_rate": 8.877968090450262e-06, + "loss": 1.1781, + "step": 19587 + }, + { + "epoch": 1.65, + "learning_rate": 8.877064861943844e-06, + "loss": 1.2262, + "step": 19588 + }, + { + "epoch": 1.65, + "learning_rate": 8.876161642715481e-06, + "loss": 1.2435, + "step": 19589 + }, + { + "epoch": 1.65, + "learning_rate": 8.875258432772626e-06, + "loss": 1.3214, + "step": 19590 + }, + { + "epoch": 1.65, + "learning_rate": 8.87435523212275e-06, + "loss": 1.1648, + "step": 19591 + }, + { + "epoch": 1.65, + "learning_rate": 8.873452040773306e-06, + "loss": 1.1445, + "step": 19592 + }, + { + "epoch": 1.65, + "learning_rate": 8.872548858731768e-06, + "loss": 1.1528, + "step": 19593 + }, + { + "epoch": 1.65, + "learning_rate": 8.871645686005587e-06, + "loss": 1.2754, + "step": 19594 + }, + { + "epoch": 1.65, + "learning_rate": 8.870742522602233e-06, + "loss": 1.1231, + "step": 19595 + }, + { + "epoch": 1.65, + "learning_rate": 8.869839368529166e-06, + "loss": 1.284, + "step": 19596 + }, + { + "epoch": 1.65, + "learning_rate": 8.868936223793847e-06, + "loss": 1.1482, + "step": 19597 + }, + { + "epoch": 1.65, + "learning_rate": 8.868033088403737e-06, + "loss": 1.1924, + "step": 19598 + }, + { + "epoch": 1.65, + "learning_rate": 8.867129962366303e-06, + "loss": 1.1883, + "step": 19599 + }, + { + "epoch": 1.65, + "learning_rate": 8.866226845689002e-06, + "loss": 1.2121, + "step": 19600 + }, + { + "epoch": 1.65, + "learning_rate": 8.865323738379295e-06, + "loss": 1.1582, + "step": 19601 + }, + { + "epoch": 1.65, + "learning_rate": 8.864420640444651e-06, + "loss": 1.2326, + "step": 19602 + }, + { + "epoch": 1.65, + "learning_rate": 8.863517551892527e-06, + "loss": 1.2271, + "step": 19603 + }, + { + "epoch": 1.65, + "learning_rate": 8.862614472730384e-06, + "loss": 1.1425, + "step": 19604 + }, + { + "epoch": 1.65, + "learning_rate": 8.861711402965681e-06, + "loss": 1.1693, + "step": 19605 + }, + { + "epoch": 1.65, + "learning_rate": 8.860808342605886e-06, + "loss": 1.2547, + "step": 19606 + }, + { + "epoch": 1.65, + "learning_rate": 8.859905291658457e-06, + "loss": 1.1823, + "step": 19607 + }, + { + "epoch": 1.65, + "learning_rate": 8.859002250130855e-06, + "loss": 1.1588, + "step": 19608 + }, + { + "epoch": 1.65, + "learning_rate": 8.85809921803054e-06, + "loss": 1.1586, + "step": 19609 + }, + { + "epoch": 1.65, + "learning_rate": 8.857196195364978e-06, + "loss": 1.1759, + "step": 19610 + }, + { + "epoch": 1.65, + "learning_rate": 8.856293182141622e-06, + "loss": 1.0769, + "step": 19611 + }, + { + "epoch": 1.65, + "learning_rate": 8.855390178367945e-06, + "loss": 1.2028, + "step": 19612 + }, + { + "epoch": 1.65, + "learning_rate": 8.854487184051392e-06, + "loss": 1.2004, + "step": 19613 + }, + { + "epoch": 1.65, + "learning_rate": 8.853584199199436e-06, + "loss": 1.1521, + "step": 19614 + }, + { + "epoch": 1.65, + "learning_rate": 8.852681223819538e-06, + "loss": 1.0891, + "step": 19615 + }, + { + "epoch": 1.65, + "learning_rate": 8.851778257919148e-06, + "loss": 1.1458, + "step": 19616 + }, + { + "epoch": 1.65, + "learning_rate": 8.85087530150574e-06, + "loss": 1.113, + "step": 19617 + }, + { + "epoch": 1.65, + "learning_rate": 8.849972354586764e-06, + "loss": 1.1744, + "step": 19618 + }, + { + "epoch": 1.65, + "learning_rate": 8.849069417169688e-06, + "loss": 1.2084, + "step": 19619 + }, + { + "epoch": 1.65, + "learning_rate": 8.848166489261963e-06, + "loss": 1.2188, + "step": 19620 + }, + { + "epoch": 1.65, + "learning_rate": 8.847263570871061e-06, + "loss": 1.1534, + "step": 19621 + }, + { + "epoch": 1.65, + "learning_rate": 8.846360662004434e-06, + "loss": 1.1795, + "step": 19622 + }, + { + "epoch": 1.65, + "learning_rate": 8.845457762669546e-06, + "loss": 1.25, + "step": 19623 + }, + { + "epoch": 1.65, + "learning_rate": 8.844554872873851e-06, + "loss": 1.1409, + "step": 19624 + }, + { + "epoch": 1.65, + "learning_rate": 8.843651992624815e-06, + "loss": 1.1676, + "step": 19625 + }, + { + "epoch": 1.65, + "learning_rate": 8.8427491219299e-06, + "loss": 1.2651, + "step": 19626 + }, + { + "epoch": 1.65, + "learning_rate": 8.841846260796557e-06, + "loss": 1.1653, + "step": 19627 + }, + { + "epoch": 1.65, + "learning_rate": 8.840943409232255e-06, + "loss": 1.1252, + "step": 19628 + }, + { + "epoch": 1.65, + "learning_rate": 8.840040567244449e-06, + "loss": 1.1788, + "step": 19629 + }, + { + "epoch": 1.65, + "learning_rate": 8.839137734840599e-06, + "loss": 1.2298, + "step": 19630 + }, + { + "epoch": 1.65, + "learning_rate": 8.838234912028161e-06, + "loss": 1.1244, + "step": 19631 + }, + { + "epoch": 1.65, + "learning_rate": 8.837332098814603e-06, + "loss": 1.1073, + "step": 19632 + }, + { + "epoch": 1.65, + "learning_rate": 8.836429295207376e-06, + "loss": 1.1931, + "step": 19633 + }, + { + "epoch": 1.65, + "learning_rate": 8.835526501213945e-06, + "loss": 1.168, + "step": 19634 + }, + { + "epoch": 1.65, + "learning_rate": 8.834623716841764e-06, + "loss": 1.1402, + "step": 19635 + }, + { + "epoch": 1.65, + "learning_rate": 8.833720942098297e-06, + "loss": 1.2788, + "step": 19636 + }, + { + "epoch": 1.65, + "learning_rate": 8.832818176991e-06, + "loss": 1.1944, + "step": 19637 + }, + { + "epoch": 1.65, + "learning_rate": 8.831915421527329e-06, + "loss": 1.3003, + "step": 19638 + }, + { + "epoch": 1.65, + "learning_rate": 8.831012675714754e-06, + "loss": 1.1848, + "step": 19639 + }, + { + "epoch": 1.65, + "learning_rate": 8.83010993956072e-06, + "loss": 1.2143, + "step": 19640 + }, + { + "epoch": 1.65, + "learning_rate": 8.829207213072697e-06, + "loss": 1.2171, + "step": 19641 + }, + { + "epoch": 1.65, + "learning_rate": 8.828304496258134e-06, + "loss": 1.1483, + "step": 19642 + }, + { + "epoch": 1.65, + "learning_rate": 8.827401789124502e-06, + "loss": 1.2024, + "step": 19643 + }, + { + "epoch": 1.65, + "learning_rate": 8.826499091679244e-06, + "loss": 1.1531, + "step": 19644 + }, + { + "epoch": 1.65, + "learning_rate": 8.82559640392983e-06, + "loss": 1.1598, + "step": 19645 + }, + { + "epoch": 1.65, + "learning_rate": 8.824693725883712e-06, + "loss": 1.174, + "step": 19646 + }, + { + "epoch": 1.65, + "learning_rate": 8.823791057548352e-06, + "loss": 1.1871, + "step": 19647 + }, + { + "epoch": 1.65, + "learning_rate": 8.822888398931205e-06, + "loss": 1.2095, + "step": 19648 + }, + { + "epoch": 1.65, + "learning_rate": 8.821985750039733e-06, + "loss": 1.223, + "step": 19649 + }, + { + "epoch": 1.65, + "learning_rate": 8.821083110881391e-06, + "loss": 1.1948, + "step": 19650 + }, + { + "epoch": 1.65, + "learning_rate": 8.820180481463635e-06, + "loss": 1.2792, + "step": 19651 + }, + { + "epoch": 1.65, + "learning_rate": 8.819277861793929e-06, + "loss": 1.194, + "step": 19652 + }, + { + "epoch": 1.65, + "learning_rate": 8.818375251879726e-06, + "loss": 1.2426, + "step": 19653 + }, + { + "epoch": 1.65, + "learning_rate": 8.817472651728486e-06, + "loss": 1.1607, + "step": 19654 + }, + { + "epoch": 1.65, + "learning_rate": 8.816570061347663e-06, + "loss": 1.1643, + "step": 19655 + }, + { + "epoch": 1.65, + "learning_rate": 8.815667480744719e-06, + "loss": 1.1834, + "step": 19656 + }, + { + "epoch": 1.65, + "learning_rate": 8.814764909927108e-06, + "loss": 1.2297, + "step": 19657 + }, + { + "epoch": 1.66, + "learning_rate": 8.813862348902291e-06, + "loss": 1.1779, + "step": 19658 + }, + { + "epoch": 1.66, + "learning_rate": 8.812959797677718e-06, + "loss": 1.1978, + "step": 19659 + }, + { + "epoch": 1.66, + "learning_rate": 8.812057256260857e-06, + "loss": 1.2154, + "step": 19660 + }, + { + "epoch": 1.66, + "learning_rate": 8.811154724659155e-06, + "loss": 1.2248, + "step": 19661 + }, + { + "epoch": 1.66, + "learning_rate": 8.810252202880076e-06, + "loss": 1.0891, + "step": 19662 + }, + { + "epoch": 1.66, + "learning_rate": 8.80934969093107e-06, + "loss": 1.2243, + "step": 19663 + }, + { + "epoch": 1.66, + "learning_rate": 8.808447188819599e-06, + "loss": 1.1521, + "step": 19664 + }, + { + "epoch": 1.66, + "learning_rate": 8.807544696553121e-06, + "loss": 1.2484, + "step": 19665 + }, + { + "epoch": 1.66, + "learning_rate": 8.806642214139087e-06, + "loss": 1.1415, + "step": 19666 + }, + { + "epoch": 1.66, + "learning_rate": 8.805739741584963e-06, + "loss": 1.1756, + "step": 19667 + }, + { + "epoch": 1.66, + "learning_rate": 8.804837278898194e-06, + "loss": 1.2113, + "step": 19668 + }, + { + "epoch": 1.66, + "learning_rate": 8.803934826086246e-06, + "loss": 1.1574, + "step": 19669 + }, + { + "epoch": 1.66, + "learning_rate": 8.803032383156566e-06, + "loss": 1.2593, + "step": 19670 + }, + { + "epoch": 1.66, + "learning_rate": 8.80212995011662e-06, + "loss": 1.1075, + "step": 19671 + }, + { + "epoch": 1.66, + "learning_rate": 8.801227526973857e-06, + "loss": 1.182, + "step": 19672 + }, + { + "epoch": 1.66, + "learning_rate": 8.80032511373574e-06, + "loss": 1.2207, + "step": 19673 + }, + { + "epoch": 1.66, + "learning_rate": 8.799422710409714e-06, + "loss": 1.2881, + "step": 19674 + }, + { + "epoch": 1.66, + "learning_rate": 8.798520317003245e-06, + "loss": 1.1776, + "step": 19675 + }, + { + "epoch": 1.66, + "learning_rate": 8.797617933523788e-06, + "loss": 1.1161, + "step": 19676 + }, + { + "epoch": 1.66, + "learning_rate": 8.796715559978792e-06, + "loss": 1.1028, + "step": 19677 + }, + { + "epoch": 1.66, + "learning_rate": 8.795813196375722e-06, + "loss": 1.1076, + "step": 19678 + }, + { + "epoch": 1.66, + "learning_rate": 8.794910842722026e-06, + "loss": 1.1978, + "step": 19679 + }, + { + "epoch": 1.66, + "learning_rate": 8.794008499025166e-06, + "loss": 1.2748, + "step": 19680 + }, + { + "epoch": 1.66, + "learning_rate": 8.793106165292588e-06, + "loss": 1.1703, + "step": 19681 + }, + { + "epoch": 1.66, + "learning_rate": 8.792203841531756e-06, + "loss": 1.1632, + "step": 19682 + }, + { + "epoch": 1.66, + "learning_rate": 8.791301527750122e-06, + "loss": 1.2913, + "step": 19683 + }, + { + "epoch": 1.66, + "learning_rate": 8.790399223955144e-06, + "loss": 1.1792, + "step": 19684 + }, + { + "epoch": 1.66, + "learning_rate": 8.78949693015427e-06, + "loss": 1.0941, + "step": 19685 + }, + { + "epoch": 1.66, + "learning_rate": 8.788594646354965e-06, + "loss": 1.158, + "step": 19686 + }, + { + "epoch": 1.66, + "learning_rate": 8.787692372564674e-06, + "loss": 1.2713, + "step": 19687 + }, + { + "epoch": 1.66, + "learning_rate": 8.786790108790856e-06, + "loss": 1.1126, + "step": 19688 + }, + { + "epoch": 1.66, + "learning_rate": 8.785887855040972e-06, + "loss": 1.2157, + "step": 19689 + }, + { + "epoch": 1.66, + "learning_rate": 8.784985611322467e-06, + "loss": 1.1919, + "step": 19690 + }, + { + "epoch": 1.66, + "learning_rate": 8.784083377642804e-06, + "loss": 1.2031, + "step": 19691 + }, + { + "epoch": 1.66, + "learning_rate": 8.783181154009427e-06, + "loss": 1.2564, + "step": 19692 + }, + { + "epoch": 1.66, + "learning_rate": 8.7822789404298e-06, + "loss": 1.2676, + "step": 19693 + }, + { + "epoch": 1.66, + "learning_rate": 8.781376736911374e-06, + "loss": 1.1931, + "step": 19694 + }, + { + "epoch": 1.66, + "learning_rate": 8.780474543461606e-06, + "loss": 1.2702, + "step": 19695 + }, + { + "epoch": 1.66, + "learning_rate": 8.779572360087943e-06, + "loss": 1.1725, + "step": 19696 + }, + { + "epoch": 1.66, + "learning_rate": 8.77867018679785e-06, + "loss": 1.2587, + "step": 19697 + }, + { + "epoch": 1.66, + "learning_rate": 8.777768023598767e-06, + "loss": 1.186, + "step": 19698 + }, + { + "epoch": 1.66, + "learning_rate": 8.776865870498162e-06, + "loss": 1.2623, + "step": 19699 + }, + { + "epoch": 1.66, + "learning_rate": 8.775963727503479e-06, + "loss": 1.1642, + "step": 19700 + }, + { + "epoch": 1.66, + "learning_rate": 8.775061594622174e-06, + "loss": 1.1273, + "step": 19701 + }, + { + "epoch": 1.66, + "learning_rate": 8.774159471861707e-06, + "loss": 1.2239, + "step": 19702 + }, + { + "epoch": 1.66, + "learning_rate": 8.773257359229523e-06, + "loss": 1.2732, + "step": 19703 + }, + { + "epoch": 1.66, + "learning_rate": 8.772355256733082e-06, + "loss": 1.1628, + "step": 19704 + }, + { + "epoch": 1.66, + "learning_rate": 8.77145316437983e-06, + "loss": 1.094, + "step": 19705 + }, + { + "epoch": 1.66, + "learning_rate": 8.77055108217723e-06, + "loss": 1.2429, + "step": 19706 + }, + { + "epoch": 1.66, + "learning_rate": 8.76964901013273e-06, + "loss": 1.2547, + "step": 19707 + }, + { + "epoch": 1.66, + "learning_rate": 8.768746948253785e-06, + "loss": 1.0512, + "step": 19708 + }, + { + "epoch": 1.66, + "learning_rate": 8.767844896547842e-06, + "loss": 1.2048, + "step": 19709 + }, + { + "epoch": 1.66, + "learning_rate": 8.766942855022363e-06, + "loss": 1.199, + "step": 19710 + }, + { + "epoch": 1.66, + "learning_rate": 8.766040823684794e-06, + "loss": 1.2217, + "step": 19711 + }, + { + "epoch": 1.66, + "learning_rate": 8.765138802542588e-06, + "loss": 1.1958, + "step": 19712 + }, + { + "epoch": 1.66, + "learning_rate": 8.764236791603206e-06, + "loss": 1.147, + "step": 19713 + }, + { + "epoch": 1.66, + "learning_rate": 8.763334790874093e-06, + "loss": 1.1364, + "step": 19714 + }, + { + "epoch": 1.66, + "learning_rate": 8.762432800362706e-06, + "loss": 1.2022, + "step": 19715 + }, + { + "epoch": 1.66, + "learning_rate": 8.761530820076489e-06, + "loss": 1.1798, + "step": 19716 + }, + { + "epoch": 1.66, + "learning_rate": 8.760628850022908e-06, + "loss": 1.1502, + "step": 19717 + }, + { + "epoch": 1.66, + "learning_rate": 8.759726890209403e-06, + "loss": 1.1136, + "step": 19718 + }, + { + "epoch": 1.66, + "learning_rate": 8.758824940643436e-06, + "loss": 1.1192, + "step": 19719 + }, + { + "epoch": 1.66, + "learning_rate": 8.75792300133245e-06, + "loss": 1.1844, + "step": 19720 + }, + { + "epoch": 1.66, + "learning_rate": 8.757021072283905e-06, + "loss": 1.1127, + "step": 19721 + }, + { + "epoch": 1.66, + "learning_rate": 8.756119153505248e-06, + "loss": 1.2115, + "step": 19722 + }, + { + "epoch": 1.66, + "learning_rate": 8.755217245003935e-06, + "loss": 1.1733, + "step": 19723 + }, + { + "epoch": 1.66, + "learning_rate": 8.75431534678741e-06, + "loss": 1.2999, + "step": 19724 + }, + { + "epoch": 1.66, + "learning_rate": 8.753413458863132e-06, + "loss": 1.1698, + "step": 19725 + }, + { + "epoch": 1.66, + "learning_rate": 8.752511581238555e-06, + "loss": 1.1306, + "step": 19726 + }, + { + "epoch": 1.66, + "learning_rate": 8.751609713921122e-06, + "loss": 1.1474, + "step": 19727 + }, + { + "epoch": 1.66, + "learning_rate": 8.750707856918293e-06, + "loss": 1.2037, + "step": 19728 + }, + { + "epoch": 1.66, + "learning_rate": 8.749806010237513e-06, + "loss": 1.1874, + "step": 19729 + }, + { + "epoch": 1.66, + "learning_rate": 8.74890417388624e-06, + "loss": 1.2454, + "step": 19730 + }, + { + "epoch": 1.66, + "learning_rate": 8.748002347871914e-06, + "loss": 1.1924, + "step": 19731 + }, + { + "epoch": 1.66, + "learning_rate": 8.747100532201998e-06, + "loss": 1.2726, + "step": 19732 + }, + { + "epoch": 1.66, + "learning_rate": 8.746198726883938e-06, + "loss": 1.1935, + "step": 19733 + }, + { + "epoch": 1.66, + "learning_rate": 8.745296931925186e-06, + "loss": 1.124, + "step": 19734 + }, + { + "epoch": 1.66, + "learning_rate": 8.744395147333189e-06, + "loss": 1.1006, + "step": 19735 + }, + { + "epoch": 1.66, + "learning_rate": 8.743493373115405e-06, + "loss": 1.2417, + "step": 19736 + }, + { + "epoch": 1.66, + "learning_rate": 8.742591609279278e-06, + "loss": 1.2077, + "step": 19737 + }, + { + "epoch": 1.66, + "learning_rate": 8.74168985583226e-06, + "loss": 1.1678, + "step": 19738 + }, + { + "epoch": 1.66, + "learning_rate": 8.740788112781808e-06, + "loss": 1.1076, + "step": 19739 + }, + { + "epoch": 1.66, + "learning_rate": 8.739886380135365e-06, + "loss": 1.1161, + "step": 19740 + }, + { + "epoch": 1.66, + "learning_rate": 8.738984657900387e-06, + "loss": 1.1532, + "step": 19741 + }, + { + "epoch": 1.66, + "learning_rate": 8.738082946084315e-06, + "loss": 1.2707, + "step": 19742 + }, + { + "epoch": 1.66, + "learning_rate": 8.737181244694611e-06, + "loss": 1.1422, + "step": 19743 + }, + { + "epoch": 1.66, + "learning_rate": 8.736279553738718e-06, + "loss": 1.1384, + "step": 19744 + }, + { + "epoch": 1.66, + "learning_rate": 8.73537787322409e-06, + "loss": 1.1697, + "step": 19745 + }, + { + "epoch": 1.66, + "learning_rate": 8.73447620315817e-06, + "loss": 1.1925, + "step": 19746 + }, + { + "epoch": 1.66, + "learning_rate": 8.73357454354842e-06, + "loss": 1.0935, + "step": 19747 + }, + { + "epoch": 1.66, + "learning_rate": 8.732672894402274e-06, + "loss": 1.2059, + "step": 19748 + }, + { + "epoch": 1.66, + "learning_rate": 8.731771255727194e-06, + "loss": 1.027, + "step": 19749 + }, + { + "epoch": 1.66, + "learning_rate": 8.730869627530627e-06, + "loss": 1.3027, + "step": 19750 + }, + { + "epoch": 1.66, + "learning_rate": 8.729968009820017e-06, + "loss": 1.2402, + "step": 19751 + }, + { + "epoch": 1.66, + "learning_rate": 8.729066402602821e-06, + "loss": 1.1913, + "step": 19752 + }, + { + "epoch": 1.66, + "learning_rate": 8.728164805886486e-06, + "loss": 1.1012, + "step": 19753 + }, + { + "epoch": 1.66, + "learning_rate": 8.72726321967846e-06, + "loss": 1.2263, + "step": 19754 + }, + { + "epoch": 1.66, + "learning_rate": 8.72636164398619e-06, + "loss": 1.1605, + "step": 19755 + }, + { + "epoch": 1.66, + "learning_rate": 8.725460078817131e-06, + "loss": 1.1232, + "step": 19756 + }, + { + "epoch": 1.66, + "learning_rate": 8.724558524178728e-06, + "loss": 1.1129, + "step": 19757 + }, + { + "epoch": 1.66, + "learning_rate": 8.72365698007843e-06, + "loss": 1.1872, + "step": 19758 + }, + { + "epoch": 1.66, + "learning_rate": 8.722755446523683e-06, + "loss": 1.1898, + "step": 19759 + }, + { + "epoch": 1.66, + "learning_rate": 8.721853923521945e-06, + "loss": 1.2333, + "step": 19760 + }, + { + "epoch": 1.66, + "learning_rate": 8.720952411080656e-06, + "loss": 1.1127, + "step": 19761 + }, + { + "epoch": 1.66, + "learning_rate": 8.720050909207266e-06, + "loss": 1.1528, + "step": 19762 + }, + { + "epoch": 1.66, + "learning_rate": 8.719149417909229e-06, + "loss": 1.2396, + "step": 19763 + }, + { + "epoch": 1.66, + "learning_rate": 8.718247937193986e-06, + "loss": 1.1763, + "step": 19764 + }, + { + "epoch": 1.66, + "learning_rate": 8.717346467068993e-06, + "loss": 1.2607, + "step": 19765 + }, + { + "epoch": 1.66, + "learning_rate": 8.71644500754169e-06, + "loss": 1.2389, + "step": 19766 + }, + { + "epoch": 1.66, + "learning_rate": 8.71554355861953e-06, + "loss": 1.2046, + "step": 19767 + }, + { + "epoch": 1.66, + "learning_rate": 8.714642120309962e-06, + "loss": 1.1747, + "step": 19768 + }, + { + "epoch": 1.66, + "learning_rate": 8.713740692620432e-06, + "loss": 1.1963, + "step": 19769 + }, + { + "epoch": 1.66, + "learning_rate": 8.712839275558385e-06, + "loss": 1.1634, + "step": 19770 + }, + { + "epoch": 1.66, + "learning_rate": 8.711937869131274e-06, + "loss": 1.1572, + "step": 19771 + }, + { + "epoch": 1.66, + "learning_rate": 8.711036473346545e-06, + "loss": 1.154, + "step": 19772 + }, + { + "epoch": 1.66, + "learning_rate": 8.710135088211646e-06, + "loss": 1.2058, + "step": 19773 + }, + { + "epoch": 1.66, + "learning_rate": 8.70923371373402e-06, + "loss": 1.1876, + "step": 19774 + }, + { + "epoch": 1.66, + "learning_rate": 8.70833234992112e-06, + "loss": 1.2508, + "step": 19775 + }, + { + "epoch": 1.66, + "learning_rate": 8.707430996780393e-06, + "loss": 1.1331, + "step": 19776 + }, + { + "epoch": 1.67, + "learning_rate": 8.70652965431928e-06, + "loss": 1.226, + "step": 19777 + }, + { + "epoch": 1.67, + "learning_rate": 8.70562832254524e-06, + "loss": 1.1461, + "step": 19778 + }, + { + "epoch": 1.67, + "learning_rate": 8.70472700146571e-06, + "loss": 1.2398, + "step": 19779 + }, + { + "epoch": 1.67, + "learning_rate": 8.70382569108814e-06, + "loss": 1.1008, + "step": 19780 + }, + { + "epoch": 1.67, + "learning_rate": 8.702924391419973e-06, + "loss": 1.2613, + "step": 19781 + }, + { + "epoch": 1.67, + "learning_rate": 8.702023102468666e-06, + "loss": 1.1456, + "step": 19782 + }, + { + "epoch": 1.67, + "learning_rate": 8.701121824241657e-06, + "loss": 1.1066, + "step": 19783 + }, + { + "epoch": 1.67, + "learning_rate": 8.700220556746398e-06, + "loss": 1.1238, + "step": 19784 + }, + { + "epoch": 1.67, + "learning_rate": 8.699319299990327e-06, + "loss": 1.1956, + "step": 19785 + }, + { + "epoch": 1.67, + "learning_rate": 8.6984180539809e-06, + "loss": 1.0733, + "step": 19786 + }, + { + "epoch": 1.67, + "learning_rate": 8.697516818725558e-06, + "loss": 1.2055, + "step": 19787 + }, + { + "epoch": 1.67, + "learning_rate": 8.696615594231748e-06, + "loss": 1.1678, + "step": 19788 + }, + { + "epoch": 1.67, + "learning_rate": 8.69571438050692e-06, + "loss": 1.1069, + "step": 19789 + }, + { + "epoch": 1.67, + "learning_rate": 8.694813177558515e-06, + "loss": 1.1309, + "step": 19790 + }, + { + "epoch": 1.67, + "learning_rate": 8.693911985393985e-06, + "loss": 1.1575, + "step": 19791 + }, + { + "epoch": 1.67, + "learning_rate": 8.693010804020767e-06, + "loss": 1.1863, + "step": 19792 + }, + { + "epoch": 1.67, + "learning_rate": 8.692109633446316e-06, + "loss": 1.2354, + "step": 19793 + }, + { + "epoch": 1.67, + "learning_rate": 8.69120847367807e-06, + "loss": 1.2729, + "step": 19794 + }, + { + "epoch": 1.67, + "learning_rate": 8.690307324723483e-06, + "loss": 1.1732, + "step": 19795 + }, + { + "epoch": 1.67, + "learning_rate": 8.689406186589991e-06, + "loss": 1.1275, + "step": 19796 + }, + { + "epoch": 1.67, + "learning_rate": 8.688505059285051e-06, + "loss": 1.1297, + "step": 19797 + }, + { + "epoch": 1.67, + "learning_rate": 8.687603942816097e-06, + "loss": 1.1926, + "step": 19798 + }, + { + "epoch": 1.67, + "learning_rate": 8.686702837190582e-06, + "loss": 1.1457, + "step": 19799 + }, + { + "epoch": 1.67, + "learning_rate": 8.685801742415945e-06, + "loss": 1.1397, + "step": 19800 + }, + { + "epoch": 1.67, + "learning_rate": 8.684900658499634e-06, + "loss": 1.2983, + "step": 19801 + }, + { + "epoch": 1.67, + "learning_rate": 8.6839995854491e-06, + "loss": 1.1561, + "step": 19802 + }, + { + "epoch": 1.67, + "learning_rate": 8.683098523271779e-06, + "loss": 1.159, + "step": 19803 + }, + { + "epoch": 1.67, + "learning_rate": 8.682197471975122e-06, + "loss": 1.1818, + "step": 19804 + }, + { + "epoch": 1.67, + "learning_rate": 8.681296431566566e-06, + "loss": 1.1781, + "step": 19805 + }, + { + "epoch": 1.67, + "learning_rate": 8.680395402053565e-06, + "loss": 1.1522, + "step": 19806 + }, + { + "epoch": 1.67, + "learning_rate": 8.679494383443559e-06, + "loss": 1.1772, + "step": 19807 + }, + { + "epoch": 1.67, + "learning_rate": 8.678593375743996e-06, + "loss": 1.0982, + "step": 19808 + }, + { + "epoch": 1.67, + "learning_rate": 8.677692378962312e-06, + "loss": 1.271, + "step": 19809 + }, + { + "epoch": 1.67, + "learning_rate": 8.67679139310596e-06, + "loss": 1.1461, + "step": 19810 + }, + { + "epoch": 1.67, + "learning_rate": 8.67589041818238e-06, + "loss": 1.142, + "step": 19811 + }, + { + "epoch": 1.67, + "learning_rate": 8.674989454199015e-06, + "loss": 1.1596, + "step": 19812 + }, + { + "epoch": 1.67, + "learning_rate": 8.674088501163314e-06, + "loss": 1.1623, + "step": 19813 + }, + { + "epoch": 1.67, + "learning_rate": 8.673187559082718e-06, + "loss": 1.1601, + "step": 19814 + }, + { + "epoch": 1.67, + "learning_rate": 8.67228662796467e-06, + "loss": 1.1702, + "step": 19815 + }, + { + "epoch": 1.67, + "learning_rate": 8.671385707816615e-06, + "loss": 1.1359, + "step": 19816 + }, + { + "epoch": 1.67, + "learning_rate": 8.670484798645998e-06, + "loss": 1.1085, + "step": 19817 + }, + { + "epoch": 1.67, + "learning_rate": 8.66958390046026e-06, + "loss": 1.1727, + "step": 19818 + }, + { + "epoch": 1.67, + "learning_rate": 8.668683013266846e-06, + "loss": 1.1917, + "step": 19819 + }, + { + "epoch": 1.67, + "learning_rate": 8.667782137073197e-06, + "loss": 1.3245, + "step": 19820 + }, + { + "epoch": 1.67, + "learning_rate": 8.66688127188676e-06, + "loss": 1.1376, + "step": 19821 + }, + { + "epoch": 1.67, + "learning_rate": 8.665980417714977e-06, + "loss": 1.236, + "step": 19822 + }, + { + "epoch": 1.67, + "learning_rate": 8.665079574565291e-06, + "loss": 1.1632, + "step": 19823 + }, + { + "epoch": 1.67, + "learning_rate": 8.664178742445143e-06, + "loss": 1.169, + "step": 19824 + }, + { + "epoch": 1.67, + "learning_rate": 8.663277921361978e-06, + "loss": 1.1867, + "step": 19825 + }, + { + "epoch": 1.67, + "learning_rate": 8.662377111323242e-06, + "loss": 1.2257, + "step": 19826 + }, + { + "epoch": 1.67, + "learning_rate": 8.661476312336368e-06, + "loss": 1.1561, + "step": 19827 + }, + { + "epoch": 1.67, + "learning_rate": 8.66057552440881e-06, + "loss": 1.2065, + "step": 19828 + }, + { + "epoch": 1.67, + "learning_rate": 8.659674747548004e-06, + "loss": 1.2115, + "step": 19829 + }, + { + "epoch": 1.67, + "learning_rate": 8.658773981761397e-06, + "loss": 1.2172, + "step": 19830 + }, + { + "epoch": 1.67, + "learning_rate": 8.657873227056424e-06, + "loss": 1.1544, + "step": 19831 + }, + { + "epoch": 1.67, + "learning_rate": 8.656972483440535e-06, + "loss": 1.2038, + "step": 19832 + }, + { + "epoch": 1.67, + "learning_rate": 8.65607175092117e-06, + "loss": 1.176, + "step": 19833 + }, + { + "epoch": 1.67, + "learning_rate": 8.65517102950577e-06, + "loss": 1.1452, + "step": 19834 + }, + { + "epoch": 1.67, + "learning_rate": 8.654270319201773e-06, + "loss": 1.2647, + "step": 19835 + }, + { + "epoch": 1.67, + "learning_rate": 8.65336962001663e-06, + "loss": 1.1004, + "step": 19836 + }, + { + "epoch": 1.67, + "learning_rate": 8.652468931957776e-06, + "loss": 1.2055, + "step": 19837 + }, + { + "epoch": 1.67, + "learning_rate": 8.651568255032653e-06, + "loss": 1.1498, + "step": 19838 + }, + { + "epoch": 1.67, + "learning_rate": 8.650667589248708e-06, + "loss": 1.2628, + "step": 19839 + }, + { + "epoch": 1.67, + "learning_rate": 8.64976693461338e-06, + "loss": 1.1203, + "step": 19840 + }, + { + "epoch": 1.67, + "learning_rate": 8.648866291134109e-06, + "loss": 1.144, + "step": 19841 + }, + { + "epoch": 1.67, + "learning_rate": 8.647965658818334e-06, + "loss": 1.1235, + "step": 19842 + }, + { + "epoch": 1.67, + "learning_rate": 8.647065037673503e-06, + "loss": 1.1844, + "step": 19843 + }, + { + "epoch": 1.67, + "learning_rate": 8.64616442770705e-06, + "loss": 1.3147, + "step": 19844 + }, + { + "epoch": 1.67, + "learning_rate": 8.645263828926426e-06, + "loss": 1.211, + "step": 19845 + }, + { + "epoch": 1.67, + "learning_rate": 8.644363241339059e-06, + "loss": 1.1772, + "step": 19846 + }, + { + "epoch": 1.67, + "learning_rate": 8.643462664952401e-06, + "loss": 1.1345, + "step": 19847 + }, + { + "epoch": 1.67, + "learning_rate": 8.642562099773886e-06, + "loss": 1.2393, + "step": 19848 + }, + { + "epoch": 1.67, + "learning_rate": 8.641661545810962e-06, + "loss": 1.2232, + "step": 19849 + }, + { + "epoch": 1.67, + "learning_rate": 8.640761003071059e-06, + "loss": 1.2113, + "step": 19850 + }, + { + "epoch": 1.67, + "learning_rate": 8.639860471561625e-06, + "loss": 1.1317, + "step": 19851 + }, + { + "epoch": 1.67, + "learning_rate": 8.638959951290102e-06, + "loss": 1.1124, + "step": 19852 + }, + { + "epoch": 1.67, + "learning_rate": 8.638059442263924e-06, + "loss": 1.136, + "step": 19853 + }, + { + "epoch": 1.67, + "learning_rate": 8.63715894449054e-06, + "loss": 1.2609, + "step": 19854 + }, + { + "epoch": 1.67, + "learning_rate": 8.63625845797738e-06, + "loss": 1.2731, + "step": 19855 + }, + { + "epoch": 1.67, + "learning_rate": 8.63535798273189e-06, + "loss": 1.1058, + "step": 19856 + }, + { + "epoch": 1.67, + "learning_rate": 8.634457518761511e-06, + "loss": 1.1551, + "step": 19857 + }, + { + "epoch": 1.67, + "learning_rate": 8.633557066073681e-06, + "loss": 1.1449, + "step": 19858 + }, + { + "epoch": 1.67, + "learning_rate": 8.632656624675837e-06, + "loss": 1.1759, + "step": 19859 + }, + { + "epoch": 1.67, + "learning_rate": 8.631756194575425e-06, + "loss": 1.2029, + "step": 19860 + }, + { + "epoch": 1.67, + "learning_rate": 8.630855775779878e-06, + "loss": 1.2324, + "step": 19861 + }, + { + "epoch": 1.67, + "learning_rate": 8.629955368296638e-06, + "loss": 1.2259, + "step": 19862 + }, + { + "epoch": 1.67, + "learning_rate": 8.629054972133148e-06, + "loss": 1.1897, + "step": 19863 + }, + { + "epoch": 1.67, + "learning_rate": 8.628154587296844e-06, + "loss": 1.1423, + "step": 19864 + }, + { + "epoch": 1.67, + "learning_rate": 8.627254213795167e-06, + "loss": 1.2224, + "step": 19865 + }, + { + "epoch": 1.67, + "learning_rate": 8.626353851635552e-06, + "loss": 1.0974, + "step": 19866 + }, + { + "epoch": 1.67, + "learning_rate": 8.625453500825444e-06, + "loss": 1.1331, + "step": 19867 + }, + { + "epoch": 1.67, + "learning_rate": 8.624553161372277e-06, + "loss": 1.1984, + "step": 19868 + }, + { + "epoch": 1.67, + "learning_rate": 8.623652833283496e-06, + "loss": 1.2265, + "step": 19869 + }, + { + "epoch": 1.67, + "learning_rate": 8.622752516566531e-06, + "loss": 1.184, + "step": 19870 + }, + { + "epoch": 1.67, + "learning_rate": 8.621852211228828e-06, + "loss": 1.1685, + "step": 19871 + }, + { + "epoch": 1.67, + "learning_rate": 8.620951917277822e-06, + "loss": 1.2455, + "step": 19872 + }, + { + "epoch": 1.67, + "learning_rate": 8.620051634720954e-06, + "loss": 1.2435, + "step": 19873 + }, + { + "epoch": 1.67, + "learning_rate": 8.619151363565658e-06, + "loss": 1.184, + "step": 19874 + }, + { + "epoch": 1.67, + "learning_rate": 8.618251103819378e-06, + "loss": 1.1161, + "step": 19875 + }, + { + "epoch": 1.67, + "learning_rate": 8.61735085548955e-06, + "loss": 1.1863, + "step": 19876 + }, + { + "epoch": 1.67, + "learning_rate": 8.616450618583607e-06, + "loss": 1.1986, + "step": 19877 + }, + { + "epoch": 1.67, + "learning_rate": 8.615550393108998e-06, + "loss": 1.1792, + "step": 19878 + }, + { + "epoch": 1.67, + "learning_rate": 8.614650179073152e-06, + "loss": 1.1874, + "step": 19879 + }, + { + "epoch": 1.67, + "learning_rate": 8.613749976483511e-06, + "loss": 1.2613, + "step": 19880 + }, + { + "epoch": 1.67, + "learning_rate": 8.612849785347508e-06, + "loss": 1.1508, + "step": 19881 + }, + { + "epoch": 1.67, + "learning_rate": 8.611949605672586e-06, + "loss": 1.1939, + "step": 19882 + }, + { + "epoch": 1.67, + "learning_rate": 8.611049437466181e-06, + "loss": 1.1316, + "step": 19883 + }, + { + "epoch": 1.67, + "learning_rate": 8.610149280735733e-06, + "loss": 1.1492, + "step": 19884 + }, + { + "epoch": 1.67, + "learning_rate": 8.609249135488669e-06, + "loss": 1.2245, + "step": 19885 + }, + { + "epoch": 1.67, + "learning_rate": 8.60834900173244e-06, + "loss": 1.1706, + "step": 19886 + }, + { + "epoch": 1.67, + "learning_rate": 8.607448879474475e-06, + "loss": 1.1382, + "step": 19887 + }, + { + "epoch": 1.67, + "learning_rate": 8.606548768722211e-06, + "loss": 1.1553, + "step": 19888 + }, + { + "epoch": 1.67, + "learning_rate": 8.60564866948309e-06, + "loss": 1.2024, + "step": 19889 + }, + { + "epoch": 1.67, + "learning_rate": 8.604748581764545e-06, + "loss": 1.2608, + "step": 19890 + }, + { + "epoch": 1.67, + "learning_rate": 8.603848505574015e-06, + "loss": 1.2448, + "step": 19891 + }, + { + "epoch": 1.67, + "learning_rate": 8.602948440918931e-06, + "loss": 1.1012, + "step": 19892 + }, + { + "epoch": 1.67, + "learning_rate": 8.60204838780674e-06, + "loss": 1.1903, + "step": 19893 + }, + { + "epoch": 1.67, + "learning_rate": 8.601148346244869e-06, + "loss": 1.1606, + "step": 19894 + }, + { + "epoch": 1.67, + "learning_rate": 8.60024831624076e-06, + "loss": 1.1938, + "step": 19895 + }, + { + "epoch": 1.68, + "learning_rate": 8.599348297801843e-06, + "loss": 1.1943, + "step": 19896 + }, + { + "epoch": 1.68, + "learning_rate": 8.598448290935565e-06, + "loss": 1.1993, + "step": 19897 + }, + { + "epoch": 1.68, + "learning_rate": 8.597548295649352e-06, + "loss": 1.1033, + "step": 19898 + }, + { + "epoch": 1.68, + "learning_rate": 8.596648311950647e-06, + "loss": 1.1769, + "step": 19899 + }, + { + "epoch": 1.68, + "learning_rate": 8.595748339846877e-06, + "loss": 1.0826, + "step": 19900 + }, + { + "epoch": 1.68, + "learning_rate": 8.594848379345487e-06, + "loss": 1.2533, + "step": 19901 + }, + { + "epoch": 1.68, + "learning_rate": 8.593948430453911e-06, + "loss": 1.1217, + "step": 19902 + }, + { + "epoch": 1.68, + "learning_rate": 8.593048493179579e-06, + "loss": 1.2056, + "step": 19903 + }, + { + "epoch": 1.68, + "learning_rate": 8.592148567529937e-06, + "loss": 1.2236, + "step": 19904 + }, + { + "epoch": 1.68, + "learning_rate": 8.591248653512407e-06, + "loss": 1.2357, + "step": 19905 + }, + { + "epoch": 1.68, + "learning_rate": 8.590348751134436e-06, + "loss": 1.1936, + "step": 19906 + }, + { + "epoch": 1.68, + "learning_rate": 8.589448860403453e-06, + "loss": 1.2312, + "step": 19907 + }, + { + "epoch": 1.68, + "learning_rate": 8.588548981326898e-06, + "loss": 1.2316, + "step": 19908 + }, + { + "epoch": 1.68, + "learning_rate": 8.587649113912197e-06, + "loss": 1.1729, + "step": 19909 + }, + { + "epoch": 1.68, + "learning_rate": 8.586749258166797e-06, + "loss": 1.1969, + "step": 19910 + }, + { + "epoch": 1.68, + "learning_rate": 8.585849414098124e-06, + "loss": 1.2245, + "step": 19911 + }, + { + "epoch": 1.68, + "learning_rate": 8.584949581713612e-06, + "loss": 1.1263, + "step": 19912 + }, + { + "epoch": 1.68, + "learning_rate": 8.584049761020706e-06, + "loss": 1.1163, + "step": 19913 + }, + { + "epoch": 1.68, + "learning_rate": 8.583149952026831e-06, + "loss": 1.1491, + "step": 19914 + }, + { + "epoch": 1.68, + "learning_rate": 8.582250154739427e-06, + "loss": 1.1916, + "step": 19915 + }, + { + "epoch": 1.68, + "learning_rate": 8.581350369165922e-06, + "loss": 1.1078, + "step": 19916 + }, + { + "epoch": 1.68, + "learning_rate": 8.580450595313758e-06, + "loss": 1.1922, + "step": 19917 + }, + { + "epoch": 1.68, + "learning_rate": 8.579550833190362e-06, + "loss": 1.2465, + "step": 19918 + }, + { + "epoch": 1.68, + "learning_rate": 8.578651082803175e-06, + "loss": 1.2157, + "step": 19919 + }, + { + "epoch": 1.68, + "learning_rate": 8.577751344159624e-06, + "loss": 1.1583, + "step": 19920 + }, + { + "epoch": 1.68, + "learning_rate": 8.576851617267151e-06, + "loss": 1.2146, + "step": 19921 + }, + { + "epoch": 1.68, + "learning_rate": 8.575951902133182e-06, + "loss": 1.1209, + "step": 19922 + }, + { + "epoch": 1.68, + "learning_rate": 8.575052198765157e-06, + "loss": 1.1647, + "step": 19923 + }, + { + "epoch": 1.68, + "learning_rate": 8.574152507170502e-06, + "loss": 1.1792, + "step": 19924 + }, + { + "epoch": 1.68, + "learning_rate": 8.573252827356658e-06, + "loss": 1.2396, + "step": 19925 + }, + { + "epoch": 1.68, + "learning_rate": 8.572353159331059e-06, + "loss": 1.1434, + "step": 19926 + }, + { + "epoch": 1.68, + "learning_rate": 8.571453503101129e-06, + "loss": 1.2649, + "step": 19927 + }, + { + "epoch": 1.68, + "learning_rate": 8.570553858674313e-06, + "loss": 1.2116, + "step": 19928 + }, + { + "epoch": 1.68, + "learning_rate": 8.569654226058035e-06, + "loss": 1.1238, + "step": 19929 + }, + { + "epoch": 1.68, + "learning_rate": 8.568754605259733e-06, + "loss": 1.1526, + "step": 19930 + }, + { + "epoch": 1.68, + "learning_rate": 8.567854996286836e-06, + "loss": 1.1066, + "step": 19931 + }, + { + "epoch": 1.68, + "learning_rate": 8.566955399146783e-06, + "loss": 1.1507, + "step": 19932 + }, + { + "epoch": 1.68, + "learning_rate": 8.566055813847002e-06, + "loss": 1.2177, + "step": 19933 + }, + { + "epoch": 1.68, + "learning_rate": 8.565156240394927e-06, + "loss": 1.1592, + "step": 19934 + }, + { + "epoch": 1.68, + "learning_rate": 8.564256678797987e-06, + "loss": 1.2215, + "step": 19935 + }, + { + "epoch": 1.68, + "learning_rate": 8.563357129063624e-06, + "loss": 1.2352, + "step": 19936 + }, + { + "epoch": 1.68, + "learning_rate": 8.562457591199258e-06, + "loss": 1.196, + "step": 19937 + }, + { + "epoch": 1.68, + "learning_rate": 8.56155806521233e-06, + "loss": 1.2007, + "step": 19938 + }, + { + "epoch": 1.68, + "learning_rate": 8.56065855111027e-06, + "loss": 1.2381, + "step": 19939 + }, + { + "epoch": 1.68, + "learning_rate": 8.559759048900508e-06, + "loss": 1.1859, + "step": 19940 + }, + { + "epoch": 1.68, + "learning_rate": 8.558859558590481e-06, + "loss": 1.1973, + "step": 19941 + }, + { + "epoch": 1.68, + "learning_rate": 8.557960080187614e-06, + "loss": 1.1078, + "step": 19942 + }, + { + "epoch": 1.68, + "learning_rate": 8.557060613699347e-06, + "loss": 1.1595, + "step": 19943 + }, + { + "epoch": 1.68, + "learning_rate": 8.556161159133102e-06, + "loss": 1.1533, + "step": 19944 + }, + { + "epoch": 1.68, + "learning_rate": 8.55526171649632e-06, + "loss": 1.2012, + "step": 19945 + }, + { + "epoch": 1.68, + "learning_rate": 8.554362285796423e-06, + "loss": 1.2059, + "step": 19946 + }, + { + "epoch": 1.68, + "learning_rate": 8.553462867040851e-06, + "loss": 1.2109, + "step": 19947 + }, + { + "epoch": 1.68, + "learning_rate": 8.552563460237031e-06, + "loss": 1.1967, + "step": 19948 + }, + { + "epoch": 1.68, + "learning_rate": 8.551664065392397e-06, + "loss": 1.0989, + "step": 19949 + }, + { + "epoch": 1.68, + "learning_rate": 8.550764682514373e-06, + "loss": 1.136, + "step": 19950 + }, + { + "epoch": 1.68, + "learning_rate": 8.549865311610397e-06, + "loss": 1.0247, + "step": 19951 + }, + { + "epoch": 1.68, + "learning_rate": 8.548965952687902e-06, + "loss": 1.1677, + "step": 19952 + }, + { + "epoch": 1.68, + "learning_rate": 8.548066605754309e-06, + "loss": 1.1034, + "step": 19953 + }, + { + "epoch": 1.68, + "learning_rate": 8.54716727081706e-06, + "loss": 1.1438, + "step": 19954 + }, + { + "epoch": 1.68, + "learning_rate": 8.546267947883573e-06, + "loss": 1.1848, + "step": 19955 + }, + { + "epoch": 1.68, + "learning_rate": 8.545368636961292e-06, + "loss": 1.1507, + "step": 19956 + }, + { + "epoch": 1.68, + "learning_rate": 8.544469338057637e-06, + "loss": 1.2256, + "step": 19957 + }, + { + "epoch": 1.68, + "learning_rate": 8.543570051180045e-06, + "loss": 1.192, + "step": 19958 + }, + { + "epoch": 1.68, + "learning_rate": 8.54267077633594e-06, + "loss": 1.1815, + "step": 19959 + }, + { + "epoch": 1.68, + "learning_rate": 8.541771513532758e-06, + "loss": 1.1768, + "step": 19960 + }, + { + "epoch": 1.68, + "learning_rate": 8.540872262777925e-06, + "loss": 1.1524, + "step": 19961 + }, + { + "epoch": 1.68, + "learning_rate": 8.53997302407887e-06, + "loss": 1.2148, + "step": 19962 + }, + { + "epoch": 1.68, + "learning_rate": 8.539073797443031e-06, + "loss": 1.2062, + "step": 19963 + }, + { + "epoch": 1.68, + "learning_rate": 8.538174582877828e-06, + "loss": 1.1847, + "step": 19964 + }, + { + "epoch": 1.68, + "learning_rate": 8.537275380390698e-06, + "loss": 1.1638, + "step": 19965 + }, + { + "epoch": 1.68, + "learning_rate": 8.536376189989061e-06, + "loss": 1.298, + "step": 19966 + }, + { + "epoch": 1.68, + "learning_rate": 8.535477011680356e-06, + "loss": 1.1058, + "step": 19967 + }, + { + "epoch": 1.68, + "learning_rate": 8.534577845472008e-06, + "loss": 1.2212, + "step": 19968 + }, + { + "epoch": 1.68, + "learning_rate": 8.53367869137145e-06, + "loss": 1.1485, + "step": 19969 + }, + { + "epoch": 1.68, + "learning_rate": 8.532779549386101e-06, + "loss": 1.1224, + "step": 19970 + }, + { + "epoch": 1.68, + "learning_rate": 8.531880419523403e-06, + "loss": 1.147, + "step": 19971 + }, + { + "epoch": 1.68, + "learning_rate": 8.530981301790776e-06, + "loss": 1.2248, + "step": 19972 + }, + { + "epoch": 1.68, + "learning_rate": 8.530082196195655e-06, + "loss": 1.189, + "step": 19973 + }, + { + "epoch": 1.68, + "learning_rate": 8.52918310274546e-06, + "loss": 1.1741, + "step": 19974 + }, + { + "epoch": 1.68, + "learning_rate": 8.528284021447627e-06, + "loss": 1.2663, + "step": 19975 + }, + { + "epoch": 1.68, + "learning_rate": 8.527384952309585e-06, + "loss": 1.1045, + "step": 19976 + }, + { + "epoch": 1.68, + "learning_rate": 8.526485895338755e-06, + "loss": 1.1804, + "step": 19977 + }, + { + "epoch": 1.68, + "learning_rate": 8.525586850542576e-06, + "loss": 1.252, + "step": 19978 + }, + { + "epoch": 1.68, + "learning_rate": 8.524687817928466e-06, + "loss": 1.1892, + "step": 19979 + }, + { + "epoch": 1.68, + "learning_rate": 8.52378879750386e-06, + "loss": 1.1493, + "step": 19980 + }, + { + "epoch": 1.68, + "learning_rate": 8.52288978927618e-06, + "loss": 1.2107, + "step": 19981 + }, + { + "epoch": 1.68, + "learning_rate": 8.52199079325286e-06, + "loss": 1.1756, + "step": 19982 + }, + { + "epoch": 1.68, + "learning_rate": 8.521091809441326e-06, + "loss": 1.1699, + "step": 19983 + }, + { + "epoch": 1.68, + "learning_rate": 8.520192837849004e-06, + "loss": 1.1978, + "step": 19984 + }, + { + "epoch": 1.68, + "learning_rate": 8.519293878483319e-06, + "loss": 1.2658, + "step": 19985 + }, + { + "epoch": 1.68, + "learning_rate": 8.518394931351708e-06, + "loss": 1.0962, + "step": 19986 + }, + { + "epoch": 1.68, + "learning_rate": 8.517495996461586e-06, + "loss": 1.1904, + "step": 19987 + }, + { + "epoch": 1.68, + "learning_rate": 8.516597073820388e-06, + "loss": 1.2094, + "step": 19988 + }, + { + "epoch": 1.68, + "learning_rate": 8.515698163435542e-06, + "loss": 1.2175, + "step": 19989 + }, + { + "epoch": 1.68, + "learning_rate": 8.514799265314472e-06, + "loss": 1.1555, + "step": 19990 + }, + { + "epoch": 1.68, + "learning_rate": 8.513900379464607e-06, + "loss": 1.2675, + "step": 19991 + }, + { + "epoch": 1.68, + "learning_rate": 8.51300150589337e-06, + "loss": 1.1777, + "step": 19992 + }, + { + "epoch": 1.68, + "learning_rate": 8.512102644608192e-06, + "loss": 1.1341, + "step": 19993 + }, + { + "epoch": 1.68, + "learning_rate": 8.511203795616498e-06, + "loss": 1.2264, + "step": 19994 + }, + { + "epoch": 1.68, + "learning_rate": 8.510304958925716e-06, + "loss": 1.1936, + "step": 19995 + }, + { + "epoch": 1.68, + "learning_rate": 8.509406134543267e-06, + "loss": 1.1754, + "step": 19996 + }, + { + "epoch": 1.68, + "learning_rate": 8.508507322476587e-06, + "loss": 1.2388, + "step": 19997 + }, + { + "epoch": 1.68, + "learning_rate": 8.507608522733093e-06, + "loss": 1.1627, + "step": 19998 + }, + { + "epoch": 1.68, + "learning_rate": 8.506709735320219e-06, + "loss": 1.1748, + "step": 19999 + }, + { + "epoch": 1.68, + "learning_rate": 8.50581096024538e-06, + "loss": 1.1142, + "step": 20000 + }, + { + "epoch": 1.68, + "learning_rate": 8.504912197516015e-06, + "loss": 1.1164, + "step": 20001 + }, + { + "epoch": 1.68, + "learning_rate": 8.504013447139542e-06, + "loss": 1.1693, + "step": 20002 + }, + { + "epoch": 1.68, + "learning_rate": 8.503114709123388e-06, + "loss": 1.3182, + "step": 20003 + }, + { + "epoch": 1.68, + "learning_rate": 8.502215983474982e-06, + "loss": 1.1908, + "step": 20004 + }, + { + "epoch": 1.68, + "learning_rate": 8.501317270201744e-06, + "loss": 1.1891, + "step": 20005 + }, + { + "epoch": 1.68, + "learning_rate": 8.500418569311107e-06, + "loss": 1.2252, + "step": 20006 + }, + { + "epoch": 1.68, + "learning_rate": 8.499519880810486e-06, + "loss": 1.2627, + "step": 20007 + }, + { + "epoch": 1.68, + "learning_rate": 8.498621204707317e-06, + "loss": 1.1227, + "step": 20008 + }, + { + "epoch": 1.68, + "learning_rate": 8.497722541009015e-06, + "loss": 1.1992, + "step": 20009 + }, + { + "epoch": 1.68, + "learning_rate": 8.496823889723014e-06, + "loss": 1.1152, + "step": 20010 + }, + { + "epoch": 1.68, + "learning_rate": 8.495925250856732e-06, + "loss": 1.2091, + "step": 20011 + }, + { + "epoch": 1.68, + "learning_rate": 8.495026624417597e-06, + "loss": 1.1362, + "step": 20012 + }, + { + "epoch": 1.68, + "learning_rate": 8.494128010413036e-06, + "loss": 1.06, + "step": 20013 + }, + { + "epoch": 1.69, + "learning_rate": 8.49322940885047e-06, + "loss": 1.2453, + "step": 20014 + }, + { + "epoch": 1.69, + "learning_rate": 8.492330819737328e-06, + "loss": 1.1579, + "step": 20015 + }, + { + "epoch": 1.69, + "learning_rate": 8.491432243081026e-06, + "loss": 1.1963, + "step": 20016 + }, + { + "epoch": 1.69, + "learning_rate": 8.490533678888997e-06, + "loss": 1.1954, + "step": 20017 + }, + { + "epoch": 1.69, + "learning_rate": 8.489635127168661e-06, + "loss": 1.0894, + "step": 20018 + }, + { + "epoch": 1.69, + "learning_rate": 8.488736587927445e-06, + "loss": 1.1487, + "step": 20019 + }, + { + "epoch": 1.69, + "learning_rate": 8.487838061172768e-06, + "loss": 1.2847, + "step": 20020 + }, + { + "epoch": 1.69, + "learning_rate": 8.486939546912059e-06, + "loss": 1.1562, + "step": 20021 + }, + { + "epoch": 1.69, + "learning_rate": 8.486041045152738e-06, + "loss": 1.2124, + "step": 20022 + }, + { + "epoch": 1.69, + "learning_rate": 8.485142555902232e-06, + "loss": 1.2293, + "step": 20023 + }, + { + "epoch": 1.69, + "learning_rate": 8.484244079167961e-06, + "loss": 1.1143, + "step": 20024 + }, + { + "epoch": 1.69, + "learning_rate": 8.48334561495735e-06, + "loss": 1.1169, + "step": 20025 + }, + { + "epoch": 1.69, + "learning_rate": 8.482447163277827e-06, + "loss": 1.2129, + "step": 20026 + }, + { + "epoch": 1.69, + "learning_rate": 8.481548724136806e-06, + "loss": 1.2986, + "step": 20027 + }, + { + "epoch": 1.69, + "learning_rate": 8.480650297541719e-06, + "loss": 1.1259, + "step": 20028 + }, + { + "epoch": 1.69, + "learning_rate": 8.479751883499985e-06, + "loss": 1.2678, + "step": 20029 + }, + { + "epoch": 1.69, + "learning_rate": 8.47885348201903e-06, + "loss": 1.1251, + "step": 20030 + }, + { + "epoch": 1.69, + "learning_rate": 8.477955093106268e-06, + "loss": 1.2716, + "step": 20031 + }, + { + "epoch": 1.69, + "learning_rate": 8.477056716769133e-06, + "loss": 1.1423, + "step": 20032 + }, + { + "epoch": 1.69, + "learning_rate": 8.47615835301504e-06, + "loss": 1.1534, + "step": 20033 + }, + { + "epoch": 1.69, + "learning_rate": 8.475260001851418e-06, + "loss": 1.1337, + "step": 20034 + }, + { + "epoch": 1.69, + "learning_rate": 8.474361663285681e-06, + "loss": 1.2127, + "step": 20035 + }, + { + "epoch": 1.69, + "learning_rate": 8.473463337325262e-06, + "loss": 1.1526, + "step": 20036 + }, + { + "epoch": 1.69, + "learning_rate": 8.472565023977573e-06, + "loss": 1.2237, + "step": 20037 + }, + { + "epoch": 1.69, + "learning_rate": 8.47166672325004e-06, + "loss": 1.2518, + "step": 20038 + }, + { + "epoch": 1.69, + "learning_rate": 8.470768435150088e-06, + "loss": 1.1202, + "step": 20039 + }, + { + "epoch": 1.69, + "learning_rate": 8.469870159685136e-06, + "loss": 1.1944, + "step": 20040 + }, + { + "epoch": 1.69, + "learning_rate": 8.468971896862608e-06, + "loss": 1.1921, + "step": 20041 + }, + { + "epoch": 1.69, + "learning_rate": 8.468073646689922e-06, + "loss": 1.389, + "step": 20042 + }, + { + "epoch": 1.69, + "learning_rate": 8.467175409174504e-06, + "loss": 1.2588, + "step": 20043 + }, + { + "epoch": 1.69, + "learning_rate": 8.46627718432377e-06, + "loss": 1.1385, + "step": 20044 + }, + { + "epoch": 1.69, + "learning_rate": 8.46537897214515e-06, + "loss": 1.1522, + "step": 20045 + }, + { + "epoch": 1.69, + "learning_rate": 8.464480772646055e-06, + "loss": 1.1905, + "step": 20046 + }, + { + "epoch": 1.69, + "learning_rate": 8.463582585833915e-06, + "loss": 1.1805, + "step": 20047 + }, + { + "epoch": 1.69, + "learning_rate": 8.462684411716145e-06, + "loss": 1.1662, + "step": 20048 + }, + { + "epoch": 1.69, + "learning_rate": 8.46178625030017e-06, + "loss": 1.041, + "step": 20049 + }, + { + "epoch": 1.69, + "learning_rate": 8.460888101593406e-06, + "loss": 1.1843, + "step": 20050 + }, + { + "epoch": 1.69, + "learning_rate": 8.459989965603277e-06, + "loss": 1.125, + "step": 20051 + }, + { + "epoch": 1.69, + "learning_rate": 8.459091842337209e-06, + "loss": 1.1845, + "step": 20052 + }, + { + "epoch": 1.69, + "learning_rate": 8.458193731802611e-06, + "loss": 1.2074, + "step": 20053 + }, + { + "epoch": 1.69, + "learning_rate": 8.457295634006916e-06, + "loss": 1.2373, + "step": 20054 + }, + { + "epoch": 1.69, + "learning_rate": 8.456397548957534e-06, + "loss": 1.1612, + "step": 20055 + }, + { + "epoch": 1.69, + "learning_rate": 8.455499476661892e-06, + "loss": 1.1981, + "step": 20056 + }, + { + "epoch": 1.69, + "learning_rate": 8.454601417127405e-06, + "loss": 1.1242, + "step": 20057 + }, + { + "epoch": 1.69, + "learning_rate": 8.453703370361502e-06, + "loss": 1.1599, + "step": 20058 + }, + { + "epoch": 1.69, + "learning_rate": 8.452805336371588e-06, + "loss": 1.2116, + "step": 20059 + }, + { + "epoch": 1.69, + "learning_rate": 8.451907315165098e-06, + "loss": 1.1992, + "step": 20060 + }, + { + "epoch": 1.69, + "learning_rate": 8.451009306749442e-06, + "loss": 1.1922, + "step": 20061 + }, + { + "epoch": 1.69, + "learning_rate": 8.45011131113204e-06, + "loss": 1.1414, + "step": 20062 + }, + { + "epoch": 1.69, + "learning_rate": 8.449213328320319e-06, + "loss": 1.1633, + "step": 20063 + }, + { + "epoch": 1.69, + "learning_rate": 8.448315358321692e-06, + "loss": 1.2637, + "step": 20064 + }, + { + "epoch": 1.69, + "learning_rate": 8.447417401143582e-06, + "loss": 1.2093, + "step": 20065 + }, + { + "epoch": 1.69, + "learning_rate": 8.446519456793405e-06, + "loss": 1.1041, + "step": 20066 + }, + { + "epoch": 1.69, + "learning_rate": 8.445621525278582e-06, + "loss": 1.1474, + "step": 20067 + }, + { + "epoch": 1.69, + "learning_rate": 8.44472360660653e-06, + "loss": 1.1878, + "step": 20068 + }, + { + "epoch": 1.69, + "learning_rate": 8.443825700784671e-06, + "loss": 1.1668, + "step": 20069 + }, + { + "epoch": 1.69, + "learning_rate": 8.44292780782042e-06, + "loss": 1.2292, + "step": 20070 + }, + { + "epoch": 1.69, + "learning_rate": 8.4420299277212e-06, + "loss": 1.1759, + "step": 20071 + }, + { + "epoch": 1.69, + "learning_rate": 8.441132060494426e-06, + "loss": 1.1488, + "step": 20072 + }, + { + "epoch": 1.69, + "learning_rate": 8.44023420614752e-06, + "loss": 1.2393, + "step": 20073 + }, + { + "epoch": 1.69, + "learning_rate": 8.439336364687893e-06, + "loss": 1.0906, + "step": 20074 + }, + { + "epoch": 1.69, + "learning_rate": 8.438438536122973e-06, + "loss": 1.1771, + "step": 20075 + }, + { + "epoch": 1.69, + "learning_rate": 8.437540720460173e-06, + "loss": 1.1393, + "step": 20076 + }, + { + "epoch": 1.69, + "learning_rate": 8.43664291770691e-06, + "loss": 1.1549, + "step": 20077 + }, + { + "epoch": 1.69, + "learning_rate": 8.435745127870605e-06, + "loss": 1.1634, + "step": 20078 + }, + { + "epoch": 1.69, + "learning_rate": 8.434847350958674e-06, + "loss": 1.2375, + "step": 20079 + }, + { + "epoch": 1.69, + "learning_rate": 8.433949586978536e-06, + "loss": 1.1833, + "step": 20080 + }, + { + "epoch": 1.69, + "learning_rate": 8.433051835937605e-06, + "loss": 1.1808, + "step": 20081 + }, + { + "epoch": 1.69, + "learning_rate": 8.432154097843304e-06, + "loss": 1.1789, + "step": 20082 + }, + { + "epoch": 1.69, + "learning_rate": 8.431256372703046e-06, + "loss": 1.218, + "step": 20083 + }, + { + "epoch": 1.69, + "learning_rate": 8.430358660524253e-06, + "loss": 1.1342, + "step": 20084 + }, + { + "epoch": 1.69, + "learning_rate": 8.429460961314333e-06, + "loss": 1.1103, + "step": 20085 + }, + { + "epoch": 1.69, + "learning_rate": 8.428563275080714e-06, + "loss": 1.2536, + "step": 20086 + }, + { + "epoch": 1.69, + "learning_rate": 8.427665601830807e-06, + "loss": 1.2019, + "step": 20087 + }, + { + "epoch": 1.69, + "learning_rate": 8.426767941572028e-06, + "loss": 1.0446, + "step": 20088 + }, + { + "epoch": 1.69, + "learning_rate": 8.4258702943118e-06, + "loss": 1.1966, + "step": 20089 + }, + { + "epoch": 1.69, + "learning_rate": 8.424972660057532e-06, + "loss": 1.2583, + "step": 20090 + }, + { + "epoch": 1.69, + "learning_rate": 8.424075038816646e-06, + "loss": 1.1824, + "step": 20091 + }, + { + "epoch": 1.69, + "learning_rate": 8.423177430596552e-06, + "loss": 1.2126, + "step": 20092 + }, + { + "epoch": 1.69, + "learning_rate": 8.422279835404677e-06, + "loss": 1.2206, + "step": 20093 + }, + { + "epoch": 1.69, + "learning_rate": 8.421382253248427e-06, + "loss": 1.1915, + "step": 20094 + }, + { + "epoch": 1.69, + "learning_rate": 8.420484684135226e-06, + "loss": 1.2098, + "step": 20095 + }, + { + "epoch": 1.69, + "learning_rate": 8.41958712807248e-06, + "loss": 1.2509, + "step": 20096 + }, + { + "epoch": 1.69, + "learning_rate": 8.418689585067616e-06, + "loss": 1.2693, + "step": 20097 + }, + { + "epoch": 1.69, + "learning_rate": 8.417792055128042e-06, + "loss": 1.2583, + "step": 20098 + }, + { + "epoch": 1.69, + "learning_rate": 8.41689453826118e-06, + "loss": 1.1655, + "step": 20099 + }, + { + "epoch": 1.69, + "learning_rate": 8.415997034474436e-06, + "loss": 1.1744, + "step": 20100 + }, + { + "epoch": 1.69, + "learning_rate": 8.415099543775234e-06, + "loss": 1.1814, + "step": 20101 + }, + { + "epoch": 1.69, + "learning_rate": 8.41420206617099e-06, + "loss": 1.152, + "step": 20102 + }, + { + "epoch": 1.69, + "learning_rate": 8.41330460166911e-06, + "loss": 1.1562, + "step": 20103 + }, + { + "epoch": 1.69, + "learning_rate": 8.412407150277021e-06, + "loss": 1.1507, + "step": 20104 + }, + { + "epoch": 1.69, + "learning_rate": 8.41150971200213e-06, + "loss": 1.1736, + "step": 20105 + }, + { + "epoch": 1.69, + "learning_rate": 8.410612286851856e-06, + "loss": 1.0823, + "step": 20106 + }, + { + "epoch": 1.69, + "learning_rate": 8.409714874833607e-06, + "loss": 1.184, + "step": 20107 + }, + { + "epoch": 1.69, + "learning_rate": 8.40881747595481e-06, + "loss": 1.2046, + "step": 20108 + }, + { + "epoch": 1.69, + "learning_rate": 8.407920090222866e-06, + "loss": 1.0951, + "step": 20109 + }, + { + "epoch": 1.69, + "learning_rate": 8.4070227176452e-06, + "loss": 1.2711, + "step": 20110 + }, + { + "epoch": 1.69, + "learning_rate": 8.406125358229218e-06, + "loss": 1.1785, + "step": 20111 + }, + { + "epoch": 1.69, + "learning_rate": 8.405228011982338e-06, + "loss": 1.2421, + "step": 20112 + }, + { + "epoch": 1.69, + "learning_rate": 8.404330678911978e-06, + "loss": 1.1841, + "step": 20113 + }, + { + "epoch": 1.69, + "learning_rate": 8.403433359025547e-06, + "loss": 1.1247, + "step": 20114 + }, + { + "epoch": 1.69, + "learning_rate": 8.402536052330463e-06, + "loss": 1.1736, + "step": 20115 + }, + { + "epoch": 1.69, + "learning_rate": 8.401638758834132e-06, + "loss": 1.2173, + "step": 20116 + }, + { + "epoch": 1.69, + "learning_rate": 8.400741478543979e-06, + "loss": 1.1572, + "step": 20117 + }, + { + "epoch": 1.69, + "learning_rate": 8.399844211467409e-06, + "loss": 1.1695, + "step": 20118 + }, + { + "epoch": 1.69, + "learning_rate": 8.398946957611839e-06, + "loss": 1.1931, + "step": 20119 + }, + { + "epoch": 1.69, + "learning_rate": 8.39804971698468e-06, + "loss": 1.1459, + "step": 20120 + }, + { + "epoch": 1.69, + "learning_rate": 8.397152489593348e-06, + "loss": 1.2025, + "step": 20121 + }, + { + "epoch": 1.69, + "learning_rate": 8.396255275445255e-06, + "loss": 1.1902, + "step": 20122 + }, + { + "epoch": 1.69, + "learning_rate": 8.395358074547816e-06, + "loss": 1.1286, + "step": 20123 + }, + { + "epoch": 1.69, + "learning_rate": 8.394460886908439e-06, + "loss": 1.1689, + "step": 20124 + }, + { + "epoch": 1.69, + "learning_rate": 8.39356371253454e-06, + "loss": 1.1006, + "step": 20125 + }, + { + "epoch": 1.69, + "learning_rate": 8.392666551433535e-06, + "loss": 1.2634, + "step": 20126 + }, + { + "epoch": 1.69, + "learning_rate": 8.391769403612829e-06, + "loss": 1.3248, + "step": 20127 + }, + { + "epoch": 1.69, + "learning_rate": 8.390872269079844e-06, + "loss": 1.2171, + "step": 20128 + }, + { + "epoch": 1.69, + "learning_rate": 8.389975147841984e-06, + "loss": 1.2402, + "step": 20129 + }, + { + "epoch": 1.69, + "learning_rate": 8.389078039906667e-06, + "loss": 1.2426, + "step": 20130 + }, + { + "epoch": 1.69, + "learning_rate": 8.388180945281297e-06, + "loss": 1.2471, + "step": 20131 + }, + { + "epoch": 1.69, + "learning_rate": 8.387283863973298e-06, + "loss": 1.159, + "step": 20132 + }, + { + "epoch": 1.7, + "learning_rate": 8.386386795990074e-06, + "loss": 1.048, + "step": 20133 + }, + { + "epoch": 1.7, + "learning_rate": 8.38548974133904e-06, + "loss": 1.1836, + "step": 20134 + }, + { + "epoch": 1.7, + "learning_rate": 8.384592700027603e-06, + "loss": 1.0962, + "step": 20135 + }, + { + "epoch": 1.7, + "learning_rate": 8.383695672063181e-06, + "loss": 1.2852, + "step": 20136 + }, + { + "epoch": 1.7, + "learning_rate": 8.382798657453182e-06, + "loss": 1.195, + "step": 20137 + }, + { + "epoch": 1.7, + "learning_rate": 8.381901656205016e-06, + "loss": 1.1963, + "step": 20138 + }, + { + "epoch": 1.7, + "learning_rate": 8.3810046683261e-06, + "loss": 1.1895, + "step": 20139 + }, + { + "epoch": 1.7, + "learning_rate": 8.380107693823839e-06, + "loss": 1.1429, + "step": 20140 + }, + { + "epoch": 1.7, + "learning_rate": 8.37921073270565e-06, + "loss": 1.1735, + "step": 20141 + }, + { + "epoch": 1.7, + "learning_rate": 8.378313784978934e-06, + "loss": 1.1982, + "step": 20142 + }, + { + "epoch": 1.7, + "learning_rate": 8.377416850651114e-06, + "loss": 1.1621, + "step": 20143 + }, + { + "epoch": 1.7, + "learning_rate": 8.376519929729594e-06, + "loss": 1.1928, + "step": 20144 + }, + { + "epoch": 1.7, + "learning_rate": 8.375623022221789e-06, + "loss": 1.127, + "step": 20145 + }, + { + "epoch": 1.7, + "learning_rate": 8.3747261281351e-06, + "loss": 1.1954, + "step": 20146 + }, + { + "epoch": 1.7, + "learning_rate": 8.373829247476947e-06, + "loss": 1.1752, + "step": 20147 + }, + { + "epoch": 1.7, + "learning_rate": 8.372932380254738e-06, + "loss": 1.1704, + "step": 20148 + }, + { + "epoch": 1.7, + "learning_rate": 8.372035526475884e-06, + "loss": 1.1524, + "step": 20149 + }, + { + "epoch": 1.7, + "learning_rate": 8.371138686147788e-06, + "loss": 1.1417, + "step": 20150 + }, + { + "epoch": 1.7, + "learning_rate": 8.37024185927787e-06, + "loss": 1.1527, + "step": 20151 + }, + { + "epoch": 1.7, + "learning_rate": 8.369345045873534e-06, + "loss": 1.0956, + "step": 20152 + }, + { + "epoch": 1.7, + "learning_rate": 8.368448245942188e-06, + "loss": 1.1484, + "step": 20153 + }, + { + "epoch": 1.7, + "learning_rate": 8.36755145949125e-06, + "loss": 1.2114, + "step": 20154 + }, + { + "epoch": 1.7, + "learning_rate": 8.366654686528122e-06, + "loss": 1.1954, + "step": 20155 + }, + { + "epoch": 1.7, + "learning_rate": 8.365757927060218e-06, + "loss": 1.18, + "step": 20156 + }, + { + "epoch": 1.7, + "learning_rate": 8.364861181094941e-06, + "loss": 1.1831, + "step": 20157 + }, + { + "epoch": 1.7, + "learning_rate": 8.363964448639708e-06, + "loss": 1.2314, + "step": 20158 + }, + { + "epoch": 1.7, + "learning_rate": 8.36306772970192e-06, + "loss": 1.2201, + "step": 20159 + }, + { + "epoch": 1.7, + "learning_rate": 8.362171024288997e-06, + "loss": 1.1876, + "step": 20160 + }, + { + "epoch": 1.7, + "learning_rate": 8.361274332408334e-06, + "loss": 1.2285, + "step": 20161 + }, + { + "epoch": 1.7, + "learning_rate": 8.360377654067349e-06, + "loss": 1.2082, + "step": 20162 + }, + { + "epoch": 1.7, + "learning_rate": 8.35948098927345e-06, + "loss": 1.1946, + "step": 20163 + }, + { + "epoch": 1.7, + "learning_rate": 8.358584338034044e-06, + "loss": 1.1265, + "step": 20164 + }, + { + "epoch": 1.7, + "learning_rate": 8.357687700356542e-06, + "loss": 1.257, + "step": 20165 + }, + { + "epoch": 1.7, + "learning_rate": 8.356791076248342e-06, + "loss": 1.1922, + "step": 20166 + }, + { + "epoch": 1.7, + "learning_rate": 8.355894465716869e-06, + "loss": 1.1968, + "step": 20167 + }, + { + "epoch": 1.7, + "learning_rate": 8.354997868769516e-06, + "loss": 1.1413, + "step": 20168 + }, + { + "epoch": 1.7, + "learning_rate": 8.354101285413702e-06, + "loss": 1.2638, + "step": 20169 + }, + { + "epoch": 1.7, + "learning_rate": 8.353204715656823e-06, + "loss": 1.1573, + "step": 20170 + }, + { + "epoch": 1.7, + "learning_rate": 8.3523081595063e-06, + "loss": 1.2373, + "step": 20171 + }, + { + "epoch": 1.7, + "learning_rate": 8.351411616969532e-06, + "loss": 1.1627, + "step": 20172 + }, + { + "epoch": 1.7, + "learning_rate": 8.35051508805393e-06, + "loss": 1.1764, + "step": 20173 + }, + { + "epoch": 1.7, + "learning_rate": 8.349618572766897e-06, + "loss": 1.1434, + "step": 20174 + }, + { + "epoch": 1.7, + "learning_rate": 8.348722071115846e-06, + "loss": 1.1744, + "step": 20175 + }, + { + "epoch": 1.7, + "learning_rate": 8.347825583108182e-06, + "loss": 1.1855, + "step": 20176 + }, + { + "epoch": 1.7, + "learning_rate": 8.34692910875131e-06, + "loss": 1.1642, + "step": 20177 + }, + { + "epoch": 1.7, + "learning_rate": 8.346032648052641e-06, + "loss": 1.2145, + "step": 20178 + }, + { + "epoch": 1.7, + "learning_rate": 8.345136201019578e-06, + "loss": 1.1591, + "step": 20179 + }, + { + "epoch": 1.7, + "learning_rate": 8.344239767659532e-06, + "loss": 1.1674, + "step": 20180 + }, + { + "epoch": 1.7, + "learning_rate": 8.343343347979903e-06, + "loss": 1.2483, + "step": 20181 + }, + { + "epoch": 1.7, + "learning_rate": 8.342446941988104e-06, + "loss": 1.2065, + "step": 20182 + }, + { + "epoch": 1.7, + "learning_rate": 8.341550549691538e-06, + "loss": 1.234, + "step": 20183 + }, + { + "epoch": 1.7, + "learning_rate": 8.340654171097613e-06, + "loss": 1.2074, + "step": 20184 + }, + { + "epoch": 1.7, + "learning_rate": 8.339757806213732e-06, + "loss": 1.3029, + "step": 20185 + }, + { + "epoch": 1.7, + "learning_rate": 8.338861455047306e-06, + "loss": 1.1992, + "step": 20186 + }, + { + "epoch": 1.7, + "learning_rate": 8.337965117605737e-06, + "loss": 1.0906, + "step": 20187 + }, + { + "epoch": 1.7, + "learning_rate": 8.33706879389643e-06, + "loss": 1.135, + "step": 20188 + }, + { + "epoch": 1.7, + "learning_rate": 8.336172483926797e-06, + "loss": 1.2051, + "step": 20189 + }, + { + "epoch": 1.7, + "learning_rate": 8.335276187704238e-06, + "loss": 1.1246, + "step": 20190 + }, + { + "epoch": 1.7, + "learning_rate": 8.334379905236161e-06, + "loss": 1.1122, + "step": 20191 + }, + { + "epoch": 1.7, + "learning_rate": 8.333483636529965e-06, + "loss": 1.1922, + "step": 20192 + }, + { + "epoch": 1.7, + "learning_rate": 8.332587381593067e-06, + "loss": 1.1374, + "step": 20193 + }, + { + "epoch": 1.7, + "learning_rate": 8.331691140432862e-06, + "loss": 1.2299, + "step": 20194 + }, + { + "epoch": 1.7, + "learning_rate": 8.330794913056762e-06, + "loss": 1.124, + "step": 20195 + }, + { + "epoch": 1.7, + "learning_rate": 8.329898699472163e-06, + "loss": 1.2112, + "step": 20196 + }, + { + "epoch": 1.7, + "learning_rate": 8.329002499686482e-06, + "loss": 1.2141, + "step": 20197 + }, + { + "epoch": 1.7, + "learning_rate": 8.328106313707113e-06, + "loss": 1.1072, + "step": 20198 + }, + { + "epoch": 1.7, + "learning_rate": 8.327210141541468e-06, + "loss": 1.1155, + "step": 20199 + }, + { + "epoch": 1.7, + "learning_rate": 8.326313983196943e-06, + "loss": 1.111, + "step": 20200 + }, + { + "epoch": 1.7, + "learning_rate": 8.32541783868095e-06, + "loss": 1.1759, + "step": 20201 + }, + { + "epoch": 1.7, + "learning_rate": 8.324521708000892e-06, + "loss": 1.0858, + "step": 20202 + }, + { + "epoch": 1.7, + "learning_rate": 8.323625591164168e-06, + "loss": 1.1398, + "step": 20203 + }, + { + "epoch": 1.7, + "learning_rate": 8.322729488178191e-06, + "loss": 1.2276, + "step": 20204 + }, + { + "epoch": 1.7, + "learning_rate": 8.321833399050358e-06, + "loss": 1.1757, + "step": 20205 + }, + { + "epoch": 1.7, + "learning_rate": 8.320937323788076e-06, + "loss": 1.1328, + "step": 20206 + }, + { + "epoch": 1.7, + "learning_rate": 8.320041262398742e-06, + "loss": 1.1047, + "step": 20207 + }, + { + "epoch": 1.7, + "learning_rate": 8.31914521488977e-06, + "loss": 1.1966, + "step": 20208 + }, + { + "epoch": 1.7, + "learning_rate": 8.318249181268556e-06, + "loss": 1.1617, + "step": 20209 + }, + { + "epoch": 1.7, + "learning_rate": 8.317353161542508e-06, + "loss": 1.2195, + "step": 20210 + }, + { + "epoch": 1.7, + "learning_rate": 8.316457155719021e-06, + "loss": 1.1567, + "step": 20211 + }, + { + "epoch": 1.7, + "learning_rate": 8.315561163805508e-06, + "loss": 1.1259, + "step": 20212 + }, + { + "epoch": 1.7, + "learning_rate": 8.314665185809368e-06, + "loss": 1.1259, + "step": 20213 + }, + { + "epoch": 1.7, + "learning_rate": 8.313769221738001e-06, + "loss": 1.1302, + "step": 20214 + }, + { + "epoch": 1.7, + "learning_rate": 8.312873271598818e-06, + "loss": 1.1194, + "step": 20215 + }, + { + "epoch": 1.7, + "learning_rate": 8.311977335399209e-06, + "loss": 1.2609, + "step": 20216 + }, + { + "epoch": 1.7, + "learning_rate": 8.311081413146589e-06, + "loss": 1.1086, + "step": 20217 + }, + { + "epoch": 1.7, + "learning_rate": 8.310185504848351e-06, + "loss": 1.1534, + "step": 20218 + }, + { + "epoch": 1.7, + "learning_rate": 8.309289610511904e-06, + "loss": 1.2007, + "step": 20219 + }, + { + "epoch": 1.7, + "learning_rate": 8.308393730144644e-06, + "loss": 1.2449, + "step": 20220 + }, + { + "epoch": 1.7, + "learning_rate": 8.307497863753979e-06, + "loss": 1.1538, + "step": 20221 + }, + { + "epoch": 1.7, + "learning_rate": 8.306602011347307e-06, + "loss": 1.1733, + "step": 20222 + }, + { + "epoch": 1.7, + "learning_rate": 8.305706172932034e-06, + "loss": 1.2217, + "step": 20223 + }, + { + "epoch": 1.7, + "learning_rate": 8.304810348515555e-06, + "loss": 1.2032, + "step": 20224 + }, + { + "epoch": 1.7, + "learning_rate": 8.303914538105276e-06, + "loss": 1.2036, + "step": 20225 + }, + { + "epoch": 1.7, + "learning_rate": 8.3030187417086e-06, + "loss": 1.1922, + "step": 20226 + }, + { + "epoch": 1.7, + "learning_rate": 8.302122959332923e-06, + "loss": 1.2493, + "step": 20227 + }, + { + "epoch": 1.7, + "learning_rate": 8.301227190985653e-06, + "loss": 1.2699, + "step": 20228 + }, + { + "epoch": 1.7, + "learning_rate": 8.300331436674186e-06, + "loss": 1.2015, + "step": 20229 + }, + { + "epoch": 1.7, + "learning_rate": 8.299435696405927e-06, + "loss": 1.1604, + "step": 20230 + }, + { + "epoch": 1.7, + "learning_rate": 8.29853997018827e-06, + "loss": 1.1487, + "step": 20231 + }, + { + "epoch": 1.7, + "learning_rate": 8.297644258028625e-06, + "loss": 1.1797, + "step": 20232 + }, + { + "epoch": 1.7, + "learning_rate": 8.296748559934385e-06, + "loss": 1.1625, + "step": 20233 + }, + { + "epoch": 1.7, + "learning_rate": 8.295852875912956e-06, + "loss": 1.1544, + "step": 20234 + }, + { + "epoch": 1.7, + "learning_rate": 8.294957205971731e-06, + "loss": 1.1328, + "step": 20235 + }, + { + "epoch": 1.7, + "learning_rate": 8.29406155011812e-06, + "loss": 1.1703, + "step": 20236 + }, + { + "epoch": 1.7, + "learning_rate": 8.293165908359516e-06, + "loss": 1.2313, + "step": 20237 + }, + { + "epoch": 1.7, + "learning_rate": 8.29227028070332e-06, + "loss": 1.1943, + "step": 20238 + }, + { + "epoch": 1.7, + "learning_rate": 8.291374667156938e-06, + "loss": 1.2413, + "step": 20239 + }, + { + "epoch": 1.7, + "learning_rate": 8.290479067727762e-06, + "loss": 1.2233, + "step": 20240 + }, + { + "epoch": 1.7, + "learning_rate": 8.289583482423198e-06, + "loss": 1.1975, + "step": 20241 + }, + { + "epoch": 1.7, + "learning_rate": 8.288687911250639e-06, + "loss": 1.1679, + "step": 20242 + }, + { + "epoch": 1.7, + "learning_rate": 8.287792354217491e-06, + "loss": 1.2559, + "step": 20243 + }, + { + "epoch": 1.7, + "learning_rate": 8.28689681133115e-06, + "loss": 1.2315, + "step": 20244 + }, + { + "epoch": 1.7, + "learning_rate": 8.286001282599017e-06, + "loss": 1.2064, + "step": 20245 + }, + { + "epoch": 1.7, + "learning_rate": 8.285105768028485e-06, + "loss": 1.1372, + "step": 20246 + }, + { + "epoch": 1.7, + "learning_rate": 8.284210267626965e-06, + "loss": 1.1725, + "step": 20247 + }, + { + "epoch": 1.7, + "learning_rate": 8.283314781401844e-06, + "loss": 1.2132, + "step": 20248 + }, + { + "epoch": 1.7, + "learning_rate": 8.28241930936053e-06, + "loss": 1.1411, + "step": 20249 + }, + { + "epoch": 1.7, + "learning_rate": 8.28152385151041e-06, + "loss": 1.1838, + "step": 20250 + }, + { + "epoch": 1.7, + "learning_rate": 8.280628407858893e-06, + "loss": 1.0514, + "step": 20251 + }, + { + "epoch": 1.71, + "learning_rate": 8.279732978413379e-06, + "loss": 1.2198, + "step": 20252 + }, + { + "epoch": 1.71, + "learning_rate": 8.278837563181255e-06, + "loss": 1.1735, + "step": 20253 + }, + { + "epoch": 1.71, + "learning_rate": 8.27794216216993e-06, + "loss": 1.1998, + "step": 20254 + }, + { + "epoch": 1.71, + "learning_rate": 8.277046775386797e-06, + "loss": 1.1803, + "step": 20255 + }, + { + "epoch": 1.71, + "learning_rate": 8.276151402839257e-06, + "loss": 1.1517, + "step": 20256 + }, + { + "epoch": 1.71, + "learning_rate": 8.275256044534702e-06, + "loss": 1.2553, + "step": 20257 + }, + { + "epoch": 1.71, + "learning_rate": 8.274360700480537e-06, + "loss": 1.1998, + "step": 20258 + }, + { + "epoch": 1.71, + "learning_rate": 8.273465370684154e-06, + "loss": 1.1565, + "step": 20259 + }, + { + "epoch": 1.71, + "learning_rate": 8.272570055152955e-06, + "loss": 1.2809, + "step": 20260 + }, + { + "epoch": 1.71, + "learning_rate": 8.27167475389433e-06, + "loss": 1.1259, + "step": 20261 + }, + { + "epoch": 1.71, + "learning_rate": 8.270779466915685e-06, + "loss": 1.2051, + "step": 20262 + }, + { + "epoch": 1.71, + "learning_rate": 8.269884194224416e-06, + "loss": 1.2661, + "step": 20263 + }, + { + "epoch": 1.71, + "learning_rate": 8.268988935827911e-06, + "loss": 1.2971, + "step": 20264 + }, + { + "epoch": 1.71, + "learning_rate": 8.26809369173358e-06, + "loss": 1.1957, + "step": 20265 + }, + { + "epoch": 1.71, + "learning_rate": 8.267198461948809e-06, + "loss": 1.1738, + "step": 20266 + }, + { + "epoch": 1.71, + "learning_rate": 8.266303246481003e-06, + "loss": 1.164, + "step": 20267 + }, + { + "epoch": 1.71, + "learning_rate": 8.26540804533755e-06, + "loss": 1.2142, + "step": 20268 + }, + { + "epoch": 1.71, + "learning_rate": 8.264512858525854e-06, + "loss": 1.1596, + "step": 20269 + }, + { + "epoch": 1.71, + "learning_rate": 8.263617686053305e-06, + "loss": 1.3109, + "step": 20270 + }, + { + "epoch": 1.71, + "learning_rate": 8.262722527927304e-06, + "loss": 1.1634, + "step": 20271 + }, + { + "epoch": 1.71, + "learning_rate": 8.261827384155246e-06, + "loss": 1.1932, + "step": 20272 + }, + { + "epoch": 1.71, + "learning_rate": 8.260932254744527e-06, + "loss": 1.1077, + "step": 20273 + }, + { + "epoch": 1.71, + "learning_rate": 8.26003713970254e-06, + "loss": 1.2353, + "step": 20274 + }, + { + "epoch": 1.71, + "learning_rate": 8.259142039036682e-06, + "loss": 1.2794, + "step": 20275 + }, + { + "epoch": 1.71, + "learning_rate": 8.258246952754354e-06, + "loss": 1.2867, + "step": 20276 + }, + { + "epoch": 1.71, + "learning_rate": 8.257351880862943e-06, + "loss": 1.2106, + "step": 20277 + }, + { + "epoch": 1.71, + "learning_rate": 8.256456823369851e-06, + "loss": 1.1457, + "step": 20278 + }, + { + "epoch": 1.71, + "learning_rate": 8.255561780282471e-06, + "loss": 1.2373, + "step": 20279 + }, + { + "epoch": 1.71, + "learning_rate": 8.254666751608199e-06, + "loss": 1.1677, + "step": 20280 + }, + { + "epoch": 1.71, + "learning_rate": 8.253771737354423e-06, + "loss": 1.1104, + "step": 20281 + }, + { + "epoch": 1.71, + "learning_rate": 8.25287673752855e-06, + "loss": 1.1451, + "step": 20282 + }, + { + "epoch": 1.71, + "learning_rate": 8.251981752137965e-06, + "loss": 1.1958, + "step": 20283 + }, + { + "epoch": 1.71, + "learning_rate": 8.251086781190069e-06, + "loss": 1.2157, + "step": 20284 + }, + { + "epoch": 1.71, + "learning_rate": 8.250191824692251e-06, + "loss": 1.246, + "step": 20285 + }, + { + "epoch": 1.71, + "learning_rate": 8.24929688265191e-06, + "loss": 1.1964, + "step": 20286 + }, + { + "epoch": 1.71, + "learning_rate": 8.248401955076438e-06, + "loss": 1.1735, + "step": 20287 + }, + { + "epoch": 1.71, + "learning_rate": 8.247507041973226e-06, + "loss": 1.2679, + "step": 20288 + }, + { + "epoch": 1.71, + "learning_rate": 8.246612143349677e-06, + "loss": 1.0842, + "step": 20289 + }, + { + "epoch": 1.71, + "learning_rate": 8.245717259213178e-06, + "loss": 1.2937, + "step": 20290 + }, + { + "epoch": 1.71, + "learning_rate": 8.244822389571125e-06, + "loss": 1.1729, + "step": 20291 + }, + { + "epoch": 1.71, + "learning_rate": 8.24392753443091e-06, + "loss": 1.2885, + "step": 20292 + }, + { + "epoch": 1.71, + "learning_rate": 8.243032693799928e-06, + "loss": 1.1421, + "step": 20293 + }, + { + "epoch": 1.71, + "learning_rate": 8.242137867685572e-06, + "loss": 1.2751, + "step": 20294 + }, + { + "epoch": 1.71, + "learning_rate": 8.241243056095237e-06, + "loss": 1.1379, + "step": 20295 + }, + { + "epoch": 1.71, + "learning_rate": 8.240348259036312e-06, + "loss": 1.2824, + "step": 20296 + }, + { + "epoch": 1.71, + "learning_rate": 8.239453476516195e-06, + "loss": 1.1814, + "step": 20297 + }, + { + "epoch": 1.71, + "learning_rate": 8.238558708542275e-06, + "loss": 1.1176, + "step": 20298 + }, + { + "epoch": 1.71, + "learning_rate": 8.23766395512195e-06, + "loss": 1.1263, + "step": 20299 + }, + { + "epoch": 1.71, + "learning_rate": 8.236769216262606e-06, + "loss": 1.1297, + "step": 20300 + }, + { + "epoch": 1.71, + "learning_rate": 8.23587449197164e-06, + "loss": 1.1608, + "step": 20301 + }, + { + "epoch": 1.71, + "learning_rate": 8.234979782256444e-06, + "loss": 1.1224, + "step": 20302 + }, + { + "epoch": 1.71, + "learning_rate": 8.234085087124408e-06, + "loss": 1.2383, + "step": 20303 + }, + { + "epoch": 1.71, + "learning_rate": 8.23319040658293e-06, + "loss": 1.1845, + "step": 20304 + }, + { + "epoch": 1.71, + "learning_rate": 8.232295740639396e-06, + "loss": 1.2141, + "step": 20305 + }, + { + "epoch": 1.71, + "learning_rate": 8.231401089301201e-06, + "loss": 1.2213, + "step": 20306 + }, + { + "epoch": 1.71, + "learning_rate": 8.230506452575735e-06, + "loss": 1.2142, + "step": 20307 + }, + { + "epoch": 1.71, + "learning_rate": 8.229611830470392e-06, + "loss": 1.1665, + "step": 20308 + }, + { + "epoch": 1.71, + "learning_rate": 8.228717222992563e-06, + "loss": 1.1724, + "step": 20309 + }, + { + "epoch": 1.71, + "learning_rate": 8.22782263014964e-06, + "loss": 1.216, + "step": 20310 + }, + { + "epoch": 1.71, + "learning_rate": 8.22692805194901e-06, + "loss": 1.187, + "step": 20311 + }, + { + "epoch": 1.71, + "learning_rate": 8.226033488398068e-06, + "loss": 1.1935, + "step": 20312 + }, + { + "epoch": 1.71, + "learning_rate": 8.225138939504211e-06, + "loss": 1.196, + "step": 20313 + }, + { + "epoch": 1.71, + "learning_rate": 8.224244405274816e-06, + "loss": 1.1763, + "step": 20314 + }, + { + "epoch": 1.71, + "learning_rate": 8.22334988571729e-06, + "loss": 1.2344, + "step": 20315 + }, + { + "epoch": 1.71, + "learning_rate": 8.22245538083901e-06, + "loss": 1.2283, + "step": 20316 + }, + { + "epoch": 1.71, + "learning_rate": 8.221560890647378e-06, + "loss": 1.1998, + "step": 20317 + }, + { + "epoch": 1.71, + "learning_rate": 8.220666415149775e-06, + "loss": 1.1929, + "step": 20318 + }, + { + "epoch": 1.71, + "learning_rate": 8.219771954353597e-06, + "loss": 1.1196, + "step": 20319 + }, + { + "epoch": 1.71, + "learning_rate": 8.21887750826623e-06, + "loss": 1.1767, + "step": 20320 + }, + { + "epoch": 1.71, + "learning_rate": 8.21798307689507e-06, + "loss": 1.2262, + "step": 20321 + }, + { + "epoch": 1.71, + "learning_rate": 8.217088660247503e-06, + "loss": 1.207, + "step": 20322 + }, + { + "epoch": 1.71, + "learning_rate": 8.216194258330923e-06, + "loss": 1.2102, + "step": 20323 + }, + { + "epoch": 1.71, + "learning_rate": 8.215299871152712e-06, + "loss": 1.1996, + "step": 20324 + }, + { + "epoch": 1.71, + "learning_rate": 8.214405498720268e-06, + "loss": 1.1589, + "step": 20325 + }, + { + "epoch": 1.71, + "learning_rate": 8.213511141040978e-06, + "loss": 1.2313, + "step": 20326 + }, + { + "epoch": 1.71, + "learning_rate": 8.212616798122224e-06, + "loss": 1.2003, + "step": 20327 + }, + { + "epoch": 1.71, + "learning_rate": 8.211722469971411e-06, + "loss": 1.2372, + "step": 20328 + }, + { + "epoch": 1.71, + "learning_rate": 8.210828156595913e-06, + "loss": 1.246, + "step": 20329 + }, + { + "epoch": 1.71, + "learning_rate": 8.209933858003131e-06, + "loss": 1.2938, + "step": 20330 + }, + { + "epoch": 1.71, + "learning_rate": 8.209039574200442e-06, + "loss": 1.0946, + "step": 20331 + }, + { + "epoch": 1.71, + "learning_rate": 8.208145305195246e-06, + "loss": 1.1762, + "step": 20332 + }, + { + "epoch": 1.71, + "learning_rate": 8.207251050994925e-06, + "loss": 1.1797, + "step": 20333 + }, + { + "epoch": 1.71, + "learning_rate": 8.206356811606871e-06, + "loss": 1.2221, + "step": 20334 + }, + { + "epoch": 1.71, + "learning_rate": 8.205462587038467e-06, + "loss": 1.2944, + "step": 20335 + }, + { + "epoch": 1.71, + "learning_rate": 8.204568377297111e-06, + "loss": 1.1754, + "step": 20336 + }, + { + "epoch": 1.71, + "learning_rate": 8.203674182390183e-06, + "loss": 1.1311, + "step": 20337 + }, + { + "epoch": 1.71, + "learning_rate": 8.20278000232507e-06, + "loss": 1.1489, + "step": 20338 + }, + { + "epoch": 1.71, + "learning_rate": 8.20188583710917e-06, + "loss": 1.1772, + "step": 20339 + }, + { + "epoch": 1.71, + "learning_rate": 8.200991686749862e-06, + "loss": 1.1383, + "step": 20340 + }, + { + "epoch": 1.71, + "learning_rate": 8.200097551254537e-06, + "loss": 1.057, + "step": 20341 + }, + { + "epoch": 1.71, + "learning_rate": 8.19920343063058e-06, + "loss": 1.2042, + "step": 20342 + }, + { + "epoch": 1.71, + "learning_rate": 8.198309324885384e-06, + "loss": 1.2238, + "step": 20343 + }, + { + "epoch": 1.71, + "learning_rate": 8.197415234026331e-06, + "loss": 1.2211, + "step": 20344 + }, + { + "epoch": 1.71, + "learning_rate": 8.196521158060813e-06, + "loss": 1.1699, + "step": 20345 + }, + { + "epoch": 1.71, + "learning_rate": 8.195627096996208e-06, + "loss": 1.1514, + "step": 20346 + }, + { + "epoch": 1.71, + "learning_rate": 8.194733050839915e-06, + "loss": 1.143, + "step": 20347 + }, + { + "epoch": 1.71, + "learning_rate": 8.193839019599315e-06, + "loss": 1.182, + "step": 20348 + }, + { + "epoch": 1.71, + "learning_rate": 8.192945003281794e-06, + "loss": 1.2336, + "step": 20349 + }, + { + "epoch": 1.71, + "learning_rate": 8.192051001894738e-06, + "loss": 1.1973, + "step": 20350 + }, + { + "epoch": 1.71, + "learning_rate": 8.191157015445535e-06, + "loss": 1.1271, + "step": 20351 + }, + { + "epoch": 1.71, + "learning_rate": 8.190263043941575e-06, + "loss": 1.1983, + "step": 20352 + }, + { + "epoch": 1.71, + "learning_rate": 8.189369087390237e-06, + "loss": 1.1579, + "step": 20353 + }, + { + "epoch": 1.71, + "learning_rate": 8.188475145798915e-06, + "loss": 1.2222, + "step": 20354 + }, + { + "epoch": 1.71, + "learning_rate": 8.187581219174988e-06, + "loss": 1.2422, + "step": 20355 + }, + { + "epoch": 1.71, + "learning_rate": 8.186687307525849e-06, + "loss": 1.2229, + "step": 20356 + }, + { + "epoch": 1.71, + "learning_rate": 8.185793410858875e-06, + "loss": 1.1727, + "step": 20357 + }, + { + "epoch": 1.71, + "learning_rate": 8.18489952918146e-06, + "loss": 1.1485, + "step": 20358 + }, + { + "epoch": 1.71, + "learning_rate": 8.184005662500984e-06, + "loss": 1.0735, + "step": 20359 + }, + { + "epoch": 1.71, + "learning_rate": 8.183111810824838e-06, + "loss": 1.1319, + "step": 20360 + }, + { + "epoch": 1.71, + "learning_rate": 8.182217974160398e-06, + "loss": 1.2441, + "step": 20361 + }, + { + "epoch": 1.71, + "learning_rate": 8.181324152515056e-06, + "loss": 1.195, + "step": 20362 + }, + { + "epoch": 1.71, + "learning_rate": 8.1804303458962e-06, + "loss": 1.2074, + "step": 20363 + }, + { + "epoch": 1.71, + "learning_rate": 8.179536554311205e-06, + "loss": 1.2185, + "step": 20364 + }, + { + "epoch": 1.71, + "learning_rate": 8.178642777767468e-06, + "loss": 1.1037, + "step": 20365 + }, + { + "epoch": 1.71, + "learning_rate": 8.177749016272364e-06, + "loss": 1.1749, + "step": 20366 + }, + { + "epoch": 1.71, + "learning_rate": 8.176855269833282e-06, + "loss": 1.1468, + "step": 20367 + }, + { + "epoch": 1.71, + "learning_rate": 8.175961538457602e-06, + "loss": 1.167, + "step": 20368 + }, + { + "epoch": 1.71, + "learning_rate": 8.175067822152719e-06, + "loss": 1.1104, + "step": 20369 + }, + { + "epoch": 1.71, + "learning_rate": 8.174174120926e-06, + "loss": 1.1888, + "step": 20370 + }, + { + "epoch": 1.72, + "learning_rate": 8.173280434784846e-06, + "loss": 1.182, + "step": 20371 + }, + { + "epoch": 1.72, + "learning_rate": 8.17238676373663e-06, + "loss": 1.0963, + "step": 20372 + }, + { + "epoch": 1.72, + "learning_rate": 8.171493107788743e-06, + "loss": 1.1476, + "step": 20373 + }, + { + "epoch": 1.72, + "learning_rate": 8.17059946694856e-06, + "loss": 1.1106, + "step": 20374 + }, + { + "epoch": 1.72, + "learning_rate": 8.169705841223471e-06, + "loss": 1.1428, + "step": 20375 + }, + { + "epoch": 1.72, + "learning_rate": 8.168812230620861e-06, + "loss": 1.2067, + "step": 20376 + }, + { + "epoch": 1.72, + "learning_rate": 8.167918635148107e-06, + "loss": 1.2576, + "step": 20377 + }, + { + "epoch": 1.72, + "learning_rate": 8.1670250548126e-06, + "loss": 1.1012, + "step": 20378 + }, + { + "epoch": 1.72, + "learning_rate": 8.166131489621714e-06, + "loss": 1.1393, + "step": 20379 + }, + { + "epoch": 1.72, + "learning_rate": 8.16523793958284e-06, + "loss": 1.2006, + "step": 20380 + }, + { + "epoch": 1.72, + "learning_rate": 8.164344404703355e-06, + "loss": 1.0859, + "step": 20381 + }, + { + "epoch": 1.72, + "learning_rate": 8.163450884990646e-06, + "loss": 1.2081, + "step": 20382 + }, + { + "epoch": 1.72, + "learning_rate": 8.162557380452092e-06, + "loss": 1.1209, + "step": 20383 + }, + { + "epoch": 1.72, + "learning_rate": 8.16166389109508e-06, + "loss": 1.2065, + "step": 20384 + }, + { + "epoch": 1.72, + "learning_rate": 8.160770416926984e-06, + "loss": 1.1632, + "step": 20385 + }, + { + "epoch": 1.72, + "learning_rate": 8.159876957955195e-06, + "loss": 1.0926, + "step": 20386 + }, + { + "epoch": 1.72, + "learning_rate": 8.158983514187091e-06, + "loss": 1.2044, + "step": 20387 + }, + { + "epoch": 1.72, + "learning_rate": 8.15809008563005e-06, + "loss": 1.2054, + "step": 20388 + }, + { + "epoch": 1.72, + "learning_rate": 8.157196672291466e-06, + "loss": 1.1971, + "step": 20389 + }, + { + "epoch": 1.72, + "learning_rate": 8.156303274178707e-06, + "loss": 1.2724, + "step": 20390 + }, + { + "epoch": 1.72, + "learning_rate": 8.155409891299165e-06, + "loss": 1.2622, + "step": 20391 + }, + { + "epoch": 1.72, + "learning_rate": 8.154516523660212e-06, + "loss": 1.2216, + "step": 20392 + }, + { + "epoch": 1.72, + "learning_rate": 8.153623171269238e-06, + "loss": 1.1731, + "step": 20393 + }, + { + "epoch": 1.72, + "learning_rate": 8.152729834133618e-06, + "loss": 1.0971, + "step": 20394 + }, + { + "epoch": 1.72, + "learning_rate": 8.151836512260738e-06, + "loss": 1.2852, + "step": 20395 + }, + { + "epoch": 1.72, + "learning_rate": 8.150943205657971e-06, + "loss": 1.2537, + "step": 20396 + }, + { + "epoch": 1.72, + "learning_rate": 8.150049914332709e-06, + "loss": 1.1991, + "step": 20397 + }, + { + "epoch": 1.72, + "learning_rate": 8.149156638292324e-06, + "loss": 1.1499, + "step": 20398 + }, + { + "epoch": 1.72, + "learning_rate": 8.1482633775442e-06, + "loss": 1.1696, + "step": 20399 + }, + { + "epoch": 1.72, + "learning_rate": 8.147370132095714e-06, + "loss": 1.1923, + "step": 20400 + }, + { + "epoch": 1.72, + "learning_rate": 8.14647690195425e-06, + "loss": 1.1414, + "step": 20401 + }, + { + "epoch": 1.72, + "learning_rate": 8.145583687127188e-06, + "loss": 1.1375, + "step": 20402 + }, + { + "epoch": 1.72, + "learning_rate": 8.144690487621905e-06, + "loss": 1.0385, + "step": 20403 + }, + { + "epoch": 1.72, + "learning_rate": 8.143797303445786e-06, + "loss": 1.2281, + "step": 20404 + }, + { + "epoch": 1.72, + "learning_rate": 8.142904134606206e-06, + "loss": 1.2362, + "step": 20405 + }, + { + "epoch": 1.72, + "learning_rate": 8.14201098111055e-06, + "loss": 1.2104, + "step": 20406 + }, + { + "epoch": 1.72, + "learning_rate": 8.141117842966187e-06, + "loss": 1.1958, + "step": 20407 + }, + { + "epoch": 1.72, + "learning_rate": 8.14022472018051e-06, + "loss": 1.2152, + "step": 20408 + }, + { + "epoch": 1.72, + "learning_rate": 8.139331612760888e-06, + "loss": 1.248, + "step": 20409 + }, + { + "epoch": 1.72, + "learning_rate": 8.138438520714706e-06, + "loss": 1.2317, + "step": 20410 + }, + { + "epoch": 1.72, + "learning_rate": 8.137545444049336e-06, + "loss": 1.1429, + "step": 20411 + }, + { + "epoch": 1.72, + "learning_rate": 8.136652382772164e-06, + "loss": 1.2225, + "step": 20412 + }, + { + "epoch": 1.72, + "learning_rate": 8.13575933689057e-06, + "loss": 1.2039, + "step": 20413 + }, + { + "epoch": 1.72, + "learning_rate": 8.134866306411923e-06, + "loss": 1.2696, + "step": 20414 + }, + { + "epoch": 1.72, + "learning_rate": 8.133973291343613e-06, + "loss": 1.1418, + "step": 20415 + }, + { + "epoch": 1.72, + "learning_rate": 8.13308029169301e-06, + "loss": 1.1403, + "step": 20416 + }, + { + "epoch": 1.72, + "learning_rate": 8.132187307467496e-06, + "loss": 1.2277, + "step": 20417 + }, + { + "epoch": 1.72, + "learning_rate": 8.131294338674445e-06, + "loss": 1.0659, + "step": 20418 + }, + { + "epoch": 1.72, + "learning_rate": 8.130401385321244e-06, + "loss": 1.1408, + "step": 20419 + }, + { + "epoch": 1.72, + "learning_rate": 8.12950844741526e-06, + "loss": 1.1511, + "step": 20420 + }, + { + "epoch": 1.72, + "learning_rate": 8.12861552496388e-06, + "loss": 1.1533, + "step": 20421 + }, + { + "epoch": 1.72, + "learning_rate": 8.127722617974474e-06, + "loss": 1.2266, + "step": 20422 + }, + { + "epoch": 1.72, + "learning_rate": 8.126829726454427e-06, + "loss": 1.2276, + "step": 20423 + }, + { + "epoch": 1.72, + "learning_rate": 8.125936850411107e-06, + "loss": 1.1474, + "step": 20424 + }, + { + "epoch": 1.72, + "learning_rate": 8.125043989851898e-06, + "loss": 1.1685, + "step": 20425 + }, + { + "epoch": 1.72, + "learning_rate": 8.124151144784179e-06, + "loss": 1.2242, + "step": 20426 + }, + { + "epoch": 1.72, + "learning_rate": 8.12325831521532e-06, + "loss": 1.0641, + "step": 20427 + }, + { + "epoch": 1.72, + "learning_rate": 8.122365501152704e-06, + "loss": 1.2065, + "step": 20428 + }, + { + "epoch": 1.72, + "learning_rate": 8.121472702603703e-06, + "loss": 1.1533, + "step": 20429 + }, + { + "epoch": 1.72, + "learning_rate": 8.120579919575699e-06, + "loss": 1.242, + "step": 20430 + }, + { + "epoch": 1.72, + "learning_rate": 8.119687152076059e-06, + "loss": 1.1658, + "step": 20431 + }, + { + "epoch": 1.72, + "learning_rate": 8.118794400112172e-06, + "loss": 1.1361, + "step": 20432 + }, + { + "epoch": 1.72, + "learning_rate": 8.117901663691404e-06, + "loss": 1.0939, + "step": 20433 + }, + { + "epoch": 1.72, + "learning_rate": 8.117008942821139e-06, + "loss": 1.1355, + "step": 20434 + }, + { + "epoch": 1.72, + "learning_rate": 8.116116237508742e-06, + "loss": 1.1703, + "step": 20435 + }, + { + "epoch": 1.72, + "learning_rate": 8.115223547761602e-06, + "loss": 1.1684, + "step": 20436 + }, + { + "epoch": 1.72, + "learning_rate": 8.114330873587085e-06, + "loss": 1.1409, + "step": 20437 + }, + { + "epoch": 1.72, + "learning_rate": 8.113438214992566e-06, + "loss": 1.2505, + "step": 20438 + }, + { + "epoch": 1.72, + "learning_rate": 8.11254557198543e-06, + "loss": 1.1736, + "step": 20439 + }, + { + "epoch": 1.72, + "learning_rate": 8.111652944573046e-06, + "loss": 1.1476, + "step": 20440 + }, + { + "epoch": 1.72, + "learning_rate": 8.11076033276279e-06, + "loss": 1.256, + "step": 20441 + }, + { + "epoch": 1.72, + "learning_rate": 8.109867736562034e-06, + "loss": 1.2347, + "step": 20442 + }, + { + "epoch": 1.72, + "learning_rate": 8.108975155978157e-06, + "loss": 1.1315, + "step": 20443 + }, + { + "epoch": 1.72, + "learning_rate": 8.108082591018533e-06, + "loss": 1.212, + "step": 20444 + }, + { + "epoch": 1.72, + "learning_rate": 8.107190041690538e-06, + "loss": 1.1471, + "step": 20445 + }, + { + "epoch": 1.72, + "learning_rate": 8.10629750800154e-06, + "loss": 1.1252, + "step": 20446 + }, + { + "epoch": 1.72, + "learning_rate": 8.105404989958923e-06, + "loss": 1.2302, + "step": 20447 + }, + { + "epoch": 1.72, + "learning_rate": 8.104512487570053e-06, + "loss": 1.1942, + "step": 20448 + }, + { + "epoch": 1.72, + "learning_rate": 8.10362000084231e-06, + "loss": 1.1347, + "step": 20449 + }, + { + "epoch": 1.72, + "learning_rate": 8.10272752978306e-06, + "loss": 1.2222, + "step": 20450 + }, + { + "epoch": 1.72, + "learning_rate": 8.101835074399686e-06, + "loss": 1.1726, + "step": 20451 + }, + { + "epoch": 1.72, + "learning_rate": 8.10094263469956e-06, + "loss": 1.1753, + "step": 20452 + }, + { + "epoch": 1.72, + "learning_rate": 8.100050210690049e-06, + "loss": 1.1517, + "step": 20453 + }, + { + "epoch": 1.72, + "learning_rate": 8.099157802378537e-06, + "loss": 1.1761, + "step": 20454 + }, + { + "epoch": 1.72, + "learning_rate": 8.098265409772388e-06, + "loss": 1.176, + "step": 20455 + }, + { + "epoch": 1.72, + "learning_rate": 8.09737303287898e-06, + "loss": 1.1227, + "step": 20456 + }, + { + "epoch": 1.72, + "learning_rate": 8.096480671705684e-06, + "loss": 1.1466, + "step": 20457 + }, + { + "epoch": 1.72, + "learning_rate": 8.095588326259876e-06, + "loss": 1.182, + "step": 20458 + }, + { + "epoch": 1.72, + "learning_rate": 8.094695996548925e-06, + "loss": 1.1237, + "step": 20459 + }, + { + "epoch": 1.72, + "learning_rate": 8.093803682580206e-06, + "loss": 1.2113, + "step": 20460 + }, + { + "epoch": 1.72, + "learning_rate": 8.092911384361088e-06, + "loss": 1.2094, + "step": 20461 + }, + { + "epoch": 1.72, + "learning_rate": 8.09201910189895e-06, + "loss": 1.1765, + "step": 20462 + }, + { + "epoch": 1.72, + "learning_rate": 8.091126835201161e-06, + "loss": 1.2379, + "step": 20463 + }, + { + "epoch": 1.72, + "learning_rate": 8.09023458427509e-06, + "loss": 1.1685, + "step": 20464 + }, + { + "epoch": 1.72, + "learning_rate": 8.089342349128115e-06, + "loss": 1.2184, + "step": 20465 + }, + { + "epoch": 1.72, + "learning_rate": 8.088450129767605e-06, + "loss": 1.2989, + "step": 20466 + }, + { + "epoch": 1.72, + "learning_rate": 8.08755792620093e-06, + "loss": 1.1409, + "step": 20467 + }, + { + "epoch": 1.72, + "learning_rate": 8.086665738435462e-06, + "loss": 1.1995, + "step": 20468 + }, + { + "epoch": 1.72, + "learning_rate": 8.085773566478578e-06, + "loss": 1.0878, + "step": 20469 + }, + { + "epoch": 1.72, + "learning_rate": 8.084881410337645e-06, + "loss": 1.1616, + "step": 20470 + }, + { + "epoch": 1.72, + "learning_rate": 8.083989270020033e-06, + "loss": 1.1062, + "step": 20471 + }, + { + "epoch": 1.72, + "learning_rate": 8.083097145533116e-06, + "loss": 1.18, + "step": 20472 + }, + { + "epoch": 1.72, + "learning_rate": 8.082205036884265e-06, + "loss": 1.1731, + "step": 20473 + }, + { + "epoch": 1.72, + "learning_rate": 8.081312944080845e-06, + "loss": 1.1691, + "step": 20474 + }, + { + "epoch": 1.72, + "learning_rate": 8.080420867130235e-06, + "loss": 1.1445, + "step": 20475 + }, + { + "epoch": 1.72, + "learning_rate": 8.079528806039803e-06, + "loss": 1.1246, + "step": 20476 + }, + { + "epoch": 1.72, + "learning_rate": 8.078636760816915e-06, + "loss": 1.1236, + "step": 20477 + }, + { + "epoch": 1.72, + "learning_rate": 8.077744731468948e-06, + "loss": 1.2121, + "step": 20478 + }, + { + "epoch": 1.72, + "learning_rate": 8.076852718003269e-06, + "loss": 1.187, + "step": 20479 + }, + { + "epoch": 1.72, + "learning_rate": 8.07596072042725e-06, + "loss": 1.1377, + "step": 20480 + }, + { + "epoch": 1.72, + "learning_rate": 8.075068738748253e-06, + "loss": 1.194, + "step": 20481 + }, + { + "epoch": 1.72, + "learning_rate": 8.07417677297366e-06, + "loss": 1.2418, + "step": 20482 + }, + { + "epoch": 1.72, + "learning_rate": 8.073284823110833e-06, + "loss": 1.152, + "step": 20483 + }, + { + "epoch": 1.72, + "learning_rate": 8.072392889167145e-06, + "loss": 1.0818, + "step": 20484 + }, + { + "epoch": 1.72, + "learning_rate": 8.07150097114996e-06, + "loss": 1.0975, + "step": 20485 + }, + { + "epoch": 1.72, + "learning_rate": 8.070609069066655e-06, + "loss": 1.1493, + "step": 20486 + }, + { + "epoch": 1.72, + "learning_rate": 8.069717182924593e-06, + "loss": 1.1523, + "step": 20487 + }, + { + "epoch": 1.72, + "learning_rate": 8.068825312731144e-06, + "loss": 1.2322, + "step": 20488 + }, + { + "epoch": 1.73, + "learning_rate": 8.067933458493682e-06, + "loss": 1.1598, + "step": 20489 + }, + { + "epoch": 1.73, + "learning_rate": 8.067041620219572e-06, + "loss": 1.2123, + "step": 20490 + }, + { + "epoch": 1.73, + "learning_rate": 8.066149797916181e-06, + "loss": 1.2169, + "step": 20491 + }, + { + "epoch": 1.73, + "learning_rate": 8.065257991590878e-06, + "loss": 1.2077, + "step": 20492 + }, + { + "epoch": 1.73, + "learning_rate": 8.064366201251036e-06, + "loss": 1.2007, + "step": 20493 + }, + { + "epoch": 1.73, + "learning_rate": 8.063474426904018e-06, + "loss": 1.1896, + "step": 20494 + }, + { + "epoch": 1.73, + "learning_rate": 8.062582668557197e-06, + "loss": 1.2232, + "step": 20495 + }, + { + "epoch": 1.73, + "learning_rate": 8.061690926217933e-06, + "loss": 1.2986, + "step": 20496 + }, + { + "epoch": 1.73, + "learning_rate": 8.060799199893602e-06, + "loss": 1.1801, + "step": 20497 + }, + { + "epoch": 1.73, + "learning_rate": 8.059907489591567e-06, + "loss": 1.0147, + "step": 20498 + }, + { + "epoch": 1.73, + "learning_rate": 8.0590157953192e-06, + "loss": 1.1842, + "step": 20499 + }, + { + "epoch": 1.73, + "learning_rate": 8.058124117083862e-06, + "loss": 1.1215, + "step": 20500 + }, + { + "epoch": 1.73, + "learning_rate": 8.057232454892924e-06, + "loss": 1.2977, + "step": 20501 + }, + { + "epoch": 1.73, + "learning_rate": 8.056340808753759e-06, + "loss": 1.1556, + "step": 20502 + }, + { + "epoch": 1.73, + "learning_rate": 8.055449178673721e-06, + "loss": 1.1339, + "step": 20503 + }, + { + "epoch": 1.73, + "learning_rate": 8.054557564660189e-06, + "loss": 1.1902, + "step": 20504 + }, + { + "epoch": 1.73, + "learning_rate": 8.053665966720524e-06, + "loss": 1.0812, + "step": 20505 + }, + { + "epoch": 1.73, + "learning_rate": 8.052774384862094e-06, + "loss": 1.1384, + "step": 20506 + }, + { + "epoch": 1.73, + "learning_rate": 8.051882819092263e-06, + "loss": 1.2294, + "step": 20507 + }, + { + "epoch": 1.73, + "learning_rate": 8.050991269418402e-06, + "loss": 1.1523, + "step": 20508 + }, + { + "epoch": 1.73, + "learning_rate": 8.050099735847874e-06, + "loss": 1.1355, + "step": 20509 + }, + { + "epoch": 1.73, + "learning_rate": 8.049208218388048e-06, + "loss": 1.168, + "step": 20510 + }, + { + "epoch": 1.73, + "learning_rate": 8.048316717046284e-06, + "loss": 1.1899, + "step": 20511 + }, + { + "epoch": 1.73, + "learning_rate": 8.047425231829953e-06, + "loss": 1.202, + "step": 20512 + }, + { + "epoch": 1.73, + "learning_rate": 8.046533762746422e-06, + "loss": 1.186, + "step": 20513 + }, + { + "epoch": 1.73, + "learning_rate": 8.04564230980305e-06, + "loss": 1.2284, + "step": 20514 + }, + { + "epoch": 1.73, + "learning_rate": 8.044750873007212e-06, + "loss": 1.1787, + "step": 20515 + }, + { + "epoch": 1.73, + "learning_rate": 8.043859452366266e-06, + "loss": 1.2385, + "step": 20516 + }, + { + "epoch": 1.73, + "learning_rate": 8.04296804788758e-06, + "loss": 1.1633, + "step": 20517 + }, + { + "epoch": 1.73, + "learning_rate": 8.042076659578515e-06, + "loss": 1.193, + "step": 20518 + }, + { + "epoch": 1.73, + "learning_rate": 8.041185287446444e-06, + "loss": 1.1104, + "step": 20519 + }, + { + "epoch": 1.73, + "learning_rate": 8.040293931498725e-06, + "loss": 1.216, + "step": 20520 + }, + { + "epoch": 1.73, + "learning_rate": 8.039402591742724e-06, + "loss": 1.1759, + "step": 20521 + }, + { + "epoch": 1.73, + "learning_rate": 8.038511268185804e-06, + "loss": 1.1875, + "step": 20522 + }, + { + "epoch": 1.73, + "learning_rate": 8.037619960835338e-06, + "loss": 1.148, + "step": 20523 + }, + { + "epoch": 1.73, + "learning_rate": 8.036728669698679e-06, + "loss": 1.2379, + "step": 20524 + }, + { + "epoch": 1.73, + "learning_rate": 8.035837394783195e-06, + "loss": 1.1285, + "step": 20525 + }, + { + "epoch": 1.73, + "learning_rate": 8.034946136096257e-06, + "loss": 1.1191, + "step": 20526 + }, + { + "epoch": 1.73, + "learning_rate": 8.034054893645215e-06, + "loss": 1.2411, + "step": 20527 + }, + { + "epoch": 1.73, + "learning_rate": 8.033163667437448e-06, + "loss": 1.193, + "step": 20528 + }, + { + "epoch": 1.73, + "learning_rate": 8.032272457480307e-06, + "loss": 1.2602, + "step": 20529 + }, + { + "epoch": 1.73, + "learning_rate": 8.031381263781166e-06, + "loss": 1.1829, + "step": 20530 + }, + { + "epoch": 1.73, + "learning_rate": 8.030490086347377e-06, + "loss": 1.1844, + "step": 20531 + }, + { + "epoch": 1.73, + "learning_rate": 8.029598925186314e-06, + "loss": 1.1638, + "step": 20532 + }, + { + "epoch": 1.73, + "learning_rate": 8.028707780305332e-06, + "loss": 1.1717, + "step": 20533 + }, + { + "epoch": 1.73, + "learning_rate": 8.027816651711802e-06, + "loss": 1.1419, + "step": 20534 + }, + { + "epoch": 1.73, + "learning_rate": 8.026925539413076e-06, + "loss": 1.2219, + "step": 20535 + }, + { + "epoch": 1.73, + "learning_rate": 8.026034443416525e-06, + "loss": 1.2227, + "step": 20536 + }, + { + "epoch": 1.73, + "learning_rate": 8.02514336372951e-06, + "loss": 1.1812, + "step": 20537 + }, + { + "epoch": 1.73, + "learning_rate": 8.02425230035939e-06, + "loss": 1.2225, + "step": 20538 + }, + { + "epoch": 1.73, + "learning_rate": 8.023361253313533e-06, + "loss": 1.1703, + "step": 20539 + }, + { + "epoch": 1.73, + "learning_rate": 8.022470222599297e-06, + "loss": 1.1039, + "step": 20540 + }, + { + "epoch": 1.73, + "learning_rate": 8.021579208224045e-06, + "loss": 1.2273, + "step": 20541 + }, + { + "epoch": 1.73, + "learning_rate": 8.020688210195137e-06, + "loss": 1.2095, + "step": 20542 + }, + { + "epoch": 1.73, + "learning_rate": 8.019797228519939e-06, + "loss": 1.2588, + "step": 20543 + }, + { + "epoch": 1.73, + "learning_rate": 8.018906263205809e-06, + "loss": 1.3085, + "step": 20544 + }, + { + "epoch": 1.73, + "learning_rate": 8.01801531426011e-06, + "loss": 1.1825, + "step": 20545 + }, + { + "epoch": 1.73, + "learning_rate": 8.0171243816902e-06, + "loss": 1.1332, + "step": 20546 + }, + { + "epoch": 1.73, + "learning_rate": 8.016233465503446e-06, + "loss": 1.0728, + "step": 20547 + }, + { + "epoch": 1.73, + "learning_rate": 8.015342565707206e-06, + "loss": 1.1495, + "step": 20548 + }, + { + "epoch": 1.73, + "learning_rate": 8.014451682308841e-06, + "loss": 1.1518, + "step": 20549 + }, + { + "epoch": 1.73, + "learning_rate": 8.013560815315708e-06, + "loss": 1.2178, + "step": 20550 + }, + { + "epoch": 1.73, + "learning_rate": 8.012669964735174e-06, + "loss": 1.1479, + "step": 20551 + }, + { + "epoch": 1.73, + "learning_rate": 8.011779130574599e-06, + "loss": 1.1633, + "step": 20552 + }, + { + "epoch": 1.73, + "learning_rate": 8.010888312841336e-06, + "loss": 1.2718, + "step": 20553 + }, + { + "epoch": 1.73, + "learning_rate": 8.009997511542755e-06, + "loss": 1.1883, + "step": 20554 + }, + { + "epoch": 1.73, + "learning_rate": 8.009106726686209e-06, + "loss": 1.196, + "step": 20555 + }, + { + "epoch": 1.73, + "learning_rate": 8.008215958279063e-06, + "loss": 1.2359, + "step": 20556 + }, + { + "epoch": 1.73, + "learning_rate": 8.00732520632867e-06, + "loss": 1.198, + "step": 20557 + }, + { + "epoch": 1.73, + "learning_rate": 8.006434470842396e-06, + "loss": 1.1675, + "step": 20558 + }, + { + "epoch": 1.73, + "learning_rate": 8.0055437518276e-06, + "loss": 1.1618, + "step": 20559 + }, + { + "epoch": 1.73, + "learning_rate": 8.00465304929164e-06, + "loss": 1.2424, + "step": 20560 + }, + { + "epoch": 1.73, + "learning_rate": 8.00376236324187e-06, + "loss": 1.0857, + "step": 20561 + }, + { + "epoch": 1.73, + "learning_rate": 8.002871693685657e-06, + "loss": 1.1606, + "step": 20562 + }, + { + "epoch": 1.73, + "learning_rate": 8.001981040630362e-06, + "loss": 1.1955, + "step": 20563 + }, + { + "epoch": 1.73, + "learning_rate": 8.001090404083332e-06, + "loss": 1.1791, + "step": 20564 + }, + { + "epoch": 1.73, + "learning_rate": 8.000199784051937e-06, + "loss": 1.1516, + "step": 20565 + }, + { + "epoch": 1.73, + "learning_rate": 7.99930918054353e-06, + "loss": 1.1331, + "step": 20566 + }, + { + "epoch": 1.73, + "learning_rate": 7.998418593565472e-06, + "loss": 1.3296, + "step": 20567 + }, + { + "epoch": 1.73, + "learning_rate": 7.997528023125117e-06, + "loss": 1.1763, + "step": 20568 + }, + { + "epoch": 1.73, + "learning_rate": 7.99663746922983e-06, + "loss": 1.2193, + "step": 20569 + }, + { + "epoch": 1.73, + "learning_rate": 7.995746931886964e-06, + "loss": 1.1842, + "step": 20570 + }, + { + "epoch": 1.73, + "learning_rate": 7.99485641110388e-06, + "loss": 1.1025, + "step": 20571 + }, + { + "epoch": 1.73, + "learning_rate": 7.99396590688793e-06, + "loss": 1.2447, + "step": 20572 + }, + { + "epoch": 1.73, + "learning_rate": 7.993075419246482e-06, + "loss": 1.1455, + "step": 20573 + }, + { + "epoch": 1.73, + "learning_rate": 7.992184948186881e-06, + "loss": 1.1213, + "step": 20574 + }, + { + "epoch": 1.73, + "learning_rate": 7.99129449371649e-06, + "loss": 1.134, + "step": 20575 + }, + { + "epoch": 1.73, + "learning_rate": 7.990404055842673e-06, + "loss": 1.2099, + "step": 20576 + }, + { + "epoch": 1.73, + "learning_rate": 7.989513634572775e-06, + "loss": 1.2147, + "step": 20577 + }, + { + "epoch": 1.73, + "learning_rate": 7.988623229914162e-06, + "loss": 1.1284, + "step": 20578 + }, + { + "epoch": 1.73, + "learning_rate": 7.987732841874187e-06, + "loss": 1.1333, + "step": 20579 + }, + { + "epoch": 1.73, + "learning_rate": 7.98684247046021e-06, + "loss": 1.2049, + "step": 20580 + }, + { + "epoch": 1.73, + "learning_rate": 7.985952115679579e-06, + "loss": 1.1991, + "step": 20581 + }, + { + "epoch": 1.73, + "learning_rate": 7.98506177753966e-06, + "loss": 1.284, + "step": 20582 + }, + { + "epoch": 1.73, + "learning_rate": 7.984171456047805e-06, + "loss": 1.1366, + "step": 20583 + }, + { + "epoch": 1.73, + "learning_rate": 7.983281151211371e-06, + "loss": 1.2016, + "step": 20584 + }, + { + "epoch": 1.73, + "learning_rate": 7.98239086303771e-06, + "loss": 1.1072, + "step": 20585 + }, + { + "epoch": 1.73, + "learning_rate": 7.981500591534186e-06, + "loss": 1.1761, + "step": 20586 + }, + { + "epoch": 1.73, + "learning_rate": 7.980610336708148e-06, + "loss": 1.1922, + "step": 20587 + }, + { + "epoch": 1.73, + "learning_rate": 7.979720098566951e-06, + "loss": 1.0945, + "step": 20588 + }, + { + "epoch": 1.73, + "learning_rate": 7.978829877117957e-06, + "loss": 1.2474, + "step": 20589 + }, + { + "epoch": 1.73, + "learning_rate": 7.977939672368518e-06, + "loss": 1.1854, + "step": 20590 + }, + { + "epoch": 1.73, + "learning_rate": 7.977049484325988e-06, + "loss": 1.1871, + "step": 20591 + }, + { + "epoch": 1.73, + "learning_rate": 7.976159312997718e-06, + "loss": 1.1443, + "step": 20592 + }, + { + "epoch": 1.73, + "learning_rate": 7.975269158391073e-06, + "loss": 1.1058, + "step": 20593 + }, + { + "epoch": 1.73, + "learning_rate": 7.9743790205134e-06, + "loss": 1.1432, + "step": 20594 + }, + { + "epoch": 1.73, + "learning_rate": 7.973488899372058e-06, + "loss": 1.182, + "step": 20595 + }, + { + "epoch": 1.73, + "learning_rate": 7.972598794974394e-06, + "loss": 1.2746, + "step": 20596 + }, + { + "epoch": 1.73, + "learning_rate": 7.971708707327774e-06, + "loss": 1.1752, + "step": 20597 + }, + { + "epoch": 1.73, + "learning_rate": 7.970818636439541e-06, + "loss": 1.0894, + "step": 20598 + }, + { + "epoch": 1.73, + "learning_rate": 7.969928582317058e-06, + "loss": 1.1913, + "step": 20599 + }, + { + "epoch": 1.73, + "learning_rate": 7.96903854496767e-06, + "loss": 1.2562, + "step": 20600 + }, + { + "epoch": 1.73, + "learning_rate": 7.968148524398738e-06, + "loss": 1.1525, + "step": 20601 + }, + { + "epoch": 1.73, + "learning_rate": 7.967258520617613e-06, + "loss": 1.2537, + "step": 20602 + }, + { + "epoch": 1.73, + "learning_rate": 7.966368533631647e-06, + "loss": 1.1903, + "step": 20603 + }, + { + "epoch": 1.73, + "learning_rate": 7.965478563448199e-06, + "loss": 1.1485, + "step": 20604 + }, + { + "epoch": 1.73, + "learning_rate": 7.964588610074615e-06, + "loss": 1.2436, + "step": 20605 + }, + { + "epoch": 1.73, + "learning_rate": 7.963698673518254e-06, + "loss": 1.1184, + "step": 20606 + }, + { + "epoch": 1.73, + "learning_rate": 7.962808753786463e-06, + "loss": 1.3072, + "step": 20607 + }, + { + "epoch": 1.74, + "learning_rate": 7.961918850886602e-06, + "loss": 1.128, + "step": 20608 + }, + { + "epoch": 1.74, + "learning_rate": 7.961028964826015e-06, + "loss": 1.1369, + "step": 20609 + }, + { + "epoch": 1.74, + "learning_rate": 7.960139095612065e-06, + "loss": 1.2267, + "step": 20610 + }, + { + "epoch": 1.74, + "learning_rate": 7.959249243252091e-06, + "loss": 1.2139, + "step": 20611 + }, + { + "epoch": 1.74, + "learning_rate": 7.958359407753456e-06, + "loss": 1.188, + "step": 20612 + }, + { + "epoch": 1.74, + "learning_rate": 7.957469589123512e-06, + "loss": 1.2197, + "step": 20613 + }, + { + "epoch": 1.74, + "learning_rate": 7.956579787369603e-06, + "loss": 1.1687, + "step": 20614 + }, + { + "epoch": 1.74, + "learning_rate": 7.955690002499091e-06, + "loss": 1.1496, + "step": 20615 + }, + { + "epoch": 1.74, + "learning_rate": 7.95480023451932e-06, + "loss": 1.1394, + "step": 20616 + }, + { + "epoch": 1.74, + "learning_rate": 7.953910483437647e-06, + "loss": 1.0823, + "step": 20617 + }, + { + "epoch": 1.74, + "learning_rate": 7.953020749261414e-06, + "loss": 1.1479, + "step": 20618 + }, + { + "epoch": 1.74, + "learning_rate": 7.952131031997985e-06, + "loss": 1.1261, + "step": 20619 + }, + { + "epoch": 1.74, + "learning_rate": 7.951241331654702e-06, + "loss": 1.1761, + "step": 20620 + }, + { + "epoch": 1.74, + "learning_rate": 7.95035164823892e-06, + "loss": 1.1051, + "step": 20621 + }, + { + "epoch": 1.74, + "learning_rate": 7.949461981757985e-06, + "loss": 1.2031, + "step": 20622 + }, + { + "epoch": 1.74, + "learning_rate": 7.948572332219255e-06, + "loss": 1.1937, + "step": 20623 + }, + { + "epoch": 1.74, + "learning_rate": 7.947682699630076e-06, + "loss": 1.151, + "step": 20624 + }, + { + "epoch": 1.74, + "learning_rate": 7.946793083997797e-06, + "loss": 1.1172, + "step": 20625 + }, + { + "epoch": 1.74, + "learning_rate": 7.945903485329776e-06, + "loss": 1.2188, + "step": 20626 + }, + { + "epoch": 1.74, + "learning_rate": 7.945013903633352e-06, + "loss": 1.1588, + "step": 20627 + }, + { + "epoch": 1.74, + "learning_rate": 7.944124338915886e-06, + "loss": 1.1683, + "step": 20628 + }, + { + "epoch": 1.74, + "learning_rate": 7.94323479118472e-06, + "loss": 1.082, + "step": 20629 + }, + { + "epoch": 1.74, + "learning_rate": 7.942345260447208e-06, + "loss": 1.1651, + "step": 20630 + }, + { + "epoch": 1.74, + "learning_rate": 7.941455746710694e-06, + "loss": 1.1196, + "step": 20631 + }, + { + "epoch": 1.74, + "learning_rate": 7.940566249982537e-06, + "loss": 1.1226, + "step": 20632 + }, + { + "epoch": 1.74, + "learning_rate": 7.939676770270077e-06, + "loss": 1.1308, + "step": 20633 + }, + { + "epoch": 1.74, + "learning_rate": 7.938787307580669e-06, + "loss": 1.1549, + "step": 20634 + }, + { + "epoch": 1.74, + "learning_rate": 7.937897861921656e-06, + "loss": 1.2717, + "step": 20635 + }, + { + "epoch": 1.74, + "learning_rate": 7.937008433300394e-06, + "loss": 1.1626, + "step": 20636 + }, + { + "epoch": 1.74, + "learning_rate": 7.936119021724229e-06, + "loss": 1.1521, + "step": 20637 + }, + { + "epoch": 1.74, + "learning_rate": 7.935229627200505e-06, + "loss": 1.1497, + "step": 20638 + }, + { + "epoch": 1.74, + "learning_rate": 7.934340249736581e-06, + "loss": 1.1195, + "step": 20639 + }, + { + "epoch": 1.74, + "learning_rate": 7.933450889339794e-06, + "loss": 1.2097, + "step": 20640 + }, + { + "epoch": 1.74, + "learning_rate": 7.932561546017502e-06, + "loss": 1.1542, + "step": 20641 + }, + { + "epoch": 1.74, + "learning_rate": 7.931672219777043e-06, + "loss": 1.1978, + "step": 20642 + }, + { + "epoch": 1.74, + "learning_rate": 7.930782910625773e-06, + "loss": 1.2321, + "step": 20643 + }, + { + "epoch": 1.74, + "learning_rate": 7.929893618571035e-06, + "loss": 1.1519, + "step": 20644 + }, + { + "epoch": 1.74, + "learning_rate": 7.929004343620181e-06, + "loss": 1.0529, + "step": 20645 + }, + { + "epoch": 1.74, + "learning_rate": 7.928115085780552e-06, + "loss": 1.2085, + "step": 20646 + }, + { + "epoch": 1.74, + "learning_rate": 7.927225845059504e-06, + "loss": 1.2356, + "step": 20647 + }, + { + "epoch": 1.74, + "learning_rate": 7.926336621464376e-06, + "loss": 1.2572, + "step": 20648 + }, + { + "epoch": 1.74, + "learning_rate": 7.925447415002523e-06, + "loss": 1.2196, + "step": 20649 + }, + { + "epoch": 1.74, + "learning_rate": 7.92455822568128e-06, + "loss": 1.2283, + "step": 20650 + }, + { + "epoch": 1.74, + "learning_rate": 7.923669053508005e-06, + "loss": 1.1561, + "step": 20651 + }, + { + "epoch": 1.74, + "learning_rate": 7.922779898490044e-06, + "loss": 1.1403, + "step": 20652 + }, + { + "epoch": 1.74, + "learning_rate": 7.921890760634733e-06, + "loss": 1.2168, + "step": 20653 + }, + { + "epoch": 1.74, + "learning_rate": 7.921001639949431e-06, + "loss": 1.102, + "step": 20654 + }, + { + "epoch": 1.74, + "learning_rate": 7.920112536441478e-06, + "loss": 1.1911, + "step": 20655 + }, + { + "epoch": 1.74, + "learning_rate": 7.919223450118223e-06, + "loss": 1.2323, + "step": 20656 + }, + { + "epoch": 1.74, + "learning_rate": 7.918334380987003e-06, + "loss": 1.336, + "step": 20657 + }, + { + "epoch": 1.74, + "learning_rate": 7.917445329055176e-06, + "loss": 1.1653, + "step": 20658 + }, + { + "epoch": 1.74, + "learning_rate": 7.916556294330081e-06, + "loss": 1.0991, + "step": 20659 + }, + { + "epoch": 1.74, + "learning_rate": 7.915667276819067e-06, + "loss": 1.1741, + "step": 20660 + }, + { + "epoch": 1.74, + "learning_rate": 7.914778276529473e-06, + "loss": 1.1641, + "step": 20661 + }, + { + "epoch": 1.74, + "learning_rate": 7.913889293468648e-06, + "loss": 1.1719, + "step": 20662 + }, + { + "epoch": 1.74, + "learning_rate": 7.913000327643942e-06, + "loss": 1.1194, + "step": 20663 + }, + { + "epoch": 1.74, + "learning_rate": 7.912111379062689e-06, + "loss": 1.2443, + "step": 20664 + }, + { + "epoch": 1.74, + "learning_rate": 7.911222447732246e-06, + "loss": 1.1998, + "step": 20665 + }, + { + "epoch": 1.74, + "learning_rate": 7.910333533659949e-06, + "loss": 1.2521, + "step": 20666 + }, + { + "epoch": 1.74, + "learning_rate": 7.909444636853146e-06, + "loss": 1.2259, + "step": 20667 + }, + { + "epoch": 1.74, + "learning_rate": 7.908555757319178e-06, + "loss": 1.0998, + "step": 20668 + }, + { + "epoch": 1.74, + "learning_rate": 7.907666895065395e-06, + "loss": 1.1627, + "step": 20669 + }, + { + "epoch": 1.74, + "learning_rate": 7.906778050099136e-06, + "loss": 1.1773, + "step": 20670 + }, + { + "epoch": 1.74, + "learning_rate": 7.905889222427747e-06, + "loss": 1.0625, + "step": 20671 + }, + { + "epoch": 1.74, + "learning_rate": 7.90500041205857e-06, + "loss": 1.1633, + "step": 20672 + }, + { + "epoch": 1.74, + "learning_rate": 7.904111618998953e-06, + "loss": 1.253, + "step": 20673 + }, + { + "epoch": 1.74, + "learning_rate": 7.903222843256233e-06, + "loss": 1.3046, + "step": 20674 + }, + { + "epoch": 1.74, + "learning_rate": 7.902334084837758e-06, + "loss": 1.2786, + "step": 20675 + }, + { + "epoch": 1.74, + "learning_rate": 7.901445343750873e-06, + "loss": 1.1915, + "step": 20676 + }, + { + "epoch": 1.74, + "learning_rate": 7.900556620002915e-06, + "loss": 1.1462, + "step": 20677 + }, + { + "epoch": 1.74, + "learning_rate": 7.899667913601232e-06, + "loss": 1.175, + "step": 20678 + }, + { + "epoch": 1.74, + "learning_rate": 7.898779224553163e-06, + "loss": 1.1909, + "step": 20679 + }, + { + "epoch": 1.74, + "learning_rate": 7.897890552866057e-06, + "loss": 1.2172, + "step": 20680 + }, + { + "epoch": 1.74, + "learning_rate": 7.897001898547247e-06, + "loss": 1.2524, + "step": 20681 + }, + { + "epoch": 1.74, + "learning_rate": 7.896113261604082e-06, + "loss": 1.1788, + "step": 20682 + }, + { + "epoch": 1.74, + "learning_rate": 7.895224642043902e-06, + "loss": 1.1702, + "step": 20683 + }, + { + "epoch": 1.74, + "learning_rate": 7.894336039874052e-06, + "loss": 1.1937, + "step": 20684 + }, + { + "epoch": 1.74, + "learning_rate": 7.893447455101868e-06, + "loss": 1.1065, + "step": 20685 + }, + { + "epoch": 1.74, + "learning_rate": 7.8925588877347e-06, + "loss": 1.1734, + "step": 20686 + }, + { + "epoch": 1.74, + "learning_rate": 7.89167033777988e-06, + "loss": 1.1819, + "step": 20687 + }, + { + "epoch": 1.74, + "learning_rate": 7.890781805244753e-06, + "loss": 1.185, + "step": 20688 + }, + { + "epoch": 1.74, + "learning_rate": 7.889893290136668e-06, + "loss": 1.1609, + "step": 20689 + }, + { + "epoch": 1.74, + "learning_rate": 7.889004792462958e-06, + "loss": 1.2439, + "step": 20690 + }, + { + "epoch": 1.74, + "learning_rate": 7.888116312230966e-06, + "loss": 1.2487, + "step": 20691 + }, + { + "epoch": 1.74, + "learning_rate": 7.88722784944803e-06, + "loss": 1.1789, + "step": 20692 + }, + { + "epoch": 1.74, + "learning_rate": 7.886339404121497e-06, + "loss": 1.2145, + "step": 20693 + }, + { + "epoch": 1.74, + "learning_rate": 7.885450976258704e-06, + "loss": 1.1762, + "step": 20694 + }, + { + "epoch": 1.74, + "learning_rate": 7.884562565866991e-06, + "loss": 1.1029, + "step": 20695 + }, + { + "epoch": 1.74, + "learning_rate": 7.883674172953697e-06, + "loss": 1.2181, + "step": 20696 + }, + { + "epoch": 1.74, + "learning_rate": 7.882785797526168e-06, + "loss": 1.2497, + "step": 20697 + }, + { + "epoch": 1.74, + "learning_rate": 7.881897439591738e-06, + "loss": 1.1875, + "step": 20698 + }, + { + "epoch": 1.74, + "learning_rate": 7.881009099157752e-06, + "loss": 1.2191, + "step": 20699 + }, + { + "epoch": 1.74, + "learning_rate": 7.88012077623154e-06, + "loss": 1.1443, + "step": 20700 + }, + { + "epoch": 1.74, + "learning_rate": 7.879232470820455e-06, + "loss": 1.2335, + "step": 20701 + }, + { + "epoch": 1.74, + "learning_rate": 7.87834418293183e-06, + "loss": 1.1619, + "step": 20702 + }, + { + "epoch": 1.74, + "learning_rate": 7.877455912573e-06, + "loss": 1.2144, + "step": 20703 + }, + { + "epoch": 1.74, + "learning_rate": 7.876567659751313e-06, + "loss": 1.1479, + "step": 20704 + }, + { + "epoch": 1.74, + "learning_rate": 7.875679424474103e-06, + "loss": 1.2253, + "step": 20705 + }, + { + "epoch": 1.74, + "learning_rate": 7.87479120674871e-06, + "loss": 1.1359, + "step": 20706 + }, + { + "epoch": 1.74, + "learning_rate": 7.873903006582467e-06, + "loss": 1.2117, + "step": 20707 + }, + { + "epoch": 1.74, + "learning_rate": 7.873014823982723e-06, + "loss": 1.0994, + "step": 20708 + }, + { + "epoch": 1.74, + "learning_rate": 7.87212665895681e-06, + "loss": 1.2345, + "step": 20709 + }, + { + "epoch": 1.74, + "learning_rate": 7.871238511512068e-06, + "loss": 1.1832, + "step": 20710 + }, + { + "epoch": 1.74, + "learning_rate": 7.870350381655831e-06, + "loss": 1.2451, + "step": 20711 + }, + { + "epoch": 1.74, + "learning_rate": 7.869462269395443e-06, + "loss": 1.2032, + "step": 20712 + }, + { + "epoch": 1.74, + "learning_rate": 7.868574174738243e-06, + "loss": 1.2124, + "step": 20713 + }, + { + "epoch": 1.74, + "learning_rate": 7.86768609769156e-06, + "loss": 1.1414, + "step": 20714 + }, + { + "epoch": 1.74, + "learning_rate": 7.86679803826274e-06, + "loss": 1.2568, + "step": 20715 + }, + { + "epoch": 1.74, + "learning_rate": 7.865909996459119e-06, + "loss": 1.1476, + "step": 20716 + }, + { + "epoch": 1.74, + "learning_rate": 7.86502197228803e-06, + "loss": 1.2512, + "step": 20717 + }, + { + "epoch": 1.74, + "learning_rate": 7.86413396575681e-06, + "loss": 1.1627, + "step": 20718 + }, + { + "epoch": 1.74, + "learning_rate": 7.863245976872804e-06, + "loss": 1.2273, + "step": 20719 + }, + { + "epoch": 1.74, + "learning_rate": 7.862358005643341e-06, + "loss": 1.1802, + "step": 20720 + }, + { + "epoch": 1.74, + "learning_rate": 7.861470052075764e-06, + "loss": 1.152, + "step": 20721 + }, + { + "epoch": 1.74, + "learning_rate": 7.860582116177399e-06, + "loss": 1.1771, + "step": 20722 + }, + { + "epoch": 1.74, + "learning_rate": 7.859694197955593e-06, + "loss": 1.1571, + "step": 20723 + }, + { + "epoch": 1.74, + "learning_rate": 7.858806297417679e-06, + "loss": 1.1863, + "step": 20724 + }, + { + "epoch": 1.74, + "learning_rate": 7.85791841457099e-06, + "loss": 1.1496, + "step": 20725 + }, + { + "epoch": 1.74, + "learning_rate": 7.857030549422868e-06, + "loss": 1.1527, + "step": 20726 + }, + { + "epoch": 1.75, + "learning_rate": 7.856142701980643e-06, + "loss": 1.1342, + "step": 20727 + }, + { + "epoch": 1.75, + "learning_rate": 7.855254872251656e-06, + "loss": 1.2358, + "step": 20728 + }, + { + "epoch": 1.75, + "learning_rate": 7.854367060243237e-06, + "loss": 1.2221, + "step": 20729 + }, + { + "epoch": 1.75, + "learning_rate": 7.853479265962727e-06, + "loss": 1.224, + "step": 20730 + }, + { + "epoch": 1.75, + "learning_rate": 7.852591489417453e-06, + "loss": 1.2138, + "step": 20731 + }, + { + "epoch": 1.75, + "learning_rate": 7.851703730614762e-06, + "loss": 1.1193, + "step": 20732 + }, + { + "epoch": 1.75, + "learning_rate": 7.850815989561979e-06, + "loss": 1.1341, + "step": 20733 + }, + { + "epoch": 1.75, + "learning_rate": 7.849928266266443e-06, + "loss": 1.2043, + "step": 20734 + }, + { + "epoch": 1.75, + "learning_rate": 7.849040560735484e-06, + "loss": 1.1584, + "step": 20735 + }, + { + "epoch": 1.75, + "learning_rate": 7.848152872976444e-06, + "loss": 1.1793, + "step": 20736 + }, + { + "epoch": 1.75, + "learning_rate": 7.847265202996653e-06, + "loss": 1.0785, + "step": 20737 + }, + { + "epoch": 1.75, + "learning_rate": 7.846377550803444e-06, + "loss": 1.2049, + "step": 20738 + }, + { + "epoch": 1.75, + "learning_rate": 7.845489916404156e-06, + "loss": 1.1166, + "step": 20739 + }, + { + "epoch": 1.75, + "learning_rate": 7.844602299806117e-06, + "loss": 1.2218, + "step": 20740 + }, + { + "epoch": 1.75, + "learning_rate": 7.843714701016667e-06, + "loss": 1.1622, + "step": 20741 + }, + { + "epoch": 1.75, + "learning_rate": 7.84282712004313e-06, + "loss": 1.1396, + "step": 20742 + }, + { + "epoch": 1.75, + "learning_rate": 7.841939556892852e-06, + "loss": 1.2349, + "step": 20743 + }, + { + "epoch": 1.75, + "learning_rate": 7.841052011573157e-06, + "loss": 1.0972, + "step": 20744 + }, + { + "epoch": 1.75, + "learning_rate": 7.840164484091384e-06, + "loss": 1.1808, + "step": 20745 + }, + { + "epoch": 1.75, + "learning_rate": 7.83927697445486e-06, + "loss": 1.2689, + "step": 20746 + }, + { + "epoch": 1.75, + "learning_rate": 7.838389482670925e-06, + "loss": 1.1052, + "step": 20747 + }, + { + "epoch": 1.75, + "learning_rate": 7.837502008746904e-06, + "loss": 1.1332, + "step": 20748 + }, + { + "epoch": 1.75, + "learning_rate": 7.836614552690137e-06, + "loss": 1.2204, + "step": 20749 + }, + { + "epoch": 1.75, + "learning_rate": 7.835727114507949e-06, + "loss": 1.0605, + "step": 20750 + }, + { + "epoch": 1.75, + "learning_rate": 7.834839694207677e-06, + "loss": 1.164, + "step": 20751 + }, + { + "epoch": 1.75, + "learning_rate": 7.833952291796655e-06, + "loss": 1.2869, + "step": 20752 + }, + { + "epoch": 1.75, + "learning_rate": 7.833064907282209e-06, + "loss": 1.118, + "step": 20753 + }, + { + "epoch": 1.75, + "learning_rate": 7.832177540671677e-06, + "loss": 1.2252, + "step": 20754 + }, + { + "epoch": 1.75, + "learning_rate": 7.831290191972387e-06, + "loss": 1.234, + "step": 20755 + }, + { + "epoch": 1.75, + "learning_rate": 7.830402861191673e-06, + "loss": 1.2465, + "step": 20756 + }, + { + "epoch": 1.75, + "learning_rate": 7.829515548336861e-06, + "loss": 1.2233, + "step": 20757 + }, + { + "epoch": 1.75, + "learning_rate": 7.828628253415292e-06, + "loss": 1.2385, + "step": 20758 + }, + { + "epoch": 1.75, + "learning_rate": 7.827740976434287e-06, + "loss": 1.2106, + "step": 20759 + }, + { + "epoch": 1.75, + "learning_rate": 7.826853717401185e-06, + "loss": 1.1282, + "step": 20760 + }, + { + "epoch": 1.75, + "learning_rate": 7.825966476323306e-06, + "loss": 1.1931, + "step": 20761 + }, + { + "epoch": 1.75, + "learning_rate": 7.825079253207992e-06, + "loss": 1.2046, + "step": 20762 + }, + { + "epoch": 1.75, + "learning_rate": 7.82419204806257e-06, + "loss": 1.255, + "step": 20763 + }, + { + "epoch": 1.75, + "learning_rate": 7.823304860894368e-06, + "loss": 1.1883, + "step": 20764 + }, + { + "epoch": 1.75, + "learning_rate": 7.82241769171072e-06, + "loss": 1.1758, + "step": 20765 + }, + { + "epoch": 1.75, + "learning_rate": 7.821530540518952e-06, + "loss": 1.2652, + "step": 20766 + }, + { + "epoch": 1.75, + "learning_rate": 7.8206434073264e-06, + "loss": 1.2735, + "step": 20767 + }, + { + "epoch": 1.75, + "learning_rate": 7.819756292140382e-06, + "loss": 1.1655, + "step": 20768 + }, + { + "epoch": 1.75, + "learning_rate": 7.818869194968242e-06, + "loss": 1.0515, + "step": 20769 + }, + { + "epoch": 1.75, + "learning_rate": 7.8179821158173e-06, + "loss": 1.0686, + "step": 20770 + }, + { + "epoch": 1.75, + "learning_rate": 7.817095054694889e-06, + "loss": 1.1289, + "step": 20771 + }, + { + "epoch": 1.75, + "learning_rate": 7.816208011608334e-06, + "loss": 1.1737, + "step": 20772 + }, + { + "epoch": 1.75, + "learning_rate": 7.815320986564974e-06, + "loss": 1.2203, + "step": 20773 + }, + { + "epoch": 1.75, + "learning_rate": 7.814433979572127e-06, + "loss": 1.1993, + "step": 20774 + }, + { + "epoch": 1.75, + "learning_rate": 7.813546990637122e-06, + "loss": 1.3002, + "step": 20775 + }, + { + "epoch": 1.75, + "learning_rate": 7.812660019767299e-06, + "loss": 1.1514, + "step": 20776 + }, + { + "epoch": 1.75, + "learning_rate": 7.811773066969974e-06, + "loss": 1.1574, + "step": 20777 + }, + { + "epoch": 1.75, + "learning_rate": 7.810886132252484e-06, + "loss": 1.1549, + "step": 20778 + }, + { + "epoch": 1.75, + "learning_rate": 7.809999215622149e-06, + "loss": 1.2569, + "step": 20779 + }, + { + "epoch": 1.75, + "learning_rate": 7.809112317086307e-06, + "loss": 1.199, + "step": 20780 + }, + { + "epoch": 1.75, + "learning_rate": 7.808225436652275e-06, + "loss": 1.2065, + "step": 20781 + }, + { + "epoch": 1.75, + "learning_rate": 7.80733857432739e-06, + "loss": 1.1326, + "step": 20782 + }, + { + "epoch": 1.75, + "learning_rate": 7.806451730118973e-06, + "loss": 1.1868, + "step": 20783 + }, + { + "epoch": 1.75, + "learning_rate": 7.805564904034355e-06, + "loss": 1.1274, + "step": 20784 + }, + { + "epoch": 1.75, + "learning_rate": 7.80467809608086e-06, + "loss": 1.205, + "step": 20785 + }, + { + "epoch": 1.75, + "learning_rate": 7.80379130626582e-06, + "loss": 1.1579, + "step": 20786 + }, + { + "epoch": 1.75, + "learning_rate": 7.802904534596556e-06, + "loss": 1.1738, + "step": 20787 + }, + { + "epoch": 1.75, + "learning_rate": 7.802017781080396e-06, + "loss": 1.1575, + "step": 20788 + }, + { + "epoch": 1.75, + "learning_rate": 7.801131045724675e-06, + "loss": 1.1589, + "step": 20789 + }, + { + "epoch": 1.75, + "learning_rate": 7.800244328536708e-06, + "loss": 1.1746, + "step": 20790 + }, + { + "epoch": 1.75, + "learning_rate": 7.79935762952383e-06, + "loss": 1.2396, + "step": 20791 + }, + { + "epoch": 1.75, + "learning_rate": 7.79847094869336e-06, + "loss": 1.1943, + "step": 20792 + }, + { + "epoch": 1.75, + "learning_rate": 7.797584286052631e-06, + "loss": 1.3256, + "step": 20793 + }, + { + "epoch": 1.75, + "learning_rate": 7.796697641608963e-06, + "loss": 1.1902, + "step": 20794 + }, + { + "epoch": 1.75, + "learning_rate": 7.795811015369688e-06, + "loss": 1.2243, + "step": 20795 + }, + { + "epoch": 1.75, + "learning_rate": 7.794924407342122e-06, + "loss": 1.2476, + "step": 20796 + }, + { + "epoch": 1.75, + "learning_rate": 7.7940378175336e-06, + "loss": 1.0737, + "step": 20797 + }, + { + "epoch": 1.75, + "learning_rate": 7.793151245951442e-06, + "loss": 1.1889, + "step": 20798 + }, + { + "epoch": 1.75, + "learning_rate": 7.792264692602978e-06, + "loss": 1.193, + "step": 20799 + }, + { + "epoch": 1.75, + "learning_rate": 7.791378157495525e-06, + "loss": 1.2031, + "step": 20800 + }, + { + "epoch": 1.75, + "learning_rate": 7.790491640636413e-06, + "loss": 1.1594, + "step": 20801 + }, + { + "epoch": 1.75, + "learning_rate": 7.78960514203297e-06, + "loss": 1.1966, + "step": 20802 + }, + { + "epoch": 1.75, + "learning_rate": 7.788718661692512e-06, + "loss": 1.2236, + "step": 20803 + }, + { + "epoch": 1.75, + "learning_rate": 7.787832199622373e-06, + "loss": 1.1753, + "step": 20804 + }, + { + "epoch": 1.75, + "learning_rate": 7.786945755829869e-06, + "loss": 1.1651, + "step": 20805 + }, + { + "epoch": 1.75, + "learning_rate": 7.78605933032233e-06, + "loss": 1.2226, + "step": 20806 + }, + { + "epoch": 1.75, + "learning_rate": 7.785172923107073e-06, + "loss": 1.2144, + "step": 20807 + }, + { + "epoch": 1.75, + "learning_rate": 7.784286534191433e-06, + "loss": 1.1136, + "step": 20808 + }, + { + "epoch": 1.75, + "learning_rate": 7.783400163582722e-06, + "loss": 1.1654, + "step": 20809 + }, + { + "epoch": 1.75, + "learning_rate": 7.78251381128827e-06, + "loss": 1.2202, + "step": 20810 + }, + { + "epoch": 1.75, + "learning_rate": 7.781627477315394e-06, + "loss": 1.1462, + "step": 20811 + }, + { + "epoch": 1.75, + "learning_rate": 7.780741161671426e-06, + "loss": 1.2272, + "step": 20812 + }, + { + "epoch": 1.75, + "learning_rate": 7.779854864363686e-06, + "loss": 1.2872, + "step": 20813 + }, + { + "epoch": 1.75, + "learning_rate": 7.778968585399491e-06, + "loss": 1.2004, + "step": 20814 + }, + { + "epoch": 1.75, + "learning_rate": 7.778082324786173e-06, + "loss": 1.2018, + "step": 20815 + }, + { + "epoch": 1.75, + "learning_rate": 7.777196082531048e-06, + "loss": 1.0354, + "step": 20816 + }, + { + "epoch": 1.75, + "learning_rate": 7.776309858641442e-06, + "loss": 1.0624, + "step": 20817 + }, + { + "epoch": 1.75, + "learning_rate": 7.77542365312467e-06, + "loss": 1.2522, + "step": 20818 + }, + { + "epoch": 1.75, + "learning_rate": 7.774537465988067e-06, + "loss": 1.1865, + "step": 20819 + }, + { + "epoch": 1.75, + "learning_rate": 7.773651297238943e-06, + "loss": 1.2097, + "step": 20820 + }, + { + "epoch": 1.75, + "learning_rate": 7.77276514688463e-06, + "loss": 1.1665, + "step": 20821 + }, + { + "epoch": 1.75, + "learning_rate": 7.771879014932435e-06, + "loss": 1.1609, + "step": 20822 + }, + { + "epoch": 1.75, + "learning_rate": 7.770992901389696e-06, + "loss": 1.1455, + "step": 20823 + }, + { + "epoch": 1.75, + "learning_rate": 7.770106806263724e-06, + "loss": 1.2533, + "step": 20824 + }, + { + "epoch": 1.75, + "learning_rate": 7.769220729561842e-06, + "loss": 1.1781, + "step": 20825 + }, + { + "epoch": 1.75, + "learning_rate": 7.768334671291376e-06, + "loss": 1.2012, + "step": 20826 + }, + { + "epoch": 1.75, + "learning_rate": 7.767448631459643e-06, + "loss": 1.2575, + "step": 20827 + }, + { + "epoch": 1.75, + "learning_rate": 7.766562610073963e-06, + "loss": 1.2962, + "step": 20828 + }, + { + "epoch": 1.75, + "learning_rate": 7.765676607141655e-06, + "loss": 1.1512, + "step": 20829 + }, + { + "epoch": 1.75, + "learning_rate": 7.764790622670044e-06, + "loss": 1.1748, + "step": 20830 + }, + { + "epoch": 1.75, + "learning_rate": 7.763904656666447e-06, + "loss": 1.1961, + "step": 20831 + }, + { + "epoch": 1.75, + "learning_rate": 7.763018709138188e-06, + "loss": 1.159, + "step": 20832 + }, + { + "epoch": 1.75, + "learning_rate": 7.762132780092581e-06, + "loss": 1.1495, + "step": 20833 + }, + { + "epoch": 1.75, + "learning_rate": 7.761246869536956e-06, + "loss": 1.1997, + "step": 20834 + }, + { + "epoch": 1.75, + "learning_rate": 7.760360977478618e-06, + "loss": 1.14, + "step": 20835 + }, + { + "epoch": 1.75, + "learning_rate": 7.759475103924899e-06, + "loss": 1.1876, + "step": 20836 + }, + { + "epoch": 1.75, + "learning_rate": 7.758589248883112e-06, + "loss": 1.2479, + "step": 20837 + }, + { + "epoch": 1.75, + "learning_rate": 7.757703412360574e-06, + "loss": 1.2327, + "step": 20838 + }, + { + "epoch": 1.75, + "learning_rate": 7.756817594364615e-06, + "loss": 1.1928, + "step": 20839 + }, + { + "epoch": 1.75, + "learning_rate": 7.755931794902543e-06, + "loss": 1.1431, + "step": 20840 + }, + { + "epoch": 1.75, + "learning_rate": 7.755046013981683e-06, + "loss": 1.1402, + "step": 20841 + }, + { + "epoch": 1.75, + "learning_rate": 7.754160251609348e-06, + "loss": 1.2096, + "step": 20842 + }, + { + "epoch": 1.75, + "learning_rate": 7.753274507792863e-06, + "loss": 1.2132, + "step": 20843 + }, + { + "epoch": 1.75, + "learning_rate": 7.752388782539541e-06, + "loss": 1.1255, + "step": 20844 + }, + { + "epoch": 1.75, + "learning_rate": 7.751503075856705e-06, + "loss": 1.1265, + "step": 20845 + }, + { + "epoch": 1.76, + "learning_rate": 7.750617387751666e-06, + "loss": 1.1437, + "step": 20846 + }, + { + "epoch": 1.76, + "learning_rate": 7.74973171823175e-06, + "loss": 1.2255, + "step": 20847 + }, + { + "epoch": 1.76, + "learning_rate": 7.74884606730427e-06, + "loss": 1.2499, + "step": 20848 + }, + { + "epoch": 1.76, + "learning_rate": 7.747960434976543e-06, + "loss": 1.2317, + "step": 20849 + }, + { + "epoch": 1.76, + "learning_rate": 7.747074821255886e-06, + "loss": 1.1362, + "step": 20850 + }, + { + "epoch": 1.76, + "learning_rate": 7.746189226149618e-06, + "loss": 1.1351, + "step": 20851 + }, + { + "epoch": 1.76, + "learning_rate": 7.74530364966506e-06, + "loss": 1.1774, + "step": 20852 + }, + { + "epoch": 1.76, + "learning_rate": 7.74441809180952e-06, + "loss": 1.165, + "step": 20853 + }, + { + "epoch": 1.76, + "learning_rate": 7.743532552590324e-06, + "loss": 1.1734, + "step": 20854 + }, + { + "epoch": 1.76, + "learning_rate": 7.742647032014782e-06, + "loss": 1.256, + "step": 20855 + }, + { + "epoch": 1.76, + "learning_rate": 7.741761530090214e-06, + "loss": 1.1799, + "step": 20856 + }, + { + "epoch": 1.76, + "learning_rate": 7.740876046823932e-06, + "loss": 1.1246, + "step": 20857 + }, + { + "epoch": 1.76, + "learning_rate": 7.739990582223258e-06, + "loss": 1.2859, + "step": 20858 + }, + { + "epoch": 1.76, + "learning_rate": 7.739105136295505e-06, + "loss": 1.1864, + "step": 20859 + }, + { + "epoch": 1.76, + "learning_rate": 7.738219709047989e-06, + "loss": 1.2953, + "step": 20860 + }, + { + "epoch": 1.76, + "learning_rate": 7.737334300488022e-06, + "loss": 1.1515, + "step": 20861 + }, + { + "epoch": 1.76, + "learning_rate": 7.736448910622925e-06, + "loss": 1.1453, + "step": 20862 + }, + { + "epoch": 1.76, + "learning_rate": 7.735563539460015e-06, + "loss": 1.1298, + "step": 20863 + }, + { + "epoch": 1.76, + "learning_rate": 7.734678187006598e-06, + "loss": 1.206, + "step": 20864 + }, + { + "epoch": 1.76, + "learning_rate": 7.733792853269998e-06, + "loss": 1.2109, + "step": 20865 + }, + { + "epoch": 1.76, + "learning_rate": 7.732907538257528e-06, + "loss": 1.0988, + "step": 20866 + }, + { + "epoch": 1.76, + "learning_rate": 7.732022241976503e-06, + "loss": 1.2061, + "step": 20867 + }, + { + "epoch": 1.76, + "learning_rate": 7.731136964434229e-06, + "loss": 1.2146, + "step": 20868 + }, + { + "epoch": 1.76, + "learning_rate": 7.730251705638034e-06, + "loss": 1.1284, + "step": 20869 + }, + { + "epoch": 1.76, + "learning_rate": 7.729366465595223e-06, + "loss": 1.1608, + "step": 20870 + }, + { + "epoch": 1.76, + "learning_rate": 7.728481244313115e-06, + "loss": 1.1864, + "step": 20871 + }, + { + "epoch": 1.76, + "learning_rate": 7.727596041799019e-06, + "loss": 1.1033, + "step": 20872 + }, + { + "epoch": 1.76, + "learning_rate": 7.726710858060255e-06, + "loss": 1.1399, + "step": 20873 + }, + { + "epoch": 1.76, + "learning_rate": 7.725825693104131e-06, + "loss": 1.2178, + "step": 20874 + }, + { + "epoch": 1.76, + "learning_rate": 7.72494054693796e-06, + "loss": 1.231, + "step": 20875 + }, + { + "epoch": 1.76, + "learning_rate": 7.724055419569064e-06, + "loss": 1.1903, + "step": 20876 + }, + { + "epoch": 1.76, + "learning_rate": 7.723170311004749e-06, + "loss": 1.1836, + "step": 20877 + }, + { + "epoch": 1.76, + "learning_rate": 7.72228522125233e-06, + "loss": 1.1337, + "step": 20878 + }, + { + "epoch": 1.76, + "learning_rate": 7.721400150319116e-06, + "loss": 1.2448, + "step": 20879 + }, + { + "epoch": 1.76, + "learning_rate": 7.720515098212428e-06, + "loss": 1.1403, + "step": 20880 + }, + { + "epoch": 1.76, + "learning_rate": 7.71963006493957e-06, + "loss": 1.2606, + "step": 20881 + }, + { + "epoch": 1.76, + "learning_rate": 7.718745050507862e-06, + "loss": 1.1463, + "step": 20882 + }, + { + "epoch": 1.76, + "learning_rate": 7.717860054924608e-06, + "loss": 1.206, + "step": 20883 + }, + { + "epoch": 1.76, + "learning_rate": 7.71697507819713e-06, + "loss": 1.2522, + "step": 20884 + }, + { + "epoch": 1.76, + "learning_rate": 7.71609012033273e-06, + "loss": 1.1626, + "step": 20885 + }, + { + "epoch": 1.76, + "learning_rate": 7.715205181338726e-06, + "loss": 1.1667, + "step": 20886 + }, + { + "epoch": 1.76, + "learning_rate": 7.714320261222428e-06, + "loss": 1.2562, + "step": 20887 + }, + { + "epoch": 1.76, + "learning_rate": 7.713435359991144e-06, + "loss": 1.1564, + "step": 20888 + }, + { + "epoch": 1.76, + "learning_rate": 7.712550477652194e-06, + "loss": 1.1685, + "step": 20889 + }, + { + "epoch": 1.76, + "learning_rate": 7.711665614212882e-06, + "loss": 1.1644, + "step": 20890 + }, + { + "epoch": 1.76, + "learning_rate": 7.710780769680522e-06, + "loss": 1.1979, + "step": 20891 + }, + { + "epoch": 1.76, + "learning_rate": 7.709895944062421e-06, + "loss": 1.1065, + "step": 20892 + }, + { + "epoch": 1.76, + "learning_rate": 7.709011137365896e-06, + "loss": 1.2261, + "step": 20893 + }, + { + "epoch": 1.76, + "learning_rate": 7.708126349598252e-06, + "loss": 1.1532, + "step": 20894 + }, + { + "epoch": 1.76, + "learning_rate": 7.707241580766806e-06, + "loss": 1.2337, + "step": 20895 + }, + { + "epoch": 1.76, + "learning_rate": 7.706356830878858e-06, + "loss": 1.1832, + "step": 20896 + }, + { + "epoch": 1.76, + "learning_rate": 7.705472099941727e-06, + "loss": 1.0824, + "step": 20897 + }, + { + "epoch": 1.76, + "learning_rate": 7.704587387962718e-06, + "loss": 1.2172, + "step": 20898 + }, + { + "epoch": 1.76, + "learning_rate": 7.703702694949146e-06, + "loss": 1.1437, + "step": 20899 + }, + { + "epoch": 1.76, + "learning_rate": 7.702818020908313e-06, + "loss": 1.1702, + "step": 20900 + }, + { + "epoch": 1.76, + "learning_rate": 7.701933365847535e-06, + "loss": 1.1549, + "step": 20901 + }, + { + "epoch": 1.76, + "learning_rate": 7.70104872977412e-06, + "loss": 1.2054, + "step": 20902 + }, + { + "epoch": 1.76, + "learning_rate": 7.700164112695372e-06, + "loss": 1.184, + "step": 20903 + }, + { + "epoch": 1.76, + "learning_rate": 7.69927951461861e-06, + "loss": 1.2528, + "step": 20904 + }, + { + "epoch": 1.76, + "learning_rate": 7.698394935551132e-06, + "loss": 1.1753, + "step": 20905 + }, + { + "epoch": 1.76, + "learning_rate": 7.697510375500257e-06, + "loss": 1.0898, + "step": 20906 + }, + { + "epoch": 1.76, + "learning_rate": 7.696625834473282e-06, + "loss": 1.1329, + "step": 20907 + }, + { + "epoch": 1.76, + "learning_rate": 7.695741312477526e-06, + "loss": 1.1923, + "step": 20908 + }, + { + "epoch": 1.76, + "learning_rate": 7.69485680952029e-06, + "loss": 1.1339, + "step": 20909 + }, + { + "epoch": 1.76, + "learning_rate": 7.693972325608887e-06, + "loss": 1.2429, + "step": 20910 + }, + { + "epoch": 1.76, + "learning_rate": 7.693087860750619e-06, + "loss": 1.2765, + "step": 20911 + }, + { + "epoch": 1.76, + "learning_rate": 7.6922034149528e-06, + "loss": 1.2224, + "step": 20912 + }, + { + "epoch": 1.76, + "learning_rate": 7.691318988222736e-06, + "loss": 1.0898, + "step": 20913 + }, + { + "epoch": 1.76, + "learning_rate": 7.69043458056773e-06, + "loss": 1.2279, + "step": 20914 + }, + { + "epoch": 1.76, + "learning_rate": 7.689550191995098e-06, + "loss": 1.1121, + "step": 20915 + }, + { + "epoch": 1.76, + "learning_rate": 7.688665822512139e-06, + "loss": 1.204, + "step": 20916 + }, + { + "epoch": 1.76, + "learning_rate": 7.687781472126164e-06, + "loss": 1.1318, + "step": 20917 + }, + { + "epoch": 1.76, + "learning_rate": 7.686897140844473e-06, + "loss": 1.2225, + "step": 20918 + }, + { + "epoch": 1.76, + "learning_rate": 7.686012828674384e-06, + "loss": 1.1386, + "step": 20919 + }, + { + "epoch": 1.76, + "learning_rate": 7.685128535623195e-06, + "loss": 1.1829, + "step": 20920 + }, + { + "epoch": 1.76, + "learning_rate": 7.684244261698217e-06, + "loss": 1.194, + "step": 20921 + }, + { + "epoch": 1.76, + "learning_rate": 7.683360006906751e-06, + "loss": 1.2469, + "step": 20922 + }, + { + "epoch": 1.76, + "learning_rate": 7.682475771256109e-06, + "loss": 1.2483, + "step": 20923 + }, + { + "epoch": 1.76, + "learning_rate": 7.681591554753592e-06, + "loss": 1.2199, + "step": 20924 + }, + { + "epoch": 1.76, + "learning_rate": 7.680707357406504e-06, + "loss": 1.1859, + "step": 20925 + }, + { + "epoch": 1.76, + "learning_rate": 7.679823179222162e-06, + "loss": 1.184, + "step": 20926 + }, + { + "epoch": 1.76, + "learning_rate": 7.678939020207858e-06, + "loss": 1.2138, + "step": 20927 + }, + { + "epoch": 1.76, + "learning_rate": 7.678054880370905e-06, + "loss": 1.2457, + "step": 20928 + }, + { + "epoch": 1.76, + "learning_rate": 7.677170759718604e-06, + "loss": 1.2226, + "step": 20929 + }, + { + "epoch": 1.76, + "learning_rate": 7.676286658258263e-06, + "loss": 1.194, + "step": 20930 + }, + { + "epoch": 1.76, + "learning_rate": 7.675402575997185e-06, + "loss": 1.1102, + "step": 20931 + }, + { + "epoch": 1.76, + "learning_rate": 7.674518512942675e-06, + "loss": 1.2123, + "step": 20932 + }, + { + "epoch": 1.76, + "learning_rate": 7.673634469102035e-06, + "loss": 1.0745, + "step": 20933 + }, + { + "epoch": 1.76, + "learning_rate": 7.672750444482576e-06, + "loss": 1.1555, + "step": 20934 + }, + { + "epoch": 1.76, + "learning_rate": 7.671866439091594e-06, + "loss": 1.1769, + "step": 20935 + }, + { + "epoch": 1.76, + "learning_rate": 7.6709824529364e-06, + "loss": 1.203, + "step": 20936 + }, + { + "epoch": 1.76, + "learning_rate": 7.67009848602429e-06, + "loss": 1.1444, + "step": 20937 + }, + { + "epoch": 1.76, + "learning_rate": 7.669214538362574e-06, + "loss": 1.2141, + "step": 20938 + }, + { + "epoch": 1.76, + "learning_rate": 7.668330609958554e-06, + "loss": 1.2072, + "step": 20939 + }, + { + "epoch": 1.76, + "learning_rate": 7.667446700819532e-06, + "loss": 1.1535, + "step": 20940 + }, + { + "epoch": 1.76, + "learning_rate": 7.666562810952813e-06, + "loss": 1.1245, + "step": 20941 + }, + { + "epoch": 1.76, + "learning_rate": 7.665678940365697e-06, + "loss": 1.2208, + "step": 20942 + }, + { + "epoch": 1.76, + "learning_rate": 7.66479508906549e-06, + "loss": 1.0926, + "step": 20943 + }, + { + "epoch": 1.76, + "learning_rate": 7.663911257059493e-06, + "loss": 1.1724, + "step": 20944 + }, + { + "epoch": 1.76, + "learning_rate": 7.66302744435501e-06, + "loss": 1.1871, + "step": 20945 + }, + { + "epoch": 1.76, + "learning_rate": 7.66214365095934e-06, + "loss": 1.178, + "step": 20946 + }, + { + "epoch": 1.76, + "learning_rate": 7.661259876879788e-06, + "loss": 1.1977, + "step": 20947 + }, + { + "epoch": 1.76, + "learning_rate": 7.660376122123657e-06, + "loss": 1.2041, + "step": 20948 + }, + { + "epoch": 1.76, + "learning_rate": 7.659492386698247e-06, + "loss": 1.1962, + "step": 20949 + }, + { + "epoch": 1.76, + "learning_rate": 7.658608670610857e-06, + "loss": 1.1478, + "step": 20950 + }, + { + "epoch": 1.76, + "learning_rate": 7.657724973868793e-06, + "loss": 1.0984, + "step": 20951 + }, + { + "epoch": 1.76, + "learning_rate": 7.656841296479356e-06, + "loss": 1.1711, + "step": 20952 + }, + { + "epoch": 1.76, + "learning_rate": 7.655957638449844e-06, + "loss": 1.1539, + "step": 20953 + }, + { + "epoch": 1.76, + "learning_rate": 7.655073999787562e-06, + "loss": 1.2475, + "step": 20954 + }, + { + "epoch": 1.76, + "learning_rate": 7.654190380499808e-06, + "loss": 1.1443, + "step": 20955 + }, + { + "epoch": 1.76, + "learning_rate": 7.653306780593888e-06, + "loss": 1.1388, + "step": 20956 + }, + { + "epoch": 1.76, + "learning_rate": 7.652423200077091e-06, + "loss": 1.1199, + "step": 20957 + }, + { + "epoch": 1.76, + "learning_rate": 7.651539638956732e-06, + "loss": 1.1264, + "step": 20958 + }, + { + "epoch": 1.76, + "learning_rate": 7.6506560972401e-06, + "loss": 1.1153, + "step": 20959 + }, + { + "epoch": 1.76, + "learning_rate": 7.649772574934502e-06, + "loss": 1.1628, + "step": 20960 + }, + { + "epoch": 1.76, + "learning_rate": 7.64888907204723e-06, + "loss": 1.1707, + "step": 20961 + }, + { + "epoch": 1.76, + "learning_rate": 7.648005588585593e-06, + "loss": 1.2325, + "step": 20962 + }, + { + "epoch": 1.76, + "learning_rate": 7.647122124556886e-06, + "loss": 1.2499, + "step": 20963 + }, + { + "epoch": 1.76, + "learning_rate": 7.646238679968406e-06, + "loss": 1.162, + "step": 20964 + }, + { + "epoch": 1.77, + "learning_rate": 7.64535525482746e-06, + "loss": 1.2863, + "step": 20965 + }, + { + "epoch": 1.77, + "learning_rate": 7.644471849141341e-06, + "loss": 1.148, + "step": 20966 + }, + { + "epoch": 1.77, + "learning_rate": 7.643588462917351e-06, + "loss": 1.1725, + "step": 20967 + }, + { + "epoch": 1.77, + "learning_rate": 7.642705096162782e-06, + "loss": 1.139, + "step": 20968 + }, + { + "epoch": 1.77, + "learning_rate": 7.641821748884944e-06, + "loss": 1.2396, + "step": 20969 + }, + { + "epoch": 1.77, + "learning_rate": 7.640938421091126e-06, + "loss": 1.2668, + "step": 20970 + }, + { + "epoch": 1.77, + "learning_rate": 7.64005511278863e-06, + "loss": 1.1182, + "step": 20971 + }, + { + "epoch": 1.77, + "learning_rate": 7.639171823984754e-06, + "loss": 1.1309, + "step": 20972 + }, + { + "epoch": 1.77, + "learning_rate": 7.638288554686796e-06, + "loss": 1.1872, + "step": 20973 + }, + { + "epoch": 1.77, + "learning_rate": 7.637405304902055e-06, + "loss": 1.2913, + "step": 20974 + }, + { + "epoch": 1.77, + "learning_rate": 7.636522074637825e-06, + "loss": 1.2476, + "step": 20975 + }, + { + "epoch": 1.77, + "learning_rate": 7.63563886390141e-06, + "loss": 1.1555, + "step": 20976 + }, + { + "epoch": 1.77, + "learning_rate": 7.6347556727001e-06, + "loss": 1.188, + "step": 20977 + }, + { + "epoch": 1.77, + "learning_rate": 7.633872501041198e-06, + "loss": 1.1671, + "step": 20978 + }, + { + "epoch": 1.77, + "learning_rate": 7.632989348931993e-06, + "loss": 1.1508, + "step": 20979 + }, + { + "epoch": 1.77, + "learning_rate": 7.632106216379795e-06, + "loss": 1.2089, + "step": 20980 + }, + { + "epoch": 1.77, + "learning_rate": 7.63122310339189e-06, + "loss": 1.1987, + "step": 20981 + }, + { + "epoch": 1.77, + "learning_rate": 7.630340009975578e-06, + "loss": 1.2362, + "step": 20982 + }, + { + "epoch": 1.77, + "learning_rate": 7.629456936138154e-06, + "loss": 1.1148, + "step": 20983 + }, + { + "epoch": 1.77, + "learning_rate": 7.628573881886919e-06, + "loss": 1.2477, + "step": 20984 + }, + { + "epoch": 1.77, + "learning_rate": 7.627690847229161e-06, + "loss": 1.14, + "step": 20985 + }, + { + "epoch": 1.77, + "learning_rate": 7.626807832172185e-06, + "loss": 1.1427, + "step": 20986 + }, + { + "epoch": 1.77, + "learning_rate": 7.625924836723277e-06, + "loss": 1.1584, + "step": 20987 + }, + { + "epoch": 1.77, + "learning_rate": 7.62504186088974e-06, + "loss": 1.2009, + "step": 20988 + }, + { + "epoch": 1.77, + "learning_rate": 7.62415890467887e-06, + "loss": 1.1469, + "step": 20989 + }, + { + "epoch": 1.77, + "learning_rate": 7.623275968097955e-06, + "loss": 1.2491, + "step": 20990 + }, + { + "epoch": 1.77, + "learning_rate": 7.622393051154301e-06, + "loss": 1.171, + "step": 20991 + }, + { + "epoch": 1.77, + "learning_rate": 7.6215101538551895e-06, + "loss": 1.1645, + "step": 20992 + }, + { + "epoch": 1.77, + "learning_rate": 7.620627276207927e-06, + "loss": 1.2348, + "step": 20993 + }, + { + "epoch": 1.77, + "learning_rate": 7.619744418219801e-06, + "loss": 1.0616, + "step": 20994 + }, + { + "epoch": 1.77, + "learning_rate": 7.618861579898111e-06, + "loss": 1.2705, + "step": 20995 + }, + { + "epoch": 1.77, + "learning_rate": 7.617978761250144e-06, + "loss": 1.2092, + "step": 20996 + }, + { + "epoch": 1.77, + "learning_rate": 7.6170959622832045e-06, + "loss": 1.1515, + "step": 20997 + }, + { + "epoch": 1.77, + "learning_rate": 7.6162131830045775e-06, + "loss": 1.1154, + "step": 20998 + }, + { + "epoch": 1.77, + "learning_rate": 7.615330423421562e-06, + "loss": 1.198, + "step": 20999 + }, + { + "epoch": 1.77, + "learning_rate": 7.614447683541446e-06, + "loss": 1.2482, + "step": 21000 + }, + { + "epoch": 1.77, + "learning_rate": 7.6135649633715285e-06, + "loss": 1.2454, + "step": 21001 + }, + { + "epoch": 1.77, + "learning_rate": 7.612682262919103e-06, + "loss": 1.1602, + "step": 21002 + }, + { + "epoch": 1.77, + "learning_rate": 7.611799582191457e-06, + "loss": 1.3202, + "step": 21003 + }, + { + "epoch": 1.77, + "learning_rate": 7.610916921195891e-06, + "loss": 1.0478, + "step": 21004 + }, + { + "epoch": 1.77, + "learning_rate": 7.610034279939692e-06, + "loss": 1.1881, + "step": 21005 + }, + { + "epoch": 1.77, + "learning_rate": 7.609151658430156e-06, + "loss": 1.2343, + "step": 21006 + }, + { + "epoch": 1.77, + "learning_rate": 7.608269056674571e-06, + "loss": 1.2253, + "step": 21007 + }, + { + "epoch": 1.77, + "learning_rate": 7.607386474680236e-06, + "loss": 1.1941, + "step": 21008 + }, + { + "epoch": 1.77, + "learning_rate": 7.606503912454438e-06, + "loss": 1.116, + "step": 21009 + }, + { + "epoch": 1.77, + "learning_rate": 7.605621370004473e-06, + "loss": 1.1178, + "step": 21010 + }, + { + "epoch": 1.77, + "learning_rate": 7.604738847337626e-06, + "loss": 1.0832, + "step": 21011 + }, + { + "epoch": 1.77, + "learning_rate": 7.603856344461196e-06, + "loss": 1.1387, + "step": 21012 + }, + { + "epoch": 1.77, + "learning_rate": 7.602973861382473e-06, + "loss": 1.213, + "step": 21013 + }, + { + "epoch": 1.77, + "learning_rate": 7.602091398108743e-06, + "loss": 1.2353, + "step": 21014 + }, + { + "epoch": 1.77, + "learning_rate": 7.601208954647305e-06, + "loss": 1.218, + "step": 21015 + }, + { + "epoch": 1.77, + "learning_rate": 7.600326531005446e-06, + "loss": 1.1246, + "step": 21016 + }, + { + "epoch": 1.77, + "learning_rate": 7.599444127190458e-06, + "loss": 1.2437, + "step": 21017 + }, + { + "epoch": 1.77, + "learning_rate": 7.598561743209627e-06, + "loss": 1.1573, + "step": 21018 + }, + { + "epoch": 1.77, + "learning_rate": 7.597679379070252e-06, + "loss": 1.1716, + "step": 21019 + }, + { + "epoch": 1.77, + "learning_rate": 7.596797034779616e-06, + "loss": 1.1782, + "step": 21020 + }, + { + "epoch": 1.77, + "learning_rate": 7.595914710345014e-06, + "loss": 1.1869, + "step": 21021 + }, + { + "epoch": 1.77, + "learning_rate": 7.5950324057737305e-06, + "loss": 1.1836, + "step": 21022 + }, + { + "epoch": 1.77, + "learning_rate": 7.594150121073063e-06, + "loss": 1.2254, + "step": 21023 + }, + { + "epoch": 1.77, + "learning_rate": 7.593267856250295e-06, + "loss": 1.1277, + "step": 21024 + }, + { + "epoch": 1.77, + "learning_rate": 7.592385611312716e-06, + "loss": 1.1935, + "step": 21025 + }, + { + "epoch": 1.77, + "learning_rate": 7.591503386267622e-06, + "loss": 1.2068, + "step": 21026 + }, + { + "epoch": 1.77, + "learning_rate": 7.590621181122295e-06, + "loss": 1.1885, + "step": 21027 + }, + { + "epoch": 1.77, + "learning_rate": 7.589738995884031e-06, + "loss": 1.183, + "step": 21028 + }, + { + "epoch": 1.77, + "learning_rate": 7.588856830560109e-06, + "loss": 1.2035, + "step": 21029 + }, + { + "epoch": 1.77, + "learning_rate": 7.587974685157827e-06, + "loss": 1.2052, + "step": 21030 + }, + { + "epoch": 1.77, + "learning_rate": 7.587092559684469e-06, + "loss": 1.1905, + "step": 21031 + }, + { + "epoch": 1.77, + "learning_rate": 7.586210454147327e-06, + "loss": 1.1763, + "step": 21032 + }, + { + "epoch": 1.77, + "learning_rate": 7.585328368553681e-06, + "loss": 1.253, + "step": 21033 + }, + { + "epoch": 1.77, + "learning_rate": 7.584446302910832e-06, + "loss": 1.1993, + "step": 21034 + }, + { + "epoch": 1.77, + "learning_rate": 7.583564257226056e-06, + "loss": 1.1581, + "step": 21035 + }, + { + "epoch": 1.77, + "learning_rate": 7.582682231506648e-06, + "loss": 1.1466, + "step": 21036 + }, + { + "epoch": 1.77, + "learning_rate": 7.58180022575989e-06, + "loss": 1.1156, + "step": 21037 + }, + { + "epoch": 1.77, + "learning_rate": 7.580918239993074e-06, + "loss": 1.2215, + "step": 21038 + }, + { + "epoch": 1.77, + "learning_rate": 7.580036274213487e-06, + "loss": 1.126, + "step": 21039 + }, + { + "epoch": 1.77, + "learning_rate": 7.579154328428411e-06, + "loss": 1.1912, + "step": 21040 + }, + { + "epoch": 1.77, + "learning_rate": 7.578272402645144e-06, + "loss": 1.2761, + "step": 21041 + }, + { + "epoch": 1.77, + "learning_rate": 7.577390496870958e-06, + "loss": 1.1454, + "step": 21042 + }, + { + "epoch": 1.77, + "learning_rate": 7.576508611113151e-06, + "loss": 1.194, + "step": 21043 + }, + { + "epoch": 1.77, + "learning_rate": 7.575626745379003e-06, + "loss": 1.2585, + "step": 21044 + }, + { + "epoch": 1.77, + "learning_rate": 7.574744899675807e-06, + "loss": 1.148, + "step": 21045 + }, + { + "epoch": 1.77, + "learning_rate": 7.573863074010839e-06, + "loss": 1.1366, + "step": 21046 + }, + { + "epoch": 1.77, + "learning_rate": 7.572981268391395e-06, + "loss": 1.1312, + "step": 21047 + }, + { + "epoch": 1.77, + "learning_rate": 7.5720994828247555e-06, + "loss": 1.2217, + "step": 21048 + }, + { + "epoch": 1.77, + "learning_rate": 7.571217717318203e-06, + "loss": 1.1808, + "step": 21049 + }, + { + "epoch": 1.77, + "learning_rate": 7.5703359718790325e-06, + "loss": 1.2049, + "step": 21050 + }, + { + "epoch": 1.77, + "learning_rate": 7.569454246514523e-06, + "loss": 1.2377, + "step": 21051 + }, + { + "epoch": 1.77, + "learning_rate": 7.56857254123196e-06, + "loss": 1.1496, + "step": 21052 + }, + { + "epoch": 1.77, + "learning_rate": 7.567690856038627e-06, + "loss": 1.2393, + "step": 21053 + }, + { + "epoch": 1.77, + "learning_rate": 7.566809190941813e-06, + "loss": 1.227, + "step": 21054 + }, + { + "epoch": 1.77, + "learning_rate": 7.565927545948798e-06, + "loss": 1.24, + "step": 21055 + }, + { + "epoch": 1.77, + "learning_rate": 7.565045921066871e-06, + "loss": 1.2597, + "step": 21056 + }, + { + "epoch": 1.77, + "learning_rate": 7.564164316303311e-06, + "loss": 1.2091, + "step": 21057 + }, + { + "epoch": 1.77, + "learning_rate": 7.563282731665408e-06, + "loss": 1.1702, + "step": 21058 + }, + { + "epoch": 1.77, + "learning_rate": 7.56240116716044e-06, + "loss": 1.231, + "step": 21059 + }, + { + "epoch": 1.77, + "learning_rate": 7.5615196227956974e-06, + "loss": 1.1868, + "step": 21060 + }, + { + "epoch": 1.77, + "learning_rate": 7.560638098578456e-06, + "loss": 1.1489, + "step": 21061 + }, + { + "epoch": 1.77, + "learning_rate": 7.559756594516005e-06, + "loss": 1.144, + "step": 21062 + }, + { + "epoch": 1.77, + "learning_rate": 7.558875110615626e-06, + "loss": 1.1204, + "step": 21063 + }, + { + "epoch": 1.77, + "learning_rate": 7.5579936468846e-06, + "loss": 1.1251, + "step": 21064 + }, + { + "epoch": 1.77, + "learning_rate": 7.557112203330217e-06, + "loss": 1.0699, + "step": 21065 + }, + { + "epoch": 1.77, + "learning_rate": 7.5562307799597516e-06, + "loss": 1.2109, + "step": 21066 + }, + { + "epoch": 1.77, + "learning_rate": 7.555349376780492e-06, + "loss": 1.246, + "step": 21067 + }, + { + "epoch": 1.77, + "learning_rate": 7.554467993799713e-06, + "loss": 1.2438, + "step": 21068 + }, + { + "epoch": 1.77, + "learning_rate": 7.553586631024707e-06, + "loss": 1.2114, + "step": 21069 + }, + { + "epoch": 1.77, + "learning_rate": 7.55270528846275e-06, + "loss": 1.2034, + "step": 21070 + }, + { + "epoch": 1.77, + "learning_rate": 7.551823966121126e-06, + "loss": 1.2006, + "step": 21071 + }, + { + "epoch": 1.77, + "learning_rate": 7.550942664007112e-06, + "loss": 1.1615, + "step": 21072 + }, + { + "epoch": 1.77, + "learning_rate": 7.5500613821279975e-06, + "loss": 1.1551, + "step": 21073 + }, + { + "epoch": 1.77, + "learning_rate": 7.549180120491058e-06, + "loss": 1.247, + "step": 21074 + }, + { + "epoch": 1.77, + "learning_rate": 7.548298879103576e-06, + "loss": 1.1619, + "step": 21075 + }, + { + "epoch": 1.77, + "learning_rate": 7.547417657972837e-06, + "loss": 1.126, + "step": 21076 + }, + { + "epoch": 1.77, + "learning_rate": 7.546536457106115e-06, + "loss": 1.174, + "step": 21077 + }, + { + "epoch": 1.77, + "learning_rate": 7.545655276510696e-06, + "loss": 1.2397, + "step": 21078 + }, + { + "epoch": 1.77, + "learning_rate": 7.544774116193855e-06, + "loss": 1.224, + "step": 21079 + }, + { + "epoch": 1.77, + "learning_rate": 7.54389297616288e-06, + "loss": 1.2411, + "step": 21080 + }, + { + "epoch": 1.77, + "learning_rate": 7.543011856425046e-06, + "loss": 1.222, + "step": 21081 + }, + { + "epoch": 1.77, + "learning_rate": 7.542130756987636e-06, + "loss": 1.2548, + "step": 21082 + }, + { + "epoch": 1.78, + "learning_rate": 7.541249677857924e-06, + "loss": 1.2107, + "step": 21083 + }, + { + "epoch": 1.78, + "learning_rate": 7.540368619043198e-06, + "loss": 1.2325, + "step": 21084 + }, + { + "epoch": 1.78, + "learning_rate": 7.539487580550731e-06, + "loss": 1.2395, + "step": 21085 + }, + { + "epoch": 1.78, + "learning_rate": 7.538606562387808e-06, + "loss": 1.2587, + "step": 21086 + }, + { + "epoch": 1.78, + "learning_rate": 7.537725564561701e-06, + "loss": 1.2228, + "step": 21087 + }, + { + "epoch": 1.78, + "learning_rate": 7.536844587079696e-06, + "loss": 1.1229, + "step": 21088 + }, + { + "epoch": 1.78, + "learning_rate": 7.535963629949071e-06, + "loss": 1.2299, + "step": 21089 + }, + { + "epoch": 1.78, + "learning_rate": 7.535082693177098e-06, + "loss": 1.0815, + "step": 21090 + }, + { + "epoch": 1.78, + "learning_rate": 7.534201776771069e-06, + "loss": 1.2222, + "step": 21091 + }, + { + "epoch": 1.78, + "learning_rate": 7.533320880738247e-06, + "loss": 1.2622, + "step": 21092 + }, + { + "epoch": 1.78, + "learning_rate": 7.532440005085921e-06, + "loss": 1.2278, + "step": 21093 + }, + { + "epoch": 1.78, + "learning_rate": 7.5315591498213636e-06, + "loss": 1.1923, + "step": 21094 + }, + { + "epoch": 1.78, + "learning_rate": 7.5306783149518555e-06, + "loss": 1.1759, + "step": 21095 + }, + { + "epoch": 1.78, + "learning_rate": 7.52979750048467e-06, + "loss": 1.1472, + "step": 21096 + }, + { + "epoch": 1.78, + "learning_rate": 7.528916706427093e-06, + "loss": 1.258, + "step": 21097 + }, + { + "epoch": 1.78, + "learning_rate": 7.528035932786395e-06, + "loss": 1.1406, + "step": 21098 + }, + { + "epoch": 1.78, + "learning_rate": 7.527155179569851e-06, + "loss": 1.1409, + "step": 21099 + }, + { + "epoch": 1.78, + "learning_rate": 7.526274446784749e-06, + "loss": 1.1839, + "step": 21100 + }, + { + "epoch": 1.78, + "learning_rate": 7.525393734438357e-06, + "loss": 1.112, + "step": 21101 + }, + { + "epoch": 1.78, + "learning_rate": 7.524513042537955e-06, + "loss": 1.219, + "step": 21102 + }, + { + "epoch": 1.78, + "learning_rate": 7.523632371090815e-06, + "loss": 1.1755, + "step": 21103 + }, + { + "epoch": 1.78, + "learning_rate": 7.52275172010422e-06, + "loss": 1.1693, + "step": 21104 + }, + { + "epoch": 1.78, + "learning_rate": 7.521871089585442e-06, + "loss": 1.1895, + "step": 21105 + }, + { + "epoch": 1.78, + "learning_rate": 7.52099047954176e-06, + "loss": 1.1585, + "step": 21106 + }, + { + "epoch": 1.78, + "learning_rate": 7.520109889980442e-06, + "loss": 1.084, + "step": 21107 + }, + { + "epoch": 1.78, + "learning_rate": 7.519229320908776e-06, + "loss": 1.2751, + "step": 21108 + }, + { + "epoch": 1.78, + "learning_rate": 7.51834877233403e-06, + "loss": 1.1847, + "step": 21109 + }, + { + "epoch": 1.78, + "learning_rate": 7.5174682442634815e-06, + "loss": 1.2646, + "step": 21110 + }, + { + "epoch": 1.78, + "learning_rate": 7.516587736704402e-06, + "loss": 1.2376, + "step": 21111 + }, + { + "epoch": 1.78, + "learning_rate": 7.515707249664069e-06, + "loss": 1.1521, + "step": 21112 + }, + { + "epoch": 1.78, + "learning_rate": 7.514826783149762e-06, + "loss": 1.0886, + "step": 21113 + }, + { + "epoch": 1.78, + "learning_rate": 7.513946337168746e-06, + "loss": 1.1916, + "step": 21114 + }, + { + "epoch": 1.78, + "learning_rate": 7.513065911728306e-06, + "loss": 1.0866, + "step": 21115 + }, + { + "epoch": 1.78, + "learning_rate": 7.5121855068357085e-06, + "loss": 1.075, + "step": 21116 + }, + { + "epoch": 1.78, + "learning_rate": 7.511305122498233e-06, + "loss": 1.1861, + "step": 21117 + }, + { + "epoch": 1.78, + "learning_rate": 7.510424758723147e-06, + "loss": 1.1229, + "step": 21118 + }, + { + "epoch": 1.78, + "learning_rate": 7.509544415517733e-06, + "loss": 1.2436, + "step": 21119 + }, + { + "epoch": 1.78, + "learning_rate": 7.508664092889257e-06, + "loss": 1.1064, + "step": 21120 + }, + { + "epoch": 1.78, + "learning_rate": 7.5077837908449994e-06, + "loss": 1.2336, + "step": 21121 + }, + { + "epoch": 1.78, + "learning_rate": 7.506903509392224e-06, + "loss": 1.1543, + "step": 21122 + }, + { + "epoch": 1.78, + "learning_rate": 7.506023248538214e-06, + "loss": 1.1759, + "step": 21123 + }, + { + "epoch": 1.78, + "learning_rate": 7.505143008290236e-06, + "loss": 1.199, + "step": 21124 + }, + { + "epoch": 1.78, + "learning_rate": 7.504262788655564e-06, + "loss": 1.1962, + "step": 21125 + }, + { + "epoch": 1.78, + "learning_rate": 7.5033825896414744e-06, + "loss": 1.1815, + "step": 21126 + }, + { + "epoch": 1.78, + "learning_rate": 7.502502411255236e-06, + "loss": 1.2504, + "step": 21127 + }, + { + "epoch": 1.78, + "learning_rate": 7.501622253504124e-06, + "loss": 1.1685, + "step": 21128 + }, + { + "epoch": 1.78, + "learning_rate": 7.500742116395403e-06, + "loss": 1.2114, + "step": 21129 + }, + { + "epoch": 1.78, + "learning_rate": 7.499861999936355e-06, + "loss": 1.1899, + "step": 21130 + }, + { + "epoch": 1.78, + "learning_rate": 7.498981904134246e-06, + "loss": 1.0936, + "step": 21131 + }, + { + "epoch": 1.78, + "learning_rate": 7.49810182899635e-06, + "loss": 1.2073, + "step": 21132 + }, + { + "epoch": 1.78, + "learning_rate": 7.497221774529935e-06, + "loss": 1.1498, + "step": 21133 + }, + { + "epoch": 1.78, + "learning_rate": 7.496341740742277e-06, + "loss": 1.229, + "step": 21134 + }, + { + "epoch": 1.78, + "learning_rate": 7.495461727640643e-06, + "loss": 1.1808, + "step": 21135 + }, + { + "epoch": 1.78, + "learning_rate": 7.4945817352323095e-06, + "loss": 1.3255, + "step": 21136 + }, + { + "epoch": 1.78, + "learning_rate": 7.493701763524537e-06, + "loss": 1.2346, + "step": 21137 + }, + { + "epoch": 1.78, + "learning_rate": 7.492821812524607e-06, + "loss": 1.2359, + "step": 21138 + }, + { + "epoch": 1.78, + "learning_rate": 7.491941882239786e-06, + "loss": 1.2614, + "step": 21139 + }, + { + "epoch": 1.78, + "learning_rate": 7.491061972677341e-06, + "loss": 1.0892, + "step": 21140 + }, + { + "epoch": 1.78, + "learning_rate": 7.490182083844548e-06, + "loss": 1.2658, + "step": 21141 + }, + { + "epoch": 1.78, + "learning_rate": 7.489302215748672e-06, + "loss": 1.1624, + "step": 21142 + }, + { + "epoch": 1.78, + "learning_rate": 7.488422368396986e-06, + "loss": 1.1557, + "step": 21143 + }, + { + "epoch": 1.78, + "learning_rate": 7.487542541796755e-06, + "loss": 1.1181, + "step": 21144 + }, + { + "epoch": 1.78, + "learning_rate": 7.486662735955258e-06, + "loss": 1.2097, + "step": 21145 + }, + { + "epoch": 1.78, + "learning_rate": 7.485782950879752e-06, + "loss": 1.2844, + "step": 21146 + }, + { + "epoch": 1.78, + "learning_rate": 7.484903186577514e-06, + "loss": 1.1654, + "step": 21147 + }, + { + "epoch": 1.78, + "learning_rate": 7.484023443055811e-06, + "loss": 1.1831, + "step": 21148 + }, + { + "epoch": 1.78, + "learning_rate": 7.483143720321908e-06, + "loss": 1.1452, + "step": 21149 + }, + { + "epoch": 1.78, + "learning_rate": 7.482264018383083e-06, + "loss": 1.1572, + "step": 21150 + }, + { + "epoch": 1.78, + "learning_rate": 7.4813843372465945e-06, + "loss": 1.1826, + "step": 21151 + }, + { + "epoch": 1.78, + "learning_rate": 7.480504676919718e-06, + "loss": 1.2097, + "step": 21152 + }, + { + "epoch": 1.78, + "learning_rate": 7.479625037409712e-06, + "loss": 1.0624, + "step": 21153 + }, + { + "epoch": 1.78, + "learning_rate": 7.478745418723858e-06, + "loss": 1.121, + "step": 21154 + }, + { + "epoch": 1.78, + "learning_rate": 7.4778658208694125e-06, + "loss": 1.1756, + "step": 21155 + }, + { + "epoch": 1.78, + "learning_rate": 7.476986243853648e-06, + "loss": 1.1476, + "step": 21156 + }, + { + "epoch": 1.78, + "learning_rate": 7.476106687683826e-06, + "loss": 1.1291, + "step": 21157 + }, + { + "epoch": 1.78, + "learning_rate": 7.475227152367225e-06, + "loss": 1.0623, + "step": 21158 + }, + { + "epoch": 1.78, + "learning_rate": 7.474347637911101e-06, + "loss": 1.1753, + "step": 21159 + }, + { + "epoch": 1.78, + "learning_rate": 7.473468144322728e-06, + "loss": 1.1958, + "step": 21160 + }, + { + "epoch": 1.78, + "learning_rate": 7.472588671609367e-06, + "loss": 1.2009, + "step": 21161 + }, + { + "epoch": 1.78, + "learning_rate": 7.471709219778287e-06, + "loss": 1.2104, + "step": 21162 + }, + { + "epoch": 1.78, + "learning_rate": 7.470829788836757e-06, + "loss": 1.2323, + "step": 21163 + }, + { + "epoch": 1.78, + "learning_rate": 7.469950378792039e-06, + "loss": 1.1121, + "step": 21164 + }, + { + "epoch": 1.78, + "learning_rate": 7.469070989651402e-06, + "loss": 1.0694, + "step": 21165 + }, + { + "epoch": 1.78, + "learning_rate": 7.4681916214221095e-06, + "loss": 1.1667, + "step": 21166 + }, + { + "epoch": 1.78, + "learning_rate": 7.467312274111431e-06, + "loss": 1.1654, + "step": 21167 + }, + { + "epoch": 1.78, + "learning_rate": 7.466432947726624e-06, + "loss": 1.1876, + "step": 21168 + }, + { + "epoch": 1.78, + "learning_rate": 7.465553642274963e-06, + "loss": 1.2208, + "step": 21169 + }, + { + "epoch": 1.78, + "learning_rate": 7.464674357763707e-06, + "loss": 1.0642, + "step": 21170 + }, + { + "epoch": 1.78, + "learning_rate": 7.463795094200125e-06, + "loss": 1.2015, + "step": 21171 + }, + { + "epoch": 1.78, + "learning_rate": 7.462915851591475e-06, + "loss": 1.183, + "step": 21172 + }, + { + "epoch": 1.78, + "learning_rate": 7.462036629945031e-06, + "loss": 1.1987, + "step": 21173 + }, + { + "epoch": 1.78, + "learning_rate": 7.4611574292680515e-06, + "loss": 1.1736, + "step": 21174 + }, + { + "epoch": 1.78, + "learning_rate": 7.460278249567797e-06, + "loss": 1.1713, + "step": 21175 + }, + { + "epoch": 1.78, + "learning_rate": 7.4593990908515425e-06, + "loss": 1.1979, + "step": 21176 + }, + { + "epoch": 1.78, + "learning_rate": 7.458519953126544e-06, + "loss": 1.1826, + "step": 21177 + }, + { + "epoch": 1.78, + "learning_rate": 7.457640836400069e-06, + "loss": 1.224, + "step": 21178 + }, + { + "epoch": 1.78, + "learning_rate": 7.456761740679372e-06, + "loss": 1.2332, + "step": 21179 + }, + { + "epoch": 1.78, + "learning_rate": 7.455882665971731e-06, + "loss": 1.2456, + "step": 21180 + }, + { + "epoch": 1.78, + "learning_rate": 7.4550036122843995e-06, + "loss": 1.1346, + "step": 21181 + }, + { + "epoch": 1.78, + "learning_rate": 7.454124579624643e-06, + "loss": 1.0864, + "step": 21182 + }, + { + "epoch": 1.78, + "learning_rate": 7.453245567999721e-06, + "loss": 1.1534, + "step": 21183 + }, + { + "epoch": 1.78, + "learning_rate": 7.452366577416903e-06, + "loss": 1.1362, + "step": 21184 + }, + { + "epoch": 1.78, + "learning_rate": 7.451487607883445e-06, + "loss": 1.1841, + "step": 21185 + }, + { + "epoch": 1.78, + "learning_rate": 7.450608659406615e-06, + "loss": 1.2445, + "step": 21186 + }, + { + "epoch": 1.78, + "learning_rate": 7.449729731993667e-06, + "loss": 1.2051, + "step": 21187 + }, + { + "epoch": 1.78, + "learning_rate": 7.448850825651871e-06, + "loss": 1.2521, + "step": 21188 + }, + { + "epoch": 1.78, + "learning_rate": 7.447971940388487e-06, + "loss": 1.181, + "step": 21189 + }, + { + "epoch": 1.78, + "learning_rate": 7.447093076210774e-06, + "loss": 1.207, + "step": 21190 + }, + { + "epoch": 1.78, + "learning_rate": 7.446214233125996e-06, + "loss": 1.1777, + "step": 21191 + }, + { + "epoch": 1.78, + "learning_rate": 7.445335411141414e-06, + "loss": 1.1807, + "step": 21192 + }, + { + "epoch": 1.78, + "learning_rate": 7.444456610264289e-06, + "loss": 1.2461, + "step": 21193 + }, + { + "epoch": 1.78, + "learning_rate": 7.443577830501878e-06, + "loss": 1.2205, + "step": 21194 + }, + { + "epoch": 1.78, + "learning_rate": 7.44269907186145e-06, + "loss": 1.1973, + "step": 21195 + }, + { + "epoch": 1.78, + "learning_rate": 7.441820334350256e-06, + "loss": 1.1376, + "step": 21196 + }, + { + "epoch": 1.78, + "learning_rate": 7.440941617975564e-06, + "loss": 1.2064, + "step": 21197 + }, + { + "epoch": 1.78, + "learning_rate": 7.440062922744631e-06, + "loss": 1.2277, + "step": 21198 + }, + { + "epoch": 1.78, + "learning_rate": 7.439184248664713e-06, + "loss": 1.171, + "step": 21199 + }, + { + "epoch": 1.78, + "learning_rate": 7.4383055957430805e-06, + "loss": 1.1331, + "step": 21200 + }, + { + "epoch": 1.78, + "learning_rate": 7.437426963986984e-06, + "loss": 1.038, + "step": 21201 + }, + { + "epoch": 1.79, + "learning_rate": 7.436548353403688e-06, + "loss": 1.1751, + "step": 21202 + }, + { + "epoch": 1.79, + "learning_rate": 7.435669764000446e-06, + "loss": 1.2455, + "step": 21203 + }, + { + "epoch": 1.79, + "learning_rate": 7.434791195784525e-06, + "loss": 1.1285, + "step": 21204 + }, + { + "epoch": 1.79, + "learning_rate": 7.433912648763179e-06, + "loss": 1.1894, + "step": 21205 + }, + { + "epoch": 1.79, + "learning_rate": 7.433034122943669e-06, + "loss": 1.1764, + "step": 21206 + }, + { + "epoch": 1.79, + "learning_rate": 7.432155618333248e-06, + "loss": 1.1488, + "step": 21207 + }, + { + "epoch": 1.79, + "learning_rate": 7.431277134939184e-06, + "loss": 1.151, + "step": 21208 + }, + { + "epoch": 1.79, + "learning_rate": 7.4303986727687284e-06, + "loss": 1.2041, + "step": 21209 + }, + { + "epoch": 1.79, + "learning_rate": 7.429520231829144e-06, + "loss": 1.0761, + "step": 21210 + }, + { + "epoch": 1.79, + "learning_rate": 7.428641812127682e-06, + "loss": 1.2067, + "step": 21211 + }, + { + "epoch": 1.79, + "learning_rate": 7.427763413671604e-06, + "loss": 1.2081, + "step": 21212 + }, + { + "epoch": 1.79, + "learning_rate": 7.426885036468172e-06, + "loss": 1.2534, + "step": 21213 + }, + { + "epoch": 1.79, + "learning_rate": 7.4260066805246355e-06, + "loss": 1.2015, + "step": 21214 + }, + { + "epoch": 1.79, + "learning_rate": 7.4251283458482595e-06, + "loss": 1.1267, + "step": 21215 + }, + { + "epoch": 1.79, + "learning_rate": 7.424250032446294e-06, + "loss": 1.1639, + "step": 21216 + }, + { + "epoch": 1.79, + "learning_rate": 7.423371740326002e-06, + "loss": 1.2565, + "step": 21217 + }, + { + "epoch": 1.79, + "learning_rate": 7.422493469494633e-06, + "loss": 1.1842, + "step": 21218 + }, + { + "epoch": 1.79, + "learning_rate": 7.4216152199594525e-06, + "loss": 1.1869, + "step": 21219 + }, + { + "epoch": 1.79, + "learning_rate": 7.420736991727708e-06, + "loss": 1.2034, + "step": 21220 + }, + { + "epoch": 1.79, + "learning_rate": 7.419858784806665e-06, + "loss": 1.148, + "step": 21221 + }, + { + "epoch": 1.79, + "learning_rate": 7.418980599203569e-06, + "loss": 1.2634, + "step": 21222 + }, + { + "epoch": 1.79, + "learning_rate": 7.4181024349256845e-06, + "loss": 1.2797, + "step": 21223 + }, + { + "epoch": 1.79, + "learning_rate": 7.417224291980264e-06, + "loss": 1.197, + "step": 21224 + }, + { + "epoch": 1.79, + "learning_rate": 7.41634617037456e-06, + "loss": 1.1751, + "step": 21225 + }, + { + "epoch": 1.79, + "learning_rate": 7.415468070115834e-06, + "loss": 1.1812, + "step": 21226 + }, + { + "epoch": 1.79, + "learning_rate": 7.414589991211337e-06, + "loss": 1.1387, + "step": 21227 + }, + { + "epoch": 1.79, + "learning_rate": 7.413711933668326e-06, + "loss": 1.1888, + "step": 21228 + }, + { + "epoch": 1.79, + "learning_rate": 7.4128338974940515e-06, + "loss": 1.1727, + "step": 21229 + }, + { + "epoch": 1.79, + "learning_rate": 7.411955882695776e-06, + "loss": 1.1705, + "step": 21230 + }, + { + "epoch": 1.79, + "learning_rate": 7.411077889280746e-06, + "loss": 1.1051, + "step": 21231 + }, + { + "epoch": 1.79, + "learning_rate": 7.410199917256223e-06, + "loss": 1.1909, + "step": 21232 + }, + { + "epoch": 1.79, + "learning_rate": 7.40932196662945e-06, + "loss": 1.2125, + "step": 21233 + }, + { + "epoch": 1.79, + "learning_rate": 7.4084440374076935e-06, + "loss": 1.2944, + "step": 21234 + }, + { + "epoch": 1.79, + "learning_rate": 7.4075661295982e-06, + "loss": 1.2197, + "step": 21235 + }, + { + "epoch": 1.79, + "learning_rate": 7.4066882432082265e-06, + "loss": 1.2257, + "step": 21236 + }, + { + "epoch": 1.79, + "learning_rate": 7.405810378245022e-06, + "loss": 1.2254, + "step": 21237 + }, + { + "epoch": 1.79, + "learning_rate": 7.404932534715843e-06, + "loss": 1.1305, + "step": 21238 + }, + { + "epoch": 1.79, + "learning_rate": 7.404054712627944e-06, + "loss": 1.1723, + "step": 21239 + }, + { + "epoch": 1.79, + "learning_rate": 7.4031769119885725e-06, + "loss": 1.0913, + "step": 21240 + }, + { + "epoch": 1.79, + "learning_rate": 7.402299132804987e-06, + "loss": 1.2477, + "step": 21241 + }, + { + "epoch": 1.79, + "learning_rate": 7.401421375084437e-06, + "loss": 1.1602, + "step": 21242 + }, + { + "epoch": 1.79, + "learning_rate": 7.4005436388341765e-06, + "loss": 1.1728, + "step": 21243 + }, + { + "epoch": 1.79, + "learning_rate": 7.399665924061452e-06, + "loss": 1.2116, + "step": 21244 + }, + { + "epoch": 1.79, + "learning_rate": 7.398788230773528e-06, + "loss": 1.269, + "step": 21245 + }, + { + "epoch": 1.79, + "learning_rate": 7.397910558977639e-06, + "loss": 1.2591, + "step": 21246 + }, + { + "epoch": 1.79, + "learning_rate": 7.397032908681052e-06, + "loss": 1.0934, + "step": 21247 + }, + { + "epoch": 1.79, + "learning_rate": 7.39615527989101e-06, + "loss": 1.219, + "step": 21248 + }, + { + "epoch": 1.79, + "learning_rate": 7.3952776726147634e-06, + "loss": 1.107, + "step": 21249 + }, + { + "epoch": 1.79, + "learning_rate": 7.39440008685957e-06, + "loss": 1.268, + "step": 21250 + }, + { + "epoch": 1.79, + "learning_rate": 7.393522522632677e-06, + "loss": 1.1098, + "step": 21251 + }, + { + "epoch": 1.79, + "learning_rate": 7.392644979941335e-06, + "loss": 1.099, + "step": 21252 + }, + { + "epoch": 1.79, + "learning_rate": 7.391767458792791e-06, + "loss": 1.1668, + "step": 21253 + }, + { + "epoch": 1.79, + "learning_rate": 7.3908899591943055e-06, + "loss": 1.2015, + "step": 21254 + }, + { + "epoch": 1.79, + "learning_rate": 7.390012481153117e-06, + "loss": 1.0864, + "step": 21255 + }, + { + "epoch": 1.79, + "learning_rate": 7.389135024676484e-06, + "loss": 1.1723, + "step": 21256 + }, + { + "epoch": 1.79, + "learning_rate": 7.388257589771648e-06, + "loss": 1.168, + "step": 21257 + }, + { + "epoch": 1.79, + "learning_rate": 7.387380176445869e-06, + "loss": 1.1672, + "step": 21258 + }, + { + "epoch": 1.79, + "learning_rate": 7.3865027847063885e-06, + "loss": 1.2224, + "step": 21259 + }, + { + "epoch": 1.79, + "learning_rate": 7.38562541456046e-06, + "loss": 1.1532, + "step": 21260 + }, + { + "epoch": 1.79, + "learning_rate": 7.384748066015328e-06, + "loss": 1.2537, + "step": 21261 + }, + { + "epoch": 1.79, + "learning_rate": 7.383870739078246e-06, + "loss": 1.0805, + "step": 21262 + }, + { + "epoch": 1.79, + "learning_rate": 7.382993433756464e-06, + "loss": 1.1443, + "step": 21263 + }, + { + "epoch": 1.79, + "learning_rate": 7.3821161500572225e-06, + "loss": 1.1553, + "step": 21264 + }, + { + "epoch": 1.79, + "learning_rate": 7.381238887987781e-06, + "loss": 1.1198, + "step": 21265 + }, + { + "epoch": 1.79, + "learning_rate": 7.3803616475553775e-06, + "loss": 1.1871, + "step": 21266 + }, + { + "epoch": 1.79, + "learning_rate": 7.379484428767267e-06, + "loss": 1.2129, + "step": 21267 + }, + { + "epoch": 1.79, + "learning_rate": 7.378607231630692e-06, + "loss": 1.1417, + "step": 21268 + }, + { + "epoch": 1.79, + "learning_rate": 7.377730056152907e-06, + "loss": 1.3034, + "step": 21269 + }, + { + "epoch": 1.79, + "learning_rate": 7.376852902341153e-06, + "loss": 1.1452, + "step": 21270 + }, + { + "epoch": 1.79, + "learning_rate": 7.375975770202682e-06, + "loss": 1.2145, + "step": 21271 + }, + { + "epoch": 1.79, + "learning_rate": 7.375098659744734e-06, + "loss": 1.1236, + "step": 21272 + }, + { + "epoch": 1.79, + "learning_rate": 7.374221570974565e-06, + "loss": 1.1784, + "step": 21273 + }, + { + "epoch": 1.79, + "learning_rate": 7.373344503899418e-06, + "loss": 1.1751, + "step": 21274 + }, + { + "epoch": 1.79, + "learning_rate": 7.372467458526536e-06, + "loss": 1.1237, + "step": 21275 + }, + { + "epoch": 1.79, + "learning_rate": 7.371590434863172e-06, + "loss": 1.1851, + "step": 21276 + }, + { + "epoch": 1.79, + "learning_rate": 7.370713432916568e-06, + "loss": 1.215, + "step": 21277 + }, + { + "epoch": 1.79, + "learning_rate": 7.369836452693972e-06, + "loss": 1.1938, + "step": 21278 + }, + { + "epoch": 1.79, + "learning_rate": 7.3689594942026254e-06, + "loss": 1.2206, + "step": 21279 + }, + { + "epoch": 1.79, + "learning_rate": 7.368082557449783e-06, + "loss": 1.2779, + "step": 21280 + }, + { + "epoch": 1.79, + "learning_rate": 7.367205642442681e-06, + "loss": 1.195, + "step": 21281 + }, + { + "epoch": 1.79, + "learning_rate": 7.366328749188572e-06, + "loss": 1.2216, + "step": 21282 + }, + { + "epoch": 1.79, + "learning_rate": 7.365451877694691e-06, + "loss": 1.2456, + "step": 21283 + }, + { + "epoch": 1.79, + "learning_rate": 7.364575027968297e-06, + "loss": 1.1732, + "step": 21284 + }, + { + "epoch": 1.79, + "learning_rate": 7.3636982000166244e-06, + "loss": 1.1332, + "step": 21285 + }, + { + "epoch": 1.79, + "learning_rate": 7.362821393846923e-06, + "loss": 1.3077, + "step": 21286 + }, + { + "epoch": 1.79, + "learning_rate": 7.361944609466431e-06, + "loss": 1.1161, + "step": 21287 + }, + { + "epoch": 1.79, + "learning_rate": 7.361067846882399e-06, + "loss": 1.1688, + "step": 21288 + }, + { + "epoch": 1.79, + "learning_rate": 7.36019110610207e-06, + "loss": 1.1415, + "step": 21289 + }, + { + "epoch": 1.79, + "learning_rate": 7.359314387132686e-06, + "loss": 1.1498, + "step": 21290 + }, + { + "epoch": 1.79, + "learning_rate": 7.358437689981492e-06, + "loss": 1.2323, + "step": 21291 + }, + { + "epoch": 1.79, + "learning_rate": 7.357561014655732e-06, + "loss": 1.2313, + "step": 21292 + }, + { + "epoch": 1.79, + "learning_rate": 7.3566843611626495e-06, + "loss": 1.0977, + "step": 21293 + }, + { + "epoch": 1.79, + "learning_rate": 7.355807729509482e-06, + "loss": 1.1289, + "step": 21294 + }, + { + "epoch": 1.79, + "learning_rate": 7.3549311197034825e-06, + "loss": 1.1226, + "step": 21295 + }, + { + "epoch": 1.79, + "learning_rate": 7.3540545317518865e-06, + "loss": 1.1746, + "step": 21296 + }, + { + "epoch": 1.79, + "learning_rate": 7.35317796566194e-06, + "loss": 1.1913, + "step": 21297 + }, + { + "epoch": 1.79, + "learning_rate": 7.3523014214408805e-06, + "loss": 1.2992, + "step": 21298 + }, + { + "epoch": 1.79, + "learning_rate": 7.3514248990959555e-06, + "loss": 1.2416, + "step": 21299 + }, + { + "epoch": 1.79, + "learning_rate": 7.35054839863441e-06, + "loss": 1.1831, + "step": 21300 + }, + { + "epoch": 1.79, + "learning_rate": 7.349671920063475e-06, + "loss": 1.1934, + "step": 21301 + }, + { + "epoch": 1.79, + "learning_rate": 7.348795463390406e-06, + "loss": 1.1245, + "step": 21302 + }, + { + "epoch": 1.79, + "learning_rate": 7.34791902862243e-06, + "loss": 1.0946, + "step": 21303 + }, + { + "epoch": 1.79, + "learning_rate": 7.347042615766801e-06, + "loss": 1.1763, + "step": 21304 + }, + { + "epoch": 1.79, + "learning_rate": 7.346166224830753e-06, + "loss": 1.1186, + "step": 21305 + }, + { + "epoch": 1.79, + "learning_rate": 7.345289855821531e-06, + "loss": 1.2733, + "step": 21306 + }, + { + "epoch": 1.79, + "learning_rate": 7.34441350874637e-06, + "loss": 1.2208, + "step": 21307 + }, + { + "epoch": 1.79, + "learning_rate": 7.343537183612518e-06, + "loss": 1.1403, + "step": 21308 + }, + { + "epoch": 1.79, + "learning_rate": 7.34266088042721e-06, + "loss": 1.1633, + "step": 21309 + }, + { + "epoch": 1.79, + "learning_rate": 7.34178459919769e-06, + "loss": 1.1965, + "step": 21310 + }, + { + "epoch": 1.79, + "learning_rate": 7.340908339931192e-06, + "loss": 1.241, + "step": 21311 + }, + { + "epoch": 1.79, + "learning_rate": 7.3400321026349645e-06, + "loss": 1.1644, + "step": 21312 + }, + { + "epoch": 1.79, + "learning_rate": 7.339155887316243e-06, + "loss": 1.2127, + "step": 21313 + }, + { + "epoch": 1.79, + "learning_rate": 7.338279693982264e-06, + "loss": 1.1773, + "step": 21314 + }, + { + "epoch": 1.79, + "learning_rate": 7.337403522640274e-06, + "loss": 1.1782, + "step": 21315 + }, + { + "epoch": 1.79, + "learning_rate": 7.336527373297506e-06, + "loss": 1.2312, + "step": 21316 + }, + { + "epoch": 1.79, + "learning_rate": 7.335651245961203e-06, + "loss": 1.0905, + "step": 21317 + }, + { + "epoch": 1.79, + "learning_rate": 7.334775140638598e-06, + "loss": 1.0787, + "step": 21318 + }, + { + "epoch": 1.79, + "learning_rate": 7.333899057336938e-06, + "loss": 1.2049, + "step": 21319 + }, + { + "epoch": 1.79, + "learning_rate": 7.333022996063455e-06, + "loss": 1.2188, + "step": 21320 + }, + { + "epoch": 1.8, + "learning_rate": 7.332146956825392e-06, + "loss": 1.1297, + "step": 21321 + }, + { + "epoch": 1.8, + "learning_rate": 7.331270939629981e-06, + "loss": 1.1638, + "step": 21322 + }, + { + "epoch": 1.8, + "learning_rate": 7.330394944484467e-06, + "loss": 1.1863, + "step": 21323 + }, + { + "epoch": 1.8, + "learning_rate": 7.329518971396084e-06, + "loss": 1.2527, + "step": 21324 + }, + { + "epoch": 1.8, + "learning_rate": 7.3286430203720645e-06, + "loss": 1.1638, + "step": 21325 + }, + { + "epoch": 1.8, + "learning_rate": 7.327767091419657e-06, + "loss": 1.1409, + "step": 21326 + }, + { + "epoch": 1.8, + "learning_rate": 7.326891184546092e-06, + "loss": 1.1619, + "step": 21327 + }, + { + "epoch": 1.8, + "learning_rate": 7.326015299758609e-06, + "loss": 1.1059, + "step": 21328 + }, + { + "epoch": 1.8, + "learning_rate": 7.325139437064439e-06, + "loss": 1.2013, + "step": 21329 + }, + { + "epoch": 1.8, + "learning_rate": 7.3242635964708275e-06, + "loss": 1.1567, + "step": 21330 + }, + { + "epoch": 1.8, + "learning_rate": 7.3233877779850045e-06, + "loss": 1.1033, + "step": 21331 + }, + { + "epoch": 1.8, + "learning_rate": 7.32251198161421e-06, + "loss": 1.1336, + "step": 21332 + }, + { + "epoch": 1.8, + "learning_rate": 7.321636207365674e-06, + "loss": 1.2212, + "step": 21333 + }, + { + "epoch": 1.8, + "learning_rate": 7.320760455246641e-06, + "loss": 1.1754, + "step": 21334 + }, + { + "epoch": 1.8, + "learning_rate": 7.319884725264342e-06, + "loss": 1.1397, + "step": 21335 + }, + { + "epoch": 1.8, + "learning_rate": 7.319009017426013e-06, + "loss": 1.2318, + "step": 21336 + }, + { + "epoch": 1.8, + "learning_rate": 7.318133331738887e-06, + "loss": 1.2576, + "step": 21337 + }, + { + "epoch": 1.8, + "learning_rate": 7.3172576682102035e-06, + "loss": 1.1289, + "step": 21338 + }, + { + "epoch": 1.8, + "learning_rate": 7.316382026847197e-06, + "loss": 1.2826, + "step": 21339 + }, + { + "epoch": 1.8, + "learning_rate": 7.315506407657098e-06, + "loss": 1.1676, + "step": 21340 + }, + { + "epoch": 1.8, + "learning_rate": 7.3146308106471476e-06, + "loss": 1.2107, + "step": 21341 + }, + { + "epoch": 1.8, + "learning_rate": 7.313755235824575e-06, + "loss": 1.1977, + "step": 21342 + }, + { + "epoch": 1.8, + "learning_rate": 7.312879683196618e-06, + "loss": 1.1035, + "step": 21343 + }, + { + "epoch": 1.8, + "learning_rate": 7.312004152770506e-06, + "loss": 1.1878, + "step": 21344 + }, + { + "epoch": 1.8, + "learning_rate": 7.3111286445534805e-06, + "loss": 1.1569, + "step": 21345 + }, + { + "epoch": 1.8, + "learning_rate": 7.310253158552767e-06, + "loss": 1.1637, + "step": 21346 + }, + { + "epoch": 1.8, + "learning_rate": 7.309377694775606e-06, + "loss": 1.1153, + "step": 21347 + }, + { + "epoch": 1.8, + "learning_rate": 7.3085022532292215e-06, + "loss": 1.1914, + "step": 21348 + }, + { + "epoch": 1.8, + "learning_rate": 7.3076268339208565e-06, + "loss": 1.2499, + "step": 21349 + }, + { + "epoch": 1.8, + "learning_rate": 7.306751436857742e-06, + "loss": 1.1717, + "step": 21350 + }, + { + "epoch": 1.8, + "learning_rate": 7.305876062047106e-06, + "loss": 1.0897, + "step": 21351 + }, + { + "epoch": 1.8, + "learning_rate": 7.305000709496188e-06, + "loss": 1.1353, + "step": 21352 + }, + { + "epoch": 1.8, + "learning_rate": 7.304125379212212e-06, + "loss": 1.246, + "step": 21353 + }, + { + "epoch": 1.8, + "learning_rate": 7.303250071202419e-06, + "loss": 1.2497, + "step": 21354 + }, + { + "epoch": 1.8, + "learning_rate": 7.302374785474035e-06, + "loss": 1.2365, + "step": 21355 + }, + { + "epoch": 1.8, + "learning_rate": 7.3014995220342964e-06, + "loss": 1.0871, + "step": 21356 + }, + { + "epoch": 1.8, + "learning_rate": 7.300624280890427e-06, + "loss": 1.1459, + "step": 21357 + }, + { + "epoch": 1.8, + "learning_rate": 7.299749062049668e-06, + "loss": 1.1737, + "step": 21358 + }, + { + "epoch": 1.8, + "learning_rate": 7.298873865519245e-06, + "loss": 1.0897, + "step": 21359 + }, + { + "epoch": 1.8, + "learning_rate": 7.297998691306393e-06, + "loss": 1.225, + "step": 21360 + }, + { + "epoch": 1.8, + "learning_rate": 7.297123539418336e-06, + "loss": 1.1604, + "step": 21361 + }, + { + "epoch": 1.8, + "learning_rate": 7.296248409862311e-06, + "loss": 1.1339, + "step": 21362 + }, + { + "epoch": 1.8, + "learning_rate": 7.29537330264555e-06, + "loss": 1.1601, + "step": 21363 + }, + { + "epoch": 1.8, + "learning_rate": 7.294498217775274e-06, + "loss": 1.2257, + "step": 21364 + }, + { + "epoch": 1.8, + "learning_rate": 7.293623155258726e-06, + "loss": 1.2795, + "step": 21365 + }, + { + "epoch": 1.8, + "learning_rate": 7.292748115103127e-06, + "loss": 1.1251, + "step": 21366 + }, + { + "epoch": 1.8, + "learning_rate": 7.291873097315712e-06, + "loss": 1.1909, + "step": 21367 + }, + { + "epoch": 1.8, + "learning_rate": 7.290998101903703e-06, + "loss": 1.1618, + "step": 21368 + }, + { + "epoch": 1.8, + "learning_rate": 7.290123128874339e-06, + "loss": 1.1754, + "step": 21369 + }, + { + "epoch": 1.8, + "learning_rate": 7.289248178234842e-06, + "loss": 1.1824, + "step": 21370 + }, + { + "epoch": 1.8, + "learning_rate": 7.288373249992448e-06, + "loss": 1.1206, + "step": 21371 + }, + { + "epoch": 1.8, + "learning_rate": 7.287498344154378e-06, + "loss": 1.1382, + "step": 21372 + }, + { + "epoch": 1.8, + "learning_rate": 7.286623460727869e-06, + "loss": 1.1558, + "step": 21373 + }, + { + "epoch": 1.8, + "learning_rate": 7.285748599720142e-06, + "loss": 1.1505, + "step": 21374 + }, + { + "epoch": 1.8, + "learning_rate": 7.284873761138427e-06, + "loss": 1.2767, + "step": 21375 + }, + { + "epoch": 1.8, + "learning_rate": 7.283998944989958e-06, + "loss": 1.2288, + "step": 21376 + }, + { + "epoch": 1.8, + "learning_rate": 7.283124151281958e-06, + "loss": 1.1854, + "step": 21377 + }, + { + "epoch": 1.8, + "learning_rate": 7.2822493800216556e-06, + "loss": 1.1589, + "step": 21378 + }, + { + "epoch": 1.8, + "learning_rate": 7.281374631216276e-06, + "loss": 1.1389, + "step": 21379 + }, + { + "epoch": 1.8, + "learning_rate": 7.280499904873053e-06, + "loss": 1.1848, + "step": 21380 + }, + { + "epoch": 1.8, + "learning_rate": 7.279625200999209e-06, + "loss": 1.1701, + "step": 21381 + }, + { + "epoch": 1.8, + "learning_rate": 7.278750519601974e-06, + "loss": 1.2131, + "step": 21382 + }, + { + "epoch": 1.8, + "learning_rate": 7.277875860688568e-06, + "loss": 1.1366, + "step": 21383 + }, + { + "epoch": 1.8, + "learning_rate": 7.277001224266227e-06, + "loss": 1.1373, + "step": 21384 + }, + { + "epoch": 1.8, + "learning_rate": 7.276126610342171e-06, + "loss": 1.2161, + "step": 21385 + }, + { + "epoch": 1.8, + "learning_rate": 7.275252018923633e-06, + "loss": 1.2159, + "step": 21386 + }, + { + "epoch": 1.8, + "learning_rate": 7.274377450017829e-06, + "loss": 1.1663, + "step": 21387 + }, + { + "epoch": 1.8, + "learning_rate": 7.273502903631992e-06, + "loss": 1.1822, + "step": 21388 + }, + { + "epoch": 1.8, + "learning_rate": 7.2726283797733514e-06, + "loss": 1.1923, + "step": 21389 + }, + { + "epoch": 1.8, + "learning_rate": 7.271753878449122e-06, + "loss": 1.2205, + "step": 21390 + }, + { + "epoch": 1.8, + "learning_rate": 7.270879399666539e-06, + "loss": 1.182, + "step": 21391 + }, + { + "epoch": 1.8, + "learning_rate": 7.2700049434328225e-06, + "loss": 1.1828, + "step": 21392 + }, + { + "epoch": 1.8, + "learning_rate": 7.269130509755201e-06, + "loss": 1.1464, + "step": 21393 + }, + { + "epoch": 1.8, + "learning_rate": 7.268256098640893e-06, + "loss": 1.1175, + "step": 21394 + }, + { + "epoch": 1.8, + "learning_rate": 7.267381710097132e-06, + "loss": 1.1209, + "step": 21395 + }, + { + "epoch": 1.8, + "learning_rate": 7.266507344131135e-06, + "loss": 1.269, + "step": 21396 + }, + { + "epoch": 1.8, + "learning_rate": 7.265633000750132e-06, + "loss": 1.1568, + "step": 21397 + }, + { + "epoch": 1.8, + "learning_rate": 7.264758679961339e-06, + "loss": 1.1528, + "step": 21398 + }, + { + "epoch": 1.8, + "learning_rate": 7.263884381771988e-06, + "loss": 1.0853, + "step": 21399 + }, + { + "epoch": 1.8, + "learning_rate": 7.263010106189301e-06, + "loss": 1.1536, + "step": 21400 + }, + { + "epoch": 1.8, + "learning_rate": 7.262135853220498e-06, + "loss": 1.246, + "step": 21401 + }, + { + "epoch": 1.8, + "learning_rate": 7.261261622872811e-06, + "loss": 1.1813, + "step": 21402 + }, + { + "epoch": 1.8, + "learning_rate": 7.260387415153449e-06, + "loss": 1.2363, + "step": 21403 + }, + { + "epoch": 1.8, + "learning_rate": 7.259513230069648e-06, + "loss": 1.1861, + "step": 21404 + }, + { + "epoch": 1.8, + "learning_rate": 7.258639067628624e-06, + "loss": 1.1907, + "step": 21405 + }, + { + "epoch": 1.8, + "learning_rate": 7.257764927837603e-06, + "loss": 1.2136, + "step": 21406 + }, + { + "epoch": 1.8, + "learning_rate": 7.256890810703802e-06, + "loss": 1.1498, + "step": 21407 + }, + { + "epoch": 1.8, + "learning_rate": 7.256016716234451e-06, + "loss": 1.2047, + "step": 21408 + }, + { + "epoch": 1.8, + "learning_rate": 7.255142644436765e-06, + "loss": 1.145, + "step": 21409 + }, + { + "epoch": 1.8, + "learning_rate": 7.254268595317973e-06, + "loss": 1.2205, + "step": 21410 + }, + { + "epoch": 1.8, + "learning_rate": 7.253394568885287e-06, + "loss": 1.2352, + "step": 21411 + }, + { + "epoch": 1.8, + "learning_rate": 7.2525205651459366e-06, + "loss": 1.1591, + "step": 21412 + }, + { + "epoch": 1.8, + "learning_rate": 7.251646584107143e-06, + "loss": 1.1447, + "step": 21413 + }, + { + "epoch": 1.8, + "learning_rate": 7.25077262577612e-06, + "loss": 1.1536, + "step": 21414 + }, + { + "epoch": 1.8, + "learning_rate": 7.249898690160098e-06, + "loss": 1.1272, + "step": 21415 + }, + { + "epoch": 1.8, + "learning_rate": 7.249024777266291e-06, + "loss": 1.0995, + "step": 21416 + }, + { + "epoch": 1.8, + "learning_rate": 7.248150887101923e-06, + "loss": 1.1279, + "step": 21417 + }, + { + "epoch": 1.8, + "learning_rate": 7.24727701967421e-06, + "loss": 1.139, + "step": 21418 + }, + { + "epoch": 1.8, + "learning_rate": 7.246403174990379e-06, + "loss": 1.1868, + "step": 21419 + }, + { + "epoch": 1.8, + "learning_rate": 7.245529353057645e-06, + "loss": 1.1194, + "step": 21420 + }, + { + "epoch": 1.8, + "learning_rate": 7.244655553883231e-06, + "loss": 1.1894, + "step": 21421 + }, + { + "epoch": 1.8, + "learning_rate": 7.243781777474351e-06, + "loss": 1.1356, + "step": 21422 + }, + { + "epoch": 1.8, + "learning_rate": 7.242908023838231e-06, + "loss": 1.2107, + "step": 21423 + }, + { + "epoch": 1.8, + "learning_rate": 7.242034292982085e-06, + "loss": 1.1082, + "step": 21424 + }, + { + "epoch": 1.8, + "learning_rate": 7.2411605849131316e-06, + "loss": 1.2587, + "step": 21425 + }, + { + "epoch": 1.8, + "learning_rate": 7.240286899638598e-06, + "loss": 1.1813, + "step": 21426 + }, + { + "epoch": 1.8, + "learning_rate": 7.239413237165696e-06, + "loss": 1.1426, + "step": 21427 + }, + { + "epoch": 1.8, + "learning_rate": 7.238539597501646e-06, + "loss": 1.1787, + "step": 21428 + }, + { + "epoch": 1.8, + "learning_rate": 7.237665980653661e-06, + "loss": 1.1847, + "step": 21429 + }, + { + "epoch": 1.8, + "learning_rate": 7.236792386628968e-06, + "loss": 1.0339, + "step": 21430 + }, + { + "epoch": 1.8, + "learning_rate": 7.235918815434779e-06, + "loss": 1.1094, + "step": 21431 + }, + { + "epoch": 1.8, + "learning_rate": 7.235045267078317e-06, + "loss": 1.1264, + "step": 21432 + }, + { + "epoch": 1.8, + "learning_rate": 7.234171741566789e-06, + "loss": 1.2008, + "step": 21433 + }, + { + "epoch": 1.8, + "learning_rate": 7.233298238907424e-06, + "loss": 1.0801, + "step": 21434 + }, + { + "epoch": 1.8, + "learning_rate": 7.232424759107434e-06, + "loss": 1.158, + "step": 21435 + }, + { + "epoch": 1.8, + "learning_rate": 7.231551302174038e-06, + "loss": 1.2263, + "step": 21436 + }, + { + "epoch": 1.8, + "learning_rate": 7.230677868114447e-06, + "loss": 1.1351, + "step": 21437 + }, + { + "epoch": 1.8, + "learning_rate": 7.229804456935882e-06, + "loss": 1.2891, + "step": 21438 + }, + { + "epoch": 1.8, + "learning_rate": 7.228931068645564e-06, + "loss": 1.2751, + "step": 21439 + }, + { + "epoch": 1.81, + "learning_rate": 7.228057703250698e-06, + "loss": 1.165, + "step": 21440 + }, + { + "epoch": 1.81, + "learning_rate": 7.227184360758512e-06, + "loss": 1.2582, + "step": 21441 + }, + { + "epoch": 1.81, + "learning_rate": 7.226311041176213e-06, + "loss": 1.1346, + "step": 21442 + }, + { + "epoch": 1.81, + "learning_rate": 7.225437744511023e-06, + "loss": 1.1647, + "step": 21443 + }, + { + "epoch": 1.81, + "learning_rate": 7.224564470770149e-06, + "loss": 1.1732, + "step": 21444 + }, + { + "epoch": 1.81, + "learning_rate": 7.223691219960817e-06, + "loss": 1.1867, + "step": 21445 + }, + { + "epoch": 1.81, + "learning_rate": 7.222817992090234e-06, + "loss": 1.2486, + "step": 21446 + }, + { + "epoch": 1.81, + "learning_rate": 7.221944787165619e-06, + "loss": 1.218, + "step": 21447 + }, + { + "epoch": 1.81, + "learning_rate": 7.221071605194181e-06, + "loss": 1.1984, + "step": 21448 + }, + { + "epoch": 1.81, + "learning_rate": 7.220198446183142e-06, + "loss": 1.2557, + "step": 21449 + }, + { + "epoch": 1.81, + "learning_rate": 7.219325310139716e-06, + "loss": 1.1606, + "step": 21450 + }, + { + "epoch": 1.81, + "learning_rate": 7.218452197071107e-06, + "loss": 1.231, + "step": 21451 + }, + { + "epoch": 1.81, + "learning_rate": 7.217579106984542e-06, + "loss": 1.1617, + "step": 21452 + }, + { + "epoch": 1.81, + "learning_rate": 7.216706039887227e-06, + "loss": 1.0746, + "step": 21453 + }, + { + "epoch": 1.81, + "learning_rate": 7.21583299578638e-06, + "loss": 1.2137, + "step": 21454 + }, + { + "epoch": 1.81, + "learning_rate": 7.214959974689207e-06, + "loss": 1.1439, + "step": 21455 + }, + { + "epoch": 1.81, + "learning_rate": 7.214086976602932e-06, + "loss": 1.2313, + "step": 21456 + }, + { + "epoch": 1.81, + "learning_rate": 7.2132140015347566e-06, + "loss": 1.1279, + "step": 21457 + }, + { + "epoch": 1.81, + "learning_rate": 7.212341049491902e-06, + "loss": 1.1592, + "step": 21458 + }, + { + "epoch": 1.81, + "learning_rate": 7.211468120481577e-06, + "loss": 1.2805, + "step": 21459 + }, + { + "epoch": 1.81, + "learning_rate": 7.210595214510997e-06, + "loss": 1.1861, + "step": 21460 + }, + { + "epoch": 1.81, + "learning_rate": 7.209722331587368e-06, + "loss": 1.1616, + "step": 21461 + }, + { + "epoch": 1.81, + "learning_rate": 7.208849471717907e-06, + "loss": 1.1729, + "step": 21462 + }, + { + "epoch": 1.81, + "learning_rate": 7.207976634909827e-06, + "loss": 1.1297, + "step": 21463 + }, + { + "epoch": 1.81, + "learning_rate": 7.2071038211703334e-06, + "loss": 1.0706, + "step": 21464 + }, + { + "epoch": 1.81, + "learning_rate": 7.2062310305066475e-06, + "loss": 1.1584, + "step": 21465 + }, + { + "epoch": 1.81, + "learning_rate": 7.205358262925971e-06, + "loss": 1.2395, + "step": 21466 + }, + { + "epoch": 1.81, + "learning_rate": 7.204485518435522e-06, + "loss": 1.1792, + "step": 21467 + }, + { + "epoch": 1.81, + "learning_rate": 7.203612797042504e-06, + "loss": 1.2413, + "step": 21468 + }, + { + "epoch": 1.81, + "learning_rate": 7.202740098754137e-06, + "loss": 1.2043, + "step": 21469 + }, + { + "epoch": 1.81, + "learning_rate": 7.201867423577623e-06, + "loss": 1.2334, + "step": 21470 + }, + { + "epoch": 1.81, + "learning_rate": 7.200994771520179e-06, + "loss": 1.2035, + "step": 21471 + }, + { + "epoch": 1.81, + "learning_rate": 7.200122142589008e-06, + "loss": 1.176, + "step": 21472 + }, + { + "epoch": 1.81, + "learning_rate": 7.1992495367913276e-06, + "loss": 1.1767, + "step": 21473 + }, + { + "epoch": 1.81, + "learning_rate": 7.198376954134341e-06, + "loss": 1.1831, + "step": 21474 + }, + { + "epoch": 1.81, + "learning_rate": 7.19750439462526e-06, + "loss": 1.2547, + "step": 21475 + }, + { + "epoch": 1.81, + "learning_rate": 7.196631858271298e-06, + "loss": 1.1395, + "step": 21476 + }, + { + "epoch": 1.81, + "learning_rate": 7.195759345079658e-06, + "loss": 1.1633, + "step": 21477 + }, + { + "epoch": 1.81, + "learning_rate": 7.194886855057556e-06, + "loss": 1.1322, + "step": 21478 + }, + { + "epoch": 1.81, + "learning_rate": 7.19401438821219e-06, + "loss": 1.1816, + "step": 21479 + }, + { + "epoch": 1.81, + "learning_rate": 7.193141944550781e-06, + "loss": 1.1739, + "step": 21480 + }, + { + "epoch": 1.81, + "learning_rate": 7.192269524080529e-06, + "loss": 1.2908, + "step": 21481 + }, + { + "epoch": 1.81, + "learning_rate": 7.191397126808646e-06, + "loss": 1.2048, + "step": 21482 + }, + { + "epoch": 1.81, + "learning_rate": 7.190524752742336e-06, + "loss": 1.0429, + "step": 21483 + }, + { + "epoch": 1.81, + "learning_rate": 7.189652401888813e-06, + "loss": 1.1379, + "step": 21484 + }, + { + "epoch": 1.81, + "learning_rate": 7.18878007425528e-06, + "loss": 1.1711, + "step": 21485 + }, + { + "epoch": 1.81, + "learning_rate": 7.187907769848947e-06, + "loss": 1.1569, + "step": 21486 + }, + { + "epoch": 1.81, + "learning_rate": 7.187035488677016e-06, + "loss": 1.089, + "step": 21487 + }, + { + "epoch": 1.81, + "learning_rate": 7.1861632307467e-06, + "loss": 1.1622, + "step": 21488 + }, + { + "epoch": 1.81, + "learning_rate": 7.185290996065206e-06, + "loss": 1.1985, + "step": 21489 + }, + { + "epoch": 1.81, + "learning_rate": 7.184418784639735e-06, + "loss": 1.1787, + "step": 21490 + }, + { + "epoch": 1.81, + "learning_rate": 7.1835465964774995e-06, + "loss": 1.1518, + "step": 21491 + }, + { + "epoch": 1.81, + "learning_rate": 7.182674431585703e-06, + "loss": 1.1786, + "step": 21492 + }, + { + "epoch": 1.81, + "learning_rate": 7.181802289971554e-06, + "loss": 1.2691, + "step": 21493 + }, + { + "epoch": 1.81, + "learning_rate": 7.180930171642253e-06, + "loss": 1.2422, + "step": 21494 + }, + { + "epoch": 1.81, + "learning_rate": 7.180058076605012e-06, + "loss": 1.1446, + "step": 21495 + }, + { + "epoch": 1.81, + "learning_rate": 7.179186004867032e-06, + "loss": 1.1897, + "step": 21496 + }, + { + "epoch": 1.81, + "learning_rate": 7.178313956435522e-06, + "loss": 1.1235, + "step": 21497 + }, + { + "epoch": 1.81, + "learning_rate": 7.177441931317681e-06, + "loss": 1.0859, + "step": 21498 + }, + { + "epoch": 1.81, + "learning_rate": 7.17656992952072e-06, + "loss": 1.1926, + "step": 21499 + }, + { + "epoch": 1.81, + "learning_rate": 7.175697951051844e-06, + "loss": 1.0875, + "step": 21500 + }, + { + "epoch": 1.81, + "learning_rate": 7.1748259959182505e-06, + "loss": 1.1928, + "step": 21501 + }, + { + "epoch": 1.81, + "learning_rate": 7.173954064127155e-06, + "loss": 1.2166, + "step": 21502 + }, + { + "epoch": 1.81, + "learning_rate": 7.1730821556857525e-06, + "loss": 1.2037, + "step": 21503 + }, + { + "epoch": 1.81, + "learning_rate": 7.172210270601252e-06, + "loss": 1.1617, + "step": 21504 + }, + { + "epoch": 1.81, + "learning_rate": 7.1713384088808506e-06, + "loss": 1.1662, + "step": 21505 + }, + { + "epoch": 1.81, + "learning_rate": 7.170466570531762e-06, + "loss": 1.1996, + "step": 21506 + }, + { + "epoch": 1.81, + "learning_rate": 7.16959475556118e-06, + "loss": 1.1323, + "step": 21507 + }, + { + "epoch": 1.81, + "learning_rate": 7.168722963976316e-06, + "loss": 1.1989, + "step": 21508 + }, + { + "epoch": 1.81, + "learning_rate": 7.167851195784366e-06, + "loss": 1.1488, + "step": 21509 + }, + { + "epoch": 1.81, + "learning_rate": 7.166979450992537e-06, + "loss": 1.1888, + "step": 21510 + }, + { + "epoch": 1.81, + "learning_rate": 7.166107729608028e-06, + "loss": 1.2326, + "step": 21511 + }, + { + "epoch": 1.81, + "learning_rate": 7.165236031638044e-06, + "loss": 1.1629, + "step": 21512 + }, + { + "epoch": 1.81, + "learning_rate": 7.16436435708979e-06, + "loss": 1.2144, + "step": 21513 + }, + { + "epoch": 1.81, + "learning_rate": 7.163492705970463e-06, + "loss": 1.1763, + "step": 21514 + }, + { + "epoch": 1.81, + "learning_rate": 7.162621078287268e-06, + "loss": 1.1714, + "step": 21515 + }, + { + "epoch": 1.81, + "learning_rate": 7.161749474047406e-06, + "loss": 1.2335, + "step": 21516 + }, + { + "epoch": 1.81, + "learning_rate": 7.1608778932580804e-06, + "loss": 1.1641, + "step": 21517 + }, + { + "epoch": 1.81, + "learning_rate": 7.1600063359264845e-06, + "loss": 1.1614, + "step": 21518 + }, + { + "epoch": 1.81, + "learning_rate": 7.159134802059831e-06, + "loss": 1.1553, + "step": 21519 + }, + { + "epoch": 1.81, + "learning_rate": 7.158263291665312e-06, + "loss": 1.0863, + "step": 21520 + }, + { + "epoch": 1.81, + "learning_rate": 7.157391804750133e-06, + "loss": 1.1432, + "step": 21521 + }, + { + "epoch": 1.81, + "learning_rate": 7.156520341321489e-06, + "loss": 1.2639, + "step": 21522 + }, + { + "epoch": 1.81, + "learning_rate": 7.155648901386587e-06, + "loss": 1.2191, + "step": 21523 + }, + { + "epoch": 1.81, + "learning_rate": 7.154777484952624e-06, + "loss": 1.1041, + "step": 21524 + }, + { + "epoch": 1.81, + "learning_rate": 7.153906092026797e-06, + "loss": 1.202, + "step": 21525 + }, + { + "epoch": 1.81, + "learning_rate": 7.1530347226163125e-06, + "loss": 1.1154, + "step": 21526 + }, + { + "epoch": 1.81, + "learning_rate": 7.1521633767283645e-06, + "loss": 1.2309, + "step": 21527 + }, + { + "epoch": 1.81, + "learning_rate": 7.151292054370156e-06, + "loss": 1.1906, + "step": 21528 + }, + { + "epoch": 1.81, + "learning_rate": 7.15042075554888e-06, + "loss": 1.2396, + "step": 21529 + }, + { + "epoch": 1.81, + "learning_rate": 7.149549480271745e-06, + "loss": 1.2547, + "step": 21530 + }, + { + "epoch": 1.81, + "learning_rate": 7.14867822854594e-06, + "loss": 1.1791, + "step": 21531 + }, + { + "epoch": 1.81, + "learning_rate": 7.147807000378672e-06, + "loss": 1.1019, + "step": 21532 + }, + { + "epoch": 1.81, + "learning_rate": 7.14693579577713e-06, + "loss": 1.1759, + "step": 21533 + }, + { + "epoch": 1.81, + "learning_rate": 7.146064614748524e-06, + "loss": 1.1814, + "step": 21534 + }, + { + "epoch": 1.81, + "learning_rate": 7.145193457300041e-06, + "loss": 1.1686, + "step": 21535 + }, + { + "epoch": 1.81, + "learning_rate": 7.144322323438887e-06, + "loss": 1.1933, + "step": 21536 + }, + { + "epoch": 1.81, + "learning_rate": 7.14345121317225e-06, + "loss": 1.1652, + "step": 21537 + }, + { + "epoch": 1.81, + "learning_rate": 7.142580126507337e-06, + "loss": 1.1447, + "step": 21538 + }, + { + "epoch": 1.81, + "learning_rate": 7.141709063451343e-06, + "loss": 1.1906, + "step": 21539 + }, + { + "epoch": 1.81, + "learning_rate": 7.14083802401146e-06, + "loss": 1.1404, + "step": 21540 + }, + { + "epoch": 1.81, + "learning_rate": 7.139967008194891e-06, + "loss": 1.158, + "step": 21541 + }, + { + "epoch": 1.81, + "learning_rate": 7.1390960160088285e-06, + "loss": 1.1904, + "step": 21542 + }, + { + "epoch": 1.81, + "learning_rate": 7.138225047460473e-06, + "loss": 1.1778, + "step": 21543 + }, + { + "epoch": 1.81, + "learning_rate": 7.1373541025570125e-06, + "loss": 1.1403, + "step": 21544 + }, + { + "epoch": 1.81, + "learning_rate": 7.1364831813056544e-06, + "loss": 1.1657, + "step": 21545 + }, + { + "epoch": 1.81, + "learning_rate": 7.135612283713587e-06, + "loss": 1.1858, + "step": 21546 + }, + { + "epoch": 1.81, + "learning_rate": 7.134741409788008e-06, + "loss": 1.1443, + "step": 21547 + }, + { + "epoch": 1.81, + "learning_rate": 7.133870559536109e-06, + "loss": 1.2314, + "step": 21548 + }, + { + "epoch": 1.81, + "learning_rate": 7.13299973296509e-06, + "loss": 1.1836, + "step": 21549 + }, + { + "epoch": 1.81, + "learning_rate": 7.132128930082147e-06, + "loss": 1.1618, + "step": 21550 + }, + { + "epoch": 1.81, + "learning_rate": 7.131258150894469e-06, + "loss": 1.1872, + "step": 21551 + }, + { + "epoch": 1.81, + "learning_rate": 7.13038739540926e-06, + "loss": 1.1844, + "step": 21552 + }, + { + "epoch": 1.81, + "learning_rate": 7.129516663633704e-06, + "loss": 1.1596, + "step": 21553 + }, + { + "epoch": 1.81, + "learning_rate": 7.128645955575003e-06, + "loss": 1.1346, + "step": 21554 + }, + { + "epoch": 1.81, + "learning_rate": 7.127775271240345e-06, + "loss": 1.1557, + "step": 21555 + }, + { + "epoch": 1.81, + "learning_rate": 7.126904610636931e-06, + "loss": 1.1138, + "step": 21556 + }, + { + "epoch": 1.81, + "learning_rate": 7.126033973771946e-06, + "loss": 1.1268, + "step": 21557 + }, + { + "epoch": 1.82, + "learning_rate": 7.1251633606525915e-06, + "loss": 1.2729, + "step": 21558 + }, + { + "epoch": 1.82, + "learning_rate": 7.1242927712860564e-06, + "loss": 1.2355, + "step": 21559 + }, + { + "epoch": 1.82, + "learning_rate": 7.123422205679536e-06, + "loss": 1.1458, + "step": 21560 + }, + { + "epoch": 1.82, + "learning_rate": 7.1225516638402185e-06, + "loss": 1.2022, + "step": 21561 + }, + { + "epoch": 1.82, + "learning_rate": 7.121681145775301e-06, + "loss": 1.1728, + "step": 21562 + }, + { + "epoch": 1.82, + "learning_rate": 7.120810651491977e-06, + "loss": 1.1847, + "step": 21563 + }, + { + "epoch": 1.82, + "learning_rate": 7.119940180997434e-06, + "loss": 1.1932, + "step": 21564 + }, + { + "epoch": 1.82, + "learning_rate": 7.11906973429887e-06, + "loss": 1.1087, + "step": 21565 + }, + { + "epoch": 1.82, + "learning_rate": 7.118199311403472e-06, + "loss": 1.0989, + "step": 21566 + }, + { + "epoch": 1.82, + "learning_rate": 7.117328912318436e-06, + "loss": 1.1757, + "step": 21567 + }, + { + "epoch": 1.82, + "learning_rate": 7.1164585370509476e-06, + "loss": 1.1982, + "step": 21568 + }, + { + "epoch": 1.82, + "learning_rate": 7.115588185608204e-06, + "loss": 1.202, + "step": 21569 + }, + { + "epoch": 1.82, + "learning_rate": 7.114717857997393e-06, + "loss": 1.0866, + "step": 21570 + }, + { + "epoch": 1.82, + "learning_rate": 7.113847554225709e-06, + "loss": 1.1526, + "step": 21571 + }, + { + "epoch": 1.82, + "learning_rate": 7.112977274300335e-06, + "loss": 1.1416, + "step": 21572 + }, + { + "epoch": 1.82, + "learning_rate": 7.112107018228471e-06, + "loss": 1.1284, + "step": 21573 + }, + { + "epoch": 1.82, + "learning_rate": 7.1112367860173015e-06, + "loss": 1.1993, + "step": 21574 + }, + { + "epoch": 1.82, + "learning_rate": 7.110366577674017e-06, + "loss": 1.194, + "step": 21575 + }, + { + "epoch": 1.82, + "learning_rate": 7.1094963932058114e-06, + "loss": 1.2825, + "step": 21576 + }, + { + "epoch": 1.82, + "learning_rate": 7.10862623261987e-06, + "loss": 1.1386, + "step": 21577 + }, + { + "epoch": 1.82, + "learning_rate": 7.107756095923386e-06, + "loss": 1.1138, + "step": 21578 + }, + { + "epoch": 1.82, + "learning_rate": 7.1068859831235425e-06, + "loss": 1.2027, + "step": 21579 + }, + { + "epoch": 1.82, + "learning_rate": 7.106015894227538e-06, + "loss": 1.1692, + "step": 21580 + }, + { + "epoch": 1.82, + "learning_rate": 7.105145829242552e-06, + "loss": 1.1825, + "step": 21581 + }, + { + "epoch": 1.82, + "learning_rate": 7.104275788175783e-06, + "loss": 1.2007, + "step": 21582 + }, + { + "epoch": 1.82, + "learning_rate": 7.103405771034408e-06, + "loss": 1.1779, + "step": 21583 + }, + { + "epoch": 1.82, + "learning_rate": 7.102535777825626e-06, + "loss": 1.1579, + "step": 21584 + }, + { + "epoch": 1.82, + "learning_rate": 7.101665808556618e-06, + "loss": 1.1025, + "step": 21585 + }, + { + "epoch": 1.82, + "learning_rate": 7.100795863234577e-06, + "loss": 1.1783, + "step": 21586 + }, + { + "epoch": 1.82, + "learning_rate": 7.099925941866685e-06, + "loss": 1.2034, + "step": 21587 + }, + { + "epoch": 1.82, + "learning_rate": 7.099056044460135e-06, + "loss": 1.1506, + "step": 21588 + }, + { + "epoch": 1.82, + "learning_rate": 7.098186171022114e-06, + "loss": 1.1569, + "step": 21589 + }, + { + "epoch": 1.82, + "learning_rate": 7.097316321559804e-06, + "loss": 1.1657, + "step": 21590 + }, + { + "epoch": 1.82, + "learning_rate": 7.096446496080399e-06, + "loss": 1.1707, + "step": 21591 + }, + { + "epoch": 1.82, + "learning_rate": 7.095576694591081e-06, + "loss": 1.1502, + "step": 21592 + }, + { + "epoch": 1.82, + "learning_rate": 7.094706917099039e-06, + "loss": 1.1938, + "step": 21593 + }, + { + "epoch": 1.82, + "learning_rate": 7.0938371636114544e-06, + "loss": 1.1432, + "step": 21594 + }, + { + "epoch": 1.82, + "learning_rate": 7.092967434135523e-06, + "loss": 1.2249, + "step": 21595 + }, + { + "epoch": 1.82, + "learning_rate": 7.09209772867842e-06, + "loss": 1.2425, + "step": 21596 + }, + { + "epoch": 1.82, + "learning_rate": 7.09122804724734e-06, + "loss": 1.2506, + "step": 21597 + }, + { + "epoch": 1.82, + "learning_rate": 7.090358389849461e-06, + "loss": 1.1928, + "step": 21598 + }, + { + "epoch": 1.82, + "learning_rate": 7.089488756491974e-06, + "loss": 1.089, + "step": 21599 + }, + { + "epoch": 1.82, + "learning_rate": 7.088619147182063e-06, + "loss": 1.1493, + "step": 21600 + }, + { + "epoch": 1.82, + "learning_rate": 7.087749561926911e-06, + "loss": 1.1623, + "step": 21601 + }, + { + "epoch": 1.82, + "learning_rate": 7.086880000733707e-06, + "loss": 1.2599, + "step": 21602 + }, + { + "epoch": 1.82, + "learning_rate": 7.086010463609629e-06, + "loss": 1.1954, + "step": 21603 + }, + { + "epoch": 1.82, + "learning_rate": 7.085140950561868e-06, + "loss": 1.2114, + "step": 21604 + }, + { + "epoch": 1.82, + "learning_rate": 7.084271461597603e-06, + "loss": 1.1769, + "step": 21605 + }, + { + "epoch": 1.82, + "learning_rate": 7.083401996724023e-06, + "loss": 1.2143, + "step": 21606 + }, + { + "epoch": 1.82, + "learning_rate": 7.082532555948307e-06, + "loss": 1.1204, + "step": 21607 + }, + { + "epoch": 1.82, + "learning_rate": 7.0816631392776435e-06, + "loss": 1.2066, + "step": 21608 + }, + { + "epoch": 1.82, + "learning_rate": 7.080793746719209e-06, + "loss": 1.1015, + "step": 21609 + }, + { + "epoch": 1.82, + "learning_rate": 7.079924378280197e-06, + "loss": 1.0946, + "step": 21610 + }, + { + "epoch": 1.82, + "learning_rate": 7.079055033967778e-06, + "loss": 1.1562, + "step": 21611 + }, + { + "epoch": 1.82, + "learning_rate": 7.07818571378914e-06, + "loss": 1.16, + "step": 21612 + }, + { + "epoch": 1.82, + "learning_rate": 7.077316417751474e-06, + "loss": 1.1412, + "step": 21613 + }, + { + "epoch": 1.82, + "learning_rate": 7.076447145861949e-06, + "loss": 1.244, + "step": 21614 + }, + { + "epoch": 1.82, + "learning_rate": 7.075577898127757e-06, + "loss": 1.2016, + "step": 21615 + }, + { + "epoch": 1.82, + "learning_rate": 7.074708674556075e-06, + "loss": 1.1447, + "step": 21616 + }, + { + "epoch": 1.82, + "learning_rate": 7.073839475154087e-06, + "loss": 1.151, + "step": 21617 + }, + { + "epoch": 1.82, + "learning_rate": 7.072970299928972e-06, + "loss": 1.157, + "step": 21618 + }, + { + "epoch": 1.82, + "learning_rate": 7.072101148887916e-06, + "loss": 1.2217, + "step": 21619 + }, + { + "epoch": 1.82, + "learning_rate": 7.071232022038095e-06, + "loss": 1.1415, + "step": 21620 + }, + { + "epoch": 1.82, + "learning_rate": 7.070362919386693e-06, + "loss": 1.2281, + "step": 21621 + }, + { + "epoch": 1.82, + "learning_rate": 7.069493840940888e-06, + "loss": 1.1768, + "step": 21622 + }, + { + "epoch": 1.82, + "learning_rate": 7.068624786707866e-06, + "loss": 1.1936, + "step": 21623 + }, + { + "epoch": 1.82, + "learning_rate": 7.067755756694803e-06, + "loss": 1.2132, + "step": 21624 + }, + { + "epoch": 1.82, + "learning_rate": 7.066886750908877e-06, + "loss": 1.0791, + "step": 21625 + }, + { + "epoch": 1.82, + "learning_rate": 7.066017769357276e-06, + "loss": 1.1575, + "step": 21626 + }, + { + "epoch": 1.82, + "learning_rate": 7.065148812047174e-06, + "loss": 1.261, + "step": 21627 + }, + { + "epoch": 1.82, + "learning_rate": 7.0642798789857536e-06, + "loss": 1.1278, + "step": 21628 + }, + { + "epoch": 1.82, + "learning_rate": 7.0634109701801865e-06, + "loss": 1.1677, + "step": 21629 + }, + { + "epoch": 1.82, + "learning_rate": 7.062542085637665e-06, + "loss": 1.2255, + "step": 21630 + }, + { + "epoch": 1.82, + "learning_rate": 7.061673225365356e-06, + "loss": 1.1829, + "step": 21631 + }, + { + "epoch": 1.82, + "learning_rate": 7.060804389370447e-06, + "loss": 1.1263, + "step": 21632 + }, + { + "epoch": 1.82, + "learning_rate": 7.059935577660109e-06, + "loss": 1.2829, + "step": 21633 + }, + { + "epoch": 1.82, + "learning_rate": 7.059066790241526e-06, + "loss": 1.2115, + "step": 21634 + }, + { + "epoch": 1.82, + "learning_rate": 7.058198027121874e-06, + "loss": 1.1392, + "step": 21635 + }, + { + "epoch": 1.82, + "learning_rate": 7.057329288308333e-06, + "loss": 1.1944, + "step": 21636 + }, + { + "epoch": 1.82, + "learning_rate": 7.056460573808075e-06, + "loss": 1.2277, + "step": 21637 + }, + { + "epoch": 1.82, + "learning_rate": 7.055591883628285e-06, + "loss": 1.1689, + "step": 21638 + }, + { + "epoch": 1.82, + "learning_rate": 7.054723217776138e-06, + "loss": 1.2033, + "step": 21639 + }, + { + "epoch": 1.82, + "learning_rate": 7.053854576258807e-06, + "loss": 1.1133, + "step": 21640 + }, + { + "epoch": 1.82, + "learning_rate": 7.052985959083476e-06, + "loss": 1.1991, + "step": 21641 + }, + { + "epoch": 1.82, + "learning_rate": 7.052117366257316e-06, + "loss": 1.1341, + "step": 21642 + }, + { + "epoch": 1.82, + "learning_rate": 7.051248797787509e-06, + "loss": 1.1631, + "step": 21643 + }, + { + "epoch": 1.82, + "learning_rate": 7.050380253681222e-06, + "loss": 1.0554, + "step": 21644 + }, + { + "epoch": 1.82, + "learning_rate": 7.049511733945643e-06, + "loss": 1.1346, + "step": 21645 + }, + { + "epoch": 1.82, + "learning_rate": 7.04864323858794e-06, + "loss": 1.2204, + "step": 21646 + }, + { + "epoch": 1.82, + "learning_rate": 7.047774767615293e-06, + "loss": 1.1649, + "step": 21647 + }, + { + "epoch": 1.82, + "learning_rate": 7.046906321034872e-06, + "loss": 1.1827, + "step": 21648 + }, + { + "epoch": 1.82, + "learning_rate": 7.046037898853857e-06, + "loss": 1.1984, + "step": 21649 + }, + { + "epoch": 1.82, + "learning_rate": 7.0451695010794256e-06, + "loss": 1.2143, + "step": 21650 + }, + { + "epoch": 1.82, + "learning_rate": 7.044301127718745e-06, + "loss": 1.1655, + "step": 21651 + }, + { + "epoch": 1.82, + "learning_rate": 7.043432778778998e-06, + "loss": 1.1663, + "step": 21652 + }, + { + "epoch": 1.82, + "learning_rate": 7.042564454267353e-06, + "loss": 1.1615, + "step": 21653 + }, + { + "epoch": 1.82, + "learning_rate": 7.041696154190991e-06, + "loss": 1.1422, + "step": 21654 + }, + { + "epoch": 1.82, + "learning_rate": 7.040827878557078e-06, + "loss": 1.1536, + "step": 21655 + }, + { + "epoch": 1.82, + "learning_rate": 7.039959627372794e-06, + "loss": 1.1546, + "step": 21656 + }, + { + "epoch": 1.82, + "learning_rate": 7.039091400645311e-06, + "loss": 1.1531, + "step": 21657 + }, + { + "epoch": 1.82, + "learning_rate": 7.038223198381802e-06, + "loss": 1.1551, + "step": 21658 + }, + { + "epoch": 1.82, + "learning_rate": 7.037355020589438e-06, + "loss": 1.2732, + "step": 21659 + }, + { + "epoch": 1.82, + "learning_rate": 7.036486867275403e-06, + "loss": 1.2034, + "step": 21660 + }, + { + "epoch": 1.82, + "learning_rate": 7.035618738446853e-06, + "loss": 1.1723, + "step": 21661 + }, + { + "epoch": 1.82, + "learning_rate": 7.034750634110972e-06, + "loss": 1.1876, + "step": 21662 + }, + { + "epoch": 1.82, + "learning_rate": 7.033882554274935e-06, + "loss": 1.1866, + "step": 21663 + }, + { + "epoch": 1.82, + "learning_rate": 7.033014498945903e-06, + "loss": 1.2905, + "step": 21664 + }, + { + "epoch": 1.82, + "learning_rate": 7.03214646813106e-06, + "loss": 1.1396, + "step": 21665 + }, + { + "epoch": 1.82, + "learning_rate": 7.0312784618375695e-06, + "loss": 1.2201, + "step": 21666 + }, + { + "epoch": 1.82, + "learning_rate": 7.030410480072611e-06, + "loss": 1.136, + "step": 21667 + }, + { + "epoch": 1.82, + "learning_rate": 7.029542522843344e-06, + "loss": 1.1009, + "step": 21668 + }, + { + "epoch": 1.82, + "learning_rate": 7.028674590156954e-06, + "loss": 1.2997, + "step": 21669 + }, + { + "epoch": 1.82, + "learning_rate": 7.0278066820206026e-06, + "loss": 1.1864, + "step": 21670 + }, + { + "epoch": 1.82, + "learning_rate": 7.026938798441466e-06, + "loss": 1.1638, + "step": 21671 + }, + { + "epoch": 1.82, + "learning_rate": 7.026070939426708e-06, + "loss": 1.1606, + "step": 21672 + }, + { + "epoch": 1.82, + "learning_rate": 7.025203104983508e-06, + "loss": 1.1885, + "step": 21673 + }, + { + "epoch": 1.82, + "learning_rate": 7.024335295119029e-06, + "loss": 1.2262, + "step": 21674 + }, + { + "epoch": 1.82, + "learning_rate": 7.023467509840443e-06, + "loss": 1.2049, + "step": 21675 + }, + { + "epoch": 1.82, + "learning_rate": 7.022599749154926e-06, + "loss": 1.1838, + "step": 21676 + }, + { + "epoch": 1.83, + "learning_rate": 7.02173201306964e-06, + "loss": 1.1704, + "step": 21677 + }, + { + "epoch": 1.83, + "learning_rate": 7.020864301591759e-06, + "loss": 1.1574, + "step": 21678 + }, + { + "epoch": 1.83, + "learning_rate": 7.019996614728446e-06, + "loss": 1.159, + "step": 21679 + }, + { + "epoch": 1.83, + "learning_rate": 7.01912895248688e-06, + "loss": 1.133, + "step": 21680 + }, + { + "epoch": 1.83, + "learning_rate": 7.018261314874222e-06, + "loss": 1.1792, + "step": 21681 + }, + { + "epoch": 1.83, + "learning_rate": 7.0173937018976456e-06, + "loss": 1.1568, + "step": 21682 + }, + { + "epoch": 1.83, + "learning_rate": 7.016526113564313e-06, + "loss": 1.1126, + "step": 21683 + }, + { + "epoch": 1.83, + "learning_rate": 7.015658549881399e-06, + "loss": 1.2104, + "step": 21684 + }, + { + "epoch": 1.83, + "learning_rate": 7.014791010856068e-06, + "loss": 1.1729, + "step": 21685 + }, + { + "epoch": 1.83, + "learning_rate": 7.013923496495492e-06, + "loss": 1.2805, + "step": 21686 + }, + { + "epoch": 1.83, + "learning_rate": 7.013056006806831e-06, + "loss": 1.1833, + "step": 21687 + }, + { + "epoch": 1.83, + "learning_rate": 7.01218854179726e-06, + "loss": 1.0696, + "step": 21688 + }, + { + "epoch": 1.83, + "learning_rate": 7.011321101473945e-06, + "loss": 1.1556, + "step": 21689 + }, + { + "epoch": 1.83, + "learning_rate": 7.010453685844047e-06, + "loss": 1.1613, + "step": 21690 + }, + { + "epoch": 1.83, + "learning_rate": 7.009586294914741e-06, + "loss": 1.2544, + "step": 21691 + }, + { + "epoch": 1.83, + "learning_rate": 7.00871892869319e-06, + "loss": 1.2058, + "step": 21692 + }, + { + "epoch": 1.83, + "learning_rate": 7.007851587186561e-06, + "loss": 1.1021, + "step": 21693 + }, + { + "epoch": 1.83, + "learning_rate": 7.006984270402015e-06, + "loss": 1.2412, + "step": 21694 + }, + { + "epoch": 1.83, + "learning_rate": 7.006116978346729e-06, + "loss": 1.2161, + "step": 21695 + }, + { + "epoch": 1.83, + "learning_rate": 7.00524971102786e-06, + "loss": 1.2257, + "step": 21696 + }, + { + "epoch": 1.83, + "learning_rate": 7.004382468452577e-06, + "loss": 1.1836, + "step": 21697 + }, + { + "epoch": 1.83, + "learning_rate": 7.003515250628044e-06, + "loss": 1.1811, + "step": 21698 + }, + { + "epoch": 1.83, + "learning_rate": 7.002648057561425e-06, + "loss": 1.2218, + "step": 21699 + }, + { + "epoch": 1.83, + "learning_rate": 7.001780889259892e-06, + "loss": 1.171, + "step": 21700 + }, + { + "epoch": 1.83, + "learning_rate": 7.0009137457306e-06, + "loss": 1.2899, + "step": 21701 + }, + { + "epoch": 1.83, + "learning_rate": 7.000046626980721e-06, + "loss": 1.1461, + "step": 21702 + }, + { + "epoch": 1.83, + "learning_rate": 6.999179533017416e-06, + "loss": 1.1335, + "step": 21703 + }, + { + "epoch": 1.83, + "learning_rate": 6.9983124638478524e-06, + "loss": 1.1855, + "step": 21704 + }, + { + "epoch": 1.83, + "learning_rate": 6.9974454194791875e-06, + "loss": 1.1491, + "step": 21705 + }, + { + "epoch": 1.83, + "learning_rate": 6.996578399918595e-06, + "loss": 1.2507, + "step": 21706 + }, + { + "epoch": 1.83, + "learning_rate": 6.995711405173229e-06, + "loss": 1.0516, + "step": 21707 + }, + { + "epoch": 1.83, + "learning_rate": 6.994844435250258e-06, + "loss": 1.1635, + "step": 21708 + }, + { + "epoch": 1.83, + "learning_rate": 6.993977490156842e-06, + "loss": 1.1258, + "step": 21709 + }, + { + "epoch": 1.83, + "learning_rate": 6.993110569900153e-06, + "loss": 1.2481, + "step": 21710 + }, + { + "epoch": 1.83, + "learning_rate": 6.992243674487338e-06, + "loss": 1.1626, + "step": 21711 + }, + { + "epoch": 1.83, + "learning_rate": 6.99137680392557e-06, + "loss": 1.2104, + "step": 21712 + }, + { + "epoch": 1.83, + "learning_rate": 6.990509958222016e-06, + "loss": 1.15, + "step": 21713 + }, + { + "epoch": 1.83, + "learning_rate": 6.9896431373838246e-06, + "loss": 1.1159, + "step": 21714 + }, + { + "epoch": 1.83, + "learning_rate": 6.988776341418171e-06, + "loss": 1.1353, + "step": 21715 + }, + { + "epoch": 1.83, + "learning_rate": 6.987909570332207e-06, + "loss": 1.246, + "step": 21716 + }, + { + "epoch": 1.83, + "learning_rate": 6.987042824133102e-06, + "loss": 1.1488, + "step": 21717 + }, + { + "epoch": 1.83, + "learning_rate": 6.986176102828006e-06, + "loss": 1.1855, + "step": 21718 + }, + { + "epoch": 1.83, + "learning_rate": 6.9853094064240945e-06, + "loss": 1.1024, + "step": 21719 + }, + { + "epoch": 1.83, + "learning_rate": 6.984442734928519e-06, + "loss": 1.0737, + "step": 21720 + }, + { + "epoch": 1.83, + "learning_rate": 6.983576088348445e-06, + "loss": 1.1719, + "step": 21721 + }, + { + "epoch": 1.83, + "learning_rate": 6.982709466691027e-06, + "loss": 1.1834, + "step": 21722 + }, + { + "epoch": 1.83, + "learning_rate": 6.981842869963431e-06, + "loss": 1.1037, + "step": 21723 + }, + { + "epoch": 1.83, + "learning_rate": 6.980976298172815e-06, + "loss": 1.1098, + "step": 21724 + }, + { + "epoch": 1.83, + "learning_rate": 6.980109751326336e-06, + "loss": 1.2074, + "step": 21725 + }, + { + "epoch": 1.83, + "learning_rate": 6.979243229431163e-06, + "loss": 1.159, + "step": 21726 + }, + { + "epoch": 1.83, + "learning_rate": 6.978376732494445e-06, + "loss": 1.2083, + "step": 21727 + }, + { + "epoch": 1.83, + "learning_rate": 6.977510260523348e-06, + "loss": 1.1399, + "step": 21728 + }, + { + "epoch": 1.83, + "learning_rate": 6.976643813525024e-06, + "loss": 1.1683, + "step": 21729 + }, + { + "epoch": 1.83, + "learning_rate": 6.975777391506642e-06, + "loss": 1.1885, + "step": 21730 + }, + { + "epoch": 1.83, + "learning_rate": 6.974910994475351e-06, + "loss": 1.0913, + "step": 21731 + }, + { + "epoch": 1.83, + "learning_rate": 6.974044622438317e-06, + "loss": 1.1614, + "step": 21732 + }, + { + "epoch": 1.83, + "learning_rate": 6.97317827540269e-06, + "loss": 1.1408, + "step": 21733 + }, + { + "epoch": 1.83, + "learning_rate": 6.972311953375637e-06, + "loss": 1.2226, + "step": 21734 + }, + { + "epoch": 1.83, + "learning_rate": 6.971445656364311e-06, + "loss": 1.2221, + "step": 21735 + }, + { + "epoch": 1.83, + "learning_rate": 6.9705793843758704e-06, + "loss": 1.2243, + "step": 21736 + }, + { + "epoch": 1.83, + "learning_rate": 6.969713137417469e-06, + "loss": 1.1543, + "step": 21737 + }, + { + "epoch": 1.83, + "learning_rate": 6.968846915496269e-06, + "loss": 1.2533, + "step": 21738 + }, + { + "epoch": 1.83, + "learning_rate": 6.967980718619429e-06, + "loss": 1.2184, + "step": 21739 + }, + { + "epoch": 1.83, + "learning_rate": 6.967114546794098e-06, + "loss": 1.1435, + "step": 21740 + }, + { + "epoch": 1.83, + "learning_rate": 6.966248400027441e-06, + "loss": 1.198, + "step": 21741 + }, + { + "epoch": 1.83, + "learning_rate": 6.965382278326609e-06, + "loss": 1.1429, + "step": 21742 + }, + { + "epoch": 1.83, + "learning_rate": 6.964516181698762e-06, + "loss": 1.1825, + "step": 21743 + }, + { + "epoch": 1.83, + "learning_rate": 6.963650110151049e-06, + "loss": 1.1773, + "step": 21744 + }, + { + "epoch": 1.83, + "learning_rate": 6.962784063690635e-06, + "loss": 1.1234, + "step": 21745 + }, + { + "epoch": 1.83, + "learning_rate": 6.961918042324669e-06, + "loss": 1.2258, + "step": 21746 + }, + { + "epoch": 1.83, + "learning_rate": 6.961052046060311e-06, + "loss": 1.2767, + "step": 21747 + }, + { + "epoch": 1.83, + "learning_rate": 6.960186074904709e-06, + "loss": 1.1597, + "step": 21748 + }, + { + "epoch": 1.83, + "learning_rate": 6.959320128865026e-06, + "loss": 1.1976, + "step": 21749 + }, + { + "epoch": 1.83, + "learning_rate": 6.958454207948413e-06, + "loss": 1.1916, + "step": 21750 + }, + { + "epoch": 1.83, + "learning_rate": 6.957588312162022e-06, + "loss": 1.1435, + "step": 21751 + }, + { + "epoch": 1.83, + "learning_rate": 6.956722441513013e-06, + "loss": 1.0691, + "step": 21752 + }, + { + "epoch": 1.83, + "learning_rate": 6.955856596008537e-06, + "loss": 1.2595, + "step": 21753 + }, + { + "epoch": 1.83, + "learning_rate": 6.954990775655749e-06, + "loss": 1.1516, + "step": 21754 + }, + { + "epoch": 1.83, + "learning_rate": 6.954124980461799e-06, + "loss": 1.1463, + "step": 21755 + }, + { + "epoch": 1.83, + "learning_rate": 6.953259210433844e-06, + "loss": 1.1638, + "step": 21756 + }, + { + "epoch": 1.83, + "learning_rate": 6.952393465579037e-06, + "loss": 1.1737, + "step": 21757 + }, + { + "epoch": 1.83, + "learning_rate": 6.951527745904532e-06, + "loss": 1.2473, + "step": 21758 + }, + { + "epoch": 1.83, + "learning_rate": 6.950662051417476e-06, + "loss": 1.0918, + "step": 21759 + }, + { + "epoch": 1.83, + "learning_rate": 6.949796382125031e-06, + "loss": 1.1332, + "step": 21760 + }, + { + "epoch": 1.83, + "learning_rate": 6.948930738034342e-06, + "loss": 1.1595, + "step": 21761 + }, + { + "epoch": 1.83, + "learning_rate": 6.9480651191525614e-06, + "loss": 1.1182, + "step": 21762 + }, + { + "epoch": 1.83, + "learning_rate": 6.9471995254868475e-06, + "loss": 1.1586, + "step": 21763 + }, + { + "epoch": 1.83, + "learning_rate": 6.9463339570443465e-06, + "loss": 1.0195, + "step": 21764 + }, + { + "epoch": 1.83, + "learning_rate": 6.9454684138322144e-06, + "loss": 1.1744, + "step": 21765 + }, + { + "epoch": 1.83, + "learning_rate": 6.944602895857595e-06, + "loss": 1.2421, + "step": 21766 + }, + { + "epoch": 1.83, + "learning_rate": 6.943737403127651e-06, + "loss": 1.1308, + "step": 21767 + }, + { + "epoch": 1.83, + "learning_rate": 6.942871935649519e-06, + "loss": 1.2336, + "step": 21768 + }, + { + "epoch": 1.83, + "learning_rate": 6.9420064934303625e-06, + "loss": 1.0934, + "step": 21769 + }, + { + "epoch": 1.83, + "learning_rate": 6.941141076477325e-06, + "loss": 1.1435, + "step": 21770 + }, + { + "epoch": 1.83, + "learning_rate": 6.940275684797562e-06, + "loss": 1.2016, + "step": 21771 + }, + { + "epoch": 1.83, + "learning_rate": 6.939410318398214e-06, + "loss": 1.2858, + "step": 21772 + }, + { + "epoch": 1.83, + "learning_rate": 6.938544977286443e-06, + "loss": 1.159, + "step": 21773 + }, + { + "epoch": 1.83, + "learning_rate": 6.937679661469392e-06, + "loss": 1.2849, + "step": 21774 + }, + { + "epoch": 1.83, + "learning_rate": 6.936814370954208e-06, + "loss": 1.227, + "step": 21775 + }, + { + "epoch": 1.83, + "learning_rate": 6.93594910574805e-06, + "loss": 1.3146, + "step": 21776 + }, + { + "epoch": 1.83, + "learning_rate": 6.935083865858056e-06, + "loss": 1.1597, + "step": 21777 + }, + { + "epoch": 1.83, + "learning_rate": 6.934218651291385e-06, + "loss": 1.1875, + "step": 21778 + }, + { + "epoch": 1.83, + "learning_rate": 6.933353462055176e-06, + "loss": 1.073, + "step": 21779 + }, + { + "epoch": 1.83, + "learning_rate": 6.932488298156586e-06, + "loss": 1.2029, + "step": 21780 + }, + { + "epoch": 1.83, + "learning_rate": 6.931623159602756e-06, + "loss": 1.1032, + "step": 21781 + }, + { + "epoch": 1.83, + "learning_rate": 6.930758046400841e-06, + "loss": 1.1464, + "step": 21782 + }, + { + "epoch": 1.83, + "learning_rate": 6.929892958557981e-06, + "loss": 1.1798, + "step": 21783 + }, + { + "epoch": 1.83, + "learning_rate": 6.9290278960813326e-06, + "loss": 1.1676, + "step": 21784 + }, + { + "epoch": 1.83, + "learning_rate": 6.928162858978036e-06, + "loss": 1.1881, + "step": 21785 + }, + { + "epoch": 1.83, + "learning_rate": 6.9272978472552424e-06, + "loss": 1.1915, + "step": 21786 + }, + { + "epoch": 1.83, + "learning_rate": 6.926432860920093e-06, + "loss": 1.1756, + "step": 21787 + }, + { + "epoch": 1.83, + "learning_rate": 6.925567899979742e-06, + "loss": 1.1944, + "step": 21788 + }, + { + "epoch": 1.83, + "learning_rate": 6.924702964441335e-06, + "loss": 1.1491, + "step": 21789 + }, + { + "epoch": 1.83, + "learning_rate": 6.923838054312012e-06, + "loss": 1.1288, + "step": 21790 + }, + { + "epoch": 1.83, + "learning_rate": 6.922973169598928e-06, + "loss": 1.1809, + "step": 21791 + }, + { + "epoch": 1.83, + "learning_rate": 6.922108310309221e-06, + "loss": 1.2181, + "step": 21792 + }, + { + "epoch": 1.83, + "learning_rate": 6.9212434764500435e-06, + "loss": 1.2133, + "step": 21793 + }, + { + "epoch": 1.83, + "learning_rate": 6.920378668028533e-06, + "loss": 1.2209, + "step": 21794 + }, + { + "epoch": 1.83, + "learning_rate": 6.919513885051845e-06, + "loss": 1.1781, + "step": 21795 + }, + { + "epoch": 1.84, + "learning_rate": 6.918649127527116e-06, + "loss": 1.1613, + "step": 21796 + }, + { + "epoch": 1.84, + "learning_rate": 6.917784395461495e-06, + "loss": 1.1394, + "step": 21797 + }, + { + "epoch": 1.84, + "learning_rate": 6.916919688862124e-06, + "loss": 1.1944, + "step": 21798 + }, + { + "epoch": 1.84, + "learning_rate": 6.91605500773615e-06, + "loss": 1.1866, + "step": 21799 + }, + { + "epoch": 1.84, + "learning_rate": 6.915190352090719e-06, + "loss": 1.2541, + "step": 21800 + }, + { + "epoch": 1.84, + "learning_rate": 6.914325721932968e-06, + "loss": 1.1169, + "step": 21801 + }, + { + "epoch": 1.84, + "learning_rate": 6.9134611172700505e-06, + "loss": 1.2614, + "step": 21802 + }, + { + "epoch": 1.84, + "learning_rate": 6.912596538109102e-06, + "loss": 1.1909, + "step": 21803 + }, + { + "epoch": 1.84, + "learning_rate": 6.911731984457271e-06, + "loss": 1.1103, + "step": 21804 + }, + { + "epoch": 1.84, + "learning_rate": 6.9108674563216945e-06, + "loss": 1.1907, + "step": 21805 + }, + { + "epoch": 1.84, + "learning_rate": 6.910002953709525e-06, + "loss": 1.2101, + "step": 21806 + }, + { + "epoch": 1.84, + "learning_rate": 6.909138476627897e-06, + "loss": 1.2808, + "step": 21807 + }, + { + "epoch": 1.84, + "learning_rate": 6.9082740250839584e-06, + "loss": 1.2062, + "step": 21808 + }, + { + "epoch": 1.84, + "learning_rate": 6.907409599084845e-06, + "loss": 1.2322, + "step": 21809 + }, + { + "epoch": 1.84, + "learning_rate": 6.906545198637707e-06, + "loss": 1.0995, + "step": 21810 + }, + { + "epoch": 1.84, + "learning_rate": 6.905680823749682e-06, + "loss": 1.2112, + "step": 21811 + }, + { + "epoch": 1.84, + "learning_rate": 6.904816474427909e-06, + "loss": 1.2135, + "step": 21812 + }, + { + "epoch": 1.84, + "learning_rate": 6.903952150679538e-06, + "loss": 1.1108, + "step": 21813 + }, + { + "epoch": 1.84, + "learning_rate": 6.9030878525117015e-06, + "loss": 1.0922, + "step": 21814 + }, + { + "epoch": 1.84, + "learning_rate": 6.902223579931548e-06, + "loss": 1.2226, + "step": 21815 + }, + { + "epoch": 1.84, + "learning_rate": 6.9013593329462095e-06, + "loss": 1.2031, + "step": 21816 + }, + { + "epoch": 1.84, + "learning_rate": 6.900495111562839e-06, + "loss": 1.1718, + "step": 21817 + }, + { + "epoch": 1.84, + "learning_rate": 6.899630915788562e-06, + "loss": 1.2298, + "step": 21818 + }, + { + "epoch": 1.84, + "learning_rate": 6.898766745630533e-06, + "loss": 1.155, + "step": 21819 + }, + { + "epoch": 1.84, + "learning_rate": 6.8979026010958825e-06, + "loss": 1.2047, + "step": 21820 + }, + { + "epoch": 1.84, + "learning_rate": 6.897038482191755e-06, + "loss": 1.1697, + "step": 21821 + }, + { + "epoch": 1.84, + "learning_rate": 6.896174388925285e-06, + "loss": 1.1054, + "step": 21822 + }, + { + "epoch": 1.84, + "learning_rate": 6.895310321303621e-06, + "loss": 1.1777, + "step": 21823 + }, + { + "epoch": 1.84, + "learning_rate": 6.894446279333893e-06, + "loss": 1.2059, + "step": 21824 + }, + { + "epoch": 1.84, + "learning_rate": 6.893582263023242e-06, + "loss": 1.1322, + "step": 21825 + }, + { + "epoch": 1.84, + "learning_rate": 6.892718272378813e-06, + "loss": 1.1277, + "step": 21826 + }, + { + "epoch": 1.84, + "learning_rate": 6.8918543074077375e-06, + "loss": 1.1386, + "step": 21827 + }, + { + "epoch": 1.84, + "learning_rate": 6.890990368117159e-06, + "loss": 1.1724, + "step": 21828 + }, + { + "epoch": 1.84, + "learning_rate": 6.890126454514207e-06, + "loss": 1.1774, + "step": 21829 + }, + { + "epoch": 1.84, + "learning_rate": 6.889262566606033e-06, + "loss": 1.2161, + "step": 21830 + }, + { + "epoch": 1.84, + "learning_rate": 6.8883987043997635e-06, + "loss": 1.1694, + "step": 21831 + }, + { + "epoch": 1.84, + "learning_rate": 6.8875348679025425e-06, + "loss": 1.1038, + "step": 21832 + }, + { + "epoch": 1.84, + "learning_rate": 6.886671057121499e-06, + "loss": 1.1544, + "step": 21833 + }, + { + "epoch": 1.84, + "learning_rate": 6.88580727206378e-06, + "loss": 1.079, + "step": 21834 + }, + { + "epoch": 1.84, + "learning_rate": 6.884943512736517e-06, + "loss": 1.1527, + "step": 21835 + }, + { + "epoch": 1.84, + "learning_rate": 6.8840797791468496e-06, + "loss": 1.1627, + "step": 21836 + }, + { + "epoch": 1.84, + "learning_rate": 6.883216071301908e-06, + "loss": 1.1951, + "step": 21837 + }, + { + "epoch": 1.84, + "learning_rate": 6.882352389208834e-06, + "loss": 1.193, + "step": 21838 + }, + { + "epoch": 1.84, + "learning_rate": 6.881488732874766e-06, + "loss": 1.1365, + "step": 21839 + }, + { + "epoch": 1.84, + "learning_rate": 6.8806251023068305e-06, + "loss": 1.0826, + "step": 21840 + }, + { + "epoch": 1.84, + "learning_rate": 6.879761497512173e-06, + "loss": 1.13, + "step": 21841 + }, + { + "epoch": 1.84, + "learning_rate": 6.878897918497922e-06, + "loss": 1.191, + "step": 21842 + }, + { + "epoch": 1.84, + "learning_rate": 6.878034365271219e-06, + "loss": 1.1932, + "step": 21843 + }, + { + "epoch": 1.84, + "learning_rate": 6.877170837839191e-06, + "loss": 1.1285, + "step": 21844 + }, + { + "epoch": 1.84, + "learning_rate": 6.876307336208981e-06, + "loss": 1.2717, + "step": 21845 + }, + { + "epoch": 1.84, + "learning_rate": 6.875443860387716e-06, + "loss": 1.1343, + "step": 21846 + }, + { + "epoch": 1.84, + "learning_rate": 6.874580410382536e-06, + "loss": 1.1897, + "step": 21847 + }, + { + "epoch": 1.84, + "learning_rate": 6.873716986200569e-06, + "loss": 1.2201, + "step": 21848 + }, + { + "epoch": 1.84, + "learning_rate": 6.872853587848956e-06, + "loss": 1.1699, + "step": 21849 + }, + { + "epoch": 1.84, + "learning_rate": 6.87199021533483e-06, + "loss": 1.2742, + "step": 21850 + }, + { + "epoch": 1.84, + "learning_rate": 6.871126868665316e-06, + "loss": 1.187, + "step": 21851 + }, + { + "epoch": 1.84, + "learning_rate": 6.870263547847557e-06, + "loss": 1.1089, + "step": 21852 + }, + { + "epoch": 1.84, + "learning_rate": 6.869400252888683e-06, + "loss": 1.1575, + "step": 21853 + }, + { + "epoch": 1.84, + "learning_rate": 6.868536983795826e-06, + "loss": 1.1911, + "step": 21854 + }, + { + "epoch": 1.84, + "learning_rate": 6.867673740576114e-06, + "loss": 1.2243, + "step": 21855 + }, + { + "epoch": 1.84, + "learning_rate": 6.866810523236689e-06, + "loss": 1.2489, + "step": 21856 + }, + { + "epoch": 1.84, + "learning_rate": 6.865947331784676e-06, + "loss": 1.1515, + "step": 21857 + }, + { + "epoch": 1.84, + "learning_rate": 6.865084166227212e-06, + "loss": 1.1992, + "step": 21858 + }, + { + "epoch": 1.84, + "learning_rate": 6.86422102657142e-06, + "loss": 1.2164, + "step": 21859 + }, + { + "epoch": 1.84, + "learning_rate": 6.8633579128244434e-06, + "loss": 1.2363, + "step": 21860 + }, + { + "epoch": 1.84, + "learning_rate": 6.862494824993404e-06, + "loss": 1.2289, + "step": 21861 + }, + { + "epoch": 1.84, + "learning_rate": 6.861631763085436e-06, + "loss": 1.2175, + "step": 21862 + }, + { + "epoch": 1.84, + "learning_rate": 6.860768727107674e-06, + "loss": 1.2346, + "step": 21863 + }, + { + "epoch": 1.84, + "learning_rate": 6.859905717067243e-06, + "loss": 1.2008, + "step": 21864 + }, + { + "epoch": 1.84, + "learning_rate": 6.8590427329712794e-06, + "loss": 1.1977, + "step": 21865 + }, + { + "epoch": 1.84, + "learning_rate": 6.858179774826904e-06, + "loss": 1.1792, + "step": 21866 + }, + { + "epoch": 1.84, + "learning_rate": 6.857316842641259e-06, + "loss": 1.2311, + "step": 21867 + }, + { + "epoch": 1.84, + "learning_rate": 6.856453936421463e-06, + "loss": 1.1462, + "step": 21868 + }, + { + "epoch": 1.84, + "learning_rate": 6.8555910561746555e-06, + "loss": 1.2337, + "step": 21869 + }, + { + "epoch": 1.84, + "learning_rate": 6.854728201907958e-06, + "loss": 1.126, + "step": 21870 + }, + { + "epoch": 1.84, + "learning_rate": 6.8538653736285054e-06, + "loss": 1.1296, + "step": 21871 + }, + { + "epoch": 1.84, + "learning_rate": 6.853002571343419e-06, + "loss": 1.3065, + "step": 21872 + }, + { + "epoch": 1.84, + "learning_rate": 6.8521397950598366e-06, + "loss": 1.234, + "step": 21873 + }, + { + "epoch": 1.84, + "learning_rate": 6.851277044784881e-06, + "loss": 1.1765, + "step": 21874 + }, + { + "epoch": 1.84, + "learning_rate": 6.8504143205256804e-06, + "loss": 1.2246, + "step": 21875 + }, + { + "epoch": 1.84, + "learning_rate": 6.849551622289367e-06, + "loss": 1.1418, + "step": 21876 + }, + { + "epoch": 1.84, + "learning_rate": 6.848688950083067e-06, + "loss": 1.1347, + "step": 21877 + }, + { + "epoch": 1.84, + "learning_rate": 6.847826303913908e-06, + "loss": 1.1858, + "step": 21878 + }, + { + "epoch": 1.84, + "learning_rate": 6.846963683789013e-06, + "loss": 1.1095, + "step": 21879 + }, + { + "epoch": 1.84, + "learning_rate": 6.846101089715518e-06, + "loss": 1.1463, + "step": 21880 + }, + { + "epoch": 1.84, + "learning_rate": 6.8452385217005416e-06, + "loss": 1.1989, + "step": 21881 + }, + { + "epoch": 1.84, + "learning_rate": 6.844375979751217e-06, + "loss": 1.1389, + "step": 21882 + }, + { + "epoch": 1.84, + "learning_rate": 6.843513463874665e-06, + "loss": 1.1685, + "step": 21883 + }, + { + "epoch": 1.84, + "learning_rate": 6.842650974078018e-06, + "loss": 1.2067, + "step": 21884 + }, + { + "epoch": 1.84, + "learning_rate": 6.841788510368397e-06, + "loss": 1.2402, + "step": 21885 + }, + { + "epoch": 1.84, + "learning_rate": 6.8409260727529315e-06, + "loss": 1.2272, + "step": 21886 + }, + { + "epoch": 1.84, + "learning_rate": 6.840063661238743e-06, + "loss": 1.0828, + "step": 21887 + }, + { + "epoch": 1.84, + "learning_rate": 6.8392012758329605e-06, + "loss": 1.0999, + "step": 21888 + }, + { + "epoch": 1.84, + "learning_rate": 6.838338916542711e-06, + "loss": 1.2289, + "step": 21889 + }, + { + "epoch": 1.84, + "learning_rate": 6.837476583375113e-06, + "loss": 1.1873, + "step": 21890 + }, + { + "epoch": 1.84, + "learning_rate": 6.836614276337301e-06, + "loss": 1.3168, + "step": 21891 + }, + { + "epoch": 1.84, + "learning_rate": 6.8357519954363925e-06, + "loss": 1.3652, + "step": 21892 + }, + { + "epoch": 1.84, + "learning_rate": 6.8348897406795144e-06, + "loss": 1.2002, + "step": 21893 + }, + { + "epoch": 1.84, + "learning_rate": 6.834027512073787e-06, + "loss": 1.1208, + "step": 21894 + }, + { + "epoch": 1.84, + "learning_rate": 6.8331653096263405e-06, + "loss": 1.0983, + "step": 21895 + }, + { + "epoch": 1.84, + "learning_rate": 6.832303133344294e-06, + "loss": 1.1526, + "step": 21896 + }, + { + "epoch": 1.84, + "learning_rate": 6.831440983234776e-06, + "loss": 1.2124, + "step": 21897 + }, + { + "epoch": 1.84, + "learning_rate": 6.830578859304901e-06, + "loss": 1.2363, + "step": 21898 + }, + { + "epoch": 1.84, + "learning_rate": 6.829716761561801e-06, + "loss": 1.1643, + "step": 21899 + }, + { + "epoch": 1.84, + "learning_rate": 6.828854690012598e-06, + "loss": 1.2558, + "step": 21900 + }, + { + "epoch": 1.84, + "learning_rate": 6.827992644664407e-06, + "loss": 1.1956, + "step": 21901 + }, + { + "epoch": 1.84, + "learning_rate": 6.827130625524361e-06, + "loss": 1.0864, + "step": 21902 + }, + { + "epoch": 1.84, + "learning_rate": 6.8262686325995756e-06, + "loss": 1.1499, + "step": 21903 + }, + { + "epoch": 1.84, + "learning_rate": 6.825406665897176e-06, + "loss": 1.2247, + "step": 21904 + }, + { + "epoch": 1.84, + "learning_rate": 6.824544725424278e-06, + "loss": 1.1378, + "step": 21905 + }, + { + "epoch": 1.84, + "learning_rate": 6.8236828111880125e-06, + "loss": 1.2083, + "step": 21906 + }, + { + "epoch": 1.84, + "learning_rate": 6.822820923195493e-06, + "loss": 1.18, + "step": 21907 + }, + { + "epoch": 1.84, + "learning_rate": 6.821959061453847e-06, + "loss": 1.1237, + "step": 21908 + }, + { + "epoch": 1.84, + "learning_rate": 6.821097225970188e-06, + "loss": 1.1145, + "step": 21909 + }, + { + "epoch": 1.84, + "learning_rate": 6.820235416751645e-06, + "loss": 1.1464, + "step": 21910 + }, + { + "epoch": 1.84, + "learning_rate": 6.8193736338053325e-06, + "loss": 1.2641, + "step": 21911 + }, + { + "epoch": 1.84, + "learning_rate": 6.818511877138372e-06, + "loss": 1.1588, + "step": 21912 + }, + { + "epoch": 1.84, + "learning_rate": 6.817650146757887e-06, + "loss": 1.2446, + "step": 21913 + }, + { + "epoch": 1.84, + "learning_rate": 6.816788442670994e-06, + "loss": 1.1481, + "step": 21914 + }, + { + "epoch": 1.85, + "learning_rate": 6.815926764884815e-06, + "loss": 1.222, + "step": 21915 + }, + { + "epoch": 1.85, + "learning_rate": 6.815065113406465e-06, + "loss": 1.162, + "step": 21916 + }, + { + "epoch": 1.85, + "learning_rate": 6.8142034882430695e-06, + "loss": 1.2289, + "step": 21917 + }, + { + "epoch": 1.85, + "learning_rate": 6.813341889401743e-06, + "loss": 1.1123, + "step": 21918 + }, + { + "epoch": 1.85, + "learning_rate": 6.812480316889607e-06, + "loss": 1.1604, + "step": 21919 + }, + { + "epoch": 1.85, + "learning_rate": 6.811618770713775e-06, + "loss": 1.1142, + "step": 21920 + }, + { + "epoch": 1.85, + "learning_rate": 6.8107572508813765e-06, + "loss": 1.1683, + "step": 21921 + }, + { + "epoch": 1.85, + "learning_rate": 6.809895757399514e-06, + "loss": 1.2494, + "step": 21922 + }, + { + "epoch": 1.85, + "learning_rate": 6.80903429027532e-06, + "loss": 1.1257, + "step": 21923 + }, + { + "epoch": 1.85, + "learning_rate": 6.808172849515902e-06, + "loss": 1.2472, + "step": 21924 + }, + { + "epoch": 1.85, + "learning_rate": 6.807311435128381e-06, + "loss": 1.1228, + "step": 21925 + }, + { + "epoch": 1.85, + "learning_rate": 6.806450047119877e-06, + "loss": 1.2421, + "step": 21926 + }, + { + "epoch": 1.85, + "learning_rate": 6.805588685497504e-06, + "loss": 1.2411, + "step": 21927 + }, + { + "epoch": 1.85, + "learning_rate": 6.8047273502683804e-06, + "loss": 1.2081, + "step": 21928 + }, + { + "epoch": 1.85, + "learning_rate": 6.8038660414396196e-06, + "loss": 1.1515, + "step": 21929 + }, + { + "epoch": 1.85, + "learning_rate": 6.803004759018343e-06, + "loss": 1.2102, + "step": 21930 + }, + { + "epoch": 1.85, + "learning_rate": 6.802143503011663e-06, + "loss": 1.2445, + "step": 21931 + }, + { + "epoch": 1.85, + "learning_rate": 6.8012822734267e-06, + "loss": 1.1439, + "step": 21932 + }, + { + "epoch": 1.85, + "learning_rate": 6.800421070270561e-06, + "loss": 1.1313, + "step": 21933 + }, + { + "epoch": 1.85, + "learning_rate": 6.799559893550372e-06, + "loss": 1.1767, + "step": 21934 + }, + { + "epoch": 1.85, + "learning_rate": 6.79869874327324e-06, + "loss": 1.1672, + "step": 21935 + }, + { + "epoch": 1.85, + "learning_rate": 6.797837619446287e-06, + "loss": 1.202, + "step": 21936 + }, + { + "epoch": 1.85, + "learning_rate": 6.79697652207662e-06, + "loss": 1.175, + "step": 21937 + }, + { + "epoch": 1.85, + "learning_rate": 6.796115451171359e-06, + "loss": 1.2777, + "step": 21938 + }, + { + "epoch": 1.85, + "learning_rate": 6.795254406737622e-06, + "loss": 1.2445, + "step": 21939 + }, + { + "epoch": 1.85, + "learning_rate": 6.794393388782512e-06, + "loss": 1.1071, + "step": 21940 + }, + { + "epoch": 1.85, + "learning_rate": 6.7935323973131565e-06, + "loss": 1.2139, + "step": 21941 + }, + { + "epoch": 1.85, + "learning_rate": 6.792671432336658e-06, + "loss": 1.1432, + "step": 21942 + }, + { + "epoch": 1.85, + "learning_rate": 6.791810493860138e-06, + "loss": 1.2199, + "step": 21943 + }, + { + "epoch": 1.85, + "learning_rate": 6.790949581890702e-06, + "loss": 1.2368, + "step": 21944 + }, + { + "epoch": 1.85, + "learning_rate": 6.790088696435471e-06, + "loss": 1.2422, + "step": 21945 + }, + { + "epoch": 1.85, + "learning_rate": 6.789227837501553e-06, + "loss": 1.2729, + "step": 21946 + }, + { + "epoch": 1.85, + "learning_rate": 6.788367005096065e-06, + "loss": 1.1892, + "step": 21947 + }, + { + "epoch": 1.85, + "learning_rate": 6.787506199226113e-06, + "loss": 1.1322, + "step": 21948 + }, + { + "epoch": 1.85, + "learning_rate": 6.786645419898813e-06, + "loss": 1.1474, + "step": 21949 + }, + { + "epoch": 1.85, + "learning_rate": 6.78578466712128e-06, + "loss": 1.202, + "step": 21950 + }, + { + "epoch": 1.85, + "learning_rate": 6.784923940900619e-06, + "loss": 1.1948, + "step": 21951 + }, + { + "epoch": 1.85, + "learning_rate": 6.784063241243948e-06, + "loss": 1.1209, + "step": 21952 + }, + { + "epoch": 1.85, + "learning_rate": 6.783202568158375e-06, + "loss": 1.1384, + "step": 21953 + }, + { + "epoch": 1.85, + "learning_rate": 6.782341921651014e-06, + "loss": 1.101, + "step": 21954 + }, + { + "epoch": 1.85, + "learning_rate": 6.781481301728969e-06, + "loss": 1.1341, + "step": 21955 + }, + { + "epoch": 1.85, + "learning_rate": 6.780620708399359e-06, + "loss": 1.1986, + "step": 21956 + }, + { + "epoch": 1.85, + "learning_rate": 6.7797601416692916e-06, + "loss": 1.2445, + "step": 21957 + }, + { + "epoch": 1.85, + "learning_rate": 6.7788996015458765e-06, + "loss": 1.1991, + "step": 21958 + }, + { + "epoch": 1.85, + "learning_rate": 6.7780390880362205e-06, + "loss": 1.1879, + "step": 21959 + }, + { + "epoch": 1.85, + "learning_rate": 6.77717860114744e-06, + "loss": 1.154, + "step": 21960 + }, + { + "epoch": 1.85, + "learning_rate": 6.776318140886639e-06, + "loss": 1.1697, + "step": 21961 + }, + { + "epoch": 1.85, + "learning_rate": 6.775457707260928e-06, + "loss": 1.0943, + "step": 21962 + }, + { + "epoch": 1.85, + "learning_rate": 6.774597300277421e-06, + "loss": 1.121, + "step": 21963 + }, + { + "epoch": 1.85, + "learning_rate": 6.7737369199432235e-06, + "loss": 1.2671, + "step": 21964 + }, + { + "epoch": 1.85, + "learning_rate": 6.772876566265444e-06, + "loss": 1.2379, + "step": 21965 + }, + { + "epoch": 1.85, + "learning_rate": 6.772016239251187e-06, + "loss": 1.1905, + "step": 21966 + }, + { + "epoch": 1.85, + "learning_rate": 6.77115593890757e-06, + "loss": 1.2012, + "step": 21967 + }, + { + "epoch": 1.85, + "learning_rate": 6.770295665241693e-06, + "loss": 1.2004, + "step": 21968 + }, + { + "epoch": 1.85, + "learning_rate": 6.76943541826067e-06, + "loss": 1.2767, + "step": 21969 + }, + { + "epoch": 1.85, + "learning_rate": 6.7685751979716e-06, + "loss": 1.1176, + "step": 21970 + }, + { + "epoch": 1.85, + "learning_rate": 6.767715004381603e-06, + "loss": 1.2279, + "step": 21971 + }, + { + "epoch": 1.85, + "learning_rate": 6.7668548374977715e-06, + "loss": 1.105, + "step": 21972 + }, + { + "epoch": 1.85, + "learning_rate": 6.7659946973272264e-06, + "loss": 1.1215, + "step": 21973 + }, + { + "epoch": 1.85, + "learning_rate": 6.765134583877064e-06, + "loss": 1.1236, + "step": 21974 + }, + { + "epoch": 1.85, + "learning_rate": 6.764274497154395e-06, + "loss": 1.1706, + "step": 21975 + }, + { + "epoch": 1.85, + "learning_rate": 6.763414437166329e-06, + "loss": 1.2022, + "step": 21976 + }, + { + "epoch": 1.85, + "learning_rate": 6.762554403919965e-06, + "loss": 1.2257, + "step": 21977 + }, + { + "epoch": 1.85, + "learning_rate": 6.761694397422417e-06, + "loss": 1.1763, + "step": 21978 + }, + { + "epoch": 1.85, + "learning_rate": 6.760834417680782e-06, + "loss": 1.2323, + "step": 21979 + }, + { + "epoch": 1.85, + "learning_rate": 6.759974464702173e-06, + "loss": 1.2005, + "step": 21980 + }, + { + "epoch": 1.85, + "learning_rate": 6.75911453849369e-06, + "loss": 1.1256, + "step": 21981 + }, + { + "epoch": 1.85, + "learning_rate": 6.75825463906244e-06, + "loss": 1.2027, + "step": 21982 + }, + { + "epoch": 1.85, + "learning_rate": 6.757394766415526e-06, + "loss": 1.1723, + "step": 21983 + }, + { + "epoch": 1.85, + "learning_rate": 6.756534920560058e-06, + "loss": 1.1673, + "step": 21984 + }, + { + "epoch": 1.85, + "learning_rate": 6.755675101503134e-06, + "loss": 1.1699, + "step": 21985 + }, + { + "epoch": 1.85, + "learning_rate": 6.754815309251863e-06, + "loss": 1.1601, + "step": 21986 + }, + { + "epoch": 1.85, + "learning_rate": 6.753955543813342e-06, + "loss": 1.175, + "step": 21987 + }, + { + "epoch": 1.85, + "learning_rate": 6.753095805194683e-06, + "loss": 1.164, + "step": 21988 + }, + { + "epoch": 1.85, + "learning_rate": 6.752236093402985e-06, + "loss": 1.1732, + "step": 21989 + }, + { + "epoch": 1.85, + "learning_rate": 6.751376408445349e-06, + "loss": 1.1522, + "step": 21990 + }, + { + "epoch": 1.85, + "learning_rate": 6.750516750328885e-06, + "loss": 1.1557, + "step": 21991 + }, + { + "epoch": 1.85, + "learning_rate": 6.749657119060688e-06, + "loss": 1.2604, + "step": 21992 + }, + { + "epoch": 1.85, + "learning_rate": 6.748797514647869e-06, + "loss": 1.0982, + "step": 21993 + }, + { + "epoch": 1.85, + "learning_rate": 6.74793793709752e-06, + "loss": 1.2353, + "step": 21994 + }, + { + "epoch": 1.85, + "learning_rate": 6.747078386416752e-06, + "loss": 1.2207, + "step": 21995 + }, + { + "epoch": 1.85, + "learning_rate": 6.746218862612661e-06, + "loss": 1.1562, + "step": 21996 + }, + { + "epoch": 1.85, + "learning_rate": 6.745359365692354e-06, + "loss": 1.132, + "step": 21997 + }, + { + "epoch": 1.85, + "learning_rate": 6.744499895662926e-06, + "loss": 1.1174, + "step": 21998 + }, + { + "epoch": 1.85, + "learning_rate": 6.743640452531483e-06, + "loss": 1.2412, + "step": 21999 + }, + { + "epoch": 1.85, + "learning_rate": 6.742781036305129e-06, + "loss": 1.1743, + "step": 22000 + }, + { + "epoch": 1.85, + "learning_rate": 6.741921646990953e-06, + "loss": 1.2141, + "step": 22001 + }, + { + "epoch": 1.85, + "learning_rate": 6.7410622845960695e-06, + "loss": 1.1743, + "step": 22002 + }, + { + "epoch": 1.85, + "learning_rate": 6.740202949127569e-06, + "loss": 1.2042, + "step": 22003 + }, + { + "epoch": 1.85, + "learning_rate": 6.739343640592557e-06, + "loss": 1.1341, + "step": 22004 + }, + { + "epoch": 1.85, + "learning_rate": 6.7384843589981285e-06, + "loss": 1.2756, + "step": 22005 + }, + { + "epoch": 1.85, + "learning_rate": 6.73762510435139e-06, + "loss": 1.3216, + "step": 22006 + }, + { + "epoch": 1.85, + "learning_rate": 6.7367658766594345e-06, + "loss": 1.1119, + "step": 22007 + }, + { + "epoch": 1.85, + "learning_rate": 6.7359066759293645e-06, + "loss": 1.2361, + "step": 22008 + }, + { + "epoch": 1.85, + "learning_rate": 6.735047502168276e-06, + "loss": 1.1805, + "step": 22009 + }, + { + "epoch": 1.85, + "learning_rate": 6.734188355383273e-06, + "loss": 1.1998, + "step": 22010 + }, + { + "epoch": 1.85, + "learning_rate": 6.73332923558145e-06, + "loss": 1.1564, + "step": 22011 + }, + { + "epoch": 1.85, + "learning_rate": 6.732470142769904e-06, + "loss": 1.2676, + "step": 22012 + }, + { + "epoch": 1.85, + "learning_rate": 6.731611076955739e-06, + "loss": 1.1781, + "step": 22013 + }, + { + "epoch": 1.85, + "learning_rate": 6.730752038146047e-06, + "loss": 1.3179, + "step": 22014 + }, + { + "epoch": 1.85, + "learning_rate": 6.72989302634793e-06, + "loss": 1.192, + "step": 22015 + }, + { + "epoch": 1.85, + "learning_rate": 6.72903404156848e-06, + "loss": 1.2675, + "step": 22016 + }, + { + "epoch": 1.85, + "learning_rate": 6.7281750838148005e-06, + "loss": 1.183, + "step": 22017 + }, + { + "epoch": 1.85, + "learning_rate": 6.7273161530939855e-06, + "loss": 1.1364, + "step": 22018 + }, + { + "epoch": 1.85, + "learning_rate": 6.726457249413132e-06, + "loss": 1.1183, + "step": 22019 + }, + { + "epoch": 1.85, + "learning_rate": 6.725598372779333e-06, + "loss": 1.1666, + "step": 22020 + }, + { + "epoch": 1.85, + "learning_rate": 6.7247395231996944e-06, + "loss": 1.191, + "step": 22021 + }, + { + "epoch": 1.85, + "learning_rate": 6.7238807006813e-06, + "loss": 1.2103, + "step": 22022 + }, + { + "epoch": 1.85, + "learning_rate": 6.723021905231255e-06, + "loss": 1.2267, + "step": 22023 + }, + { + "epoch": 1.85, + "learning_rate": 6.722163136856651e-06, + "loss": 1.2385, + "step": 22024 + }, + { + "epoch": 1.85, + "learning_rate": 6.721304395564582e-06, + "loss": 1.2594, + "step": 22025 + }, + { + "epoch": 1.85, + "learning_rate": 6.720445681362149e-06, + "loss": 1.1391, + "step": 22026 + }, + { + "epoch": 1.85, + "learning_rate": 6.719586994256442e-06, + "loss": 1.186, + "step": 22027 + }, + { + "epoch": 1.85, + "learning_rate": 6.718728334254558e-06, + "loss": 1.2243, + "step": 22028 + }, + { + "epoch": 1.85, + "learning_rate": 6.717869701363586e-06, + "loss": 1.1521, + "step": 22029 + }, + { + "epoch": 1.85, + "learning_rate": 6.7170110955906305e-06, + "loss": 1.0702, + "step": 22030 + }, + { + "epoch": 1.85, + "learning_rate": 6.716152516942778e-06, + "loss": 1.1228, + "step": 22031 + }, + { + "epoch": 1.85, + "learning_rate": 6.7152939654271264e-06, + "loss": 1.2631, + "step": 22032 + }, + { + "epoch": 1.85, + "learning_rate": 6.7144354410507624e-06, + "loss": 1.1663, + "step": 22033 + }, + { + "epoch": 1.86, + "learning_rate": 6.71357694382079e-06, + "loss": 1.1582, + "step": 22034 + }, + { + "epoch": 1.86, + "learning_rate": 6.712718473744293e-06, + "loss": 1.1341, + "step": 22035 + }, + { + "epoch": 1.86, + "learning_rate": 6.7118600308283695e-06, + "loss": 1.1066, + "step": 22036 + }, + { + "epoch": 1.86, + "learning_rate": 6.711001615080108e-06, + "loss": 1.205, + "step": 22037 + }, + { + "epoch": 1.86, + "learning_rate": 6.710143226506607e-06, + "loss": 1.1021, + "step": 22038 + }, + { + "epoch": 1.86, + "learning_rate": 6.709284865114956e-06, + "loss": 1.2393, + "step": 22039 + }, + { + "epoch": 1.86, + "learning_rate": 6.708426530912242e-06, + "loss": 1.1776, + "step": 22040 + }, + { + "epoch": 1.86, + "learning_rate": 6.707568223905566e-06, + "loss": 1.1936, + "step": 22041 + }, + { + "epoch": 1.86, + "learning_rate": 6.7067099441020134e-06, + "loss": 1.0727, + "step": 22042 + }, + { + "epoch": 1.86, + "learning_rate": 6.7058516915086795e-06, + "loss": 1.2271, + "step": 22043 + }, + { + "epoch": 1.86, + "learning_rate": 6.704993466132648e-06, + "loss": 1.2501, + "step": 22044 + }, + { + "epoch": 1.86, + "learning_rate": 6.70413526798102e-06, + "loss": 1.1235, + "step": 22045 + }, + { + "epoch": 1.86, + "learning_rate": 6.703277097060879e-06, + "loss": 1.0609, + "step": 22046 + }, + { + "epoch": 1.86, + "learning_rate": 6.702418953379319e-06, + "loss": 1.3291, + "step": 22047 + }, + { + "epoch": 1.86, + "learning_rate": 6.701560836943426e-06, + "loss": 1.1619, + "step": 22048 + }, + { + "epoch": 1.86, + "learning_rate": 6.700702747760295e-06, + "loss": 1.1974, + "step": 22049 + }, + { + "epoch": 1.86, + "learning_rate": 6.699844685837016e-06, + "loss": 1.1852, + "step": 22050 + }, + { + "epoch": 1.86, + "learning_rate": 6.698986651180673e-06, + "loss": 1.1438, + "step": 22051 + }, + { + "epoch": 1.86, + "learning_rate": 6.698128643798361e-06, + "loss": 1.1694, + "step": 22052 + }, + { + "epoch": 1.86, + "learning_rate": 6.697270663697164e-06, + "loss": 1.1694, + "step": 22053 + }, + { + "epoch": 1.86, + "learning_rate": 6.696412710884178e-06, + "loss": 1.1641, + "step": 22054 + }, + { + "epoch": 1.86, + "learning_rate": 6.695554785366483e-06, + "loss": 1.168, + "step": 22055 + }, + { + "epoch": 1.86, + "learning_rate": 6.694696887151175e-06, + "loss": 1.1769, + "step": 22056 + }, + { + "epoch": 1.86, + "learning_rate": 6.693839016245338e-06, + "loss": 1.1742, + "step": 22057 + }, + { + "epoch": 1.86, + "learning_rate": 6.6929811726560635e-06, + "loss": 1.1802, + "step": 22058 + }, + { + "epoch": 1.86, + "learning_rate": 6.692123356390432e-06, + "loss": 1.2466, + "step": 22059 + }, + { + "epoch": 1.86, + "learning_rate": 6.69126556745554e-06, + "loss": 1.1781, + "step": 22060 + }, + { + "epoch": 1.86, + "learning_rate": 6.69040780585847e-06, + "loss": 1.1168, + "step": 22061 + }, + { + "epoch": 1.86, + "learning_rate": 6.689550071606306e-06, + "loss": 1.106, + "step": 22062 + }, + { + "epoch": 1.86, + "learning_rate": 6.688692364706144e-06, + "loss": 1.1657, + "step": 22063 + }, + { + "epoch": 1.86, + "learning_rate": 6.687834685165065e-06, + "loss": 1.1061, + "step": 22064 + }, + { + "epoch": 1.86, + "learning_rate": 6.686977032990156e-06, + "loss": 1.2203, + "step": 22065 + }, + { + "epoch": 1.86, + "learning_rate": 6.686119408188498e-06, + "loss": 1.2191, + "step": 22066 + }, + { + "epoch": 1.86, + "learning_rate": 6.685261810767188e-06, + "loss": 1.1856, + "step": 22067 + }, + { + "epoch": 1.86, + "learning_rate": 6.6844042407333036e-06, + "loss": 1.2483, + "step": 22068 + }, + { + "epoch": 1.86, + "learning_rate": 6.683546698093934e-06, + "loss": 1.0707, + "step": 22069 + }, + { + "epoch": 1.86, + "learning_rate": 6.682689182856159e-06, + "loss": 1.1727, + "step": 22070 + }, + { + "epoch": 1.86, + "learning_rate": 6.681831695027072e-06, + "loss": 1.1924, + "step": 22071 + }, + { + "epoch": 1.86, + "learning_rate": 6.68097423461375e-06, + "loss": 1.1384, + "step": 22072 + }, + { + "epoch": 1.86, + "learning_rate": 6.680116801623284e-06, + "loss": 1.2209, + "step": 22073 + }, + { + "epoch": 1.86, + "learning_rate": 6.679259396062752e-06, + "loss": 1.0948, + "step": 22074 + }, + { + "epoch": 1.86, + "learning_rate": 6.678402017939243e-06, + "loss": 1.1397, + "step": 22075 + }, + { + "epoch": 1.86, + "learning_rate": 6.677544667259842e-06, + "loss": 1.1894, + "step": 22076 + }, + { + "epoch": 1.86, + "learning_rate": 6.676687344031626e-06, + "loss": 1.2613, + "step": 22077 + }, + { + "epoch": 1.86, + "learning_rate": 6.6758300482616885e-06, + "loss": 1.1449, + "step": 22078 + }, + { + "epoch": 1.86, + "learning_rate": 6.6749727799571016e-06, + "loss": 1.1235, + "step": 22079 + }, + { + "epoch": 1.86, + "learning_rate": 6.674115539124958e-06, + "loss": 1.0977, + "step": 22080 + }, + { + "epoch": 1.86, + "learning_rate": 6.6732583257723336e-06, + "loss": 1.1433, + "step": 22081 + }, + { + "epoch": 1.86, + "learning_rate": 6.672401139906316e-06, + "loss": 1.1143, + "step": 22082 + }, + { + "epoch": 1.86, + "learning_rate": 6.671543981533981e-06, + "loss": 1.188, + "step": 22083 + }, + { + "epoch": 1.86, + "learning_rate": 6.670686850662419e-06, + "loss": 1.1835, + "step": 22084 + }, + { + "epoch": 1.86, + "learning_rate": 6.669829747298705e-06, + "loss": 1.1372, + "step": 22085 + }, + { + "epoch": 1.86, + "learning_rate": 6.668972671449927e-06, + "loss": 1.1889, + "step": 22086 + }, + { + "epoch": 1.86, + "learning_rate": 6.668115623123157e-06, + "loss": 1.1374, + "step": 22087 + }, + { + "epoch": 1.86, + "learning_rate": 6.667258602325485e-06, + "loss": 1.2258, + "step": 22088 + }, + { + "epoch": 1.86, + "learning_rate": 6.66640160906399e-06, + "loss": 1.0788, + "step": 22089 + }, + { + "epoch": 1.86, + "learning_rate": 6.66554464334575e-06, + "loss": 1.1556, + "step": 22090 + }, + { + "epoch": 1.86, + "learning_rate": 6.66468770517785e-06, + "loss": 1.1615, + "step": 22091 + }, + { + "epoch": 1.86, + "learning_rate": 6.663830794567365e-06, + "loss": 1.2288, + "step": 22092 + }, + { + "epoch": 1.86, + "learning_rate": 6.662973911521381e-06, + "loss": 1.1064, + "step": 22093 + }, + { + "epoch": 1.86, + "learning_rate": 6.66211705604697e-06, + "loss": 1.2351, + "step": 22094 + }, + { + "epoch": 1.86, + "learning_rate": 6.661260228151219e-06, + "loss": 1.2555, + "step": 22095 + }, + { + "epoch": 1.86, + "learning_rate": 6.660403427841204e-06, + "loss": 1.1761, + "step": 22096 + }, + { + "epoch": 1.86, + "learning_rate": 6.659546655124007e-06, + "loss": 1.2433, + "step": 22097 + }, + { + "epoch": 1.86, + "learning_rate": 6.6586899100066995e-06, + "loss": 1.2058, + "step": 22098 + }, + { + "epoch": 1.86, + "learning_rate": 6.657833192496368e-06, + "loss": 1.1411, + "step": 22099 + }, + { + "epoch": 1.86, + "learning_rate": 6.65697650260009e-06, + "loss": 1.19, + "step": 22100 + }, + { + "epoch": 1.86, + "learning_rate": 6.656119840324938e-06, + "loss": 1.175, + "step": 22101 + }, + { + "epoch": 1.86, + "learning_rate": 6.655263205677997e-06, + "loss": 1.1348, + "step": 22102 + }, + { + "epoch": 1.86, + "learning_rate": 6.654406598666341e-06, + "loss": 1.2118, + "step": 22103 + }, + { + "epoch": 1.86, + "learning_rate": 6.65355001929705e-06, + "loss": 1.2263, + "step": 22104 + }, + { + "epoch": 1.86, + "learning_rate": 6.652693467577197e-06, + "loss": 1.2028, + "step": 22105 + }, + { + "epoch": 1.86, + "learning_rate": 6.6518369435138655e-06, + "loss": 1.2612, + "step": 22106 + }, + { + "epoch": 1.86, + "learning_rate": 6.650980447114126e-06, + "loss": 1.167, + "step": 22107 + }, + { + "epoch": 1.86, + "learning_rate": 6.650123978385059e-06, + "loss": 1.0976, + "step": 22108 + }, + { + "epoch": 1.86, + "learning_rate": 6.649267537333736e-06, + "loss": 1.2054, + "step": 22109 + }, + { + "epoch": 1.86, + "learning_rate": 6.648411123967243e-06, + "loss": 1.1892, + "step": 22110 + }, + { + "epoch": 1.86, + "learning_rate": 6.647554738292648e-06, + "loss": 1.151, + "step": 22111 + }, + { + "epoch": 1.86, + "learning_rate": 6.646698380317025e-06, + "loss": 1.2547, + "step": 22112 + }, + { + "epoch": 1.86, + "learning_rate": 6.645842050047457e-06, + "loss": 1.075, + "step": 22113 + }, + { + "epoch": 1.86, + "learning_rate": 6.644985747491014e-06, + "loss": 1.1961, + "step": 22114 + }, + { + "epoch": 1.86, + "learning_rate": 6.644129472654774e-06, + "loss": 1.1967, + "step": 22115 + }, + { + "epoch": 1.86, + "learning_rate": 6.643273225545808e-06, + "loss": 1.2294, + "step": 22116 + }, + { + "epoch": 1.86, + "learning_rate": 6.642417006171196e-06, + "loss": 1.2685, + "step": 22117 + }, + { + "epoch": 1.86, + "learning_rate": 6.641560814538006e-06, + "loss": 1.082, + "step": 22118 + }, + { + "epoch": 1.86, + "learning_rate": 6.640704650653318e-06, + "loss": 1.1631, + "step": 22119 + }, + { + "epoch": 1.86, + "learning_rate": 6.6398485145242e-06, + "loss": 1.1692, + "step": 22120 + }, + { + "epoch": 1.86, + "learning_rate": 6.638992406157733e-06, + "loss": 1.2033, + "step": 22121 + }, + { + "epoch": 1.86, + "learning_rate": 6.638136325560984e-06, + "loss": 1.1542, + "step": 22122 + }, + { + "epoch": 1.86, + "learning_rate": 6.6372802727410306e-06, + "loss": 1.1992, + "step": 22123 + }, + { + "epoch": 1.86, + "learning_rate": 6.63642424770494e-06, + "loss": 1.2191, + "step": 22124 + }, + { + "epoch": 1.86, + "learning_rate": 6.635568250459789e-06, + "loss": 1.2116, + "step": 22125 + }, + { + "epoch": 1.86, + "learning_rate": 6.634712281012654e-06, + "loss": 1.161, + "step": 22126 + }, + { + "epoch": 1.86, + "learning_rate": 6.633856339370599e-06, + "loss": 1.1435, + "step": 22127 + }, + { + "epoch": 1.86, + "learning_rate": 6.633000425540705e-06, + "loss": 1.1981, + "step": 22128 + }, + { + "epoch": 1.86, + "learning_rate": 6.6321445395300346e-06, + "loss": 1.1943, + "step": 22129 + }, + { + "epoch": 1.86, + "learning_rate": 6.631288681345668e-06, + "loss": 1.2043, + "step": 22130 + }, + { + "epoch": 1.86, + "learning_rate": 6.6304328509946695e-06, + "loss": 1.2095, + "step": 22131 + }, + { + "epoch": 1.86, + "learning_rate": 6.629577048484115e-06, + "loss": 1.1513, + "step": 22132 + }, + { + "epoch": 1.86, + "learning_rate": 6.628721273821071e-06, + "loss": 1.1718, + "step": 22133 + }, + { + "epoch": 1.86, + "learning_rate": 6.627865527012613e-06, + "loss": 1.2223, + "step": 22134 + }, + { + "epoch": 1.86, + "learning_rate": 6.627009808065807e-06, + "loss": 1.1281, + "step": 22135 + }, + { + "epoch": 1.86, + "learning_rate": 6.626154116987729e-06, + "loss": 1.2839, + "step": 22136 + }, + { + "epoch": 1.86, + "learning_rate": 6.62529845378544e-06, + "loss": 1.168, + "step": 22137 + }, + { + "epoch": 1.86, + "learning_rate": 6.6244428184660195e-06, + "loss": 1.1119, + "step": 22138 + }, + { + "epoch": 1.86, + "learning_rate": 6.623587211036533e-06, + "loss": 1.274, + "step": 22139 + }, + { + "epoch": 1.86, + "learning_rate": 6.622731631504045e-06, + "loss": 1.154, + "step": 22140 + }, + { + "epoch": 1.86, + "learning_rate": 6.621876079875634e-06, + "loss": 1.1766, + "step": 22141 + }, + { + "epoch": 1.86, + "learning_rate": 6.6210205561583615e-06, + "loss": 1.1174, + "step": 22142 + }, + { + "epoch": 1.86, + "learning_rate": 6.6201650603593025e-06, + "loss": 1.2362, + "step": 22143 + }, + { + "epoch": 1.86, + "learning_rate": 6.619309592485515e-06, + "loss": 1.2516, + "step": 22144 + }, + { + "epoch": 1.86, + "learning_rate": 6.618454152544078e-06, + "loss": 1.2002, + "step": 22145 + }, + { + "epoch": 1.86, + "learning_rate": 6.617598740542055e-06, + "loss": 1.1547, + "step": 22146 + }, + { + "epoch": 1.86, + "learning_rate": 6.616743356486514e-06, + "loss": 1.117, + "step": 22147 + }, + { + "epoch": 1.86, + "learning_rate": 6.615888000384518e-06, + "loss": 1.1149, + "step": 22148 + }, + { + "epoch": 1.86, + "learning_rate": 6.615032672243141e-06, + "loss": 1.1857, + "step": 22149 + }, + { + "epoch": 1.86, + "learning_rate": 6.6141773720694506e-06, + "loss": 1.0967, + "step": 22150 + }, + { + "epoch": 1.86, + "learning_rate": 6.613322099870504e-06, + "loss": 1.1609, + "step": 22151 + }, + { + "epoch": 1.87, + "learning_rate": 6.612466855653379e-06, + "loss": 1.1798, + "step": 22152 + }, + { + "epoch": 1.87, + "learning_rate": 6.611611639425135e-06, + "loss": 1.1832, + "step": 22153 + }, + { + "epoch": 1.87, + "learning_rate": 6.610756451192842e-06, + "loss": 1.1668, + "step": 22154 + }, + { + "epoch": 1.87, + "learning_rate": 6.60990129096356e-06, + "loss": 1.1738, + "step": 22155 + }, + { + "epoch": 1.87, + "learning_rate": 6.609046158744363e-06, + "loss": 1.2022, + "step": 22156 + }, + { + "epoch": 1.87, + "learning_rate": 6.60819105454231e-06, + "loss": 1.1507, + "step": 22157 + }, + { + "epoch": 1.87, + "learning_rate": 6.607335978364468e-06, + "loss": 1.172, + "step": 22158 + }, + { + "epoch": 1.87, + "learning_rate": 6.6064809302179e-06, + "loss": 1.1637, + "step": 22159 + }, + { + "epoch": 1.87, + "learning_rate": 6.6056259101096765e-06, + "loss": 1.2045, + "step": 22160 + }, + { + "epoch": 1.87, + "learning_rate": 6.604770918046856e-06, + "loss": 1.2584, + "step": 22161 + }, + { + "epoch": 1.87, + "learning_rate": 6.603915954036502e-06, + "loss": 1.1684, + "step": 22162 + }, + { + "epoch": 1.87, + "learning_rate": 6.603061018085686e-06, + "loss": 1.2516, + "step": 22163 + }, + { + "epoch": 1.87, + "learning_rate": 6.602206110201463e-06, + "loss": 1.2209, + "step": 22164 + }, + { + "epoch": 1.87, + "learning_rate": 6.601351230390905e-06, + "loss": 1.1146, + "step": 22165 + }, + { + "epoch": 1.87, + "learning_rate": 6.600496378661067e-06, + "loss": 1.1764, + "step": 22166 + }, + { + "epoch": 1.87, + "learning_rate": 6.599641555019019e-06, + "loss": 1.1872, + "step": 22167 + }, + { + "epoch": 1.87, + "learning_rate": 6.598786759471818e-06, + "loss": 1.2736, + "step": 22168 + }, + { + "epoch": 1.87, + "learning_rate": 6.597931992026532e-06, + "loss": 1.2352, + "step": 22169 + }, + { + "epoch": 1.87, + "learning_rate": 6.5970772526902175e-06, + "loss": 1.1326, + "step": 22170 + }, + { + "epoch": 1.87, + "learning_rate": 6.596222541469943e-06, + "loss": 1.2001, + "step": 22171 + }, + { + "epoch": 1.87, + "learning_rate": 6.595367858372765e-06, + "loss": 1.1799, + "step": 22172 + }, + { + "epoch": 1.87, + "learning_rate": 6.59451320340575e-06, + "loss": 1.1988, + "step": 22173 + }, + { + "epoch": 1.87, + "learning_rate": 6.593658576575952e-06, + "loss": 1.2503, + "step": 22174 + }, + { + "epoch": 1.87, + "learning_rate": 6.5928039778904384e-06, + "loss": 1.1316, + "step": 22175 + }, + { + "epoch": 1.87, + "learning_rate": 6.591949407356273e-06, + "loss": 1.2249, + "step": 22176 + }, + { + "epoch": 1.87, + "learning_rate": 6.591094864980507e-06, + "loss": 1.1578, + "step": 22177 + }, + { + "epoch": 1.87, + "learning_rate": 6.590240350770211e-06, + "loss": 1.1465, + "step": 22178 + }, + { + "epoch": 1.87, + "learning_rate": 6.589385864732436e-06, + "loss": 1.1194, + "step": 22179 + }, + { + "epoch": 1.87, + "learning_rate": 6.588531406874251e-06, + "loss": 1.1391, + "step": 22180 + }, + { + "epoch": 1.87, + "learning_rate": 6.587676977202708e-06, + "loss": 1.1547, + "step": 22181 + }, + { + "epoch": 1.87, + "learning_rate": 6.5868225757248735e-06, + "loss": 1.1216, + "step": 22182 + }, + { + "epoch": 1.87, + "learning_rate": 6.585968202447799e-06, + "loss": 1.0861, + "step": 22183 + }, + { + "epoch": 1.87, + "learning_rate": 6.585113857378551e-06, + "loss": 1.2409, + "step": 22184 + }, + { + "epoch": 1.87, + "learning_rate": 6.584259540524183e-06, + "loss": 1.1073, + "step": 22185 + }, + { + "epoch": 1.87, + "learning_rate": 6.583405251891758e-06, + "loss": 1.0969, + "step": 22186 + }, + { + "epoch": 1.87, + "learning_rate": 6.582550991488328e-06, + "loss": 1.1311, + "step": 22187 + }, + { + "epoch": 1.87, + "learning_rate": 6.581696759320959e-06, + "loss": 1.1666, + "step": 22188 + }, + { + "epoch": 1.87, + "learning_rate": 6.580842555396706e-06, + "loss": 1.1525, + "step": 22189 + }, + { + "epoch": 1.87, + "learning_rate": 6.579988379722622e-06, + "loss": 1.1776, + "step": 22190 + }, + { + "epoch": 1.87, + "learning_rate": 6.579134232305773e-06, + "loss": 1.1955, + "step": 22191 + }, + { + "epoch": 1.87, + "learning_rate": 6.57828011315321e-06, + "loss": 1.1519, + "step": 22192 + }, + { + "epoch": 1.87, + "learning_rate": 6.577426022271993e-06, + "loss": 1.1317, + "step": 22193 + }, + { + "epoch": 1.87, + "learning_rate": 6.576571959669173e-06, + "loss": 1.1848, + "step": 22194 + }, + { + "epoch": 1.87, + "learning_rate": 6.575717925351817e-06, + "loss": 1.2892, + "step": 22195 + }, + { + "epoch": 1.87, + "learning_rate": 6.574863919326973e-06, + "loss": 1.2228, + "step": 22196 + }, + { + "epoch": 1.87, + "learning_rate": 6.574009941601701e-06, + "loss": 1.2308, + "step": 22197 + }, + { + "epoch": 1.87, + "learning_rate": 6.573155992183051e-06, + "loss": 1.2312, + "step": 22198 + }, + { + "epoch": 1.87, + "learning_rate": 6.572302071078085e-06, + "loss": 1.2195, + "step": 22199 + }, + { + "epoch": 1.87, + "learning_rate": 6.57144817829386e-06, + "loss": 1.1895, + "step": 22200 + }, + { + "epoch": 1.87, + "learning_rate": 6.570594313837422e-06, + "loss": 1.1318, + "step": 22201 + }, + { + "epoch": 1.87, + "learning_rate": 6.5697404777158345e-06, + "loss": 1.1431, + "step": 22202 + }, + { + "epoch": 1.87, + "learning_rate": 6.568886669936147e-06, + "loss": 1.1871, + "step": 22203 + }, + { + "epoch": 1.87, + "learning_rate": 6.5680328905054205e-06, + "loss": 1.2912, + "step": 22204 + }, + { + "epoch": 1.87, + "learning_rate": 6.567179139430699e-06, + "loss": 1.2026, + "step": 22205 + }, + { + "epoch": 1.87, + "learning_rate": 6.566325416719045e-06, + "loss": 1.1509, + "step": 22206 + }, + { + "epoch": 1.87, + "learning_rate": 6.565471722377508e-06, + "loss": 1.1938, + "step": 22207 + }, + { + "epoch": 1.87, + "learning_rate": 6.5646180564131455e-06, + "loss": 1.171, + "step": 22208 + }, + { + "epoch": 1.87, + "learning_rate": 6.563764418833003e-06, + "loss": 1.189, + "step": 22209 + }, + { + "epoch": 1.87, + "learning_rate": 6.562910809644144e-06, + "loss": 1.1622, + "step": 22210 + }, + { + "epoch": 1.87, + "learning_rate": 6.562057228853613e-06, + "loss": 1.2011, + "step": 22211 + }, + { + "epoch": 1.87, + "learning_rate": 6.561203676468464e-06, + "loss": 1.1442, + "step": 22212 + }, + { + "epoch": 1.87, + "learning_rate": 6.560350152495754e-06, + "loss": 1.1452, + "step": 22213 + }, + { + "epoch": 1.87, + "learning_rate": 6.55949665694253e-06, + "loss": 1.3117, + "step": 22214 + }, + { + "epoch": 1.87, + "learning_rate": 6.558643189815849e-06, + "loss": 1.2201, + "step": 22215 + }, + { + "epoch": 1.87, + "learning_rate": 6.557789751122754e-06, + "loss": 1.1867, + "step": 22216 + }, + { + "epoch": 1.87, + "learning_rate": 6.556936340870306e-06, + "loss": 1.1866, + "step": 22217 + }, + { + "epoch": 1.87, + "learning_rate": 6.5560829590655505e-06, + "loss": 1.196, + "step": 22218 + }, + { + "epoch": 1.87, + "learning_rate": 6.5552296057155405e-06, + "loss": 1.2164, + "step": 22219 + }, + { + "epoch": 1.87, + "learning_rate": 6.554376280827324e-06, + "loss": 1.0596, + "step": 22220 + }, + { + "epoch": 1.87, + "learning_rate": 6.553522984407956e-06, + "loss": 1.2312, + "step": 22221 + }, + { + "epoch": 1.87, + "learning_rate": 6.552669716464481e-06, + "loss": 1.1313, + "step": 22222 + }, + { + "epoch": 1.87, + "learning_rate": 6.551816477003956e-06, + "loss": 1.1689, + "step": 22223 + }, + { + "epoch": 1.87, + "learning_rate": 6.550963266033423e-06, + "loss": 1.1136, + "step": 22224 + }, + { + "epoch": 1.87, + "learning_rate": 6.550110083559935e-06, + "loss": 1.0511, + "step": 22225 + }, + { + "epoch": 1.87, + "learning_rate": 6.549256929590546e-06, + "loss": 1.1577, + "step": 22226 + }, + { + "epoch": 1.87, + "learning_rate": 6.548403804132295e-06, + "loss": 1.2029, + "step": 22227 + }, + { + "epoch": 1.87, + "learning_rate": 6.547550707192243e-06, + "loss": 1.1537, + "step": 22228 + }, + { + "epoch": 1.87, + "learning_rate": 6.546697638777428e-06, + "loss": 1.0982, + "step": 22229 + }, + { + "epoch": 1.87, + "learning_rate": 6.545844598894904e-06, + "loss": 1.251, + "step": 22230 + }, + { + "epoch": 1.87, + "learning_rate": 6.544991587551713e-06, + "loss": 1.0874, + "step": 22231 + }, + { + "epoch": 1.87, + "learning_rate": 6.544138604754915e-06, + "loss": 1.1789, + "step": 22232 + }, + { + "epoch": 1.87, + "learning_rate": 6.543285650511543e-06, + "loss": 1.1103, + "step": 22233 + }, + { + "epoch": 1.87, + "learning_rate": 6.542432724828656e-06, + "loss": 1.1512, + "step": 22234 + }, + { + "epoch": 1.87, + "learning_rate": 6.541579827713292e-06, + "loss": 1.1937, + "step": 22235 + }, + { + "epoch": 1.87, + "learning_rate": 6.540726959172507e-06, + "loss": 1.2012, + "step": 22236 + }, + { + "epoch": 1.87, + "learning_rate": 6.539874119213337e-06, + "loss": 1.209, + "step": 22237 + }, + { + "epoch": 1.87, + "learning_rate": 6.5390213078428375e-06, + "loss": 1.2212, + "step": 22238 + }, + { + "epoch": 1.87, + "learning_rate": 6.538168525068054e-06, + "loss": 1.2008, + "step": 22239 + }, + { + "epoch": 1.87, + "learning_rate": 6.537315770896025e-06, + "loss": 1.2529, + "step": 22240 + }, + { + "epoch": 1.87, + "learning_rate": 6.536463045333805e-06, + "loss": 1.2584, + "step": 22241 + }, + { + "epoch": 1.87, + "learning_rate": 6.5356103483884345e-06, + "loss": 1.2769, + "step": 22242 + }, + { + "epoch": 1.87, + "learning_rate": 6.534757680066963e-06, + "loss": 1.1817, + "step": 22243 + }, + { + "epoch": 1.87, + "learning_rate": 6.533905040376427e-06, + "loss": 1.2235, + "step": 22244 + }, + { + "epoch": 1.87, + "learning_rate": 6.533052429323883e-06, + "loss": 1.1295, + "step": 22245 + }, + { + "epoch": 1.87, + "learning_rate": 6.532199846916365e-06, + "loss": 1.2704, + "step": 22246 + }, + { + "epoch": 1.87, + "learning_rate": 6.531347293160924e-06, + "loss": 1.2332, + "step": 22247 + }, + { + "epoch": 1.87, + "learning_rate": 6.530494768064599e-06, + "loss": 1.1652, + "step": 22248 + }, + { + "epoch": 1.87, + "learning_rate": 6.529642271634438e-06, + "loss": 1.1882, + "step": 22249 + }, + { + "epoch": 1.87, + "learning_rate": 6.528789803877484e-06, + "loss": 1.1419, + "step": 22250 + }, + { + "epoch": 1.87, + "learning_rate": 6.527937364800778e-06, + "loss": 1.1563, + "step": 22251 + }, + { + "epoch": 1.87, + "learning_rate": 6.527084954411367e-06, + "loss": 1.126, + "step": 22252 + }, + { + "epoch": 1.87, + "learning_rate": 6.526232572716291e-06, + "loss": 1.1858, + "step": 22253 + }, + { + "epoch": 1.87, + "learning_rate": 6.525380219722594e-06, + "loss": 1.175, + "step": 22254 + }, + { + "epoch": 1.87, + "learning_rate": 6.524527895437314e-06, + "loss": 1.1341, + "step": 22255 + }, + { + "epoch": 1.87, + "learning_rate": 6.523675599867501e-06, + "loss": 1.2196, + "step": 22256 + }, + { + "epoch": 1.87, + "learning_rate": 6.522823333020189e-06, + "loss": 1.238, + "step": 22257 + }, + { + "epoch": 1.87, + "learning_rate": 6.521971094902428e-06, + "loss": 1.1952, + "step": 22258 + }, + { + "epoch": 1.87, + "learning_rate": 6.52111888552125e-06, + "loss": 1.1441, + "step": 22259 + }, + { + "epoch": 1.87, + "learning_rate": 6.520266704883704e-06, + "loss": 1.1937, + "step": 22260 + }, + { + "epoch": 1.87, + "learning_rate": 6.519414552996827e-06, + "loss": 1.219, + "step": 22261 + }, + { + "epoch": 1.87, + "learning_rate": 6.5185624298676604e-06, + "loss": 1.1545, + "step": 22262 + }, + { + "epoch": 1.87, + "learning_rate": 6.517710335503248e-06, + "loss": 1.2141, + "step": 22263 + }, + { + "epoch": 1.87, + "learning_rate": 6.516858269910624e-06, + "loss": 1.1105, + "step": 22264 + }, + { + "epoch": 1.87, + "learning_rate": 6.516006233096836e-06, + "loss": 1.2737, + "step": 22265 + }, + { + "epoch": 1.87, + "learning_rate": 6.515154225068914e-06, + "loss": 1.1243, + "step": 22266 + }, + { + "epoch": 1.87, + "learning_rate": 6.514302245833908e-06, + "loss": 1.178, + "step": 22267 + }, + { + "epoch": 1.87, + "learning_rate": 6.51345029539885e-06, + "loss": 1.2105, + "step": 22268 + }, + { + "epoch": 1.87, + "learning_rate": 6.512598373770783e-06, + "loss": 1.1166, + "step": 22269 + }, + { + "epoch": 1.87, + "learning_rate": 6.511746480956741e-06, + "loss": 1.1275, + "step": 22270 + }, + { + "epoch": 1.88, + "learning_rate": 6.51089461696377e-06, + "loss": 1.233, + "step": 22271 + }, + { + "epoch": 1.88, + "learning_rate": 6.510042781798902e-06, + "loss": 1.1945, + "step": 22272 + }, + { + "epoch": 1.88, + "learning_rate": 6.5091909754691794e-06, + "loss": 1.1409, + "step": 22273 + }, + { + "epoch": 1.88, + "learning_rate": 6.508339197981635e-06, + "loss": 1.1382, + "step": 22274 + }, + { + "epoch": 1.88, + "learning_rate": 6.5074874493433106e-06, + "loss": 1.1803, + "step": 22275 + }, + { + "epoch": 1.88, + "learning_rate": 6.506635729561245e-06, + "loss": 1.1704, + "step": 22276 + }, + { + "epoch": 1.88, + "learning_rate": 6.505784038642469e-06, + "loss": 1.2104, + "step": 22277 + }, + { + "epoch": 1.88, + "learning_rate": 6.504932376594027e-06, + "loss": 1.2524, + "step": 22278 + }, + { + "epoch": 1.88, + "learning_rate": 6.504080743422952e-06, + "loss": 1.2369, + "step": 22279 + }, + { + "epoch": 1.88, + "learning_rate": 6.50322913913628e-06, + "loss": 1.1693, + "step": 22280 + }, + { + "epoch": 1.88, + "learning_rate": 6.502377563741046e-06, + "loss": 1.1602, + "step": 22281 + }, + { + "epoch": 1.88, + "learning_rate": 6.501526017244293e-06, + "loss": 1.1593, + "step": 22282 + }, + { + "epoch": 1.88, + "learning_rate": 6.500674499653046e-06, + "loss": 1.1455, + "step": 22283 + }, + { + "epoch": 1.88, + "learning_rate": 6.499823010974351e-06, + "loss": 1.0942, + "step": 22284 + }, + { + "epoch": 1.88, + "learning_rate": 6.498971551215236e-06, + "loss": 1.2491, + "step": 22285 + }, + { + "epoch": 1.88, + "learning_rate": 6.49812012038274e-06, + "loss": 1.1623, + "step": 22286 + }, + { + "epoch": 1.88, + "learning_rate": 6.497268718483894e-06, + "loss": 1.2593, + "step": 22287 + }, + { + "epoch": 1.88, + "learning_rate": 6.496417345525734e-06, + "loss": 1.1719, + "step": 22288 + }, + { + "epoch": 1.88, + "learning_rate": 6.495566001515299e-06, + "loss": 1.1707, + "step": 22289 + }, + { + "epoch": 1.88, + "learning_rate": 6.494714686459615e-06, + "loss": 1.2751, + "step": 22290 + }, + { + "epoch": 1.88, + "learning_rate": 6.493863400365724e-06, + "loss": 1.2563, + "step": 22291 + }, + { + "epoch": 1.88, + "learning_rate": 6.493012143240653e-06, + "loss": 1.1983, + "step": 22292 + }, + { + "epoch": 1.88, + "learning_rate": 6.492160915091442e-06, + "loss": 1.1856, + "step": 22293 + }, + { + "epoch": 1.88, + "learning_rate": 6.491309715925115e-06, + "loss": 1.2381, + "step": 22294 + }, + { + "epoch": 1.88, + "learning_rate": 6.490458545748714e-06, + "loss": 1.178, + "step": 22295 + }, + { + "epoch": 1.88, + "learning_rate": 6.489607404569266e-06, + "loss": 1.1801, + "step": 22296 + }, + { + "epoch": 1.88, + "learning_rate": 6.488756292393807e-06, + "loss": 1.2132, + "step": 22297 + }, + { + "epoch": 1.88, + "learning_rate": 6.487905209229363e-06, + "loss": 1.2686, + "step": 22298 + }, + { + "epoch": 1.88, + "learning_rate": 6.487054155082973e-06, + "loss": 1.1387, + "step": 22299 + }, + { + "epoch": 1.88, + "learning_rate": 6.486203129961666e-06, + "loss": 1.1046, + "step": 22300 + }, + { + "epoch": 1.88, + "learning_rate": 6.4853521338724716e-06, + "loss": 1.2429, + "step": 22301 + }, + { + "epoch": 1.88, + "learning_rate": 6.484501166822425e-06, + "loss": 1.2162, + "step": 22302 + }, + { + "epoch": 1.88, + "learning_rate": 6.483650228818552e-06, + "loss": 1.1669, + "step": 22303 + }, + { + "epoch": 1.88, + "learning_rate": 6.482799319867891e-06, + "loss": 1.2149, + "step": 22304 + }, + { + "epoch": 1.88, + "learning_rate": 6.481948439977461e-06, + "loss": 1.2844, + "step": 22305 + }, + { + "epoch": 1.88, + "learning_rate": 6.481097589154305e-06, + "loss": 1.2009, + "step": 22306 + }, + { + "epoch": 1.88, + "learning_rate": 6.4802467674054435e-06, + "loss": 1.1639, + "step": 22307 + }, + { + "epoch": 1.88, + "learning_rate": 6.479395974737912e-06, + "loss": 1.1473, + "step": 22308 + }, + { + "epoch": 1.88, + "learning_rate": 6.4785452111587345e-06, + "loss": 1.1716, + "step": 22309 + }, + { + "epoch": 1.88, + "learning_rate": 6.477694476674947e-06, + "loss": 1.1338, + "step": 22310 + }, + { + "epoch": 1.88, + "learning_rate": 6.476843771293573e-06, + "loss": 1.0716, + "step": 22311 + }, + { + "epoch": 1.88, + "learning_rate": 6.475993095021641e-06, + "loss": 1.2273, + "step": 22312 + }, + { + "epoch": 1.88, + "learning_rate": 6.4751424478661865e-06, + "loss": 1.2505, + "step": 22313 + }, + { + "epoch": 1.88, + "learning_rate": 6.474291829834231e-06, + "loss": 1.1929, + "step": 22314 + }, + { + "epoch": 1.88, + "learning_rate": 6.473441240932806e-06, + "loss": 1.1452, + "step": 22315 + }, + { + "epoch": 1.88, + "learning_rate": 6.472590681168936e-06, + "loss": 1.1174, + "step": 22316 + }, + { + "epoch": 1.88, + "learning_rate": 6.471740150549653e-06, + "loss": 1.1973, + "step": 22317 + }, + { + "epoch": 1.88, + "learning_rate": 6.4708896490819794e-06, + "loss": 1.2176, + "step": 22318 + }, + { + "epoch": 1.88, + "learning_rate": 6.470039176772949e-06, + "loss": 1.1699, + "step": 22319 + }, + { + "epoch": 1.88, + "learning_rate": 6.469188733629579e-06, + "loss": 1.1475, + "step": 22320 + }, + { + "epoch": 1.88, + "learning_rate": 6.468338319658907e-06, + "loss": 1.127, + "step": 22321 + }, + { + "epoch": 1.88, + "learning_rate": 6.467487934867951e-06, + "loss": 1.1899, + "step": 22322 + }, + { + "epoch": 1.88, + "learning_rate": 6.4666375792637424e-06, + "loss": 1.1152, + "step": 22323 + }, + { + "epoch": 1.88, + "learning_rate": 6.465787252853302e-06, + "loss": 1.0336, + "step": 22324 + }, + { + "epoch": 1.88, + "learning_rate": 6.4649369556436595e-06, + "loss": 1.1505, + "step": 22325 + }, + { + "epoch": 1.88, + "learning_rate": 6.464086687641842e-06, + "loss": 1.2458, + "step": 22326 + }, + { + "epoch": 1.88, + "learning_rate": 6.463236448854866e-06, + "loss": 1.1749, + "step": 22327 + }, + { + "epoch": 1.88, + "learning_rate": 6.462386239289768e-06, + "loss": 1.1356, + "step": 22328 + }, + { + "epoch": 1.88, + "learning_rate": 6.461536058953564e-06, + "loss": 1.2346, + "step": 22329 + }, + { + "epoch": 1.88, + "learning_rate": 6.460685907853285e-06, + "loss": 1.1781, + "step": 22330 + }, + { + "epoch": 1.88, + "learning_rate": 6.459835785995947e-06, + "loss": 1.1976, + "step": 22331 + }, + { + "epoch": 1.88, + "learning_rate": 6.458985693388584e-06, + "loss": 1.1884, + "step": 22332 + }, + { + "epoch": 1.88, + "learning_rate": 6.45813563003821e-06, + "loss": 1.234, + "step": 22333 + }, + { + "epoch": 1.88, + "learning_rate": 6.457285595951856e-06, + "loss": 1.2461, + "step": 22334 + }, + { + "epoch": 1.88, + "learning_rate": 6.456435591136541e-06, + "loss": 1.1993, + "step": 22335 + }, + { + "epoch": 1.88, + "learning_rate": 6.45558561559929e-06, + "loss": 1.0724, + "step": 22336 + }, + { + "epoch": 1.88, + "learning_rate": 6.454735669347122e-06, + "loss": 1.2739, + "step": 22337 + }, + { + "epoch": 1.88, + "learning_rate": 6.453885752387064e-06, + "loss": 1.2512, + "step": 22338 + }, + { + "epoch": 1.88, + "learning_rate": 6.45303586472614e-06, + "loss": 1.2377, + "step": 22339 + }, + { + "epoch": 1.88, + "learning_rate": 6.452186006371364e-06, + "loss": 1.1924, + "step": 22340 + }, + { + "epoch": 1.88, + "learning_rate": 6.451336177329766e-06, + "loss": 1.2084, + "step": 22341 + }, + { + "epoch": 1.88, + "learning_rate": 6.450486377608362e-06, + "loss": 1.1105, + "step": 22342 + }, + { + "epoch": 1.88, + "learning_rate": 6.449636607214178e-06, + "loss": 1.1763, + "step": 22343 + }, + { + "epoch": 1.88, + "learning_rate": 6.44878686615423e-06, + "loss": 1.1436, + "step": 22344 + }, + { + "epoch": 1.88, + "learning_rate": 6.447937154435544e-06, + "loss": 1.163, + "step": 22345 + }, + { + "epoch": 1.88, + "learning_rate": 6.447087472065137e-06, + "loss": 1.0768, + "step": 22346 + }, + { + "epoch": 1.88, + "learning_rate": 6.446237819050032e-06, + "loss": 1.0638, + "step": 22347 + }, + { + "epoch": 1.88, + "learning_rate": 6.445388195397242e-06, + "loss": 1.1092, + "step": 22348 + }, + { + "epoch": 1.88, + "learning_rate": 6.444538601113797e-06, + "loss": 1.1973, + "step": 22349 + }, + { + "epoch": 1.88, + "learning_rate": 6.443689036206712e-06, + "loss": 1.1298, + "step": 22350 + }, + { + "epoch": 1.88, + "learning_rate": 6.4428395006830026e-06, + "loss": 1.2079, + "step": 22351 + }, + { + "epoch": 1.88, + "learning_rate": 6.441989994549696e-06, + "loss": 1.1984, + "step": 22352 + }, + { + "epoch": 1.88, + "learning_rate": 6.441140517813804e-06, + "loss": 1.1636, + "step": 22353 + }, + { + "epoch": 1.88, + "learning_rate": 6.440291070482352e-06, + "loss": 1.1649, + "step": 22354 + }, + { + "epoch": 1.88, + "learning_rate": 6.439441652562348e-06, + "loss": 1.1698, + "step": 22355 + }, + { + "epoch": 1.88, + "learning_rate": 6.438592264060823e-06, + "loss": 1.1195, + "step": 22356 + }, + { + "epoch": 1.88, + "learning_rate": 6.437742904984785e-06, + "loss": 1.2038, + "step": 22357 + }, + { + "epoch": 1.88, + "learning_rate": 6.436893575341258e-06, + "loss": 1.1461, + "step": 22358 + }, + { + "epoch": 1.88, + "learning_rate": 6.436044275137252e-06, + "loss": 1.166, + "step": 22359 + }, + { + "epoch": 1.88, + "learning_rate": 6.4351950043797925e-06, + "loss": 1.2, + "step": 22360 + }, + { + "epoch": 1.88, + "learning_rate": 6.434345763075891e-06, + "loss": 1.1689, + "step": 22361 + }, + { + "epoch": 1.88, + "learning_rate": 6.433496551232564e-06, + "loss": 1.1316, + "step": 22362 + }, + { + "epoch": 1.88, + "learning_rate": 6.432647368856835e-06, + "loss": 1.1609, + "step": 22363 + }, + { + "epoch": 1.88, + "learning_rate": 6.4317982159557115e-06, + "loss": 1.0051, + "step": 22364 + }, + { + "epoch": 1.88, + "learning_rate": 6.430949092536216e-06, + "loss": 1.1851, + "step": 22365 + }, + { + "epoch": 1.88, + "learning_rate": 6.430099998605357e-06, + "loss": 1.2467, + "step": 22366 + }, + { + "epoch": 1.88, + "learning_rate": 6.429250934170159e-06, + "loss": 1.1022, + "step": 22367 + }, + { + "epoch": 1.88, + "learning_rate": 6.42840189923763e-06, + "loss": 1.1323, + "step": 22368 + }, + { + "epoch": 1.88, + "learning_rate": 6.42755289381479e-06, + "loss": 1.16, + "step": 22369 + }, + { + "epoch": 1.88, + "learning_rate": 6.4267039179086465e-06, + "loss": 1.1698, + "step": 22370 + }, + { + "epoch": 1.88, + "learning_rate": 6.4258549715262234e-06, + "loss": 1.1749, + "step": 22371 + }, + { + "epoch": 1.88, + "learning_rate": 6.4250060546745275e-06, + "loss": 1.1829, + "step": 22372 + }, + { + "epoch": 1.88, + "learning_rate": 6.424157167360578e-06, + "loss": 1.1365, + "step": 22373 + }, + { + "epoch": 1.88, + "learning_rate": 6.423308309591384e-06, + "loss": 1.1069, + "step": 22374 + }, + { + "epoch": 1.88, + "learning_rate": 6.422459481373962e-06, + "loss": 1.2392, + "step": 22375 + }, + { + "epoch": 1.88, + "learning_rate": 6.421610682715327e-06, + "loss": 1.109, + "step": 22376 + }, + { + "epoch": 1.88, + "learning_rate": 6.420761913622486e-06, + "loss": 1.1591, + "step": 22377 + }, + { + "epoch": 1.88, + "learning_rate": 6.41991317410246e-06, + "loss": 1.1463, + "step": 22378 + }, + { + "epoch": 1.88, + "learning_rate": 6.419064464162254e-06, + "loss": 1.172, + "step": 22379 + }, + { + "epoch": 1.88, + "learning_rate": 6.418215783808887e-06, + "loss": 1.1808, + "step": 22380 + }, + { + "epoch": 1.88, + "learning_rate": 6.4173671330493615e-06, + "loss": 1.1344, + "step": 22381 + }, + { + "epoch": 1.88, + "learning_rate": 6.416518511890701e-06, + "loss": 1.2734, + "step": 22382 + }, + { + "epoch": 1.88, + "learning_rate": 6.415669920339909e-06, + "loss": 1.2053, + "step": 22383 + }, + { + "epoch": 1.88, + "learning_rate": 6.414821358404e-06, + "loss": 1.2196, + "step": 22384 + }, + { + "epoch": 1.88, + "learning_rate": 6.413972826089982e-06, + "loss": 1.1416, + "step": 22385 + }, + { + "epoch": 1.88, + "learning_rate": 6.413124323404869e-06, + "loss": 1.2825, + "step": 22386 + }, + { + "epoch": 1.88, + "learning_rate": 6.412275850355674e-06, + "loss": 1.2374, + "step": 22387 + }, + { + "epoch": 1.88, + "learning_rate": 6.4114274069494e-06, + "loss": 1.1783, + "step": 22388 + }, + { + "epoch": 1.88, + "learning_rate": 6.410578993193066e-06, + "loss": 1.2056, + "step": 22389 + }, + { + "epoch": 1.89, + "learning_rate": 6.409730609093672e-06, + "loss": 1.1645, + "step": 22390 + }, + { + "epoch": 1.89, + "learning_rate": 6.408882254658236e-06, + "loss": 1.088, + "step": 22391 + }, + { + "epoch": 1.89, + "learning_rate": 6.408033929893763e-06, + "loss": 1.1467, + "step": 22392 + }, + { + "epoch": 1.89, + "learning_rate": 6.407185634807264e-06, + "loss": 1.1616, + "step": 22393 + }, + { + "epoch": 1.89, + "learning_rate": 6.406337369405745e-06, + "loss": 1.1992, + "step": 22394 + }, + { + "epoch": 1.89, + "learning_rate": 6.40548913369622e-06, + "loss": 1.2287, + "step": 22395 + }, + { + "epoch": 1.89, + "learning_rate": 6.404640927685692e-06, + "loss": 1.0554, + "step": 22396 + }, + { + "epoch": 1.89, + "learning_rate": 6.4037927513811725e-06, + "loss": 1.0808, + "step": 22397 + }, + { + "epoch": 1.89, + "learning_rate": 6.402944604789666e-06, + "loss": 1.2146, + "step": 22398 + }, + { + "epoch": 1.89, + "learning_rate": 6.4020964879181834e-06, + "loss": 0.9973, + "step": 22399 + }, + { + "epoch": 1.89, + "learning_rate": 6.401248400773733e-06, + "loss": 1.3001, + "step": 22400 + }, + { + "epoch": 1.89, + "learning_rate": 6.400400343363315e-06, + "loss": 1.1163, + "step": 22401 + }, + { + "epoch": 1.89, + "learning_rate": 6.3995523156939474e-06, + "loss": 1.1178, + "step": 22402 + }, + { + "epoch": 1.89, + "learning_rate": 6.3987043177726284e-06, + "loss": 1.1755, + "step": 22403 + }, + { + "epoch": 1.89, + "learning_rate": 6.397856349606368e-06, + "loss": 1.2223, + "step": 22404 + }, + { + "epoch": 1.89, + "learning_rate": 6.397008411202168e-06, + "loss": 1.1819, + "step": 22405 + }, + { + "epoch": 1.89, + "learning_rate": 6.396160502567041e-06, + "loss": 1.2218, + "step": 22406 + }, + { + "epoch": 1.89, + "learning_rate": 6.3953126237079876e-06, + "loss": 1.2225, + "step": 22407 + }, + { + "epoch": 1.89, + "learning_rate": 6.3944647746320186e-06, + "loss": 1.1984, + "step": 22408 + }, + { + "epoch": 1.89, + "learning_rate": 6.393616955346129e-06, + "loss": 1.1875, + "step": 22409 + }, + { + "epoch": 1.89, + "learning_rate": 6.392769165857335e-06, + "loss": 1.1747, + "step": 22410 + }, + { + "epoch": 1.89, + "learning_rate": 6.391921406172635e-06, + "loss": 1.132, + "step": 22411 + }, + { + "epoch": 1.89, + "learning_rate": 6.3910736762990334e-06, + "loss": 1.1441, + "step": 22412 + }, + { + "epoch": 1.89, + "learning_rate": 6.39022597624354e-06, + "loss": 1.1526, + "step": 22413 + }, + { + "epoch": 1.89, + "learning_rate": 6.389378306013152e-06, + "loss": 1.1932, + "step": 22414 + }, + { + "epoch": 1.89, + "learning_rate": 6.388530665614878e-06, + "loss": 1.1042, + "step": 22415 + }, + { + "epoch": 1.89, + "learning_rate": 6.387683055055717e-06, + "loss": 1.1843, + "step": 22416 + }, + { + "epoch": 1.89, + "learning_rate": 6.3868354743426775e-06, + "loss": 1.0752, + "step": 22417 + }, + { + "epoch": 1.89, + "learning_rate": 6.385987923482758e-06, + "loss": 1.2256, + "step": 22418 + }, + { + "epoch": 1.89, + "learning_rate": 6.385140402482965e-06, + "loss": 1.0515, + "step": 22419 + }, + { + "epoch": 1.89, + "learning_rate": 6.384292911350294e-06, + "loss": 1.1105, + "step": 22420 + }, + { + "epoch": 1.89, + "learning_rate": 6.3834454500917575e-06, + "loss": 1.1954, + "step": 22421 + }, + { + "epoch": 1.89, + "learning_rate": 6.38259801871435e-06, + "loss": 1.1984, + "step": 22422 + }, + { + "epoch": 1.89, + "learning_rate": 6.381750617225076e-06, + "loss": 1.2836, + "step": 22423 + }, + { + "epoch": 1.89, + "learning_rate": 6.380903245630934e-06, + "loss": 1.0994, + "step": 22424 + }, + { + "epoch": 1.89, + "learning_rate": 6.380055903938928e-06, + "loss": 1.1702, + "step": 22425 + }, + { + "epoch": 1.89, + "learning_rate": 6.379208592156061e-06, + "loss": 1.1164, + "step": 22426 + }, + { + "epoch": 1.89, + "learning_rate": 6.37836131028933e-06, + "loss": 1.1921, + "step": 22427 + }, + { + "epoch": 1.89, + "learning_rate": 6.377514058345737e-06, + "loss": 1.1859, + "step": 22428 + }, + { + "epoch": 1.89, + "learning_rate": 6.376666836332282e-06, + "loss": 1.1387, + "step": 22429 + }, + { + "epoch": 1.89, + "learning_rate": 6.375819644255967e-06, + "loss": 1.2166, + "step": 22430 + }, + { + "epoch": 1.89, + "learning_rate": 6.374972482123787e-06, + "loss": 1.1181, + "step": 22431 + }, + { + "epoch": 1.89, + "learning_rate": 6.374125349942748e-06, + "loss": 1.1987, + "step": 22432 + }, + { + "epoch": 1.89, + "learning_rate": 6.373278247719843e-06, + "loss": 1.0678, + "step": 22433 + }, + { + "epoch": 1.89, + "learning_rate": 6.372431175462077e-06, + "loss": 1.1468, + "step": 22434 + }, + { + "epoch": 1.89, + "learning_rate": 6.37158413317644e-06, + "loss": 1.1151, + "step": 22435 + }, + { + "epoch": 1.89, + "learning_rate": 6.37073712086994e-06, + "loss": 1.1574, + "step": 22436 + }, + { + "epoch": 1.89, + "learning_rate": 6.369890138549573e-06, + "loss": 1.1762, + "step": 22437 + }, + { + "epoch": 1.89, + "learning_rate": 6.36904318622233e-06, + "loss": 1.1943, + "step": 22438 + }, + { + "epoch": 1.89, + "learning_rate": 6.3681962638952215e-06, + "loss": 1.1629, + "step": 22439 + }, + { + "epoch": 1.89, + "learning_rate": 6.367349371575231e-06, + "loss": 1.2148, + "step": 22440 + }, + { + "epoch": 1.89, + "learning_rate": 6.366502509269368e-06, + "loss": 1.1249, + "step": 22441 + }, + { + "epoch": 1.89, + "learning_rate": 6.365655676984621e-06, + "loss": 1.1057, + "step": 22442 + }, + { + "epoch": 1.89, + "learning_rate": 6.364808874727992e-06, + "loss": 1.1599, + "step": 22443 + }, + { + "epoch": 1.89, + "learning_rate": 6.363962102506471e-06, + "loss": 1.1306, + "step": 22444 + }, + { + "epoch": 1.89, + "learning_rate": 6.363115360327063e-06, + "loss": 1.1584, + "step": 22445 + }, + { + "epoch": 1.89, + "learning_rate": 6.362268648196759e-06, + "loss": 1.1898, + "step": 22446 + }, + { + "epoch": 1.89, + "learning_rate": 6.361421966122557e-06, + "loss": 1.1161, + "step": 22447 + }, + { + "epoch": 1.89, + "learning_rate": 6.360575314111448e-06, + "loss": 1.1871, + "step": 22448 + }, + { + "epoch": 1.89, + "learning_rate": 6.359728692170431e-06, + "loss": 1.1627, + "step": 22449 + }, + { + "epoch": 1.89, + "learning_rate": 6.358882100306504e-06, + "loss": 1.2025, + "step": 22450 + }, + { + "epoch": 1.89, + "learning_rate": 6.358035538526654e-06, + "loss": 1.2779, + "step": 22451 + }, + { + "epoch": 1.89, + "learning_rate": 6.3571890068378825e-06, + "loss": 1.2273, + "step": 22452 + }, + { + "epoch": 1.89, + "learning_rate": 6.35634250524718e-06, + "loss": 1.1045, + "step": 22453 + }, + { + "epoch": 1.89, + "learning_rate": 6.355496033761546e-06, + "loss": 1.1927, + "step": 22454 + }, + { + "epoch": 1.89, + "learning_rate": 6.3546495923879635e-06, + "loss": 1.1547, + "step": 22455 + }, + { + "epoch": 1.89, + "learning_rate": 6.353803181133438e-06, + "loss": 1.2333, + "step": 22456 + }, + { + "epoch": 1.89, + "learning_rate": 6.352956800004954e-06, + "loss": 1.0985, + "step": 22457 + }, + { + "epoch": 1.89, + "learning_rate": 6.352110449009512e-06, + "loss": 1.2206, + "step": 22458 + }, + { + "epoch": 1.89, + "learning_rate": 6.351264128154096e-06, + "loss": 1.1868, + "step": 22459 + }, + { + "epoch": 1.89, + "learning_rate": 6.350417837445708e-06, + "loss": 1.1905, + "step": 22460 + }, + { + "epoch": 1.89, + "learning_rate": 6.349571576891333e-06, + "loss": 1.1437, + "step": 22461 + }, + { + "epoch": 1.89, + "learning_rate": 6.348725346497966e-06, + "loss": 1.2675, + "step": 22462 + }, + { + "epoch": 1.89, + "learning_rate": 6.3478791462726e-06, + "loss": 1.2241, + "step": 22463 + }, + { + "epoch": 1.89, + "learning_rate": 6.347032976222225e-06, + "loss": 1.1607, + "step": 22464 + }, + { + "epoch": 1.89, + "learning_rate": 6.346186836353834e-06, + "loss": 1.1567, + "step": 22465 + }, + { + "epoch": 1.89, + "learning_rate": 6.345340726674414e-06, + "loss": 1.1908, + "step": 22466 + }, + { + "epoch": 1.89, + "learning_rate": 6.344494647190961e-06, + "loss": 1.1691, + "step": 22467 + }, + { + "epoch": 1.89, + "learning_rate": 6.343648597910461e-06, + "loss": 1.133, + "step": 22468 + }, + { + "epoch": 1.89, + "learning_rate": 6.34280257883991e-06, + "loss": 1.2189, + "step": 22469 + }, + { + "epoch": 1.89, + "learning_rate": 6.341956589986289e-06, + "loss": 1.1573, + "step": 22470 + }, + { + "epoch": 1.89, + "learning_rate": 6.341110631356599e-06, + "loss": 1.1797, + "step": 22471 + }, + { + "epoch": 1.89, + "learning_rate": 6.34026470295782e-06, + "loss": 1.1347, + "step": 22472 + }, + { + "epoch": 1.89, + "learning_rate": 6.339418804796949e-06, + "loss": 1.1413, + "step": 22473 + }, + { + "epoch": 1.89, + "learning_rate": 6.338572936880967e-06, + "loss": 1.1632, + "step": 22474 + }, + { + "epoch": 1.89, + "learning_rate": 6.337727099216868e-06, + "loss": 1.1002, + "step": 22475 + }, + { + "epoch": 1.89, + "learning_rate": 6.336881291811643e-06, + "loss": 1.237, + "step": 22476 + }, + { + "epoch": 1.89, + "learning_rate": 6.336035514672272e-06, + "loss": 1.2463, + "step": 22477 + }, + { + "epoch": 1.89, + "learning_rate": 6.335189767805754e-06, + "loss": 1.0968, + "step": 22478 + }, + { + "epoch": 1.89, + "learning_rate": 6.334344051219068e-06, + "loss": 1.2169, + "step": 22479 + }, + { + "epoch": 1.89, + "learning_rate": 6.333498364919206e-06, + "loss": 1.2827, + "step": 22480 + }, + { + "epoch": 1.89, + "learning_rate": 6.332652708913151e-06, + "loss": 1.2073, + "step": 22481 + }, + { + "epoch": 1.89, + "learning_rate": 6.331807083207897e-06, + "loss": 1.2035, + "step": 22482 + }, + { + "epoch": 1.89, + "learning_rate": 6.330961487810423e-06, + "loss": 1.1809, + "step": 22483 + }, + { + "epoch": 1.89, + "learning_rate": 6.330115922727724e-06, + "loss": 1.144, + "step": 22484 + }, + { + "epoch": 1.89, + "learning_rate": 6.329270387966776e-06, + "loss": 1.1774, + "step": 22485 + }, + { + "epoch": 1.89, + "learning_rate": 6.328424883534574e-06, + "loss": 1.1119, + "step": 22486 + }, + { + "epoch": 1.89, + "learning_rate": 6.327579409438102e-06, + "loss": 1.236, + "step": 22487 + }, + { + "epoch": 1.89, + "learning_rate": 6.326733965684339e-06, + "loss": 1.1973, + "step": 22488 + }, + { + "epoch": 1.89, + "learning_rate": 6.325888552280284e-06, + "loss": 1.1391, + "step": 22489 + }, + { + "epoch": 1.89, + "learning_rate": 6.325043169232908e-06, + "loss": 1.0816, + "step": 22490 + }, + { + "epoch": 1.89, + "learning_rate": 6.324197816549203e-06, + "loss": 1.1287, + "step": 22491 + }, + { + "epoch": 1.89, + "learning_rate": 6.32335249423615e-06, + "loss": 1.2354, + "step": 22492 + }, + { + "epoch": 1.89, + "learning_rate": 6.32250720230074e-06, + "loss": 1.2321, + "step": 22493 + }, + { + "epoch": 1.89, + "learning_rate": 6.321661940749946e-06, + "loss": 1.141, + "step": 22494 + }, + { + "epoch": 1.89, + "learning_rate": 6.320816709590762e-06, + "loss": 1.1571, + "step": 22495 + }, + { + "epoch": 1.89, + "learning_rate": 6.3199715088301686e-06, + "loss": 1.2232, + "step": 22496 + }, + { + "epoch": 1.89, + "learning_rate": 6.319126338475148e-06, + "loss": 1.2468, + "step": 22497 + }, + { + "epoch": 1.89, + "learning_rate": 6.318281198532681e-06, + "loss": 1.1489, + "step": 22498 + }, + { + "epoch": 1.89, + "learning_rate": 6.317436089009754e-06, + "loss": 1.1875, + "step": 22499 + }, + { + "epoch": 1.89, + "learning_rate": 6.316591009913351e-06, + "loss": 1.1503, + "step": 22500 + }, + { + "epoch": 1.89, + "learning_rate": 6.315745961250448e-06, + "loss": 1.186, + "step": 22501 + }, + { + "epoch": 1.89, + "learning_rate": 6.314900943028035e-06, + "loss": 1.2573, + "step": 22502 + }, + { + "epoch": 1.89, + "learning_rate": 6.314055955253088e-06, + "loss": 1.2415, + "step": 22503 + }, + { + "epoch": 1.89, + "learning_rate": 6.313210997932592e-06, + "loss": 1.2245, + "step": 22504 + }, + { + "epoch": 1.89, + "learning_rate": 6.312366071073523e-06, + "loss": 1.1573, + "step": 22505 + }, + { + "epoch": 1.89, + "learning_rate": 6.311521174682868e-06, + "loss": 1.1613, + "step": 22506 + }, + { + "epoch": 1.89, + "learning_rate": 6.310676308767606e-06, + "loss": 1.1572, + "step": 22507 + }, + { + "epoch": 1.89, + "learning_rate": 6.309831473334718e-06, + "loss": 1.171, + "step": 22508 + }, + { + "epoch": 1.9, + "learning_rate": 6.30898666839118e-06, + "loss": 1.1073, + "step": 22509 + }, + { + "epoch": 1.9, + "learning_rate": 6.308141893943978e-06, + "loss": 1.0769, + "step": 22510 + }, + { + "epoch": 1.9, + "learning_rate": 6.307297150000089e-06, + "loss": 1.2108, + "step": 22511 + }, + { + "epoch": 1.9, + "learning_rate": 6.306452436566489e-06, + "loss": 1.1444, + "step": 22512 + }, + { + "epoch": 1.9, + "learning_rate": 6.305607753650166e-06, + "loss": 1.1899, + "step": 22513 + }, + { + "epoch": 1.9, + "learning_rate": 6.304763101258091e-06, + "loss": 1.17, + "step": 22514 + }, + { + "epoch": 1.9, + "learning_rate": 6.30391847939725e-06, + "loss": 1.205, + "step": 22515 + }, + { + "epoch": 1.9, + "learning_rate": 6.303073888074611e-06, + "loss": 1.0937, + "step": 22516 + }, + { + "epoch": 1.9, + "learning_rate": 6.302229327297164e-06, + "loss": 1.1, + "step": 22517 + }, + { + "epoch": 1.9, + "learning_rate": 6.301384797071881e-06, + "loss": 1.2302, + "step": 22518 + }, + { + "epoch": 1.9, + "learning_rate": 6.3005402974057415e-06, + "loss": 1.0952, + "step": 22519 + }, + { + "epoch": 1.9, + "learning_rate": 6.299695828305717e-06, + "loss": 1.1734, + "step": 22520 + }, + { + "epoch": 1.9, + "learning_rate": 6.298851389778795e-06, + "loss": 1.1001, + "step": 22521 + }, + { + "epoch": 1.9, + "learning_rate": 6.298006981831944e-06, + "loss": 1.2347, + "step": 22522 + }, + { + "epoch": 1.9, + "learning_rate": 6.297162604472148e-06, + "loss": 1.1339, + "step": 22523 + }, + { + "epoch": 1.9, + "learning_rate": 6.2963182577063745e-06, + "loss": 1.1457, + "step": 22524 + }, + { + "epoch": 1.9, + "learning_rate": 6.295473941541607e-06, + "loss": 1.2612, + "step": 22525 + }, + { + "epoch": 1.9, + "learning_rate": 6.294629655984821e-06, + "loss": 1.1644, + "step": 22526 + }, + { + "epoch": 1.9, + "learning_rate": 6.2937854010429866e-06, + "loss": 1.1896, + "step": 22527 + }, + { + "epoch": 1.9, + "learning_rate": 6.292941176723089e-06, + "loss": 1.2231, + "step": 22528 + }, + { + "epoch": 1.9, + "learning_rate": 6.292096983032094e-06, + "loss": 1.3108, + "step": 22529 + }, + { + "epoch": 1.9, + "learning_rate": 6.2912528199769825e-06, + "loss": 1.2209, + "step": 22530 + }, + { + "epoch": 1.9, + "learning_rate": 6.290408687564724e-06, + "loss": 1.2118, + "step": 22531 + }, + { + "epoch": 1.9, + "learning_rate": 6.2895645858023e-06, + "loss": 1.1956, + "step": 22532 + }, + { + "epoch": 1.9, + "learning_rate": 6.288720514696679e-06, + "loss": 1.2188, + "step": 22533 + }, + { + "epoch": 1.9, + "learning_rate": 6.2878764742548396e-06, + "loss": 1.1681, + "step": 22534 + }, + { + "epoch": 1.9, + "learning_rate": 6.287032464483747e-06, + "loss": 1.1564, + "step": 22535 + }, + { + "epoch": 1.9, + "learning_rate": 6.286188485390383e-06, + "loss": 1.2049, + "step": 22536 + }, + { + "epoch": 1.9, + "learning_rate": 6.285344536981721e-06, + "loss": 1.2015, + "step": 22537 + }, + { + "epoch": 1.9, + "learning_rate": 6.284500619264726e-06, + "loss": 1.1141, + "step": 22538 + }, + { + "epoch": 1.9, + "learning_rate": 6.283656732246381e-06, + "loss": 1.1495, + "step": 22539 + }, + { + "epoch": 1.9, + "learning_rate": 6.282812875933652e-06, + "loss": 1.2105, + "step": 22540 + }, + { + "epoch": 1.9, + "learning_rate": 6.281969050333514e-06, + "loss": 1.191, + "step": 22541 + }, + { + "epoch": 1.9, + "learning_rate": 6.281125255452934e-06, + "loss": 1.1639, + "step": 22542 + }, + { + "epoch": 1.9, + "learning_rate": 6.280281491298892e-06, + "loss": 1.1858, + "step": 22543 + }, + { + "epoch": 1.9, + "learning_rate": 6.27943775787835e-06, + "loss": 1.1607, + "step": 22544 + }, + { + "epoch": 1.9, + "learning_rate": 6.278594055198288e-06, + "loss": 1.1063, + "step": 22545 + }, + { + "epoch": 1.9, + "learning_rate": 6.2777503832656706e-06, + "loss": 1.153, + "step": 22546 + }, + { + "epoch": 1.9, + "learning_rate": 6.276906742087473e-06, + "loss": 1.0827, + "step": 22547 + }, + { + "epoch": 1.9, + "learning_rate": 6.276063131670659e-06, + "loss": 1.209, + "step": 22548 + }, + { + "epoch": 1.9, + "learning_rate": 6.275219552022204e-06, + "loss": 1.1346, + "step": 22549 + }, + { + "epoch": 1.9, + "learning_rate": 6.274376003149082e-06, + "loss": 1.2534, + "step": 22550 + }, + { + "epoch": 1.9, + "learning_rate": 6.273532485058252e-06, + "loss": 1.1146, + "step": 22551 + }, + { + "epoch": 1.9, + "learning_rate": 6.272688997756694e-06, + "loss": 1.1944, + "step": 22552 + }, + { + "epoch": 1.9, + "learning_rate": 6.27184554125137e-06, + "loss": 1.0964, + "step": 22553 + }, + { + "epoch": 1.9, + "learning_rate": 6.271002115549254e-06, + "loss": 1.0846, + "step": 22554 + }, + { + "epoch": 1.9, + "learning_rate": 6.270158720657307e-06, + "loss": 1.2351, + "step": 22555 + }, + { + "epoch": 1.9, + "learning_rate": 6.269315356582506e-06, + "loss": 1.1505, + "step": 22556 + }, + { + "epoch": 1.9, + "learning_rate": 6.268472023331815e-06, + "loss": 1.1269, + "step": 22557 + }, + { + "epoch": 1.9, + "learning_rate": 6.267628720912204e-06, + "loss": 1.1677, + "step": 22558 + }, + { + "epoch": 1.9, + "learning_rate": 6.266785449330635e-06, + "loss": 1.1151, + "step": 22559 + }, + { + "epoch": 1.9, + "learning_rate": 6.265942208594082e-06, + "loss": 1.1472, + "step": 22560 + }, + { + "epoch": 1.9, + "learning_rate": 6.265098998709509e-06, + "loss": 1.2007, + "step": 22561 + }, + { + "epoch": 1.9, + "learning_rate": 6.26425581968388e-06, + "loss": 1.106, + "step": 22562 + }, + { + "epoch": 1.9, + "learning_rate": 6.263412671524169e-06, + "loss": 1.2905, + "step": 22563 + }, + { + "epoch": 1.9, + "learning_rate": 6.262569554237339e-06, + "loss": 1.2217, + "step": 22564 + }, + { + "epoch": 1.9, + "learning_rate": 6.261726467830356e-06, + "loss": 1.2739, + "step": 22565 + }, + { + "epoch": 1.9, + "learning_rate": 6.26088341231018e-06, + "loss": 1.1901, + "step": 22566 + }, + { + "epoch": 1.9, + "learning_rate": 6.260040387683788e-06, + "loss": 1.1659, + "step": 22567 + }, + { + "epoch": 1.9, + "learning_rate": 6.259197393958136e-06, + "loss": 1.2317, + "step": 22568 + }, + { + "epoch": 1.9, + "learning_rate": 6.258354431140195e-06, + "loss": 1.129, + "step": 22569 + }, + { + "epoch": 1.9, + "learning_rate": 6.257511499236921e-06, + "loss": 1.2322, + "step": 22570 + }, + { + "epoch": 1.9, + "learning_rate": 6.256668598255291e-06, + "loss": 1.2964, + "step": 22571 + }, + { + "epoch": 1.9, + "learning_rate": 6.255825728202261e-06, + "loss": 1.2523, + "step": 22572 + }, + { + "epoch": 1.9, + "learning_rate": 6.2549828890847985e-06, + "loss": 1.2008, + "step": 22573 + }, + { + "epoch": 1.9, + "learning_rate": 6.254140080909863e-06, + "loss": 1.1219, + "step": 22574 + }, + { + "epoch": 1.9, + "learning_rate": 6.253297303684421e-06, + "loss": 1.1715, + "step": 22575 + }, + { + "epoch": 1.9, + "learning_rate": 6.252454557415439e-06, + "loss": 1.1867, + "step": 22576 + }, + { + "epoch": 1.9, + "learning_rate": 6.251611842109873e-06, + "loss": 1.1369, + "step": 22577 + }, + { + "epoch": 1.9, + "learning_rate": 6.250769157774693e-06, + "loss": 1.1662, + "step": 22578 + }, + { + "epoch": 1.9, + "learning_rate": 6.249926504416857e-06, + "loss": 1.1338, + "step": 22579 + }, + { + "epoch": 1.9, + "learning_rate": 6.24908388204333e-06, + "loss": 1.2361, + "step": 22580 + }, + { + "epoch": 1.9, + "learning_rate": 6.248241290661068e-06, + "loss": 1.2195, + "step": 22581 + }, + { + "epoch": 1.9, + "learning_rate": 6.2473987302770416e-06, + "loss": 1.1069, + "step": 22582 + }, + { + "epoch": 1.9, + "learning_rate": 6.246556200898205e-06, + "loss": 1.3136, + "step": 22583 + }, + { + "epoch": 1.9, + "learning_rate": 6.245713702531525e-06, + "loss": 1.2662, + "step": 22584 + }, + { + "epoch": 1.9, + "learning_rate": 6.244871235183956e-06, + "loss": 1.0706, + "step": 22585 + }, + { + "epoch": 1.9, + "learning_rate": 6.2440287988624645e-06, + "loss": 1.1642, + "step": 22586 + }, + { + "epoch": 1.9, + "learning_rate": 6.243186393574011e-06, + "loss": 1.1095, + "step": 22587 + }, + { + "epoch": 1.9, + "learning_rate": 6.24234401932555e-06, + "loss": 1.2566, + "step": 22588 + }, + { + "epoch": 1.9, + "learning_rate": 6.24150167612405e-06, + "loss": 1.1701, + "step": 22589 + }, + { + "epoch": 1.9, + "learning_rate": 6.2406593639764624e-06, + "loss": 1.1451, + "step": 22590 + }, + { + "epoch": 1.9, + "learning_rate": 6.239817082889754e-06, + "loss": 1.2042, + "step": 22591 + }, + { + "epoch": 1.9, + "learning_rate": 6.238974832870875e-06, + "loss": 1.1191, + "step": 22592 + }, + { + "epoch": 1.9, + "learning_rate": 6.238132613926797e-06, + "loss": 1.1258, + "step": 22593 + }, + { + "epoch": 1.9, + "learning_rate": 6.237290426064464e-06, + "loss": 1.1872, + "step": 22594 + }, + { + "epoch": 1.9, + "learning_rate": 6.236448269290847e-06, + "loss": 1.2094, + "step": 22595 + }, + { + "epoch": 1.9, + "learning_rate": 6.235606143612895e-06, + "loss": 1.1174, + "step": 22596 + }, + { + "epoch": 1.9, + "learning_rate": 6.234764049037574e-06, + "loss": 1.295, + "step": 22597 + }, + { + "epoch": 1.9, + "learning_rate": 6.233921985571831e-06, + "loss": 1.2375, + "step": 22598 + }, + { + "epoch": 1.9, + "learning_rate": 6.233079953222633e-06, + "loss": 1.176, + "step": 22599 + }, + { + "epoch": 1.9, + "learning_rate": 6.232237951996937e-06, + "loss": 1.1401, + "step": 22600 + }, + { + "epoch": 1.9, + "learning_rate": 6.231395981901691e-06, + "loss": 1.2034, + "step": 22601 + }, + { + "epoch": 1.9, + "learning_rate": 6.230554042943864e-06, + "loss": 1.2015, + "step": 22602 + }, + { + "epoch": 1.9, + "learning_rate": 6.229712135130401e-06, + "loss": 1.0804, + "step": 22603 + }, + { + "epoch": 1.9, + "learning_rate": 6.228870258468267e-06, + "loss": 1.1222, + "step": 22604 + }, + { + "epoch": 1.9, + "learning_rate": 6.2280284129644085e-06, + "loss": 1.0589, + "step": 22605 + }, + { + "epoch": 1.9, + "learning_rate": 6.2271865986257915e-06, + "loss": 1.1812, + "step": 22606 + }, + { + "epoch": 1.9, + "learning_rate": 6.226344815459364e-06, + "loss": 1.1518, + "step": 22607 + }, + { + "epoch": 1.9, + "learning_rate": 6.225503063472086e-06, + "loss": 1.2142, + "step": 22608 + }, + { + "epoch": 1.9, + "learning_rate": 6.224661342670905e-06, + "loss": 1.1537, + "step": 22609 + }, + { + "epoch": 1.9, + "learning_rate": 6.223819653062783e-06, + "loss": 1.1667, + "step": 22610 + }, + { + "epoch": 1.9, + "learning_rate": 6.222977994654672e-06, + "loss": 1.2368, + "step": 22611 + }, + { + "epoch": 1.9, + "learning_rate": 6.222136367453522e-06, + "loss": 1.1327, + "step": 22612 + }, + { + "epoch": 1.9, + "learning_rate": 6.221294771466295e-06, + "loss": 1.2014, + "step": 22613 + }, + { + "epoch": 1.9, + "learning_rate": 6.220453206699938e-06, + "loss": 1.153, + "step": 22614 + }, + { + "epoch": 1.9, + "learning_rate": 6.219611673161407e-06, + "loss": 1.1532, + "step": 22615 + }, + { + "epoch": 1.9, + "learning_rate": 6.218770170857652e-06, + "loss": 1.1545, + "step": 22616 + }, + { + "epoch": 1.9, + "learning_rate": 6.217928699795631e-06, + "loss": 1.191, + "step": 22617 + }, + { + "epoch": 1.9, + "learning_rate": 6.217087259982293e-06, + "loss": 1.2321, + "step": 22618 + }, + { + "epoch": 1.9, + "learning_rate": 6.216245851424591e-06, + "loss": 1.1951, + "step": 22619 + }, + { + "epoch": 1.9, + "learning_rate": 6.215404474129473e-06, + "loss": 1.1627, + "step": 22620 + }, + { + "epoch": 1.9, + "learning_rate": 6.214563128103899e-06, + "loss": 1.2524, + "step": 22621 + }, + { + "epoch": 1.9, + "learning_rate": 6.2137218133548125e-06, + "loss": 1.2313, + "step": 22622 + }, + { + "epoch": 1.9, + "learning_rate": 6.212880529889171e-06, + "loss": 1.2204, + "step": 22623 + }, + { + "epoch": 1.9, + "learning_rate": 6.212039277713919e-06, + "loss": 1.1312, + "step": 22624 + }, + { + "epoch": 1.9, + "learning_rate": 6.211198056836013e-06, + "loss": 1.1688, + "step": 22625 + }, + { + "epoch": 1.9, + "learning_rate": 6.210356867262403e-06, + "loss": 1.2103, + "step": 22626 + }, + { + "epoch": 1.91, + "learning_rate": 6.209515709000034e-06, + "loss": 1.1891, + "step": 22627 + }, + { + "epoch": 1.91, + "learning_rate": 6.2086745820558625e-06, + "loss": 1.0807, + "step": 22628 + }, + { + "epoch": 1.91, + "learning_rate": 6.207833486436832e-06, + "loss": 1.1803, + "step": 22629 + }, + { + "epoch": 1.91, + "learning_rate": 6.206992422149898e-06, + "loss": 1.2265, + "step": 22630 + }, + { + "epoch": 1.91, + "learning_rate": 6.206151389202002e-06, + "loss": 1.1599, + "step": 22631 + }, + { + "epoch": 1.91, + "learning_rate": 6.205310387600103e-06, + "loss": 1.1105, + "step": 22632 + }, + { + "epoch": 1.91, + "learning_rate": 6.20446941735114e-06, + "loss": 1.1448, + "step": 22633 + }, + { + "epoch": 1.91, + "learning_rate": 6.2036284784620696e-06, + "loss": 1.1494, + "step": 22634 + }, + { + "epoch": 1.91, + "learning_rate": 6.20278757093983e-06, + "loss": 1.0835, + "step": 22635 + }, + { + "epoch": 1.91, + "learning_rate": 6.201946694791376e-06, + "loss": 1.2594, + "step": 22636 + }, + { + "epoch": 1.91, + "learning_rate": 6.201105850023657e-06, + "loss": 1.137, + "step": 22637 + }, + { + "epoch": 1.91, + "learning_rate": 6.200265036643613e-06, + "loss": 1.1599, + "step": 22638 + }, + { + "epoch": 1.91, + "learning_rate": 6.199424254658202e-06, + "loss": 1.2625, + "step": 22639 + }, + { + "epoch": 1.91, + "learning_rate": 6.198583504074358e-06, + "loss": 1.2547, + "step": 22640 + }, + { + "epoch": 1.91, + "learning_rate": 6.197742784899039e-06, + "loss": 1.1609, + "step": 22641 + }, + { + "epoch": 1.91, + "learning_rate": 6.196902097139179e-06, + "loss": 1.2055, + "step": 22642 + }, + { + "epoch": 1.91, + "learning_rate": 6.196061440801739e-06, + "loss": 1.1922, + "step": 22643 + }, + { + "epoch": 1.91, + "learning_rate": 6.19522081589365e-06, + "loss": 1.076, + "step": 22644 + }, + { + "epoch": 1.91, + "learning_rate": 6.19438022242187e-06, + "loss": 1.2499, + "step": 22645 + }, + { + "epoch": 1.91, + "learning_rate": 6.193539660393335e-06, + "loss": 1.1731, + "step": 22646 + }, + { + "epoch": 1.91, + "learning_rate": 6.192699129814996e-06, + "loss": 1.1584, + "step": 22647 + }, + { + "epoch": 1.91, + "learning_rate": 6.19185863069379e-06, + "loss": 1.2116, + "step": 22648 + }, + { + "epoch": 1.91, + "learning_rate": 6.19101816303667e-06, + "loss": 1.1584, + "step": 22649 + }, + { + "epoch": 1.91, + "learning_rate": 6.190177726850579e-06, + "loss": 1.2561, + "step": 22650 + }, + { + "epoch": 1.91, + "learning_rate": 6.189337322142455e-06, + "loss": 1.1452, + "step": 22651 + }, + { + "epoch": 1.91, + "learning_rate": 6.188496948919249e-06, + "loss": 1.1451, + "step": 22652 + }, + { + "epoch": 1.91, + "learning_rate": 6.1876566071879e-06, + "loss": 1.2295, + "step": 22653 + }, + { + "epoch": 1.91, + "learning_rate": 6.186816296955354e-06, + "loss": 1.2446, + "step": 22654 + }, + { + "epoch": 1.91, + "learning_rate": 6.185976018228547e-06, + "loss": 1.1478, + "step": 22655 + }, + { + "epoch": 1.91, + "learning_rate": 6.1851357710144315e-06, + "loss": 1.2673, + "step": 22656 + }, + { + "epoch": 1.91, + "learning_rate": 6.184295555319943e-06, + "loss": 1.2324, + "step": 22657 + }, + { + "epoch": 1.91, + "learning_rate": 6.183455371152027e-06, + "loss": 1.2074, + "step": 22658 + }, + { + "epoch": 1.91, + "learning_rate": 6.1826152185176225e-06, + "loss": 1.1016, + "step": 22659 + }, + { + "epoch": 1.91, + "learning_rate": 6.1817750974236745e-06, + "loss": 1.1619, + "step": 22660 + }, + { + "epoch": 1.91, + "learning_rate": 6.180935007877122e-06, + "loss": 1.1727, + "step": 22661 + }, + { + "epoch": 1.91, + "learning_rate": 6.1800949498849035e-06, + "loss": 1.1611, + "step": 22662 + }, + { + "epoch": 1.91, + "learning_rate": 6.179254923453968e-06, + "loss": 1.1337, + "step": 22663 + }, + { + "epoch": 1.91, + "learning_rate": 6.17841492859125e-06, + "loss": 1.1685, + "step": 22664 + }, + { + "epoch": 1.91, + "learning_rate": 6.177574965303691e-06, + "loss": 1.1603, + "step": 22665 + }, + { + "epoch": 1.91, + "learning_rate": 6.176735033598228e-06, + "loss": 1.1586, + "step": 22666 + }, + { + "epoch": 1.91, + "learning_rate": 6.175895133481807e-06, + "loss": 1.1835, + "step": 22667 + }, + { + "epoch": 1.91, + "learning_rate": 6.175055264961363e-06, + "loss": 1.1951, + "step": 22668 + }, + { + "epoch": 1.91, + "learning_rate": 6.17421542804384e-06, + "loss": 1.2405, + "step": 22669 + }, + { + "epoch": 1.91, + "learning_rate": 6.173375622736166e-06, + "loss": 1.0479, + "step": 22670 + }, + { + "epoch": 1.91, + "learning_rate": 6.172535849045294e-06, + "loss": 1.0752, + "step": 22671 + }, + { + "epoch": 1.91, + "learning_rate": 6.171696106978153e-06, + "loss": 1.1798, + "step": 22672 + }, + { + "epoch": 1.91, + "learning_rate": 6.170856396541687e-06, + "loss": 1.2049, + "step": 22673 + }, + { + "epoch": 1.91, + "learning_rate": 6.170016717742825e-06, + "loss": 1.1989, + "step": 22674 + }, + { + "epoch": 1.91, + "learning_rate": 6.169177070588514e-06, + "loss": 1.2072, + "step": 22675 + }, + { + "epoch": 1.91, + "learning_rate": 6.168337455085689e-06, + "loss": 1.1382, + "step": 22676 + }, + { + "epoch": 1.91, + "learning_rate": 6.167497871241283e-06, + "loss": 1.0757, + "step": 22677 + }, + { + "epoch": 1.91, + "learning_rate": 6.16665831906224e-06, + "loss": 1.0771, + "step": 22678 + }, + { + "epoch": 1.91, + "learning_rate": 6.165818798555491e-06, + "loss": 1.1538, + "step": 22679 + }, + { + "epoch": 1.91, + "learning_rate": 6.164979309727975e-06, + "loss": 1.2162, + "step": 22680 + }, + { + "epoch": 1.91, + "learning_rate": 6.164139852586625e-06, + "loss": 1.0956, + "step": 22681 + }, + { + "epoch": 1.91, + "learning_rate": 6.163300427138382e-06, + "loss": 1.2173, + "step": 22682 + }, + { + "epoch": 1.91, + "learning_rate": 6.162461033390177e-06, + "loss": 1.0669, + "step": 22683 + }, + { + "epoch": 1.91, + "learning_rate": 6.1616216713489496e-06, + "loss": 1.0884, + "step": 22684 + }, + { + "epoch": 1.91, + "learning_rate": 6.160782341021629e-06, + "loss": 1.1838, + "step": 22685 + }, + { + "epoch": 1.91, + "learning_rate": 6.159943042415153e-06, + "loss": 1.1945, + "step": 22686 + }, + { + "epoch": 1.91, + "learning_rate": 6.159103775536461e-06, + "loss": 1.2195, + "step": 22687 + }, + { + "epoch": 1.91, + "learning_rate": 6.158264540392478e-06, + "loss": 1.1532, + "step": 22688 + }, + { + "epoch": 1.91, + "learning_rate": 6.157425336990146e-06, + "loss": 1.2726, + "step": 22689 + }, + { + "epoch": 1.91, + "learning_rate": 6.156586165336396e-06, + "loss": 1.1608, + "step": 22690 + }, + { + "epoch": 1.91, + "learning_rate": 6.155747025438162e-06, + "loss": 1.1355, + "step": 22691 + }, + { + "epoch": 1.91, + "learning_rate": 6.15490791730237e-06, + "loss": 1.2284, + "step": 22692 + }, + { + "epoch": 1.91, + "learning_rate": 6.154068840935965e-06, + "loss": 1.1903, + "step": 22693 + }, + { + "epoch": 1.91, + "learning_rate": 6.153229796345873e-06, + "loss": 1.1738, + "step": 22694 + }, + { + "epoch": 1.91, + "learning_rate": 6.15239078353903e-06, + "loss": 1.2554, + "step": 22695 + }, + { + "epoch": 1.91, + "learning_rate": 6.15155180252236e-06, + "loss": 1.1432, + "step": 22696 + }, + { + "epoch": 1.91, + "learning_rate": 6.150712853302807e-06, + "loss": 1.1725, + "step": 22697 + }, + { + "epoch": 1.91, + "learning_rate": 6.149873935887292e-06, + "loss": 1.1543, + "step": 22698 + }, + { + "epoch": 1.91, + "learning_rate": 6.149035050282748e-06, + "loss": 1.1984, + "step": 22699 + }, + { + "epoch": 1.91, + "learning_rate": 6.1481961964961165e-06, + "loss": 1.1517, + "step": 22700 + }, + { + "epoch": 1.91, + "learning_rate": 6.1473573745343155e-06, + "loss": 1.1876, + "step": 22701 + }, + { + "epoch": 1.91, + "learning_rate": 6.146518584404284e-06, + "loss": 1.2309, + "step": 22702 + }, + { + "epoch": 1.91, + "learning_rate": 6.1456798261129465e-06, + "loss": 1.1617, + "step": 22703 + }, + { + "epoch": 1.91, + "learning_rate": 6.1448410996672385e-06, + "loss": 1.1286, + "step": 22704 + }, + { + "epoch": 1.91, + "learning_rate": 6.144002405074083e-06, + "loss": 1.2626, + "step": 22705 + }, + { + "epoch": 1.91, + "learning_rate": 6.143163742340419e-06, + "loss": 1.1286, + "step": 22706 + }, + { + "epoch": 1.91, + "learning_rate": 6.142325111473167e-06, + "loss": 1.1957, + "step": 22707 + }, + { + "epoch": 1.91, + "learning_rate": 6.141486512479262e-06, + "loss": 1.1199, + "step": 22708 + }, + { + "epoch": 1.91, + "learning_rate": 6.140647945365625e-06, + "loss": 1.1558, + "step": 22709 + }, + { + "epoch": 1.91, + "learning_rate": 6.139809410139197e-06, + "loss": 1.1847, + "step": 22710 + }, + { + "epoch": 1.91, + "learning_rate": 6.138970906806894e-06, + "loss": 1.1733, + "step": 22711 + }, + { + "epoch": 1.91, + "learning_rate": 6.138132435375648e-06, + "loss": 1.1102, + "step": 22712 + }, + { + "epoch": 1.91, + "learning_rate": 6.137293995852392e-06, + "loss": 1.187, + "step": 22713 + }, + { + "epoch": 1.91, + "learning_rate": 6.136455588244048e-06, + "loss": 1.0877, + "step": 22714 + }, + { + "epoch": 1.91, + "learning_rate": 6.135617212557545e-06, + "loss": 1.1075, + "step": 22715 + }, + { + "epoch": 1.91, + "learning_rate": 6.134778868799807e-06, + "loss": 1.1613, + "step": 22716 + }, + { + "epoch": 1.91, + "learning_rate": 6.133940556977766e-06, + "loss": 1.1285, + "step": 22717 + }, + { + "epoch": 1.91, + "learning_rate": 6.1331022770983425e-06, + "loss": 1.1846, + "step": 22718 + }, + { + "epoch": 1.91, + "learning_rate": 6.1322640291684685e-06, + "loss": 1.1241, + "step": 22719 + }, + { + "epoch": 1.91, + "learning_rate": 6.131425813195063e-06, + "loss": 1.224, + "step": 22720 + }, + { + "epoch": 1.91, + "learning_rate": 6.13058762918506e-06, + "loss": 1.19, + "step": 22721 + }, + { + "epoch": 1.91, + "learning_rate": 6.129749477145377e-06, + "loss": 1.1146, + "step": 22722 + }, + { + "epoch": 1.91, + "learning_rate": 6.128911357082945e-06, + "loss": 1.1956, + "step": 22723 + }, + { + "epoch": 1.91, + "learning_rate": 6.128073269004684e-06, + "loss": 1.1807, + "step": 22724 + }, + { + "epoch": 1.91, + "learning_rate": 6.12723521291752e-06, + "loss": 1.1557, + "step": 22725 + }, + { + "epoch": 1.91, + "learning_rate": 6.126397188828381e-06, + "loss": 1.1703, + "step": 22726 + }, + { + "epoch": 1.91, + "learning_rate": 6.125559196744186e-06, + "loss": 1.2299, + "step": 22727 + }, + { + "epoch": 1.91, + "learning_rate": 6.1247212366718635e-06, + "loss": 1.2007, + "step": 22728 + }, + { + "epoch": 1.91, + "learning_rate": 6.123883308618332e-06, + "loss": 1.2177, + "step": 22729 + }, + { + "epoch": 1.91, + "learning_rate": 6.1230454125905205e-06, + "loss": 1.168, + "step": 22730 + }, + { + "epoch": 1.91, + "learning_rate": 6.1222075485953425e-06, + "loss": 1.2234, + "step": 22731 + }, + { + "epoch": 1.91, + "learning_rate": 6.121369716639732e-06, + "loss": 1.1405, + "step": 22732 + }, + { + "epoch": 1.91, + "learning_rate": 6.120531916730604e-06, + "loss": 1.2096, + "step": 22733 + }, + { + "epoch": 1.91, + "learning_rate": 6.1196941488748855e-06, + "loss": 1.2625, + "step": 22734 + }, + { + "epoch": 1.91, + "learning_rate": 6.118856413079492e-06, + "loss": 1.1963, + "step": 22735 + }, + { + "epoch": 1.91, + "learning_rate": 6.11801870935135e-06, + "loss": 1.1494, + "step": 22736 + }, + { + "epoch": 1.91, + "learning_rate": 6.117181037697382e-06, + "loss": 1.3479, + "step": 22737 + }, + { + "epoch": 1.91, + "learning_rate": 6.116343398124502e-06, + "loss": 1.2658, + "step": 22738 + }, + { + "epoch": 1.91, + "learning_rate": 6.115505790639642e-06, + "loss": 1.1868, + "step": 22739 + }, + { + "epoch": 1.91, + "learning_rate": 6.114668215249713e-06, + "loss": 1.1361, + "step": 22740 + }, + { + "epoch": 1.91, + "learning_rate": 6.1138306719616405e-06, + "loss": 1.1537, + "step": 22741 + }, + { + "epoch": 1.91, + "learning_rate": 6.112993160782339e-06, + "loss": 1.1982, + "step": 22742 + }, + { + "epoch": 1.91, + "learning_rate": 6.112155681718736e-06, + "loss": 1.1611, + "step": 22743 + }, + { + "epoch": 1.91, + "learning_rate": 6.111318234777746e-06, + "loss": 1.1697, + "step": 22744 + }, + { + "epoch": 1.91, + "learning_rate": 6.1104808199662894e-06, + "loss": 1.1852, + "step": 22745 + }, + { + "epoch": 1.92, + "learning_rate": 6.109643437291282e-06, + "loss": 1.1618, + "step": 22746 + }, + { + "epoch": 1.92, + "learning_rate": 6.1088060867596515e-06, + "loss": 1.0821, + "step": 22747 + }, + { + "epoch": 1.92, + "learning_rate": 6.107968768378305e-06, + "loss": 1.1697, + "step": 22748 + }, + { + "epoch": 1.92, + "learning_rate": 6.107131482154165e-06, + "loss": 1.0722, + "step": 22749 + }, + { + "epoch": 1.92, + "learning_rate": 6.106294228094157e-06, + "loss": 1.1256, + "step": 22750 + }, + { + "epoch": 1.92, + "learning_rate": 6.1054570062051865e-06, + "loss": 1.1595, + "step": 22751 + }, + { + "epoch": 1.92, + "learning_rate": 6.104619816494182e-06, + "loss": 1.2052, + "step": 22752 + }, + { + "epoch": 1.92, + "learning_rate": 6.103782658968051e-06, + "loss": 1.1796, + "step": 22753 + }, + { + "epoch": 1.92, + "learning_rate": 6.102945533633717e-06, + "loss": 1.2135, + "step": 22754 + }, + { + "epoch": 1.92, + "learning_rate": 6.1021084404980915e-06, + "loss": 1.1696, + "step": 22755 + }, + { + "epoch": 1.92, + "learning_rate": 6.1012713795680965e-06, + "loss": 1.1141, + "step": 22756 + }, + { + "epoch": 1.92, + "learning_rate": 6.100434350850643e-06, + "loss": 1.2171, + "step": 22757 + }, + { + "epoch": 1.92, + "learning_rate": 6.099597354352651e-06, + "loss": 1.1781, + "step": 22758 + }, + { + "epoch": 1.92, + "learning_rate": 6.098760390081031e-06, + "loss": 1.1329, + "step": 22759 + }, + { + "epoch": 1.92, + "learning_rate": 6.097923458042705e-06, + "loss": 1.1566, + "step": 22760 + }, + { + "epoch": 1.92, + "learning_rate": 6.09708655824458e-06, + "loss": 1.1392, + "step": 22761 + }, + { + "epoch": 1.92, + "learning_rate": 6.096249690693577e-06, + "loss": 1.2735, + "step": 22762 + }, + { + "epoch": 1.92, + "learning_rate": 6.0954128553966095e-06, + "loss": 1.2653, + "step": 22763 + }, + { + "epoch": 1.92, + "learning_rate": 6.094576052360588e-06, + "loss": 1.1096, + "step": 22764 + }, + { + "epoch": 1.92, + "learning_rate": 6.093739281592432e-06, + "loss": 1.2362, + "step": 22765 + }, + { + "epoch": 1.92, + "learning_rate": 6.092902543099049e-06, + "loss": 1.0941, + "step": 22766 + }, + { + "epoch": 1.92, + "learning_rate": 6.09206583688736e-06, + "loss": 1.2637, + "step": 22767 + }, + { + "epoch": 1.92, + "learning_rate": 6.09122916296427e-06, + "loss": 1.213, + "step": 22768 + }, + { + "epoch": 1.92, + "learning_rate": 6.090392521336699e-06, + "loss": 1.0853, + "step": 22769 + }, + { + "epoch": 1.92, + "learning_rate": 6.089555912011551e-06, + "loss": 1.2452, + "step": 22770 + }, + { + "epoch": 1.92, + "learning_rate": 6.088719334995749e-06, + "loss": 1.1846, + "step": 22771 + }, + { + "epoch": 1.92, + "learning_rate": 6.087882790296196e-06, + "loss": 1.2356, + "step": 22772 + }, + { + "epoch": 1.92, + "learning_rate": 6.08704627791981e-06, + "loss": 1.1846, + "step": 22773 + }, + { + "epoch": 1.92, + "learning_rate": 6.086209797873497e-06, + "loss": 1.1643, + "step": 22774 + }, + { + "epoch": 1.92, + "learning_rate": 6.085373350164172e-06, + "loss": 1.1368, + "step": 22775 + }, + { + "epoch": 1.92, + "learning_rate": 6.084536934798748e-06, + "loss": 1.1454, + "step": 22776 + }, + { + "epoch": 1.92, + "learning_rate": 6.083700551784129e-06, + "loss": 1.1365, + "step": 22777 + }, + { + "epoch": 1.92, + "learning_rate": 6.082864201127232e-06, + "loss": 1.2206, + "step": 22778 + }, + { + "epoch": 1.92, + "learning_rate": 6.082027882834964e-06, + "loss": 1.1544, + "step": 22779 + }, + { + "epoch": 1.92, + "learning_rate": 6.081191596914237e-06, + "loss": 1.2092, + "step": 22780 + }, + { + "epoch": 1.92, + "learning_rate": 6.080355343371956e-06, + "loss": 1.1788, + "step": 22781 + }, + { + "epoch": 1.92, + "learning_rate": 6.079519122215036e-06, + "loss": 1.1941, + "step": 22782 + }, + { + "epoch": 1.92, + "learning_rate": 6.078682933450384e-06, + "loss": 1.1799, + "step": 22783 + }, + { + "epoch": 1.92, + "learning_rate": 6.0778467770849085e-06, + "loss": 1.2184, + "step": 22784 + }, + { + "epoch": 1.92, + "learning_rate": 6.077010653125515e-06, + "loss": 1.1783, + "step": 22785 + }, + { + "epoch": 1.92, + "learning_rate": 6.076174561579118e-06, + "loss": 1.2654, + "step": 22786 + }, + { + "epoch": 1.92, + "learning_rate": 6.075338502452623e-06, + "loss": 1.2179, + "step": 22787 + }, + { + "epoch": 1.92, + "learning_rate": 6.074502475752933e-06, + "loss": 1.1538, + "step": 22788 + }, + { + "epoch": 1.92, + "learning_rate": 6.073666481486966e-06, + "loss": 1.2035, + "step": 22789 + }, + { + "epoch": 1.92, + "learning_rate": 6.072830519661619e-06, + "loss": 1.1672, + "step": 22790 + }, + { + "epoch": 1.92, + "learning_rate": 6.0719945902838066e-06, + "loss": 1.2411, + "step": 22791 + }, + { + "epoch": 1.92, + "learning_rate": 6.071158693360427e-06, + "loss": 1.1073, + "step": 22792 + }, + { + "epoch": 1.92, + "learning_rate": 6.0703228288983965e-06, + "loss": 1.1469, + "step": 22793 + }, + { + "epoch": 1.92, + "learning_rate": 6.069486996904614e-06, + "loss": 1.2264, + "step": 22794 + }, + { + "epoch": 1.92, + "learning_rate": 6.0686511973859894e-06, + "loss": 1.1507, + "step": 22795 + }, + { + "epoch": 1.92, + "learning_rate": 6.067815430349424e-06, + "loss": 1.1392, + "step": 22796 + }, + { + "epoch": 1.92, + "learning_rate": 6.066979695801831e-06, + "loss": 1.2127, + "step": 22797 + }, + { + "epoch": 1.92, + "learning_rate": 6.0661439937501045e-06, + "loss": 1.2392, + "step": 22798 + }, + { + "epoch": 1.92, + "learning_rate": 6.065308324201155e-06, + "loss": 1.1062, + "step": 22799 + }, + { + "epoch": 1.92, + "learning_rate": 6.064472687161894e-06, + "loss": 1.19, + "step": 22800 + }, + { + "epoch": 1.92, + "learning_rate": 6.063637082639213e-06, + "loss": 1.2019, + "step": 22801 + }, + { + "epoch": 1.92, + "learning_rate": 6.062801510640026e-06, + "loss": 1.1978, + "step": 22802 + }, + { + "epoch": 1.92, + "learning_rate": 6.061965971171232e-06, + "loss": 1.2018, + "step": 22803 + }, + { + "epoch": 1.92, + "learning_rate": 6.061130464239737e-06, + "loss": 1.1841, + "step": 22804 + }, + { + "epoch": 1.92, + "learning_rate": 6.060294989852438e-06, + "loss": 1.1835, + "step": 22805 + }, + { + "epoch": 1.92, + "learning_rate": 6.059459548016247e-06, + "loss": 1.2293, + "step": 22806 + }, + { + "epoch": 1.92, + "learning_rate": 6.0586241387380605e-06, + "loss": 1.1655, + "step": 22807 + }, + { + "epoch": 1.92, + "learning_rate": 6.057788762024785e-06, + "loss": 1.2848, + "step": 22808 + }, + { + "epoch": 1.92, + "learning_rate": 6.056953417883317e-06, + "loss": 1.2643, + "step": 22809 + }, + { + "epoch": 1.92, + "learning_rate": 6.056118106320565e-06, + "loss": 1.136, + "step": 22810 + }, + { + "epoch": 1.92, + "learning_rate": 6.055282827343425e-06, + "loss": 1.1703, + "step": 22811 + }, + { + "epoch": 1.92, + "learning_rate": 6.0544475809588e-06, + "loss": 1.1934, + "step": 22812 + }, + { + "epoch": 1.92, + "learning_rate": 6.053612367173594e-06, + "loss": 1.1645, + "step": 22813 + }, + { + "epoch": 1.92, + "learning_rate": 6.052777185994705e-06, + "loss": 1.1798, + "step": 22814 + }, + { + "epoch": 1.92, + "learning_rate": 6.051942037429035e-06, + "loss": 1.1426, + "step": 22815 + }, + { + "epoch": 1.92, + "learning_rate": 6.051106921483481e-06, + "loss": 1.2489, + "step": 22816 + }, + { + "epoch": 1.92, + "learning_rate": 6.050271838164949e-06, + "loss": 1.1154, + "step": 22817 + }, + { + "epoch": 1.92, + "learning_rate": 6.049436787480332e-06, + "loss": 1.2212, + "step": 22818 + }, + { + "epoch": 1.92, + "learning_rate": 6.048601769436535e-06, + "loss": 1.1918, + "step": 22819 + }, + { + "epoch": 1.92, + "learning_rate": 6.047766784040452e-06, + "loss": 1.147, + "step": 22820 + }, + { + "epoch": 1.92, + "learning_rate": 6.046931831298988e-06, + "loss": 1.1844, + "step": 22821 + }, + { + "epoch": 1.92, + "learning_rate": 6.046096911219036e-06, + "loss": 1.1962, + "step": 22822 + }, + { + "epoch": 1.92, + "learning_rate": 6.045262023807499e-06, + "loss": 1.2266, + "step": 22823 + }, + { + "epoch": 1.92, + "learning_rate": 6.044427169071268e-06, + "loss": 1.1518, + "step": 22824 + }, + { + "epoch": 1.92, + "learning_rate": 6.0435923470172485e-06, + "loss": 1.2266, + "step": 22825 + }, + { + "epoch": 1.92, + "learning_rate": 6.042757557652338e-06, + "loss": 1.1752, + "step": 22826 + }, + { + "epoch": 1.92, + "learning_rate": 6.041922800983427e-06, + "loss": 1.2205, + "step": 22827 + }, + { + "epoch": 1.92, + "learning_rate": 6.041088077017419e-06, + "loss": 1.1154, + "step": 22828 + }, + { + "epoch": 1.92, + "learning_rate": 6.0402533857612075e-06, + "loss": 1.1943, + "step": 22829 + }, + { + "epoch": 1.92, + "learning_rate": 6.0394187272216905e-06, + "loss": 1.1912, + "step": 22830 + }, + { + "epoch": 1.92, + "learning_rate": 6.038584101405761e-06, + "loss": 1.2724, + "step": 22831 + }, + { + "epoch": 1.92, + "learning_rate": 6.037749508320321e-06, + "loss": 1.1652, + "step": 22832 + }, + { + "epoch": 1.92, + "learning_rate": 6.036914947972261e-06, + "loss": 1.0815, + "step": 22833 + }, + { + "epoch": 1.92, + "learning_rate": 6.03608042036848e-06, + "loss": 1.1948, + "step": 22834 + }, + { + "epoch": 1.92, + "learning_rate": 6.035245925515866e-06, + "loss": 1.2175, + "step": 22835 + }, + { + "epoch": 1.92, + "learning_rate": 6.034411463421321e-06, + "loss": 1.1797, + "step": 22836 + }, + { + "epoch": 1.92, + "learning_rate": 6.033577034091742e-06, + "loss": 1.1706, + "step": 22837 + }, + { + "epoch": 1.92, + "learning_rate": 6.032742637534012e-06, + "loss": 1.1976, + "step": 22838 + }, + { + "epoch": 1.92, + "learning_rate": 6.031908273755038e-06, + "loss": 1.2151, + "step": 22839 + }, + { + "epoch": 1.92, + "learning_rate": 6.031073942761705e-06, + "loss": 1.1141, + "step": 22840 + }, + { + "epoch": 1.92, + "learning_rate": 6.0302396445609125e-06, + "loss": 1.1356, + "step": 22841 + }, + { + "epoch": 1.92, + "learning_rate": 6.029405379159545e-06, + "loss": 1.218, + "step": 22842 + }, + { + "epoch": 1.92, + "learning_rate": 6.0285711465645054e-06, + "loss": 1.1925, + "step": 22843 + }, + { + "epoch": 1.92, + "learning_rate": 6.02773694678268e-06, + "loss": 1.1938, + "step": 22844 + }, + { + "epoch": 1.92, + "learning_rate": 6.026902779820967e-06, + "loss": 1.2011, + "step": 22845 + }, + { + "epoch": 1.92, + "learning_rate": 6.026068645686249e-06, + "loss": 1.1146, + "step": 22846 + }, + { + "epoch": 1.92, + "learning_rate": 6.025234544385427e-06, + "loss": 1.1857, + "step": 22847 + }, + { + "epoch": 1.92, + "learning_rate": 6.024400475925389e-06, + "loss": 1.2326, + "step": 22848 + }, + { + "epoch": 1.92, + "learning_rate": 6.023566440313023e-06, + "loss": 1.2232, + "step": 22849 + }, + { + "epoch": 1.92, + "learning_rate": 6.02273243755523e-06, + "loss": 1.0656, + "step": 22850 + }, + { + "epoch": 1.92, + "learning_rate": 6.0218984676588895e-06, + "loss": 1.1166, + "step": 22851 + }, + { + "epoch": 1.92, + "learning_rate": 6.021064530630901e-06, + "loss": 1.2374, + "step": 22852 + }, + { + "epoch": 1.92, + "learning_rate": 6.020230626478146e-06, + "loss": 1.1208, + "step": 22853 + }, + { + "epoch": 1.92, + "learning_rate": 6.019396755207526e-06, + "loss": 1.1117, + "step": 22854 + }, + { + "epoch": 1.92, + "learning_rate": 6.018562916825917e-06, + "loss": 1.2589, + "step": 22855 + }, + { + "epoch": 1.92, + "learning_rate": 6.01772911134022e-06, + "loss": 1.1407, + "step": 22856 + }, + { + "epoch": 1.92, + "learning_rate": 6.016895338757318e-06, + "loss": 1.1617, + "step": 22857 + }, + { + "epoch": 1.92, + "learning_rate": 6.016061599084103e-06, + "loss": 1.1203, + "step": 22858 + }, + { + "epoch": 1.92, + "learning_rate": 6.015227892327457e-06, + "loss": 1.1127, + "step": 22859 + }, + { + "epoch": 1.92, + "learning_rate": 6.01439421849428e-06, + "loss": 1.1142, + "step": 22860 + }, + { + "epoch": 1.92, + "learning_rate": 6.01356057759145e-06, + "loss": 1.1342, + "step": 22861 + }, + { + "epoch": 1.92, + "learning_rate": 6.012726969625855e-06, + "loss": 1.1966, + "step": 22862 + }, + { + "epoch": 1.92, + "learning_rate": 6.011893394604392e-06, + "loss": 1.1527, + "step": 22863 + }, + { + "epoch": 1.92, + "learning_rate": 6.0110598525339395e-06, + "loss": 1.1703, + "step": 22864 + }, + { + "epoch": 1.93, + "learning_rate": 6.010226343421388e-06, + "loss": 1.2178, + "step": 22865 + }, + { + "epoch": 1.93, + "learning_rate": 6.009392867273621e-06, + "loss": 1.265, + "step": 22866 + }, + { + "epoch": 1.93, + "learning_rate": 6.0085594240975305e-06, + "loss": 1.1497, + "step": 22867 + }, + { + "epoch": 1.93, + "learning_rate": 6.007726013899997e-06, + "loss": 1.1472, + "step": 22868 + }, + { + "epoch": 1.93, + "learning_rate": 6.006892636687912e-06, + "loss": 1.1264, + "step": 22869 + }, + { + "epoch": 1.93, + "learning_rate": 6.0060592924681535e-06, + "loss": 1.1863, + "step": 22870 + }, + { + "epoch": 1.93, + "learning_rate": 6.0052259812476155e-06, + "loss": 1.1765, + "step": 22871 + }, + { + "epoch": 1.93, + "learning_rate": 6.004392703033177e-06, + "loss": 1.1764, + "step": 22872 + }, + { + "epoch": 1.93, + "learning_rate": 6.003559457831727e-06, + "loss": 1.1677, + "step": 22873 + }, + { + "epoch": 1.93, + "learning_rate": 6.002726245650142e-06, + "loss": 1.1887, + "step": 22874 + }, + { + "epoch": 1.93, + "learning_rate": 6.001893066495315e-06, + "loss": 1.1228, + "step": 22875 + }, + { + "epoch": 1.93, + "learning_rate": 6.0010599203741305e-06, + "loss": 1.1688, + "step": 22876 + }, + { + "epoch": 1.93, + "learning_rate": 6.000226807293463e-06, + "loss": 1.1909, + "step": 22877 + }, + { + "epoch": 1.93, + "learning_rate": 5.999393727260206e-06, + "loss": 1.3081, + "step": 22878 + }, + { + "epoch": 1.93, + "learning_rate": 5.998560680281238e-06, + "loss": 1.1462, + "step": 22879 + }, + { + "epoch": 1.93, + "learning_rate": 5.997727666363442e-06, + "loss": 1.1982, + "step": 22880 + }, + { + "epoch": 1.93, + "learning_rate": 5.996894685513699e-06, + "loss": 1.2226, + "step": 22881 + }, + { + "epoch": 1.93, + "learning_rate": 5.996061737738895e-06, + "loss": 1.1894, + "step": 22882 + }, + { + "epoch": 1.93, + "learning_rate": 5.99522882304591e-06, + "loss": 1.1, + "step": 22883 + }, + { + "epoch": 1.93, + "learning_rate": 5.994395941441628e-06, + "loss": 1.1203, + "step": 22884 + }, + { + "epoch": 1.93, + "learning_rate": 5.993563092932924e-06, + "loss": 1.1758, + "step": 22885 + }, + { + "epoch": 1.93, + "learning_rate": 5.992730277526686e-06, + "loss": 1.1292, + "step": 22886 + }, + { + "epoch": 1.93, + "learning_rate": 5.991897495229795e-06, + "loss": 1.1832, + "step": 22887 + }, + { + "epoch": 1.93, + "learning_rate": 5.991064746049127e-06, + "loss": 1.1426, + "step": 22888 + }, + { + "epoch": 1.93, + "learning_rate": 5.990232029991567e-06, + "loss": 1.103, + "step": 22889 + }, + { + "epoch": 1.93, + "learning_rate": 5.9893993470639925e-06, + "loss": 1.1147, + "step": 22890 + }, + { + "epoch": 1.93, + "learning_rate": 5.988566697273286e-06, + "loss": 1.168, + "step": 22891 + }, + { + "epoch": 1.93, + "learning_rate": 5.987734080626322e-06, + "loss": 1.1304, + "step": 22892 + }, + { + "epoch": 1.93, + "learning_rate": 5.986901497129986e-06, + "loss": 1.1638, + "step": 22893 + }, + { + "epoch": 1.93, + "learning_rate": 5.986068946791152e-06, + "loss": 1.1786, + "step": 22894 + }, + { + "epoch": 1.93, + "learning_rate": 5.985236429616703e-06, + "loss": 1.1828, + "step": 22895 + }, + { + "epoch": 1.93, + "learning_rate": 5.984403945613513e-06, + "loss": 1.2962, + "step": 22896 + }, + { + "epoch": 1.93, + "learning_rate": 5.983571494788465e-06, + "loss": 1.2617, + "step": 22897 + }, + { + "epoch": 1.93, + "learning_rate": 5.982739077148434e-06, + "loss": 1.075, + "step": 22898 + }, + { + "epoch": 1.93, + "learning_rate": 5.981906692700294e-06, + "loss": 1.1467, + "step": 22899 + }, + { + "epoch": 1.93, + "learning_rate": 5.981074341450934e-06, + "loss": 1.2146, + "step": 22900 + }, + { + "epoch": 1.93, + "learning_rate": 5.980242023407221e-06, + "loss": 1.1645, + "step": 22901 + }, + { + "epoch": 1.93, + "learning_rate": 5.979409738576036e-06, + "loss": 1.1822, + "step": 22902 + }, + { + "epoch": 1.93, + "learning_rate": 5.978577486964252e-06, + "loss": 1.1534, + "step": 22903 + }, + { + "epoch": 1.93, + "learning_rate": 5.977745268578752e-06, + "loss": 1.1991, + "step": 22904 + }, + { + "epoch": 1.93, + "learning_rate": 5.976913083426404e-06, + "loss": 1.1738, + "step": 22905 + }, + { + "epoch": 1.93, + "learning_rate": 5.97608093151409e-06, + "loss": 1.1495, + "step": 22906 + }, + { + "epoch": 1.93, + "learning_rate": 5.975248812848682e-06, + "loss": 1.1179, + "step": 22907 + }, + { + "epoch": 1.93, + "learning_rate": 5.974416727437058e-06, + "loss": 1.2027, + "step": 22908 + }, + { + "epoch": 1.93, + "learning_rate": 5.973584675286086e-06, + "loss": 1.1947, + "step": 22909 + }, + { + "epoch": 1.93, + "learning_rate": 5.9727526564026515e-06, + "loss": 1.1668, + "step": 22910 + }, + { + "epoch": 1.93, + "learning_rate": 5.971920670793623e-06, + "loss": 1.2182, + "step": 22911 + }, + { + "epoch": 1.93, + "learning_rate": 5.971088718465871e-06, + "loss": 1.1521, + "step": 22912 + }, + { + "epoch": 1.93, + "learning_rate": 5.970256799426277e-06, + "loss": 1.1752, + "step": 22913 + }, + { + "epoch": 1.93, + "learning_rate": 5.96942491368171e-06, + "loss": 1.1882, + "step": 22914 + }, + { + "epoch": 1.93, + "learning_rate": 5.968593061239046e-06, + "loss": 1.1459, + "step": 22915 + }, + { + "epoch": 1.93, + "learning_rate": 5.967761242105152e-06, + "loss": 1.1949, + "step": 22916 + }, + { + "epoch": 1.93, + "learning_rate": 5.966929456286911e-06, + "loss": 1.2093, + "step": 22917 + }, + { + "epoch": 1.93, + "learning_rate": 5.966097703791185e-06, + "loss": 1.1132, + "step": 22918 + }, + { + "epoch": 1.93, + "learning_rate": 5.965265984624855e-06, + "loss": 1.0727, + "step": 22919 + }, + { + "epoch": 1.93, + "learning_rate": 5.9644342987947834e-06, + "loss": 1.1745, + "step": 22920 + }, + { + "epoch": 1.93, + "learning_rate": 5.963602646307851e-06, + "loss": 1.1625, + "step": 22921 + }, + { + "epoch": 1.93, + "learning_rate": 5.9627710271709236e-06, + "loss": 1.2135, + "step": 22922 + }, + { + "epoch": 1.93, + "learning_rate": 5.961939441390877e-06, + "loss": 1.1875, + "step": 22923 + }, + { + "epoch": 1.93, + "learning_rate": 5.961107888974574e-06, + "loss": 1.1411, + "step": 22924 + }, + { + "epoch": 1.93, + "learning_rate": 5.960276369928892e-06, + "loss": 1.1527, + "step": 22925 + }, + { + "epoch": 1.93, + "learning_rate": 5.959444884260703e-06, + "loss": 1.1243, + "step": 22926 + }, + { + "epoch": 1.93, + "learning_rate": 5.958613431976869e-06, + "loss": 1.189, + "step": 22927 + }, + { + "epoch": 1.93, + "learning_rate": 5.957782013084268e-06, + "loss": 1.2465, + "step": 22928 + }, + { + "epoch": 1.93, + "learning_rate": 5.956950627589764e-06, + "loss": 1.1867, + "step": 22929 + }, + { + "epoch": 1.93, + "learning_rate": 5.956119275500228e-06, + "loss": 1.1402, + "step": 22930 + }, + { + "epoch": 1.93, + "learning_rate": 5.955287956822526e-06, + "loss": 1.0954, + "step": 22931 + }, + { + "epoch": 1.93, + "learning_rate": 5.954456671563533e-06, + "loss": 1.2297, + "step": 22932 + }, + { + "epoch": 1.93, + "learning_rate": 5.953625419730114e-06, + "loss": 1.2175, + "step": 22933 + }, + { + "epoch": 1.93, + "learning_rate": 5.952794201329136e-06, + "loss": 1.1693, + "step": 22934 + }, + { + "epoch": 1.93, + "learning_rate": 5.9519630163674645e-06, + "loss": 1.178, + "step": 22935 + }, + { + "epoch": 1.93, + "learning_rate": 5.951131864851972e-06, + "loss": 1.2162, + "step": 22936 + }, + { + "epoch": 1.93, + "learning_rate": 5.950300746789526e-06, + "loss": 1.1389, + "step": 22937 + }, + { + "epoch": 1.93, + "learning_rate": 5.949469662186988e-06, + "loss": 1.0676, + "step": 22938 + }, + { + "epoch": 1.93, + "learning_rate": 5.948638611051232e-06, + "loss": 1.1723, + "step": 22939 + }, + { + "epoch": 1.93, + "learning_rate": 5.947807593389117e-06, + "loss": 1.2242, + "step": 22940 + }, + { + "epoch": 1.93, + "learning_rate": 5.946976609207516e-06, + "loss": 1.1251, + "step": 22941 + }, + { + "epoch": 1.93, + "learning_rate": 5.946145658513285e-06, + "loss": 1.2277, + "step": 22942 + }, + { + "epoch": 1.93, + "learning_rate": 5.945314741313302e-06, + "loss": 1.119, + "step": 22943 + }, + { + "epoch": 1.93, + "learning_rate": 5.944483857614425e-06, + "loss": 1.2023, + "step": 22944 + }, + { + "epoch": 1.93, + "learning_rate": 5.94365300742352e-06, + "loss": 1.1668, + "step": 22945 + }, + { + "epoch": 1.93, + "learning_rate": 5.942822190747449e-06, + "loss": 1.1488, + "step": 22946 + }, + { + "epoch": 1.93, + "learning_rate": 5.941991407593084e-06, + "loss": 1.0882, + "step": 22947 + }, + { + "epoch": 1.93, + "learning_rate": 5.941160657967282e-06, + "loss": 1.1972, + "step": 22948 + }, + { + "epoch": 1.93, + "learning_rate": 5.940329941876909e-06, + "loss": 1.3123, + "step": 22949 + }, + { + "epoch": 1.93, + "learning_rate": 5.939499259328831e-06, + "loss": 1.1822, + "step": 22950 + }, + { + "epoch": 1.93, + "learning_rate": 5.93866861032991e-06, + "loss": 1.1685, + "step": 22951 + }, + { + "epoch": 1.93, + "learning_rate": 5.93783799488701e-06, + "loss": 1.1319, + "step": 22952 + }, + { + "epoch": 1.93, + "learning_rate": 5.937007413006988e-06, + "loss": 1.1927, + "step": 22953 + }, + { + "epoch": 1.93, + "learning_rate": 5.936176864696718e-06, + "loss": 1.14, + "step": 22954 + }, + { + "epoch": 1.93, + "learning_rate": 5.9353463499630486e-06, + "loss": 1.2122, + "step": 22955 + }, + { + "epoch": 1.93, + "learning_rate": 5.934515868812853e-06, + "loss": 1.1908, + "step": 22956 + }, + { + "epoch": 1.93, + "learning_rate": 5.933685421252986e-06, + "loss": 1.2009, + "step": 22957 + }, + { + "epoch": 1.93, + "learning_rate": 5.9328550072903124e-06, + "loss": 1.0958, + "step": 22958 + }, + { + "epoch": 1.93, + "learning_rate": 5.932024626931689e-06, + "loss": 1.3059, + "step": 22959 + }, + { + "epoch": 1.93, + "learning_rate": 5.931194280183985e-06, + "loss": 1.2685, + "step": 22960 + }, + { + "epoch": 1.93, + "learning_rate": 5.930363967054054e-06, + "loss": 1.2132, + "step": 22961 + }, + { + "epoch": 1.93, + "learning_rate": 5.929533687548757e-06, + "loss": 1.1983, + "step": 22962 + }, + { + "epoch": 1.93, + "learning_rate": 5.928703441674957e-06, + "loss": 1.1776, + "step": 22963 + }, + { + "epoch": 1.93, + "learning_rate": 5.927873229439512e-06, + "loss": 1.1444, + "step": 22964 + }, + { + "epoch": 1.93, + "learning_rate": 5.927043050849283e-06, + "loss": 1.1295, + "step": 22965 + }, + { + "epoch": 1.93, + "learning_rate": 5.926212905911124e-06, + "loss": 1.132, + "step": 22966 + }, + { + "epoch": 1.93, + "learning_rate": 5.925382794631902e-06, + "loss": 1.227, + "step": 22967 + }, + { + "epoch": 1.93, + "learning_rate": 5.924552717018468e-06, + "loss": 1.2888, + "step": 22968 + }, + { + "epoch": 1.93, + "learning_rate": 5.9237226730776874e-06, + "loss": 1.2303, + "step": 22969 + }, + { + "epoch": 1.93, + "learning_rate": 5.92289266281641e-06, + "loss": 1.1274, + "step": 22970 + }, + { + "epoch": 1.93, + "learning_rate": 5.922062686241502e-06, + "loss": 1.1563, + "step": 22971 + }, + { + "epoch": 1.93, + "learning_rate": 5.9212327433598145e-06, + "loss": 1.1233, + "step": 22972 + }, + { + "epoch": 1.93, + "learning_rate": 5.9204028341782115e-06, + "loss": 1.1649, + "step": 22973 + }, + { + "epoch": 1.93, + "learning_rate": 5.919572958703541e-06, + "loss": 1.1822, + "step": 22974 + }, + { + "epoch": 1.93, + "learning_rate": 5.918743116942667e-06, + "loss": 1.1582, + "step": 22975 + }, + { + "epoch": 1.93, + "learning_rate": 5.917913308902445e-06, + "loss": 1.2063, + "step": 22976 + }, + { + "epoch": 1.93, + "learning_rate": 5.917083534589725e-06, + "loss": 1.2378, + "step": 22977 + }, + { + "epoch": 1.93, + "learning_rate": 5.916253794011373e-06, + "loss": 1.222, + "step": 22978 + }, + { + "epoch": 1.93, + "learning_rate": 5.9154240871742355e-06, + "loss": 1.2057, + "step": 22979 + }, + { + "epoch": 1.93, + "learning_rate": 5.9145944140851755e-06, + "loss": 1.1172, + "step": 22980 + }, + { + "epoch": 1.93, + "learning_rate": 5.913764774751038e-06, + "loss": 1.1938, + "step": 22981 + }, + { + "epoch": 1.93, + "learning_rate": 5.912935169178688e-06, + "loss": 1.1542, + "step": 22982 + }, + { + "epoch": 1.93, + "learning_rate": 5.912105597374975e-06, + "loss": 1.2076, + "step": 22983 + }, + { + "epoch": 1.94, + "learning_rate": 5.911276059346754e-06, + "loss": 1.1634, + "step": 22984 + }, + { + "epoch": 1.94, + "learning_rate": 5.910446555100876e-06, + "loss": 1.1995, + "step": 22985 + }, + { + "epoch": 1.94, + "learning_rate": 5.909617084644199e-06, + "loss": 1.0966, + "step": 22986 + }, + { + "epoch": 1.94, + "learning_rate": 5.908787647983577e-06, + "loss": 1.268, + "step": 22987 + }, + { + "epoch": 1.94, + "learning_rate": 5.9079582451258565e-06, + "loss": 1.1582, + "step": 22988 + }, + { + "epoch": 1.94, + "learning_rate": 5.907128876077899e-06, + "loss": 1.2183, + "step": 22989 + }, + { + "epoch": 1.94, + "learning_rate": 5.906299540846549e-06, + "loss": 1.1138, + "step": 22990 + }, + { + "epoch": 1.94, + "learning_rate": 5.905470239438666e-06, + "loss": 1.1284, + "step": 22991 + }, + { + "epoch": 1.94, + "learning_rate": 5.904640971861095e-06, + "loss": 1.215, + "step": 22992 + }, + { + "epoch": 1.94, + "learning_rate": 5.903811738120694e-06, + "loss": 1.2494, + "step": 22993 + }, + { + "epoch": 1.94, + "learning_rate": 5.902982538224309e-06, + "loss": 1.2147, + "step": 22994 + }, + { + "epoch": 1.94, + "learning_rate": 5.902153372178796e-06, + "loss": 1.1559, + "step": 22995 + }, + { + "epoch": 1.94, + "learning_rate": 5.901324239991e-06, + "loss": 1.2493, + "step": 22996 + }, + { + "epoch": 1.94, + "learning_rate": 5.900495141667777e-06, + "loss": 1.2438, + "step": 22997 + }, + { + "epoch": 1.94, + "learning_rate": 5.899666077215975e-06, + "loss": 1.2558, + "step": 22998 + }, + { + "epoch": 1.94, + "learning_rate": 5.898837046642441e-06, + "loss": 1.1928, + "step": 22999 + }, + { + "epoch": 1.94, + "learning_rate": 5.898008049954032e-06, + "loss": 1.0889, + "step": 23000 + }, + { + "epoch": 1.94, + "learning_rate": 5.8971790871575914e-06, + "loss": 1.1576, + "step": 23001 + }, + { + "epoch": 1.94, + "learning_rate": 5.896350158259973e-06, + "loss": 1.127, + "step": 23002 + }, + { + "epoch": 1.94, + "learning_rate": 5.8955212632680185e-06, + "loss": 1.1028, + "step": 23003 + }, + { + "epoch": 1.94, + "learning_rate": 5.894692402188584e-06, + "loss": 1.1216, + "step": 23004 + }, + { + "epoch": 1.94, + "learning_rate": 5.893863575028513e-06, + "loss": 1.1977, + "step": 23005 + }, + { + "epoch": 1.94, + "learning_rate": 5.8930347817946575e-06, + "loss": 1.1395, + "step": 23006 + }, + { + "epoch": 1.94, + "learning_rate": 5.892206022493859e-06, + "loss": 1.1099, + "step": 23007 + }, + { + "epoch": 1.94, + "learning_rate": 5.891377297132977e-06, + "loss": 1.1457, + "step": 23008 + }, + { + "epoch": 1.94, + "learning_rate": 5.890548605718842e-06, + "loss": 1.177, + "step": 23009 + }, + { + "epoch": 1.94, + "learning_rate": 5.889719948258314e-06, + "loss": 1.1851, + "step": 23010 + }, + { + "epoch": 1.94, + "learning_rate": 5.888891324758235e-06, + "loss": 1.273, + "step": 23011 + }, + { + "epoch": 1.94, + "learning_rate": 5.888062735225447e-06, + "loss": 1.1962, + "step": 23012 + }, + { + "epoch": 1.94, + "learning_rate": 5.887234179666806e-06, + "loss": 1.1876, + "step": 23013 + }, + { + "epoch": 1.94, + "learning_rate": 5.886405658089151e-06, + "loss": 1.3607, + "step": 23014 + }, + { + "epoch": 1.94, + "learning_rate": 5.88557717049933e-06, + "loss": 1.2036, + "step": 23015 + }, + { + "epoch": 1.94, + "learning_rate": 5.884748716904184e-06, + "loss": 1.1863, + "step": 23016 + }, + { + "epoch": 1.94, + "learning_rate": 5.8839202973105646e-06, + "loss": 1.1685, + "step": 23017 + }, + { + "epoch": 1.94, + "learning_rate": 5.883091911725311e-06, + "loss": 1.1065, + "step": 23018 + }, + { + "epoch": 1.94, + "learning_rate": 5.8822635601552725e-06, + "loss": 1.227, + "step": 23019 + }, + { + "epoch": 1.94, + "learning_rate": 5.881435242607285e-06, + "loss": 1.2451, + "step": 23020 + }, + { + "epoch": 1.94, + "learning_rate": 5.880606959088202e-06, + "loss": 1.0803, + "step": 23021 + }, + { + "epoch": 1.94, + "learning_rate": 5.879778709604862e-06, + "loss": 1.2692, + "step": 23022 + }, + { + "epoch": 1.94, + "learning_rate": 5.878950494164109e-06, + "loss": 1.1756, + "step": 23023 + }, + { + "epoch": 1.94, + "learning_rate": 5.878122312772784e-06, + "loss": 1.2466, + "step": 23024 + }, + { + "epoch": 1.94, + "learning_rate": 5.877294165437733e-06, + "loss": 1.2111, + "step": 23025 + }, + { + "epoch": 1.94, + "learning_rate": 5.876466052165797e-06, + "loss": 1.1653, + "step": 23026 + }, + { + "epoch": 1.94, + "learning_rate": 5.875637972963817e-06, + "loss": 1.0911, + "step": 23027 + }, + { + "epoch": 1.94, + "learning_rate": 5.874809927838638e-06, + "loss": 1.1683, + "step": 23028 + }, + { + "epoch": 1.94, + "learning_rate": 5.873981916797098e-06, + "loss": 1.2462, + "step": 23029 + }, + { + "epoch": 1.94, + "learning_rate": 5.873153939846043e-06, + "loss": 1.2099, + "step": 23030 + }, + { + "epoch": 1.94, + "learning_rate": 5.872325996992305e-06, + "loss": 1.2595, + "step": 23031 + }, + { + "epoch": 1.94, + "learning_rate": 5.871498088242736e-06, + "loss": 1.1939, + "step": 23032 + }, + { + "epoch": 1.94, + "learning_rate": 5.870670213604169e-06, + "loss": 1.1463, + "step": 23033 + }, + { + "epoch": 1.94, + "learning_rate": 5.8698423730834475e-06, + "loss": 1.2448, + "step": 23034 + }, + { + "epoch": 1.94, + "learning_rate": 5.869014566687408e-06, + "loss": 1.1117, + "step": 23035 + }, + { + "epoch": 1.94, + "learning_rate": 5.868186794422892e-06, + "loss": 1.1775, + "step": 23036 + }, + { + "epoch": 1.94, + "learning_rate": 5.867359056296742e-06, + "loss": 1.1316, + "step": 23037 + }, + { + "epoch": 1.94, + "learning_rate": 5.866531352315791e-06, + "loss": 1.2165, + "step": 23038 + }, + { + "epoch": 1.94, + "learning_rate": 5.865703682486883e-06, + "loss": 1.1454, + "step": 23039 + }, + { + "epoch": 1.94, + "learning_rate": 5.864876046816854e-06, + "loss": 1.2714, + "step": 23040 + }, + { + "epoch": 1.94, + "learning_rate": 5.864048445312544e-06, + "loss": 1.2234, + "step": 23041 + }, + { + "epoch": 1.94, + "learning_rate": 5.863220877980784e-06, + "loss": 1.172, + "step": 23042 + }, + { + "epoch": 1.94, + "learning_rate": 5.862393344828423e-06, + "loss": 1.1946, + "step": 23043 + }, + { + "epoch": 1.94, + "learning_rate": 5.861565845862289e-06, + "loss": 1.2503, + "step": 23044 + }, + { + "epoch": 1.94, + "learning_rate": 5.860738381089226e-06, + "loss": 1.2062, + "step": 23045 + }, + { + "epoch": 1.94, + "learning_rate": 5.859910950516062e-06, + "loss": 1.2118, + "step": 23046 + }, + { + "epoch": 1.94, + "learning_rate": 5.8590835541496426e-06, + "loss": 1.1919, + "step": 23047 + }, + { + "epoch": 1.94, + "learning_rate": 5.858256191996798e-06, + "loss": 1.2017, + "step": 23048 + }, + { + "epoch": 1.94, + "learning_rate": 5.857428864064365e-06, + "loss": 1.1342, + "step": 23049 + }, + { + "epoch": 1.94, + "learning_rate": 5.856601570359185e-06, + "loss": 1.175, + "step": 23050 + }, + { + "epoch": 1.94, + "learning_rate": 5.855774310888087e-06, + "loss": 1.1807, + "step": 23051 + }, + { + "epoch": 1.94, + "learning_rate": 5.854947085657909e-06, + "loss": 1.1922, + "step": 23052 + }, + { + "epoch": 1.94, + "learning_rate": 5.854119894675481e-06, + "loss": 1.1543, + "step": 23053 + }, + { + "epoch": 1.94, + "learning_rate": 5.853292737947644e-06, + "loss": 1.2277, + "step": 23054 + }, + { + "epoch": 1.94, + "learning_rate": 5.85246561548123e-06, + "loss": 1.232, + "step": 23055 + }, + { + "epoch": 1.94, + "learning_rate": 5.851638527283072e-06, + "loss": 1.1876, + "step": 23056 + }, + { + "epoch": 1.94, + "learning_rate": 5.850811473360003e-06, + "loss": 1.1582, + "step": 23057 + }, + { + "epoch": 1.94, + "learning_rate": 5.849984453718861e-06, + "loss": 1.1609, + "step": 23058 + }, + { + "epoch": 1.94, + "learning_rate": 5.849157468366471e-06, + "loss": 1.1467, + "step": 23059 + }, + { + "epoch": 1.94, + "learning_rate": 5.848330517309674e-06, + "loss": 1.2028, + "step": 23060 + }, + { + "epoch": 1.94, + "learning_rate": 5.847503600555297e-06, + "loss": 1.1622, + "step": 23061 + }, + { + "epoch": 1.94, + "learning_rate": 5.846676718110171e-06, + "loss": 1.1274, + "step": 23062 + }, + { + "epoch": 1.94, + "learning_rate": 5.845849869981137e-06, + "loss": 1.1817, + "step": 23063 + }, + { + "epoch": 1.94, + "learning_rate": 5.845023056175019e-06, + "loss": 1.0384, + "step": 23064 + }, + { + "epoch": 1.94, + "learning_rate": 5.84419627669865e-06, + "loss": 1.058, + "step": 23065 + }, + { + "epoch": 1.94, + "learning_rate": 5.84336953155886e-06, + "loss": 1.2373, + "step": 23066 + }, + { + "epoch": 1.94, + "learning_rate": 5.842542820762485e-06, + "loss": 1.2326, + "step": 23067 + }, + { + "epoch": 1.94, + "learning_rate": 5.841716144316347e-06, + "loss": 1.206, + "step": 23068 + }, + { + "epoch": 1.94, + "learning_rate": 5.840889502227285e-06, + "loss": 1.2118, + "step": 23069 + }, + { + "epoch": 1.94, + "learning_rate": 5.840062894502121e-06, + "loss": 1.1275, + "step": 23070 + }, + { + "epoch": 1.94, + "learning_rate": 5.839236321147694e-06, + "loss": 1.1175, + "step": 23071 + }, + { + "epoch": 1.94, + "learning_rate": 5.838409782170822e-06, + "loss": 1.2098, + "step": 23072 + }, + { + "epoch": 1.94, + "learning_rate": 5.837583277578347e-06, + "loss": 1.1954, + "step": 23073 + }, + { + "epoch": 1.94, + "learning_rate": 5.836756807377087e-06, + "loss": 1.2799, + "step": 23074 + }, + { + "epoch": 1.94, + "learning_rate": 5.83593037157387e-06, + "loss": 1.1026, + "step": 23075 + }, + { + "epoch": 1.94, + "learning_rate": 5.835103970175538e-06, + "loss": 1.1692, + "step": 23076 + }, + { + "epoch": 1.94, + "learning_rate": 5.834277603188906e-06, + "loss": 1.2752, + "step": 23077 + }, + { + "epoch": 1.94, + "learning_rate": 5.833451270620805e-06, + "loss": 1.1583, + "step": 23078 + }, + { + "epoch": 1.94, + "learning_rate": 5.832624972478064e-06, + "loss": 1.2631, + "step": 23079 + }, + { + "epoch": 1.94, + "learning_rate": 5.831798708767509e-06, + "loss": 1.1914, + "step": 23080 + }, + { + "epoch": 1.94, + "learning_rate": 5.830972479495967e-06, + "loss": 1.1919, + "step": 23081 + }, + { + "epoch": 1.94, + "learning_rate": 5.830146284670267e-06, + "loss": 1.1718, + "step": 23082 + }, + { + "epoch": 1.94, + "learning_rate": 5.829320124297226e-06, + "loss": 1.1968, + "step": 23083 + }, + { + "epoch": 1.94, + "learning_rate": 5.828493998383686e-06, + "loss": 1.3044, + "step": 23084 + }, + { + "epoch": 1.94, + "learning_rate": 5.827667906936458e-06, + "loss": 1.1449, + "step": 23085 + }, + { + "epoch": 1.94, + "learning_rate": 5.82684184996237e-06, + "loss": 1.1829, + "step": 23086 + }, + { + "epoch": 1.94, + "learning_rate": 5.826015827468259e-06, + "loss": 1.1165, + "step": 23087 + }, + { + "epoch": 1.94, + "learning_rate": 5.8251898394609365e-06, + "loss": 1.1407, + "step": 23088 + }, + { + "epoch": 1.94, + "learning_rate": 5.824363885947231e-06, + "loss": 1.1504, + "step": 23089 + }, + { + "epoch": 1.94, + "learning_rate": 5.8235379669339674e-06, + "loss": 1.1684, + "step": 23090 + }, + { + "epoch": 1.94, + "learning_rate": 5.822712082427971e-06, + "loss": 1.2331, + "step": 23091 + }, + { + "epoch": 1.94, + "learning_rate": 5.821886232436063e-06, + "loss": 1.1791, + "step": 23092 + }, + { + "epoch": 1.94, + "learning_rate": 5.821060416965072e-06, + "loss": 1.2532, + "step": 23093 + }, + { + "epoch": 1.94, + "learning_rate": 5.820234636021809e-06, + "loss": 1.1573, + "step": 23094 + }, + { + "epoch": 1.94, + "learning_rate": 5.819408889613114e-06, + "loss": 1.1766, + "step": 23095 + }, + { + "epoch": 1.94, + "learning_rate": 5.818583177745797e-06, + "loss": 1.1338, + "step": 23096 + }, + { + "epoch": 1.94, + "learning_rate": 5.817757500426684e-06, + "loss": 1.1191, + "step": 23097 + }, + { + "epoch": 1.94, + "learning_rate": 5.816931857662596e-06, + "loss": 1.1609, + "step": 23098 + }, + { + "epoch": 1.94, + "learning_rate": 5.816106249460356e-06, + "loss": 1.1324, + "step": 23099 + }, + { + "epoch": 1.94, + "learning_rate": 5.815280675826787e-06, + "loss": 1.1623, + "step": 23100 + }, + { + "epoch": 1.94, + "learning_rate": 5.814455136768706e-06, + "loss": 1.2305, + "step": 23101 + }, + { + "epoch": 1.94, + "learning_rate": 5.813629632292937e-06, + "loss": 1.1803, + "step": 23102 + }, + { + "epoch": 1.95, + "learning_rate": 5.812804162406301e-06, + "loss": 1.2261, + "step": 23103 + }, + { + "epoch": 1.95, + "learning_rate": 5.811978727115619e-06, + "loss": 1.1766, + "step": 23104 + }, + { + "epoch": 1.95, + "learning_rate": 5.811153326427702e-06, + "loss": 1.1111, + "step": 23105 + }, + { + "epoch": 1.95, + "learning_rate": 5.810327960349385e-06, + "loss": 1.1845, + "step": 23106 + }, + { + "epoch": 1.95, + "learning_rate": 5.809502628887476e-06, + "loss": 1.3098, + "step": 23107 + }, + { + "epoch": 1.95, + "learning_rate": 5.808677332048795e-06, + "loss": 1.2202, + "step": 23108 + }, + { + "epoch": 1.95, + "learning_rate": 5.807852069840166e-06, + "loss": 1.1195, + "step": 23109 + }, + { + "epoch": 1.95, + "learning_rate": 5.807026842268407e-06, + "loss": 1.2011, + "step": 23110 + }, + { + "epoch": 1.95, + "learning_rate": 5.806201649340328e-06, + "loss": 1.2066, + "step": 23111 + }, + { + "epoch": 1.95, + "learning_rate": 5.805376491062755e-06, + "loss": 1.1503, + "step": 23112 + }, + { + "epoch": 1.95, + "learning_rate": 5.804551367442509e-06, + "loss": 1.1542, + "step": 23113 + }, + { + "epoch": 1.95, + "learning_rate": 5.803726278486395e-06, + "loss": 1.1063, + "step": 23114 + }, + { + "epoch": 1.95, + "learning_rate": 5.802901224201244e-06, + "loss": 1.1513, + "step": 23115 + }, + { + "epoch": 1.95, + "learning_rate": 5.802076204593863e-06, + "loss": 1.1504, + "step": 23116 + }, + { + "epoch": 1.95, + "learning_rate": 5.801251219671071e-06, + "loss": 1.2762, + "step": 23117 + }, + { + "epoch": 1.95, + "learning_rate": 5.800426269439687e-06, + "loss": 1.2283, + "step": 23118 + }, + { + "epoch": 1.95, + "learning_rate": 5.799601353906524e-06, + "loss": 1.1087, + "step": 23119 + }, + { + "epoch": 1.95, + "learning_rate": 5.7987764730784005e-06, + "loss": 1.2076, + "step": 23120 + }, + { + "epoch": 1.95, + "learning_rate": 5.797951626962133e-06, + "loss": 1.2273, + "step": 23121 + }, + { + "epoch": 1.95, + "learning_rate": 5.797126815564524e-06, + "loss": 1.3057, + "step": 23122 + }, + { + "epoch": 1.95, + "learning_rate": 5.796302038892409e-06, + "loss": 1.202, + "step": 23123 + }, + { + "epoch": 1.95, + "learning_rate": 5.795477296952587e-06, + "loss": 1.0911, + "step": 23124 + }, + { + "epoch": 1.95, + "learning_rate": 5.794652589751874e-06, + "loss": 1.1299, + "step": 23125 + }, + { + "epoch": 1.95, + "learning_rate": 5.793827917297095e-06, + "loss": 1.2404, + "step": 23126 + }, + { + "epoch": 1.95, + "learning_rate": 5.793003279595051e-06, + "loss": 1.1652, + "step": 23127 + }, + { + "epoch": 1.95, + "learning_rate": 5.792178676652561e-06, + "loss": 1.0683, + "step": 23128 + }, + { + "epoch": 1.95, + "learning_rate": 5.791354108476438e-06, + "loss": 1.1833, + "step": 23129 + }, + { + "epoch": 1.95, + "learning_rate": 5.790529575073494e-06, + "loss": 1.1157, + "step": 23130 + }, + { + "epoch": 1.95, + "learning_rate": 5.78970507645054e-06, + "loss": 1.2144, + "step": 23131 + }, + { + "epoch": 1.95, + "learning_rate": 5.788880612614396e-06, + "loss": 1.0332, + "step": 23132 + }, + { + "epoch": 1.95, + "learning_rate": 5.7880561835718596e-06, + "loss": 1.1343, + "step": 23133 + }, + { + "epoch": 1.95, + "learning_rate": 5.7872317893297595e-06, + "loss": 1.2245, + "step": 23134 + }, + { + "epoch": 1.95, + "learning_rate": 5.7864074298948945e-06, + "loss": 1.2119, + "step": 23135 + }, + { + "epoch": 1.95, + "learning_rate": 5.7855831052740755e-06, + "loss": 1.1277, + "step": 23136 + }, + { + "epoch": 1.95, + "learning_rate": 5.784758815474125e-06, + "loss": 1.18, + "step": 23137 + }, + { + "epoch": 1.95, + "learning_rate": 5.783934560501844e-06, + "loss": 1.2086, + "step": 23138 + }, + { + "epoch": 1.95, + "learning_rate": 5.783110340364044e-06, + "loss": 1.0318, + "step": 23139 + }, + { + "epoch": 1.95, + "learning_rate": 5.782286155067536e-06, + "loss": 1.1513, + "step": 23140 + }, + { + "epoch": 1.95, + "learning_rate": 5.78146200461913e-06, + "loss": 1.1588, + "step": 23141 + }, + { + "epoch": 1.95, + "learning_rate": 5.780637889025633e-06, + "loss": 1.0882, + "step": 23142 + }, + { + "epoch": 1.95, + "learning_rate": 5.779813808293862e-06, + "loss": 1.1125, + "step": 23143 + }, + { + "epoch": 1.95, + "learning_rate": 5.778989762430611e-06, + "loss": 1.1742, + "step": 23144 + }, + { + "epoch": 1.95, + "learning_rate": 5.778165751442707e-06, + "loss": 1.175, + "step": 23145 + }, + { + "epoch": 1.95, + "learning_rate": 5.777341775336944e-06, + "loss": 1.1241, + "step": 23146 + }, + { + "epoch": 1.95, + "learning_rate": 5.7765178341201345e-06, + "loss": 1.1486, + "step": 23147 + }, + { + "epoch": 1.95, + "learning_rate": 5.775693927799087e-06, + "loss": 1.1667, + "step": 23148 + }, + { + "epoch": 1.95, + "learning_rate": 5.774870056380607e-06, + "loss": 1.2568, + "step": 23149 + }, + { + "epoch": 1.95, + "learning_rate": 5.774046219871503e-06, + "loss": 1.0398, + "step": 23150 + }, + { + "epoch": 1.95, + "learning_rate": 5.773222418278582e-06, + "loss": 1.1536, + "step": 23151 + }, + { + "epoch": 1.95, + "learning_rate": 5.77239865160865e-06, + "loss": 1.1349, + "step": 23152 + }, + { + "epoch": 1.95, + "learning_rate": 5.771574919868515e-06, + "loss": 1.1309, + "step": 23153 + }, + { + "epoch": 1.95, + "learning_rate": 5.770751223064983e-06, + "loss": 1.2597, + "step": 23154 + }, + { + "epoch": 1.95, + "learning_rate": 5.76992756120485e-06, + "loss": 1.0799, + "step": 23155 + }, + { + "epoch": 1.95, + "learning_rate": 5.769103934294937e-06, + "loss": 1.1474, + "step": 23156 + }, + { + "epoch": 1.95, + "learning_rate": 5.768280342342038e-06, + "loss": 1.1454, + "step": 23157 + }, + { + "epoch": 1.95, + "learning_rate": 5.7674567853529625e-06, + "loss": 1.2728, + "step": 23158 + }, + { + "epoch": 1.95, + "learning_rate": 5.766633263334514e-06, + "loss": 1.0313, + "step": 23159 + }, + { + "epoch": 1.95, + "learning_rate": 5.765809776293498e-06, + "loss": 1.0994, + "step": 23160 + }, + { + "epoch": 1.95, + "learning_rate": 5.76498632423671e-06, + "loss": 1.1492, + "step": 23161 + }, + { + "epoch": 1.95, + "learning_rate": 5.764162907170963e-06, + "loss": 1.1458, + "step": 23162 + }, + { + "epoch": 1.95, + "learning_rate": 5.763339525103063e-06, + "loss": 1.2208, + "step": 23163 + }, + { + "epoch": 1.95, + "learning_rate": 5.7625161780398e-06, + "loss": 1.2358, + "step": 23164 + }, + { + "epoch": 1.95, + "learning_rate": 5.761692865987991e-06, + "loss": 1.2709, + "step": 23165 + }, + { + "epoch": 1.95, + "learning_rate": 5.7608695889544295e-06, + "loss": 1.1518, + "step": 23166 + }, + { + "epoch": 1.95, + "learning_rate": 5.76004634694592e-06, + "loss": 1.2218, + "step": 23167 + }, + { + "epoch": 1.95, + "learning_rate": 5.759223139969263e-06, + "loss": 1.2331, + "step": 23168 + }, + { + "epoch": 1.95, + "learning_rate": 5.758399968031263e-06, + "loss": 1.2204, + "step": 23169 + }, + { + "epoch": 1.95, + "learning_rate": 5.75757683113872e-06, + "loss": 1.2282, + "step": 23170 + }, + { + "epoch": 1.95, + "learning_rate": 5.756753729298439e-06, + "loss": 1.134, + "step": 23171 + }, + { + "epoch": 1.95, + "learning_rate": 5.755930662517208e-06, + "loss": 1.2263, + "step": 23172 + }, + { + "epoch": 1.95, + "learning_rate": 5.755107630801844e-06, + "loss": 1.1655, + "step": 23173 + }, + { + "epoch": 1.95, + "learning_rate": 5.754284634159136e-06, + "loss": 1.1396, + "step": 23174 + }, + { + "epoch": 1.95, + "learning_rate": 5.753461672595883e-06, + "loss": 1.1622, + "step": 23175 + }, + { + "epoch": 1.95, + "learning_rate": 5.752638746118897e-06, + "loss": 1.2085, + "step": 23176 + }, + { + "epoch": 1.95, + "learning_rate": 5.751815854734965e-06, + "loss": 1.1346, + "step": 23177 + }, + { + "epoch": 1.95, + "learning_rate": 5.750992998450889e-06, + "loss": 1.1788, + "step": 23178 + }, + { + "epoch": 1.95, + "learning_rate": 5.750170177273471e-06, + "loss": 1.1851, + "step": 23179 + }, + { + "epoch": 1.95, + "learning_rate": 5.749347391209504e-06, + "loss": 1.21, + "step": 23180 + }, + { + "epoch": 1.95, + "learning_rate": 5.748524640265791e-06, + "loss": 1.1327, + "step": 23181 + }, + { + "epoch": 1.95, + "learning_rate": 5.747701924449131e-06, + "loss": 1.1331, + "step": 23182 + }, + { + "epoch": 1.95, + "learning_rate": 5.746879243766313e-06, + "loss": 1.1723, + "step": 23183 + }, + { + "epoch": 1.95, + "learning_rate": 5.7460565982241455e-06, + "loss": 1.2291, + "step": 23184 + }, + { + "epoch": 1.95, + "learning_rate": 5.745233987829417e-06, + "loss": 1.3018, + "step": 23185 + }, + { + "epoch": 1.95, + "learning_rate": 5.7444114125889224e-06, + "loss": 1.1484, + "step": 23186 + }, + { + "epoch": 1.95, + "learning_rate": 5.7435888725094715e-06, + "loss": 1.1534, + "step": 23187 + }, + { + "epoch": 1.95, + "learning_rate": 5.742766367597848e-06, + "loss": 1.1327, + "step": 23188 + }, + { + "epoch": 1.95, + "learning_rate": 5.741943897860849e-06, + "loss": 1.1996, + "step": 23189 + }, + { + "epoch": 1.95, + "learning_rate": 5.741121463305273e-06, + "loss": 1.1205, + "step": 23190 + }, + { + "epoch": 1.95, + "learning_rate": 5.740299063937915e-06, + "loss": 1.224, + "step": 23191 + }, + { + "epoch": 1.95, + "learning_rate": 5.73947669976557e-06, + "loss": 1.1129, + "step": 23192 + }, + { + "epoch": 1.95, + "learning_rate": 5.7386543707950345e-06, + "loss": 1.1897, + "step": 23193 + }, + { + "epoch": 1.95, + "learning_rate": 5.737832077033094e-06, + "loss": 1.1006, + "step": 23194 + }, + { + "epoch": 1.95, + "learning_rate": 5.737009818486556e-06, + "loss": 1.2452, + "step": 23195 + }, + { + "epoch": 1.95, + "learning_rate": 5.736187595162204e-06, + "loss": 1.1526, + "step": 23196 + }, + { + "epoch": 1.95, + "learning_rate": 5.735365407066834e-06, + "loss": 1.3023, + "step": 23197 + }, + { + "epoch": 1.95, + "learning_rate": 5.73454325420724e-06, + "loss": 1.1262, + "step": 23198 + }, + { + "epoch": 1.95, + "learning_rate": 5.7337211365902145e-06, + "loss": 1.206, + "step": 23199 + }, + { + "epoch": 1.95, + "learning_rate": 5.732899054222552e-06, + "loss": 1.1873, + "step": 23200 + }, + { + "epoch": 1.95, + "learning_rate": 5.732077007111041e-06, + "loss": 1.1407, + "step": 23201 + }, + { + "epoch": 1.95, + "learning_rate": 5.731254995262476e-06, + "loss": 1.2317, + "step": 23202 + }, + { + "epoch": 1.95, + "learning_rate": 5.730433018683649e-06, + "loss": 1.1506, + "step": 23203 + }, + { + "epoch": 1.95, + "learning_rate": 5.729611077381354e-06, + "loss": 1.1781, + "step": 23204 + }, + { + "epoch": 1.95, + "learning_rate": 5.7287891713623724e-06, + "loss": 1.2066, + "step": 23205 + }, + { + "epoch": 1.95, + "learning_rate": 5.727967300633509e-06, + "loss": 1.1032, + "step": 23206 + }, + { + "epoch": 1.95, + "learning_rate": 5.727145465201542e-06, + "loss": 1.1838, + "step": 23207 + }, + { + "epoch": 1.95, + "learning_rate": 5.726323665073268e-06, + "loss": 1.2581, + "step": 23208 + }, + { + "epoch": 1.95, + "learning_rate": 5.7255019002554755e-06, + "loss": 1.2018, + "step": 23209 + }, + { + "epoch": 1.95, + "learning_rate": 5.724680170754955e-06, + "loss": 1.1145, + "step": 23210 + }, + { + "epoch": 1.95, + "learning_rate": 5.723858476578493e-06, + "loss": 1.2343, + "step": 23211 + }, + { + "epoch": 1.95, + "learning_rate": 5.723036817732882e-06, + "loss": 1.113, + "step": 23212 + }, + { + "epoch": 1.95, + "learning_rate": 5.722215194224913e-06, + "loss": 1.1608, + "step": 23213 + }, + { + "epoch": 1.95, + "learning_rate": 5.721393606061365e-06, + "loss": 1.1674, + "step": 23214 + }, + { + "epoch": 1.95, + "learning_rate": 5.720572053249038e-06, + "loss": 1.1147, + "step": 23215 + }, + { + "epoch": 1.95, + "learning_rate": 5.719750535794712e-06, + "loss": 1.1381, + "step": 23216 + }, + { + "epoch": 1.95, + "learning_rate": 5.718929053705177e-06, + "loss": 1.2374, + "step": 23217 + }, + { + "epoch": 1.95, + "learning_rate": 5.71810760698722e-06, + "loss": 1.1446, + "step": 23218 + }, + { + "epoch": 1.95, + "learning_rate": 5.7172861956476264e-06, + "loss": 1.2878, + "step": 23219 + }, + { + "epoch": 1.95, + "learning_rate": 5.716464819693187e-06, + "loss": 1.1619, + "step": 23220 + }, + { + "epoch": 1.96, + "learning_rate": 5.71564347913069e-06, + "loss": 1.1429, + "step": 23221 + }, + { + "epoch": 1.96, + "learning_rate": 5.71482217396691e-06, + "loss": 1.2449, + "step": 23222 + }, + { + "epoch": 1.96, + "learning_rate": 5.714000904208648e-06, + "loss": 1.1405, + "step": 23223 + }, + { + "epoch": 1.96, + "learning_rate": 5.713179669862677e-06, + "loss": 1.2115, + "step": 23224 + }, + { + "epoch": 1.96, + "learning_rate": 5.712358470935786e-06, + "loss": 1.2053, + "step": 23225 + }, + { + "epoch": 1.96, + "learning_rate": 5.711537307434767e-06, + "loss": 1.1566, + "step": 23226 + }, + { + "epoch": 1.96, + "learning_rate": 5.710716179366398e-06, + "loss": 1.2372, + "step": 23227 + }, + { + "epoch": 1.96, + "learning_rate": 5.709895086737463e-06, + "loss": 1.1782, + "step": 23228 + }, + { + "epoch": 1.96, + "learning_rate": 5.709074029554748e-06, + "loss": 1.1374, + "step": 23229 + }, + { + "epoch": 1.96, + "learning_rate": 5.708253007825035e-06, + "loss": 1.2073, + "step": 23230 + }, + { + "epoch": 1.96, + "learning_rate": 5.707432021555111e-06, + "loss": 1.2059, + "step": 23231 + }, + { + "epoch": 1.96, + "learning_rate": 5.7066110707517615e-06, + "loss": 1.1808, + "step": 23232 + }, + { + "epoch": 1.96, + "learning_rate": 5.705790155421758e-06, + "loss": 1.1726, + "step": 23233 + }, + { + "epoch": 1.96, + "learning_rate": 5.704969275571896e-06, + "loss": 1.2426, + "step": 23234 + }, + { + "epoch": 1.96, + "learning_rate": 5.704148431208951e-06, + "loss": 1.1268, + "step": 23235 + }, + { + "epoch": 1.96, + "learning_rate": 5.703327622339703e-06, + "loss": 1.1529, + "step": 23236 + }, + { + "epoch": 1.96, + "learning_rate": 5.702506848970943e-06, + "loss": 1.2326, + "step": 23237 + }, + { + "epoch": 1.96, + "learning_rate": 5.7016861111094435e-06, + "loss": 1.0717, + "step": 23238 + }, + { + "epoch": 1.96, + "learning_rate": 5.70086540876199e-06, + "loss": 1.09, + "step": 23239 + }, + { + "epoch": 1.96, + "learning_rate": 5.700044741935361e-06, + "loss": 1.2481, + "step": 23240 + }, + { + "epoch": 1.96, + "learning_rate": 5.699224110636339e-06, + "loss": 1.0925, + "step": 23241 + }, + { + "epoch": 1.96, + "learning_rate": 5.698403514871705e-06, + "loss": 1.178, + "step": 23242 + }, + { + "epoch": 1.96, + "learning_rate": 5.6975829546482395e-06, + "loss": 1.2105, + "step": 23243 + }, + { + "epoch": 1.96, + "learning_rate": 5.696762429972715e-06, + "loss": 1.1712, + "step": 23244 + }, + { + "epoch": 1.96, + "learning_rate": 5.695941940851924e-06, + "loss": 1.1881, + "step": 23245 + }, + { + "epoch": 1.96, + "learning_rate": 5.695121487292632e-06, + "loss": 1.2607, + "step": 23246 + }, + { + "epoch": 1.96, + "learning_rate": 5.694301069301626e-06, + "loss": 1.3007, + "step": 23247 + }, + { + "epoch": 1.96, + "learning_rate": 5.693480686885682e-06, + "loss": 1.2266, + "step": 23248 + }, + { + "epoch": 1.96, + "learning_rate": 5.692660340051579e-06, + "loss": 1.2706, + "step": 23249 + }, + { + "epoch": 1.96, + "learning_rate": 5.691840028806095e-06, + "loss": 1.1097, + "step": 23250 + }, + { + "epoch": 1.96, + "learning_rate": 5.691019753156006e-06, + "loss": 1.1604, + "step": 23251 + }, + { + "epoch": 1.96, + "learning_rate": 5.690199513108094e-06, + "loss": 1.1751, + "step": 23252 + }, + { + "epoch": 1.96, + "learning_rate": 5.68937930866913e-06, + "loss": 1.2021, + "step": 23253 + }, + { + "epoch": 1.96, + "learning_rate": 5.6885591398459e-06, + "loss": 1.209, + "step": 23254 + }, + { + "epoch": 1.96, + "learning_rate": 5.687739006645163e-06, + "loss": 1.1003, + "step": 23255 + }, + { + "epoch": 1.96, + "learning_rate": 5.686918909073718e-06, + "loss": 1.1007, + "step": 23256 + }, + { + "epoch": 1.96, + "learning_rate": 5.686098847138322e-06, + "loss": 1.194, + "step": 23257 + }, + { + "epoch": 1.96, + "learning_rate": 5.685278820845761e-06, + "loss": 1.1808, + "step": 23258 + }, + { + "epoch": 1.96, + "learning_rate": 5.684458830202806e-06, + "loss": 1.1273, + "step": 23259 + }, + { + "epoch": 1.96, + "learning_rate": 5.683638875216233e-06, + "loss": 1.126, + "step": 23260 + }, + { + "epoch": 1.96, + "learning_rate": 5.682818955892817e-06, + "loss": 1.1927, + "step": 23261 + }, + { + "epoch": 1.96, + "learning_rate": 5.6819990722393325e-06, + "loss": 1.1237, + "step": 23262 + }, + { + "epoch": 1.96, + "learning_rate": 5.681179224262554e-06, + "loss": 1.1101, + "step": 23263 + }, + { + "epoch": 1.96, + "learning_rate": 5.680359411969255e-06, + "loss": 1.1069, + "step": 23264 + }, + { + "epoch": 1.96, + "learning_rate": 5.6795396353662125e-06, + "loss": 1.2233, + "step": 23265 + }, + { + "epoch": 1.96, + "learning_rate": 5.678719894460189e-06, + "loss": 1.2342, + "step": 23266 + }, + { + "epoch": 1.96, + "learning_rate": 5.677900189257972e-06, + "loss": 1.1896, + "step": 23267 + }, + { + "epoch": 1.96, + "learning_rate": 5.677080519766323e-06, + "loss": 1.1221, + "step": 23268 + }, + { + "epoch": 1.96, + "learning_rate": 5.6762608859920175e-06, + "loss": 1.2763, + "step": 23269 + }, + { + "epoch": 1.96, + "learning_rate": 5.67544128794183e-06, + "loss": 1.1649, + "step": 23270 + }, + { + "epoch": 1.96, + "learning_rate": 5.674621725622533e-06, + "loss": 1.2014, + "step": 23271 + }, + { + "epoch": 1.96, + "learning_rate": 5.6738021990408895e-06, + "loss": 1.2955, + "step": 23272 + }, + { + "epoch": 1.96, + "learning_rate": 5.6729827082036836e-06, + "loss": 1.1423, + "step": 23273 + }, + { + "epoch": 1.96, + "learning_rate": 5.672163253117675e-06, + "loss": 1.1614, + "step": 23274 + }, + { + "epoch": 1.96, + "learning_rate": 5.671343833789636e-06, + "loss": 1.1641, + "step": 23275 + }, + { + "epoch": 1.96, + "learning_rate": 5.670524450226346e-06, + "loss": 1.1101, + "step": 23276 + }, + { + "epoch": 1.96, + "learning_rate": 5.669705102434566e-06, + "loss": 1.2384, + "step": 23277 + }, + { + "epoch": 1.96, + "learning_rate": 5.668885790421068e-06, + "loss": 1.3393, + "step": 23278 + }, + { + "epoch": 1.96, + "learning_rate": 5.668066514192621e-06, + "loss": 1.2091, + "step": 23279 + }, + { + "epoch": 1.96, + "learning_rate": 5.667247273755996e-06, + "loss": 1.1503, + "step": 23280 + }, + { + "epoch": 1.96, + "learning_rate": 5.666428069117959e-06, + "loss": 1.2818, + "step": 23281 + }, + { + "epoch": 1.96, + "learning_rate": 5.665608900285287e-06, + "loss": 1.157, + "step": 23282 + }, + { + "epoch": 1.96, + "learning_rate": 5.664789767264731e-06, + "loss": 1.1593, + "step": 23283 + }, + { + "epoch": 1.96, + "learning_rate": 5.663970670063077e-06, + "loss": 1.1748, + "step": 23284 + }, + { + "epoch": 1.96, + "learning_rate": 5.663151608687083e-06, + "loss": 1.2076, + "step": 23285 + }, + { + "epoch": 1.96, + "learning_rate": 5.662332583143513e-06, + "loss": 1.1811, + "step": 23286 + }, + { + "epoch": 1.96, + "learning_rate": 5.661513593439147e-06, + "loss": 1.1872, + "step": 23287 + }, + { + "epoch": 1.96, + "learning_rate": 5.6606946395807415e-06, + "loss": 1.1629, + "step": 23288 + }, + { + "epoch": 1.96, + "learning_rate": 5.659875721575065e-06, + "loss": 1.1197, + "step": 23289 + }, + { + "epoch": 1.96, + "learning_rate": 5.6590568394288845e-06, + "loss": 1.2147, + "step": 23290 + }, + { + "epoch": 1.96, + "learning_rate": 5.658237993148966e-06, + "loss": 1.078, + "step": 23291 + }, + { + "epoch": 1.96, + "learning_rate": 5.657419182742074e-06, + "loss": 1.2222, + "step": 23292 + }, + { + "epoch": 1.96, + "learning_rate": 5.6566004082149795e-06, + "loss": 1.1505, + "step": 23293 + }, + { + "epoch": 1.96, + "learning_rate": 5.655781669574434e-06, + "loss": 1.1698, + "step": 23294 + }, + { + "epoch": 1.96, + "learning_rate": 5.654962966827219e-06, + "loss": 1.1598, + "step": 23295 + }, + { + "epoch": 1.96, + "learning_rate": 5.654144299980086e-06, + "loss": 1.1827, + "step": 23296 + }, + { + "epoch": 1.96, + "learning_rate": 5.653325669039804e-06, + "loss": 1.1373, + "step": 23297 + }, + { + "epoch": 1.96, + "learning_rate": 5.652507074013136e-06, + "loss": 1.1569, + "step": 23298 + }, + { + "epoch": 1.96, + "learning_rate": 5.651688514906845e-06, + "loss": 1.1868, + "step": 23299 + }, + { + "epoch": 1.96, + "learning_rate": 5.6508699917276975e-06, + "loss": 1.1784, + "step": 23300 + }, + { + "epoch": 1.96, + "learning_rate": 5.650051504482452e-06, + "loss": 1.0766, + "step": 23301 + }, + { + "epoch": 1.96, + "learning_rate": 5.649233053177875e-06, + "loss": 1.2724, + "step": 23302 + }, + { + "epoch": 1.96, + "learning_rate": 5.6484146378207246e-06, + "loss": 1.18, + "step": 23303 + }, + { + "epoch": 1.96, + "learning_rate": 5.647596258417772e-06, + "loss": 1.1957, + "step": 23304 + }, + { + "epoch": 1.96, + "learning_rate": 5.6467779149757625e-06, + "loss": 1.0652, + "step": 23305 + }, + { + "epoch": 1.96, + "learning_rate": 5.645959607501474e-06, + "loss": 1.2237, + "step": 23306 + }, + { + "epoch": 1.96, + "learning_rate": 5.645141336001658e-06, + "loss": 1.1306, + "step": 23307 + }, + { + "epoch": 1.96, + "learning_rate": 5.644323100483078e-06, + "loss": 1.1672, + "step": 23308 + }, + { + "epoch": 1.96, + "learning_rate": 5.643504900952494e-06, + "loss": 1.1778, + "step": 23309 + }, + { + "epoch": 1.96, + "learning_rate": 5.642686737416667e-06, + "loss": 1.1071, + "step": 23310 + }, + { + "epoch": 1.96, + "learning_rate": 5.641868609882357e-06, + "loss": 1.2102, + "step": 23311 + }, + { + "epoch": 1.96, + "learning_rate": 5.641050518356323e-06, + "loss": 1.1555, + "step": 23312 + }, + { + "epoch": 1.96, + "learning_rate": 5.640232462845325e-06, + "loss": 1.1659, + "step": 23313 + }, + { + "epoch": 1.96, + "learning_rate": 5.639414443356121e-06, + "loss": 1.2074, + "step": 23314 + }, + { + "epoch": 1.96, + "learning_rate": 5.638596459895475e-06, + "loss": 1.2578, + "step": 23315 + }, + { + "epoch": 1.96, + "learning_rate": 5.637778512470133e-06, + "loss": 1.2414, + "step": 23316 + }, + { + "epoch": 1.96, + "learning_rate": 5.636960601086869e-06, + "loss": 1.2791, + "step": 23317 + }, + { + "epoch": 1.96, + "learning_rate": 5.6361427257524295e-06, + "loss": 1.1914, + "step": 23318 + }, + { + "epoch": 1.96, + "learning_rate": 5.635324886473574e-06, + "loss": 1.0906, + "step": 23319 + }, + { + "epoch": 1.96, + "learning_rate": 5.634507083257061e-06, + "loss": 1.186, + "step": 23320 + }, + { + "epoch": 1.96, + "learning_rate": 5.633689316109654e-06, + "loss": 1.1859, + "step": 23321 + }, + { + "epoch": 1.96, + "learning_rate": 5.632871585038093e-06, + "loss": 1.2911, + "step": 23322 + }, + { + "epoch": 1.96, + "learning_rate": 5.632053890049154e-06, + "loss": 1.1964, + "step": 23323 + }, + { + "epoch": 1.96, + "learning_rate": 5.631236231149579e-06, + "loss": 1.1718, + "step": 23324 + }, + { + "epoch": 1.96, + "learning_rate": 5.630418608346124e-06, + "loss": 1.1316, + "step": 23325 + }, + { + "epoch": 1.96, + "learning_rate": 5.6296010216455565e-06, + "loss": 1.1473, + "step": 23326 + }, + { + "epoch": 1.96, + "learning_rate": 5.628783471054622e-06, + "loss": 1.1691, + "step": 23327 + }, + { + "epoch": 1.96, + "learning_rate": 5.627965956580075e-06, + "loss": 1.1183, + "step": 23328 + }, + { + "epoch": 1.96, + "learning_rate": 5.627148478228674e-06, + "loss": 1.2177, + "step": 23329 + }, + { + "epoch": 1.96, + "learning_rate": 5.626331036007172e-06, + "loss": 1.1073, + "step": 23330 + }, + { + "epoch": 1.96, + "learning_rate": 5.625513629922322e-06, + "loss": 1.0937, + "step": 23331 + }, + { + "epoch": 1.96, + "learning_rate": 5.6246962599808826e-06, + "loss": 1.1962, + "step": 23332 + }, + { + "epoch": 1.96, + "learning_rate": 5.623878926189595e-06, + "loss": 1.1209, + "step": 23333 + }, + { + "epoch": 1.96, + "learning_rate": 5.623061628555229e-06, + "loss": 1.1657, + "step": 23334 + }, + { + "epoch": 1.96, + "learning_rate": 5.6222443670845236e-06, + "loss": 1.1347, + "step": 23335 + }, + { + "epoch": 1.96, + "learning_rate": 5.621427141784232e-06, + "loss": 1.1407, + "step": 23336 + }, + { + "epoch": 1.96, + "learning_rate": 5.62060995266112e-06, + "loss": 1.2201, + "step": 23337 + }, + { + "epoch": 1.96, + "learning_rate": 5.619792799721926e-06, + "loss": 1.1488, + "step": 23338 + }, + { + "epoch": 1.96, + "learning_rate": 5.618975682973405e-06, + "loss": 1.1993, + "step": 23339 + }, + { + "epoch": 1.97, + "learning_rate": 5.618158602422309e-06, + "loss": 1.2059, + "step": 23340 + }, + { + "epoch": 1.97, + "learning_rate": 5.617341558075389e-06, + "loss": 1.0916, + "step": 23341 + }, + { + "epoch": 1.97, + "learning_rate": 5.616524549939396e-06, + "loss": 1.1773, + "step": 23342 + }, + { + "epoch": 1.97, + "learning_rate": 5.615707578021083e-06, + "loss": 1.1592, + "step": 23343 + }, + { + "epoch": 1.97, + "learning_rate": 5.614890642327191e-06, + "loss": 1.1685, + "step": 23344 + }, + { + "epoch": 1.97, + "learning_rate": 5.6140737428644835e-06, + "loss": 1.1411, + "step": 23345 + }, + { + "epoch": 1.97, + "learning_rate": 5.613256879639697e-06, + "loss": 1.1432, + "step": 23346 + }, + { + "epoch": 1.97, + "learning_rate": 5.612440052659587e-06, + "loss": 1.2199, + "step": 23347 + }, + { + "epoch": 1.97, + "learning_rate": 5.611623261930902e-06, + "loss": 1.1823, + "step": 23348 + }, + { + "epoch": 1.97, + "learning_rate": 5.610806507460389e-06, + "loss": 1.1641, + "step": 23349 + }, + { + "epoch": 1.97, + "learning_rate": 5.609989789254799e-06, + "loss": 1.1733, + "step": 23350 + }, + { + "epoch": 1.97, + "learning_rate": 5.609173107320878e-06, + "loss": 1.182, + "step": 23351 + }, + { + "epoch": 1.97, + "learning_rate": 5.608356461665372e-06, + "loss": 1.1494, + "step": 23352 + }, + { + "epoch": 1.97, + "learning_rate": 5.607539852295032e-06, + "loss": 1.1515, + "step": 23353 + }, + { + "epoch": 1.97, + "learning_rate": 5.606723279216607e-06, + "loss": 1.0889, + "step": 23354 + }, + { + "epoch": 1.97, + "learning_rate": 5.605906742436833e-06, + "loss": 1.1394, + "step": 23355 + }, + { + "epoch": 1.97, + "learning_rate": 5.605090241962472e-06, + "loss": 1.2354, + "step": 23356 + }, + { + "epoch": 1.97, + "learning_rate": 5.604273777800256e-06, + "loss": 1.2009, + "step": 23357 + }, + { + "epoch": 1.97, + "learning_rate": 5.603457349956938e-06, + "loss": 1.2083, + "step": 23358 + }, + { + "epoch": 1.97, + "learning_rate": 5.602640958439263e-06, + "loss": 1.2624, + "step": 23359 + }, + { + "epoch": 1.97, + "learning_rate": 5.601824603253976e-06, + "loss": 1.0934, + "step": 23360 + }, + { + "epoch": 1.97, + "learning_rate": 5.60100828440782e-06, + "loss": 1.1361, + "step": 23361 + }, + { + "epoch": 1.97, + "learning_rate": 5.600192001907543e-06, + "loss": 1.183, + "step": 23362 + }, + { + "epoch": 1.97, + "learning_rate": 5.599375755759888e-06, + "loss": 1.1458, + "step": 23363 + }, + { + "epoch": 1.97, + "learning_rate": 5.598559545971598e-06, + "loss": 1.298, + "step": 23364 + }, + { + "epoch": 1.97, + "learning_rate": 5.597743372549421e-06, + "loss": 1.1088, + "step": 23365 + }, + { + "epoch": 1.97, + "learning_rate": 5.59692723550009e-06, + "loss": 1.1418, + "step": 23366 + }, + { + "epoch": 1.97, + "learning_rate": 5.596111134830365e-06, + "loss": 1.1314, + "step": 23367 + }, + { + "epoch": 1.97, + "learning_rate": 5.595295070546972e-06, + "loss": 1.1897, + "step": 23368 + }, + { + "epoch": 1.97, + "learning_rate": 5.594479042656662e-06, + "loss": 1.1826, + "step": 23369 + }, + { + "epoch": 1.97, + "learning_rate": 5.593663051166177e-06, + "loss": 1.121, + "step": 23370 + }, + { + "epoch": 1.97, + "learning_rate": 5.592847096082261e-06, + "loss": 1.1622, + "step": 23371 + }, + { + "epoch": 1.97, + "learning_rate": 5.592031177411646e-06, + "loss": 1.1395, + "step": 23372 + }, + { + "epoch": 1.97, + "learning_rate": 5.591215295161088e-06, + "loss": 1.2224, + "step": 23373 + }, + { + "epoch": 1.97, + "learning_rate": 5.5903994493373155e-06, + "loss": 1.2167, + "step": 23374 + }, + { + "epoch": 1.97, + "learning_rate": 5.5895836399470695e-06, + "loss": 1.1432, + "step": 23375 + }, + { + "epoch": 1.97, + "learning_rate": 5.588767866997105e-06, + "loss": 1.236, + "step": 23376 + }, + { + "epoch": 1.97, + "learning_rate": 5.587952130494145e-06, + "loss": 1.2059, + "step": 23377 + }, + { + "epoch": 1.97, + "learning_rate": 5.587136430444939e-06, + "loss": 1.151, + "step": 23378 + }, + { + "epoch": 1.97, + "learning_rate": 5.586320766856224e-06, + "loss": 1.28, + "step": 23379 + }, + { + "epoch": 1.97, + "learning_rate": 5.5855051397347395e-06, + "loss": 1.2103, + "step": 23380 + }, + { + "epoch": 1.97, + "learning_rate": 5.584689549087223e-06, + "loss": 1.1493, + "step": 23381 + }, + { + "epoch": 1.97, + "learning_rate": 5.583873994920419e-06, + "loss": 1.1453, + "step": 23382 + }, + { + "epoch": 1.97, + "learning_rate": 5.5830584772410546e-06, + "loss": 1.2318, + "step": 23383 + }, + { + "epoch": 1.97, + "learning_rate": 5.582242996055881e-06, + "loss": 1.1094, + "step": 23384 + }, + { + "epoch": 1.97, + "learning_rate": 5.5814275513716265e-06, + "loss": 1.1369, + "step": 23385 + }, + { + "epoch": 1.97, + "learning_rate": 5.580612143195028e-06, + "loss": 1.1092, + "step": 23386 + }, + { + "epoch": 1.97, + "learning_rate": 5.579796771532833e-06, + "loss": 1.1418, + "step": 23387 + }, + { + "epoch": 1.97, + "learning_rate": 5.578981436391768e-06, + "loss": 1.1813, + "step": 23388 + }, + { + "epoch": 1.97, + "learning_rate": 5.5781661377785746e-06, + "loss": 1.1699, + "step": 23389 + }, + { + "epoch": 1.97, + "learning_rate": 5.577350875699987e-06, + "loss": 1.2315, + "step": 23390 + }, + { + "epoch": 1.97, + "learning_rate": 5.576535650162742e-06, + "loss": 1.1155, + "step": 23391 + }, + { + "epoch": 1.97, + "learning_rate": 5.575720461173575e-06, + "loss": 1.1097, + "step": 23392 + }, + { + "epoch": 1.97, + "learning_rate": 5.574905308739226e-06, + "loss": 1.105, + "step": 23393 + }, + { + "epoch": 1.97, + "learning_rate": 5.574090192866417e-06, + "loss": 1.1375, + "step": 23394 + }, + { + "epoch": 1.97, + "learning_rate": 5.573275113561899e-06, + "loss": 1.1823, + "step": 23395 + }, + { + "epoch": 1.97, + "learning_rate": 5.572460070832395e-06, + "loss": 1.2072, + "step": 23396 + }, + { + "epoch": 1.97, + "learning_rate": 5.571645064684643e-06, + "loss": 1.2195, + "step": 23397 + }, + { + "epoch": 1.97, + "learning_rate": 5.570830095125377e-06, + "loss": 1.1542, + "step": 23398 + }, + { + "epoch": 1.97, + "learning_rate": 5.570015162161331e-06, + "loss": 1.2547, + "step": 23399 + }, + { + "epoch": 1.97, + "learning_rate": 5.569200265799236e-06, + "loss": 1.2531, + "step": 23400 + }, + { + "epoch": 1.97, + "learning_rate": 5.568385406045827e-06, + "loss": 1.2456, + "step": 23401 + }, + { + "epoch": 1.97, + "learning_rate": 5.567570582907836e-06, + "loss": 1.1185, + "step": 23402 + }, + { + "epoch": 1.97, + "learning_rate": 5.566755796391996e-06, + "loss": 1.1908, + "step": 23403 + }, + { + "epoch": 1.97, + "learning_rate": 5.565941046505041e-06, + "loss": 1.1517, + "step": 23404 + }, + { + "epoch": 1.97, + "learning_rate": 5.565126333253693e-06, + "loss": 1.2011, + "step": 23405 + }, + { + "epoch": 1.97, + "learning_rate": 5.564311656644699e-06, + "loss": 1.1403, + "step": 23406 + }, + { + "epoch": 1.97, + "learning_rate": 5.563497016684778e-06, + "loss": 1.1537, + "step": 23407 + }, + { + "epoch": 1.97, + "learning_rate": 5.562682413380663e-06, + "loss": 1.2442, + "step": 23408 + }, + { + "epoch": 1.97, + "learning_rate": 5.561867846739086e-06, + "loss": 1.2902, + "step": 23409 + }, + { + "epoch": 1.97, + "learning_rate": 5.561053316766779e-06, + "loss": 1.2117, + "step": 23410 + }, + { + "epoch": 1.97, + "learning_rate": 5.56023882347047e-06, + "loss": 1.102, + "step": 23411 + }, + { + "epoch": 1.97, + "learning_rate": 5.559424366856887e-06, + "loss": 1.1877, + "step": 23412 + }, + { + "epoch": 1.97, + "learning_rate": 5.558609946932763e-06, + "loss": 1.1981, + "step": 23413 + }, + { + "epoch": 1.97, + "learning_rate": 5.557795563704824e-06, + "loss": 1.1429, + "step": 23414 + }, + { + "epoch": 1.97, + "learning_rate": 5.556981217179803e-06, + "loss": 1.2583, + "step": 23415 + }, + { + "epoch": 1.97, + "learning_rate": 5.556166907364418e-06, + "loss": 1.1325, + "step": 23416 + }, + { + "epoch": 1.97, + "learning_rate": 5.555352634265412e-06, + "loss": 1.2465, + "step": 23417 + }, + { + "epoch": 1.97, + "learning_rate": 5.554538397889502e-06, + "loss": 1.1394, + "step": 23418 + }, + { + "epoch": 1.97, + "learning_rate": 5.5537241982434176e-06, + "loss": 1.1324, + "step": 23419 + }, + { + "epoch": 1.97, + "learning_rate": 5.5529100353338876e-06, + "loss": 1.2909, + "step": 23420 + }, + { + "epoch": 1.97, + "learning_rate": 5.5520959091676406e-06, + "loss": 1.1395, + "step": 23421 + }, + { + "epoch": 1.97, + "learning_rate": 5.551281819751393e-06, + "loss": 1.2159, + "step": 23422 + }, + { + "epoch": 1.97, + "learning_rate": 5.550467767091887e-06, + "loss": 1.1373, + "step": 23423 + }, + { + "epoch": 1.97, + "learning_rate": 5.549653751195838e-06, + "loss": 1.1811, + "step": 23424 + }, + { + "epoch": 1.97, + "learning_rate": 5.548839772069968e-06, + "loss": 1.2025, + "step": 23425 + }, + { + "epoch": 1.97, + "learning_rate": 5.548025829721018e-06, + "loss": 1.1837, + "step": 23426 + }, + { + "epoch": 1.97, + "learning_rate": 5.547211924155698e-06, + "loss": 1.1634, + "step": 23427 + }, + { + "epoch": 1.97, + "learning_rate": 5.546398055380738e-06, + "loss": 1.1463, + "step": 23428 + }, + { + "epoch": 1.97, + "learning_rate": 5.545584223402863e-06, + "loss": 1.1752, + "step": 23429 + }, + { + "epoch": 1.97, + "learning_rate": 5.5447704282287986e-06, + "loss": 1.1392, + "step": 23430 + }, + { + "epoch": 1.97, + "learning_rate": 5.5439566698652645e-06, + "loss": 1.2143, + "step": 23431 + }, + { + "epoch": 1.97, + "learning_rate": 5.543142948318992e-06, + "loss": 1.1847, + "step": 23432 + }, + { + "epoch": 1.97, + "learning_rate": 5.542329263596688e-06, + "loss": 1.2478, + "step": 23433 + }, + { + "epoch": 1.97, + "learning_rate": 5.541515615705098e-06, + "loss": 1.146, + "step": 23434 + }, + { + "epoch": 1.97, + "learning_rate": 5.540702004650926e-06, + "loss": 1.1986, + "step": 23435 + }, + { + "epoch": 1.97, + "learning_rate": 5.539888430440898e-06, + "loss": 1.2411, + "step": 23436 + }, + { + "epoch": 1.97, + "learning_rate": 5.539074893081748e-06, + "loss": 1.201, + "step": 23437 + }, + { + "epoch": 1.97, + "learning_rate": 5.538261392580184e-06, + "loss": 1.2168, + "step": 23438 + }, + { + "epoch": 1.97, + "learning_rate": 5.537447928942934e-06, + "loss": 1.2105, + "step": 23439 + }, + { + "epoch": 1.97, + "learning_rate": 5.536634502176715e-06, + "loss": 1.163, + "step": 23440 + }, + { + "epoch": 1.97, + "learning_rate": 5.535821112288251e-06, + "loss": 1.1163, + "step": 23441 + }, + { + "epoch": 1.97, + "learning_rate": 5.535007759284264e-06, + "loss": 1.1658, + "step": 23442 + }, + { + "epoch": 1.97, + "learning_rate": 5.534194443171472e-06, + "loss": 1.2022, + "step": 23443 + }, + { + "epoch": 1.97, + "learning_rate": 5.53338116395659e-06, + "loss": 1.1248, + "step": 23444 + }, + { + "epoch": 1.97, + "learning_rate": 5.532567921646348e-06, + "loss": 1.2125, + "step": 23445 + }, + { + "epoch": 1.97, + "learning_rate": 5.531754716247457e-06, + "loss": 1.1234, + "step": 23446 + }, + { + "epoch": 1.97, + "learning_rate": 5.530941547766638e-06, + "loss": 1.0726, + "step": 23447 + }, + { + "epoch": 1.97, + "learning_rate": 5.530128416210611e-06, + "loss": 1.1221, + "step": 23448 + }, + { + "epoch": 1.97, + "learning_rate": 5.529315321586093e-06, + "loss": 1.1993, + "step": 23449 + }, + { + "epoch": 1.97, + "learning_rate": 5.528502263899802e-06, + "loss": 1.1379, + "step": 23450 + }, + { + "epoch": 1.97, + "learning_rate": 5.5276892431584575e-06, + "loss": 1.1372, + "step": 23451 + }, + { + "epoch": 1.97, + "learning_rate": 5.526876259368774e-06, + "loss": 1.1251, + "step": 23452 + }, + { + "epoch": 1.97, + "learning_rate": 5.526063312537473e-06, + "loss": 1.2009, + "step": 23453 + }, + { + "epoch": 1.97, + "learning_rate": 5.525250402671272e-06, + "loss": 1.0977, + "step": 23454 + }, + { + "epoch": 1.97, + "learning_rate": 5.524437529776875e-06, + "loss": 1.1255, + "step": 23455 + }, + { + "epoch": 1.97, + "learning_rate": 5.523624693861016e-06, + "loss": 1.1083, + "step": 23456 + }, + { + "epoch": 1.97, + "learning_rate": 5.522811894930398e-06, + "loss": 1.172, + "step": 23457 + }, + { + "epoch": 1.97, + "learning_rate": 5.521999132991742e-06, + "loss": 1.0726, + "step": 23458 + }, + { + "epoch": 1.98, + "learning_rate": 5.521186408051762e-06, + "loss": 1.1265, + "step": 23459 + }, + { + "epoch": 1.98, + "learning_rate": 5.5203737201171735e-06, + "loss": 1.1426, + "step": 23460 + }, + { + "epoch": 1.98, + "learning_rate": 5.51956106919469e-06, + "loss": 1.1579, + "step": 23461 + }, + { + "epoch": 1.98, + "learning_rate": 5.518748455291027e-06, + "loss": 1.1788, + "step": 23462 + }, + { + "epoch": 1.98, + "learning_rate": 5.517935878412899e-06, + "loss": 1.1712, + "step": 23463 + }, + { + "epoch": 1.98, + "learning_rate": 5.51712333856702e-06, + "loss": 1.116, + "step": 23464 + }, + { + "epoch": 1.98, + "learning_rate": 5.516310835760106e-06, + "loss": 1.1873, + "step": 23465 + }, + { + "epoch": 1.98, + "learning_rate": 5.515498369998859e-06, + "loss": 1.186, + "step": 23466 + }, + { + "epoch": 1.98, + "learning_rate": 5.514685941290007e-06, + "loss": 1.2225, + "step": 23467 + }, + { + "epoch": 1.98, + "learning_rate": 5.513873549640251e-06, + "loss": 1.1854, + "step": 23468 + }, + { + "epoch": 1.98, + "learning_rate": 5.513061195056309e-06, + "loss": 1.1701, + "step": 23469 + }, + { + "epoch": 1.98, + "learning_rate": 5.51224887754489e-06, + "loss": 1.2087, + "step": 23470 + }, + { + "epoch": 1.98, + "learning_rate": 5.511436597112711e-06, + "loss": 1.215, + "step": 23471 + }, + { + "epoch": 1.98, + "learning_rate": 5.5106243537664724e-06, + "loss": 1.099, + "step": 23472 + }, + { + "epoch": 1.98, + "learning_rate": 5.509812147512901e-06, + "loss": 1.1163, + "step": 23473 + }, + { + "epoch": 1.98, + "learning_rate": 5.508999978358692e-06, + "loss": 1.2432, + "step": 23474 + }, + { + "epoch": 1.98, + "learning_rate": 5.50818784631056e-06, + "loss": 1.1653, + "step": 23475 + }, + { + "epoch": 1.98, + "learning_rate": 5.507375751375226e-06, + "loss": 1.1988, + "step": 23476 + }, + { + "epoch": 1.98, + "learning_rate": 5.506563693559387e-06, + "loss": 1.2546, + "step": 23477 + }, + { + "epoch": 1.98, + "learning_rate": 5.505751672869757e-06, + "loss": 1.1747, + "step": 23478 + }, + { + "epoch": 1.98, + "learning_rate": 5.504939689313045e-06, + "loss": 1.1461, + "step": 23479 + }, + { + "epoch": 1.98, + "learning_rate": 5.50412774289596e-06, + "loss": 1.2447, + "step": 23480 + }, + { + "epoch": 1.98, + "learning_rate": 5.503315833625209e-06, + "loss": 1.2312, + "step": 23481 + }, + { + "epoch": 1.98, + "learning_rate": 5.502503961507506e-06, + "loss": 1.226, + "step": 23482 + }, + { + "epoch": 1.98, + "learning_rate": 5.501692126549547e-06, + "loss": 1.1888, + "step": 23483 + }, + { + "epoch": 1.98, + "learning_rate": 5.500880328758056e-06, + "loss": 1.198, + "step": 23484 + }, + { + "epoch": 1.98, + "learning_rate": 5.500068568139728e-06, + "loss": 1.1823, + "step": 23485 + }, + { + "epoch": 1.98, + "learning_rate": 5.4992568447012685e-06, + "loss": 1.1819, + "step": 23486 + }, + { + "epoch": 1.98, + "learning_rate": 5.498445158449398e-06, + "loss": 1.155, + "step": 23487 + }, + { + "epoch": 1.98, + "learning_rate": 5.497633509390809e-06, + "loss": 1.2567, + "step": 23488 + }, + { + "epoch": 1.98, + "learning_rate": 5.496821897532215e-06, + "loss": 1.2225, + "step": 23489 + }, + { + "epoch": 1.98, + "learning_rate": 5.496010322880318e-06, + "loss": 1.1798, + "step": 23490 + }, + { + "epoch": 1.98, + "learning_rate": 5.4951987854418264e-06, + "loss": 1.1245, + "step": 23491 + }, + { + "epoch": 1.98, + "learning_rate": 5.494387285223444e-06, + "loss": 1.2075, + "step": 23492 + }, + { + "epoch": 1.98, + "learning_rate": 5.49357582223188e-06, + "loss": 1.1815, + "step": 23493 + }, + { + "epoch": 1.98, + "learning_rate": 5.492764396473826e-06, + "loss": 1.2575, + "step": 23494 + }, + { + "epoch": 1.98, + "learning_rate": 5.491953007956003e-06, + "loss": 1.2351, + "step": 23495 + }, + { + "epoch": 1.98, + "learning_rate": 5.491141656685104e-06, + "loss": 1.1859, + "step": 23496 + }, + { + "epoch": 1.98, + "learning_rate": 5.490330342667835e-06, + "loss": 1.2107, + "step": 23497 + }, + { + "epoch": 1.98, + "learning_rate": 5.4895190659109e-06, + "loss": 1.0987, + "step": 23498 + }, + { + "epoch": 1.98, + "learning_rate": 5.4887078264210015e-06, + "loss": 1.1649, + "step": 23499 + }, + { + "epoch": 1.98, + "learning_rate": 5.487896624204842e-06, + "loss": 1.1432, + "step": 23500 + }, + { + "epoch": 1.98, + "learning_rate": 5.487085459269127e-06, + "loss": 1.1422, + "step": 23501 + }, + { + "epoch": 1.98, + "learning_rate": 5.4862743316205554e-06, + "loss": 1.2154, + "step": 23502 + }, + { + "epoch": 1.98, + "learning_rate": 5.48546324126583e-06, + "loss": 1.1241, + "step": 23503 + }, + { + "epoch": 1.98, + "learning_rate": 5.484652188211654e-06, + "loss": 1.2088, + "step": 23504 + }, + { + "epoch": 1.98, + "learning_rate": 5.483841172464721e-06, + "loss": 1.2357, + "step": 23505 + }, + { + "epoch": 1.98, + "learning_rate": 5.483030194031745e-06, + "loss": 1.2745, + "step": 23506 + }, + { + "epoch": 1.98, + "learning_rate": 5.482219252919416e-06, + "loss": 1.2108, + "step": 23507 + }, + { + "epoch": 1.98, + "learning_rate": 5.481408349134436e-06, + "loss": 1.1327, + "step": 23508 + }, + { + "epoch": 1.98, + "learning_rate": 5.480597482683507e-06, + "loss": 1.2537, + "step": 23509 + }, + { + "epoch": 1.98, + "learning_rate": 5.479786653573328e-06, + "loss": 1.2395, + "step": 23510 + }, + { + "epoch": 1.98, + "learning_rate": 5.478975861810598e-06, + "loss": 1.129, + "step": 23511 + }, + { + "epoch": 1.98, + "learning_rate": 5.478165107402016e-06, + "loss": 1.1695, + "step": 23512 + }, + { + "epoch": 1.98, + "learning_rate": 5.477354390354283e-06, + "loss": 1.197, + "step": 23513 + }, + { + "epoch": 1.98, + "learning_rate": 5.476543710674094e-06, + "loss": 1.1093, + "step": 23514 + }, + { + "epoch": 1.98, + "learning_rate": 5.475733068368152e-06, + "loss": 1.2266, + "step": 23515 + }, + { + "epoch": 1.98, + "learning_rate": 5.4749224634431445e-06, + "loss": 1.1294, + "step": 23516 + }, + { + "epoch": 1.98, + "learning_rate": 5.474111895905783e-06, + "loss": 1.1247, + "step": 23517 + }, + { + "epoch": 1.98, + "learning_rate": 5.4733013657627535e-06, + "loss": 1.1692, + "step": 23518 + }, + { + "epoch": 1.98, + "learning_rate": 5.472490873020757e-06, + "loss": 1.2022, + "step": 23519 + }, + { + "epoch": 1.98, + "learning_rate": 5.47168041768649e-06, + "loss": 1.2635, + "step": 23520 + }, + { + "epoch": 1.98, + "learning_rate": 5.470869999766653e-06, + "loss": 1.1695, + "step": 23521 + }, + { + "epoch": 1.98, + "learning_rate": 5.470059619267929e-06, + "loss": 1.201, + "step": 23522 + }, + { + "epoch": 1.98, + "learning_rate": 5.4692492761970306e-06, + "loss": 1.2383, + "step": 23523 + }, + { + "epoch": 1.98, + "learning_rate": 5.4684389705606415e-06, + "loss": 1.2181, + "step": 23524 + }, + { + "epoch": 1.98, + "learning_rate": 5.467628702365454e-06, + "loss": 1.1351, + "step": 23525 + }, + { + "epoch": 1.98, + "learning_rate": 5.466818471618178e-06, + "loss": 1.1756, + "step": 23526 + }, + { + "epoch": 1.98, + "learning_rate": 5.466008278325494e-06, + "loss": 1.2071, + "step": 23527 + }, + { + "epoch": 1.98, + "learning_rate": 5.465198122494101e-06, + "loss": 1.1387, + "step": 23528 + }, + { + "epoch": 1.98, + "learning_rate": 5.464388004130692e-06, + "loss": 1.2342, + "step": 23529 + }, + { + "epoch": 1.98, + "learning_rate": 5.46357792324196e-06, + "loss": 1.1532, + "step": 23530 + }, + { + "epoch": 1.98, + "learning_rate": 5.4627678798346015e-06, + "loss": 1.1115, + "step": 23531 + }, + { + "epoch": 1.98, + "learning_rate": 5.461957873915309e-06, + "loss": 1.1482, + "step": 23532 + }, + { + "epoch": 1.98, + "learning_rate": 5.461147905490766e-06, + "loss": 1.2015, + "step": 23533 + }, + { + "epoch": 1.98, + "learning_rate": 5.46033797456768e-06, + "loss": 1.2394, + "step": 23534 + }, + { + "epoch": 1.98, + "learning_rate": 5.45952808115273e-06, + "loss": 1.0743, + "step": 23535 + }, + { + "epoch": 1.98, + "learning_rate": 5.4587182252526075e-06, + "loss": 1.2147, + "step": 23536 + }, + { + "epoch": 1.98, + "learning_rate": 5.457908406874016e-06, + "loss": 1.1323, + "step": 23537 + }, + { + "epoch": 1.98, + "learning_rate": 5.457098626023636e-06, + "loss": 1.1665, + "step": 23538 + }, + { + "epoch": 1.98, + "learning_rate": 5.456288882708161e-06, + "loss": 1.1453, + "step": 23539 + }, + { + "epoch": 1.98, + "learning_rate": 5.455479176934281e-06, + "loss": 1.1923, + "step": 23540 + }, + { + "epoch": 1.98, + "learning_rate": 5.454669508708687e-06, + "loss": 1.1814, + "step": 23541 + }, + { + "epoch": 1.98, + "learning_rate": 5.453859878038068e-06, + "loss": 1.2537, + "step": 23542 + }, + { + "epoch": 1.98, + "learning_rate": 5.453050284929116e-06, + "loss": 1.1835, + "step": 23543 + }, + { + "epoch": 1.98, + "learning_rate": 5.452240729388512e-06, + "loss": 1.1424, + "step": 23544 + }, + { + "epoch": 1.98, + "learning_rate": 5.451431211422957e-06, + "loss": 1.1714, + "step": 23545 + }, + { + "epoch": 1.98, + "learning_rate": 5.4506217310391305e-06, + "loss": 1.1764, + "step": 23546 + }, + { + "epoch": 1.98, + "learning_rate": 5.449812288243723e-06, + "loss": 1.1591, + "step": 23547 + }, + { + "epoch": 1.98, + "learning_rate": 5.449002883043422e-06, + "loss": 1.1195, + "step": 23548 + }, + { + "epoch": 1.98, + "learning_rate": 5.448193515444916e-06, + "loss": 1.1624, + "step": 23549 + }, + { + "epoch": 1.98, + "learning_rate": 5.447384185454892e-06, + "loss": 1.1901, + "step": 23550 + }, + { + "epoch": 1.98, + "learning_rate": 5.446574893080037e-06, + "loss": 1.2354, + "step": 23551 + }, + { + "epoch": 1.98, + "learning_rate": 5.4457656383270366e-06, + "loss": 1.1167, + "step": 23552 + }, + { + "epoch": 1.98, + "learning_rate": 5.444956421202579e-06, + "loss": 1.1657, + "step": 23553 + }, + { + "epoch": 1.98, + "learning_rate": 5.444147241713353e-06, + "loss": 1.1583, + "step": 23554 + }, + { + "epoch": 1.98, + "learning_rate": 5.443338099866032e-06, + "loss": 1.2189, + "step": 23555 + }, + { + "epoch": 1.98, + "learning_rate": 5.442528995667319e-06, + "loss": 1.1878, + "step": 23556 + }, + { + "epoch": 1.98, + "learning_rate": 5.441719929123885e-06, + "loss": 1.152, + "step": 23557 + }, + { + "epoch": 1.98, + "learning_rate": 5.44091090024242e-06, + "loss": 1.1777, + "step": 23558 + }, + { + "epoch": 1.98, + "learning_rate": 5.44010190902961e-06, + "loss": 1.1738, + "step": 23559 + }, + { + "epoch": 1.98, + "learning_rate": 5.439292955492135e-06, + "loss": 1.0769, + "step": 23560 + }, + { + "epoch": 1.98, + "learning_rate": 5.438484039636682e-06, + "loss": 1.2101, + "step": 23561 + }, + { + "epoch": 1.98, + "learning_rate": 5.437675161469933e-06, + "loss": 1.2258, + "step": 23562 + }, + { + "epoch": 1.98, + "learning_rate": 5.4368663209985724e-06, + "loss": 1.2012, + "step": 23563 + }, + { + "epoch": 1.98, + "learning_rate": 5.436057518229284e-06, + "loss": 1.1342, + "step": 23564 + }, + { + "epoch": 1.98, + "learning_rate": 5.435248753168751e-06, + "loss": 1.1622, + "step": 23565 + }, + { + "epoch": 1.98, + "learning_rate": 5.434440025823647e-06, + "loss": 1.2316, + "step": 23566 + }, + { + "epoch": 1.98, + "learning_rate": 5.433631336200669e-06, + "loss": 1.2246, + "step": 23567 + }, + { + "epoch": 1.98, + "learning_rate": 5.432822684306486e-06, + "loss": 1.1271, + "step": 23568 + }, + { + "epoch": 1.98, + "learning_rate": 5.432014070147783e-06, + "loss": 1.2175, + "step": 23569 + }, + { + "epoch": 1.98, + "learning_rate": 5.431205493731242e-06, + "loss": 1.1502, + "step": 23570 + }, + { + "epoch": 1.98, + "learning_rate": 5.430396955063544e-06, + "loss": 1.2365, + "step": 23571 + }, + { + "epoch": 1.98, + "learning_rate": 5.429588454151368e-06, + "loss": 1.1978, + "step": 23572 + }, + { + "epoch": 1.98, + "learning_rate": 5.4287799910014e-06, + "loss": 1.1498, + "step": 23573 + }, + { + "epoch": 1.98, + "learning_rate": 5.427971565620305e-06, + "loss": 1.2507, + "step": 23574 + }, + { + "epoch": 1.98, + "learning_rate": 5.427163178014777e-06, + "loss": 1.2345, + "step": 23575 + }, + { + "epoch": 1.98, + "learning_rate": 5.426354828191495e-06, + "loss": 1.0871, + "step": 23576 + }, + { + "epoch": 1.98, + "learning_rate": 5.425546516157124e-06, + "loss": 1.1658, + "step": 23577 + }, + { + "epoch": 1.99, + "learning_rate": 5.42473824191836e-06, + "loss": 1.2353, + "step": 23578 + }, + { + "epoch": 1.99, + "learning_rate": 5.42393000548187e-06, + "loss": 1.1347, + "step": 23579 + }, + { + "epoch": 1.99, + "learning_rate": 5.423121806854333e-06, + "loss": 1.2248, + "step": 23580 + }, + { + "epoch": 1.99, + "learning_rate": 5.42231364604243e-06, + "loss": 1.1053, + "step": 23581 + }, + { + "epoch": 1.99, + "learning_rate": 5.42150552305284e-06, + "loss": 1.1506, + "step": 23582 + }, + { + "epoch": 1.99, + "learning_rate": 5.4206974378922285e-06, + "loss": 1.1907, + "step": 23583 + }, + { + "epoch": 1.99, + "learning_rate": 5.419889390567289e-06, + "loss": 1.1772, + "step": 23584 + }, + { + "epoch": 1.99, + "learning_rate": 5.419081381084685e-06, + "loss": 1.1886, + "step": 23585 + }, + { + "epoch": 1.99, + "learning_rate": 5.418273409451092e-06, + "loss": 1.19, + "step": 23586 + }, + { + "epoch": 1.99, + "learning_rate": 5.417465475673199e-06, + "loss": 1.1885, + "step": 23587 + }, + { + "epoch": 1.99, + "learning_rate": 5.416657579757669e-06, + "loss": 1.1999, + "step": 23588 + }, + { + "epoch": 1.99, + "learning_rate": 5.415849721711182e-06, + "loss": 1.2704, + "step": 23589 + }, + { + "epoch": 1.99, + "learning_rate": 5.41504190154041e-06, + "loss": 1.1681, + "step": 23590 + }, + { + "epoch": 1.99, + "learning_rate": 5.414234119252028e-06, + "loss": 1.1154, + "step": 23591 + }, + { + "epoch": 1.99, + "learning_rate": 5.413426374852714e-06, + "loss": 1.1407, + "step": 23592 + }, + { + "epoch": 1.99, + "learning_rate": 5.412618668349142e-06, + "loss": 1.2036, + "step": 23593 + }, + { + "epoch": 1.99, + "learning_rate": 5.411810999747975e-06, + "loss": 1.1592, + "step": 23594 + }, + { + "epoch": 1.99, + "learning_rate": 5.411003369055902e-06, + "loss": 1.1944, + "step": 23595 + }, + { + "epoch": 1.99, + "learning_rate": 5.410195776279583e-06, + "loss": 1.1968, + "step": 23596 + }, + { + "epoch": 1.99, + "learning_rate": 5.409388221425696e-06, + "loss": 1.1944, + "step": 23597 + }, + { + "epoch": 1.99, + "learning_rate": 5.408580704500913e-06, + "loss": 1.1292, + "step": 23598 + }, + { + "epoch": 1.99, + "learning_rate": 5.407773225511905e-06, + "loss": 1.14, + "step": 23599 + }, + { + "epoch": 1.99, + "learning_rate": 5.406965784465346e-06, + "loss": 1.174, + "step": 23600 + }, + { + "epoch": 1.99, + "learning_rate": 5.406158381367903e-06, + "loss": 1.1832, + "step": 23601 + }, + { + "epoch": 1.99, + "learning_rate": 5.405351016226251e-06, + "loss": 1.0847, + "step": 23602 + }, + { + "epoch": 1.99, + "learning_rate": 5.4045436890470595e-06, + "loss": 1.1084, + "step": 23603 + }, + { + "epoch": 1.99, + "learning_rate": 5.403736399837003e-06, + "loss": 1.2757, + "step": 23604 + }, + { + "epoch": 1.99, + "learning_rate": 5.402929148602739e-06, + "loss": 1.1357, + "step": 23605 + }, + { + "epoch": 1.99, + "learning_rate": 5.402121935350954e-06, + "loss": 1.2254, + "step": 23606 + }, + { + "epoch": 1.99, + "learning_rate": 5.401314760088305e-06, + "loss": 1.2248, + "step": 23607 + }, + { + "epoch": 1.99, + "learning_rate": 5.4005076228214645e-06, + "loss": 1.2487, + "step": 23608 + }, + { + "epoch": 1.99, + "learning_rate": 5.399700523557103e-06, + "loss": 1.1634, + "step": 23609 + }, + { + "epoch": 1.99, + "learning_rate": 5.398893462301886e-06, + "loss": 1.1908, + "step": 23610 + }, + { + "epoch": 1.99, + "learning_rate": 5.398086439062486e-06, + "loss": 1.207, + "step": 23611 + }, + { + "epoch": 1.99, + "learning_rate": 5.397279453845568e-06, + "loss": 1.2642, + "step": 23612 + }, + { + "epoch": 1.99, + "learning_rate": 5.396472506657801e-06, + "loss": 1.267, + "step": 23613 + }, + { + "epoch": 1.99, + "learning_rate": 5.39566559750585e-06, + "loss": 1.253, + "step": 23614 + }, + { + "epoch": 1.99, + "learning_rate": 5.394858726396388e-06, + "loss": 1.2355, + "step": 23615 + }, + { + "epoch": 1.99, + "learning_rate": 5.39405189333607e-06, + "loss": 1.1503, + "step": 23616 + }, + { + "epoch": 1.99, + "learning_rate": 5.393245098331576e-06, + "loss": 1.1845, + "step": 23617 + }, + { + "epoch": 1.99, + "learning_rate": 5.392438341389561e-06, + "loss": 1.1931, + "step": 23618 + }, + { + "epoch": 1.99, + "learning_rate": 5.391631622516697e-06, + "loss": 1.222, + "step": 23619 + }, + { + "epoch": 1.99, + "learning_rate": 5.390824941719646e-06, + "loss": 1.2272, + "step": 23620 + }, + { + "epoch": 1.99, + "learning_rate": 5.390018299005074e-06, + "loss": 1.1854, + "step": 23621 + }, + { + "epoch": 1.99, + "learning_rate": 5.389211694379648e-06, + "loss": 1.1749, + "step": 23622 + }, + { + "epoch": 1.99, + "learning_rate": 5.388405127850033e-06, + "loss": 1.2469, + "step": 23623 + }, + { + "epoch": 1.99, + "learning_rate": 5.387598599422883e-06, + "loss": 1.151, + "step": 23624 + }, + { + "epoch": 1.99, + "learning_rate": 5.386792109104874e-06, + "loss": 1.2387, + "step": 23625 + }, + { + "epoch": 1.99, + "learning_rate": 5.385985656902668e-06, + "loss": 1.1237, + "step": 23626 + }, + { + "epoch": 1.99, + "learning_rate": 5.3851792428229175e-06, + "loss": 1.1918, + "step": 23627 + }, + { + "epoch": 1.99, + "learning_rate": 5.384372866872303e-06, + "loss": 1.2625, + "step": 23628 + }, + { + "epoch": 1.99, + "learning_rate": 5.383566529057472e-06, + "loss": 1.1126, + "step": 23629 + }, + { + "epoch": 1.99, + "learning_rate": 5.382760229385091e-06, + "loss": 1.2065, + "step": 23630 + }, + { + "epoch": 1.99, + "learning_rate": 5.381953967861821e-06, + "loss": 1.1783, + "step": 23631 + }, + { + "epoch": 1.99, + "learning_rate": 5.381147744494333e-06, + "loss": 1.1732, + "step": 23632 + }, + { + "epoch": 1.99, + "learning_rate": 5.380341559289271e-06, + "loss": 1.1546, + "step": 23633 + }, + { + "epoch": 1.99, + "learning_rate": 5.379535412253313e-06, + "loss": 1.136, + "step": 23634 + }, + { + "epoch": 1.99, + "learning_rate": 5.378729303393109e-06, + "loss": 1.1921, + "step": 23635 + }, + { + "epoch": 1.99, + "learning_rate": 5.377923232715318e-06, + "loss": 1.1969, + "step": 23636 + }, + { + "epoch": 1.99, + "learning_rate": 5.377117200226611e-06, + "loss": 1.149, + "step": 23637 + }, + { + "epoch": 1.99, + "learning_rate": 5.37631120593364e-06, + "loss": 1.1213, + "step": 23638 + }, + { + "epoch": 1.99, + "learning_rate": 5.375505249843064e-06, + "loss": 1.2969, + "step": 23639 + }, + { + "epoch": 1.99, + "learning_rate": 5.374699331961544e-06, + "loss": 1.1719, + "step": 23640 + }, + { + "epoch": 1.99, + "learning_rate": 5.373893452295739e-06, + "loss": 1.209, + "step": 23641 + }, + { + "epoch": 1.99, + "learning_rate": 5.373087610852306e-06, + "loss": 1.1616, + "step": 23642 + }, + { + "epoch": 1.99, + "learning_rate": 5.372281807637908e-06, + "loss": 1.1603, + "step": 23643 + }, + { + "epoch": 1.99, + "learning_rate": 5.37147604265919e-06, + "loss": 1.1659, + "step": 23644 + }, + { + "epoch": 1.99, + "learning_rate": 5.370670315922828e-06, + "loss": 1.1689, + "step": 23645 + }, + { + "epoch": 1.99, + "learning_rate": 5.369864627435464e-06, + "loss": 1.1561, + "step": 23646 + }, + { + "epoch": 1.99, + "learning_rate": 5.369058977203762e-06, + "loss": 1.1602, + "step": 23647 + }, + { + "epoch": 1.99, + "learning_rate": 5.368253365234375e-06, + "loss": 1.2795, + "step": 23648 + }, + { + "epoch": 1.99, + "learning_rate": 5.367447791533962e-06, + "loss": 1.2016, + "step": 23649 + }, + { + "epoch": 1.99, + "learning_rate": 5.366642256109178e-06, + "loss": 1.2302, + "step": 23650 + }, + { + "epoch": 1.99, + "learning_rate": 5.365836758966678e-06, + "loss": 1.1623, + "step": 23651 + }, + { + "epoch": 1.99, + "learning_rate": 5.365031300113119e-06, + "loss": 1.3072, + "step": 23652 + }, + { + "epoch": 1.99, + "learning_rate": 5.364225879555153e-06, + "loss": 1.1632, + "step": 23653 + }, + { + "epoch": 1.99, + "learning_rate": 5.363420497299441e-06, + "loss": 1.238, + "step": 23654 + }, + { + "epoch": 1.99, + "learning_rate": 5.362615153352626e-06, + "loss": 1.2284, + "step": 23655 + }, + { + "epoch": 1.99, + "learning_rate": 5.3618098477213755e-06, + "loss": 1.1975, + "step": 23656 + }, + { + "epoch": 1.99, + "learning_rate": 5.361004580412332e-06, + "loss": 1.1484, + "step": 23657 + }, + { + "epoch": 1.99, + "learning_rate": 5.3601993514321535e-06, + "loss": 0.9946, + "step": 23658 + }, + { + "epoch": 1.99, + "learning_rate": 5.359394160787494e-06, + "loss": 1.1652, + "step": 23659 + }, + { + "epoch": 1.99, + "learning_rate": 5.358589008485005e-06, + "loss": 1.1715, + "step": 23660 + }, + { + "epoch": 1.99, + "learning_rate": 5.357783894531339e-06, + "loss": 1.1901, + "step": 23661 + }, + { + "epoch": 1.99, + "learning_rate": 5.356978818933147e-06, + "loss": 1.1984, + "step": 23662 + }, + { + "epoch": 1.99, + "learning_rate": 5.356173781697083e-06, + "loss": 1.1473, + "step": 23663 + }, + { + "epoch": 1.99, + "learning_rate": 5.3553687828297976e-06, + "loss": 1.1548, + "step": 23664 + }, + { + "epoch": 1.99, + "learning_rate": 5.354563822337945e-06, + "loss": 1.1951, + "step": 23665 + }, + { + "epoch": 1.99, + "learning_rate": 5.353758900228165e-06, + "loss": 1.2262, + "step": 23666 + }, + { + "epoch": 1.99, + "learning_rate": 5.352954016507124e-06, + "loss": 1.193, + "step": 23667 + }, + { + "epoch": 1.99, + "learning_rate": 5.35214917118146e-06, + "loss": 1.1883, + "step": 23668 + }, + { + "epoch": 1.99, + "learning_rate": 5.351344364257829e-06, + "loss": 1.108, + "step": 23669 + }, + { + "epoch": 1.99, + "learning_rate": 5.350539595742876e-06, + "loss": 1.1917, + "step": 23670 + }, + { + "epoch": 1.99, + "learning_rate": 5.349734865643255e-06, + "loss": 1.1494, + "step": 23671 + }, + { + "epoch": 1.99, + "learning_rate": 5.348930173965611e-06, + "loss": 1.2851, + "step": 23672 + }, + { + "epoch": 1.99, + "learning_rate": 5.348125520716595e-06, + "loss": 1.1075, + "step": 23673 + }, + { + "epoch": 1.99, + "learning_rate": 5.347320905902855e-06, + "loss": 1.1709, + "step": 23674 + }, + { + "epoch": 1.99, + "learning_rate": 5.346516329531039e-06, + "loss": 1.2374, + "step": 23675 + }, + { + "epoch": 1.99, + "learning_rate": 5.345711791607797e-06, + "loss": 1.2101, + "step": 23676 + }, + { + "epoch": 1.99, + "learning_rate": 5.344907292139767e-06, + "loss": 1.1287, + "step": 23677 + }, + { + "epoch": 1.99, + "learning_rate": 5.344102831133611e-06, + "loss": 1.0961, + "step": 23678 + }, + { + "epoch": 1.99, + "learning_rate": 5.3432984085959625e-06, + "loss": 1.1456, + "step": 23679 + }, + { + "epoch": 1.99, + "learning_rate": 5.342494024533474e-06, + "loss": 1.1845, + "step": 23680 + }, + { + "epoch": 1.99, + "learning_rate": 5.3416896789527885e-06, + "loss": 1.1928, + "step": 23681 + }, + { + "epoch": 1.99, + "learning_rate": 5.3408853718605604e-06, + "loss": 1.2086, + "step": 23682 + }, + { + "epoch": 1.99, + "learning_rate": 5.3400811032634184e-06, + "loss": 1.0939, + "step": 23683 + }, + { + "epoch": 1.99, + "learning_rate": 5.339276873168027e-06, + "loss": 1.2162, + "step": 23684 + }, + { + "epoch": 1.99, + "learning_rate": 5.3384726815810164e-06, + "loss": 1.1582, + "step": 23685 + }, + { + "epoch": 1.99, + "learning_rate": 5.337668528509032e-06, + "loss": 1.1918, + "step": 23686 + }, + { + "epoch": 1.99, + "learning_rate": 5.336864413958733e-06, + "loss": 1.2475, + "step": 23687 + }, + { + "epoch": 1.99, + "learning_rate": 5.336060337936746e-06, + "loss": 1.2148, + "step": 23688 + }, + { + "epoch": 1.99, + "learning_rate": 5.335256300449723e-06, + "loss": 1.1772, + "step": 23689 + }, + { + "epoch": 1.99, + "learning_rate": 5.334452301504305e-06, + "loss": 1.1734, + "step": 23690 + }, + { + "epoch": 1.99, + "learning_rate": 5.3336483411071336e-06, + "loss": 1.214, + "step": 23691 + }, + { + "epoch": 1.99, + "learning_rate": 5.332844419264856e-06, + "loss": 1.1589, + "step": 23692 + }, + { + "epoch": 1.99, + "learning_rate": 5.332040535984112e-06, + "loss": 1.2135, + "step": 23693 + }, + { + "epoch": 1.99, + "learning_rate": 5.331236691271538e-06, + "loss": 1.1813, + "step": 23694 + }, + { + "epoch": 1.99, + "learning_rate": 5.330432885133788e-06, + "loss": 1.2102, + "step": 23695 + }, + { + "epoch": 2.0, + "learning_rate": 5.32962911757749e-06, + "loss": 1.1721, + "step": 23696 + }, + { + "epoch": 2.0, + "learning_rate": 5.328825388609292e-06, + "loss": 1.1582, + "step": 23697 + }, + { + "epoch": 2.0, + "learning_rate": 5.3280216982358345e-06, + "loss": 1.1714, + "step": 23698 + }, + { + "epoch": 2.0, + "learning_rate": 5.327218046463756e-06, + "loss": 1.1542, + "step": 23699 + }, + { + "epoch": 2.0, + "learning_rate": 5.326414433299698e-06, + "loss": 1.2294, + "step": 23700 + }, + { + "epoch": 2.0, + "learning_rate": 5.325610858750299e-06, + "loss": 1.1855, + "step": 23701 + }, + { + "epoch": 2.0, + "learning_rate": 5.3248073228222e-06, + "loss": 1.1801, + "step": 23702 + }, + { + "epoch": 2.0, + "learning_rate": 5.324003825522039e-06, + "loss": 1.1648, + "step": 23703 + }, + { + "epoch": 2.0, + "learning_rate": 5.323200366856458e-06, + "loss": 1.1626, + "step": 23704 + }, + { + "epoch": 2.0, + "learning_rate": 5.322396946832084e-06, + "loss": 1.2005, + "step": 23705 + }, + { + "epoch": 2.0, + "learning_rate": 5.3215935654555736e-06, + "loss": 1.1833, + "step": 23706 + }, + { + "epoch": 2.0, + "learning_rate": 5.3207902227335494e-06, + "loss": 1.2487, + "step": 23707 + }, + { + "epoch": 2.0, + "learning_rate": 5.3199869186726535e-06, + "loss": 1.1461, + "step": 23708 + }, + { + "epoch": 2.0, + "learning_rate": 5.319183653279524e-06, + "loss": 1.147, + "step": 23709 + }, + { + "epoch": 2.0, + "learning_rate": 5.3183804265607965e-06, + "loss": 1.0059, + "step": 23710 + }, + { + "epoch": 2.0, + "learning_rate": 5.317577238523109e-06, + "loss": 1.1558, + "step": 23711 + }, + { + "epoch": 2.0, + "learning_rate": 5.316774089173097e-06, + "loss": 1.102, + "step": 23712 + }, + { + "epoch": 2.0, + "learning_rate": 5.315970978517395e-06, + "loss": 1.1619, + "step": 23713 + }, + { + "epoch": 2.0, + "learning_rate": 5.315167906562641e-06, + "loss": 1.1445, + "step": 23714 + }, + { + "epoch": 2.0, + "learning_rate": 5.314364873315474e-06, + "loss": 1.1871, + "step": 23715 + }, + { + "epoch": 2.0, + "learning_rate": 5.313561878782514e-06, + "loss": 1.1808, + "step": 23716 + }, + { + "epoch": 2.0, + "learning_rate": 5.312758922970414e-06, + "loss": 1.2124, + "step": 23717 + }, + { + "epoch": 2.0, + "learning_rate": 5.311956005885796e-06, + "loss": 1.2159, + "step": 23718 + }, + { + "epoch": 2.0, + "learning_rate": 5.3111531275353e-06, + "loss": 1.1468, + "step": 23719 + }, + { + "epoch": 2.0, + "learning_rate": 5.310350287925555e-06, + "loss": 1.1789, + "step": 23720 + }, + { + "epoch": 2.0, + "learning_rate": 5.309547487063198e-06, + "loss": 1.0939, + "step": 23721 + }, + { + "epoch": 2.0, + "learning_rate": 5.30874472495486e-06, + "loss": 1.1475, + "step": 23722 + }, + { + "epoch": 2.0, + "learning_rate": 5.307942001607175e-06, + "loss": 1.157, + "step": 23723 + }, + { + "epoch": 2.0, + "learning_rate": 5.3071393170267745e-06, + "loss": 1.1013, + "step": 23724 + }, + { + "epoch": 2.0, + "learning_rate": 5.306336671220291e-06, + "loss": 1.1282, + "step": 23725 + }, + { + "epoch": 2.0, + "learning_rate": 5.3055340641943595e-06, + "loss": 1.2109, + "step": 23726 + }, + { + "epoch": 2.0, + "learning_rate": 5.304731495955602e-06, + "loss": 1.0501, + "step": 23727 + }, + { + "epoch": 2.0, + "learning_rate": 5.3039289665106605e-06, + "loss": 1.2605, + "step": 23728 + }, + { + "epoch": 2.0, + "learning_rate": 5.3031264758661595e-06, + "loss": 1.0927, + "step": 23729 + }, + { + "epoch": 2.0, + "learning_rate": 5.302324024028729e-06, + "loss": 1.1618, + "step": 23730 + }, + { + "epoch": 2.0, + "learning_rate": 5.301521611005003e-06, + "loss": 1.1125, + "step": 23731 + }, + { + "epoch": 2.0, + "learning_rate": 5.300719236801611e-06, + "loss": 1.1652, + "step": 23732 + }, + { + "epoch": 2.0, + "learning_rate": 5.2999169014251725e-06, + "loss": 1.156, + "step": 23733 + }, + { + "epoch": 2.0, + "learning_rate": 5.299114604882334e-06, + "loss": 1.1893, + "step": 23734 + }, + { + "epoch": 2.0, + "learning_rate": 5.2983123471797105e-06, + "loss": 1.1685, + "step": 23735 + }, + { + "epoch": 2.0, + "learning_rate": 5.2975101283239325e-06, + "loss": 1.171, + "step": 23736 + }, + { + "epoch": 2.0, + "learning_rate": 5.296707948321638e-06, + "loss": 1.1006, + "step": 23737 + }, + { + "epoch": 2.0, + "learning_rate": 5.295905807179442e-06, + "loss": 1.1851, + "step": 23738 + }, + { + "epoch": 2.0, + "learning_rate": 5.295103704903979e-06, + "loss": 1.2453, + "step": 23739 + }, + { + "epoch": 2.0, + "learning_rate": 5.294301641501875e-06, + "loss": 1.1459, + "step": 23740 + }, + { + "epoch": 2.0, + "learning_rate": 5.2934996169797585e-06, + "loss": 1.1168, + "step": 23741 + }, + { + "epoch": 2.0, + "learning_rate": 5.2926976313442525e-06, + "loss": 1.1414, + "step": 23742 + }, + { + "epoch": 2.0, + "learning_rate": 5.291895684601989e-06, + "loss": 1.1592, + "step": 23743 + }, + { + "epoch": 2.0, + "learning_rate": 5.291093776759581e-06, + "loss": 1.1824, + "step": 23744 + }, + { + "epoch": 2.0, + "learning_rate": 5.290291907823675e-06, + "loss": 1.1406, + "step": 23745 + }, + { + "epoch": 2.0, + "learning_rate": 5.289490077800878e-06, + "loss": 1.1275, + "step": 23746 + }, + { + "epoch": 2.0, + "learning_rate": 5.288688286697822e-06, + "loss": 1.1024, + "step": 23747 + }, + { + "epoch": 2.0, + "learning_rate": 5.287886534521131e-06, + "loss": 1.2086, + "step": 23748 + }, + { + "epoch": 2.0, + "learning_rate": 5.287084821277429e-06, + "loss": 1.1092, + "step": 23749 + }, + { + "epoch": 2.0, + "learning_rate": 5.286283146973341e-06, + "loss": 1.2697, + "step": 23750 + }, + { + "epoch": 2.0, + "learning_rate": 5.285481511615491e-06, + "loss": 1.1628, + "step": 23751 + }, + { + "epoch": 2.0, + "learning_rate": 5.284679915210502e-06, + "loss": 1.1737, + "step": 23752 + }, + { + "epoch": 2.0, + "learning_rate": 5.283878357764996e-06, + "loss": 1.2778, + "step": 23753 + }, + { + "epoch": 2.0, + "learning_rate": 5.283076839285599e-06, + "loss": 1.2244, + "step": 23754 + }, + { + "epoch": 2.0, + "learning_rate": 5.282275359778923e-06, + "loss": 1.1515, + "step": 23755 + }, + { + "epoch": 2.0, + "learning_rate": 5.281473919251607e-06, + "loss": 1.1337, + "step": 23756 + }, + { + "epoch": 2.0, + "learning_rate": 5.280672517710259e-06, + "loss": 1.0572, + "step": 23757 + }, + { + "epoch": 2.0, + "learning_rate": 5.279871155161507e-06, + "loss": 1.1062, + "step": 23758 + }, + { + "epoch": 2.0, + "learning_rate": 5.279069831611967e-06, + "loss": 1.1058, + "step": 23759 + }, + { + "epoch": 2.0, + "learning_rate": 5.278268547068265e-06, + "loss": 1.1917, + "step": 23760 + }, + { + "epoch": 2.0, + "learning_rate": 5.277467301537018e-06, + "loss": 1.1492, + "step": 23761 + }, + { + "epoch": 2.0, + "learning_rate": 5.2766660950248496e-06, + "loss": 1.0908, + "step": 23762 + }, + { + "epoch": 2.0, + "learning_rate": 5.275864927538377e-06, + "loss": 1.1143, + "step": 23763 + }, + { + "epoch": 2.0, + "learning_rate": 5.275063799084219e-06, + "loss": 1.1309, + "step": 23764 + }, + { + "epoch": 2.0, + "learning_rate": 5.274262709669e-06, + "loss": 1.0736, + "step": 23765 + }, + { + "epoch": 2.0, + "learning_rate": 5.273461659299328e-06, + "loss": 1.0758, + "step": 23766 + }, + { + "epoch": 2.0, + "learning_rate": 5.272660647981835e-06, + "loss": 1.094, + "step": 23767 + }, + { + "epoch": 2.0, + "learning_rate": 5.27185967572313e-06, + "loss": 1.0968, + "step": 23768 + }, + { + "epoch": 2.0, + "learning_rate": 5.271058742529833e-06, + "loss": 1.1249, + "step": 23769 + }, + { + "epoch": 2.0, + "learning_rate": 5.2702578484085635e-06, + "loss": 1.0825, + "step": 23770 + }, + { + "epoch": 2.0, + "learning_rate": 5.2694569933659355e-06, + "loss": 1.0713, + "step": 23771 + }, + { + "epoch": 2.0, + "learning_rate": 5.268656177408569e-06, + "loss": 1.0884, + "step": 23772 + }, + { + "epoch": 2.0, + "learning_rate": 5.267855400543079e-06, + "loss": 1.0753, + "step": 23773 + }, + { + "epoch": 2.0, + "learning_rate": 5.267054662776082e-06, + "loss": 1.1973, + "step": 23774 + }, + { + "epoch": 2.0, + "learning_rate": 5.266253964114194e-06, + "loss": 1.0372, + "step": 23775 + }, + { + "epoch": 2.0, + "learning_rate": 5.265453304564036e-06, + "loss": 1.0973, + "step": 23776 + }, + { + "epoch": 2.0, + "learning_rate": 5.264652684132209e-06, + "loss": 1.1551, + "step": 23777 + }, + { + "epoch": 2.0, + "learning_rate": 5.2638521028253445e-06, + "loss": 1.2006, + "step": 23778 + }, + { + "epoch": 2.0, + "learning_rate": 5.263051560650046e-06, + "loss": 1.0958, + "step": 23779 + }, + { + "epoch": 2.0, + "learning_rate": 5.262251057612933e-06, + "loss": 1.054, + "step": 23780 + }, + { + "epoch": 2.0, + "learning_rate": 5.261450593720616e-06, + "loss": 1.0861, + "step": 23781 + }, + { + "epoch": 2.0, + "learning_rate": 5.2606501689797155e-06, + "loss": 1.156, + "step": 23782 + }, + { + "epoch": 2.0, + "learning_rate": 5.259849783396832e-06, + "loss": 1.1457, + "step": 23783 + }, + { + "epoch": 2.0, + "learning_rate": 5.259049436978595e-06, + "loss": 1.144, + "step": 23784 + }, + { + "epoch": 2.0, + "learning_rate": 5.258249129731604e-06, + "loss": 1.081, + "step": 23785 + }, + { + "epoch": 2.0, + "learning_rate": 5.257448861662473e-06, + "loss": 1.105, + "step": 23786 + }, + { + "epoch": 2.0, + "learning_rate": 5.256648632777825e-06, + "loss": 1.03, + "step": 23787 + }, + { + "epoch": 2.0, + "learning_rate": 5.2558484430842605e-06, + "loss": 1.1026, + "step": 23788 + }, + { + "epoch": 2.0, + "learning_rate": 5.255048292588394e-06, + "loss": 1.1524, + "step": 23789 + }, + { + "epoch": 2.0, + "learning_rate": 5.254248181296837e-06, + "loss": 1.1128, + "step": 23790 + }, + { + "epoch": 2.0, + "learning_rate": 5.253448109216201e-06, + "loss": 1.0851, + "step": 23791 + }, + { + "epoch": 2.0, + "learning_rate": 5.252648076353095e-06, + "loss": 1.1572, + "step": 23792 + }, + { + "epoch": 2.0, + "learning_rate": 5.251848082714135e-06, + "loss": 1.1098, + "step": 23793 + }, + { + "epoch": 2.0, + "learning_rate": 5.251048128305918e-06, + "loss": 1.069, + "step": 23794 + }, + { + "epoch": 2.0, + "learning_rate": 5.2502482131350675e-06, + "loss": 1.1241, + "step": 23795 + }, + { + "epoch": 2.0, + "learning_rate": 5.2494483372081825e-06, + "loss": 1.0925, + "step": 23796 + }, + { + "epoch": 2.0, + "learning_rate": 5.248648500531877e-06, + "loss": 1.0347, + "step": 23797 + }, + { + "epoch": 2.0, + "learning_rate": 5.247848703112758e-06, + "loss": 1.1704, + "step": 23798 + }, + { + "epoch": 2.0, + "learning_rate": 5.247048944957434e-06, + "loss": 1.1286, + "step": 23799 + }, + { + "epoch": 2.0, + "learning_rate": 5.2462492260725114e-06, + "loss": 1.1191, + "step": 23800 + }, + { + "epoch": 2.0, + "learning_rate": 5.245449546464601e-06, + "loss": 1.185, + "step": 23801 + }, + { + "epoch": 2.0, + "learning_rate": 5.244649906140307e-06, + "loss": 1.0418, + "step": 23802 + }, + { + "epoch": 2.0, + "learning_rate": 5.243850305106237e-06, + "loss": 1.1249, + "step": 23803 + }, + { + "epoch": 2.0, + "learning_rate": 5.243050743369003e-06, + "loss": 1.0966, + "step": 23804 + }, + { + "epoch": 2.0, + "learning_rate": 5.2422512209351975e-06, + "loss": 1.1162, + "step": 23805 + }, + { + "epoch": 2.0, + "learning_rate": 5.241451737811443e-06, + "loss": 1.037, + "step": 23806 + }, + { + "epoch": 2.0, + "learning_rate": 5.240652294004334e-06, + "loss": 1.1051, + "step": 23807 + }, + { + "epoch": 2.0, + "learning_rate": 5.239852889520478e-06, + "loss": 1.1218, + "step": 23808 + }, + { + "epoch": 2.0, + "learning_rate": 5.239053524366481e-06, + "loss": 1.095, + "step": 23809 + }, + { + "epoch": 2.0, + "learning_rate": 5.238254198548948e-06, + "loss": 1.1697, + "step": 23810 + }, + { + "epoch": 2.0, + "learning_rate": 5.237454912074482e-06, + "loss": 1.08, + "step": 23811 + }, + { + "epoch": 2.0, + "learning_rate": 5.236655664949688e-06, + "loss": 1.1304, + "step": 23812 + }, + { + "epoch": 2.0, + "learning_rate": 5.23585645718117e-06, + "loss": 1.1786, + "step": 23813 + }, + { + "epoch": 2.0, + "learning_rate": 5.235057288775529e-06, + "loss": 1.1244, + "step": 23814 + }, + { + "epoch": 2.01, + "learning_rate": 5.234258159739376e-06, + "loss": 1.035, + "step": 23815 + }, + { + "epoch": 2.01, + "learning_rate": 5.233459070079298e-06, + "loss": 1.1632, + "step": 23816 + }, + { + "epoch": 2.01, + "learning_rate": 5.232660019801914e-06, + "loss": 1.0995, + "step": 23817 + }, + { + "epoch": 2.01, + "learning_rate": 5.231861008913816e-06, + "loss": 1.0687, + "step": 23818 + }, + { + "epoch": 2.01, + "learning_rate": 5.231062037421608e-06, + "loss": 1.0205, + "step": 23819 + }, + { + "epoch": 2.01, + "learning_rate": 5.230263105331892e-06, + "loss": 1.2134, + "step": 23820 + }, + { + "epoch": 2.01, + "learning_rate": 5.229464212651267e-06, + "loss": 1.1426, + "step": 23821 + }, + { + "epoch": 2.01, + "learning_rate": 5.228665359386338e-06, + "loss": 1.0339, + "step": 23822 + }, + { + "epoch": 2.01, + "learning_rate": 5.227866545543701e-06, + "loss": 1.0838, + "step": 23823 + }, + { + "epoch": 2.01, + "learning_rate": 5.227067771129959e-06, + "loss": 1.122, + "step": 23824 + }, + { + "epoch": 2.01, + "learning_rate": 5.22626903615171e-06, + "loss": 1.1157, + "step": 23825 + }, + { + "epoch": 2.01, + "learning_rate": 5.225470340615557e-06, + "loss": 1.1088, + "step": 23826 + }, + { + "epoch": 2.01, + "learning_rate": 5.22467168452809e-06, + "loss": 1.0637, + "step": 23827 + }, + { + "epoch": 2.01, + "learning_rate": 5.22387306789592e-06, + "loss": 1.0762, + "step": 23828 + }, + { + "epoch": 2.01, + "learning_rate": 5.223074490725637e-06, + "loss": 1.084, + "step": 23829 + }, + { + "epoch": 2.01, + "learning_rate": 5.222275953023841e-06, + "loss": 1.1436, + "step": 23830 + }, + { + "epoch": 2.01, + "learning_rate": 5.221477454797129e-06, + "loss": 1.1579, + "step": 23831 + }, + { + "epoch": 2.01, + "learning_rate": 5.220678996052105e-06, + "loss": 1.1923, + "step": 23832 + }, + { + "epoch": 2.01, + "learning_rate": 5.219880576795351e-06, + "loss": 1.1694, + "step": 23833 + }, + { + "epoch": 2.01, + "learning_rate": 5.219082197033483e-06, + "loss": 1.0962, + "step": 23834 + }, + { + "epoch": 2.01, + "learning_rate": 5.218283856773083e-06, + "loss": 1.2223, + "step": 23835 + }, + { + "epoch": 2.01, + "learning_rate": 5.217485556020749e-06, + "loss": 1.2042, + "step": 23836 + }, + { + "epoch": 2.01, + "learning_rate": 5.216687294783087e-06, + "loss": 1.0702, + "step": 23837 + }, + { + "epoch": 2.01, + "learning_rate": 5.215889073066683e-06, + "loss": 1.1773, + "step": 23838 + }, + { + "epoch": 2.01, + "learning_rate": 5.2150908908781335e-06, + "loss": 1.0982, + "step": 23839 + }, + { + "epoch": 2.01, + "learning_rate": 5.214292748224035e-06, + "loss": 1.0753, + "step": 23840 + }, + { + "epoch": 2.01, + "learning_rate": 5.21349464511098e-06, + "loss": 1.112, + "step": 23841 + }, + { + "epoch": 2.01, + "learning_rate": 5.212696581545566e-06, + "loss": 1.1966, + "step": 23842 + }, + { + "epoch": 2.01, + "learning_rate": 5.211898557534387e-06, + "loss": 1.1253, + "step": 23843 + }, + { + "epoch": 2.01, + "learning_rate": 5.2111005730840285e-06, + "loss": 1.1388, + "step": 23844 + }, + { + "epoch": 2.01, + "learning_rate": 5.210302628201096e-06, + "loss": 1.1806, + "step": 23845 + }, + { + "epoch": 2.01, + "learning_rate": 5.209504722892174e-06, + "loss": 1.1028, + "step": 23846 + }, + { + "epoch": 2.01, + "learning_rate": 5.2087068571638555e-06, + "loss": 1.1013, + "step": 23847 + }, + { + "epoch": 2.01, + "learning_rate": 5.207909031022735e-06, + "loss": 1.1296, + "step": 23848 + }, + { + "epoch": 2.01, + "learning_rate": 5.207111244475403e-06, + "loss": 1.1366, + "step": 23849 + }, + { + "epoch": 2.01, + "learning_rate": 5.206313497528453e-06, + "loss": 1.0582, + "step": 23850 + }, + { + "epoch": 2.01, + "learning_rate": 5.205515790188474e-06, + "loss": 1.0809, + "step": 23851 + }, + { + "epoch": 2.01, + "learning_rate": 5.204718122462058e-06, + "loss": 1.0973, + "step": 23852 + }, + { + "epoch": 2.01, + "learning_rate": 5.2039204943557965e-06, + "loss": 1.083, + "step": 23853 + }, + { + "epoch": 2.01, + "learning_rate": 5.203122905876283e-06, + "loss": 1.0923, + "step": 23854 + }, + { + "epoch": 2.01, + "learning_rate": 5.202325357030094e-06, + "loss": 1.1058, + "step": 23855 + }, + { + "epoch": 2.01, + "learning_rate": 5.201527847823837e-06, + "loss": 1.1876, + "step": 23856 + }, + { + "epoch": 2.01, + "learning_rate": 5.2007303782640895e-06, + "loss": 1.0619, + "step": 23857 + }, + { + "epoch": 2.01, + "learning_rate": 5.1999329483574425e-06, + "loss": 1.1661, + "step": 23858 + }, + { + "epoch": 2.01, + "learning_rate": 5.199135558110485e-06, + "loss": 1.1518, + "step": 23859 + }, + { + "epoch": 2.01, + "learning_rate": 5.198338207529807e-06, + "loss": 1.116, + "step": 23860 + }, + { + "epoch": 2.01, + "learning_rate": 5.197540896621996e-06, + "loss": 1.1692, + "step": 23861 + }, + { + "epoch": 2.01, + "learning_rate": 5.196743625393637e-06, + "loss": 1.0692, + "step": 23862 + }, + { + "epoch": 2.01, + "learning_rate": 5.195946393851321e-06, + "loss": 1.0913, + "step": 23863 + }, + { + "epoch": 2.01, + "learning_rate": 5.195149202001634e-06, + "loss": 1.1024, + "step": 23864 + }, + { + "epoch": 2.01, + "learning_rate": 5.1943520498511646e-06, + "loss": 1.1204, + "step": 23865 + }, + { + "epoch": 2.01, + "learning_rate": 5.1935549374064895e-06, + "loss": 1.1726, + "step": 23866 + }, + { + "epoch": 2.01, + "learning_rate": 5.1927578646742095e-06, + "loss": 1.1183, + "step": 23867 + }, + { + "epoch": 2.01, + "learning_rate": 5.191960831660899e-06, + "loss": 1.1445, + "step": 23868 + }, + { + "epoch": 2.01, + "learning_rate": 5.191163838373146e-06, + "loss": 1.1445, + "step": 23869 + }, + { + "epoch": 2.01, + "learning_rate": 5.190366884817537e-06, + "loss": 1.1259, + "step": 23870 + }, + { + "epoch": 2.01, + "learning_rate": 5.1895699710006555e-06, + "loss": 1.081, + "step": 23871 + }, + { + "epoch": 2.01, + "learning_rate": 5.188773096929088e-06, + "loss": 1.1963, + "step": 23872 + }, + { + "epoch": 2.01, + "learning_rate": 5.187976262609415e-06, + "loss": 1.1512, + "step": 23873 + }, + { + "epoch": 2.01, + "learning_rate": 5.187179468048222e-06, + "loss": 1.0911, + "step": 23874 + }, + { + "epoch": 2.01, + "learning_rate": 5.186382713252094e-06, + "loss": 1.1692, + "step": 23875 + }, + { + "epoch": 2.01, + "learning_rate": 5.185585998227616e-06, + "loss": 0.9936, + "step": 23876 + }, + { + "epoch": 2.01, + "learning_rate": 5.18478932298136e-06, + "loss": 1.1819, + "step": 23877 + }, + { + "epoch": 2.01, + "learning_rate": 5.183992687519922e-06, + "loss": 1.1801, + "step": 23878 + }, + { + "epoch": 2.01, + "learning_rate": 5.183196091849875e-06, + "loss": 1.1524, + "step": 23879 + }, + { + "epoch": 2.01, + "learning_rate": 5.182399535977802e-06, + "loss": 1.066, + "step": 23880 + }, + { + "epoch": 2.01, + "learning_rate": 5.181603019910288e-06, + "loss": 1.1342, + "step": 23881 + }, + { + "epoch": 2.01, + "learning_rate": 5.180806543653911e-06, + "loss": 1.1297, + "step": 23882 + }, + { + "epoch": 2.01, + "learning_rate": 5.180010107215253e-06, + "loss": 1.0104, + "step": 23883 + }, + { + "epoch": 2.01, + "learning_rate": 5.179213710600897e-06, + "loss": 1.1914, + "step": 23884 + }, + { + "epoch": 2.01, + "learning_rate": 5.178417353817413e-06, + "loss": 1.0925, + "step": 23885 + }, + { + "epoch": 2.01, + "learning_rate": 5.177621036871392e-06, + "loss": 1.1444, + "step": 23886 + }, + { + "epoch": 2.01, + "learning_rate": 5.176824759769414e-06, + "loss": 1.0901, + "step": 23887 + }, + { + "epoch": 2.01, + "learning_rate": 5.176028522518044e-06, + "loss": 1.0924, + "step": 23888 + }, + { + "epoch": 2.01, + "learning_rate": 5.175232325123879e-06, + "loss": 0.9567, + "step": 23889 + }, + { + "epoch": 2.01, + "learning_rate": 5.174436167593484e-06, + "loss": 1.0477, + "step": 23890 + }, + { + "epoch": 2.01, + "learning_rate": 5.173640049933443e-06, + "loss": 1.1205, + "step": 23891 + }, + { + "epoch": 2.01, + "learning_rate": 5.172843972150332e-06, + "loss": 1.1203, + "step": 23892 + }, + { + "epoch": 2.01, + "learning_rate": 5.172047934250732e-06, + "loss": 1.0814, + "step": 23893 + }, + { + "epoch": 2.01, + "learning_rate": 5.1712519362412105e-06, + "loss": 1.0814, + "step": 23894 + }, + { + "epoch": 2.01, + "learning_rate": 5.170455978128358e-06, + "loss": 1.0272, + "step": 23895 + }, + { + "epoch": 2.01, + "learning_rate": 5.169660059918741e-06, + "loss": 1.144, + "step": 23896 + }, + { + "epoch": 2.01, + "learning_rate": 5.168864181618938e-06, + "loss": 1.0937, + "step": 23897 + }, + { + "epoch": 2.01, + "learning_rate": 5.168068343235524e-06, + "loss": 1.1838, + "step": 23898 + }, + { + "epoch": 2.01, + "learning_rate": 5.1672725447750775e-06, + "loss": 1.1357, + "step": 23899 + }, + { + "epoch": 2.01, + "learning_rate": 5.16647678624417e-06, + "loss": 1.2078, + "step": 23900 + }, + { + "epoch": 2.01, + "learning_rate": 5.16568106764938e-06, + "loss": 1.1013, + "step": 23901 + }, + { + "epoch": 2.01, + "learning_rate": 5.164885388997279e-06, + "loss": 1.1477, + "step": 23902 + }, + { + "epoch": 2.01, + "learning_rate": 5.164089750294443e-06, + "loss": 1.1133, + "step": 23903 + }, + { + "epoch": 2.01, + "learning_rate": 5.163294151547449e-06, + "loss": 1.1616, + "step": 23904 + }, + { + "epoch": 2.01, + "learning_rate": 5.162498592762859e-06, + "loss": 1.0614, + "step": 23905 + }, + { + "epoch": 2.01, + "learning_rate": 5.161703073947262e-06, + "loss": 1.144, + "step": 23906 + }, + { + "epoch": 2.01, + "learning_rate": 5.160907595107219e-06, + "loss": 1.0159, + "step": 23907 + }, + { + "epoch": 2.01, + "learning_rate": 5.160112156249304e-06, + "loss": 1.1841, + "step": 23908 + }, + { + "epoch": 2.01, + "learning_rate": 5.159316757380093e-06, + "loss": 1.2142, + "step": 23909 + }, + { + "epoch": 2.01, + "learning_rate": 5.158521398506155e-06, + "loss": 1.1255, + "step": 23910 + }, + { + "epoch": 2.01, + "learning_rate": 5.157726079634064e-06, + "loss": 1.1553, + "step": 23911 + }, + { + "epoch": 2.01, + "learning_rate": 5.15693080077039e-06, + "loss": 1.1484, + "step": 23912 + }, + { + "epoch": 2.01, + "learning_rate": 5.1561355619217025e-06, + "loss": 1.1353, + "step": 23913 + }, + { + "epoch": 2.01, + "learning_rate": 5.155340363094574e-06, + "loss": 1.0938, + "step": 23914 + }, + { + "epoch": 2.01, + "learning_rate": 5.154545204295578e-06, + "loss": 1.1319, + "step": 23915 + }, + { + "epoch": 2.01, + "learning_rate": 5.153750085531272e-06, + "loss": 1.0864, + "step": 23916 + }, + { + "epoch": 2.01, + "learning_rate": 5.152955006808242e-06, + "loss": 1.1541, + "step": 23917 + }, + { + "epoch": 2.01, + "learning_rate": 5.152159968133044e-06, + "loss": 1.0699, + "step": 23918 + }, + { + "epoch": 2.01, + "learning_rate": 5.151364969512254e-06, + "loss": 1.1175, + "step": 23919 + }, + { + "epoch": 2.01, + "learning_rate": 5.150570010952436e-06, + "loss": 1.2147, + "step": 23920 + }, + { + "epoch": 2.01, + "learning_rate": 5.149775092460162e-06, + "loss": 1.0873, + "step": 23921 + }, + { + "epoch": 2.01, + "learning_rate": 5.148980214041999e-06, + "loss": 1.1023, + "step": 23922 + }, + { + "epoch": 2.01, + "learning_rate": 5.148185375704513e-06, + "loss": 1.1634, + "step": 23923 + }, + { + "epoch": 2.01, + "learning_rate": 5.147390577454272e-06, + "loss": 1.1666, + "step": 23924 + }, + { + "epoch": 2.01, + "learning_rate": 5.146595819297844e-06, + "loss": 1.12, + "step": 23925 + }, + { + "epoch": 2.01, + "learning_rate": 5.145801101241796e-06, + "loss": 1.1316, + "step": 23926 + }, + { + "epoch": 2.01, + "learning_rate": 5.145006423292688e-06, + "loss": 1.1295, + "step": 23927 + }, + { + "epoch": 2.01, + "learning_rate": 5.144211785457098e-06, + "loss": 1.0969, + "step": 23928 + }, + { + "epoch": 2.01, + "learning_rate": 5.143417187741579e-06, + "loss": 1.2199, + "step": 23929 + }, + { + "epoch": 2.01, + "learning_rate": 5.142622630152703e-06, + "loss": 1.1148, + "step": 23930 + }, + { + "epoch": 2.01, + "learning_rate": 5.141828112697032e-06, + "loss": 1.1618, + "step": 23931 + }, + { + "epoch": 2.01, + "learning_rate": 5.141033635381133e-06, + "loss": 1.1205, + "step": 23932 + }, + { + "epoch": 2.01, + "learning_rate": 5.14023919821157e-06, + "loss": 1.123, + "step": 23933 + }, + { + "epoch": 2.02, + "learning_rate": 5.139444801194909e-06, + "loss": 1.0656, + "step": 23934 + }, + { + "epoch": 2.02, + "learning_rate": 5.138650444337701e-06, + "loss": 1.0933, + "step": 23935 + }, + { + "epoch": 2.02, + "learning_rate": 5.137856127646525e-06, + "loss": 1.0155, + "step": 23936 + }, + { + "epoch": 2.02, + "learning_rate": 5.13706185112794e-06, + "loss": 1.1585, + "step": 23937 + }, + { + "epoch": 2.02, + "learning_rate": 5.1362676147885e-06, + "loss": 1.1902, + "step": 23938 + }, + { + "epoch": 2.02, + "learning_rate": 5.135473418634782e-06, + "loss": 0.9914, + "step": 23939 + }, + { + "epoch": 2.02, + "learning_rate": 5.134679262673334e-06, + "loss": 1.0897, + "step": 23940 + }, + { + "epoch": 2.02, + "learning_rate": 5.133885146910724e-06, + "loss": 1.1346, + "step": 23941 + }, + { + "epoch": 2.02, + "learning_rate": 5.133091071353511e-06, + "loss": 1.1255, + "step": 23942 + }, + { + "epoch": 2.02, + "learning_rate": 5.132297036008263e-06, + "loss": 1.098, + "step": 23943 + }, + { + "epoch": 2.02, + "learning_rate": 5.131503040881525e-06, + "loss": 1.1406, + "step": 23944 + }, + { + "epoch": 2.02, + "learning_rate": 5.130709085979876e-06, + "loss": 1.1348, + "step": 23945 + }, + { + "epoch": 2.02, + "learning_rate": 5.129915171309862e-06, + "loss": 1.1049, + "step": 23946 + }, + { + "epoch": 2.02, + "learning_rate": 5.129121296878048e-06, + "loss": 1.0999, + "step": 23947 + }, + { + "epoch": 2.02, + "learning_rate": 5.128327462690994e-06, + "loss": 1.1219, + "step": 23948 + }, + { + "epoch": 2.02, + "learning_rate": 5.1275336687552554e-06, + "loss": 1.12, + "step": 23949 + }, + { + "epoch": 2.02, + "learning_rate": 5.126739915077395e-06, + "loss": 1.1314, + "step": 23950 + }, + { + "epoch": 2.02, + "learning_rate": 5.125946201663967e-06, + "loss": 1.2083, + "step": 23951 + }, + { + "epoch": 2.02, + "learning_rate": 5.125152528521533e-06, + "loss": 1.067, + "step": 23952 + }, + { + "epoch": 2.02, + "learning_rate": 5.124358895656649e-06, + "loss": 1.1268, + "step": 23953 + }, + { + "epoch": 2.02, + "learning_rate": 5.123565303075875e-06, + "loss": 1.1589, + "step": 23954 + }, + { + "epoch": 2.02, + "learning_rate": 5.122771750785759e-06, + "loss": 1.1304, + "step": 23955 + }, + { + "epoch": 2.02, + "learning_rate": 5.1219782387928706e-06, + "loss": 1.1493, + "step": 23956 + }, + { + "epoch": 2.02, + "learning_rate": 5.121184767103755e-06, + "loss": 1.0919, + "step": 23957 + }, + { + "epoch": 2.02, + "learning_rate": 5.120391335724973e-06, + "loss": 0.9666, + "step": 23958 + }, + { + "epoch": 2.02, + "learning_rate": 5.119597944663077e-06, + "loss": 1.2153, + "step": 23959 + }, + { + "epoch": 2.02, + "learning_rate": 5.118804593924626e-06, + "loss": 1.1495, + "step": 23960 + }, + { + "epoch": 2.02, + "learning_rate": 5.1180112835161754e-06, + "loss": 1.1293, + "step": 23961 + }, + { + "epoch": 2.02, + "learning_rate": 5.117218013444277e-06, + "loss": 1.0578, + "step": 23962 + }, + { + "epoch": 2.02, + "learning_rate": 5.116424783715486e-06, + "loss": 1.1178, + "step": 23963 + }, + { + "epoch": 2.02, + "learning_rate": 5.115631594336357e-06, + "loss": 1.1404, + "step": 23964 + }, + { + "epoch": 2.02, + "learning_rate": 5.1148384453134455e-06, + "loss": 1.1674, + "step": 23965 + }, + { + "epoch": 2.02, + "learning_rate": 5.114045336653295e-06, + "loss": 1.2077, + "step": 23966 + }, + { + "epoch": 2.02, + "learning_rate": 5.113252268362474e-06, + "loss": 1.1475, + "step": 23967 + }, + { + "epoch": 2.02, + "learning_rate": 5.1124592404475215e-06, + "loss": 1.1438, + "step": 23968 + }, + { + "epoch": 2.02, + "learning_rate": 5.111666252914995e-06, + "loss": 1.001, + "step": 23969 + }, + { + "epoch": 2.02, + "learning_rate": 5.1108733057714485e-06, + "loss": 1.0668, + "step": 23970 + }, + { + "epoch": 2.02, + "learning_rate": 5.110080399023429e-06, + "loss": 1.1002, + "step": 23971 + }, + { + "epoch": 2.02, + "learning_rate": 5.109287532677491e-06, + "loss": 1.1551, + "step": 23972 + }, + { + "epoch": 2.02, + "learning_rate": 5.108494706740186e-06, + "loss": 1.1293, + "step": 23973 + }, + { + "epoch": 2.02, + "learning_rate": 5.107701921218063e-06, + "loss": 1.2106, + "step": 23974 + }, + { + "epoch": 2.02, + "learning_rate": 5.106909176117671e-06, + "loss": 1.2495, + "step": 23975 + }, + { + "epoch": 2.02, + "learning_rate": 5.106116471445566e-06, + "loss": 1.1626, + "step": 23976 + }, + { + "epoch": 2.02, + "learning_rate": 5.105323807208284e-06, + "loss": 1.0557, + "step": 23977 + }, + { + "epoch": 2.02, + "learning_rate": 5.104531183412393e-06, + "loss": 1.0952, + "step": 23978 + }, + { + "epoch": 2.02, + "learning_rate": 5.103738600064427e-06, + "loss": 1.1489, + "step": 23979 + }, + { + "epoch": 2.02, + "learning_rate": 5.10294605717094e-06, + "loss": 1.0927, + "step": 23980 + }, + { + "epoch": 2.02, + "learning_rate": 5.1021535547384805e-06, + "loss": 1.2388, + "step": 23981 + }, + { + "epoch": 2.02, + "learning_rate": 5.101361092773594e-06, + "loss": 1.0743, + "step": 23982 + }, + { + "epoch": 2.02, + "learning_rate": 5.100568671282833e-06, + "loss": 1.1028, + "step": 23983 + }, + { + "epoch": 2.02, + "learning_rate": 5.099776290272742e-06, + "loss": 1.0875, + "step": 23984 + }, + { + "epoch": 2.02, + "learning_rate": 5.098983949749863e-06, + "loss": 1.2085, + "step": 23985 + }, + { + "epoch": 2.02, + "learning_rate": 5.09819164972075e-06, + "loss": 1.1211, + "step": 23986 + }, + { + "epoch": 2.02, + "learning_rate": 5.097399390191948e-06, + "loss": 1.0973, + "step": 23987 + }, + { + "epoch": 2.02, + "learning_rate": 5.0966071711699965e-06, + "loss": 1.1581, + "step": 23988 + }, + { + "epoch": 2.02, + "learning_rate": 5.095814992661452e-06, + "loss": 1.0947, + "step": 23989 + }, + { + "epoch": 2.02, + "learning_rate": 5.0950228546728495e-06, + "loss": 1.1037, + "step": 23990 + }, + { + "epoch": 2.02, + "learning_rate": 5.094230757210738e-06, + "loss": 1.099, + "step": 23991 + }, + { + "epoch": 2.02, + "learning_rate": 5.0934387002816634e-06, + "loss": 1.0716, + "step": 23992 + }, + { + "epoch": 2.02, + "learning_rate": 5.0926466838921715e-06, + "loss": 1.2256, + "step": 23993 + }, + { + "epoch": 2.02, + "learning_rate": 5.091854708048796e-06, + "loss": 1.0518, + "step": 23994 + }, + { + "epoch": 2.02, + "learning_rate": 5.091062772758096e-06, + "loss": 1.1396, + "step": 23995 + }, + { + "epoch": 2.02, + "learning_rate": 5.090270878026601e-06, + "loss": 1.1081, + "step": 23996 + }, + { + "epoch": 2.02, + "learning_rate": 5.089479023860859e-06, + "loss": 1.0311, + "step": 23997 + }, + { + "epoch": 2.02, + "learning_rate": 5.088687210267414e-06, + "loss": 1.176, + "step": 23998 + }, + { + "epoch": 2.02, + "learning_rate": 5.0878954372528075e-06, + "loss": 1.1499, + "step": 23999 + }, + { + "epoch": 2.02, + "learning_rate": 5.087103704823579e-06, + "loss": 1.0365, + "step": 24000 + }, + { + "epoch": 2.02, + "learning_rate": 5.086312012986274e-06, + "loss": 1.2388, + "step": 24001 + }, + { + "epoch": 2.02, + "learning_rate": 5.0855203617474315e-06, + "loss": 1.1461, + "step": 24002 + }, + { + "epoch": 2.02, + "learning_rate": 5.084728751113593e-06, + "loss": 1.129, + "step": 24003 + }, + { + "epoch": 2.02, + "learning_rate": 5.083937181091301e-06, + "loss": 1.0948, + "step": 24004 + }, + { + "epoch": 2.02, + "learning_rate": 5.083145651687087e-06, + "loss": 1.0507, + "step": 24005 + }, + { + "epoch": 2.02, + "learning_rate": 5.082354162907504e-06, + "loss": 1.0909, + "step": 24006 + }, + { + "epoch": 2.02, + "learning_rate": 5.081562714759081e-06, + "loss": 1.1778, + "step": 24007 + }, + { + "epoch": 2.02, + "learning_rate": 5.080771307248362e-06, + "loss": 1.0571, + "step": 24008 + }, + { + "epoch": 2.02, + "learning_rate": 5.079979940381882e-06, + "loss": 1.0647, + "step": 24009 + }, + { + "epoch": 2.02, + "learning_rate": 5.079188614166186e-06, + "loss": 1.1415, + "step": 24010 + }, + { + "epoch": 2.02, + "learning_rate": 5.078397328607807e-06, + "loss": 1.0054, + "step": 24011 + }, + { + "epoch": 2.02, + "learning_rate": 5.077606083713284e-06, + "loss": 1.1016, + "step": 24012 + }, + { + "epoch": 2.02, + "learning_rate": 5.0768148794891544e-06, + "loss": 1.1515, + "step": 24013 + }, + { + "epoch": 2.02, + "learning_rate": 5.076023715941958e-06, + "loss": 1.2491, + "step": 24014 + }, + { + "epoch": 2.02, + "learning_rate": 5.075232593078232e-06, + "loss": 1.2365, + "step": 24015 + }, + { + "epoch": 2.02, + "learning_rate": 5.074441510904502e-06, + "loss": 1.0581, + "step": 24016 + }, + { + "epoch": 2.02, + "learning_rate": 5.0736504694273205e-06, + "loss": 1.211, + "step": 24017 + }, + { + "epoch": 2.02, + "learning_rate": 5.072859468653213e-06, + "loss": 1.1567, + "step": 24018 + }, + { + "epoch": 2.02, + "learning_rate": 5.0720685085887165e-06, + "loss": 1.141, + "step": 24019 + }, + { + "epoch": 2.02, + "learning_rate": 5.071277589240367e-06, + "loss": 1.1381, + "step": 24020 + }, + { + "epoch": 2.02, + "learning_rate": 5.0704867106147e-06, + "loss": 1.1713, + "step": 24021 + }, + { + "epoch": 2.02, + "learning_rate": 5.06969587271825e-06, + "loss": 1.1207, + "step": 24022 + }, + { + "epoch": 2.02, + "learning_rate": 5.068905075557549e-06, + "loss": 1.132, + "step": 24023 + }, + { + "epoch": 2.02, + "learning_rate": 5.0681143191391325e-06, + "loss": 1.1131, + "step": 24024 + }, + { + "epoch": 2.02, + "learning_rate": 5.067323603469535e-06, + "loss": 1.0827, + "step": 24025 + }, + { + "epoch": 2.02, + "learning_rate": 5.0665329285552925e-06, + "loss": 1.2487, + "step": 24026 + }, + { + "epoch": 2.02, + "learning_rate": 5.065742294402924e-06, + "loss": 1.0544, + "step": 24027 + }, + { + "epoch": 2.02, + "learning_rate": 5.064951701018983e-06, + "loss": 1.1759, + "step": 24028 + }, + { + "epoch": 2.02, + "learning_rate": 5.064161148409984e-06, + "loss": 1.1879, + "step": 24029 + }, + { + "epoch": 2.02, + "learning_rate": 5.063370636582467e-06, + "loss": 1.1171, + "step": 24030 + }, + { + "epoch": 2.02, + "learning_rate": 5.06258016554296e-06, + "loss": 1.1268, + "step": 24031 + }, + { + "epoch": 2.02, + "learning_rate": 5.061789735297997e-06, + "loss": 1.0591, + "step": 24032 + }, + { + "epoch": 2.02, + "learning_rate": 5.060999345854107e-06, + "loss": 1.2051, + "step": 24033 + }, + { + "epoch": 2.02, + "learning_rate": 5.0602089972178245e-06, + "loss": 1.0555, + "step": 24034 + }, + { + "epoch": 2.02, + "learning_rate": 5.05941868939567e-06, + "loss": 1.1536, + "step": 24035 + }, + { + "epoch": 2.02, + "learning_rate": 5.058628422394183e-06, + "loss": 1.1272, + "step": 24036 + }, + { + "epoch": 2.02, + "learning_rate": 5.057838196219892e-06, + "loss": 1.1232, + "step": 24037 + }, + { + "epoch": 2.02, + "learning_rate": 5.057048010879317e-06, + "loss": 1.0612, + "step": 24038 + }, + { + "epoch": 2.02, + "learning_rate": 5.056257866378999e-06, + "loss": 1.1737, + "step": 24039 + }, + { + "epoch": 2.02, + "learning_rate": 5.055467762725459e-06, + "loss": 1.16, + "step": 24040 + }, + { + "epoch": 2.02, + "learning_rate": 5.0546776999252265e-06, + "loss": 1.1019, + "step": 24041 + }, + { + "epoch": 2.02, + "learning_rate": 5.05388767798483e-06, + "loss": 1.0462, + "step": 24042 + }, + { + "epoch": 2.02, + "learning_rate": 5.053097696910799e-06, + "loss": 1.1073, + "step": 24043 + }, + { + "epoch": 2.02, + "learning_rate": 5.0523077567096515e-06, + "loss": 1.1316, + "step": 24044 + }, + { + "epoch": 2.02, + "learning_rate": 5.051517857387927e-06, + "loss": 1.1011, + "step": 24045 + }, + { + "epoch": 2.02, + "learning_rate": 5.050727998952143e-06, + "loss": 1.0832, + "step": 24046 + }, + { + "epoch": 2.02, + "learning_rate": 5.049938181408829e-06, + "loss": 1.108, + "step": 24047 + }, + { + "epoch": 2.02, + "learning_rate": 5.049148404764508e-06, + "loss": 1.1012, + "step": 24048 + }, + { + "epoch": 2.02, + "learning_rate": 5.048358669025708e-06, + "loss": 1.1569, + "step": 24049 + }, + { + "epoch": 2.02, + "learning_rate": 5.0475689741989546e-06, + "loss": 1.1802, + "step": 24050 + }, + { + "epoch": 2.02, + "learning_rate": 5.04677932029077e-06, + "loss": 1.2066, + "step": 24051 + }, + { + "epoch": 2.02, + "learning_rate": 5.04598970730768e-06, + "loss": 1.1823, + "step": 24052 + }, + { + "epoch": 2.03, + "learning_rate": 5.045200135256209e-06, + "loss": 1.11, + "step": 24053 + }, + { + "epoch": 2.03, + "learning_rate": 5.0444106041428846e-06, + "loss": 1.0944, + "step": 24054 + }, + { + "epoch": 2.03, + "learning_rate": 5.043621113974218e-06, + "loss": 1.2154, + "step": 24055 + }, + { + "epoch": 2.03, + "learning_rate": 5.042831664756747e-06, + "loss": 1.1118, + "step": 24056 + }, + { + "epoch": 2.03, + "learning_rate": 5.042042256496983e-06, + "loss": 1.1222, + "step": 24057 + }, + { + "epoch": 2.03, + "learning_rate": 5.041252889201453e-06, + "loss": 1.1445, + "step": 24058 + }, + { + "epoch": 2.03, + "learning_rate": 5.040463562876678e-06, + "loss": 1.044, + "step": 24059 + }, + { + "epoch": 2.03, + "learning_rate": 5.039674277529182e-06, + "loss": 1.0858, + "step": 24060 + }, + { + "epoch": 2.03, + "learning_rate": 5.0388850331654835e-06, + "loss": 1.0713, + "step": 24061 + }, + { + "epoch": 2.03, + "learning_rate": 5.038095829792105e-06, + "loss": 1.1263, + "step": 24062 + }, + { + "epoch": 2.03, + "learning_rate": 5.037306667415567e-06, + "loss": 1.1357, + "step": 24063 + }, + { + "epoch": 2.03, + "learning_rate": 5.036517546042389e-06, + "loss": 1.0995, + "step": 24064 + }, + { + "epoch": 2.03, + "learning_rate": 5.035728465679095e-06, + "loss": 1.1563, + "step": 24065 + }, + { + "epoch": 2.03, + "learning_rate": 5.034939426332195e-06, + "loss": 1.0554, + "step": 24066 + }, + { + "epoch": 2.03, + "learning_rate": 5.03415042800822e-06, + "loss": 1.1307, + "step": 24067 + }, + { + "epoch": 2.03, + "learning_rate": 5.033361470713681e-06, + "loss": 1.1322, + "step": 24068 + }, + { + "epoch": 2.03, + "learning_rate": 5.0325725544551e-06, + "loss": 1.0916, + "step": 24069 + }, + { + "epoch": 2.03, + "learning_rate": 5.0317836792389925e-06, + "loss": 1.1113, + "step": 24070 + }, + { + "epoch": 2.03, + "learning_rate": 5.03099484507188e-06, + "loss": 1.1859, + "step": 24071 + }, + { + "epoch": 2.03, + "learning_rate": 5.030206051960276e-06, + "loss": 1.1586, + "step": 24072 + }, + { + "epoch": 2.03, + "learning_rate": 5.029417299910702e-06, + "loss": 1.165, + "step": 24073 + }, + { + "epoch": 2.03, + "learning_rate": 5.028628588929672e-06, + "loss": 1.0495, + "step": 24074 + }, + { + "epoch": 2.03, + "learning_rate": 5.027839919023705e-06, + "loss": 1.1958, + "step": 24075 + }, + { + "epoch": 2.03, + "learning_rate": 5.0270512901993185e-06, + "loss": 1.1443, + "step": 24076 + }, + { + "epoch": 2.03, + "learning_rate": 5.026262702463019e-06, + "loss": 1.0597, + "step": 24077 + }, + { + "epoch": 2.03, + "learning_rate": 5.025474155821335e-06, + "loss": 1.1823, + "step": 24078 + }, + { + "epoch": 2.03, + "learning_rate": 5.024685650280774e-06, + "loss": 1.1143, + "step": 24079 + }, + { + "epoch": 2.03, + "learning_rate": 5.02389718584785e-06, + "loss": 1.1078, + "step": 24080 + }, + { + "epoch": 2.03, + "learning_rate": 5.023108762529082e-06, + "loss": 1.1521, + "step": 24081 + }, + { + "epoch": 2.03, + "learning_rate": 5.0223203803309815e-06, + "loss": 1.0409, + "step": 24082 + }, + { + "epoch": 2.03, + "learning_rate": 5.021532039260063e-06, + "loss": 1.1985, + "step": 24083 + }, + { + "epoch": 2.03, + "learning_rate": 5.0207437393228445e-06, + "loss": 1.0786, + "step": 24084 + }, + { + "epoch": 2.03, + "learning_rate": 5.019955480525827e-06, + "loss": 1.1629, + "step": 24085 + }, + { + "epoch": 2.03, + "learning_rate": 5.019167262875534e-06, + "loss": 1.1137, + "step": 24086 + }, + { + "epoch": 2.03, + "learning_rate": 5.0183790863784796e-06, + "loss": 1.1782, + "step": 24087 + }, + { + "epoch": 2.03, + "learning_rate": 5.017590951041163e-06, + "loss": 1.0091, + "step": 24088 + }, + { + "epoch": 2.03, + "learning_rate": 5.016802856870113e-06, + "loss": 1.0873, + "step": 24089 + }, + { + "epoch": 2.03, + "learning_rate": 5.016014803871829e-06, + "loss": 1.0228, + "step": 24090 + }, + { + "epoch": 2.03, + "learning_rate": 5.0152267920528265e-06, + "loss": 1.1405, + "step": 24091 + }, + { + "epoch": 2.03, + "learning_rate": 5.014438821419614e-06, + "loss": 1.104, + "step": 24092 + }, + { + "epoch": 2.03, + "learning_rate": 5.013650891978708e-06, + "loss": 1.0561, + "step": 24093 + }, + { + "epoch": 2.03, + "learning_rate": 5.0128630037366074e-06, + "loss": 1.1653, + "step": 24094 + }, + { + "epoch": 2.03, + "learning_rate": 5.012075156699836e-06, + "loss": 1.1743, + "step": 24095 + }, + { + "epoch": 2.03, + "learning_rate": 5.011287350874892e-06, + "loss": 1.0685, + "step": 24096 + }, + { + "epoch": 2.03, + "learning_rate": 5.010499586268287e-06, + "loss": 1.1265, + "step": 24097 + }, + { + "epoch": 2.03, + "learning_rate": 5.009711862886532e-06, + "loss": 1.0322, + "step": 24098 + }, + { + "epoch": 2.03, + "learning_rate": 5.0089241807361365e-06, + "loss": 1.1588, + "step": 24099 + }, + { + "epoch": 2.03, + "learning_rate": 5.008136539823605e-06, + "loss": 1.1162, + "step": 24100 + }, + { + "epoch": 2.03, + "learning_rate": 5.007348940155448e-06, + "loss": 1.0826, + "step": 24101 + }, + { + "epoch": 2.03, + "learning_rate": 5.006561381738171e-06, + "loss": 1.1502, + "step": 24102 + }, + { + "epoch": 2.03, + "learning_rate": 5.005773864578283e-06, + "loss": 1.1091, + "step": 24103 + }, + { + "epoch": 2.03, + "learning_rate": 5.0049863886822915e-06, + "loss": 1.1221, + "step": 24104 + }, + { + "epoch": 2.03, + "learning_rate": 5.004198954056695e-06, + "loss": 1.1882, + "step": 24105 + }, + { + "epoch": 2.03, + "learning_rate": 5.0034115607080135e-06, + "loss": 1.1145, + "step": 24106 + }, + { + "epoch": 2.03, + "learning_rate": 5.00262420864274e-06, + "loss": 1.1187, + "step": 24107 + }, + { + "epoch": 2.03, + "learning_rate": 5.001836897867385e-06, + "loss": 1.0669, + "step": 24108 + }, + { + "epoch": 2.03, + "learning_rate": 5.001049628388454e-06, + "loss": 1.0603, + "step": 24109 + }, + { + "epoch": 2.03, + "learning_rate": 5.00026240021245e-06, + "loss": 1.113, + "step": 24110 + }, + { + "epoch": 2.03, + "learning_rate": 4.999475213345878e-06, + "loss": 1.1593, + "step": 24111 + }, + { + "epoch": 2.03, + "learning_rate": 4.998688067795242e-06, + "loss": 1.16, + "step": 24112 + }, + { + "epoch": 2.03, + "learning_rate": 4.997900963567047e-06, + "loss": 1.1809, + "step": 24113 + }, + { + "epoch": 2.03, + "learning_rate": 4.997113900667794e-06, + "loss": 1.1045, + "step": 24114 + }, + { + "epoch": 2.03, + "learning_rate": 4.9963268791039925e-06, + "loss": 1.0754, + "step": 24115 + }, + { + "epoch": 2.03, + "learning_rate": 4.99553989888213e-06, + "loss": 1.1298, + "step": 24116 + }, + { + "epoch": 2.03, + "learning_rate": 4.9947529600087275e-06, + "loss": 1.1723, + "step": 24117 + }, + { + "epoch": 2.03, + "learning_rate": 4.993966062490273e-06, + "loss": 1.1417, + "step": 24118 + }, + { + "epoch": 2.03, + "learning_rate": 4.993179206333274e-06, + "loss": 1.0486, + "step": 24119 + }, + { + "epoch": 2.03, + "learning_rate": 4.99239239154423e-06, + "loss": 1.1541, + "step": 24120 + }, + { + "epoch": 2.03, + "learning_rate": 4.991605618129643e-06, + "loss": 1.1853, + "step": 24121 + }, + { + "epoch": 2.03, + "learning_rate": 4.9908188860960136e-06, + "loss": 1.0709, + "step": 24122 + }, + { + "epoch": 2.03, + "learning_rate": 4.990032195449842e-06, + "loss": 0.9041, + "step": 24123 + }, + { + "epoch": 2.03, + "learning_rate": 4.9892455461976275e-06, + "loss": 1.0292, + "step": 24124 + }, + { + "epoch": 2.03, + "learning_rate": 4.9884589383458695e-06, + "loss": 1.1615, + "step": 24125 + }, + { + "epoch": 2.03, + "learning_rate": 4.9876723719010715e-06, + "loss": 1.1418, + "step": 24126 + }, + { + "epoch": 2.03, + "learning_rate": 4.986885846869721e-06, + "loss": 1.0846, + "step": 24127 + }, + { + "epoch": 2.03, + "learning_rate": 4.986099363258332e-06, + "loss": 1.0233, + "step": 24128 + }, + { + "epoch": 2.03, + "learning_rate": 4.9853129210733904e-06, + "loss": 1.1677, + "step": 24129 + }, + { + "epoch": 2.03, + "learning_rate": 4.984526520321399e-06, + "loss": 1.1249, + "step": 24130 + }, + { + "epoch": 2.03, + "learning_rate": 4.983740161008855e-06, + "loss": 1.2297, + "step": 24131 + }, + { + "epoch": 2.03, + "learning_rate": 4.982953843142253e-06, + "loss": 1.0891, + "step": 24132 + }, + { + "epoch": 2.03, + "learning_rate": 4.982167566728093e-06, + "loss": 1.0387, + "step": 24133 + }, + { + "epoch": 2.03, + "learning_rate": 4.981381331772876e-06, + "loss": 1.0808, + "step": 24134 + }, + { + "epoch": 2.03, + "learning_rate": 4.980595138283083e-06, + "loss": 1.1923, + "step": 24135 + }, + { + "epoch": 2.03, + "learning_rate": 4.979808986265224e-06, + "loss": 1.1087, + "step": 24136 + }, + { + "epoch": 2.03, + "learning_rate": 4.979022875725794e-06, + "loss": 1.1737, + "step": 24137 + }, + { + "epoch": 2.03, + "learning_rate": 4.978236806671276e-06, + "loss": 1.1615, + "step": 24138 + }, + { + "epoch": 2.03, + "learning_rate": 4.977450779108179e-06, + "loss": 1.1312, + "step": 24139 + }, + { + "epoch": 2.03, + "learning_rate": 4.976664793042989e-06, + "loss": 1.0334, + "step": 24140 + }, + { + "epoch": 2.03, + "learning_rate": 4.975878848482202e-06, + "loss": 1.2065, + "step": 24141 + }, + { + "epoch": 2.03, + "learning_rate": 4.975092945432311e-06, + "loss": 1.1446, + "step": 24142 + }, + { + "epoch": 2.03, + "learning_rate": 4.974307083899815e-06, + "loss": 1.1143, + "step": 24143 + }, + { + "epoch": 2.03, + "learning_rate": 4.973521263891194e-06, + "loss": 1.1105, + "step": 24144 + }, + { + "epoch": 2.03, + "learning_rate": 4.972735485412957e-06, + "loss": 1.125, + "step": 24145 + }, + { + "epoch": 2.03, + "learning_rate": 4.971949748471584e-06, + "loss": 1.0365, + "step": 24146 + }, + { + "epoch": 2.03, + "learning_rate": 4.9711640530735715e-06, + "loss": 1.1333, + "step": 24147 + }, + { + "epoch": 2.03, + "learning_rate": 4.9703783992254115e-06, + "loss": 1.1641, + "step": 24148 + }, + { + "epoch": 2.03, + "learning_rate": 4.969592786933594e-06, + "loss": 1.0935, + "step": 24149 + }, + { + "epoch": 2.03, + "learning_rate": 4.968807216204612e-06, + "loss": 1.1573, + "step": 24150 + }, + { + "epoch": 2.03, + "learning_rate": 4.968021687044955e-06, + "loss": 1.1893, + "step": 24151 + }, + { + "epoch": 2.03, + "learning_rate": 4.967236199461111e-06, + "loss": 1.0934, + "step": 24152 + }, + { + "epoch": 2.03, + "learning_rate": 4.966450753459575e-06, + "loss": 1.2103, + "step": 24153 + }, + { + "epoch": 2.03, + "learning_rate": 4.965665349046837e-06, + "loss": 1.0792, + "step": 24154 + }, + { + "epoch": 2.03, + "learning_rate": 4.964879986229375e-06, + "loss": 1.0947, + "step": 24155 + }, + { + "epoch": 2.03, + "learning_rate": 4.964094665013693e-06, + "loss": 1.2362, + "step": 24156 + }, + { + "epoch": 2.03, + "learning_rate": 4.96330938540627e-06, + "loss": 1.0565, + "step": 24157 + }, + { + "epoch": 2.03, + "learning_rate": 4.962524147413595e-06, + "loss": 1.047, + "step": 24158 + }, + { + "epoch": 2.03, + "learning_rate": 4.96173895104216e-06, + "loss": 1.131, + "step": 24159 + }, + { + "epoch": 2.03, + "learning_rate": 4.960953796298449e-06, + "loss": 1.1292, + "step": 24160 + }, + { + "epoch": 2.03, + "learning_rate": 4.9601686831889506e-06, + "loss": 1.0315, + "step": 24161 + }, + { + "epoch": 2.03, + "learning_rate": 4.959383611720151e-06, + "loss": 1.1001, + "step": 24162 + }, + { + "epoch": 2.03, + "learning_rate": 4.958598581898537e-06, + "loss": 1.1073, + "step": 24163 + }, + { + "epoch": 2.03, + "learning_rate": 4.957813593730596e-06, + "loss": 1.1517, + "step": 24164 + }, + { + "epoch": 2.03, + "learning_rate": 4.957028647222817e-06, + "loss": 1.0373, + "step": 24165 + }, + { + "epoch": 2.03, + "learning_rate": 4.956243742381672e-06, + "loss": 1.0946, + "step": 24166 + }, + { + "epoch": 2.03, + "learning_rate": 4.955458879213664e-06, + "loss": 1.1789, + "step": 24167 + }, + { + "epoch": 2.03, + "learning_rate": 4.954674057725265e-06, + "loss": 1.128, + "step": 24168 + }, + { + "epoch": 2.03, + "learning_rate": 4.9538892779229645e-06, + "loss": 1.076, + "step": 24169 + }, + { + "epoch": 2.03, + "learning_rate": 4.9531045398132445e-06, + "loss": 1.069, + "step": 24170 + }, + { + "epoch": 2.03, + "learning_rate": 4.952319843402591e-06, + "loss": 1.0878, + "step": 24171 + }, + { + "epoch": 2.04, + "learning_rate": 4.951535188697486e-06, + "loss": 1.054, + "step": 24172 + }, + { + "epoch": 2.04, + "learning_rate": 4.950750575704413e-06, + "loss": 1.1677, + "step": 24173 + }, + { + "epoch": 2.04, + "learning_rate": 4.949966004429855e-06, + "loss": 1.066, + "step": 24174 + }, + { + "epoch": 2.04, + "learning_rate": 4.949181474880294e-06, + "loss": 1.1156, + "step": 24175 + }, + { + "epoch": 2.04, + "learning_rate": 4.948396987062216e-06, + "loss": 1.0517, + "step": 24176 + }, + { + "epoch": 2.04, + "learning_rate": 4.947612540982091e-06, + "loss": 1.0761, + "step": 24177 + }, + { + "epoch": 2.04, + "learning_rate": 4.946828136646415e-06, + "loss": 1.2115, + "step": 24178 + }, + { + "epoch": 2.04, + "learning_rate": 4.946043774061659e-06, + "loss": 1.1691, + "step": 24179 + }, + { + "epoch": 2.04, + "learning_rate": 4.945259453234308e-06, + "loss": 1.0611, + "step": 24180 + }, + { + "epoch": 2.04, + "learning_rate": 4.9444751741708395e-06, + "loss": 1.0977, + "step": 24181 + }, + { + "epoch": 2.04, + "learning_rate": 4.943690936877735e-06, + "loss": 1.1945, + "step": 24182 + }, + { + "epoch": 2.04, + "learning_rate": 4.942906741361475e-06, + "loss": 1.1119, + "step": 24183 + }, + { + "epoch": 2.04, + "learning_rate": 4.942122587628542e-06, + "loss": 1.0328, + "step": 24184 + }, + { + "epoch": 2.04, + "learning_rate": 4.941338475685403e-06, + "loss": 1.1516, + "step": 24185 + }, + { + "epoch": 2.04, + "learning_rate": 4.940554405538548e-06, + "loss": 1.1732, + "step": 24186 + }, + { + "epoch": 2.04, + "learning_rate": 4.939770377194456e-06, + "loss": 1.0887, + "step": 24187 + }, + { + "epoch": 2.04, + "learning_rate": 4.938986390659593e-06, + "loss": 1.0655, + "step": 24188 + }, + { + "epoch": 2.04, + "learning_rate": 4.938202445940451e-06, + "loss": 1.1488, + "step": 24189 + }, + { + "epoch": 2.04, + "learning_rate": 4.937418543043497e-06, + "loss": 1.1786, + "step": 24190 + }, + { + "epoch": 2.04, + "learning_rate": 4.936634681975212e-06, + "loss": 1.1876, + "step": 24191 + }, + { + "epoch": 2.04, + "learning_rate": 4.935850862742071e-06, + "loss": 1.1375, + "step": 24192 + }, + { + "epoch": 2.04, + "learning_rate": 4.935067085350551e-06, + "loss": 1.1443, + "step": 24193 + }, + { + "epoch": 2.04, + "learning_rate": 4.9342833498071275e-06, + "loss": 1.0948, + "step": 24194 + }, + { + "epoch": 2.04, + "learning_rate": 4.93349965611828e-06, + "loss": 1.2009, + "step": 24195 + }, + { + "epoch": 2.04, + "learning_rate": 4.932716004290473e-06, + "loss": 1.0884, + "step": 24196 + }, + { + "epoch": 2.04, + "learning_rate": 4.931932394330195e-06, + "loss": 1.092, + "step": 24197 + }, + { + "epoch": 2.04, + "learning_rate": 4.93114882624391e-06, + "loss": 1.1575, + "step": 24198 + }, + { + "epoch": 2.04, + "learning_rate": 4.930365300038092e-06, + "loss": 1.1543, + "step": 24199 + }, + { + "epoch": 2.04, + "learning_rate": 4.929581815719226e-06, + "loss": 1.0672, + "step": 24200 + }, + { + "epoch": 2.04, + "learning_rate": 4.928798373293774e-06, + "loss": 1.2081, + "step": 24201 + }, + { + "epoch": 2.04, + "learning_rate": 4.928014972768213e-06, + "loss": 1.189, + "step": 24202 + }, + { + "epoch": 2.04, + "learning_rate": 4.927231614149014e-06, + "loss": 1.1087, + "step": 24203 + }, + { + "epoch": 2.04, + "learning_rate": 4.926448297442656e-06, + "loss": 1.1776, + "step": 24204 + }, + { + "epoch": 2.04, + "learning_rate": 4.925665022655598e-06, + "loss": 1.0874, + "step": 24205 + }, + { + "epoch": 2.04, + "learning_rate": 4.924881789794327e-06, + "loss": 1.1065, + "step": 24206 + }, + { + "epoch": 2.04, + "learning_rate": 4.924098598865304e-06, + "loss": 1.0972, + "step": 24207 + }, + { + "epoch": 2.04, + "learning_rate": 4.923315449875002e-06, + "loss": 1.1521, + "step": 24208 + }, + { + "epoch": 2.04, + "learning_rate": 4.922532342829892e-06, + "loss": 1.1325, + "step": 24209 + }, + { + "epoch": 2.04, + "learning_rate": 4.921749277736446e-06, + "loss": 1.1128, + "step": 24210 + }, + { + "epoch": 2.04, + "learning_rate": 4.920966254601133e-06, + "loss": 1.0293, + "step": 24211 + }, + { + "epoch": 2.04, + "learning_rate": 4.920183273430421e-06, + "loss": 1.0494, + "step": 24212 + }, + { + "epoch": 2.04, + "learning_rate": 4.9194003342307805e-06, + "loss": 1.1011, + "step": 24213 + }, + { + "epoch": 2.04, + "learning_rate": 4.9186174370086815e-06, + "loss": 1.0754, + "step": 24214 + }, + { + "epoch": 2.04, + "learning_rate": 4.9178345817705956e-06, + "loss": 1.122, + "step": 24215 + }, + { + "epoch": 2.04, + "learning_rate": 4.917051768522978e-06, + "loss": 1.0925, + "step": 24216 + }, + { + "epoch": 2.04, + "learning_rate": 4.916268997272313e-06, + "loss": 1.1033, + "step": 24217 + }, + { + "epoch": 2.04, + "learning_rate": 4.915486268025058e-06, + "loss": 1.0361, + "step": 24218 + }, + { + "epoch": 2.04, + "learning_rate": 4.914703580787682e-06, + "loss": 1.1259, + "step": 24219 + }, + { + "epoch": 2.04, + "learning_rate": 4.913920935566651e-06, + "loss": 1.1407, + "step": 24220 + }, + { + "epoch": 2.04, + "learning_rate": 4.913138332368436e-06, + "loss": 1.1743, + "step": 24221 + }, + { + "epoch": 2.04, + "learning_rate": 4.912355771199499e-06, + "loss": 1.1177, + "step": 24222 + }, + { + "epoch": 2.04, + "learning_rate": 4.911573252066306e-06, + "loss": 1.0426, + "step": 24223 + }, + { + "epoch": 2.04, + "learning_rate": 4.910790774975324e-06, + "loss": 1.0901, + "step": 24224 + }, + { + "epoch": 2.04, + "learning_rate": 4.910008339933016e-06, + "loss": 1.1817, + "step": 24225 + }, + { + "epoch": 2.04, + "learning_rate": 4.909225946945854e-06, + "loss": 1.0364, + "step": 24226 + }, + { + "epoch": 2.04, + "learning_rate": 4.908443596020288e-06, + "loss": 1.0699, + "step": 24227 + }, + { + "epoch": 2.04, + "learning_rate": 4.907661287162798e-06, + "loss": 1.0656, + "step": 24228 + }, + { + "epoch": 2.04, + "learning_rate": 4.906879020379837e-06, + "loss": 1.0622, + "step": 24229 + }, + { + "epoch": 2.04, + "learning_rate": 4.90609679567787e-06, + "loss": 1.1003, + "step": 24230 + }, + { + "epoch": 2.04, + "learning_rate": 4.905314613063363e-06, + "loss": 1.0911, + "step": 24231 + }, + { + "epoch": 2.04, + "learning_rate": 4.9045324725427764e-06, + "loss": 1.1106, + "step": 24232 + }, + { + "epoch": 2.04, + "learning_rate": 4.9037503741225734e-06, + "loss": 1.1317, + "step": 24233 + }, + { + "epoch": 2.04, + "learning_rate": 4.90296831780922e-06, + "loss": 1.1774, + "step": 24234 + }, + { + "epoch": 2.04, + "learning_rate": 4.902186303609165e-06, + "loss": 1.1295, + "step": 24235 + }, + { + "epoch": 2.04, + "learning_rate": 4.901404331528884e-06, + "loss": 1.2011, + "step": 24236 + }, + { + "epoch": 2.04, + "learning_rate": 4.900622401574835e-06, + "loss": 1.129, + "step": 24237 + }, + { + "epoch": 2.04, + "learning_rate": 4.899840513753468e-06, + "loss": 1.0523, + "step": 24238 + }, + { + "epoch": 2.04, + "learning_rate": 4.899058668071258e-06, + "loss": 1.1528, + "step": 24239 + }, + { + "epoch": 2.04, + "learning_rate": 4.898276864534655e-06, + "loss": 1.088, + "step": 24240 + }, + { + "epoch": 2.04, + "learning_rate": 4.89749510315012e-06, + "loss": 1.181, + "step": 24241 + }, + { + "epoch": 2.04, + "learning_rate": 4.896713383924115e-06, + "loss": 1.2107, + "step": 24242 + }, + { + "epoch": 2.04, + "learning_rate": 4.8959317068630965e-06, + "loss": 1.2126, + "step": 24243 + }, + { + "epoch": 2.04, + "learning_rate": 4.895150071973524e-06, + "loss": 1.0857, + "step": 24244 + }, + { + "epoch": 2.04, + "learning_rate": 4.894368479261858e-06, + "loss": 1.0358, + "step": 24245 + }, + { + "epoch": 2.04, + "learning_rate": 4.893586928734547e-06, + "loss": 1.1256, + "step": 24246 + }, + { + "epoch": 2.04, + "learning_rate": 4.892805420398064e-06, + "loss": 1.0654, + "step": 24247 + }, + { + "epoch": 2.04, + "learning_rate": 4.892023954258851e-06, + "loss": 1.1497, + "step": 24248 + }, + { + "epoch": 2.04, + "learning_rate": 4.891242530323368e-06, + "loss": 1.1512, + "step": 24249 + }, + { + "epoch": 2.04, + "learning_rate": 4.8904611485980825e-06, + "loss": 1.0834, + "step": 24250 + }, + { + "epoch": 2.04, + "learning_rate": 4.889679809089438e-06, + "loss": 1.1238, + "step": 24251 + }, + { + "epoch": 2.04, + "learning_rate": 4.888898511803894e-06, + "loss": 1.0824, + "step": 24252 + }, + { + "epoch": 2.04, + "learning_rate": 4.888117256747906e-06, + "loss": 1.1748, + "step": 24253 + }, + { + "epoch": 2.04, + "learning_rate": 4.887336043927934e-06, + "loss": 1.0524, + "step": 24254 + }, + { + "epoch": 2.04, + "learning_rate": 4.88655487335042e-06, + "loss": 1.1231, + "step": 24255 + }, + { + "epoch": 2.04, + "learning_rate": 4.885773745021834e-06, + "loss": 1.0975, + "step": 24256 + }, + { + "epoch": 2.04, + "learning_rate": 4.884992658948618e-06, + "loss": 1.1569, + "step": 24257 + }, + { + "epoch": 2.04, + "learning_rate": 4.884211615137229e-06, + "loss": 1.0839, + "step": 24258 + }, + { + "epoch": 2.04, + "learning_rate": 4.883430613594121e-06, + "loss": 1.1474, + "step": 24259 + }, + { + "epoch": 2.04, + "learning_rate": 4.882649654325747e-06, + "loss": 1.057, + "step": 24260 + }, + { + "epoch": 2.04, + "learning_rate": 4.88186873733856e-06, + "loss": 1.0889, + "step": 24261 + }, + { + "epoch": 2.04, + "learning_rate": 4.881087862639011e-06, + "loss": 1.0785, + "step": 24262 + }, + { + "epoch": 2.04, + "learning_rate": 4.880307030233552e-06, + "loss": 1.1837, + "step": 24263 + }, + { + "epoch": 2.04, + "learning_rate": 4.8795262401286354e-06, + "loss": 1.1469, + "step": 24264 + }, + { + "epoch": 2.04, + "learning_rate": 4.878745492330715e-06, + "loss": 1.1665, + "step": 24265 + }, + { + "epoch": 2.04, + "learning_rate": 4.877964786846231e-06, + "loss": 1.0931, + "step": 24266 + }, + { + "epoch": 2.04, + "learning_rate": 4.8771841236816495e-06, + "loss": 1.1917, + "step": 24267 + }, + { + "epoch": 2.04, + "learning_rate": 4.8764035028434074e-06, + "loss": 1.1315, + "step": 24268 + }, + { + "epoch": 2.04, + "learning_rate": 4.87562292433796e-06, + "loss": 1.0764, + "step": 24269 + }, + { + "epoch": 2.04, + "learning_rate": 4.874842388171756e-06, + "loss": 1.1657, + "step": 24270 + }, + { + "epoch": 2.04, + "learning_rate": 4.874061894351244e-06, + "loss": 1.0957, + "step": 24271 + }, + { + "epoch": 2.04, + "learning_rate": 4.873281442882873e-06, + "loss": 1.0485, + "step": 24272 + }, + { + "epoch": 2.04, + "learning_rate": 4.872501033773093e-06, + "loss": 1.1469, + "step": 24273 + }, + { + "epoch": 2.04, + "learning_rate": 4.871720667028349e-06, + "loss": 1.0831, + "step": 24274 + }, + { + "epoch": 2.04, + "learning_rate": 4.8709403426550905e-06, + "loss": 1.1589, + "step": 24275 + }, + { + "epoch": 2.04, + "learning_rate": 4.8701600606597675e-06, + "loss": 1.0149, + "step": 24276 + }, + { + "epoch": 2.04, + "learning_rate": 4.869379821048818e-06, + "loss": 1.1507, + "step": 24277 + }, + { + "epoch": 2.04, + "learning_rate": 4.8685996238287e-06, + "loss": 1.1125, + "step": 24278 + }, + { + "epoch": 2.04, + "learning_rate": 4.867819469005851e-06, + "loss": 1.0901, + "step": 24279 + }, + { + "epoch": 2.04, + "learning_rate": 4.867039356586721e-06, + "loss": 1.1055, + "step": 24280 + }, + { + "epoch": 2.04, + "learning_rate": 4.866259286577754e-06, + "loss": 1.0672, + "step": 24281 + }, + { + "epoch": 2.04, + "learning_rate": 4.865479258985396e-06, + "loss": 1.1348, + "step": 24282 + }, + { + "epoch": 2.04, + "learning_rate": 4.864699273816091e-06, + "loss": 1.1189, + "step": 24283 + }, + { + "epoch": 2.04, + "learning_rate": 4.863919331076288e-06, + "loss": 1.0967, + "step": 24284 + }, + { + "epoch": 2.04, + "learning_rate": 4.86313943077242e-06, + "loss": 1.0915, + "step": 24285 + }, + { + "epoch": 2.04, + "learning_rate": 4.862359572910942e-06, + "loss": 1.1303, + "step": 24286 + }, + { + "epoch": 2.04, + "learning_rate": 4.861579757498297e-06, + "loss": 1.0741, + "step": 24287 + }, + { + "epoch": 2.04, + "learning_rate": 4.860799984540917e-06, + "loss": 1.2214, + "step": 24288 + }, + { + "epoch": 2.04, + "learning_rate": 4.860020254045259e-06, + "loss": 1.1646, + "step": 24289 + }, + { + "epoch": 2.05, + "learning_rate": 4.859240566017756e-06, + "loss": 1.1127, + "step": 24290 + }, + { + "epoch": 2.05, + "learning_rate": 4.858460920464852e-06, + "loss": 1.0584, + "step": 24291 + }, + { + "epoch": 2.05, + "learning_rate": 4.85768131739299e-06, + "loss": 1.0491, + "step": 24292 + }, + { + "epoch": 2.05, + "learning_rate": 4.856901756808611e-06, + "loss": 1.0978, + "step": 24293 + }, + { + "epoch": 2.05, + "learning_rate": 4.856122238718155e-06, + "loss": 1.1126, + "step": 24294 + }, + { + "epoch": 2.05, + "learning_rate": 4.855342763128067e-06, + "loss": 1.2294, + "step": 24295 + }, + { + "epoch": 2.05, + "learning_rate": 4.854563330044777e-06, + "loss": 1.0901, + "step": 24296 + }, + { + "epoch": 2.05, + "learning_rate": 4.853783939474738e-06, + "loss": 1.0764, + "step": 24297 + }, + { + "epoch": 2.05, + "learning_rate": 4.85300459142438e-06, + "loss": 1.0971, + "step": 24298 + }, + { + "epoch": 2.05, + "learning_rate": 4.85222528590014e-06, + "loss": 1.0907, + "step": 24299 + }, + { + "epoch": 2.05, + "learning_rate": 4.851446022908472e-06, + "loss": 1.1037, + "step": 24300 + }, + { + "epoch": 2.05, + "learning_rate": 4.850666802455799e-06, + "loss": 1.1095, + "step": 24301 + }, + { + "epoch": 2.05, + "learning_rate": 4.8498876245485656e-06, + "loss": 1.0785, + "step": 24302 + }, + { + "epoch": 2.05, + "learning_rate": 4.8491084891932096e-06, + "loss": 1.0918, + "step": 24303 + }, + { + "epoch": 2.05, + "learning_rate": 4.848329396396171e-06, + "loss": 1.107, + "step": 24304 + }, + { + "epoch": 2.05, + "learning_rate": 4.847550346163876e-06, + "loss": 1.1546, + "step": 24305 + }, + { + "epoch": 2.05, + "learning_rate": 4.846771338502777e-06, + "loss": 1.0898, + "step": 24306 + }, + { + "epoch": 2.05, + "learning_rate": 4.845992373419298e-06, + "loss": 1.1029, + "step": 24307 + }, + { + "epoch": 2.05, + "learning_rate": 4.84521345091988e-06, + "loss": 1.1489, + "step": 24308 + }, + { + "epoch": 2.05, + "learning_rate": 4.844434571010959e-06, + "loss": 1.072, + "step": 24309 + }, + { + "epoch": 2.05, + "learning_rate": 4.8436557336989675e-06, + "loss": 1.1098, + "step": 24310 + }, + { + "epoch": 2.05, + "learning_rate": 4.842876938990344e-06, + "loss": 1.1912, + "step": 24311 + }, + { + "epoch": 2.05, + "learning_rate": 4.842098186891521e-06, + "loss": 1.1056, + "step": 24312 + }, + { + "epoch": 2.05, + "learning_rate": 4.841319477408935e-06, + "loss": 1.0794, + "step": 24313 + }, + { + "epoch": 2.05, + "learning_rate": 4.840540810549017e-06, + "loss": 1.1016, + "step": 24314 + }, + { + "epoch": 2.05, + "learning_rate": 4.839762186318206e-06, + "loss": 1.1402, + "step": 24315 + }, + { + "epoch": 2.05, + "learning_rate": 4.838983604722924e-06, + "loss": 1.1418, + "step": 24316 + }, + { + "epoch": 2.05, + "learning_rate": 4.838205065769619e-06, + "loss": 1.1444, + "step": 24317 + }, + { + "epoch": 2.05, + "learning_rate": 4.837426569464712e-06, + "loss": 1.1016, + "step": 24318 + }, + { + "epoch": 2.05, + "learning_rate": 4.836648115814639e-06, + "loss": 1.1523, + "step": 24319 + }, + { + "epoch": 2.05, + "learning_rate": 4.835869704825831e-06, + "loss": 1.1987, + "step": 24320 + }, + { + "epoch": 2.05, + "learning_rate": 4.83509133650472e-06, + "loss": 1.0654, + "step": 24321 + }, + { + "epoch": 2.05, + "learning_rate": 4.834313010857739e-06, + "loss": 1.0406, + "step": 24322 + }, + { + "epoch": 2.05, + "learning_rate": 4.833534727891316e-06, + "loss": 1.0842, + "step": 24323 + }, + { + "epoch": 2.05, + "learning_rate": 4.832756487611881e-06, + "loss": 1.0391, + "step": 24324 + }, + { + "epoch": 2.05, + "learning_rate": 4.831978290025869e-06, + "loss": 1.1583, + "step": 24325 + }, + { + "epoch": 2.05, + "learning_rate": 4.831200135139708e-06, + "loss": 1.0898, + "step": 24326 + }, + { + "epoch": 2.05, + "learning_rate": 4.830422022959819e-06, + "loss": 1.1196, + "step": 24327 + }, + { + "epoch": 2.05, + "learning_rate": 4.8296439534926445e-06, + "loss": 1.1482, + "step": 24328 + }, + { + "epoch": 2.05, + "learning_rate": 4.828865926744603e-06, + "loss": 1.0941, + "step": 24329 + }, + { + "epoch": 2.05, + "learning_rate": 4.828087942722126e-06, + "loss": 1.075, + "step": 24330 + }, + { + "epoch": 2.05, + "learning_rate": 4.827310001431642e-06, + "loss": 1.1368, + "step": 24331 + }, + { + "epoch": 2.05, + "learning_rate": 4.826532102879578e-06, + "loss": 1.228, + "step": 24332 + }, + { + "epoch": 2.05, + "learning_rate": 4.8257542470723605e-06, + "loss": 1.1864, + "step": 24333 + }, + { + "epoch": 2.05, + "learning_rate": 4.8249764340164225e-06, + "loss": 1.1272, + "step": 24334 + }, + { + "epoch": 2.05, + "learning_rate": 4.824198663718178e-06, + "loss": 1.1296, + "step": 24335 + }, + { + "epoch": 2.05, + "learning_rate": 4.823420936184064e-06, + "loss": 1.1803, + "step": 24336 + }, + { + "epoch": 2.05, + "learning_rate": 4.822643251420507e-06, + "loss": 1.1363, + "step": 24337 + }, + { + "epoch": 2.05, + "learning_rate": 4.82186560943392e-06, + "loss": 1.0783, + "step": 24338 + }, + { + "epoch": 2.05, + "learning_rate": 4.821088010230744e-06, + "loss": 1.2341, + "step": 24339 + }, + { + "epoch": 2.05, + "learning_rate": 4.8203104538173935e-06, + "loss": 1.2046, + "step": 24340 + }, + { + "epoch": 2.05, + "learning_rate": 4.819532940200295e-06, + "loss": 0.9722, + "step": 24341 + }, + { + "epoch": 2.05, + "learning_rate": 4.818755469385874e-06, + "loss": 1.1622, + "step": 24342 + }, + { + "epoch": 2.05, + "learning_rate": 4.817978041380553e-06, + "loss": 1.1514, + "step": 24343 + }, + { + "epoch": 2.05, + "learning_rate": 4.817200656190757e-06, + "loss": 1.1298, + "step": 24344 + }, + { + "epoch": 2.05, + "learning_rate": 4.816423313822911e-06, + "loss": 1.2248, + "step": 24345 + }, + { + "epoch": 2.05, + "learning_rate": 4.815646014283426e-06, + "loss": 1.106, + "step": 24346 + }, + { + "epoch": 2.05, + "learning_rate": 4.8148687575787415e-06, + "loss": 1.0941, + "step": 24347 + }, + { + "epoch": 2.05, + "learning_rate": 4.8140915437152665e-06, + "loss": 1.15, + "step": 24348 + }, + { + "epoch": 2.05, + "learning_rate": 4.813314372699423e-06, + "loss": 1.1294, + "step": 24349 + }, + { + "epoch": 2.05, + "learning_rate": 4.812537244537644e-06, + "loss": 1.1175, + "step": 24350 + }, + { + "epoch": 2.05, + "learning_rate": 4.811760159236339e-06, + "loss": 1.1754, + "step": 24351 + }, + { + "epoch": 2.05, + "learning_rate": 4.810983116801931e-06, + "loss": 1.2173, + "step": 24352 + }, + { + "epoch": 2.05, + "learning_rate": 4.810206117240842e-06, + "loss": 1.2094, + "step": 24353 + }, + { + "epoch": 2.05, + "learning_rate": 4.809429160559496e-06, + "loss": 1.1089, + "step": 24354 + }, + { + "epoch": 2.05, + "learning_rate": 4.808652246764298e-06, + "loss": 1.1156, + "step": 24355 + }, + { + "epoch": 2.05, + "learning_rate": 4.807875375861684e-06, + "loss": 1.121, + "step": 24356 + }, + { + "epoch": 2.05, + "learning_rate": 4.807098547858062e-06, + "loss": 1.0708, + "step": 24357 + }, + { + "epoch": 2.05, + "learning_rate": 4.806321762759855e-06, + "loss": 1.1463, + "step": 24358 + }, + { + "epoch": 2.05, + "learning_rate": 4.805545020573479e-06, + "loss": 1.15, + "step": 24359 + }, + { + "epoch": 2.05, + "learning_rate": 4.804768321305352e-06, + "loss": 1.1096, + "step": 24360 + }, + { + "epoch": 2.05, + "learning_rate": 4.803991664961891e-06, + "loss": 1.1788, + "step": 24361 + }, + { + "epoch": 2.05, + "learning_rate": 4.803215051549516e-06, + "loss": 1.1676, + "step": 24362 + }, + { + "epoch": 2.05, + "learning_rate": 4.80243848107464e-06, + "loss": 1.0994, + "step": 24363 + }, + { + "epoch": 2.05, + "learning_rate": 4.8016619535436804e-06, + "loss": 1.0531, + "step": 24364 + }, + { + "epoch": 2.05, + "learning_rate": 4.800885468963058e-06, + "loss": 1.1287, + "step": 24365 + }, + { + "epoch": 2.05, + "learning_rate": 4.8001090273391755e-06, + "loss": 1.0838, + "step": 24366 + }, + { + "epoch": 2.05, + "learning_rate": 4.799332628678465e-06, + "loss": 1.0796, + "step": 24367 + }, + { + "epoch": 2.05, + "learning_rate": 4.798556272987328e-06, + "loss": 1.0949, + "step": 24368 + }, + { + "epoch": 2.05, + "learning_rate": 4.797779960272184e-06, + "loss": 1.1195, + "step": 24369 + }, + { + "epoch": 2.05, + "learning_rate": 4.797003690539448e-06, + "loss": 1.0419, + "step": 24370 + }, + { + "epoch": 2.05, + "learning_rate": 4.796227463795532e-06, + "loss": 1.0431, + "step": 24371 + }, + { + "epoch": 2.05, + "learning_rate": 4.795451280046849e-06, + "loss": 1.1626, + "step": 24372 + }, + { + "epoch": 2.05, + "learning_rate": 4.794675139299815e-06, + "loss": 1.131, + "step": 24373 + }, + { + "epoch": 2.05, + "learning_rate": 4.7938990415608385e-06, + "loss": 1.1843, + "step": 24374 + }, + { + "epoch": 2.05, + "learning_rate": 4.7931229868363375e-06, + "loss": 1.1643, + "step": 24375 + }, + { + "epoch": 2.05, + "learning_rate": 4.792346975132722e-06, + "loss": 1.1541, + "step": 24376 + }, + { + "epoch": 2.05, + "learning_rate": 4.791571006456396e-06, + "loss": 1.1615, + "step": 24377 + }, + { + "epoch": 2.05, + "learning_rate": 4.790795080813783e-06, + "loss": 1.1574, + "step": 24378 + }, + { + "epoch": 2.05, + "learning_rate": 4.7900191982112876e-06, + "loss": 1.1681, + "step": 24379 + }, + { + "epoch": 2.05, + "learning_rate": 4.789243358655319e-06, + "loss": 1.0518, + "step": 24380 + }, + { + "epoch": 2.05, + "learning_rate": 4.7884675621522904e-06, + "loss": 1.0701, + "step": 24381 + }, + { + "epoch": 2.05, + "learning_rate": 4.78769180870861e-06, + "loss": 1.1765, + "step": 24382 + }, + { + "epoch": 2.05, + "learning_rate": 4.786916098330687e-06, + "loss": 1.0973, + "step": 24383 + }, + { + "epoch": 2.05, + "learning_rate": 4.786140431024937e-06, + "loss": 1.1317, + "step": 24384 + }, + { + "epoch": 2.05, + "learning_rate": 4.7853648067977555e-06, + "loss": 1.1339, + "step": 24385 + }, + { + "epoch": 2.05, + "learning_rate": 4.784589225655563e-06, + "loss": 1.1121, + "step": 24386 + }, + { + "epoch": 2.05, + "learning_rate": 4.7838136876047655e-06, + "loss": 1.1143, + "step": 24387 + }, + { + "epoch": 2.05, + "learning_rate": 4.783038192651763e-06, + "loss": 1.1218, + "step": 24388 + }, + { + "epoch": 2.05, + "learning_rate": 4.782262740802975e-06, + "loss": 1.078, + "step": 24389 + }, + { + "epoch": 2.05, + "learning_rate": 4.781487332064799e-06, + "loss": 1.1278, + "step": 24390 + }, + { + "epoch": 2.05, + "learning_rate": 4.780711966443646e-06, + "loss": 1.1299, + "step": 24391 + }, + { + "epoch": 2.05, + "learning_rate": 4.77993664394592e-06, + "loss": 1.1083, + "step": 24392 + }, + { + "epoch": 2.05, + "learning_rate": 4.779161364578026e-06, + "loss": 1.1531, + "step": 24393 + }, + { + "epoch": 2.05, + "learning_rate": 4.778386128346374e-06, + "loss": 1.1019, + "step": 24394 + }, + { + "epoch": 2.05, + "learning_rate": 4.777610935257371e-06, + "loss": 1.2379, + "step": 24395 + }, + { + "epoch": 2.05, + "learning_rate": 4.77683578531741e-06, + "loss": 1.0621, + "step": 24396 + }, + { + "epoch": 2.05, + "learning_rate": 4.77606067853291e-06, + "loss": 1.1666, + "step": 24397 + }, + { + "epoch": 2.05, + "learning_rate": 4.775285614910264e-06, + "loss": 1.099, + "step": 24398 + }, + { + "epoch": 2.05, + "learning_rate": 4.774510594455879e-06, + "loss": 1.1064, + "step": 24399 + }, + { + "epoch": 2.05, + "learning_rate": 4.773735617176166e-06, + "loss": 1.0873, + "step": 24400 + }, + { + "epoch": 2.05, + "learning_rate": 4.772960683077518e-06, + "loss": 1.0993, + "step": 24401 + }, + { + "epoch": 2.05, + "learning_rate": 4.772185792166343e-06, + "loss": 1.1732, + "step": 24402 + }, + { + "epoch": 2.05, + "learning_rate": 4.771410944449041e-06, + "loss": 1.106, + "step": 24403 + }, + { + "epoch": 2.05, + "learning_rate": 4.770636139932019e-06, + "loss": 1.1362, + "step": 24404 + }, + { + "epoch": 2.05, + "learning_rate": 4.769861378621667e-06, + "loss": 1.0789, + "step": 24405 + }, + { + "epoch": 2.05, + "learning_rate": 4.7690866605244024e-06, + "loss": 1.1793, + "step": 24406 + }, + { + "epoch": 2.05, + "learning_rate": 4.768311985646614e-06, + "loss": 1.0877, + "step": 24407 + }, + { + "epoch": 2.05, + "learning_rate": 4.7675373539947055e-06, + "loss": 1.1586, + "step": 24408 + }, + { + "epoch": 2.06, + "learning_rate": 4.766762765575079e-06, + "loss": 1.0363, + "step": 24409 + }, + { + "epoch": 2.06, + "learning_rate": 4.7659882203941336e-06, + "loss": 1.0758, + "step": 24410 + }, + { + "epoch": 2.06, + "learning_rate": 4.765213718458268e-06, + "loss": 1.0814, + "step": 24411 + }, + { + "epoch": 2.06, + "learning_rate": 4.764439259773882e-06, + "loss": 1.1158, + "step": 24412 + }, + { + "epoch": 2.06, + "learning_rate": 4.763664844347374e-06, + "loss": 1.1376, + "step": 24413 + }, + { + "epoch": 2.06, + "learning_rate": 4.7628904721851435e-06, + "loss": 1.1587, + "step": 24414 + }, + { + "epoch": 2.06, + "learning_rate": 4.762116143293591e-06, + "loss": 1.114, + "step": 24415 + }, + { + "epoch": 2.06, + "learning_rate": 4.761341857679105e-06, + "loss": 1.1488, + "step": 24416 + }, + { + "epoch": 2.06, + "learning_rate": 4.760567615348096e-06, + "loss": 1.0124, + "step": 24417 + }, + { + "epoch": 2.06, + "learning_rate": 4.759793416306951e-06, + "loss": 1.1542, + "step": 24418 + }, + { + "epoch": 2.06, + "learning_rate": 4.759019260562069e-06, + "loss": 1.1327, + "step": 24419 + }, + { + "epoch": 2.06, + "learning_rate": 4.758245148119848e-06, + "loss": 1.1221, + "step": 24420 + }, + { + "epoch": 2.06, + "learning_rate": 4.7574710789866825e-06, + "loss": 1.1376, + "step": 24421 + }, + { + "epoch": 2.06, + "learning_rate": 4.7566970531689694e-06, + "loss": 1.0858, + "step": 24422 + }, + { + "epoch": 2.06, + "learning_rate": 4.755923070673103e-06, + "loss": 1.1054, + "step": 24423 + }, + { + "epoch": 2.06, + "learning_rate": 4.755149131505478e-06, + "loss": 1.1739, + "step": 24424 + }, + { + "epoch": 2.06, + "learning_rate": 4.754375235672491e-06, + "loss": 1.195, + "step": 24425 + }, + { + "epoch": 2.06, + "learning_rate": 4.753601383180536e-06, + "loss": 1.1394, + "step": 24426 + }, + { + "epoch": 2.06, + "learning_rate": 4.752827574036e-06, + "loss": 1.0722, + "step": 24427 + }, + { + "epoch": 2.06, + "learning_rate": 4.752053808245288e-06, + "loss": 1.1241, + "step": 24428 + }, + { + "epoch": 2.06, + "learning_rate": 4.751280085814783e-06, + "loss": 1.1566, + "step": 24429 + }, + { + "epoch": 2.06, + "learning_rate": 4.750506406750883e-06, + "loss": 1.0978, + "step": 24430 + }, + { + "epoch": 2.06, + "learning_rate": 4.749732771059979e-06, + "loss": 1.1098, + "step": 24431 + }, + { + "epoch": 2.06, + "learning_rate": 4.7489591787484626e-06, + "loss": 1.0759, + "step": 24432 + }, + { + "epoch": 2.06, + "learning_rate": 4.748185629822727e-06, + "loss": 1.2661, + "step": 24433 + }, + { + "epoch": 2.06, + "learning_rate": 4.747412124289164e-06, + "loss": 1.1257, + "step": 24434 + }, + { + "epoch": 2.06, + "learning_rate": 4.746638662154159e-06, + "loss": 1.0859, + "step": 24435 + }, + { + "epoch": 2.06, + "learning_rate": 4.7458652434241085e-06, + "loss": 1.1394, + "step": 24436 + }, + { + "epoch": 2.06, + "learning_rate": 4.745091868105403e-06, + "loss": 1.139, + "step": 24437 + }, + { + "epoch": 2.06, + "learning_rate": 4.744318536204425e-06, + "loss": 1.0758, + "step": 24438 + }, + { + "epoch": 2.06, + "learning_rate": 4.743545247727577e-06, + "loss": 1.1437, + "step": 24439 + }, + { + "epoch": 2.06, + "learning_rate": 4.742772002681235e-06, + "loss": 1.1819, + "step": 24440 + }, + { + "epoch": 2.06, + "learning_rate": 4.741998801071794e-06, + "loss": 1.1691, + "step": 24441 + }, + { + "epoch": 2.06, + "learning_rate": 4.741225642905643e-06, + "loss": 1.1548, + "step": 24442 + }, + { + "epoch": 2.06, + "learning_rate": 4.7404525281891665e-06, + "loss": 1.1472, + "step": 24443 + }, + { + "epoch": 2.06, + "learning_rate": 4.739679456928756e-06, + "loss": 1.1636, + "step": 24444 + }, + { + "epoch": 2.06, + "learning_rate": 4.7389064291308005e-06, + "loss": 1.1332, + "step": 24445 + }, + { + "epoch": 2.06, + "learning_rate": 4.738133444801677e-06, + "loss": 1.1373, + "step": 24446 + }, + { + "epoch": 2.06, + "learning_rate": 4.737360503947785e-06, + "loss": 1.1042, + "step": 24447 + }, + { + "epoch": 2.06, + "learning_rate": 4.736587606575502e-06, + "loss": 1.1441, + "step": 24448 + }, + { + "epoch": 2.06, + "learning_rate": 4.735814752691213e-06, + "loss": 1.2035, + "step": 24449 + }, + { + "epoch": 2.06, + "learning_rate": 4.735041942301316e-06, + "loss": 1.0988, + "step": 24450 + }, + { + "epoch": 2.06, + "learning_rate": 4.734269175412182e-06, + "loss": 1.1568, + "step": 24451 + }, + { + "epoch": 2.06, + "learning_rate": 4.733496452030202e-06, + "loss": 1.1246, + "step": 24452 + }, + { + "epoch": 2.06, + "learning_rate": 4.7327237721617594e-06, + "loss": 1.1925, + "step": 24453 + }, + { + "epoch": 2.06, + "learning_rate": 4.731951135813243e-06, + "loss": 1.0969, + "step": 24454 + }, + { + "epoch": 2.06, + "learning_rate": 4.731178542991024e-06, + "loss": 1.084, + "step": 24455 + }, + { + "epoch": 2.06, + "learning_rate": 4.730405993701502e-06, + "loss": 1.0639, + "step": 24456 + }, + { + "epoch": 2.06, + "learning_rate": 4.729633487951048e-06, + "loss": 1.1132, + "step": 24457 + }, + { + "epoch": 2.06, + "learning_rate": 4.728861025746051e-06, + "loss": 1.086, + "step": 24458 + }, + { + "epoch": 2.06, + "learning_rate": 4.7280886070928886e-06, + "loss": 1.1425, + "step": 24459 + }, + { + "epoch": 2.06, + "learning_rate": 4.727316231997947e-06, + "loss": 1.0696, + "step": 24460 + }, + { + "epoch": 2.06, + "learning_rate": 4.726543900467605e-06, + "loss": 1.115, + "step": 24461 + }, + { + "epoch": 2.06, + "learning_rate": 4.725771612508246e-06, + "loss": 1.113, + "step": 24462 + }, + { + "epoch": 2.06, + "learning_rate": 4.724999368126249e-06, + "loss": 1.1964, + "step": 24463 + }, + { + "epoch": 2.06, + "learning_rate": 4.724227167327996e-06, + "loss": 1.1425, + "step": 24464 + }, + { + "epoch": 2.06, + "learning_rate": 4.72345501011987e-06, + "loss": 1.0531, + "step": 24465 + }, + { + "epoch": 2.06, + "learning_rate": 4.7226828965082404e-06, + "loss": 1.1971, + "step": 24466 + }, + { + "epoch": 2.06, + "learning_rate": 4.721910826499501e-06, + "loss": 1.0627, + "step": 24467 + }, + { + "epoch": 2.06, + "learning_rate": 4.72113880010002e-06, + "loss": 1.1111, + "step": 24468 + }, + { + "epoch": 2.06, + "learning_rate": 4.720366817316179e-06, + "loss": 1.1688, + "step": 24469 + }, + { + "epoch": 2.06, + "learning_rate": 4.719594878154357e-06, + "loss": 1.0328, + "step": 24470 + }, + { + "epoch": 2.06, + "learning_rate": 4.718822982620933e-06, + "loss": 1.2233, + "step": 24471 + }, + { + "epoch": 2.06, + "learning_rate": 4.718051130722283e-06, + "loss": 1.0955, + "step": 24472 + }, + { + "epoch": 2.06, + "learning_rate": 4.717279322464786e-06, + "loss": 1.106, + "step": 24473 + }, + { + "epoch": 2.06, + "learning_rate": 4.716507557854817e-06, + "loss": 1.0658, + "step": 24474 + }, + { + "epoch": 2.06, + "learning_rate": 4.715735836898754e-06, + "loss": 1.1577, + "step": 24475 + }, + { + "epoch": 2.06, + "learning_rate": 4.714964159602976e-06, + "loss": 1.0493, + "step": 24476 + }, + { + "epoch": 2.06, + "learning_rate": 4.714192525973848e-06, + "loss": 1.0779, + "step": 24477 + }, + { + "epoch": 2.06, + "learning_rate": 4.713420936017761e-06, + "loss": 1.07, + "step": 24478 + }, + { + "epoch": 2.06, + "learning_rate": 4.7126493897410775e-06, + "loss": 1.0721, + "step": 24479 + }, + { + "epoch": 2.06, + "learning_rate": 4.711877887150177e-06, + "loss": 1.1228, + "step": 24480 + }, + { + "epoch": 2.06, + "learning_rate": 4.7111064282514355e-06, + "loss": 1.2463, + "step": 24481 + }, + { + "epoch": 2.06, + "learning_rate": 4.710335013051224e-06, + "loss": 1.102, + "step": 24482 + }, + { + "epoch": 2.06, + "learning_rate": 4.7095636415559165e-06, + "loss": 1.2077, + "step": 24483 + }, + { + "epoch": 2.06, + "learning_rate": 4.708792313771893e-06, + "loss": 1.0266, + "step": 24484 + }, + { + "epoch": 2.06, + "learning_rate": 4.708021029705512e-06, + "loss": 1.1154, + "step": 24485 + }, + { + "epoch": 2.06, + "learning_rate": 4.70724978936316e-06, + "loss": 1.0301, + "step": 24486 + }, + { + "epoch": 2.06, + "learning_rate": 4.706478592751207e-06, + "loss": 1.0771, + "step": 24487 + }, + { + "epoch": 2.06, + "learning_rate": 4.705707439876014e-06, + "loss": 1.0501, + "step": 24488 + }, + { + "epoch": 2.06, + "learning_rate": 4.7049363307439684e-06, + "loss": 1.0824, + "step": 24489 + }, + { + "epoch": 2.06, + "learning_rate": 4.70416526536143e-06, + "loss": 1.0442, + "step": 24490 + }, + { + "epoch": 2.06, + "learning_rate": 4.703394243734773e-06, + "loss": 1.1959, + "step": 24491 + }, + { + "epoch": 2.06, + "learning_rate": 4.702623265870368e-06, + "loss": 1.1501, + "step": 24492 + }, + { + "epoch": 2.06, + "learning_rate": 4.701852331774586e-06, + "loss": 1.034, + "step": 24493 + }, + { + "epoch": 2.06, + "learning_rate": 4.701081441453794e-06, + "loss": 1.1484, + "step": 24494 + }, + { + "epoch": 2.06, + "learning_rate": 4.700310594914367e-06, + "loss": 1.1277, + "step": 24495 + }, + { + "epoch": 2.06, + "learning_rate": 4.6995397921626625e-06, + "loss": 1.038, + "step": 24496 + }, + { + "epoch": 2.06, + "learning_rate": 4.6987690332050654e-06, + "loss": 1.0636, + "step": 24497 + }, + { + "epoch": 2.06, + "learning_rate": 4.697998318047929e-06, + "loss": 1.1399, + "step": 24498 + }, + { + "epoch": 2.06, + "learning_rate": 4.697227646697627e-06, + "loss": 1.0939, + "step": 24499 + }, + { + "epoch": 2.06, + "learning_rate": 4.696457019160532e-06, + "loss": 1.1226, + "step": 24500 + }, + { + "epoch": 2.06, + "learning_rate": 4.695686435443003e-06, + "loss": 1.2671, + "step": 24501 + }, + { + "epoch": 2.06, + "learning_rate": 4.69491589555141e-06, + "loss": 1.1239, + "step": 24502 + }, + { + "epoch": 2.06, + "learning_rate": 4.694145399492122e-06, + "loss": 1.0868, + "step": 24503 + }, + { + "epoch": 2.06, + "learning_rate": 4.693374947271501e-06, + "loss": 1.0821, + "step": 24504 + }, + { + "epoch": 2.06, + "learning_rate": 4.692604538895914e-06, + "loss": 1.1493, + "step": 24505 + }, + { + "epoch": 2.06, + "learning_rate": 4.6918341743717325e-06, + "loss": 1.0879, + "step": 24506 + }, + { + "epoch": 2.06, + "learning_rate": 4.6910638537053065e-06, + "loss": 1.1498, + "step": 24507 + }, + { + "epoch": 2.06, + "learning_rate": 4.690293576903019e-06, + "loss": 1.1095, + "step": 24508 + }, + { + "epoch": 2.06, + "learning_rate": 4.689523343971221e-06, + "loss": 1.1003, + "step": 24509 + }, + { + "epoch": 2.06, + "learning_rate": 4.6887531549162804e-06, + "loss": 1.1558, + "step": 24510 + }, + { + "epoch": 2.06, + "learning_rate": 4.687983009744561e-06, + "loss": 1.07, + "step": 24511 + }, + { + "epoch": 2.06, + "learning_rate": 4.687212908462426e-06, + "loss": 1.1062, + "step": 24512 + }, + { + "epoch": 2.06, + "learning_rate": 4.686442851076238e-06, + "loss": 1.1362, + "step": 24513 + }, + { + "epoch": 2.06, + "learning_rate": 4.685672837592361e-06, + "loss": 1.1921, + "step": 24514 + }, + { + "epoch": 2.06, + "learning_rate": 4.684902868017158e-06, + "loss": 1.0381, + "step": 24515 + }, + { + "epoch": 2.06, + "learning_rate": 4.684132942356983e-06, + "loss": 1.2034, + "step": 24516 + }, + { + "epoch": 2.06, + "learning_rate": 4.683363060618209e-06, + "loss": 1.1138, + "step": 24517 + }, + { + "epoch": 2.06, + "learning_rate": 4.682593222807187e-06, + "loss": 1.0664, + "step": 24518 + }, + { + "epoch": 2.06, + "learning_rate": 4.681823428930282e-06, + "loss": 1.1203, + "step": 24519 + }, + { + "epoch": 2.06, + "learning_rate": 4.681053678993854e-06, + "loss": 1.1108, + "step": 24520 + }, + { + "epoch": 2.06, + "learning_rate": 4.680283973004262e-06, + "loss": 1.1011, + "step": 24521 + }, + { + "epoch": 2.06, + "learning_rate": 4.6795143109678676e-06, + "loss": 1.3575, + "step": 24522 + }, + { + "epoch": 2.06, + "learning_rate": 4.678744692891028e-06, + "loss": 1.1202, + "step": 24523 + }, + { + "epoch": 2.06, + "learning_rate": 4.6779751187801035e-06, + "loss": 1.0801, + "step": 24524 + }, + { + "epoch": 2.06, + "learning_rate": 4.6772055886414515e-06, + "loss": 1.1511, + "step": 24525 + }, + { + "epoch": 2.06, + "learning_rate": 4.676436102481435e-06, + "loss": 1.1485, + "step": 24526 + }, + { + "epoch": 2.06, + "learning_rate": 4.675666660306397e-06, + "loss": 1.1884, + "step": 24527 + }, + { + "epoch": 2.07, + "learning_rate": 4.674897262122714e-06, + "loss": 1.0345, + "step": 24528 + }, + { + "epoch": 2.07, + "learning_rate": 4.674127907936731e-06, + "loss": 1.1021, + "step": 24529 + }, + { + "epoch": 2.07, + "learning_rate": 4.673358597754807e-06, + "loss": 1.1475, + "step": 24530 + }, + { + "epoch": 2.07, + "learning_rate": 4.672589331583299e-06, + "loss": 1.1175, + "step": 24531 + }, + { + "epoch": 2.07, + "learning_rate": 4.671820109428564e-06, + "loss": 1.0512, + "step": 24532 + }, + { + "epoch": 2.07, + "learning_rate": 4.671050931296955e-06, + "loss": 1.0811, + "step": 24533 + }, + { + "epoch": 2.07, + "learning_rate": 4.670281797194834e-06, + "loss": 1.0978, + "step": 24534 + }, + { + "epoch": 2.07, + "learning_rate": 4.669512707128542e-06, + "loss": 1.0879, + "step": 24535 + }, + { + "epoch": 2.07, + "learning_rate": 4.668743661104446e-06, + "loss": 1.1277, + "step": 24536 + }, + { + "epoch": 2.07, + "learning_rate": 4.667974659128899e-06, + "loss": 1.1641, + "step": 24537 + }, + { + "epoch": 2.07, + "learning_rate": 4.667205701208245e-06, + "loss": 1.0373, + "step": 24538 + }, + { + "epoch": 2.07, + "learning_rate": 4.666436787348853e-06, + "loss": 1.178, + "step": 24539 + }, + { + "epoch": 2.07, + "learning_rate": 4.665667917557062e-06, + "loss": 1.1332, + "step": 24540 + }, + { + "epoch": 2.07, + "learning_rate": 4.664899091839229e-06, + "loss": 1.1522, + "step": 24541 + }, + { + "epoch": 2.07, + "learning_rate": 4.664130310201709e-06, + "loss": 1.0722, + "step": 24542 + }, + { + "epoch": 2.07, + "learning_rate": 4.663361572650851e-06, + "loss": 1.1122, + "step": 24543 + }, + { + "epoch": 2.07, + "learning_rate": 4.6625928791930085e-06, + "loss": 1.1594, + "step": 24544 + }, + { + "epoch": 2.07, + "learning_rate": 4.6618242298345355e-06, + "loss": 1.1193, + "step": 24545 + }, + { + "epoch": 2.07, + "learning_rate": 4.66105562458177e-06, + "loss": 1.0963, + "step": 24546 + }, + { + "epoch": 2.07, + "learning_rate": 4.66028706344108e-06, + "loss": 1.0444, + "step": 24547 + }, + { + "epoch": 2.07, + "learning_rate": 4.659518546418805e-06, + "loss": 1.0617, + "step": 24548 + }, + { + "epoch": 2.07, + "learning_rate": 4.658750073521291e-06, + "loss": 1.1793, + "step": 24549 + }, + { + "epoch": 2.07, + "learning_rate": 4.657981644754902e-06, + "loss": 1.2336, + "step": 24550 + }, + { + "epoch": 2.07, + "learning_rate": 4.6572132601259744e-06, + "loss": 1.1048, + "step": 24551 + }, + { + "epoch": 2.07, + "learning_rate": 4.656444919640861e-06, + "loss": 1.2468, + "step": 24552 + }, + { + "epoch": 2.07, + "learning_rate": 4.655676623305909e-06, + "loss": 1.0233, + "step": 24553 + }, + { + "epoch": 2.07, + "learning_rate": 4.6549083711274675e-06, + "loss": 1.0768, + "step": 24554 + }, + { + "epoch": 2.07, + "learning_rate": 4.654140163111885e-06, + "loss": 1.1502, + "step": 24555 + }, + { + "epoch": 2.07, + "learning_rate": 4.653371999265509e-06, + "loss": 1.043, + "step": 24556 + }, + { + "epoch": 2.07, + "learning_rate": 4.6526038795946785e-06, + "loss": 1.0722, + "step": 24557 + }, + { + "epoch": 2.07, + "learning_rate": 4.651835804105753e-06, + "loss": 1.2124, + "step": 24558 + }, + { + "epoch": 2.07, + "learning_rate": 4.651067772805068e-06, + "loss": 1.1389, + "step": 24559 + }, + { + "epoch": 2.07, + "learning_rate": 4.650299785698974e-06, + "loss": 1.0361, + "step": 24560 + }, + { + "epoch": 2.07, + "learning_rate": 4.649531842793814e-06, + "loss": 1.1305, + "step": 24561 + }, + { + "epoch": 2.07, + "learning_rate": 4.648763944095934e-06, + "loss": 1.1512, + "step": 24562 + }, + { + "epoch": 2.07, + "learning_rate": 4.6479960896116805e-06, + "loss": 1.0455, + "step": 24563 + }, + { + "epoch": 2.07, + "learning_rate": 4.647228279347395e-06, + "loss": 1.1992, + "step": 24564 + }, + { + "epoch": 2.07, + "learning_rate": 4.646460513309427e-06, + "loss": 1.2084, + "step": 24565 + }, + { + "epoch": 2.07, + "learning_rate": 4.6456927915041075e-06, + "loss": 1.0745, + "step": 24566 + }, + { + "epoch": 2.07, + "learning_rate": 4.644925113937796e-06, + "loss": 1.1395, + "step": 24567 + }, + { + "epoch": 2.07, + "learning_rate": 4.644157480616822e-06, + "loss": 1.1999, + "step": 24568 + }, + { + "epoch": 2.07, + "learning_rate": 4.643389891547534e-06, + "loss": 1.061, + "step": 24569 + }, + { + "epoch": 2.07, + "learning_rate": 4.642622346736273e-06, + "loss": 1.1135, + "step": 24570 + }, + { + "epoch": 2.07, + "learning_rate": 4.641854846189381e-06, + "loss": 1.0586, + "step": 24571 + }, + { + "epoch": 2.07, + "learning_rate": 4.641087389913198e-06, + "loss": 1.0894, + "step": 24572 + }, + { + "epoch": 2.07, + "learning_rate": 4.640319977914067e-06, + "loss": 1.1562, + "step": 24573 + }, + { + "epoch": 2.07, + "learning_rate": 4.639552610198327e-06, + "loss": 1.1849, + "step": 24574 + }, + { + "epoch": 2.07, + "learning_rate": 4.638785286772319e-06, + "loss": 1.1648, + "step": 24575 + }, + { + "epoch": 2.07, + "learning_rate": 4.638018007642387e-06, + "loss": 1.0918, + "step": 24576 + }, + { + "epoch": 2.07, + "learning_rate": 4.637250772814859e-06, + "loss": 1.0952, + "step": 24577 + }, + { + "epoch": 2.07, + "learning_rate": 4.636483582296088e-06, + "loss": 1.194, + "step": 24578 + }, + { + "epoch": 2.07, + "learning_rate": 4.635716436092403e-06, + "loss": 1.1065, + "step": 24579 + }, + { + "epoch": 2.07, + "learning_rate": 4.634949334210146e-06, + "loss": 1.0435, + "step": 24580 + }, + { + "epoch": 2.07, + "learning_rate": 4.6341822766556544e-06, + "loss": 1.0405, + "step": 24581 + }, + { + "epoch": 2.07, + "learning_rate": 4.633415263435266e-06, + "loss": 1.1394, + "step": 24582 + }, + { + "epoch": 2.07, + "learning_rate": 4.632648294555318e-06, + "loss": 1.0969, + "step": 24583 + }, + { + "epoch": 2.07, + "learning_rate": 4.631881370022151e-06, + "loss": 1.1121, + "step": 24584 + }, + { + "epoch": 2.07, + "learning_rate": 4.631114489842091e-06, + "loss": 1.1551, + "step": 24585 + }, + { + "epoch": 2.07, + "learning_rate": 4.630347654021486e-06, + "loss": 1.133, + "step": 24586 + }, + { + "epoch": 2.07, + "learning_rate": 4.629580862566669e-06, + "loss": 1.0833, + "step": 24587 + }, + { + "epoch": 2.07, + "learning_rate": 4.628814115483967e-06, + "loss": 1.0796, + "step": 24588 + }, + { + "epoch": 2.07, + "learning_rate": 4.62804741277973e-06, + "loss": 1.0719, + "step": 24589 + }, + { + "epoch": 2.07, + "learning_rate": 4.62728075446028e-06, + "loss": 1.0329, + "step": 24590 + }, + { + "epoch": 2.07, + "learning_rate": 4.626514140531957e-06, + "loss": 1.0654, + "step": 24591 + }, + { + "epoch": 2.07, + "learning_rate": 4.625747571001092e-06, + "loss": 1.0714, + "step": 24592 + }, + { + "epoch": 2.07, + "learning_rate": 4.6249810458740215e-06, + "loss": 1.1764, + "step": 24593 + }, + { + "epoch": 2.07, + "learning_rate": 4.624214565157079e-06, + "loss": 1.0943, + "step": 24594 + }, + { + "epoch": 2.07, + "learning_rate": 4.623448128856599e-06, + "loss": 1.0871, + "step": 24595 + }, + { + "epoch": 2.07, + "learning_rate": 4.622681736978904e-06, + "loss": 1.0212, + "step": 24596 + }, + { + "epoch": 2.07, + "learning_rate": 4.6219153895303404e-06, + "loss": 1.1216, + "step": 24597 + }, + { + "epoch": 2.07, + "learning_rate": 4.621149086517229e-06, + "loss": 1.0758, + "step": 24598 + }, + { + "epoch": 2.07, + "learning_rate": 4.620382827945903e-06, + "loss": 1.1182, + "step": 24599 + }, + { + "epoch": 2.07, + "learning_rate": 4.619616613822702e-06, + "loss": 1.1806, + "step": 24600 + }, + { + "epoch": 2.07, + "learning_rate": 4.6188504441539476e-06, + "loss": 1.0777, + "step": 24601 + }, + { + "epoch": 2.07, + "learning_rate": 4.618084318945972e-06, + "loss": 1.1094, + "step": 24602 + }, + { + "epoch": 2.07, + "learning_rate": 4.6173182382051075e-06, + "loss": 1.0776, + "step": 24603 + }, + { + "epoch": 2.07, + "learning_rate": 4.61655220193768e-06, + "loss": 1.1585, + "step": 24604 + }, + { + "epoch": 2.07, + "learning_rate": 4.615786210150024e-06, + "loss": 1.1152, + "step": 24605 + }, + { + "epoch": 2.07, + "learning_rate": 4.615020262848468e-06, + "loss": 1.1504, + "step": 24606 + }, + { + "epoch": 2.07, + "learning_rate": 4.614254360039331e-06, + "loss": 1.1709, + "step": 24607 + }, + { + "epoch": 2.07, + "learning_rate": 4.6134885017289545e-06, + "loss": 1.173, + "step": 24608 + }, + { + "epoch": 2.07, + "learning_rate": 4.612722687923658e-06, + "loss": 1.165, + "step": 24609 + }, + { + "epoch": 2.07, + "learning_rate": 4.611956918629769e-06, + "loss": 1.1224, + "step": 24610 + }, + { + "epoch": 2.07, + "learning_rate": 4.611191193853617e-06, + "loss": 1.0665, + "step": 24611 + }, + { + "epoch": 2.07, + "learning_rate": 4.610425513601528e-06, + "loss": 1.0905, + "step": 24612 + }, + { + "epoch": 2.07, + "learning_rate": 4.609659877879828e-06, + "loss": 1.0697, + "step": 24613 + }, + { + "epoch": 2.07, + "learning_rate": 4.608894286694844e-06, + "loss": 1.1815, + "step": 24614 + }, + { + "epoch": 2.07, + "learning_rate": 4.608128740052903e-06, + "loss": 1.0641, + "step": 24615 + }, + { + "epoch": 2.07, + "learning_rate": 4.6073632379603216e-06, + "loss": 1.1321, + "step": 24616 + }, + { + "epoch": 2.07, + "learning_rate": 4.606597780423439e-06, + "loss": 1.0795, + "step": 24617 + }, + { + "epoch": 2.07, + "learning_rate": 4.605832367448567e-06, + "loss": 1.0031, + "step": 24618 + }, + { + "epoch": 2.07, + "learning_rate": 4.605066999042035e-06, + "loss": 1.1199, + "step": 24619 + }, + { + "epoch": 2.07, + "learning_rate": 4.604301675210165e-06, + "loss": 1.1452, + "step": 24620 + }, + { + "epoch": 2.07, + "learning_rate": 4.603536395959282e-06, + "loss": 1.1299, + "step": 24621 + }, + { + "epoch": 2.07, + "learning_rate": 4.602771161295709e-06, + "loss": 1.1701, + "step": 24622 + }, + { + "epoch": 2.07, + "learning_rate": 4.602005971225767e-06, + "loss": 1.179, + "step": 24623 + }, + { + "epoch": 2.07, + "learning_rate": 4.60124082575578e-06, + "loss": 1.0776, + "step": 24624 + }, + { + "epoch": 2.07, + "learning_rate": 4.600475724892069e-06, + "loss": 1.1228, + "step": 24625 + }, + { + "epoch": 2.07, + "learning_rate": 4.599710668640959e-06, + "loss": 1.1178, + "step": 24626 + }, + { + "epoch": 2.07, + "learning_rate": 4.59894565700876e-06, + "loss": 1.1269, + "step": 24627 + }, + { + "epoch": 2.07, + "learning_rate": 4.598180690001809e-06, + "loss": 1.1188, + "step": 24628 + }, + { + "epoch": 2.07, + "learning_rate": 4.597415767626413e-06, + "loss": 1.1026, + "step": 24629 + }, + { + "epoch": 2.07, + "learning_rate": 4.596650889888898e-06, + "loss": 1.1169, + "step": 24630 + }, + { + "epoch": 2.07, + "learning_rate": 4.595886056795582e-06, + "loss": 1.152, + "step": 24631 + }, + { + "epoch": 2.07, + "learning_rate": 4.595121268352786e-06, + "loss": 1.0993, + "step": 24632 + }, + { + "epoch": 2.07, + "learning_rate": 4.594356524566827e-06, + "loss": 1.1187, + "step": 24633 + }, + { + "epoch": 2.07, + "learning_rate": 4.593591825444028e-06, + "loss": 1.1499, + "step": 24634 + }, + { + "epoch": 2.07, + "learning_rate": 4.592827170990697e-06, + "loss": 1.221, + "step": 24635 + }, + { + "epoch": 2.07, + "learning_rate": 4.592062561213161e-06, + "loss": 1.0702, + "step": 24636 + }, + { + "epoch": 2.07, + "learning_rate": 4.59129799611774e-06, + "loss": 1.0936, + "step": 24637 + }, + { + "epoch": 2.07, + "learning_rate": 4.590533475710737e-06, + "loss": 1.1312, + "step": 24638 + }, + { + "epoch": 2.07, + "learning_rate": 4.589768999998485e-06, + "loss": 1.2162, + "step": 24639 + }, + { + "epoch": 2.07, + "learning_rate": 4.589004568987291e-06, + "loss": 1.085, + "step": 24640 + }, + { + "epoch": 2.07, + "learning_rate": 4.588240182683472e-06, + "loss": 1.1189, + "step": 24641 + }, + { + "epoch": 2.07, + "learning_rate": 4.587475841093345e-06, + "loss": 1.1141, + "step": 24642 + }, + { + "epoch": 2.07, + "learning_rate": 4.5867115442232245e-06, + "loss": 1.0984, + "step": 24643 + }, + { + "epoch": 2.07, + "learning_rate": 4.585947292079426e-06, + "loss": 1.1626, + "step": 24644 + }, + { + "epoch": 2.07, + "learning_rate": 4.5851830846682664e-06, + "loss": 1.2012, + "step": 24645 + }, + { + "epoch": 2.07, + "learning_rate": 4.58441892199605e-06, + "loss": 1.0121, + "step": 24646 + }, + { + "epoch": 2.08, + "learning_rate": 4.583654804069107e-06, + "loss": 1.1322, + "step": 24647 + }, + { + "epoch": 2.08, + "learning_rate": 4.582890730893736e-06, + "loss": 1.0805, + "step": 24648 + }, + { + "epoch": 2.08, + "learning_rate": 4.582126702476251e-06, + "loss": 1.0869, + "step": 24649 + }, + { + "epoch": 2.08, + "learning_rate": 4.5813627188229784e-06, + "loss": 1.107, + "step": 24650 + }, + { + "epoch": 2.08, + "learning_rate": 4.5805987799402165e-06, + "loss": 1.1499, + "step": 24651 + }, + { + "epoch": 2.08, + "learning_rate": 4.5798348858342825e-06, + "loss": 1.1047, + "step": 24652 + }, + { + "epoch": 2.08, + "learning_rate": 4.579071036511486e-06, + "loss": 1.1675, + "step": 24653 + }, + { + "epoch": 2.08, + "learning_rate": 4.5783072319781405e-06, + "loss": 1.1307, + "step": 24654 + }, + { + "epoch": 2.08, + "learning_rate": 4.577543472240556e-06, + "loss": 1.0293, + "step": 24655 + }, + { + "epoch": 2.08, + "learning_rate": 4.576779757305046e-06, + "loss": 1.0574, + "step": 24656 + }, + { + "epoch": 2.08, + "learning_rate": 4.57601608717791e-06, + "loss": 1.0699, + "step": 24657 + }, + { + "epoch": 2.08, + "learning_rate": 4.575252461865472e-06, + "loss": 1.1172, + "step": 24658 + }, + { + "epoch": 2.08, + "learning_rate": 4.574488881374032e-06, + "loss": 1.1408, + "step": 24659 + }, + { + "epoch": 2.08, + "learning_rate": 4.573725345709901e-06, + "loss": 1.1016, + "step": 24660 + }, + { + "epoch": 2.08, + "learning_rate": 4.572961854879386e-06, + "loss": 1.0975, + "step": 24661 + }, + { + "epoch": 2.08, + "learning_rate": 4.572198408888799e-06, + "loss": 1.1595, + "step": 24662 + }, + { + "epoch": 2.08, + "learning_rate": 4.571435007744445e-06, + "loss": 1.0371, + "step": 24663 + }, + { + "epoch": 2.08, + "learning_rate": 4.5706716514526314e-06, + "loss": 1.099, + "step": 24664 + }, + { + "epoch": 2.08, + "learning_rate": 4.569908340019672e-06, + "loss": 1.1425, + "step": 24665 + }, + { + "epoch": 2.08, + "learning_rate": 4.5691450734518585e-06, + "loss": 1.0274, + "step": 24666 + }, + { + "epoch": 2.08, + "learning_rate": 4.568381851755515e-06, + "loss": 1.0748, + "step": 24667 + }, + { + "epoch": 2.08, + "learning_rate": 4.567618674936934e-06, + "loss": 1.1588, + "step": 24668 + }, + { + "epoch": 2.08, + "learning_rate": 4.566855543002427e-06, + "loss": 1.1379, + "step": 24669 + }, + { + "epoch": 2.08, + "learning_rate": 4.5660924559583e-06, + "loss": 1.0273, + "step": 24670 + }, + { + "epoch": 2.08, + "learning_rate": 4.565329413810854e-06, + "loss": 1.1274, + "step": 24671 + }, + { + "epoch": 2.08, + "learning_rate": 4.564566416566396e-06, + "loss": 1.1989, + "step": 24672 + }, + { + "epoch": 2.08, + "learning_rate": 4.56380346423123e-06, + "loss": 1.1318, + "step": 24673 + }, + { + "epoch": 2.08, + "learning_rate": 4.5630405568116595e-06, + "loss": 1.1648, + "step": 24674 + }, + { + "epoch": 2.08, + "learning_rate": 4.562277694313989e-06, + "loss": 1.0836, + "step": 24675 + }, + { + "epoch": 2.08, + "learning_rate": 4.561514876744523e-06, + "loss": 1.0705, + "step": 24676 + }, + { + "epoch": 2.08, + "learning_rate": 4.560752104109555e-06, + "loss": 1.1341, + "step": 24677 + }, + { + "epoch": 2.08, + "learning_rate": 4.5599893764154e-06, + "loss": 1.0527, + "step": 24678 + }, + { + "epoch": 2.08, + "learning_rate": 4.559226693668352e-06, + "loss": 1.0584, + "step": 24679 + }, + { + "epoch": 2.08, + "learning_rate": 4.5584640558747125e-06, + "loss": 1.0892, + "step": 24680 + }, + { + "epoch": 2.08, + "learning_rate": 4.557701463040786e-06, + "loss": 1.1395, + "step": 24681 + }, + { + "epoch": 2.08, + "learning_rate": 4.55693891517287e-06, + "loss": 1.201, + "step": 24682 + }, + { + "epoch": 2.08, + "learning_rate": 4.556176412277268e-06, + "loss": 1.0843, + "step": 24683 + }, + { + "epoch": 2.08, + "learning_rate": 4.555413954360283e-06, + "loss": 1.1333, + "step": 24684 + }, + { + "epoch": 2.08, + "learning_rate": 4.5546515414282015e-06, + "loss": 1.1052, + "step": 24685 + }, + { + "epoch": 2.08, + "learning_rate": 4.5538891734873355e-06, + "loss": 1.1233, + "step": 24686 + }, + { + "epoch": 2.08, + "learning_rate": 4.553126850543985e-06, + "loss": 1.0893, + "step": 24687 + }, + { + "epoch": 2.08, + "learning_rate": 4.5523645726044356e-06, + "loss": 1.0397, + "step": 24688 + }, + { + "epoch": 2.08, + "learning_rate": 4.551602339675001e-06, + "loss": 1.1302, + "step": 24689 + }, + { + "epoch": 2.08, + "learning_rate": 4.550840151761968e-06, + "loss": 1.0618, + "step": 24690 + }, + { + "epoch": 2.08, + "learning_rate": 4.550078008871638e-06, + "loss": 1.1672, + "step": 24691 + }, + { + "epoch": 2.08, + "learning_rate": 4.549315911010307e-06, + "loss": 1.0649, + "step": 24692 + }, + { + "epoch": 2.08, + "learning_rate": 4.548553858184273e-06, + "loss": 1.017, + "step": 24693 + }, + { + "epoch": 2.08, + "learning_rate": 4.547791850399831e-06, + "loss": 1.0963, + "step": 24694 + }, + { + "epoch": 2.08, + "learning_rate": 4.547029887663281e-06, + "loss": 1.0617, + "step": 24695 + }, + { + "epoch": 2.08, + "learning_rate": 4.546267969980908e-06, + "loss": 1.1029, + "step": 24696 + }, + { + "epoch": 2.08, + "learning_rate": 4.545506097359023e-06, + "loss": 1.1087, + "step": 24697 + }, + { + "epoch": 2.08, + "learning_rate": 4.544744269803908e-06, + "loss": 1.0883, + "step": 24698 + }, + { + "epoch": 2.08, + "learning_rate": 4.543982487321859e-06, + "loss": 1.1601, + "step": 24699 + }, + { + "epoch": 2.08, + "learning_rate": 4.543220749919178e-06, + "loss": 1.0902, + "step": 24700 + }, + { + "epoch": 2.08, + "learning_rate": 4.542459057602151e-06, + "loss": 1.0449, + "step": 24701 + }, + { + "epoch": 2.08, + "learning_rate": 4.541697410377074e-06, + "loss": 1.0863, + "step": 24702 + }, + { + "epoch": 2.08, + "learning_rate": 4.540935808250241e-06, + "loss": 1.1455, + "step": 24703 + }, + { + "epoch": 2.08, + "learning_rate": 4.540174251227941e-06, + "loss": 1.1859, + "step": 24704 + }, + { + "epoch": 2.08, + "learning_rate": 4.539412739316471e-06, + "loss": 1.1402, + "step": 24705 + }, + { + "epoch": 2.08, + "learning_rate": 4.5386512725221235e-06, + "loss": 1.0946, + "step": 24706 + }, + { + "epoch": 2.08, + "learning_rate": 4.53788985085118e-06, + "loss": 1.1323, + "step": 24707 + }, + { + "epoch": 2.08, + "learning_rate": 4.5371284743099445e-06, + "loss": 1.1508, + "step": 24708 + }, + { + "epoch": 2.08, + "learning_rate": 4.5363671429047e-06, + "loss": 1.0459, + "step": 24709 + }, + { + "epoch": 2.08, + "learning_rate": 4.535605856641737e-06, + "loss": 1.0857, + "step": 24710 + }, + { + "epoch": 2.08, + "learning_rate": 4.534844615527348e-06, + "loss": 1.1243, + "step": 24711 + }, + { + "epoch": 2.08, + "learning_rate": 4.5340834195678225e-06, + "loss": 1.1236, + "step": 24712 + }, + { + "epoch": 2.08, + "learning_rate": 4.533322268769449e-06, + "loss": 1.0928, + "step": 24713 + }, + { + "epoch": 2.08, + "learning_rate": 4.5325611631385156e-06, + "loss": 1.2066, + "step": 24714 + }, + { + "epoch": 2.08, + "learning_rate": 4.531800102681316e-06, + "loss": 1.1482, + "step": 24715 + }, + { + "epoch": 2.08, + "learning_rate": 4.5310390874041265e-06, + "loss": 1.1577, + "step": 24716 + }, + { + "epoch": 2.08, + "learning_rate": 4.53027811731325e-06, + "loss": 1.1366, + "step": 24717 + }, + { + "epoch": 2.08, + "learning_rate": 4.5295171924149615e-06, + "loss": 1.0817, + "step": 24718 + }, + { + "epoch": 2.08, + "learning_rate": 4.5287563127155545e-06, + "loss": 1.0905, + "step": 24719 + }, + { + "epoch": 2.08, + "learning_rate": 4.527995478221313e-06, + "loss": 1.1578, + "step": 24720 + }, + { + "epoch": 2.08, + "learning_rate": 4.527234688938524e-06, + "loss": 1.1336, + "step": 24721 + }, + { + "epoch": 2.08, + "learning_rate": 4.526473944873473e-06, + "loss": 1.156, + "step": 24722 + }, + { + "epoch": 2.08, + "learning_rate": 4.525713246032447e-06, + "loss": 1.0798, + "step": 24723 + }, + { + "epoch": 2.08, + "learning_rate": 4.52495259242173e-06, + "loss": 1.0919, + "step": 24724 + }, + { + "epoch": 2.08, + "learning_rate": 4.524191984047605e-06, + "loss": 1.156, + "step": 24725 + }, + { + "epoch": 2.08, + "learning_rate": 4.523431420916364e-06, + "loss": 1.1242, + "step": 24726 + }, + { + "epoch": 2.08, + "learning_rate": 4.5226709030342785e-06, + "loss": 1.0971, + "step": 24727 + }, + { + "epoch": 2.08, + "learning_rate": 4.521910430407645e-06, + "loss": 1.0265, + "step": 24728 + }, + { + "epoch": 2.08, + "learning_rate": 4.521150003042738e-06, + "loss": 1.1278, + "step": 24729 + }, + { + "epoch": 2.08, + "learning_rate": 4.520389620945842e-06, + "loss": 1.1705, + "step": 24730 + }, + { + "epoch": 2.08, + "learning_rate": 4.519629284123241e-06, + "loss": 1.1117, + "step": 24731 + }, + { + "epoch": 2.08, + "learning_rate": 4.518868992581218e-06, + "loss": 1.126, + "step": 24732 + }, + { + "epoch": 2.08, + "learning_rate": 4.518108746326052e-06, + "loss": 1.1068, + "step": 24733 + }, + { + "epoch": 2.08, + "learning_rate": 4.517348545364031e-06, + "loss": 1.0636, + "step": 24734 + }, + { + "epoch": 2.08, + "learning_rate": 4.516588389701423e-06, + "loss": 1.1688, + "step": 24735 + }, + { + "epoch": 2.08, + "learning_rate": 4.5158282793445194e-06, + "loss": 1.0977, + "step": 24736 + }, + { + "epoch": 2.08, + "learning_rate": 4.515068214299603e-06, + "loss": 1.0725, + "step": 24737 + }, + { + "epoch": 2.08, + "learning_rate": 4.51430819457294e-06, + "loss": 1.0512, + "step": 24738 + }, + { + "epoch": 2.08, + "learning_rate": 4.5135482201708254e-06, + "loss": 1.1098, + "step": 24739 + }, + { + "epoch": 2.08, + "learning_rate": 4.512788291099528e-06, + "loss": 1.1646, + "step": 24740 + }, + { + "epoch": 2.08, + "learning_rate": 4.512028407365331e-06, + "loss": 1.1268, + "step": 24741 + }, + { + "epoch": 2.08, + "learning_rate": 4.51126856897451e-06, + "loss": 1.1795, + "step": 24742 + }, + { + "epoch": 2.08, + "learning_rate": 4.510508775933345e-06, + "loss": 1.0789, + "step": 24743 + }, + { + "epoch": 2.08, + "learning_rate": 4.509749028248115e-06, + "loss": 1.1269, + "step": 24744 + }, + { + "epoch": 2.08, + "learning_rate": 4.508989325925097e-06, + "loss": 1.0589, + "step": 24745 + }, + { + "epoch": 2.08, + "learning_rate": 4.508229668970561e-06, + "loss": 1.121, + "step": 24746 + }, + { + "epoch": 2.08, + "learning_rate": 4.507470057390796e-06, + "loss": 1.0901, + "step": 24747 + }, + { + "epoch": 2.08, + "learning_rate": 4.506710491192067e-06, + "loss": 1.0555, + "step": 24748 + }, + { + "epoch": 2.08, + "learning_rate": 4.50595097038065e-06, + "loss": 1.1202, + "step": 24749 + }, + { + "epoch": 2.08, + "learning_rate": 4.505191494962833e-06, + "loss": 1.1375, + "step": 24750 + }, + { + "epoch": 2.08, + "learning_rate": 4.504432064944878e-06, + "loss": 1.112, + "step": 24751 + }, + { + "epoch": 2.08, + "learning_rate": 4.503672680333064e-06, + "loss": 1.0948, + "step": 24752 + }, + { + "epoch": 2.08, + "learning_rate": 4.502913341133666e-06, + "loss": 1.178, + "step": 24753 + }, + { + "epoch": 2.08, + "learning_rate": 4.502154047352957e-06, + "loss": 1.1303, + "step": 24754 + }, + { + "epoch": 2.08, + "learning_rate": 4.501394798997211e-06, + "loss": 1.0418, + "step": 24755 + }, + { + "epoch": 2.08, + "learning_rate": 4.500635596072703e-06, + "loss": 1.1137, + "step": 24756 + }, + { + "epoch": 2.08, + "learning_rate": 4.499876438585697e-06, + "loss": 1.067, + "step": 24757 + }, + { + "epoch": 2.08, + "learning_rate": 4.499117326542482e-06, + "loss": 1.0858, + "step": 24758 + }, + { + "epoch": 2.08, + "learning_rate": 4.498358259949313e-06, + "loss": 1.1017, + "step": 24759 + }, + { + "epoch": 2.08, + "learning_rate": 4.497599238812471e-06, + "loss": 1.1671, + "step": 24760 + }, + { + "epoch": 2.08, + "learning_rate": 4.496840263138225e-06, + "loss": 1.0615, + "step": 24761 + }, + { + "epoch": 2.08, + "learning_rate": 4.496081332932845e-06, + "loss": 1.1132, + "step": 24762 + }, + { + "epoch": 2.08, + "learning_rate": 4.495322448202603e-06, + "loss": 1.1483, + "step": 24763 + }, + { + "epoch": 2.08, + "learning_rate": 4.494563608953768e-06, + "loss": 1.0355, + "step": 24764 + }, + { + "epoch": 2.09, + "learning_rate": 4.493804815192615e-06, + "loss": 1.0288, + "step": 24765 + }, + { + "epoch": 2.09, + "learning_rate": 4.493046066925401e-06, + "loss": 1.0808, + "step": 24766 + }, + { + "epoch": 2.09, + "learning_rate": 4.492287364158411e-06, + "loss": 1.1684, + "step": 24767 + }, + { + "epoch": 2.09, + "learning_rate": 4.4915287068979005e-06, + "loss": 1.132, + "step": 24768 + }, + { + "epoch": 2.09, + "learning_rate": 4.490770095150144e-06, + "loss": 1.0612, + "step": 24769 + }, + { + "epoch": 2.09, + "learning_rate": 4.490011528921407e-06, + "loss": 1.0394, + "step": 24770 + }, + { + "epoch": 2.09, + "learning_rate": 4.4892530082179595e-06, + "loss": 1.1117, + "step": 24771 + }, + { + "epoch": 2.09, + "learning_rate": 4.488494533046066e-06, + "loss": 1.1691, + "step": 24772 + }, + { + "epoch": 2.09, + "learning_rate": 4.487736103411996e-06, + "loss": 1.1812, + "step": 24773 + }, + { + "epoch": 2.09, + "learning_rate": 4.486977719322013e-06, + "loss": 1.04, + "step": 24774 + }, + { + "epoch": 2.09, + "learning_rate": 4.486219380782385e-06, + "loss": 1.1359, + "step": 24775 + }, + { + "epoch": 2.09, + "learning_rate": 4.485461087799381e-06, + "loss": 1.0702, + "step": 24776 + }, + { + "epoch": 2.09, + "learning_rate": 4.4847028403792545e-06, + "loss": 1.1339, + "step": 24777 + }, + { + "epoch": 2.09, + "learning_rate": 4.483944638528287e-06, + "loss": 1.1312, + "step": 24778 + }, + { + "epoch": 2.09, + "learning_rate": 4.48318648225273e-06, + "loss": 1.1673, + "step": 24779 + }, + { + "epoch": 2.09, + "learning_rate": 4.482428371558851e-06, + "loss": 1.1572, + "step": 24780 + }, + { + "epoch": 2.09, + "learning_rate": 4.481670306452916e-06, + "loss": 1.0717, + "step": 24781 + }, + { + "epoch": 2.09, + "learning_rate": 4.4809122869411845e-06, + "loss": 1.0453, + "step": 24782 + }, + { + "epoch": 2.09, + "learning_rate": 4.4801543130299245e-06, + "loss": 1.1403, + "step": 24783 + }, + { + "epoch": 2.09, + "learning_rate": 4.4793963847253984e-06, + "loss": 1.1454, + "step": 24784 + }, + { + "epoch": 2.09, + "learning_rate": 4.47863850203386e-06, + "loss": 1.1079, + "step": 24785 + }, + { + "epoch": 2.09, + "learning_rate": 4.47788066496158e-06, + "loss": 1.0491, + "step": 24786 + }, + { + "epoch": 2.09, + "learning_rate": 4.477122873514822e-06, + "loss": 1.0645, + "step": 24787 + }, + { + "epoch": 2.09, + "learning_rate": 4.476365127699834e-06, + "loss": 1.2404, + "step": 24788 + }, + { + "epoch": 2.09, + "learning_rate": 4.475607427522894e-06, + "loss": 1.1831, + "step": 24789 + }, + { + "epoch": 2.09, + "learning_rate": 4.474849772990247e-06, + "loss": 1.1382, + "step": 24790 + }, + { + "epoch": 2.09, + "learning_rate": 4.474092164108161e-06, + "loss": 1.0524, + "step": 24791 + }, + { + "epoch": 2.09, + "learning_rate": 4.473334600882895e-06, + "loss": 1.0291, + "step": 24792 + }, + { + "epoch": 2.09, + "learning_rate": 4.472577083320706e-06, + "loss": 1.2398, + "step": 24793 + }, + { + "epoch": 2.09, + "learning_rate": 4.471819611427855e-06, + "loss": 1.1427, + "step": 24794 + }, + { + "epoch": 2.09, + "learning_rate": 4.471062185210603e-06, + "loss": 1.0578, + "step": 24795 + }, + { + "epoch": 2.09, + "learning_rate": 4.470304804675197e-06, + "loss": 1.1509, + "step": 24796 + }, + { + "epoch": 2.09, + "learning_rate": 4.469547469827911e-06, + "loss": 1.1003, + "step": 24797 + }, + { + "epoch": 2.09, + "learning_rate": 4.4687901806749885e-06, + "loss": 1.1744, + "step": 24798 + }, + { + "epoch": 2.09, + "learning_rate": 4.468032937222688e-06, + "loss": 1.1541, + "step": 24799 + }, + { + "epoch": 2.09, + "learning_rate": 4.467275739477278e-06, + "loss": 1.078, + "step": 24800 + }, + { + "epoch": 2.09, + "learning_rate": 4.466518587445003e-06, + "loss": 1.0846, + "step": 24801 + }, + { + "epoch": 2.09, + "learning_rate": 4.465761481132123e-06, + "loss": 1.1153, + "step": 24802 + }, + { + "epoch": 2.09, + "learning_rate": 4.465004420544892e-06, + "loss": 1.048, + "step": 24803 + }, + { + "epoch": 2.09, + "learning_rate": 4.464247405689565e-06, + "loss": 1.2353, + "step": 24804 + }, + { + "epoch": 2.09, + "learning_rate": 4.463490436572399e-06, + "loss": 1.1468, + "step": 24805 + }, + { + "epoch": 2.09, + "learning_rate": 4.4627335131996495e-06, + "loss": 1.092, + "step": 24806 + }, + { + "epoch": 2.09, + "learning_rate": 4.46197663557756e-06, + "loss": 1.0975, + "step": 24807 + }, + { + "epoch": 2.09, + "learning_rate": 4.461219803712401e-06, + "loss": 1.22, + "step": 24808 + }, + { + "epoch": 2.09, + "learning_rate": 4.4604630176104125e-06, + "loss": 1.0328, + "step": 24809 + }, + { + "epoch": 2.09, + "learning_rate": 4.459706277277851e-06, + "loss": 1.1912, + "step": 24810 + }, + { + "epoch": 2.09, + "learning_rate": 4.45894958272097e-06, + "loss": 1.1389, + "step": 24811 + }, + { + "epoch": 2.09, + "learning_rate": 4.45819293394602e-06, + "loss": 1.0655, + "step": 24812 + }, + { + "epoch": 2.09, + "learning_rate": 4.457436330959255e-06, + "loss": 1.2229, + "step": 24813 + }, + { + "epoch": 2.09, + "learning_rate": 4.456679773766923e-06, + "loss": 1.0914, + "step": 24814 + }, + { + "epoch": 2.09, + "learning_rate": 4.455923262375278e-06, + "loss": 1.1828, + "step": 24815 + }, + { + "epoch": 2.09, + "learning_rate": 4.45516679679057e-06, + "loss": 1.1073, + "step": 24816 + }, + { + "epoch": 2.09, + "learning_rate": 4.4544103770190505e-06, + "loss": 1.1327, + "step": 24817 + }, + { + "epoch": 2.09, + "learning_rate": 4.453654003066961e-06, + "loss": 1.0166, + "step": 24818 + }, + { + "epoch": 2.09, + "learning_rate": 4.452897674940564e-06, + "loss": 1.0834, + "step": 24819 + }, + { + "epoch": 2.09, + "learning_rate": 4.4521413926460984e-06, + "loss": 1.1942, + "step": 24820 + }, + { + "epoch": 2.09, + "learning_rate": 4.451385156189816e-06, + "loss": 1.096, + "step": 24821 + }, + { + "epoch": 2.09, + "learning_rate": 4.4506289655779655e-06, + "loss": 1.0268, + "step": 24822 + }, + { + "epoch": 2.09, + "learning_rate": 4.449872820816794e-06, + "loss": 1.0938, + "step": 24823 + }, + { + "epoch": 2.09, + "learning_rate": 4.4491167219125496e-06, + "loss": 1.1333, + "step": 24824 + }, + { + "epoch": 2.09, + "learning_rate": 4.44836066887148e-06, + "loss": 1.1421, + "step": 24825 + }, + { + "epoch": 2.09, + "learning_rate": 4.447604661699833e-06, + "loss": 1.0797, + "step": 24826 + }, + { + "epoch": 2.09, + "learning_rate": 4.446848700403849e-06, + "loss": 1.1537, + "step": 24827 + }, + { + "epoch": 2.09, + "learning_rate": 4.446092784989784e-06, + "loss": 1.0638, + "step": 24828 + }, + { + "epoch": 2.09, + "learning_rate": 4.445336915463874e-06, + "loss": 1.087, + "step": 24829 + }, + { + "epoch": 2.09, + "learning_rate": 4.444581091832368e-06, + "loss": 1.0466, + "step": 24830 + }, + { + "epoch": 2.09, + "learning_rate": 4.443825314101512e-06, + "loss": 1.1562, + "step": 24831 + }, + { + "epoch": 2.09, + "learning_rate": 4.443069582277548e-06, + "loss": 1.1446, + "step": 24832 + }, + { + "epoch": 2.09, + "learning_rate": 4.442313896366722e-06, + "loss": 1.1882, + "step": 24833 + }, + { + "epoch": 2.09, + "learning_rate": 4.441558256375282e-06, + "loss": 1.0992, + "step": 24834 + }, + { + "epoch": 2.09, + "learning_rate": 4.440802662309459e-06, + "loss": 1.136, + "step": 24835 + }, + { + "epoch": 2.09, + "learning_rate": 4.440047114175506e-06, + "loss": 1.1263, + "step": 24836 + }, + { + "epoch": 2.09, + "learning_rate": 4.439291611979668e-06, + "loss": 1.049, + "step": 24837 + }, + { + "epoch": 2.09, + "learning_rate": 4.438536155728175e-06, + "loss": 1.0402, + "step": 24838 + }, + { + "epoch": 2.09, + "learning_rate": 4.437780745427283e-06, + "loss": 1.1054, + "step": 24839 + }, + { + "epoch": 2.09, + "learning_rate": 4.437025381083223e-06, + "loss": 1.0523, + "step": 24840 + }, + { + "epoch": 2.09, + "learning_rate": 4.4362700627022405e-06, + "loss": 1.1503, + "step": 24841 + }, + { + "epoch": 2.09, + "learning_rate": 4.4355147902905744e-06, + "loss": 1.1654, + "step": 24842 + }, + { + "epoch": 2.09, + "learning_rate": 4.434759563854466e-06, + "loss": 1.1587, + "step": 24843 + }, + { + "epoch": 2.09, + "learning_rate": 4.434004383400156e-06, + "loss": 1.1633, + "step": 24844 + }, + { + "epoch": 2.09, + "learning_rate": 4.433249248933886e-06, + "loss": 1.1841, + "step": 24845 + }, + { + "epoch": 2.09, + "learning_rate": 4.432494160461886e-06, + "loss": 1.1362, + "step": 24846 + }, + { + "epoch": 2.09, + "learning_rate": 4.431739117990407e-06, + "loss": 1.0276, + "step": 24847 + }, + { + "epoch": 2.09, + "learning_rate": 4.430984121525679e-06, + "loss": 1.1386, + "step": 24848 + }, + { + "epoch": 2.09, + "learning_rate": 4.430229171073938e-06, + "loss": 1.0995, + "step": 24849 + }, + { + "epoch": 2.09, + "learning_rate": 4.429474266641434e-06, + "loss": 1.1192, + "step": 24850 + }, + { + "epoch": 2.09, + "learning_rate": 4.428719408234392e-06, + "loss": 1.1243, + "step": 24851 + }, + { + "epoch": 2.09, + "learning_rate": 4.427964595859053e-06, + "loss": 1.0896, + "step": 24852 + }, + { + "epoch": 2.09, + "learning_rate": 4.427209829521655e-06, + "loss": 1.1502, + "step": 24853 + }, + { + "epoch": 2.09, + "learning_rate": 4.4264551092284325e-06, + "loss": 1.104, + "step": 24854 + }, + { + "epoch": 2.09, + "learning_rate": 4.4257004349856214e-06, + "loss": 1.1533, + "step": 24855 + }, + { + "epoch": 2.09, + "learning_rate": 4.424945806799462e-06, + "loss": 1.1499, + "step": 24856 + }, + { + "epoch": 2.09, + "learning_rate": 4.4241912246761765e-06, + "loss": 1.0742, + "step": 24857 + }, + { + "epoch": 2.09, + "learning_rate": 4.423436688622015e-06, + "loss": 1.1769, + "step": 24858 + }, + { + "epoch": 2.09, + "learning_rate": 4.4226821986432004e-06, + "loss": 1.1189, + "step": 24859 + }, + { + "epoch": 2.09, + "learning_rate": 4.4219277547459705e-06, + "loss": 1.12, + "step": 24860 + }, + { + "epoch": 2.09, + "learning_rate": 4.421173356936559e-06, + "loss": 1.1409, + "step": 24861 + }, + { + "epoch": 2.09, + "learning_rate": 4.420419005221198e-06, + "loss": 1.1059, + "step": 24862 + }, + { + "epoch": 2.09, + "learning_rate": 4.419664699606121e-06, + "loss": 1.0756, + "step": 24863 + }, + { + "epoch": 2.09, + "learning_rate": 4.418910440097561e-06, + "loss": 1.1695, + "step": 24864 + }, + { + "epoch": 2.09, + "learning_rate": 4.418156226701749e-06, + "loss": 1.1326, + "step": 24865 + }, + { + "epoch": 2.09, + "learning_rate": 4.417402059424916e-06, + "loss": 1.105, + "step": 24866 + }, + { + "epoch": 2.09, + "learning_rate": 4.416647938273297e-06, + "loss": 1.1205, + "step": 24867 + }, + { + "epoch": 2.09, + "learning_rate": 4.415893863253114e-06, + "loss": 1.1801, + "step": 24868 + }, + { + "epoch": 2.09, + "learning_rate": 4.41513983437061e-06, + "loss": 1.1152, + "step": 24869 + }, + { + "epoch": 2.09, + "learning_rate": 4.4143858516320035e-06, + "loss": 1.1337, + "step": 24870 + }, + { + "epoch": 2.09, + "learning_rate": 4.413631915043529e-06, + "loss": 1.1479, + "step": 24871 + }, + { + "epoch": 2.09, + "learning_rate": 4.412878024611416e-06, + "loss": 1.1205, + "step": 24872 + }, + { + "epoch": 2.09, + "learning_rate": 4.412124180341893e-06, + "loss": 1.2228, + "step": 24873 + }, + { + "epoch": 2.09, + "learning_rate": 4.411370382241187e-06, + "loss": 1.1329, + "step": 24874 + }, + { + "epoch": 2.09, + "learning_rate": 4.410616630315528e-06, + "loss": 1.1942, + "step": 24875 + }, + { + "epoch": 2.09, + "learning_rate": 4.409862924571148e-06, + "loss": 1.1802, + "step": 24876 + }, + { + "epoch": 2.09, + "learning_rate": 4.4091092650142605e-06, + "loss": 1.1476, + "step": 24877 + }, + { + "epoch": 2.09, + "learning_rate": 4.40835565165111e-06, + "loss": 1.0084, + "step": 24878 + }, + { + "epoch": 2.09, + "learning_rate": 4.407602084487911e-06, + "loss": 1.0252, + "step": 24879 + }, + { + "epoch": 2.09, + "learning_rate": 4.406848563530893e-06, + "loss": 1.0635, + "step": 24880 + }, + { + "epoch": 2.09, + "learning_rate": 4.406095088786283e-06, + "loss": 1.1346, + "step": 24881 + }, + { + "epoch": 2.09, + "learning_rate": 4.405341660260305e-06, + "loss": 1.1015, + "step": 24882 + }, + { + "epoch": 2.09, + "learning_rate": 4.404588277959185e-06, + "loss": 1.1633, + "step": 24883 + }, + { + "epoch": 2.1, + "learning_rate": 4.403834941889151e-06, + "loss": 1.0976, + "step": 24884 + }, + { + "epoch": 2.1, + "learning_rate": 4.403081652056417e-06, + "loss": 1.0376, + "step": 24885 + }, + { + "epoch": 2.1, + "learning_rate": 4.402328408467216e-06, + "loss": 1.1454, + "step": 24886 + }, + { + "epoch": 2.1, + "learning_rate": 4.4015752111277735e-06, + "loss": 1.0196, + "step": 24887 + }, + { + "epoch": 2.1, + "learning_rate": 4.400822060044301e-06, + "loss": 1.1494, + "step": 24888 + }, + { + "epoch": 2.1, + "learning_rate": 4.400068955223035e-06, + "loss": 1.0136, + "step": 24889 + }, + { + "epoch": 2.1, + "learning_rate": 4.39931589667019e-06, + "loss": 1.177, + "step": 24890 + }, + { + "epoch": 2.1, + "learning_rate": 4.398562884391988e-06, + "loss": 1.1322, + "step": 24891 + }, + { + "epoch": 2.1, + "learning_rate": 4.397809918394652e-06, + "loss": 1.0476, + "step": 24892 + }, + { + "epoch": 2.1, + "learning_rate": 4.397056998684403e-06, + "loss": 1.0894, + "step": 24893 + }, + { + "epoch": 2.1, + "learning_rate": 4.396304125267464e-06, + "loss": 1.089, + "step": 24894 + }, + { + "epoch": 2.1, + "learning_rate": 4.395551298150056e-06, + "loss": 1.0818, + "step": 24895 + }, + { + "epoch": 2.1, + "learning_rate": 4.39479851733839e-06, + "loss": 1.1513, + "step": 24896 + }, + { + "epoch": 2.1, + "learning_rate": 4.3940457828387e-06, + "loss": 1.1733, + "step": 24897 + }, + { + "epoch": 2.1, + "learning_rate": 4.393293094657194e-06, + "loss": 1.0561, + "step": 24898 + }, + { + "epoch": 2.1, + "learning_rate": 4.392540452800092e-06, + "loss": 1.1811, + "step": 24899 + }, + { + "epoch": 2.1, + "learning_rate": 4.391787857273621e-06, + "loss": 1.0489, + "step": 24900 + }, + { + "epoch": 2.1, + "learning_rate": 4.391035308083991e-06, + "loss": 1.0854, + "step": 24901 + }, + { + "epoch": 2.1, + "learning_rate": 4.390282805237421e-06, + "loss": 1.1637, + "step": 24902 + }, + { + "epoch": 2.1, + "learning_rate": 4.389530348740132e-06, + "loss": 1.2435, + "step": 24903 + }, + { + "epoch": 2.1, + "learning_rate": 4.388777938598338e-06, + "loss": 1.1678, + "step": 24904 + }, + { + "epoch": 2.1, + "learning_rate": 4.388025574818257e-06, + "loss": 1.0535, + "step": 24905 + }, + { + "epoch": 2.1, + "learning_rate": 4.387273257406108e-06, + "loss": 1.0673, + "step": 24906 + }, + { + "epoch": 2.1, + "learning_rate": 4.386520986368095e-06, + "loss": 1.0532, + "step": 24907 + }, + { + "epoch": 2.1, + "learning_rate": 4.385768761710451e-06, + "loss": 1.1326, + "step": 24908 + }, + { + "epoch": 2.1, + "learning_rate": 4.385016583439377e-06, + "loss": 1.1267, + "step": 24909 + }, + { + "epoch": 2.1, + "learning_rate": 4.3842644515610945e-06, + "loss": 1.0635, + "step": 24910 + }, + { + "epoch": 2.1, + "learning_rate": 4.3835123660818156e-06, + "loss": 1.1521, + "step": 24911 + }, + { + "epoch": 2.1, + "learning_rate": 4.382760327007756e-06, + "loss": 1.1518, + "step": 24912 + }, + { + "epoch": 2.1, + "learning_rate": 4.382008334345128e-06, + "loss": 1.1514, + "step": 24913 + }, + { + "epoch": 2.1, + "learning_rate": 4.381256388100144e-06, + "loss": 1.0779, + "step": 24914 + }, + { + "epoch": 2.1, + "learning_rate": 4.3805044882790194e-06, + "loss": 1.0622, + "step": 24915 + }, + { + "epoch": 2.1, + "learning_rate": 4.379752634887965e-06, + "loss": 1.0273, + "step": 24916 + }, + { + "epoch": 2.1, + "learning_rate": 4.379000827933196e-06, + "loss": 1.1365, + "step": 24917 + }, + { + "epoch": 2.1, + "learning_rate": 4.3782490674209145e-06, + "loss": 1.0939, + "step": 24918 + }, + { + "epoch": 2.1, + "learning_rate": 4.3774973533573455e-06, + "loss": 1.1143, + "step": 24919 + }, + { + "epoch": 2.1, + "learning_rate": 4.376745685748689e-06, + "loss": 1.0842, + "step": 24920 + }, + { + "epoch": 2.1, + "learning_rate": 4.3759940646011585e-06, + "loss": 1.1882, + "step": 24921 + }, + { + "epoch": 2.1, + "learning_rate": 4.3752424899209666e-06, + "loss": 1.0887, + "step": 24922 + }, + { + "epoch": 2.1, + "learning_rate": 4.374490961714321e-06, + "loss": 1.1708, + "step": 24923 + }, + { + "epoch": 2.1, + "learning_rate": 4.37373947998743e-06, + "loss": 1.0628, + "step": 24924 + }, + { + "epoch": 2.1, + "learning_rate": 4.372988044746506e-06, + "loss": 1.1175, + "step": 24925 + }, + { + "epoch": 2.1, + "learning_rate": 4.372236655997758e-06, + "loss": 1.2769, + "step": 24926 + }, + { + "epoch": 2.1, + "learning_rate": 4.3714853137473846e-06, + "loss": 1.2134, + "step": 24927 + }, + { + "epoch": 2.1, + "learning_rate": 4.370734018001609e-06, + "loss": 1.0727, + "step": 24928 + }, + { + "epoch": 2.1, + "learning_rate": 4.3699827687666265e-06, + "loss": 1.2199, + "step": 24929 + }, + { + "epoch": 2.1, + "learning_rate": 4.369231566048648e-06, + "loss": 1.0363, + "step": 24930 + }, + { + "epoch": 2.1, + "learning_rate": 4.3684804098538805e-06, + "loss": 1.0669, + "step": 24931 + }, + { + "epoch": 2.1, + "learning_rate": 4.36772930018853e-06, + "loss": 1.127, + "step": 24932 + }, + { + "epoch": 2.1, + "learning_rate": 4.366978237058803e-06, + "loss": 1.1139, + "step": 24933 + }, + { + "epoch": 2.1, + "learning_rate": 4.366227220470908e-06, + "loss": 1.0975, + "step": 24934 + }, + { + "epoch": 2.1, + "learning_rate": 4.365476250431039e-06, + "loss": 1.1546, + "step": 24935 + }, + { + "epoch": 2.1, + "learning_rate": 4.364725326945413e-06, + "loss": 1.01, + "step": 24936 + }, + { + "epoch": 2.1, + "learning_rate": 4.363974450020233e-06, + "loss": 1.1055, + "step": 24937 + }, + { + "epoch": 2.1, + "learning_rate": 4.363223619661693e-06, + "loss": 1.0805, + "step": 24938 + }, + { + "epoch": 2.1, + "learning_rate": 4.362472835876011e-06, + "loss": 1.1544, + "step": 24939 + }, + { + "epoch": 2.1, + "learning_rate": 4.3617220986693775e-06, + "loss": 1.0905, + "step": 24940 + }, + { + "epoch": 2.1, + "learning_rate": 4.360971408048001e-06, + "loss": 1.0714, + "step": 24941 + }, + { + "epoch": 2.1, + "learning_rate": 4.360220764018085e-06, + "loss": 1.2181, + "step": 24942 + }, + { + "epoch": 2.1, + "learning_rate": 4.359470166585828e-06, + "loss": 1.1425, + "step": 24943 + }, + { + "epoch": 2.1, + "learning_rate": 4.3587196157574355e-06, + "loss": 1.1389, + "step": 24944 + }, + { + "epoch": 2.1, + "learning_rate": 4.357969111539108e-06, + "loss": 1.066, + "step": 24945 + }, + { + "epoch": 2.1, + "learning_rate": 4.3572186539370406e-06, + "loss": 1.1087, + "step": 24946 + }, + { + "epoch": 2.1, + "learning_rate": 4.356468242957446e-06, + "loss": 1.0904, + "step": 24947 + }, + { + "epoch": 2.1, + "learning_rate": 4.355717878606512e-06, + "loss": 1.0735, + "step": 24948 + }, + { + "epoch": 2.1, + "learning_rate": 4.35496756089044e-06, + "loss": 1.1454, + "step": 24949 + }, + { + "epoch": 2.1, + "learning_rate": 4.35421728981544e-06, + "loss": 1.0435, + "step": 24950 + }, + { + "epoch": 2.1, + "learning_rate": 4.3534670653877e-06, + "loss": 1.0331, + "step": 24951 + }, + { + "epoch": 2.1, + "learning_rate": 4.352716887613422e-06, + "loss": 1.1533, + "step": 24952 + }, + { + "epoch": 2.1, + "learning_rate": 4.351966756498804e-06, + "loss": 1.1199, + "step": 24953 + }, + { + "epoch": 2.1, + "learning_rate": 4.3512166720500455e-06, + "loss": 0.9139, + "step": 24954 + }, + { + "epoch": 2.1, + "learning_rate": 4.350466634273343e-06, + "loss": 1.1546, + "step": 24955 + }, + { + "epoch": 2.1, + "learning_rate": 4.349716643174895e-06, + "loss": 1.1584, + "step": 24956 + }, + { + "epoch": 2.1, + "learning_rate": 4.34896669876089e-06, + "loss": 1.1331, + "step": 24957 + }, + { + "epoch": 2.1, + "learning_rate": 4.348216801037538e-06, + "loss": 1.1661, + "step": 24958 + }, + { + "epoch": 2.1, + "learning_rate": 4.347466950011024e-06, + "loss": 1.1713, + "step": 24959 + }, + { + "epoch": 2.1, + "learning_rate": 4.346717145687548e-06, + "loss": 1.0877, + "step": 24960 + }, + { + "epoch": 2.1, + "learning_rate": 4.345967388073303e-06, + "loss": 1.173, + "step": 24961 + }, + { + "epoch": 2.1, + "learning_rate": 4.3452176771744845e-06, + "loss": 1.1517, + "step": 24962 + }, + { + "epoch": 2.1, + "learning_rate": 4.344468012997289e-06, + "loss": 1.1787, + "step": 24963 + }, + { + "epoch": 2.1, + "learning_rate": 4.343718395547908e-06, + "loss": 1.0646, + "step": 24964 + }, + { + "epoch": 2.1, + "learning_rate": 4.342968824832535e-06, + "loss": 1.0543, + "step": 24965 + }, + { + "epoch": 2.1, + "learning_rate": 4.342219300857366e-06, + "loss": 1.2513, + "step": 24966 + }, + { + "epoch": 2.1, + "learning_rate": 4.3414698236285935e-06, + "loss": 1.1403, + "step": 24967 + }, + { + "epoch": 2.1, + "learning_rate": 4.340720393152402e-06, + "loss": 1.0789, + "step": 24968 + }, + { + "epoch": 2.1, + "learning_rate": 4.3399710094349964e-06, + "loss": 1.0688, + "step": 24969 + }, + { + "epoch": 2.1, + "learning_rate": 4.3392216724825595e-06, + "loss": 1.072, + "step": 24970 + }, + { + "epoch": 2.1, + "learning_rate": 4.3384723823012846e-06, + "loss": 1.1448, + "step": 24971 + }, + { + "epoch": 2.1, + "learning_rate": 4.337723138897362e-06, + "loss": 1.1554, + "step": 24972 + }, + { + "epoch": 2.1, + "learning_rate": 4.336973942276983e-06, + "loss": 1.1221, + "step": 24973 + }, + { + "epoch": 2.1, + "learning_rate": 4.3362247924463375e-06, + "loss": 1.0355, + "step": 24974 + }, + { + "epoch": 2.1, + "learning_rate": 4.335475689411615e-06, + "loss": 1.0944, + "step": 24975 + }, + { + "epoch": 2.1, + "learning_rate": 4.334726633179011e-06, + "loss": 1.0867, + "step": 24976 + }, + { + "epoch": 2.1, + "learning_rate": 4.333977623754699e-06, + "loss": 1.0782, + "step": 24977 + }, + { + "epoch": 2.1, + "learning_rate": 4.333228661144886e-06, + "loss": 1.2082, + "step": 24978 + }, + { + "epoch": 2.1, + "learning_rate": 4.332479745355746e-06, + "loss": 1.1288, + "step": 24979 + }, + { + "epoch": 2.1, + "learning_rate": 4.331730876393474e-06, + "loss": 1.0955, + "step": 24980 + }, + { + "epoch": 2.1, + "learning_rate": 4.330982054264256e-06, + "loss": 1.0995, + "step": 24981 + }, + { + "epoch": 2.1, + "learning_rate": 4.330233278974277e-06, + "loss": 1.0569, + "step": 24982 + }, + { + "epoch": 2.1, + "learning_rate": 4.329484550529726e-06, + "loss": 1.0488, + "step": 24983 + }, + { + "epoch": 2.1, + "learning_rate": 4.328735868936792e-06, + "loss": 1.1592, + "step": 24984 + }, + { + "epoch": 2.1, + "learning_rate": 4.3279872342016506e-06, + "loss": 1.0799, + "step": 24985 + }, + { + "epoch": 2.1, + "learning_rate": 4.327238646330496e-06, + "loss": 1.1222, + "step": 24986 + }, + { + "epoch": 2.1, + "learning_rate": 4.326490105329516e-06, + "loss": 1.2107, + "step": 24987 + }, + { + "epoch": 2.1, + "learning_rate": 4.325741611204883e-06, + "loss": 1.1263, + "step": 24988 + }, + { + "epoch": 2.1, + "learning_rate": 4.324993163962796e-06, + "loss": 1.1649, + "step": 24989 + }, + { + "epoch": 2.1, + "learning_rate": 4.324244763609429e-06, + "loss": 1.1594, + "step": 24990 + }, + { + "epoch": 2.1, + "learning_rate": 4.323496410150967e-06, + "loss": 1.2155, + "step": 24991 + }, + { + "epoch": 2.1, + "learning_rate": 4.322748103593595e-06, + "loss": 1.159, + "step": 24992 + }, + { + "epoch": 2.1, + "learning_rate": 4.321999843943494e-06, + "loss": 1.108, + "step": 24993 + }, + { + "epoch": 2.1, + "learning_rate": 4.321251631206849e-06, + "loss": 1.1062, + "step": 24994 + }, + { + "epoch": 2.1, + "learning_rate": 4.320503465389843e-06, + "loss": 1.1515, + "step": 24995 + }, + { + "epoch": 2.1, + "learning_rate": 4.3197553464986484e-06, + "loss": 1.1705, + "step": 24996 + }, + { + "epoch": 2.1, + "learning_rate": 4.3190072745394604e-06, + "loss": 1.1505, + "step": 24997 + }, + { + "epoch": 2.1, + "learning_rate": 4.318259249518448e-06, + "loss": 1.0969, + "step": 24998 + }, + { + "epoch": 2.1, + "learning_rate": 4.317511271441793e-06, + "loss": 1.1598, + "step": 24999 + }, + { + "epoch": 2.1, + "learning_rate": 4.316763340315686e-06, + "loss": 1.0501, + "step": 25000 + }, + { + "epoch": 2.1, + "learning_rate": 4.316015456146295e-06, + "loss": 1.1158, + "step": 25001 + }, + { + "epoch": 2.1, + "learning_rate": 4.315267618939804e-06, + "loss": 1.1041, + "step": 25002 + }, + { + "epoch": 2.11, + "learning_rate": 4.314519828702391e-06, + "loss": 1.1515, + "step": 25003 + }, + { + "epoch": 2.11, + "learning_rate": 4.313772085440236e-06, + "loss": 1.2011, + "step": 25004 + }, + { + "epoch": 2.11, + "learning_rate": 4.3130243891595136e-06, + "loss": 1.1499, + "step": 25005 + }, + { + "epoch": 2.11, + "learning_rate": 4.312276739866408e-06, + "loss": 1.0812, + "step": 25006 + }, + { + "epoch": 2.11, + "learning_rate": 4.3115291375670854e-06, + "loss": 1.0879, + "step": 25007 + }, + { + "epoch": 2.11, + "learning_rate": 4.310781582267737e-06, + "loss": 1.1198, + "step": 25008 + }, + { + "epoch": 2.11, + "learning_rate": 4.310034073974528e-06, + "loss": 1.0772, + "step": 25009 + }, + { + "epoch": 2.11, + "learning_rate": 4.309286612693635e-06, + "loss": 1.0618, + "step": 25010 + }, + { + "epoch": 2.11, + "learning_rate": 4.308539198431245e-06, + "loss": 1.0991, + "step": 25011 + }, + { + "epoch": 2.11, + "learning_rate": 4.307791831193522e-06, + "loss": 1.0857, + "step": 25012 + }, + { + "epoch": 2.11, + "learning_rate": 4.307044510986645e-06, + "loss": 1.1372, + "step": 25013 + }, + { + "epoch": 2.11, + "learning_rate": 4.306297237816787e-06, + "loss": 1.0884, + "step": 25014 + }, + { + "epoch": 2.11, + "learning_rate": 4.305550011690123e-06, + "loss": 1.0953, + "step": 25015 + }, + { + "epoch": 2.11, + "learning_rate": 4.304802832612829e-06, + "loss": 1.1428, + "step": 25016 + }, + { + "epoch": 2.11, + "learning_rate": 4.304055700591079e-06, + "loss": 1.1912, + "step": 25017 + }, + { + "epoch": 2.11, + "learning_rate": 4.3033086156310365e-06, + "loss": 1.3145, + "step": 25018 + }, + { + "epoch": 2.11, + "learning_rate": 4.30256157773889e-06, + "loss": 1.1144, + "step": 25019 + }, + { + "epoch": 2.11, + "learning_rate": 4.301814586920799e-06, + "loss": 1.0358, + "step": 25020 + }, + { + "epoch": 2.11, + "learning_rate": 4.301067643182939e-06, + "loss": 1.0784, + "step": 25021 + }, + { + "epoch": 2.11, + "learning_rate": 4.300320746531481e-06, + "loss": 1.189, + "step": 25022 + }, + { + "epoch": 2.11, + "learning_rate": 4.2995738969725984e-06, + "loss": 1.0498, + "step": 25023 + }, + { + "epoch": 2.11, + "learning_rate": 4.298827094512461e-06, + "loss": 1.0847, + "step": 25024 + }, + { + "epoch": 2.11, + "learning_rate": 4.298080339157239e-06, + "loss": 1.0887, + "step": 25025 + }, + { + "epoch": 2.11, + "learning_rate": 4.297333630913104e-06, + "loss": 1.1476, + "step": 25026 + }, + { + "epoch": 2.11, + "learning_rate": 4.296586969786219e-06, + "loss": 1.1809, + "step": 25027 + }, + { + "epoch": 2.11, + "learning_rate": 4.295840355782762e-06, + "loss": 1.1232, + "step": 25028 + }, + { + "epoch": 2.11, + "learning_rate": 4.2950937889088965e-06, + "loss": 1.0501, + "step": 25029 + }, + { + "epoch": 2.11, + "learning_rate": 4.29434726917079e-06, + "loss": 1.1943, + "step": 25030 + }, + { + "epoch": 2.11, + "learning_rate": 4.2936007965746125e-06, + "loss": 1.0274, + "step": 25031 + }, + { + "epoch": 2.11, + "learning_rate": 4.292854371126532e-06, + "loss": 1.1387, + "step": 25032 + }, + { + "epoch": 2.11, + "learning_rate": 4.292107992832716e-06, + "loss": 1.1386, + "step": 25033 + }, + { + "epoch": 2.11, + "learning_rate": 4.291361661699334e-06, + "loss": 1.1636, + "step": 25034 + }, + { + "epoch": 2.11, + "learning_rate": 4.290615377732541e-06, + "loss": 1.1036, + "step": 25035 + }, + { + "epoch": 2.11, + "learning_rate": 4.2898691409385155e-06, + "loss": 1.0398, + "step": 25036 + }, + { + "epoch": 2.11, + "learning_rate": 4.289122951323421e-06, + "loss": 1.0891, + "step": 25037 + }, + { + "epoch": 2.11, + "learning_rate": 4.288376808893414e-06, + "loss": 1.1554, + "step": 25038 + }, + { + "epoch": 2.11, + "learning_rate": 4.287630713654673e-06, + "loss": 1.0679, + "step": 25039 + }, + { + "epoch": 2.11, + "learning_rate": 4.286884665613351e-06, + "loss": 1.0645, + "step": 25040 + }, + { + "epoch": 2.11, + "learning_rate": 4.286138664775616e-06, + "loss": 1.0386, + "step": 25041 + }, + { + "epoch": 2.11, + "learning_rate": 4.285392711147634e-06, + "loss": 1.1445, + "step": 25042 + }, + { + "epoch": 2.11, + "learning_rate": 4.284646804735565e-06, + "loss": 1.1231, + "step": 25043 + }, + { + "epoch": 2.11, + "learning_rate": 4.283900945545572e-06, + "loss": 1.0425, + "step": 25044 + }, + { + "epoch": 2.11, + "learning_rate": 4.283155133583825e-06, + "loss": 1.0185, + "step": 25045 + }, + { + "epoch": 2.11, + "learning_rate": 4.2824093688564705e-06, + "loss": 1.0628, + "step": 25046 + }, + { + "epoch": 2.11, + "learning_rate": 4.281663651369689e-06, + "loss": 1.1346, + "step": 25047 + }, + { + "epoch": 2.11, + "learning_rate": 4.280917981129626e-06, + "loss": 1.1237, + "step": 25048 + }, + { + "epoch": 2.11, + "learning_rate": 4.280172358142448e-06, + "loss": 1.137, + "step": 25049 + }, + { + "epoch": 2.11, + "learning_rate": 4.279426782414323e-06, + "loss": 1.0793, + "step": 25050 + }, + { + "epoch": 2.11, + "learning_rate": 4.278681253951401e-06, + "loss": 1.1226, + "step": 25051 + }, + { + "epoch": 2.11, + "learning_rate": 4.277935772759844e-06, + "loss": 1.1097, + "step": 25052 + }, + { + "epoch": 2.11, + "learning_rate": 4.2771903388458155e-06, + "loss": 1.0835, + "step": 25053 + }, + { + "epoch": 2.11, + "learning_rate": 4.276444952215471e-06, + "loss": 1.1655, + "step": 25054 + }, + { + "epoch": 2.11, + "learning_rate": 4.275699612874969e-06, + "loss": 1.0847, + "step": 25055 + }, + { + "epoch": 2.11, + "learning_rate": 4.274954320830474e-06, + "loss": 1.1821, + "step": 25056 + }, + { + "epoch": 2.11, + "learning_rate": 4.27420907608813e-06, + "loss": 1.061, + "step": 25057 + }, + { + "epoch": 2.11, + "learning_rate": 4.273463878654111e-06, + "loss": 1.1164, + "step": 25058 + }, + { + "epoch": 2.11, + "learning_rate": 4.272718728534564e-06, + "loss": 1.1288, + "step": 25059 + }, + { + "epoch": 2.11, + "learning_rate": 4.2719736257356425e-06, + "loss": 1.1841, + "step": 25060 + }, + { + "epoch": 2.11, + "learning_rate": 4.271228570263516e-06, + "loss": 1.0892, + "step": 25061 + }, + { + "epoch": 2.11, + "learning_rate": 4.270483562124329e-06, + "loss": 0.9946, + "step": 25062 + }, + { + "epoch": 2.11, + "learning_rate": 4.269738601324239e-06, + "loss": 1.1575, + "step": 25063 + }, + { + "epoch": 2.11, + "learning_rate": 4.268993687869404e-06, + "loss": 1.1067, + "step": 25064 + }, + { + "epoch": 2.11, + "learning_rate": 4.268248821765977e-06, + "loss": 1.1124, + "step": 25065 + }, + { + "epoch": 2.11, + "learning_rate": 4.267504003020113e-06, + "loss": 1.0847, + "step": 25066 + }, + { + "epoch": 2.11, + "learning_rate": 4.2667592316379695e-06, + "loss": 1.1744, + "step": 25067 + }, + { + "epoch": 2.11, + "learning_rate": 4.266014507625688e-06, + "loss": 1.0692, + "step": 25068 + }, + { + "epoch": 2.11, + "learning_rate": 4.265269830989437e-06, + "loss": 1.0881, + "step": 25069 + }, + { + "epoch": 2.11, + "learning_rate": 4.264525201735357e-06, + "loss": 1.0733, + "step": 25070 + }, + { + "epoch": 2.11, + "learning_rate": 4.263780619869607e-06, + "loss": 1.083, + "step": 25071 + }, + { + "epoch": 2.11, + "learning_rate": 4.263036085398338e-06, + "loss": 1.1469, + "step": 25072 + }, + { + "epoch": 2.11, + "learning_rate": 4.2622915983277e-06, + "loss": 1.0533, + "step": 25073 + }, + { + "epoch": 2.11, + "learning_rate": 4.261547158663846e-06, + "loss": 1.1377, + "step": 25074 + }, + { + "epoch": 2.11, + "learning_rate": 4.260802766412925e-06, + "loss": 1.0577, + "step": 25075 + }, + { + "epoch": 2.11, + "learning_rate": 4.2600584215810925e-06, + "loss": 1.2154, + "step": 25076 + }, + { + "epoch": 2.11, + "learning_rate": 4.259314124174486e-06, + "loss": 1.0276, + "step": 25077 + }, + { + "epoch": 2.11, + "learning_rate": 4.258569874199272e-06, + "loss": 1.1202, + "step": 25078 + }, + { + "epoch": 2.11, + "learning_rate": 4.257825671661586e-06, + "loss": 1.1947, + "step": 25079 + }, + { + "epoch": 2.11, + "learning_rate": 4.257081516567584e-06, + "loss": 1.0317, + "step": 25080 + }, + { + "epoch": 2.11, + "learning_rate": 4.256337408923411e-06, + "loss": 1.0272, + "step": 25081 + }, + { + "epoch": 2.11, + "learning_rate": 4.255593348735217e-06, + "loss": 1.0747, + "step": 25082 + }, + { + "epoch": 2.11, + "learning_rate": 4.254849336009149e-06, + "loss": 1.0763, + "step": 25083 + }, + { + "epoch": 2.11, + "learning_rate": 4.254105370751359e-06, + "loss": 1.1159, + "step": 25084 + }, + { + "epoch": 2.11, + "learning_rate": 4.253361452967981e-06, + "loss": 1.0657, + "step": 25085 + }, + { + "epoch": 2.11, + "learning_rate": 4.2526175826651735e-06, + "loss": 1.1423, + "step": 25086 + }, + { + "epoch": 2.11, + "learning_rate": 4.251873759849084e-06, + "loss": 1.112, + "step": 25087 + }, + { + "epoch": 2.11, + "learning_rate": 4.251129984525845e-06, + "loss": 1.1241, + "step": 25088 + }, + { + "epoch": 2.11, + "learning_rate": 4.250386256701616e-06, + "loss": 1.0595, + "step": 25089 + }, + { + "epoch": 2.11, + "learning_rate": 4.249642576382535e-06, + "loss": 1.0504, + "step": 25090 + }, + { + "epoch": 2.11, + "learning_rate": 4.248898943574745e-06, + "loss": 1.0426, + "step": 25091 + }, + { + "epoch": 2.11, + "learning_rate": 4.2481553582843925e-06, + "loss": 1.0786, + "step": 25092 + }, + { + "epoch": 2.11, + "learning_rate": 4.247411820517622e-06, + "loss": 1.1294, + "step": 25093 + }, + { + "epoch": 2.11, + "learning_rate": 4.246668330280577e-06, + "loss": 1.078, + "step": 25094 + }, + { + "epoch": 2.11, + "learning_rate": 4.2459248875794015e-06, + "loss": 1.1473, + "step": 25095 + }, + { + "epoch": 2.11, + "learning_rate": 4.24518149242023e-06, + "loss": 1.0445, + "step": 25096 + }, + { + "epoch": 2.11, + "learning_rate": 4.244438144809218e-06, + "loss": 1.1379, + "step": 25097 + }, + { + "epoch": 2.11, + "learning_rate": 4.243694844752495e-06, + "loss": 1.2197, + "step": 25098 + }, + { + "epoch": 2.11, + "learning_rate": 4.242951592256205e-06, + "loss": 1.111, + "step": 25099 + }, + { + "epoch": 2.11, + "learning_rate": 4.242208387326499e-06, + "loss": 1.1805, + "step": 25100 + }, + { + "epoch": 2.11, + "learning_rate": 4.241465229969506e-06, + "loss": 1.1003, + "step": 25101 + }, + { + "epoch": 2.11, + "learning_rate": 4.24072212019137e-06, + "loss": 1.0512, + "step": 25102 + }, + { + "epoch": 2.11, + "learning_rate": 4.2399790579982315e-06, + "loss": 1.1504, + "step": 25103 + }, + { + "epoch": 2.11, + "learning_rate": 4.239236043396229e-06, + "loss": 1.1221, + "step": 25104 + }, + { + "epoch": 2.11, + "learning_rate": 4.238493076391502e-06, + "loss": 1.1404, + "step": 25105 + }, + { + "epoch": 2.11, + "learning_rate": 4.237750156990194e-06, + "loss": 1.085, + "step": 25106 + }, + { + "epoch": 2.11, + "learning_rate": 4.237007285198429e-06, + "loss": 1.11, + "step": 25107 + }, + { + "epoch": 2.11, + "learning_rate": 4.236264461022363e-06, + "loss": 1.0157, + "step": 25108 + }, + { + "epoch": 2.11, + "learning_rate": 4.23552168446812e-06, + "loss": 1.1668, + "step": 25109 + }, + { + "epoch": 2.11, + "learning_rate": 4.23477895554184e-06, + "loss": 1.0401, + "step": 25110 + }, + { + "epoch": 2.11, + "learning_rate": 4.234036274249667e-06, + "loss": 1.198, + "step": 25111 + }, + { + "epoch": 2.11, + "learning_rate": 4.233293640597728e-06, + "loss": 1.1706, + "step": 25112 + }, + { + "epoch": 2.11, + "learning_rate": 4.232551054592162e-06, + "loss": 1.1216, + "step": 25113 + }, + { + "epoch": 2.11, + "learning_rate": 4.231808516239106e-06, + "loss": 1.1377, + "step": 25114 + }, + { + "epoch": 2.11, + "learning_rate": 4.231066025544695e-06, + "loss": 1.1128, + "step": 25115 + }, + { + "epoch": 2.11, + "learning_rate": 4.2303235825150614e-06, + "loss": 1.157, + "step": 25116 + }, + { + "epoch": 2.11, + "learning_rate": 4.229581187156344e-06, + "loss": 1.1401, + "step": 25117 + }, + { + "epoch": 2.11, + "learning_rate": 4.228838839474667e-06, + "loss": 1.1554, + "step": 25118 + }, + { + "epoch": 2.11, + "learning_rate": 4.228096539476179e-06, + "loss": 1.0645, + "step": 25119 + }, + { + "epoch": 2.11, + "learning_rate": 4.227354287167e-06, + "loss": 1.1134, + "step": 25120 + }, + { + "epoch": 2.11, + "learning_rate": 4.2266120825532666e-06, + "loss": 1.0669, + "step": 25121 + }, + { + "epoch": 2.12, + "learning_rate": 4.225869925641113e-06, + "loss": 1.1759, + "step": 25122 + }, + { + "epoch": 2.12, + "learning_rate": 4.22512781643667e-06, + "loss": 1.1539, + "step": 25123 + }, + { + "epoch": 2.12, + "learning_rate": 4.224385754946068e-06, + "loss": 1.0835, + "step": 25124 + }, + { + "epoch": 2.12, + "learning_rate": 4.223643741175441e-06, + "loss": 1.1708, + "step": 25125 + }, + { + "epoch": 2.12, + "learning_rate": 4.222901775130917e-06, + "loss": 1.1293, + "step": 25126 + }, + { + "epoch": 2.12, + "learning_rate": 4.222159856818627e-06, + "loss": 1.142, + "step": 25127 + }, + { + "epoch": 2.12, + "learning_rate": 4.221417986244706e-06, + "loss": 1.1463, + "step": 25128 + }, + { + "epoch": 2.12, + "learning_rate": 4.220676163415272e-06, + "loss": 1.1269, + "step": 25129 + }, + { + "epoch": 2.12, + "learning_rate": 4.219934388336468e-06, + "loss": 1.2618, + "step": 25130 + }, + { + "epoch": 2.12, + "learning_rate": 4.219192661014412e-06, + "loss": 1.123, + "step": 25131 + }, + { + "epoch": 2.12, + "learning_rate": 4.2184509814552365e-06, + "loss": 1.0995, + "step": 25132 + }, + { + "epoch": 2.12, + "learning_rate": 4.217709349665069e-06, + "loss": 1.1223, + "step": 25133 + }, + { + "epoch": 2.12, + "learning_rate": 4.216967765650042e-06, + "loss": 1.1067, + "step": 25134 + }, + { + "epoch": 2.12, + "learning_rate": 4.216226229416269e-06, + "loss": 1.1675, + "step": 25135 + }, + { + "epoch": 2.12, + "learning_rate": 4.2154847409698915e-06, + "loss": 1.0481, + "step": 25136 + }, + { + "epoch": 2.12, + "learning_rate": 4.214743300317034e-06, + "loss": 1.1646, + "step": 25137 + }, + { + "epoch": 2.12, + "learning_rate": 4.214001907463811e-06, + "loss": 1.1496, + "step": 25138 + }, + { + "epoch": 2.12, + "learning_rate": 4.213260562416362e-06, + "loss": 1.0995, + "step": 25139 + }, + { + "epoch": 2.12, + "learning_rate": 4.212519265180804e-06, + "loss": 1.0987, + "step": 25140 + }, + { + "epoch": 2.12, + "learning_rate": 4.2117780157632635e-06, + "loss": 1.1245, + "step": 25141 + }, + { + "epoch": 2.12, + "learning_rate": 4.2110368141698665e-06, + "loss": 1.1727, + "step": 25142 + }, + { + "epoch": 2.12, + "learning_rate": 4.210295660406734e-06, + "loss": 1.1215, + "step": 25143 + }, + { + "epoch": 2.12, + "learning_rate": 4.2095545544799945e-06, + "loss": 1.1452, + "step": 25144 + }, + { + "epoch": 2.12, + "learning_rate": 4.20881349639577e-06, + "loss": 1.1519, + "step": 25145 + }, + { + "epoch": 2.12, + "learning_rate": 4.2080724861601756e-06, + "loss": 1.1422, + "step": 25146 + }, + { + "epoch": 2.12, + "learning_rate": 4.207331523779347e-06, + "loss": 1.0405, + "step": 25147 + }, + { + "epoch": 2.12, + "learning_rate": 4.206590609259395e-06, + "loss": 1.1382, + "step": 25148 + }, + { + "epoch": 2.12, + "learning_rate": 4.205849742606442e-06, + "loss": 1.1334, + "step": 25149 + }, + { + "epoch": 2.12, + "learning_rate": 4.20510892382662e-06, + "loss": 1.1044, + "step": 25150 + }, + { + "epoch": 2.12, + "learning_rate": 4.20436815292604e-06, + "loss": 1.0794, + "step": 25151 + }, + { + "epoch": 2.12, + "learning_rate": 4.203627429910825e-06, + "loss": 1.0566, + "step": 25152 + }, + { + "epoch": 2.12, + "learning_rate": 4.202886754787094e-06, + "loss": 1.1052, + "step": 25153 + }, + { + "epoch": 2.12, + "learning_rate": 4.20214612756097e-06, + "loss": 1.0787, + "step": 25154 + }, + { + "epoch": 2.12, + "learning_rate": 4.201405548238569e-06, + "loss": 1.1365, + "step": 25155 + }, + { + "epoch": 2.12, + "learning_rate": 4.200665016826014e-06, + "loss": 1.1779, + "step": 25156 + }, + { + "epoch": 2.12, + "learning_rate": 4.199924533329415e-06, + "loss": 1.0699, + "step": 25157 + }, + { + "epoch": 2.12, + "learning_rate": 4.199184097754902e-06, + "loss": 1.1846, + "step": 25158 + }, + { + "epoch": 2.12, + "learning_rate": 4.198443710108583e-06, + "loss": 1.1125, + "step": 25159 + }, + { + "epoch": 2.12, + "learning_rate": 4.1977033703965755e-06, + "loss": 1.1277, + "step": 25160 + }, + { + "epoch": 2.12, + "learning_rate": 4.196963078625007e-06, + "loss": 1.0282, + "step": 25161 + }, + { + "epoch": 2.12, + "learning_rate": 4.196222834799982e-06, + "loss": 1.1647, + "step": 25162 + }, + { + "epoch": 2.12, + "learning_rate": 4.195482638927623e-06, + "loss": 1.0986, + "step": 25163 + }, + { + "epoch": 2.12, + "learning_rate": 4.194742491014043e-06, + "loss": 1.1149, + "step": 25164 + }, + { + "epoch": 2.12, + "learning_rate": 4.1940023910653595e-06, + "loss": 1.1409, + "step": 25165 + }, + { + "epoch": 2.12, + "learning_rate": 4.193262339087685e-06, + "loss": 1.0501, + "step": 25166 + }, + { + "epoch": 2.12, + "learning_rate": 4.1925223350871405e-06, + "loss": 1.1356, + "step": 25167 + }, + { + "epoch": 2.12, + "learning_rate": 4.191782379069827e-06, + "loss": 1.1527, + "step": 25168 + }, + { + "epoch": 2.12, + "learning_rate": 4.191042471041873e-06, + "loss": 1.2166, + "step": 25169 + }, + { + "epoch": 2.12, + "learning_rate": 4.190302611009382e-06, + "loss": 1.1011, + "step": 25170 + }, + { + "epoch": 2.12, + "learning_rate": 4.189562798978472e-06, + "loss": 1.0941, + "step": 25171 + }, + { + "epoch": 2.12, + "learning_rate": 4.1888230349552515e-06, + "loss": 1.0855, + "step": 25172 + }, + { + "epoch": 2.12, + "learning_rate": 4.1880833189458356e-06, + "loss": 1.1062, + "step": 25173 + }, + { + "epoch": 2.12, + "learning_rate": 4.187343650956336e-06, + "loss": 1.0834, + "step": 25174 + }, + { + "epoch": 2.12, + "learning_rate": 4.186604030992863e-06, + "loss": 1.1829, + "step": 25175 + }, + { + "epoch": 2.12, + "learning_rate": 4.185864459061528e-06, + "loss": 1.1459, + "step": 25176 + }, + { + "epoch": 2.12, + "learning_rate": 4.185124935168441e-06, + "loss": 1.1147, + "step": 25177 + }, + { + "epoch": 2.12, + "learning_rate": 4.184385459319717e-06, + "loss": 1.0679, + "step": 25178 + }, + { + "epoch": 2.12, + "learning_rate": 4.183646031521454e-06, + "loss": 1.0768, + "step": 25179 + }, + { + "epoch": 2.12, + "learning_rate": 4.182906651779777e-06, + "loss": 1.1041, + "step": 25180 + }, + { + "epoch": 2.12, + "learning_rate": 4.182167320100782e-06, + "loss": 1.0739, + "step": 25181 + }, + { + "epoch": 2.12, + "learning_rate": 4.181428036490584e-06, + "loss": 1.0938, + "step": 25182 + }, + { + "epoch": 2.12, + "learning_rate": 4.1806888009552885e-06, + "loss": 1.0362, + "step": 25183 + }, + { + "epoch": 2.12, + "learning_rate": 4.179949613501009e-06, + "loss": 1.0876, + "step": 25184 + }, + { + "epoch": 2.12, + "learning_rate": 4.17921047413384e-06, + "loss": 1.159, + "step": 25185 + }, + { + "epoch": 2.12, + "learning_rate": 4.178471382859902e-06, + "loss": 1.0571, + "step": 25186 + }, + { + "epoch": 2.12, + "learning_rate": 4.1777323396853e-06, + "loss": 1.1136, + "step": 25187 + }, + { + "epoch": 2.12, + "learning_rate": 4.17699334461613e-06, + "loss": 1.1362, + "step": 25188 + }, + { + "epoch": 2.12, + "learning_rate": 4.17625439765851e-06, + "loss": 1.1986, + "step": 25189 + }, + { + "epoch": 2.12, + "learning_rate": 4.175515498818537e-06, + "loss": 1.0648, + "step": 25190 + }, + { + "epoch": 2.12, + "learning_rate": 4.17477664810232e-06, + "loss": 1.1589, + "step": 25191 + }, + { + "epoch": 2.12, + "learning_rate": 4.174037845515961e-06, + "loss": 1.1087, + "step": 25192 + }, + { + "epoch": 2.12, + "learning_rate": 4.173299091065568e-06, + "loss": 1.1635, + "step": 25193 + }, + { + "epoch": 2.12, + "learning_rate": 4.1725603847572415e-06, + "loss": 1.1504, + "step": 25194 + }, + { + "epoch": 2.12, + "learning_rate": 4.1718217265970895e-06, + "loss": 1.0333, + "step": 25195 + }, + { + "epoch": 2.12, + "learning_rate": 4.171083116591204e-06, + "loss": 1.1255, + "step": 25196 + }, + { + "epoch": 2.12, + "learning_rate": 4.170344554745704e-06, + "loss": 1.1784, + "step": 25197 + }, + { + "epoch": 2.12, + "learning_rate": 4.169606041066679e-06, + "loss": 1.1588, + "step": 25198 + }, + { + "epoch": 2.12, + "learning_rate": 4.168867575560232e-06, + "loss": 1.063, + "step": 25199 + }, + { + "epoch": 2.12, + "learning_rate": 4.1681291582324725e-06, + "loss": 1.1226, + "step": 25200 + }, + { + "epoch": 2.12, + "learning_rate": 4.167390789089494e-06, + "loss": 1.1245, + "step": 25201 + }, + { + "epoch": 2.12, + "learning_rate": 4.166652468137399e-06, + "loss": 1.0761, + "step": 25202 + }, + { + "epoch": 2.12, + "learning_rate": 4.165914195382288e-06, + "loss": 1.186, + "step": 25203 + }, + { + "epoch": 2.12, + "learning_rate": 4.165175970830262e-06, + "loss": 1.1653, + "step": 25204 + }, + { + "epoch": 2.12, + "learning_rate": 4.164437794487419e-06, + "loss": 1.1816, + "step": 25205 + }, + { + "epoch": 2.12, + "learning_rate": 4.163699666359862e-06, + "loss": 1.1766, + "step": 25206 + }, + { + "epoch": 2.12, + "learning_rate": 4.1629615864536785e-06, + "loss": 1.0151, + "step": 25207 + }, + { + "epoch": 2.12, + "learning_rate": 4.162223554774982e-06, + "loss": 1.1335, + "step": 25208 + }, + { + "epoch": 2.12, + "learning_rate": 4.161485571329861e-06, + "loss": 1.0703, + "step": 25209 + }, + { + "epoch": 2.12, + "learning_rate": 4.160747636124409e-06, + "loss": 1.0427, + "step": 25210 + }, + { + "epoch": 2.12, + "learning_rate": 4.1600097491647366e-06, + "loss": 1.1393, + "step": 25211 + }, + { + "epoch": 2.12, + "learning_rate": 4.159271910456929e-06, + "loss": 1.1578, + "step": 25212 + }, + { + "epoch": 2.12, + "learning_rate": 4.158534120007087e-06, + "loss": 1.1523, + "step": 25213 + }, + { + "epoch": 2.12, + "learning_rate": 4.1577963778213044e-06, + "loss": 1.1119, + "step": 25214 + }, + { + "epoch": 2.12, + "learning_rate": 4.1570586839056795e-06, + "loss": 1.1758, + "step": 25215 + }, + { + "epoch": 2.12, + "learning_rate": 4.156321038266304e-06, + "loss": 1.2055, + "step": 25216 + }, + { + "epoch": 2.12, + "learning_rate": 4.1555834409092786e-06, + "loss": 1.129, + "step": 25217 + }, + { + "epoch": 2.12, + "learning_rate": 4.154845891840686e-06, + "loss": 0.9984, + "step": 25218 + }, + { + "epoch": 2.12, + "learning_rate": 4.154108391066635e-06, + "loss": 1.0636, + "step": 25219 + }, + { + "epoch": 2.12, + "learning_rate": 4.153370938593206e-06, + "loss": 1.0339, + "step": 25220 + }, + { + "epoch": 2.12, + "learning_rate": 4.1526335344265e-06, + "loss": 1.1195, + "step": 25221 + }, + { + "epoch": 2.12, + "learning_rate": 4.151896178572606e-06, + "loss": 1.1598, + "step": 25222 + }, + { + "epoch": 2.12, + "learning_rate": 4.151158871037617e-06, + "loss": 1.0835, + "step": 25223 + }, + { + "epoch": 2.12, + "learning_rate": 4.1504216118276256e-06, + "loss": 1.0525, + "step": 25224 + }, + { + "epoch": 2.12, + "learning_rate": 4.149684400948724e-06, + "loss": 1.1148, + "step": 25225 + }, + { + "epoch": 2.12, + "learning_rate": 4.148947238407001e-06, + "loss": 1.1126, + "step": 25226 + }, + { + "epoch": 2.12, + "learning_rate": 4.148210124208548e-06, + "loss": 1.1176, + "step": 25227 + }, + { + "epoch": 2.12, + "learning_rate": 4.1474730583594604e-06, + "loss": 1.1953, + "step": 25228 + }, + { + "epoch": 2.12, + "learning_rate": 4.146736040865815e-06, + "loss": 1.1206, + "step": 25229 + }, + { + "epoch": 2.12, + "learning_rate": 4.145999071733718e-06, + "loss": 1.0152, + "step": 25230 + }, + { + "epoch": 2.12, + "learning_rate": 4.145262150969247e-06, + "loss": 1.1382, + "step": 25231 + }, + { + "epoch": 2.12, + "learning_rate": 4.1445252785784925e-06, + "loss": 1.1987, + "step": 25232 + }, + { + "epoch": 2.12, + "learning_rate": 4.143788454567546e-06, + "loss": 1.12, + "step": 25233 + }, + { + "epoch": 2.12, + "learning_rate": 4.143051678942496e-06, + "loss": 1.0379, + "step": 25234 + }, + { + "epoch": 2.12, + "learning_rate": 4.14231495170942e-06, + "loss": 1.0989, + "step": 25235 + }, + { + "epoch": 2.12, + "learning_rate": 4.141578272874417e-06, + "loss": 1.1204, + "step": 25236 + }, + { + "epoch": 2.12, + "learning_rate": 4.140841642443572e-06, + "loss": 1.1316, + "step": 25237 + }, + { + "epoch": 2.12, + "learning_rate": 4.140105060422962e-06, + "loss": 1.1251, + "step": 25238 + }, + { + "epoch": 2.12, + "learning_rate": 4.139368526818685e-06, + "loss": 1.0701, + "step": 25239 + }, + { + "epoch": 2.12, + "learning_rate": 4.13863204163682e-06, + "loss": 1.0273, + "step": 25240 + }, + { + "epoch": 2.13, + "learning_rate": 4.137895604883451e-06, + "loss": 1.019, + "step": 25241 + }, + { + "epoch": 2.13, + "learning_rate": 4.137159216564665e-06, + "loss": 1.1372, + "step": 25242 + }, + { + "epoch": 2.13, + "learning_rate": 4.136422876686546e-06, + "loss": 1.1036, + "step": 25243 + }, + { + "epoch": 2.13, + "learning_rate": 4.135686585255177e-06, + "loss": 1.1774, + "step": 25244 + }, + { + "epoch": 2.13, + "learning_rate": 4.134950342276648e-06, + "loss": 1.0621, + "step": 25245 + }, + { + "epoch": 2.13, + "learning_rate": 4.134214147757027e-06, + "loss": 1.1088, + "step": 25246 + }, + { + "epoch": 2.13, + "learning_rate": 4.133478001702414e-06, + "loss": 1.1212, + "step": 25247 + }, + { + "epoch": 2.13, + "learning_rate": 4.13274190411888e-06, + "loss": 1.2043, + "step": 25248 + }, + { + "epoch": 2.13, + "learning_rate": 4.132005855012507e-06, + "loss": 1.178, + "step": 25249 + }, + { + "epoch": 2.13, + "learning_rate": 4.131269854389385e-06, + "loss": 0.9903, + "step": 25250 + }, + { + "epoch": 2.13, + "learning_rate": 4.130533902255587e-06, + "loss": 1.0015, + "step": 25251 + }, + { + "epoch": 2.13, + "learning_rate": 4.129797998617197e-06, + "loss": 1.0759, + "step": 25252 + }, + { + "epoch": 2.13, + "learning_rate": 4.1290621434802935e-06, + "loss": 1.0962, + "step": 25253 + }, + { + "epoch": 2.13, + "learning_rate": 4.128326336850959e-06, + "loss": 1.1639, + "step": 25254 + }, + { + "epoch": 2.13, + "learning_rate": 4.12759057873527e-06, + "loss": 1.1213, + "step": 25255 + }, + { + "epoch": 2.13, + "learning_rate": 4.126854869139311e-06, + "loss": 1.1498, + "step": 25256 + }, + { + "epoch": 2.13, + "learning_rate": 4.1261192080691505e-06, + "loss": 1.0789, + "step": 25257 + }, + { + "epoch": 2.13, + "learning_rate": 4.12538359553088e-06, + "loss": 1.1074, + "step": 25258 + }, + { + "epoch": 2.13, + "learning_rate": 4.124648031530566e-06, + "loss": 1.0361, + "step": 25259 + }, + { + "epoch": 2.13, + "learning_rate": 4.123912516074286e-06, + "loss": 1.1818, + "step": 25260 + }, + { + "epoch": 2.13, + "learning_rate": 4.12317704916813e-06, + "loss": 1.0719, + "step": 25261 + }, + { + "epoch": 2.13, + "learning_rate": 4.122441630818163e-06, + "loss": 0.9833, + "step": 25262 + }, + { + "epoch": 2.13, + "learning_rate": 4.121706261030462e-06, + "loss": 1.0948, + "step": 25263 + }, + { + "epoch": 2.13, + "learning_rate": 4.120970939811107e-06, + "loss": 1.1426, + "step": 25264 + }, + { + "epoch": 2.13, + "learning_rate": 4.120235667166172e-06, + "loss": 1.06, + "step": 25265 + }, + { + "epoch": 2.13, + "learning_rate": 4.11950044310173e-06, + "loss": 1.0505, + "step": 25266 + }, + { + "epoch": 2.13, + "learning_rate": 4.118765267623861e-06, + "loss": 1.1173, + "step": 25267 + }, + { + "epoch": 2.13, + "learning_rate": 4.1180301407386295e-06, + "loss": 1.0891, + "step": 25268 + }, + { + "epoch": 2.13, + "learning_rate": 4.117295062452122e-06, + "loss": 1.0761, + "step": 25269 + }, + { + "epoch": 2.13, + "learning_rate": 4.116560032770402e-06, + "loss": 1.1152, + "step": 25270 + }, + { + "epoch": 2.13, + "learning_rate": 4.115825051699547e-06, + "loss": 1.1073, + "step": 25271 + }, + { + "epoch": 2.13, + "learning_rate": 4.115090119245627e-06, + "loss": 1.1568, + "step": 25272 + }, + { + "epoch": 2.13, + "learning_rate": 4.114355235414716e-06, + "loss": 1.0455, + "step": 25273 + }, + { + "epoch": 2.13, + "learning_rate": 4.113620400212887e-06, + "loss": 1.1681, + "step": 25274 + }, + { + "epoch": 2.13, + "learning_rate": 4.112885613646209e-06, + "loss": 1.0733, + "step": 25275 + }, + { + "epoch": 2.13, + "learning_rate": 4.112150875720754e-06, + "loss": 1.1575, + "step": 25276 + }, + { + "epoch": 2.13, + "learning_rate": 4.1114161864425935e-06, + "loss": 1.0319, + "step": 25277 + }, + { + "epoch": 2.13, + "learning_rate": 4.110681545817801e-06, + "loss": 1.1056, + "step": 25278 + }, + { + "epoch": 2.13, + "learning_rate": 4.109946953852434e-06, + "loss": 1.1408, + "step": 25279 + }, + { + "epoch": 2.13, + "learning_rate": 4.1092124105525775e-06, + "loss": 1.1325, + "step": 25280 + }, + { + "epoch": 2.13, + "learning_rate": 4.108477915924292e-06, + "loss": 1.0677, + "step": 25281 + }, + { + "epoch": 2.13, + "learning_rate": 4.107743469973645e-06, + "loss": 1.0888, + "step": 25282 + }, + { + "epoch": 2.13, + "learning_rate": 4.107009072706708e-06, + "loss": 1.1435, + "step": 25283 + }, + { + "epoch": 2.13, + "learning_rate": 4.106274724129552e-06, + "loss": 1.107, + "step": 25284 + }, + { + "epoch": 2.13, + "learning_rate": 4.105540424248233e-06, + "loss": 1.1058, + "step": 25285 + }, + { + "epoch": 2.13, + "learning_rate": 4.104806173068829e-06, + "loss": 1.0482, + "step": 25286 + }, + { + "epoch": 2.13, + "learning_rate": 4.104071970597406e-06, + "loss": 1.1691, + "step": 25287 + }, + { + "epoch": 2.13, + "learning_rate": 4.103337816840021e-06, + "loss": 1.0566, + "step": 25288 + }, + { + "epoch": 2.13, + "learning_rate": 4.102603711802754e-06, + "loss": 1.1471, + "step": 25289 + }, + { + "epoch": 2.13, + "learning_rate": 4.101869655491658e-06, + "loss": 1.2195, + "step": 25290 + }, + { + "epoch": 2.13, + "learning_rate": 4.101135647912803e-06, + "loss": 1.2261, + "step": 25291 + }, + { + "epoch": 2.13, + "learning_rate": 4.100401689072253e-06, + "loss": 1.0985, + "step": 25292 + }, + { + "epoch": 2.13, + "learning_rate": 4.099667778976072e-06, + "loss": 1.0735, + "step": 25293 + }, + { + "epoch": 2.13, + "learning_rate": 4.098933917630325e-06, + "loss": 1.1248, + "step": 25294 + }, + { + "epoch": 2.13, + "learning_rate": 4.098200105041077e-06, + "loss": 1.1146, + "step": 25295 + }, + { + "epoch": 2.13, + "learning_rate": 4.097466341214382e-06, + "loss": 1.1359, + "step": 25296 + }, + { + "epoch": 2.13, + "learning_rate": 4.096732626156317e-06, + "loss": 1.0921, + "step": 25297 + }, + { + "epoch": 2.13, + "learning_rate": 4.095998959872931e-06, + "loss": 1.1469, + "step": 25298 + }, + { + "epoch": 2.13, + "learning_rate": 4.095265342370288e-06, + "loss": 1.0358, + "step": 25299 + }, + { + "epoch": 2.13, + "learning_rate": 4.094531773654461e-06, + "loss": 1.1307, + "step": 25300 + }, + { + "epoch": 2.13, + "learning_rate": 4.093798253731498e-06, + "loss": 1.1407, + "step": 25301 + }, + { + "epoch": 2.13, + "learning_rate": 4.093064782607464e-06, + "loss": 1.1151, + "step": 25302 + }, + { + "epoch": 2.13, + "learning_rate": 4.09233136028842e-06, + "loss": 1.0761, + "step": 25303 + }, + { + "epoch": 2.13, + "learning_rate": 4.091597986780423e-06, + "loss": 1.0611, + "step": 25304 + }, + { + "epoch": 2.13, + "learning_rate": 4.0908646620895365e-06, + "loss": 1.123, + "step": 25305 + }, + { + "epoch": 2.13, + "learning_rate": 4.09013138622182e-06, + "loss": 1.1498, + "step": 25306 + }, + { + "epoch": 2.13, + "learning_rate": 4.0893981591833225e-06, + "loss": 1.0758, + "step": 25307 + }, + { + "epoch": 2.13, + "learning_rate": 4.088664980980116e-06, + "loss": 1.0734, + "step": 25308 + }, + { + "epoch": 2.13, + "learning_rate": 4.0879318516182475e-06, + "loss": 1.1579, + "step": 25309 + }, + { + "epoch": 2.13, + "learning_rate": 4.087198771103775e-06, + "loss": 1.1238, + "step": 25310 + }, + { + "epoch": 2.13, + "learning_rate": 4.086465739442765e-06, + "loss": 1.1757, + "step": 25311 + }, + { + "epoch": 2.13, + "learning_rate": 4.085732756641265e-06, + "loss": 1.1034, + "step": 25312 + }, + { + "epoch": 2.13, + "learning_rate": 4.084999822705335e-06, + "loss": 1.0884, + "step": 25313 + }, + { + "epoch": 2.13, + "learning_rate": 4.084266937641027e-06, + "loss": 1.1538, + "step": 25314 + }, + { + "epoch": 2.13, + "learning_rate": 4.083534101454399e-06, + "loss": 1.1532, + "step": 25315 + }, + { + "epoch": 2.13, + "learning_rate": 4.0828013141515075e-06, + "loss": 1.1968, + "step": 25316 + }, + { + "epoch": 2.13, + "learning_rate": 4.082068575738407e-06, + "loss": 1.1088, + "step": 25317 + }, + { + "epoch": 2.13, + "learning_rate": 4.0813358862211435e-06, + "loss": 1.0414, + "step": 25318 + }, + { + "epoch": 2.13, + "learning_rate": 4.0806032456057854e-06, + "loss": 1.1537, + "step": 25319 + }, + { + "epoch": 2.13, + "learning_rate": 4.079870653898373e-06, + "loss": 0.999, + "step": 25320 + }, + { + "epoch": 2.13, + "learning_rate": 4.079138111104965e-06, + "loss": 1.0541, + "step": 25321 + }, + { + "epoch": 2.13, + "learning_rate": 4.078405617231611e-06, + "loss": 1.1352, + "step": 25322 + }, + { + "epoch": 2.13, + "learning_rate": 4.077673172284366e-06, + "loss": 1.0678, + "step": 25323 + }, + { + "epoch": 2.13, + "learning_rate": 4.076940776269281e-06, + "loss": 1.05, + "step": 25324 + }, + { + "epoch": 2.13, + "learning_rate": 4.076208429192406e-06, + "loss": 1.1247, + "step": 25325 + }, + { + "epoch": 2.13, + "learning_rate": 4.075476131059794e-06, + "loss": 1.1365, + "step": 25326 + }, + { + "epoch": 2.13, + "learning_rate": 4.0747438818774935e-06, + "loss": 1.1082, + "step": 25327 + }, + { + "epoch": 2.13, + "learning_rate": 4.074011681651559e-06, + "loss": 1.0985, + "step": 25328 + }, + { + "epoch": 2.13, + "learning_rate": 4.073279530388029e-06, + "loss": 1.0936, + "step": 25329 + }, + { + "epoch": 2.13, + "learning_rate": 4.072547428092968e-06, + "loss": 1.1646, + "step": 25330 + }, + { + "epoch": 2.13, + "learning_rate": 4.071815374772413e-06, + "loss": 1.118, + "step": 25331 + }, + { + "epoch": 2.13, + "learning_rate": 4.071083370432417e-06, + "loss": 1.1543, + "step": 25332 + }, + { + "epoch": 2.13, + "learning_rate": 4.070351415079028e-06, + "loss": 1.1116, + "step": 25333 + }, + { + "epoch": 2.13, + "learning_rate": 4.069619508718297e-06, + "loss": 1.2022, + "step": 25334 + }, + { + "epoch": 2.13, + "learning_rate": 4.068887651356261e-06, + "loss": 1.2275, + "step": 25335 + }, + { + "epoch": 2.13, + "learning_rate": 4.068155842998975e-06, + "loss": 1.1374, + "step": 25336 + }, + { + "epoch": 2.13, + "learning_rate": 4.067424083652489e-06, + "loss": 1.1435, + "step": 25337 + }, + { + "epoch": 2.13, + "learning_rate": 4.066692373322837e-06, + "loss": 1.1169, + "step": 25338 + }, + { + "epoch": 2.13, + "learning_rate": 4.065960712016077e-06, + "loss": 1.1875, + "step": 25339 + }, + { + "epoch": 2.13, + "learning_rate": 4.065229099738247e-06, + "loss": 1.172, + "step": 25340 + }, + { + "epoch": 2.13, + "learning_rate": 4.064497536495392e-06, + "loss": 1.0981, + "step": 25341 + }, + { + "epoch": 2.13, + "learning_rate": 4.063766022293559e-06, + "loss": 1.1601, + "step": 25342 + }, + { + "epoch": 2.13, + "learning_rate": 4.063034557138791e-06, + "loss": 1.0272, + "step": 25343 + }, + { + "epoch": 2.13, + "learning_rate": 4.062303141037131e-06, + "loss": 1.1032, + "step": 25344 + }, + { + "epoch": 2.13, + "learning_rate": 4.061571773994626e-06, + "loss": 1.1502, + "step": 25345 + }, + { + "epoch": 2.13, + "learning_rate": 4.0608404560173095e-06, + "loss": 1.0933, + "step": 25346 + }, + { + "epoch": 2.13, + "learning_rate": 4.0601091871112365e-06, + "loss": 1.1389, + "step": 25347 + }, + { + "epoch": 2.13, + "learning_rate": 4.059377967282439e-06, + "loss": 1.1151, + "step": 25348 + }, + { + "epoch": 2.13, + "learning_rate": 4.058646796536958e-06, + "loss": 1.1586, + "step": 25349 + }, + { + "epoch": 2.13, + "learning_rate": 4.057915674880846e-06, + "loss": 1.059, + "step": 25350 + }, + { + "epoch": 2.13, + "learning_rate": 4.0571846023201324e-06, + "loss": 1.0919, + "step": 25351 + }, + { + "epoch": 2.13, + "learning_rate": 4.0564535788608624e-06, + "loss": 1.1778, + "step": 25352 + }, + { + "epoch": 2.13, + "learning_rate": 4.055722604509075e-06, + "loss": 1.1184, + "step": 25353 + }, + { + "epoch": 2.13, + "learning_rate": 4.054991679270808e-06, + "loss": 1.0978, + "step": 25354 + }, + { + "epoch": 2.13, + "learning_rate": 4.054260803152106e-06, + "loss": 1.0473, + "step": 25355 + }, + { + "epoch": 2.13, + "learning_rate": 4.053529976159006e-06, + "loss": 1.1271, + "step": 25356 + }, + { + "epoch": 2.13, + "learning_rate": 4.052799198297536e-06, + "loss": 1.1752, + "step": 25357 + }, + { + "epoch": 2.13, + "learning_rate": 4.05206846957375e-06, + "loss": 1.1241, + "step": 25358 + }, + { + "epoch": 2.14, + "learning_rate": 4.051337789993676e-06, + "loss": 0.9861, + "step": 25359 + }, + { + "epoch": 2.14, + "learning_rate": 4.050607159563348e-06, + "loss": 1.0927, + "step": 25360 + }, + { + "epoch": 2.14, + "learning_rate": 4.049876578288815e-06, + "loss": 1.1169, + "step": 25361 + }, + { + "epoch": 2.14, + "learning_rate": 4.049146046176103e-06, + "loss": 1.1054, + "step": 25362 + }, + { + "epoch": 2.14, + "learning_rate": 4.048415563231249e-06, + "loss": 1.1585, + "step": 25363 + }, + { + "epoch": 2.14, + "learning_rate": 4.047685129460293e-06, + "loss": 1.1446, + "step": 25364 + }, + { + "epoch": 2.14, + "learning_rate": 4.046954744869265e-06, + "loss": 1.2174, + "step": 25365 + }, + { + "epoch": 2.14, + "learning_rate": 4.046224409464203e-06, + "loss": 1.1532, + "step": 25366 + }, + { + "epoch": 2.14, + "learning_rate": 4.045494123251145e-06, + "loss": 1.1425, + "step": 25367 + }, + { + "epoch": 2.14, + "learning_rate": 4.044763886236112e-06, + "loss": 1.1084, + "step": 25368 + }, + { + "epoch": 2.14, + "learning_rate": 4.044033698425154e-06, + "loss": 1.112, + "step": 25369 + }, + { + "epoch": 2.14, + "learning_rate": 4.043303559824291e-06, + "loss": 1.1249, + "step": 25370 + }, + { + "epoch": 2.14, + "learning_rate": 4.04257347043956e-06, + "loss": 1.1225, + "step": 25371 + }, + { + "epoch": 2.14, + "learning_rate": 4.041843430276995e-06, + "loss": 1.1449, + "step": 25372 + }, + { + "epoch": 2.14, + "learning_rate": 4.041113439342625e-06, + "loss": 1.075, + "step": 25373 + }, + { + "epoch": 2.14, + "learning_rate": 4.040383497642484e-06, + "loss": 1.1022, + "step": 25374 + }, + { + "epoch": 2.14, + "learning_rate": 4.0396536051826015e-06, + "loss": 1.1501, + "step": 25375 + }, + { + "epoch": 2.14, + "learning_rate": 4.038923761969009e-06, + "loss": 1.0928, + "step": 25376 + }, + { + "epoch": 2.14, + "learning_rate": 4.038193968007735e-06, + "loss": 1.1219, + "step": 25377 + }, + { + "epoch": 2.14, + "learning_rate": 4.037464223304814e-06, + "loss": 1.1172, + "step": 25378 + }, + { + "epoch": 2.14, + "learning_rate": 4.036734527866266e-06, + "loss": 1.1195, + "step": 25379 + }, + { + "epoch": 2.14, + "learning_rate": 4.036004881698132e-06, + "loss": 1.1678, + "step": 25380 + }, + { + "epoch": 2.14, + "learning_rate": 4.035275284806431e-06, + "loss": 1.0407, + "step": 25381 + }, + { + "epoch": 2.14, + "learning_rate": 4.0345457371971955e-06, + "loss": 1.0867, + "step": 25382 + }, + { + "epoch": 2.14, + "learning_rate": 4.033816238876451e-06, + "loss": 1.0978, + "step": 25383 + }, + { + "epoch": 2.14, + "learning_rate": 4.033086789850231e-06, + "loss": 1.1751, + "step": 25384 + }, + { + "epoch": 2.14, + "learning_rate": 4.03235739012455e-06, + "loss": 1.1444, + "step": 25385 + }, + { + "epoch": 2.14, + "learning_rate": 4.031628039705445e-06, + "loss": 1.0906, + "step": 25386 + }, + { + "epoch": 2.14, + "learning_rate": 4.030898738598943e-06, + "loss": 1.1981, + "step": 25387 + }, + { + "epoch": 2.14, + "learning_rate": 4.030169486811058e-06, + "loss": 1.1556, + "step": 25388 + }, + { + "epoch": 2.14, + "learning_rate": 4.029440284347833e-06, + "loss": 1.0864, + "step": 25389 + }, + { + "epoch": 2.14, + "learning_rate": 4.028711131215278e-06, + "loss": 1.0441, + "step": 25390 + }, + { + "epoch": 2.14, + "learning_rate": 4.027982027419422e-06, + "loss": 1.1692, + "step": 25391 + }, + { + "epoch": 2.14, + "learning_rate": 4.027252972966291e-06, + "loss": 1.1031, + "step": 25392 + }, + { + "epoch": 2.14, + "learning_rate": 4.026523967861907e-06, + "loss": 0.9906, + "step": 25393 + }, + { + "epoch": 2.14, + "learning_rate": 4.0257950121122935e-06, + "loss": 1.1073, + "step": 25394 + }, + { + "epoch": 2.14, + "learning_rate": 4.025066105723477e-06, + "loss": 1.1399, + "step": 25395 + }, + { + "epoch": 2.14, + "learning_rate": 4.0243372487014705e-06, + "loss": 1.2237, + "step": 25396 + }, + { + "epoch": 2.14, + "learning_rate": 4.023608441052309e-06, + "loss": 1.1254, + "step": 25397 + }, + { + "epoch": 2.14, + "learning_rate": 4.022879682782003e-06, + "loss": 1.1777, + "step": 25398 + }, + { + "epoch": 2.14, + "learning_rate": 4.022150973896574e-06, + "loss": 1.1573, + "step": 25399 + }, + { + "epoch": 2.14, + "learning_rate": 4.021422314402056e-06, + "loss": 1.1121, + "step": 25400 + }, + { + "epoch": 2.14, + "learning_rate": 4.0206937043044545e-06, + "loss": 1.1324, + "step": 25401 + }, + { + "epoch": 2.14, + "learning_rate": 4.019965143609797e-06, + "loss": 1.0868, + "step": 25402 + }, + { + "epoch": 2.14, + "learning_rate": 4.0192366323241e-06, + "loss": 1.1192, + "step": 25403 + }, + { + "epoch": 2.14, + "learning_rate": 4.018508170453385e-06, + "loss": 1.2166, + "step": 25404 + }, + { + "epoch": 2.14, + "learning_rate": 4.017779758003668e-06, + "loss": 1.0725, + "step": 25405 + }, + { + "epoch": 2.14, + "learning_rate": 4.017051394980974e-06, + "loss": 1.1068, + "step": 25406 + }, + { + "epoch": 2.14, + "learning_rate": 4.01632308139131e-06, + "loss": 1.0537, + "step": 25407 + }, + { + "epoch": 2.14, + "learning_rate": 4.015594817240706e-06, + "loss": 1.0727, + "step": 25408 + }, + { + "epoch": 2.14, + "learning_rate": 4.014866602535169e-06, + "loss": 1.1144, + "step": 25409 + }, + { + "epoch": 2.14, + "learning_rate": 4.014138437280716e-06, + "loss": 1.1131, + "step": 25410 + }, + { + "epoch": 2.14, + "learning_rate": 4.013410321483374e-06, + "loss": 1.1097, + "step": 25411 + }, + { + "epoch": 2.14, + "learning_rate": 4.012682255149148e-06, + "loss": 1.1421, + "step": 25412 + }, + { + "epoch": 2.14, + "learning_rate": 4.011954238284059e-06, + "loss": 1.1845, + "step": 25413 + }, + { + "epoch": 2.14, + "learning_rate": 4.01122627089412e-06, + "loss": 1.0433, + "step": 25414 + }, + { + "epoch": 2.14, + "learning_rate": 4.010498352985346e-06, + "loss": 1.0114, + "step": 25415 + }, + { + "epoch": 2.14, + "learning_rate": 4.009770484563751e-06, + "loss": 1.0826, + "step": 25416 + }, + { + "epoch": 2.14, + "learning_rate": 4.009042665635353e-06, + "loss": 1.1343, + "step": 25417 + }, + { + "epoch": 2.14, + "learning_rate": 4.008314896206154e-06, + "loss": 1.1747, + "step": 25418 + }, + { + "epoch": 2.14, + "learning_rate": 4.0075871762821836e-06, + "loss": 1.1483, + "step": 25419 + }, + { + "epoch": 2.14, + "learning_rate": 4.006859505869442e-06, + "loss": 1.082, + "step": 25420 + }, + { + "epoch": 2.14, + "learning_rate": 4.0061318849739436e-06, + "loss": 1.0965, + "step": 25421 + }, + { + "epoch": 2.14, + "learning_rate": 4.005404313601704e-06, + "loss": 1.2406, + "step": 25422 + }, + { + "epoch": 2.14, + "learning_rate": 4.004676791758732e-06, + "loss": 1.0858, + "step": 25423 + }, + { + "epoch": 2.14, + "learning_rate": 4.003949319451038e-06, + "loss": 1.1825, + "step": 25424 + }, + { + "epoch": 2.14, + "learning_rate": 4.003221896684634e-06, + "loss": 1.209, + "step": 25425 + }, + { + "epoch": 2.14, + "learning_rate": 4.00249452346553e-06, + "loss": 1.1141, + "step": 25426 + }, + { + "epoch": 2.14, + "learning_rate": 4.001767199799737e-06, + "loss": 1.1452, + "step": 25427 + }, + { + "epoch": 2.14, + "learning_rate": 4.001039925693265e-06, + "loss": 1.0814, + "step": 25428 + }, + { + "epoch": 2.14, + "learning_rate": 4.000312701152115e-06, + "loss": 1.1418, + "step": 25429 + }, + { + "epoch": 2.14, + "learning_rate": 3.999585526182308e-06, + "loss": 1.114, + "step": 25430 + }, + { + "epoch": 2.14, + "learning_rate": 3.998858400789842e-06, + "loss": 1.0805, + "step": 25431 + }, + { + "epoch": 2.14, + "learning_rate": 3.998131324980728e-06, + "loss": 1.1487, + "step": 25432 + }, + { + "epoch": 2.14, + "learning_rate": 3.9974042987609754e-06, + "loss": 1.0834, + "step": 25433 + }, + { + "epoch": 2.14, + "learning_rate": 3.99667732213659e-06, + "loss": 1.1205, + "step": 25434 + }, + { + "epoch": 2.14, + "learning_rate": 3.995950395113576e-06, + "loss": 1.1493, + "step": 25435 + }, + { + "epoch": 2.14, + "learning_rate": 3.9952235176979416e-06, + "loss": 1.0433, + "step": 25436 + }, + { + "epoch": 2.14, + "learning_rate": 3.994496689895693e-06, + "loss": 1.1236, + "step": 25437 + }, + { + "epoch": 2.14, + "learning_rate": 3.9937699117128346e-06, + "loss": 1.1491, + "step": 25438 + }, + { + "epoch": 2.14, + "learning_rate": 3.993043183155375e-06, + "loss": 1.1657, + "step": 25439 + }, + { + "epoch": 2.14, + "learning_rate": 3.9923165042293075e-06, + "loss": 1.1226, + "step": 25440 + }, + { + "epoch": 2.14, + "learning_rate": 3.9915898749406525e-06, + "loss": 1.1456, + "step": 25441 + }, + { + "epoch": 2.14, + "learning_rate": 3.9908632952954e-06, + "loss": 1.0877, + "step": 25442 + }, + { + "epoch": 2.14, + "learning_rate": 3.990136765299558e-06, + "loss": 1.2019, + "step": 25443 + }, + { + "epoch": 2.14, + "learning_rate": 3.98941028495913e-06, + "loss": 1.1125, + "step": 25444 + }, + { + "epoch": 2.14, + "learning_rate": 3.988683854280121e-06, + "loss": 1.0317, + "step": 25445 + }, + { + "epoch": 2.14, + "learning_rate": 3.987957473268522e-06, + "loss": 1.1695, + "step": 25446 + }, + { + "epoch": 2.14, + "learning_rate": 3.987231141930351e-06, + "loss": 1.1071, + "step": 25447 + }, + { + "epoch": 2.14, + "learning_rate": 3.986504860271596e-06, + "loss": 1.0978, + "step": 25448 + }, + { + "epoch": 2.14, + "learning_rate": 3.985778628298261e-06, + "loss": 1.0889, + "step": 25449 + }, + { + "epoch": 2.14, + "learning_rate": 3.985052446016352e-06, + "loss": 1.1269, + "step": 25450 + }, + { + "epoch": 2.14, + "learning_rate": 3.9843263134318625e-06, + "loss": 1.1872, + "step": 25451 + }, + { + "epoch": 2.14, + "learning_rate": 3.983600230550794e-06, + "loss": 1.0577, + "step": 25452 + }, + { + "epoch": 2.14, + "learning_rate": 3.982874197379146e-06, + "loss": 1.1704, + "step": 25453 + }, + { + "epoch": 2.14, + "learning_rate": 3.982148213922918e-06, + "loss": 1.0687, + "step": 25454 + }, + { + "epoch": 2.14, + "learning_rate": 3.981422280188107e-06, + "loss": 1.1382, + "step": 25455 + }, + { + "epoch": 2.14, + "learning_rate": 3.9806963961807135e-06, + "loss": 1.1955, + "step": 25456 + }, + { + "epoch": 2.14, + "learning_rate": 3.979970561906726e-06, + "loss": 0.9759, + "step": 25457 + }, + { + "epoch": 2.14, + "learning_rate": 3.979244777372157e-06, + "loss": 1.1904, + "step": 25458 + }, + { + "epoch": 2.14, + "learning_rate": 3.97851904258299e-06, + "loss": 1.2201, + "step": 25459 + }, + { + "epoch": 2.14, + "learning_rate": 3.977793357545222e-06, + "loss": 1.0524, + "step": 25460 + }, + { + "epoch": 2.14, + "learning_rate": 3.9770677222648604e-06, + "loss": 1.1512, + "step": 25461 + }, + { + "epoch": 2.14, + "learning_rate": 3.976342136747889e-06, + "loss": 1.0877, + "step": 25462 + }, + { + "epoch": 2.14, + "learning_rate": 3.975616601000306e-06, + "loss": 1.1629, + "step": 25463 + }, + { + "epoch": 2.14, + "learning_rate": 3.974891115028108e-06, + "loss": 1.118, + "step": 25464 + }, + { + "epoch": 2.14, + "learning_rate": 3.974165678837287e-06, + "loss": 1.0672, + "step": 25465 + }, + { + "epoch": 2.14, + "learning_rate": 3.9734402924338376e-06, + "loss": 1.0732, + "step": 25466 + }, + { + "epoch": 2.14, + "learning_rate": 3.9727149558237576e-06, + "loss": 1.0434, + "step": 25467 + }, + { + "epoch": 2.14, + "learning_rate": 3.971989669013028e-06, + "loss": 1.2079, + "step": 25468 + }, + { + "epoch": 2.14, + "learning_rate": 3.9712644320076565e-06, + "loss": 1.1183, + "step": 25469 + }, + { + "epoch": 2.14, + "learning_rate": 3.970539244813623e-06, + "loss": 1.0632, + "step": 25470 + }, + { + "epoch": 2.14, + "learning_rate": 3.969814107436926e-06, + "loss": 1.0178, + "step": 25471 + }, + { + "epoch": 2.14, + "learning_rate": 3.969089019883553e-06, + "loss": 1.0706, + "step": 25472 + }, + { + "epoch": 2.14, + "learning_rate": 3.968363982159496e-06, + "loss": 1.0856, + "step": 25473 + }, + { + "epoch": 2.14, + "learning_rate": 3.967638994270748e-06, + "loss": 1.0606, + "step": 25474 + }, + { + "epoch": 2.14, + "learning_rate": 3.966914056223297e-06, + "loss": 1.1478, + "step": 25475 + }, + { + "epoch": 2.14, + "learning_rate": 3.966189168023132e-06, + "loss": 1.085, + "step": 25476 + }, + { + "epoch": 2.14, + "learning_rate": 3.965464329676244e-06, + "loss": 1.1843, + "step": 25477 + }, + { + "epoch": 2.15, + "learning_rate": 3.964739541188624e-06, + "loss": 1.0782, + "step": 25478 + }, + { + "epoch": 2.15, + "learning_rate": 3.964014802566251e-06, + "loss": 1.0237, + "step": 25479 + }, + { + "epoch": 2.15, + "learning_rate": 3.963290113815126e-06, + "loss": 1.1564, + "step": 25480 + }, + { + "epoch": 2.15, + "learning_rate": 3.9625654749412254e-06, + "loss": 1.107, + "step": 25481 + }, + { + "epoch": 2.15, + "learning_rate": 3.961840885950543e-06, + "loss": 1.1936, + "step": 25482 + }, + { + "epoch": 2.15, + "learning_rate": 3.961116346849063e-06, + "loss": 1.1949, + "step": 25483 + }, + { + "epoch": 2.15, + "learning_rate": 3.9603918576427725e-06, + "loss": 1.1244, + "step": 25484 + }, + { + "epoch": 2.15, + "learning_rate": 3.959667418337657e-06, + "loss": 1.1357, + "step": 25485 + }, + { + "epoch": 2.15, + "learning_rate": 3.958943028939703e-06, + "loss": 1.1731, + "step": 25486 + }, + { + "epoch": 2.15, + "learning_rate": 3.958218689454895e-06, + "loss": 1.087, + "step": 25487 + }, + { + "epoch": 2.15, + "learning_rate": 3.957494399889216e-06, + "loss": 1.1019, + "step": 25488 + }, + { + "epoch": 2.15, + "learning_rate": 3.956770160248658e-06, + "loss": 1.0354, + "step": 25489 + }, + { + "epoch": 2.15, + "learning_rate": 3.956045970539191e-06, + "loss": 1.1279, + "step": 25490 + }, + { + "epoch": 2.15, + "learning_rate": 3.955321830766814e-06, + "loss": 1.0665, + "step": 25491 + }, + { + "epoch": 2.15, + "learning_rate": 3.954597740937498e-06, + "loss": 1.1014, + "step": 25492 + }, + { + "epoch": 2.15, + "learning_rate": 3.953873701057231e-06, + "loss": 1.1377, + "step": 25493 + }, + { + "epoch": 2.15, + "learning_rate": 3.953149711131995e-06, + "loss": 1.117, + "step": 25494 + }, + { + "epoch": 2.15, + "learning_rate": 3.9524257711677736e-06, + "loss": 1.1628, + "step": 25495 + }, + { + "epoch": 2.15, + "learning_rate": 3.95170188117054e-06, + "loss": 1.133, + "step": 25496 + }, + { + "epoch": 2.15, + "learning_rate": 3.950978041146289e-06, + "loss": 1.0272, + "step": 25497 + }, + { + "epoch": 2.15, + "learning_rate": 3.950254251100988e-06, + "loss": 1.0484, + "step": 25498 + }, + { + "epoch": 2.15, + "learning_rate": 3.94953051104062e-06, + "loss": 1.084, + "step": 25499 + }, + { + "epoch": 2.15, + "learning_rate": 3.948806820971174e-06, + "loss": 1.1687, + "step": 25500 + }, + { + "epoch": 2.15, + "learning_rate": 3.948083180898619e-06, + "loss": 1.0832, + "step": 25501 + }, + { + "epoch": 2.15, + "learning_rate": 3.947359590828938e-06, + "loss": 1.0387, + "step": 25502 + }, + { + "epoch": 2.15, + "learning_rate": 3.946636050768109e-06, + "loss": 1.1108, + "step": 25503 + }, + { + "epoch": 2.15, + "learning_rate": 3.9459125607221104e-06, + "loss": 1.1271, + "step": 25504 + }, + { + "epoch": 2.15, + "learning_rate": 3.9451891206969195e-06, + "loss": 1.175, + "step": 25505 + }, + { + "epoch": 2.15, + "learning_rate": 3.944465730698518e-06, + "loss": 1.0952, + "step": 25506 + }, + { + "epoch": 2.15, + "learning_rate": 3.9437423907328706e-06, + "loss": 1.1633, + "step": 25507 + }, + { + "epoch": 2.15, + "learning_rate": 3.943019100805969e-06, + "loss": 1.0914, + "step": 25508 + }, + { + "epoch": 2.15, + "learning_rate": 3.942295860923779e-06, + "loss": 1.0605, + "step": 25509 + }, + { + "epoch": 2.15, + "learning_rate": 3.941572671092277e-06, + "loss": 1.1117, + "step": 25510 + }, + { + "epoch": 2.15, + "learning_rate": 3.940849531317446e-06, + "loss": 1.0798, + "step": 25511 + }, + { + "epoch": 2.15, + "learning_rate": 3.940126441605252e-06, + "loss": 1.1354, + "step": 25512 + }, + { + "epoch": 2.15, + "learning_rate": 3.939403401961672e-06, + "loss": 1.0979, + "step": 25513 + }, + { + "epoch": 2.15, + "learning_rate": 3.9386804123926825e-06, + "loss": 1.1564, + "step": 25514 + }, + { + "epoch": 2.15, + "learning_rate": 3.937957472904254e-06, + "loss": 1.1103, + "step": 25515 + }, + { + "epoch": 2.15, + "learning_rate": 3.937234583502362e-06, + "loss": 1.056, + "step": 25516 + }, + { + "epoch": 2.15, + "learning_rate": 3.936511744192981e-06, + "loss": 1.0911, + "step": 25517 + }, + { + "epoch": 2.15, + "learning_rate": 3.935788954982074e-06, + "loss": 1.1655, + "step": 25518 + }, + { + "epoch": 2.15, + "learning_rate": 3.935066215875626e-06, + "loss": 1.1907, + "step": 25519 + }, + { + "epoch": 2.15, + "learning_rate": 3.934343526879598e-06, + "loss": 1.1446, + "step": 25520 + }, + { + "epoch": 2.15, + "learning_rate": 3.933620887999966e-06, + "loss": 1.037, + "step": 25521 + }, + { + "epoch": 2.15, + "learning_rate": 3.932898299242699e-06, + "loss": 1.0676, + "step": 25522 + }, + { + "epoch": 2.15, + "learning_rate": 3.932175760613768e-06, + "loss": 1.065, + "step": 25523 + }, + { + "epoch": 2.15, + "learning_rate": 3.931453272119143e-06, + "loss": 1.1928, + "step": 25524 + }, + { + "epoch": 2.15, + "learning_rate": 3.930730833764794e-06, + "loss": 1.2707, + "step": 25525 + }, + { + "epoch": 2.15, + "learning_rate": 3.9300084455566875e-06, + "loss": 1.0754, + "step": 25526 + }, + { + "epoch": 2.15, + "learning_rate": 3.929286107500796e-06, + "loss": 1.1548, + "step": 25527 + }, + { + "epoch": 2.15, + "learning_rate": 3.928563819603088e-06, + "loss": 1.1692, + "step": 25528 + }, + { + "epoch": 2.15, + "learning_rate": 3.9278415818695215e-06, + "loss": 1.0619, + "step": 25529 + }, + { + "epoch": 2.15, + "learning_rate": 3.9271193943060795e-06, + "loss": 0.999, + "step": 25530 + }, + { + "epoch": 2.15, + "learning_rate": 3.926397256918716e-06, + "loss": 1.14, + "step": 25531 + }, + { + "epoch": 2.15, + "learning_rate": 3.925675169713402e-06, + "loss": 1.1547, + "step": 25532 + }, + { + "epoch": 2.15, + "learning_rate": 3.924953132696105e-06, + "loss": 1.0688, + "step": 25533 + }, + { + "epoch": 2.15, + "learning_rate": 3.924231145872788e-06, + "loss": 1.1554, + "step": 25534 + }, + { + "epoch": 2.15, + "learning_rate": 3.923509209249418e-06, + "loss": 1.1375, + "step": 25535 + }, + { + "epoch": 2.15, + "learning_rate": 3.92278732283196e-06, + "loss": 1.0392, + "step": 25536 + }, + { + "epoch": 2.15, + "learning_rate": 3.922065486626379e-06, + "loss": 1.1003, + "step": 25537 + }, + { + "epoch": 2.15, + "learning_rate": 3.921343700638637e-06, + "loss": 1.1667, + "step": 25538 + }, + { + "epoch": 2.15, + "learning_rate": 3.920621964874702e-06, + "loss": 1.1792, + "step": 25539 + }, + { + "epoch": 2.15, + "learning_rate": 3.919900279340527e-06, + "loss": 1.0843, + "step": 25540 + }, + { + "epoch": 2.15, + "learning_rate": 3.91917864404209e-06, + "loss": 1.1579, + "step": 25541 + }, + { + "epoch": 2.15, + "learning_rate": 3.918457058985341e-06, + "loss": 1.0708, + "step": 25542 + }, + { + "epoch": 2.15, + "learning_rate": 3.917735524176246e-06, + "loss": 1.0901, + "step": 25543 + }, + { + "epoch": 2.15, + "learning_rate": 3.9170140396207665e-06, + "loss": 1.0808, + "step": 25544 + }, + { + "epoch": 2.15, + "learning_rate": 3.916292605324867e-06, + "loss": 1.1704, + "step": 25545 + }, + { + "epoch": 2.15, + "learning_rate": 3.9155712212945e-06, + "loss": 1.0735, + "step": 25546 + }, + { + "epoch": 2.15, + "learning_rate": 3.914849887535635e-06, + "loss": 1.0977, + "step": 25547 + }, + { + "epoch": 2.15, + "learning_rate": 3.914128604054226e-06, + "loss": 1.0886, + "step": 25548 + }, + { + "epoch": 2.15, + "learning_rate": 3.91340737085623e-06, + "loss": 1.0725, + "step": 25549 + }, + { + "epoch": 2.15, + "learning_rate": 3.912686187947619e-06, + "loss": 1.1202, + "step": 25550 + }, + { + "epoch": 2.15, + "learning_rate": 3.9119650553343366e-06, + "loss": 1.1448, + "step": 25551 + }, + { + "epoch": 2.15, + "learning_rate": 3.91124397302235e-06, + "loss": 1.142, + "step": 25552 + }, + { + "epoch": 2.15, + "learning_rate": 3.9105229410176125e-06, + "loss": 1.1719, + "step": 25553 + }, + { + "epoch": 2.15, + "learning_rate": 3.909801959326084e-06, + "loss": 1.1435, + "step": 25554 + }, + { + "epoch": 2.15, + "learning_rate": 3.9090810279537204e-06, + "loss": 1.1542, + "step": 25555 + }, + { + "epoch": 2.15, + "learning_rate": 3.9083601469064826e-06, + "loss": 1.136, + "step": 25556 + }, + { + "epoch": 2.15, + "learning_rate": 3.907639316190316e-06, + "loss": 1.1699, + "step": 25557 + }, + { + "epoch": 2.15, + "learning_rate": 3.90691853581119e-06, + "loss": 1.1541, + "step": 25558 + }, + { + "epoch": 2.15, + "learning_rate": 3.906197805775049e-06, + "loss": 0.9679, + "step": 25559 + }, + { + "epoch": 2.15, + "learning_rate": 3.9054771260878496e-06, + "loss": 1.1899, + "step": 25560 + }, + { + "epoch": 2.15, + "learning_rate": 3.904756496755555e-06, + "loss": 1.0917, + "step": 25561 + }, + { + "epoch": 2.15, + "learning_rate": 3.904035917784109e-06, + "loss": 1.1486, + "step": 25562 + }, + { + "epoch": 2.15, + "learning_rate": 3.90331538917947e-06, + "loss": 1.1182, + "step": 25563 + }, + { + "epoch": 2.15, + "learning_rate": 3.90259491094759e-06, + "loss": 1.0687, + "step": 25564 + }, + { + "epoch": 2.15, + "learning_rate": 3.9018744830944236e-06, + "loss": 1.1389, + "step": 25565 + }, + { + "epoch": 2.15, + "learning_rate": 3.901154105625921e-06, + "loss": 1.0456, + "step": 25566 + }, + { + "epoch": 2.15, + "learning_rate": 3.900433778548037e-06, + "loss": 1.0957, + "step": 25567 + }, + { + "epoch": 2.15, + "learning_rate": 3.899713501866716e-06, + "loss": 1.0926, + "step": 25568 + }, + { + "epoch": 2.15, + "learning_rate": 3.898993275587921e-06, + "loss": 1.0722, + "step": 25569 + }, + { + "epoch": 2.15, + "learning_rate": 3.898273099717592e-06, + "loss": 1.121, + "step": 25570 + }, + { + "epoch": 2.15, + "learning_rate": 3.8975529742616845e-06, + "loss": 1.0394, + "step": 25571 + }, + { + "epoch": 2.15, + "learning_rate": 3.896832899226147e-06, + "loss": 1.1192, + "step": 25572 + }, + { + "epoch": 2.15, + "learning_rate": 3.896112874616929e-06, + "loss": 1.0801, + "step": 25573 + }, + { + "epoch": 2.15, + "learning_rate": 3.89539290043998e-06, + "loss": 1.1338, + "step": 25574 + }, + { + "epoch": 2.15, + "learning_rate": 3.894672976701249e-06, + "loss": 1.1407, + "step": 25575 + }, + { + "epoch": 2.15, + "learning_rate": 3.8939531034066835e-06, + "loss": 1.0909, + "step": 25576 + }, + { + "epoch": 2.15, + "learning_rate": 3.8932332805622315e-06, + "loss": 1.0427, + "step": 25577 + }, + { + "epoch": 2.15, + "learning_rate": 3.892513508173845e-06, + "loss": 1.2257, + "step": 25578 + }, + { + "epoch": 2.15, + "learning_rate": 3.891793786247459e-06, + "loss": 1.0669, + "step": 25579 + }, + { + "epoch": 2.15, + "learning_rate": 3.891074114789034e-06, + "loss": 1.0687, + "step": 25580 + }, + { + "epoch": 2.15, + "learning_rate": 3.890354493804507e-06, + "loss": 1.1524, + "step": 25581 + }, + { + "epoch": 2.15, + "learning_rate": 3.889634923299826e-06, + "loss": 1.0513, + "step": 25582 + }, + { + "epoch": 2.15, + "learning_rate": 3.888915403280937e-06, + "loss": 1.1255, + "step": 25583 + }, + { + "epoch": 2.15, + "learning_rate": 3.888195933753785e-06, + "loss": 1.0448, + "step": 25584 + }, + { + "epoch": 2.15, + "learning_rate": 3.887476514724313e-06, + "loss": 1.0617, + "step": 25585 + }, + { + "epoch": 2.15, + "learning_rate": 3.886757146198467e-06, + "loss": 1.1214, + "step": 25586 + }, + { + "epoch": 2.15, + "learning_rate": 3.886037828182191e-06, + "loss": 1.0999, + "step": 25587 + }, + { + "epoch": 2.15, + "learning_rate": 3.885318560681427e-06, + "loss": 1.0176, + "step": 25588 + }, + { + "epoch": 2.15, + "learning_rate": 3.88459934370212e-06, + "loss": 1.2109, + "step": 25589 + }, + { + "epoch": 2.15, + "learning_rate": 3.883880177250204e-06, + "loss": 1.1833, + "step": 25590 + }, + { + "epoch": 2.15, + "learning_rate": 3.883161061331635e-06, + "loss": 1.1591, + "step": 25591 + }, + { + "epoch": 2.15, + "learning_rate": 3.882441995952344e-06, + "loss": 1.0852, + "step": 25592 + }, + { + "epoch": 2.15, + "learning_rate": 3.8817229811182745e-06, + "loss": 1.16, + "step": 25593 + }, + { + "epoch": 2.15, + "learning_rate": 3.881004016835367e-06, + "loss": 1.1555, + "step": 25594 + }, + { + "epoch": 2.15, + "learning_rate": 3.880285103109567e-06, + "loss": 1.0596, + "step": 25595 + }, + { + "epoch": 2.15, + "learning_rate": 3.8795662399468025e-06, + "loss": 1.1615, + "step": 25596 + }, + { + "epoch": 2.16, + "learning_rate": 3.8788474273530275e-06, + "loss": 1.094, + "step": 25597 + }, + { + "epoch": 2.16, + "learning_rate": 3.878128665334171e-06, + "loss": 1.0609, + "step": 25598 + }, + { + "epoch": 2.16, + "learning_rate": 3.8774099538961705e-06, + "loss": 1.1285, + "step": 25599 + }, + { + "epoch": 2.16, + "learning_rate": 3.876691293044976e-06, + "loss": 1.1368, + "step": 25600 + }, + { + "epoch": 2.16, + "learning_rate": 3.8759726827865154e-06, + "loss": 1.1689, + "step": 25601 + }, + { + "epoch": 2.16, + "learning_rate": 3.875254123126727e-06, + "loss": 1.0801, + "step": 25602 + }, + { + "epoch": 2.16, + "learning_rate": 3.8745356140715494e-06, + "loss": 1.0853, + "step": 25603 + }, + { + "epoch": 2.16, + "learning_rate": 3.8738171556269185e-06, + "loss": 1.2606, + "step": 25604 + }, + { + "epoch": 2.16, + "learning_rate": 3.873098747798772e-06, + "loss": 1.1255, + "step": 25605 + }, + { + "epoch": 2.16, + "learning_rate": 3.8723803905930465e-06, + "loss": 1.0546, + "step": 25606 + }, + { + "epoch": 2.16, + "learning_rate": 3.871662084015671e-06, + "loss": 1.0079, + "step": 25607 + }, + { + "epoch": 2.16, + "learning_rate": 3.870943828072589e-06, + "loss": 1.0734, + "step": 25608 + }, + { + "epoch": 2.16, + "learning_rate": 3.8702256227697274e-06, + "loss": 1.1024, + "step": 25609 + }, + { + "epoch": 2.16, + "learning_rate": 3.869507468113022e-06, + "loss": 1.1298, + "step": 25610 + }, + { + "epoch": 2.16, + "learning_rate": 3.868789364108413e-06, + "loss": 1.1349, + "step": 25611 + }, + { + "epoch": 2.16, + "learning_rate": 3.8680713107618254e-06, + "loss": 1.1697, + "step": 25612 + }, + { + "epoch": 2.16, + "learning_rate": 3.867353308079196e-06, + "loss": 1.0995, + "step": 25613 + }, + { + "epoch": 2.16, + "learning_rate": 3.866635356066455e-06, + "loss": 1.1251, + "step": 25614 + }, + { + "epoch": 2.16, + "learning_rate": 3.8659174547295365e-06, + "loss": 1.0994, + "step": 25615 + }, + { + "epoch": 2.16, + "learning_rate": 3.86519960407437e-06, + "loss": 1.1154, + "step": 25616 + }, + { + "epoch": 2.16, + "learning_rate": 3.864481804106892e-06, + "loss": 1.2225, + "step": 25617 + }, + { + "epoch": 2.16, + "learning_rate": 3.863764054833023e-06, + "loss": 1.0617, + "step": 25618 + }, + { + "epoch": 2.16, + "learning_rate": 3.863046356258705e-06, + "loss": 1.0904, + "step": 25619 + }, + { + "epoch": 2.16, + "learning_rate": 3.8623287083898586e-06, + "loss": 1.1048, + "step": 25620 + }, + { + "epoch": 2.16, + "learning_rate": 3.861611111232417e-06, + "loss": 1.1247, + "step": 25621 + }, + { + "epoch": 2.16, + "learning_rate": 3.860893564792308e-06, + "loss": 1.0917, + "step": 25622 + }, + { + "epoch": 2.16, + "learning_rate": 3.860176069075463e-06, + "loss": 1.1508, + "step": 25623 + }, + { + "epoch": 2.16, + "learning_rate": 3.8594586240878076e-06, + "loss": 1.1444, + "step": 25624 + }, + { + "epoch": 2.16, + "learning_rate": 3.858741229835269e-06, + "loss": 1.1896, + "step": 25625 + }, + { + "epoch": 2.16, + "learning_rate": 3.858023886323777e-06, + "loss": 1.1073, + "step": 25626 + }, + { + "epoch": 2.16, + "learning_rate": 3.857306593559257e-06, + "loss": 1.1265, + "step": 25627 + }, + { + "epoch": 2.16, + "learning_rate": 3.8565893515476395e-06, + "loss": 1.194, + "step": 25628 + }, + { + "epoch": 2.16, + "learning_rate": 3.85587216029484e-06, + "loss": 1.1344, + "step": 25629 + }, + { + "epoch": 2.16, + "learning_rate": 3.8551550198067976e-06, + "loss": 1.1349, + "step": 25630 + }, + { + "epoch": 2.16, + "learning_rate": 3.854437930089429e-06, + "loss": 1.0947, + "step": 25631 + }, + { + "epoch": 2.16, + "learning_rate": 3.853720891148659e-06, + "loss": 1.1689, + "step": 25632 + }, + { + "epoch": 2.16, + "learning_rate": 3.8530039029904155e-06, + "loss": 1.1198, + "step": 25633 + }, + { + "epoch": 2.16, + "learning_rate": 3.852286965620621e-06, + "loss": 1.1431, + "step": 25634 + }, + { + "epoch": 2.16, + "learning_rate": 3.851570079045199e-06, + "loss": 1.1207, + "step": 25635 + }, + { + "epoch": 2.16, + "learning_rate": 3.850853243270072e-06, + "loss": 1.0746, + "step": 25636 + }, + { + "epoch": 2.16, + "learning_rate": 3.850136458301164e-06, + "loss": 1.1075, + "step": 25637 + }, + { + "epoch": 2.16, + "learning_rate": 3.849419724144396e-06, + "loss": 1.0441, + "step": 25638 + }, + { + "epoch": 2.16, + "learning_rate": 3.8487030408056945e-06, + "loss": 1.018, + "step": 25639 + }, + { + "epoch": 2.16, + "learning_rate": 3.847986408290971e-06, + "loss": 1.1162, + "step": 25640 + }, + { + "epoch": 2.16, + "learning_rate": 3.847269826606159e-06, + "loss": 1.1096, + "step": 25641 + }, + { + "epoch": 2.16, + "learning_rate": 3.846553295757171e-06, + "loss": 1.078, + "step": 25642 + }, + { + "epoch": 2.16, + "learning_rate": 3.845836815749927e-06, + "loss": 1.0847, + "step": 25643 + }, + { + "epoch": 2.16, + "learning_rate": 3.845120386590349e-06, + "loss": 1.0669, + "step": 25644 + }, + { + "epoch": 2.16, + "learning_rate": 3.84440400828436e-06, + "loss": 1.0543, + "step": 25645 + }, + { + "epoch": 2.16, + "learning_rate": 3.843687680837868e-06, + "loss": 1.1339, + "step": 25646 + }, + { + "epoch": 2.16, + "learning_rate": 3.8429714042568055e-06, + "loss": 1.1093, + "step": 25647 + }, + { + "epoch": 2.16, + "learning_rate": 3.842255178547081e-06, + "loss": 1.1161, + "step": 25648 + }, + { + "epoch": 2.16, + "learning_rate": 3.841539003714612e-06, + "loss": 1.1706, + "step": 25649 + }, + { + "epoch": 2.16, + "learning_rate": 3.840822879765323e-06, + "loss": 1.1437, + "step": 25650 + }, + { + "epoch": 2.16, + "learning_rate": 3.840106806705126e-06, + "loss": 1.0872, + "step": 25651 + }, + { + "epoch": 2.16, + "learning_rate": 3.8393907845399366e-06, + "loss": 1.129, + "step": 25652 + }, + { + "epoch": 2.16, + "learning_rate": 3.838674813275672e-06, + "loss": 1.1142, + "step": 25653 + }, + { + "epoch": 2.16, + "learning_rate": 3.837958892918249e-06, + "loss": 1.0999, + "step": 25654 + }, + { + "epoch": 2.16, + "learning_rate": 3.8372430234735805e-06, + "loss": 1.2254, + "step": 25655 + }, + { + "epoch": 2.16, + "learning_rate": 3.836527204947586e-06, + "loss": 1.1171, + "step": 25656 + }, + { + "epoch": 2.16, + "learning_rate": 3.835811437346171e-06, + "loss": 1.106, + "step": 25657 + }, + { + "epoch": 2.16, + "learning_rate": 3.83509572067526e-06, + "loss": 1.1401, + "step": 25658 + }, + { + "epoch": 2.16, + "learning_rate": 3.834380054940759e-06, + "loss": 1.1047, + "step": 25659 + }, + { + "epoch": 2.16, + "learning_rate": 3.833664440148578e-06, + "loss": 1.083, + "step": 25660 + }, + { + "epoch": 2.16, + "learning_rate": 3.832948876304643e-06, + "loss": 1.0535, + "step": 25661 + }, + { + "epoch": 2.16, + "learning_rate": 3.832233363414854e-06, + "loss": 1.0806, + "step": 25662 + }, + { + "epoch": 2.16, + "learning_rate": 3.831517901485128e-06, + "loss": 1.0974, + "step": 25663 + }, + { + "epoch": 2.16, + "learning_rate": 3.830802490521374e-06, + "loss": 1.1405, + "step": 25664 + }, + { + "epoch": 2.16, + "learning_rate": 3.830087130529505e-06, + "loss": 1.0557, + "step": 25665 + }, + { + "epoch": 2.16, + "learning_rate": 3.829371821515431e-06, + "loss": 1.1641, + "step": 25666 + }, + { + "epoch": 2.16, + "learning_rate": 3.828656563485065e-06, + "loss": 1.1234, + "step": 25667 + }, + { + "epoch": 2.16, + "learning_rate": 3.827941356444307e-06, + "loss": 1.1385, + "step": 25668 + }, + { + "epoch": 2.16, + "learning_rate": 3.827226200399079e-06, + "loss": 1.0438, + "step": 25669 + }, + { + "epoch": 2.16, + "learning_rate": 3.82651109535528e-06, + "loss": 1.1187, + "step": 25670 + }, + { + "epoch": 2.16, + "learning_rate": 3.8257960413188226e-06, + "loss": 1.0814, + "step": 25671 + }, + { + "epoch": 2.16, + "learning_rate": 3.825081038295614e-06, + "loss": 1.0968, + "step": 25672 + }, + { + "epoch": 2.16, + "learning_rate": 3.824366086291563e-06, + "loss": 1.0314, + "step": 25673 + }, + { + "epoch": 2.16, + "learning_rate": 3.823651185312574e-06, + "loss": 1.1122, + "step": 25674 + }, + { + "epoch": 2.16, + "learning_rate": 3.8229363353645575e-06, + "loss": 1.0822, + "step": 25675 + }, + { + "epoch": 2.16, + "learning_rate": 3.822221536453417e-06, + "loss": 1.094, + "step": 25676 + }, + { + "epoch": 2.16, + "learning_rate": 3.8215067885850594e-06, + "loss": 1.0905, + "step": 25677 + }, + { + "epoch": 2.16, + "learning_rate": 3.820792091765394e-06, + "loss": 1.2185, + "step": 25678 + }, + { + "epoch": 2.16, + "learning_rate": 3.820077446000316e-06, + "loss": 1.1309, + "step": 25679 + }, + { + "epoch": 2.16, + "learning_rate": 3.81936285129574e-06, + "loss": 1.1292, + "step": 25680 + }, + { + "epoch": 2.16, + "learning_rate": 3.818648307657565e-06, + "loss": 1.085, + "step": 25681 + }, + { + "epoch": 2.16, + "learning_rate": 3.817933815091695e-06, + "loss": 1.1076, + "step": 25682 + }, + { + "epoch": 2.16, + "learning_rate": 3.817219373604036e-06, + "loss": 1.0812, + "step": 25683 + }, + { + "epoch": 2.16, + "learning_rate": 3.8165049832004865e-06, + "loss": 1.1027, + "step": 25684 + }, + { + "epoch": 2.16, + "learning_rate": 3.815790643886953e-06, + "loss": 1.199, + "step": 25685 + }, + { + "epoch": 2.16, + "learning_rate": 3.815076355669337e-06, + "loss": 1.0714, + "step": 25686 + }, + { + "epoch": 2.16, + "learning_rate": 3.8143621185535397e-06, + "loss": 1.1736, + "step": 25687 + }, + { + "epoch": 2.16, + "learning_rate": 3.8136479325454613e-06, + "loss": 1.1352, + "step": 25688 + }, + { + "epoch": 2.16, + "learning_rate": 3.812933797651007e-06, + "loss": 1.0201, + "step": 25689 + }, + { + "epoch": 2.16, + "learning_rate": 3.8122197138760675e-06, + "loss": 1.1585, + "step": 25690 + }, + { + "epoch": 2.16, + "learning_rate": 3.811505681226556e-06, + "loss": 1.1641, + "step": 25691 + }, + { + "epoch": 2.16, + "learning_rate": 3.8107916997083614e-06, + "loss": 1.0182, + "step": 25692 + }, + { + "epoch": 2.16, + "learning_rate": 3.810077769327386e-06, + "loss": 1.0839, + "step": 25693 + }, + { + "epoch": 2.16, + "learning_rate": 3.80936389008953e-06, + "loss": 1.0489, + "step": 25694 + }, + { + "epoch": 2.16, + "learning_rate": 3.8086500620006937e-06, + "loss": 1.1058, + "step": 25695 + }, + { + "epoch": 2.16, + "learning_rate": 3.8079362850667655e-06, + "loss": 1.1855, + "step": 25696 + }, + { + "epoch": 2.16, + "learning_rate": 3.8072225592936574e-06, + "loss": 1.0375, + "step": 25697 + }, + { + "epoch": 2.16, + "learning_rate": 3.8065088846872532e-06, + "loss": 1.0772, + "step": 25698 + }, + { + "epoch": 2.16, + "learning_rate": 3.8057952612534523e-06, + "loss": 1.1268, + "step": 25699 + }, + { + "epoch": 2.16, + "learning_rate": 3.80508168899816e-06, + "loss": 1.1116, + "step": 25700 + }, + { + "epoch": 2.16, + "learning_rate": 3.8043681679272616e-06, + "loss": 1.1407, + "step": 25701 + }, + { + "epoch": 2.16, + "learning_rate": 3.803654698046657e-06, + "loss": 1.0751, + "step": 25702 + }, + { + "epoch": 2.16, + "learning_rate": 3.80294127936224e-06, + "loss": 1.0807, + "step": 25703 + }, + { + "epoch": 2.16, + "learning_rate": 3.8022279118799054e-06, + "loss": 1.0254, + "step": 25704 + }, + { + "epoch": 2.16, + "learning_rate": 3.8015145956055464e-06, + "loss": 1.138, + "step": 25705 + }, + { + "epoch": 2.16, + "learning_rate": 3.800801330545062e-06, + "loss": 1.0389, + "step": 25706 + }, + { + "epoch": 2.16, + "learning_rate": 3.8000881167043345e-06, + "loss": 1.0448, + "step": 25707 + }, + { + "epoch": 2.16, + "learning_rate": 3.79937495408927e-06, + "loss": 1.1517, + "step": 25708 + }, + { + "epoch": 2.16, + "learning_rate": 3.79866184270575e-06, + "loss": 1.173, + "step": 25709 + }, + { + "epoch": 2.16, + "learning_rate": 3.797948782559667e-06, + "loss": 1.0849, + "step": 25710 + }, + { + "epoch": 2.16, + "learning_rate": 3.7972357736569243e-06, + "loss": 1.145, + "step": 25711 + }, + { + "epoch": 2.16, + "learning_rate": 3.7965228160033997e-06, + "loss": 1.0842, + "step": 25712 + }, + { + "epoch": 2.16, + "learning_rate": 3.7958099096049895e-06, + "loss": 1.041, + "step": 25713 + }, + { + "epoch": 2.16, + "learning_rate": 3.7950970544675838e-06, + "loss": 1.1121, + "step": 25714 + }, + { + "epoch": 2.16, + "learning_rate": 3.7943842505970717e-06, + "loss": 1.1649, + "step": 25715 + }, + { + "epoch": 2.17, + "learning_rate": 3.7936714979993426e-06, + "loss": 1.0132, + "step": 25716 + }, + { + "epoch": 2.17, + "learning_rate": 3.7929587966802907e-06, + "loss": 1.1668, + "step": 25717 + }, + { + "epoch": 2.17, + "learning_rate": 3.7922461466457917e-06, + "loss": 1.0832, + "step": 25718 + }, + { + "epoch": 2.17, + "learning_rate": 3.791533547901749e-06, + "loss": 1.1488, + "step": 25719 + }, + { + "epoch": 2.17, + "learning_rate": 3.790821000454039e-06, + "loss": 1.133, + "step": 25720 + }, + { + "epoch": 2.17, + "learning_rate": 3.790108504308554e-06, + "loss": 1.1095, + "step": 25721 + }, + { + "epoch": 2.17, + "learning_rate": 3.7893960594711797e-06, + "loss": 1.0858, + "step": 25722 + }, + { + "epoch": 2.17, + "learning_rate": 3.7886836659478034e-06, + "loss": 1.039, + "step": 25723 + }, + { + "epoch": 2.17, + "learning_rate": 3.7879713237443093e-06, + "loss": 1.0537, + "step": 25724 + }, + { + "epoch": 2.17, + "learning_rate": 3.7872590328665848e-06, + "loss": 1.1132, + "step": 25725 + }, + { + "epoch": 2.17, + "learning_rate": 3.7865467933205146e-06, + "loss": 1.1054, + "step": 25726 + }, + { + "epoch": 2.17, + "learning_rate": 3.785834605111983e-06, + "loss": 1.1646, + "step": 25727 + }, + { + "epoch": 2.17, + "learning_rate": 3.785122468246878e-06, + "loss": 1.1467, + "step": 25728 + }, + { + "epoch": 2.17, + "learning_rate": 3.784410382731074e-06, + "loss": 1.1155, + "step": 25729 + }, + { + "epoch": 2.17, + "learning_rate": 3.783698348570467e-06, + "loss": 1.1261, + "step": 25730 + }, + { + "epoch": 2.17, + "learning_rate": 3.782986365770931e-06, + "loss": 1.1267, + "step": 25731 + }, + { + "epoch": 2.17, + "learning_rate": 3.78227443433835e-06, + "loss": 1.0741, + "step": 25732 + }, + { + "epoch": 2.17, + "learning_rate": 3.7815625542786084e-06, + "loss": 1.1518, + "step": 25733 + }, + { + "epoch": 2.17, + "learning_rate": 3.780850725597588e-06, + "loss": 1.1946, + "step": 25734 + }, + { + "epoch": 2.17, + "learning_rate": 3.7801389483011675e-06, + "loss": 1.1169, + "step": 25735 + }, + { + "epoch": 2.17, + "learning_rate": 3.779427222395231e-06, + "loss": 1.1405, + "step": 25736 + }, + { + "epoch": 2.17, + "learning_rate": 3.7787155478856573e-06, + "loss": 1.1333, + "step": 25737 + }, + { + "epoch": 2.17, + "learning_rate": 3.778003924778326e-06, + "loss": 1.1852, + "step": 25738 + }, + { + "epoch": 2.17, + "learning_rate": 3.7772923530791217e-06, + "loss": 1.0852, + "step": 25739 + }, + { + "epoch": 2.17, + "learning_rate": 3.7765808327939124e-06, + "loss": 1.1225, + "step": 25740 + }, + { + "epoch": 2.17, + "learning_rate": 3.7758693639285917e-06, + "loss": 1.0971, + "step": 25741 + }, + { + "epoch": 2.17, + "learning_rate": 3.7751579464890264e-06, + "loss": 1.1083, + "step": 25742 + }, + { + "epoch": 2.17, + "learning_rate": 3.774446580481098e-06, + "loss": 1.1421, + "step": 25743 + }, + { + "epoch": 2.17, + "learning_rate": 3.773735265910684e-06, + "loss": 1.0959, + "step": 25744 + }, + { + "epoch": 2.17, + "learning_rate": 3.7730240027836628e-06, + "loss": 1.1954, + "step": 25745 + }, + { + "epoch": 2.17, + "learning_rate": 3.77231279110591e-06, + "loss": 1.1697, + "step": 25746 + }, + { + "epoch": 2.17, + "learning_rate": 3.771601630883305e-06, + "loss": 1.1131, + "step": 25747 + }, + { + "epoch": 2.17, + "learning_rate": 3.770890522121714e-06, + "loss": 1.1917, + "step": 25748 + }, + { + "epoch": 2.17, + "learning_rate": 3.770179464827023e-06, + "loss": 1.0562, + "step": 25749 + }, + { + "epoch": 2.17, + "learning_rate": 3.7694684590051056e-06, + "loss": 1.1403, + "step": 25750 + }, + { + "epoch": 2.17, + "learning_rate": 3.7687575046618273e-06, + "loss": 1.0744, + "step": 25751 + }, + { + "epoch": 2.17, + "learning_rate": 3.768046601803076e-06, + "loss": 1.1109, + "step": 25752 + }, + { + "epoch": 2.17, + "learning_rate": 3.767335750434714e-06, + "loss": 1.1491, + "step": 25753 + }, + { + "epoch": 2.17, + "learning_rate": 3.7666249505626198e-06, + "loss": 1.1829, + "step": 25754 + }, + { + "epoch": 2.17, + "learning_rate": 3.7659142021926643e-06, + "loss": 1.093, + "step": 25755 + }, + { + "epoch": 2.17, + "learning_rate": 3.7652035053307247e-06, + "loss": 1.1286, + "step": 25756 + }, + { + "epoch": 2.17, + "learning_rate": 3.7644928599826626e-06, + "loss": 1.0396, + "step": 25757 + }, + { + "epoch": 2.17, + "learning_rate": 3.763782266154362e-06, + "loss": 1.1805, + "step": 25758 + }, + { + "epoch": 2.17, + "learning_rate": 3.7630717238516857e-06, + "loss": 1.0836, + "step": 25759 + }, + { + "epoch": 2.17, + "learning_rate": 3.7623612330805025e-06, + "loss": 1.0523, + "step": 25760 + }, + { + "epoch": 2.17, + "learning_rate": 3.761650793846695e-06, + "loss": 1.0377, + "step": 25761 + }, + { + "epoch": 2.17, + "learning_rate": 3.7609404061561206e-06, + "loss": 1.0863, + "step": 25762 + }, + { + "epoch": 2.17, + "learning_rate": 3.7602300700146546e-06, + "loss": 1.1233, + "step": 25763 + }, + { + "epoch": 2.17, + "learning_rate": 3.759519785428164e-06, + "loss": 1.116, + "step": 25764 + }, + { + "epoch": 2.17, + "learning_rate": 3.758809552402518e-06, + "loss": 1.0734, + "step": 25765 + }, + { + "epoch": 2.17, + "learning_rate": 3.7580993709435842e-06, + "loss": 1.1539, + "step": 25766 + }, + { + "epoch": 2.17, + "learning_rate": 3.7573892410572366e-06, + "loss": 1.1633, + "step": 25767 + }, + { + "epoch": 2.17, + "learning_rate": 3.756679162749328e-06, + "loss": 1.0305, + "step": 25768 + }, + { + "epoch": 2.17, + "learning_rate": 3.755969136025742e-06, + "loss": 1.0968, + "step": 25769 + }, + { + "epoch": 2.17, + "learning_rate": 3.755259160892334e-06, + "loss": 1.2306, + "step": 25770 + }, + { + "epoch": 2.17, + "learning_rate": 3.754549237354974e-06, + "loss": 1.1248, + "step": 25771 + }, + { + "epoch": 2.17, + "learning_rate": 3.753839365419526e-06, + "loss": 1.0996, + "step": 25772 + }, + { + "epoch": 2.17, + "learning_rate": 3.753129545091857e-06, + "loss": 1.1337, + "step": 25773 + }, + { + "epoch": 2.17, + "learning_rate": 3.75241977637783e-06, + "loss": 1.1115, + "step": 25774 + }, + { + "epoch": 2.17, + "learning_rate": 3.751710059283311e-06, + "loss": 1.1413, + "step": 25775 + }, + { + "epoch": 2.17, + "learning_rate": 3.7510003938141628e-06, + "loss": 1.1532, + "step": 25776 + }, + { + "epoch": 2.17, + "learning_rate": 3.75029077997625e-06, + "loss": 1.1882, + "step": 25777 + }, + { + "epoch": 2.17, + "learning_rate": 3.7495812177754375e-06, + "loss": 1.1126, + "step": 25778 + }, + { + "epoch": 2.17, + "learning_rate": 3.7488717072175784e-06, + "loss": 1.048, + "step": 25779 + }, + { + "epoch": 2.17, + "learning_rate": 3.748162248308549e-06, + "loss": 1.1329, + "step": 25780 + }, + { + "epoch": 2.17, + "learning_rate": 3.747452841054201e-06, + "loss": 1.125, + "step": 25781 + }, + { + "epoch": 2.17, + "learning_rate": 3.746743485460399e-06, + "loss": 1.1144, + "step": 25782 + }, + { + "epoch": 2.17, + "learning_rate": 3.7460341815330027e-06, + "loss": 1.1199, + "step": 25783 + }, + { + "epoch": 2.17, + "learning_rate": 3.745324929277874e-06, + "loss": 1.1417, + "step": 25784 + }, + { + "epoch": 2.17, + "learning_rate": 3.7446157287008723e-06, + "loss": 1.0932, + "step": 25785 + }, + { + "epoch": 2.17, + "learning_rate": 3.7439065798078565e-06, + "loss": 1.1177, + "step": 25786 + }, + { + "epoch": 2.17, + "learning_rate": 3.743197482604688e-06, + "loss": 1.1086, + "step": 25787 + }, + { + "epoch": 2.17, + "learning_rate": 3.7424884370972246e-06, + "loss": 1.1484, + "step": 25788 + }, + { + "epoch": 2.17, + "learning_rate": 3.741779443291327e-06, + "loss": 1.0237, + "step": 25789 + }, + { + "epoch": 2.17, + "learning_rate": 3.7410705011928437e-06, + "loss": 1.2153, + "step": 25790 + }, + { + "epoch": 2.17, + "learning_rate": 3.740361610807647e-06, + "loss": 1.1873, + "step": 25791 + }, + { + "epoch": 2.17, + "learning_rate": 3.739652772141581e-06, + "loss": 1.1873, + "step": 25792 + }, + { + "epoch": 2.17, + "learning_rate": 3.7389439852005085e-06, + "loss": 1.1532, + "step": 25793 + }, + { + "epoch": 2.17, + "learning_rate": 3.738235249990284e-06, + "loss": 1.1759, + "step": 25794 + }, + { + "epoch": 2.17, + "learning_rate": 3.7375265665167636e-06, + "loss": 1.1065, + "step": 25795 + }, + { + "epoch": 2.17, + "learning_rate": 3.7368179347858035e-06, + "loss": 1.1677, + "step": 25796 + }, + { + "epoch": 2.17, + "learning_rate": 3.736109354803261e-06, + "loss": 1.0859, + "step": 25797 + }, + { + "epoch": 2.17, + "learning_rate": 3.7354008265749817e-06, + "loss": 0.993, + "step": 25798 + }, + { + "epoch": 2.17, + "learning_rate": 3.7346923501068286e-06, + "loss": 1.1228, + "step": 25799 + }, + { + "epoch": 2.17, + "learning_rate": 3.733983925404656e-06, + "loss": 1.0779, + "step": 25800 + }, + { + "epoch": 2.17, + "learning_rate": 3.733275552474307e-06, + "loss": 1.1767, + "step": 25801 + }, + { + "epoch": 2.17, + "learning_rate": 3.7325672313216475e-06, + "loss": 1.0946, + "step": 25802 + }, + { + "epoch": 2.17, + "learning_rate": 3.73185896195252e-06, + "loss": 1.1749, + "step": 25803 + }, + { + "epoch": 2.17, + "learning_rate": 3.73115074437278e-06, + "loss": 1.1354, + "step": 25804 + }, + { + "epoch": 2.17, + "learning_rate": 3.730442578588278e-06, + "loss": 1.1096, + "step": 25805 + }, + { + "epoch": 2.17, + "learning_rate": 3.7297344646048706e-06, + "loss": 1.081, + "step": 25806 + }, + { + "epoch": 2.17, + "learning_rate": 3.7290264024283963e-06, + "loss": 1.1429, + "step": 25807 + }, + { + "epoch": 2.17, + "learning_rate": 3.7283183920647194e-06, + "loss": 1.1064, + "step": 25808 + }, + { + "epoch": 2.17, + "learning_rate": 3.7276104335196807e-06, + "loss": 1.1355, + "step": 25809 + }, + { + "epoch": 2.17, + "learning_rate": 3.7269025267991276e-06, + "loss": 1.0839, + "step": 25810 + }, + { + "epoch": 2.17, + "learning_rate": 3.7261946719089204e-06, + "loss": 1.0545, + "step": 25811 + }, + { + "epoch": 2.17, + "learning_rate": 3.725486868854897e-06, + "loss": 1.0356, + "step": 25812 + }, + { + "epoch": 2.17, + "learning_rate": 3.7247791176429106e-06, + "loss": 1.1022, + "step": 25813 + }, + { + "epoch": 2.17, + "learning_rate": 3.7240714182788063e-06, + "loss": 1.0186, + "step": 25814 + }, + { + "epoch": 2.17, + "learning_rate": 3.7233637707684332e-06, + "loss": 1.1977, + "step": 25815 + }, + { + "epoch": 2.17, + "learning_rate": 3.722656175117637e-06, + "loss": 1.0583, + "step": 25816 + }, + { + "epoch": 2.17, + "learning_rate": 3.721948631332267e-06, + "loss": 1.1522, + "step": 25817 + }, + { + "epoch": 2.17, + "learning_rate": 3.7212411394181614e-06, + "loss": 1.0862, + "step": 25818 + }, + { + "epoch": 2.17, + "learning_rate": 3.720533699381177e-06, + "loss": 1.0488, + "step": 25819 + }, + { + "epoch": 2.17, + "learning_rate": 3.7198263112271495e-06, + "loss": 1.1178, + "step": 25820 + }, + { + "epoch": 2.17, + "learning_rate": 3.719118974961927e-06, + "loss": 1.1358, + "step": 25821 + }, + { + "epoch": 2.17, + "learning_rate": 3.7184116905913536e-06, + "loss": 1.0964, + "step": 25822 + }, + { + "epoch": 2.17, + "learning_rate": 3.717704458121274e-06, + "loss": 1.0455, + "step": 25823 + }, + { + "epoch": 2.17, + "learning_rate": 3.7169972775575293e-06, + "loss": 1.0756, + "step": 25824 + }, + { + "epoch": 2.17, + "learning_rate": 3.7162901489059646e-06, + "loss": 1.1144, + "step": 25825 + }, + { + "epoch": 2.17, + "learning_rate": 3.7155830721724217e-06, + "loss": 0.9958, + "step": 25826 + }, + { + "epoch": 2.17, + "learning_rate": 3.7148760473627434e-06, + "loss": 1.0803, + "step": 25827 + }, + { + "epoch": 2.17, + "learning_rate": 3.714169074482773e-06, + "loss": 1.1489, + "step": 25828 + }, + { + "epoch": 2.17, + "learning_rate": 3.713462153538343e-06, + "loss": 1.0407, + "step": 25829 + }, + { + "epoch": 2.17, + "learning_rate": 3.7127552845353077e-06, + "loss": 1.0784, + "step": 25830 + }, + { + "epoch": 2.17, + "learning_rate": 3.7120484674794964e-06, + "loss": 1.0831, + "step": 25831 + }, + { + "epoch": 2.17, + "learning_rate": 3.711341702376754e-06, + "loss": 1.1074, + "step": 25832 + }, + { + "epoch": 2.17, + "learning_rate": 3.7106349892329187e-06, + "loss": 1.0967, + "step": 25833 + }, + { + "epoch": 2.18, + "learning_rate": 3.70992832805383e-06, + "loss": 1.1525, + "step": 25834 + }, + { + "epoch": 2.18, + "learning_rate": 3.709221718845326e-06, + "loss": 1.1942, + "step": 25835 + }, + { + "epoch": 2.18, + "learning_rate": 3.7085151616132463e-06, + "loss": 1.0051, + "step": 25836 + }, + { + "epoch": 2.18, + "learning_rate": 3.7078086563634275e-06, + "loss": 1.0608, + "step": 25837 + }, + { + "epoch": 2.18, + "learning_rate": 3.7071022031017067e-06, + "loss": 1.0661, + "step": 25838 + }, + { + "epoch": 2.18, + "learning_rate": 3.7063958018339265e-06, + "loss": 1.1191, + "step": 25839 + }, + { + "epoch": 2.18, + "learning_rate": 3.7056894525659093e-06, + "loss": 1.137, + "step": 25840 + }, + { + "epoch": 2.18, + "learning_rate": 3.7049831553035085e-06, + "loss": 1.0358, + "step": 25841 + }, + { + "epoch": 2.18, + "learning_rate": 3.704276910052549e-06, + "loss": 1.1326, + "step": 25842 + }, + { + "epoch": 2.18, + "learning_rate": 3.703570716818867e-06, + "loss": 1.0716, + "step": 25843 + }, + { + "epoch": 2.18, + "learning_rate": 3.7028645756082993e-06, + "loss": 1.0544, + "step": 25844 + }, + { + "epoch": 2.18, + "learning_rate": 3.70215848642668e-06, + "loss": 1.0217, + "step": 25845 + }, + { + "epoch": 2.18, + "learning_rate": 3.701452449279843e-06, + "loss": 1.0351, + "step": 25846 + }, + { + "epoch": 2.18, + "learning_rate": 3.700746464173625e-06, + "loss": 1.1543, + "step": 25847 + }, + { + "epoch": 2.18, + "learning_rate": 3.7000405311138487e-06, + "loss": 1.199, + "step": 25848 + }, + { + "epoch": 2.18, + "learning_rate": 3.6993346501063564e-06, + "loss": 1.1203, + "step": 25849 + }, + { + "epoch": 2.18, + "learning_rate": 3.698628821156982e-06, + "loss": 1.1184, + "step": 25850 + }, + { + "epoch": 2.18, + "learning_rate": 3.6979230442715464e-06, + "loss": 1.135, + "step": 25851 + }, + { + "epoch": 2.18, + "learning_rate": 3.6972173194558937e-06, + "loss": 1.1924, + "step": 25852 + }, + { + "epoch": 2.18, + "learning_rate": 3.6965116467158455e-06, + "loss": 1.0908, + "step": 25853 + }, + { + "epoch": 2.18, + "learning_rate": 3.695806026057236e-06, + "loss": 1.0583, + "step": 25854 + }, + { + "epoch": 2.18, + "learning_rate": 3.695100457485894e-06, + "loss": 1.1449, + "step": 25855 + }, + { + "epoch": 2.18, + "learning_rate": 3.694394941007654e-06, + "loss": 1.1514, + "step": 25856 + }, + { + "epoch": 2.18, + "learning_rate": 3.693689476628334e-06, + "loss": 1.1946, + "step": 25857 + }, + { + "epoch": 2.18, + "learning_rate": 3.6929840643537763e-06, + "loss": 1.0971, + "step": 25858 + }, + { + "epoch": 2.18, + "learning_rate": 3.692278704189799e-06, + "loss": 1.1702, + "step": 25859 + }, + { + "epoch": 2.18, + "learning_rate": 3.691573396142231e-06, + "loss": 1.0256, + "step": 25860 + }, + { + "epoch": 2.18, + "learning_rate": 3.6908681402169088e-06, + "loss": 1.0893, + "step": 25861 + }, + { + "epoch": 2.18, + "learning_rate": 3.6901629364196503e-06, + "loss": 1.0468, + "step": 25862 + }, + { + "epoch": 2.18, + "learning_rate": 3.689457784756285e-06, + "loss": 1.2058, + "step": 25863 + }, + { + "epoch": 2.18, + "learning_rate": 3.6887526852326384e-06, + "loss": 1.1131, + "step": 25864 + }, + { + "epoch": 2.18, + "learning_rate": 3.688047637854538e-06, + "loss": 1.0974, + "step": 25865 + }, + { + "epoch": 2.18, + "learning_rate": 3.6873426426278083e-06, + "loss": 1.0407, + "step": 25866 + }, + { + "epoch": 2.18, + "learning_rate": 3.6866376995582765e-06, + "loss": 1.1368, + "step": 25867 + }, + { + "epoch": 2.18, + "learning_rate": 3.6859328086517577e-06, + "loss": 1.0785, + "step": 25868 + }, + { + "epoch": 2.18, + "learning_rate": 3.6852279699140904e-06, + "loss": 1.1407, + "step": 25869 + }, + { + "epoch": 2.18, + "learning_rate": 3.6845231833510862e-06, + "loss": 1.08, + "step": 25870 + }, + { + "epoch": 2.18, + "learning_rate": 3.6838184489685735e-06, + "loss": 1.1472, + "step": 25871 + }, + { + "epoch": 2.18, + "learning_rate": 3.683113766772374e-06, + "loss": 1.0474, + "step": 25872 + }, + { + "epoch": 2.18, + "learning_rate": 3.6824091367683103e-06, + "loss": 1.1299, + "step": 25873 + }, + { + "epoch": 2.18, + "learning_rate": 3.6817045589622037e-06, + "loss": 1.062, + "step": 25874 + }, + { + "epoch": 2.18, + "learning_rate": 3.681000033359876e-06, + "loss": 1.0544, + "step": 25875 + }, + { + "epoch": 2.18, + "learning_rate": 3.6802955599671484e-06, + "loss": 1.1928, + "step": 25876 + }, + { + "epoch": 2.18, + "learning_rate": 3.679591138789843e-06, + "loss": 1.1426, + "step": 25877 + }, + { + "epoch": 2.18, + "learning_rate": 3.6788867698337805e-06, + "loss": 1.1096, + "step": 25878 + }, + { + "epoch": 2.18, + "learning_rate": 3.6781824531047716e-06, + "loss": 1.1041, + "step": 25879 + }, + { + "epoch": 2.18, + "learning_rate": 3.6774781886086495e-06, + "loss": 1.2343, + "step": 25880 + }, + { + "epoch": 2.18, + "learning_rate": 3.676773976351223e-06, + "loss": 1.1281, + "step": 25881 + }, + { + "epoch": 2.18, + "learning_rate": 3.6760698163383135e-06, + "loss": 1.0873, + "step": 25882 + }, + { + "epoch": 2.18, + "learning_rate": 3.675365708575739e-06, + "loss": 1.1205, + "step": 25883 + }, + { + "epoch": 2.18, + "learning_rate": 3.674661653069317e-06, + "loss": 1.0655, + "step": 25884 + }, + { + "epoch": 2.18, + "learning_rate": 3.673957649824865e-06, + "loss": 1.1464, + "step": 25885 + }, + { + "epoch": 2.18, + "learning_rate": 3.6732536988481993e-06, + "loss": 1.0785, + "step": 25886 + }, + { + "epoch": 2.18, + "learning_rate": 3.6725498001451364e-06, + "loss": 1.1262, + "step": 25887 + }, + { + "epoch": 2.18, + "learning_rate": 3.6718459537214924e-06, + "loss": 1.1446, + "step": 25888 + }, + { + "epoch": 2.18, + "learning_rate": 3.6711421595830852e-06, + "loss": 1.2364, + "step": 25889 + }, + { + "epoch": 2.18, + "learning_rate": 3.670438417735721e-06, + "loss": 1.0919, + "step": 25890 + }, + { + "epoch": 2.18, + "learning_rate": 3.669734728185227e-06, + "loss": 1.1275, + "step": 25891 + }, + { + "epoch": 2.18, + "learning_rate": 3.669031090937406e-06, + "loss": 1.071, + "step": 25892 + }, + { + "epoch": 2.18, + "learning_rate": 3.6683275059980774e-06, + "loss": 1.0807, + "step": 25893 + }, + { + "epoch": 2.18, + "learning_rate": 3.6676239733730535e-06, + "loss": 1.1355, + "step": 25894 + }, + { + "epoch": 2.18, + "learning_rate": 3.666920493068147e-06, + "loss": 1.1234, + "step": 25895 + }, + { + "epoch": 2.18, + "learning_rate": 3.6662170650891694e-06, + "loss": 1.1875, + "step": 25896 + }, + { + "epoch": 2.18, + "learning_rate": 3.665513689441937e-06, + "loss": 1.2018, + "step": 25897 + }, + { + "epoch": 2.18, + "learning_rate": 3.664810366132251e-06, + "loss": 1.0704, + "step": 25898 + }, + { + "epoch": 2.18, + "learning_rate": 3.664107095165933e-06, + "loss": 1.0317, + "step": 25899 + }, + { + "epoch": 2.18, + "learning_rate": 3.6634038765487935e-06, + "loss": 1.207, + "step": 25900 + }, + { + "epoch": 2.18, + "learning_rate": 3.662700710286632e-06, + "loss": 1.0654, + "step": 25901 + }, + { + "epoch": 2.18, + "learning_rate": 3.661997596385273e-06, + "loss": 1.213, + "step": 25902 + }, + { + "epoch": 2.18, + "learning_rate": 3.6612945348505137e-06, + "loss": 1.1341, + "step": 25903 + }, + { + "epoch": 2.18, + "learning_rate": 3.6605915256881688e-06, + "loss": 1.072, + "step": 25904 + }, + { + "epoch": 2.18, + "learning_rate": 3.659888568904044e-06, + "loss": 1.1517, + "step": 25905 + }, + { + "epoch": 2.18, + "learning_rate": 3.659185664503954e-06, + "loss": 1.0592, + "step": 25906 + }, + { + "epoch": 2.18, + "learning_rate": 3.6584828124936933e-06, + "loss": 1.1164, + "step": 25907 + }, + { + "epoch": 2.18, + "learning_rate": 3.6577800128790854e-06, + "loss": 1.1033, + "step": 25908 + }, + { + "epoch": 2.18, + "learning_rate": 3.657077265665924e-06, + "loss": 1.1019, + "step": 25909 + }, + { + "epoch": 2.18, + "learning_rate": 3.6563745708600172e-06, + "loss": 1.1909, + "step": 25910 + }, + { + "epoch": 2.18, + "learning_rate": 3.6556719284671804e-06, + "loss": 1.1411, + "step": 25911 + }, + { + "epoch": 2.18, + "learning_rate": 3.6549693384932096e-06, + "loss": 1.1643, + "step": 25912 + }, + { + "epoch": 2.18, + "learning_rate": 3.654266800943913e-06, + "loss": 1.1944, + "step": 25913 + }, + { + "epoch": 2.18, + "learning_rate": 3.6535643158250943e-06, + "loss": 1.091, + "step": 25914 + }, + { + "epoch": 2.18, + "learning_rate": 3.6528618831425587e-06, + "loss": 1.099, + "step": 25915 + }, + { + "epoch": 2.18, + "learning_rate": 3.6521595029021106e-06, + "loss": 1.1267, + "step": 25916 + }, + { + "epoch": 2.18, + "learning_rate": 3.651457175109554e-06, + "loss": 1.1093, + "step": 25917 + }, + { + "epoch": 2.18, + "learning_rate": 3.650754899770684e-06, + "loss": 1.1044, + "step": 25918 + }, + { + "epoch": 2.18, + "learning_rate": 3.6500526768913157e-06, + "loss": 1.1548, + "step": 25919 + }, + { + "epoch": 2.18, + "learning_rate": 3.6493505064772416e-06, + "loss": 1.0998, + "step": 25920 + }, + { + "epoch": 2.18, + "learning_rate": 3.648648388534265e-06, + "loss": 1.203, + "step": 25921 + }, + { + "epoch": 2.18, + "learning_rate": 3.647946323068189e-06, + "loss": 1.1112, + "step": 25922 + }, + { + "epoch": 2.18, + "learning_rate": 3.647244310084813e-06, + "loss": 1.1249, + "step": 25923 + }, + { + "epoch": 2.18, + "learning_rate": 3.6465423495899376e-06, + "loss": 1.2059, + "step": 25924 + }, + { + "epoch": 2.18, + "learning_rate": 3.6458404415893634e-06, + "loss": 1.1528, + "step": 25925 + }, + { + "epoch": 2.18, + "learning_rate": 3.645138586088889e-06, + "loss": 1.1193, + "step": 25926 + }, + { + "epoch": 2.18, + "learning_rate": 3.6444367830943115e-06, + "loss": 1.1593, + "step": 25927 + }, + { + "epoch": 2.18, + "learning_rate": 3.6437350326114375e-06, + "loss": 1.1005, + "step": 25928 + }, + { + "epoch": 2.18, + "learning_rate": 3.64303333464605e-06, + "loss": 1.0773, + "step": 25929 + }, + { + "epoch": 2.18, + "learning_rate": 3.642331689203963e-06, + "loss": 1.1855, + "step": 25930 + }, + { + "epoch": 2.18, + "learning_rate": 3.6416300962909633e-06, + "loss": 1.0525, + "step": 25931 + }, + { + "epoch": 2.18, + "learning_rate": 3.6409285559128505e-06, + "loss": 1.0571, + "step": 25932 + }, + { + "epoch": 2.18, + "learning_rate": 3.6402270680754213e-06, + "loss": 1.1077, + "step": 25933 + }, + { + "epoch": 2.18, + "learning_rate": 3.639525632784471e-06, + "loss": 1.0301, + "step": 25934 + }, + { + "epoch": 2.18, + "learning_rate": 3.638824250045796e-06, + "loss": 1.1075, + "step": 25935 + }, + { + "epoch": 2.18, + "learning_rate": 3.63812291986519e-06, + "loss": 1.0632, + "step": 25936 + }, + { + "epoch": 2.18, + "learning_rate": 3.6374216422484496e-06, + "loss": 1.1242, + "step": 25937 + }, + { + "epoch": 2.18, + "learning_rate": 3.636720417201367e-06, + "loss": 1.1411, + "step": 25938 + }, + { + "epoch": 2.18, + "learning_rate": 3.6360192447297404e-06, + "loss": 1.1333, + "step": 25939 + }, + { + "epoch": 2.18, + "learning_rate": 3.635318124839353e-06, + "loss": 1.0466, + "step": 25940 + }, + { + "epoch": 2.18, + "learning_rate": 3.6346170575360116e-06, + "loss": 1.1158, + "step": 25941 + }, + { + "epoch": 2.18, + "learning_rate": 3.633916042825497e-06, + "loss": 1.0794, + "step": 25942 + }, + { + "epoch": 2.18, + "learning_rate": 3.6332150807136056e-06, + "loss": 1.1425, + "step": 25943 + }, + { + "epoch": 2.18, + "learning_rate": 3.6325141712061286e-06, + "loss": 1.1515, + "step": 25944 + }, + { + "epoch": 2.18, + "learning_rate": 3.6318133143088586e-06, + "loss": 1.079, + "step": 25945 + }, + { + "epoch": 2.18, + "learning_rate": 3.6311125100275833e-06, + "loss": 1.0633, + "step": 25946 + }, + { + "epoch": 2.18, + "learning_rate": 3.6304117583681e-06, + "loss": 1.1481, + "step": 25947 + }, + { + "epoch": 2.18, + "learning_rate": 3.629711059336185e-06, + "loss": 1.1488, + "step": 25948 + }, + { + "epoch": 2.18, + "learning_rate": 3.62901041293764e-06, + "loss": 1.0941, + "step": 25949 + }, + { + "epoch": 2.18, + "learning_rate": 3.6283098191782528e-06, + "loss": 1.058, + "step": 25950 + }, + { + "epoch": 2.18, + "learning_rate": 3.627609278063803e-06, + "loss": 1.0852, + "step": 25951 + }, + { + "epoch": 2.18, + "learning_rate": 3.626908789600091e-06, + "loss": 1.1031, + "step": 25952 + }, + { + "epoch": 2.19, + "learning_rate": 3.626208353792895e-06, + "loss": 1.1179, + "step": 25953 + }, + { + "epoch": 2.19, + "learning_rate": 3.6255079706480058e-06, + "loss": 1.1093, + "step": 25954 + }, + { + "epoch": 2.19, + "learning_rate": 3.6248076401712097e-06, + "loss": 1.0857, + "step": 25955 + }, + { + "epoch": 2.19, + "learning_rate": 3.6241073623682967e-06, + "loss": 1.1136, + "step": 25956 + }, + { + "epoch": 2.19, + "learning_rate": 3.623407137245042e-06, + "loss": 1.0998, + "step": 25957 + }, + { + "epoch": 2.19, + "learning_rate": 3.6227069648072456e-06, + "loss": 1.0721, + "step": 25958 + }, + { + "epoch": 2.19, + "learning_rate": 3.6220068450606814e-06, + "loss": 1.0279, + "step": 25959 + }, + { + "epoch": 2.19, + "learning_rate": 3.6213067780111344e-06, + "loss": 1.111, + "step": 25960 + }, + { + "epoch": 2.19, + "learning_rate": 3.6206067636643993e-06, + "loss": 1.1094, + "step": 25961 + }, + { + "epoch": 2.19, + "learning_rate": 3.6199068020262497e-06, + "loss": 1.0735, + "step": 25962 + }, + { + "epoch": 2.19, + "learning_rate": 3.6192068931024714e-06, + "loss": 1.1224, + "step": 25963 + }, + { + "epoch": 2.19, + "learning_rate": 3.6185070368988484e-06, + "loss": 1.1827, + "step": 25964 + }, + { + "epoch": 2.19, + "learning_rate": 3.6178072334211623e-06, + "loss": 1.1824, + "step": 25965 + }, + { + "epoch": 2.19, + "learning_rate": 3.6171074826751953e-06, + "loss": 1.0275, + "step": 25966 + }, + { + "epoch": 2.19, + "learning_rate": 3.6164077846667323e-06, + "loss": 1.113, + "step": 25967 + }, + { + "epoch": 2.19, + "learning_rate": 3.615708139401545e-06, + "loss": 1.2247, + "step": 25968 + }, + { + "epoch": 2.19, + "learning_rate": 3.615008546885427e-06, + "loss": 1.2021, + "step": 25969 + }, + { + "epoch": 2.19, + "learning_rate": 3.6143090071241484e-06, + "loss": 1.092, + "step": 25970 + }, + { + "epoch": 2.19, + "learning_rate": 3.613609520123492e-06, + "loss": 1.1112, + "step": 25971 + }, + { + "epoch": 2.19, + "learning_rate": 3.6129100858892374e-06, + "loss": 1.0959, + "step": 25972 + }, + { + "epoch": 2.19, + "learning_rate": 3.612210704427165e-06, + "loss": 1.0599, + "step": 25973 + }, + { + "epoch": 2.19, + "learning_rate": 3.6115113757430507e-06, + "loss": 1.0843, + "step": 25974 + }, + { + "epoch": 2.19, + "learning_rate": 3.6108120998426756e-06, + "loss": 1.1551, + "step": 25975 + }, + { + "epoch": 2.19, + "learning_rate": 3.610112876731815e-06, + "loss": 1.0452, + "step": 25976 + }, + { + "epoch": 2.19, + "learning_rate": 3.609413706416246e-06, + "loss": 1.2019, + "step": 25977 + }, + { + "epoch": 2.19, + "learning_rate": 3.6087145889017507e-06, + "loss": 1.0681, + "step": 25978 + }, + { + "epoch": 2.19, + "learning_rate": 3.608015524194094e-06, + "loss": 1.2007, + "step": 25979 + }, + { + "epoch": 2.19, + "learning_rate": 3.607316512299065e-06, + "loss": 1.1907, + "step": 25980 + }, + { + "epoch": 2.19, + "learning_rate": 3.60661755322243e-06, + "loss": 1.0534, + "step": 25981 + }, + { + "epoch": 2.19, + "learning_rate": 3.605918646969967e-06, + "loss": 1.1228, + "step": 25982 + }, + { + "epoch": 2.19, + "learning_rate": 3.6052197935474496e-06, + "loss": 1.1764, + "step": 25983 + }, + { + "epoch": 2.19, + "learning_rate": 3.604520992960654e-06, + "loss": 1.1339, + "step": 25984 + }, + { + "epoch": 2.19, + "learning_rate": 3.6038222452153516e-06, + "loss": 1.1269, + "step": 25985 + }, + { + "epoch": 2.19, + "learning_rate": 3.603123550317318e-06, + "loss": 1.0705, + "step": 25986 + }, + { + "epoch": 2.19, + "learning_rate": 3.602424908272324e-06, + "loss": 1.166, + "step": 25987 + }, + { + "epoch": 2.19, + "learning_rate": 3.601726319086144e-06, + "loss": 1.1981, + "step": 25988 + }, + { + "epoch": 2.19, + "learning_rate": 3.601027782764551e-06, + "loss": 1.159, + "step": 25989 + }, + { + "epoch": 2.19, + "learning_rate": 3.6003292993133087e-06, + "loss": 1.0581, + "step": 25990 + }, + { + "epoch": 2.19, + "learning_rate": 3.5996308687381998e-06, + "loss": 1.0663, + "step": 25991 + }, + { + "epoch": 2.19, + "learning_rate": 3.598932491044986e-06, + "loss": 1.0703, + "step": 25992 + }, + { + "epoch": 2.19, + "learning_rate": 3.5982341662394403e-06, + "loss": 1.1113, + "step": 25993 + }, + { + "epoch": 2.19, + "learning_rate": 3.597535894327333e-06, + "loss": 1.1718, + "step": 25994 + }, + { + "epoch": 2.19, + "learning_rate": 3.596837675314433e-06, + "loss": 1.1706, + "step": 25995 + }, + { + "epoch": 2.19, + "learning_rate": 3.59613950920651e-06, + "loss": 1.1954, + "step": 25996 + }, + { + "epoch": 2.19, + "learning_rate": 3.595441396009335e-06, + "loss": 1.0632, + "step": 25997 + }, + { + "epoch": 2.19, + "learning_rate": 3.5947433357286654e-06, + "loss": 1.0251, + "step": 25998 + }, + { + "epoch": 2.19, + "learning_rate": 3.5940453283702804e-06, + "loss": 1.0246, + "step": 25999 + }, + { + "epoch": 2.19, + "learning_rate": 3.593347373939946e-06, + "loss": 1.0961, + "step": 26000 + }, + { + "epoch": 2.19, + "learning_rate": 3.59264947244342e-06, + "loss": 1.0226, + "step": 26001 + }, + { + "epoch": 2.19, + "learning_rate": 3.5919516238864814e-06, + "loss": 1.0493, + "step": 26002 + }, + { + "epoch": 2.19, + "learning_rate": 3.5912538282748855e-06, + "loss": 1.0905, + "step": 26003 + }, + { + "epoch": 2.19, + "learning_rate": 3.5905560856144015e-06, + "loss": 1.1336, + "step": 26004 + }, + { + "epoch": 2.19, + "learning_rate": 3.5898583959107947e-06, + "loss": 1.056, + "step": 26005 + }, + { + "epoch": 2.19, + "learning_rate": 3.5891607591698317e-06, + "loss": 1.1598, + "step": 26006 + }, + { + "epoch": 2.19, + "learning_rate": 3.5884631753972686e-06, + "loss": 1.213, + "step": 26007 + }, + { + "epoch": 2.19, + "learning_rate": 3.5877656445988807e-06, + "loss": 1.1618, + "step": 26008 + }, + { + "epoch": 2.19, + "learning_rate": 3.587068166780423e-06, + "loss": 1.1223, + "step": 26009 + }, + { + "epoch": 2.19, + "learning_rate": 3.5863707419476555e-06, + "loss": 1.0923, + "step": 26010 + }, + { + "epoch": 2.19, + "learning_rate": 3.585673370106354e-06, + "loss": 1.0368, + "step": 26011 + }, + { + "epoch": 2.19, + "learning_rate": 3.5849760512622677e-06, + "loss": 1.1078, + "step": 26012 + }, + { + "epoch": 2.19, + "learning_rate": 3.5842787854211623e-06, + "loss": 1.1052, + "step": 26013 + }, + { + "epoch": 2.19, + "learning_rate": 3.5835815725887992e-06, + "loss": 1.1146, + "step": 26014 + }, + { + "epoch": 2.19, + "learning_rate": 3.5828844127709383e-06, + "loss": 1.0879, + "step": 26015 + }, + { + "epoch": 2.19, + "learning_rate": 3.58218730597334e-06, + "loss": 1.1314, + "step": 26016 + }, + { + "epoch": 2.19, + "learning_rate": 3.581490252201768e-06, + "loss": 1.1182, + "step": 26017 + }, + { + "epoch": 2.19, + "learning_rate": 3.5807932514619704e-06, + "loss": 1.0711, + "step": 26018 + }, + { + "epoch": 2.19, + "learning_rate": 3.5800963037597204e-06, + "loss": 1.0373, + "step": 26019 + }, + { + "epoch": 2.19, + "learning_rate": 3.5793994091007667e-06, + "loss": 1.0654, + "step": 26020 + }, + { + "epoch": 2.19, + "learning_rate": 3.5787025674908694e-06, + "loss": 1.092, + "step": 26021 + }, + { + "epoch": 2.19, + "learning_rate": 3.5780057789357858e-06, + "loss": 1.1302, + "step": 26022 + }, + { + "epoch": 2.19, + "learning_rate": 3.577309043441275e-06, + "loss": 1.0547, + "step": 26023 + }, + { + "epoch": 2.19, + "learning_rate": 3.576612361013092e-06, + "loss": 1.2134, + "step": 26024 + }, + { + "epoch": 2.19, + "learning_rate": 3.5759157316569926e-06, + "loss": 1.1357, + "step": 26025 + }, + { + "epoch": 2.19, + "learning_rate": 3.575219155378734e-06, + "loss": 1.0829, + "step": 26026 + }, + { + "epoch": 2.19, + "learning_rate": 3.574522632184071e-06, + "loss": 1.1477, + "step": 26027 + }, + { + "epoch": 2.19, + "learning_rate": 3.5738261620787616e-06, + "loss": 1.1279, + "step": 26028 + }, + { + "epoch": 2.19, + "learning_rate": 3.5731297450685507e-06, + "loss": 1.1205, + "step": 26029 + }, + { + "epoch": 2.19, + "learning_rate": 3.572433381159206e-06, + "loss": 1.0586, + "step": 26030 + }, + { + "epoch": 2.19, + "learning_rate": 3.571737070356469e-06, + "loss": 1.2187, + "step": 26031 + }, + { + "epoch": 2.19, + "learning_rate": 3.571040812666099e-06, + "loss": 1.1161, + "step": 26032 + }, + { + "epoch": 2.19, + "learning_rate": 3.5703446080938466e-06, + "loss": 1.0731, + "step": 26033 + }, + { + "epoch": 2.19, + "learning_rate": 3.5696484566454646e-06, + "loss": 1.1298, + "step": 26034 + }, + { + "epoch": 2.19, + "learning_rate": 3.568952358326705e-06, + "loss": 1.1854, + "step": 26035 + }, + { + "epoch": 2.19, + "learning_rate": 3.56825631314332e-06, + "loss": 1.0638, + "step": 26036 + }, + { + "epoch": 2.19, + "learning_rate": 3.567560321101059e-06, + "loss": 1.0936, + "step": 26037 + }, + { + "epoch": 2.19, + "learning_rate": 3.566864382205674e-06, + "loss": 1.163, + "step": 26038 + }, + { + "epoch": 2.19, + "learning_rate": 3.5661684964629174e-06, + "loss": 1.0786, + "step": 26039 + }, + { + "epoch": 2.19, + "learning_rate": 3.5654726638785286e-06, + "loss": 1.1623, + "step": 26040 + }, + { + "epoch": 2.19, + "learning_rate": 3.564776884458271e-06, + "loss": 1.1382, + "step": 26041 + }, + { + "epoch": 2.19, + "learning_rate": 3.5640811582078826e-06, + "loss": 1.1339, + "step": 26042 + }, + { + "epoch": 2.19, + "learning_rate": 3.563385485133116e-06, + "loss": 1.0968, + "step": 26043 + }, + { + "epoch": 2.19, + "learning_rate": 3.562689865239718e-06, + "loss": 1.1233, + "step": 26044 + }, + { + "epoch": 2.19, + "learning_rate": 3.5619942985334366e-06, + "loss": 1.0189, + "step": 26045 + }, + { + "epoch": 2.19, + "learning_rate": 3.5612987850200186e-06, + "loss": 1.1411, + "step": 26046 + }, + { + "epoch": 2.19, + "learning_rate": 3.560603324705215e-06, + "loss": 1.1381, + "step": 26047 + }, + { + "epoch": 2.19, + "learning_rate": 3.5599079175947605e-06, + "loss": 1.1334, + "step": 26048 + }, + { + "epoch": 2.19, + "learning_rate": 3.559212563694411e-06, + "loss": 1.224, + "step": 26049 + }, + { + "epoch": 2.19, + "learning_rate": 3.558517263009912e-06, + "loss": 1.072, + "step": 26050 + }, + { + "epoch": 2.19, + "learning_rate": 3.5578220155469978e-06, + "loss": 1.1307, + "step": 26051 + }, + { + "epoch": 2.19, + "learning_rate": 3.5571268213114273e-06, + "loss": 1.077, + "step": 26052 + }, + { + "epoch": 2.19, + "learning_rate": 3.556431680308933e-06, + "loss": 1.0977, + "step": 26053 + }, + { + "epoch": 2.19, + "learning_rate": 3.5557365925452625e-06, + "loss": 1.05, + "step": 26054 + }, + { + "epoch": 2.19, + "learning_rate": 3.555041558026159e-06, + "loss": 1.0798, + "step": 26055 + }, + { + "epoch": 2.19, + "learning_rate": 3.5543465767573647e-06, + "loss": 1.1886, + "step": 26056 + }, + { + "epoch": 2.19, + "learning_rate": 3.5536516487446215e-06, + "loss": 1.1082, + "step": 26057 + }, + { + "epoch": 2.19, + "learning_rate": 3.5529567739936754e-06, + "loss": 1.1347, + "step": 26058 + }, + { + "epoch": 2.19, + "learning_rate": 3.5522619525102564e-06, + "loss": 1.0761, + "step": 26059 + }, + { + "epoch": 2.19, + "learning_rate": 3.5515671843001165e-06, + "loss": 1.1448, + "step": 26060 + }, + { + "epoch": 2.19, + "learning_rate": 3.5508724693689957e-06, + "loss": 1.1276, + "step": 26061 + }, + { + "epoch": 2.19, + "learning_rate": 3.5501778077226233e-06, + "loss": 1.1691, + "step": 26062 + }, + { + "epoch": 2.19, + "learning_rate": 3.549483199366753e-06, + "loss": 1.0607, + "step": 26063 + }, + { + "epoch": 2.19, + "learning_rate": 3.548788644307115e-06, + "loss": 1.1542, + "step": 26064 + }, + { + "epoch": 2.19, + "learning_rate": 3.5480941425494488e-06, + "loss": 1.0964, + "step": 26065 + }, + { + "epoch": 2.19, + "learning_rate": 3.5473996940994926e-06, + "loss": 1.128, + "step": 26066 + }, + { + "epoch": 2.19, + "learning_rate": 3.5467052989629903e-06, + "loss": 1.1135, + "step": 26067 + }, + { + "epoch": 2.19, + "learning_rate": 3.546010957145666e-06, + "loss": 1.1244, + "step": 26068 + }, + { + "epoch": 2.19, + "learning_rate": 3.5453166686532725e-06, + "loss": 1.1253, + "step": 26069 + }, + { + "epoch": 2.19, + "learning_rate": 3.5446224334915348e-06, + "loss": 1.1347, + "step": 26070 + }, + { + "epoch": 2.19, + "learning_rate": 3.543928251666192e-06, + "loss": 1.1098, + "step": 26071 + }, + { + "epoch": 2.2, + "learning_rate": 3.5432341231829803e-06, + "loss": 1.1373, + "step": 26072 + }, + { + "epoch": 2.2, + "learning_rate": 3.5425400480476346e-06, + "loss": 1.1213, + "step": 26073 + }, + { + "epoch": 2.2, + "learning_rate": 3.5418460262658904e-06, + "loss": 1.0789, + "step": 26074 + }, + { + "epoch": 2.2, + "learning_rate": 3.541152057843481e-06, + "loss": 1.1596, + "step": 26075 + }, + { + "epoch": 2.2, + "learning_rate": 3.5404581427861396e-06, + "loss": 1.2295, + "step": 26076 + }, + { + "epoch": 2.2, + "learning_rate": 3.5397642810996012e-06, + "loss": 1.1563, + "step": 26077 + }, + { + "epoch": 2.2, + "learning_rate": 3.539070472789601e-06, + "loss": 1.2044, + "step": 26078 + }, + { + "epoch": 2.2, + "learning_rate": 3.5383767178618624e-06, + "loss": 1.0166, + "step": 26079 + }, + { + "epoch": 2.2, + "learning_rate": 3.5376830163221286e-06, + "loss": 1.172, + "step": 26080 + }, + { + "epoch": 2.2, + "learning_rate": 3.5369893681761237e-06, + "loss": 1.1198, + "step": 26081 + }, + { + "epoch": 2.2, + "learning_rate": 3.53629577342958e-06, + "loss": 1.1005, + "step": 26082 + }, + { + "epoch": 2.2, + "learning_rate": 3.53560223208823e-06, + "loss": 1.08, + "step": 26083 + }, + { + "epoch": 2.2, + "learning_rate": 3.5349087441578033e-06, + "loss": 1.1368, + "step": 26084 + }, + { + "epoch": 2.2, + "learning_rate": 3.534215309644029e-06, + "loss": 1.1581, + "step": 26085 + }, + { + "epoch": 2.2, + "learning_rate": 3.533521928552638e-06, + "loss": 1.1271, + "step": 26086 + }, + { + "epoch": 2.2, + "learning_rate": 3.5328286008893573e-06, + "loss": 1.1405, + "step": 26087 + }, + { + "epoch": 2.2, + "learning_rate": 3.5321353266599167e-06, + "loss": 1.1077, + "step": 26088 + }, + { + "epoch": 2.2, + "learning_rate": 3.531442105870049e-06, + "loss": 1.2415, + "step": 26089 + }, + { + "epoch": 2.2, + "learning_rate": 3.5307489385254677e-06, + "loss": 1.1152, + "step": 26090 + }, + { + "epoch": 2.2, + "learning_rate": 3.5300558246319172e-06, + "loss": 1.1034, + "step": 26091 + }, + { + "epoch": 2.2, + "learning_rate": 3.5293627641951124e-06, + "loss": 1.09, + "step": 26092 + }, + { + "epoch": 2.2, + "learning_rate": 3.528669757220784e-06, + "loss": 1.1018, + "step": 26093 + }, + { + "epoch": 2.2, + "learning_rate": 3.5279768037146567e-06, + "loss": 1.1224, + "step": 26094 + }, + { + "epoch": 2.2, + "learning_rate": 3.527283903682456e-06, + "loss": 1.1388, + "step": 26095 + }, + { + "epoch": 2.2, + "learning_rate": 3.5265910571299076e-06, + "loss": 1.1071, + "step": 26096 + }, + { + "epoch": 2.2, + "learning_rate": 3.5258982640627393e-06, + "loss": 1.1651, + "step": 26097 + }, + { + "epoch": 2.2, + "learning_rate": 3.5252055244866635e-06, + "loss": 1.0915, + "step": 26098 + }, + { + "epoch": 2.2, + "learning_rate": 3.5245128384074166e-06, + "loss": 1.2008, + "step": 26099 + }, + { + "epoch": 2.2, + "learning_rate": 3.5238202058307203e-06, + "loss": 1.1229, + "step": 26100 + }, + { + "epoch": 2.2, + "learning_rate": 3.523127626762286e-06, + "loss": 1.0809, + "step": 26101 + }, + { + "epoch": 2.2, + "learning_rate": 3.5224351012078527e-06, + "loss": 1.1568, + "step": 26102 + }, + { + "epoch": 2.2, + "learning_rate": 3.5217426291731293e-06, + "loss": 1.1422, + "step": 26103 + }, + { + "epoch": 2.2, + "learning_rate": 3.521050210663841e-06, + "loss": 1.1849, + "step": 26104 + }, + { + "epoch": 2.2, + "learning_rate": 3.52035784568571e-06, + "loss": 1.1912, + "step": 26105 + }, + { + "epoch": 2.2, + "learning_rate": 3.5196655342444576e-06, + "loss": 1.0884, + "step": 26106 + }, + { + "epoch": 2.2, + "learning_rate": 3.5189732763458005e-06, + "loss": 1.0813, + "step": 26107 + }, + { + "epoch": 2.2, + "learning_rate": 3.518281071995465e-06, + "loss": 1.068, + "step": 26108 + }, + { + "epoch": 2.2, + "learning_rate": 3.5175889211991586e-06, + "loss": 1.1409, + "step": 26109 + }, + { + "epoch": 2.2, + "learning_rate": 3.51689682396261e-06, + "loss": 1.0856, + "step": 26110 + }, + { + "epoch": 2.2, + "learning_rate": 3.516204780291539e-06, + "loss": 1.1247, + "step": 26111 + }, + { + "epoch": 2.2, + "learning_rate": 3.515512790191652e-06, + "loss": 1.1395, + "step": 26112 + }, + { + "epoch": 2.2, + "learning_rate": 3.51482085366868e-06, + "loss": 1.024, + "step": 26113 + }, + { + "epoch": 2.2, + "learning_rate": 3.5141289707283312e-06, + "loss": 1.1379, + "step": 26114 + }, + { + "epoch": 2.2, + "learning_rate": 3.5134371413763237e-06, + "loss": 1.1056, + "step": 26115 + }, + { + "epoch": 2.2, + "learning_rate": 3.5127453656183743e-06, + "loss": 1.1091, + "step": 26116 + }, + { + "epoch": 2.2, + "learning_rate": 3.5120536434602026e-06, + "loss": 1.1979, + "step": 26117 + }, + { + "epoch": 2.2, + "learning_rate": 3.5113619749075135e-06, + "loss": 1.1829, + "step": 26118 + }, + { + "epoch": 2.2, + "learning_rate": 3.5106703599660354e-06, + "loss": 1.1806, + "step": 26119 + }, + { + "epoch": 2.2, + "learning_rate": 3.5099787986414723e-06, + "loss": 1.074, + "step": 26120 + }, + { + "epoch": 2.2, + "learning_rate": 3.5092872909395402e-06, + "loss": 1.0954, + "step": 26121 + }, + { + "epoch": 2.2, + "learning_rate": 3.5085958368659556e-06, + "loss": 1.1918, + "step": 26122 + }, + { + "epoch": 2.2, + "learning_rate": 3.507904436426428e-06, + "loss": 1.0824, + "step": 26123 + }, + { + "epoch": 2.2, + "learning_rate": 3.507213089626673e-06, + "loss": 1.0206, + "step": 26124 + }, + { + "epoch": 2.2, + "learning_rate": 3.506521796472401e-06, + "loss": 1.1229, + "step": 26125 + }, + { + "epoch": 2.2, + "learning_rate": 3.505830556969324e-06, + "loss": 1.2246, + "step": 26126 + }, + { + "epoch": 2.2, + "learning_rate": 3.505139371123153e-06, + "loss": 1.078, + "step": 26127 + }, + { + "epoch": 2.2, + "learning_rate": 3.5044482389396028e-06, + "loss": 1.1906, + "step": 26128 + }, + { + "epoch": 2.2, + "learning_rate": 3.5037571604243735e-06, + "loss": 1.0528, + "step": 26129 + }, + { + "epoch": 2.2, + "learning_rate": 3.5030661355831886e-06, + "loss": 1.0838, + "step": 26130 + }, + { + "epoch": 2.2, + "learning_rate": 3.5023751644217473e-06, + "loss": 1.0022, + "step": 26131 + }, + { + "epoch": 2.2, + "learning_rate": 3.501684246945761e-06, + "loss": 1.1494, + "step": 26132 + }, + { + "epoch": 2.2, + "learning_rate": 3.50099338316094e-06, + "loss": 1.0885, + "step": 26133 + }, + { + "epoch": 2.2, + "learning_rate": 3.500302573072991e-06, + "loss": 1.1253, + "step": 26134 + }, + { + "epoch": 2.2, + "learning_rate": 3.4996118166876226e-06, + "loss": 1.2682, + "step": 26135 + }, + { + "epoch": 2.2, + "learning_rate": 3.4989211140105415e-06, + "loss": 1.0197, + "step": 26136 + }, + { + "epoch": 2.2, + "learning_rate": 3.498230465047455e-06, + "loss": 1.1099, + "step": 26137 + }, + { + "epoch": 2.2, + "learning_rate": 3.49753986980407e-06, + "loss": 1.1227, + "step": 26138 + }, + { + "epoch": 2.2, + "learning_rate": 3.4968493282860938e-06, + "loss": 1.141, + "step": 26139 + }, + { + "epoch": 2.2, + "learning_rate": 3.496158840499224e-06, + "loss": 1.1983, + "step": 26140 + }, + { + "epoch": 2.2, + "learning_rate": 3.495468406449177e-06, + "loss": 1.2207, + "step": 26141 + }, + { + "epoch": 2.2, + "learning_rate": 3.4947780261416486e-06, + "loss": 1.0892, + "step": 26142 + }, + { + "epoch": 2.2, + "learning_rate": 3.494087699582347e-06, + "loss": 1.1258, + "step": 26143 + }, + { + "epoch": 2.2, + "learning_rate": 3.4933974267769745e-06, + "loss": 1.0651, + "step": 26144 + }, + { + "epoch": 2.2, + "learning_rate": 3.4927072077312343e-06, + "loss": 1.0988, + "step": 26145 + }, + { + "epoch": 2.2, + "learning_rate": 3.4920170424508303e-06, + "loss": 1.0981, + "step": 26146 + }, + { + "epoch": 2.2, + "learning_rate": 3.491326930941468e-06, + "loss": 1.102, + "step": 26147 + }, + { + "epoch": 2.2, + "learning_rate": 3.490636873208838e-06, + "loss": 1.091, + "step": 26148 + }, + { + "epoch": 2.2, + "learning_rate": 3.4899468692586537e-06, + "loss": 1.1721, + "step": 26149 + }, + { + "epoch": 2.2, + "learning_rate": 3.489256919096614e-06, + "loss": 1.0318, + "step": 26150 + }, + { + "epoch": 2.2, + "learning_rate": 3.488567022728412e-06, + "loss": 1.0772, + "step": 26151 + }, + { + "epoch": 2.2, + "learning_rate": 3.4878771801597588e-06, + "loss": 1.149, + "step": 26152 + }, + { + "epoch": 2.2, + "learning_rate": 3.487187391396346e-06, + "loss": 1.0871, + "step": 26153 + }, + { + "epoch": 2.2, + "learning_rate": 3.4864976564438744e-06, + "loss": 1.0546, + "step": 26154 + }, + { + "epoch": 2.2, + "learning_rate": 3.4858079753080455e-06, + "loss": 1.137, + "step": 26155 + }, + { + "epoch": 2.2, + "learning_rate": 3.485118347994554e-06, + "loss": 1.1608, + "step": 26156 + }, + { + "epoch": 2.2, + "learning_rate": 3.484428774509101e-06, + "loss": 1.12, + "step": 26157 + }, + { + "epoch": 2.2, + "learning_rate": 3.4837392548573846e-06, + "loss": 1.1019, + "step": 26158 + }, + { + "epoch": 2.2, + "learning_rate": 3.4830497890450944e-06, + "loss": 1.0835, + "step": 26159 + }, + { + "epoch": 2.2, + "learning_rate": 3.4823603770779356e-06, + "loss": 1.2102, + "step": 26160 + }, + { + "epoch": 2.2, + "learning_rate": 3.4816710189616053e-06, + "loss": 1.1612, + "step": 26161 + }, + { + "epoch": 2.2, + "learning_rate": 3.4809817147017877e-06, + "loss": 1.1523, + "step": 26162 + }, + { + "epoch": 2.2, + "learning_rate": 3.4802924643041925e-06, + "loss": 1.1696, + "step": 26163 + }, + { + "epoch": 2.2, + "learning_rate": 3.479603267774504e-06, + "loss": 1.0542, + "step": 26164 + }, + { + "epoch": 2.2, + "learning_rate": 3.47891412511842e-06, + "loss": 1.1533, + "step": 26165 + }, + { + "epoch": 2.2, + "learning_rate": 3.478225036341635e-06, + "loss": 1.1003, + "step": 26166 + }, + { + "epoch": 2.2, + "learning_rate": 3.477536001449845e-06, + "loss": 1.1245, + "step": 26167 + }, + { + "epoch": 2.2, + "learning_rate": 3.4768470204487324e-06, + "loss": 1.0614, + "step": 26168 + }, + { + "epoch": 2.2, + "learning_rate": 3.4761580933440053e-06, + "loss": 1.1766, + "step": 26169 + }, + { + "epoch": 2.2, + "learning_rate": 3.475469220141344e-06, + "loss": 1.1554, + "step": 26170 + }, + { + "epoch": 2.2, + "learning_rate": 3.474780400846445e-06, + "loss": 1.1369, + "step": 26171 + }, + { + "epoch": 2.2, + "learning_rate": 3.474091635464998e-06, + "loss": 1.0755, + "step": 26172 + }, + { + "epoch": 2.2, + "learning_rate": 3.4734029240026945e-06, + "loss": 1.137, + "step": 26173 + }, + { + "epoch": 2.2, + "learning_rate": 3.472714266465225e-06, + "loss": 1.1598, + "step": 26174 + }, + { + "epoch": 2.2, + "learning_rate": 3.4720256628582795e-06, + "loss": 1.1889, + "step": 26175 + }, + { + "epoch": 2.2, + "learning_rate": 3.471337113187546e-06, + "loss": 1.0718, + "step": 26176 + }, + { + "epoch": 2.2, + "learning_rate": 3.470648617458716e-06, + "loss": 1.0904, + "step": 26177 + }, + { + "epoch": 2.2, + "learning_rate": 3.4699601756774793e-06, + "loss": 1.1282, + "step": 26178 + }, + { + "epoch": 2.2, + "learning_rate": 3.4692717878495154e-06, + "loss": 1.1396, + "step": 26179 + }, + { + "epoch": 2.2, + "learning_rate": 3.468583453980524e-06, + "loss": 1.0817, + "step": 26180 + }, + { + "epoch": 2.2, + "learning_rate": 3.4678951740761846e-06, + "loss": 1.1066, + "step": 26181 + }, + { + "epoch": 2.2, + "learning_rate": 3.4672069481421845e-06, + "loss": 1.0912, + "step": 26182 + }, + { + "epoch": 2.2, + "learning_rate": 3.466518776184211e-06, + "loss": 1.0674, + "step": 26183 + }, + { + "epoch": 2.2, + "learning_rate": 3.4658306582079516e-06, + "loss": 1.1356, + "step": 26184 + }, + { + "epoch": 2.2, + "learning_rate": 3.4651425942190895e-06, + "loss": 1.1724, + "step": 26185 + }, + { + "epoch": 2.2, + "learning_rate": 3.4644545842233114e-06, + "loss": 1.1035, + "step": 26186 + }, + { + "epoch": 2.2, + "learning_rate": 3.4637666282263004e-06, + "loss": 1.0422, + "step": 26187 + }, + { + "epoch": 2.2, + "learning_rate": 3.463078726233743e-06, + "loss": 1.1087, + "step": 26188 + }, + { + "epoch": 2.2, + "learning_rate": 3.4623908782513225e-06, + "loss": 1.0752, + "step": 26189 + }, + { + "epoch": 2.2, + "learning_rate": 3.4617030842847153e-06, + "loss": 1.1748, + "step": 26190 + }, + { + "epoch": 2.21, + "learning_rate": 3.4610153443396168e-06, + "loss": 1.0877, + "step": 26191 + }, + { + "epoch": 2.21, + "learning_rate": 3.460327658421698e-06, + "loss": 1.0816, + "step": 26192 + }, + { + "epoch": 2.21, + "learning_rate": 3.4596400265366458e-06, + "loss": 1.0634, + "step": 26193 + }, + { + "epoch": 2.21, + "learning_rate": 3.4589524486901406e-06, + "loss": 1.0408, + "step": 26194 + }, + { + "epoch": 2.21, + "learning_rate": 3.458264924887863e-06, + "loss": 1.1513, + "step": 26195 + }, + { + "epoch": 2.21, + "learning_rate": 3.457577455135496e-06, + "loss": 1.1173, + "step": 26196 + }, + { + "epoch": 2.21, + "learning_rate": 3.4568900394387193e-06, + "loss": 1.1732, + "step": 26197 + }, + { + "epoch": 2.21, + "learning_rate": 3.4562026778032053e-06, + "loss": 1.1021, + "step": 26198 + }, + { + "epoch": 2.21, + "learning_rate": 3.4555153702346424e-06, + "loss": 1.1752, + "step": 26199 + }, + { + "epoch": 2.21, + "learning_rate": 3.4548281167387087e-06, + "loss": 1.0845, + "step": 26200 + }, + { + "epoch": 2.21, + "learning_rate": 3.454140917321074e-06, + "loss": 1.245, + "step": 26201 + }, + { + "epoch": 2.21, + "learning_rate": 3.4534537719874284e-06, + "loss": 1.1516, + "step": 26202 + }, + { + "epoch": 2.21, + "learning_rate": 3.45276668074344e-06, + "loss": 1.1968, + "step": 26203 + }, + { + "epoch": 2.21, + "learning_rate": 3.4520796435947877e-06, + "loss": 1.0992, + "step": 26204 + }, + { + "epoch": 2.21, + "learning_rate": 3.4513926605471504e-06, + "loss": 1.0931, + "step": 26205 + }, + { + "epoch": 2.21, + "learning_rate": 3.450705731606202e-06, + "loss": 1.0638, + "step": 26206 + }, + { + "epoch": 2.21, + "learning_rate": 3.4500188567776184e-06, + "loss": 1.1537, + "step": 26207 + }, + { + "epoch": 2.21, + "learning_rate": 3.449332036067079e-06, + "loss": 1.0953, + "step": 26208 + }, + { + "epoch": 2.21, + "learning_rate": 3.4486452694802475e-06, + "loss": 1.1065, + "step": 26209 + }, + { + "epoch": 2.21, + "learning_rate": 3.4479585570228104e-06, + "loss": 1.1315, + "step": 26210 + }, + { + "epoch": 2.21, + "learning_rate": 3.447271898700438e-06, + "loss": 1.1603, + "step": 26211 + }, + { + "epoch": 2.21, + "learning_rate": 3.446585294518797e-06, + "loss": 1.1121, + "step": 26212 + }, + { + "epoch": 2.21, + "learning_rate": 3.4458987444835724e-06, + "loss": 1.0916, + "step": 26213 + }, + { + "epoch": 2.21, + "learning_rate": 3.4452122486004257e-06, + "loss": 1.1298, + "step": 26214 + }, + { + "epoch": 2.21, + "learning_rate": 3.444525806875034e-06, + "loss": 1.0624, + "step": 26215 + }, + { + "epoch": 2.21, + "learning_rate": 3.4438394193130675e-06, + "loss": 1.1323, + "step": 26216 + }, + { + "epoch": 2.21, + "learning_rate": 3.443153085920202e-06, + "loss": 1.1202, + "step": 26217 + }, + { + "epoch": 2.21, + "learning_rate": 3.4424668067020973e-06, + "loss": 1.0638, + "step": 26218 + }, + { + "epoch": 2.21, + "learning_rate": 3.4417805816644368e-06, + "loss": 1.0798, + "step": 26219 + }, + { + "epoch": 2.21, + "learning_rate": 3.441094410812881e-06, + "loss": 1.0063, + "step": 26220 + }, + { + "epoch": 2.21, + "learning_rate": 3.4404082941531016e-06, + "loss": 1.0715, + "step": 26221 + }, + { + "epoch": 2.21, + "learning_rate": 3.4397222316907686e-06, + "loss": 1.1509, + "step": 26222 + }, + { + "epoch": 2.21, + "learning_rate": 3.4390362234315512e-06, + "loss": 1.2222, + "step": 26223 + }, + { + "epoch": 2.21, + "learning_rate": 3.4383502693811143e-06, + "loss": 1.0712, + "step": 26224 + }, + { + "epoch": 2.21, + "learning_rate": 3.4376643695451296e-06, + "loss": 1.0697, + "step": 26225 + }, + { + "epoch": 2.21, + "learning_rate": 3.4369785239292608e-06, + "loss": 1.0796, + "step": 26226 + }, + { + "epoch": 2.21, + "learning_rate": 3.4362927325391758e-06, + "loss": 1.1199, + "step": 26227 + }, + { + "epoch": 2.21, + "learning_rate": 3.435606995380545e-06, + "loss": 1.122, + "step": 26228 + }, + { + "epoch": 2.21, + "learning_rate": 3.434921312459022e-06, + "loss": 1.1803, + "step": 26229 + }, + { + "epoch": 2.21, + "learning_rate": 3.434235683780288e-06, + "loss": 1.1284, + "step": 26230 + }, + { + "epoch": 2.21, + "learning_rate": 3.4335501093499968e-06, + "loss": 1.233, + "step": 26231 + }, + { + "epoch": 2.21, + "learning_rate": 3.4328645891738154e-06, + "loss": 1.1633, + "step": 26232 + }, + { + "epoch": 2.21, + "learning_rate": 3.4321791232574096e-06, + "loss": 1.1157, + "step": 26233 + }, + { + "epoch": 2.21, + "learning_rate": 3.4314937116064415e-06, + "loss": 1.1322, + "step": 26234 + }, + { + "epoch": 2.21, + "learning_rate": 3.430808354226575e-06, + "loss": 1.1562, + "step": 26235 + }, + { + "epoch": 2.21, + "learning_rate": 3.4301230511234706e-06, + "loss": 1.0845, + "step": 26236 + }, + { + "epoch": 2.21, + "learning_rate": 3.429437802302793e-06, + "loss": 1.0778, + "step": 26237 + }, + { + "epoch": 2.21, + "learning_rate": 3.428752607770204e-06, + "loss": 1.1736, + "step": 26238 + }, + { + "epoch": 2.21, + "learning_rate": 3.4280674675313663e-06, + "loss": 1.0664, + "step": 26239 + }, + { + "epoch": 2.21, + "learning_rate": 3.427382381591933e-06, + "loss": 1.1325, + "step": 26240 + }, + { + "epoch": 2.21, + "learning_rate": 3.426697349957575e-06, + "loss": 1.1135, + "step": 26241 + }, + { + "epoch": 2.21, + "learning_rate": 3.4260123726339443e-06, + "loss": 1.1437, + "step": 26242 + }, + { + "epoch": 2.21, + "learning_rate": 3.425327449626704e-06, + "loss": 1.1187, + "step": 26243 + }, + { + "epoch": 2.21, + "learning_rate": 3.4246425809415128e-06, + "loss": 1.0536, + "step": 26244 + }, + { + "epoch": 2.21, + "learning_rate": 3.4239577665840284e-06, + "loss": 1.1363, + "step": 26245 + }, + { + "epoch": 2.21, + "learning_rate": 3.4232730065599096e-06, + "loss": 1.0571, + "step": 26246 + }, + { + "epoch": 2.21, + "learning_rate": 3.4225883008748173e-06, + "loss": 1.069, + "step": 26247 + }, + { + "epoch": 2.21, + "learning_rate": 3.4219036495343984e-06, + "loss": 1.059, + "step": 26248 + }, + { + "epoch": 2.21, + "learning_rate": 3.42121905254432e-06, + "loss": 1.105, + "step": 26249 + }, + { + "epoch": 2.21, + "learning_rate": 3.4205345099102392e-06, + "loss": 1.0775, + "step": 26250 + }, + { + "epoch": 2.21, + "learning_rate": 3.4198500216378018e-06, + "loss": 1.1145, + "step": 26251 + }, + { + "epoch": 2.21, + "learning_rate": 3.419165587732676e-06, + "loss": 1.0846, + "step": 26252 + }, + { + "epoch": 2.21, + "learning_rate": 3.418481208200506e-06, + "loss": 1.1181, + "step": 26253 + }, + { + "epoch": 2.21, + "learning_rate": 3.417796883046951e-06, + "loss": 1.1724, + "step": 26254 + }, + { + "epoch": 2.21, + "learning_rate": 3.4171126122776653e-06, + "loss": 1.0626, + "step": 26255 + }, + { + "epoch": 2.21, + "learning_rate": 3.4164283958983015e-06, + "loss": 1.1873, + "step": 26256 + }, + { + "epoch": 2.21, + "learning_rate": 3.415744233914513e-06, + "loss": 1.1612, + "step": 26257 + }, + { + "epoch": 2.21, + "learning_rate": 3.415060126331957e-06, + "loss": 1.1146, + "step": 26258 + }, + { + "epoch": 2.21, + "learning_rate": 3.414376073156275e-06, + "loss": 1.1477, + "step": 26259 + }, + { + "epoch": 2.21, + "learning_rate": 3.4136920743931278e-06, + "loss": 1.005, + "step": 26260 + }, + { + "epoch": 2.21, + "learning_rate": 3.413008130048169e-06, + "loss": 1.0604, + "step": 26261 + }, + { + "epoch": 2.21, + "learning_rate": 3.4123242401270373e-06, + "loss": 1.1996, + "step": 26262 + }, + { + "epoch": 2.21, + "learning_rate": 3.411640404635399e-06, + "loss": 1.0793, + "step": 26263 + }, + { + "epoch": 2.21, + "learning_rate": 3.4109566235788915e-06, + "loss": 1.1906, + "step": 26264 + }, + { + "epoch": 2.21, + "learning_rate": 3.4102728969631706e-06, + "loss": 1.1175, + "step": 26265 + }, + { + "epoch": 2.21, + "learning_rate": 3.409589224793883e-06, + "loss": 1.0282, + "step": 26266 + }, + { + "epoch": 2.21, + "learning_rate": 3.4089056070766825e-06, + "loss": 1.0444, + "step": 26267 + }, + { + "epoch": 2.21, + "learning_rate": 3.4082220438172066e-06, + "loss": 1.1024, + "step": 26268 + }, + { + "epoch": 2.21, + "learning_rate": 3.4075385350211154e-06, + "loss": 1.127, + "step": 26269 + }, + { + "epoch": 2.21, + "learning_rate": 3.4068550806940483e-06, + "loss": 1.1841, + "step": 26270 + }, + { + "epoch": 2.21, + "learning_rate": 3.4061716808416546e-06, + "loss": 1.1485, + "step": 26271 + }, + { + "epoch": 2.21, + "learning_rate": 3.405488335469581e-06, + "loss": 1.1484, + "step": 26272 + }, + { + "epoch": 2.21, + "learning_rate": 3.4048050445834724e-06, + "loss": 1.0965, + "step": 26273 + }, + { + "epoch": 2.21, + "learning_rate": 3.404121808188976e-06, + "loss": 1.1176, + "step": 26274 + }, + { + "epoch": 2.21, + "learning_rate": 3.4034386262917364e-06, + "loss": 1.2192, + "step": 26275 + }, + { + "epoch": 2.21, + "learning_rate": 3.4027554988973967e-06, + "loss": 1.1214, + "step": 26276 + }, + { + "epoch": 2.21, + "learning_rate": 3.4020724260116032e-06, + "loss": 1.0094, + "step": 26277 + }, + { + "epoch": 2.21, + "learning_rate": 3.4013894076400024e-06, + "loss": 1.035, + "step": 26278 + }, + { + "epoch": 2.21, + "learning_rate": 3.4007064437882278e-06, + "loss": 1.0715, + "step": 26279 + }, + { + "epoch": 2.21, + "learning_rate": 3.4000235344619346e-06, + "loss": 1.1191, + "step": 26280 + }, + { + "epoch": 2.21, + "learning_rate": 3.399340679666756e-06, + "loss": 1.0821, + "step": 26281 + }, + { + "epoch": 2.21, + "learning_rate": 3.398657879408338e-06, + "loss": 1.2476, + "step": 26282 + }, + { + "epoch": 2.21, + "learning_rate": 3.3979751336923205e-06, + "loss": 1.0775, + "step": 26283 + }, + { + "epoch": 2.21, + "learning_rate": 3.397292442524346e-06, + "loss": 1.1269, + "step": 26284 + }, + { + "epoch": 2.21, + "learning_rate": 3.396609805910054e-06, + "loss": 1.1207, + "step": 26285 + }, + { + "epoch": 2.21, + "learning_rate": 3.3959272238550846e-06, + "loss": 1.1071, + "step": 26286 + }, + { + "epoch": 2.21, + "learning_rate": 3.395244696365079e-06, + "loss": 1.1705, + "step": 26287 + }, + { + "epoch": 2.21, + "learning_rate": 3.3945622234456744e-06, + "loss": 1.1511, + "step": 26288 + }, + { + "epoch": 2.21, + "learning_rate": 3.3938798051025144e-06, + "loss": 1.1042, + "step": 26289 + }, + { + "epoch": 2.21, + "learning_rate": 3.3931974413412284e-06, + "loss": 1.0749, + "step": 26290 + }, + { + "epoch": 2.21, + "learning_rate": 3.3925151321674643e-06, + "loss": 1.076, + "step": 26291 + }, + { + "epoch": 2.21, + "learning_rate": 3.3918328775868524e-06, + "loss": 1.0712, + "step": 26292 + }, + { + "epoch": 2.21, + "learning_rate": 3.391150677605033e-06, + "loss": 1.0418, + "step": 26293 + }, + { + "epoch": 2.21, + "learning_rate": 3.390468532227641e-06, + "loss": 1.0575, + "step": 26294 + }, + { + "epoch": 2.21, + "learning_rate": 3.3897864414603133e-06, + "loss": 1.0893, + "step": 26295 + }, + { + "epoch": 2.21, + "learning_rate": 3.389104405308685e-06, + "loss": 1.1556, + "step": 26296 + }, + { + "epoch": 2.21, + "learning_rate": 3.388422423778396e-06, + "loss": 1.0915, + "step": 26297 + }, + { + "epoch": 2.21, + "learning_rate": 3.3877404968750703e-06, + "loss": 1.1461, + "step": 26298 + }, + { + "epoch": 2.21, + "learning_rate": 3.3870586246043512e-06, + "loss": 1.0966, + "step": 26299 + }, + { + "epoch": 2.21, + "learning_rate": 3.386376806971875e-06, + "loss": 1.1698, + "step": 26300 + }, + { + "epoch": 2.21, + "learning_rate": 3.3856950439832625e-06, + "loss": 1.1117, + "step": 26301 + }, + { + "epoch": 2.21, + "learning_rate": 3.3850133356441605e-06, + "loss": 1.1081, + "step": 26302 + }, + { + "epoch": 2.21, + "learning_rate": 3.3843316819601936e-06, + "loss": 1.1172, + "step": 26303 + }, + { + "epoch": 2.21, + "learning_rate": 3.3836500829369945e-06, + "loss": 1.1443, + "step": 26304 + }, + { + "epoch": 2.21, + "learning_rate": 3.3829685385801956e-06, + "loss": 1.1539, + "step": 26305 + }, + { + "epoch": 2.21, + "learning_rate": 3.3822870488954297e-06, + "loss": 1.1858, + "step": 26306 + }, + { + "epoch": 2.21, + "learning_rate": 3.3816056138883257e-06, + "loss": 1.0913, + "step": 26307 + }, + { + "epoch": 2.21, + "learning_rate": 3.3809242335645176e-06, + "loss": 1.1551, + "step": 26308 + }, + { + "epoch": 2.21, + "learning_rate": 3.380242907929625e-06, + "loss": 1.1254, + "step": 26309 + }, + { + "epoch": 2.22, + "learning_rate": 3.379561636989287e-06, + "loss": 1.166, + "step": 26310 + }, + { + "epoch": 2.22, + "learning_rate": 3.3788804207491334e-06, + "loss": 1.0483, + "step": 26311 + }, + { + "epoch": 2.22, + "learning_rate": 3.378199259214782e-06, + "loss": 1.0847, + "step": 26312 + }, + { + "epoch": 2.22, + "learning_rate": 3.3775181523918743e-06, + "loss": 1.148, + "step": 26313 + }, + { + "epoch": 2.22, + "learning_rate": 3.3768371002860277e-06, + "loss": 1.137, + "step": 26314 + }, + { + "epoch": 2.22, + "learning_rate": 3.376156102902872e-06, + "loss": 1.2005, + "step": 26315 + }, + { + "epoch": 2.22, + "learning_rate": 3.375475160248036e-06, + "loss": 1.0818, + "step": 26316 + }, + { + "epoch": 2.22, + "learning_rate": 3.3747942723271464e-06, + "loss": 1.0725, + "step": 26317 + }, + { + "epoch": 2.22, + "learning_rate": 3.3741134391458207e-06, + "loss": 1.035, + "step": 26318 + }, + { + "epoch": 2.22, + "learning_rate": 3.3734326607096968e-06, + "loss": 1.0788, + "step": 26319 + }, + { + "epoch": 2.22, + "learning_rate": 3.37275193702439e-06, + "loss": 1.065, + "step": 26320 + }, + { + "epoch": 2.22, + "learning_rate": 3.372071268095527e-06, + "loss": 1.0957, + "step": 26321 + }, + { + "epoch": 2.22, + "learning_rate": 3.3713906539287334e-06, + "loss": 1.1385, + "step": 26322 + }, + { + "epoch": 2.22, + "learning_rate": 3.3707100945296312e-06, + "loss": 1.1715, + "step": 26323 + }, + { + "epoch": 2.22, + "learning_rate": 3.370029589903844e-06, + "loss": 1.0852, + "step": 26324 + }, + { + "epoch": 2.22, + "learning_rate": 3.3693491400569935e-06, + "loss": 1.1449, + "step": 26325 + }, + { + "epoch": 2.22, + "learning_rate": 3.368668744994703e-06, + "loss": 1.1072, + "step": 26326 + }, + { + "epoch": 2.22, + "learning_rate": 3.367988404722594e-06, + "loss": 1.001, + "step": 26327 + }, + { + "epoch": 2.22, + "learning_rate": 3.36730811924629e-06, + "loss": 1.1255, + "step": 26328 + }, + { + "epoch": 2.22, + "learning_rate": 3.366627888571402e-06, + "loss": 1.0953, + "step": 26329 + }, + { + "epoch": 2.22, + "learning_rate": 3.3659477127035656e-06, + "loss": 1.1, + "step": 26330 + }, + { + "epoch": 2.22, + "learning_rate": 3.3652675916483877e-06, + "loss": 1.1315, + "step": 26331 + }, + { + "epoch": 2.22, + "learning_rate": 3.3645875254114923e-06, + "loss": 1.0601, + "step": 26332 + }, + { + "epoch": 2.22, + "learning_rate": 3.363907513998499e-06, + "loss": 1.1127, + "step": 26333 + }, + { + "epoch": 2.22, + "learning_rate": 3.363227557415025e-06, + "loss": 1.0972, + "step": 26334 + }, + { + "epoch": 2.22, + "learning_rate": 3.3625476556666905e-06, + "loss": 1.0473, + "step": 26335 + }, + { + "epoch": 2.22, + "learning_rate": 3.3618678087591093e-06, + "loss": 1.1787, + "step": 26336 + }, + { + "epoch": 2.22, + "learning_rate": 3.361188016697903e-06, + "loss": 1.137, + "step": 26337 + }, + { + "epoch": 2.22, + "learning_rate": 3.3605082794886846e-06, + "loss": 1.1471, + "step": 26338 + }, + { + "epoch": 2.22, + "learning_rate": 3.359828597137075e-06, + "loss": 1.0832, + "step": 26339 + }, + { + "epoch": 2.22, + "learning_rate": 3.35914896964868e-06, + "loss": 1.1582, + "step": 26340 + }, + { + "epoch": 2.22, + "learning_rate": 3.358469397029128e-06, + "loss": 1.1122, + "step": 26341 + }, + { + "epoch": 2.22, + "learning_rate": 3.3577898792840246e-06, + "loss": 1.0616, + "step": 26342 + }, + { + "epoch": 2.22, + "learning_rate": 3.357110416418987e-06, + "loss": 1.0903, + "step": 26343 + }, + { + "epoch": 2.22, + "learning_rate": 3.35643100843963e-06, + "loss": 1.178, + "step": 26344 + }, + { + "epoch": 2.22, + "learning_rate": 3.3557516553515646e-06, + "loss": 1.1571, + "step": 26345 + }, + { + "epoch": 2.22, + "learning_rate": 3.355072357160406e-06, + "loss": 1.1358, + "step": 26346 + }, + { + "epoch": 2.22, + "learning_rate": 3.3543931138717667e-06, + "loss": 1.1325, + "step": 26347 + }, + { + "epoch": 2.22, + "learning_rate": 3.3537139254912585e-06, + "loss": 1.0482, + "step": 26348 + }, + { + "epoch": 2.22, + "learning_rate": 3.3530347920244923e-06, + "loss": 1.0518, + "step": 26349 + }, + { + "epoch": 2.22, + "learning_rate": 3.3523557134770833e-06, + "loss": 1.1336, + "step": 26350 + }, + { + "epoch": 2.22, + "learning_rate": 3.3516766898546326e-06, + "loss": 1.1267, + "step": 26351 + }, + { + "epoch": 2.22, + "learning_rate": 3.350997721162763e-06, + "loss": 1.1298, + "step": 26352 + }, + { + "epoch": 2.22, + "learning_rate": 3.3503188074070756e-06, + "loss": 1.1409, + "step": 26353 + }, + { + "epoch": 2.22, + "learning_rate": 3.349639948593182e-06, + "loss": 1.1389, + "step": 26354 + }, + { + "epoch": 2.22, + "learning_rate": 3.3489611447266913e-06, + "loss": 1.2073, + "step": 26355 + }, + { + "epoch": 2.22, + "learning_rate": 3.3482823958132126e-06, + "loss": 1.1944, + "step": 26356 + }, + { + "epoch": 2.22, + "learning_rate": 3.347603701858353e-06, + "loss": 1.1687, + "step": 26357 + }, + { + "epoch": 2.22, + "learning_rate": 3.346925062867724e-06, + "loss": 1.1399, + "step": 26358 + }, + { + "epoch": 2.22, + "learning_rate": 3.346246478846924e-06, + "loss": 1.1081, + "step": 26359 + }, + { + "epoch": 2.22, + "learning_rate": 3.345567949801567e-06, + "loss": 1.1112, + "step": 26360 + }, + { + "epoch": 2.22, + "learning_rate": 3.344889475737261e-06, + "loss": 1.1151, + "step": 26361 + }, + { + "epoch": 2.22, + "learning_rate": 3.344211056659602e-06, + "loss": 1.0398, + "step": 26362 + }, + { + "epoch": 2.22, + "learning_rate": 3.3435326925742073e-06, + "loss": 1.1488, + "step": 26363 + }, + { + "epoch": 2.22, + "learning_rate": 3.342854383486672e-06, + "loss": 1.1165, + "step": 26364 + }, + { + "epoch": 2.22, + "learning_rate": 3.3421761294026054e-06, + "loss": 1.1315, + "step": 26365 + }, + { + "epoch": 2.22, + "learning_rate": 3.3414979303276108e-06, + "loss": 1.0857, + "step": 26366 + }, + { + "epoch": 2.22, + "learning_rate": 3.3408197862672895e-06, + "loss": 1.1476, + "step": 26367 + }, + { + "epoch": 2.22, + "learning_rate": 3.3401416972272483e-06, + "loss": 1.0714, + "step": 26368 + }, + { + "epoch": 2.22, + "learning_rate": 3.3394636632130893e-06, + "loss": 1.0976, + "step": 26369 + }, + { + "epoch": 2.22, + "learning_rate": 3.3387856842304076e-06, + "loss": 1.1115, + "step": 26370 + }, + { + "epoch": 2.22, + "learning_rate": 3.338107760284818e-06, + "loss": 1.0452, + "step": 26371 + }, + { + "epoch": 2.22, + "learning_rate": 3.337429891381909e-06, + "loss": 1.0839, + "step": 26372 + }, + { + "epoch": 2.22, + "learning_rate": 3.336752077527288e-06, + "loss": 1.1455, + "step": 26373 + }, + { + "epoch": 2.22, + "learning_rate": 3.336074318726553e-06, + "loss": 1.1644, + "step": 26374 + }, + { + "epoch": 2.22, + "learning_rate": 3.3353966149853044e-06, + "loss": 1.2359, + "step": 26375 + }, + { + "epoch": 2.22, + "learning_rate": 3.334718966309143e-06, + "loss": 1.0617, + "step": 26376 + }, + { + "epoch": 2.22, + "learning_rate": 3.3340413727036658e-06, + "loss": 1.1277, + "step": 26377 + }, + { + "epoch": 2.22, + "learning_rate": 3.3333638341744757e-06, + "loss": 1.1512, + "step": 26378 + }, + { + "epoch": 2.22, + "learning_rate": 3.33268635072716e-06, + "loss": 1.1044, + "step": 26379 + }, + { + "epoch": 2.22, + "learning_rate": 3.3320089223673315e-06, + "loss": 1.1363, + "step": 26380 + }, + { + "epoch": 2.22, + "learning_rate": 3.3313315491005748e-06, + "loss": 1.2112, + "step": 26381 + }, + { + "epoch": 2.22, + "learning_rate": 3.3306542309324917e-06, + "loss": 1.0947, + "step": 26382 + }, + { + "epoch": 2.22, + "learning_rate": 3.329976967868678e-06, + "loss": 1.1432, + "step": 26383 + }, + { + "epoch": 2.22, + "learning_rate": 3.329299759914728e-06, + "loss": 1.1394, + "step": 26384 + }, + { + "epoch": 2.22, + "learning_rate": 3.3286226070762395e-06, + "loss": 1.1704, + "step": 26385 + }, + { + "epoch": 2.22, + "learning_rate": 3.3279455093588063e-06, + "loss": 1.1468, + "step": 26386 + }, + { + "epoch": 2.22, + "learning_rate": 3.327268466768022e-06, + "loss": 1.1165, + "step": 26387 + }, + { + "epoch": 2.22, + "learning_rate": 3.326591479309481e-06, + "loss": 1.1297, + "step": 26388 + }, + { + "epoch": 2.22, + "learning_rate": 3.325914546988781e-06, + "loss": 1.0823, + "step": 26389 + }, + { + "epoch": 2.22, + "learning_rate": 3.3252376698115043e-06, + "loss": 1.1086, + "step": 26390 + }, + { + "epoch": 2.22, + "learning_rate": 3.324560847783256e-06, + "loss": 1.1253, + "step": 26391 + }, + { + "epoch": 2.22, + "learning_rate": 3.3238840809096197e-06, + "loss": 1.1162, + "step": 26392 + }, + { + "epoch": 2.22, + "learning_rate": 3.3232073691961895e-06, + "loss": 1.0807, + "step": 26393 + }, + { + "epoch": 2.22, + "learning_rate": 3.322530712648556e-06, + "loss": 1.1031, + "step": 26394 + }, + { + "epoch": 2.22, + "learning_rate": 3.321854111272312e-06, + "loss": 1.2048, + "step": 26395 + }, + { + "epoch": 2.22, + "learning_rate": 3.321177565073047e-06, + "loss": 1.143, + "step": 26396 + }, + { + "epoch": 2.22, + "learning_rate": 3.3205010740563494e-06, + "loss": 1.102, + "step": 26397 + }, + { + "epoch": 2.22, + "learning_rate": 3.3198246382278097e-06, + "loss": 1.0308, + "step": 26398 + }, + { + "epoch": 2.22, + "learning_rate": 3.3191482575930177e-06, + "loss": 1.1494, + "step": 26399 + }, + { + "epoch": 2.22, + "learning_rate": 3.318471932157564e-06, + "loss": 1.1859, + "step": 26400 + }, + { + "epoch": 2.22, + "learning_rate": 3.317795661927027e-06, + "loss": 1.0964, + "step": 26401 + }, + { + "epoch": 2.22, + "learning_rate": 3.3171194469070067e-06, + "loss": 1.0563, + "step": 26402 + }, + { + "epoch": 2.22, + "learning_rate": 3.3164432871030815e-06, + "loss": 1.1279, + "step": 26403 + }, + { + "epoch": 2.22, + "learning_rate": 3.3157671825208404e-06, + "loss": 1.0956, + "step": 26404 + }, + { + "epoch": 2.22, + "learning_rate": 3.315091133165871e-06, + "loss": 1.19, + "step": 26405 + }, + { + "epoch": 2.22, + "learning_rate": 3.3144151390437572e-06, + "loss": 1.0925, + "step": 26406 + }, + { + "epoch": 2.22, + "learning_rate": 3.3137392001600855e-06, + "loss": 1.1294, + "step": 26407 + }, + { + "epoch": 2.22, + "learning_rate": 3.313063316520444e-06, + "loss": 1.1411, + "step": 26408 + }, + { + "epoch": 2.22, + "learning_rate": 3.3123874881304064e-06, + "loss": 1.1291, + "step": 26409 + }, + { + "epoch": 2.22, + "learning_rate": 3.311711714995567e-06, + "loss": 1.1083, + "step": 26410 + }, + { + "epoch": 2.22, + "learning_rate": 3.3110359971215087e-06, + "loss": 1.148, + "step": 26411 + }, + { + "epoch": 2.22, + "learning_rate": 3.310360334513807e-06, + "loss": 1.1373, + "step": 26412 + }, + { + "epoch": 2.22, + "learning_rate": 3.309684727178054e-06, + "loss": 1.1527, + "step": 26413 + }, + { + "epoch": 2.22, + "learning_rate": 3.3090091751198237e-06, + "loss": 1.0742, + "step": 26414 + }, + { + "epoch": 2.22, + "learning_rate": 3.308333678344702e-06, + "loss": 1.0061, + "step": 26415 + }, + { + "epoch": 2.22, + "learning_rate": 3.3076582368582686e-06, + "loss": 1.1259, + "step": 26416 + }, + { + "epoch": 2.22, + "learning_rate": 3.306982850666105e-06, + "loss": 1.152, + "step": 26417 + }, + { + "epoch": 2.22, + "learning_rate": 3.3063075197737903e-06, + "loss": 1.0919, + "step": 26418 + }, + { + "epoch": 2.22, + "learning_rate": 3.3056322441869095e-06, + "loss": 1.1109, + "step": 26419 + }, + { + "epoch": 2.22, + "learning_rate": 3.3049570239110307e-06, + "loss": 1.1074, + "step": 26420 + }, + { + "epoch": 2.22, + "learning_rate": 3.3042818589517456e-06, + "loss": 1.1267, + "step": 26421 + }, + { + "epoch": 2.22, + "learning_rate": 3.303606749314624e-06, + "loss": 1.1134, + "step": 26422 + }, + { + "epoch": 2.22, + "learning_rate": 3.3029316950052428e-06, + "loss": 1.1145, + "step": 26423 + }, + { + "epoch": 2.22, + "learning_rate": 3.30225669602919e-06, + "loss": 1.159, + "step": 26424 + }, + { + "epoch": 2.22, + "learning_rate": 3.3015817523920324e-06, + "loss": 1.1667, + "step": 26425 + }, + { + "epoch": 2.22, + "learning_rate": 3.3009068640993514e-06, + "loss": 1.0935, + "step": 26426 + }, + { + "epoch": 2.22, + "learning_rate": 3.3002320311567206e-06, + "loss": 1.0722, + "step": 26427 + }, + { + "epoch": 2.23, + "learning_rate": 3.2995572535697207e-06, + "loss": 1.1121, + "step": 26428 + }, + { + "epoch": 2.23, + "learning_rate": 3.2988825313439167e-06, + "loss": 1.142, + "step": 26429 + }, + { + "epoch": 2.23, + "learning_rate": 3.2982078644848956e-06, + "loss": 1.1564, + "step": 26430 + }, + { + "epoch": 2.23, + "learning_rate": 3.297533252998223e-06, + "loss": 1.2048, + "step": 26431 + }, + { + "epoch": 2.23, + "learning_rate": 3.2968586968894767e-06, + "loss": 1.0614, + "step": 26432 + }, + { + "epoch": 2.23, + "learning_rate": 3.2961841961642284e-06, + "loss": 1.1285, + "step": 26433 + }, + { + "epoch": 2.23, + "learning_rate": 3.295509750828052e-06, + "loss": 1.1244, + "step": 26434 + }, + { + "epoch": 2.23, + "learning_rate": 3.2948353608865192e-06, + "loss": 1.13, + "step": 26435 + }, + { + "epoch": 2.23, + "learning_rate": 3.294161026345203e-06, + "loss": 1.0719, + "step": 26436 + }, + { + "epoch": 2.23, + "learning_rate": 3.2934867472096743e-06, + "loss": 1.0305, + "step": 26437 + }, + { + "epoch": 2.23, + "learning_rate": 3.2928125234855047e-06, + "loss": 1.1623, + "step": 26438 + }, + { + "epoch": 2.23, + "learning_rate": 3.2921383551782684e-06, + "loss": 1.1035, + "step": 26439 + }, + { + "epoch": 2.23, + "learning_rate": 3.291464242293524e-06, + "loss": 1.163, + "step": 26440 + }, + { + "epoch": 2.23, + "learning_rate": 3.290790184836856e-06, + "loss": 1.1162, + "step": 26441 + }, + { + "epoch": 2.23, + "learning_rate": 3.290116182813824e-06, + "loss": 1.1227, + "step": 26442 + }, + { + "epoch": 2.23, + "learning_rate": 3.289442236229999e-06, + "loss": 1.1033, + "step": 26443 + }, + { + "epoch": 2.23, + "learning_rate": 3.28876834509095e-06, + "loss": 1.1518, + "step": 26444 + }, + { + "epoch": 2.23, + "learning_rate": 3.2880945094022456e-06, + "loss": 1.2214, + "step": 26445 + }, + { + "epoch": 2.23, + "learning_rate": 3.2874207291694517e-06, + "loss": 1.1281, + "step": 26446 + }, + { + "epoch": 2.23, + "learning_rate": 3.2867470043981365e-06, + "loss": 1.0744, + "step": 26447 + }, + { + "epoch": 2.23, + "learning_rate": 3.2860733350938658e-06, + "loss": 1.0991, + "step": 26448 + }, + { + "epoch": 2.23, + "learning_rate": 3.285399721262206e-06, + "loss": 1.1223, + "step": 26449 + }, + { + "epoch": 2.23, + "learning_rate": 3.284726162908725e-06, + "loss": 1.1619, + "step": 26450 + }, + { + "epoch": 2.23, + "learning_rate": 3.28405266003898e-06, + "loss": 1.0923, + "step": 26451 + }, + { + "epoch": 2.23, + "learning_rate": 3.283379212658547e-06, + "loss": 1.0499, + "step": 26452 + }, + { + "epoch": 2.23, + "learning_rate": 3.2827058207729822e-06, + "loss": 1.148, + "step": 26453 + }, + { + "epoch": 2.23, + "learning_rate": 3.2820324843878515e-06, + "loss": 1.1019, + "step": 26454 + }, + { + "epoch": 2.23, + "learning_rate": 3.281359203508718e-06, + "loss": 1.0957, + "step": 26455 + }, + { + "epoch": 2.23, + "learning_rate": 3.280685978141146e-06, + "loss": 1.1084, + "step": 26456 + }, + { + "epoch": 2.23, + "learning_rate": 3.280012808290697e-06, + "loss": 1.079, + "step": 26457 + }, + { + "epoch": 2.23, + "learning_rate": 3.279339693962935e-06, + "loss": 1.1652, + "step": 26458 + }, + { + "epoch": 2.23, + "learning_rate": 3.2786666351634132e-06, + "loss": 1.1218, + "step": 26459 + }, + { + "epoch": 2.23, + "learning_rate": 3.2779936318977013e-06, + "loss": 1.0219, + "step": 26460 + }, + { + "epoch": 2.23, + "learning_rate": 3.2773206841713624e-06, + "loss": 1.1112, + "step": 26461 + }, + { + "epoch": 2.23, + "learning_rate": 3.276647791989943e-06, + "loss": 1.0477, + "step": 26462 + }, + { + "epoch": 2.23, + "learning_rate": 3.275974955359019e-06, + "loss": 1.1427, + "step": 26463 + }, + { + "epoch": 2.23, + "learning_rate": 3.2753021742841383e-06, + "loss": 1.1515, + "step": 26464 + }, + { + "epoch": 2.23, + "learning_rate": 3.274629448770863e-06, + "loss": 1.2046, + "step": 26465 + }, + { + "epoch": 2.23, + "learning_rate": 3.2739567788247526e-06, + "loss": 1.0922, + "step": 26466 + }, + { + "epoch": 2.23, + "learning_rate": 3.2732841644513625e-06, + "loss": 1.1111, + "step": 26467 + }, + { + "epoch": 2.23, + "learning_rate": 3.272611605656253e-06, + "loss": 1.1037, + "step": 26468 + }, + { + "epoch": 2.23, + "learning_rate": 3.271939102444981e-06, + "loss": 1.1202, + "step": 26469 + }, + { + "epoch": 2.23, + "learning_rate": 3.2712666548230965e-06, + "loss": 1.0438, + "step": 26470 + }, + { + "epoch": 2.23, + "learning_rate": 3.2705942627961663e-06, + "loss": 1.1703, + "step": 26471 + }, + { + "epoch": 2.23, + "learning_rate": 3.2699219263697358e-06, + "loss": 1.134, + "step": 26472 + }, + { + "epoch": 2.23, + "learning_rate": 3.269249645549362e-06, + "loss": 1.1524, + "step": 26473 + }, + { + "epoch": 2.23, + "learning_rate": 3.2685774203406082e-06, + "loss": 1.064, + "step": 26474 + }, + { + "epoch": 2.23, + "learning_rate": 3.2679052507490184e-06, + "loss": 1.0771, + "step": 26475 + }, + { + "epoch": 2.23, + "learning_rate": 3.26723313678015e-06, + "loss": 1.1414, + "step": 26476 + }, + { + "epoch": 2.23, + "learning_rate": 3.266561078439556e-06, + "loss": 1.1356, + "step": 26477 + }, + { + "epoch": 2.23, + "learning_rate": 3.2658890757327934e-06, + "loss": 1.1094, + "step": 26478 + }, + { + "epoch": 2.23, + "learning_rate": 3.2652171286654032e-06, + "loss": 1.1296, + "step": 26479 + }, + { + "epoch": 2.23, + "learning_rate": 3.264545237242952e-06, + "loss": 1.164, + "step": 26480 + }, + { + "epoch": 2.23, + "learning_rate": 3.2638734014709793e-06, + "loss": 1.1275, + "step": 26481 + }, + { + "epoch": 2.23, + "learning_rate": 3.263201621355041e-06, + "loss": 1.1117, + "step": 26482 + }, + { + "epoch": 2.23, + "learning_rate": 3.2625298969006867e-06, + "loss": 1.1698, + "step": 26483 + }, + { + "epoch": 2.23, + "learning_rate": 3.261858228113467e-06, + "loss": 1.125, + "step": 26484 + }, + { + "epoch": 2.23, + "learning_rate": 3.2611866149989302e-06, + "loss": 1.1779, + "step": 26485 + }, + { + "epoch": 2.23, + "learning_rate": 3.2605150575626275e-06, + "loss": 1.0562, + "step": 26486 + }, + { + "epoch": 2.23, + "learning_rate": 3.259843555810106e-06, + "loss": 1.1438, + "step": 26487 + }, + { + "epoch": 2.23, + "learning_rate": 3.2591721097469132e-06, + "loss": 1.0505, + "step": 26488 + }, + { + "epoch": 2.23, + "learning_rate": 3.258500719378602e-06, + "loss": 1.0804, + "step": 26489 + }, + { + "epoch": 2.23, + "learning_rate": 3.257829384710709e-06, + "loss": 1.1008, + "step": 26490 + }, + { + "epoch": 2.23, + "learning_rate": 3.2571581057487935e-06, + "loss": 1.0888, + "step": 26491 + }, + { + "epoch": 2.23, + "learning_rate": 3.2564868824983918e-06, + "loss": 1.124, + "step": 26492 + }, + { + "epoch": 2.23, + "learning_rate": 3.255815714965055e-06, + "loss": 1.1602, + "step": 26493 + }, + { + "epoch": 2.23, + "learning_rate": 3.2551446031543265e-06, + "loss": 1.2026, + "step": 26494 + }, + { + "epoch": 2.23, + "learning_rate": 3.2544735470717515e-06, + "loss": 1.1798, + "step": 26495 + }, + { + "epoch": 2.23, + "learning_rate": 3.253802546722875e-06, + "loss": 1.1405, + "step": 26496 + }, + { + "epoch": 2.23, + "learning_rate": 3.253131602113242e-06, + "loss": 1.0511, + "step": 26497 + }, + { + "epoch": 2.23, + "learning_rate": 3.2524607132483944e-06, + "loss": 1.1837, + "step": 26498 + }, + { + "epoch": 2.23, + "learning_rate": 3.2517898801338753e-06, + "loss": 1.1732, + "step": 26499 + }, + { + "epoch": 2.23, + "learning_rate": 3.2511191027752322e-06, + "loss": 1.2108, + "step": 26500 + }, + { + "epoch": 2.23, + "learning_rate": 3.2504483811779954e-06, + "loss": 1.093, + "step": 26501 + }, + { + "epoch": 2.23, + "learning_rate": 3.249777715347722e-06, + "loss": 1.1439, + "step": 26502 + }, + { + "epoch": 2.23, + "learning_rate": 3.2491071052899404e-06, + "loss": 1.0225, + "step": 26503 + }, + { + "epoch": 2.23, + "learning_rate": 3.2484365510101977e-06, + "loss": 1.14, + "step": 26504 + }, + { + "epoch": 2.23, + "learning_rate": 3.247766052514033e-06, + "loss": 1.1583, + "step": 26505 + }, + { + "epoch": 2.23, + "learning_rate": 3.247095609806986e-06, + "loss": 1.1052, + "step": 26506 + }, + { + "epoch": 2.23, + "learning_rate": 3.2464252228945957e-06, + "loss": 1.1959, + "step": 26507 + }, + { + "epoch": 2.23, + "learning_rate": 3.2457548917824055e-06, + "loss": 1.1302, + "step": 26508 + }, + { + "epoch": 2.23, + "learning_rate": 3.245084616475943e-06, + "loss": 1.0706, + "step": 26509 + }, + { + "epoch": 2.23, + "learning_rate": 3.2444143969807574e-06, + "loss": 1.1071, + "step": 26510 + }, + { + "epoch": 2.23, + "learning_rate": 3.243744233302385e-06, + "loss": 1.154, + "step": 26511 + }, + { + "epoch": 2.23, + "learning_rate": 3.2430741254463537e-06, + "loss": 1.1284, + "step": 26512 + }, + { + "epoch": 2.23, + "learning_rate": 3.2424040734182127e-06, + "loss": 1.1153, + "step": 26513 + }, + { + "epoch": 2.23, + "learning_rate": 3.2417340772234884e-06, + "loss": 1.1417, + "step": 26514 + }, + { + "epoch": 2.23, + "learning_rate": 3.2410641368677197e-06, + "loss": 1.0744, + "step": 26515 + }, + { + "epoch": 2.23, + "learning_rate": 3.2403942523564436e-06, + "loss": 1.1531, + "step": 26516 + }, + { + "epoch": 2.23, + "learning_rate": 3.239724423695193e-06, + "loss": 1.12, + "step": 26517 + }, + { + "epoch": 2.23, + "learning_rate": 3.239054650889503e-06, + "loss": 1.0751, + "step": 26518 + }, + { + "epoch": 2.23, + "learning_rate": 3.2383849339449094e-06, + "loss": 1.1196, + "step": 26519 + }, + { + "epoch": 2.23, + "learning_rate": 3.237715272866938e-06, + "loss": 1.0734, + "step": 26520 + }, + { + "epoch": 2.23, + "learning_rate": 3.2370456676611328e-06, + "loss": 1.1003, + "step": 26521 + }, + { + "epoch": 2.23, + "learning_rate": 3.236376118333018e-06, + "loss": 1.1633, + "step": 26522 + }, + { + "epoch": 2.23, + "learning_rate": 3.235706624888125e-06, + "loss": 1.1828, + "step": 26523 + }, + { + "epoch": 2.23, + "learning_rate": 3.2350371873319953e-06, + "loss": 1.1949, + "step": 26524 + }, + { + "epoch": 2.23, + "learning_rate": 3.234367805670151e-06, + "loss": 1.1127, + "step": 26525 + }, + { + "epoch": 2.23, + "learning_rate": 3.233698479908124e-06, + "loss": 1.1275, + "step": 26526 + }, + { + "epoch": 2.23, + "learning_rate": 3.233029210051446e-06, + "loss": 1.112, + "step": 26527 + }, + { + "epoch": 2.23, + "learning_rate": 3.2323599961056495e-06, + "loss": 1.1312, + "step": 26528 + }, + { + "epoch": 2.23, + "learning_rate": 3.2316908380762545e-06, + "loss": 1.0373, + "step": 26529 + }, + { + "epoch": 2.23, + "learning_rate": 3.2310217359688022e-06, + "loss": 1.1858, + "step": 26530 + }, + { + "epoch": 2.23, + "learning_rate": 3.2303526897888115e-06, + "loss": 1.1637, + "step": 26531 + }, + { + "epoch": 2.23, + "learning_rate": 3.2296836995418135e-06, + "loss": 1.0807, + "step": 26532 + }, + { + "epoch": 2.23, + "learning_rate": 3.229014765233336e-06, + "loss": 1.1774, + "step": 26533 + }, + { + "epoch": 2.23, + "learning_rate": 3.2283458868689045e-06, + "loss": 1.1792, + "step": 26534 + }, + { + "epoch": 2.23, + "learning_rate": 3.2276770644540468e-06, + "loss": 1.0694, + "step": 26535 + }, + { + "epoch": 2.23, + "learning_rate": 3.2270082979942897e-06, + "loss": 1.1238, + "step": 26536 + }, + { + "epoch": 2.23, + "learning_rate": 3.2263395874951565e-06, + "loss": 1.2042, + "step": 26537 + }, + { + "epoch": 2.23, + "learning_rate": 3.225670932962174e-06, + "loss": 1.201, + "step": 26538 + }, + { + "epoch": 2.23, + "learning_rate": 3.2250023344008683e-06, + "loss": 1.0943, + "step": 26539 + }, + { + "epoch": 2.23, + "learning_rate": 3.2243337918167562e-06, + "loss": 1.0854, + "step": 26540 + }, + { + "epoch": 2.23, + "learning_rate": 3.2236653052153733e-06, + "loss": 1.1126, + "step": 26541 + }, + { + "epoch": 2.23, + "learning_rate": 3.2229968746022334e-06, + "loss": 1.109, + "step": 26542 + }, + { + "epoch": 2.23, + "learning_rate": 3.222328499982862e-06, + "loss": 1.067, + "step": 26543 + }, + { + "epoch": 2.23, + "learning_rate": 3.221660181362781e-06, + "loss": 1.1368, + "step": 26544 + }, + { + "epoch": 2.23, + "learning_rate": 3.2209919187475136e-06, + "loss": 1.1152, + "step": 26545 + }, + { + "epoch": 2.23, + "learning_rate": 3.22032371214258e-06, + "loss": 1.0724, + "step": 26546 + }, + { + "epoch": 2.24, + "learning_rate": 3.219655561553502e-06, + "loss": 1.0867, + "step": 26547 + }, + { + "epoch": 2.24, + "learning_rate": 3.2189874669858003e-06, + "loss": 1.2283, + "step": 26548 + }, + { + "epoch": 2.24, + "learning_rate": 3.2183194284449938e-06, + "loss": 1.1456, + "step": 26549 + }, + { + "epoch": 2.24, + "learning_rate": 3.217651445936606e-06, + "loss": 1.0933, + "step": 26550 + }, + { + "epoch": 2.24, + "learning_rate": 3.216983519466146e-06, + "loss": 0.9794, + "step": 26551 + }, + { + "epoch": 2.24, + "learning_rate": 3.2163156490391456e-06, + "loss": 1.0885, + "step": 26552 + }, + { + "epoch": 2.24, + "learning_rate": 3.215647834661112e-06, + "loss": 1.0979, + "step": 26553 + }, + { + "epoch": 2.24, + "learning_rate": 3.2149800763375683e-06, + "loss": 1.027, + "step": 26554 + }, + { + "epoch": 2.24, + "learning_rate": 3.214312374074031e-06, + "loss": 0.9766, + "step": 26555 + }, + { + "epoch": 2.24, + "learning_rate": 3.2136447278760165e-06, + "loss": 1.1174, + "step": 26556 + }, + { + "epoch": 2.24, + "learning_rate": 3.2129771377490405e-06, + "loss": 1.1775, + "step": 26557 + }, + { + "epoch": 2.24, + "learning_rate": 3.2123096036986234e-06, + "loss": 1.1037, + "step": 26558 + }, + { + "epoch": 2.24, + "learning_rate": 3.21164212573027e-06, + "loss": 1.1182, + "step": 26559 + }, + { + "epoch": 2.24, + "learning_rate": 3.2109747038495042e-06, + "loss": 1.0832, + "step": 26560 + }, + { + "epoch": 2.24, + "learning_rate": 3.2103073380618433e-06, + "loss": 1.1271, + "step": 26561 + }, + { + "epoch": 2.24, + "learning_rate": 3.209640028372788e-06, + "loss": 1.039, + "step": 26562 + }, + { + "epoch": 2.24, + "learning_rate": 3.2089727747878675e-06, + "loss": 1.1, + "step": 26563 + }, + { + "epoch": 2.24, + "learning_rate": 3.2083055773125836e-06, + "loss": 1.0689, + "step": 26564 + }, + { + "epoch": 2.24, + "learning_rate": 3.2076384359524527e-06, + "loss": 1.0502, + "step": 26565 + }, + { + "epoch": 2.24, + "learning_rate": 3.2069713507129874e-06, + "loss": 1.1119, + "step": 26566 + }, + { + "epoch": 2.24, + "learning_rate": 3.206304321599698e-06, + "loss": 1.0609, + "step": 26567 + }, + { + "epoch": 2.24, + "learning_rate": 3.205637348618097e-06, + "loss": 1.0796, + "step": 26568 + }, + { + "epoch": 2.24, + "learning_rate": 3.2049704317736975e-06, + "loss": 1.1705, + "step": 26569 + }, + { + "epoch": 2.24, + "learning_rate": 3.2043035710720007e-06, + "loss": 1.114, + "step": 26570 + }, + { + "epoch": 2.24, + "learning_rate": 3.2036367665185297e-06, + "loss": 1.1647, + "step": 26571 + }, + { + "epoch": 2.24, + "learning_rate": 3.2029700181187828e-06, + "loss": 1.0679, + "step": 26572 + }, + { + "epoch": 2.24, + "learning_rate": 3.202303325878271e-06, + "loss": 1.0946, + "step": 26573 + }, + { + "epoch": 2.24, + "learning_rate": 3.20163668980251e-06, + "loss": 1.198, + "step": 26574 + }, + { + "epoch": 2.24, + "learning_rate": 3.2009701098969992e-06, + "loss": 1.0568, + "step": 26575 + }, + { + "epoch": 2.24, + "learning_rate": 3.2003035861672494e-06, + "loss": 1.0475, + "step": 26576 + }, + { + "epoch": 2.24, + "learning_rate": 3.199637118618768e-06, + "loss": 1.1431, + "step": 26577 + }, + { + "epoch": 2.24, + "learning_rate": 3.1989707072570643e-06, + "loss": 1.0784, + "step": 26578 + }, + { + "epoch": 2.24, + "learning_rate": 3.1983043520876345e-06, + "loss": 1.1658, + "step": 26579 + }, + { + "epoch": 2.24, + "learning_rate": 3.1976380531159977e-06, + "loss": 1.1207, + "step": 26580 + }, + { + "epoch": 2.24, + "learning_rate": 3.1969718103476487e-06, + "loss": 1.1342, + "step": 26581 + }, + { + "epoch": 2.24, + "learning_rate": 3.1963056237880964e-06, + "loss": 1.1436, + "step": 26582 + }, + { + "epoch": 2.24, + "learning_rate": 3.1956394934428436e-06, + "loss": 1.2272, + "step": 26583 + }, + { + "epoch": 2.24, + "learning_rate": 3.1949734193173953e-06, + "loss": 1.0507, + "step": 26584 + }, + { + "epoch": 2.24, + "learning_rate": 3.194307401417255e-06, + "loss": 1.0934, + "step": 26585 + }, + { + "epoch": 2.24, + "learning_rate": 3.1936414397479253e-06, + "loss": 1.1547, + "step": 26586 + }, + { + "epoch": 2.24, + "learning_rate": 3.192975534314907e-06, + "loss": 1.1215, + "step": 26587 + }, + { + "epoch": 2.24, + "learning_rate": 3.1923096851237047e-06, + "loss": 1.1416, + "step": 26588 + }, + { + "epoch": 2.24, + "learning_rate": 3.191643892179821e-06, + "loss": 1.12, + "step": 26589 + }, + { + "epoch": 2.24, + "learning_rate": 3.1909781554887487e-06, + "loss": 1.0916, + "step": 26590 + }, + { + "epoch": 2.24, + "learning_rate": 3.190312475056e-06, + "loss": 1.1115, + "step": 26591 + }, + { + "epoch": 2.24, + "learning_rate": 3.189646850887066e-06, + "loss": 1.1585, + "step": 26592 + }, + { + "epoch": 2.24, + "learning_rate": 3.188981282987449e-06, + "loss": 1.0927, + "step": 26593 + }, + { + "epoch": 2.24, + "learning_rate": 3.188315771362648e-06, + "loss": 1.0827, + "step": 26594 + }, + { + "epoch": 2.24, + "learning_rate": 3.1876503160181627e-06, + "loss": 1.1078, + "step": 26595 + }, + { + "epoch": 2.24, + "learning_rate": 3.18698491695949e-06, + "loss": 1.1285, + "step": 26596 + }, + { + "epoch": 2.24, + "learning_rate": 3.1863195741921295e-06, + "loss": 0.9521, + "step": 26597 + }, + { + "epoch": 2.24, + "learning_rate": 3.185654287721577e-06, + "loss": 1.0998, + "step": 26598 + }, + { + "epoch": 2.24, + "learning_rate": 3.1849890575533294e-06, + "loss": 1.0932, + "step": 26599 + }, + { + "epoch": 2.24, + "learning_rate": 3.184323883692887e-06, + "loss": 1.1685, + "step": 26600 + }, + { + "epoch": 2.24, + "learning_rate": 3.1836587661457352e-06, + "loss": 1.1289, + "step": 26601 + }, + { + "epoch": 2.24, + "learning_rate": 3.182993704917382e-06, + "loss": 1.217, + "step": 26602 + }, + { + "epoch": 2.24, + "learning_rate": 3.182328700013314e-06, + "loss": 1.0407, + "step": 26603 + }, + { + "epoch": 2.24, + "learning_rate": 3.181663751439028e-06, + "loss": 1.1222, + "step": 26604 + }, + { + "epoch": 2.24, + "learning_rate": 3.1809988592000187e-06, + "loss": 1.0766, + "step": 26605 + }, + { + "epoch": 2.24, + "learning_rate": 3.180334023301779e-06, + "loss": 1.179, + "step": 26606 + }, + { + "epoch": 2.24, + "learning_rate": 3.179669243749802e-06, + "loss": 1.1186, + "step": 26607 + }, + { + "epoch": 2.24, + "learning_rate": 3.1790045205495836e-06, + "loss": 1.1436, + "step": 26608 + }, + { + "epoch": 2.24, + "learning_rate": 3.178339853706607e-06, + "loss": 1.1512, + "step": 26609 + }, + { + "epoch": 2.24, + "learning_rate": 3.1776752432263714e-06, + "loss": 1.149, + "step": 26610 + }, + { + "epoch": 2.24, + "learning_rate": 3.1770106891143714e-06, + "loss": 1.0712, + "step": 26611 + }, + { + "epoch": 2.24, + "learning_rate": 3.176346191376085e-06, + "loss": 1.1356, + "step": 26612 + }, + { + "epoch": 2.24, + "learning_rate": 3.1756817500170178e-06, + "loss": 1.102, + "step": 26613 + }, + { + "epoch": 2.24, + "learning_rate": 3.1750173650426476e-06, + "loss": 1.1488, + "step": 26614 + }, + { + "epoch": 2.24, + "learning_rate": 3.1743530364584697e-06, + "loss": 1.1466, + "step": 26615 + }, + { + "epoch": 2.24, + "learning_rate": 3.17368876426997e-06, + "loss": 1.1135, + "step": 26616 + }, + { + "epoch": 2.24, + "learning_rate": 3.1730245484826394e-06, + "loss": 1.0566, + "step": 26617 + }, + { + "epoch": 2.24, + "learning_rate": 3.172360389101964e-06, + "loss": 1.0176, + "step": 26618 + }, + { + "epoch": 2.24, + "learning_rate": 3.1716962861334364e-06, + "loss": 1.1941, + "step": 26619 + }, + { + "epoch": 2.24, + "learning_rate": 3.171032239582532e-06, + "loss": 1.2347, + "step": 26620 + }, + { + "epoch": 2.24, + "learning_rate": 3.1703682494547517e-06, + "loss": 1.2029, + "step": 26621 + }, + { + "epoch": 2.24, + "learning_rate": 3.1697043157555716e-06, + "loss": 1.1111, + "step": 26622 + }, + { + "epoch": 2.24, + "learning_rate": 3.169040438490477e-06, + "loss": 1.2114, + "step": 26623 + }, + { + "epoch": 2.24, + "learning_rate": 3.1683766176649634e-06, + "loss": 1.0628, + "step": 26624 + }, + { + "epoch": 2.24, + "learning_rate": 3.167712853284505e-06, + "loss": 1.0928, + "step": 26625 + }, + { + "epoch": 2.24, + "learning_rate": 3.1670491453545894e-06, + "loss": 1.0681, + "step": 26626 + }, + { + "epoch": 2.24, + "learning_rate": 3.1663854938807013e-06, + "loss": 1.2066, + "step": 26627 + }, + { + "epoch": 2.24, + "learning_rate": 3.1657218988683257e-06, + "loss": 1.1456, + "step": 26628 + }, + { + "epoch": 2.24, + "learning_rate": 3.165058360322937e-06, + "loss": 1.174, + "step": 26629 + }, + { + "epoch": 2.24, + "learning_rate": 3.16439487825003e-06, + "loss": 1.1232, + "step": 26630 + }, + { + "epoch": 2.24, + "learning_rate": 3.1637314526550767e-06, + "loss": 1.1532, + "step": 26631 + }, + { + "epoch": 2.24, + "learning_rate": 3.1630680835435624e-06, + "loss": 1.1582, + "step": 26632 + }, + { + "epoch": 2.24, + "learning_rate": 3.162404770920967e-06, + "loss": 1.1252, + "step": 26633 + }, + { + "epoch": 2.24, + "learning_rate": 3.161741514792773e-06, + "loss": 1.116, + "step": 26634 + }, + { + "epoch": 2.24, + "learning_rate": 3.1610783151644577e-06, + "loss": 1.1304, + "step": 26635 + }, + { + "epoch": 2.24, + "learning_rate": 3.160415172041503e-06, + "loss": 1.1884, + "step": 26636 + }, + { + "epoch": 2.24, + "learning_rate": 3.159752085429387e-06, + "loss": 1.1177, + "step": 26637 + }, + { + "epoch": 2.24, + "learning_rate": 3.1590890553335884e-06, + "loss": 1.0851, + "step": 26638 + }, + { + "epoch": 2.24, + "learning_rate": 3.158426081759588e-06, + "loss": 1.0973, + "step": 26639 + }, + { + "epoch": 2.24, + "learning_rate": 3.1577631647128547e-06, + "loss": 1.2043, + "step": 26640 + }, + { + "epoch": 2.24, + "learning_rate": 3.157100304198879e-06, + "loss": 1.1083, + "step": 26641 + }, + { + "epoch": 2.24, + "learning_rate": 3.156437500223127e-06, + "loss": 1.1074, + "step": 26642 + }, + { + "epoch": 2.24, + "learning_rate": 3.1557747527910786e-06, + "loss": 1.0192, + "step": 26643 + }, + { + "epoch": 2.24, + "learning_rate": 3.15511206190821e-06, + "loss": 1.1538, + "step": 26644 + }, + { + "epoch": 2.24, + "learning_rate": 3.1544494275799953e-06, + "loss": 1.1289, + "step": 26645 + }, + { + "epoch": 2.24, + "learning_rate": 3.1537868498119105e-06, + "loss": 1.1439, + "step": 26646 + }, + { + "epoch": 2.24, + "learning_rate": 3.1531243286094306e-06, + "loss": 1.1679, + "step": 26647 + }, + { + "epoch": 2.24, + "learning_rate": 3.152461863978028e-06, + "loss": 1.0876, + "step": 26648 + }, + { + "epoch": 2.24, + "learning_rate": 3.151799455923177e-06, + "loss": 1.1674, + "step": 26649 + }, + { + "epoch": 2.24, + "learning_rate": 3.151137104450355e-06, + "loss": 1.1265, + "step": 26650 + }, + { + "epoch": 2.24, + "learning_rate": 3.1504748095650215e-06, + "loss": 1.1201, + "step": 26651 + }, + { + "epoch": 2.24, + "learning_rate": 3.1498125712726657e-06, + "loss": 1.1236, + "step": 26652 + }, + { + "epoch": 2.24, + "learning_rate": 3.149150389578747e-06, + "loss": 1.0921, + "step": 26653 + }, + { + "epoch": 2.24, + "learning_rate": 3.1484882644887403e-06, + "loss": 1.0954, + "step": 26654 + }, + { + "epoch": 2.24, + "learning_rate": 3.1478261960081157e-06, + "loss": 1.154, + "step": 26655 + }, + { + "epoch": 2.24, + "learning_rate": 3.147164184142345e-06, + "loss": 1.062, + "step": 26656 + }, + { + "epoch": 2.24, + "learning_rate": 3.146502228896896e-06, + "loss": 1.182, + "step": 26657 + }, + { + "epoch": 2.24, + "learning_rate": 3.145840330277243e-06, + "loss": 1.1497, + "step": 26658 + }, + { + "epoch": 2.24, + "learning_rate": 3.1451784882888436e-06, + "loss": 1.1665, + "step": 26659 + }, + { + "epoch": 2.24, + "learning_rate": 3.144516702937177e-06, + "loss": 1.0936, + "step": 26660 + }, + { + "epoch": 2.24, + "learning_rate": 3.1438549742277113e-06, + "loss": 1.1761, + "step": 26661 + }, + { + "epoch": 2.24, + "learning_rate": 3.143193302165903e-06, + "loss": 1.0923, + "step": 26662 + }, + { + "epoch": 2.24, + "learning_rate": 3.142531686757233e-06, + "loss": 1.1001, + "step": 26663 + }, + { + "epoch": 2.24, + "learning_rate": 3.1418701280071574e-06, + "loss": 1.1003, + "step": 26664 + }, + { + "epoch": 2.24, + "learning_rate": 3.141208625921145e-06, + "loss": 1.1564, + "step": 26665 + }, + { + "epoch": 2.25, + "learning_rate": 3.140547180504663e-06, + "loss": 1.1602, + "step": 26666 + }, + { + "epoch": 2.25, + "learning_rate": 3.139885791763174e-06, + "loss": 1.0228, + "step": 26667 + }, + { + "epoch": 2.25, + "learning_rate": 3.139224459702146e-06, + "loss": 1.1345, + "step": 26668 + }, + { + "epoch": 2.25, + "learning_rate": 3.138563184327044e-06, + "loss": 1.0952, + "step": 26669 + }, + { + "epoch": 2.25, + "learning_rate": 3.137901965643322e-06, + "loss": 1.2151, + "step": 26670 + }, + { + "epoch": 2.25, + "learning_rate": 3.137240803656456e-06, + "loss": 1.0505, + "step": 26671 + }, + { + "epoch": 2.25, + "learning_rate": 3.1365796983719e-06, + "loss": 1.0617, + "step": 26672 + }, + { + "epoch": 2.25, + "learning_rate": 3.1359186497951167e-06, + "loss": 1.0852, + "step": 26673 + }, + { + "epoch": 2.25, + "learning_rate": 3.1352576579315765e-06, + "loss": 1.0836, + "step": 26674 + }, + { + "epoch": 2.25, + "learning_rate": 3.1345967227867303e-06, + "loss": 1.1452, + "step": 26675 + }, + { + "epoch": 2.25, + "learning_rate": 3.1339358443660438e-06, + "loss": 1.2333, + "step": 26676 + }, + { + "epoch": 2.25, + "learning_rate": 3.1332750226749776e-06, + "loss": 1.1584, + "step": 26677 + }, + { + "epoch": 2.25, + "learning_rate": 3.13261425771899e-06, + "loss": 1.087, + "step": 26678 + }, + { + "epoch": 2.25, + "learning_rate": 3.13195354950354e-06, + "loss": 1.1077, + "step": 26679 + }, + { + "epoch": 2.25, + "learning_rate": 3.131292898034093e-06, + "loss": 1.1435, + "step": 26680 + }, + { + "epoch": 2.25, + "learning_rate": 3.130632303316096e-06, + "loss": 1.0748, + "step": 26681 + }, + { + "epoch": 2.25, + "learning_rate": 3.1299717653550187e-06, + "loss": 1.1065, + "step": 26682 + }, + { + "epoch": 2.25, + "learning_rate": 3.12931128415631e-06, + "loss": 1.1165, + "step": 26683 + }, + { + "epoch": 2.25, + "learning_rate": 3.1286508597254316e-06, + "loss": 1.0825, + "step": 26684 + }, + { + "epoch": 2.25, + "learning_rate": 3.127990492067838e-06, + "loss": 1.0417, + "step": 26685 + }, + { + "epoch": 2.25, + "learning_rate": 3.127330181188987e-06, + "loss": 1.1318, + "step": 26686 + }, + { + "epoch": 2.25, + "learning_rate": 3.1266699270943323e-06, + "loss": 1.2166, + "step": 26687 + }, + { + "epoch": 2.25, + "learning_rate": 3.126009729789331e-06, + "loss": 1.1719, + "step": 26688 + }, + { + "epoch": 2.25, + "learning_rate": 3.1253495892794406e-06, + "loss": 1.1767, + "step": 26689 + }, + { + "epoch": 2.25, + "learning_rate": 3.1246895055701064e-06, + "loss": 1.0769, + "step": 26690 + }, + { + "epoch": 2.25, + "learning_rate": 3.1240294786667937e-06, + "loss": 1.0797, + "step": 26691 + }, + { + "epoch": 2.25, + "learning_rate": 3.123369508574946e-06, + "loss": 1.0627, + "step": 26692 + }, + { + "epoch": 2.25, + "learning_rate": 3.122709595300022e-06, + "loss": 1.1739, + "step": 26693 + }, + { + "epoch": 2.25, + "learning_rate": 3.12204973884747e-06, + "loss": 1.09, + "step": 26694 + }, + { + "epoch": 2.25, + "learning_rate": 3.1213899392227463e-06, + "loss": 1.1573, + "step": 26695 + }, + { + "epoch": 2.25, + "learning_rate": 3.1207301964312997e-06, + "loss": 1.1643, + "step": 26696 + }, + { + "epoch": 2.25, + "learning_rate": 3.120070510478582e-06, + "loss": 1.1336, + "step": 26697 + }, + { + "epoch": 2.25, + "learning_rate": 3.119410881370043e-06, + "loss": 1.0687, + "step": 26698 + }, + { + "epoch": 2.25, + "learning_rate": 3.1187513091111345e-06, + "loss": 1.1269, + "step": 26699 + }, + { + "epoch": 2.25, + "learning_rate": 3.1180917937073073e-06, + "loss": 1.1631, + "step": 26700 + }, + { + "epoch": 2.25, + "learning_rate": 3.1174323351640013e-06, + "loss": 1.1909, + "step": 26701 + }, + { + "epoch": 2.25, + "learning_rate": 3.1167729334866792e-06, + "loss": 1.2092, + "step": 26702 + }, + { + "epoch": 2.25, + "learning_rate": 3.1161135886807793e-06, + "loss": 1.1425, + "step": 26703 + }, + { + "epoch": 2.25, + "learning_rate": 3.1154543007517514e-06, + "loss": 1.0912, + "step": 26704 + }, + { + "epoch": 2.25, + "learning_rate": 3.114795069705043e-06, + "loss": 1.1532, + "step": 26705 + }, + { + "epoch": 2.25, + "learning_rate": 3.114135895546102e-06, + "loss": 1.1941, + "step": 26706 + }, + { + "epoch": 2.25, + "learning_rate": 3.1134767782803744e-06, + "loss": 1.14, + "step": 26707 + }, + { + "epoch": 2.25, + "learning_rate": 3.112817717913308e-06, + "loss": 1.1107, + "step": 26708 + }, + { + "epoch": 2.25, + "learning_rate": 3.1121587144503394e-06, + "loss": 1.0371, + "step": 26709 + }, + { + "epoch": 2.25, + "learning_rate": 3.1114997678969226e-06, + "loss": 1.0848, + "step": 26710 + }, + { + "epoch": 2.25, + "learning_rate": 3.110840878258504e-06, + "loss": 1.2064, + "step": 26711 + }, + { + "epoch": 2.25, + "learning_rate": 3.1101820455405153e-06, + "loss": 1.1561, + "step": 26712 + }, + { + "epoch": 2.25, + "learning_rate": 3.1095232697484147e-06, + "loss": 1.1434, + "step": 26713 + }, + { + "epoch": 2.25, + "learning_rate": 3.1088645508876346e-06, + "loss": 1.0878, + "step": 26714 + }, + { + "epoch": 2.25, + "learning_rate": 3.10820588896362e-06, + "loss": 1.0987, + "step": 26715 + }, + { + "epoch": 2.25, + "learning_rate": 3.1075472839818154e-06, + "loss": 1.0138, + "step": 26716 + }, + { + "epoch": 2.25, + "learning_rate": 3.1068887359476606e-06, + "loss": 1.1282, + "step": 26717 + }, + { + "epoch": 2.25, + "learning_rate": 3.106230244866597e-06, + "loss": 1.0066, + "step": 26718 + }, + { + "epoch": 2.25, + "learning_rate": 3.1055718107440693e-06, + "loss": 1.0838, + "step": 26719 + }, + { + "epoch": 2.25, + "learning_rate": 3.104913433585507e-06, + "loss": 1.0904, + "step": 26720 + }, + { + "epoch": 2.25, + "learning_rate": 3.104255113396363e-06, + "loss": 1.1732, + "step": 26721 + }, + { + "epoch": 2.25, + "learning_rate": 3.1035968501820655e-06, + "loss": 1.053, + "step": 26722 + }, + { + "epoch": 2.25, + "learning_rate": 3.102938643948057e-06, + "loss": 1.0881, + "step": 26723 + }, + { + "epoch": 2.25, + "learning_rate": 3.102280494699782e-06, + "loss": 0.9969, + "step": 26724 + }, + { + "epoch": 2.25, + "learning_rate": 3.10162240244267e-06, + "loss": 1.0921, + "step": 26725 + }, + { + "epoch": 2.25, + "learning_rate": 3.100964367182162e-06, + "loss": 1.1035, + "step": 26726 + }, + { + "epoch": 2.25, + "learning_rate": 3.1003063889236937e-06, + "loss": 1.0581, + "step": 26727 + }, + { + "epoch": 2.25, + "learning_rate": 3.0996484676727022e-06, + "loss": 1.1121, + "step": 26728 + }, + { + "epoch": 2.25, + "learning_rate": 3.098990603434624e-06, + "loss": 1.0852, + "step": 26729 + }, + { + "epoch": 2.25, + "learning_rate": 3.098332796214897e-06, + "loss": 1.1081, + "step": 26730 + }, + { + "epoch": 2.25, + "learning_rate": 3.0976750460189463e-06, + "loss": 1.213, + "step": 26731 + }, + { + "epoch": 2.25, + "learning_rate": 3.0970173528522206e-06, + "loss": 0.981, + "step": 26732 + }, + { + "epoch": 2.25, + "learning_rate": 3.0963597167201433e-06, + "loss": 1.1176, + "step": 26733 + }, + { + "epoch": 2.25, + "learning_rate": 3.0957021376281515e-06, + "loss": 1.072, + "step": 26734 + }, + { + "epoch": 2.25, + "learning_rate": 3.095044615581678e-06, + "loss": 1.1127, + "step": 26735 + }, + { + "epoch": 2.25, + "learning_rate": 3.0943871505861567e-06, + "loss": 1.0522, + "step": 26736 + }, + { + "epoch": 2.25, + "learning_rate": 3.093729742647018e-06, + "loss": 1.0885, + "step": 26737 + }, + { + "epoch": 2.25, + "learning_rate": 3.0930723917696946e-06, + "loss": 1.0563, + "step": 26738 + }, + { + "epoch": 2.25, + "learning_rate": 3.0924150979596213e-06, + "loss": 1.0977, + "step": 26739 + }, + { + "epoch": 2.25, + "learning_rate": 3.0917578612222175e-06, + "loss": 1.1064, + "step": 26740 + }, + { + "epoch": 2.25, + "learning_rate": 3.0911006815629285e-06, + "loss": 1.1835, + "step": 26741 + }, + { + "epoch": 2.25, + "learning_rate": 3.090443558987174e-06, + "loss": 1.0657, + "step": 26742 + }, + { + "epoch": 2.25, + "learning_rate": 3.089786493500385e-06, + "loss": 1.1712, + "step": 26743 + }, + { + "epoch": 2.25, + "learning_rate": 3.0891294851079924e-06, + "loss": 1.1656, + "step": 26744 + }, + { + "epoch": 2.25, + "learning_rate": 3.0884725338154233e-06, + "loss": 1.1157, + "step": 26745 + }, + { + "epoch": 2.25, + "learning_rate": 3.0878156396281067e-06, + "loss": 1.1049, + "step": 26746 + }, + { + "epoch": 2.25, + "learning_rate": 3.0871588025514685e-06, + "loss": 1.0948, + "step": 26747 + }, + { + "epoch": 2.25, + "learning_rate": 3.086502022590938e-06, + "loss": 1.1472, + "step": 26748 + }, + { + "epoch": 2.25, + "learning_rate": 3.0858452997519394e-06, + "loss": 1.0131, + "step": 26749 + }, + { + "epoch": 2.25, + "learning_rate": 3.085188634039903e-06, + "loss": 1.1443, + "step": 26750 + }, + { + "epoch": 2.25, + "learning_rate": 3.084532025460245e-06, + "loss": 1.1764, + "step": 26751 + }, + { + "epoch": 2.25, + "learning_rate": 3.083875474018403e-06, + "loss": 1.1307, + "step": 26752 + }, + { + "epoch": 2.25, + "learning_rate": 3.083218979719792e-06, + "loss": 1.1542, + "step": 26753 + }, + { + "epoch": 2.25, + "learning_rate": 3.082562542569839e-06, + "loss": 1.2371, + "step": 26754 + }, + { + "epoch": 2.25, + "learning_rate": 3.081906162573969e-06, + "loss": 1.018, + "step": 26755 + }, + { + "epoch": 2.25, + "learning_rate": 3.0812498397376046e-06, + "loss": 1.0993, + "step": 26756 + }, + { + "epoch": 2.25, + "learning_rate": 3.0805935740661676e-06, + "loss": 1.0234, + "step": 26757 + }, + { + "epoch": 2.25, + "learning_rate": 3.0799373655650835e-06, + "loss": 1.1291, + "step": 26758 + }, + { + "epoch": 2.25, + "learning_rate": 3.0792812142397664e-06, + "loss": 1.0999, + "step": 26759 + }, + { + "epoch": 2.25, + "learning_rate": 3.0786251200956454e-06, + "loss": 1.1117, + "step": 26760 + }, + { + "epoch": 2.25, + "learning_rate": 3.0779690831381424e-06, + "loss": 1.1458, + "step": 26761 + }, + { + "epoch": 2.25, + "learning_rate": 3.0773131033726665e-06, + "loss": 1.1955, + "step": 26762 + }, + { + "epoch": 2.25, + "learning_rate": 3.076657180804653e-06, + "loss": 1.0567, + "step": 26763 + }, + { + "epoch": 2.25, + "learning_rate": 3.0760013154395106e-06, + "loss": 1.0282, + "step": 26764 + }, + { + "epoch": 2.25, + "learning_rate": 3.075345507282661e-06, + "loss": 1.0724, + "step": 26765 + }, + { + "epoch": 2.25, + "learning_rate": 3.0746897563395226e-06, + "loss": 1.133, + "step": 26766 + }, + { + "epoch": 2.25, + "learning_rate": 3.074034062615514e-06, + "loss": 1.0819, + "step": 26767 + }, + { + "epoch": 2.25, + "learning_rate": 3.073378426116053e-06, + "loss": 1.1628, + "step": 26768 + }, + { + "epoch": 2.25, + "learning_rate": 3.0727228468465585e-06, + "loss": 1.1709, + "step": 26769 + }, + { + "epoch": 2.25, + "learning_rate": 3.0720673248124387e-06, + "loss": 1.1858, + "step": 26770 + }, + { + "epoch": 2.25, + "learning_rate": 3.0714118600191232e-06, + "loss": 1.0773, + "step": 26771 + }, + { + "epoch": 2.25, + "learning_rate": 3.0707564524720166e-06, + "loss": 1.1488, + "step": 26772 + }, + { + "epoch": 2.25, + "learning_rate": 3.070101102176535e-06, + "loss": 1.2889, + "step": 26773 + }, + { + "epoch": 2.25, + "learning_rate": 3.069445809138101e-06, + "loss": 1.1301, + "step": 26774 + }, + { + "epoch": 2.25, + "learning_rate": 3.0687905733621216e-06, + "loss": 1.0375, + "step": 26775 + }, + { + "epoch": 2.25, + "learning_rate": 3.0681353948540127e-06, + "loss": 1.1899, + "step": 26776 + }, + { + "epoch": 2.25, + "learning_rate": 3.0674802736191876e-06, + "loss": 1.1892, + "step": 26777 + }, + { + "epoch": 2.25, + "learning_rate": 3.066825209663059e-06, + "loss": 1.1287, + "step": 26778 + }, + { + "epoch": 2.25, + "learning_rate": 3.06617020299104e-06, + "loss": 1.1467, + "step": 26779 + }, + { + "epoch": 2.25, + "learning_rate": 3.0655152536085443e-06, + "loss": 1.0968, + "step": 26780 + }, + { + "epoch": 2.25, + "learning_rate": 3.064860361520975e-06, + "loss": 1.178, + "step": 26781 + }, + { + "epoch": 2.25, + "learning_rate": 3.0642055267337556e-06, + "loss": 1.1196, + "step": 26782 + }, + { + "epoch": 2.25, + "learning_rate": 3.0635507492522864e-06, + "loss": 1.1333, + "step": 26783 + }, + { + "epoch": 2.25, + "learning_rate": 3.06289602908198e-06, + "loss": 1.1349, + "step": 26784 + }, + { + "epoch": 2.26, + "learning_rate": 3.062241366228248e-06, + "loss": 1.1186, + "step": 26785 + }, + { + "epoch": 2.26, + "learning_rate": 3.0615867606964977e-06, + "loss": 1.1082, + "step": 26786 + }, + { + "epoch": 2.26, + "learning_rate": 3.060932212492138e-06, + "loss": 1.0368, + "step": 26787 + }, + { + "epoch": 2.26, + "learning_rate": 3.060277721620577e-06, + "loss": 1.11, + "step": 26788 + }, + { + "epoch": 2.26, + "learning_rate": 3.059623288087226e-06, + "loss": 1.046, + "step": 26789 + }, + { + "epoch": 2.26, + "learning_rate": 3.058968911897482e-06, + "loss": 1.0249, + "step": 26790 + }, + { + "epoch": 2.26, + "learning_rate": 3.0583145930567647e-06, + "loss": 1.0853, + "step": 26791 + }, + { + "epoch": 2.26, + "learning_rate": 3.0576603315704714e-06, + "loss": 1.1852, + "step": 26792 + }, + { + "epoch": 2.26, + "learning_rate": 3.05700612744401e-06, + "loss": 1.0955, + "step": 26793 + }, + { + "epoch": 2.26, + "learning_rate": 3.0563519806827847e-06, + "loss": 1.1021, + "step": 26794 + }, + { + "epoch": 2.26, + "learning_rate": 3.055697891292203e-06, + "loss": 1.1427, + "step": 26795 + }, + { + "epoch": 2.26, + "learning_rate": 3.0550438592776675e-06, + "loss": 1.0861, + "step": 26796 + }, + { + "epoch": 2.26, + "learning_rate": 3.054389884644582e-06, + "loss": 1.1491, + "step": 26797 + }, + { + "epoch": 2.26, + "learning_rate": 3.05373596739835e-06, + "loss": 1.1253, + "step": 26798 + }, + { + "epoch": 2.26, + "learning_rate": 3.053082107544375e-06, + "loss": 1.1496, + "step": 26799 + }, + { + "epoch": 2.26, + "learning_rate": 3.052428305088062e-06, + "loss": 1.0849, + "step": 26800 + }, + { + "epoch": 2.26, + "learning_rate": 3.0517745600348035e-06, + "loss": 1.1034, + "step": 26801 + }, + { + "epoch": 2.26, + "learning_rate": 3.051120872390013e-06, + "loss": 1.0274, + "step": 26802 + }, + { + "epoch": 2.26, + "learning_rate": 3.050467242159084e-06, + "loss": 1.1163, + "step": 26803 + }, + { + "epoch": 2.26, + "learning_rate": 3.0498136693474177e-06, + "loss": 1.1403, + "step": 26804 + }, + { + "epoch": 2.26, + "learning_rate": 3.0491601539604152e-06, + "loss": 1.0178, + "step": 26805 + }, + { + "epoch": 2.26, + "learning_rate": 3.048506696003477e-06, + "loss": 1.197, + "step": 26806 + }, + { + "epoch": 2.26, + "learning_rate": 3.0478532954820005e-06, + "loss": 1.0721, + "step": 26807 + }, + { + "epoch": 2.26, + "learning_rate": 3.0471999524013886e-06, + "loss": 1.0333, + "step": 26808 + }, + { + "epoch": 2.26, + "learning_rate": 3.0465466667670286e-06, + "loss": 1.0832, + "step": 26809 + }, + { + "epoch": 2.26, + "learning_rate": 3.0458934385843295e-06, + "loss": 1.2031, + "step": 26810 + }, + { + "epoch": 2.26, + "learning_rate": 3.0452402678586866e-06, + "loss": 1.0845, + "step": 26811 + }, + { + "epoch": 2.26, + "learning_rate": 3.04458715459549e-06, + "loss": 1.1508, + "step": 26812 + }, + { + "epoch": 2.26, + "learning_rate": 3.043934098800144e-06, + "loss": 1.0365, + "step": 26813 + }, + { + "epoch": 2.26, + "learning_rate": 3.0432811004780393e-06, + "loss": 1.0956, + "step": 26814 + }, + { + "epoch": 2.26, + "learning_rate": 3.042628159634573e-06, + "loss": 1.1191, + "step": 26815 + }, + { + "epoch": 2.26, + "learning_rate": 3.0419752762751387e-06, + "loss": 1.0148, + "step": 26816 + }, + { + "epoch": 2.26, + "learning_rate": 3.0413224504051307e-06, + "loss": 1.0879, + "step": 26817 + }, + { + "epoch": 2.26, + "learning_rate": 3.0406696820299455e-06, + "loss": 1.052, + "step": 26818 + }, + { + "epoch": 2.26, + "learning_rate": 3.0400169711549765e-06, + "loss": 1.1164, + "step": 26819 + }, + { + "epoch": 2.26, + "learning_rate": 3.039364317785609e-06, + "loss": 1.1013, + "step": 26820 + }, + { + "epoch": 2.26, + "learning_rate": 3.0387117219272477e-06, + "loss": 1.2129, + "step": 26821 + }, + { + "epoch": 2.26, + "learning_rate": 3.038059183585275e-06, + "loss": 1.1238, + "step": 26822 + }, + { + "epoch": 2.26, + "learning_rate": 3.037406702765081e-06, + "loss": 1.1324, + "step": 26823 + }, + { + "epoch": 2.26, + "learning_rate": 3.0367542794720685e-06, + "loss": 1.0743, + "step": 26824 + }, + { + "epoch": 2.26, + "learning_rate": 3.0361019137116166e-06, + "loss": 1.1778, + "step": 26825 + }, + { + "epoch": 2.26, + "learning_rate": 3.03544960548912e-06, + "loss": 1.1498, + "step": 26826 + }, + { + "epoch": 2.26, + "learning_rate": 3.0347973548099686e-06, + "loss": 1.0596, + "step": 26827 + }, + { + "epoch": 2.26, + "learning_rate": 3.0341451616795492e-06, + "loss": 1.1429, + "step": 26828 + }, + { + "epoch": 2.26, + "learning_rate": 3.0334930261032526e-06, + "loss": 1.0886, + "step": 26829 + }, + { + "epoch": 2.26, + "learning_rate": 3.0328409480864697e-06, + "loss": 1.167, + "step": 26830 + }, + { + "epoch": 2.26, + "learning_rate": 3.0321889276345773e-06, + "loss": 1.0932, + "step": 26831 + }, + { + "epoch": 2.26, + "learning_rate": 3.031536964752977e-06, + "loss": 1.0601, + "step": 26832 + }, + { + "epoch": 2.26, + "learning_rate": 3.030885059447045e-06, + "loss": 1.0919, + "step": 26833 + }, + { + "epoch": 2.26, + "learning_rate": 3.030233211722171e-06, + "loss": 1.1416, + "step": 26834 + }, + { + "epoch": 2.26, + "learning_rate": 3.0295814215837417e-06, + "loss": 1.1057, + "step": 26835 + }, + { + "epoch": 2.26, + "learning_rate": 3.0289296890371402e-06, + "loss": 1.1409, + "step": 26836 + }, + { + "epoch": 2.26, + "learning_rate": 3.0282780140877534e-06, + "loss": 1.1165, + "step": 26837 + }, + { + "epoch": 2.26, + "learning_rate": 3.0276263967409646e-06, + "loss": 1.121, + "step": 26838 + }, + { + "epoch": 2.26, + "learning_rate": 3.0269748370021613e-06, + "loss": 1.1811, + "step": 26839 + }, + { + "epoch": 2.26, + "learning_rate": 3.0263233348767175e-06, + "loss": 1.1305, + "step": 26840 + }, + { + "epoch": 2.26, + "learning_rate": 3.0256718903700276e-06, + "loss": 1.1244, + "step": 26841 + }, + { + "epoch": 2.26, + "learning_rate": 3.025020503487467e-06, + "loss": 1.1036, + "step": 26842 + }, + { + "epoch": 2.26, + "learning_rate": 3.024369174234417e-06, + "loss": 1.1943, + "step": 26843 + }, + { + "epoch": 2.26, + "learning_rate": 3.0237179026162634e-06, + "loss": 1.108, + "step": 26844 + }, + { + "epoch": 2.26, + "learning_rate": 3.0230666886383844e-06, + "loss": 1.0733, + "step": 26845 + }, + { + "epoch": 2.26, + "learning_rate": 3.022415532306161e-06, + "loss": 1.1626, + "step": 26846 + }, + { + "epoch": 2.26, + "learning_rate": 3.021764433624974e-06, + "loss": 1.0823, + "step": 26847 + }, + { + "epoch": 2.26, + "learning_rate": 3.0211133926002034e-06, + "loss": 1.0966, + "step": 26848 + }, + { + "epoch": 2.26, + "learning_rate": 3.020462409237226e-06, + "loss": 1.0802, + "step": 26849 + }, + { + "epoch": 2.26, + "learning_rate": 3.019811483541426e-06, + "loss": 1.1893, + "step": 26850 + }, + { + "epoch": 2.26, + "learning_rate": 3.019160615518171e-06, + "loss": 1.1695, + "step": 26851 + }, + { + "epoch": 2.26, + "learning_rate": 3.0185098051728513e-06, + "loss": 1.1435, + "step": 26852 + }, + { + "epoch": 2.26, + "learning_rate": 3.017859052510834e-06, + "loss": 1.1145, + "step": 26853 + }, + { + "epoch": 2.26, + "learning_rate": 3.017208357537501e-06, + "loss": 1.1668, + "step": 26854 + }, + { + "epoch": 2.26, + "learning_rate": 3.0165577202582273e-06, + "loss": 1.1186, + "step": 26855 + }, + { + "epoch": 2.26, + "learning_rate": 3.015907140678388e-06, + "loss": 1.0783, + "step": 26856 + }, + { + "epoch": 2.26, + "learning_rate": 3.0152566188033592e-06, + "loss": 1.0868, + "step": 26857 + }, + { + "epoch": 2.26, + "learning_rate": 3.0146061546385197e-06, + "loss": 1.1706, + "step": 26858 + }, + { + "epoch": 2.26, + "learning_rate": 3.0139557481892325e-06, + "loss": 1.1252, + "step": 26859 + }, + { + "epoch": 2.26, + "learning_rate": 3.013305399460882e-06, + "loss": 1.1242, + "step": 26860 + }, + { + "epoch": 2.26, + "learning_rate": 3.0126551084588417e-06, + "loss": 1.0476, + "step": 26861 + }, + { + "epoch": 2.26, + "learning_rate": 3.0120048751884747e-06, + "loss": 1.1283, + "step": 26862 + }, + { + "epoch": 2.26, + "learning_rate": 3.0113546996551658e-06, + "loss": 1.1377, + "step": 26863 + }, + { + "epoch": 2.26, + "learning_rate": 3.0107045818642788e-06, + "loss": 1.1515, + "step": 26864 + }, + { + "epoch": 2.26, + "learning_rate": 3.0100545218211865e-06, + "loss": 1.0197, + "step": 26865 + }, + { + "epoch": 2.26, + "learning_rate": 3.0094045195312615e-06, + "loss": 1.1451, + "step": 26866 + }, + { + "epoch": 2.26, + "learning_rate": 3.0087545749998725e-06, + "loss": 1.1047, + "step": 26867 + }, + { + "epoch": 2.26, + "learning_rate": 3.0081046882323917e-06, + "loss": 1.0924, + "step": 26868 + }, + { + "epoch": 2.26, + "learning_rate": 3.0074548592341913e-06, + "loss": 1.0916, + "step": 26869 + }, + { + "epoch": 2.26, + "learning_rate": 3.0068050880106294e-06, + "loss": 1.1001, + "step": 26870 + }, + { + "epoch": 2.26, + "learning_rate": 3.0061553745670892e-06, + "loss": 1.1312, + "step": 26871 + }, + { + "epoch": 2.26, + "learning_rate": 3.005505718908928e-06, + "loss": 1.1342, + "step": 26872 + }, + { + "epoch": 2.26, + "learning_rate": 3.004856121041513e-06, + "loss": 1.1002, + "step": 26873 + }, + { + "epoch": 2.26, + "learning_rate": 3.0042065809702226e-06, + "loss": 1.0879, + "step": 26874 + }, + { + "epoch": 2.26, + "learning_rate": 3.0035570987004136e-06, + "loss": 1.1179, + "step": 26875 + }, + { + "epoch": 2.26, + "learning_rate": 3.002907674237454e-06, + "loss": 1.1618, + "step": 26876 + }, + { + "epoch": 2.26, + "learning_rate": 3.002258307586712e-06, + "loss": 1.0724, + "step": 26877 + }, + { + "epoch": 2.26, + "learning_rate": 3.00160899875355e-06, + "loss": 1.145, + "step": 26878 + }, + { + "epoch": 2.26, + "learning_rate": 3.000959747743335e-06, + "loss": 0.9961, + "step": 26879 + }, + { + "epoch": 2.26, + "learning_rate": 3.000310554561433e-06, + "loss": 1.0623, + "step": 26880 + }, + { + "epoch": 2.26, + "learning_rate": 2.9996614192132e-06, + "loss": 1.1116, + "step": 26881 + }, + { + "epoch": 2.26, + "learning_rate": 2.9990123417040108e-06, + "loss": 1.1834, + "step": 26882 + }, + { + "epoch": 2.26, + "learning_rate": 2.998363322039218e-06, + "loss": 1.1592, + "step": 26883 + }, + { + "epoch": 2.26, + "learning_rate": 2.99771436022419e-06, + "loss": 1.0647, + "step": 26884 + }, + { + "epoch": 2.26, + "learning_rate": 2.9970654562642863e-06, + "loss": 0.9911, + "step": 26885 + }, + { + "epoch": 2.26, + "learning_rate": 2.9964166101648683e-06, + "loss": 1.1488, + "step": 26886 + }, + { + "epoch": 2.26, + "learning_rate": 2.995767821931299e-06, + "loss": 1.0387, + "step": 26887 + }, + { + "epoch": 2.26, + "learning_rate": 2.9951190915689366e-06, + "loss": 1.0423, + "step": 26888 + }, + { + "epoch": 2.26, + "learning_rate": 2.9944704190831454e-06, + "loss": 1.0814, + "step": 26889 + }, + { + "epoch": 2.26, + "learning_rate": 2.9938218044792753e-06, + "loss": 1.0852, + "step": 26890 + }, + { + "epoch": 2.26, + "learning_rate": 2.9931732477626986e-06, + "loss": 1.114, + "step": 26891 + }, + { + "epoch": 2.26, + "learning_rate": 2.9925247489387634e-06, + "loss": 1.1426, + "step": 26892 + }, + { + "epoch": 2.26, + "learning_rate": 2.991876308012831e-06, + "loss": 1.1911, + "step": 26893 + }, + { + "epoch": 2.26, + "learning_rate": 2.9912279249902598e-06, + "loss": 1.1484, + "step": 26894 + }, + { + "epoch": 2.26, + "learning_rate": 2.9905795998764076e-06, + "loss": 1.1145, + "step": 26895 + }, + { + "epoch": 2.26, + "learning_rate": 2.9899313326766287e-06, + "loss": 1.1185, + "step": 26896 + }, + { + "epoch": 2.26, + "learning_rate": 2.9892831233962803e-06, + "loss": 1.084, + "step": 26897 + }, + { + "epoch": 2.26, + "learning_rate": 2.9886349720407206e-06, + "loss": 1.1427, + "step": 26898 + }, + { + "epoch": 2.26, + "learning_rate": 2.987986878615301e-06, + "loss": 1.0244, + "step": 26899 + }, + { + "epoch": 2.26, + "learning_rate": 2.9873388431253814e-06, + "loss": 1.0079, + "step": 26900 + }, + { + "epoch": 2.26, + "learning_rate": 2.986690865576306e-06, + "loss": 1.1751, + "step": 26901 + }, + { + "epoch": 2.26, + "learning_rate": 2.9860429459734417e-06, + "loss": 1.1037, + "step": 26902 + }, + { + "epoch": 2.27, + "learning_rate": 2.9853950843221323e-06, + "loss": 1.1511, + "step": 26903 + }, + { + "epoch": 2.27, + "learning_rate": 2.9847472806277335e-06, + "loss": 1.0713, + "step": 26904 + }, + { + "epoch": 2.27, + "learning_rate": 2.984099534895598e-06, + "loss": 1.1031, + "step": 26905 + }, + { + "epoch": 2.27, + "learning_rate": 2.983451847131078e-06, + "loss": 1.055, + "step": 26906 + }, + { + "epoch": 2.27, + "learning_rate": 2.9828042173395234e-06, + "loss": 1.1433, + "step": 26907 + }, + { + "epoch": 2.27, + "learning_rate": 2.9821566455262897e-06, + "loss": 1.1277, + "step": 26908 + }, + { + "epoch": 2.27, + "learning_rate": 2.981509131696718e-06, + "loss": 1.0694, + "step": 26909 + }, + { + "epoch": 2.27, + "learning_rate": 2.980861675856167e-06, + "loss": 1.1586, + "step": 26910 + }, + { + "epoch": 2.27, + "learning_rate": 2.9802142780099863e-06, + "loss": 1.1135, + "step": 26911 + }, + { + "epoch": 2.27, + "learning_rate": 2.979566938163515e-06, + "loss": 1.088, + "step": 26912 + }, + { + "epoch": 2.27, + "learning_rate": 2.9789196563221157e-06, + "loss": 1.0889, + "step": 26913 + }, + { + "epoch": 2.27, + "learning_rate": 2.978272432491125e-06, + "loss": 1.1237, + "step": 26914 + }, + { + "epoch": 2.27, + "learning_rate": 2.9776252666758955e-06, + "loss": 1.1845, + "step": 26915 + }, + { + "epoch": 2.27, + "learning_rate": 2.976978158881774e-06, + "loss": 1.1227, + "step": 26916 + }, + { + "epoch": 2.27, + "learning_rate": 2.976331109114106e-06, + "loss": 1.0626, + "step": 26917 + }, + { + "epoch": 2.27, + "learning_rate": 2.975684117378238e-06, + "loss": 1.1211, + "step": 26918 + }, + { + "epoch": 2.27, + "learning_rate": 2.975037183679519e-06, + "loss": 1.0865, + "step": 26919 + }, + { + "epoch": 2.27, + "learning_rate": 2.974390308023285e-06, + "loss": 1.1734, + "step": 26920 + }, + { + "epoch": 2.27, + "learning_rate": 2.9737434904148933e-06, + "loss": 1.1468, + "step": 26921 + }, + { + "epoch": 2.27, + "learning_rate": 2.973096730859678e-06, + "loss": 1.1208, + "step": 26922 + }, + { + "epoch": 2.27, + "learning_rate": 2.972450029362983e-06, + "loss": 1.0754, + "step": 26923 + }, + { + "epoch": 2.27, + "learning_rate": 2.971803385930162e-06, + "loss": 1.0481, + "step": 26924 + }, + { + "epoch": 2.27, + "learning_rate": 2.971156800566548e-06, + "loss": 1.0965, + "step": 26925 + }, + { + "epoch": 2.27, + "learning_rate": 2.9705102732774847e-06, + "loss": 1.0816, + "step": 26926 + }, + { + "epoch": 2.27, + "learning_rate": 2.969863804068316e-06, + "loss": 1.1233, + "step": 26927 + }, + { + "epoch": 2.27, + "learning_rate": 2.9692173929443823e-06, + "loss": 1.1389, + "step": 26928 + }, + { + "epoch": 2.27, + "learning_rate": 2.9685710399110245e-06, + "loss": 1.0814, + "step": 26929 + }, + { + "epoch": 2.27, + "learning_rate": 2.9679247449735873e-06, + "loss": 1.1519, + "step": 26930 + }, + { + "epoch": 2.27, + "learning_rate": 2.967278508137399e-06, + "loss": 1.1313, + "step": 26931 + }, + { + "epoch": 2.27, + "learning_rate": 2.9666323294078124e-06, + "loss": 1.1369, + "step": 26932 + }, + { + "epoch": 2.27, + "learning_rate": 2.965986208790158e-06, + "loss": 1.059, + "step": 26933 + }, + { + "epoch": 2.27, + "learning_rate": 2.9653401462897767e-06, + "loss": 1.1224, + "step": 26934 + }, + { + "epoch": 2.27, + "learning_rate": 2.9646941419120067e-06, + "loss": 1.1202, + "step": 26935 + }, + { + "epoch": 2.27, + "learning_rate": 2.964048195662185e-06, + "loss": 1.0845, + "step": 26936 + }, + { + "epoch": 2.27, + "learning_rate": 2.9634023075456488e-06, + "loss": 1.2292, + "step": 26937 + }, + { + "epoch": 2.27, + "learning_rate": 2.9627564775677342e-06, + "loss": 1.0774, + "step": 26938 + }, + { + "epoch": 2.27, + "learning_rate": 2.962110705733782e-06, + "loss": 1.0898, + "step": 26939 + }, + { + "epoch": 2.27, + "learning_rate": 2.961464992049117e-06, + "loss": 1.1617, + "step": 26940 + }, + { + "epoch": 2.27, + "learning_rate": 2.9608193365190876e-06, + "loss": 1.1461, + "step": 26941 + }, + { + "epoch": 2.27, + "learning_rate": 2.960173739149017e-06, + "loss": 1.0872, + "step": 26942 + }, + { + "epoch": 2.27, + "learning_rate": 2.959528199944245e-06, + "loss": 1.1064, + "step": 26943 + }, + { + "epoch": 2.27, + "learning_rate": 2.9588827189101042e-06, + "loss": 1.0855, + "step": 26944 + }, + { + "epoch": 2.27, + "learning_rate": 2.958237296051928e-06, + "loss": 1.0816, + "step": 26945 + }, + { + "epoch": 2.27, + "learning_rate": 2.957591931375049e-06, + "loss": 1.1178, + "step": 26946 + }, + { + "epoch": 2.27, + "learning_rate": 2.956946624884799e-06, + "loss": 1.0432, + "step": 26947 + }, + { + "epoch": 2.27, + "learning_rate": 2.95630137658651e-06, + "loss": 1.1599, + "step": 26948 + }, + { + "epoch": 2.27, + "learning_rate": 2.9556561864855125e-06, + "loss": 1.1286, + "step": 26949 + }, + { + "epoch": 2.27, + "learning_rate": 2.9550110545871426e-06, + "loss": 1.0723, + "step": 26950 + }, + { + "epoch": 2.27, + "learning_rate": 2.9543659808967194e-06, + "loss": 1.1671, + "step": 26951 + }, + { + "epoch": 2.27, + "learning_rate": 2.9537209654195853e-06, + "loss": 1.0999, + "step": 26952 + }, + { + "epoch": 2.27, + "learning_rate": 2.953076008161061e-06, + "loss": 1.1567, + "step": 26953 + }, + { + "epoch": 2.27, + "learning_rate": 2.952431109126478e-06, + "loss": 1.112, + "step": 26954 + }, + { + "epoch": 2.27, + "learning_rate": 2.951786268321164e-06, + "loss": 1.0854, + "step": 26955 + }, + { + "epoch": 2.27, + "learning_rate": 2.9511414857504484e-06, + "loss": 1.058, + "step": 26956 + }, + { + "epoch": 2.27, + "learning_rate": 2.9504967614196565e-06, + "loss": 1.1431, + "step": 26957 + }, + { + "epoch": 2.27, + "learning_rate": 2.9498520953341213e-06, + "loss": 1.1058, + "step": 26958 + }, + { + "epoch": 2.27, + "learning_rate": 2.949207487499157e-06, + "loss": 1.1295, + "step": 26959 + }, + { + "epoch": 2.27, + "learning_rate": 2.9485629379201007e-06, + "loss": 1.1302, + "step": 26960 + }, + { + "epoch": 2.27, + "learning_rate": 2.947918446602277e-06, + "loss": 1.1615, + "step": 26961 + }, + { + "epoch": 2.27, + "learning_rate": 2.9472740135510015e-06, + "loss": 1.0801, + "step": 26962 + }, + { + "epoch": 2.27, + "learning_rate": 2.946629638771612e-06, + "loss": 1.0956, + "step": 26963 + }, + { + "epoch": 2.27, + "learning_rate": 2.9459853222694223e-06, + "loss": 1.0924, + "step": 26964 + }, + { + "epoch": 2.27, + "learning_rate": 2.9453410640497605e-06, + "loss": 1.0638, + "step": 26965 + }, + { + "epoch": 2.27, + "learning_rate": 2.9446968641179472e-06, + "loss": 1.1586, + "step": 26966 + }, + { + "epoch": 2.27, + "learning_rate": 2.944052722479307e-06, + "loss": 1.0836, + "step": 26967 + }, + { + "epoch": 2.27, + "learning_rate": 2.9434086391391613e-06, + "loss": 1.1678, + "step": 26968 + }, + { + "epoch": 2.27, + "learning_rate": 2.9427646141028355e-06, + "loss": 1.0857, + "step": 26969 + }, + { + "epoch": 2.27, + "learning_rate": 2.9421206473756403e-06, + "loss": 1.0971, + "step": 26970 + }, + { + "epoch": 2.27, + "learning_rate": 2.9414767389629086e-06, + "loss": 1.1269, + "step": 26971 + }, + { + "epoch": 2.27, + "learning_rate": 2.9408328888699534e-06, + "loss": 1.0348, + "step": 26972 + }, + { + "epoch": 2.27, + "learning_rate": 2.940189097102092e-06, + "loss": 1.1131, + "step": 26973 + }, + { + "epoch": 2.27, + "learning_rate": 2.9395453636646544e-06, + "loss": 1.0792, + "step": 26974 + }, + { + "epoch": 2.27, + "learning_rate": 2.9389016885629497e-06, + "loss": 1.057, + "step": 26975 + }, + { + "epoch": 2.27, + "learning_rate": 2.938258071802299e-06, + "loss": 1.0039, + "step": 26976 + }, + { + "epoch": 2.27, + "learning_rate": 2.9376145133880195e-06, + "loss": 1.1394, + "step": 26977 + }, + { + "epoch": 2.27, + "learning_rate": 2.9369710133254294e-06, + "loss": 1.1434, + "step": 26978 + }, + { + "epoch": 2.27, + "learning_rate": 2.936327571619846e-06, + "loss": 1.0867, + "step": 26979 + }, + { + "epoch": 2.27, + "learning_rate": 2.9356841882765875e-06, + "loss": 1.1393, + "step": 26980 + }, + { + "epoch": 2.27, + "learning_rate": 2.935040863300961e-06, + "loss": 0.9705, + "step": 26981 + }, + { + "epoch": 2.27, + "learning_rate": 2.934397596698293e-06, + "loss": 1.1886, + "step": 26982 + }, + { + "epoch": 2.27, + "learning_rate": 2.9337543884738915e-06, + "loss": 1.0974, + "step": 26983 + }, + { + "epoch": 2.27, + "learning_rate": 2.933111238633073e-06, + "loss": 1.1076, + "step": 26984 + }, + { + "epoch": 2.27, + "learning_rate": 2.93246814718115e-06, + "loss": 1.0965, + "step": 26985 + }, + { + "epoch": 2.27, + "learning_rate": 2.9318251141234376e-06, + "loss": 1.0914, + "step": 26986 + }, + { + "epoch": 2.27, + "learning_rate": 2.9311821394652486e-06, + "loss": 1.0843, + "step": 26987 + }, + { + "epoch": 2.27, + "learning_rate": 2.9305392232118944e-06, + "loss": 1.0992, + "step": 26988 + }, + { + "epoch": 2.27, + "learning_rate": 2.9298963653686884e-06, + "loss": 1.1504, + "step": 26989 + }, + { + "epoch": 2.27, + "learning_rate": 2.9292535659409406e-06, + "loss": 1.1964, + "step": 26990 + }, + { + "epoch": 2.27, + "learning_rate": 2.928610824933966e-06, + "loss": 1.0952, + "step": 26991 + }, + { + "epoch": 2.27, + "learning_rate": 2.927968142353069e-06, + "loss": 1.1372, + "step": 26992 + }, + { + "epoch": 2.27, + "learning_rate": 2.927325518203563e-06, + "loss": 1.0679, + "step": 26993 + }, + { + "epoch": 2.27, + "learning_rate": 2.9266829524907568e-06, + "loss": 1.0316, + "step": 26994 + }, + { + "epoch": 2.27, + "learning_rate": 2.9260404452199597e-06, + "loss": 1.0279, + "step": 26995 + }, + { + "epoch": 2.27, + "learning_rate": 2.9253979963964808e-06, + "loss": 1.1132, + "step": 26996 + }, + { + "epoch": 2.27, + "learning_rate": 2.924755606025628e-06, + "loss": 1.0659, + "step": 26997 + }, + { + "epoch": 2.27, + "learning_rate": 2.9241132741127076e-06, + "loss": 1.136, + "step": 26998 + }, + { + "epoch": 2.27, + "learning_rate": 2.9234710006630285e-06, + "loss": 1.1077, + "step": 26999 + }, + { + "epoch": 2.27, + "learning_rate": 2.922828785681899e-06, + "loss": 1.0203, + "step": 27000 + }, + { + "epoch": 2.27, + "learning_rate": 2.9221866291746182e-06, + "loss": 1.0586, + "step": 27001 + }, + { + "epoch": 2.27, + "learning_rate": 2.921544531146503e-06, + "loss": 1.0595, + "step": 27002 + }, + { + "epoch": 2.27, + "learning_rate": 2.9209024916028483e-06, + "loss": 1.1429, + "step": 27003 + }, + { + "epoch": 2.27, + "learning_rate": 2.9202605105489624e-06, + "loss": 1.079, + "step": 27004 + }, + { + "epoch": 2.27, + "learning_rate": 2.9196185879901508e-06, + "loss": 1.1057, + "step": 27005 + }, + { + "epoch": 2.27, + "learning_rate": 2.9189767239317166e-06, + "loss": 1.1618, + "step": 27006 + }, + { + "epoch": 2.27, + "learning_rate": 2.918334918378962e-06, + "loss": 1.2658, + "step": 27007 + }, + { + "epoch": 2.27, + "learning_rate": 2.917693171337195e-06, + "loss": 1.0877, + "step": 27008 + }, + { + "epoch": 2.27, + "learning_rate": 2.917051482811707e-06, + "loss": 0.9972, + "step": 27009 + }, + { + "epoch": 2.27, + "learning_rate": 2.916409852807809e-06, + "loss": 1.1486, + "step": 27010 + }, + { + "epoch": 2.27, + "learning_rate": 2.915768281330803e-06, + "loss": 1.1958, + "step": 27011 + }, + { + "epoch": 2.27, + "learning_rate": 2.915126768385982e-06, + "loss": 1.1362, + "step": 27012 + }, + { + "epoch": 2.27, + "learning_rate": 2.9144853139786557e-06, + "loss": 1.162, + "step": 27013 + }, + { + "epoch": 2.27, + "learning_rate": 2.9138439181141163e-06, + "loss": 1.1041, + "step": 27014 + }, + { + "epoch": 2.27, + "learning_rate": 2.913202580797667e-06, + "loss": 1.0973, + "step": 27015 + }, + { + "epoch": 2.27, + "learning_rate": 2.9125613020346054e-06, + "loss": 1.0333, + "step": 27016 + }, + { + "epoch": 2.27, + "learning_rate": 2.911920081830232e-06, + "loss": 1.0908, + "step": 27017 + }, + { + "epoch": 2.27, + "learning_rate": 2.911278920189842e-06, + "loss": 1.1172, + "step": 27018 + }, + { + "epoch": 2.27, + "learning_rate": 2.910637817118738e-06, + "loss": 1.1881, + "step": 27019 + }, + { + "epoch": 2.27, + "learning_rate": 2.9099967726222065e-06, + "loss": 1.1326, + "step": 27020 + }, + { + "epoch": 2.27, + "learning_rate": 2.909355786705557e-06, + "loss": 1.0633, + "step": 27021 + }, + { + "epoch": 2.28, + "learning_rate": 2.9087148593740764e-06, + "loss": 1.1432, + "step": 27022 + }, + { + "epoch": 2.28, + "learning_rate": 2.90807399063306e-06, + "loss": 1.1142, + "step": 27023 + }, + { + "epoch": 2.28, + "learning_rate": 2.907433180487812e-06, + "loss": 1.1164, + "step": 27024 + }, + { + "epoch": 2.28, + "learning_rate": 2.9067924289436188e-06, + "loss": 1.1373, + "step": 27025 + }, + { + "epoch": 2.28, + "learning_rate": 2.9061517360057756e-06, + "loss": 1.0763, + "step": 27026 + }, + { + "epoch": 2.28, + "learning_rate": 2.905511101679578e-06, + "loss": 1.0779, + "step": 27027 + }, + { + "epoch": 2.28, + "learning_rate": 2.904870525970317e-06, + "loss": 1.0075, + "step": 27028 + }, + { + "epoch": 2.28, + "learning_rate": 2.9042300088832877e-06, + "loss": 1.0778, + "step": 27029 + }, + { + "epoch": 2.28, + "learning_rate": 2.9035895504237833e-06, + "loss": 1.1063, + "step": 27030 + }, + { + "epoch": 2.28, + "learning_rate": 2.902949150597086e-06, + "loss": 1.1509, + "step": 27031 + }, + { + "epoch": 2.28, + "learning_rate": 2.9023088094085026e-06, + "loss": 1.1138, + "step": 27032 + }, + { + "epoch": 2.28, + "learning_rate": 2.9016685268633117e-06, + "loss": 1.1127, + "step": 27033 + }, + { + "epoch": 2.28, + "learning_rate": 2.901028302966806e-06, + "loss": 1.1122, + "step": 27034 + }, + { + "epoch": 2.28, + "learning_rate": 2.900388137724277e-06, + "loss": 1.1549, + "step": 27035 + }, + { + "epoch": 2.28, + "learning_rate": 2.8997480311410143e-06, + "loss": 1.1021, + "step": 27036 + }, + { + "epoch": 2.28, + "learning_rate": 2.8991079832223044e-06, + "loss": 1.1744, + "step": 27037 + }, + { + "epoch": 2.28, + "learning_rate": 2.898467993973437e-06, + "loss": 1.1696, + "step": 27038 + }, + { + "epoch": 2.28, + "learning_rate": 2.8978280633997003e-06, + "loss": 1.0718, + "step": 27039 + }, + { + "epoch": 2.28, + "learning_rate": 2.8971881915063817e-06, + "loss": 1.0783, + "step": 27040 + }, + { + "epoch": 2.28, + "learning_rate": 2.89654837829877e-06, + "loss": 1.1822, + "step": 27041 + }, + { + "epoch": 2.28, + "learning_rate": 2.8959086237821423e-06, + "loss": 1.1448, + "step": 27042 + }, + { + "epoch": 2.28, + "learning_rate": 2.8952689279617984e-06, + "loss": 1.1527, + "step": 27043 + }, + { + "epoch": 2.28, + "learning_rate": 2.8946292908430116e-06, + "loss": 1.0444, + "step": 27044 + }, + { + "epoch": 2.28, + "learning_rate": 2.8939897124310734e-06, + "loss": 1.0785, + "step": 27045 + }, + { + "epoch": 2.28, + "learning_rate": 2.893350192731266e-06, + "loss": 1.2111, + "step": 27046 + }, + { + "epoch": 2.28, + "learning_rate": 2.8927107317488736e-06, + "loss": 1.2136, + "step": 27047 + }, + { + "epoch": 2.28, + "learning_rate": 2.8920713294891788e-06, + "loss": 1.1338, + "step": 27048 + }, + { + "epoch": 2.28, + "learning_rate": 2.891431985957467e-06, + "loss": 1.1658, + "step": 27049 + }, + { + "epoch": 2.28, + "learning_rate": 2.890792701159022e-06, + "loss": 1.1647, + "step": 27050 + }, + { + "epoch": 2.28, + "learning_rate": 2.8901534750991158e-06, + "loss": 1.1355, + "step": 27051 + }, + { + "epoch": 2.28, + "learning_rate": 2.8895143077830434e-06, + "loss": 1.1403, + "step": 27052 + }, + { + "epoch": 2.28, + "learning_rate": 2.8888751992160767e-06, + "loss": 1.1078, + "step": 27053 + }, + { + "epoch": 2.28, + "learning_rate": 2.8882361494034993e-06, + "loss": 1.1153, + "step": 27054 + }, + { + "epoch": 2.28, + "learning_rate": 2.88759715835059e-06, + "loss": 1.0963, + "step": 27055 + }, + { + "epoch": 2.28, + "learning_rate": 2.8869582260626304e-06, + "loss": 1.1341, + "step": 27056 + }, + { + "epoch": 2.28, + "learning_rate": 2.8863193525448973e-06, + "loss": 1.1131, + "step": 27057 + }, + { + "epoch": 2.28, + "learning_rate": 2.885680537802674e-06, + "loss": 1.1383, + "step": 27058 + }, + { + "epoch": 2.28, + "learning_rate": 2.8850417818412292e-06, + "loss": 1.1649, + "step": 27059 + }, + { + "epoch": 2.28, + "learning_rate": 2.884403084665849e-06, + "loss": 1.1398, + "step": 27060 + }, + { + "epoch": 2.28, + "learning_rate": 2.88376444628181e-06, + "loss": 1.0701, + "step": 27061 + }, + { + "epoch": 2.28, + "learning_rate": 2.883125866694382e-06, + "loss": 1.0634, + "step": 27062 + }, + { + "epoch": 2.28, + "learning_rate": 2.8824873459088508e-06, + "loss": 1.0856, + "step": 27063 + }, + { + "epoch": 2.28, + "learning_rate": 2.881848883930485e-06, + "loss": 1.1068, + "step": 27064 + }, + { + "epoch": 2.28, + "learning_rate": 2.8812104807645615e-06, + "loss": 1.1318, + "step": 27065 + }, + { + "epoch": 2.28, + "learning_rate": 2.880572136416354e-06, + "loss": 1.1102, + "step": 27066 + }, + { + "epoch": 2.28, + "learning_rate": 2.879933850891139e-06, + "loss": 1.1149, + "step": 27067 + }, + { + "epoch": 2.28, + "learning_rate": 2.8792956241941894e-06, + "loss": 1.1564, + "step": 27068 + }, + { + "epoch": 2.28, + "learning_rate": 2.878657456330781e-06, + "loss": 1.1857, + "step": 27069 + }, + { + "epoch": 2.28, + "learning_rate": 2.878019347306178e-06, + "loss": 1.1297, + "step": 27070 + }, + { + "epoch": 2.28, + "learning_rate": 2.8773812971256654e-06, + "loss": 1.2365, + "step": 27071 + }, + { + "epoch": 2.28, + "learning_rate": 2.876743305794504e-06, + "loss": 1.1187, + "step": 27072 + }, + { + "epoch": 2.28, + "learning_rate": 2.8761053733179658e-06, + "loss": 1.1901, + "step": 27073 + }, + { + "epoch": 2.28, + "learning_rate": 2.875467499701331e-06, + "loss": 1.0987, + "step": 27074 + }, + { + "epoch": 2.28, + "learning_rate": 2.8748296849498616e-06, + "loss": 1.1407, + "step": 27075 + }, + { + "epoch": 2.28, + "learning_rate": 2.8741919290688293e-06, + "loss": 1.2191, + "step": 27076 + }, + { + "epoch": 2.28, + "learning_rate": 2.8735542320635036e-06, + "loss": 1.0529, + "step": 27077 + }, + { + "epoch": 2.28, + "learning_rate": 2.8729165939391546e-06, + "loss": 1.124, + "step": 27078 + }, + { + "epoch": 2.28, + "learning_rate": 2.872279014701048e-06, + "loss": 1.0489, + "step": 27079 + }, + { + "epoch": 2.28, + "learning_rate": 2.871641494354458e-06, + "loss": 1.1292, + "step": 27080 + }, + { + "epoch": 2.28, + "learning_rate": 2.871004032904641e-06, + "loss": 1.0693, + "step": 27081 + }, + { + "epoch": 2.28, + "learning_rate": 2.8703666303568757e-06, + "loss": 1.0659, + "step": 27082 + }, + { + "epoch": 2.28, + "learning_rate": 2.8697292867164196e-06, + "loss": 1.0801, + "step": 27083 + }, + { + "epoch": 2.28, + "learning_rate": 2.869092001988543e-06, + "loss": 1.0783, + "step": 27084 + }, + { + "epoch": 2.28, + "learning_rate": 2.8684547761785097e-06, + "loss": 1.1171, + "step": 27085 + }, + { + "epoch": 2.28, + "learning_rate": 2.8678176092915866e-06, + "loss": 1.1695, + "step": 27086 + }, + { + "epoch": 2.28, + "learning_rate": 2.8671805013330357e-06, + "loss": 1.0863, + "step": 27087 + }, + { + "epoch": 2.28, + "learning_rate": 2.8665434523081227e-06, + "loss": 1.1022, + "step": 27088 + }, + { + "epoch": 2.28, + "learning_rate": 2.865906462222111e-06, + "loss": 0.9637, + "step": 27089 + }, + { + "epoch": 2.28, + "learning_rate": 2.8652695310802636e-06, + "loss": 1.1506, + "step": 27090 + }, + { + "epoch": 2.28, + "learning_rate": 2.864632658887845e-06, + "loss": 1.0765, + "step": 27091 + }, + { + "epoch": 2.28, + "learning_rate": 2.8639958456501093e-06, + "loss": 1.0281, + "step": 27092 + }, + { + "epoch": 2.28, + "learning_rate": 2.8633590913723298e-06, + "loss": 1.1476, + "step": 27093 + }, + { + "epoch": 2.28, + "learning_rate": 2.8627223960597582e-06, + "loss": 1.1511, + "step": 27094 + }, + { + "epoch": 2.28, + "learning_rate": 2.8620857597176586e-06, + "loss": 1.0981, + "step": 27095 + }, + { + "epoch": 2.28, + "learning_rate": 2.8614491823512914e-06, + "loss": 1.1456, + "step": 27096 + }, + { + "epoch": 2.28, + "learning_rate": 2.860812663965915e-06, + "loss": 1.1423, + "step": 27097 + }, + { + "epoch": 2.28, + "learning_rate": 2.8601762045667902e-06, + "loss": 0.996, + "step": 27098 + }, + { + "epoch": 2.28, + "learning_rate": 2.859539804159174e-06, + "loss": 1.0986, + "step": 27099 + }, + { + "epoch": 2.28, + "learning_rate": 2.858903462748328e-06, + "loss": 1.12, + "step": 27100 + }, + { + "epoch": 2.28, + "learning_rate": 2.8582671803395003e-06, + "loss": 1.06, + "step": 27101 + }, + { + "epoch": 2.28, + "learning_rate": 2.8576309569379624e-06, + "loss": 1.1832, + "step": 27102 + }, + { + "epoch": 2.28, + "learning_rate": 2.8569947925489596e-06, + "loss": 1.0815, + "step": 27103 + }, + { + "epoch": 2.28, + "learning_rate": 2.8563586871777516e-06, + "loss": 1.1456, + "step": 27104 + }, + { + "epoch": 2.28, + "learning_rate": 2.8557226408295947e-06, + "loss": 1.0595, + "step": 27105 + }, + { + "epoch": 2.28, + "learning_rate": 2.8550866535097433e-06, + "loss": 1.1373, + "step": 27106 + }, + { + "epoch": 2.28, + "learning_rate": 2.854450725223453e-06, + "loss": 1.1279, + "step": 27107 + }, + { + "epoch": 2.28, + "learning_rate": 2.8538148559759813e-06, + "loss": 1.0025, + "step": 27108 + }, + { + "epoch": 2.28, + "learning_rate": 2.8531790457725717e-06, + "loss": 1.11, + "step": 27109 + }, + { + "epoch": 2.28, + "learning_rate": 2.8525432946184873e-06, + "loss": 1.1044, + "step": 27110 + }, + { + "epoch": 2.28, + "learning_rate": 2.8519076025189805e-06, + "loss": 1.176, + "step": 27111 + }, + { + "epoch": 2.28, + "learning_rate": 2.8512719694792944e-06, + "loss": 1.1122, + "step": 27112 + }, + { + "epoch": 2.28, + "learning_rate": 2.8506363955046934e-06, + "loss": 1.0821, + "step": 27113 + }, + { + "epoch": 2.28, + "learning_rate": 2.8500008806004198e-06, + "loss": 1.0802, + "step": 27114 + }, + { + "epoch": 2.28, + "learning_rate": 2.849365424771727e-06, + "loss": 1.134, + "step": 27115 + }, + { + "epoch": 2.28, + "learning_rate": 2.848730028023866e-06, + "loss": 1.1974, + "step": 27116 + }, + { + "epoch": 2.28, + "learning_rate": 2.8480946903620866e-06, + "loss": 1.1982, + "step": 27117 + }, + { + "epoch": 2.28, + "learning_rate": 2.8474594117916367e-06, + "loss": 1.1559, + "step": 27118 + }, + { + "epoch": 2.28, + "learning_rate": 2.8468241923177696e-06, + "loss": 1.1842, + "step": 27119 + }, + { + "epoch": 2.28, + "learning_rate": 2.8461890319457242e-06, + "loss": 1.1947, + "step": 27120 + }, + { + "epoch": 2.28, + "learning_rate": 2.845553930680761e-06, + "loss": 1.1897, + "step": 27121 + }, + { + "epoch": 2.28, + "learning_rate": 2.8449188885281174e-06, + "loss": 1.1497, + "step": 27122 + }, + { + "epoch": 2.28, + "learning_rate": 2.8442839054930418e-06, + "loss": 1.1048, + "step": 27123 + }, + { + "epoch": 2.28, + "learning_rate": 2.843648981580788e-06, + "loss": 1.1259, + "step": 27124 + }, + { + "epoch": 2.28, + "learning_rate": 2.843014116796593e-06, + "loss": 1.0933, + "step": 27125 + }, + { + "epoch": 2.28, + "learning_rate": 2.8423793111457067e-06, + "loss": 1.0007, + "step": 27126 + }, + { + "epoch": 2.28, + "learning_rate": 2.841744564633373e-06, + "loss": 1.1493, + "step": 27127 + }, + { + "epoch": 2.28, + "learning_rate": 2.841109877264837e-06, + "loss": 1.1851, + "step": 27128 + }, + { + "epoch": 2.28, + "learning_rate": 2.8404752490453426e-06, + "loss": 1.0977, + "step": 27129 + }, + { + "epoch": 2.28, + "learning_rate": 2.839840679980136e-06, + "loss": 1.2208, + "step": 27130 + }, + { + "epoch": 2.28, + "learning_rate": 2.83920617007445e-06, + "loss": 1.0792, + "step": 27131 + }, + { + "epoch": 2.28, + "learning_rate": 2.8385717193335415e-06, + "loss": 1.0341, + "step": 27132 + }, + { + "epoch": 2.28, + "learning_rate": 2.837937327762642e-06, + "loss": 1.1138, + "step": 27133 + }, + { + "epoch": 2.28, + "learning_rate": 2.837302995366996e-06, + "loss": 1.1089, + "step": 27134 + }, + { + "epoch": 2.28, + "learning_rate": 2.8366687221518465e-06, + "loss": 1.0655, + "step": 27135 + }, + { + "epoch": 2.28, + "learning_rate": 2.8360345081224317e-06, + "loss": 1.1027, + "step": 27136 + }, + { + "epoch": 2.28, + "learning_rate": 2.8354003532839923e-06, + "loss": 1.1331, + "step": 27137 + }, + { + "epoch": 2.28, + "learning_rate": 2.834766257641768e-06, + "loss": 1.1557, + "step": 27138 + }, + { + "epoch": 2.28, + "learning_rate": 2.834132221200998e-06, + "loss": 1.223, + "step": 27139 + }, + { + "epoch": 2.28, + "learning_rate": 2.8334982439669213e-06, + "loss": 1.0361, + "step": 27140 + }, + { + "epoch": 2.29, + "learning_rate": 2.83286432594478e-06, + "loss": 1.1012, + "step": 27141 + }, + { + "epoch": 2.29, + "learning_rate": 2.8322304671397992e-06, + "loss": 1.0813, + "step": 27142 + }, + { + "epoch": 2.29, + "learning_rate": 2.8315966675572316e-06, + "loss": 1.2064, + "step": 27143 + }, + { + "epoch": 2.29, + "learning_rate": 2.830962927202303e-06, + "loss": 1.1004, + "step": 27144 + }, + { + "epoch": 2.29, + "learning_rate": 2.8303292460802532e-06, + "loss": 1.1325, + "step": 27145 + }, + { + "epoch": 2.29, + "learning_rate": 2.829695624196317e-06, + "loss": 1.0852, + "step": 27146 + }, + { + "epoch": 2.29, + "learning_rate": 2.829062061555732e-06, + "loss": 1.0548, + "step": 27147 + }, + { + "epoch": 2.29, + "learning_rate": 2.8284285581637293e-06, + "loss": 1.0585, + "step": 27148 + }, + { + "epoch": 2.29, + "learning_rate": 2.8277951140255466e-06, + "loss": 1.0752, + "step": 27149 + }, + { + "epoch": 2.29, + "learning_rate": 2.8271617291464183e-06, + "loss": 1.0903, + "step": 27150 + }, + { + "epoch": 2.29, + "learning_rate": 2.8265284035315697e-06, + "loss": 1.1001, + "step": 27151 + }, + { + "epoch": 2.29, + "learning_rate": 2.825895137186245e-06, + "loss": 1.1029, + "step": 27152 + }, + { + "epoch": 2.29, + "learning_rate": 2.825261930115668e-06, + "loss": 1.1204, + "step": 27153 + }, + { + "epoch": 2.29, + "learning_rate": 2.824628782325073e-06, + "loss": 1.1304, + "step": 27154 + }, + { + "epoch": 2.29, + "learning_rate": 2.8239956938196923e-06, + "loss": 1.1108, + "step": 27155 + }, + { + "epoch": 2.29, + "learning_rate": 2.823362664604754e-06, + "loss": 1.101, + "step": 27156 + }, + { + "epoch": 2.29, + "learning_rate": 2.822729694685491e-06, + "loss": 1.1654, + "step": 27157 + }, + { + "epoch": 2.29, + "learning_rate": 2.8220967840671352e-06, + "loss": 1.1268, + "step": 27158 + }, + { + "epoch": 2.29, + "learning_rate": 2.8214639327549076e-06, + "loss": 1.0964, + "step": 27159 + }, + { + "epoch": 2.29, + "learning_rate": 2.8208311407540457e-06, + "loss": 1.1694, + "step": 27160 + }, + { + "epoch": 2.29, + "learning_rate": 2.8201984080697763e-06, + "loss": 1.0576, + "step": 27161 + }, + { + "epoch": 2.29, + "learning_rate": 2.81956573470732e-06, + "loss": 1.1015, + "step": 27162 + }, + { + "epoch": 2.29, + "learning_rate": 2.818933120671916e-06, + "loss": 1.168, + "step": 27163 + }, + { + "epoch": 2.29, + "learning_rate": 2.8183005659687812e-06, + "loss": 1.1343, + "step": 27164 + }, + { + "epoch": 2.29, + "learning_rate": 2.8176680706031456e-06, + "loss": 1.0256, + "step": 27165 + }, + { + "epoch": 2.29, + "learning_rate": 2.817035634580235e-06, + "loss": 1.0699, + "step": 27166 + }, + { + "epoch": 2.29, + "learning_rate": 2.816403257905276e-06, + "loss": 1.2102, + "step": 27167 + }, + { + "epoch": 2.29, + "learning_rate": 2.815770940583491e-06, + "loss": 1.0975, + "step": 27168 + }, + { + "epoch": 2.29, + "learning_rate": 2.815138682620109e-06, + "loss": 1.1117, + "step": 27169 + }, + { + "epoch": 2.29, + "learning_rate": 2.814506484020345e-06, + "loss": 1.0976, + "step": 27170 + }, + { + "epoch": 2.29, + "learning_rate": 2.8138743447894334e-06, + "loss": 1.1019, + "step": 27171 + }, + { + "epoch": 2.29, + "learning_rate": 2.8132422649325896e-06, + "loss": 1.1196, + "step": 27172 + }, + { + "epoch": 2.29, + "learning_rate": 2.8126102444550353e-06, + "loss": 1.1192, + "step": 27173 + }, + { + "epoch": 2.29, + "learning_rate": 2.8119782833620003e-06, + "loss": 1.0988, + "step": 27174 + }, + { + "epoch": 2.29, + "learning_rate": 2.8113463816586995e-06, + "loss": 1.0863, + "step": 27175 + }, + { + "epoch": 2.29, + "learning_rate": 2.8107145393503552e-06, + "loss": 1.21, + "step": 27176 + }, + { + "epoch": 2.29, + "learning_rate": 2.8100827564421873e-06, + "loss": 1.1343, + "step": 27177 + }, + { + "epoch": 2.29, + "learning_rate": 2.8094510329394178e-06, + "loss": 1.1149, + "step": 27178 + }, + { + "epoch": 2.29, + "learning_rate": 2.8088193688472655e-06, + "loss": 1.1107, + "step": 27179 + }, + { + "epoch": 2.29, + "learning_rate": 2.808187764170951e-06, + "loss": 1.0536, + "step": 27180 + }, + { + "epoch": 2.29, + "learning_rate": 2.8075562189156848e-06, + "loss": 1.0667, + "step": 27181 + }, + { + "epoch": 2.29, + "learning_rate": 2.8069247330866976e-06, + "loss": 1.0973, + "step": 27182 + }, + { + "epoch": 2.29, + "learning_rate": 2.8062933066891975e-06, + "loss": 1.1835, + "step": 27183 + }, + { + "epoch": 2.29, + "learning_rate": 2.8056619397284036e-06, + "loss": 1.0871, + "step": 27184 + }, + { + "epoch": 2.29, + "learning_rate": 2.8050306322095343e-06, + "loss": 1.0594, + "step": 27185 + }, + { + "epoch": 2.29, + "learning_rate": 2.8043993841378035e-06, + "loss": 1.1526, + "step": 27186 + }, + { + "epoch": 2.29, + "learning_rate": 2.8037681955184283e-06, + "loss": 1.124, + "step": 27187 + }, + { + "epoch": 2.29, + "learning_rate": 2.8031370663566236e-06, + "loss": 1.1104, + "step": 27188 + }, + { + "epoch": 2.29, + "learning_rate": 2.802505996657604e-06, + "loss": 1.1382, + "step": 27189 + }, + { + "epoch": 2.29, + "learning_rate": 2.8018749864265826e-06, + "loss": 1.0828, + "step": 27190 + }, + { + "epoch": 2.29, + "learning_rate": 2.8012440356687775e-06, + "loss": 1.1575, + "step": 27191 + }, + { + "epoch": 2.29, + "learning_rate": 2.8006131443893914e-06, + "loss": 1.1086, + "step": 27192 + }, + { + "epoch": 2.29, + "learning_rate": 2.79998231259365e-06, + "loss": 1.1196, + "step": 27193 + }, + { + "epoch": 2.29, + "learning_rate": 2.7993515402867556e-06, + "loss": 1.0801, + "step": 27194 + }, + { + "epoch": 2.29, + "learning_rate": 2.798720827473924e-06, + "loss": 1.1259, + "step": 27195 + }, + { + "epoch": 2.29, + "learning_rate": 2.798090174160365e-06, + "loss": 1.1327, + "step": 27196 + }, + { + "epoch": 2.29, + "learning_rate": 2.7974595803512894e-06, + "loss": 1.1808, + "step": 27197 + }, + { + "epoch": 2.29, + "learning_rate": 2.796829046051909e-06, + "loss": 1.058, + "step": 27198 + }, + { + "epoch": 2.29, + "learning_rate": 2.796198571267432e-06, + "loss": 1.1152, + "step": 27199 + }, + { + "epoch": 2.29, + "learning_rate": 2.795568156003071e-06, + "loss": 1.0952, + "step": 27200 + }, + { + "epoch": 2.29, + "learning_rate": 2.794937800264025e-06, + "loss": 1.1321, + "step": 27201 + }, + { + "epoch": 2.29, + "learning_rate": 2.7943075040555156e-06, + "loss": 1.3034, + "step": 27202 + }, + { + "epoch": 2.29, + "learning_rate": 2.7936772673827406e-06, + "loss": 1.1086, + "step": 27203 + }, + { + "epoch": 2.29, + "learning_rate": 2.79304709025091e-06, + "loss": 1.1772, + "step": 27204 + }, + { + "epoch": 2.29, + "learning_rate": 2.7924169726652308e-06, + "loss": 1.1128, + "step": 27205 + }, + { + "epoch": 2.29, + "learning_rate": 2.7917869146309094e-06, + "loss": 1.1439, + "step": 27206 + }, + { + "epoch": 2.29, + "learning_rate": 2.7911569161531517e-06, + "loss": 1.0954, + "step": 27207 + }, + { + "epoch": 2.29, + "learning_rate": 2.7905269772371647e-06, + "loss": 1.0636, + "step": 27208 + }, + { + "epoch": 2.29, + "learning_rate": 2.7898970978881455e-06, + "loss": 1.1, + "step": 27209 + }, + { + "epoch": 2.29, + "learning_rate": 2.7892672781113084e-06, + "loss": 1.1375, + "step": 27210 + }, + { + "epoch": 2.29, + "learning_rate": 2.788637517911854e-06, + "loss": 1.0658, + "step": 27211 + }, + { + "epoch": 2.29, + "learning_rate": 2.7880078172949798e-06, + "loss": 1.1057, + "step": 27212 + }, + { + "epoch": 2.29, + "learning_rate": 2.787378176265898e-06, + "loss": 1.1386, + "step": 27213 + }, + { + "epoch": 2.29, + "learning_rate": 2.7867485948298033e-06, + "loss": 1.123, + "step": 27214 + }, + { + "epoch": 2.29, + "learning_rate": 2.7861190729919006e-06, + "loss": 1.1438, + "step": 27215 + }, + { + "epoch": 2.29, + "learning_rate": 2.7854896107573902e-06, + "loss": 1.0171, + "step": 27216 + }, + { + "epoch": 2.29, + "learning_rate": 2.7848602081314746e-06, + "loss": 1.1652, + "step": 27217 + }, + { + "epoch": 2.29, + "learning_rate": 2.784230865119353e-06, + "loss": 1.1485, + "step": 27218 + }, + { + "epoch": 2.29, + "learning_rate": 2.7836015817262274e-06, + "loss": 1.1468, + "step": 27219 + }, + { + "epoch": 2.29, + "learning_rate": 2.78297235795729e-06, + "loss": 1.0565, + "step": 27220 + }, + { + "epoch": 2.29, + "learning_rate": 2.7823431938177505e-06, + "loss": 1.0536, + "step": 27221 + }, + { + "epoch": 2.29, + "learning_rate": 2.781714089312798e-06, + "loss": 1.1384, + "step": 27222 + }, + { + "epoch": 2.29, + "learning_rate": 2.781085044447631e-06, + "loss": 1.2071, + "step": 27223 + }, + { + "epoch": 2.29, + "learning_rate": 2.7804560592274566e-06, + "loss": 1.0421, + "step": 27224 + }, + { + "epoch": 2.29, + "learning_rate": 2.77982713365746e-06, + "loss": 1.1438, + "step": 27225 + }, + { + "epoch": 2.29, + "learning_rate": 2.7791982677428443e-06, + "loss": 1.1363, + "step": 27226 + }, + { + "epoch": 2.29, + "learning_rate": 2.7785694614888024e-06, + "loss": 1.1604, + "step": 27227 + }, + { + "epoch": 2.29, + "learning_rate": 2.777940714900531e-06, + "loss": 1.0993, + "step": 27228 + }, + { + "epoch": 2.29, + "learning_rate": 2.7773120279832246e-06, + "loss": 1.2022, + "step": 27229 + }, + { + "epoch": 2.29, + "learning_rate": 2.776683400742081e-06, + "loss": 1.0965, + "step": 27230 + }, + { + "epoch": 2.29, + "learning_rate": 2.7760548331822844e-06, + "loss": 1.1082, + "step": 27231 + }, + { + "epoch": 2.29, + "learning_rate": 2.7754263253090406e-06, + "loss": 1.0225, + "step": 27232 + }, + { + "epoch": 2.29, + "learning_rate": 2.774797877127535e-06, + "loss": 1.17, + "step": 27233 + }, + { + "epoch": 2.29, + "learning_rate": 2.7741694886429606e-06, + "loss": 1.1631, + "step": 27234 + }, + { + "epoch": 2.29, + "learning_rate": 2.773541159860511e-06, + "loss": 1.1586, + "step": 27235 + }, + { + "epoch": 2.29, + "learning_rate": 2.7729128907853763e-06, + "loss": 1.0132, + "step": 27236 + }, + { + "epoch": 2.29, + "learning_rate": 2.772284681422749e-06, + "loss": 1.1384, + "step": 27237 + }, + { + "epoch": 2.29, + "learning_rate": 2.771656531777819e-06, + "loss": 1.1245, + "step": 27238 + }, + { + "epoch": 2.29, + "learning_rate": 2.7710284418557756e-06, + "loss": 1.1848, + "step": 27239 + }, + { + "epoch": 2.29, + "learning_rate": 2.770400411661809e-06, + "loss": 1.1432, + "step": 27240 + }, + { + "epoch": 2.29, + "learning_rate": 2.7697724412011106e-06, + "loss": 1.1263, + "step": 27241 + }, + { + "epoch": 2.29, + "learning_rate": 2.76914453047886e-06, + "loss": 1.0552, + "step": 27242 + }, + { + "epoch": 2.29, + "learning_rate": 2.7685166795002584e-06, + "loss": 1.1118, + "step": 27243 + }, + { + "epoch": 2.29, + "learning_rate": 2.7678888882704836e-06, + "loss": 1.09, + "step": 27244 + }, + { + "epoch": 2.29, + "learning_rate": 2.767261156794724e-06, + "loss": 1.1303, + "step": 27245 + }, + { + "epoch": 2.29, + "learning_rate": 2.7666334850781686e-06, + "loss": 1.0962, + "step": 27246 + }, + { + "epoch": 2.29, + "learning_rate": 2.766005873126002e-06, + "loss": 1.0522, + "step": 27247 + }, + { + "epoch": 2.29, + "learning_rate": 2.765378320943409e-06, + "loss": 1.0854, + "step": 27248 + }, + { + "epoch": 2.29, + "learning_rate": 2.7647508285355774e-06, + "loss": 1.2165, + "step": 27249 + }, + { + "epoch": 2.29, + "learning_rate": 2.7641233959076917e-06, + "loss": 1.118, + "step": 27250 + }, + { + "epoch": 2.29, + "learning_rate": 2.76349602306493e-06, + "loss": 1.0864, + "step": 27251 + }, + { + "epoch": 2.29, + "learning_rate": 2.762868710012484e-06, + "loss": 1.1806, + "step": 27252 + }, + { + "epoch": 2.29, + "learning_rate": 2.762241456755531e-06, + "loss": 1.1588, + "step": 27253 + }, + { + "epoch": 2.29, + "learning_rate": 2.761614263299256e-06, + "loss": 1.0905, + "step": 27254 + }, + { + "epoch": 2.29, + "learning_rate": 2.76098712964884e-06, + "loss": 1.0394, + "step": 27255 + }, + { + "epoch": 2.29, + "learning_rate": 2.7603600558094644e-06, + "loss": 1.1052, + "step": 27256 + }, + { + "epoch": 2.29, + "learning_rate": 2.759733041786312e-06, + "loss": 1.1328, + "step": 27257 + }, + { + "epoch": 2.29, + "learning_rate": 2.7591060875845655e-06, + "loss": 1.0888, + "step": 27258 + }, + { + "epoch": 2.29, + "learning_rate": 2.7584791932093947e-06, + "loss": 1.1749, + "step": 27259 + }, + { + "epoch": 2.3, + "learning_rate": 2.75785235866599e-06, + "loss": 1.1515, + "step": 27260 + }, + { + "epoch": 2.3, + "learning_rate": 2.7572255839595306e-06, + "loss": 1.1306, + "step": 27261 + }, + { + "epoch": 2.3, + "learning_rate": 2.756598869095185e-06, + "loss": 1.1585, + "step": 27262 + }, + { + "epoch": 2.3, + "learning_rate": 2.755972214078143e-06, + "loss": 1.1345, + "step": 27263 + }, + { + "epoch": 2.3, + "learning_rate": 2.7553456189135752e-06, + "loss": 1.0538, + "step": 27264 + }, + { + "epoch": 2.3, + "learning_rate": 2.7547190836066593e-06, + "loss": 1.1806, + "step": 27265 + }, + { + "epoch": 2.3, + "learning_rate": 2.754092608162574e-06, + "loss": 1.1514, + "step": 27266 + }, + { + "epoch": 2.3, + "learning_rate": 2.753466192586494e-06, + "loss": 1.052, + "step": 27267 + }, + { + "epoch": 2.3, + "learning_rate": 2.752839836883595e-06, + "loss": 1.0514, + "step": 27268 + }, + { + "epoch": 2.3, + "learning_rate": 2.752213541059057e-06, + "loss": 1.0586, + "step": 27269 + }, + { + "epoch": 2.3, + "learning_rate": 2.751587305118043e-06, + "loss": 1.0935, + "step": 27270 + }, + { + "epoch": 2.3, + "learning_rate": 2.750961129065741e-06, + "loss": 1.1285, + "step": 27271 + }, + { + "epoch": 2.3, + "learning_rate": 2.750335012907316e-06, + "loss": 1.1379, + "step": 27272 + }, + { + "epoch": 2.3, + "learning_rate": 2.7497089566479395e-06, + "loss": 1.0837, + "step": 27273 + }, + { + "epoch": 2.3, + "learning_rate": 2.7490829602927935e-06, + "loss": 0.9454, + "step": 27274 + }, + { + "epoch": 2.3, + "learning_rate": 2.748457023847042e-06, + "loss": 1.0656, + "step": 27275 + }, + { + "epoch": 2.3, + "learning_rate": 2.7478311473158605e-06, + "loss": 1.2287, + "step": 27276 + }, + { + "epoch": 2.3, + "learning_rate": 2.747205330704418e-06, + "loss": 1.1399, + "step": 27277 + }, + { + "epoch": 2.3, + "learning_rate": 2.746579574017886e-06, + "loss": 1.1528, + "step": 27278 + }, + { + "epoch": 2.3, + "learning_rate": 2.745953877261436e-06, + "loss": 1.1507, + "step": 27279 + }, + { + "epoch": 2.3, + "learning_rate": 2.7453282404402383e-06, + "loss": 1.1508, + "step": 27280 + }, + { + "epoch": 2.3, + "learning_rate": 2.7447026635594553e-06, + "loss": 1.0781, + "step": 27281 + }, + { + "epoch": 2.3, + "learning_rate": 2.7440771466242667e-06, + "loss": 1.0607, + "step": 27282 + }, + { + "epoch": 2.3, + "learning_rate": 2.7434516896398322e-06, + "loss": 1.1418, + "step": 27283 + }, + { + "epoch": 2.3, + "learning_rate": 2.7428262926113214e-06, + "loss": 1.1574, + "step": 27284 + }, + { + "epoch": 2.3, + "learning_rate": 2.7422009555439035e-06, + "loss": 1.1105, + "step": 27285 + }, + { + "epoch": 2.3, + "learning_rate": 2.741575678442744e-06, + "loss": 1.1384, + "step": 27286 + }, + { + "epoch": 2.3, + "learning_rate": 2.7409504613130078e-06, + "loss": 1.1919, + "step": 27287 + }, + { + "epoch": 2.3, + "learning_rate": 2.7403253041598632e-06, + "loss": 1.0837, + "step": 27288 + }, + { + "epoch": 2.3, + "learning_rate": 2.7397002069884746e-06, + "loss": 1.1139, + "step": 27289 + }, + { + "epoch": 2.3, + "learning_rate": 2.7390751698040064e-06, + "loss": 1.0372, + "step": 27290 + }, + { + "epoch": 2.3, + "learning_rate": 2.7384501926116246e-06, + "loss": 1.1154, + "step": 27291 + }, + { + "epoch": 2.3, + "learning_rate": 2.7378252754164867e-06, + "loss": 1.067, + "step": 27292 + }, + { + "epoch": 2.3, + "learning_rate": 2.7372004182237665e-06, + "loss": 1.121, + "step": 27293 + }, + { + "epoch": 2.3, + "learning_rate": 2.7365756210386172e-06, + "loss": 1.2091, + "step": 27294 + }, + { + "epoch": 2.3, + "learning_rate": 2.735950883866205e-06, + "loss": 1.0747, + "step": 27295 + }, + { + "epoch": 2.3, + "learning_rate": 2.7353262067116913e-06, + "loss": 1.1422, + "step": 27296 + }, + { + "epoch": 2.3, + "learning_rate": 2.7347015895802375e-06, + "loss": 1.1026, + "step": 27297 + }, + { + "epoch": 2.3, + "learning_rate": 2.734077032477005e-06, + "loss": 1.1179, + "step": 27298 + }, + { + "epoch": 2.3, + "learning_rate": 2.7334525354071527e-06, + "loss": 1.063, + "step": 27299 + }, + { + "epoch": 2.3, + "learning_rate": 2.732828098375845e-06, + "loss": 1.0747, + "step": 27300 + }, + { + "epoch": 2.3, + "learning_rate": 2.73220372138823e-06, + "loss": 1.0457, + "step": 27301 + }, + { + "epoch": 2.3, + "learning_rate": 2.7315794044494815e-06, + "loss": 1.14, + "step": 27302 + }, + { + "epoch": 2.3, + "learning_rate": 2.7309551475647455e-06, + "loss": 1.1186, + "step": 27303 + }, + { + "epoch": 2.3, + "learning_rate": 2.730330950739185e-06, + "loss": 1.1298, + "step": 27304 + }, + { + "epoch": 2.3, + "learning_rate": 2.729706813977957e-06, + "loss": 1.1768, + "step": 27305 + }, + { + "epoch": 2.3, + "learning_rate": 2.729082737286218e-06, + "loss": 1.0286, + "step": 27306 + }, + { + "epoch": 2.3, + "learning_rate": 2.7284587206691247e-06, + "loss": 1.0823, + "step": 27307 + }, + { + "epoch": 2.3, + "learning_rate": 2.727834764131836e-06, + "loss": 1.1919, + "step": 27308 + }, + { + "epoch": 2.3, + "learning_rate": 2.727210867679497e-06, + "loss": 1.1507, + "step": 27309 + }, + { + "epoch": 2.3, + "learning_rate": 2.7265870313172726e-06, + "loss": 1.1541, + "step": 27310 + }, + { + "epoch": 2.3, + "learning_rate": 2.7259632550503168e-06, + "loss": 1.155, + "step": 27311 + }, + { + "epoch": 2.3, + "learning_rate": 2.7253395388837743e-06, + "loss": 1.0449, + "step": 27312 + }, + { + "epoch": 2.3, + "learning_rate": 2.7247158828228114e-06, + "loss": 1.0841, + "step": 27313 + }, + { + "epoch": 2.3, + "learning_rate": 2.7240922868725715e-06, + "loss": 1.1407, + "step": 27314 + }, + { + "epoch": 2.3, + "learning_rate": 2.7234687510382098e-06, + "loss": 1.209, + "step": 27315 + }, + { + "epoch": 2.3, + "learning_rate": 2.722845275324877e-06, + "loss": 1.0809, + "step": 27316 + }, + { + "epoch": 2.3, + "learning_rate": 2.722221859737726e-06, + "loss": 1.0988, + "step": 27317 + }, + { + "epoch": 2.3, + "learning_rate": 2.721598504281907e-06, + "loss": 1.0842, + "step": 27318 + }, + { + "epoch": 2.3, + "learning_rate": 2.720975208962574e-06, + "loss": 1.1392, + "step": 27319 + }, + { + "epoch": 2.3, + "learning_rate": 2.720351973784867e-06, + "loss": 1.1125, + "step": 27320 + }, + { + "epoch": 2.3, + "learning_rate": 2.7197287987539487e-06, + "loss": 1.0961, + "step": 27321 + }, + { + "epoch": 2.3, + "learning_rate": 2.7191056838749574e-06, + "loss": 1.1274, + "step": 27322 + }, + { + "epoch": 2.3, + "learning_rate": 2.718482629153042e-06, + "loss": 1.2646, + "step": 27323 + }, + { + "epoch": 2.3, + "learning_rate": 2.7178596345933595e-06, + "loss": 1.0634, + "step": 27324 + }, + { + "epoch": 2.3, + "learning_rate": 2.717236700201048e-06, + "loss": 1.1075, + "step": 27325 + }, + { + "epoch": 2.3, + "learning_rate": 2.716613825981258e-06, + "loss": 1.1471, + "step": 27326 + }, + { + "epoch": 2.3, + "learning_rate": 2.715991011939135e-06, + "loss": 1.1155, + "step": 27327 + }, + { + "epoch": 2.3, + "learning_rate": 2.7153682580798266e-06, + "loss": 1.1238, + "step": 27328 + }, + { + "epoch": 2.3, + "learning_rate": 2.7147455644084763e-06, + "loss": 1.0642, + "step": 27329 + }, + { + "epoch": 2.3, + "learning_rate": 2.714122930930233e-06, + "loss": 1.1395, + "step": 27330 + }, + { + "epoch": 2.3, + "learning_rate": 2.7135003576502315e-06, + "loss": 1.1813, + "step": 27331 + }, + { + "epoch": 2.3, + "learning_rate": 2.712877844573628e-06, + "loss": 1.0403, + "step": 27332 + }, + { + "epoch": 2.3, + "learning_rate": 2.712255391705557e-06, + "loss": 1.1156, + "step": 27333 + }, + { + "epoch": 2.3, + "learning_rate": 2.7116329990511646e-06, + "loss": 1.089, + "step": 27334 + }, + { + "epoch": 2.3, + "learning_rate": 2.7110106666155934e-06, + "loss": 1.0554, + "step": 27335 + }, + { + "epoch": 2.3, + "learning_rate": 2.7103883944039843e-06, + "loss": 1.2145, + "step": 27336 + }, + { + "epoch": 2.3, + "learning_rate": 2.709766182421479e-06, + "loss": 1.1627, + "step": 27337 + }, + { + "epoch": 2.3, + "learning_rate": 2.7091440306732187e-06, + "loss": 1.0927, + "step": 27338 + }, + { + "epoch": 2.3, + "learning_rate": 2.7085219391643448e-06, + "loss": 1.1333, + "step": 27339 + }, + { + "epoch": 2.3, + "learning_rate": 2.7078999078999947e-06, + "loss": 1.0428, + "step": 27340 + }, + { + "epoch": 2.3, + "learning_rate": 2.7072779368853132e-06, + "loss": 1.2086, + "step": 27341 + }, + { + "epoch": 2.3, + "learning_rate": 2.7066560261254293e-06, + "loss": 1.0401, + "step": 27342 + }, + { + "epoch": 2.3, + "learning_rate": 2.7060341756254936e-06, + "loss": 1.1291, + "step": 27343 + }, + { + "epoch": 2.3, + "learning_rate": 2.7054123853906345e-06, + "loss": 1.1431, + "step": 27344 + }, + { + "epoch": 2.3, + "learning_rate": 2.7047906554259928e-06, + "loss": 1.1195, + "step": 27345 + }, + { + "epoch": 2.3, + "learning_rate": 2.704168985736706e-06, + "loss": 1.177, + "step": 27346 + }, + { + "epoch": 2.3, + "learning_rate": 2.7035473763279097e-06, + "loss": 1.1155, + "step": 27347 + }, + { + "epoch": 2.3, + "learning_rate": 2.70292582720474e-06, + "loss": 1.1579, + "step": 27348 + }, + { + "epoch": 2.3, + "learning_rate": 2.7023043383723326e-06, + "loss": 1.0944, + "step": 27349 + }, + { + "epoch": 2.3, + "learning_rate": 2.701682909835822e-06, + "loss": 1.1821, + "step": 27350 + }, + { + "epoch": 2.3, + "learning_rate": 2.701061541600344e-06, + "loss": 1.0994, + "step": 27351 + }, + { + "epoch": 2.3, + "learning_rate": 2.7004402336710332e-06, + "loss": 1.1025, + "step": 27352 + }, + { + "epoch": 2.3, + "learning_rate": 2.6998189860530167e-06, + "loss": 1.1506, + "step": 27353 + }, + { + "epoch": 2.3, + "learning_rate": 2.6991977987514374e-06, + "loss": 1.0412, + "step": 27354 + }, + { + "epoch": 2.3, + "learning_rate": 2.6985766717714192e-06, + "loss": 1.0558, + "step": 27355 + }, + { + "epoch": 2.3, + "learning_rate": 2.6979556051180968e-06, + "loss": 1.1655, + "step": 27356 + }, + { + "epoch": 2.3, + "learning_rate": 2.6973345987966026e-06, + "loss": 1.0711, + "step": 27357 + }, + { + "epoch": 2.3, + "learning_rate": 2.6967136528120707e-06, + "loss": 1.2028, + "step": 27358 + }, + { + "epoch": 2.3, + "learning_rate": 2.6960927671696213e-06, + "loss": 1.1051, + "step": 27359 + }, + { + "epoch": 2.3, + "learning_rate": 2.6954719418743947e-06, + "loss": 1.1885, + "step": 27360 + }, + { + "epoch": 2.3, + "learning_rate": 2.6948511769315187e-06, + "loss": 1.1643, + "step": 27361 + }, + { + "epoch": 2.3, + "learning_rate": 2.6942304723461154e-06, + "loss": 1.1192, + "step": 27362 + }, + { + "epoch": 2.3, + "learning_rate": 2.6936098281233238e-06, + "loss": 1.1146, + "step": 27363 + }, + { + "epoch": 2.3, + "learning_rate": 2.6929892442682628e-06, + "loss": 1.1897, + "step": 27364 + }, + { + "epoch": 2.3, + "learning_rate": 2.6923687207860625e-06, + "loss": 1.1844, + "step": 27365 + }, + { + "epoch": 2.3, + "learning_rate": 2.691748257681851e-06, + "loss": 1.2057, + "step": 27366 + }, + { + "epoch": 2.3, + "learning_rate": 2.6911278549607556e-06, + "loss": 1.0988, + "step": 27367 + }, + { + "epoch": 2.3, + "learning_rate": 2.690507512627899e-06, + "loss": 1.0701, + "step": 27368 + }, + { + "epoch": 2.3, + "learning_rate": 2.689887230688413e-06, + "loss": 1.0747, + "step": 27369 + }, + { + "epoch": 2.3, + "learning_rate": 2.6892670091474116e-06, + "loss": 1.1644, + "step": 27370 + }, + { + "epoch": 2.3, + "learning_rate": 2.6886468480100323e-06, + "loss": 1.2352, + "step": 27371 + }, + { + "epoch": 2.3, + "learning_rate": 2.6880267472813903e-06, + "loss": 1.1494, + "step": 27372 + }, + { + "epoch": 2.3, + "learning_rate": 2.6874067069666076e-06, + "loss": 1.2037, + "step": 27373 + }, + { + "epoch": 2.3, + "learning_rate": 2.686786727070817e-06, + "loss": 1.1974, + "step": 27374 + }, + { + "epoch": 2.3, + "learning_rate": 2.686166807599132e-06, + "loss": 1.1089, + "step": 27375 + }, + { + "epoch": 2.3, + "learning_rate": 2.6855469485566777e-06, + "loss": 1.1546, + "step": 27376 + }, + { + "epoch": 2.3, + "learning_rate": 2.6849271499485753e-06, + "loss": 1.1125, + "step": 27377 + }, + { + "epoch": 2.3, + "learning_rate": 2.684307411779946e-06, + "loss": 1.0788, + "step": 27378 + }, + { + "epoch": 2.31, + "learning_rate": 2.683687734055911e-06, + "loss": 1.0854, + "step": 27379 + }, + { + "epoch": 2.31, + "learning_rate": 2.683068116781591e-06, + "loss": 1.1199, + "step": 27380 + }, + { + "epoch": 2.31, + "learning_rate": 2.6824485599620986e-06, + "loss": 1.1052, + "step": 27381 + }, + { + "epoch": 2.31, + "learning_rate": 2.6818290636025644e-06, + "loss": 1.1104, + "step": 27382 + }, + { + "epoch": 2.31, + "learning_rate": 2.681209627708097e-06, + "loss": 1.1885, + "step": 27383 + }, + { + "epoch": 2.31, + "learning_rate": 2.680590252283819e-06, + "loss": 1.0538, + "step": 27384 + }, + { + "epoch": 2.31, + "learning_rate": 2.6799709373348458e-06, + "loss": 1.2285, + "step": 27385 + }, + { + "epoch": 2.31, + "learning_rate": 2.6793516828662957e-06, + "loss": 1.1442, + "step": 27386 + }, + { + "epoch": 2.31, + "learning_rate": 2.6787324888832842e-06, + "loss": 1.1075, + "step": 27387 + }, + { + "epoch": 2.31, + "learning_rate": 2.67811335539093e-06, + "loss": 1.1352, + "step": 27388 + }, + { + "epoch": 2.31, + "learning_rate": 2.677494282394345e-06, + "loss": 1.1539, + "step": 27389 + }, + { + "epoch": 2.31, + "learning_rate": 2.6768752698986458e-06, + "loss": 1.1175, + "step": 27390 + }, + { + "epoch": 2.31, + "learning_rate": 2.6762563179089506e-06, + "loss": 1.1643, + "step": 27391 + }, + { + "epoch": 2.31, + "learning_rate": 2.675637426430363e-06, + "loss": 1.0576, + "step": 27392 + }, + { + "epoch": 2.31, + "learning_rate": 2.6750185954680086e-06, + "loss": 1.1287, + "step": 27393 + }, + { + "epoch": 2.31, + "learning_rate": 2.674399825026993e-06, + "loss": 1.1459, + "step": 27394 + }, + { + "epoch": 2.31, + "learning_rate": 2.673781115112429e-06, + "loss": 1.1731, + "step": 27395 + }, + { + "epoch": 2.31, + "learning_rate": 2.673162465729432e-06, + "loss": 1.1086, + "step": 27396 + }, + { + "epoch": 2.31, + "learning_rate": 2.6725438768831104e-06, + "loss": 1.1083, + "step": 27397 + }, + { + "epoch": 2.31, + "learning_rate": 2.6719253485785767e-06, + "loss": 1.1097, + "step": 27398 + }, + { + "epoch": 2.31, + "learning_rate": 2.6713068808209406e-06, + "loss": 1.0802, + "step": 27399 + }, + { + "epoch": 2.31, + "learning_rate": 2.670688473615313e-06, + "loss": 1.1072, + "step": 27400 + }, + { + "epoch": 2.31, + "learning_rate": 2.670070126966803e-06, + "loss": 1.166, + "step": 27401 + }, + { + "epoch": 2.31, + "learning_rate": 2.6694518408805227e-06, + "loss": 1.0934, + "step": 27402 + }, + { + "epoch": 2.31, + "learning_rate": 2.668833615361571e-06, + "loss": 1.0727, + "step": 27403 + }, + { + "epoch": 2.31, + "learning_rate": 2.6682154504150682e-06, + "loss": 1.0911, + "step": 27404 + }, + { + "epoch": 2.31, + "learning_rate": 2.6675973460461125e-06, + "loss": 1.1261, + "step": 27405 + }, + { + "epoch": 2.31, + "learning_rate": 2.6669793022598144e-06, + "loss": 1.1904, + "step": 27406 + }, + { + "epoch": 2.31, + "learning_rate": 2.6663613190612803e-06, + "loss": 1.1369, + "step": 27407 + }, + { + "epoch": 2.31, + "learning_rate": 2.665743396455619e-06, + "loss": 1.1996, + "step": 27408 + }, + { + "epoch": 2.31, + "learning_rate": 2.6651255344479277e-06, + "loss": 1.1625, + "step": 27409 + }, + { + "epoch": 2.31, + "learning_rate": 2.664507733043319e-06, + "loss": 1.1142, + "step": 27410 + }, + { + "epoch": 2.31, + "learning_rate": 2.6638899922468984e-06, + "loss": 1.1509, + "step": 27411 + }, + { + "epoch": 2.31, + "learning_rate": 2.6632723120637603e-06, + "loss": 1.0442, + "step": 27412 + }, + { + "epoch": 2.31, + "learning_rate": 2.6626546924990205e-06, + "loss": 1.0735, + "step": 27413 + }, + { + "epoch": 2.31, + "learning_rate": 2.662037133557772e-06, + "loss": 1.0697, + "step": 27414 + }, + { + "epoch": 2.31, + "learning_rate": 2.6614196352451227e-06, + "loss": 1.2273, + "step": 27415 + }, + { + "epoch": 2.31, + "learning_rate": 2.6608021975661723e-06, + "loss": 1.2052, + "step": 27416 + }, + { + "epoch": 2.31, + "learning_rate": 2.6601848205260227e-06, + "loss": 1.0927, + "step": 27417 + }, + { + "epoch": 2.31, + "learning_rate": 2.6595675041297754e-06, + "loss": 1.1629, + "step": 27418 + }, + { + "epoch": 2.31, + "learning_rate": 2.6589502483825334e-06, + "loss": 1.1, + "step": 27419 + }, + { + "epoch": 2.31, + "learning_rate": 2.658333053289388e-06, + "loss": 1.117, + "step": 27420 + }, + { + "epoch": 2.31, + "learning_rate": 2.6577159188554515e-06, + "loss": 1.1135, + "step": 27421 + }, + { + "epoch": 2.31, + "learning_rate": 2.6570988450858115e-06, + "loss": 1.1339, + "step": 27422 + }, + { + "epoch": 2.31, + "learning_rate": 2.656481831985569e-06, + "loss": 1.0602, + "step": 27423 + }, + { + "epoch": 2.31, + "learning_rate": 2.6558648795598287e-06, + "loss": 1.0956, + "step": 27424 + }, + { + "epoch": 2.31, + "learning_rate": 2.6552479878136807e-06, + "loss": 1.0421, + "step": 27425 + }, + { + "epoch": 2.31, + "learning_rate": 2.6546311567522244e-06, + "loss": 1.0548, + "step": 27426 + }, + { + "epoch": 2.31, + "learning_rate": 2.654014386380557e-06, + "loss": 1.1473, + "step": 27427 + }, + { + "epoch": 2.31, + "learning_rate": 2.653397676703773e-06, + "loss": 1.0967, + "step": 27428 + }, + { + "epoch": 2.31, + "learning_rate": 2.6527810277269683e-06, + "loss": 1.087, + "step": 27429 + }, + { + "epoch": 2.31, + "learning_rate": 2.6521644394552413e-06, + "loss": 1.1428, + "step": 27430 + }, + { + "epoch": 2.31, + "learning_rate": 2.6515479118936784e-06, + "loss": 1.0419, + "step": 27431 + }, + { + "epoch": 2.31, + "learning_rate": 2.6509314450473834e-06, + "loss": 1.1517, + "step": 27432 + }, + { + "epoch": 2.31, + "learning_rate": 2.6503150389214427e-06, + "loss": 1.0629, + "step": 27433 + }, + { + "epoch": 2.31, + "learning_rate": 2.6496986935209503e-06, + "loss": 1.0893, + "step": 27434 + }, + { + "epoch": 2.31, + "learning_rate": 2.6490824088510003e-06, + "loss": 1.1189, + "step": 27435 + }, + { + "epoch": 2.31, + "learning_rate": 2.648466184916684e-06, + "loss": 1.1661, + "step": 27436 + }, + { + "epoch": 2.31, + "learning_rate": 2.647850021723093e-06, + "loss": 1.1792, + "step": 27437 + }, + { + "epoch": 2.31, + "learning_rate": 2.647233919275317e-06, + "loss": 1.1239, + "step": 27438 + }, + { + "epoch": 2.31, + "learning_rate": 2.646617877578449e-06, + "loss": 1.1068, + "step": 27439 + }, + { + "epoch": 2.31, + "learning_rate": 2.6460018966375768e-06, + "loss": 1.192, + "step": 27440 + }, + { + "epoch": 2.31, + "learning_rate": 2.645385976457794e-06, + "loss": 1.0446, + "step": 27441 + }, + { + "epoch": 2.31, + "learning_rate": 2.6447701170441795e-06, + "loss": 1.1917, + "step": 27442 + }, + { + "epoch": 2.31, + "learning_rate": 2.644154318401834e-06, + "loss": 1.1423, + "step": 27443 + }, + { + "epoch": 2.31, + "learning_rate": 2.6435385805358372e-06, + "loss": 1.1137, + "step": 27444 + }, + { + "epoch": 2.31, + "learning_rate": 2.642922903451279e-06, + "loss": 1.1131, + "step": 27445 + }, + { + "epoch": 2.31, + "learning_rate": 2.6423072871532464e-06, + "loss": 1.0793, + "step": 27446 + }, + { + "epoch": 2.31, + "learning_rate": 2.641691731646826e-06, + "loss": 1.0766, + "step": 27447 + }, + { + "epoch": 2.31, + "learning_rate": 2.6410762369371024e-06, + "loss": 1.1245, + "step": 27448 + }, + { + "epoch": 2.31, + "learning_rate": 2.6404608030291635e-06, + "loss": 1.1236, + "step": 27449 + }, + { + "epoch": 2.31, + "learning_rate": 2.639845429928092e-06, + "loss": 1.1105, + "step": 27450 + }, + { + "epoch": 2.31, + "learning_rate": 2.6392301176389724e-06, + "loss": 1.1562, + "step": 27451 + }, + { + "epoch": 2.31, + "learning_rate": 2.638614866166893e-06, + "loss": 1.1529, + "step": 27452 + }, + { + "epoch": 2.31, + "learning_rate": 2.637999675516928e-06, + "loss": 1.1055, + "step": 27453 + }, + { + "epoch": 2.31, + "learning_rate": 2.6373845456941716e-06, + "loss": 1.1195, + "step": 27454 + }, + { + "epoch": 2.31, + "learning_rate": 2.6367694767036965e-06, + "loss": 1.0859, + "step": 27455 + }, + { + "epoch": 2.31, + "learning_rate": 2.6361544685505903e-06, + "loss": 1.1013, + "step": 27456 + }, + { + "epoch": 2.31, + "learning_rate": 2.6355395212399315e-06, + "loss": 1.1049, + "step": 27457 + }, + { + "epoch": 2.31, + "learning_rate": 2.634924634776805e-06, + "loss": 1.0778, + "step": 27458 + }, + { + "epoch": 2.31, + "learning_rate": 2.634309809166282e-06, + "loss": 1.1621, + "step": 27459 + }, + { + "epoch": 2.31, + "learning_rate": 2.6336950444134526e-06, + "loss": 1.0726, + "step": 27460 + }, + { + "epoch": 2.31, + "learning_rate": 2.6330803405233953e-06, + "loss": 1.1125, + "step": 27461 + }, + { + "epoch": 2.31, + "learning_rate": 2.6324656975011786e-06, + "loss": 1.1132, + "step": 27462 + }, + { + "epoch": 2.31, + "learning_rate": 2.6318511153518956e-06, + "loss": 1.1177, + "step": 27463 + }, + { + "epoch": 2.31, + "learning_rate": 2.631236594080613e-06, + "loss": 1.1252, + "step": 27464 + }, + { + "epoch": 2.31, + "learning_rate": 2.630622133692413e-06, + "loss": 1.0899, + "step": 27465 + }, + { + "epoch": 2.31, + "learning_rate": 2.63000773419237e-06, + "loss": 1.083, + "step": 27466 + }, + { + "epoch": 2.31, + "learning_rate": 2.6293933955855632e-06, + "loss": 1.0263, + "step": 27467 + }, + { + "epoch": 2.31, + "learning_rate": 2.6287791178770663e-06, + "loss": 1.1575, + "step": 27468 + }, + { + "epoch": 2.31, + "learning_rate": 2.6281649010719577e-06, + "loss": 1.133, + "step": 27469 + }, + { + "epoch": 2.31, + "learning_rate": 2.6275507451753057e-06, + "loss": 1.1152, + "step": 27470 + }, + { + "epoch": 2.31, + "learning_rate": 2.6269366501921934e-06, + "loss": 1.1552, + "step": 27471 + }, + { + "epoch": 2.31, + "learning_rate": 2.6263226161276877e-06, + "loss": 1.1504, + "step": 27472 + }, + { + "epoch": 2.31, + "learning_rate": 2.6257086429868615e-06, + "loss": 1.097, + "step": 27473 + }, + { + "epoch": 2.31, + "learning_rate": 2.6250947307747965e-06, + "loss": 1.1212, + "step": 27474 + }, + { + "epoch": 2.31, + "learning_rate": 2.624480879496556e-06, + "loss": 1.1156, + "step": 27475 + }, + { + "epoch": 2.31, + "learning_rate": 2.623867089157216e-06, + "loss": 1.1718, + "step": 27476 + }, + { + "epoch": 2.31, + "learning_rate": 2.623253359761846e-06, + "loss": 1.1273, + "step": 27477 + }, + { + "epoch": 2.31, + "learning_rate": 2.6226396913155174e-06, + "loss": 1.1288, + "step": 27478 + }, + { + "epoch": 2.31, + "learning_rate": 2.6220260838233003e-06, + "loss": 1.0932, + "step": 27479 + }, + { + "epoch": 2.31, + "learning_rate": 2.621412537290269e-06, + "loss": 1.0732, + "step": 27480 + }, + { + "epoch": 2.31, + "learning_rate": 2.620799051721482e-06, + "loss": 1.1167, + "step": 27481 + }, + { + "epoch": 2.31, + "learning_rate": 2.6201856271220206e-06, + "loss": 1.1021, + "step": 27482 + }, + { + "epoch": 2.31, + "learning_rate": 2.6195722634969455e-06, + "loss": 1.1204, + "step": 27483 + }, + { + "epoch": 2.31, + "learning_rate": 2.6189589608513253e-06, + "loss": 1.0364, + "step": 27484 + }, + { + "epoch": 2.31, + "learning_rate": 2.6183457191902295e-06, + "loss": 1.1204, + "step": 27485 + }, + { + "epoch": 2.31, + "learning_rate": 2.6177325385187226e-06, + "loss": 1.0699, + "step": 27486 + }, + { + "epoch": 2.31, + "learning_rate": 2.6171194188418726e-06, + "loss": 1.0552, + "step": 27487 + }, + { + "epoch": 2.31, + "learning_rate": 2.6165063601647454e-06, + "loss": 1.153, + "step": 27488 + }, + { + "epoch": 2.31, + "learning_rate": 2.6158933624924053e-06, + "loss": 1.1249, + "step": 27489 + }, + { + "epoch": 2.31, + "learning_rate": 2.6152804258299168e-06, + "loss": 1.0403, + "step": 27490 + }, + { + "epoch": 2.31, + "learning_rate": 2.6146675501823484e-06, + "loss": 1.095, + "step": 27491 + }, + { + "epoch": 2.31, + "learning_rate": 2.614054735554754e-06, + "loss": 1.0676, + "step": 27492 + }, + { + "epoch": 2.31, + "learning_rate": 2.6134419819522094e-06, + "loss": 1.1316, + "step": 27493 + }, + { + "epoch": 2.31, + "learning_rate": 2.612829289379769e-06, + "loss": 1.0682, + "step": 27494 + }, + { + "epoch": 2.31, + "learning_rate": 2.6122166578424957e-06, + "loss": 1.1299, + "step": 27495 + }, + { + "epoch": 2.31, + "learning_rate": 2.611604087345454e-06, + "loss": 1.1407, + "step": 27496 + }, + { + "epoch": 2.32, + "learning_rate": 2.6109915778937035e-06, + "loss": 1.0928, + "step": 27497 + }, + { + "epoch": 2.32, + "learning_rate": 2.6103791294923053e-06, + "loss": 1.096, + "step": 27498 + }, + { + "epoch": 2.32, + "learning_rate": 2.6097667421463193e-06, + "loss": 1.1241, + "step": 27499 + }, + { + "epoch": 2.32, + "learning_rate": 2.6091544158608064e-06, + "loss": 1.1152, + "step": 27500 + }, + { + "epoch": 2.32, + "learning_rate": 2.608542150640825e-06, + "loss": 1.1065, + "step": 27501 + }, + { + "epoch": 2.32, + "learning_rate": 2.6079299464914363e-06, + "loss": 1.109, + "step": 27502 + }, + { + "epoch": 2.32, + "learning_rate": 2.6073178034176905e-06, + "loss": 1.032, + "step": 27503 + }, + { + "epoch": 2.32, + "learning_rate": 2.6067057214246573e-06, + "loss": 1.0871, + "step": 27504 + }, + { + "epoch": 2.32, + "learning_rate": 2.6060937005173838e-06, + "loss": 1.0935, + "step": 27505 + }, + { + "epoch": 2.32, + "learning_rate": 2.6054817407009305e-06, + "loss": 1.1379, + "step": 27506 + }, + { + "epoch": 2.32, + "learning_rate": 2.6048698419803542e-06, + "loss": 1.0915, + "step": 27507 + }, + { + "epoch": 2.32, + "learning_rate": 2.6042580043607123e-06, + "loss": 1.0971, + "step": 27508 + }, + { + "epoch": 2.32, + "learning_rate": 2.6036462278470533e-06, + "loss": 1.1426, + "step": 27509 + }, + { + "epoch": 2.32, + "learning_rate": 2.603034512444438e-06, + "loss": 1.1662, + "step": 27510 + }, + { + "epoch": 2.32, + "learning_rate": 2.6024228581579225e-06, + "loss": 1.1178, + "step": 27511 + }, + { + "epoch": 2.32, + "learning_rate": 2.60181126499255e-06, + "loss": 1.1384, + "step": 27512 + }, + { + "epoch": 2.32, + "learning_rate": 2.601199732953387e-06, + "loss": 1.1477, + "step": 27513 + }, + { + "epoch": 2.32, + "learning_rate": 2.600588262045477e-06, + "loss": 1.0989, + "step": 27514 + }, + { + "epoch": 2.32, + "learning_rate": 2.5999768522738745e-06, + "loss": 1.1209, + "step": 27515 + }, + { + "epoch": 2.32, + "learning_rate": 2.5993655036436315e-06, + "loss": 1.1171, + "step": 27516 + }, + { + "epoch": 2.32, + "learning_rate": 2.5987542161598e-06, + "loss": 1.1264, + "step": 27517 + }, + { + "epoch": 2.32, + "learning_rate": 2.598142989827429e-06, + "loss": 1.0755, + "step": 27518 + }, + { + "epoch": 2.32, + "learning_rate": 2.5975318246515737e-06, + "loss": 1.1549, + "step": 27519 + }, + { + "epoch": 2.32, + "learning_rate": 2.5969207206372714e-06, + "loss": 1.115, + "step": 27520 + }, + { + "epoch": 2.32, + "learning_rate": 2.596309677789587e-06, + "loss": 1.0204, + "step": 27521 + }, + { + "epoch": 2.32, + "learning_rate": 2.5956986961135587e-06, + "loss": 1.1252, + "step": 27522 + }, + { + "epoch": 2.32, + "learning_rate": 2.5950877756142333e-06, + "loss": 1.1148, + "step": 27523 + }, + { + "epoch": 2.32, + "learning_rate": 2.59447691629667e-06, + "loss": 1.2102, + "step": 27524 + }, + { + "epoch": 2.32, + "learning_rate": 2.593866118165904e-06, + "loss": 1.1646, + "step": 27525 + }, + { + "epoch": 2.32, + "learning_rate": 2.593255381226988e-06, + "loss": 1.1113, + "step": 27526 + }, + { + "epoch": 2.32, + "learning_rate": 2.5926447054849656e-06, + "loss": 1.1054, + "step": 27527 + }, + { + "epoch": 2.32, + "learning_rate": 2.5920340909448827e-06, + "loss": 1.1159, + "step": 27528 + }, + { + "epoch": 2.32, + "learning_rate": 2.5914235376117857e-06, + "loss": 1.1256, + "step": 27529 + }, + { + "epoch": 2.32, + "learning_rate": 2.5908130454907222e-06, + "loss": 1.0636, + "step": 27530 + }, + { + "epoch": 2.32, + "learning_rate": 2.5902026145867255e-06, + "loss": 1.1656, + "step": 27531 + }, + { + "epoch": 2.32, + "learning_rate": 2.5895922449048527e-06, + "loss": 1.2169, + "step": 27532 + }, + { + "epoch": 2.32, + "learning_rate": 2.588981936450137e-06, + "loss": 1.0705, + "step": 27533 + }, + { + "epoch": 2.32, + "learning_rate": 2.5883716892276246e-06, + "loss": 1.1504, + "step": 27534 + }, + { + "epoch": 2.32, + "learning_rate": 2.5877615032423575e-06, + "loss": 1.0913, + "step": 27535 + }, + { + "epoch": 2.32, + "learning_rate": 2.5871513784993763e-06, + "loss": 1.1073, + "step": 27536 + }, + { + "epoch": 2.32, + "learning_rate": 2.5865413150037246e-06, + "loss": 1.1308, + "step": 27537 + }, + { + "epoch": 2.32, + "learning_rate": 2.5859313127604393e-06, + "loss": 1.0537, + "step": 27538 + }, + { + "epoch": 2.32, + "learning_rate": 2.5853213717745627e-06, + "loss": 1.0758, + "step": 27539 + }, + { + "epoch": 2.32, + "learning_rate": 2.584711492051134e-06, + "loss": 1.1299, + "step": 27540 + }, + { + "epoch": 2.32, + "learning_rate": 2.584101673595196e-06, + "loss": 1.1607, + "step": 27541 + }, + { + "epoch": 2.32, + "learning_rate": 2.583491916411777e-06, + "loss": 1.1167, + "step": 27542 + }, + { + "epoch": 2.32, + "learning_rate": 2.5828822205059266e-06, + "loss": 1.1696, + "step": 27543 + }, + { + "epoch": 2.32, + "learning_rate": 2.5822725858826746e-06, + "loss": 1.0976, + "step": 27544 + }, + { + "epoch": 2.32, + "learning_rate": 2.581663012547061e-06, + "loss": 1.1587, + "step": 27545 + }, + { + "epoch": 2.32, + "learning_rate": 2.5810535005041206e-06, + "loss": 1.1252, + "step": 27546 + }, + { + "epoch": 2.32, + "learning_rate": 2.580444049758891e-06, + "loss": 1.1294, + "step": 27547 + }, + { + "epoch": 2.32, + "learning_rate": 2.579834660316408e-06, + "loss": 1.0962, + "step": 27548 + }, + { + "epoch": 2.32, + "learning_rate": 2.5792253321817053e-06, + "loss": 1.194, + "step": 27549 + }, + { + "epoch": 2.32, + "learning_rate": 2.578616065359818e-06, + "loss": 1.128, + "step": 27550 + }, + { + "epoch": 2.32, + "learning_rate": 2.5780068598557795e-06, + "loss": 1.0817, + "step": 27551 + }, + { + "epoch": 2.32, + "learning_rate": 2.5773977156746265e-06, + "loss": 0.9544, + "step": 27552 + }, + { + "epoch": 2.32, + "learning_rate": 2.576788632821383e-06, + "loss": 1.1427, + "step": 27553 + }, + { + "epoch": 2.32, + "learning_rate": 2.576179611301094e-06, + "loss": 1.0887, + "step": 27554 + }, + { + "epoch": 2.32, + "learning_rate": 2.5755706511187827e-06, + "loss": 1.0942, + "step": 27555 + }, + { + "epoch": 2.32, + "learning_rate": 2.574961752279481e-06, + "loss": 1.1784, + "step": 27556 + }, + { + "epoch": 2.32, + "learning_rate": 2.574352914788223e-06, + "loss": 1.1445, + "step": 27557 + }, + { + "epoch": 2.32, + "learning_rate": 2.5737441386500404e-06, + "loss": 1.1458, + "step": 27558 + }, + { + "epoch": 2.32, + "learning_rate": 2.5731354238699548e-06, + "loss": 1.0602, + "step": 27559 + }, + { + "epoch": 2.32, + "learning_rate": 2.5725267704530045e-06, + "loss": 1.1156, + "step": 27560 + }, + { + "epoch": 2.32, + "learning_rate": 2.571918178404217e-06, + "loss": 1.1179, + "step": 27561 + }, + { + "epoch": 2.32, + "learning_rate": 2.5713096477286137e-06, + "loss": 1.1877, + "step": 27562 + }, + { + "epoch": 2.32, + "learning_rate": 2.570701178431233e-06, + "loss": 1.0333, + "step": 27563 + }, + { + "epoch": 2.32, + "learning_rate": 2.570092770517094e-06, + "loss": 1.1169, + "step": 27564 + }, + { + "epoch": 2.32, + "learning_rate": 2.5694844239912264e-06, + "loss": 1.1052, + "step": 27565 + }, + { + "epoch": 2.32, + "learning_rate": 2.5688761388586556e-06, + "loss": 1.1549, + "step": 27566 + }, + { + "epoch": 2.32, + "learning_rate": 2.5682679151244107e-06, + "loss": 1.1362, + "step": 27567 + }, + { + "epoch": 2.32, + "learning_rate": 2.5676597527935133e-06, + "loss": 1.11, + "step": 27568 + }, + { + "epoch": 2.32, + "learning_rate": 2.567051651870992e-06, + "loss": 1.1477, + "step": 27569 + }, + { + "epoch": 2.32, + "learning_rate": 2.5664436123618643e-06, + "loss": 1.0761, + "step": 27570 + }, + { + "epoch": 2.32, + "learning_rate": 2.565835634271163e-06, + "loss": 1.0924, + "step": 27571 + }, + { + "epoch": 2.32, + "learning_rate": 2.5652277176039054e-06, + "loss": 1.1016, + "step": 27572 + }, + { + "epoch": 2.32, + "learning_rate": 2.564619862365112e-06, + "loss": 1.0861, + "step": 27573 + }, + { + "epoch": 2.32, + "learning_rate": 2.5640120685598147e-06, + "loss": 1.1667, + "step": 27574 + }, + { + "epoch": 2.32, + "learning_rate": 2.5634043361930273e-06, + "loss": 1.1714, + "step": 27575 + }, + { + "epoch": 2.32, + "learning_rate": 2.562796665269773e-06, + "loss": 1.0364, + "step": 27576 + }, + { + "epoch": 2.32, + "learning_rate": 2.5621890557950726e-06, + "loss": 1.0921, + "step": 27577 + }, + { + "epoch": 2.32, + "learning_rate": 2.561581507773947e-06, + "loss": 1.1085, + "step": 27578 + }, + { + "epoch": 2.32, + "learning_rate": 2.5609740212114154e-06, + "loss": 1.1115, + "step": 27579 + }, + { + "epoch": 2.32, + "learning_rate": 2.5603665961125e-06, + "loss": 1.0433, + "step": 27580 + }, + { + "epoch": 2.32, + "learning_rate": 2.5597592324822107e-06, + "loss": 1.1108, + "step": 27581 + }, + { + "epoch": 2.32, + "learning_rate": 2.5591519303255786e-06, + "loss": 1.0371, + "step": 27582 + }, + { + "epoch": 2.32, + "learning_rate": 2.558544689647611e-06, + "loss": 1.1852, + "step": 27583 + }, + { + "epoch": 2.32, + "learning_rate": 2.5579375104533287e-06, + "loss": 1.1129, + "step": 27584 + }, + { + "epoch": 2.32, + "learning_rate": 2.557330392747749e-06, + "loss": 1.051, + "step": 27585 + }, + { + "epoch": 2.32, + "learning_rate": 2.556723336535887e-06, + "loss": 1.1875, + "step": 27586 + }, + { + "epoch": 2.32, + "learning_rate": 2.5561163418227585e-06, + "loss": 1.1188, + "step": 27587 + }, + { + "epoch": 2.32, + "learning_rate": 2.5555094086133804e-06, + "loss": 1.1062, + "step": 27588 + }, + { + "epoch": 2.32, + "learning_rate": 2.5549025369127654e-06, + "loss": 1.158, + "step": 27589 + }, + { + "epoch": 2.32, + "learning_rate": 2.554295726725928e-06, + "loss": 1.0488, + "step": 27590 + }, + { + "epoch": 2.32, + "learning_rate": 2.5536889780578856e-06, + "loss": 1.1259, + "step": 27591 + }, + { + "epoch": 2.32, + "learning_rate": 2.553082290913642e-06, + "loss": 1.1302, + "step": 27592 + }, + { + "epoch": 2.32, + "learning_rate": 2.552475665298222e-06, + "loss": 1.1108, + "step": 27593 + }, + { + "epoch": 2.32, + "learning_rate": 2.5518691012166284e-06, + "loss": 1.1407, + "step": 27594 + }, + { + "epoch": 2.32, + "learning_rate": 2.551262598673876e-06, + "loss": 1.0806, + "step": 27595 + }, + { + "epoch": 2.32, + "learning_rate": 2.5506561576749765e-06, + "loss": 1.1635, + "step": 27596 + }, + { + "epoch": 2.32, + "learning_rate": 2.550049778224939e-06, + "loss": 1.1983, + "step": 27597 + }, + { + "epoch": 2.32, + "learning_rate": 2.5494434603287756e-06, + "loss": 1.0943, + "step": 27598 + }, + { + "epoch": 2.32, + "learning_rate": 2.548837203991493e-06, + "loss": 1.1161, + "step": 27599 + }, + { + "epoch": 2.32, + "learning_rate": 2.548231009218104e-06, + "loss": 1.074, + "step": 27600 + }, + { + "epoch": 2.32, + "learning_rate": 2.5476248760136148e-06, + "loss": 1.1639, + "step": 27601 + }, + { + "epoch": 2.32, + "learning_rate": 2.547018804383036e-06, + "loss": 1.1345, + "step": 27602 + }, + { + "epoch": 2.32, + "learning_rate": 2.5464127943313677e-06, + "loss": 1.1851, + "step": 27603 + }, + { + "epoch": 2.32, + "learning_rate": 2.545806845863629e-06, + "loss": 1.1266, + "step": 27604 + }, + { + "epoch": 2.32, + "learning_rate": 2.545200958984816e-06, + "loss": 1.1793, + "step": 27605 + }, + { + "epoch": 2.32, + "learning_rate": 2.5445951336999387e-06, + "loss": 1.1972, + "step": 27606 + }, + { + "epoch": 2.32, + "learning_rate": 2.5439893700140018e-06, + "loss": 1.1692, + "step": 27607 + }, + { + "epoch": 2.32, + "learning_rate": 2.543383667932012e-06, + "loss": 1.184, + "step": 27608 + }, + { + "epoch": 2.32, + "learning_rate": 2.542778027458972e-06, + "loss": 1.0066, + "step": 27609 + }, + { + "epoch": 2.32, + "learning_rate": 2.5421724485998865e-06, + "loss": 1.1316, + "step": 27610 + }, + { + "epoch": 2.32, + "learning_rate": 2.5415669313597624e-06, + "loss": 1.02, + "step": 27611 + }, + { + "epoch": 2.32, + "learning_rate": 2.540961475743594e-06, + "loss": 1.1205, + "step": 27612 + }, + { + "epoch": 2.32, + "learning_rate": 2.540356081756394e-06, + "loss": 1.09, + "step": 27613 + }, + { + "epoch": 2.32, + "learning_rate": 2.539750749403157e-06, + "loss": 1.1035, + "step": 27614 + }, + { + "epoch": 2.32, + "learning_rate": 2.539145478688887e-06, + "loss": 1.1262, + "step": 27615 + }, + { + "epoch": 2.33, + "learning_rate": 2.5385402696185846e-06, + "loss": 1.0655, + "step": 27616 + }, + { + "epoch": 2.33, + "learning_rate": 2.5379351221972503e-06, + "loss": 1.237, + "step": 27617 + }, + { + "epoch": 2.33, + "learning_rate": 2.537330036429885e-06, + "loss": 1.129, + "step": 27618 + }, + { + "epoch": 2.33, + "learning_rate": 2.536725012321489e-06, + "loss": 1.0717, + "step": 27619 + }, + { + "epoch": 2.33, + "learning_rate": 2.5361200498770543e-06, + "loss": 1.1861, + "step": 27620 + }, + { + "epoch": 2.33, + "learning_rate": 2.5355151491015905e-06, + "loss": 1.1979, + "step": 27621 + }, + { + "epoch": 2.33, + "learning_rate": 2.534910310000086e-06, + "loss": 1.0987, + "step": 27622 + }, + { + "epoch": 2.33, + "learning_rate": 2.534305532577539e-06, + "loss": 1.1455, + "step": 27623 + }, + { + "epoch": 2.33, + "learning_rate": 2.5337008168389556e-06, + "loss": 1.1245, + "step": 27624 + }, + { + "epoch": 2.33, + "learning_rate": 2.5330961627893214e-06, + "loss": 1.2124, + "step": 27625 + }, + { + "epoch": 2.33, + "learning_rate": 2.5324915704336373e-06, + "loss": 1.0856, + "step": 27626 + }, + { + "epoch": 2.33, + "learning_rate": 2.531887039776897e-06, + "loss": 1.1304, + "step": 27627 + }, + { + "epoch": 2.33, + "learning_rate": 2.531282570824096e-06, + "loss": 1.1765, + "step": 27628 + }, + { + "epoch": 2.33, + "learning_rate": 2.53067816358023e-06, + "loss": 1.2384, + "step": 27629 + }, + { + "epoch": 2.33, + "learning_rate": 2.5300738180502937e-06, + "loss": 1.0448, + "step": 27630 + }, + { + "epoch": 2.33, + "learning_rate": 2.5294695342392715e-06, + "loss": 1.1062, + "step": 27631 + }, + { + "epoch": 2.33, + "learning_rate": 2.52886531215217e-06, + "loss": 1.1483, + "step": 27632 + }, + { + "epoch": 2.33, + "learning_rate": 2.5282611517939716e-06, + "loss": 1.0563, + "step": 27633 + }, + { + "epoch": 2.33, + "learning_rate": 2.52765705316967e-06, + "loss": 1.0629, + "step": 27634 + }, + { + "epoch": 2.33, + "learning_rate": 2.5270530162842576e-06, + "loss": 1.1285, + "step": 27635 + }, + { + "epoch": 2.33, + "learning_rate": 2.5264490411427246e-06, + "loss": 1.1269, + "step": 27636 + }, + { + "epoch": 2.33, + "learning_rate": 2.5258451277500616e-06, + "loss": 1.0949, + "step": 27637 + }, + { + "epoch": 2.33, + "learning_rate": 2.5252412761112576e-06, + "loss": 1.0562, + "step": 27638 + }, + { + "epoch": 2.33, + "learning_rate": 2.524637486231303e-06, + "loss": 1.0577, + "step": 27639 + }, + { + "epoch": 2.33, + "learning_rate": 2.5240337581151865e-06, + "loss": 1.1086, + "step": 27640 + }, + { + "epoch": 2.33, + "learning_rate": 2.523430091767898e-06, + "loss": 1.0735, + "step": 27641 + }, + { + "epoch": 2.33, + "learning_rate": 2.5228264871944164e-06, + "loss": 1.1047, + "step": 27642 + }, + { + "epoch": 2.33, + "learning_rate": 2.5222229443997428e-06, + "loss": 0.9821, + "step": 27643 + }, + { + "epoch": 2.33, + "learning_rate": 2.5216194633888513e-06, + "loss": 1.1698, + "step": 27644 + }, + { + "epoch": 2.33, + "learning_rate": 2.5210160441667353e-06, + "loss": 1.1193, + "step": 27645 + }, + { + "epoch": 2.33, + "learning_rate": 2.520412686738377e-06, + "loss": 0.9982, + "step": 27646 + }, + { + "epoch": 2.33, + "learning_rate": 2.519809391108764e-06, + "loss": 1.1028, + "step": 27647 + }, + { + "epoch": 2.33, + "learning_rate": 2.5192061572828787e-06, + "loss": 1.1081, + "step": 27648 + }, + { + "epoch": 2.33, + "learning_rate": 2.5186029852657067e-06, + "loss": 1.1406, + "step": 27649 + }, + { + "epoch": 2.33, + "learning_rate": 2.5179998750622314e-06, + "loss": 1.1499, + "step": 27650 + }, + { + "epoch": 2.33, + "learning_rate": 2.5173968266774353e-06, + "loss": 1.1712, + "step": 27651 + }, + { + "epoch": 2.33, + "learning_rate": 2.516793840116304e-06, + "loss": 1.0788, + "step": 27652 + }, + { + "epoch": 2.33, + "learning_rate": 2.5161909153838117e-06, + "loss": 1.1958, + "step": 27653 + }, + { + "epoch": 2.33, + "learning_rate": 2.515588052484951e-06, + "loss": 1.1049, + "step": 27654 + }, + { + "epoch": 2.33, + "learning_rate": 2.5149852514246943e-06, + "loss": 1.156, + "step": 27655 + }, + { + "epoch": 2.33, + "learning_rate": 2.5143825122080246e-06, + "loss": 1.086, + "step": 27656 + }, + { + "epoch": 2.33, + "learning_rate": 2.5137798348399224e-06, + "loss": 1.0529, + "step": 27657 + }, + { + "epoch": 2.33, + "learning_rate": 2.513177219325367e-06, + "loss": 1.1033, + "step": 27658 + }, + { + "epoch": 2.33, + "learning_rate": 2.512574665669337e-06, + "loss": 1.0475, + "step": 27659 + }, + { + "epoch": 2.33, + "learning_rate": 2.511972173876812e-06, + "loss": 1.1404, + "step": 27660 + }, + { + "epoch": 2.33, + "learning_rate": 2.5113697439527696e-06, + "loss": 1.0388, + "step": 27661 + }, + { + "epoch": 2.33, + "learning_rate": 2.5107673759021876e-06, + "loss": 1.1529, + "step": 27662 + }, + { + "epoch": 2.33, + "learning_rate": 2.510165069730044e-06, + "loss": 1.128, + "step": 27663 + }, + { + "epoch": 2.33, + "learning_rate": 2.5095628254413086e-06, + "loss": 1.1475, + "step": 27664 + }, + { + "epoch": 2.33, + "learning_rate": 2.5089606430409685e-06, + "loss": 1.167, + "step": 27665 + }, + { + "epoch": 2.33, + "learning_rate": 2.508358522533989e-06, + "loss": 1.1456, + "step": 27666 + }, + { + "epoch": 2.33, + "learning_rate": 2.5077564639253493e-06, + "loss": 1.0023, + "step": 27667 + }, + { + "epoch": 2.33, + "learning_rate": 2.5071544672200244e-06, + "loss": 1.0257, + "step": 27668 + }, + { + "epoch": 2.33, + "learning_rate": 2.50655253242299e-06, + "loss": 1.0994, + "step": 27669 + }, + { + "epoch": 2.33, + "learning_rate": 2.5059506595392102e-06, + "loss": 1.1371, + "step": 27670 + }, + { + "epoch": 2.33, + "learning_rate": 2.505348848573671e-06, + "loss": 1.0461, + "step": 27671 + }, + { + "epoch": 2.33, + "learning_rate": 2.504747099531335e-06, + "loss": 1.1062, + "step": 27672 + }, + { + "epoch": 2.33, + "learning_rate": 2.5041454124171738e-06, + "loss": 1.097, + "step": 27673 + }, + { + "epoch": 2.33, + "learning_rate": 2.503543787236169e-06, + "loss": 1.1198, + "step": 27674 + }, + { + "epoch": 2.33, + "learning_rate": 2.502942223993281e-06, + "loss": 1.1402, + "step": 27675 + }, + { + "epoch": 2.33, + "learning_rate": 2.5023407226934835e-06, + "loss": 1.1178, + "step": 27676 + }, + { + "epoch": 2.33, + "learning_rate": 2.5017392833417477e-06, + "loss": 1.122, + "step": 27677 + }, + { + "epoch": 2.33, + "learning_rate": 2.5011379059430406e-06, + "loss": 1.1566, + "step": 27678 + }, + { + "epoch": 2.33, + "learning_rate": 2.5005365905023325e-06, + "loss": 1.0971, + "step": 27679 + }, + { + "epoch": 2.33, + "learning_rate": 2.4999353370245938e-06, + "loss": 1.0621, + "step": 27680 + }, + { + "epoch": 2.33, + "learning_rate": 2.4993341455147835e-06, + "loss": 1.1628, + "step": 27681 + }, + { + "epoch": 2.33, + "learning_rate": 2.498733015977882e-06, + "loss": 1.0929, + "step": 27682 + }, + { + "epoch": 2.33, + "learning_rate": 2.4981319484188448e-06, + "loss": 1.1304, + "step": 27683 + }, + { + "epoch": 2.33, + "learning_rate": 2.4975309428426407e-06, + "loss": 1.1156, + "step": 27684 + }, + { + "epoch": 2.33, + "learning_rate": 2.4969299992542427e-06, + "loss": 1.1534, + "step": 27685 + }, + { + "epoch": 2.33, + "learning_rate": 2.496329117658607e-06, + "loss": 1.1262, + "step": 27686 + }, + { + "epoch": 2.33, + "learning_rate": 2.495728298060701e-06, + "loss": 1.0712, + "step": 27687 + }, + { + "epoch": 2.33, + "learning_rate": 2.4951275404654908e-06, + "loss": 1.1336, + "step": 27688 + }, + { + "epoch": 2.33, + "learning_rate": 2.494526844877938e-06, + "loss": 1.1487, + "step": 27689 + }, + { + "epoch": 2.33, + "learning_rate": 2.493926211303006e-06, + "loss": 1.0384, + "step": 27690 + }, + { + "epoch": 2.33, + "learning_rate": 2.4933256397456616e-06, + "loss": 1.1095, + "step": 27691 + }, + { + "epoch": 2.33, + "learning_rate": 2.4927251302108568e-06, + "loss": 1.0566, + "step": 27692 + }, + { + "epoch": 2.33, + "learning_rate": 2.492124682703566e-06, + "loss": 1.1446, + "step": 27693 + }, + { + "epoch": 2.33, + "learning_rate": 2.4915242972287414e-06, + "loss": 1.1203, + "step": 27694 + }, + { + "epoch": 2.33, + "learning_rate": 2.490923973791345e-06, + "loss": 1.1361, + "step": 27695 + }, + { + "epoch": 2.33, + "learning_rate": 2.4903237123963387e-06, + "loss": 1.1248, + "step": 27696 + }, + { + "epoch": 2.33, + "learning_rate": 2.489723513048681e-06, + "loss": 1.1064, + "step": 27697 + }, + { + "epoch": 2.33, + "learning_rate": 2.4891233757533306e-06, + "loss": 1.1348, + "step": 27698 + }, + { + "epoch": 2.33, + "learning_rate": 2.4885233005152474e-06, + "loss": 1.0739, + "step": 27699 + }, + { + "epoch": 2.33, + "learning_rate": 2.487923287339389e-06, + "loss": 1.038, + "step": 27700 + }, + { + "epoch": 2.33, + "learning_rate": 2.4873233362307113e-06, + "loss": 1.1633, + "step": 27701 + }, + { + "epoch": 2.33, + "learning_rate": 2.486723447194177e-06, + "loss": 1.1051, + "step": 27702 + }, + { + "epoch": 2.33, + "learning_rate": 2.486123620234732e-06, + "loss": 1.0428, + "step": 27703 + }, + { + "epoch": 2.33, + "learning_rate": 2.4855238553573437e-06, + "loss": 1.1447, + "step": 27704 + }, + { + "epoch": 2.33, + "learning_rate": 2.4849241525669586e-06, + "loss": 1.1567, + "step": 27705 + }, + { + "epoch": 2.33, + "learning_rate": 2.4843245118685367e-06, + "loss": 1.0545, + "step": 27706 + }, + { + "epoch": 2.33, + "learning_rate": 2.48372493326703e-06, + "loss": 1.1041, + "step": 27707 + }, + { + "epoch": 2.33, + "learning_rate": 2.4831254167673948e-06, + "loss": 1.1709, + "step": 27708 + }, + { + "epoch": 2.33, + "learning_rate": 2.482525962374581e-06, + "loss": 1.1943, + "step": 27709 + }, + { + "epoch": 2.33, + "learning_rate": 2.4819265700935446e-06, + "loss": 1.1442, + "step": 27710 + }, + { + "epoch": 2.33, + "learning_rate": 2.4813272399292376e-06, + "loss": 1.1293, + "step": 27711 + }, + { + "epoch": 2.33, + "learning_rate": 2.4807279718866107e-06, + "loss": 1.0963, + "step": 27712 + }, + { + "epoch": 2.33, + "learning_rate": 2.4801287659706187e-06, + "loss": 1.2197, + "step": 27713 + }, + { + "epoch": 2.33, + "learning_rate": 2.4795296221862043e-06, + "loss": 1.2132, + "step": 27714 + }, + { + "epoch": 2.33, + "learning_rate": 2.478930540538328e-06, + "loss": 1.1608, + "step": 27715 + }, + { + "epoch": 2.33, + "learning_rate": 2.4783315210319314e-06, + "loss": 1.0532, + "step": 27716 + }, + { + "epoch": 2.33, + "learning_rate": 2.4777325636719683e-06, + "loss": 1.1368, + "step": 27717 + }, + { + "epoch": 2.33, + "learning_rate": 2.477133668463385e-06, + "loss": 1.1103, + "step": 27718 + }, + { + "epoch": 2.33, + "learning_rate": 2.476534835411135e-06, + "loss": 1.0865, + "step": 27719 + }, + { + "epoch": 2.33, + "learning_rate": 2.4759360645201547e-06, + "loss": 1.0971, + "step": 27720 + }, + { + "epoch": 2.33, + "learning_rate": 2.475337355795406e-06, + "loss": 1.1263, + "step": 27721 + }, + { + "epoch": 2.33, + "learning_rate": 2.474738709241824e-06, + "loss": 1.0698, + "step": 27722 + }, + { + "epoch": 2.33, + "learning_rate": 2.474140124864357e-06, + "loss": 1.0746, + "step": 27723 + }, + { + "epoch": 2.33, + "learning_rate": 2.4735416026679592e-06, + "loss": 1.1229, + "step": 27724 + }, + { + "epoch": 2.33, + "learning_rate": 2.4729431426575657e-06, + "loss": 1.1476, + "step": 27725 + }, + { + "epoch": 2.33, + "learning_rate": 2.4723447448381257e-06, + "loss": 1.0945, + "step": 27726 + }, + { + "epoch": 2.33, + "learning_rate": 2.4717464092145815e-06, + "loss": 1.2201, + "step": 27727 + }, + { + "epoch": 2.33, + "learning_rate": 2.4711481357918786e-06, + "loss": 1.1633, + "step": 27728 + }, + { + "epoch": 2.33, + "learning_rate": 2.4705499245749585e-06, + "loss": 1.1824, + "step": 27729 + }, + { + "epoch": 2.33, + "learning_rate": 2.4699517755687684e-06, + "loss": 1.1599, + "step": 27730 + }, + { + "epoch": 2.33, + "learning_rate": 2.469353688778241e-06, + "loss": 1.0771, + "step": 27731 + }, + { + "epoch": 2.33, + "learning_rate": 2.4687556642083287e-06, + "loss": 1.1066, + "step": 27732 + }, + { + "epoch": 2.33, + "learning_rate": 2.468157701863965e-06, + "loss": 1.1673, + "step": 27733 + }, + { + "epoch": 2.33, + "learning_rate": 2.46755980175009e-06, + "loss": 1.0835, + "step": 27734 + }, + { + "epoch": 2.34, + "learning_rate": 2.4669619638716524e-06, + "loss": 1.2288, + "step": 27735 + }, + { + "epoch": 2.34, + "learning_rate": 2.466364188233582e-06, + "loss": 1.1269, + "step": 27736 + }, + { + "epoch": 2.34, + "learning_rate": 2.465766474840823e-06, + "loss": 1.1165, + "step": 27737 + }, + { + "epoch": 2.34, + "learning_rate": 2.465168823698312e-06, + "loss": 1.1756, + "step": 27738 + }, + { + "epoch": 2.34, + "learning_rate": 2.4645712348109884e-06, + "loss": 1.1036, + "step": 27739 + }, + { + "epoch": 2.34, + "learning_rate": 2.463973708183788e-06, + "loss": 1.1132, + "step": 27740 + }, + { + "epoch": 2.34, + "learning_rate": 2.4633762438216515e-06, + "loss": 1.1448, + "step": 27741 + }, + { + "epoch": 2.34, + "learning_rate": 2.4627788417295073e-06, + "loss": 1.1413, + "step": 27742 + }, + { + "epoch": 2.34, + "learning_rate": 2.462181501912302e-06, + "loss": 1.13, + "step": 27743 + }, + { + "epoch": 2.34, + "learning_rate": 2.461584224374963e-06, + "loss": 1.0751, + "step": 27744 + }, + { + "epoch": 2.34, + "learning_rate": 2.4609870091224276e-06, + "loss": 1.0769, + "step": 27745 + }, + { + "epoch": 2.34, + "learning_rate": 2.4603898561596295e-06, + "loss": 1.0799, + "step": 27746 + }, + { + "epoch": 2.34, + "learning_rate": 2.4597927654915045e-06, + "loss": 1.1189, + "step": 27747 + }, + { + "epoch": 2.34, + "learning_rate": 2.4591957371229847e-06, + "loss": 1.1263, + "step": 27748 + }, + { + "epoch": 2.34, + "learning_rate": 2.458598771059003e-06, + "loss": 1.0735, + "step": 27749 + }, + { + "epoch": 2.34, + "learning_rate": 2.4580018673044915e-06, + "loss": 1.1352, + "step": 27750 + }, + { + "epoch": 2.34, + "learning_rate": 2.457405025864382e-06, + "loss": 1.2048, + "step": 27751 + }, + { + "epoch": 2.34, + "learning_rate": 2.4568082467436106e-06, + "loss": 1.1673, + "step": 27752 + }, + { + "epoch": 2.34, + "learning_rate": 2.4562115299470956e-06, + "loss": 1.1441, + "step": 27753 + }, + { + "epoch": 2.34, + "learning_rate": 2.4556148754797824e-06, + "loss": 1.161, + "step": 27754 + }, + { + "epoch": 2.34, + "learning_rate": 2.4550182833465908e-06, + "loss": 1.2001, + "step": 27755 + }, + { + "epoch": 2.34, + "learning_rate": 2.454421753552453e-06, + "loss": 1.0448, + "step": 27756 + }, + { + "epoch": 2.34, + "learning_rate": 2.453825286102297e-06, + "loss": 1.1012, + "step": 27757 + }, + { + "epoch": 2.34, + "learning_rate": 2.453228881001051e-06, + "loss": 1.0333, + "step": 27758 + }, + { + "epoch": 2.34, + "learning_rate": 2.452632538253644e-06, + "loss": 1.1011, + "step": 27759 + }, + { + "epoch": 2.34, + "learning_rate": 2.4520362578650024e-06, + "loss": 1.1315, + "step": 27760 + }, + { + "epoch": 2.34, + "learning_rate": 2.451440039840053e-06, + "loss": 1.0297, + "step": 27761 + }, + { + "epoch": 2.34, + "learning_rate": 2.4508438841837213e-06, + "loss": 1.0885, + "step": 27762 + }, + { + "epoch": 2.34, + "learning_rate": 2.450247790900937e-06, + "loss": 1.117, + "step": 27763 + }, + { + "epoch": 2.34, + "learning_rate": 2.449651759996615e-06, + "loss": 1.0007, + "step": 27764 + }, + { + "epoch": 2.34, + "learning_rate": 2.449055791475694e-06, + "loss": 1.0571, + "step": 27765 + }, + { + "epoch": 2.34, + "learning_rate": 2.4484598853430865e-06, + "loss": 1.1083, + "step": 27766 + }, + { + "epoch": 2.34, + "learning_rate": 2.4478640416037204e-06, + "loss": 1.1189, + "step": 27767 + }, + { + "epoch": 2.34, + "learning_rate": 2.4472682602625187e-06, + "loss": 1.1285, + "step": 27768 + }, + { + "epoch": 2.34, + "learning_rate": 2.4466725413244065e-06, + "loss": 1.153, + "step": 27769 + }, + { + "epoch": 2.34, + "learning_rate": 2.4460768847942986e-06, + "loss": 1.1939, + "step": 27770 + }, + { + "epoch": 2.34, + "learning_rate": 2.445481290677125e-06, + "loss": 1.1347, + "step": 27771 + }, + { + "epoch": 2.34, + "learning_rate": 2.4448857589778007e-06, + "loss": 1.2344, + "step": 27772 + }, + { + "epoch": 2.34, + "learning_rate": 2.4442902897012453e-06, + "loss": 1.1347, + "step": 27773 + }, + { + "epoch": 2.34, + "learning_rate": 2.4436948828523877e-06, + "loss": 1.043, + "step": 27774 + }, + { + "epoch": 2.34, + "learning_rate": 2.443099538436138e-06, + "loss": 1.0649, + "step": 27775 + }, + { + "epoch": 2.34, + "learning_rate": 2.4425042564574186e-06, + "loss": 1.1286, + "step": 27776 + }, + { + "epoch": 2.34, + "learning_rate": 2.4419090369211475e-06, + "loss": 1.1519, + "step": 27777 + }, + { + "epoch": 2.34, + "learning_rate": 2.441313879832243e-06, + "loss": 1.1453, + "step": 27778 + }, + { + "epoch": 2.34, + "learning_rate": 2.4407187851956227e-06, + "loss": 1.0541, + "step": 27779 + }, + { + "epoch": 2.34, + "learning_rate": 2.440123753016205e-06, + "loss": 1.0613, + "step": 27780 + }, + { + "epoch": 2.34, + "learning_rate": 2.439528783298899e-06, + "loss": 1.0691, + "step": 27781 + }, + { + "epoch": 2.34, + "learning_rate": 2.4389338760486313e-06, + "loss": 1.111, + "step": 27782 + }, + { + "epoch": 2.34, + "learning_rate": 2.4383390312703105e-06, + "loss": 1.1963, + "step": 27783 + }, + { + "epoch": 2.34, + "learning_rate": 2.4377442489688485e-06, + "loss": 1.2047, + "step": 27784 + }, + { + "epoch": 2.34, + "learning_rate": 2.4371495291491688e-06, + "loss": 1.1279, + "step": 27785 + }, + { + "epoch": 2.34, + "learning_rate": 2.436554871816178e-06, + "loss": 1.0866, + "step": 27786 + }, + { + "epoch": 2.34, + "learning_rate": 2.4359602769747926e-06, + "loss": 1.1619, + "step": 27787 + }, + { + "epoch": 2.34, + "learning_rate": 2.435365744629923e-06, + "loss": 1.2597, + "step": 27788 + }, + { + "epoch": 2.34, + "learning_rate": 2.4347712747864825e-06, + "loss": 1.0042, + "step": 27789 + }, + { + "epoch": 2.34, + "learning_rate": 2.4341768674493827e-06, + "loss": 1.1558, + "step": 27790 + }, + { + "epoch": 2.34, + "learning_rate": 2.433582522623539e-06, + "loss": 1.1802, + "step": 27791 + }, + { + "epoch": 2.34, + "learning_rate": 2.4329882403138515e-06, + "loss": 1.2103, + "step": 27792 + }, + { + "epoch": 2.34, + "learning_rate": 2.4323940205252427e-06, + "loss": 1.1186, + "step": 27793 + }, + { + "epoch": 2.34, + "learning_rate": 2.431799863262614e-06, + "loss": 1.1468, + "step": 27794 + }, + { + "epoch": 2.34, + "learning_rate": 2.4312057685308765e-06, + "loss": 1.1593, + "step": 27795 + }, + { + "epoch": 2.34, + "learning_rate": 2.430611736334939e-06, + "loss": 1.165, + "step": 27796 + }, + { + "epoch": 2.34, + "learning_rate": 2.43001776667971e-06, + "loss": 1.0785, + "step": 27797 + }, + { + "epoch": 2.34, + "learning_rate": 2.429423859570097e-06, + "loss": 1.0971, + "step": 27798 + }, + { + "epoch": 2.34, + "learning_rate": 2.4288300150110077e-06, + "loss": 0.9616, + "step": 27799 + }, + { + "epoch": 2.34, + "learning_rate": 2.428236233007347e-06, + "loss": 1.1127, + "step": 27800 + }, + { + "epoch": 2.34, + "learning_rate": 2.4276425135640225e-06, + "loss": 1.1781, + "step": 27801 + }, + { + "epoch": 2.34, + "learning_rate": 2.4270488566859417e-06, + "loss": 1.1003, + "step": 27802 + }, + { + "epoch": 2.34, + "learning_rate": 2.4264552623780004e-06, + "loss": 1.0939, + "step": 27803 + }, + { + "epoch": 2.34, + "learning_rate": 2.425861730645116e-06, + "loss": 1.0653, + "step": 27804 + }, + { + "epoch": 2.34, + "learning_rate": 2.425268261492183e-06, + "loss": 1.1612, + "step": 27805 + }, + { + "epoch": 2.34, + "learning_rate": 2.424674854924108e-06, + "loss": 1.1823, + "step": 27806 + }, + { + "epoch": 2.34, + "learning_rate": 2.424081510945794e-06, + "loss": 1.109, + "step": 27807 + }, + { + "epoch": 2.34, + "learning_rate": 2.423488229562142e-06, + "loss": 1.1283, + "step": 27808 + }, + { + "epoch": 2.34, + "learning_rate": 2.4228950107780557e-06, + "loss": 1.0957, + "step": 27809 + }, + { + "epoch": 2.34, + "learning_rate": 2.4223018545984366e-06, + "loss": 1.086, + "step": 27810 + }, + { + "epoch": 2.34, + "learning_rate": 2.421708761028184e-06, + "loss": 1.1807, + "step": 27811 + }, + { + "epoch": 2.34, + "learning_rate": 2.4211157300721986e-06, + "loss": 1.0821, + "step": 27812 + }, + { + "epoch": 2.34, + "learning_rate": 2.4205227617353845e-06, + "loss": 1.0793, + "step": 27813 + }, + { + "epoch": 2.34, + "learning_rate": 2.4199298560226304e-06, + "loss": 1.0658, + "step": 27814 + }, + { + "epoch": 2.34, + "learning_rate": 2.4193370129388482e-06, + "loss": 1.1844, + "step": 27815 + }, + { + "epoch": 2.34, + "learning_rate": 2.4187442324889267e-06, + "loss": 1.2027, + "step": 27816 + }, + { + "epoch": 2.34, + "learning_rate": 2.4181515146777666e-06, + "loss": 1.053, + "step": 27817 + }, + { + "epoch": 2.34, + "learning_rate": 2.4175588595102652e-06, + "loss": 1.1379, + "step": 27818 + }, + { + "epoch": 2.34, + "learning_rate": 2.4169662669913218e-06, + "loss": 1.1446, + "step": 27819 + }, + { + "epoch": 2.34, + "learning_rate": 2.4163737371258244e-06, + "loss": 1.17, + "step": 27820 + }, + { + "epoch": 2.34, + "learning_rate": 2.41578126991868e-06, + "loss": 1.0912, + "step": 27821 + }, + { + "epoch": 2.34, + "learning_rate": 2.4151888653747756e-06, + "loss": 1.189, + "step": 27822 + }, + { + "epoch": 2.34, + "learning_rate": 2.4145965234990044e-06, + "loss": 1.1954, + "step": 27823 + }, + { + "epoch": 2.34, + "learning_rate": 2.4140042442962708e-06, + "loss": 1.1451, + "step": 27824 + }, + { + "epoch": 2.34, + "learning_rate": 2.413412027771459e-06, + "loss": 1.1247, + "step": 27825 + }, + { + "epoch": 2.34, + "learning_rate": 2.412819873929465e-06, + "loss": 1.1375, + "step": 27826 + }, + { + "epoch": 2.34, + "learning_rate": 2.412227782775182e-06, + "loss": 1.021, + "step": 27827 + }, + { + "epoch": 2.34, + "learning_rate": 2.4116357543135007e-06, + "loss": 1.0169, + "step": 27828 + }, + { + "epoch": 2.34, + "learning_rate": 2.411043788549313e-06, + "loss": 1.1201, + "step": 27829 + }, + { + "epoch": 2.34, + "learning_rate": 2.4104518854875135e-06, + "loss": 1.1513, + "step": 27830 + }, + { + "epoch": 2.34, + "learning_rate": 2.4098600451329844e-06, + "loss": 1.0383, + "step": 27831 + }, + { + "epoch": 2.34, + "learning_rate": 2.4092682674906253e-06, + "loss": 1.0967, + "step": 27832 + }, + { + "epoch": 2.34, + "learning_rate": 2.4086765525653187e-06, + "loss": 1.086, + "step": 27833 + }, + { + "epoch": 2.34, + "learning_rate": 2.4080849003619543e-06, + "loss": 1.11, + "step": 27834 + }, + { + "epoch": 2.34, + "learning_rate": 2.407493310885427e-06, + "loss": 1.1259, + "step": 27835 + }, + { + "epoch": 2.34, + "learning_rate": 2.406901784140617e-06, + "loss": 1.0355, + "step": 27836 + }, + { + "epoch": 2.34, + "learning_rate": 2.4063103201324144e-06, + "loss": 1.0969, + "step": 27837 + }, + { + "epoch": 2.34, + "learning_rate": 2.405718918865706e-06, + "loss": 1.1195, + "step": 27838 + }, + { + "epoch": 2.34, + "learning_rate": 2.4051275803453787e-06, + "loss": 1.1228, + "step": 27839 + }, + { + "epoch": 2.34, + "learning_rate": 2.4045363045763182e-06, + "loss": 1.1915, + "step": 27840 + }, + { + "epoch": 2.34, + "learning_rate": 2.403945091563412e-06, + "loss": 1.0862, + "step": 27841 + }, + { + "epoch": 2.34, + "learning_rate": 2.403353941311538e-06, + "loss": 1.104, + "step": 27842 + }, + { + "epoch": 2.34, + "learning_rate": 2.40276285382559e-06, + "loss": 1.1297, + "step": 27843 + }, + { + "epoch": 2.34, + "learning_rate": 2.4021718291104436e-06, + "loss": 1.0414, + "step": 27844 + }, + { + "epoch": 2.34, + "learning_rate": 2.401580867170985e-06, + "loss": 1.1058, + "step": 27845 + }, + { + "epoch": 2.34, + "learning_rate": 2.4009899680120973e-06, + "loss": 1.1827, + "step": 27846 + }, + { + "epoch": 2.34, + "learning_rate": 2.4003991316386623e-06, + "loss": 1.0139, + "step": 27847 + }, + { + "epoch": 2.34, + "learning_rate": 2.399808358055562e-06, + "loss": 1.1325, + "step": 27848 + }, + { + "epoch": 2.34, + "learning_rate": 2.3992176472676777e-06, + "loss": 1.1691, + "step": 27849 + }, + { + "epoch": 2.34, + "learning_rate": 2.3986269992798893e-06, + "loss": 1.1775, + "step": 27850 + }, + { + "epoch": 2.34, + "learning_rate": 2.398036414097078e-06, + "loss": 1.1194, + "step": 27851 + }, + { + "epoch": 2.34, + "learning_rate": 2.397445891724125e-06, + "loss": 1.1312, + "step": 27852 + }, + { + "epoch": 2.34, + "learning_rate": 2.396855432165902e-06, + "loss": 1.0221, + "step": 27853 + }, + { + "epoch": 2.35, + "learning_rate": 2.3962650354272978e-06, + "loss": 1.1808, + "step": 27854 + }, + { + "epoch": 2.35, + "learning_rate": 2.395674701513183e-06, + "loss": 1.1052, + "step": 27855 + }, + { + "epoch": 2.35, + "learning_rate": 2.395084430428437e-06, + "loss": 1.168, + "step": 27856 + }, + { + "epoch": 2.35, + "learning_rate": 2.394494222177939e-06, + "loss": 1.109, + "step": 27857 + }, + { + "epoch": 2.35, + "learning_rate": 2.3939040767665623e-06, + "loss": 1.1164, + "step": 27858 + }, + { + "epoch": 2.35, + "learning_rate": 2.3933139941991846e-06, + "loss": 1.0966, + "step": 27859 + }, + { + "epoch": 2.35, + "learning_rate": 2.3927239744806805e-06, + "loss": 1.128, + "step": 27860 + }, + { + "epoch": 2.35, + "learning_rate": 2.3921340176159268e-06, + "loss": 1.0743, + "step": 27861 + }, + { + "epoch": 2.35, + "learning_rate": 2.391544123609796e-06, + "loss": 1.0918, + "step": 27862 + }, + { + "epoch": 2.35, + "learning_rate": 2.3909542924671656e-06, + "loss": 1.1146, + "step": 27863 + }, + { + "epoch": 2.35, + "learning_rate": 2.3903645241928995e-06, + "loss": 1.0912, + "step": 27864 + }, + { + "epoch": 2.35, + "learning_rate": 2.3897748187918846e-06, + "loss": 1.137, + "step": 27865 + }, + { + "epoch": 2.35, + "learning_rate": 2.389185176268982e-06, + "loss": 1.0746, + "step": 27866 + }, + { + "epoch": 2.35, + "learning_rate": 2.3885955966290677e-06, + "loss": 1.1059, + "step": 27867 + }, + { + "epoch": 2.35, + "learning_rate": 2.388006079877011e-06, + "loss": 1.1198, + "step": 27868 + }, + { + "epoch": 2.35, + "learning_rate": 2.387416626017688e-06, + "loss": 1.2013, + "step": 27869 + }, + { + "epoch": 2.35, + "learning_rate": 2.3868272350559596e-06, + "loss": 1.1037, + "step": 27870 + }, + { + "epoch": 2.35, + "learning_rate": 2.386237906996707e-06, + "loss": 1.1543, + "step": 27871 + }, + { + "epoch": 2.35, + "learning_rate": 2.385648641844791e-06, + "loss": 1.1487, + "step": 27872 + }, + { + "epoch": 2.35, + "learning_rate": 2.3850594396050786e-06, + "loss": 1.1736, + "step": 27873 + }, + { + "epoch": 2.35, + "learning_rate": 2.3844703002824487e-06, + "loss": 1.0924, + "step": 27874 + }, + { + "epoch": 2.35, + "learning_rate": 2.38388122388176e-06, + "loss": 1.0333, + "step": 27875 + }, + { + "epoch": 2.35, + "learning_rate": 2.3832922104078815e-06, + "loss": 1.139, + "step": 27876 + }, + { + "epoch": 2.35, + "learning_rate": 2.382703259865681e-06, + "loss": 1.0914, + "step": 27877 + }, + { + "epoch": 2.35, + "learning_rate": 2.382114372260024e-06, + "loss": 1.1427, + "step": 27878 + }, + { + "epoch": 2.35, + "learning_rate": 2.3815255475957755e-06, + "loss": 1.0789, + "step": 27879 + }, + { + "epoch": 2.35, + "learning_rate": 2.3809367858778042e-06, + "loss": 1.0833, + "step": 27880 + }, + { + "epoch": 2.35, + "learning_rate": 2.3803480871109663e-06, + "loss": 1.0312, + "step": 27881 + }, + { + "epoch": 2.35, + "learning_rate": 2.3797594513001364e-06, + "loss": 1.117, + "step": 27882 + }, + { + "epoch": 2.35, + "learning_rate": 2.37917087845017e-06, + "loss": 1.1108, + "step": 27883 + }, + { + "epoch": 2.35, + "learning_rate": 2.3785823685659294e-06, + "loss": 1.081, + "step": 27884 + }, + { + "epoch": 2.35, + "learning_rate": 2.3779939216522863e-06, + "loss": 1.1313, + "step": 27885 + }, + { + "epoch": 2.35, + "learning_rate": 2.377405537714095e-06, + "loss": 1.1145, + "step": 27886 + }, + { + "epoch": 2.35, + "learning_rate": 2.3768172167562175e-06, + "loss": 1.1458, + "step": 27887 + }, + { + "epoch": 2.35, + "learning_rate": 2.3762289587835165e-06, + "loss": 1.0583, + "step": 27888 + }, + { + "epoch": 2.35, + "learning_rate": 2.3756407638008518e-06, + "loss": 1.1425, + "step": 27889 + }, + { + "epoch": 2.35, + "learning_rate": 2.3750526318130827e-06, + "loss": 1.0374, + "step": 27890 + }, + { + "epoch": 2.35, + "learning_rate": 2.374464562825072e-06, + "loss": 1.1488, + "step": 27891 + }, + { + "epoch": 2.35, + "learning_rate": 2.37387655684167e-06, + "loss": 1.1188, + "step": 27892 + }, + { + "epoch": 2.35, + "learning_rate": 2.373288613867747e-06, + "loss": 1.0749, + "step": 27893 + }, + { + "epoch": 2.35, + "learning_rate": 2.37270073390815e-06, + "loss": 1.1299, + "step": 27894 + }, + { + "epoch": 2.35, + "learning_rate": 2.3721129169677415e-06, + "loss": 1.1426, + "step": 27895 + }, + { + "epoch": 2.35, + "learning_rate": 2.3715251630513768e-06, + "loss": 1.2159, + "step": 27896 + }, + { + "epoch": 2.35, + "learning_rate": 2.370937472163913e-06, + "loss": 1.089, + "step": 27897 + }, + { + "epoch": 2.35, + "learning_rate": 2.370349844310206e-06, + "loss": 1.1779, + "step": 27898 + }, + { + "epoch": 2.35, + "learning_rate": 2.369762279495109e-06, + "loss": 1.2454, + "step": 27899 + }, + { + "epoch": 2.35, + "learning_rate": 2.3691747777234796e-06, + "loss": 1.1081, + "step": 27900 + }, + { + "epoch": 2.35, + "learning_rate": 2.368587339000169e-06, + "loss": 1.1201, + "step": 27901 + }, + { + "epoch": 2.35, + "learning_rate": 2.3679999633300366e-06, + "loss": 1.1237, + "step": 27902 + }, + { + "epoch": 2.35, + "learning_rate": 2.367412650717924e-06, + "loss": 1.06, + "step": 27903 + }, + { + "epoch": 2.35, + "learning_rate": 2.3668254011686964e-06, + "loss": 1.0931, + "step": 27904 + }, + { + "epoch": 2.35, + "learning_rate": 2.366238214687199e-06, + "loss": 1.0603, + "step": 27905 + }, + { + "epoch": 2.35, + "learning_rate": 2.3656510912782827e-06, + "loss": 1.1056, + "step": 27906 + }, + { + "epoch": 2.35, + "learning_rate": 2.365064030946801e-06, + "loss": 1.1239, + "step": 27907 + }, + { + "epoch": 2.35, + "learning_rate": 2.364477033697604e-06, + "loss": 1.173, + "step": 27908 + }, + { + "epoch": 2.35, + "learning_rate": 2.3638900995355418e-06, + "loss": 1.1122, + "step": 27909 + }, + { + "epoch": 2.35, + "learning_rate": 2.3633032284654634e-06, + "loss": 1.1276, + "step": 27910 + }, + { + "epoch": 2.35, + "learning_rate": 2.362716420492217e-06, + "loss": 1.0966, + "step": 27911 + }, + { + "epoch": 2.35, + "learning_rate": 2.3621296756206524e-06, + "loss": 1.0908, + "step": 27912 + }, + { + "epoch": 2.35, + "learning_rate": 2.361542993855619e-06, + "loss": 1.0936, + "step": 27913 + }, + { + "epoch": 2.35, + "learning_rate": 2.360956375201957e-06, + "loss": 1.0742, + "step": 27914 + }, + { + "epoch": 2.35, + "learning_rate": 2.3603698196645242e-06, + "loss": 1.1514, + "step": 27915 + }, + { + "epoch": 2.35, + "learning_rate": 2.3597833272481574e-06, + "loss": 1.0188, + "step": 27916 + }, + { + "epoch": 2.35, + "learning_rate": 2.359196897957706e-06, + "loss": 1.134, + "step": 27917 + }, + { + "epoch": 2.35, + "learning_rate": 2.3586105317980155e-06, + "loss": 1.1112, + "step": 27918 + }, + { + "epoch": 2.35, + "learning_rate": 2.358024228773933e-06, + "loss": 1.1256, + "step": 27919 + }, + { + "epoch": 2.35, + "learning_rate": 2.357437988890294e-06, + "loss": 1.1261, + "step": 27920 + }, + { + "epoch": 2.35, + "learning_rate": 2.3568518121519547e-06, + "loss": 1.0182, + "step": 27921 + }, + { + "epoch": 2.35, + "learning_rate": 2.3562656985637488e-06, + "loss": 1.0791, + "step": 27922 + }, + { + "epoch": 2.35, + "learning_rate": 2.355679648130519e-06, + "loss": 1.1336, + "step": 27923 + }, + { + "epoch": 2.35, + "learning_rate": 2.355093660857116e-06, + "loss": 1.0496, + "step": 27924 + }, + { + "epoch": 2.35, + "learning_rate": 2.3545077367483737e-06, + "loss": 1.1439, + "step": 27925 + }, + { + "epoch": 2.35, + "learning_rate": 2.3539218758091353e-06, + "loss": 1.1363, + "step": 27926 + }, + { + "epoch": 2.35, + "learning_rate": 2.3533360780442403e-06, + "loss": 1.1272, + "step": 27927 + }, + { + "epoch": 2.35, + "learning_rate": 2.352750343458531e-06, + "loss": 1.1907, + "step": 27928 + }, + { + "epoch": 2.35, + "learning_rate": 2.352164672056846e-06, + "loss": 1.1856, + "step": 27929 + }, + { + "epoch": 2.35, + "learning_rate": 2.351579063844026e-06, + "loss": 1.1251, + "step": 27930 + }, + { + "epoch": 2.35, + "learning_rate": 2.350993518824902e-06, + "loss": 1.179, + "step": 27931 + }, + { + "epoch": 2.35, + "learning_rate": 2.3504080370043235e-06, + "loss": 1.158, + "step": 27932 + }, + { + "epoch": 2.35, + "learning_rate": 2.3498226183871186e-06, + "loss": 1.2103, + "step": 27933 + }, + { + "epoch": 2.35, + "learning_rate": 2.349237262978126e-06, + "loss": 1.1464, + "step": 27934 + }, + { + "epoch": 2.35, + "learning_rate": 2.3486519707821876e-06, + "loss": 1.2148, + "step": 27935 + }, + { + "epoch": 2.35, + "learning_rate": 2.3480667418041337e-06, + "loss": 1.154, + "step": 27936 + }, + { + "epoch": 2.35, + "learning_rate": 2.3474815760488014e-06, + "loss": 1.0587, + "step": 27937 + }, + { + "epoch": 2.35, + "learning_rate": 2.3468964735210254e-06, + "loss": 1.185, + "step": 27938 + }, + { + "epoch": 2.35, + "learning_rate": 2.34631143422564e-06, + "loss": 1.0639, + "step": 27939 + }, + { + "epoch": 2.35, + "learning_rate": 2.3457264581674786e-06, + "loss": 1.0614, + "step": 27940 + }, + { + "epoch": 2.35, + "learning_rate": 2.345141545351378e-06, + "loss": 1.1835, + "step": 27941 + }, + { + "epoch": 2.35, + "learning_rate": 2.3445566957821618e-06, + "loss": 1.0984, + "step": 27942 + }, + { + "epoch": 2.35, + "learning_rate": 2.343971909464674e-06, + "loss": 1.0449, + "step": 27943 + }, + { + "epoch": 2.35, + "learning_rate": 2.3433871864037384e-06, + "loss": 1.2133, + "step": 27944 + }, + { + "epoch": 2.35, + "learning_rate": 2.3428025266041876e-06, + "loss": 1.1018, + "step": 27945 + }, + { + "epoch": 2.35, + "learning_rate": 2.3422179300708537e-06, + "loss": 1.1463, + "step": 27946 + }, + { + "epoch": 2.35, + "learning_rate": 2.3416333968085647e-06, + "loss": 1.0955, + "step": 27947 + }, + { + "epoch": 2.35, + "learning_rate": 2.3410489268221526e-06, + "loss": 1.2011, + "step": 27948 + }, + { + "epoch": 2.35, + "learning_rate": 2.3404645201164453e-06, + "loss": 1.1308, + "step": 27949 + }, + { + "epoch": 2.35, + "learning_rate": 2.3398801766962707e-06, + "loss": 1.1313, + "step": 27950 + }, + { + "epoch": 2.35, + "learning_rate": 2.3392958965664594e-06, + "loss": 1.1632, + "step": 27951 + }, + { + "epoch": 2.35, + "learning_rate": 2.338711679731839e-06, + "loss": 1.0262, + "step": 27952 + }, + { + "epoch": 2.35, + "learning_rate": 2.338127526197228e-06, + "loss": 1.0816, + "step": 27953 + }, + { + "epoch": 2.35, + "learning_rate": 2.3375434359674664e-06, + "loss": 1.1809, + "step": 27954 + }, + { + "epoch": 2.35, + "learning_rate": 2.33695940904737e-06, + "loss": 1.1243, + "step": 27955 + }, + { + "epoch": 2.35, + "learning_rate": 2.3363754454417674e-06, + "loss": 1.1548, + "step": 27956 + }, + { + "epoch": 2.35, + "learning_rate": 2.3357915451554834e-06, + "loss": 1.1749, + "step": 27957 + }, + { + "epoch": 2.35, + "learning_rate": 2.3352077081933434e-06, + "loss": 1.0511, + "step": 27958 + }, + { + "epoch": 2.35, + "learning_rate": 2.3346239345601694e-06, + "loss": 1.1568, + "step": 27959 + }, + { + "epoch": 2.35, + "learning_rate": 2.3340402242607864e-06, + "loss": 1.145, + "step": 27960 + }, + { + "epoch": 2.35, + "learning_rate": 2.3334565773000162e-06, + "loss": 1.0933, + "step": 27961 + }, + { + "epoch": 2.35, + "learning_rate": 2.332872993682681e-06, + "loss": 1.0073, + "step": 27962 + }, + { + "epoch": 2.35, + "learning_rate": 2.3322894734136057e-06, + "loss": 1.1506, + "step": 27963 + }, + { + "epoch": 2.35, + "learning_rate": 2.3317060164976033e-06, + "loss": 1.1969, + "step": 27964 + }, + { + "epoch": 2.35, + "learning_rate": 2.3311226229395058e-06, + "loss": 1.1166, + "step": 27965 + }, + { + "epoch": 2.35, + "learning_rate": 2.330539292744125e-06, + "loss": 1.07, + "step": 27966 + }, + { + "epoch": 2.35, + "learning_rate": 2.3299560259162823e-06, + "loss": 1.091, + "step": 27967 + }, + { + "epoch": 2.35, + "learning_rate": 2.329372822460798e-06, + "loss": 0.9933, + "step": 27968 + }, + { + "epoch": 2.35, + "learning_rate": 2.3287896823824905e-06, + "loss": 1.0485, + "step": 27969 + }, + { + "epoch": 2.35, + "learning_rate": 2.3282066056861786e-06, + "loss": 1.2064, + "step": 27970 + }, + { + "epoch": 2.35, + "learning_rate": 2.327623592376681e-06, + "loss": 1.0996, + "step": 27971 + }, + { + "epoch": 2.36, + "learning_rate": 2.3270406424588066e-06, + "loss": 1.068, + "step": 27972 + }, + { + "epoch": 2.36, + "learning_rate": 2.326457755937381e-06, + "loss": 1.1984, + "step": 27973 + }, + { + "epoch": 2.36, + "learning_rate": 2.3258749328172205e-06, + "loss": 1.1835, + "step": 27974 + }, + { + "epoch": 2.36, + "learning_rate": 2.325292173103132e-06, + "loss": 1.1025, + "step": 27975 + }, + { + "epoch": 2.36, + "learning_rate": 2.3247094767999413e-06, + "loss": 1.0737, + "step": 27976 + }, + { + "epoch": 2.36, + "learning_rate": 2.3241268439124544e-06, + "loss": 1.1394, + "step": 27977 + }, + { + "epoch": 2.36, + "learning_rate": 2.3235442744454874e-06, + "loss": 1.104, + "step": 27978 + }, + { + "epoch": 2.36, + "learning_rate": 2.3229617684038554e-06, + "loss": 1.1172, + "step": 27979 + }, + { + "epoch": 2.36, + "learning_rate": 2.322379325792372e-06, + "loss": 1.1203, + "step": 27980 + }, + { + "epoch": 2.36, + "learning_rate": 2.321796946615843e-06, + "loss": 1.0522, + "step": 27981 + }, + { + "epoch": 2.36, + "learning_rate": 2.3212146308790904e-06, + "loss": 1.0714, + "step": 27982 + }, + { + "epoch": 2.36, + "learning_rate": 2.3206323785869165e-06, + "loss": 1.0659, + "step": 27983 + }, + { + "epoch": 2.36, + "learning_rate": 2.320050189744134e-06, + "loss": 1.0819, + "step": 27984 + }, + { + "epoch": 2.36, + "learning_rate": 2.31946806435556e-06, + "loss": 1.1, + "step": 27985 + }, + { + "epoch": 2.36, + "learning_rate": 2.318886002425995e-06, + "loss": 1.1163, + "step": 27986 + }, + { + "epoch": 2.36, + "learning_rate": 2.3183040039602534e-06, + "loss": 1.1495, + "step": 27987 + }, + { + "epoch": 2.36, + "learning_rate": 2.317722068963142e-06, + "loss": 1.179, + "step": 27988 + }, + { + "epoch": 2.36, + "learning_rate": 2.31714019743947e-06, + "loss": 1.1542, + "step": 27989 + }, + { + "epoch": 2.36, + "learning_rate": 2.3165583893940445e-06, + "loss": 1.138, + "step": 27990 + }, + { + "epoch": 2.36, + "learning_rate": 2.3159766448316745e-06, + "loss": 1.1233, + "step": 27991 + }, + { + "epoch": 2.36, + "learning_rate": 2.3153949637571606e-06, + "loss": 1.0004, + "step": 27992 + }, + { + "epoch": 2.36, + "learning_rate": 2.3148133461753174e-06, + "loss": 1.028, + "step": 27993 + }, + { + "epoch": 2.36, + "learning_rate": 2.314231792090943e-06, + "loss": 1.0013, + "step": 27994 + }, + { + "epoch": 2.36, + "learning_rate": 2.3136503015088453e-06, + "loss": 1.0861, + "step": 27995 + }, + { + "epoch": 2.36, + "learning_rate": 2.3130688744338294e-06, + "loss": 1.0553, + "step": 27996 + }, + { + "epoch": 2.36, + "learning_rate": 2.312487510870698e-06, + "loss": 1.0807, + "step": 27997 + }, + { + "epoch": 2.36, + "learning_rate": 2.311906210824255e-06, + "loss": 1.0923, + "step": 27998 + }, + { + "epoch": 2.36, + "learning_rate": 2.311324974299305e-06, + "loss": 1.0437, + "step": 27999 + }, + { + "epoch": 2.36, + "learning_rate": 2.310743801300647e-06, + "loss": 1.227, + "step": 28000 + }, + { + "epoch": 2.36, + "learning_rate": 2.3101626918330855e-06, + "loss": 1.1499, + "step": 28001 + }, + { + "epoch": 2.36, + "learning_rate": 2.309581645901423e-06, + "loss": 1.1344, + "step": 28002 + }, + { + "epoch": 2.36, + "learning_rate": 2.3090006635104523e-06, + "loss": 1.0597, + "step": 28003 + }, + { + "epoch": 2.36, + "learning_rate": 2.308419744664986e-06, + "loss": 1.1354, + "step": 28004 + }, + { + "epoch": 2.36, + "learning_rate": 2.3078388893698147e-06, + "loss": 1.0597, + "step": 28005 + }, + { + "epoch": 2.36, + "learning_rate": 2.30725809762974e-06, + "loss": 1.0587, + "step": 28006 + }, + { + "epoch": 2.36, + "learning_rate": 2.3066773694495605e-06, + "loss": 1.201, + "step": 28007 + }, + { + "epoch": 2.36, + "learning_rate": 2.3060967048340756e-06, + "loss": 1.0981, + "step": 28008 + }, + { + "epoch": 2.36, + "learning_rate": 2.3055161037880814e-06, + "loss": 1.0829, + "step": 28009 + }, + { + "epoch": 2.36, + "learning_rate": 2.304935566316375e-06, + "loss": 1.1156, + "step": 28010 + }, + { + "epoch": 2.36, + "learning_rate": 2.3043550924237545e-06, + "loss": 1.1054, + "step": 28011 + }, + { + "epoch": 2.36, + "learning_rate": 2.3037746821150154e-06, + "loss": 1.1736, + "step": 28012 + }, + { + "epoch": 2.36, + "learning_rate": 2.303194335394956e-06, + "loss": 1.1396, + "step": 28013 + }, + { + "epoch": 2.36, + "learning_rate": 2.302614052268363e-06, + "loss": 1.1496, + "step": 28014 + }, + { + "epoch": 2.36, + "learning_rate": 2.3020338327400405e-06, + "loss": 1.1735, + "step": 28015 + }, + { + "epoch": 2.36, + "learning_rate": 2.3014536768147754e-06, + "loss": 1.0658, + "step": 28016 + }, + { + "epoch": 2.36, + "learning_rate": 2.3008735844973652e-06, + "loss": 1.0789, + "step": 28017 + }, + { + "epoch": 2.36, + "learning_rate": 2.3002935557926e-06, + "loss": 1.0657, + "step": 28018 + }, + { + "epoch": 2.36, + "learning_rate": 2.299713590705275e-06, + "loss": 1.1435, + "step": 28019 + }, + { + "epoch": 2.36, + "learning_rate": 2.2991336892401794e-06, + "loss": 1.1235, + "step": 28020 + }, + { + "epoch": 2.36, + "learning_rate": 2.29855385140211e-06, + "loss": 1.1373, + "step": 28021 + }, + { + "epoch": 2.36, + "learning_rate": 2.297974077195847e-06, + "loss": 1.0899, + "step": 28022 + }, + { + "epoch": 2.36, + "learning_rate": 2.2973943666261898e-06, + "loss": 1.1751, + "step": 28023 + }, + { + "epoch": 2.36, + "learning_rate": 2.2968147196979284e-06, + "loss": 1.047, + "step": 28024 + }, + { + "epoch": 2.36, + "learning_rate": 2.296235136415844e-06, + "loss": 1.098, + "step": 28025 + }, + { + "epoch": 2.36, + "learning_rate": 2.2956556167847356e-06, + "loss": 1.1668, + "step": 28026 + }, + { + "epoch": 2.36, + "learning_rate": 2.295076160809383e-06, + "loss": 1.1244, + "step": 28027 + }, + { + "epoch": 2.36, + "learning_rate": 2.2944967684945787e-06, + "loss": 1.0739, + "step": 28028 + }, + { + "epoch": 2.36, + "learning_rate": 2.293917439845106e-06, + "loss": 1.0328, + "step": 28029 + }, + { + "epoch": 2.36, + "learning_rate": 2.293338174865758e-06, + "loss": 1.0761, + "step": 28030 + }, + { + "epoch": 2.36, + "learning_rate": 2.2927589735613097e-06, + "loss": 1.13, + "step": 28031 + }, + { + "epoch": 2.36, + "learning_rate": 2.2921798359365597e-06, + "loss": 1.2666, + "step": 28032 + }, + { + "epoch": 2.36, + "learning_rate": 2.2916007619962843e-06, + "loss": 1.0829, + "step": 28033 + }, + { + "epoch": 2.36, + "learning_rate": 2.291021751745267e-06, + "loss": 1.0718, + "step": 28034 + }, + { + "epoch": 2.36, + "learning_rate": 2.290442805188301e-06, + "loss": 1.0519, + "step": 28035 + }, + { + "epoch": 2.36, + "learning_rate": 2.2898639223301612e-06, + "loss": 1.0836, + "step": 28036 + }, + { + "epoch": 2.36, + "learning_rate": 2.2892851031756337e-06, + "loss": 1.0547, + "step": 28037 + }, + { + "epoch": 2.36, + "learning_rate": 2.288706347729499e-06, + "loss": 1.0313, + "step": 28038 + }, + { + "epoch": 2.36, + "learning_rate": 2.2881276559965417e-06, + "loss": 1.0625, + "step": 28039 + }, + { + "epoch": 2.36, + "learning_rate": 2.2875490279815416e-06, + "loss": 1.13, + "step": 28040 + }, + { + "epoch": 2.36, + "learning_rate": 2.286970463689282e-06, + "loss": 1.1203, + "step": 28041 + }, + { + "epoch": 2.36, + "learning_rate": 2.286391963124536e-06, + "loss": 1.0829, + "step": 28042 + }, + { + "epoch": 2.36, + "learning_rate": 2.2858135262920933e-06, + "loss": 1.0894, + "step": 28043 + }, + { + "epoch": 2.36, + "learning_rate": 2.285235153196725e-06, + "loss": 1.0483, + "step": 28044 + }, + { + "epoch": 2.36, + "learning_rate": 2.284656843843214e-06, + "loss": 1.0871, + "step": 28045 + }, + { + "epoch": 2.36, + "learning_rate": 2.2840785982363365e-06, + "loss": 1.1175, + "step": 28046 + }, + { + "epoch": 2.36, + "learning_rate": 2.283500416380872e-06, + "loss": 1.1166, + "step": 28047 + }, + { + "epoch": 2.36, + "learning_rate": 2.282922298281595e-06, + "loss": 1.1062, + "step": 28048 + }, + { + "epoch": 2.36, + "learning_rate": 2.2823442439432854e-06, + "loss": 1.0679, + "step": 28049 + }, + { + "epoch": 2.36, + "learning_rate": 2.2817662533707165e-06, + "loss": 1.1313, + "step": 28050 + }, + { + "epoch": 2.36, + "learning_rate": 2.2811883265686654e-06, + "loss": 1.1062, + "step": 28051 + }, + { + "epoch": 2.36, + "learning_rate": 2.280610463541909e-06, + "loss": 1.0881, + "step": 28052 + }, + { + "epoch": 2.36, + "learning_rate": 2.2800326642952144e-06, + "loss": 1.147, + "step": 28053 + }, + { + "epoch": 2.36, + "learning_rate": 2.279454928833367e-06, + "loss": 1.1274, + "step": 28054 + }, + { + "epoch": 2.36, + "learning_rate": 2.2788772571611297e-06, + "loss": 1.0831, + "step": 28055 + }, + { + "epoch": 2.36, + "learning_rate": 2.2782996492832798e-06, + "loss": 1.0865, + "step": 28056 + }, + { + "epoch": 2.36, + "learning_rate": 2.2777221052045896e-06, + "loss": 1.0911, + "step": 28057 + }, + { + "epoch": 2.36, + "learning_rate": 2.277144624929831e-06, + "loss": 1.031, + "step": 28058 + }, + { + "epoch": 2.36, + "learning_rate": 2.276567208463776e-06, + "loss": 1.1129, + "step": 28059 + }, + { + "epoch": 2.36, + "learning_rate": 2.2759898558111938e-06, + "loss": 1.1507, + "step": 28060 + }, + { + "epoch": 2.36, + "learning_rate": 2.2754125669768546e-06, + "loss": 1.1776, + "step": 28061 + }, + { + "epoch": 2.36, + "learning_rate": 2.27483534196553e-06, + "loss": 1.088, + "step": 28062 + }, + { + "epoch": 2.36, + "learning_rate": 2.274258180781991e-06, + "loss": 1.1369, + "step": 28063 + }, + { + "epoch": 2.36, + "learning_rate": 2.2736810834309962e-06, + "loss": 1.1125, + "step": 28064 + }, + { + "epoch": 2.36, + "learning_rate": 2.2731040499173283e-06, + "loss": 1.1211, + "step": 28065 + }, + { + "epoch": 2.36, + "learning_rate": 2.2725270802457444e-06, + "loss": 1.0729, + "step": 28066 + }, + { + "epoch": 2.36, + "learning_rate": 2.2719501744210147e-06, + "loss": 1.1884, + "step": 28067 + }, + { + "epoch": 2.36, + "learning_rate": 2.271373332447907e-06, + "loss": 1.138, + "step": 28068 + }, + { + "epoch": 2.36, + "learning_rate": 2.270796554331185e-06, + "loss": 1.016, + "step": 28069 + }, + { + "epoch": 2.36, + "learning_rate": 2.2702198400756158e-06, + "loss": 1.1029, + "step": 28070 + }, + { + "epoch": 2.36, + "learning_rate": 2.269643189685967e-06, + "loss": 1.1687, + "step": 28071 + }, + { + "epoch": 2.36, + "learning_rate": 2.269066603166995e-06, + "loss": 1.0823, + "step": 28072 + }, + { + "epoch": 2.36, + "learning_rate": 2.2684900805234723e-06, + "loss": 1.1187, + "step": 28073 + }, + { + "epoch": 2.36, + "learning_rate": 2.2679136217601605e-06, + "loss": 1.1978, + "step": 28074 + }, + { + "epoch": 2.36, + "learning_rate": 2.267337226881816e-06, + "loss": 1.1836, + "step": 28075 + }, + { + "epoch": 2.36, + "learning_rate": 2.266760895893211e-06, + "loss": 1.1639, + "step": 28076 + }, + { + "epoch": 2.36, + "learning_rate": 2.2661846287991008e-06, + "loss": 0.9979, + "step": 28077 + }, + { + "epoch": 2.36, + "learning_rate": 2.265608425604248e-06, + "loss": 1.141, + "step": 28078 + }, + { + "epoch": 2.36, + "learning_rate": 2.2650322863134123e-06, + "loss": 1.1396, + "step": 28079 + }, + { + "epoch": 2.36, + "learning_rate": 2.2644562109313596e-06, + "loss": 1.201, + "step": 28080 + }, + { + "epoch": 2.36, + "learning_rate": 2.263880199462839e-06, + "loss": 1.1759, + "step": 28081 + }, + { + "epoch": 2.36, + "learning_rate": 2.2633042519126224e-06, + "loss": 1.0972, + "step": 28082 + }, + { + "epoch": 2.36, + "learning_rate": 2.2627283682854585e-06, + "loss": 1.08, + "step": 28083 + }, + { + "epoch": 2.36, + "learning_rate": 2.2621525485861074e-06, + "loss": 1.1398, + "step": 28084 + }, + { + "epoch": 2.36, + "learning_rate": 2.2615767928193335e-06, + "loss": 1.0922, + "step": 28085 + }, + { + "epoch": 2.36, + "learning_rate": 2.2610011009898857e-06, + "loss": 1.0269, + "step": 28086 + }, + { + "epoch": 2.36, + "learning_rate": 2.260425473102523e-06, + "loss": 1.2101, + "step": 28087 + }, + { + "epoch": 2.36, + "learning_rate": 2.259849909162003e-06, + "loss": 1.1407, + "step": 28088 + }, + { + "epoch": 2.36, + "learning_rate": 2.25927440917308e-06, + "loss": 1.1632, + "step": 28089 + }, + { + "epoch": 2.36, + "learning_rate": 2.2586989731405097e-06, + "loss": 1.0296, + "step": 28090 + }, + { + "epoch": 2.37, + "learning_rate": 2.2581236010690476e-06, + "loss": 1.1156, + "step": 28091 + }, + { + "epoch": 2.37, + "learning_rate": 2.2575482929634405e-06, + "loss": 1.0843, + "step": 28092 + }, + { + "epoch": 2.37, + "learning_rate": 2.2569730488284535e-06, + "loss": 1.11, + "step": 28093 + }, + { + "epoch": 2.37, + "learning_rate": 2.2563978686688303e-06, + "loss": 1.2144, + "step": 28094 + }, + { + "epoch": 2.37, + "learning_rate": 2.255822752489327e-06, + "loss": 1.1076, + "step": 28095 + }, + { + "epoch": 2.37, + "learning_rate": 2.2552477002946936e-06, + "loss": 1.0737, + "step": 28096 + }, + { + "epoch": 2.37, + "learning_rate": 2.254672712089683e-06, + "loss": 1.0875, + "step": 28097 + }, + { + "epoch": 2.37, + "learning_rate": 2.2540977878790447e-06, + "loss": 1.1152, + "step": 28098 + }, + { + "epoch": 2.37, + "learning_rate": 2.2535229276675297e-06, + "loss": 1.1929, + "step": 28099 + }, + { + "epoch": 2.37, + "learning_rate": 2.252948131459888e-06, + "loss": 1.1909, + "step": 28100 + }, + { + "epoch": 2.37, + "learning_rate": 2.2523733992608667e-06, + "loss": 1.0209, + "step": 28101 + }, + { + "epoch": 2.37, + "learning_rate": 2.25179873107522e-06, + "loss": 1.0924, + "step": 28102 + }, + { + "epoch": 2.37, + "learning_rate": 2.2512241269076862e-06, + "loss": 1.1031, + "step": 28103 + }, + { + "epoch": 2.37, + "learning_rate": 2.250649586763024e-06, + "loss": 1.1575, + "step": 28104 + }, + { + "epoch": 2.37, + "learning_rate": 2.250075110645972e-06, + "loss": 1.0436, + "step": 28105 + }, + { + "epoch": 2.37, + "learning_rate": 2.24950069856128e-06, + "loss": 1.1537, + "step": 28106 + }, + { + "epoch": 2.37, + "learning_rate": 2.248926350513694e-06, + "loss": 1.0793, + "step": 28107 + }, + { + "epoch": 2.37, + "learning_rate": 2.2483520665079584e-06, + "loss": 1.1033, + "step": 28108 + }, + { + "epoch": 2.37, + "learning_rate": 2.2477778465488187e-06, + "loss": 1.2362, + "step": 28109 + }, + { + "epoch": 2.37, + "learning_rate": 2.24720369064102e-06, + "loss": 1.1045, + "step": 28110 + }, + { + "epoch": 2.37, + "learning_rate": 2.246629598789306e-06, + "loss": 1.0384, + "step": 28111 + }, + { + "epoch": 2.37, + "learning_rate": 2.246055570998419e-06, + "loss": 1.072, + "step": 28112 + }, + { + "epoch": 2.37, + "learning_rate": 2.2454816072731057e-06, + "loss": 1.0521, + "step": 28113 + }, + { + "epoch": 2.37, + "learning_rate": 2.244907707618099e-06, + "loss": 1.1241, + "step": 28114 + }, + { + "epoch": 2.37, + "learning_rate": 2.2443338720381536e-06, + "loss": 1.1905, + "step": 28115 + }, + { + "epoch": 2.37, + "learning_rate": 2.2437601005379994e-06, + "loss": 1.1382, + "step": 28116 + }, + { + "epoch": 2.37, + "learning_rate": 2.2431863931223828e-06, + "loss": 1.1258, + "step": 28117 + }, + { + "epoch": 2.37, + "learning_rate": 2.2426127497960427e-06, + "loss": 1.1174, + "step": 28118 + }, + { + "epoch": 2.37, + "learning_rate": 2.242039170563718e-06, + "loss": 1.1418, + "step": 28119 + }, + { + "epoch": 2.37, + "learning_rate": 2.2414656554301484e-06, + "loss": 1.0618, + "step": 28120 + }, + { + "epoch": 2.37, + "learning_rate": 2.240892204400077e-06, + "loss": 1.0537, + "step": 28121 + }, + { + "epoch": 2.37, + "learning_rate": 2.2403188174782296e-06, + "loss": 1.141, + "step": 28122 + }, + { + "epoch": 2.37, + "learning_rate": 2.2397454946693563e-06, + "loss": 1.0933, + "step": 28123 + }, + { + "epoch": 2.37, + "learning_rate": 2.2391722359781908e-06, + "loss": 1.0702, + "step": 28124 + }, + { + "epoch": 2.37, + "learning_rate": 2.2385990414094628e-06, + "loss": 1.1945, + "step": 28125 + }, + { + "epoch": 2.37, + "learning_rate": 2.238025910967919e-06, + "loss": 1.0892, + "step": 28126 + }, + { + "epoch": 2.37, + "learning_rate": 2.2374528446582875e-06, + "loss": 1.1741, + "step": 28127 + }, + { + "epoch": 2.37, + "learning_rate": 2.2368798424853046e-06, + "loss": 1.1803, + "step": 28128 + }, + { + "epoch": 2.37, + "learning_rate": 2.2363069044537044e-06, + "loss": 1.0564, + "step": 28129 + }, + { + "epoch": 2.37, + "learning_rate": 2.235734030568224e-06, + "loss": 1.1846, + "step": 28130 + }, + { + "epoch": 2.37, + "learning_rate": 2.235161220833588e-06, + "loss": 1.1465, + "step": 28131 + }, + { + "epoch": 2.37, + "learning_rate": 2.234588475254541e-06, + "loss": 1.0685, + "step": 28132 + }, + { + "epoch": 2.37, + "learning_rate": 2.2340157938358065e-06, + "loss": 1.1369, + "step": 28133 + }, + { + "epoch": 2.37, + "learning_rate": 2.233443176582116e-06, + "loss": 1.1428, + "step": 28134 + }, + { + "epoch": 2.37, + "learning_rate": 2.23287062349821e-06, + "loss": 1.0324, + "step": 28135 + }, + { + "epoch": 2.37, + "learning_rate": 2.2322981345888095e-06, + "loss": 1.0742, + "step": 28136 + }, + { + "epoch": 2.37, + "learning_rate": 2.2317257098586476e-06, + "loss": 1.094, + "step": 28137 + }, + { + "epoch": 2.37, + "learning_rate": 2.231153349312454e-06, + "loss": 1.0947, + "step": 28138 + }, + { + "epoch": 2.37, + "learning_rate": 2.230581052954958e-06, + "loss": 1.0296, + "step": 28139 + }, + { + "epoch": 2.37, + "learning_rate": 2.230008820790888e-06, + "loss": 1.1112, + "step": 28140 + }, + { + "epoch": 2.37, + "learning_rate": 2.2294366528249743e-06, + "loss": 1.1044, + "step": 28141 + }, + { + "epoch": 2.37, + "learning_rate": 2.2288645490619363e-06, + "loss": 1.137, + "step": 28142 + }, + { + "epoch": 2.37, + "learning_rate": 2.2282925095065123e-06, + "loss": 1.132, + "step": 28143 + }, + { + "epoch": 2.37, + "learning_rate": 2.22772053416342e-06, + "loss": 1.1405, + "step": 28144 + }, + { + "epoch": 2.37, + "learning_rate": 2.2271486230373873e-06, + "loss": 1.1731, + "step": 28145 + }, + { + "epoch": 2.37, + "learning_rate": 2.2265767761331413e-06, + "loss": 1.0412, + "step": 28146 + }, + { + "epoch": 2.37, + "learning_rate": 2.226004993455406e-06, + "loss": 1.0814, + "step": 28147 + }, + { + "epoch": 2.37, + "learning_rate": 2.225433275008905e-06, + "loss": 1.1133, + "step": 28148 + }, + { + "epoch": 2.37, + "learning_rate": 2.2248616207983618e-06, + "loss": 1.0387, + "step": 28149 + }, + { + "epoch": 2.37, + "learning_rate": 2.224290030828501e-06, + "loss": 1.0873, + "step": 28150 + }, + { + "epoch": 2.37, + "learning_rate": 2.2237185051040445e-06, + "loss": 1.1661, + "step": 28151 + }, + { + "epoch": 2.37, + "learning_rate": 2.2231470436297178e-06, + "loss": 1.1679, + "step": 28152 + }, + { + "epoch": 2.37, + "learning_rate": 2.222575646410232e-06, + "loss": 1.0271, + "step": 28153 + }, + { + "epoch": 2.37, + "learning_rate": 2.222004313450322e-06, + "loss": 1.0596, + "step": 28154 + }, + { + "epoch": 2.37, + "learning_rate": 2.2214330447546983e-06, + "loss": 1.1413, + "step": 28155 + }, + { + "epoch": 2.37, + "learning_rate": 2.2208618403280845e-06, + "loss": 1.1165, + "step": 28156 + }, + { + "epoch": 2.37, + "learning_rate": 2.2202907001752006e-06, + "loss": 1.0629, + "step": 28157 + }, + { + "epoch": 2.37, + "learning_rate": 2.2197196243007633e-06, + "loss": 1.1358, + "step": 28158 + }, + { + "epoch": 2.37, + "learning_rate": 2.219148612709493e-06, + "loss": 1.0579, + "step": 28159 + }, + { + "epoch": 2.37, + "learning_rate": 2.2185776654061074e-06, + "loss": 1.1522, + "step": 28160 + }, + { + "epoch": 2.37, + "learning_rate": 2.2180067823953234e-06, + "loss": 1.168, + "step": 28161 + }, + { + "epoch": 2.37, + "learning_rate": 2.2174359636818576e-06, + "loss": 1.0559, + "step": 28162 + }, + { + "epoch": 2.37, + "learning_rate": 2.2168652092704303e-06, + "loss": 0.9809, + "step": 28163 + }, + { + "epoch": 2.37, + "learning_rate": 2.216294519165747e-06, + "loss": 1.2134, + "step": 28164 + }, + { + "epoch": 2.37, + "learning_rate": 2.2157238933725357e-06, + "loss": 1.0358, + "step": 28165 + }, + { + "epoch": 2.37, + "learning_rate": 2.215153331895502e-06, + "loss": 1.2325, + "step": 28166 + }, + { + "epoch": 2.37, + "learning_rate": 2.214582834739363e-06, + "loss": 1.0459, + "step": 28167 + }, + { + "epoch": 2.37, + "learning_rate": 2.2140124019088315e-06, + "loss": 1.1738, + "step": 28168 + }, + { + "epoch": 2.37, + "learning_rate": 2.2134420334086216e-06, + "loss": 1.1198, + "step": 28169 + }, + { + "epoch": 2.37, + "learning_rate": 2.2128717292434464e-06, + "loss": 1.1804, + "step": 28170 + }, + { + "epoch": 2.37, + "learning_rate": 2.2123014894180194e-06, + "loss": 1.1226, + "step": 28171 + }, + { + "epoch": 2.37, + "learning_rate": 2.2117313139370445e-06, + "loss": 1.2305, + "step": 28172 + }, + { + "epoch": 2.37, + "learning_rate": 2.2111612028052397e-06, + "loss": 1.084, + "step": 28173 + }, + { + "epoch": 2.37, + "learning_rate": 2.2105911560273187e-06, + "loss": 1.1155, + "step": 28174 + }, + { + "epoch": 2.37, + "learning_rate": 2.210021173607979e-06, + "loss": 1.0772, + "step": 28175 + }, + { + "epoch": 2.37, + "learning_rate": 2.209451255551944e-06, + "loss": 1.0486, + "step": 28176 + }, + { + "epoch": 2.37, + "learning_rate": 2.208881401863914e-06, + "loss": 1.0854, + "step": 28177 + }, + { + "epoch": 2.37, + "learning_rate": 2.208311612548598e-06, + "loss": 1.1538, + "step": 28178 + }, + { + "epoch": 2.37, + "learning_rate": 2.207741887610705e-06, + "loss": 1.2341, + "step": 28179 + }, + { + "epoch": 2.37, + "learning_rate": 2.2071722270549457e-06, + "loss": 1.1163, + "step": 28180 + }, + { + "epoch": 2.37, + "learning_rate": 2.206602630886018e-06, + "loss": 1.0673, + "step": 28181 + }, + { + "epoch": 2.37, + "learning_rate": 2.2060330991086388e-06, + "loss": 1.1442, + "step": 28182 + }, + { + "epoch": 2.37, + "learning_rate": 2.205463631727507e-06, + "loss": 1.0515, + "step": 28183 + }, + { + "epoch": 2.37, + "learning_rate": 2.204894228747325e-06, + "loss": 1.0613, + "step": 28184 + }, + { + "epoch": 2.37, + "learning_rate": 2.2043248901728075e-06, + "loss": 1.0529, + "step": 28185 + }, + { + "epoch": 2.37, + "learning_rate": 2.2037556160086502e-06, + "loss": 1.1778, + "step": 28186 + }, + { + "epoch": 2.37, + "learning_rate": 2.203186406259559e-06, + "loss": 1.1322, + "step": 28187 + }, + { + "epoch": 2.37, + "learning_rate": 2.2026172609302364e-06, + "loss": 1.1191, + "step": 28188 + }, + { + "epoch": 2.37, + "learning_rate": 2.202048180025387e-06, + "loss": 1.0652, + "step": 28189 + }, + { + "epoch": 2.37, + "learning_rate": 2.2014791635497103e-06, + "loss": 1.0955, + "step": 28190 + }, + { + "epoch": 2.37, + "learning_rate": 2.2009102115079104e-06, + "loss": 1.162, + "step": 28191 + }, + { + "epoch": 2.37, + "learning_rate": 2.2003413239046824e-06, + "loss": 1.1498, + "step": 28192 + }, + { + "epoch": 2.37, + "learning_rate": 2.199772500744736e-06, + "loss": 1.1043, + "step": 28193 + }, + { + "epoch": 2.37, + "learning_rate": 2.199203742032762e-06, + "loss": 1.2003, + "step": 28194 + }, + { + "epoch": 2.37, + "learning_rate": 2.1986350477734632e-06, + "loss": 1.1612, + "step": 28195 + }, + { + "epoch": 2.37, + "learning_rate": 2.1980664179715394e-06, + "loss": 1.106, + "step": 28196 + }, + { + "epoch": 2.37, + "learning_rate": 2.197497852631687e-06, + "loss": 1.032, + "step": 28197 + }, + { + "epoch": 2.37, + "learning_rate": 2.1969293517586043e-06, + "loss": 1.1963, + "step": 28198 + }, + { + "epoch": 2.37, + "learning_rate": 2.196360915356989e-06, + "loss": 1.224, + "step": 28199 + }, + { + "epoch": 2.37, + "learning_rate": 2.1957925434315374e-06, + "loss": 1.0999, + "step": 28200 + }, + { + "epoch": 2.37, + "learning_rate": 2.1952242359869447e-06, + "loss": 1.0756, + "step": 28201 + }, + { + "epoch": 2.37, + "learning_rate": 2.194655993027911e-06, + "loss": 1.0452, + "step": 28202 + }, + { + "epoch": 2.37, + "learning_rate": 2.1940878145591217e-06, + "loss": 1.0477, + "step": 28203 + }, + { + "epoch": 2.37, + "learning_rate": 2.1935197005852825e-06, + "loss": 1.0934, + "step": 28204 + }, + { + "epoch": 2.37, + "learning_rate": 2.192951651111079e-06, + "loss": 1.1653, + "step": 28205 + }, + { + "epoch": 2.37, + "learning_rate": 2.192383666141208e-06, + "loss": 1.1327, + "step": 28206 + }, + { + "epoch": 2.37, + "learning_rate": 2.191815745680361e-06, + "loss": 1.2074, + "step": 28207 + }, + { + "epoch": 2.37, + "learning_rate": 2.1912478897332322e-06, + "loss": 1.1547, + "step": 28208 + }, + { + "epoch": 2.37, + "learning_rate": 2.190680098304512e-06, + "loss": 1.1309, + "step": 28209 + }, + { + "epoch": 2.38, + "learning_rate": 2.1901123713988925e-06, + "loss": 1.0973, + "step": 28210 + }, + { + "epoch": 2.38, + "learning_rate": 2.189544709021063e-06, + "loss": 1.1439, + "step": 28211 + }, + { + "epoch": 2.38, + "learning_rate": 2.1889771111757164e-06, + "loss": 1.1174, + "step": 28212 + }, + { + "epoch": 2.38, + "learning_rate": 2.188409577867543e-06, + "loss": 1.1748, + "step": 28213 + }, + { + "epoch": 2.38, + "learning_rate": 2.1878421091012236e-06, + "loss": 1.0195, + "step": 28214 + }, + { + "epoch": 2.38, + "learning_rate": 2.187274704881458e-06, + "loss": 1.1082, + "step": 28215 + }, + { + "epoch": 2.38, + "learning_rate": 2.1867073652129277e-06, + "loss": 1.1182, + "step": 28216 + }, + { + "epoch": 2.38, + "learning_rate": 2.1861400901003205e-06, + "loss": 1.0899, + "step": 28217 + }, + { + "epoch": 2.38, + "learning_rate": 2.1855728795483254e-06, + "loss": 1.1976, + "step": 28218 + }, + { + "epoch": 2.38, + "learning_rate": 2.1850057335616284e-06, + "loss": 1.1589, + "step": 28219 + }, + { + "epoch": 2.38, + "learning_rate": 2.1844386521449144e-06, + "loss": 1.1548, + "step": 28220 + }, + { + "epoch": 2.38, + "learning_rate": 2.1838716353028732e-06, + "loss": 1.0928, + "step": 28221 + }, + { + "epoch": 2.38, + "learning_rate": 2.1833046830401795e-06, + "loss": 1.1617, + "step": 28222 + }, + { + "epoch": 2.38, + "learning_rate": 2.182737795361527e-06, + "loss": 1.0174, + "step": 28223 + }, + { + "epoch": 2.38, + "learning_rate": 2.1821709722716e-06, + "loss": 1.0818, + "step": 28224 + }, + { + "epoch": 2.38, + "learning_rate": 2.1816042137750728e-06, + "loss": 1.1437, + "step": 28225 + }, + { + "epoch": 2.38, + "learning_rate": 2.18103751987664e-06, + "loss": 1.0607, + "step": 28226 + }, + { + "epoch": 2.38, + "learning_rate": 2.1804708905809733e-06, + "loss": 1.0625, + "step": 28227 + }, + { + "epoch": 2.38, + "learning_rate": 2.17990432589276e-06, + "loss": 1.0478, + "step": 28228 + }, + { + "epoch": 2.38, + "learning_rate": 2.1793378258166786e-06, + "loss": 1.013, + "step": 28229 + }, + { + "epoch": 2.38, + "learning_rate": 2.178771390357415e-06, + "loss": 1.1975, + "step": 28230 + }, + { + "epoch": 2.38, + "learning_rate": 2.1782050195196393e-06, + "loss": 1.2044, + "step": 28231 + }, + { + "epoch": 2.38, + "learning_rate": 2.1776387133080424e-06, + "loss": 1.1417, + "step": 28232 + }, + { + "epoch": 2.38, + "learning_rate": 2.177072471727296e-06, + "loss": 1.0751, + "step": 28233 + }, + { + "epoch": 2.38, + "learning_rate": 2.176506294782077e-06, + "loss": 1.0797, + "step": 28234 + }, + { + "epoch": 2.38, + "learning_rate": 2.175940182477072e-06, + "loss": 1.1131, + "step": 28235 + }, + { + "epoch": 2.38, + "learning_rate": 2.175374134816951e-06, + "loss": 1.1086, + "step": 28236 + }, + { + "epoch": 2.38, + "learning_rate": 2.1748081518063935e-06, + "loss": 1.1125, + "step": 28237 + }, + { + "epoch": 2.38, + "learning_rate": 2.1742422334500744e-06, + "loss": 1.1502, + "step": 28238 + }, + { + "epoch": 2.38, + "learning_rate": 2.173676379752672e-06, + "loss": 1.0016, + "step": 28239 + }, + { + "epoch": 2.38, + "learning_rate": 2.1731105907188587e-06, + "loss": 1.1534, + "step": 28240 + }, + { + "epoch": 2.38, + "learning_rate": 2.1725448663533134e-06, + "loss": 1.1198, + "step": 28241 + }, + { + "epoch": 2.38, + "learning_rate": 2.1719792066607025e-06, + "loss": 1.0553, + "step": 28242 + }, + { + "epoch": 2.38, + "learning_rate": 2.1714136116457097e-06, + "loss": 1.1843, + "step": 28243 + }, + { + "epoch": 2.38, + "learning_rate": 2.170848081313002e-06, + "loss": 1.0988, + "step": 28244 + }, + { + "epoch": 2.38, + "learning_rate": 2.1702826156672506e-06, + "loss": 1.1975, + "step": 28245 + }, + { + "epoch": 2.38, + "learning_rate": 2.169717214713132e-06, + "loss": 1.1374, + "step": 28246 + }, + { + "epoch": 2.38, + "learning_rate": 2.1691518784553146e-06, + "loss": 1.2333, + "step": 28247 + }, + { + "epoch": 2.38, + "learning_rate": 2.1685866068984718e-06, + "loss": 1.0343, + "step": 28248 + }, + { + "epoch": 2.38, + "learning_rate": 2.168021400047272e-06, + "loss": 1.0091, + "step": 28249 + }, + { + "epoch": 2.38, + "learning_rate": 2.1674562579063863e-06, + "loss": 1.1696, + "step": 28250 + }, + { + "epoch": 2.38, + "learning_rate": 2.1668911804804826e-06, + "loss": 1.0882, + "step": 28251 + }, + { + "epoch": 2.38, + "learning_rate": 2.166326167774235e-06, + "loss": 1.1115, + "step": 28252 + }, + { + "epoch": 2.38, + "learning_rate": 2.1657612197923016e-06, + "loss": 1.0301, + "step": 28253 + }, + { + "epoch": 2.38, + "learning_rate": 2.1651963365393625e-06, + "loss": 1.1247, + "step": 28254 + }, + { + "epoch": 2.38, + "learning_rate": 2.164631518020075e-06, + "loss": 1.1371, + "step": 28255 + }, + { + "epoch": 2.38, + "learning_rate": 2.16406676423911e-06, + "loss": 1.0804, + "step": 28256 + }, + { + "epoch": 2.38, + "learning_rate": 2.1635020752011328e-06, + "loss": 1.1803, + "step": 28257 + }, + { + "epoch": 2.38, + "learning_rate": 2.1629374509108105e-06, + "loss": 1.1704, + "step": 28258 + }, + { + "epoch": 2.38, + "learning_rate": 2.1623728913728057e-06, + "loss": 0.9875, + "step": 28259 + }, + { + "epoch": 2.38, + "learning_rate": 2.1618083965917857e-06, + "loss": 1.0973, + "step": 28260 + }, + { + "epoch": 2.38, + "learning_rate": 2.1612439665724128e-06, + "loss": 1.1438, + "step": 28261 + }, + { + "epoch": 2.38, + "learning_rate": 2.160679601319351e-06, + "loss": 1.1197, + "step": 28262 + }, + { + "epoch": 2.38, + "learning_rate": 2.1601153008372657e-06, + "loss": 1.1009, + "step": 28263 + }, + { + "epoch": 2.38, + "learning_rate": 2.1595510651308114e-06, + "loss": 1.2249, + "step": 28264 + }, + { + "epoch": 2.38, + "learning_rate": 2.158986894204661e-06, + "loss": 1.0879, + "step": 28265 + }, + { + "epoch": 2.38, + "learning_rate": 2.1584227880634678e-06, + "loss": 1.1687, + "step": 28266 + }, + { + "epoch": 2.38, + "learning_rate": 2.1578587467118962e-06, + "loss": 1.1802, + "step": 28267 + }, + { + "epoch": 2.38, + "learning_rate": 2.1572947701546034e-06, + "loss": 1.1607, + "step": 28268 + }, + { + "epoch": 2.38, + "learning_rate": 2.156730858396252e-06, + "loss": 1.0238, + "step": 28269 + }, + { + "epoch": 2.38, + "learning_rate": 2.156167011441501e-06, + "loss": 1.027, + "step": 28270 + }, + { + "epoch": 2.38, + "learning_rate": 2.1556032292950113e-06, + "loss": 1.1994, + "step": 28271 + }, + { + "epoch": 2.38, + "learning_rate": 2.155039511961432e-06, + "loss": 1.1559, + "step": 28272 + }, + { + "epoch": 2.38, + "learning_rate": 2.1544758594454297e-06, + "loss": 1.2556, + "step": 28273 + }, + { + "epoch": 2.38, + "learning_rate": 2.1539122717516612e-06, + "loss": 1.1742, + "step": 28274 + }, + { + "epoch": 2.38, + "learning_rate": 2.153348748884776e-06, + "loss": 1.0132, + "step": 28275 + }, + { + "epoch": 2.38, + "learning_rate": 2.1527852908494395e-06, + "loss": 1.1251, + "step": 28276 + }, + { + "epoch": 2.38, + "learning_rate": 2.1522218976503007e-06, + "loss": 1.1588, + "step": 28277 + }, + { + "epoch": 2.38, + "learning_rate": 2.1516585692920155e-06, + "loss": 1.1462, + "step": 28278 + }, + { + "epoch": 2.38, + "learning_rate": 2.1510953057792384e-06, + "loss": 1.0846, + "step": 28279 + }, + { + "epoch": 2.38, + "learning_rate": 2.1505321071166243e-06, + "loss": 1.0665, + "step": 28280 + }, + { + "epoch": 2.38, + "learning_rate": 2.149968973308826e-06, + "loss": 1.0802, + "step": 28281 + }, + { + "epoch": 2.38, + "learning_rate": 2.1494059043604997e-06, + "loss": 1.0883, + "step": 28282 + }, + { + "epoch": 2.38, + "learning_rate": 2.148842900276288e-06, + "loss": 1.0942, + "step": 28283 + }, + { + "epoch": 2.38, + "learning_rate": 2.148279961060853e-06, + "loss": 1.1405, + "step": 28284 + }, + { + "epoch": 2.38, + "learning_rate": 2.1477170867188435e-06, + "loss": 1.0088, + "step": 28285 + }, + { + "epoch": 2.38, + "learning_rate": 2.1471542772549027e-06, + "loss": 1.1505, + "step": 28286 + }, + { + "epoch": 2.38, + "learning_rate": 2.146591532673693e-06, + "loss": 1.1302, + "step": 28287 + }, + { + "epoch": 2.38, + "learning_rate": 2.1460288529798536e-06, + "loss": 1.0733, + "step": 28288 + }, + { + "epoch": 2.38, + "learning_rate": 2.1454662381780376e-06, + "loss": 1.1, + "step": 28289 + }, + { + "epoch": 2.38, + "learning_rate": 2.1449036882728948e-06, + "loss": 1.1373, + "step": 28290 + }, + { + "epoch": 2.38, + "learning_rate": 2.1443412032690723e-06, + "loss": 1.1631, + "step": 28291 + }, + { + "epoch": 2.38, + "learning_rate": 2.1437787831712132e-06, + "loss": 1.1337, + "step": 28292 + }, + { + "epoch": 2.38, + "learning_rate": 2.143216427983973e-06, + "loss": 1.0601, + "step": 28293 + }, + { + "epoch": 2.38, + "learning_rate": 2.1426541377119903e-06, + "loss": 1.1563, + "step": 28294 + }, + { + "epoch": 2.38, + "learning_rate": 2.1420919123599137e-06, + "loss": 1.1106, + "step": 28295 + }, + { + "epoch": 2.38, + "learning_rate": 2.1415297519323887e-06, + "loss": 1.0738, + "step": 28296 + }, + { + "epoch": 2.38, + "learning_rate": 2.14096765643406e-06, + "loss": 1.1086, + "step": 28297 + }, + { + "epoch": 2.38, + "learning_rate": 2.1404056258695706e-06, + "loss": 1.2202, + "step": 28298 + }, + { + "epoch": 2.38, + "learning_rate": 2.1398436602435667e-06, + "loss": 1.1935, + "step": 28299 + }, + { + "epoch": 2.38, + "learning_rate": 2.1392817595606897e-06, + "loss": 1.0932, + "step": 28300 + }, + { + "epoch": 2.38, + "learning_rate": 2.1387199238255817e-06, + "loss": 1.1491, + "step": 28301 + }, + { + "epoch": 2.38, + "learning_rate": 2.1381581530428898e-06, + "loss": 1.044, + "step": 28302 + }, + { + "epoch": 2.38, + "learning_rate": 2.137596447217245e-06, + "loss": 1.1694, + "step": 28303 + }, + { + "epoch": 2.38, + "learning_rate": 2.1370348063533007e-06, + "loss": 1.1884, + "step": 28304 + }, + { + "epoch": 2.38, + "learning_rate": 2.136473230455688e-06, + "loss": 1.1804, + "step": 28305 + }, + { + "epoch": 2.38, + "learning_rate": 2.13591171952905e-06, + "loss": 1.14, + "step": 28306 + }, + { + "epoch": 2.38, + "learning_rate": 2.1353502735780274e-06, + "loss": 1.0544, + "step": 28307 + }, + { + "epoch": 2.38, + "learning_rate": 2.134788892607257e-06, + "loss": 0.9848, + "step": 28308 + }, + { + "epoch": 2.38, + "learning_rate": 2.1342275766213774e-06, + "loss": 1.1353, + "step": 28309 + }, + { + "epoch": 2.38, + "learning_rate": 2.133666325625028e-06, + "loss": 1.1467, + "step": 28310 + }, + { + "epoch": 2.38, + "learning_rate": 2.133105139622844e-06, + "loss": 1.0552, + "step": 28311 + }, + { + "epoch": 2.38, + "learning_rate": 2.1325440186194633e-06, + "loss": 1.1271, + "step": 28312 + }, + { + "epoch": 2.38, + "learning_rate": 2.1319829626195253e-06, + "loss": 1.0973, + "step": 28313 + }, + { + "epoch": 2.38, + "learning_rate": 2.1314219716276563e-06, + "loss": 1.0744, + "step": 28314 + }, + { + "epoch": 2.38, + "learning_rate": 2.1308610456485022e-06, + "loss": 1.117, + "step": 28315 + }, + { + "epoch": 2.38, + "learning_rate": 2.13030018468669e-06, + "loss": 1.0763, + "step": 28316 + }, + { + "epoch": 2.38, + "learning_rate": 2.1297393887468563e-06, + "loss": 1.1659, + "step": 28317 + }, + { + "epoch": 2.38, + "learning_rate": 2.129178657833635e-06, + "loss": 1.1802, + "step": 28318 + }, + { + "epoch": 2.38, + "learning_rate": 2.128617991951658e-06, + "loss": 1.1477, + "step": 28319 + }, + { + "epoch": 2.38, + "learning_rate": 2.128057391105559e-06, + "loss": 1.171, + "step": 28320 + }, + { + "epoch": 2.38, + "learning_rate": 2.127496855299972e-06, + "loss": 1.1037, + "step": 28321 + }, + { + "epoch": 2.38, + "learning_rate": 2.1269363845395195e-06, + "loss": 1.0712, + "step": 28322 + }, + { + "epoch": 2.38, + "learning_rate": 2.1263759788288416e-06, + "loss": 1.0864, + "step": 28323 + }, + { + "epoch": 2.38, + "learning_rate": 2.125815638172567e-06, + "loss": 1.0501, + "step": 28324 + }, + { + "epoch": 2.38, + "learning_rate": 2.125255362575319e-06, + "loss": 1.0954, + "step": 28325 + }, + { + "epoch": 2.38, + "learning_rate": 2.124695152041737e-06, + "loss": 1.203, + "step": 28326 + }, + { + "epoch": 2.38, + "learning_rate": 2.124135006576441e-06, + "loss": 1.0933, + "step": 28327 + }, + { + "epoch": 2.38, + "learning_rate": 2.123574926184062e-06, + "loss": 1.1167, + "step": 28328 + }, + { + "epoch": 2.39, + "learning_rate": 2.1230149108692277e-06, + "loss": 1.0191, + "step": 28329 + }, + { + "epoch": 2.39, + "learning_rate": 2.1224549606365653e-06, + "loss": 1.0906, + "step": 28330 + }, + { + "epoch": 2.39, + "learning_rate": 2.121895075490702e-06, + "loss": 1.1781, + "step": 28331 + }, + { + "epoch": 2.39, + "learning_rate": 2.1213352554362643e-06, + "loss": 1.1073, + "step": 28332 + }, + { + "epoch": 2.39, + "learning_rate": 2.120775500477872e-06, + "loss": 1.1293, + "step": 28333 + }, + { + "epoch": 2.39, + "learning_rate": 2.120215810620155e-06, + "loss": 1.148, + "step": 28334 + }, + { + "epoch": 2.39, + "learning_rate": 2.119656185867741e-06, + "loss": 1.0709, + "step": 28335 + }, + { + "epoch": 2.39, + "learning_rate": 2.1190966262252443e-06, + "loss": 1.0838, + "step": 28336 + }, + { + "epoch": 2.39, + "learning_rate": 2.118537131697299e-06, + "loss": 1.1314, + "step": 28337 + }, + { + "epoch": 2.39, + "learning_rate": 2.1179777022885194e-06, + "loss": 1.2156, + "step": 28338 + }, + { + "epoch": 2.39, + "learning_rate": 2.1174183380035305e-06, + "loss": 1.1085, + "step": 28339 + }, + { + "epoch": 2.39, + "learning_rate": 2.1168590388469536e-06, + "loss": 1.1353, + "step": 28340 + }, + { + "epoch": 2.39, + "learning_rate": 2.116299804823413e-06, + "loss": 1.1549, + "step": 28341 + }, + { + "epoch": 2.39, + "learning_rate": 2.115740635937522e-06, + "loss": 1.0952, + "step": 28342 + }, + { + "epoch": 2.39, + "learning_rate": 2.1151815321939096e-06, + "loss": 1.0944, + "step": 28343 + }, + { + "epoch": 2.39, + "learning_rate": 2.114622493597188e-06, + "loss": 1.0707, + "step": 28344 + }, + { + "epoch": 2.39, + "learning_rate": 2.1140635201519798e-06, + "loss": 1.2288, + "step": 28345 + }, + { + "epoch": 2.39, + "learning_rate": 2.113504611862902e-06, + "loss": 1.1338, + "step": 28346 + }, + { + "epoch": 2.39, + "learning_rate": 2.1129457687345746e-06, + "loss": 1.1128, + "step": 28347 + }, + { + "epoch": 2.39, + "learning_rate": 2.112386990771612e-06, + "loss": 1.0374, + "step": 28348 + }, + { + "epoch": 2.39, + "learning_rate": 2.1118282779786325e-06, + "loss": 1.0698, + "step": 28349 + }, + { + "epoch": 2.39, + "learning_rate": 2.111269630360252e-06, + "loss": 1.1832, + "step": 28350 + }, + { + "epoch": 2.39, + "learning_rate": 2.110711047921088e-06, + "loss": 1.1978, + "step": 28351 + }, + { + "epoch": 2.39, + "learning_rate": 2.1101525306657557e-06, + "loss": 1.0925, + "step": 28352 + }, + { + "epoch": 2.39, + "learning_rate": 2.1095940785988633e-06, + "loss": 1.1655, + "step": 28353 + }, + { + "epoch": 2.39, + "learning_rate": 2.1090356917250344e-06, + "loss": 1.0639, + "step": 28354 + }, + { + "epoch": 2.39, + "learning_rate": 2.108477370048876e-06, + "loss": 1.2785, + "step": 28355 + }, + { + "epoch": 2.39, + "learning_rate": 2.107919113575003e-06, + "loss": 1.1525, + "step": 28356 + }, + { + "epoch": 2.39, + "learning_rate": 2.1073609223080293e-06, + "loss": 1.1079, + "step": 28357 + }, + { + "epoch": 2.39, + "learning_rate": 2.106802796252564e-06, + "loss": 1.1307, + "step": 28358 + }, + { + "epoch": 2.39, + "learning_rate": 2.106244735413221e-06, + "loss": 1.0529, + "step": 28359 + }, + { + "epoch": 2.39, + "learning_rate": 2.10568673979461e-06, + "loss": 1.1205, + "step": 28360 + }, + { + "epoch": 2.39, + "learning_rate": 2.105128809401342e-06, + "loss": 1.0976, + "step": 28361 + }, + { + "epoch": 2.39, + "learning_rate": 2.1045709442380258e-06, + "loss": 1.1065, + "step": 28362 + }, + { + "epoch": 2.39, + "learning_rate": 2.104013144309275e-06, + "loss": 1.1371, + "step": 28363 + }, + { + "epoch": 2.39, + "learning_rate": 2.103455409619688e-06, + "loss": 1.1607, + "step": 28364 + }, + { + "epoch": 2.39, + "learning_rate": 2.1028977401738858e-06, + "loss": 1.116, + "step": 28365 + }, + { + "epoch": 2.39, + "learning_rate": 2.1023401359764674e-06, + "loss": 1.1113, + "step": 28366 + }, + { + "epoch": 2.39, + "learning_rate": 2.1017825970320426e-06, + "loss": 1.1412, + "step": 28367 + }, + { + "epoch": 2.39, + "learning_rate": 2.1012251233452164e-06, + "loss": 1.1483, + "step": 28368 + }, + { + "epoch": 2.39, + "learning_rate": 2.1006677149205965e-06, + "loss": 1.0892, + "step": 28369 + }, + { + "epoch": 2.39, + "learning_rate": 2.1001103717627892e-06, + "loss": 1.1486, + "step": 28370 + }, + { + "epoch": 2.39, + "learning_rate": 2.099553093876401e-06, + "loss": 1.0869, + "step": 28371 + }, + { + "epoch": 2.39, + "learning_rate": 2.098995881266027e-06, + "loss": 1.127, + "step": 28372 + }, + { + "epoch": 2.39, + "learning_rate": 2.098438733936281e-06, + "loss": 1.0385, + "step": 28373 + }, + { + "epoch": 2.39, + "learning_rate": 2.0978816518917655e-06, + "loss": 1.0761, + "step": 28374 + }, + { + "epoch": 2.39, + "learning_rate": 2.0973246351370757e-06, + "loss": 1.1329, + "step": 28375 + }, + { + "epoch": 2.39, + "learning_rate": 2.0967676836768235e-06, + "loss": 1.1262, + "step": 28376 + }, + { + "epoch": 2.39, + "learning_rate": 2.0962107975156042e-06, + "loss": 1.0651, + "step": 28377 + }, + { + "epoch": 2.39, + "learning_rate": 2.095653976658021e-06, + "loss": 1.1564, + "step": 28378 + }, + { + "epoch": 2.39, + "learning_rate": 2.095097221108674e-06, + "loss": 1.1276, + "step": 28379 + }, + { + "epoch": 2.39, + "learning_rate": 2.0945405308721634e-06, + "loss": 1.1629, + "step": 28380 + }, + { + "epoch": 2.39, + "learning_rate": 2.093983905953089e-06, + "loss": 1.0133, + "step": 28381 + }, + { + "epoch": 2.39, + "learning_rate": 2.093427346356053e-06, + "loss": 1.1261, + "step": 28382 + }, + { + "epoch": 2.39, + "learning_rate": 2.0928708520856444e-06, + "loss": 1.1241, + "step": 28383 + }, + { + "epoch": 2.39, + "learning_rate": 2.092314423146471e-06, + "loss": 1.0487, + "step": 28384 + }, + { + "epoch": 2.39, + "learning_rate": 2.0917580595431286e-06, + "loss": 1.1421, + "step": 28385 + }, + { + "epoch": 2.39, + "learning_rate": 2.0912017612802073e-06, + "loss": 1.191, + "step": 28386 + }, + { + "epoch": 2.39, + "learning_rate": 2.090645528362314e-06, + "loss": 1.1885, + "step": 28387 + }, + { + "epoch": 2.39, + "learning_rate": 2.0900893607940354e-06, + "loss": 1.0681, + "step": 28388 + }, + { + "epoch": 2.39, + "learning_rate": 2.08953325857997e-06, + "loss": 1.1069, + "step": 28389 + }, + { + "epoch": 2.39, + "learning_rate": 2.0889772217247117e-06, + "loss": 1.1246, + "step": 28390 + }, + { + "epoch": 2.39, + "learning_rate": 2.0884212502328595e-06, + "loss": 1.0615, + "step": 28391 + }, + { + "epoch": 2.39, + "learning_rate": 2.087865344108997e-06, + "loss": 1.1532, + "step": 28392 + }, + { + "epoch": 2.39, + "learning_rate": 2.0873095033577284e-06, + "loss": 1.1624, + "step": 28393 + }, + { + "epoch": 2.39, + "learning_rate": 2.086753727983639e-06, + "loss": 1.1096, + "step": 28394 + }, + { + "epoch": 2.39, + "learning_rate": 2.086198017991322e-06, + "loss": 1.0604, + "step": 28395 + }, + { + "epoch": 2.39, + "learning_rate": 2.08564237338537e-06, + "loss": 1.0897, + "step": 28396 + }, + { + "epoch": 2.39, + "learning_rate": 2.0850867941703747e-06, + "loss": 1.0832, + "step": 28397 + }, + { + "epoch": 2.39, + "learning_rate": 2.084531280350924e-06, + "loss": 1.2184, + "step": 28398 + }, + { + "epoch": 2.39, + "learning_rate": 2.0839758319316096e-06, + "loss": 1.0187, + "step": 28399 + }, + { + "epoch": 2.39, + "learning_rate": 2.0834204489170206e-06, + "loss": 1.151, + "step": 28400 + }, + { + "epoch": 2.39, + "learning_rate": 2.0828651313117452e-06, + "loss": 1.1347, + "step": 28401 + }, + { + "epoch": 2.39, + "learning_rate": 2.0823098791203743e-06, + "loss": 1.1598, + "step": 28402 + }, + { + "epoch": 2.39, + "learning_rate": 2.0817546923474886e-06, + "loss": 1.1855, + "step": 28403 + }, + { + "epoch": 2.39, + "learning_rate": 2.081199570997684e-06, + "loss": 1.1201, + "step": 28404 + }, + { + "epoch": 2.39, + "learning_rate": 2.0806445150755416e-06, + "loss": 1.14, + "step": 28405 + }, + { + "epoch": 2.39, + "learning_rate": 2.080089524585648e-06, + "loss": 1.0794, + "step": 28406 + }, + { + "epoch": 2.39, + "learning_rate": 2.079534599532589e-06, + "loss": 1.1571, + "step": 28407 + }, + { + "epoch": 2.39, + "learning_rate": 2.078979739920951e-06, + "loss": 1.102, + "step": 28408 + }, + { + "epoch": 2.39, + "learning_rate": 2.0784249457553163e-06, + "loss": 1.1641, + "step": 28409 + }, + { + "epoch": 2.39, + "learning_rate": 2.077870217040271e-06, + "loss": 1.089, + "step": 28410 + }, + { + "epoch": 2.39, + "learning_rate": 2.077315553780397e-06, + "loss": 1.1431, + "step": 28411 + }, + { + "epoch": 2.39, + "learning_rate": 2.0767609559802773e-06, + "loss": 1.1886, + "step": 28412 + }, + { + "epoch": 2.39, + "learning_rate": 2.0762064236444966e-06, + "loss": 1.1047, + "step": 28413 + }, + { + "epoch": 2.39, + "learning_rate": 2.07565195677763e-06, + "loss": 1.0768, + "step": 28414 + }, + { + "epoch": 2.39, + "learning_rate": 2.075097555384268e-06, + "loss": 1.116, + "step": 28415 + }, + { + "epoch": 2.39, + "learning_rate": 2.074543219468983e-06, + "loss": 1.0717, + "step": 28416 + }, + { + "epoch": 2.39, + "learning_rate": 2.0739889490363597e-06, + "loss": 1.1756, + "step": 28417 + }, + { + "epoch": 2.39, + "learning_rate": 2.0734347440909753e-06, + "loss": 1.0986, + "step": 28418 + }, + { + "epoch": 2.39, + "learning_rate": 2.0728806046374105e-06, + "loss": 1.0794, + "step": 28419 + }, + { + "epoch": 2.39, + "learning_rate": 2.072326530680242e-06, + "loss": 1.1608, + "step": 28420 + }, + { + "epoch": 2.39, + "learning_rate": 2.0717725222240524e-06, + "loss": 1.1306, + "step": 28421 + }, + { + "epoch": 2.39, + "learning_rate": 2.0712185792734107e-06, + "loss": 1.1061, + "step": 28422 + }, + { + "epoch": 2.39, + "learning_rate": 2.0706647018329007e-06, + "loss": 0.9844, + "step": 28423 + }, + { + "epoch": 2.39, + "learning_rate": 2.0701108899070998e-06, + "loss": 1.0651, + "step": 28424 + }, + { + "epoch": 2.39, + "learning_rate": 2.0695571435005746e-06, + "loss": 1.2795, + "step": 28425 + }, + { + "epoch": 2.39, + "learning_rate": 2.0690034626179123e-06, + "loss": 1.1685, + "step": 28426 + }, + { + "epoch": 2.39, + "learning_rate": 2.0684498472636784e-06, + "loss": 1.1343, + "step": 28427 + }, + { + "epoch": 2.39, + "learning_rate": 2.067896297442451e-06, + "loss": 1.0799, + "step": 28428 + }, + { + "epoch": 2.39, + "learning_rate": 2.0673428131588024e-06, + "loss": 1.1789, + "step": 28429 + }, + { + "epoch": 2.39, + "learning_rate": 2.0667893944173057e-06, + "loss": 1.1123, + "step": 28430 + }, + { + "epoch": 2.39, + "learning_rate": 2.0662360412225346e-06, + "loss": 1.1425, + "step": 28431 + }, + { + "epoch": 2.39, + "learning_rate": 2.0656827535790635e-06, + "loss": 1.1081, + "step": 28432 + }, + { + "epoch": 2.39, + "learning_rate": 2.0651295314914553e-06, + "loss": 1.189, + "step": 28433 + }, + { + "epoch": 2.39, + "learning_rate": 2.0645763749642887e-06, + "loss": 1.1391, + "step": 28434 + }, + { + "epoch": 2.39, + "learning_rate": 2.0640232840021348e-06, + "loss": 1.1003, + "step": 28435 + }, + { + "epoch": 2.39, + "learning_rate": 2.0634702586095547e-06, + "loss": 1.146, + "step": 28436 + }, + { + "epoch": 2.39, + "learning_rate": 2.0629172987911284e-06, + "loss": 1.0843, + "step": 28437 + }, + { + "epoch": 2.39, + "learning_rate": 2.062364404551418e-06, + "loss": 1.1061, + "step": 28438 + }, + { + "epoch": 2.39, + "learning_rate": 2.061811575894992e-06, + "loss": 1.072, + "step": 28439 + }, + { + "epoch": 2.39, + "learning_rate": 2.06125881282642e-06, + "loss": 1.0586, + "step": 28440 + }, + { + "epoch": 2.39, + "learning_rate": 2.0607061153502717e-06, + "loss": 1.075, + "step": 28441 + }, + { + "epoch": 2.39, + "learning_rate": 2.0601534834711033e-06, + "loss": 1.1786, + "step": 28442 + }, + { + "epoch": 2.39, + "learning_rate": 2.059600917193495e-06, + "loss": 0.9727, + "step": 28443 + }, + { + "epoch": 2.39, + "learning_rate": 2.0590484165220014e-06, + "loss": 1.0411, + "step": 28444 + }, + { + "epoch": 2.39, + "learning_rate": 2.0584959814611916e-06, + "loss": 1.204, + "step": 28445 + }, + { + "epoch": 2.39, + "learning_rate": 2.0579436120156295e-06, + "loss": 1.0351, + "step": 28446 + }, + { + "epoch": 2.39, + "learning_rate": 2.0573913081898787e-06, + "loss": 1.198, + "step": 28447 + }, + { + "epoch": 2.4, + "learning_rate": 2.0568390699885045e-06, + "loss": 1.098, + "step": 28448 + }, + { + "epoch": 2.4, + "learning_rate": 2.0562868974160667e-06, + "loss": 1.1741, + "step": 28449 + }, + { + "epoch": 2.4, + "learning_rate": 2.0557347904771284e-06, + "loss": 1.1288, + "step": 28450 + }, + { + "epoch": 2.4, + "learning_rate": 2.055182749176252e-06, + "loss": 1.0533, + "step": 28451 + }, + { + "epoch": 2.4, + "learning_rate": 2.0546307735180016e-06, + "loss": 1.1775, + "step": 28452 + }, + { + "epoch": 2.4, + "learning_rate": 2.0540788635069298e-06, + "loss": 1.0712, + "step": 28453 + }, + { + "epoch": 2.4, + "learning_rate": 2.0535270191476054e-06, + "loss": 1.1715, + "step": 28454 + }, + { + "epoch": 2.4, + "learning_rate": 2.0529752404445822e-06, + "loss": 1.1189, + "step": 28455 + }, + { + "epoch": 2.4, + "learning_rate": 2.0524235274024206e-06, + "loss": 1.2127, + "step": 28456 + }, + { + "epoch": 2.4, + "learning_rate": 2.0518718800256798e-06, + "loss": 1.1392, + "step": 28457 + }, + { + "epoch": 2.4, + "learning_rate": 2.051320298318917e-06, + "loss": 1.1052, + "step": 28458 + }, + { + "epoch": 2.4, + "learning_rate": 2.0507687822866897e-06, + "loss": 1.149, + "step": 28459 + }, + { + "epoch": 2.4, + "learning_rate": 2.050217331933555e-06, + "loss": 1.1303, + "step": 28460 + }, + { + "epoch": 2.4, + "learning_rate": 2.049665947264068e-06, + "loss": 1.0593, + "step": 28461 + }, + { + "epoch": 2.4, + "learning_rate": 2.0491146282827867e-06, + "loss": 1.0312, + "step": 28462 + }, + { + "epoch": 2.4, + "learning_rate": 2.048563374994267e-06, + "loss": 1.1323, + "step": 28463 + }, + { + "epoch": 2.4, + "learning_rate": 2.0480121874030555e-06, + "loss": 1.1679, + "step": 28464 + }, + { + "epoch": 2.4, + "learning_rate": 2.0474610655137195e-06, + "loss": 1.0667, + "step": 28465 + }, + { + "epoch": 2.4, + "learning_rate": 2.0469100093308016e-06, + "loss": 1.0567, + "step": 28466 + }, + { + "epoch": 2.4, + "learning_rate": 2.0463590188588577e-06, + "loss": 1.0485, + "step": 28467 + }, + { + "epoch": 2.4, + "learning_rate": 2.0458080941024417e-06, + "loss": 1.1024, + "step": 28468 + }, + { + "epoch": 2.4, + "learning_rate": 2.0452572350661047e-06, + "loss": 1.175, + "step": 28469 + }, + { + "epoch": 2.4, + "learning_rate": 2.0447064417543994e-06, + "loss": 1.1651, + "step": 28470 + }, + { + "epoch": 2.4, + "learning_rate": 2.044155714171877e-06, + "loss": 0.9753, + "step": 28471 + }, + { + "epoch": 2.4, + "learning_rate": 2.043605052323081e-06, + "loss": 1.0742, + "step": 28472 + }, + { + "epoch": 2.4, + "learning_rate": 2.0430544562125696e-06, + "loss": 1.0988, + "step": 28473 + }, + { + "epoch": 2.4, + "learning_rate": 2.042503925844891e-06, + "loss": 0.9633, + "step": 28474 + }, + { + "epoch": 2.4, + "learning_rate": 2.0419534612245863e-06, + "loss": 1.1239, + "step": 28475 + }, + { + "epoch": 2.4, + "learning_rate": 2.0414030623562154e-06, + "loss": 1.0963, + "step": 28476 + }, + { + "epoch": 2.4, + "learning_rate": 2.040852729244316e-06, + "loss": 1.1792, + "step": 28477 + }, + { + "epoch": 2.4, + "learning_rate": 2.0403024618934386e-06, + "loss": 1.1787, + "step": 28478 + }, + { + "epoch": 2.4, + "learning_rate": 2.03975226030813e-06, + "loss": 1.1297, + "step": 28479 + }, + { + "epoch": 2.4, + "learning_rate": 2.039202124492935e-06, + "loss": 1.1178, + "step": 28480 + }, + { + "epoch": 2.4, + "learning_rate": 2.0386520544524003e-06, + "loss": 1.1137, + "step": 28481 + }, + { + "epoch": 2.4, + "learning_rate": 2.038102050191073e-06, + "loss": 1.065, + "step": 28482 + }, + { + "epoch": 2.4, + "learning_rate": 2.03755211171349e-06, + "loss": 1.1281, + "step": 28483 + }, + { + "epoch": 2.4, + "learning_rate": 2.0370022390242018e-06, + "loss": 1.138, + "step": 28484 + }, + { + "epoch": 2.4, + "learning_rate": 2.036452432127752e-06, + "loss": 1.1426, + "step": 28485 + }, + { + "epoch": 2.4, + "learning_rate": 2.0359026910286763e-06, + "loss": 1.0438, + "step": 28486 + }, + { + "epoch": 2.4, + "learning_rate": 2.035353015731526e-06, + "loss": 1.092, + "step": 28487 + }, + { + "epoch": 2.4, + "learning_rate": 2.034803406240835e-06, + "loss": 1.0993, + "step": 28488 + }, + { + "epoch": 2.4, + "learning_rate": 2.0342538625611486e-06, + "loss": 1.151, + "step": 28489 + }, + { + "epoch": 2.4, + "learning_rate": 2.033704384697006e-06, + "loss": 1.121, + "step": 28490 + }, + { + "epoch": 2.4, + "learning_rate": 2.0331549726529497e-06, + "loss": 1.09, + "step": 28491 + }, + { + "epoch": 2.4, + "learning_rate": 2.0326056264335116e-06, + "loss": 0.9894, + "step": 28492 + }, + { + "epoch": 2.4, + "learning_rate": 2.032056346043241e-06, + "loss": 1.0349, + "step": 28493 + }, + { + "epoch": 2.4, + "learning_rate": 2.0315071314866675e-06, + "loss": 1.0795, + "step": 28494 + }, + { + "epoch": 2.4, + "learning_rate": 2.0309579827683334e-06, + "loss": 1.007, + "step": 28495 + }, + { + "epoch": 2.4, + "learning_rate": 2.030408899892774e-06, + "loss": 1.1674, + "step": 28496 + }, + { + "epoch": 2.4, + "learning_rate": 2.0298598828645265e-06, + "loss": 1.1001, + "step": 28497 + }, + { + "epoch": 2.4, + "learning_rate": 2.029310931688129e-06, + "loss": 1.1213, + "step": 28498 + }, + { + "epoch": 2.4, + "learning_rate": 2.0287620463681147e-06, + "loss": 1.1345, + "step": 28499 + }, + { + "epoch": 2.4, + "learning_rate": 2.0282132269090195e-06, + "loss": 1.0707, + "step": 28500 + }, + { + "epoch": 2.4, + "learning_rate": 2.0276644733153773e-06, + "loss": 1.0743, + "step": 28501 + }, + { + "epoch": 2.4, + "learning_rate": 2.0271157855917266e-06, + "loss": 1.12, + "step": 28502 + }, + { + "epoch": 2.4, + "learning_rate": 2.026567163742591e-06, + "loss": 1.0527, + "step": 28503 + }, + { + "epoch": 2.4, + "learning_rate": 2.026018607772514e-06, + "loss": 1.0771, + "step": 28504 + }, + { + "epoch": 2.4, + "learning_rate": 2.0254701176860216e-06, + "loss": 1.0897, + "step": 28505 + }, + { + "epoch": 2.4, + "learning_rate": 2.024921693487647e-06, + "loss": 1.0916, + "step": 28506 + }, + { + "epoch": 2.4, + "learning_rate": 2.0243733351819215e-06, + "loss": 1.1839, + "step": 28507 + }, + { + "epoch": 2.4, + "learning_rate": 2.0238250427733762e-06, + "loss": 1.132, + "step": 28508 + }, + { + "epoch": 2.4, + "learning_rate": 2.0232768162665405e-06, + "loss": 1.0175, + "step": 28509 + }, + { + "epoch": 2.4, + "learning_rate": 2.0227286556659453e-06, + "loss": 1.1073, + "step": 28510 + }, + { + "epoch": 2.4, + "learning_rate": 2.022180560976119e-06, + "loss": 1.106, + "step": 28511 + }, + { + "epoch": 2.4, + "learning_rate": 2.0216325322015906e-06, + "loss": 1.1866, + "step": 28512 + }, + { + "epoch": 2.4, + "learning_rate": 2.0210845693468895e-06, + "loss": 1.1595, + "step": 28513 + }, + { + "epoch": 2.4, + "learning_rate": 2.0205366724165352e-06, + "loss": 1.074, + "step": 28514 + }, + { + "epoch": 2.4, + "learning_rate": 2.019988841415068e-06, + "loss": 1.0971, + "step": 28515 + }, + { + "epoch": 2.4, + "learning_rate": 2.0194410763470027e-06, + "loss": 1.0709, + "step": 28516 + }, + { + "epoch": 2.4, + "learning_rate": 2.0188933772168695e-06, + "loss": 1.2012, + "step": 28517 + }, + { + "epoch": 2.4, + "learning_rate": 2.0183457440291944e-06, + "loss": 1.1601, + "step": 28518 + }, + { + "epoch": 2.4, + "learning_rate": 2.0177981767885e-06, + "loss": 1.0808, + "step": 28519 + }, + { + "epoch": 2.4, + "learning_rate": 2.017250675499313e-06, + "loss": 1.1148, + "step": 28520 + }, + { + "epoch": 2.4, + "learning_rate": 2.016703240166158e-06, + "loss": 1.2148, + "step": 28521 + }, + { + "epoch": 2.4, + "learning_rate": 2.01615587079355e-06, + "loss": 1.0576, + "step": 28522 + }, + { + "epoch": 2.4, + "learning_rate": 2.0156085673860205e-06, + "loss": 1.1123, + "step": 28523 + }, + { + "epoch": 2.4, + "learning_rate": 2.0150613299480905e-06, + "loss": 1.1254, + "step": 28524 + }, + { + "epoch": 2.4, + "learning_rate": 2.0145141584842744e-06, + "loss": 1.1566, + "step": 28525 + }, + { + "epoch": 2.4, + "learning_rate": 2.013967052999103e-06, + "loss": 0.9916, + "step": 28526 + }, + { + "epoch": 2.4, + "learning_rate": 2.013420013497088e-06, + "loss": 1.1531, + "step": 28527 + }, + { + "epoch": 2.4, + "learning_rate": 2.0128730399827547e-06, + "loss": 1.096, + "step": 28528 + }, + { + "epoch": 2.4, + "learning_rate": 2.0123261324606205e-06, + "loss": 1.2089, + "step": 28529 + }, + { + "epoch": 2.4, + "learning_rate": 2.011779290935203e-06, + "loss": 1.1466, + "step": 28530 + }, + { + "epoch": 2.4, + "learning_rate": 2.0112325154110225e-06, + "loss": 1.1762, + "step": 28531 + }, + { + "epoch": 2.4, + "learning_rate": 2.010685805892597e-06, + "loss": 1.1235, + "step": 28532 + }, + { + "epoch": 2.4, + "learning_rate": 2.0101391623844378e-06, + "loss": 1.1155, + "step": 28533 + }, + { + "epoch": 2.4, + "learning_rate": 2.0095925848910693e-06, + "loss": 1.1876, + "step": 28534 + }, + { + "epoch": 2.4, + "learning_rate": 2.0090460734170057e-06, + "loss": 1.0911, + "step": 28535 + }, + { + "epoch": 2.4, + "learning_rate": 2.0084996279667568e-06, + "loss": 1.1269, + "step": 28536 + }, + { + "epoch": 2.4, + "learning_rate": 2.0079532485448462e-06, + "loss": 1.0912, + "step": 28537 + }, + { + "epoch": 2.4, + "learning_rate": 2.0074069351557803e-06, + "loss": 1.094, + "step": 28538 + }, + { + "epoch": 2.4, + "learning_rate": 2.0068606878040774e-06, + "loss": 1.1208, + "step": 28539 + }, + { + "epoch": 2.4, + "learning_rate": 2.006314506494249e-06, + "loss": 1.0364, + "step": 28540 + }, + { + "epoch": 2.4, + "learning_rate": 2.005768391230811e-06, + "loss": 1.1279, + "step": 28541 + }, + { + "epoch": 2.4, + "learning_rate": 2.005222342018268e-06, + "loss": 1.046, + "step": 28542 + }, + { + "epoch": 2.4, + "learning_rate": 2.0046763588611416e-06, + "loss": 1.1509, + "step": 28543 + }, + { + "epoch": 2.4, + "learning_rate": 2.0041304417639363e-06, + "loss": 1.0893, + "step": 28544 + }, + { + "epoch": 2.4, + "learning_rate": 2.0035845907311635e-06, + "loss": 1.1075, + "step": 28545 + }, + { + "epoch": 2.4, + "learning_rate": 2.0030388057673345e-06, + "loss": 1.0275, + "step": 28546 + }, + { + "epoch": 2.4, + "learning_rate": 2.0024930868769578e-06, + "loss": 1.0925, + "step": 28547 + }, + { + "epoch": 2.4, + "learning_rate": 2.0019474340645415e-06, + "loss": 1.0778, + "step": 28548 + }, + { + "epoch": 2.4, + "learning_rate": 2.0014018473345975e-06, + "loss": 1.0875, + "step": 28549 + }, + { + "epoch": 2.4, + "learning_rate": 2.0008563266916303e-06, + "loss": 1.237, + "step": 28550 + }, + { + "epoch": 2.4, + "learning_rate": 2.0003108721401477e-06, + "loss": 1.0931, + "step": 28551 + }, + { + "epoch": 2.4, + "learning_rate": 1.9997654836846603e-06, + "loss": 1.1258, + "step": 28552 + }, + { + "epoch": 2.4, + "learning_rate": 1.9992201613296657e-06, + "loss": 1.1634, + "step": 28553 + }, + { + "epoch": 2.4, + "learning_rate": 1.9986749050796795e-06, + "loss": 1.1248, + "step": 28554 + }, + { + "epoch": 2.4, + "learning_rate": 1.9981297149391986e-06, + "loss": 1.0675, + "step": 28555 + }, + { + "epoch": 2.4, + "learning_rate": 1.997584590912732e-06, + "loss": 1.147, + "step": 28556 + }, + { + "epoch": 2.4, + "learning_rate": 1.997039533004783e-06, + "loss": 1.1218, + "step": 28557 + }, + { + "epoch": 2.4, + "learning_rate": 1.9964945412198545e-06, + "loss": 1.1059, + "step": 28558 + }, + { + "epoch": 2.4, + "learning_rate": 1.995949615562448e-06, + "loss": 1.1794, + "step": 28559 + }, + { + "epoch": 2.4, + "learning_rate": 1.9954047560370694e-06, + "loss": 1.1248, + "step": 28560 + }, + { + "epoch": 2.4, + "learning_rate": 1.994859962648217e-06, + "loss": 1.1921, + "step": 28561 + }, + { + "epoch": 2.4, + "learning_rate": 1.9943152354003946e-06, + "loss": 1.1298, + "step": 28562 + }, + { + "epoch": 2.4, + "learning_rate": 1.993770574298104e-06, + "loss": 1.1101, + "step": 28563 + }, + { + "epoch": 2.4, + "learning_rate": 1.9932259793458373e-06, + "loss": 1.0985, + "step": 28564 + }, + { + "epoch": 2.4, + "learning_rate": 1.9926814505481064e-06, + "loss": 1.1619, + "step": 28565 + }, + { + "epoch": 2.41, + "learning_rate": 1.9921369879094012e-06, + "loss": 1.1797, + "step": 28566 + }, + { + "epoch": 2.41, + "learning_rate": 1.9915925914342226e-06, + "loss": 1.0721, + "step": 28567 + }, + { + "epoch": 2.41, + "learning_rate": 1.99104826112707e-06, + "loss": 1.0677, + "step": 28568 + }, + { + "epoch": 2.41, + "learning_rate": 1.9905039969924388e-06, + "loss": 1.1248, + "step": 28569 + }, + { + "epoch": 2.41, + "learning_rate": 1.9899597990348274e-06, + "loss": 1.1262, + "step": 28570 + }, + { + "epoch": 2.41, + "learning_rate": 1.989415667258734e-06, + "loss": 1.2336, + "step": 28571 + }, + { + "epoch": 2.41, + "learning_rate": 1.9888716016686473e-06, + "loss": 1.0815, + "step": 28572 + }, + { + "epoch": 2.41, + "learning_rate": 1.9883276022690702e-06, + "loss": 1.0892, + "step": 28573 + }, + { + "epoch": 2.41, + "learning_rate": 1.987783669064496e-06, + "loss": 1.1839, + "step": 28574 + }, + { + "epoch": 2.41, + "learning_rate": 1.987239802059413e-06, + "loss": 1.0811, + "step": 28575 + }, + { + "epoch": 2.41, + "learning_rate": 1.986696001258325e-06, + "loss": 1.0992, + "step": 28576 + }, + { + "epoch": 2.41, + "learning_rate": 1.986152266665715e-06, + "loss": 1.1383, + "step": 28577 + }, + { + "epoch": 2.41, + "learning_rate": 1.9856085982860805e-06, + "loss": 1.0568, + "step": 28578 + }, + { + "epoch": 2.41, + "learning_rate": 1.9850649961239132e-06, + "loss": 1.1875, + "step": 28579 + }, + { + "epoch": 2.41, + "learning_rate": 1.9845214601837025e-06, + "loss": 1.0903, + "step": 28580 + }, + { + "epoch": 2.41, + "learning_rate": 1.983977990469943e-06, + "loss": 1.1292, + "step": 28581 + }, + { + "epoch": 2.41, + "learning_rate": 1.983434586987124e-06, + "loss": 1.0356, + "step": 28582 + }, + { + "epoch": 2.41, + "learning_rate": 1.982891249739729e-06, + "loss": 1.0576, + "step": 28583 + }, + { + "epoch": 2.41, + "learning_rate": 1.9823479787322543e-06, + "loss": 1.0812, + "step": 28584 + }, + { + "epoch": 2.41, + "learning_rate": 1.9818047739691904e-06, + "loss": 1.0982, + "step": 28585 + }, + { + "epoch": 2.41, + "learning_rate": 1.9812616354550153e-06, + "loss": 1.0966, + "step": 28586 + }, + { + "epoch": 2.41, + "learning_rate": 1.9807185631942294e-06, + "loss": 1.1475, + "step": 28587 + }, + { + "epoch": 2.41, + "learning_rate": 1.9801755571913094e-06, + "loss": 1.1559, + "step": 28588 + }, + { + "epoch": 2.41, + "learning_rate": 1.979632617450746e-06, + "loss": 1.0551, + "step": 28589 + }, + { + "epoch": 2.41, + "learning_rate": 1.979089743977025e-06, + "loss": 1.0069, + "step": 28590 + }, + { + "epoch": 2.41, + "learning_rate": 1.9785469367746313e-06, + "loss": 1.0869, + "step": 28591 + }, + { + "epoch": 2.41, + "learning_rate": 1.97800419584805e-06, + "loss": 1.2048, + "step": 28592 + }, + { + "epoch": 2.41, + "learning_rate": 1.977461521201769e-06, + "loss": 1.1288, + "step": 28593 + }, + { + "epoch": 2.41, + "learning_rate": 1.9769189128402623e-06, + "loss": 1.0825, + "step": 28594 + }, + { + "epoch": 2.41, + "learning_rate": 1.9763763707680237e-06, + "loss": 1.174, + "step": 28595 + }, + { + "epoch": 2.41, + "learning_rate": 1.9758338949895294e-06, + "loss": 1.0917, + "step": 28596 + }, + { + "epoch": 2.41, + "learning_rate": 1.975291485509262e-06, + "loss": 1.1096, + "step": 28597 + }, + { + "epoch": 2.41, + "learning_rate": 1.9747491423317088e-06, + "loss": 1.0721, + "step": 28598 + }, + { + "epoch": 2.41, + "learning_rate": 1.9742068654613444e-06, + "loss": 1.1204, + "step": 28599 + }, + { + "epoch": 2.41, + "learning_rate": 1.9736646549026506e-06, + "loss": 1.0707, + "step": 28600 + }, + { + "epoch": 2.41, + "learning_rate": 1.9731225106601092e-06, + "loss": 1.1133, + "step": 28601 + }, + { + "epoch": 2.41, + "learning_rate": 1.9725804327382004e-06, + "loss": 1.1152, + "step": 28602 + }, + { + "epoch": 2.41, + "learning_rate": 1.9720384211413968e-06, + "loss": 1.0329, + "step": 28603 + }, + { + "epoch": 2.41, + "learning_rate": 1.971496475874185e-06, + "loss": 1.175, + "step": 28604 + }, + { + "epoch": 2.41, + "learning_rate": 1.970954596941037e-06, + "loss": 1.1012, + "step": 28605 + }, + { + "epoch": 2.41, + "learning_rate": 1.9704127843464314e-06, + "loss": 1.0516, + "step": 28606 + }, + { + "epoch": 2.41, + "learning_rate": 1.969871038094846e-06, + "loss": 1.0936, + "step": 28607 + }, + { + "epoch": 2.41, + "learning_rate": 1.969329358190756e-06, + "loss": 1.0306, + "step": 28608 + }, + { + "epoch": 2.41, + "learning_rate": 1.9687877446386363e-06, + "loss": 1.1501, + "step": 28609 + }, + { + "epoch": 2.41, + "learning_rate": 1.9682461974429625e-06, + "loss": 1.1659, + "step": 28610 + }, + { + "epoch": 2.41, + "learning_rate": 1.96770471660821e-06, + "loss": 1.1321, + "step": 28611 + }, + { + "epoch": 2.41, + "learning_rate": 1.9671633021388504e-06, + "loss": 1.2084, + "step": 28612 + }, + { + "epoch": 2.41, + "learning_rate": 1.966621954039362e-06, + "loss": 1.0753, + "step": 28613 + }, + { + "epoch": 2.41, + "learning_rate": 1.966080672314208e-06, + "loss": 1.0501, + "step": 28614 + }, + { + "epoch": 2.41, + "learning_rate": 1.9655394569678733e-06, + "loss": 1.0346, + "step": 28615 + }, + { + "epoch": 2.41, + "learning_rate": 1.9649983080048184e-06, + "loss": 1.1301, + "step": 28616 + }, + { + "epoch": 2.41, + "learning_rate": 1.9644572254295204e-06, + "loss": 1.2182, + "step": 28617 + }, + { + "epoch": 2.41, + "learning_rate": 1.9639162092464482e-06, + "loss": 1.1539, + "step": 28618 + }, + { + "epoch": 2.41, + "learning_rate": 1.963375259460073e-06, + "loss": 1.0735, + "step": 28619 + }, + { + "epoch": 2.41, + "learning_rate": 1.9628343760748623e-06, + "loss": 1.0811, + "step": 28620 + }, + { + "epoch": 2.41, + "learning_rate": 1.96229355909529e-06, + "loss": 1.0896, + "step": 28621 + }, + { + "epoch": 2.41, + "learning_rate": 1.9617528085258143e-06, + "loss": 1.1368, + "step": 28622 + }, + { + "epoch": 2.41, + "learning_rate": 1.9612121243709114e-06, + "loss": 1.1364, + "step": 28623 + }, + { + "epoch": 2.41, + "learning_rate": 1.9606715066350514e-06, + "loss": 1.1799, + "step": 28624 + }, + { + "epoch": 2.41, + "learning_rate": 1.9601309553226897e-06, + "loss": 1.0793, + "step": 28625 + }, + { + "epoch": 2.41, + "learning_rate": 1.9595904704383038e-06, + "loss": 1.0734, + "step": 28626 + }, + { + "epoch": 2.41, + "learning_rate": 1.959050051986353e-06, + "loss": 1.1607, + "step": 28627 + }, + { + "epoch": 2.41, + "learning_rate": 1.958509699971305e-06, + "loss": 1.1618, + "step": 28628 + }, + { + "epoch": 2.41, + "learning_rate": 1.9579694143976226e-06, + "loss": 1.1416, + "step": 28629 + }, + { + "epoch": 2.41, + "learning_rate": 1.95742919526977e-06, + "loss": 1.0403, + "step": 28630 + }, + { + "epoch": 2.41, + "learning_rate": 1.9568890425922117e-06, + "loss": 1.046, + "step": 28631 + }, + { + "epoch": 2.41, + "learning_rate": 1.956348956369414e-06, + "loss": 1.0899, + "step": 28632 + }, + { + "epoch": 2.41, + "learning_rate": 1.9558089366058287e-06, + "loss": 1.1114, + "step": 28633 + }, + { + "epoch": 2.41, + "learning_rate": 1.955268983305928e-06, + "loss": 1.2157, + "step": 28634 + }, + { + "epoch": 2.41, + "learning_rate": 1.9547290964741718e-06, + "loss": 1.1398, + "step": 28635 + }, + { + "epoch": 2.41, + "learning_rate": 1.954189276115014e-06, + "loss": 1.151, + "step": 28636 + }, + { + "epoch": 2.41, + "learning_rate": 1.9536495222329244e-06, + "loss": 1.0565, + "step": 28637 + }, + { + "epoch": 2.41, + "learning_rate": 1.953109834832354e-06, + "loss": 1.1169, + "step": 28638 + }, + { + "epoch": 2.41, + "learning_rate": 1.952570213917766e-06, + "loss": 1.0381, + "step": 28639 + }, + { + "epoch": 2.41, + "learning_rate": 1.952030659493619e-06, + "loss": 1.1761, + "step": 28640 + }, + { + "epoch": 2.41, + "learning_rate": 1.95149117156437e-06, + "loss": 1.167, + "step": 28641 + }, + { + "epoch": 2.41, + "learning_rate": 1.9509517501344768e-06, + "loss": 1.1155, + "step": 28642 + }, + { + "epoch": 2.41, + "learning_rate": 1.9504123952084e-06, + "loss": 1.1796, + "step": 28643 + }, + { + "epoch": 2.41, + "learning_rate": 1.949873106790585e-06, + "loss": 1.1571, + "step": 28644 + }, + { + "epoch": 2.41, + "learning_rate": 1.9493338848855016e-06, + "loss": 1.1212, + "step": 28645 + }, + { + "epoch": 2.41, + "learning_rate": 1.948794729497594e-06, + "loss": 1.0387, + "step": 28646 + }, + { + "epoch": 2.41, + "learning_rate": 1.948255640631319e-06, + "loss": 1.1421, + "step": 28647 + }, + { + "epoch": 2.41, + "learning_rate": 1.9477166182911387e-06, + "loss": 1.0775, + "step": 28648 + }, + { + "epoch": 2.41, + "learning_rate": 1.9471776624814974e-06, + "loss": 1.0668, + "step": 28649 + }, + { + "epoch": 2.41, + "learning_rate": 1.9466387732068527e-06, + "loss": 1.126, + "step": 28650 + }, + { + "epoch": 2.41, + "learning_rate": 1.946099950471655e-06, + "loss": 1.1798, + "step": 28651 + }, + { + "epoch": 2.41, + "learning_rate": 1.9455611942803587e-06, + "loss": 1.0095, + "step": 28652 + }, + { + "epoch": 2.41, + "learning_rate": 1.9450225046374096e-06, + "loss": 1.1477, + "step": 28653 + }, + { + "epoch": 2.41, + "learning_rate": 1.9444838815472676e-06, + "loss": 1.0954, + "step": 28654 + }, + { + "epoch": 2.41, + "learning_rate": 1.9439453250143747e-06, + "loss": 1.14, + "step": 28655 + }, + { + "epoch": 2.41, + "learning_rate": 1.9434068350431845e-06, + "loss": 1.1369, + "step": 28656 + }, + { + "epoch": 2.41, + "learning_rate": 1.9428684116381457e-06, + "loss": 1.1065, + "step": 28657 + }, + { + "epoch": 2.41, + "learning_rate": 1.942330054803706e-06, + "loss": 1.1275, + "step": 28658 + }, + { + "epoch": 2.41, + "learning_rate": 1.941791764544314e-06, + "loss": 1.1378, + "step": 28659 + }, + { + "epoch": 2.41, + "learning_rate": 1.9412535408644174e-06, + "loss": 1.1057, + "step": 28660 + }, + { + "epoch": 2.41, + "learning_rate": 1.9407153837684633e-06, + "loss": 1.0717, + "step": 28661 + }, + { + "epoch": 2.41, + "learning_rate": 1.9401772932608986e-06, + "loss": 1.1233, + "step": 28662 + }, + { + "epoch": 2.41, + "learning_rate": 1.9396392693461706e-06, + "loss": 1.0855, + "step": 28663 + }, + { + "epoch": 2.41, + "learning_rate": 1.939101312028718e-06, + "loss": 1.1343, + "step": 28664 + }, + { + "epoch": 2.41, + "learning_rate": 1.9385634213129945e-06, + "loss": 1.0848, + "step": 28665 + }, + { + "epoch": 2.41, + "learning_rate": 1.9380255972034377e-06, + "loss": 1.2261, + "step": 28666 + }, + { + "epoch": 2.41, + "learning_rate": 1.937487839704494e-06, + "loss": 1.128, + "step": 28667 + }, + { + "epoch": 2.41, + "learning_rate": 1.9369501488206066e-06, + "loss": 1.1577, + "step": 28668 + }, + { + "epoch": 2.41, + "learning_rate": 1.936412524556216e-06, + "loss": 1.0846, + "step": 28669 + }, + { + "epoch": 2.41, + "learning_rate": 1.935874966915767e-06, + "loss": 1.1299, + "step": 28670 + }, + { + "epoch": 2.41, + "learning_rate": 1.9353374759037024e-06, + "loss": 1.0605, + "step": 28671 + }, + { + "epoch": 2.41, + "learning_rate": 1.934800051524455e-06, + "loss": 1.1029, + "step": 28672 + }, + { + "epoch": 2.41, + "learning_rate": 1.9342626937824737e-06, + "loss": 1.1304, + "step": 28673 + }, + { + "epoch": 2.41, + "learning_rate": 1.933725402682197e-06, + "loss": 1.134, + "step": 28674 + }, + { + "epoch": 2.41, + "learning_rate": 1.933188178228058e-06, + "loss": 1.1684, + "step": 28675 + }, + { + "epoch": 2.41, + "learning_rate": 1.9326510204245053e-06, + "loss": 1.1213, + "step": 28676 + }, + { + "epoch": 2.41, + "learning_rate": 1.9321139292759694e-06, + "loss": 1.0677, + "step": 28677 + }, + { + "epoch": 2.41, + "learning_rate": 1.9315769047868906e-06, + "loss": 1.0812, + "step": 28678 + }, + { + "epoch": 2.41, + "learning_rate": 1.9310399469617048e-06, + "loss": 1.1392, + "step": 28679 + }, + { + "epoch": 2.41, + "learning_rate": 1.93050305580485e-06, + "loss": 1.12, + "step": 28680 + }, + { + "epoch": 2.41, + "learning_rate": 1.92996623132076e-06, + "loss": 1.0263, + "step": 28681 + }, + { + "epoch": 2.41, + "learning_rate": 1.9294294735138764e-06, + "loss": 1.1978, + "step": 28682 + }, + { + "epoch": 2.41, + "learning_rate": 1.9288927823886227e-06, + "loss": 1.0293, + "step": 28683 + }, + { + "epoch": 2.41, + "learning_rate": 1.9283561579494436e-06, + "loss": 1.1067, + "step": 28684 + }, + { + "epoch": 2.42, + "learning_rate": 1.927819600200771e-06, + "loss": 1.005, + "step": 28685 + }, + { + "epoch": 2.42, + "learning_rate": 1.927283109147031e-06, + "loss": 1.0476, + "step": 28686 + }, + { + "epoch": 2.42, + "learning_rate": 1.9267466847926675e-06, + "loss": 1.1332, + "step": 28687 + }, + { + "epoch": 2.42, + "learning_rate": 1.926210327142104e-06, + "loss": 1.0861, + "step": 28688 + }, + { + "epoch": 2.42, + "learning_rate": 1.925674036199774e-06, + "loss": 1.0699, + "step": 28689 + }, + { + "epoch": 2.42, + "learning_rate": 1.925137811970109e-06, + "loss": 1.1239, + "step": 28690 + }, + { + "epoch": 2.42, + "learning_rate": 1.924601654457541e-06, + "loss": 1.1195, + "step": 28691 + }, + { + "epoch": 2.42, + "learning_rate": 1.9240655636664972e-06, + "loss": 1.1233, + "step": 28692 + }, + { + "epoch": 2.42, + "learning_rate": 1.923529539601412e-06, + "loss": 1.0633, + "step": 28693 + }, + { + "epoch": 2.42, + "learning_rate": 1.9229935822667055e-06, + "loss": 1.2298, + "step": 28694 + }, + { + "epoch": 2.42, + "learning_rate": 1.922457691666816e-06, + "loss": 1.0959, + "step": 28695 + }, + { + "epoch": 2.42, + "learning_rate": 1.9219218678061623e-06, + "loss": 1.1733, + "step": 28696 + }, + { + "epoch": 2.42, + "learning_rate": 1.921386110689174e-06, + "loss": 1.1051, + "step": 28697 + }, + { + "epoch": 2.42, + "learning_rate": 1.920850420320285e-06, + "loss": 1.0531, + "step": 28698 + }, + { + "epoch": 2.42, + "learning_rate": 1.9203147967039114e-06, + "loss": 1.2093, + "step": 28699 + }, + { + "epoch": 2.42, + "learning_rate": 1.919779239844484e-06, + "loss": 1.0908, + "step": 28700 + }, + { + "epoch": 2.42, + "learning_rate": 1.9192437497464257e-06, + "loss": 1.1, + "step": 28701 + }, + { + "epoch": 2.42, + "learning_rate": 1.9187083264141647e-06, + "loss": 1.0971, + "step": 28702 + }, + { + "epoch": 2.42, + "learning_rate": 1.9181729698521158e-06, + "loss": 1.1809, + "step": 28703 + }, + { + "epoch": 2.42, + "learning_rate": 1.9176376800647145e-06, + "loss": 1.1646, + "step": 28704 + }, + { + "epoch": 2.42, + "learning_rate": 1.9171024570563744e-06, + "loss": 1.2499, + "step": 28705 + }, + { + "epoch": 2.42, + "learning_rate": 1.916567300831521e-06, + "loss": 1.0999, + "step": 28706 + }, + { + "epoch": 2.42, + "learning_rate": 1.916032211394575e-06, + "loss": 1.1924, + "step": 28707 + }, + { + "epoch": 2.42, + "learning_rate": 1.9154971887499584e-06, + "loss": 1.0785, + "step": 28708 + }, + { + "epoch": 2.42, + "learning_rate": 1.9149622329020923e-06, + "loss": 1.1331, + "step": 28709 + }, + { + "epoch": 2.42, + "learning_rate": 1.914427343855395e-06, + "loss": 1.1287, + "step": 28710 + }, + { + "epoch": 2.42, + "learning_rate": 1.913892521614287e-06, + "loss": 1.0895, + "step": 28711 + }, + { + "epoch": 2.42, + "learning_rate": 1.9133577661831858e-06, + "loss": 1.2629, + "step": 28712 + }, + { + "epoch": 2.42, + "learning_rate": 1.9128230775665137e-06, + "loss": 1.0455, + "step": 28713 + }, + { + "epoch": 2.42, + "learning_rate": 1.9122884557686816e-06, + "loss": 1.131, + "step": 28714 + }, + { + "epoch": 2.42, + "learning_rate": 1.9117539007941143e-06, + "loss": 1.1144, + "step": 28715 + }, + { + "epoch": 2.42, + "learning_rate": 1.9112194126472227e-06, + "loss": 1.091, + "step": 28716 + }, + { + "epoch": 2.42, + "learning_rate": 1.9106849913324253e-06, + "loss": 1.1687, + "step": 28717 + }, + { + "epoch": 2.42, + "learning_rate": 1.910150636854138e-06, + "loss": 1.1185, + "step": 28718 + }, + { + "epoch": 2.42, + "learning_rate": 1.909616349216775e-06, + "loss": 1.1901, + "step": 28719 + }, + { + "epoch": 2.42, + "learning_rate": 1.9090821284247506e-06, + "loss": 1.0364, + "step": 28720 + }, + { + "epoch": 2.42, + "learning_rate": 1.9085479744824807e-06, + "loss": 1.114, + "step": 28721 + }, + { + "epoch": 2.42, + "learning_rate": 1.9080138873943733e-06, + "loss": 1.1952, + "step": 28722 + }, + { + "epoch": 2.42, + "learning_rate": 1.9074798671648477e-06, + "loss": 1.1273, + "step": 28723 + }, + { + "epoch": 2.42, + "learning_rate": 1.906945913798315e-06, + "loss": 1.1122, + "step": 28724 + }, + { + "epoch": 2.42, + "learning_rate": 1.9064120272991805e-06, + "loss": 1.1267, + "step": 28725 + }, + { + "epoch": 2.42, + "learning_rate": 1.9058782076718651e-06, + "loss": 1.0542, + "step": 28726 + }, + { + "epoch": 2.42, + "learning_rate": 1.9053444549207724e-06, + "loss": 1.1239, + "step": 28727 + }, + { + "epoch": 2.42, + "learning_rate": 1.9048107690503137e-06, + "loss": 1.1581, + "step": 28728 + }, + { + "epoch": 2.42, + "learning_rate": 1.9042771500648993e-06, + "loss": 1.0734, + "step": 28729 + }, + { + "epoch": 2.42, + "learning_rate": 1.9037435979689379e-06, + "loss": 1.2303, + "step": 28730 + }, + { + "epoch": 2.42, + "learning_rate": 1.903210112766839e-06, + "loss": 1.0785, + "step": 28731 + }, + { + "epoch": 2.42, + "learning_rate": 1.9026766944630115e-06, + "loss": 1.1147, + "step": 28732 + }, + { + "epoch": 2.42, + "learning_rate": 1.9021433430618552e-06, + "loss": 1.0675, + "step": 28733 + }, + { + "epoch": 2.42, + "learning_rate": 1.9016100585677844e-06, + "loss": 1.1028, + "step": 28734 + }, + { + "epoch": 2.42, + "learning_rate": 1.9010768409852064e-06, + "loss": 1.1174, + "step": 28735 + }, + { + "epoch": 2.42, + "learning_rate": 1.9005436903185182e-06, + "loss": 1.1196, + "step": 28736 + }, + { + "epoch": 2.42, + "learning_rate": 1.9000106065721358e-06, + "loss": 1.1631, + "step": 28737 + }, + { + "epoch": 2.42, + "learning_rate": 1.8994775897504548e-06, + "loss": 1.1284, + "step": 28738 + }, + { + "epoch": 2.42, + "learning_rate": 1.8989446398578826e-06, + "loss": 1.1887, + "step": 28739 + }, + { + "epoch": 2.42, + "learning_rate": 1.8984117568988237e-06, + "loss": 1.0788, + "step": 28740 + }, + { + "epoch": 2.42, + "learning_rate": 1.8978789408776777e-06, + "loss": 1.1277, + "step": 28741 + }, + { + "epoch": 2.42, + "learning_rate": 1.89734619179885e-06, + "loss": 1.0574, + "step": 28742 + }, + { + "epoch": 2.42, + "learning_rate": 1.8968135096667439e-06, + "loss": 1.145, + "step": 28743 + }, + { + "epoch": 2.42, + "learning_rate": 1.8962808944857525e-06, + "loss": 1.0516, + "step": 28744 + }, + { + "epoch": 2.42, + "learning_rate": 1.8957483462602866e-06, + "loss": 1.0302, + "step": 28745 + }, + { + "epoch": 2.42, + "learning_rate": 1.8952158649947394e-06, + "loss": 1.0914, + "step": 28746 + }, + { + "epoch": 2.42, + "learning_rate": 1.8946834506935107e-06, + "loss": 1.0707, + "step": 28747 + }, + { + "epoch": 2.42, + "learning_rate": 1.894151103361006e-06, + "loss": 1.1766, + "step": 28748 + }, + { + "epoch": 2.42, + "learning_rate": 1.893618823001616e-06, + "loss": 1.0691, + "step": 28749 + }, + { + "epoch": 2.42, + "learning_rate": 1.893086609619742e-06, + "loss": 1.0715, + "step": 28750 + }, + { + "epoch": 2.42, + "learning_rate": 1.892554463219781e-06, + "loss": 1.1421, + "step": 28751 + }, + { + "epoch": 2.42, + "learning_rate": 1.8920223838061325e-06, + "loss": 1.1133, + "step": 28752 + }, + { + "epoch": 2.42, + "learning_rate": 1.8914903713831845e-06, + "loss": 1.033, + "step": 28753 + }, + { + "epoch": 2.42, + "learning_rate": 1.8909584259553437e-06, + "loss": 1.13, + "step": 28754 + }, + { + "epoch": 2.42, + "learning_rate": 1.8904265475269968e-06, + "loss": 1.175, + "step": 28755 + }, + { + "epoch": 2.42, + "learning_rate": 1.889894736102541e-06, + "loss": 1.1003, + "step": 28756 + }, + { + "epoch": 2.42, + "learning_rate": 1.8893629916863709e-06, + "loss": 1.1915, + "step": 28757 + }, + { + "epoch": 2.42, + "learning_rate": 1.8888313142828784e-06, + "loss": 1.146, + "step": 28758 + }, + { + "epoch": 2.42, + "learning_rate": 1.888299703896459e-06, + "loss": 1.166, + "step": 28759 + }, + { + "epoch": 2.42, + "learning_rate": 1.8877681605315024e-06, + "loss": 1.126, + "step": 28760 + }, + { + "epoch": 2.42, + "learning_rate": 1.8872366841924016e-06, + "loss": 1.1135, + "step": 28761 + }, + { + "epoch": 2.42, + "learning_rate": 1.8867052748835478e-06, + "loss": 1.131, + "step": 28762 + }, + { + "epoch": 2.42, + "learning_rate": 1.8861739326093343e-06, + "loss": 1.1838, + "step": 28763 + }, + { + "epoch": 2.42, + "learning_rate": 1.8856426573741438e-06, + "loss": 1.1808, + "step": 28764 + }, + { + "epoch": 2.42, + "learning_rate": 1.8851114491823752e-06, + "loss": 1.1114, + "step": 28765 + }, + { + "epoch": 2.42, + "learning_rate": 1.8845803080384107e-06, + "loss": 1.1126, + "step": 28766 + }, + { + "epoch": 2.42, + "learning_rate": 1.884049233946641e-06, + "loss": 1.1058, + "step": 28767 + }, + { + "epoch": 2.42, + "learning_rate": 1.8835182269114537e-06, + "loss": 1.0658, + "step": 28768 + }, + { + "epoch": 2.42, + "learning_rate": 1.8829872869372367e-06, + "loss": 1.1509, + "step": 28769 + }, + { + "epoch": 2.42, + "learning_rate": 1.8824564140283763e-06, + "loss": 1.1268, + "step": 28770 + }, + { + "epoch": 2.42, + "learning_rate": 1.8819256081892611e-06, + "loss": 1.0772, + "step": 28771 + }, + { + "epoch": 2.42, + "learning_rate": 1.8813948694242701e-06, + "loss": 1.0813, + "step": 28772 + }, + { + "epoch": 2.42, + "learning_rate": 1.880864197737795e-06, + "loss": 1.0939, + "step": 28773 + }, + { + "epoch": 2.42, + "learning_rate": 1.8803335931342215e-06, + "loss": 1.0655, + "step": 28774 + }, + { + "epoch": 2.42, + "learning_rate": 1.8798030556179247e-06, + "loss": 1.1117, + "step": 28775 + }, + { + "epoch": 2.42, + "learning_rate": 1.8792725851932992e-06, + "loss": 1.0969, + "step": 28776 + }, + { + "epoch": 2.42, + "learning_rate": 1.878742181864721e-06, + "loss": 1.1224, + "step": 28777 + }, + { + "epoch": 2.42, + "learning_rate": 1.8782118456365727e-06, + "loss": 1.0958, + "step": 28778 + }, + { + "epoch": 2.42, + "learning_rate": 1.8776815765132383e-06, + "loss": 1.0718, + "step": 28779 + }, + { + "epoch": 2.42, + "learning_rate": 1.8771513744990977e-06, + "loss": 1.0659, + "step": 28780 + }, + { + "epoch": 2.42, + "learning_rate": 1.8766212395985317e-06, + "loss": 1.0783, + "step": 28781 + }, + { + "epoch": 2.42, + "learning_rate": 1.8760911718159248e-06, + "loss": 1.0612, + "step": 28782 + }, + { + "epoch": 2.42, + "learning_rate": 1.8755611711556465e-06, + "loss": 1.1357, + "step": 28783 + }, + { + "epoch": 2.42, + "learning_rate": 1.8750312376220847e-06, + "loss": 1.0235, + "step": 28784 + }, + { + "epoch": 2.42, + "learning_rate": 1.874501371219617e-06, + "loss": 1.0746, + "step": 28785 + }, + { + "epoch": 2.42, + "learning_rate": 1.8739715719526153e-06, + "loss": 1.1068, + "step": 28786 + }, + { + "epoch": 2.42, + "learning_rate": 1.8734418398254662e-06, + "loss": 1.0235, + "step": 28787 + }, + { + "epoch": 2.42, + "learning_rate": 1.8729121748425384e-06, + "loss": 1.0819, + "step": 28788 + }, + { + "epoch": 2.42, + "learning_rate": 1.8723825770082116e-06, + "loss": 1.2048, + "step": 28789 + }, + { + "epoch": 2.42, + "learning_rate": 1.8718530463268613e-06, + "loss": 1.0848, + "step": 28790 + }, + { + "epoch": 2.42, + "learning_rate": 1.871323582802862e-06, + "loss": 1.0508, + "step": 28791 + }, + { + "epoch": 2.42, + "learning_rate": 1.8707941864405887e-06, + "loss": 1.1799, + "step": 28792 + }, + { + "epoch": 2.42, + "learning_rate": 1.8702648572444181e-06, + "loss": 1.0574, + "step": 28793 + }, + { + "epoch": 2.42, + "learning_rate": 1.8697355952187167e-06, + "loss": 1.046, + "step": 28794 + }, + { + "epoch": 2.42, + "learning_rate": 1.8692064003678655e-06, + "loss": 1.0368, + "step": 28795 + }, + { + "epoch": 2.42, + "learning_rate": 1.8686772726962322e-06, + "loss": 1.0947, + "step": 28796 + }, + { + "epoch": 2.42, + "learning_rate": 1.8681482122081862e-06, + "loss": 1.1609, + "step": 28797 + }, + { + "epoch": 2.42, + "learning_rate": 1.8676192189081078e-06, + "loss": 1.0643, + "step": 28798 + }, + { + "epoch": 2.42, + "learning_rate": 1.8670902928003587e-06, + "loss": 1.1098, + "step": 28799 + }, + { + "epoch": 2.42, + "learning_rate": 1.8665614338893135e-06, + "loss": 1.0643, + "step": 28800 + }, + { + "epoch": 2.42, + "learning_rate": 1.8660326421793406e-06, + "loss": 1.0961, + "step": 28801 + }, + { + "epoch": 2.42, + "learning_rate": 1.8655039176748113e-06, + "loss": 1.1198, + "step": 28802 + }, + { + "epoch": 2.42, + "learning_rate": 1.8649752603800874e-06, + "loss": 1.0843, + "step": 28803 + }, + { + "epoch": 2.43, + "learning_rate": 1.8644466702995456e-06, + "loss": 1.1305, + "step": 28804 + }, + { + "epoch": 2.43, + "learning_rate": 1.863918147437548e-06, + "loss": 1.1347, + "step": 28805 + }, + { + "epoch": 2.43, + "learning_rate": 1.8633896917984618e-06, + "loss": 1.1357, + "step": 28806 + }, + { + "epoch": 2.43, + "learning_rate": 1.862861303386654e-06, + "loss": 1.0988, + "step": 28807 + }, + { + "epoch": 2.43, + "learning_rate": 1.8623329822064894e-06, + "loss": 1.2238, + "step": 28808 + }, + { + "epoch": 2.43, + "learning_rate": 1.8618047282623353e-06, + "loss": 1.0349, + "step": 28809 + }, + { + "epoch": 2.43, + "learning_rate": 1.8612765415585543e-06, + "loss": 1.137, + "step": 28810 + }, + { + "epoch": 2.43, + "learning_rate": 1.8607484220995121e-06, + "loss": 1.1202, + "step": 28811 + }, + { + "epoch": 2.43, + "learning_rate": 1.8602203698895715e-06, + "loss": 1.1144, + "step": 28812 + }, + { + "epoch": 2.43, + "learning_rate": 1.8596923849330973e-06, + "loss": 1.065, + "step": 28813 + }, + { + "epoch": 2.43, + "learning_rate": 1.8591644672344444e-06, + "loss": 1.1441, + "step": 28814 + }, + { + "epoch": 2.43, + "learning_rate": 1.8586366167979852e-06, + "loss": 1.158, + "step": 28815 + }, + { + "epoch": 2.43, + "learning_rate": 1.8581088336280739e-06, + "loss": 1.1766, + "step": 28816 + }, + { + "epoch": 2.43, + "learning_rate": 1.8575811177290726e-06, + "loss": 1.1205, + "step": 28817 + }, + { + "epoch": 2.43, + "learning_rate": 1.8570534691053422e-06, + "loss": 1.0439, + "step": 28818 + }, + { + "epoch": 2.43, + "learning_rate": 1.8565258877612424e-06, + "loss": 1.1882, + "step": 28819 + }, + { + "epoch": 2.43, + "learning_rate": 1.8559983737011311e-06, + "loss": 1.1598, + "step": 28820 + }, + { + "epoch": 2.43, + "learning_rate": 1.8554709269293714e-06, + "loss": 1.1191, + "step": 28821 + }, + { + "epoch": 2.43, + "learning_rate": 1.8549435474503118e-06, + "loss": 1.0235, + "step": 28822 + }, + { + "epoch": 2.43, + "learning_rate": 1.8544162352683181e-06, + "loss": 1.1374, + "step": 28823 + }, + { + "epoch": 2.43, + "learning_rate": 1.8538889903877465e-06, + "loss": 1.0487, + "step": 28824 + }, + { + "epoch": 2.43, + "learning_rate": 1.853361812812946e-06, + "loss": 1.0675, + "step": 28825 + }, + { + "epoch": 2.43, + "learning_rate": 1.852834702548283e-06, + "loss": 1.1381, + "step": 28826 + }, + { + "epoch": 2.43, + "learning_rate": 1.852307659598105e-06, + "loss": 1.0872, + "step": 28827 + }, + { + "epoch": 2.43, + "learning_rate": 1.8517806839667673e-06, + "loss": 1.1273, + "step": 28828 + }, + { + "epoch": 2.43, + "learning_rate": 1.8512537756586269e-06, + "loss": 1.0835, + "step": 28829 + }, + { + "epoch": 2.43, + "learning_rate": 1.8507269346780344e-06, + "loss": 1.0503, + "step": 28830 + }, + { + "epoch": 2.43, + "learning_rate": 1.8502001610293452e-06, + "loss": 1.1161, + "step": 28831 + }, + { + "epoch": 2.43, + "learning_rate": 1.8496734547169127e-06, + "loss": 1.1566, + "step": 28832 + }, + { + "epoch": 2.43, + "learning_rate": 1.849146815745081e-06, + "loss": 1.0848, + "step": 28833 + }, + { + "epoch": 2.43, + "learning_rate": 1.84862024411821e-06, + "loss": 1.1167, + "step": 28834 + }, + { + "epoch": 2.43, + "learning_rate": 1.8480937398406495e-06, + "loss": 1.0809, + "step": 28835 + }, + { + "epoch": 2.43, + "learning_rate": 1.847567302916743e-06, + "loss": 1.0973, + "step": 28836 + }, + { + "epoch": 2.43, + "learning_rate": 1.8470409333508489e-06, + "loss": 1.0874, + "step": 28837 + }, + { + "epoch": 2.43, + "learning_rate": 1.8465146311473092e-06, + "loss": 1.143, + "step": 28838 + }, + { + "epoch": 2.43, + "learning_rate": 1.8459883963104764e-06, + "loss": 1.0676, + "step": 28839 + }, + { + "epoch": 2.43, + "learning_rate": 1.8454622288446955e-06, + "loss": 1.1319, + "step": 28840 + }, + { + "epoch": 2.43, + "learning_rate": 1.8449361287543166e-06, + "loss": 1.1387, + "step": 28841 + }, + { + "epoch": 2.43, + "learning_rate": 1.844410096043685e-06, + "loss": 1.0581, + "step": 28842 + }, + { + "epoch": 2.43, + "learning_rate": 1.8438841307171496e-06, + "loss": 1.0911, + "step": 28843 + }, + { + "epoch": 2.43, + "learning_rate": 1.843358232779049e-06, + "loss": 1.1476, + "step": 28844 + }, + { + "epoch": 2.43, + "learning_rate": 1.8428324022337374e-06, + "loss": 1.0717, + "step": 28845 + }, + { + "epoch": 2.43, + "learning_rate": 1.8423066390855537e-06, + "loss": 1.1399, + "step": 28846 + }, + { + "epoch": 2.43, + "learning_rate": 1.84178094333884e-06, + "loss": 1.0621, + "step": 28847 + }, + { + "epoch": 2.43, + "learning_rate": 1.8412553149979483e-06, + "loss": 1.0392, + "step": 28848 + }, + { + "epoch": 2.43, + "learning_rate": 1.8407297540672141e-06, + "loss": 1.1671, + "step": 28849 + }, + { + "epoch": 2.43, + "learning_rate": 1.8402042605509828e-06, + "loss": 1.2228, + "step": 28850 + }, + { + "epoch": 2.43, + "learning_rate": 1.8396788344535942e-06, + "loss": 1.094, + "step": 28851 + }, + { + "epoch": 2.43, + "learning_rate": 1.8391534757793938e-06, + "loss": 1.0868, + "step": 28852 + }, + { + "epoch": 2.43, + "learning_rate": 1.8386281845327137e-06, + "loss": 1.167, + "step": 28853 + }, + { + "epoch": 2.43, + "learning_rate": 1.8381029607179046e-06, + "loss": 1.048, + "step": 28854 + }, + { + "epoch": 2.43, + "learning_rate": 1.837577804339299e-06, + "loss": 1.1419, + "step": 28855 + }, + { + "epoch": 2.43, + "learning_rate": 1.8370527154012374e-06, + "loss": 1.0993, + "step": 28856 + }, + { + "epoch": 2.43, + "learning_rate": 1.836527693908059e-06, + "loss": 1.17, + "step": 28857 + }, + { + "epoch": 2.43, + "learning_rate": 1.8360027398641013e-06, + "loss": 1.1264, + "step": 28858 + }, + { + "epoch": 2.43, + "learning_rate": 1.8354778532737016e-06, + "loss": 1.2094, + "step": 28859 + }, + { + "epoch": 2.43, + "learning_rate": 1.834953034141197e-06, + "loss": 1.1311, + "step": 28860 + }, + { + "epoch": 2.43, + "learning_rate": 1.8344282824709237e-06, + "loss": 1.1092, + "step": 28861 + }, + { + "epoch": 2.43, + "learning_rate": 1.8339035982672171e-06, + "loss": 1.1021, + "step": 28862 + }, + { + "epoch": 2.43, + "learning_rate": 1.8333789815344149e-06, + "loss": 1.0987, + "step": 28863 + }, + { + "epoch": 2.43, + "learning_rate": 1.8328544322768438e-06, + "loss": 1.0328, + "step": 28864 + }, + { + "epoch": 2.43, + "learning_rate": 1.832329950498849e-06, + "loss": 1.0662, + "step": 28865 + }, + { + "epoch": 2.43, + "learning_rate": 1.831805536204755e-06, + "loss": 1.1154, + "step": 28866 + }, + { + "epoch": 2.43, + "learning_rate": 1.8312811893988979e-06, + "loss": 1.1335, + "step": 28867 + }, + { + "epoch": 2.43, + "learning_rate": 1.8307569100856103e-06, + "loss": 1.0809, + "step": 28868 + }, + { + "epoch": 2.43, + "learning_rate": 1.8302326982692242e-06, + "loss": 1.0932, + "step": 28869 + }, + { + "epoch": 2.43, + "learning_rate": 1.8297085539540694e-06, + "loss": 1.1219, + "step": 28870 + }, + { + "epoch": 2.43, + "learning_rate": 1.82918447714448e-06, + "loss": 1.1173, + "step": 28871 + }, + { + "epoch": 2.43, + "learning_rate": 1.8286604678447784e-06, + "loss": 1.1131, + "step": 28872 + }, + { + "epoch": 2.43, + "learning_rate": 1.8281365260593008e-06, + "loss": 1.0065, + "step": 28873 + }, + { + "epoch": 2.43, + "learning_rate": 1.8276126517923787e-06, + "loss": 1.0311, + "step": 28874 + }, + { + "epoch": 2.43, + "learning_rate": 1.8270888450483304e-06, + "loss": 1.0715, + "step": 28875 + }, + { + "epoch": 2.43, + "learning_rate": 1.8265651058314947e-06, + "loss": 1.0486, + "step": 28876 + }, + { + "epoch": 2.43, + "learning_rate": 1.8260414341461919e-06, + "loss": 1.1418, + "step": 28877 + }, + { + "epoch": 2.43, + "learning_rate": 1.8255178299967502e-06, + "loss": 1.1397, + "step": 28878 + }, + { + "epoch": 2.43, + "learning_rate": 1.8249942933874975e-06, + "loss": 1.0374, + "step": 28879 + }, + { + "epoch": 2.43, + "learning_rate": 1.8244708243227571e-06, + "loss": 1.1476, + "step": 28880 + }, + { + "epoch": 2.43, + "learning_rate": 1.8239474228068555e-06, + "loss": 1.1998, + "step": 28881 + }, + { + "epoch": 2.43, + "learning_rate": 1.8234240888441201e-06, + "loss": 1.0819, + "step": 28882 + }, + { + "epoch": 2.43, + "learning_rate": 1.8229008224388655e-06, + "loss": 1.1269, + "step": 28883 + }, + { + "epoch": 2.43, + "learning_rate": 1.822377623595425e-06, + "loss": 1.017, + "step": 28884 + }, + { + "epoch": 2.43, + "learning_rate": 1.821854492318119e-06, + "loss": 1.071, + "step": 28885 + }, + { + "epoch": 2.43, + "learning_rate": 1.8213314286112648e-06, + "loss": 1.2135, + "step": 28886 + }, + { + "epoch": 2.43, + "learning_rate": 1.8208084324791918e-06, + "loss": 1.0972, + "step": 28887 + }, + { + "epoch": 2.43, + "learning_rate": 1.8202855039262157e-06, + "loss": 1.1286, + "step": 28888 + }, + { + "epoch": 2.43, + "learning_rate": 1.8197626429566584e-06, + "loss": 1.1667, + "step": 28889 + }, + { + "epoch": 2.43, + "learning_rate": 1.8192398495748399e-06, + "loss": 1.152, + "step": 28890 + }, + { + "epoch": 2.43, + "learning_rate": 1.8187171237850797e-06, + "loss": 1.1797, + "step": 28891 + }, + { + "epoch": 2.43, + "learning_rate": 1.8181944655916971e-06, + "loss": 1.1842, + "step": 28892 + }, + { + "epoch": 2.43, + "learning_rate": 1.817671874999013e-06, + "loss": 1.1532, + "step": 28893 + }, + { + "epoch": 2.43, + "learning_rate": 1.8171493520113382e-06, + "loss": 1.2163, + "step": 28894 + }, + { + "epoch": 2.43, + "learning_rate": 1.8166268966329981e-06, + "loss": 1.1253, + "step": 28895 + }, + { + "epoch": 2.43, + "learning_rate": 1.8161045088683038e-06, + "loss": 1.1473, + "step": 28896 + }, + { + "epoch": 2.43, + "learning_rate": 1.8155821887215709e-06, + "loss": 1.0837, + "step": 28897 + }, + { + "epoch": 2.43, + "learning_rate": 1.815059936197121e-06, + "loss": 1.0997, + "step": 28898 + }, + { + "epoch": 2.43, + "learning_rate": 1.8145377512992646e-06, + "loss": 1.1569, + "step": 28899 + }, + { + "epoch": 2.43, + "learning_rate": 1.8140156340323157e-06, + "loss": 1.0978, + "step": 28900 + }, + { + "epoch": 2.43, + "learning_rate": 1.8134935844005908e-06, + "loss": 1.0063, + "step": 28901 + }, + { + "epoch": 2.43, + "learning_rate": 1.812971602408401e-06, + "loss": 1.1211, + "step": 28902 + }, + { + "epoch": 2.43, + "learning_rate": 1.8124496880600595e-06, + "loss": 1.0373, + "step": 28903 + }, + { + "epoch": 2.43, + "learning_rate": 1.811927841359883e-06, + "loss": 1.0868, + "step": 28904 + }, + { + "epoch": 2.43, + "learning_rate": 1.8114060623121733e-06, + "loss": 1.1586, + "step": 28905 + }, + { + "epoch": 2.43, + "learning_rate": 1.8108843509212526e-06, + "loss": 1.1691, + "step": 28906 + }, + { + "epoch": 2.43, + "learning_rate": 1.810362707191423e-06, + "loss": 1.1263, + "step": 28907 + }, + { + "epoch": 2.43, + "learning_rate": 1.8098411311269993e-06, + "loss": 1.1161, + "step": 28908 + }, + { + "epoch": 2.43, + "learning_rate": 1.8093196227322873e-06, + "loss": 1.1935, + "step": 28909 + }, + { + "epoch": 2.43, + "learning_rate": 1.8087981820115996e-06, + "loss": 1.1393, + "step": 28910 + }, + { + "epoch": 2.43, + "learning_rate": 1.8082768089692427e-06, + "loss": 1.1499, + "step": 28911 + }, + { + "epoch": 2.43, + "learning_rate": 1.807755503609524e-06, + "loss": 1.0938, + "step": 28912 + }, + { + "epoch": 2.43, + "learning_rate": 1.8072342659367536e-06, + "loss": 1.1484, + "step": 28913 + }, + { + "epoch": 2.43, + "learning_rate": 1.8067130959552304e-06, + "loss": 1.1169, + "step": 28914 + }, + { + "epoch": 2.43, + "learning_rate": 1.8061919936692717e-06, + "loss": 1.184, + "step": 28915 + }, + { + "epoch": 2.43, + "learning_rate": 1.8056709590831745e-06, + "loss": 1.2159, + "step": 28916 + }, + { + "epoch": 2.43, + "learning_rate": 1.805149992201246e-06, + "loss": 1.0338, + "step": 28917 + }, + { + "epoch": 2.43, + "learning_rate": 1.804629093027791e-06, + "loss": 1.1164, + "step": 28918 + }, + { + "epoch": 2.43, + "learning_rate": 1.8041082615671123e-06, + "loss": 1.1327, + "step": 28919 + }, + { + "epoch": 2.43, + "learning_rate": 1.8035874978235158e-06, + "loss": 1.1067, + "step": 28920 + }, + { + "epoch": 2.43, + "learning_rate": 1.8030668018013043e-06, + "loss": 1.1129, + "step": 28921 + }, + { + "epoch": 2.43, + "learning_rate": 1.8025461735047734e-06, + "loss": 1.1038, + "step": 28922 + }, + { + "epoch": 2.44, + "learning_rate": 1.802025612938232e-06, + "loss": 1.0416, + "step": 28923 + }, + { + "epoch": 2.44, + "learning_rate": 1.8015051201059807e-06, + "loss": 1.1088, + "step": 28924 + }, + { + "epoch": 2.44, + "learning_rate": 1.8009846950123144e-06, + "loss": 1.1663, + "step": 28925 + }, + { + "epoch": 2.44, + "learning_rate": 1.8004643376615404e-06, + "loss": 1.233, + "step": 28926 + }, + { + "epoch": 2.44, + "learning_rate": 1.7999440480579521e-06, + "loss": 1.1308, + "step": 28927 + }, + { + "epoch": 2.44, + "learning_rate": 1.7994238262058495e-06, + "loss": 1.0535, + "step": 28928 + }, + { + "epoch": 2.44, + "learning_rate": 1.7989036721095332e-06, + "loss": 1.0497, + "step": 28929 + }, + { + "epoch": 2.44, + "learning_rate": 1.798383585773298e-06, + "loss": 1.1752, + "step": 28930 + }, + { + "epoch": 2.44, + "learning_rate": 1.7978635672014433e-06, + "loss": 1.0607, + "step": 28931 + }, + { + "epoch": 2.44, + "learning_rate": 1.7973436163982672e-06, + "loss": 1.0869, + "step": 28932 + }, + { + "epoch": 2.44, + "learning_rate": 1.7968237333680582e-06, + "loss": 1.1038, + "step": 28933 + }, + { + "epoch": 2.44, + "learning_rate": 1.79630391811512e-06, + "loss": 1.0725, + "step": 28934 + }, + { + "epoch": 2.44, + "learning_rate": 1.7957841706437463e-06, + "loss": 1.1105, + "step": 28935 + }, + { + "epoch": 2.44, + "learning_rate": 1.795264490958225e-06, + "loss": 1.1141, + "step": 28936 + }, + { + "epoch": 2.44, + "learning_rate": 1.7947448790628585e-06, + "loss": 1.1283, + "step": 28937 + }, + { + "epoch": 2.44, + "learning_rate": 1.7942253349619344e-06, + "loss": 1.1076, + "step": 28938 + }, + { + "epoch": 2.44, + "learning_rate": 1.7937058586597456e-06, + "loss": 1.0856, + "step": 28939 + }, + { + "epoch": 2.44, + "learning_rate": 1.7931864501605856e-06, + "loss": 1.1248, + "step": 28940 + }, + { + "epoch": 2.44, + "learning_rate": 1.7926671094687464e-06, + "loss": 1.1405, + "step": 28941 + }, + { + "epoch": 2.44, + "learning_rate": 1.7921478365885181e-06, + "loss": 1.0581, + "step": 28942 + }, + { + "epoch": 2.44, + "learning_rate": 1.7916286315241926e-06, + "loss": 1.1232, + "step": 28943 + }, + { + "epoch": 2.44, + "learning_rate": 1.7911094942800545e-06, + "loss": 1.198, + "step": 28944 + }, + { + "epoch": 2.44, + "learning_rate": 1.790590424860401e-06, + "loss": 1.2363, + "step": 28945 + }, + { + "epoch": 2.44, + "learning_rate": 1.7900714232695149e-06, + "loss": 1.0199, + "step": 28946 + }, + { + "epoch": 2.44, + "learning_rate": 1.7895524895116845e-06, + "loss": 1.1836, + "step": 28947 + }, + { + "epoch": 2.44, + "learning_rate": 1.7890336235912032e-06, + "loss": 1.1436, + "step": 28948 + }, + { + "epoch": 2.44, + "learning_rate": 1.7885148255123507e-06, + "loss": 1.0859, + "step": 28949 + }, + { + "epoch": 2.44, + "learning_rate": 1.7879960952794173e-06, + "loss": 1.0868, + "step": 28950 + }, + { + "epoch": 2.44, + "learning_rate": 1.7874774328966893e-06, + "loss": 1.1448, + "step": 28951 + }, + { + "epoch": 2.44, + "learning_rate": 1.78695883836845e-06, + "loss": 1.0199, + "step": 28952 + }, + { + "epoch": 2.44, + "learning_rate": 1.786440311698986e-06, + "loss": 1.0845, + "step": 28953 + }, + { + "epoch": 2.44, + "learning_rate": 1.7859218528925826e-06, + "loss": 1.074, + "step": 28954 + }, + { + "epoch": 2.44, + "learning_rate": 1.7854034619535177e-06, + "loss": 1.1281, + "step": 28955 + }, + { + "epoch": 2.44, + "learning_rate": 1.7848851388860822e-06, + "loss": 1.0775, + "step": 28956 + }, + { + "epoch": 2.44, + "learning_rate": 1.7843668836945527e-06, + "loss": 1.108, + "step": 28957 + }, + { + "epoch": 2.44, + "learning_rate": 1.7838486963832136e-06, + "loss": 1.1735, + "step": 28958 + }, + { + "epoch": 2.44, + "learning_rate": 1.7833305769563458e-06, + "loss": 1.0983, + "step": 28959 + }, + { + "epoch": 2.44, + "learning_rate": 1.7828125254182315e-06, + "loss": 1.1329, + "step": 28960 + }, + { + "epoch": 2.44, + "learning_rate": 1.7822945417731484e-06, + "loss": 1.1235, + "step": 28961 + }, + { + "epoch": 2.44, + "learning_rate": 1.7817766260253787e-06, + "loss": 1.0521, + "step": 28962 + }, + { + "epoch": 2.44, + "learning_rate": 1.7812587781792035e-06, + "loss": 1.0458, + "step": 28963 + }, + { + "epoch": 2.44, + "learning_rate": 1.7807409982388923e-06, + "loss": 1.0656, + "step": 28964 + }, + { + "epoch": 2.44, + "learning_rate": 1.7802232862087355e-06, + "loss": 1.0845, + "step": 28965 + }, + { + "epoch": 2.44, + "learning_rate": 1.7797056420930025e-06, + "loss": 1.0951, + "step": 28966 + }, + { + "epoch": 2.44, + "learning_rate": 1.7791880658959714e-06, + "loss": 1.1247, + "step": 28967 + }, + { + "epoch": 2.44, + "learning_rate": 1.7786705576219198e-06, + "loss": 1.1535, + "step": 28968 + }, + { + "epoch": 2.44, + "learning_rate": 1.7781531172751231e-06, + "loss": 1.1248, + "step": 28969 + }, + { + "epoch": 2.44, + "learning_rate": 1.7776357448598568e-06, + "loss": 1.0608, + "step": 28970 + }, + { + "epoch": 2.44, + "learning_rate": 1.7771184403803977e-06, + "loss": 1.1808, + "step": 28971 + }, + { + "epoch": 2.44, + "learning_rate": 1.776601203841013e-06, + "loss": 1.0544, + "step": 28972 + }, + { + "epoch": 2.44, + "learning_rate": 1.7760840352459841e-06, + "loss": 1.0572, + "step": 28973 + }, + { + "epoch": 2.44, + "learning_rate": 1.775566934599582e-06, + "loss": 1.1717, + "step": 28974 + }, + { + "epoch": 2.44, + "learning_rate": 1.7750499019060741e-06, + "loss": 1.063, + "step": 28975 + }, + { + "epoch": 2.44, + "learning_rate": 1.7745329371697406e-06, + "loss": 1.1865, + "step": 28976 + }, + { + "epoch": 2.44, + "learning_rate": 1.774016040394846e-06, + "loss": 1.1777, + "step": 28977 + }, + { + "epoch": 2.44, + "learning_rate": 1.7734992115856642e-06, + "loss": 1.0724, + "step": 28978 + }, + { + "epoch": 2.44, + "learning_rate": 1.7729824507464644e-06, + "loss": 1.1134, + "step": 28979 + }, + { + "epoch": 2.44, + "learning_rate": 1.7724657578815162e-06, + "loss": 1.0801, + "step": 28980 + }, + { + "epoch": 2.44, + "learning_rate": 1.77194913299509e-06, + "loss": 1.1551, + "step": 28981 + }, + { + "epoch": 2.44, + "learning_rate": 1.7714325760914553e-06, + "loss": 1.1953, + "step": 28982 + }, + { + "epoch": 2.44, + "learning_rate": 1.7709160871748721e-06, + "loss": 1.189, + "step": 28983 + }, + { + "epoch": 2.44, + "learning_rate": 1.7703996662496182e-06, + "loss": 1.1337, + "step": 28984 + }, + { + "epoch": 2.44, + "learning_rate": 1.7698833133199567e-06, + "loss": 1.0691, + "step": 28985 + }, + { + "epoch": 2.44, + "learning_rate": 1.76936702839015e-06, + "loss": 1.1159, + "step": 28986 + }, + { + "epoch": 2.44, + "learning_rate": 1.7688508114644708e-06, + "loss": 1.1696, + "step": 28987 + }, + { + "epoch": 2.44, + "learning_rate": 1.7683346625471787e-06, + "loss": 1.0993, + "step": 28988 + }, + { + "epoch": 2.44, + "learning_rate": 1.7678185816425397e-06, + "loss": 1.1069, + "step": 28989 + }, + { + "epoch": 2.44, + "learning_rate": 1.7673025687548184e-06, + "loss": 1.0487, + "step": 28990 + }, + { + "epoch": 2.44, + "learning_rate": 1.766786623888278e-06, + "loss": 1.0031, + "step": 28991 + }, + { + "epoch": 2.44, + "learning_rate": 1.7662707470471819e-06, + "loss": 1.0239, + "step": 28992 + }, + { + "epoch": 2.44, + "learning_rate": 1.7657549382357931e-06, + "loss": 1.0927, + "step": 28993 + }, + { + "epoch": 2.44, + "learning_rate": 1.7652391974583683e-06, + "loss": 1.1348, + "step": 28994 + }, + { + "epoch": 2.44, + "learning_rate": 1.7647235247191784e-06, + "loss": 1.1821, + "step": 28995 + }, + { + "epoch": 2.44, + "learning_rate": 1.7642079200224749e-06, + "loss": 1.052, + "step": 28996 + }, + { + "epoch": 2.44, + "learning_rate": 1.7636923833725195e-06, + "loss": 1.1833, + "step": 28997 + }, + { + "epoch": 2.44, + "learning_rate": 1.7631769147735778e-06, + "loss": 1.1366, + "step": 28998 + }, + { + "epoch": 2.44, + "learning_rate": 1.762661514229903e-06, + "loss": 1.0961, + "step": 28999 + }, + { + "epoch": 2.44, + "learning_rate": 1.7621461817457542e-06, + "loss": 1.1226, + "step": 29000 + }, + { + "epoch": 2.44, + "learning_rate": 1.761630917325391e-06, + "loss": 1.1493, + "step": 29001 + }, + { + "epoch": 2.44, + "learning_rate": 1.761115720973069e-06, + "loss": 1.0857, + "step": 29002 + }, + { + "epoch": 2.44, + "learning_rate": 1.760600592693046e-06, + "loss": 1.1299, + "step": 29003 + }, + { + "epoch": 2.44, + "learning_rate": 1.7600855324895795e-06, + "loss": 1.2554, + "step": 29004 + }, + { + "epoch": 2.44, + "learning_rate": 1.7595705403669193e-06, + "loss": 1.1573, + "step": 29005 + }, + { + "epoch": 2.44, + "learning_rate": 1.7590556163293293e-06, + "loss": 1.0449, + "step": 29006 + }, + { + "epoch": 2.44, + "learning_rate": 1.758540760381058e-06, + "loss": 1.145, + "step": 29007 + }, + { + "epoch": 2.44, + "learning_rate": 1.7580259725263583e-06, + "loss": 1.0791, + "step": 29008 + }, + { + "epoch": 2.44, + "learning_rate": 1.7575112527694871e-06, + "loss": 1.0991, + "step": 29009 + }, + { + "epoch": 2.44, + "learning_rate": 1.7569966011146956e-06, + "loss": 1.1271, + "step": 29010 + }, + { + "epoch": 2.44, + "learning_rate": 1.7564820175662368e-06, + "loss": 1.1144, + "step": 29011 + }, + { + "epoch": 2.44, + "learning_rate": 1.7559675021283618e-06, + "loss": 1.0188, + "step": 29012 + }, + { + "epoch": 2.44, + "learning_rate": 1.755453054805324e-06, + "loss": 1.053, + "step": 29013 + }, + { + "epoch": 2.44, + "learning_rate": 1.7549386756013654e-06, + "loss": 1.1495, + "step": 29014 + }, + { + "epoch": 2.44, + "learning_rate": 1.7544243645207492e-06, + "loss": 1.1532, + "step": 29015 + }, + { + "epoch": 2.44, + "learning_rate": 1.7539101215677135e-06, + "loss": 1.107, + "step": 29016 + }, + { + "epoch": 2.44, + "learning_rate": 1.7533959467465134e-06, + "loss": 1.1299, + "step": 29017 + }, + { + "epoch": 2.44, + "learning_rate": 1.7528818400613933e-06, + "loss": 1.1337, + "step": 29018 + }, + { + "epoch": 2.44, + "learning_rate": 1.7523678015166045e-06, + "loss": 1.19, + "step": 29019 + }, + { + "epoch": 2.44, + "learning_rate": 1.7518538311163913e-06, + "loss": 1.1483, + "step": 29020 + }, + { + "epoch": 2.44, + "learning_rate": 1.7513399288650024e-06, + "loss": 1.1329, + "step": 29021 + }, + { + "epoch": 2.44, + "learning_rate": 1.7508260947666822e-06, + "loss": 1.1608, + "step": 29022 + }, + { + "epoch": 2.44, + "learning_rate": 1.7503123288256774e-06, + "loss": 1.0893, + "step": 29023 + }, + { + "epoch": 2.44, + "learning_rate": 1.7497986310462344e-06, + "loss": 1.1425, + "step": 29024 + }, + { + "epoch": 2.44, + "learning_rate": 1.7492850014325912e-06, + "loss": 1.0772, + "step": 29025 + }, + { + "epoch": 2.44, + "learning_rate": 1.7487714399889998e-06, + "loss": 1.1023, + "step": 29026 + }, + { + "epoch": 2.44, + "learning_rate": 1.748257946719698e-06, + "loss": 1.1444, + "step": 29027 + }, + { + "epoch": 2.44, + "learning_rate": 1.74774452162893e-06, + "loss": 1.0587, + "step": 29028 + }, + { + "epoch": 2.44, + "learning_rate": 1.7472311647209372e-06, + "loss": 1.0968, + "step": 29029 + }, + { + "epoch": 2.44, + "learning_rate": 1.7467178759999614e-06, + "loss": 1.1566, + "step": 29030 + }, + { + "epoch": 2.44, + "learning_rate": 1.746204655470245e-06, + "loss": 1.0575, + "step": 29031 + }, + { + "epoch": 2.44, + "learning_rate": 1.745691503136029e-06, + "loss": 1.1756, + "step": 29032 + }, + { + "epoch": 2.44, + "learning_rate": 1.7451784190015465e-06, + "loss": 1.206, + "step": 29033 + }, + { + "epoch": 2.44, + "learning_rate": 1.7446654030710453e-06, + "loss": 1.0706, + "step": 29034 + }, + { + "epoch": 2.44, + "learning_rate": 1.7441524553487622e-06, + "loss": 1.1513, + "step": 29035 + }, + { + "epoch": 2.44, + "learning_rate": 1.743639575838929e-06, + "loss": 1.0055, + "step": 29036 + }, + { + "epoch": 2.44, + "learning_rate": 1.7431267645457929e-06, + "loss": 1.1197, + "step": 29037 + }, + { + "epoch": 2.44, + "learning_rate": 1.742614021473583e-06, + "loss": 1.1022, + "step": 29038 + }, + { + "epoch": 2.44, + "learning_rate": 1.742101346626539e-06, + "loss": 1.1074, + "step": 29039 + }, + { + "epoch": 2.44, + "learning_rate": 1.741588740008897e-06, + "loss": 1.0701, + "step": 29040 + }, + { + "epoch": 2.45, + "learning_rate": 1.7410762016248917e-06, + "loss": 1.1427, + "step": 29041 + }, + { + "epoch": 2.45, + "learning_rate": 1.7405637314787582e-06, + "loss": 1.066, + "step": 29042 + }, + { + "epoch": 2.45, + "learning_rate": 1.7400513295747324e-06, + "loss": 1.1683, + "step": 29043 + }, + { + "epoch": 2.45, + "learning_rate": 1.7395389959170416e-06, + "loss": 1.1414, + "step": 29044 + }, + { + "epoch": 2.45, + "learning_rate": 1.7390267305099274e-06, + "loss": 1.1263, + "step": 29045 + }, + { + "epoch": 2.45, + "learning_rate": 1.738514533357617e-06, + "loss": 1.2118, + "step": 29046 + }, + { + "epoch": 2.45, + "learning_rate": 1.7380024044643407e-06, + "loss": 1.1318, + "step": 29047 + }, + { + "epoch": 2.45, + "learning_rate": 1.737490343834337e-06, + "loss": 1.122, + "step": 29048 + }, + { + "epoch": 2.45, + "learning_rate": 1.7369783514718308e-06, + "loss": 1.0308, + "step": 29049 + }, + { + "epoch": 2.45, + "learning_rate": 1.7364664273810527e-06, + "loss": 1.0846, + "step": 29050 + }, + { + "epoch": 2.45, + "learning_rate": 1.7359545715662352e-06, + "loss": 1.0648, + "step": 29051 + }, + { + "epoch": 2.45, + "learning_rate": 1.7354427840316046e-06, + "loss": 1.0776, + "step": 29052 + }, + { + "epoch": 2.45, + "learning_rate": 1.734931064781391e-06, + "loss": 1.0396, + "step": 29053 + }, + { + "epoch": 2.45, + "learning_rate": 1.7344194138198257e-06, + "loss": 1.0726, + "step": 29054 + }, + { + "epoch": 2.45, + "learning_rate": 1.7339078311511259e-06, + "loss": 1.1528, + "step": 29055 + }, + { + "epoch": 2.45, + "learning_rate": 1.733396316779531e-06, + "loss": 1.0712, + "step": 29056 + }, + { + "epoch": 2.45, + "learning_rate": 1.7328848707092594e-06, + "loss": 1.1612, + "step": 29057 + }, + { + "epoch": 2.45, + "learning_rate": 1.732373492944538e-06, + "loss": 1.1783, + "step": 29058 + }, + { + "epoch": 2.45, + "learning_rate": 1.731862183489592e-06, + "loss": 1.02, + "step": 29059 + }, + { + "epoch": 2.45, + "learning_rate": 1.7313509423486485e-06, + "loss": 1.1466, + "step": 29060 + }, + { + "epoch": 2.45, + "learning_rate": 1.7308397695259293e-06, + "loss": 1.0653, + "step": 29061 + }, + { + "epoch": 2.45, + "learning_rate": 1.7303286650256578e-06, + "loss": 1.1181, + "step": 29062 + }, + { + "epoch": 2.45, + "learning_rate": 1.7298176288520607e-06, + "loss": 1.0679, + "step": 29063 + }, + { + "epoch": 2.45, + "learning_rate": 1.7293066610093513e-06, + "loss": 1.1576, + "step": 29064 + }, + { + "epoch": 2.45, + "learning_rate": 1.7287957615017625e-06, + "loss": 1.2568, + "step": 29065 + }, + { + "epoch": 2.45, + "learning_rate": 1.728284930333508e-06, + "loss": 1.1689, + "step": 29066 + }, + { + "epoch": 2.45, + "learning_rate": 1.7277741675088112e-06, + "loss": 1.1588, + "step": 29067 + }, + { + "epoch": 2.45, + "learning_rate": 1.7272634730318905e-06, + "loss": 1.0441, + "step": 29068 + }, + { + "epoch": 2.45, + "learning_rate": 1.7267528469069673e-06, + "loss": 1.0863, + "step": 29069 + }, + { + "epoch": 2.45, + "learning_rate": 1.7262422891382602e-06, + "loss": 1.0502, + "step": 29070 + }, + { + "epoch": 2.45, + "learning_rate": 1.725731799729987e-06, + "loss": 1.2059, + "step": 29071 + }, + { + "epoch": 2.45, + "learning_rate": 1.7252213786863659e-06, + "loss": 1.0945, + "step": 29072 + }, + { + "epoch": 2.45, + "learning_rate": 1.724711026011613e-06, + "loss": 1.0837, + "step": 29073 + }, + { + "epoch": 2.45, + "learning_rate": 1.7242007417099493e-06, + "loss": 1.2132, + "step": 29074 + }, + { + "epoch": 2.45, + "learning_rate": 1.723690525785583e-06, + "loss": 1.1051, + "step": 29075 + }, + { + "epoch": 2.45, + "learning_rate": 1.7231803782427392e-06, + "loss": 1.1506, + "step": 29076 + }, + { + "epoch": 2.45, + "learning_rate": 1.7226702990856248e-06, + "loss": 1.1501, + "step": 29077 + }, + { + "epoch": 2.45, + "learning_rate": 1.7221602883184585e-06, + "loss": 1.1612, + "step": 29078 + }, + { + "epoch": 2.45, + "learning_rate": 1.7216503459454526e-06, + "loss": 1.1151, + "step": 29079 + }, + { + "epoch": 2.45, + "learning_rate": 1.7211404719708214e-06, + "loss": 1.0502, + "step": 29080 + }, + { + "epoch": 2.45, + "learning_rate": 1.7206306663987759e-06, + "loss": 1.216, + "step": 29081 + }, + { + "epoch": 2.45, + "learning_rate": 1.720120929233533e-06, + "loss": 1.1765, + "step": 29082 + }, + { + "epoch": 2.45, + "learning_rate": 1.719611260479296e-06, + "loss": 1.0629, + "step": 29083 + }, + { + "epoch": 2.45, + "learning_rate": 1.7191016601402831e-06, + "loss": 1.1034, + "step": 29084 + }, + { + "epoch": 2.45, + "learning_rate": 1.7185921282207052e-06, + "loss": 1.1655, + "step": 29085 + }, + { + "epoch": 2.45, + "learning_rate": 1.7180826647247651e-06, + "loss": 1.1485, + "step": 29086 + }, + { + "epoch": 2.45, + "learning_rate": 1.7175732696566805e-06, + "loss": 1.1074, + "step": 29087 + }, + { + "epoch": 2.45, + "learning_rate": 1.7170639430206536e-06, + "loss": 1.0508, + "step": 29088 + }, + { + "epoch": 2.45, + "learning_rate": 1.7165546848208958e-06, + "loss": 1.1981, + "step": 29089 + }, + { + "epoch": 2.45, + "learning_rate": 1.7160454950616146e-06, + "loss": 1.0907, + "step": 29090 + }, + { + "epoch": 2.45, + "learning_rate": 1.7155363737470166e-06, + "loss": 1.1649, + "step": 29091 + }, + { + "epoch": 2.45, + "learning_rate": 1.7150273208813084e-06, + "loss": 1.234, + "step": 29092 + }, + { + "epoch": 2.45, + "learning_rate": 1.714518336468698e-06, + "loss": 1.1129, + "step": 29093 + }, + { + "epoch": 2.45, + "learning_rate": 1.714009420513384e-06, + "loss": 1.2047, + "step": 29094 + }, + { + "epoch": 2.45, + "learning_rate": 1.713500573019582e-06, + "loss": 1.0693, + "step": 29095 + }, + { + "epoch": 2.45, + "learning_rate": 1.7129917939914865e-06, + "loss": 1.1169, + "step": 29096 + }, + { + "epoch": 2.45, + "learning_rate": 1.7124830834333027e-06, + "loss": 1.1753, + "step": 29097 + }, + { + "epoch": 2.45, + "learning_rate": 1.7119744413492412e-06, + "loss": 1.1054, + "step": 29098 + }, + { + "epoch": 2.45, + "learning_rate": 1.711465867743497e-06, + "loss": 1.0891, + "step": 29099 + }, + { + "epoch": 2.45, + "learning_rate": 1.7109573626202747e-06, + "loss": 1.1659, + "step": 29100 + }, + { + "epoch": 2.45, + "learning_rate": 1.7104489259837753e-06, + "loss": 1.1656, + "step": 29101 + }, + { + "epoch": 2.45, + "learning_rate": 1.7099405578382e-06, + "loss": 1.0787, + "step": 29102 + }, + { + "epoch": 2.45, + "learning_rate": 1.7094322581877487e-06, + "loss": 1.2361, + "step": 29103 + }, + { + "epoch": 2.45, + "learning_rate": 1.7089240270366237e-06, + "loss": 1.053, + "step": 29104 + }, + { + "epoch": 2.45, + "learning_rate": 1.708415864389017e-06, + "loss": 1.1074, + "step": 29105 + }, + { + "epoch": 2.45, + "learning_rate": 1.7079077702491376e-06, + "loss": 1.138, + "step": 29106 + }, + { + "epoch": 2.45, + "learning_rate": 1.7073997446211742e-06, + "loss": 1.1873, + "step": 29107 + }, + { + "epoch": 2.45, + "learning_rate": 1.7068917875093295e-06, + "loss": 1.0869, + "step": 29108 + }, + { + "epoch": 2.45, + "learning_rate": 1.7063838989177982e-06, + "loss": 1.0801, + "step": 29109 + }, + { + "epoch": 2.45, + "learning_rate": 1.7058760788507778e-06, + "loss": 1.0545, + "step": 29110 + }, + { + "epoch": 2.45, + "learning_rate": 1.7053683273124633e-06, + "loss": 1.0988, + "step": 29111 + }, + { + "epoch": 2.45, + "learning_rate": 1.7048606443070515e-06, + "loss": 1.2016, + "step": 29112 + }, + { + "epoch": 2.45, + "learning_rate": 1.7043530298387368e-06, + "loss": 1.0718, + "step": 29113 + }, + { + "epoch": 2.45, + "learning_rate": 1.703845483911708e-06, + "loss": 1.1343, + "step": 29114 + }, + { + "epoch": 2.45, + "learning_rate": 1.7033380065301686e-06, + "loss": 1.1811, + "step": 29115 + }, + { + "epoch": 2.45, + "learning_rate": 1.7028305976983017e-06, + "loss": 1.1437, + "step": 29116 + }, + { + "epoch": 2.45, + "learning_rate": 1.7023232574203053e-06, + "loss": 1.1241, + "step": 29117 + }, + { + "epoch": 2.45, + "learning_rate": 1.701815985700369e-06, + "loss": 1.1728, + "step": 29118 + }, + { + "epoch": 2.45, + "learning_rate": 1.7013087825426855e-06, + "loss": 1.2041, + "step": 29119 + }, + { + "epoch": 2.45, + "learning_rate": 1.7008016479514433e-06, + "loss": 1.0864, + "step": 29120 + }, + { + "epoch": 2.45, + "learning_rate": 1.7002945819308358e-06, + "loss": 1.0201, + "step": 29121 + }, + { + "epoch": 2.45, + "learning_rate": 1.6997875844850497e-06, + "loss": 1.1067, + "step": 29122 + }, + { + "epoch": 2.45, + "learning_rate": 1.699280655618275e-06, + "loss": 1.1304, + "step": 29123 + }, + { + "epoch": 2.45, + "learning_rate": 1.6987737953347028e-06, + "loss": 1.1078, + "step": 29124 + }, + { + "epoch": 2.45, + "learning_rate": 1.6982670036385129e-06, + "loss": 1.1081, + "step": 29125 + }, + { + "epoch": 2.45, + "learning_rate": 1.6977602805339033e-06, + "loss": 1.113, + "step": 29126 + }, + { + "epoch": 2.45, + "learning_rate": 1.6972536260250526e-06, + "loss": 1.1356, + "step": 29127 + }, + { + "epoch": 2.45, + "learning_rate": 1.6967470401161502e-06, + "loss": 1.1562, + "step": 29128 + }, + { + "epoch": 2.45, + "learning_rate": 1.6962405228113798e-06, + "loss": 1.0627, + "step": 29129 + }, + { + "epoch": 2.45, + "learning_rate": 1.6957340741149287e-06, + "loss": 1.0915, + "step": 29130 + }, + { + "epoch": 2.45, + "learning_rate": 1.6952276940309808e-06, + "loss": 1.0773, + "step": 29131 + }, + { + "epoch": 2.45, + "learning_rate": 1.694721382563721e-06, + "loss": 1.1333, + "step": 29132 + }, + { + "epoch": 2.45, + "learning_rate": 1.6942151397173268e-06, + "loss": 1.1392, + "step": 29133 + }, + { + "epoch": 2.45, + "learning_rate": 1.6937089654959882e-06, + "loss": 1.0778, + "step": 29134 + }, + { + "epoch": 2.45, + "learning_rate": 1.6932028599038863e-06, + "loss": 1.0993, + "step": 29135 + }, + { + "epoch": 2.45, + "learning_rate": 1.6926968229451967e-06, + "loss": 1.0487, + "step": 29136 + }, + { + "epoch": 2.45, + "learning_rate": 1.6921908546241084e-06, + "loss": 1.1654, + "step": 29137 + }, + { + "epoch": 2.45, + "learning_rate": 1.6916849549447967e-06, + "loss": 1.1148, + "step": 29138 + }, + { + "epoch": 2.45, + "learning_rate": 1.6911791239114428e-06, + "loss": 1.0684, + "step": 29139 + }, + { + "epoch": 2.45, + "learning_rate": 1.6906733615282256e-06, + "loss": 1.1003, + "step": 29140 + }, + { + "epoch": 2.45, + "learning_rate": 1.6901676677993239e-06, + "loss": 1.105, + "step": 29141 + }, + { + "epoch": 2.45, + "learning_rate": 1.6896620427289179e-06, + "loss": 1.1183, + "step": 29142 + }, + { + "epoch": 2.45, + "learning_rate": 1.6891564863211851e-06, + "loss": 1.1009, + "step": 29143 + }, + { + "epoch": 2.45, + "learning_rate": 1.6886509985802967e-06, + "loss": 1.1464, + "step": 29144 + }, + { + "epoch": 2.45, + "learning_rate": 1.6881455795104384e-06, + "loss": 1.0077, + "step": 29145 + }, + { + "epoch": 2.45, + "learning_rate": 1.687640229115779e-06, + "loss": 1.2095, + "step": 29146 + }, + { + "epoch": 2.45, + "learning_rate": 1.6871349474004938e-06, + "loss": 1.1348, + "step": 29147 + }, + { + "epoch": 2.45, + "learning_rate": 1.6866297343687655e-06, + "loss": 1.1893, + "step": 29148 + }, + { + "epoch": 2.45, + "learning_rate": 1.6861245900247603e-06, + "loss": 1.1652, + "step": 29149 + }, + { + "epoch": 2.45, + "learning_rate": 1.685619514372655e-06, + "loss": 1.0192, + "step": 29150 + }, + { + "epoch": 2.45, + "learning_rate": 1.685114507416621e-06, + "loss": 1.1019, + "step": 29151 + }, + { + "epoch": 2.45, + "learning_rate": 1.6846095691608334e-06, + "loss": 1.047, + "step": 29152 + }, + { + "epoch": 2.45, + "learning_rate": 1.6841046996094623e-06, + "loss": 1.0782, + "step": 29153 + }, + { + "epoch": 2.45, + "learning_rate": 1.6835998987666812e-06, + "loss": 1.0899, + "step": 29154 + }, + { + "epoch": 2.45, + "learning_rate": 1.6830951666366557e-06, + "loss": 1.1075, + "step": 29155 + }, + { + "epoch": 2.45, + "learning_rate": 1.6825905032235635e-06, + "loss": 1.0845, + "step": 29156 + }, + { + "epoch": 2.45, + "learning_rate": 1.6820859085315687e-06, + "loss": 1.0529, + "step": 29157 + }, + { + "epoch": 2.45, + "learning_rate": 1.681581382564842e-06, + "loss": 1.0778, + "step": 29158 + }, + { + "epoch": 2.45, + "learning_rate": 1.6810769253275526e-06, + "loss": 1.1099, + "step": 29159 + }, + { + "epoch": 2.46, + "learning_rate": 1.6805725368238668e-06, + "loss": 1.0568, + "step": 29160 + }, + { + "epoch": 2.46, + "learning_rate": 1.680068217057954e-06, + "loss": 1.1781, + "step": 29161 + }, + { + "epoch": 2.46, + "learning_rate": 1.6795639660339814e-06, + "loss": 1.129, + "step": 29162 + }, + { + "epoch": 2.46, + "learning_rate": 1.679059783756115e-06, + "loss": 1.1945, + "step": 29163 + }, + { + "epoch": 2.46, + "learning_rate": 1.6785556702285155e-06, + "loss": 1.1341, + "step": 29164 + }, + { + "epoch": 2.46, + "learning_rate": 1.678051625455357e-06, + "loss": 1.0941, + "step": 29165 + }, + { + "epoch": 2.46, + "learning_rate": 1.6775476494407972e-06, + "loss": 1.1714, + "step": 29166 + }, + { + "epoch": 2.46, + "learning_rate": 1.6770437421890017e-06, + "loss": 1.0621, + "step": 29167 + }, + { + "epoch": 2.46, + "learning_rate": 1.6765399037041353e-06, + "loss": 1.1291, + "step": 29168 + }, + { + "epoch": 2.46, + "learning_rate": 1.6760361339903596e-06, + "loss": 1.1629, + "step": 29169 + }, + { + "epoch": 2.46, + "learning_rate": 1.6755324330518375e-06, + "loss": 1.0595, + "step": 29170 + }, + { + "epoch": 2.46, + "learning_rate": 1.6750288008927307e-06, + "loss": 1.121, + "step": 29171 + }, + { + "epoch": 2.46, + "learning_rate": 1.6745252375171994e-06, + "loss": 1.1764, + "step": 29172 + }, + { + "epoch": 2.46, + "learning_rate": 1.674021742929406e-06, + "loss": 1.0066, + "step": 29173 + }, + { + "epoch": 2.46, + "learning_rate": 1.6735183171335112e-06, + "loss": 1.1098, + "step": 29174 + }, + { + "epoch": 2.46, + "learning_rate": 1.6730149601336688e-06, + "loss": 1.1847, + "step": 29175 + }, + { + "epoch": 2.46, + "learning_rate": 1.6725116719340463e-06, + "loss": 1.2588, + "step": 29176 + }, + { + "epoch": 2.46, + "learning_rate": 1.6720084525387957e-06, + "loss": 1.1265, + "step": 29177 + }, + { + "epoch": 2.46, + "learning_rate": 1.6715053019520755e-06, + "loss": 1.0548, + "step": 29178 + }, + { + "epoch": 2.46, + "learning_rate": 1.6710022201780452e-06, + "loss": 1.0725, + "step": 29179 + }, + { + "epoch": 2.46, + "learning_rate": 1.6704992072208593e-06, + "loss": 1.0874, + "step": 29180 + }, + { + "epoch": 2.46, + "learning_rate": 1.6699962630846745e-06, + "loss": 1.0525, + "step": 29181 + }, + { + "epoch": 2.46, + "learning_rate": 1.6694933877736497e-06, + "loss": 1.1235, + "step": 29182 + }, + { + "epoch": 2.46, + "learning_rate": 1.6689905812919317e-06, + "loss": 1.1633, + "step": 29183 + }, + { + "epoch": 2.46, + "learning_rate": 1.6684878436436813e-06, + "loss": 1.0798, + "step": 29184 + }, + { + "epoch": 2.46, + "learning_rate": 1.6679851748330545e-06, + "loss": 1.0897, + "step": 29185 + }, + { + "epoch": 2.46, + "learning_rate": 1.6674825748641953e-06, + "loss": 1.0342, + "step": 29186 + }, + { + "epoch": 2.46, + "learning_rate": 1.6669800437412665e-06, + "loss": 1.0235, + "step": 29187 + }, + { + "epoch": 2.46, + "learning_rate": 1.6664775814684132e-06, + "loss": 1.0332, + "step": 29188 + }, + { + "epoch": 2.46, + "learning_rate": 1.6659751880497888e-06, + "loss": 1.1035, + "step": 29189 + }, + { + "epoch": 2.46, + "learning_rate": 1.6654728634895456e-06, + "loss": 1.041, + "step": 29190 + }, + { + "epoch": 2.46, + "learning_rate": 1.6649706077918326e-06, + "loss": 1.1687, + "step": 29191 + }, + { + "epoch": 2.46, + "learning_rate": 1.6644684209607998e-06, + "loss": 1.1068, + "step": 29192 + }, + { + "epoch": 2.46, + "learning_rate": 1.6639663030005981e-06, + "loss": 1.0686, + "step": 29193 + }, + { + "epoch": 2.46, + "learning_rate": 1.6634642539153712e-06, + "loss": 1.1103, + "step": 29194 + }, + { + "epoch": 2.46, + "learning_rate": 1.6629622737092754e-06, + "loss": 1.124, + "step": 29195 + }, + { + "epoch": 2.46, + "learning_rate": 1.6624603623864511e-06, + "loss": 1.0475, + "step": 29196 + }, + { + "epoch": 2.46, + "learning_rate": 1.6619585199510447e-06, + "loss": 1.0818, + "step": 29197 + }, + { + "epoch": 2.46, + "learning_rate": 1.6614567464072095e-06, + "loss": 1.1281, + "step": 29198 + }, + { + "epoch": 2.46, + "learning_rate": 1.660955041759087e-06, + "loss": 1.1745, + "step": 29199 + }, + { + "epoch": 2.46, + "learning_rate": 1.6604534060108213e-06, + "loss": 1.0709, + "step": 29200 + }, + { + "epoch": 2.46, + "learning_rate": 1.6599518391665592e-06, + "loss": 1.1615, + "step": 29201 + }, + { + "epoch": 2.46, + "learning_rate": 1.6594503412304441e-06, + "loss": 1.1364, + "step": 29202 + }, + { + "epoch": 2.46, + "learning_rate": 1.6589489122066194e-06, + "loss": 1.1374, + "step": 29203 + }, + { + "epoch": 2.46, + "learning_rate": 1.6584475520992305e-06, + "loss": 1.111, + "step": 29204 + }, + { + "epoch": 2.46, + "learning_rate": 1.657946260912413e-06, + "loss": 1.0816, + "step": 29205 + }, + { + "epoch": 2.46, + "learning_rate": 1.6574450386503183e-06, + "loss": 1.0274, + "step": 29206 + }, + { + "epoch": 2.46, + "learning_rate": 1.6569438853170794e-06, + "loss": 1.1032, + "step": 29207 + }, + { + "epoch": 2.46, + "learning_rate": 1.656442800916841e-06, + "loss": 1.0814, + "step": 29208 + }, + { + "epoch": 2.46, + "learning_rate": 1.655941785453742e-06, + "loss": 1.1638, + "step": 29209 + }, + { + "epoch": 2.46, + "learning_rate": 1.6554408389319232e-06, + "loss": 1.1294, + "step": 29210 + }, + { + "epoch": 2.46, + "learning_rate": 1.6549399613555218e-06, + "loss": 1.0607, + "step": 29211 + }, + { + "epoch": 2.46, + "learning_rate": 1.6544391527286784e-06, + "loss": 1.04, + "step": 29212 + }, + { + "epoch": 2.46, + "learning_rate": 1.6539384130555292e-06, + "loss": 1.1535, + "step": 29213 + }, + { + "epoch": 2.46, + "learning_rate": 1.6534377423402114e-06, + "loss": 1.0998, + "step": 29214 + }, + { + "epoch": 2.46, + "learning_rate": 1.6529371405868644e-06, + "loss": 1.118, + "step": 29215 + }, + { + "epoch": 2.46, + "learning_rate": 1.652436607799618e-06, + "loss": 1.1418, + "step": 29216 + }, + { + "epoch": 2.46, + "learning_rate": 1.6519361439826164e-06, + "loss": 1.0793, + "step": 29217 + }, + { + "epoch": 2.46, + "learning_rate": 1.651435749139988e-06, + "loss": 1.0452, + "step": 29218 + }, + { + "epoch": 2.46, + "learning_rate": 1.650935423275869e-06, + "loss": 1.2099, + "step": 29219 + }, + { + "epoch": 2.46, + "learning_rate": 1.650435166394394e-06, + "loss": 1.1088, + "step": 29220 + }, + { + "epoch": 2.46, + "learning_rate": 1.6499349784996953e-06, + "loss": 1.0608, + "step": 29221 + }, + { + "epoch": 2.46, + "learning_rate": 1.6494348595959065e-06, + "loss": 1.0526, + "step": 29222 + }, + { + "epoch": 2.46, + "learning_rate": 1.6489348096871595e-06, + "loss": 1.1535, + "step": 29223 + }, + { + "epoch": 2.46, + "learning_rate": 1.6484348287775886e-06, + "loss": 1.2015, + "step": 29224 + }, + { + "epoch": 2.46, + "learning_rate": 1.6479349168713165e-06, + "loss": 1.0365, + "step": 29225 + }, + { + "epoch": 2.46, + "learning_rate": 1.6474350739724842e-06, + "loss": 1.0369, + "step": 29226 + }, + { + "epoch": 2.46, + "learning_rate": 1.646935300085214e-06, + "loss": 1.1745, + "step": 29227 + }, + { + "epoch": 2.46, + "learning_rate": 1.6464355952136368e-06, + "loss": 1.0568, + "step": 29228 + }, + { + "epoch": 2.46, + "learning_rate": 1.645935959361884e-06, + "loss": 1.0247, + "step": 29229 + }, + { + "epoch": 2.46, + "learning_rate": 1.64543639253408e-06, + "loss": 1.1204, + "step": 29230 + }, + { + "epoch": 2.46, + "learning_rate": 1.6449368947343559e-06, + "loss": 1.1209, + "step": 29231 + }, + { + "epoch": 2.46, + "learning_rate": 1.6444374659668393e-06, + "loss": 1.0917, + "step": 29232 + }, + { + "epoch": 2.46, + "learning_rate": 1.6439381062356486e-06, + "loss": 1.068, + "step": 29233 + }, + { + "epoch": 2.46, + "learning_rate": 1.6434388155449188e-06, + "loss": 1.0151, + "step": 29234 + }, + { + "epoch": 2.46, + "learning_rate": 1.6429395938987747e-06, + "loss": 1.213, + "step": 29235 + }, + { + "epoch": 2.46, + "learning_rate": 1.6424404413013328e-06, + "loss": 1.1367, + "step": 29236 + }, + { + "epoch": 2.46, + "learning_rate": 1.641941357756728e-06, + "loss": 1.0958, + "step": 29237 + }, + { + "epoch": 2.46, + "learning_rate": 1.641442343269075e-06, + "loss": 1.1092, + "step": 29238 + }, + { + "epoch": 2.46, + "learning_rate": 1.6409433978425015e-06, + "loss": 1.1805, + "step": 29239 + }, + { + "epoch": 2.46, + "learning_rate": 1.6404445214811294e-06, + "loss": 1.0608, + "step": 29240 + }, + { + "epoch": 2.46, + "learning_rate": 1.6399457141890795e-06, + "loss": 1.1019, + "step": 29241 + }, + { + "epoch": 2.46, + "learning_rate": 1.6394469759704734e-06, + "loss": 1.16, + "step": 29242 + }, + { + "epoch": 2.46, + "learning_rate": 1.6389483068294343e-06, + "loss": 1.191, + "step": 29243 + }, + { + "epoch": 2.46, + "learning_rate": 1.6384497067700766e-06, + "loss": 1.0929, + "step": 29244 + }, + { + "epoch": 2.46, + "learning_rate": 1.6379511757965272e-06, + "loss": 1.1857, + "step": 29245 + }, + { + "epoch": 2.46, + "learning_rate": 1.6374527139128994e-06, + "loss": 1.1015, + "step": 29246 + }, + { + "epoch": 2.46, + "learning_rate": 1.6369543211233108e-06, + "loss": 1.182, + "step": 29247 + }, + { + "epoch": 2.46, + "learning_rate": 1.6364559974318871e-06, + "loss": 1.1481, + "step": 29248 + }, + { + "epoch": 2.46, + "learning_rate": 1.6359577428427386e-06, + "loss": 1.1767, + "step": 29249 + }, + { + "epoch": 2.46, + "learning_rate": 1.635459557359983e-06, + "loss": 1.2163, + "step": 29250 + }, + { + "epoch": 2.46, + "learning_rate": 1.634961440987739e-06, + "loss": 1.1688, + "step": 29251 + }, + { + "epoch": 2.46, + "learning_rate": 1.6344633937301202e-06, + "loss": 1.0792, + "step": 29252 + }, + { + "epoch": 2.46, + "learning_rate": 1.633965415591242e-06, + "loss": 1.097, + "step": 29253 + }, + { + "epoch": 2.46, + "learning_rate": 1.6334675065752215e-06, + "loss": 1.0264, + "step": 29254 + }, + { + "epoch": 2.46, + "learning_rate": 1.632969666686165e-06, + "loss": 1.1649, + "step": 29255 + }, + { + "epoch": 2.46, + "learning_rate": 1.632471895928196e-06, + "loss": 1.104, + "step": 29256 + }, + { + "epoch": 2.46, + "learning_rate": 1.631974194305419e-06, + "loss": 1.1645, + "step": 29257 + }, + { + "epoch": 2.46, + "learning_rate": 1.6314765618219508e-06, + "loss": 1.0705, + "step": 29258 + }, + { + "epoch": 2.46, + "learning_rate": 1.6309789984819003e-06, + "loss": 1.0248, + "step": 29259 + }, + { + "epoch": 2.46, + "learning_rate": 1.6304815042893796e-06, + "loss": 1.145, + "step": 29260 + }, + { + "epoch": 2.46, + "learning_rate": 1.6299840792485e-06, + "loss": 1.1536, + "step": 29261 + }, + { + "epoch": 2.46, + "learning_rate": 1.6294867233633704e-06, + "loss": 1.1491, + "step": 29262 + }, + { + "epoch": 2.46, + "learning_rate": 1.628989436638101e-06, + "loss": 1.0518, + "step": 29263 + }, + { + "epoch": 2.46, + "learning_rate": 1.6284922190767994e-06, + "loss": 1.1551, + "step": 29264 + }, + { + "epoch": 2.46, + "learning_rate": 1.6279950706835768e-06, + "loss": 1.1357, + "step": 29265 + }, + { + "epoch": 2.46, + "learning_rate": 1.6274979914625332e-06, + "loss": 1.1754, + "step": 29266 + }, + { + "epoch": 2.46, + "learning_rate": 1.6270009814177856e-06, + "loss": 1.1255, + "step": 29267 + }, + { + "epoch": 2.46, + "learning_rate": 1.6265040405534338e-06, + "loss": 1.1455, + "step": 29268 + }, + { + "epoch": 2.46, + "learning_rate": 1.6260071688735857e-06, + "loss": 1.0444, + "step": 29269 + }, + { + "epoch": 2.46, + "learning_rate": 1.6255103663823456e-06, + "loss": 1.011, + "step": 29270 + }, + { + "epoch": 2.46, + "learning_rate": 1.6250136330838185e-06, + "loss": 1.041, + "step": 29271 + }, + { + "epoch": 2.46, + "learning_rate": 1.6245169689821106e-06, + "loss": 1.1632, + "step": 29272 + }, + { + "epoch": 2.46, + "learning_rate": 1.6240203740813232e-06, + "loss": 1.0934, + "step": 29273 + }, + { + "epoch": 2.46, + "learning_rate": 1.6235238483855619e-06, + "loss": 1.0922, + "step": 29274 + }, + { + "epoch": 2.46, + "learning_rate": 1.6230273918989236e-06, + "loss": 0.9965, + "step": 29275 + }, + { + "epoch": 2.46, + "learning_rate": 1.6225310046255183e-06, + "loss": 1.1042, + "step": 29276 + }, + { + "epoch": 2.46, + "learning_rate": 1.6220346865694403e-06, + "loss": 1.1795, + "step": 29277 + }, + { + "epoch": 2.46, + "learning_rate": 1.6215384377347941e-06, + "loss": 1.142, + "step": 29278 + }, + { + "epoch": 2.47, + "learning_rate": 1.6210422581256779e-06, + "loss": 1.1576, + "step": 29279 + }, + { + "epoch": 2.47, + "learning_rate": 1.6205461477461926e-06, + "loss": 1.0685, + "step": 29280 + }, + { + "epoch": 2.47, + "learning_rate": 1.6200501066004358e-06, + "loss": 1.1554, + "step": 29281 + }, + { + "epoch": 2.47, + "learning_rate": 1.6195541346925104e-06, + "loss": 1.1511, + "step": 29282 + }, + { + "epoch": 2.47, + "learning_rate": 1.6190582320265058e-06, + "loss": 1.1575, + "step": 29283 + }, + { + "epoch": 2.47, + "learning_rate": 1.6185623986065269e-06, + "loss": 1.0401, + "step": 29284 + }, + { + "epoch": 2.47, + "learning_rate": 1.618066634436669e-06, + "loss": 1.24, + "step": 29285 + }, + { + "epoch": 2.47, + "learning_rate": 1.6175709395210236e-06, + "loss": 1.1389, + "step": 29286 + }, + { + "epoch": 2.47, + "learning_rate": 1.6170753138636952e-06, + "loss": 1.0913, + "step": 29287 + }, + { + "epoch": 2.47, + "learning_rate": 1.6165797574687693e-06, + "loss": 1.1824, + "step": 29288 + }, + { + "epoch": 2.47, + "learning_rate": 1.6160842703403457e-06, + "loss": 1.0972, + "step": 29289 + }, + { + "epoch": 2.47, + "learning_rate": 1.615588852482517e-06, + "loss": 1.1496, + "step": 29290 + }, + { + "epoch": 2.47, + "learning_rate": 1.6150935038993765e-06, + "loss": 1.1021, + "step": 29291 + }, + { + "epoch": 2.47, + "learning_rate": 1.6145982245950164e-06, + "loss": 1.1856, + "step": 29292 + }, + { + "epoch": 2.47, + "learning_rate": 1.6141030145735326e-06, + "loss": 1.0391, + "step": 29293 + }, + { + "epoch": 2.47, + "learning_rate": 1.6136078738390081e-06, + "loss": 1.0819, + "step": 29294 + }, + { + "epoch": 2.47, + "learning_rate": 1.6131128023955444e-06, + "loss": 1.0996, + "step": 29295 + }, + { + "epoch": 2.47, + "learning_rate": 1.6126178002472237e-06, + "loss": 1.0314, + "step": 29296 + }, + { + "epoch": 2.47, + "learning_rate": 1.6121228673981381e-06, + "loss": 1.1681, + "step": 29297 + }, + { + "epoch": 2.47, + "learning_rate": 1.6116280038523813e-06, + "loss": 1.1203, + "step": 29298 + }, + { + "epoch": 2.47, + "learning_rate": 1.6111332096140375e-06, + "loss": 1.1901, + "step": 29299 + }, + { + "epoch": 2.47, + "learning_rate": 1.6106384846871948e-06, + "loss": 1.0715, + "step": 29300 + }, + { + "epoch": 2.47, + "learning_rate": 1.610143829075942e-06, + "loss": 1.1072, + "step": 29301 + }, + { + "epoch": 2.47, + "learning_rate": 1.6096492427843668e-06, + "loss": 1.178, + "step": 29302 + }, + { + "epoch": 2.47, + "learning_rate": 1.6091547258165541e-06, + "loss": 1.1608, + "step": 29303 + }, + { + "epoch": 2.47, + "learning_rate": 1.6086602781765925e-06, + "loss": 1.1395, + "step": 29304 + }, + { + "epoch": 2.47, + "learning_rate": 1.6081658998685611e-06, + "loss": 1.1383, + "step": 29305 + }, + { + "epoch": 2.47, + "learning_rate": 1.6076715908965534e-06, + "loss": 1.0991, + "step": 29306 + }, + { + "epoch": 2.47, + "learning_rate": 1.6071773512646471e-06, + "loss": 1.082, + "step": 29307 + }, + { + "epoch": 2.47, + "learning_rate": 1.6066831809769267e-06, + "loss": 1.1533, + "step": 29308 + }, + { + "epoch": 2.47, + "learning_rate": 1.6061890800374757e-06, + "loss": 1.102, + "step": 29309 + }, + { + "epoch": 2.47, + "learning_rate": 1.6056950484503786e-06, + "loss": 1.1415, + "step": 29310 + }, + { + "epoch": 2.47, + "learning_rate": 1.6052010862197142e-06, + "loss": 1.1459, + "step": 29311 + }, + { + "epoch": 2.47, + "learning_rate": 1.604707193349565e-06, + "loss": 1.1424, + "step": 29312 + }, + { + "epoch": 2.47, + "learning_rate": 1.6042133698440121e-06, + "loss": 1.1855, + "step": 29313 + }, + { + "epoch": 2.47, + "learning_rate": 1.6037196157071356e-06, + "loss": 1.114, + "step": 29314 + }, + { + "epoch": 2.47, + "learning_rate": 1.6032259309430177e-06, + "loss": 1.066, + "step": 29315 + }, + { + "epoch": 2.47, + "learning_rate": 1.6027323155557283e-06, + "loss": 1.0615, + "step": 29316 + }, + { + "epoch": 2.47, + "learning_rate": 1.602238769549358e-06, + "loss": 1.1163, + "step": 29317 + }, + { + "epoch": 2.47, + "learning_rate": 1.6017452929279752e-06, + "loss": 1.1417, + "step": 29318 + }, + { + "epoch": 2.47, + "learning_rate": 1.6012518856956617e-06, + "loss": 1.1402, + "step": 29319 + }, + { + "epoch": 2.47, + "learning_rate": 1.6007585478564925e-06, + "loss": 1.1263, + "step": 29320 + }, + { + "epoch": 2.47, + "learning_rate": 1.6002652794145435e-06, + "loss": 1.1735, + "step": 29321 + }, + { + "epoch": 2.47, + "learning_rate": 1.5997720803738914e-06, + "loss": 1.0266, + "step": 29322 + }, + { + "epoch": 2.47, + "learning_rate": 1.5992789507386118e-06, + "loss": 1.1392, + "step": 29323 + }, + { + "epoch": 2.47, + "learning_rate": 1.5987858905127795e-06, + "loss": 1.0972, + "step": 29324 + }, + { + "epoch": 2.47, + "learning_rate": 1.5982928997004621e-06, + "loss": 1.0812, + "step": 29325 + }, + { + "epoch": 2.47, + "learning_rate": 1.597799978305743e-06, + "loss": 1.0364, + "step": 29326 + }, + { + "epoch": 2.47, + "learning_rate": 1.5973071263326868e-06, + "loss": 1.1944, + "step": 29327 + }, + { + "epoch": 2.47, + "learning_rate": 1.5968143437853689e-06, + "loss": 1.1368, + "step": 29328 + }, + { + "epoch": 2.47, + "learning_rate": 1.5963216306678596e-06, + "loss": 1.2231, + "step": 29329 + }, + { + "epoch": 2.47, + "learning_rate": 1.5958289869842302e-06, + "loss": 1.0951, + "step": 29330 + }, + { + "epoch": 2.47, + "learning_rate": 1.5953364127385517e-06, + "loss": 1.1312, + "step": 29331 + }, + { + "epoch": 2.47, + "learning_rate": 1.5948439079348966e-06, + "loss": 1.0595, + "step": 29332 + }, + { + "epoch": 2.47, + "learning_rate": 1.5943514725773257e-06, + "loss": 1.1305, + "step": 29333 + }, + { + "epoch": 2.47, + "learning_rate": 1.593859106669916e-06, + "loss": 1.0878, + "step": 29334 + }, + { + "epoch": 2.47, + "learning_rate": 1.5933668102167343e-06, + "loss": 1.114, + "step": 29335 + }, + { + "epoch": 2.47, + "learning_rate": 1.5928745832218428e-06, + "loss": 1.1132, + "step": 29336 + }, + { + "epoch": 2.47, + "learning_rate": 1.5923824256893161e-06, + "loss": 1.0806, + "step": 29337 + }, + { + "epoch": 2.47, + "learning_rate": 1.5918903376232142e-06, + "loss": 1.1269, + "step": 29338 + }, + { + "epoch": 2.47, + "learning_rate": 1.591398319027605e-06, + "loss": 1.1058, + "step": 29339 + }, + { + "epoch": 2.47, + "learning_rate": 1.5909063699065552e-06, + "loss": 1.0315, + "step": 29340 + }, + { + "epoch": 2.47, + "learning_rate": 1.5904144902641272e-06, + "loss": 1.1479, + "step": 29341 + }, + { + "epoch": 2.47, + "learning_rate": 1.5899226801043866e-06, + "loss": 1.0985, + "step": 29342 + }, + { + "epoch": 2.47, + "learning_rate": 1.5894309394313978e-06, + "loss": 1.1316, + "step": 29343 + }, + { + "epoch": 2.47, + "learning_rate": 1.5889392682492188e-06, + "loss": 1.1349, + "step": 29344 + }, + { + "epoch": 2.47, + "learning_rate": 1.5884476665619198e-06, + "loss": 1.1689, + "step": 29345 + }, + { + "epoch": 2.47, + "learning_rate": 1.587956134373555e-06, + "loss": 1.1619, + "step": 29346 + }, + { + "epoch": 2.47, + "learning_rate": 1.587464671688187e-06, + "loss": 1.1688, + "step": 29347 + }, + { + "epoch": 2.47, + "learning_rate": 1.5869732785098834e-06, + "loss": 1.1106, + "step": 29348 + }, + { + "epoch": 2.47, + "learning_rate": 1.586481954842697e-06, + "loss": 1.0589, + "step": 29349 + }, + { + "epoch": 2.47, + "learning_rate": 1.5859907006906883e-06, + "loss": 1.2169, + "step": 29350 + }, + { + "epoch": 2.47, + "learning_rate": 1.5854995160579178e-06, + "loss": 1.207, + "step": 29351 + }, + { + "epoch": 2.47, + "learning_rate": 1.5850084009484434e-06, + "loss": 1.1078, + "step": 29352 + }, + { + "epoch": 2.47, + "learning_rate": 1.5845173553663218e-06, + "loss": 0.9947, + "step": 29353 + }, + { + "epoch": 2.47, + "learning_rate": 1.584026379315614e-06, + "loss": 1.1434, + "step": 29354 + }, + { + "epoch": 2.47, + "learning_rate": 1.5835354728003694e-06, + "loss": 1.1431, + "step": 29355 + }, + { + "epoch": 2.47, + "learning_rate": 1.5830446358246532e-06, + "loss": 1.2354, + "step": 29356 + }, + { + "epoch": 2.47, + "learning_rate": 1.5825538683925134e-06, + "loss": 1.1288, + "step": 29357 + }, + { + "epoch": 2.47, + "learning_rate": 1.582063170508007e-06, + "loss": 1.1833, + "step": 29358 + }, + { + "epoch": 2.47, + "learning_rate": 1.5815725421751892e-06, + "loss": 1.1064, + "step": 29359 + }, + { + "epoch": 2.47, + "learning_rate": 1.581081983398114e-06, + "loss": 1.0417, + "step": 29360 + }, + { + "epoch": 2.47, + "learning_rate": 1.580591494180832e-06, + "loss": 1.1665, + "step": 29361 + }, + { + "epoch": 2.47, + "learning_rate": 1.5801010745273993e-06, + "loss": 1.0246, + "step": 29362 + }, + { + "epoch": 2.47, + "learning_rate": 1.5796107244418658e-06, + "loss": 1.1223, + "step": 29363 + }, + { + "epoch": 2.47, + "learning_rate": 1.5791204439282827e-06, + "loss": 0.9871, + "step": 29364 + }, + { + "epoch": 2.47, + "learning_rate": 1.5786302329907043e-06, + "loss": 1.1036, + "step": 29365 + }, + { + "epoch": 2.47, + "learning_rate": 1.5781400916331735e-06, + "loss": 1.1883, + "step": 29366 + }, + { + "epoch": 2.47, + "learning_rate": 1.5776500198597499e-06, + "loss": 1.1463, + "step": 29367 + }, + { + "epoch": 2.47, + "learning_rate": 1.577160017674474e-06, + "loss": 1.1372, + "step": 29368 + }, + { + "epoch": 2.47, + "learning_rate": 1.5766700850813987e-06, + "loss": 1.1202, + "step": 29369 + }, + { + "epoch": 2.47, + "learning_rate": 1.5761802220845702e-06, + "loss": 1.213, + "step": 29370 + }, + { + "epoch": 2.47, + "learning_rate": 1.575690428688036e-06, + "loss": 1.0469, + "step": 29371 + }, + { + "epoch": 2.47, + "learning_rate": 1.5752007048958451e-06, + "loss": 1.1779, + "step": 29372 + }, + { + "epoch": 2.47, + "learning_rate": 1.574711050712041e-06, + "loss": 1.1399, + "step": 29373 + }, + { + "epoch": 2.47, + "learning_rate": 1.574221466140674e-06, + "loss": 1.0801, + "step": 29374 + }, + { + "epoch": 2.47, + "learning_rate": 1.5737319511857806e-06, + "loss": 1.0713, + "step": 29375 + }, + { + "epoch": 2.47, + "learning_rate": 1.5732425058514157e-06, + "loss": 1.1403, + "step": 29376 + }, + { + "epoch": 2.47, + "learning_rate": 1.5727531301416155e-06, + "loss": 1.0806, + "step": 29377 + }, + { + "epoch": 2.47, + "learning_rate": 1.572263824060426e-06, + "loss": 1.148, + "step": 29378 + }, + { + "epoch": 2.47, + "learning_rate": 1.5717745876118896e-06, + "loss": 0.9884, + "step": 29379 + }, + { + "epoch": 2.47, + "learning_rate": 1.5712854208000495e-06, + "loss": 1.1523, + "step": 29380 + }, + { + "epoch": 2.47, + "learning_rate": 1.5707963236289459e-06, + "loss": 1.103, + "step": 29381 + }, + { + "epoch": 2.47, + "learning_rate": 1.5703072961026244e-06, + "loss": 1.0619, + "step": 29382 + }, + { + "epoch": 2.47, + "learning_rate": 1.5698183382251164e-06, + "loss": 1.155, + "step": 29383 + }, + { + "epoch": 2.47, + "learning_rate": 1.5693294500004697e-06, + "loss": 1.1617, + "step": 29384 + }, + { + "epoch": 2.47, + "learning_rate": 1.5688406314327231e-06, + "loss": 1.0902, + "step": 29385 + }, + { + "epoch": 2.47, + "learning_rate": 1.5683518825259103e-06, + "loss": 1.1352, + "step": 29386 + }, + { + "epoch": 2.47, + "learning_rate": 1.5678632032840757e-06, + "loss": 1.0917, + "step": 29387 + }, + { + "epoch": 2.47, + "learning_rate": 1.5673745937112528e-06, + "loss": 1.0644, + "step": 29388 + }, + { + "epoch": 2.47, + "learning_rate": 1.5668860538114795e-06, + "loss": 1.1785, + "step": 29389 + }, + { + "epoch": 2.47, + "learning_rate": 1.5663975835887923e-06, + "loss": 1.116, + "step": 29390 + }, + { + "epoch": 2.47, + "learning_rate": 1.5659091830472272e-06, + "loss": 1.1314, + "step": 29391 + }, + { + "epoch": 2.47, + "learning_rate": 1.565420852190821e-06, + "loss": 1.1717, + "step": 29392 + }, + { + "epoch": 2.47, + "learning_rate": 1.5649325910236079e-06, + "loss": 1.0709, + "step": 29393 + }, + { + "epoch": 2.47, + "learning_rate": 1.5644443995496173e-06, + "loss": 1.0774, + "step": 29394 + }, + { + "epoch": 2.47, + "learning_rate": 1.5639562777728912e-06, + "loss": 1.1373, + "step": 29395 + }, + { + "epoch": 2.47, + "learning_rate": 1.5634682256974564e-06, + "loss": 1.0864, + "step": 29396 + }, + { + "epoch": 2.47, + "learning_rate": 1.5629802433273445e-06, + "loss": 1.0482, + "step": 29397 + }, + { + "epoch": 2.48, + "learning_rate": 1.562492330666594e-06, + "loss": 1.1227, + "step": 29398 + }, + { + "epoch": 2.48, + "learning_rate": 1.5620044877192308e-06, + "loss": 1.134, + "step": 29399 + }, + { + "epoch": 2.48, + "learning_rate": 1.5615167144892874e-06, + "loss": 1.1172, + "step": 29400 + }, + { + "epoch": 2.48, + "learning_rate": 1.5610290109807923e-06, + "loss": 1.0304, + "step": 29401 + }, + { + "epoch": 2.48, + "learning_rate": 1.5605413771977773e-06, + "loss": 1.0494, + "step": 29402 + }, + { + "epoch": 2.48, + "learning_rate": 1.56005381314427e-06, + "loss": 1.1776, + "step": 29403 + }, + { + "epoch": 2.48, + "learning_rate": 1.5595663188243016e-06, + "loss": 1.1139, + "step": 29404 + }, + { + "epoch": 2.48, + "learning_rate": 1.5590788942418933e-06, + "loss": 1.1491, + "step": 29405 + }, + { + "epoch": 2.48, + "learning_rate": 1.5585915394010808e-06, + "loss": 1.102, + "step": 29406 + }, + { + "epoch": 2.48, + "learning_rate": 1.5581042543058845e-06, + "loss": 1.117, + "step": 29407 + }, + { + "epoch": 2.48, + "learning_rate": 1.5576170389603318e-06, + "loss": 1.0535, + "step": 29408 + }, + { + "epoch": 2.48, + "learning_rate": 1.5571298933684498e-06, + "loss": 1.0851, + "step": 29409 + }, + { + "epoch": 2.48, + "learning_rate": 1.556642817534263e-06, + "loss": 1.1046, + "step": 29410 + }, + { + "epoch": 2.48, + "learning_rate": 1.5561558114617947e-06, + "loss": 1.1567, + "step": 29411 + }, + { + "epoch": 2.48, + "learning_rate": 1.5556688751550698e-06, + "loss": 1.1847, + "step": 29412 + }, + { + "epoch": 2.48, + "learning_rate": 1.5551820086181114e-06, + "loss": 1.1324, + "step": 29413 + }, + { + "epoch": 2.48, + "learning_rate": 1.554695211854942e-06, + "loss": 1.0847, + "step": 29414 + }, + { + "epoch": 2.48, + "learning_rate": 1.554208484869586e-06, + "loss": 1.0734, + "step": 29415 + }, + { + "epoch": 2.48, + "learning_rate": 1.553721827666057e-06, + "loss": 1.1533, + "step": 29416 + }, + { + "epoch": 2.48, + "learning_rate": 1.5532352402483864e-06, + "loss": 1.0684, + "step": 29417 + }, + { + "epoch": 2.48, + "learning_rate": 1.5527487226205873e-06, + "loss": 1.1352, + "step": 29418 + }, + { + "epoch": 2.48, + "learning_rate": 1.5522622747866822e-06, + "loss": 1.1429, + "step": 29419 + }, + { + "epoch": 2.48, + "learning_rate": 1.551775896750689e-06, + "loss": 1.1762, + "step": 29420 + }, + { + "epoch": 2.48, + "learning_rate": 1.5512895885166278e-06, + "loss": 1.0998, + "step": 29421 + }, + { + "epoch": 2.48, + "learning_rate": 1.5508033500885155e-06, + "loss": 1.1542, + "step": 29422 + }, + { + "epoch": 2.48, + "learning_rate": 1.5503171814703711e-06, + "loss": 1.1256, + "step": 29423 + }, + { + "epoch": 2.48, + "learning_rate": 1.5498310826662112e-06, + "loss": 1.0531, + "step": 29424 + }, + { + "epoch": 2.48, + "learning_rate": 1.549345053680048e-06, + "loss": 1.1384, + "step": 29425 + }, + { + "epoch": 2.48, + "learning_rate": 1.5488590945159044e-06, + "loss": 1.1297, + "step": 29426 + }, + { + "epoch": 2.48, + "learning_rate": 1.5483732051777888e-06, + "loss": 1.1651, + "step": 29427 + }, + { + "epoch": 2.48, + "learning_rate": 1.5478873856697196e-06, + "loss": 1.1534, + "step": 29428 + }, + { + "epoch": 2.48, + "learning_rate": 1.547401635995709e-06, + "loss": 1.1533, + "step": 29429 + }, + { + "epoch": 2.48, + "learning_rate": 1.5469159561597725e-06, + "loss": 1.0881, + "step": 29430 + }, + { + "epoch": 2.48, + "learning_rate": 1.5464303461659203e-06, + "loss": 1.1027, + "step": 29431 + }, + { + "epoch": 2.48, + "learning_rate": 1.5459448060181693e-06, + "loss": 1.135, + "step": 29432 + }, + { + "epoch": 2.48, + "learning_rate": 1.5454593357205227e-06, + "loss": 1.0844, + "step": 29433 + }, + { + "epoch": 2.48, + "learning_rate": 1.544973935277e-06, + "loss": 1.0723, + "step": 29434 + }, + { + "epoch": 2.48, + "learning_rate": 1.5444886046916108e-06, + "loss": 1.0332, + "step": 29435 + }, + { + "epoch": 2.48, + "learning_rate": 1.5440033439683576e-06, + "loss": 1.1605, + "step": 29436 + }, + { + "epoch": 2.48, + "learning_rate": 1.5435181531112597e-06, + "loss": 1.0574, + "step": 29437 + }, + { + "epoch": 2.48, + "learning_rate": 1.5430330321243193e-06, + "loss": 0.9843, + "step": 29438 + }, + { + "epoch": 2.48, + "learning_rate": 1.5425479810115463e-06, + "loss": 1.1332, + "step": 29439 + }, + { + "epoch": 2.48, + "learning_rate": 1.5420629997769488e-06, + "loss": 1.0155, + "step": 29440 + }, + { + "epoch": 2.48, + "learning_rate": 1.5415780884245345e-06, + "loss": 1.1909, + "step": 29441 + }, + { + "epoch": 2.48, + "learning_rate": 1.5410932469583073e-06, + "loss": 1.0428, + "step": 29442 + }, + { + "epoch": 2.48, + "learning_rate": 1.5406084753822792e-06, + "loss": 1.1325, + "step": 29443 + }, + { + "epoch": 2.48, + "learning_rate": 1.5401237737004448e-06, + "loss": 1.2005, + "step": 29444 + }, + { + "epoch": 2.48, + "learning_rate": 1.539639141916821e-06, + "loss": 1.2239, + "step": 29445 + }, + { + "epoch": 2.48, + "learning_rate": 1.5391545800354034e-06, + "loss": 1.1267, + "step": 29446 + }, + { + "epoch": 2.48, + "learning_rate": 1.5386700880601968e-06, + "loss": 1.0729, + "step": 29447 + }, + { + "epoch": 2.48, + "learning_rate": 1.5381856659952099e-06, + "loss": 1.2073, + "step": 29448 + }, + { + "epoch": 2.48, + "learning_rate": 1.5377013138444386e-06, + "loss": 1.1138, + "step": 29449 + }, + { + "epoch": 2.48, + "learning_rate": 1.5372170316118885e-06, + "loss": 1.0977, + "step": 29450 + }, + { + "epoch": 2.48, + "learning_rate": 1.5367328193015585e-06, + "loss": 1.0917, + "step": 29451 + }, + { + "epoch": 2.48, + "learning_rate": 1.5362486769174512e-06, + "loss": 1.0657, + "step": 29452 + }, + { + "epoch": 2.48, + "learning_rate": 1.5357646044635654e-06, + "loss": 1.0737, + "step": 29453 + }, + { + "epoch": 2.48, + "learning_rate": 1.535280601943905e-06, + "loss": 1.1161, + "step": 29454 + }, + { + "epoch": 2.48, + "learning_rate": 1.5347966693624594e-06, + "loss": 1.0991, + "step": 29455 + }, + { + "epoch": 2.48, + "learning_rate": 1.5343128067232382e-06, + "loss": 1.1081, + "step": 29456 + }, + { + "epoch": 2.48, + "learning_rate": 1.5338290140302315e-06, + "loss": 1.1195, + "step": 29457 + }, + { + "epoch": 2.48, + "learning_rate": 1.5333452912874392e-06, + "loss": 1.0399, + "step": 29458 + }, + { + "epoch": 2.48, + "learning_rate": 1.532861638498857e-06, + "loss": 1.1294, + "step": 29459 + }, + { + "epoch": 2.48, + "learning_rate": 1.5323780556684831e-06, + "loss": 1.0724, + "step": 29460 + }, + { + "epoch": 2.48, + "learning_rate": 1.5318945428003106e-06, + "loss": 1.1557, + "step": 29461 + }, + { + "epoch": 2.48, + "learning_rate": 1.5314110998983356e-06, + "loss": 1.0598, + "step": 29462 + }, + { + "epoch": 2.48, + "learning_rate": 1.5309277269665523e-06, + "loss": 1.1126, + "step": 29463 + }, + { + "epoch": 2.48, + "learning_rate": 1.5304444240089555e-06, + "loss": 1.0786, + "step": 29464 + }, + { + "epoch": 2.48, + "learning_rate": 1.5299611910295387e-06, + "loss": 1.1164, + "step": 29465 + }, + { + "epoch": 2.48, + "learning_rate": 1.5294780280322896e-06, + "loss": 1.0879, + "step": 29466 + }, + { + "epoch": 2.48, + "learning_rate": 1.5289949350212085e-06, + "loss": 1.1342, + "step": 29467 + }, + { + "epoch": 2.48, + "learning_rate": 1.528511912000279e-06, + "loss": 1.0298, + "step": 29468 + }, + { + "epoch": 2.48, + "learning_rate": 1.5280289589734965e-06, + "loss": 1.1055, + "step": 29469 + }, + { + "epoch": 2.48, + "learning_rate": 1.527546075944849e-06, + "loss": 1.1432, + "step": 29470 + }, + { + "epoch": 2.48, + "learning_rate": 1.527063262918328e-06, + "loss": 1.1756, + "step": 29471 + }, + { + "epoch": 2.48, + "learning_rate": 1.5265805198979221e-06, + "loss": 1.0623, + "step": 29472 + }, + { + "epoch": 2.48, + "learning_rate": 1.5260978468876197e-06, + "loss": 1.0403, + "step": 29473 + }, + { + "epoch": 2.48, + "learning_rate": 1.5256152438914107e-06, + "loss": 1.0639, + "step": 29474 + }, + { + "epoch": 2.48, + "learning_rate": 1.5251327109132774e-06, + "loss": 1.1127, + "step": 29475 + }, + { + "epoch": 2.48, + "learning_rate": 1.5246502479572133e-06, + "loss": 1.0939, + "step": 29476 + }, + { + "epoch": 2.48, + "learning_rate": 1.5241678550271988e-06, + "loss": 1.1665, + "step": 29477 + }, + { + "epoch": 2.48, + "learning_rate": 1.5236855321272226e-06, + "loss": 1.143, + "step": 29478 + }, + { + "epoch": 2.48, + "learning_rate": 1.5232032792612694e-06, + "loss": 1.1852, + "step": 29479 + }, + { + "epoch": 2.48, + "learning_rate": 1.522721096433324e-06, + "loss": 1.1149, + "step": 29480 + }, + { + "epoch": 2.48, + "learning_rate": 1.5222389836473694e-06, + "loss": 1.1485, + "step": 29481 + }, + { + "epoch": 2.48, + "learning_rate": 1.5217569409073917e-06, + "loss": 1.0416, + "step": 29482 + }, + { + "epoch": 2.48, + "learning_rate": 1.5212749682173677e-06, + "loss": 1.0682, + "step": 29483 + }, + { + "epoch": 2.48, + "learning_rate": 1.5207930655812853e-06, + "loss": 1.1332, + "step": 29484 + }, + { + "epoch": 2.48, + "learning_rate": 1.5203112330031267e-06, + "loss": 1.0798, + "step": 29485 + }, + { + "epoch": 2.48, + "learning_rate": 1.5198294704868654e-06, + "loss": 1.1298, + "step": 29486 + }, + { + "epoch": 2.48, + "learning_rate": 1.5193477780364928e-06, + "loss": 1.0651, + "step": 29487 + }, + { + "epoch": 2.48, + "learning_rate": 1.5188661556559803e-06, + "loss": 1.0771, + "step": 29488 + }, + { + "epoch": 2.48, + "learning_rate": 1.5183846033493089e-06, + "loss": 1.1008, + "step": 29489 + }, + { + "epoch": 2.48, + "learning_rate": 1.5179031211204598e-06, + "loss": 1.1206, + "step": 29490 + }, + { + "epoch": 2.48, + "learning_rate": 1.5174217089734088e-06, + "loss": 1.1261, + "step": 29491 + }, + { + "epoch": 2.48, + "learning_rate": 1.5169403669121341e-06, + "loss": 1.0742, + "step": 29492 + }, + { + "epoch": 2.48, + "learning_rate": 1.5164590949406166e-06, + "loss": 1.0957, + "step": 29493 + }, + { + "epoch": 2.48, + "learning_rate": 1.5159778930628233e-06, + "loss": 1.0791, + "step": 29494 + }, + { + "epoch": 2.48, + "learning_rate": 1.515496761282741e-06, + "loss": 1.1399, + "step": 29495 + }, + { + "epoch": 2.48, + "learning_rate": 1.5150156996043375e-06, + "loss": 1.1429, + "step": 29496 + }, + { + "epoch": 2.48, + "learning_rate": 1.5145347080315875e-06, + "loss": 1.2041, + "step": 29497 + }, + { + "epoch": 2.48, + "learning_rate": 1.5140537865684723e-06, + "loss": 1.2402, + "step": 29498 + }, + { + "epoch": 2.48, + "learning_rate": 1.5135729352189576e-06, + "loss": 1.2106, + "step": 29499 + }, + { + "epoch": 2.48, + "learning_rate": 1.5130921539870202e-06, + "loss": 0.9863, + "step": 29500 + }, + { + "epoch": 2.48, + "learning_rate": 1.5126114428766314e-06, + "loss": 1.1439, + "step": 29501 + }, + { + "epoch": 2.48, + "learning_rate": 1.5121308018917623e-06, + "loss": 1.0705, + "step": 29502 + }, + { + "epoch": 2.48, + "learning_rate": 1.5116502310363856e-06, + "loss": 1.0231, + "step": 29503 + }, + { + "epoch": 2.48, + "learning_rate": 1.5111697303144734e-06, + "loss": 1.1396, + "step": 29504 + }, + { + "epoch": 2.48, + "learning_rate": 1.5106892997299893e-06, + "loss": 1.158, + "step": 29505 + }, + { + "epoch": 2.48, + "learning_rate": 1.5102089392869124e-06, + "loss": 1.146, + "step": 29506 + }, + { + "epoch": 2.48, + "learning_rate": 1.5097286489892026e-06, + "loss": 1.144, + "step": 29507 + }, + { + "epoch": 2.48, + "learning_rate": 1.5092484288408327e-06, + "loss": 1.2657, + "step": 29508 + }, + { + "epoch": 2.48, + "learning_rate": 1.5087682788457702e-06, + "loss": 1.0964, + "step": 29509 + }, + { + "epoch": 2.48, + "learning_rate": 1.5082881990079813e-06, + "loss": 1.2247, + "step": 29510 + }, + { + "epoch": 2.48, + "learning_rate": 1.5078081893314323e-06, + "loss": 1.0633, + "step": 29511 + }, + { + "epoch": 2.48, + "learning_rate": 1.5073282498200904e-06, + "loss": 1.043, + "step": 29512 + }, + { + "epoch": 2.48, + "learning_rate": 1.506848380477921e-06, + "loss": 1.0896, + "step": 29513 + }, + { + "epoch": 2.48, + "learning_rate": 1.5063685813088879e-06, + "loss": 0.9547, + "step": 29514 + }, + { + "epoch": 2.48, + "learning_rate": 1.5058888523169579e-06, + "loss": 1.02, + "step": 29515 + }, + { + "epoch": 2.48, + "learning_rate": 1.5054091935060889e-06, + "loss": 1.079, + "step": 29516 + }, + { + "epoch": 2.49, + "learning_rate": 1.504929604880252e-06, + "loss": 1.1292, + "step": 29517 + }, + { + "epoch": 2.49, + "learning_rate": 1.504450086443403e-06, + "loss": 1.0776, + "step": 29518 + }, + { + "epoch": 2.49, + "learning_rate": 1.5039706381995056e-06, + "loss": 1.1442, + "step": 29519 + }, + { + "epoch": 2.49, + "learning_rate": 1.503491260152523e-06, + "loss": 1.1884, + "step": 29520 + }, + { + "epoch": 2.49, + "learning_rate": 1.5030119523064146e-06, + "loss": 1.0293, + "step": 29521 + }, + { + "epoch": 2.49, + "learning_rate": 1.5025327146651403e-06, + "loss": 1.1971, + "step": 29522 + }, + { + "epoch": 2.49, + "learning_rate": 1.5020535472326615e-06, + "loss": 1.1363, + "step": 29523 + }, + { + "epoch": 2.49, + "learning_rate": 1.5015744500129348e-06, + "loss": 1.2295, + "step": 29524 + }, + { + "epoch": 2.49, + "learning_rate": 1.5010954230099206e-06, + "loss": 1.0569, + "step": 29525 + }, + { + "epoch": 2.49, + "learning_rate": 1.500616466227578e-06, + "loss": 1.1251, + "step": 29526 + }, + { + "epoch": 2.49, + "learning_rate": 1.500137579669857e-06, + "loss": 1.0739, + "step": 29527 + }, + { + "epoch": 2.49, + "learning_rate": 1.4996587633407245e-06, + "loss": 1.1164, + "step": 29528 + }, + { + "epoch": 2.49, + "learning_rate": 1.4991800172441306e-06, + "loss": 1.0825, + "step": 29529 + }, + { + "epoch": 2.49, + "learning_rate": 1.4987013413840312e-06, + "loss": 1.0845, + "step": 29530 + }, + { + "epoch": 2.49, + "learning_rate": 1.498222735764382e-06, + "loss": 1.1248, + "step": 29531 + }, + { + "epoch": 2.49, + "learning_rate": 1.4977442003891395e-06, + "loss": 1.1039, + "step": 29532 + }, + { + "epoch": 2.49, + "learning_rate": 1.497265735262251e-06, + "loss": 1.0957, + "step": 29533 + }, + { + "epoch": 2.49, + "learning_rate": 1.4967873403876775e-06, + "loss": 1.0904, + "step": 29534 + }, + { + "epoch": 2.49, + "learning_rate": 1.4963090157693693e-06, + "loss": 1.1411, + "step": 29535 + }, + { + "epoch": 2.49, + "learning_rate": 1.4958307614112733e-06, + "loss": 1.0983, + "step": 29536 + }, + { + "epoch": 2.49, + "learning_rate": 1.4953525773173506e-06, + "loss": 1.1263, + "step": 29537 + }, + { + "epoch": 2.49, + "learning_rate": 1.4948744634915435e-06, + "loss": 1.0705, + "step": 29538 + }, + { + "epoch": 2.49, + "learning_rate": 1.494396419937807e-06, + "loss": 1.1105, + "step": 29539 + }, + { + "epoch": 2.49, + "learning_rate": 1.4939184466600875e-06, + "loss": 1.1188, + "step": 29540 + }, + { + "epoch": 2.49, + "learning_rate": 1.4934405436623378e-06, + "loss": 1.1418, + "step": 29541 + }, + { + "epoch": 2.49, + "learning_rate": 1.4929627109485035e-06, + "loss": 1.1286, + "step": 29542 + }, + { + "epoch": 2.49, + "learning_rate": 1.4924849485225369e-06, + "loss": 1.0189, + "step": 29543 + }, + { + "epoch": 2.49, + "learning_rate": 1.492007256388378e-06, + "loss": 1.1212, + "step": 29544 + }, + { + "epoch": 2.49, + "learning_rate": 1.4915296345499818e-06, + "loss": 1.0748, + "step": 29545 + }, + { + "epoch": 2.49, + "learning_rate": 1.4910520830112886e-06, + "loss": 1.1129, + "step": 29546 + }, + { + "epoch": 2.49, + "learning_rate": 1.4905746017762446e-06, + "loss": 1.0371, + "step": 29547 + }, + { + "epoch": 2.49, + "learning_rate": 1.4900971908488015e-06, + "loss": 1.134, + "step": 29548 + }, + { + "epoch": 2.49, + "learning_rate": 1.4896198502328963e-06, + "loss": 1.1197, + "step": 29549 + }, + { + "epoch": 2.49, + "learning_rate": 1.4891425799324743e-06, + "loss": 1.1072, + "step": 29550 + }, + { + "epoch": 2.49, + "learning_rate": 1.4886653799514816e-06, + "loss": 1.1606, + "step": 29551 + }, + { + "epoch": 2.49, + "learning_rate": 1.4881882502938594e-06, + "loss": 1.0984, + "step": 29552 + }, + { + "epoch": 2.49, + "learning_rate": 1.48771119096355e-06, + "loss": 1.0167, + "step": 29553 + }, + { + "epoch": 2.49, + "learning_rate": 1.487234201964497e-06, + "loss": 1.0988, + "step": 29554 + }, + { + "epoch": 2.49, + "learning_rate": 1.486757283300635e-06, + "loss": 1.0919, + "step": 29555 + }, + { + "epoch": 2.49, + "learning_rate": 1.486280434975913e-06, + "loss": 1.0506, + "step": 29556 + }, + { + "epoch": 2.49, + "learning_rate": 1.4858036569942646e-06, + "loss": 1.0869, + "step": 29557 + }, + { + "epoch": 2.49, + "learning_rate": 1.4853269493596312e-06, + "loss": 1.0803, + "step": 29558 + }, + { + "epoch": 2.49, + "learning_rate": 1.4848503120759527e-06, + "loss": 1.1053, + "step": 29559 + }, + { + "epoch": 2.49, + "learning_rate": 1.484373745147165e-06, + "loss": 1.1454, + "step": 29560 + }, + { + "epoch": 2.49, + "learning_rate": 1.4838972485772075e-06, + "loss": 1.0659, + "step": 29561 + }, + { + "epoch": 2.49, + "learning_rate": 1.4834208223700164e-06, + "loss": 1.193, + "step": 29562 + }, + { + "epoch": 2.49, + "learning_rate": 1.4829444665295266e-06, + "loss": 1.0638, + "step": 29563 + }, + { + "epoch": 2.49, + "learning_rate": 1.4824681810596776e-06, + "loss": 1.1642, + "step": 29564 + }, + { + "epoch": 2.49, + "learning_rate": 1.4819919659644033e-06, + "loss": 1.0717, + "step": 29565 + }, + { + "epoch": 2.49, + "learning_rate": 1.4815158212476333e-06, + "loss": 1.0866, + "step": 29566 + }, + { + "epoch": 2.49, + "learning_rate": 1.4810397469133108e-06, + "loss": 1.0909, + "step": 29567 + }, + { + "epoch": 2.49, + "learning_rate": 1.4805637429653618e-06, + "loss": 1.1288, + "step": 29568 + }, + { + "epoch": 2.49, + "learning_rate": 1.480087809407722e-06, + "loss": 1.0669, + "step": 29569 + }, + { + "epoch": 2.49, + "learning_rate": 1.4796119462443225e-06, + "loss": 1.1039, + "step": 29570 + }, + { + "epoch": 2.49, + "learning_rate": 1.4791361534790972e-06, + "loss": 1.2292, + "step": 29571 + }, + { + "epoch": 2.49, + "learning_rate": 1.4786604311159748e-06, + "loss": 1.1558, + "step": 29572 + }, + { + "epoch": 2.49, + "learning_rate": 1.4781847791588878e-06, + "loss": 1.0424, + "step": 29573 + }, + { + "epoch": 2.49, + "learning_rate": 1.4777091976117652e-06, + "loss": 1.1418, + "step": 29574 + }, + { + "epoch": 2.49, + "learning_rate": 1.4772336864785374e-06, + "loss": 1.0095, + "step": 29575 + }, + { + "epoch": 2.49, + "learning_rate": 1.4767582457631335e-06, + "loss": 1.0019, + "step": 29576 + }, + { + "epoch": 2.49, + "learning_rate": 1.4762828754694769e-06, + "loss": 1.1031, + "step": 29577 + }, + { + "epoch": 2.49, + "learning_rate": 1.4758075756015045e-06, + "loss": 1.1555, + "step": 29578 + }, + { + "epoch": 2.49, + "learning_rate": 1.4753323461631353e-06, + "loss": 1.0451, + "step": 29579 + }, + { + "epoch": 2.49, + "learning_rate": 1.4748571871582972e-06, + "loss": 1.0056, + "step": 29580 + }, + { + "epoch": 2.49, + "learning_rate": 1.4743820985909184e-06, + "loss": 1.1406, + "step": 29581 + }, + { + "epoch": 2.49, + "learning_rate": 1.4739070804649258e-06, + "loss": 1.0907, + "step": 29582 + }, + { + "epoch": 2.49, + "learning_rate": 1.4734321327842371e-06, + "loss": 1.1439, + "step": 29583 + }, + { + "epoch": 2.49, + "learning_rate": 1.4729572555527828e-06, + "loss": 1.1586, + "step": 29584 + }, + { + "epoch": 2.49, + "learning_rate": 1.472482448774487e-06, + "loss": 1.1693, + "step": 29585 + }, + { + "epoch": 2.49, + "learning_rate": 1.472007712453266e-06, + "loss": 1.0994, + "step": 29586 + }, + { + "epoch": 2.49, + "learning_rate": 1.471533046593051e-06, + "loss": 1.1689, + "step": 29587 + }, + { + "epoch": 2.49, + "learning_rate": 1.4710584511977565e-06, + "loss": 1.1823, + "step": 29588 + }, + { + "epoch": 2.49, + "learning_rate": 1.4705839262713073e-06, + "loss": 1.1448, + "step": 29589 + }, + { + "epoch": 2.49, + "learning_rate": 1.4701094718176234e-06, + "loss": 1.1387, + "step": 29590 + }, + { + "epoch": 2.49, + "learning_rate": 1.4696350878406251e-06, + "loss": 1.1462, + "step": 29591 + }, + { + "epoch": 2.49, + "learning_rate": 1.4691607743442326e-06, + "loss": 1.028, + "step": 29592 + }, + { + "epoch": 2.49, + "learning_rate": 1.4686865313323651e-06, + "loss": 1.1171, + "step": 29593 + }, + { + "epoch": 2.49, + "learning_rate": 1.4682123588089359e-06, + "loss": 1.1697, + "step": 29594 + }, + { + "epoch": 2.49, + "learning_rate": 1.4677382567778708e-06, + "loss": 1.1216, + "step": 29595 + }, + { + "epoch": 2.49, + "learning_rate": 1.4672642252430803e-06, + "loss": 1.1794, + "step": 29596 + }, + { + "epoch": 2.49, + "learning_rate": 1.4667902642084818e-06, + "loss": 1.152, + "step": 29597 + }, + { + "epoch": 2.49, + "learning_rate": 1.4663163736779985e-06, + "loss": 1.0665, + "step": 29598 + }, + { + "epoch": 2.49, + "learning_rate": 1.4658425536555365e-06, + "loss": 1.1312, + "step": 29599 + }, + { + "epoch": 2.49, + "learning_rate": 1.4653688041450153e-06, + "loss": 1.0882, + "step": 29600 + }, + { + "epoch": 2.49, + "learning_rate": 1.4648951251503485e-06, + "loss": 1.1101, + "step": 29601 + }, + { + "epoch": 2.49, + "learning_rate": 1.4644215166754506e-06, + "loss": 1.1504, + "step": 29602 + }, + { + "epoch": 2.49, + "learning_rate": 1.4639479787242327e-06, + "loss": 1.0639, + "step": 29603 + }, + { + "epoch": 2.49, + "learning_rate": 1.463474511300611e-06, + "loss": 1.054, + "step": 29604 + }, + { + "epoch": 2.49, + "learning_rate": 1.463001114408491e-06, + "loss": 1.0417, + "step": 29605 + }, + { + "epoch": 2.49, + "learning_rate": 1.4625277880517918e-06, + "loss": 1.099, + "step": 29606 + }, + { + "epoch": 2.49, + "learning_rate": 1.4620545322344182e-06, + "loss": 1.0185, + "step": 29607 + }, + { + "epoch": 2.49, + "learning_rate": 1.4615813469602836e-06, + "loss": 1.1868, + "step": 29608 + }, + { + "epoch": 2.49, + "learning_rate": 1.4611082322332958e-06, + "loss": 1.1694, + "step": 29609 + }, + { + "epoch": 2.49, + "learning_rate": 1.4606351880573645e-06, + "loss": 1.121, + "step": 29610 + }, + { + "epoch": 2.49, + "learning_rate": 1.460162214436398e-06, + "loss": 1.1193, + "step": 29611 + }, + { + "epoch": 2.49, + "learning_rate": 1.4596893113743037e-06, + "loss": 1.1218, + "step": 29612 + }, + { + "epoch": 2.49, + "learning_rate": 1.4592164788749907e-06, + "loss": 1.0614, + "step": 29613 + }, + { + "epoch": 2.49, + "learning_rate": 1.4587437169423647e-06, + "loss": 1.0451, + "step": 29614 + }, + { + "epoch": 2.49, + "learning_rate": 1.4582710255803333e-06, + "loss": 1.1778, + "step": 29615 + }, + { + "epoch": 2.49, + "learning_rate": 1.4577984047927961e-06, + "loss": 1.0449, + "step": 29616 + }, + { + "epoch": 2.49, + "learning_rate": 1.4573258545836667e-06, + "loss": 1.1332, + "step": 29617 + }, + { + "epoch": 2.49, + "learning_rate": 1.4568533749568436e-06, + "loss": 1.1321, + "step": 29618 + }, + { + "epoch": 2.49, + "learning_rate": 1.456380965916231e-06, + "loss": 1.0712, + "step": 29619 + }, + { + "epoch": 2.49, + "learning_rate": 1.455908627465734e-06, + "loss": 1.0783, + "step": 29620 + }, + { + "epoch": 2.49, + "learning_rate": 1.4554363596092536e-06, + "loss": 1.1908, + "step": 29621 + }, + { + "epoch": 2.49, + "learning_rate": 1.4549641623506927e-06, + "loss": 1.0995, + "step": 29622 + }, + { + "epoch": 2.49, + "learning_rate": 1.454492035693953e-06, + "loss": 1.1363, + "step": 29623 + }, + { + "epoch": 2.49, + "learning_rate": 1.4540199796429354e-06, + "loss": 1.1146, + "step": 29624 + }, + { + "epoch": 2.49, + "learning_rate": 1.4535479942015396e-06, + "loss": 1.1571, + "step": 29625 + }, + { + "epoch": 2.49, + "learning_rate": 1.4530760793736676e-06, + "loss": 1.1173, + "step": 29626 + }, + { + "epoch": 2.49, + "learning_rate": 1.452604235163212e-06, + "loss": 1.0258, + "step": 29627 + }, + { + "epoch": 2.49, + "learning_rate": 1.4521324615740795e-06, + "loss": 1.0992, + "step": 29628 + }, + { + "epoch": 2.49, + "learning_rate": 1.4516607586101627e-06, + "loss": 1.027, + "step": 29629 + }, + { + "epoch": 2.49, + "learning_rate": 1.4511891262753597e-06, + "loss": 1.0158, + "step": 29630 + }, + { + "epoch": 2.49, + "learning_rate": 1.4507175645735693e-06, + "loss": 1.1114, + "step": 29631 + }, + { + "epoch": 2.49, + "learning_rate": 1.4502460735086887e-06, + "loss": 1.227, + "step": 29632 + }, + { + "epoch": 2.49, + "learning_rate": 1.4497746530846058e-06, + "loss": 1.1679, + "step": 29633 + }, + { + "epoch": 2.49, + "learning_rate": 1.449303303305224e-06, + "loss": 1.0688, + "step": 29634 + }, + { + "epoch": 2.5, + "learning_rate": 1.4488320241744368e-06, + "loss": 1.0101, + "step": 29635 + }, + { + "epoch": 2.5, + "learning_rate": 1.4483608156961326e-06, + "loss": 1.1586, + "step": 29636 + }, + { + "epoch": 2.5, + "learning_rate": 1.4478896778742113e-06, + "loss": 1.0955, + "step": 29637 + }, + { + "epoch": 2.5, + "learning_rate": 1.447418610712561e-06, + "loss": 1.1988, + "step": 29638 + }, + { + "epoch": 2.5, + "learning_rate": 1.4469476142150752e-06, + "loss": 1.1489, + "step": 29639 + }, + { + "epoch": 2.5, + "learning_rate": 1.4464766883856462e-06, + "loss": 1.1082, + "step": 29640 + }, + { + "epoch": 2.5, + "learning_rate": 1.4460058332281635e-06, + "loss": 1.122, + "step": 29641 + }, + { + "epoch": 2.5, + "learning_rate": 1.4455350487465181e-06, + "loss": 1.1673, + "step": 29642 + }, + { + "epoch": 2.5, + "learning_rate": 1.4450643349446026e-06, + "loss": 1.156, + "step": 29643 + }, + { + "epoch": 2.5, + "learning_rate": 1.4445936918262993e-06, + "loss": 1.1681, + "step": 29644 + }, + { + "epoch": 2.5, + "learning_rate": 1.4441231193955053e-06, + "loss": 1.0977, + "step": 29645 + }, + { + "epoch": 2.5, + "learning_rate": 1.4436526176561016e-06, + "loss": 1.1184, + "step": 29646 + }, + { + "epoch": 2.5, + "learning_rate": 1.4431821866119766e-06, + "loss": 1.1358, + "step": 29647 + }, + { + "epoch": 2.5, + "learning_rate": 1.4427118262670226e-06, + "loss": 1.1037, + "step": 29648 + }, + { + "epoch": 2.5, + "learning_rate": 1.4422415366251207e-06, + "loss": 1.1223, + "step": 29649 + }, + { + "epoch": 2.5, + "learning_rate": 1.4417713176901581e-06, + "loss": 1.0847, + "step": 29650 + }, + { + "epoch": 2.5, + "learning_rate": 1.4413011694660195e-06, + "loss": 1.1537, + "step": 29651 + }, + { + "epoch": 2.5, + "learning_rate": 1.4408310919565904e-06, + "loss": 1.2249, + "step": 29652 + }, + { + "epoch": 2.5, + "learning_rate": 1.4403610851657535e-06, + "loss": 1.0177, + "step": 29653 + }, + { + "epoch": 2.5, + "learning_rate": 1.4398911490973943e-06, + "loss": 1.117, + "step": 29654 + }, + { + "epoch": 2.5, + "learning_rate": 1.4394212837553911e-06, + "loss": 1.1455, + "step": 29655 + }, + { + "epoch": 2.5, + "learning_rate": 1.4389514891436329e-06, + "loss": 1.1802, + "step": 29656 + }, + { + "epoch": 2.5, + "learning_rate": 1.4384817652659943e-06, + "loss": 1.0536, + "step": 29657 + }, + { + "epoch": 2.5, + "learning_rate": 1.43801211212636e-06, + "loss": 1.2254, + "step": 29658 + }, + { + "epoch": 2.5, + "learning_rate": 1.4375425297286095e-06, + "loss": 1.1226, + "step": 29659 + }, + { + "epoch": 2.5, + "learning_rate": 1.4370730180766234e-06, + "loss": 1.0582, + "step": 29660 + }, + { + "epoch": 2.5, + "learning_rate": 1.4366035771742803e-06, + "loss": 1.0992, + "step": 29661 + }, + { + "epoch": 2.5, + "learning_rate": 1.436134207025459e-06, + "loss": 1.0413, + "step": 29662 + }, + { + "epoch": 2.5, + "learning_rate": 1.435664907634039e-06, + "loss": 1.1301, + "step": 29663 + }, + { + "epoch": 2.5, + "learning_rate": 1.4351956790038946e-06, + "loss": 1.0748, + "step": 29664 + }, + { + "epoch": 2.5, + "learning_rate": 1.4347265211389083e-06, + "loss": 1.1234, + "step": 29665 + }, + { + "epoch": 2.5, + "learning_rate": 1.434257434042947e-06, + "loss": 1.1243, + "step": 29666 + }, + { + "epoch": 2.5, + "learning_rate": 1.4337884177198969e-06, + "loss": 1.0537, + "step": 29667 + }, + { + "epoch": 2.5, + "learning_rate": 1.4333194721736265e-06, + "loss": 1.1155, + "step": 29668 + }, + { + "epoch": 2.5, + "learning_rate": 1.432850597408012e-06, + "loss": 1.0294, + "step": 29669 + }, + { + "epoch": 2.5, + "learning_rate": 1.4323817934269279e-06, + "loss": 1.1023, + "step": 29670 + }, + { + "epoch": 2.5, + "learning_rate": 1.431913060234248e-06, + "loss": 1.0743, + "step": 29671 + }, + { + "epoch": 2.5, + "learning_rate": 1.4314443978338444e-06, + "loss": 1.1492, + "step": 29672 + }, + { + "epoch": 2.5, + "learning_rate": 1.4309758062295885e-06, + "loss": 1.2397, + "step": 29673 + }, + { + "epoch": 2.5, + "learning_rate": 1.4305072854253532e-06, + "loss": 1.1216, + "step": 29674 + }, + { + "epoch": 2.5, + "learning_rate": 1.4300388354250104e-06, + "loss": 1.1745, + "step": 29675 + }, + { + "epoch": 2.5, + "learning_rate": 1.429570456232431e-06, + "loss": 1.1363, + "step": 29676 + }, + { + "epoch": 2.5, + "learning_rate": 1.4291021478514789e-06, + "loss": 1.1309, + "step": 29677 + }, + { + "epoch": 2.5, + "learning_rate": 1.4286339102860313e-06, + "loss": 0.9978, + "step": 29678 + }, + { + "epoch": 2.5, + "learning_rate": 1.428165743539953e-06, + "loss": 1.0934, + "step": 29679 + }, + { + "epoch": 2.5, + "learning_rate": 1.4276976476171124e-06, + "loss": 1.0878, + "step": 29680 + }, + { + "epoch": 2.5, + "learning_rate": 1.4272296225213767e-06, + "loss": 1.1772, + "step": 29681 + }, + { + "epoch": 2.5, + "learning_rate": 1.4267616682566155e-06, + "loss": 1.054, + "step": 29682 + }, + { + "epoch": 2.5, + "learning_rate": 1.4262937848266888e-06, + "loss": 1.1744, + "step": 29683 + }, + { + "epoch": 2.5, + "learning_rate": 1.4258259722354694e-06, + "loss": 1.0593, + "step": 29684 + }, + { + "epoch": 2.5, + "learning_rate": 1.4253582304868218e-06, + "loss": 1.0905, + "step": 29685 + }, + { + "epoch": 2.5, + "learning_rate": 1.4248905595846053e-06, + "loss": 1.0919, + "step": 29686 + }, + { + "epoch": 2.5, + "learning_rate": 1.4244229595326908e-06, + "loss": 1.1139, + "step": 29687 + }, + { + "epoch": 2.5, + "learning_rate": 1.4239554303349367e-06, + "loss": 1.1861, + "step": 29688 + }, + { + "epoch": 2.5, + "learning_rate": 1.4234879719952066e-06, + "loss": 1.0374, + "step": 29689 + }, + { + "epoch": 2.5, + "learning_rate": 1.423020584517365e-06, + "loss": 1.0719, + "step": 29690 + }, + { + "epoch": 2.5, + "learning_rate": 1.422553267905271e-06, + "loss": 1.0648, + "step": 29691 + }, + { + "epoch": 2.5, + "learning_rate": 1.4220860221627885e-06, + "loss": 1.1006, + "step": 29692 + }, + { + "epoch": 2.5, + "learning_rate": 1.4216188472937774e-06, + "loss": 1.1253, + "step": 29693 + }, + { + "epoch": 2.5, + "learning_rate": 1.4211517433020926e-06, + "loss": 1.0858, + "step": 29694 + }, + { + "epoch": 2.5, + "learning_rate": 1.4206847101916033e-06, + "loss": 1.187, + "step": 29695 + }, + { + "epoch": 2.5, + "learning_rate": 1.4202177479661593e-06, + "loss": 1.1217, + "step": 29696 + }, + { + "epoch": 2.5, + "learning_rate": 1.4197508566296203e-06, + "loss": 1.1339, + "step": 29697 + }, + { + "epoch": 2.5, + "learning_rate": 1.4192840361858507e-06, + "loss": 1.0021, + "step": 29698 + }, + { + "epoch": 2.5, + "learning_rate": 1.4188172866386996e-06, + "loss": 1.0852, + "step": 29699 + }, + { + "epoch": 2.5, + "learning_rate": 1.4183506079920263e-06, + "loss": 1.0544, + "step": 29700 + }, + { + "epoch": 2.5, + "learning_rate": 1.4178840002496875e-06, + "loss": 1.2154, + "step": 29701 + }, + { + "epoch": 2.5, + "learning_rate": 1.417417463415537e-06, + "loss": 1.1806, + "step": 29702 + }, + { + "epoch": 2.5, + "learning_rate": 1.4169509974934303e-06, + "loss": 1.0962, + "step": 29703 + }, + { + "epoch": 2.5, + "learning_rate": 1.4164846024872237e-06, + "loss": 1.0792, + "step": 29704 + }, + { + "epoch": 2.5, + "learning_rate": 1.4160182784007636e-06, + "loss": 1.0796, + "step": 29705 + }, + { + "epoch": 2.5, + "learning_rate": 1.415552025237913e-06, + "loss": 1.0267, + "step": 29706 + }, + { + "epoch": 2.5, + "learning_rate": 1.4150858430025161e-06, + "loss": 1.1515, + "step": 29707 + }, + { + "epoch": 2.5, + "learning_rate": 1.4146197316984279e-06, + "loss": 1.1626, + "step": 29708 + }, + { + "epoch": 2.5, + "learning_rate": 1.4141536913294996e-06, + "loss": 1.1858, + "step": 29709 + }, + { + "epoch": 2.5, + "learning_rate": 1.4136877218995803e-06, + "loss": 1.2192, + "step": 29710 + }, + { + "epoch": 2.5, + "learning_rate": 1.4132218234125217e-06, + "loss": 1.0887, + "step": 29711 + }, + { + "epoch": 2.5, + "learning_rate": 1.4127559958721716e-06, + "loss": 1.1687, + "step": 29712 + }, + { + "epoch": 2.5, + "learning_rate": 1.4122902392823801e-06, + "loss": 1.0529, + "step": 29713 + }, + { + "epoch": 2.5, + "learning_rate": 1.4118245536469955e-06, + "loss": 1.2277, + "step": 29714 + }, + { + "epoch": 2.5, + "learning_rate": 1.4113589389698668e-06, + "loss": 1.1366, + "step": 29715 + }, + { + "epoch": 2.5, + "learning_rate": 1.4108933952548354e-06, + "loss": 1.1814, + "step": 29716 + }, + { + "epoch": 2.5, + "learning_rate": 1.410427922505756e-06, + "loss": 1.1525, + "step": 29717 + }, + { + "epoch": 2.5, + "learning_rate": 1.4099625207264678e-06, + "loss": 1.0975, + "step": 29718 + }, + { + "epoch": 2.5, + "learning_rate": 1.4094971899208177e-06, + "loss": 1.1221, + "step": 29719 + }, + { + "epoch": 2.5, + "learning_rate": 1.4090319300926514e-06, + "loss": 1.1426, + "step": 29720 + }, + { + "epoch": 2.5, + "learning_rate": 1.4085667412458137e-06, + "loss": 1.1543, + "step": 29721 + }, + { + "epoch": 2.5, + "learning_rate": 1.408101623384146e-06, + "loss": 1.1091, + "step": 29722 + }, + { + "epoch": 2.5, + "learning_rate": 1.407636576511493e-06, + "loss": 1.1936, + "step": 29723 + }, + { + "epoch": 2.5, + "learning_rate": 1.407171600631696e-06, + "loss": 1.1385, + "step": 29724 + }, + { + "epoch": 2.5, + "learning_rate": 1.4067066957485987e-06, + "loss": 1.1622, + "step": 29725 + }, + { + "epoch": 2.5, + "learning_rate": 1.4062418618660424e-06, + "loss": 1.0625, + "step": 29726 + }, + { + "epoch": 2.5, + "learning_rate": 1.4057770989878605e-06, + "loss": 1.1997, + "step": 29727 + }, + { + "epoch": 2.5, + "learning_rate": 1.4053124071179048e-06, + "loss": 1.0076, + "step": 29728 + }, + { + "epoch": 2.5, + "learning_rate": 1.4048477862600052e-06, + "loss": 1.0617, + "step": 29729 + }, + { + "epoch": 2.5, + "learning_rate": 1.4043832364180055e-06, + "loss": 1.1933, + "step": 29730 + }, + { + "epoch": 2.5, + "learning_rate": 1.4039187575957413e-06, + "loss": 1.1725, + "step": 29731 + }, + { + "epoch": 2.5, + "learning_rate": 1.4034543497970532e-06, + "loss": 1.1099, + "step": 29732 + }, + { + "epoch": 2.5, + "learning_rate": 1.4029900130257733e-06, + "loss": 1.0297, + "step": 29733 + }, + { + "epoch": 2.5, + "learning_rate": 1.4025257472857424e-06, + "loss": 1.0389, + "step": 29734 + }, + { + "epoch": 2.5, + "learning_rate": 1.402061552580799e-06, + "loss": 1.1446, + "step": 29735 + }, + { + "epoch": 2.5, + "learning_rate": 1.4015974289147694e-06, + "loss": 1.0776, + "step": 29736 + }, + { + "epoch": 2.5, + "learning_rate": 1.401133376291498e-06, + "loss": 1.1244, + "step": 29737 + }, + { + "epoch": 2.5, + "learning_rate": 1.400669394714812e-06, + "loss": 1.0634, + "step": 29738 + }, + { + "epoch": 2.5, + "learning_rate": 1.4002054841885482e-06, + "loss": 1.1757, + "step": 29739 + }, + { + "epoch": 2.5, + "learning_rate": 1.399741644716539e-06, + "loss": 1.1173, + "step": 29740 + }, + { + "epoch": 2.5, + "learning_rate": 1.3992778763026172e-06, + "loss": 1.0947, + "step": 29741 + }, + { + "epoch": 2.5, + "learning_rate": 1.3988141789506137e-06, + "loss": 1.1101, + "step": 29742 + }, + { + "epoch": 2.5, + "learning_rate": 1.3983505526643627e-06, + "loss": 1.1384, + "step": 29743 + }, + { + "epoch": 2.5, + "learning_rate": 1.3978869974476871e-06, + "loss": 1.1372, + "step": 29744 + }, + { + "epoch": 2.5, + "learning_rate": 1.3974235133044266e-06, + "loss": 1.0284, + "step": 29745 + }, + { + "epoch": 2.5, + "learning_rate": 1.3969601002384049e-06, + "loss": 1.1593, + "step": 29746 + }, + { + "epoch": 2.5, + "learning_rate": 1.3964967582534494e-06, + "loss": 1.0155, + "step": 29747 + }, + { + "epoch": 2.5, + "learning_rate": 1.3960334873533954e-06, + "loss": 1.1941, + "step": 29748 + }, + { + "epoch": 2.5, + "learning_rate": 1.3955702875420652e-06, + "loss": 1.0875, + "step": 29749 + }, + { + "epoch": 2.5, + "learning_rate": 1.3951071588232857e-06, + "loss": 1.0765, + "step": 29750 + }, + { + "epoch": 2.5, + "learning_rate": 1.3946441012008848e-06, + "loss": 1.0966, + "step": 29751 + }, + { + "epoch": 2.5, + "learning_rate": 1.3941811146786898e-06, + "loss": 1.1498, + "step": 29752 + }, + { + "epoch": 2.5, + "learning_rate": 1.3937181992605231e-06, + "loss": 1.1417, + "step": 29753 + }, + { + "epoch": 2.51, + "learning_rate": 1.3932553549502148e-06, + "loss": 1.1009, + "step": 29754 + }, + { + "epoch": 2.51, + "learning_rate": 1.3927925817515797e-06, + "loss": 0.9404, + "step": 29755 + }, + { + "epoch": 2.51, + "learning_rate": 1.3923298796684514e-06, + "loss": 1.132, + "step": 29756 + }, + { + "epoch": 2.51, + "learning_rate": 1.3918672487046469e-06, + "loss": 1.0188, + "step": 29757 + }, + { + "epoch": 2.51, + "learning_rate": 1.39140468886399e-06, + "loss": 1.02, + "step": 29758 + }, + { + "epoch": 2.51, + "learning_rate": 1.3909422001503014e-06, + "loss": 1.0824, + "step": 29759 + }, + { + "epoch": 2.51, + "learning_rate": 1.3904797825674056e-06, + "loss": 1.1719, + "step": 29760 + }, + { + "epoch": 2.51, + "learning_rate": 1.39001743611912e-06, + "loss": 1.1, + "step": 29761 + }, + { + "epoch": 2.51, + "learning_rate": 1.389555160809266e-06, + "loss": 1.0918, + "step": 29762 + }, + { + "epoch": 2.51, + "learning_rate": 1.3890929566416622e-06, + "loss": 1.0891, + "step": 29763 + }, + { + "epoch": 2.51, + "learning_rate": 1.3886308236201296e-06, + "loss": 1.156, + "step": 29764 + }, + { + "epoch": 2.51, + "learning_rate": 1.3881687617484874e-06, + "loss": 1.141, + "step": 29765 + }, + { + "epoch": 2.51, + "learning_rate": 1.3877067710305459e-06, + "loss": 1.1001, + "step": 29766 + }, + { + "epoch": 2.51, + "learning_rate": 1.387244851470132e-06, + "loss": 1.0437, + "step": 29767 + }, + { + "epoch": 2.51, + "learning_rate": 1.386783003071055e-06, + "loss": 1.2, + "step": 29768 + }, + { + "epoch": 2.51, + "learning_rate": 1.3863212258371339e-06, + "loss": 1.0533, + "step": 29769 + }, + { + "epoch": 2.51, + "learning_rate": 1.3858595197721825e-06, + "loss": 1.075, + "step": 29770 + }, + { + "epoch": 2.51, + "learning_rate": 1.3853978848800175e-06, + "loss": 1.1236, + "step": 29771 + }, + { + "epoch": 2.51, + "learning_rate": 1.384936321164453e-06, + "loss": 1.1178, + "step": 29772 + }, + { + "epoch": 2.51, + "learning_rate": 1.3844748286293008e-06, + "loss": 1.2508, + "step": 29773 + }, + { + "epoch": 2.51, + "learning_rate": 1.384013407278375e-06, + "loss": 1.1054, + "step": 29774 + }, + { + "epoch": 2.51, + "learning_rate": 1.383552057115487e-06, + "loss": 1.0569, + "step": 29775 + }, + { + "epoch": 2.51, + "learning_rate": 1.3830907781444536e-06, + "loss": 1.1178, + "step": 29776 + }, + { + "epoch": 2.51, + "learning_rate": 1.3826295703690762e-06, + "loss": 1.1759, + "step": 29777 + }, + { + "epoch": 2.51, + "learning_rate": 1.3821684337931763e-06, + "loss": 0.9828, + "step": 29778 + }, + { + "epoch": 2.51, + "learning_rate": 1.3817073684205562e-06, + "loss": 1.0472, + "step": 29779 + }, + { + "epoch": 2.51, + "learning_rate": 1.3812463742550286e-06, + "loss": 1.181, + "step": 29780 + }, + { + "epoch": 2.51, + "learning_rate": 1.3807854513004005e-06, + "loss": 1.0782, + "step": 29781 + }, + { + "epoch": 2.51, + "learning_rate": 1.3803245995604852e-06, + "loss": 1.0969, + "step": 29782 + }, + { + "epoch": 2.51, + "learning_rate": 1.379863819039081e-06, + "loss": 1.0964, + "step": 29783 + }, + { + "epoch": 2.51, + "learning_rate": 1.379403109740004e-06, + "loss": 1.13, + "step": 29784 + }, + { + "epoch": 2.51, + "learning_rate": 1.3789424716670597e-06, + "loss": 1.1832, + "step": 29785 + }, + { + "epoch": 2.51, + "learning_rate": 1.3784819048240472e-06, + "loss": 1.0831, + "step": 29786 + }, + { + "epoch": 2.51, + "learning_rate": 1.3780214092147803e-06, + "loss": 1.0528, + "step": 29787 + }, + { + "epoch": 2.51, + "learning_rate": 1.3775609848430583e-06, + "loss": 1.1185, + "step": 29788 + }, + { + "epoch": 2.51, + "learning_rate": 1.3771006317126867e-06, + "loss": 1.1132, + "step": 29789 + }, + { + "epoch": 2.51, + "learning_rate": 1.3766403498274695e-06, + "loss": 1.111, + "step": 29790 + }, + { + "epoch": 2.51, + "learning_rate": 1.376180139191209e-06, + "loss": 1.0176, + "step": 29791 + }, + { + "epoch": 2.51, + "learning_rate": 1.3757199998077086e-06, + "loss": 1.1235, + "step": 29792 + }, + { + "epoch": 2.51, + "learning_rate": 1.3752599316807725e-06, + "loss": 1.0903, + "step": 29793 + }, + { + "epoch": 2.51, + "learning_rate": 1.3747999348141938e-06, + "loss": 1.0719, + "step": 29794 + }, + { + "epoch": 2.51, + "learning_rate": 1.3743400092117831e-06, + "loss": 1.0721, + "step": 29795 + }, + { + "epoch": 2.51, + "learning_rate": 1.373880154877334e-06, + "loss": 1.0458, + "step": 29796 + }, + { + "epoch": 2.51, + "learning_rate": 1.3734203718146445e-06, + "loss": 1.111, + "step": 29797 + }, + { + "epoch": 2.51, + "learning_rate": 1.3729606600275225e-06, + "loss": 1.2047, + "step": 29798 + }, + { + "epoch": 2.51, + "learning_rate": 1.3725010195197574e-06, + "loss": 1.0797, + "step": 29799 + }, + { + "epoch": 2.51, + "learning_rate": 1.3720414502951507e-06, + "loss": 1.1146, + "step": 29800 + }, + { + "epoch": 2.51, + "learning_rate": 1.371581952357498e-06, + "loss": 1.0635, + "step": 29801 + }, + { + "epoch": 2.51, + "learning_rate": 1.3711225257105975e-06, + "loss": 1.0258, + "step": 29802 + }, + { + "epoch": 2.51, + "learning_rate": 1.370663170358244e-06, + "loss": 1.1437, + "step": 29803 + }, + { + "epoch": 2.51, + "learning_rate": 1.3702038863042345e-06, + "loss": 1.0971, + "step": 29804 + }, + { + "epoch": 2.51, + "learning_rate": 1.369744673552359e-06, + "loss": 1.1706, + "step": 29805 + }, + { + "epoch": 2.51, + "learning_rate": 1.369285532106419e-06, + "loss": 1.1955, + "step": 29806 + }, + { + "epoch": 2.51, + "learning_rate": 1.3688264619702018e-06, + "loss": 1.1043, + "step": 29807 + }, + { + "epoch": 2.51, + "learning_rate": 1.368367463147503e-06, + "loss": 1.0472, + "step": 29808 + }, + { + "epoch": 2.51, + "learning_rate": 1.3679085356421151e-06, + "loss": 1.134, + "step": 29809 + }, + { + "epoch": 2.51, + "learning_rate": 1.3674496794578295e-06, + "loss": 1.1225, + "step": 29810 + }, + { + "epoch": 2.51, + "learning_rate": 1.3669908945984366e-06, + "loss": 1.1086, + "step": 29811 + }, + { + "epoch": 2.51, + "learning_rate": 1.3665321810677278e-06, + "loss": 1.1219, + "step": 29812 + }, + { + "epoch": 2.51, + "learning_rate": 1.3660735388694946e-06, + "loss": 1.1213, + "step": 29813 + }, + { + "epoch": 2.51, + "learning_rate": 1.3656149680075237e-06, + "loss": 1.1279, + "step": 29814 + }, + { + "epoch": 2.51, + "learning_rate": 1.3651564684856077e-06, + "loss": 1.1248, + "step": 29815 + }, + { + "epoch": 2.51, + "learning_rate": 1.3646980403075284e-06, + "loss": 1.1262, + "step": 29816 + }, + { + "epoch": 2.51, + "learning_rate": 1.364239683477082e-06, + "loss": 1.1009, + "step": 29817 + }, + { + "epoch": 2.51, + "learning_rate": 1.3637813979980485e-06, + "loss": 1.097, + "step": 29818 + }, + { + "epoch": 2.51, + "learning_rate": 1.3633231838742166e-06, + "loss": 1.0348, + "step": 29819 + }, + { + "epoch": 2.51, + "learning_rate": 1.3628650411093736e-06, + "loss": 1.1116, + "step": 29820 + }, + { + "epoch": 2.51, + "learning_rate": 1.362406969707304e-06, + "loss": 1.1053, + "step": 29821 + }, + { + "epoch": 2.51, + "learning_rate": 1.3619489696717913e-06, + "loss": 1.0614, + "step": 29822 + }, + { + "epoch": 2.51, + "learning_rate": 1.3614910410066217e-06, + "loss": 1.1042, + "step": 29823 + }, + { + "epoch": 2.51, + "learning_rate": 1.3610331837155776e-06, + "loss": 1.1648, + "step": 29824 + }, + { + "epoch": 2.51, + "learning_rate": 1.3605753978024417e-06, + "loss": 1.0766, + "step": 29825 + }, + { + "epoch": 2.51, + "learning_rate": 1.3601176832709995e-06, + "loss": 1.1102, + "step": 29826 + }, + { + "epoch": 2.51, + "learning_rate": 1.3596600401250258e-06, + "loss": 1.1513, + "step": 29827 + }, + { + "epoch": 2.51, + "learning_rate": 1.3592024683683104e-06, + "loss": 1.0565, + "step": 29828 + }, + { + "epoch": 2.51, + "learning_rate": 1.3587449680046272e-06, + "loss": 1.074, + "step": 29829 + }, + { + "epoch": 2.51, + "learning_rate": 1.3582875390377592e-06, + "loss": 1.1826, + "step": 29830 + }, + { + "epoch": 2.51, + "learning_rate": 1.3578301814714844e-06, + "loss": 1.0903, + "step": 29831 + }, + { + "epoch": 2.51, + "learning_rate": 1.3573728953095833e-06, + "loss": 1.0422, + "step": 29832 + }, + { + "epoch": 2.51, + "learning_rate": 1.3569156805558326e-06, + "loss": 1.0668, + "step": 29833 + }, + { + "epoch": 2.51, + "learning_rate": 1.3564585372140114e-06, + "loss": 1.054, + "step": 29834 + }, + { + "epoch": 2.51, + "learning_rate": 1.356001465287895e-06, + "loss": 1.0721, + "step": 29835 + }, + { + "epoch": 2.51, + "learning_rate": 1.355544464781262e-06, + "loss": 1.0793, + "step": 29836 + }, + { + "epoch": 2.51, + "learning_rate": 1.3550875356978888e-06, + "loss": 1.1492, + "step": 29837 + }, + { + "epoch": 2.51, + "learning_rate": 1.3546306780415453e-06, + "loss": 1.0856, + "step": 29838 + }, + { + "epoch": 2.51, + "learning_rate": 1.3541738918160141e-06, + "loss": 1.0964, + "step": 29839 + }, + { + "epoch": 2.51, + "learning_rate": 1.3537171770250624e-06, + "loss": 1.0883, + "step": 29840 + }, + { + "epoch": 2.51, + "learning_rate": 1.353260533672467e-06, + "loss": 1.0932, + "step": 29841 + }, + { + "epoch": 2.51, + "learning_rate": 1.3528039617620004e-06, + "loss": 1.1437, + "step": 29842 + }, + { + "epoch": 2.51, + "learning_rate": 1.3523474612974374e-06, + "loss": 1.1659, + "step": 29843 + }, + { + "epoch": 2.51, + "learning_rate": 1.3518910322825429e-06, + "loss": 1.171, + "step": 29844 + }, + { + "epoch": 2.51, + "learning_rate": 1.351434674721097e-06, + "loss": 1.1714, + "step": 29845 + }, + { + "epoch": 2.51, + "learning_rate": 1.3509783886168626e-06, + "loss": 1.0991, + "step": 29846 + }, + { + "epoch": 2.51, + "learning_rate": 1.3505221739736117e-06, + "loss": 1.1273, + "step": 29847 + }, + { + "epoch": 2.51, + "learning_rate": 1.3500660307951185e-06, + "loss": 1.1098, + "step": 29848 + }, + { + "epoch": 2.51, + "learning_rate": 1.3496099590851463e-06, + "loss": 1.0472, + "step": 29849 + }, + { + "epoch": 2.51, + "learning_rate": 1.3491539588474645e-06, + "loss": 1.0661, + "step": 29850 + }, + { + "epoch": 2.51, + "learning_rate": 1.348698030085841e-06, + "loss": 1.1928, + "step": 29851 + }, + { + "epoch": 2.51, + "learning_rate": 1.348242172804044e-06, + "loss": 1.1063, + "step": 29852 + }, + { + "epoch": 2.51, + "learning_rate": 1.3477863870058393e-06, + "loss": 1.0573, + "step": 29853 + }, + { + "epoch": 2.51, + "learning_rate": 1.3473306726949942e-06, + "loss": 1.0702, + "step": 29854 + }, + { + "epoch": 2.51, + "learning_rate": 1.3468750298752674e-06, + "loss": 1.1527, + "step": 29855 + }, + { + "epoch": 2.51, + "learning_rate": 1.3464194585504342e-06, + "loss": 1.1796, + "step": 29856 + }, + { + "epoch": 2.51, + "learning_rate": 1.34596395872425e-06, + "loss": 1.0788, + "step": 29857 + }, + { + "epoch": 2.51, + "learning_rate": 1.345508530400481e-06, + "loss": 1.1227, + "step": 29858 + }, + { + "epoch": 2.51, + "learning_rate": 1.345053173582891e-06, + "loss": 1.0754, + "step": 29859 + }, + { + "epoch": 2.51, + "learning_rate": 1.3445978882752408e-06, + "loss": 1.2026, + "step": 29860 + }, + { + "epoch": 2.51, + "learning_rate": 1.3441426744812924e-06, + "loss": 1.1582, + "step": 29861 + }, + { + "epoch": 2.51, + "learning_rate": 1.3436875322048092e-06, + "loss": 1.1491, + "step": 29862 + }, + { + "epoch": 2.51, + "learning_rate": 1.3432324614495484e-06, + "loss": 1.1098, + "step": 29863 + }, + { + "epoch": 2.51, + "learning_rate": 1.3427774622192723e-06, + "loss": 1.1315, + "step": 29864 + }, + { + "epoch": 2.51, + "learning_rate": 1.3423225345177415e-06, + "loss": 1.0476, + "step": 29865 + }, + { + "epoch": 2.51, + "learning_rate": 1.3418676783487083e-06, + "loss": 1.1359, + "step": 29866 + }, + { + "epoch": 2.51, + "learning_rate": 1.3414128937159398e-06, + "loss": 1.1251, + "step": 29867 + }, + { + "epoch": 2.51, + "learning_rate": 1.3409581806231874e-06, + "loss": 1.1873, + "step": 29868 + }, + { + "epoch": 2.51, + "learning_rate": 1.3405035390742093e-06, + "loss": 1.0551, + "step": 29869 + }, + { + "epoch": 2.51, + "learning_rate": 1.3400489690727624e-06, + "loss": 1.1125, + "step": 29870 + }, + { + "epoch": 2.51, + "learning_rate": 1.3395944706226028e-06, + "loss": 1.01, + "step": 29871 + }, + { + "epoch": 2.51, + "learning_rate": 1.339140043727485e-06, + "loss": 1.0668, + "step": 29872 + }, + { + "epoch": 2.52, + "learning_rate": 1.3386856883911636e-06, + "loss": 1.0469, + "step": 29873 + }, + { + "epoch": 2.52, + "learning_rate": 1.3382314046173927e-06, + "loss": 1.1053, + "step": 29874 + }, + { + "epoch": 2.52, + "learning_rate": 1.337777192409927e-06, + "loss": 1.1346, + "step": 29875 + }, + { + "epoch": 2.52, + "learning_rate": 1.33732305177252e-06, + "loss": 1.1164, + "step": 29876 + }, + { + "epoch": 2.52, + "learning_rate": 1.3368689827089187e-06, + "loss": 1.0932, + "step": 29877 + }, + { + "epoch": 2.52, + "learning_rate": 1.3364149852228824e-06, + "loss": 1.1249, + "step": 29878 + }, + { + "epoch": 2.52, + "learning_rate": 1.335961059318156e-06, + "loss": 1.1237, + "step": 29879 + }, + { + "epoch": 2.52, + "learning_rate": 1.3355072049984918e-06, + "loss": 1.0487, + "step": 29880 + }, + { + "epoch": 2.52, + "learning_rate": 1.3350534222676403e-06, + "loss": 1.1017, + "step": 29881 + }, + { + "epoch": 2.52, + "learning_rate": 1.3345997111293506e-06, + "loss": 1.1797, + "step": 29882 + }, + { + "epoch": 2.52, + "learning_rate": 1.3341460715873723e-06, + "loss": 1.0653, + "step": 29883 + }, + { + "epoch": 2.52, + "learning_rate": 1.3336925036454519e-06, + "loss": 1.0552, + "step": 29884 + }, + { + "epoch": 2.52, + "learning_rate": 1.3332390073073365e-06, + "loss": 1.1654, + "step": 29885 + }, + { + "epoch": 2.52, + "learning_rate": 1.3327855825767754e-06, + "loss": 0.9937, + "step": 29886 + }, + { + "epoch": 2.52, + "learning_rate": 1.3323322294575158e-06, + "loss": 1.1354, + "step": 29887 + }, + { + "epoch": 2.52, + "learning_rate": 1.3318789479532967e-06, + "loss": 1.066, + "step": 29888 + }, + { + "epoch": 2.52, + "learning_rate": 1.331425738067872e-06, + "loss": 1.1174, + "step": 29889 + }, + { + "epoch": 2.52, + "learning_rate": 1.3309725998049806e-06, + "loss": 1.1887, + "step": 29890 + }, + { + "epoch": 2.52, + "learning_rate": 1.3305195331683673e-06, + "loss": 1.1713, + "step": 29891 + }, + { + "epoch": 2.52, + "learning_rate": 1.3300665381617773e-06, + "loss": 1.1417, + "step": 29892 + }, + { + "epoch": 2.52, + "learning_rate": 1.3296136147889527e-06, + "loss": 1.1402, + "step": 29893 + }, + { + "epoch": 2.52, + "learning_rate": 1.329160763053633e-06, + "loss": 1.1869, + "step": 29894 + }, + { + "epoch": 2.52, + "learning_rate": 1.3287079829595651e-06, + "loss": 1.2034, + "step": 29895 + }, + { + "epoch": 2.52, + "learning_rate": 1.328255274510485e-06, + "loss": 1.0749, + "step": 29896 + }, + { + "epoch": 2.52, + "learning_rate": 1.3278026377101327e-06, + "loss": 1.1741, + "step": 29897 + }, + { + "epoch": 2.52, + "learning_rate": 1.3273500725622557e-06, + "loss": 1.1332, + "step": 29898 + }, + { + "epoch": 2.52, + "learning_rate": 1.3268975790705851e-06, + "loss": 1.0395, + "step": 29899 + }, + { + "epoch": 2.52, + "learning_rate": 1.3264451572388626e-06, + "loss": 1.1313, + "step": 29900 + }, + { + "epoch": 2.52, + "learning_rate": 1.325992807070826e-06, + "loss": 1.1093, + "step": 29901 + }, + { + "epoch": 2.52, + "learning_rate": 1.3255405285702138e-06, + "loss": 1.0316, + "step": 29902 + }, + { + "epoch": 2.52, + "learning_rate": 1.3250883217407607e-06, + "loss": 1.0946, + "step": 29903 + }, + { + "epoch": 2.52, + "learning_rate": 1.3246361865862077e-06, + "loss": 1.0656, + "step": 29904 + }, + { + "epoch": 2.52, + "learning_rate": 1.3241841231102814e-06, + "loss": 1.125, + "step": 29905 + }, + { + "epoch": 2.52, + "learning_rate": 1.3237321313167273e-06, + "loss": 1.113, + "step": 29906 + }, + { + "epoch": 2.52, + "learning_rate": 1.3232802112092735e-06, + "loss": 1.0255, + "step": 29907 + }, + { + "epoch": 2.52, + "learning_rate": 1.322828362791655e-06, + "loss": 1.1083, + "step": 29908 + }, + { + "epoch": 2.52, + "learning_rate": 1.3223765860676053e-06, + "loss": 1.0998, + "step": 29909 + }, + { + "epoch": 2.52, + "learning_rate": 1.321924881040857e-06, + "loss": 1.1231, + "step": 29910 + }, + { + "epoch": 2.52, + "learning_rate": 1.3214732477151438e-06, + "loss": 1.0975, + "step": 29911 + }, + { + "epoch": 2.52, + "learning_rate": 1.321021686094196e-06, + "loss": 1.1559, + "step": 29912 + }, + { + "epoch": 2.52, + "learning_rate": 1.320570196181744e-06, + "loss": 1.0812, + "step": 29913 + }, + { + "epoch": 2.52, + "learning_rate": 1.3201187779815194e-06, + "loss": 1.0949, + "step": 29914 + }, + { + "epoch": 2.52, + "learning_rate": 1.3196674314972536e-06, + "loss": 1.1377, + "step": 29915 + }, + { + "epoch": 2.52, + "learning_rate": 1.3192161567326678e-06, + "loss": 1.1041, + "step": 29916 + }, + { + "epoch": 2.52, + "learning_rate": 1.3187649536915027e-06, + "loss": 1.1065, + "step": 29917 + }, + { + "epoch": 2.52, + "learning_rate": 1.318313822377476e-06, + "loss": 1.0735, + "step": 29918 + }, + { + "epoch": 2.52, + "learning_rate": 1.3178627627943197e-06, + "loss": 1.0629, + "step": 29919 + }, + { + "epoch": 2.52, + "learning_rate": 1.3174117749457604e-06, + "loss": 1.1612, + "step": 29920 + }, + { + "epoch": 2.52, + "learning_rate": 1.3169608588355232e-06, + "loss": 1.1904, + "step": 29921 + }, + { + "epoch": 2.52, + "learning_rate": 1.3165100144673349e-06, + "loss": 1.2338, + "step": 29922 + }, + { + "epoch": 2.52, + "learning_rate": 1.3160592418449192e-06, + "loss": 1.1179, + "step": 29923 + }, + { + "epoch": 2.52, + "learning_rate": 1.315608540972002e-06, + "loss": 1.1386, + "step": 29924 + }, + { + "epoch": 2.52, + "learning_rate": 1.3151579118523051e-06, + "loss": 1.1975, + "step": 29925 + }, + { + "epoch": 2.52, + "learning_rate": 1.3147073544895561e-06, + "loss": 1.0632, + "step": 29926 + }, + { + "epoch": 2.52, + "learning_rate": 1.3142568688874702e-06, + "loss": 1.133, + "step": 29927 + }, + { + "epoch": 2.52, + "learning_rate": 1.3138064550497787e-06, + "loss": 1.1073, + "step": 29928 + }, + { + "epoch": 2.52, + "learning_rate": 1.3133561129801952e-06, + "loss": 1.0889, + "step": 29929 + }, + { + "epoch": 2.52, + "learning_rate": 1.3129058426824449e-06, + "loss": 1.0538, + "step": 29930 + }, + { + "epoch": 2.52, + "learning_rate": 1.3124556441602455e-06, + "loss": 0.9997, + "step": 29931 + }, + { + "epoch": 2.52, + "learning_rate": 1.3120055174173174e-06, + "loss": 1.0709, + "step": 29932 + }, + { + "epoch": 2.52, + "learning_rate": 1.3115554624573811e-06, + "loss": 1.0916, + "step": 29933 + }, + { + "epoch": 2.52, + "learning_rate": 1.3111054792841549e-06, + "loss": 1.1526, + "step": 29934 + }, + { + "epoch": 2.52, + "learning_rate": 1.3106555679013556e-06, + "loss": 1.0616, + "step": 29935 + }, + { + "epoch": 2.52, + "learning_rate": 1.3102057283127e-06, + "loss": 1.1068, + "step": 29936 + }, + { + "epoch": 2.52, + "learning_rate": 1.309755960521909e-06, + "loss": 1.1855, + "step": 29937 + }, + { + "epoch": 2.52, + "learning_rate": 1.3093062645326903e-06, + "loss": 1.0676, + "step": 29938 + }, + { + "epoch": 2.52, + "learning_rate": 1.3088566403487691e-06, + "loss": 1.0866, + "step": 29939 + }, + { + "epoch": 2.52, + "learning_rate": 1.3084070879738541e-06, + "loss": 1.0922, + "step": 29940 + }, + { + "epoch": 2.52, + "learning_rate": 1.3079576074116617e-06, + "loss": 1.0491, + "step": 29941 + }, + { + "epoch": 2.52, + "learning_rate": 1.3075081986659043e-06, + "loss": 1.0745, + "step": 29942 + }, + { + "epoch": 2.52, + "learning_rate": 1.307058861740299e-06, + "loss": 1.0031, + "step": 29943 + }, + { + "epoch": 2.52, + "learning_rate": 1.3066095966385517e-06, + "loss": 1.1342, + "step": 29944 + }, + { + "epoch": 2.52, + "learning_rate": 1.3061604033643816e-06, + "loss": 1.1508, + "step": 29945 + }, + { + "epoch": 2.52, + "learning_rate": 1.3057112819214946e-06, + "loss": 1.0964, + "step": 29946 + }, + { + "epoch": 2.52, + "learning_rate": 1.3052622323136021e-06, + "loss": 1.1826, + "step": 29947 + }, + { + "epoch": 2.52, + "learning_rate": 1.3048132545444192e-06, + "loss": 1.1847, + "step": 29948 + }, + { + "epoch": 2.52, + "learning_rate": 1.3043643486176494e-06, + "loss": 1.1116, + "step": 29949 + }, + { + "epoch": 2.52, + "learning_rate": 1.3039155145370052e-06, + "loss": 1.1853, + "step": 29950 + }, + { + "epoch": 2.52, + "learning_rate": 1.303466752306194e-06, + "loss": 1.0977, + "step": 29951 + }, + { + "epoch": 2.52, + "learning_rate": 1.3030180619289235e-06, + "loss": 1.1344, + "step": 29952 + }, + { + "epoch": 2.52, + "learning_rate": 1.302569443408902e-06, + "loss": 1.1007, + "step": 29953 + }, + { + "epoch": 2.52, + "learning_rate": 1.3021208967498367e-06, + "loss": 1.2002, + "step": 29954 + }, + { + "epoch": 2.52, + "learning_rate": 1.3016724219554277e-06, + "loss": 1.1289, + "step": 29955 + }, + { + "epoch": 2.52, + "learning_rate": 1.301224019029389e-06, + "loss": 1.1745, + "step": 29956 + }, + { + "epoch": 2.52, + "learning_rate": 1.3007756879754208e-06, + "loss": 1.1179, + "step": 29957 + }, + { + "epoch": 2.52, + "learning_rate": 1.3003274287972268e-06, + "loss": 1.1409, + "step": 29958 + }, + { + "epoch": 2.52, + "learning_rate": 1.2998792414985117e-06, + "loss": 1.0853, + "step": 29959 + }, + { + "epoch": 2.52, + "learning_rate": 1.2994311260829794e-06, + "loss": 1.0595, + "step": 29960 + }, + { + "epoch": 2.52, + "learning_rate": 1.2989830825543314e-06, + "loss": 1.1334, + "step": 29961 + }, + { + "epoch": 2.52, + "learning_rate": 1.2985351109162692e-06, + "loss": 1.0403, + "step": 29962 + }, + { + "epoch": 2.52, + "learning_rate": 1.298087211172495e-06, + "loss": 1.0595, + "step": 29963 + }, + { + "epoch": 2.52, + "learning_rate": 1.2976393833267098e-06, + "loss": 1.1287, + "step": 29964 + }, + { + "epoch": 2.52, + "learning_rate": 1.2971916273826156e-06, + "loss": 1.0911, + "step": 29965 + }, + { + "epoch": 2.52, + "learning_rate": 1.2967439433439043e-06, + "loss": 1.114, + "step": 29966 + }, + { + "epoch": 2.52, + "learning_rate": 1.296296331214284e-06, + "loss": 1.1034, + "step": 29967 + }, + { + "epoch": 2.52, + "learning_rate": 1.295848790997447e-06, + "loss": 1.0432, + "step": 29968 + }, + { + "epoch": 2.52, + "learning_rate": 1.295401322697094e-06, + "loss": 1.1413, + "step": 29969 + }, + { + "epoch": 2.52, + "learning_rate": 1.2949539263169197e-06, + "loss": 1.0827, + "step": 29970 + }, + { + "epoch": 2.52, + "learning_rate": 1.2945066018606233e-06, + "loss": 1.0586, + "step": 29971 + }, + { + "epoch": 2.52, + "learning_rate": 1.2940593493318987e-06, + "loss": 1.1967, + "step": 29972 + }, + { + "epoch": 2.52, + "learning_rate": 1.2936121687344427e-06, + "loss": 1.1594, + "step": 29973 + }, + { + "epoch": 2.52, + "learning_rate": 1.293165060071948e-06, + "loss": 1.1009, + "step": 29974 + }, + { + "epoch": 2.52, + "learning_rate": 1.2927180233481119e-06, + "loss": 1.1507, + "step": 29975 + }, + { + "epoch": 2.52, + "learning_rate": 1.2922710585666264e-06, + "loss": 1.1665, + "step": 29976 + }, + { + "epoch": 2.52, + "learning_rate": 1.2918241657311814e-06, + "loss": 1.1763, + "step": 29977 + }, + { + "epoch": 2.52, + "learning_rate": 1.291377344845476e-06, + "loss": 1.0619, + "step": 29978 + }, + { + "epoch": 2.52, + "learning_rate": 1.2909305959131957e-06, + "loss": 1.036, + "step": 29979 + }, + { + "epoch": 2.52, + "learning_rate": 1.2904839189380346e-06, + "loss": 1.0374, + "step": 29980 + }, + { + "epoch": 2.52, + "learning_rate": 1.290037313923682e-06, + "loss": 1.1071, + "step": 29981 + }, + { + "epoch": 2.52, + "learning_rate": 1.2895907808738296e-06, + "loss": 1.0885, + "step": 29982 + }, + { + "epoch": 2.52, + "learning_rate": 1.289144319792165e-06, + "loss": 1.1975, + "step": 29983 + }, + { + "epoch": 2.52, + "learning_rate": 1.2886979306823777e-06, + "loss": 1.1263, + "step": 29984 + }, + { + "epoch": 2.52, + "learning_rate": 1.288251613548157e-06, + "loss": 1.1347, + "step": 29985 + }, + { + "epoch": 2.52, + "learning_rate": 1.287805368393189e-06, + "loss": 1.0943, + "step": 29986 + }, + { + "epoch": 2.52, + "learning_rate": 1.287359195221164e-06, + "loss": 1.0651, + "step": 29987 + }, + { + "epoch": 2.52, + "learning_rate": 1.286913094035761e-06, + "loss": 0.9665, + "step": 29988 + }, + { + "epoch": 2.52, + "learning_rate": 1.2864670648406752e-06, + "loss": 1.0435, + "step": 29989 + }, + { + "epoch": 2.52, + "learning_rate": 1.2860211076395845e-06, + "loss": 1.1755, + "step": 29990 + }, + { + "epoch": 2.52, + "learning_rate": 1.2855752224361762e-06, + "loss": 1.1043, + "step": 29991 + }, + { + "epoch": 2.53, + "learning_rate": 1.2851294092341348e-06, + "loss": 1.1665, + "step": 29992 + }, + { + "epoch": 2.53, + "learning_rate": 1.2846836680371456e-06, + "loss": 1.0897, + "step": 29993 + }, + { + "epoch": 2.53, + "learning_rate": 1.284237998848884e-06, + "loss": 1.1266, + "step": 29994 + }, + { + "epoch": 2.53, + "learning_rate": 1.283792401673043e-06, + "loss": 1.0945, + "step": 29995 + }, + { + "epoch": 2.53, + "learning_rate": 1.283346876513296e-06, + "loss": 1.129, + "step": 29996 + }, + { + "epoch": 2.53, + "learning_rate": 1.2829014233733238e-06, + "loss": 1.1591, + "step": 29997 + }, + { + "epoch": 2.53, + "learning_rate": 1.2824560422568155e-06, + "loss": 1.1563, + "step": 29998 + }, + { + "epoch": 2.53, + "learning_rate": 1.2820107331674426e-06, + "loss": 1.1672, + "step": 29999 + }, + { + "epoch": 2.53, + "learning_rate": 1.2815654961088875e-06, + "loss": 1.1233, + "step": 30000 + }, + { + "epoch": 2.53, + "learning_rate": 1.2811203310848285e-06, + "loss": 1.1249, + "step": 30001 + }, + { + "epoch": 2.53, + "learning_rate": 1.2806752380989428e-06, + "loss": 1.1309, + "step": 30002 + }, + { + "epoch": 2.53, + "learning_rate": 1.2802302171549097e-06, + "loss": 1.1501, + "step": 30003 + }, + { + "epoch": 2.53, + "learning_rate": 1.2797852682564072e-06, + "loss": 1.0646, + "step": 30004 + }, + { + "epoch": 2.53, + "learning_rate": 1.2793403914071057e-06, + "loss": 1.1665, + "step": 30005 + }, + { + "epoch": 2.53, + "learning_rate": 1.278895586610689e-06, + "loss": 1.0981, + "step": 30006 + }, + { + "epoch": 2.53, + "learning_rate": 1.278450853870825e-06, + "loss": 1.1954, + "step": 30007 + }, + { + "epoch": 2.53, + "learning_rate": 1.2780061931911925e-06, + "loss": 1.1394, + "step": 30008 + }, + { + "epoch": 2.53, + "learning_rate": 1.2775616045754635e-06, + "loss": 1.1245, + "step": 30009 + }, + { + "epoch": 2.53, + "learning_rate": 1.2771170880273121e-06, + "loss": 1.1954, + "step": 30010 + }, + { + "epoch": 2.53, + "learning_rate": 1.2766726435504117e-06, + "loss": 1.1066, + "step": 30011 + }, + { + "epoch": 2.53, + "learning_rate": 1.2762282711484331e-06, + "loss": 1.0444, + "step": 30012 + }, + { + "epoch": 2.53, + "learning_rate": 1.2757839708250486e-06, + "loss": 1.1396, + "step": 30013 + }, + { + "epoch": 2.53, + "learning_rate": 1.2753397425839287e-06, + "loss": 1.1635, + "step": 30014 + }, + { + "epoch": 2.53, + "learning_rate": 1.274895586428747e-06, + "loss": 1.1278, + "step": 30015 + }, + { + "epoch": 2.53, + "learning_rate": 1.2744515023631655e-06, + "loss": 1.0036, + "step": 30016 + }, + { + "epoch": 2.53, + "learning_rate": 1.2740074903908628e-06, + "loss": 1.112, + "step": 30017 + }, + { + "epoch": 2.53, + "learning_rate": 1.2735635505155007e-06, + "loss": 1.0735, + "step": 30018 + }, + { + "epoch": 2.53, + "learning_rate": 1.273119682740749e-06, + "loss": 1.151, + "step": 30019 + }, + { + "epoch": 2.53, + "learning_rate": 1.272675887070276e-06, + "loss": 1.0473, + "step": 30020 + }, + { + "epoch": 2.53, + "learning_rate": 1.2722321635077473e-06, + "loss": 1.0888, + "step": 30021 + }, + { + "epoch": 2.53, + "learning_rate": 1.2717885120568296e-06, + "loss": 1.0766, + "step": 30022 + }, + { + "epoch": 2.53, + "learning_rate": 1.271344932721189e-06, + "loss": 1.1379, + "step": 30023 + }, + { + "epoch": 2.53, + "learning_rate": 1.2709014255044905e-06, + "loss": 1.097, + "step": 30024 + }, + { + "epoch": 2.53, + "learning_rate": 1.2704579904103974e-06, + "loss": 1.1912, + "step": 30025 + }, + { + "epoch": 2.53, + "learning_rate": 1.2700146274425774e-06, + "loss": 1.1255, + "step": 30026 + }, + { + "epoch": 2.53, + "learning_rate": 1.2695713366046857e-06, + "loss": 1.0754, + "step": 30027 + }, + { + "epoch": 2.53, + "learning_rate": 1.2691281179003934e-06, + "loss": 1.0917, + "step": 30028 + }, + { + "epoch": 2.53, + "learning_rate": 1.268684971333357e-06, + "loss": 1.1263, + "step": 30029 + }, + { + "epoch": 2.53, + "learning_rate": 1.2682418969072395e-06, + "loss": 1.1061, + "step": 30030 + }, + { + "epoch": 2.53, + "learning_rate": 1.2677988946257026e-06, + "loss": 1.167, + "step": 30031 + }, + { + "epoch": 2.53, + "learning_rate": 1.2673559644924062e-06, + "loss": 1.0434, + "step": 30032 + }, + { + "epoch": 2.53, + "learning_rate": 1.2669131065110086e-06, + "loss": 1.0497, + "step": 30033 + }, + { + "epoch": 2.53, + "learning_rate": 1.2664703206851702e-06, + "loss": 1.11, + "step": 30034 + }, + { + "epoch": 2.53, + "learning_rate": 1.2660276070185495e-06, + "loss": 1.16, + "step": 30035 + }, + { + "epoch": 2.53, + "learning_rate": 1.2655849655148033e-06, + "loss": 1.1235, + "step": 30036 + }, + { + "epoch": 2.53, + "learning_rate": 1.2651423961775921e-06, + "loss": 1.1252, + "step": 30037 + }, + { + "epoch": 2.53, + "learning_rate": 1.2646998990105651e-06, + "loss": 1.0906, + "step": 30038 + }, + { + "epoch": 2.53, + "learning_rate": 1.2642574740173885e-06, + "loss": 1.0729, + "step": 30039 + }, + { + "epoch": 2.53, + "learning_rate": 1.2638151212017092e-06, + "loss": 1.0636, + "step": 30040 + }, + { + "epoch": 2.53, + "learning_rate": 1.2633728405671853e-06, + "loss": 1.0749, + "step": 30041 + }, + { + "epoch": 2.53, + "learning_rate": 1.2629306321174718e-06, + "loss": 1.1693, + "step": 30042 + }, + { + "epoch": 2.53, + "learning_rate": 1.2624884958562233e-06, + "loss": 1.1089, + "step": 30043 + }, + { + "epoch": 2.53, + "learning_rate": 1.2620464317870873e-06, + "loss": 1.1059, + "step": 30044 + }, + { + "epoch": 2.53, + "learning_rate": 1.261604439913725e-06, + "loss": 1.0969, + "step": 30045 + }, + { + "epoch": 2.53, + "learning_rate": 1.2611625202397803e-06, + "loss": 1.0563, + "step": 30046 + }, + { + "epoch": 2.53, + "learning_rate": 1.2607206727689059e-06, + "loss": 1.1079, + "step": 30047 + }, + { + "epoch": 2.53, + "learning_rate": 1.2602788975047586e-06, + "loss": 1.0922, + "step": 30048 + }, + { + "epoch": 2.53, + "learning_rate": 1.2598371944509813e-06, + "loss": 1.1518, + "step": 30049 + }, + { + "epoch": 2.53, + "learning_rate": 1.2593955636112275e-06, + "loss": 1.1026, + "step": 30050 + }, + { + "epoch": 2.53, + "learning_rate": 1.2589540049891436e-06, + "loss": 1.0774, + "step": 30051 + }, + { + "epoch": 2.53, + "learning_rate": 1.2585125185883796e-06, + "loss": 1.0883, + "step": 30052 + }, + { + "epoch": 2.53, + "learning_rate": 1.2580711044125837e-06, + "loss": 1.0416, + "step": 30053 + }, + { + "epoch": 2.53, + "learning_rate": 1.2576297624654033e-06, + "loss": 1.1023, + "step": 30054 + }, + { + "epoch": 2.53, + "learning_rate": 1.2571884927504796e-06, + "loss": 1.1305, + "step": 30055 + }, + { + "epoch": 2.53, + "learning_rate": 1.2567472952714676e-06, + "loss": 1.1159, + "step": 30056 + }, + { + "epoch": 2.53, + "learning_rate": 1.2563061700320046e-06, + "loss": 1.187, + "step": 30057 + }, + { + "epoch": 2.53, + "learning_rate": 1.2558651170357383e-06, + "loss": 1.1469, + "step": 30058 + }, + { + "epoch": 2.53, + "learning_rate": 1.255424136286314e-06, + "loss": 1.0381, + "step": 30059 + }, + { + "epoch": 2.53, + "learning_rate": 1.2549832277873731e-06, + "loss": 1.0886, + "step": 30060 + }, + { + "epoch": 2.53, + "learning_rate": 1.2545423915425603e-06, + "loss": 1.0719, + "step": 30061 + }, + { + "epoch": 2.53, + "learning_rate": 1.2541016275555162e-06, + "loss": 1.0561, + "step": 30062 + }, + { + "epoch": 2.53, + "learning_rate": 1.2536609358298846e-06, + "loss": 1.0776, + "step": 30063 + }, + { + "epoch": 2.53, + "learning_rate": 1.2532203163693046e-06, + "loss": 1.0943, + "step": 30064 + }, + { + "epoch": 2.53, + "learning_rate": 1.2527797691774202e-06, + "loss": 1.0854, + "step": 30065 + }, + { + "epoch": 2.53, + "learning_rate": 1.2523392942578649e-06, + "loss": 1.0648, + "step": 30066 + }, + { + "epoch": 2.53, + "learning_rate": 1.251898891614285e-06, + "loss": 1.1786, + "step": 30067 + }, + { + "epoch": 2.53, + "learning_rate": 1.2514585612503149e-06, + "loss": 1.12, + "step": 30068 + }, + { + "epoch": 2.53, + "learning_rate": 1.2510183031695944e-06, + "loss": 1.1644, + "step": 30069 + }, + { + "epoch": 2.53, + "learning_rate": 1.2505781173757602e-06, + "loss": 1.1385, + "step": 30070 + }, + { + "epoch": 2.53, + "learning_rate": 1.2501380038724499e-06, + "loss": 1.0119, + "step": 30071 + }, + { + "epoch": 2.53, + "learning_rate": 1.2496979626633e-06, + "loss": 1.0804, + "step": 30072 + }, + { + "epoch": 2.53, + "learning_rate": 1.2492579937519455e-06, + "loss": 1.0629, + "step": 30073 + }, + { + "epoch": 2.53, + "learning_rate": 1.2488180971420228e-06, + "loss": 1.1911, + "step": 30074 + }, + { + "epoch": 2.53, + "learning_rate": 1.2483782728371652e-06, + "loss": 1.1277, + "step": 30075 + }, + { + "epoch": 2.53, + "learning_rate": 1.2479385208410088e-06, + "loss": 1.2, + "step": 30076 + }, + { + "epoch": 2.53, + "learning_rate": 1.247498841157182e-06, + "loss": 1.0555, + "step": 30077 + }, + { + "epoch": 2.53, + "learning_rate": 1.247059233789325e-06, + "loss": 1.1036, + "step": 30078 + }, + { + "epoch": 2.53, + "learning_rate": 1.246619698741064e-06, + "loss": 1.0929, + "step": 30079 + }, + { + "epoch": 2.53, + "learning_rate": 1.2461802360160313e-06, + "loss": 1.0372, + "step": 30080 + }, + { + "epoch": 2.53, + "learning_rate": 1.24574084561786e-06, + "loss": 1.0381, + "step": 30081 + }, + { + "epoch": 2.53, + "learning_rate": 1.24530152755018e-06, + "loss": 1.0668, + "step": 30082 + }, + { + "epoch": 2.53, + "learning_rate": 1.24486228181662e-06, + "loss": 1.1702, + "step": 30083 + }, + { + "epoch": 2.53, + "learning_rate": 1.2444231084208092e-06, + "loss": 1.0777, + "step": 30084 + }, + { + "epoch": 2.53, + "learning_rate": 1.2439840073663777e-06, + "loss": 1.0577, + "step": 30085 + }, + { + "epoch": 2.53, + "learning_rate": 1.2435449786569532e-06, + "loss": 1.1217, + "step": 30086 + }, + { + "epoch": 2.53, + "learning_rate": 1.2431060222961632e-06, + "loss": 1.0811, + "step": 30087 + }, + { + "epoch": 2.53, + "learning_rate": 1.242667138287631e-06, + "loss": 1.0778, + "step": 30088 + }, + { + "epoch": 2.53, + "learning_rate": 1.2422283266349888e-06, + "loss": 1.1337, + "step": 30089 + }, + { + "epoch": 2.53, + "learning_rate": 1.2417895873418574e-06, + "loss": 1.1349, + "step": 30090 + }, + { + "epoch": 2.53, + "learning_rate": 1.2413509204118623e-06, + "loss": 1.1552, + "step": 30091 + }, + { + "epoch": 2.53, + "learning_rate": 1.24091232584863e-06, + "loss": 1.199, + "step": 30092 + }, + { + "epoch": 2.53, + "learning_rate": 1.2404738036557862e-06, + "loss": 1.1357, + "step": 30093 + }, + { + "epoch": 2.53, + "learning_rate": 1.2400353538369459e-06, + "loss": 1.0903, + "step": 30094 + }, + { + "epoch": 2.53, + "learning_rate": 1.2395969763957416e-06, + "loss": 1.0861, + "step": 30095 + }, + { + "epoch": 2.53, + "learning_rate": 1.2391586713357885e-06, + "loss": 1.0804, + "step": 30096 + }, + { + "epoch": 2.53, + "learning_rate": 1.2387204386607087e-06, + "loss": 1.1135, + "step": 30097 + }, + { + "epoch": 2.53, + "learning_rate": 1.2382822783741278e-06, + "loss": 1.039, + "step": 30098 + }, + { + "epoch": 2.53, + "learning_rate": 1.2378441904796612e-06, + "loss": 1.0905, + "step": 30099 + }, + { + "epoch": 2.53, + "learning_rate": 1.2374061749809308e-06, + "loss": 1.0257, + "step": 30100 + }, + { + "epoch": 2.53, + "learning_rate": 1.2369682318815547e-06, + "loss": 1.1305, + "step": 30101 + }, + { + "epoch": 2.53, + "learning_rate": 1.2365303611851509e-06, + "loss": 1.0979, + "step": 30102 + }, + { + "epoch": 2.53, + "learning_rate": 1.236092562895338e-06, + "loss": 1.1047, + "step": 30103 + }, + { + "epoch": 2.53, + "learning_rate": 1.2356548370157362e-06, + "loss": 1.1117, + "step": 30104 + }, + { + "epoch": 2.53, + "learning_rate": 1.2352171835499537e-06, + "loss": 1.0933, + "step": 30105 + }, + { + "epoch": 2.53, + "learning_rate": 1.2347796025016169e-06, + "loss": 1.1585, + "step": 30106 + }, + { + "epoch": 2.53, + "learning_rate": 1.2343420938743344e-06, + "loss": 1.1071, + "step": 30107 + }, + { + "epoch": 2.53, + "learning_rate": 1.2339046576717228e-06, + "loss": 1.1159, + "step": 30108 + }, + { + "epoch": 2.53, + "learning_rate": 1.2334672938973958e-06, + "loss": 1.1721, + "step": 30109 + }, + { + "epoch": 2.54, + "learning_rate": 1.233030002554968e-06, + "loss": 1.1141, + "step": 30110 + }, + { + "epoch": 2.54, + "learning_rate": 1.2325927836480522e-06, + "loss": 1.0639, + "step": 30111 + }, + { + "epoch": 2.54, + "learning_rate": 1.2321556371802602e-06, + "loss": 1.1855, + "step": 30112 + }, + { + "epoch": 2.54, + "learning_rate": 1.2317185631552053e-06, + "loss": 1.1075, + "step": 30113 + }, + { + "epoch": 2.54, + "learning_rate": 1.2312815615764973e-06, + "loss": 1.0571, + "step": 30114 + }, + { + "epoch": 2.54, + "learning_rate": 1.2308446324477497e-06, + "loss": 1.1113, + "step": 30115 + }, + { + "epoch": 2.54, + "learning_rate": 1.2304077757725652e-06, + "loss": 1.1448, + "step": 30116 + }, + { + "epoch": 2.54, + "learning_rate": 1.2299709915545633e-06, + "loss": 1.1242, + "step": 30117 + }, + { + "epoch": 2.54, + "learning_rate": 1.2295342797973464e-06, + "loss": 1.0657, + "step": 30118 + }, + { + "epoch": 2.54, + "learning_rate": 1.229097640504523e-06, + "loss": 1.0865, + "step": 30119 + }, + { + "epoch": 2.54, + "learning_rate": 1.2286610736797021e-06, + "loss": 1.1631, + "step": 30120 + }, + { + "epoch": 2.54, + "learning_rate": 1.2282245793264912e-06, + "loss": 1.1537, + "step": 30121 + }, + { + "epoch": 2.54, + "learning_rate": 1.2277881574484962e-06, + "loss": 1.1132, + "step": 30122 + }, + { + "epoch": 2.54, + "learning_rate": 1.2273518080493219e-06, + "loss": 1.0539, + "step": 30123 + }, + { + "epoch": 2.54, + "learning_rate": 1.2269155311325752e-06, + "loss": 1.0223, + "step": 30124 + }, + { + "epoch": 2.54, + "learning_rate": 1.2264793267018604e-06, + "loss": 1.0423, + "step": 30125 + }, + { + "epoch": 2.54, + "learning_rate": 1.2260431947607832e-06, + "loss": 1.0634, + "step": 30126 + }, + { + "epoch": 2.54, + "learning_rate": 1.2256071353129407e-06, + "loss": 1.1062, + "step": 30127 + }, + { + "epoch": 2.54, + "learning_rate": 1.2251711483619445e-06, + "loss": 1.08, + "step": 30128 + }, + { + "epoch": 2.54, + "learning_rate": 1.2247352339113904e-06, + "loss": 1.039, + "step": 30129 + }, + { + "epoch": 2.54, + "learning_rate": 1.2242993919648826e-06, + "loss": 1.1223, + "step": 30130 + }, + { + "epoch": 2.54, + "learning_rate": 1.2238636225260214e-06, + "loss": 1.1497, + "step": 30131 + }, + { + "epoch": 2.54, + "learning_rate": 1.2234279255984082e-06, + "loss": 1.1309, + "step": 30132 + }, + { + "epoch": 2.54, + "learning_rate": 1.2229923011856416e-06, + "loss": 1.1258, + "step": 30133 + }, + { + "epoch": 2.54, + "learning_rate": 1.2225567492913227e-06, + "loss": 1.0594, + "step": 30134 + }, + { + "epoch": 2.54, + "learning_rate": 1.2221212699190476e-06, + "loss": 1.1146, + "step": 30135 + }, + { + "epoch": 2.54, + "learning_rate": 1.221685863072417e-06, + "loss": 1.1363, + "step": 30136 + }, + { + "epoch": 2.54, + "learning_rate": 1.221250528755029e-06, + "loss": 1.0944, + "step": 30137 + }, + { + "epoch": 2.54, + "learning_rate": 1.2208152669704742e-06, + "loss": 1.1594, + "step": 30138 + }, + { + "epoch": 2.54, + "learning_rate": 1.2203800777223584e-06, + "loss": 1.1528, + "step": 30139 + }, + { + "epoch": 2.54, + "learning_rate": 1.2199449610142688e-06, + "loss": 1.077, + "step": 30140 + }, + { + "epoch": 2.54, + "learning_rate": 1.2195099168498048e-06, + "loss": 1.1253, + "step": 30141 + }, + { + "epoch": 2.54, + "learning_rate": 1.21907494523256e-06, + "loss": 1.2227, + "step": 30142 + }, + { + "epoch": 2.54, + "learning_rate": 1.2186400461661284e-06, + "loss": 1.1149, + "step": 30143 + }, + { + "epoch": 2.54, + "learning_rate": 1.2182052196541028e-06, + "loss": 1.0768, + "step": 30144 + }, + { + "epoch": 2.54, + "learning_rate": 1.2177704657000788e-06, + "loss": 1.1014, + "step": 30145 + }, + { + "epoch": 2.54, + "learning_rate": 1.2173357843076406e-06, + "loss": 1.1863, + "step": 30146 + }, + { + "epoch": 2.54, + "learning_rate": 1.2169011754803882e-06, + "loss": 1.1259, + "step": 30147 + }, + { + "epoch": 2.54, + "learning_rate": 1.2164666392219105e-06, + "loss": 1.1404, + "step": 30148 + }, + { + "epoch": 2.54, + "learning_rate": 1.2160321755357928e-06, + "loss": 1.0777, + "step": 30149 + }, + { + "epoch": 2.54, + "learning_rate": 1.2155977844256329e-06, + "loss": 1.1412, + "step": 30150 + }, + { + "epoch": 2.54, + "learning_rate": 1.2151634658950129e-06, + "loss": 1.1305, + "step": 30151 + }, + { + "epoch": 2.54, + "learning_rate": 1.2147292199475237e-06, + "loss": 1.1723, + "step": 30152 + }, + { + "epoch": 2.54, + "learning_rate": 1.2142950465867543e-06, + "loss": 1.106, + "step": 30153 + }, + { + "epoch": 2.54, + "learning_rate": 1.213860945816292e-06, + "loss": 1.1157, + "step": 30154 + }, + { + "epoch": 2.54, + "learning_rate": 1.2134269176397185e-06, + "loss": 1.1303, + "step": 30155 + }, + { + "epoch": 2.54, + "learning_rate": 1.2129929620606285e-06, + "loss": 1.1319, + "step": 30156 + }, + { + "epoch": 2.54, + "learning_rate": 1.2125590790826003e-06, + "loss": 1.1168, + "step": 30157 + }, + { + "epoch": 2.54, + "learning_rate": 1.2121252687092221e-06, + "loss": 1.1229, + "step": 30158 + }, + { + "epoch": 2.54, + "learning_rate": 1.2116915309440768e-06, + "loss": 1.1074, + "step": 30159 + }, + { + "epoch": 2.54, + "learning_rate": 1.2112578657907481e-06, + "loss": 1.1388, + "step": 30160 + }, + { + "epoch": 2.54, + "learning_rate": 1.2108242732528209e-06, + "loss": 1.1593, + "step": 30161 + }, + { + "epoch": 2.54, + "learning_rate": 1.2103907533338755e-06, + "loss": 1.0784, + "step": 30162 + }, + { + "epoch": 2.54, + "learning_rate": 1.2099573060374947e-06, + "loss": 1.1085, + "step": 30163 + }, + { + "epoch": 2.54, + "learning_rate": 1.2095239313672613e-06, + "loss": 1.1604, + "step": 30164 + }, + { + "epoch": 2.54, + "learning_rate": 1.2090906293267546e-06, + "loss": 1.1796, + "step": 30165 + }, + { + "epoch": 2.54, + "learning_rate": 1.2086573999195516e-06, + "loss": 1.1403, + "step": 30166 + }, + { + "epoch": 2.54, + "learning_rate": 1.2082242431492386e-06, + "loss": 1.029, + "step": 30167 + }, + { + "epoch": 2.54, + "learning_rate": 1.2077911590193891e-06, + "loss": 1.092, + "step": 30168 + }, + { + "epoch": 2.54, + "learning_rate": 1.2073581475335816e-06, + "loss": 1.0859, + "step": 30169 + }, + { + "epoch": 2.54, + "learning_rate": 1.2069252086953965e-06, + "loss": 1.1402, + "step": 30170 + }, + { + "epoch": 2.54, + "learning_rate": 1.2064923425084084e-06, + "loss": 1.0919, + "step": 30171 + }, + { + "epoch": 2.54, + "learning_rate": 1.2060595489761961e-06, + "loss": 1.0649, + "step": 30172 + }, + { + "epoch": 2.54, + "learning_rate": 1.205626828102333e-06, + "loss": 1.1249, + "step": 30173 + }, + { + "epoch": 2.54, + "learning_rate": 1.2051941798903965e-06, + "loss": 1.2221, + "step": 30174 + }, + { + "epoch": 2.54, + "learning_rate": 1.2047616043439593e-06, + "loss": 1.1629, + "step": 30175 + }, + { + "epoch": 2.54, + "learning_rate": 1.2043291014665992e-06, + "loss": 1.0794, + "step": 30176 + }, + { + "epoch": 2.54, + "learning_rate": 1.2038966712618826e-06, + "loss": 1.1383, + "step": 30177 + }, + { + "epoch": 2.54, + "learning_rate": 1.2034643137333923e-06, + "loss": 1.1299, + "step": 30178 + }, + { + "epoch": 2.54, + "learning_rate": 1.203032028884692e-06, + "loss": 1.0535, + "step": 30179 + }, + { + "epoch": 2.54, + "learning_rate": 1.2025998167193564e-06, + "loss": 1.1019, + "step": 30180 + }, + { + "epoch": 2.54, + "learning_rate": 1.2021676772409564e-06, + "loss": 1.0956, + "step": 30181 + }, + { + "epoch": 2.54, + "learning_rate": 1.2017356104530632e-06, + "loss": 1.072, + "step": 30182 + }, + { + "epoch": 2.54, + "learning_rate": 1.2013036163592462e-06, + "loss": 1.1524, + "step": 30183 + }, + { + "epoch": 2.54, + "learning_rate": 1.200871694963074e-06, + "loss": 1.0861, + "step": 30184 + }, + { + "epoch": 2.54, + "learning_rate": 1.2004398462681165e-06, + "loss": 1.1607, + "step": 30185 + }, + { + "epoch": 2.54, + "learning_rate": 1.2000080702779416e-06, + "loss": 1.1736, + "step": 30186 + }, + { + "epoch": 2.54, + "learning_rate": 1.199576366996118e-06, + "loss": 1.0281, + "step": 30187 + }, + { + "epoch": 2.54, + "learning_rate": 1.1991447364262066e-06, + "loss": 1.2307, + "step": 30188 + }, + { + "epoch": 2.54, + "learning_rate": 1.198713178571782e-06, + "loss": 1.0459, + "step": 30189 + }, + { + "epoch": 2.54, + "learning_rate": 1.198281693436405e-06, + "loss": 1.1214, + "step": 30190 + }, + { + "epoch": 2.54, + "learning_rate": 1.1978502810236402e-06, + "loss": 1.0708, + "step": 30191 + }, + { + "epoch": 2.54, + "learning_rate": 1.197418941337054e-06, + "loss": 1.0658, + "step": 30192 + }, + { + "epoch": 2.54, + "learning_rate": 1.1969876743802101e-06, + "loss": 1.0434, + "step": 30193 + }, + { + "epoch": 2.54, + "learning_rate": 1.1965564801566708e-06, + "loss": 1.0894, + "step": 30194 + }, + { + "epoch": 2.54, + "learning_rate": 1.1961253586700017e-06, + "loss": 1.0536, + "step": 30195 + }, + { + "epoch": 2.54, + "learning_rate": 1.1956943099237583e-06, + "loss": 1.0504, + "step": 30196 + }, + { + "epoch": 2.54, + "learning_rate": 1.195263333921508e-06, + "loss": 1.1775, + "step": 30197 + }, + { + "epoch": 2.54, + "learning_rate": 1.1948324306668124e-06, + "loss": 1.1393, + "step": 30198 + }, + { + "epoch": 2.54, + "learning_rate": 1.1944016001632241e-06, + "loss": 0.9801, + "step": 30199 + }, + { + "epoch": 2.54, + "learning_rate": 1.1939708424143137e-06, + "loss": 1.103, + "step": 30200 + }, + { + "epoch": 2.54, + "learning_rate": 1.1935401574236316e-06, + "loss": 1.1335, + "step": 30201 + }, + { + "epoch": 2.54, + "learning_rate": 1.1931095451947395e-06, + "loss": 1.1, + "step": 30202 + }, + { + "epoch": 2.54, + "learning_rate": 1.1926790057311944e-06, + "loss": 1.1055, + "step": 30203 + }, + { + "epoch": 2.54, + "learning_rate": 1.1922485390365556e-06, + "loss": 1.155, + "step": 30204 + }, + { + "epoch": 2.54, + "learning_rate": 1.191818145114375e-06, + "loss": 1.1343, + "step": 30205 + }, + { + "epoch": 2.54, + "learning_rate": 1.1913878239682164e-06, + "loss": 1.0908, + "step": 30206 + }, + { + "epoch": 2.54, + "learning_rate": 1.1909575756016279e-06, + "loss": 1.1399, + "step": 30207 + }, + { + "epoch": 2.54, + "learning_rate": 1.1905274000181665e-06, + "loss": 1.1652, + "step": 30208 + }, + { + "epoch": 2.54, + "learning_rate": 1.1900972972213887e-06, + "loss": 1.1381, + "step": 30209 + }, + { + "epoch": 2.54, + "learning_rate": 1.1896672672148456e-06, + "loss": 1.0509, + "step": 30210 + }, + { + "epoch": 2.54, + "learning_rate": 1.1892373100020915e-06, + "loss": 1.1188, + "step": 30211 + }, + { + "epoch": 2.54, + "learning_rate": 1.1888074255866778e-06, + "loss": 1.0724, + "step": 30212 + }, + { + "epoch": 2.54, + "learning_rate": 1.188377613972158e-06, + "loss": 1.0395, + "step": 30213 + }, + { + "epoch": 2.54, + "learning_rate": 1.187947875162082e-06, + "loss": 1.0808, + "step": 30214 + }, + { + "epoch": 2.54, + "learning_rate": 1.1875182091600034e-06, + "loss": 1.0931, + "step": 30215 + }, + { + "epoch": 2.54, + "learning_rate": 1.187088615969465e-06, + "loss": 1.0691, + "step": 30216 + }, + { + "epoch": 2.54, + "learning_rate": 1.186659095594026e-06, + "loss": 1.0569, + "step": 30217 + }, + { + "epoch": 2.54, + "learning_rate": 1.186229648037227e-06, + "loss": 1.1752, + "step": 30218 + }, + { + "epoch": 2.54, + "learning_rate": 1.1858002733026196e-06, + "loss": 1.1506, + "step": 30219 + }, + { + "epoch": 2.54, + "learning_rate": 1.1853709713937523e-06, + "loss": 1.0753, + "step": 30220 + }, + { + "epoch": 2.54, + "learning_rate": 1.1849417423141695e-06, + "loss": 1.1102, + "step": 30221 + }, + { + "epoch": 2.54, + "learning_rate": 1.18451258606742e-06, + "loss": 1.1481, + "step": 30222 + }, + { + "epoch": 2.54, + "learning_rate": 1.1840835026570497e-06, + "loss": 1.1002, + "step": 30223 + }, + { + "epoch": 2.54, + "learning_rate": 1.1836544920866012e-06, + "loss": 1.1098, + "step": 30224 + }, + { + "epoch": 2.54, + "learning_rate": 1.1832255543596215e-06, + "loss": 1.191, + "step": 30225 + }, + { + "epoch": 2.54, + "learning_rate": 1.1827966894796573e-06, + "loss": 1.1136, + "step": 30226 + }, + { + "epoch": 2.54, + "learning_rate": 1.182367897450243e-06, + "loss": 1.0866, + "step": 30227 + }, + { + "epoch": 2.54, + "learning_rate": 1.1819391782749324e-06, + "loss": 1.0477, + "step": 30228 + }, + { + "epoch": 2.55, + "learning_rate": 1.1815105319572595e-06, + "loss": 1.1321, + "step": 30229 + }, + { + "epoch": 2.55, + "learning_rate": 1.1810819585007693e-06, + "loss": 1.1177, + "step": 30230 + }, + { + "epoch": 2.55, + "learning_rate": 1.1806534579090023e-06, + "loss": 1.1909, + "step": 30231 + }, + { + "epoch": 2.55, + "learning_rate": 1.180225030185499e-06, + "loss": 1.155, + "step": 30232 + }, + { + "epoch": 2.55, + "learning_rate": 1.1797966753337996e-06, + "loss": 1.1681, + "step": 30233 + }, + { + "epoch": 2.55, + "learning_rate": 1.1793683933574419e-06, + "loss": 1.0883, + "step": 30234 + }, + { + "epoch": 2.55, + "learning_rate": 1.1789401842599668e-06, + "loss": 1.135, + "step": 30235 + }, + { + "epoch": 2.55, + "learning_rate": 1.1785120480449098e-06, + "loss": 1.0175, + "step": 30236 + }, + { + "epoch": 2.55, + "learning_rate": 1.178083984715812e-06, + "loss": 1.1028, + "step": 30237 + }, + { + "epoch": 2.55, + "learning_rate": 1.1776559942762035e-06, + "loss": 1.0883, + "step": 30238 + }, + { + "epoch": 2.55, + "learning_rate": 1.1772280767296297e-06, + "loss": 1.0933, + "step": 30239 + }, + { + "epoch": 2.55, + "learning_rate": 1.1768002320796178e-06, + "loss": 1.0554, + "step": 30240 + }, + { + "epoch": 2.55, + "learning_rate": 1.1763724603297066e-06, + "loss": 1.1591, + "step": 30241 + }, + { + "epoch": 2.55, + "learning_rate": 1.1759447614834295e-06, + "loss": 1.0409, + "step": 30242 + }, + { + "epoch": 2.55, + "learning_rate": 1.1755171355443218e-06, + "loss": 1.0725, + "step": 30243 + }, + { + "epoch": 2.55, + "learning_rate": 1.1750895825159158e-06, + "loss": 1.0878, + "step": 30244 + }, + { + "epoch": 2.55, + "learning_rate": 1.1746621024017457e-06, + "loss": 1.1394, + "step": 30245 + }, + { + "epoch": 2.55, + "learning_rate": 1.1742346952053374e-06, + "loss": 1.1781, + "step": 30246 + }, + { + "epoch": 2.55, + "learning_rate": 1.1738073609302293e-06, + "loss": 1.0696, + "step": 30247 + }, + { + "epoch": 2.55, + "learning_rate": 1.1733800995799516e-06, + "loss": 1.0493, + "step": 30248 + }, + { + "epoch": 2.55, + "learning_rate": 1.1729529111580273e-06, + "loss": 1.0758, + "step": 30249 + }, + { + "epoch": 2.55, + "learning_rate": 1.1725257956679959e-06, + "loss": 1.1428, + "step": 30250 + }, + { + "epoch": 2.55, + "learning_rate": 1.1720987531133798e-06, + "loss": 1.1369, + "step": 30251 + }, + { + "epoch": 2.55, + "learning_rate": 1.1716717834977087e-06, + "loss": 1.0831, + "step": 30252 + }, + { + "epoch": 2.55, + "learning_rate": 1.1712448868245109e-06, + "loss": 1.143, + "step": 30253 + }, + { + "epoch": 2.55, + "learning_rate": 1.1708180630973142e-06, + "loss": 1.0616, + "step": 30254 + }, + { + "epoch": 2.55, + "learning_rate": 1.170391312319642e-06, + "loss": 1.1262, + "step": 30255 + }, + { + "epoch": 2.55, + "learning_rate": 1.1699646344950256e-06, + "loss": 1.1223, + "step": 30256 + }, + { + "epoch": 2.55, + "learning_rate": 1.1695380296269865e-06, + "loss": 1.045, + "step": 30257 + }, + { + "epoch": 2.55, + "learning_rate": 1.1691114977190486e-06, + "loss": 1.113, + "step": 30258 + }, + { + "epoch": 2.55, + "learning_rate": 1.1686850387747394e-06, + "loss": 1.1127, + "step": 30259 + }, + { + "epoch": 2.55, + "learning_rate": 1.1682586527975802e-06, + "loss": 1.0893, + "step": 30260 + }, + { + "epoch": 2.55, + "learning_rate": 1.167832339791094e-06, + "loss": 1.1087, + "step": 30261 + }, + { + "epoch": 2.55, + "learning_rate": 1.1674060997588043e-06, + "loss": 1.0737, + "step": 30262 + }, + { + "epoch": 2.55, + "learning_rate": 1.166979932704232e-06, + "loss": 1.1375, + "step": 30263 + }, + { + "epoch": 2.55, + "learning_rate": 1.1665538386308983e-06, + "loss": 1.13, + "step": 30264 + }, + { + "epoch": 2.55, + "learning_rate": 1.1661278175423263e-06, + "loss": 1.1057, + "step": 30265 + }, + { + "epoch": 2.55, + "learning_rate": 1.1657018694420286e-06, + "loss": 1.1438, + "step": 30266 + }, + { + "epoch": 2.55, + "learning_rate": 1.1652759943335345e-06, + "loss": 1.2066, + "step": 30267 + }, + { + "epoch": 2.55, + "learning_rate": 1.1648501922203548e-06, + "loss": 1.0134, + "step": 30268 + }, + { + "epoch": 2.55, + "learning_rate": 1.164424463106011e-06, + "loss": 1.0798, + "step": 30269 + }, + { + "epoch": 2.55, + "learning_rate": 1.16399880699402e-06, + "loss": 1.1004, + "step": 30270 + }, + { + "epoch": 2.55, + "learning_rate": 1.1635732238878982e-06, + "loss": 1.1017, + "step": 30271 + }, + { + "epoch": 2.55, + "learning_rate": 1.163147713791163e-06, + "loss": 1.08, + "step": 30272 + }, + { + "epoch": 2.55, + "learning_rate": 1.1627222767073286e-06, + "loss": 1.1242, + "step": 30273 + }, + { + "epoch": 2.55, + "learning_rate": 1.1622969126399119e-06, + "loss": 1.0993, + "step": 30274 + }, + { + "epoch": 2.55, + "learning_rate": 1.1618716215924264e-06, + "loss": 1.0828, + "step": 30275 + }, + { + "epoch": 2.55, + "learning_rate": 1.1614464035683882e-06, + "loss": 1.1101, + "step": 30276 + }, + { + "epoch": 2.55, + "learning_rate": 1.1610212585713043e-06, + "loss": 1.0895, + "step": 30277 + }, + { + "epoch": 2.55, + "learning_rate": 1.1605961866046954e-06, + "loss": 1.2541, + "step": 30278 + }, + { + "epoch": 2.55, + "learning_rate": 1.1601711876720678e-06, + "loss": 1.1078, + "step": 30279 + }, + { + "epoch": 2.55, + "learning_rate": 1.1597462617769362e-06, + "loss": 1.061, + "step": 30280 + }, + { + "epoch": 2.55, + "learning_rate": 1.159321408922809e-06, + "loss": 1.1092, + "step": 30281 + }, + { + "epoch": 2.55, + "learning_rate": 1.1588966291131975e-06, + "loss": 1.1642, + "step": 30282 + }, + { + "epoch": 2.55, + "learning_rate": 1.1584719223516116e-06, + "loss": 1.1584, + "step": 30283 + }, + { + "epoch": 2.55, + "learning_rate": 1.1580472886415616e-06, + "loss": 1.135, + "step": 30284 + }, + { + "epoch": 2.55, + "learning_rate": 1.1576227279865538e-06, + "loss": 1.1527, + "step": 30285 + }, + { + "epoch": 2.55, + "learning_rate": 1.157198240390096e-06, + "loss": 1.1122, + "step": 30286 + }, + { + "epoch": 2.55, + "learning_rate": 1.1567738258557003e-06, + "loss": 1.1245, + "step": 30287 + }, + { + "epoch": 2.55, + "learning_rate": 1.1563494843868639e-06, + "loss": 1.1536, + "step": 30288 + }, + { + "epoch": 2.55, + "learning_rate": 1.1559252159871037e-06, + "loss": 1.1715, + "step": 30289 + }, + { + "epoch": 2.55, + "learning_rate": 1.1555010206599171e-06, + "loss": 1.1021, + "step": 30290 + }, + { + "epoch": 2.55, + "learning_rate": 1.1550768984088111e-06, + "loss": 1.136, + "step": 30291 + }, + { + "epoch": 2.55, + "learning_rate": 1.154652849237291e-06, + "loss": 1.0864, + "step": 30292 + }, + { + "epoch": 2.55, + "learning_rate": 1.1542288731488604e-06, + "loss": 1.1021, + "step": 30293 + }, + { + "epoch": 2.55, + "learning_rate": 1.153804970147021e-06, + "loss": 1.1565, + "step": 30294 + }, + { + "epoch": 2.55, + "learning_rate": 1.153381140235279e-06, + "loss": 1.1855, + "step": 30295 + }, + { + "epoch": 2.55, + "learning_rate": 1.152957383417128e-06, + "loss": 1.1106, + "step": 30296 + }, + { + "epoch": 2.55, + "learning_rate": 1.1525336996960778e-06, + "loss": 1.0791, + "step": 30297 + }, + { + "epoch": 2.55, + "learning_rate": 1.1521100890756276e-06, + "loss": 1.2904, + "step": 30298 + }, + { + "epoch": 2.55, + "learning_rate": 1.1516865515592712e-06, + "loss": 1.1435, + "step": 30299 + }, + { + "epoch": 2.55, + "learning_rate": 1.151263087150516e-06, + "loss": 1.1971, + "step": 30300 + }, + { + "epoch": 2.55, + "learning_rate": 1.1508396958528556e-06, + "loss": 1.1242, + "step": 30301 + }, + { + "epoch": 2.55, + "learning_rate": 1.1504163776697896e-06, + "loss": 1.2264, + "step": 30302 + }, + { + "epoch": 2.55, + "learning_rate": 1.1499931326048163e-06, + "loss": 1.0424, + "step": 30303 + }, + { + "epoch": 2.55, + "learning_rate": 1.149569960661433e-06, + "loss": 1.072, + "step": 30304 + }, + { + "epoch": 2.55, + "learning_rate": 1.1491468618431323e-06, + "loss": 1.0881, + "step": 30305 + }, + { + "epoch": 2.55, + "learning_rate": 1.1487238361534158e-06, + "loss": 1.0806, + "step": 30306 + }, + { + "epoch": 2.55, + "learning_rate": 1.1483008835957744e-06, + "loss": 1.163, + "step": 30307 + }, + { + "epoch": 2.55, + "learning_rate": 1.1478780041737037e-06, + "loss": 1.0566, + "step": 30308 + }, + { + "epoch": 2.55, + "learning_rate": 1.1474551978906977e-06, + "loss": 1.1266, + "step": 30309 + }, + { + "epoch": 2.55, + "learning_rate": 1.1470324647502506e-06, + "loss": 1.1728, + "step": 30310 + }, + { + "epoch": 2.55, + "learning_rate": 1.1466098047558549e-06, + "loss": 1.1249, + "step": 30311 + }, + { + "epoch": 2.55, + "learning_rate": 1.1461872179110024e-06, + "loss": 1.0516, + "step": 30312 + }, + { + "epoch": 2.55, + "learning_rate": 1.1457647042191844e-06, + "loss": 1.1544, + "step": 30313 + }, + { + "epoch": 2.55, + "learning_rate": 1.1453422636838918e-06, + "loss": 1.0694, + "step": 30314 + }, + { + "epoch": 2.55, + "learning_rate": 1.1449198963086173e-06, + "loss": 1.1152, + "step": 30315 + }, + { + "epoch": 2.55, + "learning_rate": 1.1444976020968447e-06, + "loss": 1.1229, + "step": 30316 + }, + { + "epoch": 2.55, + "learning_rate": 1.144075381052071e-06, + "loss": 1.1167, + "step": 30317 + }, + { + "epoch": 2.55, + "learning_rate": 1.1436532331777793e-06, + "loss": 1.0715, + "step": 30318 + }, + { + "epoch": 2.55, + "learning_rate": 1.143231158477458e-06, + "loss": 1.1854, + "step": 30319 + }, + { + "epoch": 2.55, + "learning_rate": 1.1428091569545962e-06, + "loss": 1.1527, + "step": 30320 + }, + { + "epoch": 2.55, + "learning_rate": 1.1423872286126792e-06, + "loss": 1.1619, + "step": 30321 + }, + { + "epoch": 2.55, + "learning_rate": 1.141965373455194e-06, + "loss": 1.2222, + "step": 30322 + }, + { + "epoch": 2.55, + "learning_rate": 1.1415435914856265e-06, + "loss": 1.051, + "step": 30323 + }, + { + "epoch": 2.55, + "learning_rate": 1.1411218827074599e-06, + "loss": 1.1754, + "step": 30324 + }, + { + "epoch": 2.55, + "learning_rate": 1.1407002471241802e-06, + "loss": 1.13, + "step": 30325 + }, + { + "epoch": 2.55, + "learning_rate": 1.1402786847392722e-06, + "loss": 1.1332, + "step": 30326 + }, + { + "epoch": 2.55, + "learning_rate": 1.1398571955562132e-06, + "loss": 1.0914, + "step": 30327 + }, + { + "epoch": 2.55, + "learning_rate": 1.1394357795784927e-06, + "loss": 1.1524, + "step": 30328 + }, + { + "epoch": 2.55, + "learning_rate": 1.139014436809589e-06, + "loss": 1.1899, + "step": 30329 + }, + { + "epoch": 2.55, + "learning_rate": 1.1385931672529826e-06, + "loss": 1.164, + "step": 30330 + }, + { + "epoch": 2.55, + "learning_rate": 1.1381719709121564e-06, + "loss": 1.0878, + "step": 30331 + }, + { + "epoch": 2.55, + "learning_rate": 1.1377508477905897e-06, + "loss": 1.0227, + "step": 30332 + }, + { + "epoch": 2.55, + "learning_rate": 1.1373297978917608e-06, + "loss": 1.0894, + "step": 30333 + }, + { + "epoch": 2.55, + "learning_rate": 1.1369088212191503e-06, + "loss": 1.166, + "step": 30334 + }, + { + "epoch": 2.55, + "learning_rate": 1.1364879177762355e-06, + "loss": 1.0885, + "step": 30335 + }, + { + "epoch": 2.55, + "learning_rate": 1.1360670875664947e-06, + "loss": 1.1609, + "step": 30336 + }, + { + "epoch": 2.55, + "learning_rate": 1.1356463305934051e-06, + "loss": 1.0633, + "step": 30337 + }, + { + "epoch": 2.55, + "learning_rate": 1.1352256468604394e-06, + "loss": 1.0205, + "step": 30338 + }, + { + "epoch": 2.55, + "learning_rate": 1.1348050363710805e-06, + "loss": 1.1746, + "step": 30339 + }, + { + "epoch": 2.55, + "learning_rate": 1.1343844991287978e-06, + "loss": 1.0835, + "step": 30340 + }, + { + "epoch": 2.55, + "learning_rate": 1.1339640351370672e-06, + "loss": 1.0721, + "step": 30341 + }, + { + "epoch": 2.55, + "learning_rate": 1.1335436443993641e-06, + "loss": 1.1106, + "step": 30342 + }, + { + "epoch": 2.55, + "learning_rate": 1.1331233269191598e-06, + "loss": 1.05, + "step": 30343 + }, + { + "epoch": 2.55, + "learning_rate": 1.1327030826999287e-06, + "loss": 1.1448, + "step": 30344 + }, + { + "epoch": 2.55, + "learning_rate": 1.132282911745145e-06, + "loss": 1.1861, + "step": 30345 + }, + { + "epoch": 2.55, + "learning_rate": 1.1318628140582743e-06, + "loss": 1.0782, + "step": 30346 + }, + { + "epoch": 2.55, + "learning_rate": 1.1314427896427927e-06, + "loss": 1.0883, + "step": 30347 + }, + { + "epoch": 2.56, + "learning_rate": 1.1310228385021704e-06, + "loss": 1.1005, + "step": 30348 + }, + { + "epoch": 2.56, + "learning_rate": 1.1306029606398728e-06, + "loss": 1.0572, + "step": 30349 + }, + { + "epoch": 2.56, + "learning_rate": 1.1301831560593756e-06, + "loss": 1.1212, + "step": 30350 + }, + { + "epoch": 2.56, + "learning_rate": 1.129763424764142e-06, + "loss": 1.0947, + "step": 30351 + }, + { + "epoch": 2.56, + "learning_rate": 1.1293437667576423e-06, + "loss": 1.0001, + "step": 30352 + }, + { + "epoch": 2.56, + "learning_rate": 1.1289241820433428e-06, + "loss": 1.0671, + "step": 30353 + }, + { + "epoch": 2.56, + "learning_rate": 1.1285046706247127e-06, + "loss": 1.1073, + "step": 30354 + }, + { + "epoch": 2.56, + "learning_rate": 1.1280852325052126e-06, + "loss": 1.1074, + "step": 30355 + }, + { + "epoch": 2.56, + "learning_rate": 1.1276658676883157e-06, + "loss": 1.1229, + "step": 30356 + }, + { + "epoch": 2.56, + "learning_rate": 1.1272465761774809e-06, + "loss": 1.1055, + "step": 30357 + }, + { + "epoch": 2.56, + "learning_rate": 1.1268273579761724e-06, + "loss": 1.0608, + "step": 30358 + }, + { + "epoch": 2.56, + "learning_rate": 1.1264082130878606e-06, + "loss": 1.1154, + "step": 30359 + }, + { + "epoch": 2.56, + "learning_rate": 1.1259891415160019e-06, + "loss": 1.1217, + "step": 30360 + }, + { + "epoch": 2.56, + "learning_rate": 1.1255701432640599e-06, + "loss": 1.2261, + "step": 30361 + }, + { + "epoch": 2.56, + "learning_rate": 1.1251512183354984e-06, + "loss": 1.0446, + "step": 30362 + }, + { + "epoch": 2.56, + "learning_rate": 1.1247323667337784e-06, + "loss": 1.1592, + "step": 30363 + }, + { + "epoch": 2.56, + "learning_rate": 1.124313588462359e-06, + "loss": 1.0473, + "step": 30364 + }, + { + "epoch": 2.56, + "learning_rate": 1.1238948835247032e-06, + "loss": 1.0316, + "step": 30365 + }, + { + "epoch": 2.56, + "learning_rate": 1.1234762519242659e-06, + "loss": 1.1207, + "step": 30366 + }, + { + "epoch": 2.56, + "learning_rate": 1.1230576936645122e-06, + "loss": 1.1342, + "step": 30367 + }, + { + "epoch": 2.56, + "learning_rate": 1.1226392087488936e-06, + "loss": 1.169, + "step": 30368 + }, + { + "epoch": 2.56, + "learning_rate": 1.1222207971808718e-06, + "loss": 1.1713, + "step": 30369 + }, + { + "epoch": 2.56, + "learning_rate": 1.1218024589639032e-06, + "loss": 1.1632, + "step": 30370 + }, + { + "epoch": 2.56, + "learning_rate": 1.1213841941014437e-06, + "loss": 1.0627, + "step": 30371 + }, + { + "epoch": 2.56, + "learning_rate": 1.1209660025969503e-06, + "loss": 1.1384, + "step": 30372 + }, + { + "epoch": 2.56, + "learning_rate": 1.1205478844538764e-06, + "loss": 1.1562, + "step": 30373 + }, + { + "epoch": 2.56, + "learning_rate": 1.1201298396756788e-06, + "loss": 1.0569, + "step": 30374 + }, + { + "epoch": 2.56, + "learning_rate": 1.1197118682658092e-06, + "loss": 1.078, + "step": 30375 + }, + { + "epoch": 2.56, + "learning_rate": 1.119293970227725e-06, + "loss": 1.0905, + "step": 30376 + }, + { + "epoch": 2.56, + "learning_rate": 1.1188761455648723e-06, + "loss": 1.1412, + "step": 30377 + }, + { + "epoch": 2.56, + "learning_rate": 1.1184583942807103e-06, + "loss": 1.1277, + "step": 30378 + }, + { + "epoch": 2.56, + "learning_rate": 1.1180407163786866e-06, + "loss": 1.059, + "step": 30379 + }, + { + "epoch": 2.56, + "learning_rate": 1.1176231118622526e-06, + "loss": 1.134, + "step": 30380 + }, + { + "epoch": 2.56, + "learning_rate": 1.117205580734859e-06, + "loss": 1.1585, + "step": 30381 + }, + { + "epoch": 2.56, + "learning_rate": 1.1167881229999556e-06, + "loss": 1.1512, + "step": 30382 + }, + { + "epoch": 2.56, + "learning_rate": 1.1163707386609923e-06, + "loss": 1.117, + "step": 30383 + }, + { + "epoch": 2.56, + "learning_rate": 1.1159534277214168e-06, + "loss": 1.131, + "step": 30384 + }, + { + "epoch": 2.56, + "learning_rate": 1.1155361901846773e-06, + "loss": 1.1328, + "step": 30385 + }, + { + "epoch": 2.56, + "learning_rate": 1.115119026054221e-06, + "loss": 1.093, + "step": 30386 + }, + { + "epoch": 2.56, + "learning_rate": 1.1147019353334977e-06, + "loss": 1.0324, + "step": 30387 + }, + { + "epoch": 2.56, + "learning_rate": 1.1142849180259452e-06, + "loss": 1.1692, + "step": 30388 + }, + { + "epoch": 2.56, + "learning_rate": 1.11386797413502e-06, + "loss": 1.05, + "step": 30389 + }, + { + "epoch": 2.56, + "learning_rate": 1.1134511036641583e-06, + "loss": 1.1175, + "step": 30390 + }, + { + "epoch": 2.56, + "learning_rate": 1.113034306616808e-06, + "loss": 1.169, + "step": 30391 + }, + { + "epoch": 2.56, + "learning_rate": 1.1126175829964125e-06, + "loss": 1.1877, + "step": 30392 + }, + { + "epoch": 2.56, + "learning_rate": 1.1122009328064153e-06, + "loss": 1.1005, + "step": 30393 + }, + { + "epoch": 2.56, + "learning_rate": 1.1117843560502584e-06, + "loss": 1.0057, + "step": 30394 + }, + { + "epoch": 2.56, + "learning_rate": 1.1113678527313853e-06, + "loss": 1.1266, + "step": 30395 + }, + { + "epoch": 2.56, + "learning_rate": 1.1109514228532326e-06, + "loss": 1.0439, + "step": 30396 + }, + { + "epoch": 2.56, + "learning_rate": 1.1105350664192448e-06, + "loss": 1.175, + "step": 30397 + }, + { + "epoch": 2.56, + "learning_rate": 1.1101187834328641e-06, + "loss": 1.0576, + "step": 30398 + }, + { + "epoch": 2.56, + "learning_rate": 1.1097025738975232e-06, + "loss": 1.0731, + "step": 30399 + }, + { + "epoch": 2.56, + "learning_rate": 1.1092864378166678e-06, + "loss": 1.1401, + "step": 30400 + }, + { + "epoch": 2.56, + "learning_rate": 1.1088703751937324e-06, + "loss": 1.0814, + "step": 30401 + }, + { + "epoch": 2.56, + "learning_rate": 1.108454386032155e-06, + "loss": 1.0531, + "step": 30402 + }, + { + "epoch": 2.56, + "learning_rate": 1.108038470335373e-06, + "loss": 1.1753, + "step": 30403 + }, + { + "epoch": 2.56, + "learning_rate": 1.1076226281068258e-06, + "loss": 1.1022, + "step": 30404 + }, + { + "epoch": 2.56, + "learning_rate": 1.1072068593499419e-06, + "loss": 1.1863, + "step": 30405 + }, + { + "epoch": 2.56, + "learning_rate": 1.106791164068165e-06, + "loss": 1.0145, + "step": 30406 + }, + { + "epoch": 2.56, + "learning_rate": 1.1063755422649236e-06, + "loss": 1.1382, + "step": 30407 + }, + { + "epoch": 2.56, + "learning_rate": 1.1059599939436516e-06, + "loss": 1.1017, + "step": 30408 + }, + { + "epoch": 2.56, + "learning_rate": 1.1055445191077884e-06, + "loss": 1.1282, + "step": 30409 + }, + { + "epoch": 2.56, + "learning_rate": 1.1051291177607614e-06, + "loss": 1.1028, + "step": 30410 + }, + { + "epoch": 2.56, + "learning_rate": 1.1047137899060046e-06, + "loss": 1.0458, + "step": 30411 + }, + { + "epoch": 2.56, + "learning_rate": 1.1042985355469493e-06, + "loss": 1.1006, + "step": 30412 + }, + { + "epoch": 2.56, + "learning_rate": 1.1038833546870254e-06, + "loss": 1.1278, + "step": 30413 + }, + { + "epoch": 2.56, + "learning_rate": 1.1034682473296654e-06, + "loss": 1.064, + "step": 30414 + }, + { + "epoch": 2.56, + "learning_rate": 1.1030532134783e-06, + "loss": 1.0658, + "step": 30415 + }, + { + "epoch": 2.56, + "learning_rate": 1.1026382531363522e-06, + "loss": 1.1218, + "step": 30416 + }, + { + "epoch": 2.56, + "learning_rate": 1.102223366307259e-06, + "loss": 1.0996, + "step": 30417 + }, + { + "epoch": 2.56, + "learning_rate": 1.1018085529944411e-06, + "loss": 1.0784, + "step": 30418 + }, + { + "epoch": 2.56, + "learning_rate": 1.1013938132013292e-06, + "loss": 1.1862, + "step": 30419 + }, + { + "epoch": 2.56, + "learning_rate": 1.100979146931349e-06, + "loss": 1.0993, + "step": 30420 + }, + { + "epoch": 2.56, + "learning_rate": 1.1005645541879285e-06, + "loss": 1.1541, + "step": 30421 + }, + { + "epoch": 2.56, + "learning_rate": 1.100150034974491e-06, + "loss": 1.2449, + "step": 30422 + }, + { + "epoch": 2.56, + "learning_rate": 1.0997355892944618e-06, + "loss": 1.1308, + "step": 30423 + }, + { + "epoch": 2.56, + "learning_rate": 1.0993212171512657e-06, + "loss": 1.2258, + "step": 30424 + }, + { + "epoch": 2.56, + "learning_rate": 1.098906918548327e-06, + "loss": 1.0965, + "step": 30425 + }, + { + "epoch": 2.56, + "learning_rate": 1.0984926934890694e-06, + "loss": 1.0666, + "step": 30426 + }, + { + "epoch": 2.56, + "learning_rate": 1.0980785419769103e-06, + "loss": 1.0961, + "step": 30427 + }, + { + "epoch": 2.56, + "learning_rate": 1.097664464015279e-06, + "loss": 1.1404, + "step": 30428 + }, + { + "epoch": 2.56, + "learning_rate": 1.0972504596075916e-06, + "loss": 1.0383, + "step": 30429 + }, + { + "epoch": 2.56, + "learning_rate": 1.0968365287572703e-06, + "loss": 1.115, + "step": 30430 + }, + { + "epoch": 2.56, + "learning_rate": 1.0964226714677339e-06, + "loss": 1.1261, + "step": 30431 + }, + { + "epoch": 2.56, + "learning_rate": 1.0960088877424035e-06, + "loss": 1.0364, + "step": 30432 + }, + { + "epoch": 2.56, + "learning_rate": 1.0955951775846974e-06, + "loss": 1.1672, + "step": 30433 + }, + { + "epoch": 2.56, + "learning_rate": 1.0951815409980338e-06, + "loss": 1.1347, + "step": 30434 + }, + { + "epoch": 2.56, + "learning_rate": 1.09476797798583e-06, + "loss": 1.1236, + "step": 30435 + }, + { + "epoch": 2.56, + "learning_rate": 1.0943544885515033e-06, + "loss": 1.1322, + "step": 30436 + }, + { + "epoch": 2.56, + "learning_rate": 1.0939410726984722e-06, + "loss": 1.1481, + "step": 30437 + }, + { + "epoch": 2.56, + "learning_rate": 1.0935277304301451e-06, + "loss": 1.1512, + "step": 30438 + }, + { + "epoch": 2.56, + "learning_rate": 1.093114461749948e-06, + "loss": 1.125, + "step": 30439 + }, + { + "epoch": 2.56, + "learning_rate": 1.0927012666612867e-06, + "loss": 1.0706, + "step": 30440 + }, + { + "epoch": 2.56, + "learning_rate": 1.0922881451675783e-06, + "loss": 1.0812, + "step": 30441 + }, + { + "epoch": 2.56, + "learning_rate": 1.0918750972722369e-06, + "loss": 1.1816, + "step": 30442 + }, + { + "epoch": 2.56, + "learning_rate": 1.0914621229786738e-06, + "loss": 1.1492, + "step": 30443 + }, + { + "epoch": 2.56, + "learning_rate": 1.0910492222903013e-06, + "loss": 0.9781, + "step": 30444 + }, + { + "epoch": 2.56, + "learning_rate": 1.0906363952105336e-06, + "loss": 1.1821, + "step": 30445 + }, + { + "epoch": 2.56, + "learning_rate": 1.0902236417427757e-06, + "loss": 1.0344, + "step": 30446 + }, + { + "epoch": 2.56, + "learning_rate": 1.089810961890444e-06, + "loss": 1.053, + "step": 30447 + }, + { + "epoch": 2.56, + "learning_rate": 1.0893983556569464e-06, + "loss": 1.1567, + "step": 30448 + }, + { + "epoch": 2.56, + "learning_rate": 1.0889858230456885e-06, + "loss": 1.0781, + "step": 30449 + }, + { + "epoch": 2.56, + "learning_rate": 1.088573364060085e-06, + "loss": 1.0885, + "step": 30450 + }, + { + "epoch": 2.56, + "learning_rate": 1.0881609787035385e-06, + "loss": 1.1819, + "step": 30451 + }, + { + "epoch": 2.56, + "learning_rate": 1.087748666979459e-06, + "loss": 1.1471, + "step": 30452 + }, + { + "epoch": 2.56, + "learning_rate": 1.0873364288912514e-06, + "loss": 1.2025, + "step": 30453 + }, + { + "epoch": 2.56, + "learning_rate": 1.0869242644423228e-06, + "loss": 1.1835, + "step": 30454 + }, + { + "epoch": 2.56, + "learning_rate": 1.0865121736360784e-06, + "loss": 1.1019, + "step": 30455 + }, + { + "epoch": 2.56, + "learning_rate": 1.0861001564759256e-06, + "loss": 1.1942, + "step": 30456 + }, + { + "epoch": 2.56, + "learning_rate": 1.0856882129652612e-06, + "loss": 1.0758, + "step": 30457 + }, + { + "epoch": 2.56, + "learning_rate": 1.0852763431074964e-06, + "loss": 1.0815, + "step": 30458 + }, + { + "epoch": 2.56, + "learning_rate": 1.0848645469060338e-06, + "loss": 1.2263, + "step": 30459 + }, + { + "epoch": 2.56, + "learning_rate": 1.0844528243642694e-06, + "loss": 1.0821, + "step": 30460 + }, + { + "epoch": 2.56, + "learning_rate": 1.0840411754856128e-06, + "loss": 1.1479, + "step": 30461 + }, + { + "epoch": 2.56, + "learning_rate": 1.0836296002734592e-06, + "loss": 1.1325, + "step": 30462 + }, + { + "epoch": 2.56, + "learning_rate": 1.0832180987312112e-06, + "loss": 1.0902, + "step": 30463 + }, + { + "epoch": 2.56, + "learning_rate": 1.0828066708622697e-06, + "loss": 1.0885, + "step": 30464 + }, + { + "epoch": 2.56, + "learning_rate": 1.0823953166700352e-06, + "loss": 1.1802, + "step": 30465 + }, + { + "epoch": 2.56, + "learning_rate": 1.0819840361579014e-06, + "loss": 1.1088, + "step": 30466 + }, + { + "epoch": 2.57, + "learning_rate": 1.0815728293292726e-06, + "loss": 1.1541, + "step": 30467 + }, + { + "epoch": 2.57, + "learning_rate": 1.0811616961875414e-06, + "loss": 1.0819, + "step": 30468 + }, + { + "epoch": 2.57, + "learning_rate": 1.0807506367361065e-06, + "loss": 1.1474, + "step": 30469 + }, + { + "epoch": 2.57, + "learning_rate": 1.0803396509783649e-06, + "loss": 1.1092, + "step": 30470 + }, + { + "epoch": 2.57, + "learning_rate": 1.0799287389177116e-06, + "loss": 1.0798, + "step": 30471 + }, + { + "epoch": 2.57, + "learning_rate": 1.0795179005575418e-06, + "loss": 1.1554, + "step": 30472 + }, + { + "epoch": 2.57, + "learning_rate": 1.0791071359012506e-06, + "loss": 1.2386, + "step": 30473 + }, + { + "epoch": 2.57, + "learning_rate": 1.0786964449522297e-06, + "loss": 1.1531, + "step": 30474 + }, + { + "epoch": 2.57, + "learning_rate": 1.0782858277138752e-06, + "loss": 1.0926, + "step": 30475 + }, + { + "epoch": 2.57, + "learning_rate": 1.0778752841895802e-06, + "loss": 1.0566, + "step": 30476 + }, + { + "epoch": 2.57, + "learning_rate": 1.0774648143827315e-06, + "loss": 1.0298, + "step": 30477 + }, + { + "epoch": 2.57, + "learning_rate": 1.077054418296728e-06, + "loss": 1.0599, + "step": 30478 + }, + { + "epoch": 2.57, + "learning_rate": 1.0766440959349545e-06, + "loss": 1.0262, + "step": 30479 + }, + { + "epoch": 2.57, + "learning_rate": 1.0762338473008028e-06, + "loss": 1.0711, + "step": 30480 + }, + { + "epoch": 2.57, + "learning_rate": 1.0758236723976634e-06, + "loss": 1.0744, + "step": 30481 + }, + { + "epoch": 2.57, + "learning_rate": 1.075413571228925e-06, + "loss": 1.0616, + "step": 30482 + }, + { + "epoch": 2.57, + "learning_rate": 1.0750035437979755e-06, + "loss": 1.1377, + "step": 30483 + }, + { + "epoch": 2.57, + "learning_rate": 1.074593590108204e-06, + "loss": 1.1291, + "step": 30484 + }, + { + "epoch": 2.57, + "learning_rate": 1.074183710162996e-06, + "loss": 1.1423, + "step": 30485 + }, + { + "epoch": 2.57, + "learning_rate": 1.0737739039657392e-06, + "loss": 1.2369, + "step": 30486 + }, + { + "epoch": 2.57, + "learning_rate": 1.0733641715198206e-06, + "loss": 1.1602, + "step": 30487 + }, + { + "epoch": 2.57, + "learning_rate": 1.07295451282862e-06, + "loss": 1.0977, + "step": 30488 + }, + { + "epoch": 2.57, + "learning_rate": 1.0725449278955314e-06, + "loss": 1.0686, + "step": 30489 + }, + { + "epoch": 2.57, + "learning_rate": 1.0721354167239307e-06, + "loss": 1.0971, + "step": 30490 + }, + { + "epoch": 2.57, + "learning_rate": 1.071725979317204e-06, + "loss": 1.1647, + "step": 30491 + }, + { + "epoch": 2.57, + "learning_rate": 1.0713166156787358e-06, + "loss": 1.0865, + "step": 30492 + }, + { + "epoch": 2.57, + "learning_rate": 1.0709073258119073e-06, + "loss": 1.0472, + "step": 30493 + }, + { + "epoch": 2.57, + "learning_rate": 1.0704981097200995e-06, + "loss": 1.11, + "step": 30494 + }, + { + "epoch": 2.57, + "learning_rate": 1.0700889674066961e-06, + "loss": 1.0861, + "step": 30495 + }, + { + "epoch": 2.57, + "learning_rate": 1.0696798988750723e-06, + "loss": 1.1389, + "step": 30496 + }, + { + "epoch": 2.57, + "learning_rate": 1.069270904128613e-06, + "loss": 1.1216, + "step": 30497 + }, + { + "epoch": 2.57, + "learning_rate": 1.068861983170697e-06, + "loss": 1.1343, + "step": 30498 + }, + { + "epoch": 2.57, + "learning_rate": 1.068453136004698e-06, + "loss": 1.0907, + "step": 30499 + }, + { + "epoch": 2.57, + "learning_rate": 1.068044362634002e-06, + "loss": 1.093, + "step": 30500 + }, + { + "epoch": 2.57, + "learning_rate": 1.06763566306198e-06, + "loss": 1.0837, + "step": 30501 + }, + { + "epoch": 2.57, + "learning_rate": 1.0672270372920113e-06, + "loss": 1.0939, + "step": 30502 + }, + { + "epoch": 2.57, + "learning_rate": 1.066818485327471e-06, + "loss": 1.1874, + "step": 30503 + }, + { + "epoch": 2.57, + "learning_rate": 1.0664100071717354e-06, + "loss": 1.1052, + "step": 30504 + }, + { + "epoch": 2.57, + "learning_rate": 1.0660016028281794e-06, + "loss": 1.0489, + "step": 30505 + }, + { + "epoch": 2.57, + "learning_rate": 1.0655932723001793e-06, + "loss": 1.1425, + "step": 30506 + }, + { + "epoch": 2.57, + "learning_rate": 1.0651850155911036e-06, + "loss": 1.1171, + "step": 30507 + }, + { + "epoch": 2.57, + "learning_rate": 1.0647768327043295e-06, + "loss": 1.0357, + "step": 30508 + }, + { + "epoch": 2.57, + "learning_rate": 1.064368723643232e-06, + "loss": 1.1255, + "step": 30509 + }, + { + "epoch": 2.57, + "learning_rate": 1.0639606884111752e-06, + "loss": 1.096, + "step": 30510 + }, + { + "epoch": 2.57, + "learning_rate": 1.0635527270115398e-06, + "loss": 1.1643, + "step": 30511 + }, + { + "epoch": 2.57, + "learning_rate": 1.0631448394476895e-06, + "loss": 1.1744, + "step": 30512 + }, + { + "epoch": 2.57, + "learning_rate": 1.0627370257229974e-06, + "loss": 1.112, + "step": 30513 + }, + { + "epoch": 2.57, + "learning_rate": 1.0623292858408308e-06, + "loss": 1.0718, + "step": 30514 + }, + { + "epoch": 2.57, + "learning_rate": 1.0619216198045635e-06, + "loss": 1.1693, + "step": 30515 + }, + { + "epoch": 2.57, + "learning_rate": 1.061514027617555e-06, + "loss": 1.0623, + "step": 30516 + }, + { + "epoch": 2.57, + "learning_rate": 1.061106509283183e-06, + "loss": 1.1993, + "step": 30517 + }, + { + "epoch": 2.57, + "learning_rate": 1.0606990648048076e-06, + "loss": 1.0866, + "step": 30518 + }, + { + "epoch": 2.57, + "learning_rate": 1.0602916941857977e-06, + "loss": 1.115, + "step": 30519 + }, + { + "epoch": 2.57, + "learning_rate": 1.0598843974295192e-06, + "loss": 1.1555, + "step": 30520 + }, + { + "epoch": 2.57, + "learning_rate": 1.0594771745393361e-06, + "loss": 1.1133, + "step": 30521 + }, + { + "epoch": 2.57, + "learning_rate": 1.0590700255186147e-06, + "loss": 1.0746, + "step": 30522 + }, + { + "epoch": 2.57, + "learning_rate": 1.058662950370718e-06, + "loss": 1.0462, + "step": 30523 + }, + { + "epoch": 2.57, + "learning_rate": 1.0582559490990095e-06, + "loss": 1.0788, + "step": 30524 + }, + { + "epoch": 2.57, + "learning_rate": 1.0578490217068527e-06, + "loss": 1.145, + "step": 30525 + }, + { + "epoch": 2.57, + "learning_rate": 1.05744216819761e-06, + "loss": 1.2197, + "step": 30526 + }, + { + "epoch": 2.57, + "learning_rate": 1.0570353885746386e-06, + "loss": 1.1567, + "step": 30527 + }, + { + "epoch": 2.57, + "learning_rate": 1.0566286828413075e-06, + "loss": 1.1739, + "step": 30528 + }, + { + "epoch": 2.57, + "learning_rate": 1.0562220510009703e-06, + "loss": 1.1787, + "step": 30529 + }, + { + "epoch": 2.57, + "learning_rate": 1.0558154930569887e-06, + "loss": 1.0553, + "step": 30530 + }, + { + "epoch": 2.57, + "learning_rate": 1.0554090090127212e-06, + "loss": 1.2032, + "step": 30531 + }, + { + "epoch": 2.57, + "learning_rate": 1.0550025988715285e-06, + "loss": 1.1426, + "step": 30532 + }, + { + "epoch": 2.57, + "learning_rate": 1.0545962626367657e-06, + "loss": 1.1865, + "step": 30533 + }, + { + "epoch": 2.57, + "learning_rate": 1.0541900003117921e-06, + "loss": 1.0997, + "step": 30534 + }, + { + "epoch": 2.57, + "learning_rate": 1.0537838118999645e-06, + "loss": 1.1205, + "step": 30535 + }, + { + "epoch": 2.57, + "learning_rate": 1.0533776974046372e-06, + "loss": 1.0955, + "step": 30536 + }, + { + "epoch": 2.57, + "learning_rate": 1.0529716568291692e-06, + "loss": 1.1236, + "step": 30537 + }, + { + "epoch": 2.57, + "learning_rate": 1.0525656901769076e-06, + "loss": 1.1827, + "step": 30538 + }, + { + "epoch": 2.57, + "learning_rate": 1.0521597974512177e-06, + "loss": 1.1067, + "step": 30539 + }, + { + "epoch": 2.57, + "learning_rate": 1.0517539786554442e-06, + "loss": 1.0849, + "step": 30540 + }, + { + "epoch": 2.57, + "learning_rate": 1.0513482337929426e-06, + "loss": 1.0699, + "step": 30541 + }, + { + "epoch": 2.57, + "learning_rate": 1.0509425628670665e-06, + "loss": 1.0967, + "step": 30542 + }, + { + "epoch": 2.57, + "learning_rate": 1.050536965881167e-06, + "loss": 1.1104, + "step": 30543 + }, + { + "epoch": 2.57, + "learning_rate": 1.0501314428385955e-06, + "loss": 1.1443, + "step": 30544 + }, + { + "epoch": 2.57, + "learning_rate": 1.049725993742704e-06, + "loss": 1.1897, + "step": 30545 + }, + { + "epoch": 2.57, + "learning_rate": 1.0493206185968362e-06, + "loss": 1.1108, + "step": 30546 + }, + { + "epoch": 2.57, + "learning_rate": 1.0489153174043498e-06, + "loss": 1.1014, + "step": 30547 + }, + { + "epoch": 2.57, + "learning_rate": 1.0485100901685907e-06, + "loss": 1.0703, + "step": 30548 + }, + { + "epoch": 2.57, + "learning_rate": 1.048104936892902e-06, + "loss": 1.215, + "step": 30549 + }, + { + "epoch": 2.57, + "learning_rate": 1.0476998575806385e-06, + "loss": 1.1369, + "step": 30550 + }, + { + "epoch": 2.57, + "learning_rate": 1.0472948522351434e-06, + "loss": 0.9908, + "step": 30551 + }, + { + "epoch": 2.57, + "learning_rate": 1.0468899208597627e-06, + "loss": 1.1645, + "step": 30552 + }, + { + "epoch": 2.57, + "learning_rate": 1.0464850634578427e-06, + "loss": 1.1975, + "step": 30553 + }, + { + "epoch": 2.57, + "learning_rate": 1.0460802800327286e-06, + "loss": 1.0486, + "step": 30554 + }, + { + "epoch": 2.57, + "learning_rate": 1.0456755705877652e-06, + "loss": 1.1897, + "step": 30555 + }, + { + "epoch": 2.57, + "learning_rate": 1.045270935126298e-06, + "loss": 1.0897, + "step": 30556 + }, + { + "epoch": 2.57, + "learning_rate": 1.044866373651664e-06, + "loss": 1.2056, + "step": 30557 + }, + { + "epoch": 2.57, + "learning_rate": 1.0444618861672119e-06, + "loss": 1.2931, + "step": 30558 + }, + { + "epoch": 2.57, + "learning_rate": 1.0440574726762842e-06, + "loss": 1.1379, + "step": 30559 + }, + { + "epoch": 2.57, + "learning_rate": 1.0436531331822152e-06, + "loss": 1.1556, + "step": 30560 + }, + { + "epoch": 2.57, + "learning_rate": 1.0432488676883557e-06, + "loss": 1.0655, + "step": 30561 + }, + { + "epoch": 2.57, + "learning_rate": 1.042844676198037e-06, + "loss": 1.1222, + "step": 30562 + }, + { + "epoch": 2.57, + "learning_rate": 1.0424405587146035e-06, + "loss": 1.0852, + "step": 30563 + }, + { + "epoch": 2.57, + "learning_rate": 1.0420365152413925e-06, + "loss": 1.1187, + "step": 30564 + }, + { + "epoch": 2.57, + "learning_rate": 1.0416325457817444e-06, + "loss": 1.0076, + "step": 30565 + }, + { + "epoch": 2.57, + "learning_rate": 1.0412286503389924e-06, + "loss": 1.0538, + "step": 30566 + }, + { + "epoch": 2.57, + "learning_rate": 1.040824828916479e-06, + "loss": 1.0967, + "step": 30567 + }, + { + "epoch": 2.57, + "learning_rate": 1.0404210815175375e-06, + "loss": 1.1154, + "step": 30568 + }, + { + "epoch": 2.57, + "learning_rate": 1.0400174081455027e-06, + "loss": 1.1158, + "step": 30569 + }, + { + "epoch": 2.57, + "learning_rate": 1.0396138088037132e-06, + "loss": 1.0827, + "step": 30570 + }, + { + "epoch": 2.57, + "learning_rate": 1.039210283495502e-06, + "loss": 0.9802, + "step": 30571 + }, + { + "epoch": 2.57, + "learning_rate": 1.038806832224203e-06, + "loss": 1.0925, + "step": 30572 + }, + { + "epoch": 2.57, + "learning_rate": 1.0384034549931488e-06, + "loss": 1.0631, + "step": 30573 + }, + { + "epoch": 2.57, + "learning_rate": 1.0380001518056747e-06, + "loss": 1.1706, + "step": 30574 + }, + { + "epoch": 2.57, + "learning_rate": 1.0375969226651117e-06, + "loss": 1.0059, + "step": 30575 + }, + { + "epoch": 2.57, + "learning_rate": 1.0371937675747923e-06, + "loss": 1.1587, + "step": 30576 + }, + { + "epoch": 2.57, + "learning_rate": 1.0367906865380428e-06, + "loss": 1.1671, + "step": 30577 + }, + { + "epoch": 2.57, + "learning_rate": 1.0363876795582007e-06, + "loss": 1.1038, + "step": 30578 + }, + { + "epoch": 2.57, + "learning_rate": 1.035984746638592e-06, + "loss": 1.1008, + "step": 30579 + }, + { + "epoch": 2.57, + "learning_rate": 1.035581887782544e-06, + "loss": 1.1368, + "step": 30580 + }, + { + "epoch": 2.57, + "learning_rate": 1.0351791029933889e-06, + "loss": 1.0858, + "step": 30581 + }, + { + "epoch": 2.57, + "learning_rate": 1.0347763922744524e-06, + "loss": 1.0106, + "step": 30582 + }, + { + "epoch": 2.57, + "learning_rate": 1.0343737556290634e-06, + "loss": 1.1711, + "step": 30583 + }, + { + "epoch": 2.57, + "learning_rate": 1.0339711930605478e-06, + "loss": 1.0816, + "step": 30584 + }, + { + "epoch": 2.58, + "learning_rate": 1.0335687045722308e-06, + "loss": 1.0623, + "step": 30585 + }, + { + "epoch": 2.58, + "learning_rate": 1.033166290167439e-06, + "loss": 1.059, + "step": 30586 + }, + { + "epoch": 2.58, + "learning_rate": 1.032763949849499e-06, + "loss": 1.1155, + "step": 30587 + }, + { + "epoch": 2.58, + "learning_rate": 1.0323616836217299e-06, + "loss": 1.1, + "step": 30588 + }, + { + "epoch": 2.58, + "learning_rate": 1.0319594914874619e-06, + "loss": 1.1356, + "step": 30589 + }, + { + "epoch": 2.58, + "learning_rate": 1.0315573734500128e-06, + "loss": 1.1558, + "step": 30590 + }, + { + "epoch": 2.58, + "learning_rate": 1.0311553295127074e-06, + "loss": 1.1451, + "step": 30591 + }, + { + "epoch": 2.58, + "learning_rate": 1.0307533596788667e-06, + "loss": 1.1875, + "step": 30592 + }, + { + "epoch": 2.58, + "learning_rate": 1.0303514639518132e-06, + "loss": 1.1474, + "step": 30593 + }, + { + "epoch": 2.58, + "learning_rate": 1.0299496423348653e-06, + "loss": 1.0571, + "step": 30594 + }, + { + "epoch": 2.58, + "learning_rate": 1.0295478948313465e-06, + "loss": 1.1459, + "step": 30595 + }, + { + "epoch": 2.58, + "learning_rate": 1.0291462214445702e-06, + "loss": 1.1907, + "step": 30596 + }, + { + "epoch": 2.58, + "learning_rate": 1.0287446221778607e-06, + "loss": 1.0891, + "step": 30597 + }, + { + "epoch": 2.58, + "learning_rate": 1.0283430970345354e-06, + "loss": 0.9902, + "step": 30598 + }, + { + "epoch": 2.58, + "learning_rate": 1.0279416460179082e-06, + "loss": 1.1134, + "step": 30599 + }, + { + "epoch": 2.58, + "learning_rate": 1.027540269131302e-06, + "loss": 1.0908, + "step": 30600 + }, + { + "epoch": 2.58, + "learning_rate": 1.0271389663780274e-06, + "loss": 1.1237, + "step": 30601 + }, + { + "epoch": 2.58, + "learning_rate": 1.026737737761403e-06, + "loss": 1.1037, + "step": 30602 + }, + { + "epoch": 2.58, + "learning_rate": 1.0263365832847427e-06, + "loss": 1.1396, + "step": 30603 + }, + { + "epoch": 2.58, + "learning_rate": 1.0259355029513618e-06, + "loss": 1.101, + "step": 30604 + }, + { + "epoch": 2.58, + "learning_rate": 1.025534496764573e-06, + "loss": 1.0879, + "step": 30605 + }, + { + "epoch": 2.58, + "learning_rate": 1.0251335647276927e-06, + "loss": 1.1096, + "step": 30606 + }, + { + "epoch": 2.58, + "learning_rate": 1.0247327068440283e-06, + "loss": 1.1424, + "step": 30607 + }, + { + "epoch": 2.58, + "learning_rate": 1.0243319231168958e-06, + "loss": 1.1061, + "step": 30608 + }, + { + "epoch": 2.58, + "learning_rate": 1.0239312135496072e-06, + "loss": 1.1326, + "step": 30609 + }, + { + "epoch": 2.58, + "learning_rate": 1.0235305781454686e-06, + "loss": 1.1788, + "step": 30610 + }, + { + "epoch": 2.58, + "learning_rate": 1.0231300169077962e-06, + "loss": 1.1048, + "step": 30611 + }, + { + "epoch": 2.58, + "learning_rate": 1.0227295298398955e-06, + "loss": 1.1007, + "step": 30612 + }, + { + "epoch": 2.58, + "learning_rate": 1.0223291169450756e-06, + "loss": 1.2448, + "step": 30613 + }, + { + "epoch": 2.58, + "learning_rate": 1.0219287782266462e-06, + "loss": 1.1346, + "step": 30614 + }, + { + "epoch": 2.58, + "learning_rate": 1.0215285136879172e-06, + "loss": 1.0444, + "step": 30615 + }, + { + "epoch": 2.58, + "learning_rate": 1.021128323332189e-06, + "loss": 1.1061, + "step": 30616 + }, + { + "epoch": 2.58, + "learning_rate": 1.0207282071627756e-06, + "loss": 1.0961, + "step": 30617 + }, + { + "epoch": 2.58, + "learning_rate": 1.0203281651829767e-06, + "loss": 1.1672, + "step": 30618 + }, + { + "epoch": 2.58, + "learning_rate": 1.0199281973961017e-06, + "loss": 1.0373, + "step": 30619 + }, + { + "epoch": 2.58, + "learning_rate": 1.0195283038054538e-06, + "loss": 1.1743, + "step": 30620 + }, + { + "epoch": 2.58, + "learning_rate": 1.0191284844143369e-06, + "loss": 1.1772, + "step": 30621 + }, + { + "epoch": 2.58, + "learning_rate": 1.0187287392260547e-06, + "loss": 1.0332, + "step": 30622 + }, + { + "epoch": 2.58, + "learning_rate": 1.0183290682439095e-06, + "loss": 1.0139, + "step": 30623 + }, + { + "epoch": 2.58, + "learning_rate": 1.0179294714712052e-06, + "loss": 1.1019, + "step": 30624 + }, + { + "epoch": 2.58, + "learning_rate": 1.0175299489112412e-06, + "loss": 1.0773, + "step": 30625 + }, + { + "epoch": 2.58, + "learning_rate": 1.0171305005673216e-06, + "loss": 1.0584, + "step": 30626 + }, + { + "epoch": 2.58, + "learning_rate": 1.0167311264427416e-06, + "loss": 1.0632, + "step": 30627 + }, + { + "epoch": 2.58, + "learning_rate": 1.0163318265408074e-06, + "loss": 1.1361, + "step": 30628 + }, + { + "epoch": 2.58, + "learning_rate": 1.015932600864813e-06, + "loss": 1.1167, + "step": 30629 + }, + { + "epoch": 2.58, + "learning_rate": 1.015533449418058e-06, + "loss": 1.1488, + "step": 30630 + }, + { + "epoch": 2.58, + "learning_rate": 1.015134372203842e-06, + "loss": 1.1461, + "step": 30631 + }, + { + "epoch": 2.58, + "learning_rate": 1.0147353692254601e-06, + "loss": 1.0308, + "step": 30632 + }, + { + "epoch": 2.58, + "learning_rate": 1.0143364404862122e-06, + "loss": 1.181, + "step": 30633 + }, + { + "epoch": 2.58, + "learning_rate": 1.0139375859893908e-06, + "loss": 1.1495, + "step": 30634 + }, + { + "epoch": 2.58, + "learning_rate": 1.0135388057382934e-06, + "loss": 1.2442, + "step": 30635 + }, + { + "epoch": 2.58, + "learning_rate": 1.0131400997362152e-06, + "loss": 1.0947, + "step": 30636 + }, + { + "epoch": 2.58, + "learning_rate": 1.01274146798645e-06, + "loss": 1.1671, + "step": 30637 + }, + { + "epoch": 2.58, + "learning_rate": 1.0123429104922888e-06, + "loss": 1.0885, + "step": 30638 + }, + { + "epoch": 2.58, + "learning_rate": 1.01194442725703e-06, + "loss": 1.1574, + "step": 30639 + }, + { + "epoch": 2.58, + "learning_rate": 1.0115460182839608e-06, + "loss": 1.1598, + "step": 30640 + }, + { + "epoch": 2.58, + "learning_rate": 1.0111476835763757e-06, + "loss": 1.1004, + "step": 30641 + }, + { + "epoch": 2.58, + "learning_rate": 1.0107494231375647e-06, + "loss": 1.1277, + "step": 30642 + }, + { + "epoch": 2.58, + "learning_rate": 1.010351236970819e-06, + "loss": 1.0798, + "step": 30643 + }, + { + "epoch": 2.58, + "learning_rate": 1.0099531250794282e-06, + "loss": 1.1009, + "step": 30644 + }, + { + "epoch": 2.58, + "learning_rate": 1.0095550874666837e-06, + "loss": 1.0784, + "step": 30645 + }, + { + "epoch": 2.58, + "learning_rate": 1.0091571241358689e-06, + "loss": 1.1842, + "step": 30646 + }, + { + "epoch": 2.58, + "learning_rate": 1.0087592350902774e-06, + "loss": 1.0897, + "step": 30647 + }, + { + "epoch": 2.58, + "learning_rate": 1.0083614203331971e-06, + "loss": 1.1453, + "step": 30648 + }, + { + "epoch": 2.58, + "learning_rate": 1.0079636798679072e-06, + "loss": 1.0368, + "step": 30649 + }, + { + "epoch": 2.58, + "learning_rate": 1.007566013697704e-06, + "loss": 1.0506, + "step": 30650 + }, + { + "epoch": 2.58, + "learning_rate": 1.007168421825866e-06, + "loss": 1.1477, + "step": 30651 + }, + { + "epoch": 2.58, + "learning_rate": 1.0067709042556807e-06, + "loss": 1.1113, + "step": 30652 + }, + { + "epoch": 2.58, + "learning_rate": 1.006373460990433e-06, + "loss": 1.1681, + "step": 30653 + }, + { + "epoch": 2.58, + "learning_rate": 1.005976092033405e-06, + "loss": 1.0509, + "step": 30654 + }, + { + "epoch": 2.58, + "learning_rate": 1.0055787973878816e-06, + "loss": 1.2201, + "step": 30655 + }, + { + "epoch": 2.58, + "learning_rate": 1.005181577057147e-06, + "loss": 1.0533, + "step": 30656 + }, + { + "epoch": 2.58, + "learning_rate": 1.0047844310444765e-06, + "loss": 1.0044, + "step": 30657 + }, + { + "epoch": 2.58, + "learning_rate": 1.0043873593531584e-06, + "loss": 1.1313, + "step": 30658 + }, + { + "epoch": 2.58, + "learning_rate": 1.003990361986472e-06, + "loss": 1.0957, + "step": 30659 + }, + { + "epoch": 2.58, + "learning_rate": 1.0035934389476932e-06, + "loss": 1.1402, + "step": 30660 + }, + { + "epoch": 2.58, + "learning_rate": 1.0031965902401085e-06, + "loss": 1.1131, + "step": 30661 + }, + { + "epoch": 2.58, + "learning_rate": 1.0027998158669905e-06, + "loss": 1.1722, + "step": 30662 + }, + { + "epoch": 2.58, + "learning_rate": 1.0024031158316194e-06, + "loss": 1.0924, + "step": 30663 + }, + { + "epoch": 2.58, + "learning_rate": 1.002006490137275e-06, + "loss": 1.1862, + "step": 30664 + }, + { + "epoch": 2.58, + "learning_rate": 1.0016099387872336e-06, + "loss": 1.0887, + "step": 30665 + }, + { + "epoch": 2.58, + "learning_rate": 1.0012134617847668e-06, + "loss": 1.084, + "step": 30666 + }, + { + "epoch": 2.58, + "learning_rate": 1.0008170591331578e-06, + "loss": 1.1298, + "step": 30667 + }, + { + "epoch": 2.58, + "learning_rate": 1.0004207308356772e-06, + "loss": 1.1819, + "step": 30668 + }, + { + "epoch": 2.58, + "learning_rate": 1.0000244768956001e-06, + "loss": 1.1529, + "step": 30669 + }, + { + "epoch": 2.58, + "learning_rate": 9.996282973162018e-07, + "loss": 1.1701, + "step": 30670 + }, + { + "epoch": 2.58, + "learning_rate": 9.992321921007552e-07, + "loss": 1.1161, + "step": 30671 + }, + { + "epoch": 2.58, + "learning_rate": 9.988361612525321e-07, + "loss": 1.1302, + "step": 30672 + }, + { + "epoch": 2.58, + "learning_rate": 9.984402047748064e-07, + "loss": 1.102, + "step": 30673 + }, + { + "epoch": 2.58, + "learning_rate": 9.98044322670848e-07, + "loss": 1.1792, + "step": 30674 + }, + { + "epoch": 2.58, + "learning_rate": 9.976485149439274e-07, + "loss": 1.1106, + "step": 30675 + }, + { + "epoch": 2.58, + "learning_rate": 9.972527815973188e-07, + "loss": 1.1273, + "step": 30676 + }, + { + "epoch": 2.58, + "learning_rate": 9.96857122634285e-07, + "loss": 1.1514, + "step": 30677 + }, + { + "epoch": 2.58, + "learning_rate": 9.964615380581032e-07, + "loss": 1.2402, + "step": 30678 + }, + { + "epoch": 2.58, + "learning_rate": 9.960660278720359e-07, + "loss": 1.0882, + "step": 30679 + }, + { + "epoch": 2.58, + "learning_rate": 9.956705920793518e-07, + "loss": 1.1194, + "step": 30680 + }, + { + "epoch": 2.58, + "learning_rate": 9.952752306833203e-07, + "loss": 1.1548, + "step": 30681 + }, + { + "epoch": 2.58, + "learning_rate": 9.948799436872048e-07, + "loss": 1.12, + "step": 30682 + }, + { + "epoch": 2.58, + "learning_rate": 9.944847310942751e-07, + "loss": 1.1156, + "step": 30683 + }, + { + "epoch": 2.58, + "learning_rate": 9.940895929077932e-07, + "loss": 1.1332, + "step": 30684 + }, + { + "epoch": 2.58, + "learning_rate": 9.936945291310263e-07, + "loss": 1.1358, + "step": 30685 + }, + { + "epoch": 2.58, + "learning_rate": 9.932995397672374e-07, + "loss": 1.0891, + "step": 30686 + }, + { + "epoch": 2.58, + "learning_rate": 9.929046248196927e-07, + "loss": 1.124, + "step": 30687 + }, + { + "epoch": 2.58, + "learning_rate": 9.925097842916476e-07, + "loss": 1.0842, + "step": 30688 + }, + { + "epoch": 2.58, + "learning_rate": 9.921150181863738e-07, + "loss": 1.0492, + "step": 30689 + }, + { + "epoch": 2.58, + "learning_rate": 9.917203265071273e-07, + "loss": 1.1354, + "step": 30690 + }, + { + "epoch": 2.58, + "learning_rate": 9.913257092571693e-07, + "loss": 1.1058, + "step": 30691 + }, + { + "epoch": 2.58, + "learning_rate": 9.909311664397624e-07, + "loss": 1.1021, + "step": 30692 + }, + { + "epoch": 2.58, + "learning_rate": 9.905366980581653e-07, + "loss": 1.0896, + "step": 30693 + }, + { + "epoch": 2.58, + "learning_rate": 9.901423041156365e-07, + "loss": 1.079, + "step": 30694 + }, + { + "epoch": 2.58, + "learning_rate": 9.89747984615439e-07, + "loss": 1.1686, + "step": 30695 + }, + { + "epoch": 2.58, + "learning_rate": 9.893537395608222e-07, + "loss": 1.0723, + "step": 30696 + }, + { + "epoch": 2.58, + "learning_rate": 9.889595689550513e-07, + "loss": 1.1286, + "step": 30697 + }, + { + "epoch": 2.58, + "learning_rate": 9.885654728013815e-07, + "loss": 1.091, + "step": 30698 + }, + { + "epoch": 2.58, + "learning_rate": 9.881714511030638e-07, + "loss": 1.1636, + "step": 30699 + }, + { + "epoch": 2.58, + "learning_rate": 9.87777503863363e-07, + "loss": 1.0784, + "step": 30700 + }, + { + "epoch": 2.58, + "learning_rate": 9.873836310855256e-07, + "loss": 1.0764, + "step": 30701 + }, + { + "epoch": 2.58, + "learning_rate": 9.86989832772809e-07, + "loss": 1.184, + "step": 30702 + }, + { + "epoch": 2.58, + "learning_rate": 9.86596108928467e-07, + "loss": 1.1268, + "step": 30703 + }, + { + "epoch": 2.59, + "learning_rate": 9.86202459555754e-07, + "loss": 1.1012, + "step": 30704 + }, + { + "epoch": 2.59, + "learning_rate": 9.858088846579195e-07, + "loss": 1.1076, + "step": 30705 + }, + { + "epoch": 2.59, + "learning_rate": 9.854153842382198e-07, + "loss": 1.2083, + "step": 30706 + }, + { + "epoch": 2.59, + "learning_rate": 9.850219582998987e-07, + "loss": 1.1474, + "step": 30707 + }, + { + "epoch": 2.59, + "learning_rate": 9.846286068462129e-07, + "loss": 1.1007, + "step": 30708 + }, + { + "epoch": 2.59, + "learning_rate": 9.84235329880413e-07, + "loss": 1.1085, + "step": 30709 + }, + { + "epoch": 2.59, + "learning_rate": 9.83842127405743e-07, + "loss": 1.1185, + "step": 30710 + }, + { + "epoch": 2.59, + "learning_rate": 9.83448999425458e-07, + "loss": 1.1594, + "step": 30711 + }, + { + "epoch": 2.59, + "learning_rate": 9.830559459428024e-07, + "loss": 1.1138, + "step": 30712 + }, + { + "epoch": 2.59, + "learning_rate": 9.826629669610233e-07, + "loss": 1.125, + "step": 30713 + }, + { + "epoch": 2.59, + "learning_rate": 9.822700624833692e-07, + "loss": 1.038, + "step": 30714 + }, + { + "epoch": 2.59, + "learning_rate": 9.818772325130865e-07, + "loss": 1.0729, + "step": 30715 + }, + { + "epoch": 2.59, + "learning_rate": 9.814844770534183e-07, + "loss": 1.0905, + "step": 30716 + }, + { + "epoch": 2.59, + "learning_rate": 9.810917961076137e-07, + "loss": 1.0706, + "step": 30717 + }, + { + "epoch": 2.59, + "learning_rate": 9.806991896789142e-07, + "loss": 1.0531, + "step": 30718 + }, + { + "epoch": 2.59, + "learning_rate": 9.803066577705644e-07, + "loss": 1.123, + "step": 30719 + }, + { + "epoch": 2.59, + "learning_rate": 9.799142003858087e-07, + "loss": 1.1091, + "step": 30720 + }, + { + "epoch": 2.59, + "learning_rate": 9.795218175278876e-07, + "loss": 1.1085, + "step": 30721 + }, + { + "epoch": 2.59, + "learning_rate": 9.79129509200044e-07, + "loss": 1.1505, + "step": 30722 + }, + { + "epoch": 2.59, + "learning_rate": 9.787372754055192e-07, + "loss": 1.0546, + "step": 30723 + }, + { + "epoch": 2.59, + "learning_rate": 9.783451161475555e-07, + "loss": 1.1313, + "step": 30724 + }, + { + "epoch": 2.59, + "learning_rate": 9.779530314293906e-07, + "loss": 1.1462, + "step": 30725 + }, + { + "epoch": 2.59, + "learning_rate": 9.775610212542674e-07, + "loss": 1.0297, + "step": 30726 + }, + { + "epoch": 2.59, + "learning_rate": 9.77169085625419e-07, + "loss": 1.1622, + "step": 30727 + }, + { + "epoch": 2.59, + "learning_rate": 9.767772245460905e-07, + "loss": 1.1902, + "step": 30728 + }, + { + "epoch": 2.59, + "learning_rate": 9.763854380195148e-07, + "loss": 1.1209, + "step": 30729 + }, + { + "epoch": 2.59, + "learning_rate": 9.759937260489305e-07, + "loss": 1.085, + "step": 30730 + }, + { + "epoch": 2.59, + "learning_rate": 9.75602088637574e-07, + "loss": 1.0915, + "step": 30731 + }, + { + "epoch": 2.59, + "learning_rate": 9.752105257886812e-07, + "loss": 1.1045, + "step": 30732 + }, + { + "epoch": 2.59, + "learning_rate": 9.748190375054878e-07, + "loss": 1.2011, + "step": 30733 + }, + { + "epoch": 2.59, + "learning_rate": 9.744276237912275e-07, + "loss": 1.0556, + "step": 30734 + }, + { + "epoch": 2.59, + "learning_rate": 9.740362846491358e-07, + "loss": 1.0312, + "step": 30735 + }, + { + "epoch": 2.59, + "learning_rate": 9.736450200824431e-07, + "loss": 1.134, + "step": 30736 + }, + { + "epoch": 2.59, + "learning_rate": 9.732538300943873e-07, + "loss": 1.1272, + "step": 30737 + }, + { + "epoch": 2.59, + "learning_rate": 9.72862714688193e-07, + "loss": 1.0678, + "step": 30738 + }, + { + "epoch": 2.59, + "learning_rate": 9.724716738671003e-07, + "loss": 1.1207, + "step": 30739 + }, + { + "epoch": 2.59, + "learning_rate": 9.72080707634334e-07, + "loss": 1.0855, + "step": 30740 + }, + { + "epoch": 2.59, + "learning_rate": 9.716898159931253e-07, + "loss": 1.1768, + "step": 30741 + }, + { + "epoch": 2.59, + "learning_rate": 9.712989989467047e-07, + "loss": 1.1795, + "step": 30742 + }, + { + "epoch": 2.59, + "learning_rate": 9.709082564983018e-07, + "loss": 1.0778, + "step": 30743 + }, + { + "epoch": 2.59, + "learning_rate": 9.705175886511453e-07, + "loss": 1.0899, + "step": 30744 + }, + { + "epoch": 2.59, + "learning_rate": 9.701269954084635e-07, + "loss": 1.1205, + "step": 30745 + }, + { + "epoch": 2.59, + "learning_rate": 9.697364767734785e-07, + "loss": 1.1362, + "step": 30746 + }, + { + "epoch": 2.59, + "learning_rate": 9.69346032749423e-07, + "loss": 1.0789, + "step": 30747 + }, + { + "epoch": 2.59, + "learning_rate": 9.689556633395225e-07, + "loss": 1.0907, + "step": 30748 + }, + { + "epoch": 2.59, + "learning_rate": 9.685653685469964e-07, + "loss": 1.0768, + "step": 30749 + }, + { + "epoch": 2.59, + "learning_rate": 9.681751483750778e-07, + "loss": 1.1103, + "step": 30750 + }, + { + "epoch": 2.59, + "learning_rate": 9.677850028269853e-07, + "loss": 1.1596, + "step": 30751 + }, + { + "epoch": 2.59, + "learning_rate": 9.673949319059428e-07, + "loss": 1.0205, + "step": 30752 + }, + { + "epoch": 2.59, + "learning_rate": 9.670049356151744e-07, + "loss": 1.1373, + "step": 30753 + }, + { + "epoch": 2.59, + "learning_rate": 9.666150139579023e-07, + "loss": 1.0752, + "step": 30754 + }, + { + "epoch": 2.59, + "learning_rate": 9.66225166937349e-07, + "loss": 1.199, + "step": 30755 + }, + { + "epoch": 2.59, + "learning_rate": 9.658353945567355e-07, + "loss": 1.1621, + "step": 30756 + }, + { + "epoch": 2.59, + "learning_rate": 9.654456968192772e-07, + "loss": 1.0848, + "step": 30757 + }, + { + "epoch": 2.59, + "learning_rate": 9.650560737282e-07, + "loss": 1.1114, + "step": 30758 + }, + { + "epoch": 2.59, + "learning_rate": 9.646665252867237e-07, + "loss": 1.0665, + "step": 30759 + }, + { + "epoch": 2.59, + "learning_rate": 9.642770514980593e-07, + "loss": 1.1806, + "step": 30760 + }, + { + "epoch": 2.59, + "learning_rate": 9.638876523654338e-07, + "loss": 1.0625, + "step": 30761 + }, + { + "epoch": 2.59, + "learning_rate": 9.634983278920595e-07, + "loss": 0.9757, + "step": 30762 + }, + { + "epoch": 2.59, + "learning_rate": 9.631090780811536e-07, + "loss": 1.1881, + "step": 30763 + }, + { + "epoch": 2.59, + "learning_rate": 9.627199029359323e-07, + "loss": 1.0845, + "step": 30764 + }, + { + "epoch": 2.59, + "learning_rate": 9.62330802459611e-07, + "loss": 1.1561, + "step": 30765 + }, + { + "epoch": 2.59, + "learning_rate": 9.619417766554061e-07, + "loss": 1.0878, + "step": 30766 + }, + { + "epoch": 2.59, + "learning_rate": 9.615528255265327e-07, + "loss": 1.0982, + "step": 30767 + }, + { + "epoch": 2.59, + "learning_rate": 9.611639490761982e-07, + "loss": 1.2197, + "step": 30768 + }, + { + "epoch": 2.59, + "learning_rate": 9.607751473076232e-07, + "loss": 1.1576, + "step": 30769 + }, + { + "epoch": 2.59, + "learning_rate": 9.603864202240165e-07, + "loss": 1.1667, + "step": 30770 + }, + { + "epoch": 2.59, + "learning_rate": 9.599977678285888e-07, + "loss": 1.1034, + "step": 30771 + }, + { + "epoch": 2.59, + "learning_rate": 9.59609190124554e-07, + "loss": 1.1175, + "step": 30772 + }, + { + "epoch": 2.59, + "learning_rate": 9.592206871151199e-07, + "loss": 1.19, + "step": 30773 + }, + { + "epoch": 2.59, + "learning_rate": 9.588322588034992e-07, + "loss": 1.1862, + "step": 30774 + }, + { + "epoch": 2.59, + "learning_rate": 9.584439051928996e-07, + "loss": 1.0566, + "step": 30775 + }, + { + "epoch": 2.59, + "learning_rate": 9.580556262865315e-07, + "loss": 1.118, + "step": 30776 + }, + { + "epoch": 2.59, + "learning_rate": 9.576674220875982e-07, + "loss": 1.0993, + "step": 30777 + }, + { + "epoch": 2.59, + "learning_rate": 9.572792925993135e-07, + "loss": 1.1162, + "step": 30778 + }, + { + "epoch": 2.59, + "learning_rate": 9.568912378248807e-07, + "loss": 1.102, + "step": 30779 + }, + { + "epoch": 2.59, + "learning_rate": 9.56503257767506e-07, + "loss": 1.1347, + "step": 30780 + }, + { + "epoch": 2.59, + "learning_rate": 9.561153524303957e-07, + "loss": 1.0069, + "step": 30781 + }, + { + "epoch": 2.59, + "learning_rate": 9.55727521816755e-07, + "loss": 1.1251, + "step": 30782 + }, + { + "epoch": 2.59, + "learning_rate": 9.55339765929788e-07, + "loss": 1.1468, + "step": 30783 + }, + { + "epoch": 2.59, + "learning_rate": 9.54952084772699e-07, + "loss": 1.0889, + "step": 30784 + }, + { + "epoch": 2.59, + "learning_rate": 9.545644783486896e-07, + "loss": 1.0772, + "step": 30785 + }, + { + "epoch": 2.59, + "learning_rate": 9.54176946660964e-07, + "loss": 0.984, + "step": 30786 + }, + { + "epoch": 2.59, + "learning_rate": 9.537894897127253e-07, + "loss": 1.0471, + "step": 30787 + }, + { + "epoch": 2.59, + "learning_rate": 9.534021075071697e-07, + "loss": 1.0558, + "step": 30788 + }, + { + "epoch": 2.59, + "learning_rate": 9.530148000475048e-07, + "loss": 1.2254, + "step": 30789 + }, + { + "epoch": 2.59, + "learning_rate": 9.526275673369245e-07, + "loss": 1.1354, + "step": 30790 + }, + { + "epoch": 2.59, + "learning_rate": 9.522404093786308e-07, + "loss": 1.163, + "step": 30791 + }, + { + "epoch": 2.59, + "learning_rate": 9.518533261758233e-07, + "loss": 1.159, + "step": 30792 + }, + { + "epoch": 2.59, + "learning_rate": 9.514663177316985e-07, + "loss": 1.1305, + "step": 30793 + }, + { + "epoch": 2.59, + "learning_rate": 9.510793840494548e-07, + "loss": 1.1467, + "step": 30794 + }, + { + "epoch": 2.59, + "learning_rate": 9.506925251322918e-07, + "loss": 1.1731, + "step": 30795 + }, + { + "epoch": 2.59, + "learning_rate": 9.503057409833993e-07, + "loss": 1.1841, + "step": 30796 + }, + { + "epoch": 2.59, + "learning_rate": 9.499190316059792e-07, + "loss": 1.0738, + "step": 30797 + }, + { + "epoch": 2.59, + "learning_rate": 9.495323970032266e-07, + "loss": 1.121, + "step": 30798 + }, + { + "epoch": 2.59, + "learning_rate": 9.491458371783291e-07, + "loss": 1.1206, + "step": 30799 + }, + { + "epoch": 2.59, + "learning_rate": 9.487593521344895e-07, + "loss": 1.1035, + "step": 30800 + }, + { + "epoch": 2.59, + "learning_rate": 9.483729418748954e-07, + "loss": 1.2065, + "step": 30801 + }, + { + "epoch": 2.59, + "learning_rate": 9.47986606402741e-07, + "loss": 1.1059, + "step": 30802 + }, + { + "epoch": 2.59, + "learning_rate": 9.47600345721218e-07, + "loss": 1.0054, + "step": 30803 + }, + { + "epoch": 2.59, + "learning_rate": 9.472141598335172e-07, + "loss": 1.1862, + "step": 30804 + }, + { + "epoch": 2.59, + "learning_rate": 9.468280487428317e-07, + "loss": 1.1874, + "step": 30805 + }, + { + "epoch": 2.59, + "learning_rate": 9.46442012452351e-07, + "loss": 1.036, + "step": 30806 + }, + { + "epoch": 2.59, + "learning_rate": 9.460560509652605e-07, + "loss": 1.0787, + "step": 30807 + }, + { + "epoch": 2.59, + "learning_rate": 9.456701642847554e-07, + "loss": 1.1718, + "step": 30808 + }, + { + "epoch": 2.59, + "learning_rate": 9.45284352414022e-07, + "loss": 1.1584, + "step": 30809 + }, + { + "epoch": 2.59, + "learning_rate": 9.448986153562434e-07, + "loss": 1.1642, + "step": 30810 + }, + { + "epoch": 2.59, + "learning_rate": 9.445129531146147e-07, + "loss": 1.0133, + "step": 30811 + }, + { + "epoch": 2.59, + "learning_rate": 9.441273656923155e-07, + "loss": 1.1472, + "step": 30812 + }, + { + "epoch": 2.59, + "learning_rate": 9.437418530925346e-07, + "loss": 1.1837, + "step": 30813 + }, + { + "epoch": 2.59, + "learning_rate": 9.433564153184571e-07, + "loss": 1.1541, + "step": 30814 + }, + { + "epoch": 2.59, + "learning_rate": 9.429710523732683e-07, + "loss": 1.1465, + "step": 30815 + }, + { + "epoch": 2.59, + "learning_rate": 9.425857642601499e-07, + "loss": 1.1129, + "step": 30816 + }, + { + "epoch": 2.59, + "learning_rate": 9.422005509822885e-07, + "loss": 1.0687, + "step": 30817 + }, + { + "epoch": 2.59, + "learning_rate": 9.418154125428625e-07, + "loss": 1.0996, + "step": 30818 + }, + { + "epoch": 2.59, + "learning_rate": 9.414303489450594e-07, + "loss": 1.0666, + "step": 30819 + }, + { + "epoch": 2.59, + "learning_rate": 9.410453601920555e-07, + "loss": 1.1243, + "step": 30820 + }, + { + "epoch": 2.59, + "learning_rate": 9.406604462870328e-07, + "loss": 1.0771, + "step": 30821 + }, + { + "epoch": 2.59, + "learning_rate": 9.402756072331765e-07, + "loss": 1.111, + "step": 30822 + }, + { + "epoch": 2.6, + "learning_rate": 9.398908430336606e-07, + "loss": 1.1405, + "step": 30823 + }, + { + "epoch": 2.6, + "learning_rate": 9.395061536916672e-07, + "loss": 1.0633, + "step": 30824 + }, + { + "epoch": 2.6, + "learning_rate": 9.391215392103725e-07, + "loss": 1.1903, + "step": 30825 + }, + { + "epoch": 2.6, + "learning_rate": 9.387369995929573e-07, + "loss": 1.0433, + "step": 30826 + }, + { + "epoch": 2.6, + "learning_rate": 9.383525348425937e-07, + "loss": 1.0859, + "step": 30827 + }, + { + "epoch": 2.6, + "learning_rate": 9.379681449624656e-07, + "loss": 1.109, + "step": 30828 + }, + { + "epoch": 2.6, + "learning_rate": 9.375838299557427e-07, + "loss": 1.1866, + "step": 30829 + }, + { + "epoch": 2.6, + "learning_rate": 9.371995898256026e-07, + "loss": 1.1135, + "step": 30830 + }, + { + "epoch": 2.6, + "learning_rate": 9.368154245752203e-07, + "loss": 1.0917, + "step": 30831 + }, + { + "epoch": 2.6, + "learning_rate": 9.364313342077691e-07, + "loss": 1.1154, + "step": 30832 + }, + { + "epoch": 2.6, + "learning_rate": 9.36047318726423e-07, + "loss": 1.2135, + "step": 30833 + }, + { + "epoch": 2.6, + "learning_rate": 9.356633781343549e-07, + "loss": 1.1356, + "step": 30834 + }, + { + "epoch": 2.6, + "learning_rate": 9.352795124347369e-07, + "loss": 1.113, + "step": 30835 + }, + { + "epoch": 2.6, + "learning_rate": 9.348957216307397e-07, + "loss": 1.1735, + "step": 30836 + }, + { + "epoch": 2.6, + "learning_rate": 9.345120057255386e-07, + "loss": 1.0463, + "step": 30837 + }, + { + "epoch": 2.6, + "learning_rate": 9.341283647222965e-07, + "loss": 1.1173, + "step": 30838 + }, + { + "epoch": 2.6, + "learning_rate": 9.33744798624191e-07, + "loss": 1.1339, + "step": 30839 + }, + { + "epoch": 2.6, + "learning_rate": 9.33361307434385e-07, + "loss": 1.0976, + "step": 30840 + }, + { + "epoch": 2.6, + "learning_rate": 9.329778911560505e-07, + "loss": 1.0329, + "step": 30841 + }, + { + "epoch": 2.6, + "learning_rate": 9.325945497923539e-07, + "loss": 1.1054, + "step": 30842 + }, + { + "epoch": 2.6, + "learning_rate": 9.322112833464625e-07, + "loss": 1.1664, + "step": 30843 + }, + { + "epoch": 2.6, + "learning_rate": 9.318280918215449e-07, + "loss": 1.1174, + "step": 30844 + }, + { + "epoch": 2.6, + "learning_rate": 9.314449752207666e-07, + "loss": 1.1746, + "step": 30845 + }, + { + "epoch": 2.6, + "learning_rate": 9.310619335472892e-07, + "loss": 1.0611, + "step": 30846 + }, + { + "epoch": 2.6, + "learning_rate": 9.306789668042826e-07, + "loss": 1.0139, + "step": 30847 + }, + { + "epoch": 2.6, + "learning_rate": 9.302960749949097e-07, + "loss": 1.1912, + "step": 30848 + }, + { + "epoch": 2.6, + "learning_rate": 9.299132581223302e-07, + "loss": 1.2002, + "step": 30849 + }, + { + "epoch": 2.6, + "learning_rate": 9.295305161897139e-07, + "loss": 1.1847, + "step": 30850 + }, + { + "epoch": 2.6, + "learning_rate": 9.291478492002182e-07, + "loss": 0.9913, + "step": 30851 + }, + { + "epoch": 2.6, + "learning_rate": 9.287652571570049e-07, + "loss": 1.0872, + "step": 30852 + }, + { + "epoch": 2.6, + "learning_rate": 9.283827400632361e-07, + "loss": 1.1151, + "step": 30853 + }, + { + "epoch": 2.6, + "learning_rate": 9.280002979220737e-07, + "loss": 1.1757, + "step": 30854 + }, + { + "epoch": 2.6, + "learning_rate": 9.276179307366751e-07, + "loss": 1.2033, + "step": 30855 + }, + { + "epoch": 2.6, + "learning_rate": 9.272356385102033e-07, + "loss": 1.1056, + "step": 30856 + }, + { + "epoch": 2.6, + "learning_rate": 9.268534212458102e-07, + "loss": 1.1321, + "step": 30857 + }, + { + "epoch": 2.6, + "learning_rate": 9.264712789466602e-07, + "loss": 1.0356, + "step": 30858 + }, + { + "epoch": 2.6, + "learning_rate": 9.260892116159093e-07, + "loss": 1.1363, + "step": 30859 + }, + { + "epoch": 2.6, + "learning_rate": 9.257072192567107e-07, + "loss": 1.1013, + "step": 30860 + }, + { + "epoch": 2.6, + "learning_rate": 9.253253018722264e-07, + "loss": 1.1304, + "step": 30861 + }, + { + "epoch": 2.6, + "learning_rate": 9.249434594656059e-07, + "loss": 1.1065, + "step": 30862 + }, + { + "epoch": 2.6, + "learning_rate": 9.24561692040008e-07, + "loss": 1.1461, + "step": 30863 + }, + { + "epoch": 2.6, + "learning_rate": 9.241799995985867e-07, + "loss": 1.1427, + "step": 30864 + }, + { + "epoch": 2.6, + "learning_rate": 9.23798382144494e-07, + "loss": 1.1385, + "step": 30865 + }, + { + "epoch": 2.6, + "learning_rate": 9.234168396808829e-07, + "loss": 1.0067, + "step": 30866 + }, + { + "epoch": 2.6, + "learning_rate": 9.230353722109098e-07, + "loss": 1.0699, + "step": 30867 + }, + { + "epoch": 2.6, + "learning_rate": 9.226539797377199e-07, + "loss": 1.1289, + "step": 30868 + }, + { + "epoch": 2.6, + "learning_rate": 9.222726622644707e-07, + "loss": 1.0789, + "step": 30869 + }, + { + "epoch": 2.6, + "learning_rate": 9.218914197943086e-07, + "loss": 1.1024, + "step": 30870 + }, + { + "epoch": 2.6, + "learning_rate": 9.215102523303832e-07, + "loss": 1.1101, + "step": 30871 + }, + { + "epoch": 2.6, + "learning_rate": 9.211291598758498e-07, + "loss": 1.1155, + "step": 30872 + }, + { + "epoch": 2.6, + "learning_rate": 9.207481424338505e-07, + "loss": 1.1252, + "step": 30873 + }, + { + "epoch": 2.6, + "learning_rate": 9.203672000075359e-07, + "loss": 1.0458, + "step": 30874 + }, + { + "epoch": 2.6, + "learning_rate": 9.199863326000524e-07, + "loss": 1.0644, + "step": 30875 + }, + { + "epoch": 2.6, + "learning_rate": 9.196055402145509e-07, + "loss": 1.1131, + "step": 30876 + }, + { + "epoch": 2.6, + "learning_rate": 9.192248228541712e-07, + "loss": 1.2067, + "step": 30877 + }, + { + "epoch": 2.6, + "learning_rate": 9.188441805220649e-07, + "loss": 1.1076, + "step": 30878 + }, + { + "epoch": 2.6, + "learning_rate": 9.18463613221372e-07, + "loss": 1.0899, + "step": 30879 + }, + { + "epoch": 2.6, + "learning_rate": 9.180831209552398e-07, + "loss": 1.0437, + "step": 30880 + }, + { + "epoch": 2.6, + "learning_rate": 9.177027037268116e-07, + "loss": 1.0936, + "step": 30881 + }, + { + "epoch": 2.6, + "learning_rate": 9.173223615392312e-07, + "loss": 1.165, + "step": 30882 + }, + { + "epoch": 2.6, + "learning_rate": 9.169420943956387e-07, + "loss": 1.1498, + "step": 30883 + }, + { + "epoch": 2.6, + "learning_rate": 9.165619022991789e-07, + "loss": 1.1628, + "step": 30884 + }, + { + "epoch": 2.6, + "learning_rate": 9.161817852529909e-07, + "loss": 1.1617, + "step": 30885 + }, + { + "epoch": 2.6, + "learning_rate": 9.158017432602162e-07, + "loss": 1.074, + "step": 30886 + }, + { + "epoch": 2.6, + "learning_rate": 9.154217763239969e-07, + "loss": 1.2547, + "step": 30887 + }, + { + "epoch": 2.6, + "learning_rate": 9.150418844474662e-07, + "loss": 1.039, + "step": 30888 + }, + { + "epoch": 2.6, + "learning_rate": 9.146620676337714e-07, + "loss": 1.0554, + "step": 30889 + }, + { + "epoch": 2.6, + "learning_rate": 9.142823258860444e-07, + "loss": 0.9617, + "step": 30890 + }, + { + "epoch": 2.6, + "learning_rate": 9.13902659207424e-07, + "loss": 1.0162, + "step": 30891 + }, + { + "epoch": 2.6, + "learning_rate": 9.135230676010487e-07, + "loss": 1.1122, + "step": 30892 + }, + { + "epoch": 2.6, + "learning_rate": 9.131435510700526e-07, + "loss": 1.084, + "step": 30893 + }, + { + "epoch": 2.6, + "learning_rate": 9.127641096175732e-07, + "loss": 1.1569, + "step": 30894 + }, + { + "epoch": 2.6, + "learning_rate": 9.12384743246747e-07, + "loss": 1.1769, + "step": 30895 + }, + { + "epoch": 2.6, + "learning_rate": 9.120054519607024e-07, + "loss": 1.1569, + "step": 30896 + }, + { + "epoch": 2.6, + "learning_rate": 9.116262357625804e-07, + "loss": 1.1626, + "step": 30897 + }, + { + "epoch": 2.6, + "learning_rate": 9.112470946555118e-07, + "loss": 1.0953, + "step": 30898 + }, + { + "epoch": 2.6, + "learning_rate": 9.108680286426253e-07, + "loss": 1.1921, + "step": 30899 + }, + { + "epoch": 2.6, + "learning_rate": 9.104890377270592e-07, + "loss": 1.1473, + "step": 30900 + }, + { + "epoch": 2.6, + "learning_rate": 9.101101219119402e-07, + "loss": 1.1034, + "step": 30901 + }, + { + "epoch": 2.6, + "learning_rate": 9.097312812004e-07, + "loss": 1.0943, + "step": 30902 + }, + { + "epoch": 2.6, + "learning_rate": 9.093525155955696e-07, + "loss": 1.1105, + "step": 30903 + }, + { + "epoch": 2.6, + "learning_rate": 9.089738251005776e-07, + "loss": 1.1211, + "step": 30904 + }, + { + "epoch": 2.6, + "learning_rate": 9.085952097185546e-07, + "loss": 1.0082, + "step": 30905 + }, + { + "epoch": 2.6, + "learning_rate": 9.082166694526284e-07, + "loss": 1.1898, + "step": 30906 + }, + { + "epoch": 2.6, + "learning_rate": 9.078382043059219e-07, + "loss": 1.144, + "step": 30907 + }, + { + "epoch": 2.6, + "learning_rate": 9.074598142815683e-07, + "loss": 1.1342, + "step": 30908 + }, + { + "epoch": 2.6, + "learning_rate": 9.070814993826937e-07, + "loss": 1.0805, + "step": 30909 + }, + { + "epoch": 2.6, + "learning_rate": 9.067032596124182e-07, + "loss": 1.0721, + "step": 30910 + }, + { + "epoch": 2.6, + "learning_rate": 9.063250949738744e-07, + "loss": 1.1062, + "step": 30911 + }, + { + "epoch": 2.6, + "learning_rate": 9.059470054701813e-07, + "loss": 1.0597, + "step": 30912 + }, + { + "epoch": 2.6, + "learning_rate": 9.055689911044641e-07, + "loss": 1.083, + "step": 30913 + }, + { + "epoch": 2.6, + "learning_rate": 9.051910518798468e-07, + "loss": 1.1415, + "step": 30914 + }, + { + "epoch": 2.6, + "learning_rate": 9.048131877994515e-07, + "loss": 1.0738, + "step": 30915 + }, + { + "epoch": 2.6, + "learning_rate": 9.044353988664001e-07, + "loss": 1.1201, + "step": 30916 + }, + { + "epoch": 2.6, + "learning_rate": 9.040576850838167e-07, + "loss": 1.1379, + "step": 30917 + }, + { + "epoch": 2.6, + "learning_rate": 9.036800464548157e-07, + "loss": 1.0387, + "step": 30918 + }, + { + "epoch": 2.6, + "learning_rate": 9.033024829825254e-07, + "loss": 1.1606, + "step": 30919 + }, + { + "epoch": 2.6, + "learning_rate": 9.029249946700591e-07, + "loss": 1.1085, + "step": 30920 + }, + { + "epoch": 2.6, + "learning_rate": 9.025475815205364e-07, + "loss": 1.0386, + "step": 30921 + }, + { + "epoch": 2.6, + "learning_rate": 9.021702435370816e-07, + "loss": 1.1161, + "step": 30922 + }, + { + "epoch": 2.6, + "learning_rate": 9.017929807228043e-07, + "loss": 1.1277, + "step": 30923 + }, + { + "epoch": 2.6, + "learning_rate": 9.014157930808265e-07, + "loss": 1.0913, + "step": 30924 + }, + { + "epoch": 2.6, + "learning_rate": 9.010386806142635e-07, + "loss": 1.1254, + "step": 30925 + }, + { + "epoch": 2.6, + "learning_rate": 9.006616433262317e-07, + "loss": 1.0455, + "step": 30926 + }, + { + "epoch": 2.6, + "learning_rate": 9.00284681219843e-07, + "loss": 1.035, + "step": 30927 + }, + { + "epoch": 2.6, + "learning_rate": 8.999077942982171e-07, + "loss": 1.0598, + "step": 30928 + }, + { + "epoch": 2.6, + "learning_rate": 8.99530982564465e-07, + "loss": 1.0869, + "step": 30929 + }, + { + "epoch": 2.6, + "learning_rate": 8.991542460216996e-07, + "loss": 1.1628, + "step": 30930 + }, + { + "epoch": 2.6, + "learning_rate": 8.98777584673034e-07, + "loss": 1.1404, + "step": 30931 + }, + { + "epoch": 2.6, + "learning_rate": 8.984009985215814e-07, + "loss": 1.0702, + "step": 30932 + }, + { + "epoch": 2.6, + "learning_rate": 8.980244875704525e-07, + "loss": 1.0803, + "step": 30933 + }, + { + "epoch": 2.6, + "learning_rate": 8.976480518227582e-07, + "loss": 1.0206, + "step": 30934 + }, + { + "epoch": 2.6, + "learning_rate": 8.972716912816093e-07, + "loss": 1.1548, + "step": 30935 + }, + { + "epoch": 2.6, + "learning_rate": 8.968954059501145e-07, + "loss": 1.0883, + "step": 30936 + }, + { + "epoch": 2.6, + "learning_rate": 8.965191958313846e-07, + "loss": 1.0946, + "step": 30937 + }, + { + "epoch": 2.6, + "learning_rate": 8.961430609285238e-07, + "loss": 1.1462, + "step": 30938 + }, + { + "epoch": 2.6, + "learning_rate": 8.957670012446462e-07, + "loss": 1.0536, + "step": 30939 + }, + { + "epoch": 2.6, + "learning_rate": 8.953910167828528e-07, + "loss": 1.0576, + "step": 30940 + }, + { + "epoch": 2.6, + "learning_rate": 8.950151075462532e-07, + "loss": 1.0704, + "step": 30941 + }, + { + "epoch": 2.61, + "learning_rate": 8.946392735379528e-07, + "loss": 1.0838, + "step": 30942 + }, + { + "epoch": 2.61, + "learning_rate": 8.942635147610567e-07, + "loss": 1.0723, + "step": 30943 + }, + { + "epoch": 2.61, + "learning_rate": 8.938878312186694e-07, + "loss": 1.0911, + "step": 30944 + }, + { + "epoch": 2.61, + "learning_rate": 8.935122229138971e-07, + "loss": 1.057, + "step": 30945 + }, + { + "epoch": 2.61, + "learning_rate": 8.931366898498372e-07, + "loss": 1.1554, + "step": 30946 + }, + { + "epoch": 2.61, + "learning_rate": 8.927612320295997e-07, + "loss": 1.1501, + "step": 30947 + }, + { + "epoch": 2.61, + "learning_rate": 8.923858494562843e-07, + "loss": 1.2456, + "step": 30948 + }, + { + "epoch": 2.61, + "learning_rate": 8.920105421329883e-07, + "loss": 1.056, + "step": 30949 + }, + { + "epoch": 2.61, + "learning_rate": 8.916353100628206e-07, + "loss": 1.0922, + "step": 30950 + }, + { + "epoch": 2.61, + "learning_rate": 8.912601532488741e-07, + "loss": 1.1766, + "step": 30951 + }, + { + "epoch": 2.61, + "learning_rate": 8.908850716942519e-07, + "loss": 1.067, + "step": 30952 + }, + { + "epoch": 2.61, + "learning_rate": 8.905100654020526e-07, + "loss": 1.1594, + "step": 30953 + }, + { + "epoch": 2.61, + "learning_rate": 8.901351343753739e-07, + "loss": 1.1298, + "step": 30954 + }, + { + "epoch": 2.61, + "learning_rate": 8.897602786173154e-07, + "loss": 1.1295, + "step": 30955 + }, + { + "epoch": 2.61, + "learning_rate": 8.893854981309746e-07, + "loss": 1.1291, + "step": 30956 + }, + { + "epoch": 2.61, + "learning_rate": 8.890107929194425e-07, + "loss": 1.073, + "step": 30957 + }, + { + "epoch": 2.61, + "learning_rate": 8.886361629858209e-07, + "loss": 1.1503, + "step": 30958 + }, + { + "epoch": 2.61, + "learning_rate": 8.882616083332052e-07, + "loss": 1.1117, + "step": 30959 + }, + { + "epoch": 2.61, + "learning_rate": 8.878871289646851e-07, + "loss": 1.1959, + "step": 30960 + }, + { + "epoch": 2.61, + "learning_rate": 8.875127248833604e-07, + "loss": 1.0279, + "step": 30961 + }, + { + "epoch": 2.61, + "learning_rate": 8.871383960923208e-07, + "loss": 1.1689, + "step": 30962 + }, + { + "epoch": 2.61, + "learning_rate": 8.867641425946605e-07, + "loss": 1.1454, + "step": 30963 + }, + { + "epoch": 2.61, + "learning_rate": 8.863899643934715e-07, + "loss": 1.0113, + "step": 30964 + }, + { + "epoch": 2.61, + "learning_rate": 8.860158614918446e-07, + "loss": 1.065, + "step": 30965 + }, + { + "epoch": 2.61, + "learning_rate": 8.856418338928718e-07, + "loss": 1.0721, + "step": 30966 + }, + { + "epoch": 2.61, + "learning_rate": 8.852678815996452e-07, + "loss": 1.0656, + "step": 30967 + }, + { + "epoch": 2.61, + "learning_rate": 8.848940046152488e-07, + "loss": 1.1033, + "step": 30968 + }, + { + "epoch": 2.61, + "learning_rate": 8.845202029427791e-07, + "loss": 1.1292, + "step": 30969 + }, + { + "epoch": 2.61, + "learning_rate": 8.841464765853192e-07, + "loss": 1.1349, + "step": 30970 + }, + { + "epoch": 2.61, + "learning_rate": 8.837728255459566e-07, + "loss": 1.1533, + "step": 30971 + }, + { + "epoch": 2.61, + "learning_rate": 8.833992498277833e-07, + "loss": 1.1783, + "step": 30972 + }, + { + "epoch": 2.61, + "learning_rate": 8.830257494338823e-07, + "loss": 1.1276, + "step": 30973 + }, + { + "epoch": 2.61, + "learning_rate": 8.82652324367339e-07, + "loss": 1.124, + "step": 30974 + }, + { + "epoch": 2.61, + "learning_rate": 8.82278974631241e-07, + "loss": 1.0889, + "step": 30975 + }, + { + "epoch": 2.61, + "learning_rate": 8.819057002286747e-07, + "loss": 1.0796, + "step": 30976 + }, + { + "epoch": 2.61, + "learning_rate": 8.815325011627174e-07, + "loss": 1.0305, + "step": 30977 + }, + { + "epoch": 2.61, + "learning_rate": 8.811593774364601e-07, + "loss": 1.1315, + "step": 30978 + }, + { + "epoch": 2.61, + "learning_rate": 8.807863290529806e-07, + "loss": 1.1077, + "step": 30979 + }, + { + "epoch": 2.61, + "learning_rate": 8.804133560153638e-07, + "loss": 1.0862, + "step": 30980 + }, + { + "epoch": 2.61, + "learning_rate": 8.800404583266898e-07, + "loss": 1.1272, + "step": 30981 + }, + { + "epoch": 2.61, + "learning_rate": 8.796676359900402e-07, + "loss": 1.0511, + "step": 30982 + }, + { + "epoch": 2.61, + "learning_rate": 8.792948890084963e-07, + "loss": 1.1044, + "step": 30983 + }, + { + "epoch": 2.61, + "learning_rate": 8.789222173851364e-07, + "loss": 1.0387, + "step": 30984 + }, + { + "epoch": 2.61, + "learning_rate": 8.785496211230405e-07, + "loss": 1.207, + "step": 30985 + }, + { + "epoch": 2.61, + "learning_rate": 8.781771002252859e-07, + "loss": 1.1293, + "step": 30986 + }, + { + "epoch": 2.61, + "learning_rate": 8.778046546949548e-07, + "loss": 1.1066, + "step": 30987 + }, + { + "epoch": 2.61, + "learning_rate": 8.774322845351158e-07, + "loss": 1.0463, + "step": 30988 + }, + { + "epoch": 2.61, + "learning_rate": 8.770599897488563e-07, + "loss": 1.0893, + "step": 30989 + }, + { + "epoch": 2.61, + "learning_rate": 8.76687770339244e-07, + "loss": 1.0388, + "step": 30990 + }, + { + "epoch": 2.61, + "learning_rate": 8.763156263093575e-07, + "loss": 1.0331, + "step": 30991 + }, + { + "epoch": 2.61, + "learning_rate": 8.75943557662271e-07, + "loss": 1.1465, + "step": 30992 + }, + { + "epoch": 2.61, + "learning_rate": 8.755715644010588e-07, + "loss": 1.1491, + "step": 30993 + }, + { + "epoch": 2.61, + "learning_rate": 8.751996465287948e-07, + "loss": 1.175, + "step": 30994 + }, + { + "epoch": 2.61, + "learning_rate": 8.748278040485537e-07, + "loss": 1.2024, + "step": 30995 + }, + { + "epoch": 2.61, + "learning_rate": 8.744560369634014e-07, + "loss": 1.0607, + "step": 30996 + }, + { + "epoch": 2.61, + "learning_rate": 8.740843452764158e-07, + "loss": 1.081, + "step": 30997 + }, + { + "epoch": 2.61, + "learning_rate": 8.737127289906689e-07, + "loss": 1.0341, + "step": 30998 + }, + { + "epoch": 2.61, + "learning_rate": 8.733411881092235e-07, + "loss": 1.1584, + "step": 30999 + }, + { + "epoch": 2.61, + "learning_rate": 8.729697226351574e-07, + "loss": 1.2028, + "step": 31000 + }, + { + "epoch": 2.61, + "learning_rate": 8.725983325715359e-07, + "loss": 1.0439, + "step": 31001 + }, + { + "epoch": 2.61, + "learning_rate": 8.722270179214276e-07, + "loss": 1.1164, + "step": 31002 + }, + { + "epoch": 2.61, + "learning_rate": 8.718557786879011e-07, + "loss": 1.0907, + "step": 31003 + }, + { + "epoch": 2.61, + "learning_rate": 8.71484614874023e-07, + "loss": 1.0522, + "step": 31004 + }, + { + "epoch": 2.61, + "learning_rate": 8.711135264828607e-07, + "loss": 1.1203, + "step": 31005 + }, + { + "epoch": 2.61, + "learning_rate": 8.707425135174819e-07, + "loss": 1.147, + "step": 31006 + }, + { + "epoch": 2.61, + "learning_rate": 8.703715759809461e-07, + "loss": 1.1432, + "step": 31007 + }, + { + "epoch": 2.61, + "learning_rate": 8.700007138763255e-07, + "loss": 1.1332, + "step": 31008 + }, + { + "epoch": 2.61, + "learning_rate": 8.696299272066811e-07, + "loss": 1.134, + "step": 31009 + }, + { + "epoch": 2.61, + "learning_rate": 8.692592159750735e-07, + "loss": 1.0418, + "step": 31010 + }, + { + "epoch": 2.61, + "learning_rate": 8.688885801845725e-07, + "loss": 1.0642, + "step": 31011 + }, + { + "epoch": 2.61, + "learning_rate": 8.685180198382337e-07, + "loss": 1.2105, + "step": 31012 + }, + { + "epoch": 2.61, + "learning_rate": 8.681475349391222e-07, + "loss": 1.0523, + "step": 31013 + }, + { + "epoch": 2.61, + "learning_rate": 8.677771254902978e-07, + "loss": 1.0575, + "step": 31014 + }, + { + "epoch": 2.61, + "learning_rate": 8.674067914948214e-07, + "loss": 1.1508, + "step": 31015 + }, + { + "epoch": 2.61, + "learning_rate": 8.67036532955754e-07, + "loss": 1.0949, + "step": 31016 + }, + { + "epoch": 2.61, + "learning_rate": 8.666663498761541e-07, + "loss": 1.0714, + "step": 31017 + }, + { + "epoch": 2.61, + "learning_rate": 8.662962422590771e-07, + "loss": 1.1513, + "step": 31018 + }, + { + "epoch": 2.61, + "learning_rate": 8.659262101075872e-07, + "loss": 1.1432, + "step": 31019 + }, + { + "epoch": 2.61, + "learning_rate": 8.655562534247363e-07, + "loss": 1.1447, + "step": 31020 + }, + { + "epoch": 2.61, + "learning_rate": 8.651863722135812e-07, + "loss": 1.1421, + "step": 31021 + }, + { + "epoch": 2.61, + "learning_rate": 8.648165664771824e-07, + "loss": 1.0003, + "step": 31022 + }, + { + "epoch": 2.61, + "learning_rate": 8.644468362185921e-07, + "loss": 1.0843, + "step": 31023 + }, + { + "epoch": 2.61, + "learning_rate": 8.640771814408655e-07, + "loss": 1.0778, + "step": 31024 + }, + { + "epoch": 2.61, + "learning_rate": 8.637076021470581e-07, + "loss": 1.1821, + "step": 31025 + }, + { + "epoch": 2.61, + "learning_rate": 8.633380983402228e-07, + "loss": 1.0214, + "step": 31026 + }, + { + "epoch": 2.61, + "learning_rate": 8.629686700234097e-07, + "loss": 1.1673, + "step": 31027 + }, + { + "epoch": 2.61, + "learning_rate": 8.625993171996772e-07, + "loss": 1.1814, + "step": 31028 + }, + { + "epoch": 2.61, + "learning_rate": 8.622300398720707e-07, + "loss": 1.1815, + "step": 31029 + }, + { + "epoch": 2.61, + "learning_rate": 8.618608380436444e-07, + "loss": 1.139, + "step": 31030 + }, + { + "epoch": 2.61, + "learning_rate": 8.614917117174493e-07, + "loss": 1.0554, + "step": 31031 + }, + { + "epoch": 2.61, + "learning_rate": 8.611226608965339e-07, + "loss": 1.139, + "step": 31032 + }, + { + "epoch": 2.61, + "learning_rate": 8.607536855839472e-07, + "loss": 1.1018, + "step": 31033 + }, + { + "epoch": 2.61, + "learning_rate": 8.603847857827397e-07, + "loss": 0.9869, + "step": 31034 + }, + { + "epoch": 2.61, + "learning_rate": 8.600159614959569e-07, + "loss": 1.1536, + "step": 31035 + }, + { + "epoch": 2.61, + "learning_rate": 8.596472127266487e-07, + "loss": 1.1131, + "step": 31036 + }, + { + "epoch": 2.61, + "learning_rate": 8.592785394778613e-07, + "loss": 1.1174, + "step": 31037 + }, + { + "epoch": 2.61, + "learning_rate": 8.58909941752637e-07, + "loss": 1.0687, + "step": 31038 + }, + { + "epoch": 2.61, + "learning_rate": 8.585414195540275e-07, + "loss": 1.0795, + "step": 31039 + }, + { + "epoch": 2.61, + "learning_rate": 8.58172972885073e-07, + "loss": 1.0835, + "step": 31040 + }, + { + "epoch": 2.61, + "learning_rate": 8.578046017488185e-07, + "loss": 1.1204, + "step": 31041 + }, + { + "epoch": 2.61, + "learning_rate": 8.574363061483082e-07, + "loss": 1.0621, + "step": 31042 + }, + { + "epoch": 2.61, + "learning_rate": 8.570680860865865e-07, + "loss": 1.145, + "step": 31043 + }, + { + "epoch": 2.61, + "learning_rate": 8.566999415666932e-07, + "loss": 1.1914, + "step": 31044 + }, + { + "epoch": 2.61, + "learning_rate": 8.563318725916736e-07, + "loss": 1.0734, + "step": 31045 + }, + { + "epoch": 2.61, + "learning_rate": 8.55963879164562e-07, + "loss": 1.0964, + "step": 31046 + }, + { + "epoch": 2.61, + "learning_rate": 8.555959612884057e-07, + "loss": 1.0869, + "step": 31047 + }, + { + "epoch": 2.61, + "learning_rate": 8.552281189662436e-07, + "loss": 1.0807, + "step": 31048 + }, + { + "epoch": 2.61, + "learning_rate": 8.548603522011101e-07, + "loss": 1.0774, + "step": 31049 + }, + { + "epoch": 2.61, + "learning_rate": 8.544926609960513e-07, + "loss": 1.0887, + "step": 31050 + }, + { + "epoch": 2.61, + "learning_rate": 8.541250453540983e-07, + "loss": 1.1033, + "step": 31051 + }, + { + "epoch": 2.61, + "learning_rate": 8.53757505278292e-07, + "loss": 1.0798, + "step": 31052 + }, + { + "epoch": 2.61, + "learning_rate": 8.533900407716678e-07, + "loss": 1.1068, + "step": 31053 + }, + { + "epoch": 2.61, + "learning_rate": 8.530226518372631e-07, + "loss": 1.1123, + "step": 31054 + }, + { + "epoch": 2.61, + "learning_rate": 8.526553384781122e-07, + "loss": 1.1025, + "step": 31055 + }, + { + "epoch": 2.61, + "learning_rate": 8.522881006972528e-07, + "loss": 1.1521, + "step": 31056 + }, + { + "epoch": 2.61, + "learning_rate": 8.519209384977123e-07, + "loss": 1.1232, + "step": 31057 + }, + { + "epoch": 2.61, + "learning_rate": 8.515538518825306e-07, + "loss": 1.0858, + "step": 31058 + }, + { + "epoch": 2.61, + "learning_rate": 8.511868408547407e-07, + "loss": 1.1352, + "step": 31059 + }, + { + "epoch": 2.61, + "learning_rate": 8.508199054173705e-07, + "loss": 1.2061, + "step": 31060 + }, + { + "epoch": 2.62, + "learning_rate": 8.504530455734572e-07, + "loss": 1.122, + "step": 31061 + }, + { + "epoch": 2.62, + "learning_rate": 8.500862613260275e-07, + "loss": 1.1552, + "step": 31062 + }, + { + "epoch": 2.62, + "learning_rate": 8.497195526781133e-07, + "loss": 1.0298, + "step": 31063 + }, + { + "epoch": 2.62, + "learning_rate": 8.493529196327444e-07, + "loss": 1.1471, + "step": 31064 + }, + { + "epoch": 2.62, + "learning_rate": 8.489863621929495e-07, + "loss": 1.0971, + "step": 31065 + }, + { + "epoch": 2.62, + "learning_rate": 8.486198803617595e-07, + "loss": 1.1075, + "step": 31066 + }, + { + "epoch": 2.62, + "learning_rate": 8.482534741422022e-07, + "loss": 1.1335, + "step": 31067 + }, + { + "epoch": 2.62, + "learning_rate": 8.478871435372993e-07, + "loss": 1.1187, + "step": 31068 + }, + { + "epoch": 2.62, + "learning_rate": 8.475208885500863e-07, + "loss": 1.1369, + "step": 31069 + }, + { + "epoch": 2.62, + "learning_rate": 8.47154709183583e-07, + "loss": 1.0812, + "step": 31070 + }, + { + "epoch": 2.62, + "learning_rate": 8.467886054408147e-07, + "loss": 1.1152, + "step": 31071 + }, + { + "epoch": 2.62, + "learning_rate": 8.464225773248125e-07, + "loss": 1.0762, + "step": 31072 + }, + { + "epoch": 2.62, + "learning_rate": 8.460566248385938e-07, + "loss": 1.1439, + "step": 31073 + }, + { + "epoch": 2.62, + "learning_rate": 8.45690747985185e-07, + "loss": 1.1274, + "step": 31074 + }, + { + "epoch": 2.62, + "learning_rate": 8.453249467676095e-07, + "loss": 1.031, + "step": 31075 + }, + { + "epoch": 2.62, + "learning_rate": 8.44959221188889e-07, + "loss": 1.1842, + "step": 31076 + }, + { + "epoch": 2.62, + "learning_rate": 8.445935712520459e-07, + "loss": 1.0381, + "step": 31077 + }, + { + "epoch": 2.62, + "learning_rate": 8.442279969601019e-07, + "loss": 1.0507, + "step": 31078 + }, + { + "epoch": 2.62, + "learning_rate": 8.438624983160737e-07, + "loss": 1.0449, + "step": 31079 + }, + { + "epoch": 2.62, + "learning_rate": 8.434970753229865e-07, + "loss": 1.2035, + "step": 31080 + }, + { + "epoch": 2.62, + "learning_rate": 8.431317279838558e-07, + "loss": 1.092, + "step": 31081 + }, + { + "epoch": 2.62, + "learning_rate": 8.427664563017002e-07, + "loss": 1.0602, + "step": 31082 + }, + { + "epoch": 2.62, + "learning_rate": 8.424012602795395e-07, + "loss": 1.1986, + "step": 31083 + }, + { + "epoch": 2.62, + "learning_rate": 8.420361399203914e-07, + "loss": 1.1685, + "step": 31084 + }, + { + "epoch": 2.62, + "learning_rate": 8.416710952272699e-07, + "loss": 1.2255, + "step": 31085 + }, + { + "epoch": 2.62, + "learning_rate": 8.413061262031941e-07, + "loss": 1.1041, + "step": 31086 + }, + { + "epoch": 2.62, + "learning_rate": 8.40941232851179e-07, + "loss": 1.0765, + "step": 31087 + }, + { + "epoch": 2.62, + "learning_rate": 8.405764151742368e-07, + "loss": 1.0571, + "step": 31088 + }, + { + "epoch": 2.62, + "learning_rate": 8.402116731753862e-07, + "loss": 1.0905, + "step": 31089 + }, + { + "epoch": 2.62, + "learning_rate": 8.398470068576358e-07, + "loss": 1.1238, + "step": 31090 + }, + { + "epoch": 2.62, + "learning_rate": 8.394824162240023e-07, + "loss": 1.0193, + "step": 31091 + }, + { + "epoch": 2.62, + "learning_rate": 8.391179012774953e-07, + "loss": 1.1552, + "step": 31092 + }, + { + "epoch": 2.62, + "learning_rate": 8.387534620211291e-07, + "loss": 1.0576, + "step": 31093 + }, + { + "epoch": 2.62, + "learning_rate": 8.383890984579146e-07, + "loss": 1.0879, + "step": 31094 + }, + { + "epoch": 2.62, + "learning_rate": 8.380248105908617e-07, + "loss": 1.0886, + "step": 31095 + }, + { + "epoch": 2.62, + "learning_rate": 8.376605984229769e-07, + "loss": 1.1872, + "step": 31096 + }, + { + "epoch": 2.62, + "learning_rate": 8.372964619572743e-07, + "loss": 1.1582, + "step": 31097 + }, + { + "epoch": 2.62, + "learning_rate": 8.369324011967628e-07, + "loss": 1.0765, + "step": 31098 + }, + { + "epoch": 2.62, + "learning_rate": 8.365684161444454e-07, + "loss": 1.1753, + "step": 31099 + }, + { + "epoch": 2.62, + "learning_rate": 8.362045068033353e-07, + "loss": 1.1465, + "step": 31100 + }, + { + "epoch": 2.62, + "learning_rate": 8.358406731764346e-07, + "loss": 1.1043, + "step": 31101 + }, + { + "epoch": 2.62, + "learning_rate": 8.354769152667508e-07, + "loss": 1.1305, + "step": 31102 + }, + { + "epoch": 2.62, + "learning_rate": 8.351132330772904e-07, + "loss": 1.1551, + "step": 31103 + }, + { + "epoch": 2.62, + "learning_rate": 8.347496266110578e-07, + "loss": 1.1145, + "step": 31104 + }, + { + "epoch": 2.62, + "learning_rate": 8.343860958710559e-07, + "loss": 1.0844, + "step": 31105 + }, + { + "epoch": 2.62, + "learning_rate": 8.340226408602924e-07, + "loss": 1.0965, + "step": 31106 + }, + { + "epoch": 2.62, + "learning_rate": 8.33659261581763e-07, + "loss": 1.0943, + "step": 31107 + }, + { + "epoch": 2.62, + "learning_rate": 8.332959580384758e-07, + "loss": 1.1061, + "step": 31108 + }, + { + "epoch": 2.62, + "learning_rate": 8.329327302334345e-07, + "loss": 1.1422, + "step": 31109 + }, + { + "epoch": 2.62, + "learning_rate": 8.32569578169632e-07, + "loss": 1.11, + "step": 31110 + }, + { + "epoch": 2.62, + "learning_rate": 8.322065018500769e-07, + "loss": 1.1265, + "step": 31111 + }, + { + "epoch": 2.62, + "learning_rate": 8.318435012777648e-07, + "loss": 1.1892, + "step": 31112 + }, + { + "epoch": 2.62, + "learning_rate": 8.314805764556954e-07, + "loss": 1.0984, + "step": 31113 + }, + { + "epoch": 2.62, + "learning_rate": 8.311177273868686e-07, + "loss": 1.0033, + "step": 31114 + }, + { + "epoch": 2.62, + "learning_rate": 8.30754954074281e-07, + "loss": 1.0966, + "step": 31115 + }, + { + "epoch": 2.62, + "learning_rate": 8.303922565209299e-07, + "loss": 1.1578, + "step": 31116 + }, + { + "epoch": 2.62, + "learning_rate": 8.300296347298154e-07, + "loss": 1.0867, + "step": 31117 + }, + { + "epoch": 2.62, + "learning_rate": 8.296670887039271e-07, + "loss": 1.1305, + "step": 31118 + }, + { + "epoch": 2.62, + "learning_rate": 8.293046184462672e-07, + "loss": 1.2083, + "step": 31119 + }, + { + "epoch": 2.62, + "learning_rate": 8.289422239598255e-07, + "loss": 1.0047, + "step": 31120 + }, + { + "epoch": 2.62, + "learning_rate": 8.285799052475973e-07, + "loss": 1.1383, + "step": 31121 + }, + { + "epoch": 2.62, + "learning_rate": 8.282176623125803e-07, + "loss": 1.0558, + "step": 31122 + }, + { + "epoch": 2.62, + "learning_rate": 8.278554951577622e-07, + "loss": 1.0437, + "step": 31123 + }, + { + "epoch": 2.62, + "learning_rate": 8.274934037861382e-07, + "loss": 1.1013, + "step": 31124 + }, + { + "epoch": 2.62, + "learning_rate": 8.271313882006993e-07, + "loss": 1.1368, + "step": 31125 + }, + { + "epoch": 2.62, + "learning_rate": 8.267694484044364e-07, + "loss": 1.1258, + "step": 31126 + }, + { + "epoch": 2.62, + "learning_rate": 8.264075844003395e-07, + "loss": 1.1037, + "step": 31127 + }, + { + "epoch": 2.62, + "learning_rate": 8.260457961914015e-07, + "loss": 1.0599, + "step": 31128 + }, + { + "epoch": 2.62, + "learning_rate": 8.256840837806057e-07, + "loss": 1.1383, + "step": 31129 + }, + { + "epoch": 2.62, + "learning_rate": 8.253224471709475e-07, + "loss": 1.1021, + "step": 31130 + }, + { + "epoch": 2.62, + "learning_rate": 8.249608863654091e-07, + "loss": 1.0359, + "step": 31131 + }, + { + "epoch": 2.62, + "learning_rate": 8.245994013669811e-07, + "loss": 1.0735, + "step": 31132 + }, + { + "epoch": 2.62, + "learning_rate": 8.242379921786491e-07, + "loss": 1.1028, + "step": 31133 + }, + { + "epoch": 2.62, + "learning_rate": 8.238766588033997e-07, + "loss": 1.109, + "step": 31134 + }, + { + "epoch": 2.62, + "learning_rate": 8.235154012442181e-07, + "loss": 1.1019, + "step": 31135 + }, + { + "epoch": 2.62, + "learning_rate": 8.231542195040886e-07, + "loss": 1.1441, + "step": 31136 + }, + { + "epoch": 2.62, + "learning_rate": 8.227931135859979e-07, + "loss": 1.0945, + "step": 31137 + }, + { + "epoch": 2.62, + "learning_rate": 8.224320834929234e-07, + "loss": 1.0957, + "step": 31138 + }, + { + "epoch": 2.62, + "learning_rate": 8.220711292278572e-07, + "loss": 1.2255, + "step": 31139 + }, + { + "epoch": 2.62, + "learning_rate": 8.217102507937736e-07, + "loss": 1.1857, + "step": 31140 + }, + { + "epoch": 2.62, + "learning_rate": 8.213494481936579e-07, + "loss": 1.052, + "step": 31141 + }, + { + "epoch": 2.62, + "learning_rate": 8.209887214304901e-07, + "loss": 1.1186, + "step": 31142 + }, + { + "epoch": 2.62, + "learning_rate": 8.206280705072522e-07, + "loss": 1.0646, + "step": 31143 + }, + { + "epoch": 2.62, + "learning_rate": 8.202674954269219e-07, + "loss": 1.0753, + "step": 31144 + }, + { + "epoch": 2.62, + "learning_rate": 8.199069961924821e-07, + "loss": 1.0181, + "step": 31145 + }, + { + "epoch": 2.62, + "learning_rate": 8.195465728069041e-07, + "loss": 1.1446, + "step": 31146 + }, + { + "epoch": 2.62, + "learning_rate": 8.19186225273173e-07, + "loss": 1.0638, + "step": 31147 + }, + { + "epoch": 2.62, + "learning_rate": 8.188259535942655e-07, + "loss": 1.1091, + "step": 31148 + }, + { + "epoch": 2.62, + "learning_rate": 8.184657577731525e-07, + "loss": 1.0659, + "step": 31149 + }, + { + "epoch": 2.62, + "learning_rate": 8.181056378128171e-07, + "loss": 1.1394, + "step": 31150 + }, + { + "epoch": 2.62, + "learning_rate": 8.177455937162293e-07, + "loss": 1.1763, + "step": 31151 + }, + { + "epoch": 2.62, + "learning_rate": 8.173856254863666e-07, + "loss": 1.1446, + "step": 31152 + }, + { + "epoch": 2.62, + "learning_rate": 8.170257331262033e-07, + "loss": 1.0639, + "step": 31153 + }, + { + "epoch": 2.62, + "learning_rate": 8.166659166387114e-07, + "loss": 1.0537, + "step": 31154 + }, + { + "epoch": 2.62, + "learning_rate": 8.163061760268653e-07, + "loss": 1.1164, + "step": 31155 + }, + { + "epoch": 2.62, + "learning_rate": 8.159465112936382e-07, + "loss": 1.1557, + "step": 31156 + }, + { + "epoch": 2.62, + "learning_rate": 8.155869224419976e-07, + "loss": 1.0284, + "step": 31157 + }, + { + "epoch": 2.62, + "learning_rate": 8.15227409474918e-07, + "loss": 1.1015, + "step": 31158 + }, + { + "epoch": 2.62, + "learning_rate": 8.148679723953711e-07, + "loss": 1.1771, + "step": 31159 + }, + { + "epoch": 2.62, + "learning_rate": 8.145086112063217e-07, + "loss": 1.1593, + "step": 31160 + }, + { + "epoch": 2.62, + "learning_rate": 8.141493259107447e-07, + "loss": 1.0417, + "step": 31161 + }, + { + "epoch": 2.62, + "learning_rate": 8.137901165116035e-07, + "loss": 1.1533, + "step": 31162 + }, + { + "epoch": 2.62, + "learning_rate": 8.134309830118692e-07, + "loss": 1.13, + "step": 31163 + }, + { + "epoch": 2.62, + "learning_rate": 8.130719254145081e-07, + "loss": 1.112, + "step": 31164 + }, + { + "epoch": 2.62, + "learning_rate": 8.127129437224867e-07, + "loss": 1.1462, + "step": 31165 + }, + { + "epoch": 2.62, + "learning_rate": 8.123540379387718e-07, + "loss": 1.0561, + "step": 31166 + }, + { + "epoch": 2.62, + "learning_rate": 8.119952080663296e-07, + "loss": 1.132, + "step": 31167 + }, + { + "epoch": 2.62, + "learning_rate": 8.116364541081201e-07, + "loss": 1.0807, + "step": 31168 + }, + { + "epoch": 2.62, + "learning_rate": 8.112777760671143e-07, + "loss": 1.1017, + "step": 31169 + }, + { + "epoch": 2.62, + "learning_rate": 8.109191739462707e-07, + "loss": 1.0914, + "step": 31170 + }, + { + "epoch": 2.62, + "learning_rate": 8.105606477485517e-07, + "loss": 1.1596, + "step": 31171 + }, + { + "epoch": 2.62, + "learning_rate": 8.102021974769258e-07, + "loss": 1.1075, + "step": 31172 + }, + { + "epoch": 2.62, + "learning_rate": 8.098438231343486e-07, + "loss": 1.0825, + "step": 31173 + }, + { + "epoch": 2.62, + "learning_rate": 8.094855247237821e-07, + "loss": 1.152, + "step": 31174 + }, + { + "epoch": 2.62, + "learning_rate": 8.091273022481882e-07, + "loss": 1.1573, + "step": 31175 + }, + { + "epoch": 2.62, + "learning_rate": 8.08769155710526e-07, + "loss": 1.0922, + "step": 31176 + }, + { + "epoch": 2.62, + "learning_rate": 8.084110851137539e-07, + "loss": 1.0598, + "step": 31177 + }, + { + "epoch": 2.62, + "learning_rate": 8.080530904608341e-07, + "loss": 1.0989, + "step": 31178 + }, + { + "epoch": 2.63, + "learning_rate": 8.076951717547166e-07, + "loss": 1.0206, + "step": 31179 + }, + { + "epoch": 2.63, + "learning_rate": 8.073373289983677e-07, + "loss": 1.0665, + "step": 31180 + }, + { + "epoch": 2.63, + "learning_rate": 8.069795621947385e-07, + "loss": 1.1617, + "step": 31181 + }, + { + "epoch": 2.63, + "learning_rate": 8.066218713467855e-07, + "loss": 1.1658, + "step": 31182 + }, + { + "epoch": 2.63, + "learning_rate": 8.062642564574652e-07, + "loss": 1.1343, + "step": 31183 + }, + { + "epoch": 2.63, + "learning_rate": 8.05906717529732e-07, + "loss": 1.1043, + "step": 31184 + }, + { + "epoch": 2.63, + "learning_rate": 8.0554925456654e-07, + "loss": 1.0675, + "step": 31185 + }, + { + "epoch": 2.63, + "learning_rate": 8.051918675708426e-07, + "loss": 1.0733, + "step": 31186 + }, + { + "epoch": 2.63, + "learning_rate": 8.048345565455951e-07, + "loss": 1.0942, + "step": 31187 + }, + { + "epoch": 2.63, + "learning_rate": 8.044773214937429e-07, + "loss": 1.058, + "step": 31188 + }, + { + "epoch": 2.63, + "learning_rate": 8.04120162418246e-07, + "loss": 1.0736, + "step": 31189 + }, + { + "epoch": 2.63, + "learning_rate": 8.037630793220485e-07, + "loss": 1.101, + "step": 31190 + }, + { + "epoch": 2.63, + "learning_rate": 8.03406072208105e-07, + "loss": 1.1431, + "step": 31191 + }, + { + "epoch": 2.63, + "learning_rate": 8.030491410793628e-07, + "loss": 1.158, + "step": 31192 + }, + { + "epoch": 2.63, + "learning_rate": 8.026922859387731e-07, + "loss": 1.0672, + "step": 31193 + }, + { + "epoch": 2.63, + "learning_rate": 8.023355067892824e-07, + "loss": 1.1692, + "step": 31194 + }, + { + "epoch": 2.63, + "learning_rate": 8.019788036338416e-07, + "loss": 1.179, + "step": 31195 + }, + { + "epoch": 2.63, + "learning_rate": 8.016221764753918e-07, + "loss": 1.1353, + "step": 31196 + }, + { + "epoch": 2.63, + "learning_rate": 8.01265625316886e-07, + "loss": 1.1066, + "step": 31197 + }, + { + "epoch": 2.63, + "learning_rate": 8.009091501612687e-07, + "loss": 1.1143, + "step": 31198 + }, + { + "epoch": 2.63, + "learning_rate": 8.005527510114808e-07, + "loss": 1.1217, + "step": 31199 + }, + { + "epoch": 2.63, + "learning_rate": 8.001964278704732e-07, + "loss": 1.0969, + "step": 31200 + }, + { + "epoch": 2.63, + "learning_rate": 7.998401807411859e-07, + "loss": 1.1523, + "step": 31201 + }, + { + "epoch": 2.63, + "learning_rate": 7.994840096265632e-07, + "loss": 1.1843, + "step": 31202 + }, + { + "epoch": 2.63, + "learning_rate": 7.991279145295483e-07, + "loss": 1.0852, + "step": 31203 + }, + { + "epoch": 2.63, + "learning_rate": 7.987718954530832e-07, + "loss": 1.1288, + "step": 31204 + }, + { + "epoch": 2.63, + "learning_rate": 7.984159524001089e-07, + "loss": 1.1591, + "step": 31205 + }, + { + "epoch": 2.63, + "learning_rate": 7.980600853735687e-07, + "loss": 1.0488, + "step": 31206 + }, + { + "epoch": 2.63, + "learning_rate": 7.97704294376398e-07, + "loss": 1.2046, + "step": 31207 + }, + { + "epoch": 2.63, + "learning_rate": 7.97348579411541e-07, + "loss": 1.1605, + "step": 31208 + }, + { + "epoch": 2.63, + "learning_rate": 7.969929404819377e-07, + "loss": 1.1007, + "step": 31209 + }, + { + "epoch": 2.63, + "learning_rate": 7.966373775905201e-07, + "loss": 1.1629, + "step": 31210 + }, + { + "epoch": 2.63, + "learning_rate": 7.962818907402337e-07, + "loss": 1.0575, + "step": 31211 + }, + { + "epoch": 2.63, + "learning_rate": 7.959264799340094e-07, + "loss": 1.1411, + "step": 31212 + }, + { + "epoch": 2.63, + "learning_rate": 7.955711451747871e-07, + "loss": 1.0845, + "step": 31213 + }, + { + "epoch": 2.63, + "learning_rate": 7.952158864655013e-07, + "loss": 1.0905, + "step": 31214 + }, + { + "epoch": 2.63, + "learning_rate": 7.948607038090883e-07, + "loss": 1.1512, + "step": 31215 + }, + { + "epoch": 2.63, + "learning_rate": 7.945055972084814e-07, + "loss": 1.1586, + "step": 31216 + }, + { + "epoch": 2.63, + "learning_rate": 7.941505666666171e-07, + "loss": 1.1146, + "step": 31217 + }, + { + "epoch": 2.63, + "learning_rate": 7.937956121864243e-07, + "loss": 1.1359, + "step": 31218 + }, + { + "epoch": 2.63, + "learning_rate": 7.934407337708416e-07, + "loss": 1.2022, + "step": 31219 + }, + { + "epoch": 2.63, + "learning_rate": 7.930859314227968e-07, + "loss": 1.0703, + "step": 31220 + }, + { + "epoch": 2.63, + "learning_rate": 7.927312051452196e-07, + "loss": 1.1123, + "step": 31221 + }, + { + "epoch": 2.63, + "learning_rate": 7.923765549410489e-07, + "loss": 1.1231, + "step": 31222 + }, + { + "epoch": 2.63, + "learning_rate": 7.920219808132068e-07, + "loss": 1.125, + "step": 31223 + }, + { + "epoch": 2.63, + "learning_rate": 7.916674827646276e-07, + "loss": 1.1315, + "step": 31224 + }, + { + "epoch": 2.63, + "learning_rate": 7.913130607982378e-07, + "loss": 1.1065, + "step": 31225 + }, + { + "epoch": 2.63, + "learning_rate": 7.909587149169662e-07, + "loss": 1.148, + "step": 31226 + }, + { + "epoch": 2.63, + "learning_rate": 7.906044451237416e-07, + "loss": 1.1742, + "step": 31227 + }, + { + "epoch": 2.63, + "learning_rate": 7.902502514214916e-07, + "loss": 1.1698, + "step": 31228 + }, + { + "epoch": 2.63, + "learning_rate": 7.898961338131384e-07, + "loss": 1.1351, + "step": 31229 + }, + { + "epoch": 2.63, + "learning_rate": 7.89542092301615e-07, + "loss": 1.2135, + "step": 31230 + }, + { + "epoch": 2.63, + "learning_rate": 7.891881268898405e-07, + "loss": 1.1401, + "step": 31231 + }, + { + "epoch": 2.63, + "learning_rate": 7.888342375807412e-07, + "loss": 1.1796, + "step": 31232 + }, + { + "epoch": 2.63, + "learning_rate": 7.884804243772415e-07, + "loss": 1.1796, + "step": 31233 + }, + { + "epoch": 2.63, + "learning_rate": 7.881266872822646e-07, + "loss": 1.1726, + "step": 31234 + }, + { + "epoch": 2.63, + "learning_rate": 7.877730262987326e-07, + "loss": 1.1027, + "step": 31235 + }, + { + "epoch": 2.63, + "learning_rate": 7.874194414295688e-07, + "loss": 1.0513, + "step": 31236 + }, + { + "epoch": 2.63, + "learning_rate": 7.870659326776964e-07, + "loss": 1.1729, + "step": 31237 + }, + { + "epoch": 2.63, + "learning_rate": 7.867125000460296e-07, + "loss": 1.1386, + "step": 31238 + }, + { + "epoch": 2.63, + "learning_rate": 7.863591435374962e-07, + "loss": 1.1237, + "step": 31239 + }, + { + "epoch": 2.63, + "learning_rate": 7.860058631550105e-07, + "loss": 1.1612, + "step": 31240 + }, + { + "epoch": 2.63, + "learning_rate": 7.856526589014934e-07, + "loss": 1.1107, + "step": 31241 + }, + { + "epoch": 2.63, + "learning_rate": 7.852995307798627e-07, + "loss": 1.1472, + "step": 31242 + }, + { + "epoch": 2.63, + "learning_rate": 7.849464787930361e-07, + "loss": 1.1311, + "step": 31243 + }, + { + "epoch": 2.63, + "learning_rate": 7.845935029439311e-07, + "loss": 1.1055, + "step": 31244 + }, + { + "epoch": 2.63, + "learning_rate": 7.842406032354665e-07, + "loss": 1.0386, + "step": 31245 + }, + { + "epoch": 2.63, + "learning_rate": 7.838877796705513e-07, + "loss": 1.1248, + "step": 31246 + }, + { + "epoch": 2.63, + "learning_rate": 7.835350322521062e-07, + "loss": 1.1311, + "step": 31247 + }, + { + "epoch": 2.63, + "learning_rate": 7.831823609830469e-07, + "loss": 1.0348, + "step": 31248 + }, + { + "epoch": 2.63, + "learning_rate": 7.82829765866281e-07, + "loss": 1.0505, + "step": 31249 + }, + { + "epoch": 2.63, + "learning_rate": 7.824772469047281e-07, + "loss": 1.1943, + "step": 31250 + }, + { + "epoch": 2.63, + "learning_rate": 7.821248041012974e-07, + "loss": 1.0841, + "step": 31251 + }, + { + "epoch": 2.63, + "learning_rate": 7.817724374589009e-07, + "loss": 1.1912, + "step": 31252 + }, + { + "epoch": 2.63, + "learning_rate": 7.814201469804506e-07, + "loss": 1.0951, + "step": 31253 + }, + { + "epoch": 2.63, + "learning_rate": 7.810679326688586e-07, + "loss": 1.1386, + "step": 31254 + }, + { + "epoch": 2.63, + "learning_rate": 7.807157945270327e-07, + "loss": 1.0248, + "step": 31255 + }, + { + "epoch": 2.63, + "learning_rate": 7.803637325578861e-07, + "loss": 1.062, + "step": 31256 + }, + { + "epoch": 2.63, + "learning_rate": 7.800117467643209e-07, + "loss": 1.1162, + "step": 31257 + }, + { + "epoch": 2.63, + "learning_rate": 7.796598371492503e-07, + "loss": 1.1341, + "step": 31258 + }, + { + "epoch": 2.63, + "learning_rate": 7.793080037155843e-07, + "loss": 1.122, + "step": 31259 + }, + { + "epoch": 2.63, + "learning_rate": 7.789562464662226e-07, + "loss": 1.0838, + "step": 31260 + }, + { + "epoch": 2.63, + "learning_rate": 7.786045654040785e-07, + "loss": 1.154, + "step": 31261 + }, + { + "epoch": 2.63, + "learning_rate": 7.782529605320532e-07, + "loss": 1.0558, + "step": 31262 + }, + { + "epoch": 2.63, + "learning_rate": 7.779014318530531e-07, + "loss": 1.1357, + "step": 31263 + }, + { + "epoch": 2.63, + "learning_rate": 7.775499793699837e-07, + "loss": 1.1752, + "step": 31264 + }, + { + "epoch": 2.63, + "learning_rate": 7.77198603085747e-07, + "loss": 1.0948, + "step": 31265 + }, + { + "epoch": 2.63, + "learning_rate": 7.768473030032475e-07, + "loss": 1.1449, + "step": 31266 + }, + { + "epoch": 2.63, + "learning_rate": 7.764960791253895e-07, + "loss": 1.0876, + "step": 31267 + }, + { + "epoch": 2.63, + "learning_rate": 7.761449314550685e-07, + "loss": 1.1065, + "step": 31268 + }, + { + "epoch": 2.63, + "learning_rate": 7.757938599951942e-07, + "loss": 1.0443, + "step": 31269 + }, + { + "epoch": 2.63, + "learning_rate": 7.754428647486612e-07, + "loss": 1.0516, + "step": 31270 + }, + { + "epoch": 2.63, + "learning_rate": 7.750919457183692e-07, + "loss": 1.0693, + "step": 31271 + }, + { + "epoch": 2.63, + "learning_rate": 7.747411029072238e-07, + "loss": 1.1421, + "step": 31272 + }, + { + "epoch": 2.63, + "learning_rate": 7.743903363181182e-07, + "loss": 1.0601, + "step": 31273 + }, + { + "epoch": 2.63, + "learning_rate": 7.740396459539523e-07, + "loss": 1.1488, + "step": 31274 + }, + { + "epoch": 2.63, + "learning_rate": 7.736890318176238e-07, + "loss": 1.1399, + "step": 31275 + }, + { + "epoch": 2.63, + "learning_rate": 7.733384939120292e-07, + "loss": 1.093, + "step": 31276 + }, + { + "epoch": 2.63, + "learning_rate": 7.72988032240064e-07, + "loss": 1.1056, + "step": 31277 + }, + { + "epoch": 2.63, + "learning_rate": 7.726376468046271e-07, + "loss": 1.1482, + "step": 31278 + }, + { + "epoch": 2.63, + "learning_rate": 7.722873376086082e-07, + "loss": 1.1051, + "step": 31279 + }, + { + "epoch": 2.63, + "learning_rate": 7.719371046549073e-07, + "loss": 1.1252, + "step": 31280 + }, + { + "epoch": 2.63, + "learning_rate": 7.715869479464133e-07, + "loss": 1.1021, + "step": 31281 + }, + { + "epoch": 2.63, + "learning_rate": 7.712368674860227e-07, + "loss": 1.0715, + "step": 31282 + }, + { + "epoch": 2.63, + "learning_rate": 7.708868632766253e-07, + "loss": 1.1164, + "step": 31283 + }, + { + "epoch": 2.63, + "learning_rate": 7.705369353211156e-07, + "loss": 1.2257, + "step": 31284 + }, + { + "epoch": 2.63, + "learning_rate": 7.701870836223823e-07, + "loss": 1.0866, + "step": 31285 + }, + { + "epoch": 2.63, + "learning_rate": 7.698373081833177e-07, + "loss": 1.1918, + "step": 31286 + }, + { + "epoch": 2.63, + "learning_rate": 7.694876090068137e-07, + "loss": 1.1042, + "step": 31287 + }, + { + "epoch": 2.63, + "learning_rate": 7.691379860957526e-07, + "loss": 1.1278, + "step": 31288 + }, + { + "epoch": 2.63, + "learning_rate": 7.68788439453032e-07, + "loss": 1.018, + "step": 31289 + }, + { + "epoch": 2.63, + "learning_rate": 7.684389690815342e-07, + "loss": 1.0438, + "step": 31290 + }, + { + "epoch": 2.63, + "learning_rate": 7.680895749841478e-07, + "loss": 1.0975, + "step": 31291 + }, + { + "epoch": 2.63, + "learning_rate": 7.677402571637593e-07, + "loss": 1.0757, + "step": 31292 + }, + { + "epoch": 2.63, + "learning_rate": 7.673910156232556e-07, + "loss": 1.0237, + "step": 31293 + }, + { + "epoch": 2.63, + "learning_rate": 7.67041850365523e-07, + "loss": 1.0932, + "step": 31294 + }, + { + "epoch": 2.63, + "learning_rate": 7.666927613934461e-07, + "loss": 1.1796, + "step": 31295 + }, + { + "epoch": 2.63, + "learning_rate": 7.663437487099057e-07, + "loss": 1.0615, + "step": 31296 + }, + { + "epoch": 2.63, + "learning_rate": 7.659948123177908e-07, + "loss": 1.1198, + "step": 31297 + }, + { + "epoch": 2.64, + "learning_rate": 7.656459522199833e-07, + "loss": 1.1561, + "step": 31298 + }, + { + "epoch": 2.64, + "learning_rate": 7.65297168419361e-07, + "loss": 1.0092, + "step": 31299 + }, + { + "epoch": 2.64, + "learning_rate": 7.649484609188118e-07, + "loss": 1.0157, + "step": 31300 + }, + { + "epoch": 2.64, + "learning_rate": 7.645998297212132e-07, + "loss": 1.0557, + "step": 31301 + }, + { + "epoch": 2.64, + "learning_rate": 7.642512748294462e-07, + "loss": 1.1228, + "step": 31302 + }, + { + "epoch": 2.64, + "learning_rate": 7.639027962463907e-07, + "loss": 1.0863, + "step": 31303 + }, + { + "epoch": 2.64, + "learning_rate": 7.635543939749257e-07, + "loss": 1.0589, + "step": 31304 + }, + { + "epoch": 2.64, + "learning_rate": 7.63206068017931e-07, + "loss": 1.2103, + "step": 31305 + }, + { + "epoch": 2.64, + "learning_rate": 7.628578183782854e-07, + "loss": 1.1305, + "step": 31306 + }, + { + "epoch": 2.64, + "learning_rate": 7.625096450588621e-07, + "loss": 1.0518, + "step": 31307 + }, + { + "epoch": 2.64, + "learning_rate": 7.62161548062541e-07, + "loss": 1.1822, + "step": 31308 + }, + { + "epoch": 2.64, + "learning_rate": 7.618135273922001e-07, + "loss": 1.0531, + "step": 31309 + }, + { + "epoch": 2.64, + "learning_rate": 7.614655830507089e-07, + "loss": 1.1468, + "step": 31310 + }, + { + "epoch": 2.64, + "learning_rate": 7.611177150409499e-07, + "loss": 1.1054, + "step": 31311 + }, + { + "epoch": 2.64, + "learning_rate": 7.607699233657906e-07, + "loss": 1.1078, + "step": 31312 + }, + { + "epoch": 2.64, + "learning_rate": 7.604222080281077e-07, + "loss": 1.0694, + "step": 31313 + }, + { + "epoch": 2.64, + "learning_rate": 7.600745690307731e-07, + "loss": 1.1093, + "step": 31314 + }, + { + "epoch": 2.64, + "learning_rate": 7.597270063766593e-07, + "loss": 1.1583, + "step": 31315 + }, + { + "epoch": 2.64, + "learning_rate": 7.593795200686383e-07, + "loss": 1.0485, + "step": 31316 + }, + { + "epoch": 2.64, + "learning_rate": 7.590321101095833e-07, + "loss": 1.076, + "step": 31317 + }, + { + "epoch": 2.64, + "learning_rate": 7.586847765023586e-07, + "loss": 1.1784, + "step": 31318 + }, + { + "epoch": 2.64, + "learning_rate": 7.583375192498422e-07, + "loss": 1.0832, + "step": 31319 + }, + { + "epoch": 2.64, + "learning_rate": 7.579903383548959e-07, + "loss": 1.1427, + "step": 31320 + }, + { + "epoch": 2.64, + "learning_rate": 7.57643233820391e-07, + "loss": 1.08, + "step": 31321 + }, + { + "epoch": 2.64, + "learning_rate": 7.572962056491972e-07, + "loss": 1.0807, + "step": 31322 + }, + { + "epoch": 2.64, + "learning_rate": 7.569492538441803e-07, + "loss": 1.0693, + "step": 31323 + }, + { + "epoch": 2.64, + "learning_rate": 7.566023784082055e-07, + "loss": 1.0933, + "step": 31324 + }, + { + "epoch": 2.64, + "learning_rate": 7.562555793441407e-07, + "loss": 1.0806, + "step": 31325 + }, + { + "epoch": 2.64, + "learning_rate": 7.559088566548511e-07, + "loss": 1.0363, + "step": 31326 + }, + { + "epoch": 2.64, + "learning_rate": 7.555622103432003e-07, + "loss": 1.0599, + "step": 31327 + }, + { + "epoch": 2.64, + "learning_rate": 7.552156404120547e-07, + "loss": 1.0471, + "step": 31328 + }, + { + "epoch": 2.64, + "learning_rate": 7.548691468642733e-07, + "loss": 1.1031, + "step": 31329 + }, + { + "epoch": 2.64, + "learning_rate": 7.545227297027258e-07, + "loss": 1.0552, + "step": 31330 + }, + { + "epoch": 2.64, + "learning_rate": 7.541763889302678e-07, + "loss": 1.1522, + "step": 31331 + }, + { + "epoch": 2.64, + "learning_rate": 7.538301245497637e-07, + "loss": 1.1113, + "step": 31332 + }, + { + "epoch": 2.64, + "learning_rate": 7.534839365640745e-07, + "loss": 1.222, + "step": 31333 + }, + { + "epoch": 2.64, + "learning_rate": 7.531378249760601e-07, + "loss": 1.1256, + "step": 31334 + }, + { + "epoch": 2.64, + "learning_rate": 7.527917897885806e-07, + "loss": 1.1341, + "step": 31335 + }, + { + "epoch": 2.64, + "learning_rate": 7.524458310044935e-07, + "loss": 1.0966, + "step": 31336 + }, + { + "epoch": 2.64, + "learning_rate": 7.520999486266623e-07, + "loss": 1.0988, + "step": 31337 + }, + { + "epoch": 2.64, + "learning_rate": 7.517541426579367e-07, + "loss": 1.1434, + "step": 31338 + }, + { + "epoch": 2.64, + "learning_rate": 7.514084131011812e-07, + "loss": 1.1181, + "step": 31339 + }, + { + "epoch": 2.64, + "learning_rate": 7.51062759959248e-07, + "loss": 1.089, + "step": 31340 + }, + { + "epoch": 2.64, + "learning_rate": 7.507171832349946e-07, + "loss": 1.1477, + "step": 31341 + }, + { + "epoch": 2.64, + "learning_rate": 7.503716829312757e-07, + "loss": 1.118, + "step": 31342 + }, + { + "epoch": 2.64, + "learning_rate": 7.500262590509455e-07, + "loss": 1.1075, + "step": 31343 + }, + { + "epoch": 2.64, + "learning_rate": 7.496809115968595e-07, + "loss": 1.1553, + "step": 31344 + }, + { + "epoch": 2.64, + "learning_rate": 7.493356405718721e-07, + "loss": 1.163, + "step": 31345 + }, + { + "epoch": 2.64, + "learning_rate": 7.4899044597883e-07, + "loss": 1.0778, + "step": 31346 + }, + { + "epoch": 2.64, + "learning_rate": 7.486453278205919e-07, + "loss": 1.0926, + "step": 31347 + }, + { + "epoch": 2.64, + "learning_rate": 7.483002861000088e-07, + "loss": 1.1111, + "step": 31348 + }, + { + "epoch": 2.64, + "learning_rate": 7.479553208199264e-07, + "loss": 1.152, + "step": 31349 + }, + { + "epoch": 2.64, + "learning_rate": 7.476104319832012e-07, + "loss": 1.1409, + "step": 31350 + }, + { + "epoch": 2.64, + "learning_rate": 7.472656195926776e-07, + "loss": 1.1129, + "step": 31351 + }, + { + "epoch": 2.64, + "learning_rate": 7.469208836512077e-07, + "loss": 1.0882, + "step": 31352 + }, + { + "epoch": 2.64, + "learning_rate": 7.465762241616382e-07, + "loss": 1.0789, + "step": 31353 + }, + { + "epoch": 2.64, + "learning_rate": 7.462316411268178e-07, + "loss": 1.1129, + "step": 31354 + }, + { + "epoch": 2.64, + "learning_rate": 7.458871345495922e-07, + "loss": 1.1414, + "step": 31355 + }, + { + "epoch": 2.64, + "learning_rate": 7.455427044328112e-07, + "loss": 1.2053, + "step": 31356 + }, + { + "epoch": 2.64, + "learning_rate": 7.451983507793148e-07, + "loss": 1.1041, + "step": 31357 + }, + { + "epoch": 2.64, + "learning_rate": 7.448540735919529e-07, + "loss": 1.1336, + "step": 31358 + }, + { + "epoch": 2.64, + "learning_rate": 7.445098728735712e-07, + "loss": 1.2336, + "step": 31359 + }, + { + "epoch": 2.64, + "learning_rate": 7.441657486270071e-07, + "loss": 1.0771, + "step": 31360 + }, + { + "epoch": 2.64, + "learning_rate": 7.438217008551107e-07, + "loss": 1.1579, + "step": 31361 + }, + { + "epoch": 2.64, + "learning_rate": 7.434777295607209e-07, + "loss": 1.2632, + "step": 31362 + }, + { + "epoch": 2.64, + "learning_rate": 7.431338347466799e-07, + "loss": 1.2259, + "step": 31363 + }, + { + "epoch": 2.64, + "learning_rate": 7.427900164158308e-07, + "loss": 1.1401, + "step": 31364 + }, + { + "epoch": 2.64, + "learning_rate": 7.424462745710126e-07, + "loss": 1.1338, + "step": 31365 + }, + { + "epoch": 2.64, + "learning_rate": 7.421026092150663e-07, + "loss": 1.141, + "step": 31366 + }, + { + "epoch": 2.64, + "learning_rate": 7.417590203508318e-07, + "loss": 1.1603, + "step": 31367 + }, + { + "epoch": 2.64, + "learning_rate": 7.414155079811447e-07, + "loss": 1.112, + "step": 31368 + }, + { + "epoch": 2.64, + "learning_rate": 7.410720721088493e-07, + "loss": 1.095, + "step": 31369 + }, + { + "epoch": 2.64, + "learning_rate": 7.40728712736778e-07, + "loss": 1.2114, + "step": 31370 + }, + { + "epoch": 2.64, + "learning_rate": 7.403854298677682e-07, + "loss": 1.1134, + "step": 31371 + }, + { + "epoch": 2.64, + "learning_rate": 7.400422235046601e-07, + "loss": 1.0987, + "step": 31372 + }, + { + "epoch": 2.64, + "learning_rate": 7.396990936502845e-07, + "loss": 1.0945, + "step": 31373 + }, + { + "epoch": 2.64, + "learning_rate": 7.393560403074796e-07, + "loss": 1.0722, + "step": 31374 + }, + { + "epoch": 2.64, + "learning_rate": 7.390130634790782e-07, + "loss": 1.1922, + "step": 31375 + }, + { + "epoch": 2.64, + "learning_rate": 7.38670163167915e-07, + "loss": 1.1259, + "step": 31376 + }, + { + "epoch": 2.64, + "learning_rate": 7.383273393768231e-07, + "loss": 1.0844, + "step": 31377 + }, + { + "epoch": 2.64, + "learning_rate": 7.37984592108637e-07, + "loss": 1.1299, + "step": 31378 + }, + { + "epoch": 2.64, + "learning_rate": 7.376419213661823e-07, + "loss": 1.1173, + "step": 31379 + }, + { + "epoch": 2.64, + "learning_rate": 7.372993271522977e-07, + "loss": 1.0667, + "step": 31380 + }, + { + "epoch": 2.64, + "learning_rate": 7.369568094698098e-07, + "loss": 1.0843, + "step": 31381 + }, + { + "epoch": 2.64, + "learning_rate": 7.366143683215487e-07, + "loss": 1.1444, + "step": 31382 + }, + { + "epoch": 2.64, + "learning_rate": 7.362720037103443e-07, + "loss": 1.0699, + "step": 31383 + }, + { + "epoch": 2.64, + "learning_rate": 7.359297156390255e-07, + "loss": 1.1225, + "step": 31384 + }, + { + "epoch": 2.64, + "learning_rate": 7.355875041104199e-07, + "loss": 1.1776, + "step": 31385 + }, + { + "epoch": 2.64, + "learning_rate": 7.352453691273553e-07, + "loss": 1.1737, + "step": 31386 + }, + { + "epoch": 2.64, + "learning_rate": 7.349033106926584e-07, + "loss": 1.0817, + "step": 31387 + }, + { + "epoch": 2.64, + "learning_rate": 7.345613288091568e-07, + "loss": 1.1276, + "step": 31388 + }, + { + "epoch": 2.64, + "learning_rate": 7.342194234796751e-07, + "loss": 1.1177, + "step": 31389 + }, + { + "epoch": 2.64, + "learning_rate": 7.338775947070364e-07, + "loss": 1.1204, + "step": 31390 + }, + { + "epoch": 2.64, + "learning_rate": 7.335358424940675e-07, + "loss": 1.1802, + "step": 31391 + }, + { + "epoch": 2.64, + "learning_rate": 7.331941668435904e-07, + "loss": 0.9938, + "step": 31392 + }, + { + "epoch": 2.64, + "learning_rate": 7.328525677584297e-07, + "loss": 1.1344, + "step": 31393 + }, + { + "epoch": 2.64, + "learning_rate": 7.325110452414064e-07, + "loss": 1.1991, + "step": 31394 + }, + { + "epoch": 2.64, + "learning_rate": 7.321695992953437e-07, + "loss": 1.0947, + "step": 31395 + }, + { + "epoch": 2.64, + "learning_rate": 7.318282299230595e-07, + "loss": 1.1865, + "step": 31396 + }, + { + "epoch": 2.64, + "learning_rate": 7.31486937127378e-07, + "loss": 1.072, + "step": 31397 + }, + { + "epoch": 2.64, + "learning_rate": 7.311457209111195e-07, + "loss": 1.0767, + "step": 31398 + }, + { + "epoch": 2.64, + "learning_rate": 7.308045812770981e-07, + "loss": 1.1408, + "step": 31399 + }, + { + "epoch": 2.64, + "learning_rate": 7.304635182281395e-07, + "loss": 1.0874, + "step": 31400 + }, + { + "epoch": 2.64, + "learning_rate": 7.301225317670557e-07, + "loss": 1.1664, + "step": 31401 + }, + { + "epoch": 2.64, + "learning_rate": 7.297816218966658e-07, + "loss": 1.1453, + "step": 31402 + }, + { + "epoch": 2.64, + "learning_rate": 7.294407886197885e-07, + "loss": 1.1809, + "step": 31403 + }, + { + "epoch": 2.64, + "learning_rate": 7.291000319392372e-07, + "loss": 1.1254, + "step": 31404 + }, + { + "epoch": 2.64, + "learning_rate": 7.287593518578284e-07, + "loss": 1.1674, + "step": 31405 + }, + { + "epoch": 2.64, + "learning_rate": 7.284187483783788e-07, + "loss": 1.1584, + "step": 31406 + }, + { + "epoch": 2.64, + "learning_rate": 7.280782215036986e-07, + "loss": 1.1366, + "step": 31407 + }, + { + "epoch": 2.64, + "learning_rate": 7.277377712366041e-07, + "loss": 1.0544, + "step": 31408 + }, + { + "epoch": 2.64, + "learning_rate": 7.273973975799098e-07, + "loss": 1.1182, + "step": 31409 + }, + { + "epoch": 2.64, + "learning_rate": 7.270571005364213e-07, + "loss": 1.158, + "step": 31410 + }, + { + "epoch": 2.64, + "learning_rate": 7.267168801089597e-07, + "loss": 1.1665, + "step": 31411 + }, + { + "epoch": 2.64, + "learning_rate": 7.263767363003283e-07, + "loss": 1.1386, + "step": 31412 + }, + { + "epoch": 2.64, + "learning_rate": 7.260366691133413e-07, + "loss": 1.0212, + "step": 31413 + }, + { + "epoch": 2.64, + "learning_rate": 7.256966785508068e-07, + "loss": 1.0501, + "step": 31414 + }, + { + "epoch": 2.64, + "learning_rate": 7.253567646155335e-07, + "loss": 1.1701, + "step": 31415 + }, + { + "epoch": 2.64, + "learning_rate": 7.250169273103325e-07, + "loss": 1.1229, + "step": 31416 + }, + { + "epoch": 2.65, + "learning_rate": 7.246771666380104e-07, + "loss": 1.1821, + "step": 31417 + }, + { + "epoch": 2.65, + "learning_rate": 7.243374826013716e-07, + "loss": 1.0573, + "step": 31418 + }, + { + "epoch": 2.65, + "learning_rate": 7.239978752032284e-07, + "loss": 1.1923, + "step": 31419 + }, + { + "epoch": 2.65, + "learning_rate": 7.236583444463818e-07, + "loss": 1.0756, + "step": 31420 + }, + { + "epoch": 2.65, + "learning_rate": 7.233188903336364e-07, + "loss": 1.0535, + "step": 31421 + }, + { + "epoch": 2.65, + "learning_rate": 7.229795128678029e-07, + "loss": 1.0611, + "step": 31422 + }, + { + "epoch": 2.65, + "learning_rate": 7.226402120516807e-07, + "loss": 1.153, + "step": 31423 + }, + { + "epoch": 2.65, + "learning_rate": 7.223009878880727e-07, + "loss": 1.1852, + "step": 31424 + }, + { + "epoch": 2.65, + "learning_rate": 7.219618403797846e-07, + "loss": 1.0351, + "step": 31425 + }, + { + "epoch": 2.65, + "learning_rate": 7.216227695296163e-07, + "loss": 1.1178, + "step": 31426 + }, + { + "epoch": 2.65, + "learning_rate": 7.212837753403712e-07, + "loss": 1.1161, + "step": 31427 + }, + { + "epoch": 2.65, + "learning_rate": 7.209448578148492e-07, + "loss": 1.0987, + "step": 31428 + }, + { + "epoch": 2.65, + "learning_rate": 7.206060169558482e-07, + "loss": 1.1845, + "step": 31429 + }, + { + "epoch": 2.65, + "learning_rate": 7.202672527661736e-07, + "loss": 1.0359, + "step": 31430 + }, + { + "epoch": 2.65, + "learning_rate": 7.199285652486187e-07, + "loss": 1.1323, + "step": 31431 + }, + { + "epoch": 2.65, + "learning_rate": 7.195899544059837e-07, + "loss": 1.1856, + "step": 31432 + }, + { + "epoch": 2.65, + "learning_rate": 7.192514202410672e-07, + "loss": 1.0599, + "step": 31433 + }, + { + "epoch": 2.65, + "learning_rate": 7.18912962756666e-07, + "loss": 1.2065, + "step": 31434 + }, + { + "epoch": 2.65, + "learning_rate": 7.185745819555756e-07, + "loss": 1.115, + "step": 31435 + }, + { + "epoch": 2.65, + "learning_rate": 7.182362778405927e-07, + "loss": 1.1422, + "step": 31436 + }, + { + "epoch": 2.65, + "learning_rate": 7.178980504145116e-07, + "loss": 1.1356, + "step": 31437 + }, + { + "epoch": 2.65, + "learning_rate": 7.175598996801281e-07, + "loss": 1.0464, + "step": 31438 + }, + { + "epoch": 2.65, + "learning_rate": 7.172218256402374e-07, + "loss": 1.1047, + "step": 31439 + }, + { + "epoch": 2.65, + "learning_rate": 7.168838282976265e-07, + "loss": 1.1501, + "step": 31440 + }, + { + "epoch": 2.65, + "learning_rate": 7.165459076550963e-07, + "loss": 1.2203, + "step": 31441 + }, + { + "epoch": 2.65, + "learning_rate": 7.162080637154334e-07, + "loss": 1.1894, + "step": 31442 + }, + { + "epoch": 2.65, + "learning_rate": 7.158702964814312e-07, + "loss": 1.0483, + "step": 31443 + }, + { + "epoch": 2.65, + "learning_rate": 7.155326059558787e-07, + "loss": 1.1949, + "step": 31444 + }, + { + "epoch": 2.65, + "learning_rate": 7.151949921415691e-07, + "loss": 1.1353, + "step": 31445 + }, + { + "epoch": 2.65, + "learning_rate": 7.148574550412878e-07, + "loss": 1.0291, + "step": 31446 + }, + { + "epoch": 2.65, + "learning_rate": 7.145199946578263e-07, + "loss": 1.1047, + "step": 31447 + }, + { + "epoch": 2.65, + "learning_rate": 7.141826109939742e-07, + "loss": 1.1659, + "step": 31448 + }, + { + "epoch": 2.65, + "learning_rate": 7.13845304052514e-07, + "loss": 1.2029, + "step": 31449 + }, + { + "epoch": 2.65, + "learning_rate": 7.135080738362398e-07, + "loss": 0.9892, + "step": 31450 + }, + { + "epoch": 2.65, + "learning_rate": 7.131709203479309e-07, + "loss": 1.1241, + "step": 31451 + }, + { + "epoch": 2.65, + "learning_rate": 7.12833843590377e-07, + "loss": 1.0927, + "step": 31452 + }, + { + "epoch": 2.65, + "learning_rate": 7.124968435663626e-07, + "loss": 1.1436, + "step": 31453 + }, + { + "epoch": 2.65, + "learning_rate": 7.121599202786711e-07, + "loss": 1.0948, + "step": 31454 + }, + { + "epoch": 2.65, + "learning_rate": 7.118230737300869e-07, + "loss": 1.0751, + "step": 31455 + }, + { + "epoch": 2.65, + "learning_rate": 7.114863039233943e-07, + "loss": 1.1019, + "step": 31456 + }, + { + "epoch": 2.65, + "learning_rate": 7.111496108613713e-07, + "loss": 1.0698, + "step": 31457 + }, + { + "epoch": 2.65, + "learning_rate": 7.108129945468057e-07, + "loss": 1.0562, + "step": 31458 + }, + { + "epoch": 2.65, + "learning_rate": 7.104764549824761e-07, + "loss": 1.1691, + "step": 31459 + }, + { + "epoch": 2.65, + "learning_rate": 7.101399921711605e-07, + "loss": 1.0434, + "step": 31460 + }, + { + "epoch": 2.65, + "learning_rate": 7.098036061156454e-07, + "loss": 1.1458, + "step": 31461 + }, + { + "epoch": 2.65, + "learning_rate": 7.094672968187033e-07, + "loss": 1.0972, + "step": 31462 + }, + { + "epoch": 2.65, + "learning_rate": 7.091310642831173e-07, + "loss": 1.0308, + "step": 31463 + }, + { + "epoch": 2.65, + "learning_rate": 7.08794908511663e-07, + "loss": 1.158, + "step": 31464 + }, + { + "epoch": 2.65, + "learning_rate": 7.084588295071182e-07, + "loss": 1.153, + "step": 31465 + }, + { + "epoch": 2.65, + "learning_rate": 7.081228272722607e-07, + "loss": 1.1425, + "step": 31466 + }, + { + "epoch": 2.65, + "learning_rate": 7.077869018098682e-07, + "loss": 0.9936, + "step": 31467 + }, + { + "epoch": 2.65, + "learning_rate": 7.074510531227108e-07, + "loss": 1.1348, + "step": 31468 + }, + { + "epoch": 2.65, + "learning_rate": 7.071152812135695e-07, + "loss": 1.124, + "step": 31469 + }, + { + "epoch": 2.65, + "learning_rate": 7.067795860852145e-07, + "loss": 1.0418, + "step": 31470 + }, + { + "epoch": 2.65, + "learning_rate": 7.064439677404189e-07, + "loss": 1.1521, + "step": 31471 + }, + { + "epoch": 2.65, + "learning_rate": 7.061084261819617e-07, + "loss": 1.0853, + "step": 31472 + }, + { + "epoch": 2.65, + "learning_rate": 7.057729614126085e-07, + "loss": 1.2107, + "step": 31473 + }, + { + "epoch": 2.65, + "learning_rate": 7.054375734351337e-07, + "loss": 1.1155, + "step": 31474 + }, + { + "epoch": 2.65, + "learning_rate": 7.051022622523096e-07, + "loss": 1.1583, + "step": 31475 + }, + { + "epoch": 2.65, + "learning_rate": 7.047670278669039e-07, + "loss": 1.0454, + "step": 31476 + }, + { + "epoch": 2.65, + "learning_rate": 7.044318702816888e-07, + "loss": 1.1294, + "step": 31477 + }, + { + "epoch": 2.65, + "learning_rate": 7.040967894994332e-07, + "loss": 1.1422, + "step": 31478 + }, + { + "epoch": 2.65, + "learning_rate": 7.037617855229029e-07, + "loss": 1.0777, + "step": 31479 + }, + { + "epoch": 2.65, + "learning_rate": 7.034268583548698e-07, + "loss": 1.173, + "step": 31480 + }, + { + "epoch": 2.65, + "learning_rate": 7.030920079980985e-07, + "loss": 1.1524, + "step": 31481 + }, + { + "epoch": 2.65, + "learning_rate": 7.027572344553568e-07, + "loss": 1.0648, + "step": 31482 + }, + { + "epoch": 2.65, + "learning_rate": 7.024225377294103e-07, + "loss": 1.0516, + "step": 31483 + }, + { + "epoch": 2.65, + "learning_rate": 7.020879178230233e-07, + "loss": 1.0769, + "step": 31484 + }, + { + "epoch": 2.65, + "learning_rate": 7.017533747389626e-07, + "loss": 1.0635, + "step": 31485 + }, + { + "epoch": 2.65, + "learning_rate": 7.014189084799905e-07, + "loss": 1.0179, + "step": 31486 + }, + { + "epoch": 2.65, + "learning_rate": 7.010845190488724e-07, + "loss": 1.0817, + "step": 31487 + }, + { + "epoch": 2.65, + "learning_rate": 7.007502064483695e-07, + "loss": 1.133, + "step": 31488 + }, + { + "epoch": 2.65, + "learning_rate": 7.004159706812464e-07, + "loss": 1.2236, + "step": 31489 + }, + { + "epoch": 2.65, + "learning_rate": 7.000818117502583e-07, + "loss": 1.0713, + "step": 31490 + }, + { + "epoch": 2.65, + "learning_rate": 6.997477296581756e-07, + "loss": 1.091, + "step": 31491 + }, + { + "epoch": 2.65, + "learning_rate": 6.994137244077514e-07, + "loss": 1.1491, + "step": 31492 + }, + { + "epoch": 2.65, + "learning_rate": 6.99079796001747e-07, + "loss": 1.0704, + "step": 31493 + }, + { + "epoch": 2.65, + "learning_rate": 6.987459444429223e-07, + "loss": 1.1207, + "step": 31494 + }, + { + "epoch": 2.65, + "learning_rate": 6.984121697340373e-07, + "loss": 1.1362, + "step": 31495 + }, + { + "epoch": 2.65, + "learning_rate": 6.980784718778444e-07, + "loss": 1.095, + "step": 31496 + }, + { + "epoch": 2.65, + "learning_rate": 6.977448508771057e-07, + "loss": 1.1288, + "step": 31497 + }, + { + "epoch": 2.65, + "learning_rate": 6.974113067345778e-07, + "loss": 1.2039, + "step": 31498 + }, + { + "epoch": 2.65, + "learning_rate": 6.97077839453012e-07, + "loss": 1.0514, + "step": 31499 + }, + { + "epoch": 2.65, + "learning_rate": 6.967444490351694e-07, + "loss": 1.2891, + "step": 31500 + }, + { + "epoch": 2.65, + "learning_rate": 6.964111354837999e-07, + "loss": 1.0684, + "step": 31501 + }, + { + "epoch": 2.65, + "learning_rate": 6.960778988016592e-07, + "loss": 1.1082, + "step": 31502 + }, + { + "epoch": 2.65, + "learning_rate": 6.957447389915006e-07, + "loss": 1.119, + "step": 31503 + }, + { + "epoch": 2.65, + "learning_rate": 6.954116560560764e-07, + "loss": 1.0611, + "step": 31504 + }, + { + "epoch": 2.65, + "learning_rate": 6.950786499981399e-07, + "loss": 1.1071, + "step": 31505 + }, + { + "epoch": 2.65, + "learning_rate": 6.947457208204422e-07, + "loss": 1.1926, + "step": 31506 + }, + { + "epoch": 2.65, + "learning_rate": 6.944128685257301e-07, + "loss": 1.0361, + "step": 31507 + }, + { + "epoch": 2.65, + "learning_rate": 6.940800931167591e-07, + "loss": 1.0902, + "step": 31508 + }, + { + "epoch": 2.65, + "learning_rate": 6.93747394596278e-07, + "loss": 1.1223, + "step": 31509 + }, + { + "epoch": 2.65, + "learning_rate": 6.934147729670315e-07, + "loss": 1.1383, + "step": 31510 + }, + { + "epoch": 2.65, + "learning_rate": 6.930822282317739e-07, + "loss": 1.0952, + "step": 31511 + }, + { + "epoch": 2.65, + "learning_rate": 6.927497603932476e-07, + "loss": 1.0941, + "step": 31512 + }, + { + "epoch": 2.65, + "learning_rate": 6.924173694542013e-07, + "loss": 1.0711, + "step": 31513 + }, + { + "epoch": 2.65, + "learning_rate": 6.92085055417382e-07, + "loss": 1.1038, + "step": 31514 + }, + { + "epoch": 2.65, + "learning_rate": 6.917528182855348e-07, + "loss": 1.1814, + "step": 31515 + }, + { + "epoch": 2.65, + "learning_rate": 6.914206580614046e-07, + "loss": 1.0726, + "step": 31516 + }, + { + "epoch": 2.65, + "learning_rate": 6.91088574747737e-07, + "loss": 1.1226, + "step": 31517 + }, + { + "epoch": 2.65, + "learning_rate": 6.907565683472728e-07, + "loss": 1.1983, + "step": 31518 + }, + { + "epoch": 2.65, + "learning_rate": 6.9042463886276e-07, + "loss": 1.1563, + "step": 31519 + }, + { + "epoch": 2.65, + "learning_rate": 6.900927862969365e-07, + "loss": 1.1865, + "step": 31520 + }, + { + "epoch": 2.65, + "learning_rate": 6.897610106525443e-07, + "loss": 1.1302, + "step": 31521 + }, + { + "epoch": 2.65, + "learning_rate": 6.894293119323292e-07, + "loss": 1.1003, + "step": 31522 + }, + { + "epoch": 2.65, + "learning_rate": 6.890976901390278e-07, + "loss": 1.1228, + "step": 31523 + }, + { + "epoch": 2.65, + "learning_rate": 6.887661452753813e-07, + "loss": 1.0066, + "step": 31524 + }, + { + "epoch": 2.65, + "learning_rate": 6.884346773441286e-07, + "loss": 1.0924, + "step": 31525 + }, + { + "epoch": 2.65, + "learning_rate": 6.881032863480086e-07, + "loss": 1.118, + "step": 31526 + }, + { + "epoch": 2.65, + "learning_rate": 6.877719722897591e-07, + "loss": 0.9331, + "step": 31527 + }, + { + "epoch": 2.65, + "learning_rate": 6.874407351721202e-07, + "loss": 1.1767, + "step": 31528 + }, + { + "epoch": 2.65, + "learning_rate": 6.87109574997823e-07, + "loss": 1.158, + "step": 31529 + }, + { + "epoch": 2.65, + "learning_rate": 6.867784917696096e-07, + "loss": 1.135, + "step": 31530 + }, + { + "epoch": 2.65, + "learning_rate": 6.864474854902126e-07, + "loss": 1.212, + "step": 31531 + }, + { + "epoch": 2.65, + "learning_rate": 6.861165561623662e-07, + "loss": 1.103, + "step": 31532 + }, + { + "epoch": 2.65, + "learning_rate": 6.857857037888061e-07, + "loss": 1.0697, + "step": 31533 + }, + { + "epoch": 2.65, + "learning_rate": 6.854549283722644e-07, + "loss": 1.0868, + "step": 31534 + }, + { + "epoch": 2.65, + "learning_rate": 6.851242299154759e-07, + "loss": 1.0736, + "step": 31535 + }, + { + "epoch": 2.66, + "learning_rate": 6.847936084211726e-07, + "loss": 1.0967, + "step": 31536 + }, + { + "epoch": 2.66, + "learning_rate": 6.844630638920846e-07, + "loss": 1.1002, + "step": 31537 + }, + { + "epoch": 2.66, + "learning_rate": 6.841325963309453e-07, + "loss": 1.0851, + "step": 31538 + }, + { + "epoch": 2.66, + "learning_rate": 6.838022057404848e-07, + "loss": 1.0827, + "step": 31539 + }, + { + "epoch": 2.66, + "learning_rate": 6.834718921234295e-07, + "loss": 1.0741, + "step": 31540 + }, + { + "epoch": 2.66, + "learning_rate": 6.831416554825133e-07, + "loss": 1.2252, + "step": 31541 + }, + { + "epoch": 2.66, + "learning_rate": 6.828114958204612e-07, + "loss": 1.1475, + "step": 31542 + }, + { + "epoch": 2.66, + "learning_rate": 6.824814131400037e-07, + "loss": 1.102, + "step": 31543 + }, + { + "epoch": 2.66, + "learning_rate": 6.821514074438651e-07, + "loss": 1.1332, + "step": 31544 + }, + { + "epoch": 2.66, + "learning_rate": 6.818214787347766e-07, + "loss": 1.1523, + "step": 31545 + }, + { + "epoch": 2.66, + "learning_rate": 6.81491627015457e-07, + "loss": 1.0956, + "step": 31546 + }, + { + "epoch": 2.66, + "learning_rate": 6.811618522886376e-07, + "loss": 1.2072, + "step": 31547 + }, + { + "epoch": 2.66, + "learning_rate": 6.808321545570429e-07, + "loss": 1.1427, + "step": 31548 + }, + { + "epoch": 2.66, + "learning_rate": 6.805025338233929e-07, + "loss": 1.0619, + "step": 31549 + }, + { + "epoch": 2.66, + "learning_rate": 6.801729900904163e-07, + "loss": 1.103, + "step": 31550 + }, + { + "epoch": 2.66, + "learning_rate": 6.798435233608325e-07, + "loss": 1.1385, + "step": 31551 + }, + { + "epoch": 2.66, + "learning_rate": 6.795141336373634e-07, + "loss": 1.2294, + "step": 31552 + }, + { + "epoch": 2.66, + "learning_rate": 6.791848209227314e-07, + "loss": 1.0972, + "step": 31553 + }, + { + "epoch": 2.66, + "learning_rate": 6.788555852196577e-07, + "loss": 1.1414, + "step": 31554 + }, + { + "epoch": 2.66, + "learning_rate": 6.785264265308622e-07, + "loss": 1.1031, + "step": 31555 + }, + { + "epoch": 2.66, + "learning_rate": 6.781973448590661e-07, + "loss": 1.1774, + "step": 31556 + }, + { + "epoch": 2.66, + "learning_rate": 6.778683402069841e-07, + "loss": 1.1505, + "step": 31557 + }, + { + "epoch": 2.66, + "learning_rate": 6.775394125773394e-07, + "loss": 1.1216, + "step": 31558 + }, + { + "epoch": 2.66, + "learning_rate": 6.772105619728486e-07, + "loss": 1.1803, + "step": 31559 + }, + { + "epoch": 2.66, + "learning_rate": 6.768817883962242e-07, + "loss": 1.0793, + "step": 31560 + }, + { + "epoch": 2.66, + "learning_rate": 6.765530918501895e-07, + "loss": 1.1524, + "step": 31561 + }, + { + "epoch": 2.66, + "learning_rate": 6.762244723374567e-07, + "loss": 1.1646, + "step": 31562 + }, + { + "epoch": 2.66, + "learning_rate": 6.758959298607404e-07, + "loss": 1.056, + "step": 31563 + }, + { + "epoch": 2.66, + "learning_rate": 6.755674644227561e-07, + "loss": 1.0754, + "step": 31564 + }, + { + "epoch": 2.66, + "learning_rate": 6.752390760262174e-07, + "loss": 1.1833, + "step": 31565 + }, + { + "epoch": 2.66, + "learning_rate": 6.749107646738373e-07, + "loss": 1.1124, + "step": 31566 + }, + { + "epoch": 2.66, + "learning_rate": 6.745825303683318e-07, + "loss": 1.191, + "step": 31567 + }, + { + "epoch": 2.66, + "learning_rate": 6.742543731124051e-07, + "loss": 1.136, + "step": 31568 + }, + { + "epoch": 2.66, + "learning_rate": 6.739262929087775e-07, + "loss": 1.025, + "step": 31569 + }, + { + "epoch": 2.66, + "learning_rate": 6.735982897601545e-07, + "loss": 0.9899, + "step": 31570 + }, + { + "epoch": 2.66, + "learning_rate": 6.73270363669245e-07, + "loss": 1.083, + "step": 31571 + }, + { + "epoch": 2.66, + "learning_rate": 6.729425146387635e-07, + "loss": 1.1545, + "step": 31572 + }, + { + "epoch": 2.66, + "learning_rate": 6.726147426714158e-07, + "loss": 1.1062, + "step": 31573 + }, + { + "epoch": 2.66, + "learning_rate": 6.722870477699095e-07, + "loss": 1.1845, + "step": 31574 + }, + { + "epoch": 2.66, + "learning_rate": 6.719594299369525e-07, + "loss": 1.0971, + "step": 31575 + }, + { + "epoch": 2.66, + "learning_rate": 6.716318891752527e-07, + "loss": 1.0709, + "step": 31576 + }, + { + "epoch": 2.66, + "learning_rate": 6.713044254875145e-07, + "loss": 1.0949, + "step": 31577 + }, + { + "epoch": 2.66, + "learning_rate": 6.70977038876447e-07, + "loss": 1.0597, + "step": 31578 + }, + { + "epoch": 2.66, + "learning_rate": 6.7064972934475e-07, + "loss": 1.0676, + "step": 31579 + }, + { + "epoch": 2.66, + "learning_rate": 6.703224968951339e-07, + "loss": 1.1844, + "step": 31580 + }, + { + "epoch": 2.66, + "learning_rate": 6.699953415302973e-07, + "loss": 1.033, + "step": 31581 + }, + { + "epoch": 2.66, + "learning_rate": 6.69668263252945e-07, + "loss": 1.0234, + "step": 31582 + }, + { + "epoch": 2.66, + "learning_rate": 6.693412620657802e-07, + "loss": 1.202, + "step": 31583 + }, + { + "epoch": 2.66, + "learning_rate": 6.690143379715042e-07, + "loss": 1.014, + "step": 31584 + }, + { + "epoch": 2.66, + "learning_rate": 6.686874909728169e-07, + "loss": 1.078, + "step": 31585 + }, + { + "epoch": 2.66, + "learning_rate": 6.683607210724207e-07, + "loss": 1.163, + "step": 31586 + }, + { + "epoch": 2.66, + "learning_rate": 6.680340282730147e-07, + "loss": 1.1844, + "step": 31587 + }, + { + "epoch": 2.66, + "learning_rate": 6.677074125772987e-07, + "loss": 1.1015, + "step": 31588 + }, + { + "epoch": 2.66, + "learning_rate": 6.673808739879717e-07, + "loss": 1.1051, + "step": 31589 + }, + { + "epoch": 2.66, + "learning_rate": 6.670544125077272e-07, + "loss": 1.1402, + "step": 31590 + }, + { + "epoch": 2.66, + "learning_rate": 6.667280281392696e-07, + "loss": 1.0701, + "step": 31591 + }, + { + "epoch": 2.66, + "learning_rate": 6.664017208852902e-07, + "loss": 1.0754, + "step": 31592 + }, + { + "epoch": 2.66, + "learning_rate": 6.660754907484879e-07, + "loss": 1.1952, + "step": 31593 + }, + { + "epoch": 2.66, + "learning_rate": 6.657493377315561e-07, + "loss": 1.0457, + "step": 31594 + }, + { + "epoch": 2.66, + "learning_rate": 6.654232618371926e-07, + "loss": 1.2109, + "step": 31595 + }, + { + "epoch": 2.66, + "learning_rate": 6.650972630680863e-07, + "loss": 1.1666, + "step": 31596 + }, + { + "epoch": 2.66, + "learning_rate": 6.647713414269364e-07, + "loss": 1.1179, + "step": 31597 + }, + { + "epoch": 2.66, + "learning_rate": 6.644454969164338e-07, + "loss": 1.1175, + "step": 31598 + }, + { + "epoch": 2.66, + "learning_rate": 6.641197295392677e-07, + "loss": 1.1341, + "step": 31599 + }, + { + "epoch": 2.66, + "learning_rate": 6.637940392981346e-07, + "loss": 1.0946, + "step": 31600 + }, + { + "epoch": 2.66, + "learning_rate": 6.634684261957225e-07, + "loss": 1.1864, + "step": 31601 + }, + { + "epoch": 2.66, + "learning_rate": 6.631428902347214e-07, + "loss": 1.2509, + "step": 31602 + }, + { + "epoch": 2.66, + "learning_rate": 6.628174314178227e-07, + "loss": 1.055, + "step": 31603 + }, + { + "epoch": 2.66, + "learning_rate": 6.62492049747715e-07, + "loss": 1.1391, + "step": 31604 + }, + { + "epoch": 2.66, + "learning_rate": 6.621667452270852e-07, + "loss": 1.1766, + "step": 31605 + }, + { + "epoch": 2.66, + "learning_rate": 6.618415178586246e-07, + "loss": 1.1427, + "step": 31606 + }, + { + "epoch": 2.66, + "learning_rate": 6.615163676450154e-07, + "loss": 1.1489, + "step": 31607 + }, + { + "epoch": 2.66, + "learning_rate": 6.611912945889476e-07, + "loss": 1.1549, + "step": 31608 + }, + { + "epoch": 2.66, + "learning_rate": 6.608662986931069e-07, + "loss": 1.1499, + "step": 31609 + }, + { + "epoch": 2.66, + "learning_rate": 6.605413799601756e-07, + "loss": 1.1299, + "step": 31610 + }, + { + "epoch": 2.66, + "learning_rate": 6.602165383928428e-07, + "loss": 1.1349, + "step": 31611 + }, + { + "epoch": 2.66, + "learning_rate": 6.598917739937894e-07, + "loss": 1.1141, + "step": 31612 + }, + { + "epoch": 2.66, + "learning_rate": 6.595670867656989e-07, + "loss": 1.0358, + "step": 31613 + }, + { + "epoch": 2.66, + "learning_rate": 6.592424767112548e-07, + "loss": 1.064, + "step": 31614 + }, + { + "epoch": 2.66, + "learning_rate": 6.589179438331384e-07, + "loss": 1.1422, + "step": 31615 + }, + { + "epoch": 2.66, + "learning_rate": 6.585934881340317e-07, + "loss": 1.0958, + "step": 31616 + }, + { + "epoch": 2.66, + "learning_rate": 6.58269109616616e-07, + "loss": 1.1189, + "step": 31617 + }, + { + "epoch": 2.66, + "learning_rate": 6.579448082835682e-07, + "loss": 1.1134, + "step": 31618 + }, + { + "epoch": 2.66, + "learning_rate": 6.576205841375738e-07, + "loss": 1.1736, + "step": 31619 + }, + { + "epoch": 2.66, + "learning_rate": 6.572964371813062e-07, + "loss": 1.0427, + "step": 31620 + }, + { + "epoch": 2.66, + "learning_rate": 6.569723674174433e-07, + "loss": 1.1176, + "step": 31621 + }, + { + "epoch": 2.66, + "learning_rate": 6.566483748486685e-07, + "loss": 1.1659, + "step": 31622 + }, + { + "epoch": 2.66, + "learning_rate": 6.563244594776541e-07, + "loss": 1.2202, + "step": 31623 + }, + { + "epoch": 2.66, + "learning_rate": 6.560006213070758e-07, + "loss": 1.0897, + "step": 31624 + }, + { + "epoch": 2.66, + "learning_rate": 6.556768603396124e-07, + "loss": 1.1721, + "step": 31625 + }, + { + "epoch": 2.66, + "learning_rate": 6.553531765779364e-07, + "loss": 1.0907, + "step": 31626 + }, + { + "epoch": 2.66, + "learning_rate": 6.550295700247233e-07, + "loss": 1.1289, + "step": 31627 + }, + { + "epoch": 2.66, + "learning_rate": 6.547060406826489e-07, + "loss": 0.9859, + "step": 31628 + }, + { + "epoch": 2.66, + "learning_rate": 6.543825885543808e-07, + "loss": 1.1416, + "step": 31629 + }, + { + "epoch": 2.66, + "learning_rate": 6.540592136425972e-07, + "loss": 1.1041, + "step": 31630 + }, + { + "epoch": 2.66, + "learning_rate": 6.537359159499657e-07, + "loss": 1.1193, + "step": 31631 + }, + { + "epoch": 2.66, + "learning_rate": 6.534126954791597e-07, + "loss": 1.1373, + "step": 31632 + }, + { + "epoch": 2.66, + "learning_rate": 6.530895522328495e-07, + "loss": 1.1084, + "step": 31633 + }, + { + "epoch": 2.66, + "learning_rate": 6.52766486213705e-07, + "loss": 1.118, + "step": 31634 + }, + { + "epoch": 2.66, + "learning_rate": 6.524434974243942e-07, + "loss": 1.1516, + "step": 31635 + }, + { + "epoch": 2.66, + "learning_rate": 6.521205858675883e-07, + "loss": 1.1419, + "step": 31636 + }, + { + "epoch": 2.66, + "learning_rate": 6.517977515459528e-07, + "loss": 1.1197, + "step": 31637 + }, + { + "epoch": 2.66, + "learning_rate": 6.514749944621568e-07, + "loss": 1.2108, + "step": 31638 + }, + { + "epoch": 2.66, + "learning_rate": 6.51152314618867e-07, + "loss": 1.1284, + "step": 31639 + }, + { + "epoch": 2.66, + "learning_rate": 6.508297120187468e-07, + "loss": 1.1287, + "step": 31640 + }, + { + "epoch": 2.66, + "learning_rate": 6.505071866644652e-07, + "loss": 1.0716, + "step": 31641 + }, + { + "epoch": 2.66, + "learning_rate": 6.501847385586857e-07, + "loss": 1.2072, + "step": 31642 + }, + { + "epoch": 2.66, + "learning_rate": 6.498623677040705e-07, + "loss": 1.1324, + "step": 31643 + }, + { + "epoch": 2.66, + "learning_rate": 6.495400741032854e-07, + "loss": 1.1436, + "step": 31644 + }, + { + "epoch": 2.66, + "learning_rate": 6.492178577589936e-07, + "loss": 1.1904, + "step": 31645 + }, + { + "epoch": 2.66, + "learning_rate": 6.488957186738553e-07, + "loss": 1.226, + "step": 31646 + }, + { + "epoch": 2.66, + "learning_rate": 6.485736568505341e-07, + "loss": 1.1344, + "step": 31647 + }, + { + "epoch": 2.66, + "learning_rate": 6.48251672291692e-07, + "loss": 1.077, + "step": 31648 + }, + { + "epoch": 2.66, + "learning_rate": 6.479297649999839e-07, + "loss": 1.1731, + "step": 31649 + }, + { + "epoch": 2.66, + "learning_rate": 6.476079349780762e-07, + "loss": 1.1928, + "step": 31650 + }, + { + "epoch": 2.66, + "learning_rate": 6.472861822286236e-07, + "loss": 1.0594, + "step": 31651 + }, + { + "epoch": 2.66, + "learning_rate": 6.469645067542852e-07, + "loss": 1.2203, + "step": 31652 + }, + { + "epoch": 2.66, + "learning_rate": 6.466429085577197e-07, + "loss": 1.0574, + "step": 31653 + }, + { + "epoch": 2.67, + "learning_rate": 6.463213876415842e-07, + "loss": 1.1523, + "step": 31654 + }, + { + "epoch": 2.67, + "learning_rate": 6.459999440085352e-07, + "loss": 1.056, + "step": 31655 + }, + { + "epoch": 2.67, + "learning_rate": 6.456785776612295e-07, + "loss": 1.1716, + "step": 31656 + }, + { + "epoch": 2.67, + "learning_rate": 6.453572886023174e-07, + "loss": 1.1533, + "step": 31657 + }, + { + "epoch": 2.67, + "learning_rate": 6.450360768344599e-07, + "loss": 1.0623, + "step": 31658 + }, + { + "epoch": 2.67, + "learning_rate": 6.447149423603094e-07, + "loss": 1.1308, + "step": 31659 + }, + { + "epoch": 2.67, + "learning_rate": 6.443938851825149e-07, + "loss": 1.071, + "step": 31660 + }, + { + "epoch": 2.67, + "learning_rate": 6.440729053037353e-07, + "loss": 1.192, + "step": 31661 + }, + { + "epoch": 2.67, + "learning_rate": 6.437520027266187e-07, + "loss": 1.156, + "step": 31662 + }, + { + "epoch": 2.67, + "learning_rate": 6.434311774538171e-07, + "loss": 1.0345, + "step": 31663 + }, + { + "epoch": 2.67, + "learning_rate": 6.431104294879808e-07, + "loss": 1.0999, + "step": 31664 + }, + { + "epoch": 2.67, + "learning_rate": 6.427897588317622e-07, + "loss": 1.0425, + "step": 31665 + }, + { + "epoch": 2.67, + "learning_rate": 6.424691654878079e-07, + "loss": 1.0947, + "step": 31666 + }, + { + "epoch": 2.67, + "learning_rate": 6.421486494587703e-07, + "loss": 1.0638, + "step": 31667 + }, + { + "epoch": 2.67, + "learning_rate": 6.418282107472928e-07, + "loss": 1.0299, + "step": 31668 + }, + { + "epoch": 2.67, + "learning_rate": 6.415078493560279e-07, + "loss": 0.9931, + "step": 31669 + }, + { + "epoch": 2.67, + "learning_rate": 6.411875652876199e-07, + "loss": 1.1323, + "step": 31670 + }, + { + "epoch": 2.67, + "learning_rate": 6.408673585447123e-07, + "loss": 1.1027, + "step": 31671 + }, + { + "epoch": 2.67, + "learning_rate": 6.405472291299586e-07, + "loss": 1.0882, + "step": 31672 + }, + { + "epoch": 2.67, + "learning_rate": 6.402271770459967e-07, + "loss": 1.0989, + "step": 31673 + }, + { + "epoch": 2.67, + "learning_rate": 6.399072022954733e-07, + "loss": 1.0763, + "step": 31674 + }, + { + "epoch": 2.67, + "learning_rate": 6.395873048810319e-07, + "loss": 1.0932, + "step": 31675 + }, + { + "epoch": 2.67, + "learning_rate": 6.392674848053171e-07, + "loss": 1.1022, + "step": 31676 + }, + { + "epoch": 2.67, + "learning_rate": 6.389477420709689e-07, + "loss": 1.0935, + "step": 31677 + }, + { + "epoch": 2.67, + "learning_rate": 6.38628076680633e-07, + "loss": 1.0047, + "step": 31678 + }, + { + "epoch": 2.67, + "learning_rate": 6.38308488636944e-07, + "loss": 1.0501, + "step": 31679 + }, + { + "epoch": 2.67, + "learning_rate": 6.379889779425497e-07, + "loss": 1.0868, + "step": 31680 + }, + { + "epoch": 2.67, + "learning_rate": 6.376695446000846e-07, + "loss": 1.0814, + "step": 31681 + }, + { + "epoch": 2.67, + "learning_rate": 6.373501886121913e-07, + "loss": 1.0351, + "step": 31682 + }, + { + "epoch": 2.67, + "learning_rate": 6.370309099815064e-07, + "loss": 1.0658, + "step": 31683 + }, + { + "epoch": 2.67, + "learning_rate": 6.367117087106689e-07, + "loss": 1.096, + "step": 31684 + }, + { + "epoch": 2.67, + "learning_rate": 6.363925848023156e-07, + "loss": 1.1593, + "step": 31685 + }, + { + "epoch": 2.67, + "learning_rate": 6.360735382590844e-07, + "loss": 1.1657, + "step": 31686 + }, + { + "epoch": 2.67, + "learning_rate": 6.357545690836097e-07, + "loss": 1.1681, + "step": 31687 + }, + { + "epoch": 2.67, + "learning_rate": 6.354356772785275e-07, + "loss": 1.157, + "step": 31688 + }, + { + "epoch": 2.67, + "learning_rate": 6.351168628464744e-07, + "loss": 1.1373, + "step": 31689 + }, + { + "epoch": 2.67, + "learning_rate": 6.347981257900804e-07, + "loss": 1.1255, + "step": 31690 + }, + { + "epoch": 2.67, + "learning_rate": 6.344794661119835e-07, + "loss": 1.1168, + "step": 31691 + }, + { + "epoch": 2.67, + "learning_rate": 6.341608838148139e-07, + "loss": 1.2, + "step": 31692 + }, + { + "epoch": 2.67, + "learning_rate": 6.338423789012049e-07, + "loss": 1.0493, + "step": 31693 + }, + { + "epoch": 2.67, + "learning_rate": 6.335239513737868e-07, + "loss": 1.1154, + "step": 31694 + }, + { + "epoch": 2.67, + "learning_rate": 6.332056012351928e-07, + "loss": 1.1555, + "step": 31695 + }, + { + "epoch": 2.67, + "learning_rate": 6.328873284880511e-07, + "loss": 1.1072, + "step": 31696 + }, + { + "epoch": 2.67, + "learning_rate": 6.325691331349915e-07, + "loss": 1.1478, + "step": 31697 + }, + { + "epoch": 2.67, + "learning_rate": 6.322510151786454e-07, + "loss": 1.065, + "step": 31698 + }, + { + "epoch": 2.67, + "learning_rate": 6.319329746216374e-07, + "loss": 1.1313, + "step": 31699 + }, + { + "epoch": 2.67, + "learning_rate": 6.316150114665998e-07, + "loss": 1.1095, + "step": 31700 + }, + { + "epoch": 2.67, + "learning_rate": 6.31297125716156e-07, + "loss": 1.0693, + "step": 31701 + }, + { + "epoch": 2.67, + "learning_rate": 6.309793173729328e-07, + "loss": 1.1449, + "step": 31702 + }, + { + "epoch": 2.67, + "learning_rate": 6.306615864395571e-07, + "loss": 1.1873, + "step": 31703 + }, + { + "epoch": 2.67, + "learning_rate": 6.303439329186545e-07, + "loss": 0.9892, + "step": 31704 + }, + { + "epoch": 2.67, + "learning_rate": 6.300263568128496e-07, + "loss": 1.1449, + "step": 31705 + }, + { + "epoch": 2.67, + "learning_rate": 6.297088581247668e-07, + "loss": 1.0775, + "step": 31706 + }, + { + "epoch": 2.67, + "learning_rate": 6.293914368570241e-07, + "loss": 1.1339, + "step": 31707 + }, + { + "epoch": 2.67, + "learning_rate": 6.290740930122507e-07, + "loss": 1.1237, + "step": 31708 + }, + { + "epoch": 2.67, + "learning_rate": 6.287568265930677e-07, + "loss": 1.1332, + "step": 31709 + }, + { + "epoch": 2.67, + "learning_rate": 6.284396376020929e-07, + "loss": 1.124, + "step": 31710 + }, + { + "epoch": 2.67, + "learning_rate": 6.28122526041951e-07, + "loss": 1.0803, + "step": 31711 + }, + { + "epoch": 2.67, + "learning_rate": 6.278054919152588e-07, + "loss": 1.2062, + "step": 31712 + }, + { + "epoch": 2.67, + "learning_rate": 6.274885352246374e-07, + "loss": 1.0363, + "step": 31713 + }, + { + "epoch": 2.67, + "learning_rate": 6.271716559727048e-07, + "loss": 1.125, + "step": 31714 + }, + { + "epoch": 2.67, + "learning_rate": 6.2685485416208e-07, + "loss": 1.048, + "step": 31715 + }, + { + "epoch": 2.67, + "learning_rate": 6.265381297953799e-07, + "loss": 1.092, + "step": 31716 + }, + { + "epoch": 2.67, + "learning_rate": 6.262214828752245e-07, + "loss": 1.0711, + "step": 31717 + }, + { + "epoch": 2.67, + "learning_rate": 6.259049134042228e-07, + "loss": 1.0679, + "step": 31718 + }, + { + "epoch": 2.67, + "learning_rate": 6.255884213849983e-07, + "loss": 1.1709, + "step": 31719 + }, + { + "epoch": 2.67, + "learning_rate": 6.2527200682016e-07, + "loss": 1.0013, + "step": 31720 + }, + { + "epoch": 2.67, + "learning_rate": 6.249556697123249e-07, + "loss": 1.0876, + "step": 31721 + }, + { + "epoch": 2.67, + "learning_rate": 6.246394100641085e-07, + "loss": 1.1287, + "step": 31722 + }, + { + "epoch": 2.67, + "learning_rate": 6.243232278781198e-07, + "loss": 1.0991, + "step": 31723 + }, + { + "epoch": 2.67, + "learning_rate": 6.240071231569745e-07, + "loss": 1.0674, + "step": 31724 + }, + { + "epoch": 2.67, + "learning_rate": 6.236910959032816e-07, + "loss": 1.193, + "step": 31725 + }, + { + "epoch": 2.67, + "learning_rate": 6.233751461196546e-07, + "loss": 1.0459, + "step": 31726 + }, + { + "epoch": 2.67, + "learning_rate": 6.230592738087027e-07, + "loss": 0.9962, + "step": 31727 + }, + { + "epoch": 2.67, + "learning_rate": 6.22743478973038e-07, + "loss": 1.1208, + "step": 31728 + }, + { + "epoch": 2.67, + "learning_rate": 6.224277616152641e-07, + "loss": 1.1103, + "step": 31729 + }, + { + "epoch": 2.67, + "learning_rate": 6.221121217379966e-07, + "loss": 1.0896, + "step": 31730 + }, + { + "epoch": 2.67, + "learning_rate": 6.217965593438391e-07, + "loss": 1.1245, + "step": 31731 + }, + { + "epoch": 2.67, + "learning_rate": 6.214810744353993e-07, + "loss": 1.0848, + "step": 31732 + }, + { + "epoch": 2.67, + "learning_rate": 6.211656670152843e-07, + "loss": 1.1021, + "step": 31733 + }, + { + "epoch": 2.67, + "learning_rate": 6.208503370860996e-07, + "loss": 1.1349, + "step": 31734 + }, + { + "epoch": 2.67, + "learning_rate": 6.205350846504521e-07, + "loss": 1.1985, + "step": 31735 + }, + { + "epoch": 2.67, + "learning_rate": 6.202199097109452e-07, + "loss": 1.0801, + "step": 31736 + }, + { + "epoch": 2.67, + "learning_rate": 6.199048122701834e-07, + "loss": 1.1281, + "step": 31737 + }, + { + "epoch": 2.67, + "learning_rate": 6.195897923307703e-07, + "loss": 1.119, + "step": 31738 + }, + { + "epoch": 2.67, + "learning_rate": 6.192748498953094e-07, + "loss": 1.1381, + "step": 31739 + }, + { + "epoch": 2.67, + "learning_rate": 6.189599849663996e-07, + "loss": 1.0877, + "step": 31740 + }, + { + "epoch": 2.67, + "learning_rate": 6.186451975466467e-07, + "loss": 1.153, + "step": 31741 + }, + { + "epoch": 2.67, + "learning_rate": 6.183304876386487e-07, + "loss": 1.0419, + "step": 31742 + }, + { + "epoch": 2.67, + "learning_rate": 6.180158552450066e-07, + "loss": 1.0976, + "step": 31743 + }, + { + "epoch": 2.67, + "learning_rate": 6.177013003683207e-07, + "loss": 1.1094, + "step": 31744 + }, + { + "epoch": 2.67, + "learning_rate": 6.173868230111891e-07, + "loss": 1.1034, + "step": 31745 + }, + { + "epoch": 2.67, + "learning_rate": 6.170724231762093e-07, + "loss": 1.0726, + "step": 31746 + }, + { + "epoch": 2.67, + "learning_rate": 6.167581008659817e-07, + "loss": 1.0851, + "step": 31747 + }, + { + "epoch": 2.67, + "learning_rate": 6.16443856083101e-07, + "loss": 1.1305, + "step": 31748 + }, + { + "epoch": 2.67, + "learning_rate": 6.161296888301638e-07, + "loss": 1.0658, + "step": 31749 + }, + { + "epoch": 2.67, + "learning_rate": 6.158155991097681e-07, + "loss": 1.139, + "step": 31750 + }, + { + "epoch": 2.67, + "learning_rate": 6.155015869245051e-07, + "loss": 1.0902, + "step": 31751 + }, + { + "epoch": 2.67, + "learning_rate": 6.151876522769728e-07, + "loss": 1.0721, + "step": 31752 + }, + { + "epoch": 2.67, + "learning_rate": 6.148737951697637e-07, + "loss": 1.1001, + "step": 31753 + }, + { + "epoch": 2.67, + "learning_rate": 6.145600156054699e-07, + "loss": 1.1314, + "step": 31754 + }, + { + "epoch": 2.67, + "learning_rate": 6.142463135866861e-07, + "loss": 1.2126, + "step": 31755 + }, + { + "epoch": 2.67, + "learning_rate": 6.139326891160036e-07, + "loss": 1.0321, + "step": 31756 + }, + { + "epoch": 2.67, + "learning_rate": 6.136191421960102e-07, + "loss": 1.0495, + "step": 31757 + }, + { + "epoch": 2.67, + "learning_rate": 6.133056728293008e-07, + "loss": 1.1866, + "step": 31758 + }, + { + "epoch": 2.67, + "learning_rate": 6.129922810184663e-07, + "loss": 1.088, + "step": 31759 + }, + { + "epoch": 2.67, + "learning_rate": 6.126789667660904e-07, + "loss": 1.0905, + "step": 31760 + }, + { + "epoch": 2.67, + "learning_rate": 6.123657300747687e-07, + "loss": 1.0323, + "step": 31761 + }, + { + "epoch": 2.67, + "learning_rate": 6.120525709470837e-07, + "loss": 1.0822, + "step": 31762 + }, + { + "epoch": 2.67, + "learning_rate": 6.117394893856243e-07, + "loss": 1.1803, + "step": 31763 + }, + { + "epoch": 2.67, + "learning_rate": 6.114264853929785e-07, + "loss": 1.0904, + "step": 31764 + }, + { + "epoch": 2.67, + "learning_rate": 6.111135589717321e-07, + "loss": 1.1245, + "step": 31765 + }, + { + "epoch": 2.67, + "learning_rate": 6.108007101244706e-07, + "loss": 1.1092, + "step": 31766 + }, + { + "epoch": 2.67, + "learning_rate": 6.104879388537799e-07, + "loss": 1.1069, + "step": 31767 + }, + { + "epoch": 2.67, + "learning_rate": 6.101752451622389e-07, + "loss": 1.0767, + "step": 31768 + }, + { + "epoch": 2.67, + "learning_rate": 6.09862629052439e-07, + "loss": 1.1138, + "step": 31769 + }, + { + "epoch": 2.67, + "learning_rate": 6.095500905269569e-07, + "loss": 1.1118, + "step": 31770 + }, + { + "epoch": 2.67, + "learning_rate": 6.092376295883773e-07, + "loss": 1.0903, + "step": 31771 + }, + { + "epoch": 2.67, + "learning_rate": 6.089252462392836e-07, + "loss": 1.0472, + "step": 31772 + }, + { + "epoch": 2.68, + "learning_rate": 6.086129404822538e-07, + "loss": 1.1796, + "step": 31773 + }, + { + "epoch": 2.68, + "learning_rate": 6.083007123198703e-07, + "loss": 1.1476, + "step": 31774 + }, + { + "epoch": 2.68, + "learning_rate": 6.07988561754711e-07, + "loss": 0.9817, + "step": 31775 + }, + { + "epoch": 2.68, + "learning_rate": 6.07676488789356e-07, + "loss": 1.0901, + "step": 31776 + }, + { + "epoch": 2.68, + "learning_rate": 6.073644934263845e-07, + "loss": 1.1466, + "step": 31777 + }, + { + "epoch": 2.68, + "learning_rate": 6.070525756683743e-07, + "loss": 1.0443, + "step": 31778 + }, + { + "epoch": 2.68, + "learning_rate": 6.067407355179e-07, + "loss": 1.2135, + "step": 31779 + }, + { + "epoch": 2.68, + "learning_rate": 6.064289729775419e-07, + "loss": 1.0473, + "step": 31780 + }, + { + "epoch": 2.68, + "learning_rate": 6.061172880498733e-07, + "loss": 1.0684, + "step": 31781 + }, + { + "epoch": 2.68, + "learning_rate": 6.058056807374691e-07, + "loss": 1.1129, + "step": 31782 + }, + { + "epoch": 2.68, + "learning_rate": 6.054941510429057e-07, + "loss": 1.1578, + "step": 31783 + }, + { + "epoch": 2.68, + "learning_rate": 6.051826989687559e-07, + "loss": 1.1269, + "step": 31784 + }, + { + "epoch": 2.68, + "learning_rate": 6.04871324517593e-07, + "loss": 1.1966, + "step": 31785 + }, + { + "epoch": 2.68, + "learning_rate": 6.045600276919905e-07, + "loss": 1.0443, + "step": 31786 + }, + { + "epoch": 2.68, + "learning_rate": 6.042488084945208e-07, + "loss": 1.1285, + "step": 31787 + }, + { + "epoch": 2.68, + "learning_rate": 6.039376669277531e-07, + "loss": 1.0771, + "step": 31788 + }, + { + "epoch": 2.68, + "learning_rate": 6.036266029942616e-07, + "loss": 1.145, + "step": 31789 + }, + { + "epoch": 2.68, + "learning_rate": 6.033156166966114e-07, + "loss": 1.1054, + "step": 31790 + }, + { + "epoch": 2.68, + "learning_rate": 6.03004708037378e-07, + "loss": 1.1337, + "step": 31791 + }, + { + "epoch": 2.68, + "learning_rate": 6.026938770191248e-07, + "loss": 1.0699, + "step": 31792 + }, + { + "epoch": 2.68, + "learning_rate": 6.023831236444233e-07, + "loss": 1.1002, + "step": 31793 + }, + { + "epoch": 2.68, + "learning_rate": 6.020724479158402e-07, + "loss": 1.1179, + "step": 31794 + }, + { + "epoch": 2.68, + "learning_rate": 6.017618498359413e-07, + "loss": 1.089, + "step": 31795 + }, + { + "epoch": 2.68, + "learning_rate": 6.014513294072943e-07, + "loss": 1.0949, + "step": 31796 + }, + { + "epoch": 2.68, + "learning_rate": 6.011408866324642e-07, + "loss": 1.1696, + "step": 31797 + }, + { + "epoch": 2.68, + "learning_rate": 6.008305215140164e-07, + "loss": 1.0621, + "step": 31798 + }, + { + "epoch": 2.68, + "learning_rate": 6.005202340545147e-07, + "loss": 1.1817, + "step": 31799 + }, + { + "epoch": 2.68, + "learning_rate": 6.002100242565245e-07, + "loss": 1.0983, + "step": 31800 + }, + { + "epoch": 2.68, + "learning_rate": 5.99899892122604e-07, + "loss": 1.202, + "step": 31801 + }, + { + "epoch": 2.68, + "learning_rate": 5.995898376553233e-07, + "loss": 1.1903, + "step": 31802 + }, + { + "epoch": 2.68, + "learning_rate": 5.99279860857237e-07, + "loss": 1.0592, + "step": 31803 + }, + { + "epoch": 2.68, + "learning_rate": 5.989699617309097e-07, + "loss": 1.0451, + "step": 31804 + }, + { + "epoch": 2.68, + "learning_rate": 5.986601402789016e-07, + "loss": 1.1009, + "step": 31805 + }, + { + "epoch": 2.68, + "learning_rate": 5.983503965037729e-07, + "loss": 1.0032, + "step": 31806 + }, + { + "epoch": 2.68, + "learning_rate": 5.980407304080804e-07, + "loss": 1.1277, + "step": 31807 + }, + { + "epoch": 2.68, + "learning_rate": 5.977311419943854e-07, + "loss": 1.144, + "step": 31808 + }, + { + "epoch": 2.68, + "learning_rate": 5.974216312652459e-07, + "loss": 1.0784, + "step": 31809 + }, + { + "epoch": 2.68, + "learning_rate": 5.971121982232153e-07, + "loss": 1.1968, + "step": 31810 + }, + { + "epoch": 2.68, + "learning_rate": 5.968028428708562e-07, + "loss": 1.134, + "step": 31811 + }, + { + "epoch": 2.68, + "learning_rate": 5.964935652107195e-07, + "loss": 1.145, + "step": 31812 + }, + { + "epoch": 2.68, + "learning_rate": 5.961843652453625e-07, + "loss": 1.0098, + "step": 31813 + }, + { + "epoch": 2.68, + "learning_rate": 5.958752429773396e-07, + "loss": 1.1228, + "step": 31814 + }, + { + "epoch": 2.68, + "learning_rate": 5.955661984092064e-07, + "loss": 1.0861, + "step": 31815 + }, + { + "epoch": 2.68, + "learning_rate": 5.952572315435134e-07, + "loss": 1.129, + "step": 31816 + }, + { + "epoch": 2.68, + "learning_rate": 5.949483423828184e-07, + "loss": 1.0415, + "step": 31817 + }, + { + "epoch": 2.68, + "learning_rate": 5.94639530929666e-07, + "loss": 1.0746, + "step": 31818 + }, + { + "epoch": 2.68, + "learning_rate": 5.943307971866152e-07, + "loss": 1.0994, + "step": 31819 + }, + { + "epoch": 2.68, + "learning_rate": 5.940221411562119e-07, + "loss": 1.0138, + "step": 31820 + }, + { + "epoch": 2.68, + "learning_rate": 5.937135628410062e-07, + "loss": 1.1665, + "step": 31821 + }, + { + "epoch": 2.68, + "learning_rate": 5.934050622435527e-07, + "loss": 1.1286, + "step": 31822 + }, + { + "epoch": 2.68, + "learning_rate": 5.930966393663961e-07, + "loss": 1.0554, + "step": 31823 + }, + { + "epoch": 2.68, + "learning_rate": 5.927882942120855e-07, + "loss": 1.0586, + "step": 31824 + }, + { + "epoch": 2.68, + "learning_rate": 5.924800267831676e-07, + "loss": 1.1399, + "step": 31825 + }, + { + "epoch": 2.68, + "learning_rate": 5.921718370821917e-07, + "loss": 1.1633, + "step": 31826 + }, + { + "epoch": 2.68, + "learning_rate": 5.918637251117032e-07, + "loss": 1.1083, + "step": 31827 + }, + { + "epoch": 2.68, + "learning_rate": 5.915556908742481e-07, + "loss": 1.1193, + "step": 31828 + }, + { + "epoch": 2.68, + "learning_rate": 5.912477343723688e-07, + "loss": 1.1509, + "step": 31829 + }, + { + "epoch": 2.68, + "learning_rate": 5.909398556086143e-07, + "loss": 1.1046, + "step": 31830 + }, + { + "epoch": 2.68, + "learning_rate": 5.906320545855249e-07, + "loss": 1.1263, + "step": 31831 + }, + { + "epoch": 2.68, + "learning_rate": 5.903243313056439e-07, + "loss": 1.1955, + "step": 31832 + }, + { + "epoch": 2.68, + "learning_rate": 5.900166857715162e-07, + "loss": 1.0432, + "step": 31833 + }, + { + "epoch": 2.68, + "learning_rate": 5.897091179856807e-07, + "loss": 1.1831, + "step": 31834 + }, + { + "epoch": 2.68, + "learning_rate": 5.894016279506809e-07, + "loss": 1.1167, + "step": 31835 + }, + { + "epoch": 2.68, + "learning_rate": 5.890942156690571e-07, + "loss": 1.063, + "step": 31836 + }, + { + "epoch": 2.68, + "learning_rate": 5.887868811433484e-07, + "loss": 1.1624, + "step": 31837 + }, + { + "epoch": 2.68, + "learning_rate": 5.884796243760949e-07, + "loss": 1.1715, + "step": 31838 + }, + { + "epoch": 2.68, + "learning_rate": 5.881724453698357e-07, + "loss": 1.1467, + "step": 31839 + }, + { + "epoch": 2.68, + "learning_rate": 5.878653441271043e-07, + "loss": 1.1206, + "step": 31840 + }, + { + "epoch": 2.68, + "learning_rate": 5.875583206504454e-07, + "loss": 1.1676, + "step": 31841 + }, + { + "epoch": 2.68, + "learning_rate": 5.872513749423914e-07, + "loss": 1.0863, + "step": 31842 + }, + { + "epoch": 2.68, + "learning_rate": 5.86944507005478e-07, + "loss": 1.1522, + "step": 31843 + }, + { + "epoch": 2.68, + "learning_rate": 5.866377168422422e-07, + "loss": 1.0644, + "step": 31844 + }, + { + "epoch": 2.68, + "learning_rate": 5.863310044552173e-07, + "loss": 1.1412, + "step": 31845 + }, + { + "epoch": 2.68, + "learning_rate": 5.860243698469392e-07, + "loss": 1.1206, + "step": 31846 + }, + { + "epoch": 2.68, + "learning_rate": 5.857178130199403e-07, + "loss": 1.0695, + "step": 31847 + }, + { + "epoch": 2.68, + "learning_rate": 5.854113339767542e-07, + "loss": 1.1427, + "step": 31848 + }, + { + "epoch": 2.68, + "learning_rate": 5.851049327199121e-07, + "loss": 1.1274, + "step": 31849 + }, + { + "epoch": 2.68, + "learning_rate": 5.847986092519475e-07, + "loss": 1.1053, + "step": 31850 + }, + { + "epoch": 2.68, + "learning_rate": 5.844923635753874e-07, + "loss": 1.1515, + "step": 31851 + }, + { + "epoch": 2.68, + "learning_rate": 5.841861956927686e-07, + "loss": 1.1519, + "step": 31852 + }, + { + "epoch": 2.68, + "learning_rate": 5.838801056066146e-07, + "loss": 1.0632, + "step": 31853 + }, + { + "epoch": 2.68, + "learning_rate": 5.835740933194567e-07, + "loss": 1.1059, + "step": 31854 + }, + { + "epoch": 2.68, + "learning_rate": 5.832681588338229e-07, + "loss": 1.1139, + "step": 31855 + }, + { + "epoch": 2.68, + "learning_rate": 5.829623021522435e-07, + "loss": 1.1581, + "step": 31856 + }, + { + "epoch": 2.68, + "learning_rate": 5.826565232772408e-07, + "loss": 1.0317, + "step": 31857 + }, + { + "epoch": 2.68, + "learning_rate": 5.823508222113439e-07, + "loss": 1.1592, + "step": 31858 + }, + { + "epoch": 2.68, + "learning_rate": 5.820451989570808e-07, + "loss": 1.0611, + "step": 31859 + }, + { + "epoch": 2.68, + "learning_rate": 5.817396535169717e-07, + "loss": 1.1508, + "step": 31860 + }, + { + "epoch": 2.68, + "learning_rate": 5.814341858935468e-07, + "loss": 1.233, + "step": 31861 + }, + { + "epoch": 2.68, + "learning_rate": 5.811287960893252e-07, + "loss": 1.1478, + "step": 31862 + }, + { + "epoch": 2.68, + "learning_rate": 5.808234841068317e-07, + "loss": 1.1128, + "step": 31863 + }, + { + "epoch": 2.68, + "learning_rate": 5.805182499485896e-07, + "loss": 1.1316, + "step": 31864 + }, + { + "epoch": 2.68, + "learning_rate": 5.802130936171202e-07, + "loss": 1.0625, + "step": 31865 + }, + { + "epoch": 2.68, + "learning_rate": 5.79908015114945e-07, + "loss": 1.1338, + "step": 31866 + }, + { + "epoch": 2.68, + "learning_rate": 5.796030144445852e-07, + "loss": 1.0673, + "step": 31867 + }, + { + "epoch": 2.68, + "learning_rate": 5.792980916085588e-07, + "loss": 1.1812, + "step": 31868 + }, + { + "epoch": 2.68, + "learning_rate": 5.789932466093884e-07, + "loss": 1.1409, + "step": 31869 + }, + { + "epoch": 2.68, + "learning_rate": 5.786884794495895e-07, + "loss": 1.0873, + "step": 31870 + }, + { + "epoch": 2.68, + "learning_rate": 5.783837901316802e-07, + "loss": 1.1402, + "step": 31871 + }, + { + "epoch": 2.68, + "learning_rate": 5.78079178658183e-07, + "loss": 1.1105, + "step": 31872 + }, + { + "epoch": 2.68, + "learning_rate": 5.777746450316091e-07, + "loss": 1.2264, + "step": 31873 + }, + { + "epoch": 2.68, + "learning_rate": 5.774701892544754e-07, + "loss": 1.1373, + "step": 31874 + }, + { + "epoch": 2.68, + "learning_rate": 5.771658113293e-07, + "loss": 1.1344, + "step": 31875 + }, + { + "epoch": 2.68, + "learning_rate": 5.768615112585963e-07, + "loss": 1.1734, + "step": 31876 + }, + { + "epoch": 2.68, + "learning_rate": 5.765572890448778e-07, + "loss": 1.0942, + "step": 31877 + }, + { + "epoch": 2.68, + "learning_rate": 5.762531446906616e-07, + "loss": 1.1038, + "step": 31878 + }, + { + "epoch": 2.68, + "learning_rate": 5.759490781984534e-07, + "loss": 1.1608, + "step": 31879 + }, + { + "epoch": 2.68, + "learning_rate": 5.756450895707743e-07, + "loss": 1.0791, + "step": 31880 + }, + { + "epoch": 2.68, + "learning_rate": 5.753411788101293e-07, + "loss": 1.0332, + "step": 31881 + }, + { + "epoch": 2.68, + "learning_rate": 5.750373459190318e-07, + "loss": 1.1621, + "step": 31882 + }, + { + "epoch": 2.68, + "learning_rate": 5.747335908999929e-07, + "loss": 1.1611, + "step": 31883 + }, + { + "epoch": 2.68, + "learning_rate": 5.744299137555209e-07, + "loss": 1.0847, + "step": 31884 + }, + { + "epoch": 2.68, + "learning_rate": 5.741263144881259e-07, + "loss": 1.1703, + "step": 31885 + }, + { + "epoch": 2.68, + "learning_rate": 5.738227931003159e-07, + "loss": 1.1672, + "step": 31886 + }, + { + "epoch": 2.68, + "learning_rate": 5.735193495945978e-07, + "loss": 1.1292, + "step": 31887 + }, + { + "epoch": 2.68, + "learning_rate": 5.732159839734808e-07, + "loss": 1.1115, + "step": 31888 + }, + { + "epoch": 2.68, + "learning_rate": 5.729126962394715e-07, + "loss": 1.2089, + "step": 31889 + }, + { + "epoch": 2.68, + "learning_rate": 5.726094863950716e-07, + "loss": 1.2096, + "step": 31890 + }, + { + "epoch": 2.68, + "learning_rate": 5.723063544427921e-07, + "loss": 1.1257, + "step": 31891 + }, + { + "epoch": 2.69, + "learning_rate": 5.720033003851333e-07, + "loss": 1.0665, + "step": 31892 + }, + { + "epoch": 2.69, + "learning_rate": 5.717003242246e-07, + "loss": 1.1131, + "step": 31893 + }, + { + "epoch": 2.69, + "learning_rate": 5.713974259636967e-07, + "loss": 1.0883, + "step": 31894 + }, + { + "epoch": 2.69, + "learning_rate": 5.71094605604926e-07, + "loss": 1.1262, + "step": 31895 + }, + { + "epoch": 2.69, + "learning_rate": 5.707918631507881e-07, + "loss": 1.0462, + "step": 31896 + }, + { + "epoch": 2.69, + "learning_rate": 5.704891986037863e-07, + "loss": 1.0759, + "step": 31897 + }, + { + "epoch": 2.69, + "learning_rate": 5.701866119664201e-07, + "loss": 1.1097, + "step": 31898 + }, + { + "epoch": 2.69, + "learning_rate": 5.698841032411906e-07, + "loss": 1.1132, + "step": 31899 + }, + { + "epoch": 2.69, + "learning_rate": 5.69581672430598e-07, + "loss": 1.1279, + "step": 31900 + }, + { + "epoch": 2.69, + "learning_rate": 5.692793195371371e-07, + "loss": 1.0973, + "step": 31901 + }, + { + "epoch": 2.69, + "learning_rate": 5.689770445633113e-07, + "loss": 1.2184, + "step": 31902 + }, + { + "epoch": 2.69, + "learning_rate": 5.686748475116133e-07, + "loss": 1.1329, + "step": 31903 + }, + { + "epoch": 2.69, + "learning_rate": 5.68372728384543e-07, + "loss": 1.1015, + "step": 31904 + }, + { + "epoch": 2.69, + "learning_rate": 5.680706871845954e-07, + "loss": 1.0266, + "step": 31905 + }, + { + "epoch": 2.69, + "learning_rate": 5.677687239142682e-07, + "loss": 1.1862, + "step": 31906 + }, + { + "epoch": 2.69, + "learning_rate": 5.674668385760506e-07, + "loss": 1.0567, + "step": 31907 + }, + { + "epoch": 2.69, + "learning_rate": 5.67165031172443e-07, + "loss": 1.1283, + "step": 31908 + }, + { + "epoch": 2.69, + "learning_rate": 5.668633017059388e-07, + "loss": 1.1015, + "step": 31909 + }, + { + "epoch": 2.69, + "learning_rate": 5.66561650179025e-07, + "loss": 1.0699, + "step": 31910 + }, + { + "epoch": 2.69, + "learning_rate": 5.662600765942017e-07, + "loss": 1.1184, + "step": 31911 + }, + { + "epoch": 2.69, + "learning_rate": 5.659585809539547e-07, + "loss": 1.0105, + "step": 31912 + }, + { + "epoch": 2.69, + "learning_rate": 5.656571632607777e-07, + "loss": 1.0817, + "step": 31913 + }, + { + "epoch": 2.69, + "learning_rate": 5.653558235171596e-07, + "loss": 1.1449, + "step": 31914 + }, + { + "epoch": 2.69, + "learning_rate": 5.650545617255921e-07, + "loss": 1.0811, + "step": 31915 + }, + { + "epoch": 2.69, + "learning_rate": 5.647533778885639e-07, + "loss": 1.1081, + "step": 31916 + }, + { + "epoch": 2.69, + "learning_rate": 5.644522720085632e-07, + "loss": 1.1408, + "step": 31917 + }, + { + "epoch": 2.69, + "learning_rate": 5.641512440880759e-07, + "loss": 1.0857, + "step": 31918 + }, + { + "epoch": 2.69, + "learning_rate": 5.638502941295931e-07, + "loss": 1.0839, + "step": 31919 + }, + { + "epoch": 2.69, + "learning_rate": 5.635494221355975e-07, + "loss": 1.0606, + "step": 31920 + }, + { + "epoch": 2.69, + "learning_rate": 5.632486281085748e-07, + "loss": 1.1432, + "step": 31921 + }, + { + "epoch": 2.69, + "learning_rate": 5.629479120510162e-07, + "loss": 1.19, + "step": 31922 + }, + { + "epoch": 2.69, + "learning_rate": 5.626472739653999e-07, + "loss": 1.2064, + "step": 31923 + }, + { + "epoch": 2.69, + "learning_rate": 5.623467138542116e-07, + "loss": 1.0877, + "step": 31924 + }, + { + "epoch": 2.69, + "learning_rate": 5.62046231719936e-07, + "loss": 1.084, + "step": 31925 + }, + { + "epoch": 2.69, + "learning_rate": 5.617458275650555e-07, + "loss": 1.1282, + "step": 31926 + }, + { + "epoch": 2.69, + "learning_rate": 5.614455013920506e-07, + "loss": 1.0176, + "step": 31927 + }, + { + "epoch": 2.69, + "learning_rate": 5.611452532034056e-07, + "loss": 1.0805, + "step": 31928 + }, + { + "epoch": 2.69, + "learning_rate": 5.608450830015966e-07, + "loss": 1.1035, + "step": 31929 + }, + { + "epoch": 2.69, + "learning_rate": 5.605449907891091e-07, + "loss": 1.0419, + "step": 31930 + }, + { + "epoch": 2.69, + "learning_rate": 5.602449765684181e-07, + "loss": 1.1555, + "step": 31931 + }, + { + "epoch": 2.69, + "learning_rate": 5.599450403420048e-07, + "loss": 1.0189, + "step": 31932 + }, + { + "epoch": 2.69, + "learning_rate": 5.596451821123461e-07, + "loss": 1.0137, + "step": 31933 + }, + { + "epoch": 2.69, + "learning_rate": 5.593454018819211e-07, + "loss": 1.1663, + "step": 31934 + }, + { + "epoch": 2.69, + "learning_rate": 5.590456996532045e-07, + "loss": 1.1387, + "step": 31935 + }, + { + "epoch": 2.69, + "learning_rate": 5.587460754286755e-07, + "loss": 1.1551, + "step": 31936 + }, + { + "epoch": 2.69, + "learning_rate": 5.584465292108065e-07, + "loss": 1.1769, + "step": 31937 + }, + { + "epoch": 2.69, + "learning_rate": 5.581470610020745e-07, + "loss": 1.0482, + "step": 31938 + }, + { + "epoch": 2.69, + "learning_rate": 5.578476708049552e-07, + "loss": 1.0278, + "step": 31939 + }, + { + "epoch": 2.69, + "learning_rate": 5.575483586219177e-07, + "loss": 1.0515, + "step": 31940 + }, + { + "epoch": 2.69, + "learning_rate": 5.57249124455439e-07, + "loss": 1.1722, + "step": 31941 + }, + { + "epoch": 2.69, + "learning_rate": 5.569499683079904e-07, + "loss": 1.0259, + "step": 31942 + }, + { + "epoch": 2.69, + "learning_rate": 5.566508901820423e-07, + "loss": 1.1154, + "step": 31943 + }, + { + "epoch": 2.69, + "learning_rate": 5.563518900800668e-07, + "loss": 1.1025, + "step": 31944 + }, + { + "epoch": 2.69, + "learning_rate": 5.560529680045346e-07, + "loss": 1.0797, + "step": 31945 + }, + { + "epoch": 2.69, + "learning_rate": 5.557541239579145e-07, + "loss": 1.1349, + "step": 31946 + }, + { + "epoch": 2.69, + "learning_rate": 5.554553579426769e-07, + "loss": 1.119, + "step": 31947 + }, + { + "epoch": 2.69, + "learning_rate": 5.551566699612898e-07, + "loss": 1.147, + "step": 31948 + }, + { + "epoch": 2.69, + "learning_rate": 5.548580600162212e-07, + "loss": 1.2096, + "step": 31949 + }, + { + "epoch": 2.69, + "learning_rate": 5.545595281099403e-07, + "loss": 1.0144, + "step": 31950 + }, + { + "epoch": 2.69, + "learning_rate": 5.542610742449073e-07, + "loss": 1.1411, + "step": 31951 + }, + { + "epoch": 2.69, + "learning_rate": 5.539626984235957e-07, + "loss": 1.1248, + "step": 31952 + }, + { + "epoch": 2.69, + "learning_rate": 5.536644006484659e-07, + "loss": 1.0372, + "step": 31953 + }, + { + "epoch": 2.69, + "learning_rate": 5.533661809219848e-07, + "loss": 1.1448, + "step": 31954 + }, + { + "epoch": 2.69, + "learning_rate": 5.530680392466147e-07, + "loss": 1.1222, + "step": 31955 + }, + { + "epoch": 2.69, + "learning_rate": 5.527699756248217e-07, + "loss": 1.1703, + "step": 31956 + }, + { + "epoch": 2.69, + "learning_rate": 5.524719900590636e-07, + "loss": 1.1391, + "step": 31957 + }, + { + "epoch": 2.69, + "learning_rate": 5.521740825518063e-07, + "loss": 1.2112, + "step": 31958 + }, + { + "epoch": 2.69, + "learning_rate": 5.518762531055133e-07, + "loss": 1.1217, + "step": 31959 + }, + { + "epoch": 2.69, + "learning_rate": 5.515785017226383e-07, + "loss": 1.1388, + "step": 31960 + }, + { + "epoch": 2.69, + "learning_rate": 5.512808284056492e-07, + "loss": 1.1202, + "step": 31961 + }, + { + "epoch": 2.69, + "learning_rate": 5.509832331570009e-07, + "loss": 1.1679, + "step": 31962 + }, + { + "epoch": 2.69, + "learning_rate": 5.506857159791523e-07, + "loss": 1.0816, + "step": 31963 + }, + { + "epoch": 2.69, + "learning_rate": 5.503882768745628e-07, + "loss": 1.1362, + "step": 31964 + }, + { + "epoch": 2.69, + "learning_rate": 5.50090915845689e-07, + "loss": 1.0854, + "step": 31965 + }, + { + "epoch": 2.69, + "learning_rate": 5.497936328949893e-07, + "loss": 1.1252, + "step": 31966 + }, + { + "epoch": 2.69, + "learning_rate": 5.494964280249204e-07, + "loss": 1.2015, + "step": 31967 + }, + { + "epoch": 2.69, + "learning_rate": 5.491993012379338e-07, + "loss": 1.1432, + "step": 31968 + }, + { + "epoch": 2.69, + "learning_rate": 5.489022525364896e-07, + "loss": 1.1069, + "step": 31969 + }, + { + "epoch": 2.69, + "learning_rate": 5.486052819230381e-07, + "loss": 1.0922, + "step": 31970 + }, + { + "epoch": 2.69, + "learning_rate": 5.48308389400033e-07, + "loss": 1.1012, + "step": 31971 + }, + { + "epoch": 2.69, + "learning_rate": 5.480115749699333e-07, + "loss": 1.0875, + "step": 31972 + }, + { + "epoch": 2.69, + "learning_rate": 5.477148386351839e-07, + "loss": 1.157, + "step": 31973 + }, + { + "epoch": 2.69, + "learning_rate": 5.474181803982403e-07, + "loss": 1.1895, + "step": 31974 + }, + { + "epoch": 2.69, + "learning_rate": 5.471216002615531e-07, + "loss": 1.0694, + "step": 31975 + }, + { + "epoch": 2.69, + "learning_rate": 5.468250982275724e-07, + "loss": 1.1089, + "step": 31976 + }, + { + "epoch": 2.69, + "learning_rate": 5.465286742987475e-07, + "loss": 1.086, + "step": 31977 + }, + { + "epoch": 2.69, + "learning_rate": 5.462323284775306e-07, + "loss": 1.0648, + "step": 31978 + }, + { + "epoch": 2.69, + "learning_rate": 5.459360607663644e-07, + "loss": 1.0738, + "step": 31979 + }, + { + "epoch": 2.69, + "learning_rate": 5.456398711677035e-07, + "loss": 1.1089, + "step": 31980 + }, + { + "epoch": 2.69, + "learning_rate": 5.453437596839895e-07, + "loss": 1.1511, + "step": 31981 + }, + { + "epoch": 2.69, + "learning_rate": 5.450477263176724e-07, + "loss": 1.1934, + "step": 31982 + }, + { + "epoch": 2.69, + "learning_rate": 5.447517710711958e-07, + "loss": 1.0974, + "step": 31983 + }, + { + "epoch": 2.69, + "learning_rate": 5.444558939470068e-07, + "loss": 1.1615, + "step": 31984 + }, + { + "epoch": 2.69, + "learning_rate": 5.44160094947549e-07, + "loss": 1.0171, + "step": 31985 + }, + { + "epoch": 2.69, + "learning_rate": 5.43864374075267e-07, + "loss": 1.1, + "step": 31986 + }, + { + "epoch": 2.69, + "learning_rate": 5.435687313326032e-07, + "loss": 1.1425, + "step": 31987 + }, + { + "epoch": 2.69, + "learning_rate": 5.432731667220025e-07, + "loss": 1.0304, + "step": 31988 + }, + { + "epoch": 2.69, + "learning_rate": 5.429776802459053e-07, + "loss": 1.0895, + "step": 31989 + }, + { + "epoch": 2.69, + "learning_rate": 5.426822719067515e-07, + "loss": 1.0953, + "step": 31990 + }, + { + "epoch": 2.69, + "learning_rate": 5.42386941706986e-07, + "loss": 1.0786, + "step": 31991 + }, + { + "epoch": 2.69, + "learning_rate": 5.420916896490447e-07, + "loss": 1.1225, + "step": 31992 + }, + { + "epoch": 2.69, + "learning_rate": 5.417965157353688e-07, + "loss": 1.0982, + "step": 31993 + }, + { + "epoch": 2.69, + "learning_rate": 5.415014199683966e-07, + "loss": 1.1238, + "step": 31994 + }, + { + "epoch": 2.69, + "learning_rate": 5.412064023505681e-07, + "loss": 0.9952, + "step": 31995 + }, + { + "epoch": 2.69, + "learning_rate": 5.409114628843181e-07, + "loss": 1.1683, + "step": 31996 + }, + { + "epoch": 2.69, + "learning_rate": 5.40616601572086e-07, + "loss": 1.1302, + "step": 31997 + }, + { + "epoch": 2.69, + "learning_rate": 5.403218184163062e-07, + "loss": 1.1006, + "step": 31998 + }, + { + "epoch": 2.69, + "learning_rate": 5.400271134194157e-07, + "loss": 1.1012, + "step": 31999 + }, + { + "epoch": 2.69, + "learning_rate": 5.397324865838493e-07, + "loss": 1.083, + "step": 32000 + }, + { + "epoch": 2.69, + "learning_rate": 5.394379379120384e-07, + "loss": 1.0677, + "step": 32001 + }, + { + "epoch": 2.69, + "learning_rate": 5.391434674064222e-07, + "loss": 1.0758, + "step": 32002 + }, + { + "epoch": 2.69, + "learning_rate": 5.388490750694286e-07, + "loss": 1.0473, + "step": 32003 + }, + { + "epoch": 2.69, + "learning_rate": 5.385547609034925e-07, + "loss": 1.0313, + "step": 32004 + }, + { + "epoch": 2.69, + "learning_rate": 5.38260524911044e-07, + "loss": 0.9919, + "step": 32005 + }, + { + "epoch": 2.69, + "learning_rate": 5.379663670945179e-07, + "loss": 1.1238, + "step": 32006 + }, + { + "epoch": 2.69, + "learning_rate": 5.376722874563378e-07, + "loss": 1.1546, + "step": 32007 + }, + { + "epoch": 2.69, + "learning_rate": 5.373782859989396e-07, + "loss": 1.2258, + "step": 32008 + }, + { + "epoch": 2.69, + "learning_rate": 5.370843627247523e-07, + "loss": 1.1498, + "step": 32009 + }, + { + "epoch": 2.69, + "learning_rate": 5.367905176361987e-07, + "loss": 1.0812, + "step": 32010 + }, + { + "epoch": 2.7, + "learning_rate": 5.364967507357144e-07, + "loss": 1.0856, + "step": 32011 + }, + { + "epoch": 2.7, + "learning_rate": 5.362030620257208e-07, + "loss": 1.1576, + "step": 32012 + }, + { + "epoch": 2.7, + "learning_rate": 5.35909451508646e-07, + "loss": 1.0804, + "step": 32013 + }, + { + "epoch": 2.7, + "learning_rate": 5.356159191869159e-07, + "loss": 1.1492, + "step": 32014 + }, + { + "epoch": 2.7, + "learning_rate": 5.353224650629563e-07, + "loss": 1.1228, + "step": 32015 + }, + { + "epoch": 2.7, + "learning_rate": 5.350290891391918e-07, + "loss": 1.0827, + "step": 32016 + }, + { + "epoch": 2.7, + "learning_rate": 5.347357914180484e-07, + "loss": 1.0608, + "step": 32017 + }, + { + "epoch": 2.7, + "learning_rate": 5.34442571901943e-07, + "loss": 1.1665, + "step": 32018 + }, + { + "epoch": 2.7, + "learning_rate": 5.34149430593306e-07, + "loss": 1.0039, + "step": 32019 + }, + { + "epoch": 2.7, + "learning_rate": 5.33856367494554e-07, + "loss": 1.0744, + "step": 32020 + }, + { + "epoch": 2.7, + "learning_rate": 5.335633826081088e-07, + "loss": 1.0959, + "step": 32021 + }, + { + "epoch": 2.7, + "learning_rate": 5.332704759363961e-07, + "loss": 1.1714, + "step": 32022 + }, + { + "epoch": 2.7, + "learning_rate": 5.329776474818305e-07, + "loss": 1.2059, + "step": 32023 + }, + { + "epoch": 2.7, + "learning_rate": 5.326848972468335e-07, + "loss": 1.1274, + "step": 32024 + }, + { + "epoch": 2.7, + "learning_rate": 5.323922252338242e-07, + "loss": 1.1681, + "step": 32025 + }, + { + "epoch": 2.7, + "learning_rate": 5.320996314452198e-07, + "loss": 1.0909, + "step": 32026 + }, + { + "epoch": 2.7, + "learning_rate": 5.318071158834381e-07, + "loss": 1.0525, + "step": 32027 + }, + { + "epoch": 2.7, + "learning_rate": 5.315146785508985e-07, + "loss": 1.1616, + "step": 32028 + }, + { + "epoch": 2.7, + "learning_rate": 5.312223194500122e-07, + "loss": 1.0677, + "step": 32029 + }, + { + "epoch": 2.7, + "learning_rate": 5.309300385831994e-07, + "loss": 1.142, + "step": 32030 + }, + { + "epoch": 2.7, + "learning_rate": 5.306378359528719e-07, + "loss": 1.1321, + "step": 32031 + }, + { + "epoch": 2.7, + "learning_rate": 5.303457115614452e-07, + "loss": 1.0797, + "step": 32032 + }, + { + "epoch": 2.7, + "learning_rate": 5.30053665411333e-07, + "loss": 1.1086, + "step": 32033 + }, + { + "epoch": 2.7, + "learning_rate": 5.297616975049491e-07, + "loss": 1.0768, + "step": 32034 + }, + { + "epoch": 2.7, + "learning_rate": 5.294698078447036e-07, + "loss": 1.1476, + "step": 32035 + }, + { + "epoch": 2.7, + "learning_rate": 5.291779964330102e-07, + "loss": 1.1029, + "step": 32036 + }, + { + "epoch": 2.7, + "learning_rate": 5.288862632722791e-07, + "loss": 1.059, + "step": 32037 + }, + { + "epoch": 2.7, + "learning_rate": 5.285946083649207e-07, + "loss": 1.0712, + "step": 32038 + }, + { + "epoch": 2.7, + "learning_rate": 5.283030317133464e-07, + "loss": 1.1413, + "step": 32039 + }, + { + "epoch": 2.7, + "learning_rate": 5.28011533319961e-07, + "loss": 1.0572, + "step": 32040 + }, + { + "epoch": 2.7, + "learning_rate": 5.277201131871778e-07, + "loss": 1.1666, + "step": 32041 + }, + { + "epoch": 2.7, + "learning_rate": 5.27428771317402e-07, + "loss": 1.0304, + "step": 32042 + }, + { + "epoch": 2.7, + "learning_rate": 5.271375077130414e-07, + "loss": 1.1105, + "step": 32043 + }, + { + "epoch": 2.7, + "learning_rate": 5.268463223765019e-07, + "loss": 1.0933, + "step": 32044 + }, + { + "epoch": 2.7, + "learning_rate": 5.265552153101905e-07, + "loss": 1.1616, + "step": 32045 + }, + { + "epoch": 2.7, + "learning_rate": 5.26264186516512e-07, + "loss": 1.0671, + "step": 32046 + }, + { + "epoch": 2.7, + "learning_rate": 5.259732359978697e-07, + "loss": 1.0788, + "step": 32047 + }, + { + "epoch": 2.7, + "learning_rate": 5.256823637566688e-07, + "loss": 1.1162, + "step": 32048 + }, + { + "epoch": 2.7, + "learning_rate": 5.253915697953127e-07, + "loss": 1.1188, + "step": 32049 + }, + { + "epoch": 2.7, + "learning_rate": 5.251008541162051e-07, + "loss": 1.0414, + "step": 32050 + }, + { + "epoch": 2.7, + "learning_rate": 5.248102167217439e-07, + "loss": 1.2597, + "step": 32051 + }, + { + "epoch": 2.7, + "learning_rate": 5.245196576143363e-07, + "loss": 1.052, + "step": 32052 + }, + { + "epoch": 2.7, + "learning_rate": 5.242291767963781e-07, + "loss": 1.1256, + "step": 32053 + }, + { + "epoch": 2.7, + "learning_rate": 5.239387742702707e-07, + "loss": 1.1644, + "step": 32054 + }, + { + "epoch": 2.7, + "learning_rate": 5.236484500384143e-07, + "loss": 1.2675, + "step": 32055 + }, + { + "epoch": 2.7, + "learning_rate": 5.23358204103207e-07, + "loss": 1.1452, + "step": 32056 + }, + { + "epoch": 2.7, + "learning_rate": 5.230680364670471e-07, + "loss": 1.1319, + "step": 32057 + }, + { + "epoch": 2.7, + "learning_rate": 5.227779471323324e-07, + "loss": 1.0116, + "step": 32058 + }, + { + "epoch": 2.7, + "learning_rate": 5.22487936101459e-07, + "loss": 1.1105, + "step": 32059 + }, + { + "epoch": 2.7, + "learning_rate": 5.221980033768237e-07, + "loss": 1.1762, + "step": 32060 + }, + { + "epoch": 2.7, + "learning_rate": 5.219081489608235e-07, + "loss": 1.0882, + "step": 32061 + }, + { + "epoch": 2.7, + "learning_rate": 5.216183728558478e-07, + "loss": 1.069, + "step": 32062 + }, + { + "epoch": 2.7, + "learning_rate": 5.213286750642987e-07, + "loss": 1.1724, + "step": 32063 + }, + { + "epoch": 2.7, + "learning_rate": 5.210390555885636e-07, + "loss": 1.1091, + "step": 32064 + }, + { + "epoch": 2.7, + "learning_rate": 5.20749514431037e-07, + "loss": 1.1075, + "step": 32065 + }, + { + "epoch": 2.7, + "learning_rate": 5.204600515941116e-07, + "loss": 1.0929, + "step": 32066 + }, + { + "epoch": 2.7, + "learning_rate": 5.201706670801809e-07, + "loss": 1.0681, + "step": 32067 + }, + { + "epoch": 2.7, + "learning_rate": 5.19881360891632e-07, + "loss": 1.1232, + "step": 32068 + }, + { + "epoch": 2.7, + "learning_rate": 5.195921330308595e-07, + "loss": 1.1148, + "step": 32069 + }, + { + "epoch": 2.7, + "learning_rate": 5.193029835002495e-07, + "loss": 1.1061, + "step": 32070 + }, + { + "epoch": 2.7, + "learning_rate": 5.190139123021909e-07, + "loss": 1.1598, + "step": 32071 + }, + { + "epoch": 2.7, + "learning_rate": 5.187249194390764e-07, + "loss": 1.1064, + "step": 32072 + }, + { + "epoch": 2.7, + "learning_rate": 5.184360049132897e-07, + "loss": 1.1784, + "step": 32073 + }, + { + "epoch": 2.7, + "learning_rate": 5.181471687272188e-07, + "loss": 1.0805, + "step": 32074 + }, + { + "epoch": 2.7, + "learning_rate": 5.178584108832518e-07, + "loss": 1.161, + "step": 32075 + }, + { + "epoch": 2.7, + "learning_rate": 5.175697313837724e-07, + "loss": 1.0783, + "step": 32076 + }, + { + "epoch": 2.7, + "learning_rate": 5.172811302311665e-07, + "loss": 1.1266, + "step": 32077 + }, + { + "epoch": 2.7, + "learning_rate": 5.169926074278208e-07, + "loss": 1.0628, + "step": 32078 + }, + { + "epoch": 2.7, + "learning_rate": 5.167041629761138e-07, + "loss": 1.0915, + "step": 32079 + }, + { + "epoch": 2.7, + "learning_rate": 5.164157968784356e-07, + "loss": 1.136, + "step": 32080 + }, + { + "epoch": 2.7, + "learning_rate": 5.161275091371632e-07, + "loss": 1.2146, + "step": 32081 + }, + { + "epoch": 2.7, + "learning_rate": 5.158392997546801e-07, + "loss": 1.1164, + "step": 32082 + }, + { + "epoch": 2.7, + "learning_rate": 5.155511687333692e-07, + "loss": 1.028, + "step": 32083 + }, + { + "epoch": 2.7, + "learning_rate": 5.152631160756094e-07, + "loss": 1.1648, + "step": 32084 + }, + { + "epoch": 2.7, + "learning_rate": 5.149751417837812e-07, + "loss": 1.1105, + "step": 32085 + }, + { + "epoch": 2.7, + "learning_rate": 5.146872458602636e-07, + "loss": 1.1928, + "step": 32086 + }, + { + "epoch": 2.7, + "learning_rate": 5.14399428307436e-07, + "loss": 1.0962, + "step": 32087 + }, + { + "epoch": 2.7, + "learning_rate": 5.141116891276754e-07, + "loss": 1.1122, + "step": 32088 + }, + { + "epoch": 2.7, + "learning_rate": 5.138240283233631e-07, + "loss": 1.1473, + "step": 32089 + }, + { + "epoch": 2.7, + "learning_rate": 5.135364458968683e-07, + "loss": 1.0905, + "step": 32090 + }, + { + "epoch": 2.7, + "learning_rate": 5.132489418505748e-07, + "loss": 1.1039, + "step": 32091 + }, + { + "epoch": 2.7, + "learning_rate": 5.129615161868529e-07, + "loss": 1.2307, + "step": 32092 + }, + { + "epoch": 2.7, + "learning_rate": 5.126741689080794e-07, + "loss": 1.0468, + "step": 32093 + }, + { + "epoch": 2.7, + "learning_rate": 5.123869000166282e-07, + "loss": 1.1357, + "step": 32094 + }, + { + "epoch": 2.7, + "learning_rate": 5.120997095148727e-07, + "loss": 1.0595, + "step": 32095 + }, + { + "epoch": 2.7, + "learning_rate": 5.118125974051869e-07, + "loss": 1.0573, + "step": 32096 + }, + { + "epoch": 2.7, + "learning_rate": 5.115255636899408e-07, + "loss": 1.1114, + "step": 32097 + }, + { + "epoch": 2.7, + "learning_rate": 5.112386083715082e-07, + "loss": 1.0321, + "step": 32098 + }, + { + "epoch": 2.7, + "learning_rate": 5.109517314522583e-07, + "loss": 1.1092, + "step": 32099 + }, + { + "epoch": 2.7, + "learning_rate": 5.106649329345636e-07, + "loss": 1.1615, + "step": 32100 + }, + { + "epoch": 2.7, + "learning_rate": 5.103782128207901e-07, + "loss": 1.0963, + "step": 32101 + }, + { + "epoch": 2.7, + "learning_rate": 5.100915711133103e-07, + "loss": 1.1533, + "step": 32102 + }, + { + "epoch": 2.7, + "learning_rate": 5.098050078144911e-07, + "loss": 1.0781, + "step": 32103 + }, + { + "epoch": 2.7, + "learning_rate": 5.095185229266997e-07, + "loss": 1.1598, + "step": 32104 + }, + { + "epoch": 2.7, + "learning_rate": 5.092321164523028e-07, + "loss": 1.007, + "step": 32105 + }, + { + "epoch": 2.7, + "learning_rate": 5.089457883936688e-07, + "loss": 1.0643, + "step": 32106 + }, + { + "epoch": 2.7, + "learning_rate": 5.086595387531612e-07, + "loss": 1.1471, + "step": 32107 + }, + { + "epoch": 2.7, + "learning_rate": 5.08373367533147e-07, + "loss": 1.1278, + "step": 32108 + }, + { + "epoch": 2.7, + "learning_rate": 5.080872747359889e-07, + "loss": 1.1047, + "step": 32109 + }, + { + "epoch": 2.7, + "learning_rate": 5.078012603640525e-07, + "loss": 1.1042, + "step": 32110 + }, + { + "epoch": 2.7, + "learning_rate": 5.075153244197006e-07, + "loss": 1.1197, + "step": 32111 + }, + { + "epoch": 2.7, + "learning_rate": 5.072294669052924e-07, + "loss": 1.0442, + "step": 32112 + }, + { + "epoch": 2.7, + "learning_rate": 5.069436878231948e-07, + "loss": 1.1854, + "step": 32113 + }, + { + "epoch": 2.7, + "learning_rate": 5.06657987175766e-07, + "loss": 1.0941, + "step": 32114 + }, + { + "epoch": 2.7, + "learning_rate": 5.063723649653673e-07, + "loss": 1.1188, + "step": 32115 + }, + { + "epoch": 2.7, + "learning_rate": 5.060868211943581e-07, + "loss": 1.185, + "step": 32116 + }, + { + "epoch": 2.7, + "learning_rate": 5.058013558650987e-07, + "loss": 1.1337, + "step": 32117 + }, + { + "epoch": 2.7, + "learning_rate": 5.05515968979946e-07, + "loss": 1.1365, + "step": 32118 + }, + { + "epoch": 2.7, + "learning_rate": 5.052306605412605e-07, + "loss": 1.1113, + "step": 32119 + }, + { + "epoch": 2.7, + "learning_rate": 5.049454305513968e-07, + "loss": 1.1322, + "step": 32120 + }, + { + "epoch": 2.7, + "learning_rate": 5.046602790127108e-07, + "loss": 1.1151, + "step": 32121 + }, + { + "epoch": 2.7, + "learning_rate": 5.043752059275641e-07, + "loss": 1.0613, + "step": 32122 + }, + { + "epoch": 2.7, + "learning_rate": 5.04090211298307e-07, + "loss": 1.1214, + "step": 32123 + }, + { + "epoch": 2.7, + "learning_rate": 5.038052951272954e-07, + "loss": 1.1314, + "step": 32124 + }, + { + "epoch": 2.7, + "learning_rate": 5.035204574168839e-07, + "loss": 1.1505, + "step": 32125 + }, + { + "epoch": 2.7, + "learning_rate": 5.032356981694253e-07, + "loss": 1.0646, + "step": 32126 + }, + { + "epoch": 2.7, + "learning_rate": 5.029510173872732e-07, + "loss": 1.1449, + "step": 32127 + }, + { + "epoch": 2.7, + "learning_rate": 5.0266641507278e-07, + "loss": 1.1137, + "step": 32128 + }, + { + "epoch": 2.7, + "learning_rate": 5.02381891228294e-07, + "loss": 1.1664, + "step": 32129 + }, + { + "epoch": 2.71, + "learning_rate": 5.020974458561711e-07, + "loss": 1.0561, + "step": 32130 + }, + { + "epoch": 2.71, + "learning_rate": 5.018130789587572e-07, + "loss": 1.0546, + "step": 32131 + }, + { + "epoch": 2.71, + "learning_rate": 5.015287905384036e-07, + "loss": 1.1125, + "step": 32132 + }, + { + "epoch": 2.71, + "learning_rate": 5.012445805974598e-07, + "loss": 1.1615, + "step": 32133 + }, + { + "epoch": 2.71, + "learning_rate": 5.009604491382725e-07, + "loss": 1.1156, + "step": 32134 + }, + { + "epoch": 2.71, + "learning_rate": 5.0067639616319e-07, + "loss": 1.0756, + "step": 32135 + }, + { + "epoch": 2.71, + "learning_rate": 5.003924216745593e-07, + "loss": 1.1292, + "step": 32136 + }, + { + "epoch": 2.71, + "learning_rate": 5.001085256747262e-07, + "loss": 1.1224, + "step": 32137 + }, + { + "epoch": 2.71, + "learning_rate": 4.998247081660379e-07, + "loss": 1.0502, + "step": 32138 + }, + { + "epoch": 2.71, + "learning_rate": 4.99540969150839e-07, + "loss": 1.154, + "step": 32139 + }, + { + "epoch": 2.71, + "learning_rate": 4.992573086314712e-07, + "loss": 1.1806, + "step": 32140 + }, + { + "epoch": 2.71, + "learning_rate": 4.989737266102824e-07, + "loss": 1.0179, + "step": 32141 + }, + { + "epoch": 2.71, + "learning_rate": 4.986902230896118e-07, + "loss": 1.0787, + "step": 32142 + }, + { + "epoch": 2.71, + "learning_rate": 4.984067980718033e-07, + "loss": 1.2228, + "step": 32143 + }, + { + "epoch": 2.71, + "learning_rate": 4.981234515591993e-07, + "loss": 1.115, + "step": 32144 + }, + { + "epoch": 2.71, + "learning_rate": 4.978401835541402e-07, + "loss": 1.1256, + "step": 32145 + }, + { + "epoch": 2.71, + "learning_rate": 4.975569940589653e-07, + "loss": 1.1912, + "step": 32146 + }, + { + "epoch": 2.71, + "learning_rate": 4.972738830760171e-07, + "loss": 1.091, + "step": 32147 + }, + { + "epoch": 2.71, + "learning_rate": 4.969908506076337e-07, + "loss": 1.1327, + "step": 32148 + }, + { + "epoch": 2.71, + "learning_rate": 4.967078966561523e-07, + "loss": 1.0182, + "step": 32149 + }, + { + "epoch": 2.71, + "learning_rate": 4.96425021223913e-07, + "loss": 1.1973, + "step": 32150 + }, + { + "epoch": 2.71, + "learning_rate": 4.961422243132497e-07, + "loss": 1.0597, + "step": 32151 + }, + { + "epoch": 2.71, + "learning_rate": 4.958595059265026e-07, + "loss": 1.0133, + "step": 32152 + }, + { + "epoch": 2.71, + "learning_rate": 4.955768660660054e-07, + "loss": 1.1538, + "step": 32153 + }, + { + "epoch": 2.71, + "learning_rate": 4.952943047340942e-07, + "loss": 1.2034, + "step": 32154 + }, + { + "epoch": 2.71, + "learning_rate": 4.950118219331024e-07, + "loss": 1.1479, + "step": 32155 + }, + { + "epoch": 2.71, + "learning_rate": 4.947294176653661e-07, + "loss": 1.1782, + "step": 32156 + }, + { + "epoch": 2.71, + "learning_rate": 4.944470919332168e-07, + "loss": 1.2279, + "step": 32157 + }, + { + "epoch": 2.71, + "learning_rate": 4.94164844738988e-07, + "loss": 1.0999, + "step": 32158 + }, + { + "epoch": 2.71, + "learning_rate": 4.938826760850113e-07, + "loss": 1.1359, + "step": 32159 + }, + { + "epoch": 2.71, + "learning_rate": 4.936005859736181e-07, + "loss": 1.1217, + "step": 32160 + }, + { + "epoch": 2.71, + "learning_rate": 4.933185744071411e-07, + "loss": 1.0393, + "step": 32161 + }, + { + "epoch": 2.71, + "learning_rate": 4.930366413879051e-07, + "loss": 1.1057, + "step": 32162 + }, + { + "epoch": 2.71, + "learning_rate": 4.927547869182458e-07, + "loss": 1.1102, + "step": 32163 + }, + { + "epoch": 2.71, + "learning_rate": 4.924730110004883e-07, + "loss": 1.1896, + "step": 32164 + }, + { + "epoch": 2.71, + "learning_rate": 4.921913136369616e-07, + "loss": 1.0763, + "step": 32165 + }, + { + "epoch": 2.71, + "learning_rate": 4.919096948299939e-07, + "loss": 1.1492, + "step": 32166 + }, + { + "epoch": 2.71, + "learning_rate": 4.916281545819113e-07, + "loss": 1.2202, + "step": 32167 + }, + { + "epoch": 2.71, + "learning_rate": 4.913466928950372e-07, + "loss": 1.1693, + "step": 32168 + }, + { + "epoch": 2.71, + "learning_rate": 4.910653097717033e-07, + "loss": 1.0691, + "step": 32169 + }, + { + "epoch": 2.71, + "learning_rate": 4.907840052142299e-07, + "loss": 1.1322, + "step": 32170 + }, + { + "epoch": 2.71, + "learning_rate": 4.905027792249406e-07, + "loss": 1.0684, + "step": 32171 + }, + { + "epoch": 2.71, + "learning_rate": 4.902216318061637e-07, + "loss": 1.1859, + "step": 32172 + }, + { + "epoch": 2.71, + "learning_rate": 4.899405629602172e-07, + "loss": 1.197, + "step": 32173 + }, + { + "epoch": 2.71, + "learning_rate": 4.89659572689426e-07, + "loss": 1.0984, + "step": 32174 + }, + { + "epoch": 2.71, + "learning_rate": 4.893786609961115e-07, + "loss": 1.1248, + "step": 32175 + }, + { + "epoch": 2.71, + "learning_rate": 4.89097827882593e-07, + "loss": 1.1403, + "step": 32176 + }, + { + "epoch": 2.71, + "learning_rate": 4.888170733511932e-07, + "loss": 1.0882, + "step": 32177 + }, + { + "epoch": 2.71, + "learning_rate": 4.885363974042324e-07, + "loss": 1.175, + "step": 32178 + }, + { + "epoch": 2.71, + "learning_rate": 4.88255800044024e-07, + "loss": 1.1139, + "step": 32179 + }, + { + "epoch": 2.71, + "learning_rate": 4.879752812728933e-07, + "loss": 1.1391, + "step": 32180 + }, + { + "epoch": 2.71, + "learning_rate": 4.876948410931548e-07, + "loss": 1.1026, + "step": 32181 + }, + { + "epoch": 2.71, + "learning_rate": 4.874144795071245e-07, + "loss": 1.0738, + "step": 32182 + }, + { + "epoch": 2.71, + "learning_rate": 4.871341965171206e-07, + "loss": 1.0957, + "step": 32183 + }, + { + "epoch": 2.71, + "learning_rate": 4.868539921254589e-07, + "loss": 1.1071, + "step": 32184 + }, + { + "epoch": 2.71, + "learning_rate": 4.865738663344543e-07, + "loss": 1.1078, + "step": 32185 + }, + { + "epoch": 2.71, + "learning_rate": 4.862938191464206e-07, + "loss": 1.1006, + "step": 32186 + }, + { + "epoch": 2.71, + "learning_rate": 4.860138505636713e-07, + "loss": 1.0622, + "step": 32187 + }, + { + "epoch": 2.71, + "learning_rate": 4.857339605885214e-07, + "loss": 1.1973, + "step": 32188 + }, + { + "epoch": 2.71, + "learning_rate": 4.854541492232845e-07, + "loss": 1.2082, + "step": 32189 + }, + { + "epoch": 2.71, + "learning_rate": 4.851744164702665e-07, + "loss": 1.1542, + "step": 32190 + }, + { + "epoch": 2.71, + "learning_rate": 4.848947623317856e-07, + "loss": 1.0913, + "step": 32191 + }, + { + "epoch": 2.71, + "learning_rate": 4.846151868101489e-07, + "loss": 1.0573, + "step": 32192 + }, + { + "epoch": 2.71, + "learning_rate": 4.843356899076656e-07, + "loss": 1.109, + "step": 32193 + }, + { + "epoch": 2.71, + "learning_rate": 4.840562716266473e-07, + "loss": 1.1253, + "step": 32194 + }, + { + "epoch": 2.71, + "learning_rate": 4.837769319694008e-07, + "loss": 1.1382, + "step": 32195 + }, + { + "epoch": 2.71, + "learning_rate": 4.834976709382356e-07, + "loss": 1.1439, + "step": 32196 + }, + { + "epoch": 2.71, + "learning_rate": 4.832184885354574e-07, + "loss": 1.0963, + "step": 32197 + }, + { + "epoch": 2.71, + "learning_rate": 4.829393847633746e-07, + "loss": 1.1673, + "step": 32198 + }, + { + "epoch": 2.71, + "learning_rate": 4.826603596242918e-07, + "loss": 1.1206, + "step": 32199 + }, + { + "epoch": 2.71, + "learning_rate": 4.823814131205174e-07, + "loss": 1.1485, + "step": 32200 + }, + { + "epoch": 2.71, + "learning_rate": 4.821025452543504e-07, + "loss": 1.1241, + "step": 32201 + }, + { + "epoch": 2.71, + "learning_rate": 4.818237560281014e-07, + "loss": 1.1328, + "step": 32202 + }, + { + "epoch": 2.71, + "learning_rate": 4.815450454440685e-07, + "loss": 1.0897, + "step": 32203 + }, + { + "epoch": 2.71, + "learning_rate": 4.812664135045575e-07, + "loss": 1.0704, + "step": 32204 + }, + { + "epoch": 2.71, + "learning_rate": 4.809878602118689e-07, + "loss": 1.0722, + "step": 32205 + }, + { + "epoch": 2.71, + "learning_rate": 4.807093855683064e-07, + "loss": 1.1056, + "step": 32206 + }, + { + "epoch": 2.71, + "learning_rate": 4.804309895761683e-07, + "loss": 1.0645, + "step": 32207 + }, + { + "epoch": 2.71, + "learning_rate": 4.80152672237757e-07, + "loss": 1.1805, + "step": 32208 + }, + { + "epoch": 2.71, + "learning_rate": 4.798744335553707e-07, + "loss": 1.1007, + "step": 32209 + }, + { + "epoch": 2.71, + "learning_rate": 4.795962735313087e-07, + "loss": 1.1245, + "step": 32210 + }, + { + "epoch": 2.71, + "learning_rate": 4.793181921678702e-07, + "loss": 1.117, + "step": 32211 + }, + { + "epoch": 2.71, + "learning_rate": 4.790401894673491e-07, + "loss": 1.0589, + "step": 32212 + }, + { + "epoch": 2.71, + "learning_rate": 4.787622654320479e-07, + "loss": 1.2152, + "step": 32213 + }, + { + "epoch": 2.71, + "learning_rate": 4.784844200642591e-07, + "loss": 1.1508, + "step": 32214 + }, + { + "epoch": 2.71, + "learning_rate": 4.782066533662788e-07, + "loss": 1.1119, + "step": 32215 + }, + { + "epoch": 2.71, + "learning_rate": 4.779289653404029e-07, + "loss": 1.1613, + "step": 32216 + }, + { + "epoch": 2.71, + "learning_rate": 4.776513559889262e-07, + "loss": 1.1168, + "step": 32217 + }, + { + "epoch": 2.71, + "learning_rate": 4.773738253141391e-07, + "loss": 1.1468, + "step": 32218 + }, + { + "epoch": 2.71, + "learning_rate": 4.770963733183398e-07, + "loss": 1.0304, + "step": 32219 + }, + { + "epoch": 2.71, + "learning_rate": 4.7681900000381644e-07, + "loss": 1.0628, + "step": 32220 + }, + { + "epoch": 2.71, + "learning_rate": 4.765417053728616e-07, + "loss": 1.1328, + "step": 32221 + }, + { + "epoch": 2.71, + "learning_rate": 4.7626448942776904e-07, + "loss": 1.1648, + "step": 32222 + }, + { + "epoch": 2.71, + "learning_rate": 4.759873521708258e-07, + "loss": 1.0876, + "step": 32223 + }, + { + "epoch": 2.71, + "learning_rate": 4.757102936043234e-07, + "loss": 1.076, + "step": 32224 + }, + { + "epoch": 2.71, + "learning_rate": 4.754333137305511e-07, + "loss": 1.1032, + "step": 32225 + }, + { + "epoch": 2.71, + "learning_rate": 4.751564125517971e-07, + "loss": 1.1519, + "step": 32226 + }, + { + "epoch": 2.71, + "learning_rate": 4.7487959007034844e-07, + "loss": 1.0421, + "step": 32227 + }, + { + "epoch": 2.71, + "learning_rate": 4.7460284628849553e-07, + "loss": 1.221, + "step": 32228 + }, + { + "epoch": 2.71, + "learning_rate": 4.743261812085187e-07, + "loss": 1.0529, + "step": 32229 + }, + { + "epoch": 2.71, + "learning_rate": 4.740495948327106e-07, + "loss": 1.0987, + "step": 32230 + }, + { + "epoch": 2.71, + "learning_rate": 4.737730871633517e-07, + "loss": 1.1572, + "step": 32231 + }, + { + "epoch": 2.71, + "learning_rate": 4.734966582027289e-07, + "loss": 1.1139, + "step": 32232 + }, + { + "epoch": 2.71, + "learning_rate": 4.73220307953125e-07, + "loss": 1.0994, + "step": 32233 + }, + { + "epoch": 2.71, + "learning_rate": 4.7294403641682473e-07, + "loss": 1.13, + "step": 32234 + }, + { + "epoch": 2.71, + "learning_rate": 4.726678435961085e-07, + "loss": 1.1628, + "step": 32235 + }, + { + "epoch": 2.71, + "learning_rate": 4.723917294932612e-07, + "loss": 1.1591, + "step": 32236 + }, + { + "epoch": 2.71, + "learning_rate": 4.72115694110562e-07, + "loss": 1.1212, + "step": 32237 + }, + { + "epoch": 2.71, + "learning_rate": 4.7183973745029253e-07, + "loss": 1.1363, + "step": 32238 + }, + { + "epoch": 2.71, + "learning_rate": 4.715638595147332e-07, + "loss": 1.0377, + "step": 32239 + }, + { + "epoch": 2.71, + "learning_rate": 4.7128806030615983e-07, + "loss": 1.1392, + "step": 32240 + }, + { + "epoch": 2.71, + "learning_rate": 4.7101233982685847e-07, + "loss": 1.2291, + "step": 32241 + }, + { + "epoch": 2.71, + "learning_rate": 4.707366980790995e-07, + "loss": 1.0997, + "step": 32242 + }, + { + "epoch": 2.71, + "learning_rate": 4.7046113506516556e-07, + "loss": 1.0316, + "step": 32243 + }, + { + "epoch": 2.71, + "learning_rate": 4.7018565078733036e-07, + "loss": 1.0885, + "step": 32244 + }, + { + "epoch": 2.71, + "learning_rate": 4.69910245247871e-07, + "loss": 1.115, + "step": 32245 + }, + { + "epoch": 2.71, + "learning_rate": 4.696349184490634e-07, + "loss": 1.0949, + "step": 32246 + }, + { + "epoch": 2.71, + "learning_rate": 4.693596703931835e-07, + "loss": 1.1594, + "step": 32247 + }, + { + "epoch": 2.72, + "learning_rate": 4.690845010825029e-07, + "loss": 1.1075, + "step": 32248 + }, + { + "epoch": 2.72, + "learning_rate": 4.688094105192964e-07, + "loss": 1.075, + "step": 32249 + }, + { + "epoch": 2.72, + "learning_rate": 4.6853439870583885e-07, + "loss": 1.166, + "step": 32250 + }, + { + "epoch": 2.72, + "learning_rate": 4.6825946564439726e-07, + "loss": 1.0624, + "step": 32251 + }, + { + "epoch": 2.72, + "learning_rate": 4.679846113372499e-07, + "loss": 1.1238, + "step": 32252 + }, + { + "epoch": 2.72, + "learning_rate": 4.677098357866627e-07, + "loss": 1.0529, + "step": 32253 + }, + { + "epoch": 2.72, + "learning_rate": 4.674351389949072e-07, + "loss": 0.9982, + "step": 32254 + }, + { + "epoch": 2.72, + "learning_rate": 4.6716052096425377e-07, + "loss": 1.1105, + "step": 32255 + }, + { + "epoch": 2.72, + "learning_rate": 4.6688598169697174e-07, + "loss": 1.1052, + "step": 32256 + }, + { + "epoch": 2.72, + "learning_rate": 4.6661152119532705e-07, + "loss": 1.1608, + "step": 32257 + }, + { + "epoch": 2.72, + "learning_rate": 4.663371394615901e-07, + "loss": 1.1932, + "step": 32258 + }, + { + "epoch": 2.72, + "learning_rate": 4.6606283649802684e-07, + "loss": 1.1127, + "step": 32259 + }, + { + "epoch": 2.72, + "learning_rate": 4.657886123069033e-07, + "loss": 1.0675, + "step": 32260 + }, + { + "epoch": 2.72, + "learning_rate": 4.6551446689048763e-07, + "loss": 1.1349, + "step": 32261 + }, + { + "epoch": 2.72, + "learning_rate": 4.652404002510391e-07, + "loss": 1.1964, + "step": 32262 + }, + { + "epoch": 2.72, + "learning_rate": 4.649664123908293e-07, + "loss": 1.1109, + "step": 32263 + }, + { + "epoch": 2.72, + "learning_rate": 4.6469250331211744e-07, + "loss": 1.0901, + "step": 32264 + }, + { + "epoch": 2.72, + "learning_rate": 4.644186730171674e-07, + "loss": 1.0569, + "step": 32265 + }, + { + "epoch": 2.72, + "learning_rate": 4.6414492150824273e-07, + "loss": 1.0745, + "step": 32266 + }, + { + "epoch": 2.72, + "learning_rate": 4.6387124878760513e-07, + "loss": 1.2016, + "step": 32267 + }, + { + "epoch": 2.72, + "learning_rate": 4.635976548575127e-07, + "loss": 1.174, + "step": 32268 + }, + { + "epoch": 2.72, + "learning_rate": 4.633241397202315e-07, + "loss": 1.1259, + "step": 32269 + }, + { + "epoch": 2.72, + "learning_rate": 4.6305070337801626e-07, + "loss": 1.0937, + "step": 32270 + }, + { + "epoch": 2.72, + "learning_rate": 4.627773458331275e-07, + "loss": 1.0165, + "step": 32271 + }, + { + "epoch": 2.72, + "learning_rate": 4.625040670878278e-07, + "loss": 1.0939, + "step": 32272 + }, + { + "epoch": 2.72, + "learning_rate": 4.6223086714436983e-07, + "loss": 1.1907, + "step": 32273 + }, + { + "epoch": 2.72, + "learning_rate": 4.6195774600501286e-07, + "loss": 1.1289, + "step": 32274 + }, + { + "epoch": 2.72, + "learning_rate": 4.61684703672014e-07, + "loss": 1.0822, + "step": 32275 + }, + { + "epoch": 2.72, + "learning_rate": 4.614117401476292e-07, + "loss": 1.0972, + "step": 32276 + }, + { + "epoch": 2.72, + "learning_rate": 4.611388554341123e-07, + "loss": 1.1362, + "step": 32277 + }, + { + "epoch": 2.72, + "learning_rate": 4.6086604953372026e-07, + "loss": 1.0645, + "step": 32278 + }, + { + "epoch": 2.72, + "learning_rate": 4.605933224487047e-07, + "loss": 1.126, + "step": 32279 + }, + { + "epoch": 2.72, + "learning_rate": 4.6032067418132153e-07, + "loss": 1.178, + "step": 32280 + }, + { + "epoch": 2.72, + "learning_rate": 4.600481047338212e-07, + "loss": 1.1513, + "step": 32281 + }, + { + "epoch": 2.72, + "learning_rate": 4.597756141084564e-07, + "loss": 1.1415, + "step": 32282 + }, + { + "epoch": 2.72, + "learning_rate": 4.595032023074786e-07, + "loss": 1.2378, + "step": 32283 + }, + { + "epoch": 2.72, + "learning_rate": 4.592308693331382e-07, + "loss": 1.0646, + "step": 32284 + }, + { + "epoch": 2.72, + "learning_rate": 4.589586151876868e-07, + "loss": 1.1724, + "step": 32285 + }, + { + "epoch": 2.72, + "learning_rate": 4.586864398733726e-07, + "loss": 1.0831, + "step": 32286 + }, + { + "epoch": 2.72, + "learning_rate": 4.584143433924437e-07, + "loss": 1.1638, + "step": 32287 + }, + { + "epoch": 2.72, + "learning_rate": 4.581423257471507e-07, + "loss": 1.1588, + "step": 32288 + }, + { + "epoch": 2.72, + "learning_rate": 4.578703869397394e-07, + "loss": 1.0936, + "step": 32289 + }, + { + "epoch": 2.72, + "learning_rate": 4.575985269724559e-07, + "loss": 1.0319, + "step": 32290 + }, + { + "epoch": 2.72, + "learning_rate": 4.5732674584754834e-07, + "loss": 1.1032, + "step": 32291 + }, + { + "epoch": 2.72, + "learning_rate": 4.570550435672605e-07, + "loss": 1.2544, + "step": 32292 + }, + { + "epoch": 2.72, + "learning_rate": 4.5678342013383836e-07, + "loss": 1.2385, + "step": 32293 + }, + { + "epoch": 2.72, + "learning_rate": 4.565118755495257e-07, + "loss": 1.0478, + "step": 32294 + }, + { + "epoch": 2.72, + "learning_rate": 4.5624040981656625e-07, + "loss": 1.1434, + "step": 32295 + }, + { + "epoch": 2.72, + "learning_rate": 4.559690229372027e-07, + "loss": 1.0903, + "step": 32296 + }, + { + "epoch": 2.72, + "learning_rate": 4.5569771491367766e-07, + "loss": 1.13, + "step": 32297 + }, + { + "epoch": 2.72, + "learning_rate": 4.5542648574823265e-07, + "loss": 1.0449, + "step": 32298 + }, + { + "epoch": 2.72, + "learning_rate": 4.551553354431093e-07, + "loss": 1.0473, + "step": 32299 + }, + { + "epoch": 2.72, + "learning_rate": 4.54884264000548e-07, + "loss": 1.142, + "step": 32300 + }, + { + "epoch": 2.72, + "learning_rate": 4.5461327142278466e-07, + "loss": 1.1507, + "step": 32301 + }, + { + "epoch": 2.72, + "learning_rate": 4.5434235771206423e-07, + "loss": 1.1458, + "step": 32302 + }, + { + "epoch": 2.72, + "learning_rate": 4.540715228706216e-07, + "loss": 1.0687, + "step": 32303 + }, + { + "epoch": 2.72, + "learning_rate": 4.5380076690069495e-07, + "loss": 1.1639, + "step": 32304 + }, + { + "epoch": 2.72, + "learning_rate": 4.535300898045214e-07, + "loss": 1.1556, + "step": 32305 + }, + { + "epoch": 2.72, + "learning_rate": 4.5325949158433693e-07, + "loss": 1.1521, + "step": 32306 + }, + { + "epoch": 2.72, + "learning_rate": 4.529889722423786e-07, + "loss": 1.0254, + "step": 32307 + }, + { + "epoch": 2.72, + "learning_rate": 4.527185317808802e-07, + "loss": 1.1103, + "step": 32308 + }, + { + "epoch": 2.72, + "learning_rate": 4.5244817020207775e-07, + "loss": 1.0375, + "step": 32309 + }, + { + "epoch": 2.72, + "learning_rate": 4.521778875082039e-07, + "loss": 1.1104, + "step": 32310 + }, + { + "epoch": 2.72, + "learning_rate": 4.519076837014924e-07, + "loss": 1.014, + "step": 32311 + }, + { + "epoch": 2.72, + "learning_rate": 4.516375587841737e-07, + "loss": 1.0868, + "step": 32312 + }, + { + "epoch": 2.72, + "learning_rate": 4.513675127584838e-07, + "loss": 1.117, + "step": 32313 + }, + { + "epoch": 2.72, + "learning_rate": 4.5109754562664974e-07, + "loss": 1.0864, + "step": 32314 + }, + { + "epoch": 2.72, + "learning_rate": 4.5082765739090427e-07, + "loss": 1.0628, + "step": 32315 + }, + { + "epoch": 2.72, + "learning_rate": 4.505578480534767e-07, + "loss": 1.0785, + "step": 32316 + }, + { + "epoch": 2.72, + "learning_rate": 4.502881176165974e-07, + "loss": 1.0935, + "step": 32317 + }, + { + "epoch": 2.72, + "learning_rate": 4.5001846608249243e-07, + "loss": 1.1769, + "step": 32318 + }, + { + "epoch": 2.72, + "learning_rate": 4.4974889345339225e-07, + "loss": 1.1158, + "step": 32319 + }, + { + "epoch": 2.72, + "learning_rate": 4.4947939973152276e-07, + "loss": 1.0795, + "step": 32320 + }, + { + "epoch": 2.72, + "learning_rate": 4.4920998491911003e-07, + "loss": 1.0668, + "step": 32321 + }, + { + "epoch": 2.72, + "learning_rate": 4.489406490183834e-07, + "loss": 1.0325, + "step": 32322 + }, + { + "epoch": 2.72, + "learning_rate": 4.486713920315644e-07, + "loss": 1.1195, + "step": 32323 + }, + { + "epoch": 2.72, + "learning_rate": 4.48402213960879e-07, + "loss": 1.1333, + "step": 32324 + }, + { + "epoch": 2.72, + "learning_rate": 4.4813311480855213e-07, + "loss": 1.0183, + "step": 32325 + }, + { + "epoch": 2.72, + "learning_rate": 4.4786409457680537e-07, + "loss": 1.1381, + "step": 32326 + }, + { + "epoch": 2.72, + "learning_rate": 4.4759515326786353e-07, + "loss": 1.1023, + "step": 32327 + }, + { + "epoch": 2.72, + "learning_rate": 4.473262908839482e-07, + "loss": 1.1671, + "step": 32328 + }, + { + "epoch": 2.72, + "learning_rate": 4.470575074272787e-07, + "loss": 1.0994, + "step": 32329 + }, + { + "epoch": 2.72, + "learning_rate": 4.4678880290007885e-07, + "loss": 1.128, + "step": 32330 + }, + { + "epoch": 2.72, + "learning_rate": 4.4652017730456685e-07, + "loss": 1.1517, + "step": 32331 + }, + { + "epoch": 2.72, + "learning_rate": 4.462516306429632e-07, + "loss": 1.0196, + "step": 32332 + }, + { + "epoch": 2.72, + "learning_rate": 4.4598316291748487e-07, + "loss": 1.1259, + "step": 32333 + }, + { + "epoch": 2.72, + "learning_rate": 4.457147741303514e-07, + "loss": 1.0592, + "step": 32334 + }, + { + "epoch": 2.72, + "learning_rate": 4.454464642837808e-07, + "loss": 1.1328, + "step": 32335 + }, + { + "epoch": 2.72, + "learning_rate": 4.451782333799892e-07, + "loss": 1.1166, + "step": 32336 + }, + { + "epoch": 2.72, + "learning_rate": 4.4491008142119376e-07, + "loss": 1.0458, + "step": 32337 + }, + { + "epoch": 2.72, + "learning_rate": 4.4464200840960816e-07, + "loss": 1.1904, + "step": 32338 + }, + { + "epoch": 2.72, + "learning_rate": 4.4437401434744953e-07, + "loss": 1.1198, + "step": 32339 + }, + { + "epoch": 2.72, + "learning_rate": 4.4410609923692947e-07, + "loss": 1.0884, + "step": 32340 + }, + { + "epoch": 2.72, + "learning_rate": 4.4383826308026514e-07, + "loss": 1.1384, + "step": 32341 + }, + { + "epoch": 2.72, + "learning_rate": 4.4357050587966576e-07, + "loss": 1.0778, + "step": 32342 + }, + { + "epoch": 2.72, + "learning_rate": 4.433028276373452e-07, + "loss": 1.1737, + "step": 32343 + }, + { + "epoch": 2.72, + "learning_rate": 4.4303522835551617e-07, + "loss": 1.1263, + "step": 32344 + }, + { + "epoch": 2.72, + "learning_rate": 4.4276770803638794e-07, + "loss": 1.0377, + "step": 32345 + }, + { + "epoch": 2.72, + "learning_rate": 4.425002666821721e-07, + "loss": 1.1296, + "step": 32346 + }, + { + "epoch": 2.72, + "learning_rate": 4.422329042950779e-07, + "loss": 1.2149, + "step": 32347 + }, + { + "epoch": 2.72, + "learning_rate": 4.4196562087731377e-07, + "loss": 1.0323, + "step": 32348 + }, + { + "epoch": 2.72, + "learning_rate": 4.416984164310889e-07, + "loss": 1.2023, + "step": 32349 + }, + { + "epoch": 2.72, + "learning_rate": 4.414312909586116e-07, + "loss": 1.0761, + "step": 32350 + }, + { + "epoch": 2.72, + "learning_rate": 4.411642444620867e-07, + "loss": 1.0528, + "step": 32351 + }, + { + "epoch": 2.72, + "learning_rate": 4.408972769437236e-07, + "loss": 1.1506, + "step": 32352 + }, + { + "epoch": 2.72, + "learning_rate": 4.4063038840572505e-07, + "loss": 1.0278, + "step": 32353 + }, + { + "epoch": 2.72, + "learning_rate": 4.403635788502991e-07, + "loss": 1.0997, + "step": 32354 + }, + { + "epoch": 2.72, + "learning_rate": 4.400968482796475e-07, + "loss": 1.097, + "step": 32355 + }, + { + "epoch": 2.72, + "learning_rate": 4.3983019669597506e-07, + "loss": 1.1206, + "step": 32356 + }, + { + "epoch": 2.72, + "learning_rate": 4.395636241014856e-07, + "loss": 1.1066, + "step": 32357 + }, + { + "epoch": 2.72, + "learning_rate": 4.3929713049838173e-07, + "loss": 1.1713, + "step": 32358 + }, + { + "epoch": 2.72, + "learning_rate": 4.39030715888864e-07, + "loss": 1.1721, + "step": 32359 + }, + { + "epoch": 2.72, + "learning_rate": 4.3876438027513515e-07, + "loss": 1.2055, + "step": 32360 + }, + { + "epoch": 2.72, + "learning_rate": 4.384981236593955e-07, + "loss": 1.0251, + "step": 32361 + }, + { + "epoch": 2.72, + "learning_rate": 4.382319460438422e-07, + "loss": 1.0852, + "step": 32362 + }, + { + "epoch": 2.72, + "learning_rate": 4.3796584743067916e-07, + "loss": 1.1283, + "step": 32363 + }, + { + "epoch": 2.72, + "learning_rate": 4.376998278221012e-07, + "loss": 1.1227, + "step": 32364 + }, + { + "epoch": 2.72, + "learning_rate": 4.3743388722030764e-07, + "loss": 1.1165, + "step": 32365 + }, + { + "epoch": 2.72, + "learning_rate": 4.3716802562749574e-07, + "loss": 1.1257, + "step": 32366 + }, + { + "epoch": 2.73, + "learning_rate": 4.369022430458625e-07, + "loss": 1.053, + "step": 32367 + }, + { + "epoch": 2.73, + "learning_rate": 4.366365394776029e-07, + "loss": 1.0088, + "step": 32368 + }, + { + "epoch": 2.73, + "learning_rate": 4.363709149249151e-07, + "loss": 1.1936, + "step": 32369 + }, + { + "epoch": 2.73, + "learning_rate": 4.361053693899886e-07, + "loss": 1.129, + "step": 32370 + }, + { + "epoch": 2.73, + "learning_rate": 4.3583990287502043e-07, + "loss": 1.1052, + "step": 32371 + }, + { + "epoch": 2.73, + "learning_rate": 4.355745153822066e-07, + "loss": 1.0169, + "step": 32372 + }, + { + "epoch": 2.73, + "learning_rate": 4.3530920691373433e-07, + "loss": 1.093, + "step": 32373 + }, + { + "epoch": 2.73, + "learning_rate": 4.350439774718007e-07, + "loss": 1.003, + "step": 32374 + }, + { + "epoch": 2.73, + "learning_rate": 4.34778827058594e-07, + "loss": 1.1622, + "step": 32375 + }, + { + "epoch": 2.73, + "learning_rate": 4.3451375567630573e-07, + "loss": 1.0794, + "step": 32376 + }, + { + "epoch": 2.73, + "learning_rate": 4.3424876332712526e-07, + "loss": 1.0701, + "step": 32377 + }, + { + "epoch": 2.73, + "learning_rate": 4.339838500132454e-07, + "loss": 1.0781, + "step": 32378 + }, + { + "epoch": 2.73, + "learning_rate": 4.337190157368498e-07, + "loss": 1.093, + "step": 32379 + }, + { + "epoch": 2.73, + "learning_rate": 4.334542605001313e-07, + "loss": 1.1805, + "step": 32380 + }, + { + "epoch": 2.73, + "learning_rate": 4.331895843052747e-07, + "loss": 1.2082, + "step": 32381 + }, + { + "epoch": 2.73, + "learning_rate": 4.329249871544672e-07, + "loss": 1.1247, + "step": 32382 + }, + { + "epoch": 2.73, + "learning_rate": 4.3266046904989587e-07, + "loss": 1.1967, + "step": 32383 + }, + { + "epoch": 2.73, + "learning_rate": 4.3239602999374464e-07, + "loss": 1.2037, + "step": 32384 + }, + { + "epoch": 2.73, + "learning_rate": 4.3213166998820053e-07, + "loss": 1.0725, + "step": 32385 + }, + { + "epoch": 2.73, + "learning_rate": 4.3186738903544634e-07, + "loss": 1.0557, + "step": 32386 + }, + { + "epoch": 2.73, + "learning_rate": 4.316031871376658e-07, + "loss": 1.1125, + "step": 32387 + }, + { + "epoch": 2.73, + "learning_rate": 4.3133906429704163e-07, + "loss": 1.0897, + "step": 32388 + }, + { + "epoch": 2.73, + "learning_rate": 4.310750205157577e-07, + "loss": 1.0836, + "step": 32389 + }, + { + "epoch": 2.73, + "learning_rate": 4.3081105579599216e-07, + "loss": 1.2178, + "step": 32390 + }, + { + "epoch": 2.73, + "learning_rate": 4.3054717013993107e-07, + "loss": 1.1235, + "step": 32391 + }, + { + "epoch": 2.73, + "learning_rate": 4.302833635497505e-07, + "loss": 1.0694, + "step": 32392 + }, + { + "epoch": 2.73, + "learning_rate": 4.30019636027631e-07, + "loss": 1.1284, + "step": 32393 + }, + { + "epoch": 2.73, + "learning_rate": 4.2975598757575286e-07, + "loss": 1.1309, + "step": 32394 + }, + { + "epoch": 2.73, + "learning_rate": 4.294924181962934e-07, + "loss": 1.0479, + "step": 32395 + }, + { + "epoch": 2.73, + "learning_rate": 4.2922892789142966e-07, + "loss": 1.0633, + "step": 32396 + }, + { + "epoch": 2.73, + "learning_rate": 4.289655166633411e-07, + "loss": 1.1122, + "step": 32397 + }, + { + "epoch": 2.73, + "learning_rate": 4.287021845142014e-07, + "loss": 1.0226, + "step": 32398 + }, + { + "epoch": 2.73, + "learning_rate": 4.2843893144618787e-07, + "loss": 1.0919, + "step": 32399 + }, + { + "epoch": 2.73, + "learning_rate": 4.2817575746147644e-07, + "loss": 1.049, + "step": 32400 + }, + { + "epoch": 2.73, + "learning_rate": 4.2791266256223765e-07, + "loss": 1.129, + "step": 32401 + }, + { + "epoch": 2.73, + "learning_rate": 4.276496467506497e-07, + "loss": 1.1933, + "step": 32402 + }, + { + "epoch": 2.73, + "learning_rate": 4.273867100288842e-07, + "loss": 1.2139, + "step": 32403 + }, + { + "epoch": 2.73, + "learning_rate": 4.271238523991117e-07, + "loss": 1.1378, + "step": 32404 + }, + { + "epoch": 2.73, + "learning_rate": 4.268610738635071e-07, + "loss": 1.0542, + "step": 32405 + }, + { + "epoch": 2.73, + "learning_rate": 4.2659837442423857e-07, + "loss": 1.1035, + "step": 32406 + }, + { + "epoch": 2.73, + "learning_rate": 4.2633575408348007e-07, + "loss": 1.1539, + "step": 32407 + }, + { + "epoch": 2.73, + "learning_rate": 4.2607321284339754e-07, + "loss": 1.1203, + "step": 32408 + }, + { + "epoch": 2.73, + "learning_rate": 4.2581075070616374e-07, + "loss": 1.0288, + "step": 32409 + }, + { + "epoch": 2.73, + "learning_rate": 4.255483676739447e-07, + "loss": 1.1876, + "step": 32410 + }, + { + "epoch": 2.73, + "learning_rate": 4.252860637489109e-07, + "loss": 1.1761, + "step": 32411 + }, + { + "epoch": 2.73, + "learning_rate": 4.250238389332251e-07, + "loss": 1.1168, + "step": 32412 + }, + { + "epoch": 2.73, + "learning_rate": 4.247616932290599e-07, + "loss": 1.1402, + "step": 32413 + }, + { + "epoch": 2.73, + "learning_rate": 4.244996266385759e-07, + "loss": 1.0842, + "step": 32414 + }, + { + "epoch": 2.73, + "learning_rate": 4.242376391639413e-07, + "loss": 1.0715, + "step": 32415 + }, + { + "epoch": 2.73, + "learning_rate": 4.2397573080732e-07, + "loss": 1.0976, + "step": 32416 + }, + { + "epoch": 2.73, + "learning_rate": 4.237139015708758e-07, + "loss": 1.1012, + "step": 32417 + }, + { + "epoch": 2.73, + "learning_rate": 4.234521514567713e-07, + "loss": 1.0293, + "step": 32418 + }, + { + "epoch": 2.73, + "learning_rate": 4.231904804671727e-07, + "loss": 1.1383, + "step": 32419 + }, + { + "epoch": 2.73, + "learning_rate": 4.2292888860423596e-07, + "loss": 1.1496, + "step": 32420 + }, + { + "epoch": 2.73, + "learning_rate": 4.2266737587012716e-07, + "loss": 1.0917, + "step": 32421 + }, + { + "epoch": 2.73, + "learning_rate": 4.2240594226700684e-07, + "loss": 1.0721, + "step": 32422 + }, + { + "epoch": 2.73, + "learning_rate": 4.2214458779703096e-07, + "loss": 1.1556, + "step": 32423 + }, + { + "epoch": 2.73, + "learning_rate": 4.218833124623656e-07, + "loss": 1.1491, + "step": 32424 + }, + { + "epoch": 2.73, + "learning_rate": 4.216221162651635e-07, + "loss": 1.1012, + "step": 32425 + }, + { + "epoch": 2.73, + "learning_rate": 4.213609992075851e-07, + "loss": 1.0789, + "step": 32426 + }, + { + "epoch": 2.73, + "learning_rate": 4.2109996129178765e-07, + "loss": 1.1238, + "step": 32427 + }, + { + "epoch": 2.73, + "learning_rate": 4.208390025199294e-07, + "loss": 1.0345, + "step": 32428 + }, + { + "epoch": 2.73, + "learning_rate": 4.205781228941619e-07, + "loss": 1.1214, + "step": 32429 + }, + { + "epoch": 2.73, + "learning_rate": 4.203173224166468e-07, + "loss": 1.081, + "step": 32430 + }, + { + "epoch": 2.73, + "learning_rate": 4.2005660108953463e-07, + "loss": 1.1175, + "step": 32431 + }, + { + "epoch": 2.73, + "learning_rate": 4.197959589149803e-07, + "loss": 1.1175, + "step": 32432 + }, + { + "epoch": 2.73, + "learning_rate": 4.195353958951387e-07, + "loss": 1.0691, + "step": 32433 + }, + { + "epoch": 2.73, + "learning_rate": 4.192749120321604e-07, + "loss": 1.0817, + "step": 32434 + }, + { + "epoch": 2.73, + "learning_rate": 4.1901450732820037e-07, + "loss": 1.0094, + "step": 32435 + }, + { + "epoch": 2.73, + "learning_rate": 4.1875418178540795e-07, + "loss": 1.1426, + "step": 32436 + }, + { + "epoch": 2.73, + "learning_rate": 4.1849393540593585e-07, + "loss": 1.1515, + "step": 32437 + }, + { + "epoch": 2.73, + "learning_rate": 4.182337681919324e-07, + "loss": 1.1518, + "step": 32438 + }, + { + "epoch": 2.73, + "learning_rate": 4.1797368014555026e-07, + "loss": 1.112, + "step": 32439 + }, + { + "epoch": 2.73, + "learning_rate": 4.1771367126893225e-07, + "loss": 1.1031, + "step": 32440 + }, + { + "epoch": 2.73, + "learning_rate": 4.1745374156423436e-07, + "loss": 1.1551, + "step": 32441 + }, + { + "epoch": 2.73, + "learning_rate": 4.1719389103359933e-07, + "loss": 1.1576, + "step": 32442 + }, + { + "epoch": 2.73, + "learning_rate": 4.169341196791743e-07, + "loss": 1.0684, + "step": 32443 + }, + { + "epoch": 2.73, + "learning_rate": 4.166744275031076e-07, + "loss": 1.0487, + "step": 32444 + }, + { + "epoch": 2.73, + "learning_rate": 4.1641481450754304e-07, + "loss": 1.1919, + "step": 32445 + }, + { + "epoch": 2.73, + "learning_rate": 4.1615528069462675e-07, + "loss": 1.105, + "step": 32446 + }, + { + "epoch": 2.73, + "learning_rate": 4.158958260665025e-07, + "loss": 1.1273, + "step": 32447 + }, + { + "epoch": 2.73, + "learning_rate": 4.156364506253141e-07, + "loss": 1.0234, + "step": 32448 + }, + { + "epoch": 2.73, + "learning_rate": 4.1537715437320547e-07, + "loss": 1.0501, + "step": 32449 + }, + { + "epoch": 2.73, + "learning_rate": 4.151179373123193e-07, + "loss": 1.085, + "step": 32450 + }, + { + "epoch": 2.73, + "learning_rate": 4.148587994447928e-07, + "loss": 1.1455, + "step": 32451 + }, + { + "epoch": 2.73, + "learning_rate": 4.145997407727742e-07, + "loss": 1.1559, + "step": 32452 + }, + { + "epoch": 2.73, + "learning_rate": 4.1434076129839853e-07, + "loss": 1.1429, + "step": 32453 + }, + { + "epoch": 2.73, + "learning_rate": 4.1408186102380733e-07, + "loss": 1.2104, + "step": 32454 + }, + { + "epoch": 2.73, + "learning_rate": 4.1382303995114e-07, + "loss": 1.2122, + "step": 32455 + }, + { + "epoch": 2.73, + "learning_rate": 4.1356429808253386e-07, + "loss": 1.0746, + "step": 32456 + }, + { + "epoch": 2.73, + "learning_rate": 4.133056354201281e-07, + "loss": 1.0643, + "step": 32457 + }, + { + "epoch": 2.73, + "learning_rate": 4.130470519660601e-07, + "loss": 1.0517, + "step": 32458 + }, + { + "epoch": 2.73, + "learning_rate": 4.127885477224647e-07, + "loss": 1.1921, + "step": 32459 + }, + { + "epoch": 2.73, + "learning_rate": 4.1253012269147796e-07, + "loss": 1.1249, + "step": 32460 + }, + { + "epoch": 2.73, + "learning_rate": 4.1227177687523825e-07, + "loss": 1.0414, + "step": 32461 + }, + { + "epoch": 2.73, + "learning_rate": 4.120135102758749e-07, + "loss": 1.1079, + "step": 32462 + }, + { + "epoch": 2.73, + "learning_rate": 4.117553228955262e-07, + "loss": 1.121, + "step": 32463 + }, + { + "epoch": 2.73, + "learning_rate": 4.114972147363228e-07, + "loss": 1.0692, + "step": 32464 + }, + { + "epoch": 2.73, + "learning_rate": 4.1123918580039836e-07, + "loss": 1.1054, + "step": 32465 + }, + { + "epoch": 2.73, + "learning_rate": 4.109812360898846e-07, + "loss": 1.1253, + "step": 32466 + }, + { + "epoch": 2.73, + "learning_rate": 4.1072336560691207e-07, + "loss": 1.1779, + "step": 32467 + }, + { + "epoch": 2.73, + "learning_rate": 4.1046557435361236e-07, + "loss": 1.0827, + "step": 32468 + }, + { + "epoch": 2.73, + "learning_rate": 4.1020786233211706e-07, + "loss": 1.0502, + "step": 32469 + }, + { + "epoch": 2.73, + "learning_rate": 4.099502295445501e-07, + "loss": 1.1966, + "step": 32470 + }, + { + "epoch": 2.73, + "learning_rate": 4.096926759930453e-07, + "loss": 1.1323, + "step": 32471 + }, + { + "epoch": 2.73, + "learning_rate": 4.094352016797298e-07, + "loss": 1.1582, + "step": 32472 + }, + { + "epoch": 2.73, + "learning_rate": 4.091778066067276e-07, + "loss": 1.1596, + "step": 32473 + }, + { + "epoch": 2.73, + "learning_rate": 4.0892049077617125e-07, + "loss": 1.1474, + "step": 32474 + }, + { + "epoch": 2.73, + "learning_rate": 4.086632541901814e-07, + "loss": 1.132, + "step": 32475 + }, + { + "epoch": 2.73, + "learning_rate": 4.0840609685088516e-07, + "loss": 1.0517, + "step": 32476 + }, + { + "epoch": 2.73, + "learning_rate": 4.0814901876040757e-07, + "loss": 1.0745, + "step": 32477 + }, + { + "epoch": 2.73, + "learning_rate": 4.078920199208736e-07, + "loss": 1.14, + "step": 32478 + }, + { + "epoch": 2.73, + "learning_rate": 4.0763510033440366e-07, + "loss": 1.1305, + "step": 32479 + }, + { + "epoch": 2.73, + "learning_rate": 4.07378260003124e-07, + "loss": 1.0475, + "step": 32480 + }, + { + "epoch": 2.73, + "learning_rate": 4.07121498929155e-07, + "loss": 1.143, + "step": 32481 + }, + { + "epoch": 2.73, + "learning_rate": 4.0686481711461836e-07, + "loss": 1.1458, + "step": 32482 + }, + { + "epoch": 2.73, + "learning_rate": 4.0660821456163345e-07, + "loss": 1.122, + "step": 32483 + }, + { + "epoch": 2.73, + "learning_rate": 4.063516912723231e-07, + "loss": 1.0591, + "step": 32484 + }, + { + "epoch": 2.73, + "learning_rate": 4.060952472488044e-07, + "loss": 1.2101, + "step": 32485 + }, + { + "epoch": 2.74, + "learning_rate": 4.05838882493198e-07, + "loss": 1.0556, + "step": 32486 + }, + { + "epoch": 2.74, + "learning_rate": 4.0558259700762103e-07, + "loss": 1.1311, + "step": 32487 + }, + { + "epoch": 2.74, + "learning_rate": 4.053263907941918e-07, + "loss": 1.2129, + "step": 32488 + }, + { + "epoch": 2.74, + "learning_rate": 4.0507026385502747e-07, + "loss": 1.1324, + "step": 32489 + }, + { + "epoch": 2.74, + "learning_rate": 4.0481421619224083e-07, + "loss": 1.1298, + "step": 32490 + }, + { + "epoch": 2.74, + "learning_rate": 4.045582478079524e-07, + "loss": 1.2061, + "step": 32491 + }, + { + "epoch": 2.74, + "learning_rate": 4.043023587042727e-07, + "loss": 1.0903, + "step": 32492 + }, + { + "epoch": 2.74, + "learning_rate": 4.0404654888331893e-07, + "loss": 1.0932, + "step": 32493 + }, + { + "epoch": 2.74, + "learning_rate": 4.037908183472039e-07, + "loss": 1.0994, + "step": 32494 + }, + { + "epoch": 2.74, + "learning_rate": 4.035351670980392e-07, + "loss": 1.1562, + "step": 32495 + }, + { + "epoch": 2.74, + "learning_rate": 4.032795951379398e-07, + "loss": 1.0967, + "step": 32496 + }, + { + "epoch": 2.74, + "learning_rate": 4.03024102469014e-07, + "loss": 1.112, + "step": 32497 + }, + { + "epoch": 2.74, + "learning_rate": 4.0276868909337576e-07, + "loss": 1.0772, + "step": 32498 + }, + { + "epoch": 2.74, + "learning_rate": 4.0251335501313326e-07, + "loss": 1.1151, + "step": 32499 + }, + { + "epoch": 2.74, + "learning_rate": 4.0225810023039825e-07, + "loss": 1.0848, + "step": 32500 + }, + { + "epoch": 2.74, + "learning_rate": 4.0200292474727676e-07, + "loss": 1.0765, + "step": 32501 + }, + { + "epoch": 2.74, + "learning_rate": 4.0174782856588155e-07, + "loss": 1.1404, + "step": 32502 + }, + { + "epoch": 2.74, + "learning_rate": 4.0149281168831545e-07, + "loss": 1.0986, + "step": 32503 + }, + { + "epoch": 2.74, + "learning_rate": 4.012378741166878e-07, + "loss": 1.2169, + "step": 32504 + }, + { + "epoch": 2.74, + "learning_rate": 4.009830158531047e-07, + "loss": 1.0869, + "step": 32505 + }, + { + "epoch": 2.74, + "learning_rate": 4.007282368996723e-07, + "loss": 1.121, + "step": 32506 + }, + { + "epoch": 2.74, + "learning_rate": 4.0047353725849556e-07, + "loss": 1.1409, + "step": 32507 + }, + { + "epoch": 2.74, + "learning_rate": 4.002189169316784e-07, + "loss": 1.0028, + "step": 32508 + }, + { + "epoch": 2.74, + "learning_rate": 3.9996437592132566e-07, + "loss": 1.1901, + "step": 32509 + }, + { + "epoch": 2.74, + "learning_rate": 3.997099142295391e-07, + "loss": 1.1426, + "step": 32510 + }, + { + "epoch": 2.74, + "learning_rate": 3.9945553185842257e-07, + "loss": 1.0449, + "step": 32511 + }, + { + "epoch": 2.74, + "learning_rate": 3.992012288100744e-07, + "loss": 1.0271, + "step": 32512 + }, + { + "epoch": 2.74, + "learning_rate": 3.9894700508660176e-07, + "loss": 1.1323, + "step": 32513 + }, + { + "epoch": 2.74, + "learning_rate": 3.986928606900997e-07, + "loss": 1.1189, + "step": 32514 + }, + { + "epoch": 2.74, + "learning_rate": 3.9843879562266984e-07, + "loss": 1.0861, + "step": 32515 + }, + { + "epoch": 2.74, + "learning_rate": 3.9818480988641274e-07, + "loss": 1.108, + "step": 32516 + }, + { + "epoch": 2.74, + "learning_rate": 3.979309034834244e-07, + "loss": 1.012, + "step": 32517 + }, + { + "epoch": 2.74, + "learning_rate": 3.9767707641580443e-07, + "loss": 1.128, + "step": 32518 + }, + { + "epoch": 2.74, + "learning_rate": 3.97423328685651e-07, + "loss": 1.0589, + "step": 32519 + }, + { + "epoch": 2.74, + "learning_rate": 3.97169660295057e-07, + "loss": 1.1105, + "step": 32520 + }, + { + "epoch": 2.74, + "learning_rate": 3.9691607124612064e-07, + "loss": 1.1577, + "step": 32521 + }, + { + "epoch": 2.74, + "learning_rate": 3.966625615409392e-07, + "loss": 1.0326, + "step": 32522 + }, + { + "epoch": 2.74, + "learning_rate": 3.964091311816032e-07, + "loss": 1.0751, + "step": 32523 + }, + { + "epoch": 2.74, + "learning_rate": 3.9615578017020985e-07, + "loss": 1.0575, + "step": 32524 + }, + { + "epoch": 2.74, + "learning_rate": 3.959025085088508e-07, + "loss": 1.1314, + "step": 32525 + }, + { + "epoch": 2.74, + "learning_rate": 3.9564931619961886e-07, + "loss": 1.0693, + "step": 32526 + }, + { + "epoch": 2.74, + "learning_rate": 3.953962032446057e-07, + "loss": 1.0946, + "step": 32527 + }, + { + "epoch": 2.74, + "learning_rate": 3.9514316964590406e-07, + "loss": 1.0118, + "step": 32528 + }, + { + "epoch": 2.74, + "learning_rate": 3.9489021540560114e-07, + "loss": 1.1537, + "step": 32529 + }, + { + "epoch": 2.74, + "learning_rate": 3.946373405257919e-07, + "loss": 0.9908, + "step": 32530 + }, + { + "epoch": 2.74, + "learning_rate": 3.943845450085615e-07, + "loss": 1.1336, + "step": 32531 + }, + { + "epoch": 2.74, + "learning_rate": 3.9413182885600034e-07, + "loss": 1.1639, + "step": 32532 + }, + { + "epoch": 2.74, + "learning_rate": 3.9387919207019675e-07, + "loss": 1.1366, + "step": 32533 + }, + { + "epoch": 2.74, + "learning_rate": 3.936266346532369e-07, + "loss": 1.1368, + "step": 32534 + }, + { + "epoch": 2.74, + "learning_rate": 3.9337415660720803e-07, + "loss": 1.1514, + "step": 32535 + }, + { + "epoch": 2.74, + "learning_rate": 3.931217579341973e-07, + "loss": 1.1106, + "step": 32536 + }, + { + "epoch": 2.74, + "learning_rate": 3.928694386362886e-07, + "loss": 1.107, + "step": 32537 + }, + { + "epoch": 2.74, + "learning_rate": 3.92617198715568e-07, + "loss": 1.1456, + "step": 32538 + }, + { + "epoch": 2.74, + "learning_rate": 3.9236503817411955e-07, + "loss": 1.1111, + "step": 32539 + }, + { + "epoch": 2.74, + "learning_rate": 3.9211295701402365e-07, + "loss": 1.0801, + "step": 32540 + }, + { + "epoch": 2.74, + "learning_rate": 3.918609552373676e-07, + "loss": 1.1363, + "step": 32541 + }, + { + "epoch": 2.74, + "learning_rate": 3.9160903284623073e-07, + "loss": 1.2271, + "step": 32542 + }, + { + "epoch": 2.74, + "learning_rate": 3.9135718984269376e-07, + "loss": 1.1856, + "step": 32543 + }, + { + "epoch": 2.74, + "learning_rate": 3.911054262288405e-07, + "loss": 1.1215, + "step": 32544 + }, + { + "epoch": 2.74, + "learning_rate": 3.908537420067482e-07, + "loss": 1.115, + "step": 32545 + }, + { + "epoch": 2.74, + "learning_rate": 3.9060213717849847e-07, + "loss": 1.0336, + "step": 32546 + }, + { + "epoch": 2.74, + "learning_rate": 3.903506117461686e-07, + "loss": 1.1815, + "step": 32547 + }, + { + "epoch": 2.74, + "learning_rate": 3.9009916571183804e-07, + "loss": 1.1661, + "step": 32548 + }, + { + "epoch": 2.74, + "learning_rate": 3.8984779907758284e-07, + "loss": 1.1604, + "step": 32549 + }, + { + "epoch": 2.74, + "learning_rate": 3.8959651184548366e-07, + "loss": 1.0672, + "step": 32550 + }, + { + "epoch": 2.74, + "learning_rate": 3.8934530401760986e-07, + "loss": 1.0363, + "step": 32551 + }, + { + "epoch": 2.74, + "learning_rate": 3.890941755960442e-07, + "loss": 1.1656, + "step": 32552 + }, + { + "epoch": 2.74, + "learning_rate": 3.888431265828563e-07, + "loss": 1.1608, + "step": 32553 + }, + { + "epoch": 2.74, + "learning_rate": 3.885921569801232e-07, + "loss": 1.0603, + "step": 32554 + }, + { + "epoch": 2.74, + "learning_rate": 3.883412667899178e-07, + "loss": 1.0722, + "step": 32555 + }, + { + "epoch": 2.74, + "learning_rate": 3.8809045601431283e-07, + "loss": 1.1419, + "step": 32556 + }, + { + "epoch": 2.74, + "learning_rate": 3.8783972465538e-07, + "loss": 1.0804, + "step": 32557 + }, + { + "epoch": 2.74, + "learning_rate": 3.8758907271519207e-07, + "loss": 1.1246, + "step": 32558 + }, + { + "epoch": 2.74, + "learning_rate": 3.8733850019581966e-07, + "loss": 1.0691, + "step": 32559 + }, + { + "epoch": 2.74, + "learning_rate": 3.870880070993333e-07, + "loss": 1.1578, + "step": 32560 + }, + { + "epoch": 2.74, + "learning_rate": 3.8683759342780346e-07, + "loss": 1.1406, + "step": 32561 + }, + { + "epoch": 2.74, + "learning_rate": 3.865872591832953e-07, + "loss": 1.1802, + "step": 32562 + }, + { + "epoch": 2.74, + "learning_rate": 3.863370043678827e-07, + "loss": 1.1173, + "step": 32563 + }, + { + "epoch": 2.74, + "learning_rate": 3.8608682898362947e-07, + "loss": 1.1659, + "step": 32564 + }, + { + "epoch": 2.74, + "learning_rate": 3.8583673303260296e-07, + "loss": 1.0762, + "step": 32565 + }, + { + "epoch": 2.74, + "learning_rate": 3.855867165168714e-07, + "loss": 1.1405, + "step": 32566 + }, + { + "epoch": 2.74, + "learning_rate": 3.853367794384988e-07, + "loss": 1.1938, + "step": 32567 + }, + { + "epoch": 2.74, + "learning_rate": 3.850869217995501e-07, + "loss": 1.1487, + "step": 32568 + }, + { + "epoch": 2.74, + "learning_rate": 3.848371436020926e-07, + "loss": 1.0809, + "step": 32569 + }, + { + "epoch": 2.74, + "learning_rate": 3.8458744484818457e-07, + "loss": 1.1904, + "step": 32570 + }, + { + "epoch": 2.74, + "learning_rate": 3.843378255398944e-07, + "loss": 1.092, + "step": 32571 + }, + { + "epoch": 2.74, + "learning_rate": 3.840882856792838e-07, + "loss": 1.0631, + "step": 32572 + }, + { + "epoch": 2.74, + "learning_rate": 3.8383882526840997e-07, + "loss": 1.1467, + "step": 32573 + }, + { + "epoch": 2.74, + "learning_rate": 3.835894443093402e-07, + "loss": 1.0693, + "step": 32574 + }, + { + "epoch": 2.74, + "learning_rate": 3.8334014280413166e-07, + "loss": 1.0396, + "step": 32575 + }, + { + "epoch": 2.74, + "learning_rate": 3.8309092075484387e-07, + "loss": 1.0883, + "step": 32576 + }, + { + "epoch": 2.74, + "learning_rate": 3.828417781635363e-07, + "loss": 1.1114, + "step": 32577 + }, + { + "epoch": 2.74, + "learning_rate": 3.8259271503226946e-07, + "loss": 1.0434, + "step": 32578 + }, + { + "epoch": 2.74, + "learning_rate": 3.8234373136309735e-07, + "loss": 1.0796, + "step": 32579 + }, + { + "epoch": 2.74, + "learning_rate": 3.8209482715808154e-07, + "loss": 1.131, + "step": 32580 + }, + { + "epoch": 2.74, + "learning_rate": 3.8184600241927495e-07, + "loss": 1.2174, + "step": 32581 + }, + { + "epoch": 2.74, + "learning_rate": 3.8159725714873584e-07, + "loss": 1.1231, + "step": 32582 + }, + { + "epoch": 2.74, + "learning_rate": 3.813485913485171e-07, + "loss": 1.113, + "step": 32583 + }, + { + "epoch": 2.74, + "learning_rate": 3.811000050206759e-07, + "loss": 1.0676, + "step": 32584 + }, + { + "epoch": 2.74, + "learning_rate": 3.8085149816726397e-07, + "loss": 1.1422, + "step": 32585 + }, + { + "epoch": 2.74, + "learning_rate": 3.806030707903363e-07, + "loss": 1.1469, + "step": 32586 + }, + { + "epoch": 2.74, + "learning_rate": 3.8035472289194466e-07, + "loss": 1.1426, + "step": 32587 + }, + { + "epoch": 2.74, + "learning_rate": 3.8010645447414065e-07, + "loss": 1.0951, + "step": 32588 + }, + { + "epoch": 2.74, + "learning_rate": 3.79858265538976e-07, + "loss": 1.175, + "step": 32589 + }, + { + "epoch": 2.74, + "learning_rate": 3.7961015608850015e-07, + "loss": 1.0823, + "step": 32590 + }, + { + "epoch": 2.74, + "learning_rate": 3.7936212612476595e-07, + "loss": 1.1057, + "step": 32591 + }, + { + "epoch": 2.74, + "learning_rate": 3.7911417564982066e-07, + "loss": 1.0484, + "step": 32592 + }, + { + "epoch": 2.74, + "learning_rate": 3.788663046657115e-07, + "loss": 1.1163, + "step": 32593 + }, + { + "epoch": 2.74, + "learning_rate": 3.7861851317448904e-07, + "loss": 1.0783, + "step": 32594 + }, + { + "epoch": 2.74, + "learning_rate": 3.7837080117819944e-07, + "loss": 1.0993, + "step": 32595 + }, + { + "epoch": 2.74, + "learning_rate": 3.7812316867889e-07, + "loss": 1.193, + "step": 32596 + }, + { + "epoch": 2.74, + "learning_rate": 3.7787561567860677e-07, + "loss": 1.0798, + "step": 32597 + }, + { + "epoch": 2.74, + "learning_rate": 3.776281421793937e-07, + "loss": 1.0149, + "step": 32598 + }, + { + "epoch": 2.74, + "learning_rate": 3.7738074818329695e-07, + "loss": 1.1594, + "step": 32599 + }, + { + "epoch": 2.74, + "learning_rate": 3.7713343369236044e-07, + "loss": 1.1022, + "step": 32600 + }, + { + "epoch": 2.74, + "learning_rate": 3.7688619870862587e-07, + "loss": 1.0798, + "step": 32601 + }, + { + "epoch": 2.74, + "learning_rate": 3.766390432341394e-07, + "loss": 1.0957, + "step": 32602 + }, + { + "epoch": 2.74, + "learning_rate": 3.763919672709393e-07, + "loss": 1.0805, + "step": 32603 + }, + { + "epoch": 2.74, + "learning_rate": 3.761449708210696e-07, + "loss": 1.1153, + "step": 32604 + }, + { + "epoch": 2.75, + "learning_rate": 3.7589805388656975e-07, + "loss": 1.0595, + "step": 32605 + }, + { + "epoch": 2.75, + "learning_rate": 3.756512164694803e-07, + "loss": 1.2198, + "step": 32606 + }, + { + "epoch": 2.75, + "learning_rate": 3.7540445857183974e-07, + "loss": 1.0658, + "step": 32607 + }, + { + "epoch": 2.75, + "learning_rate": 3.751577801956896e-07, + "loss": 1.1576, + "step": 32608 + }, + { + "epoch": 2.75, + "learning_rate": 3.7491118134306503e-07, + "loss": 1.0585, + "step": 32609 + }, + { + "epoch": 2.75, + "learning_rate": 3.746646620160044e-07, + "loss": 1.1555, + "step": 32610 + }, + { + "epoch": 2.75, + "learning_rate": 3.744182222165471e-07, + "loss": 1.1094, + "step": 32611 + }, + { + "epoch": 2.75, + "learning_rate": 3.7417186194672385e-07, + "loss": 1.1098, + "step": 32612 + }, + { + "epoch": 2.75, + "learning_rate": 3.739255812085762e-07, + "loss": 1.1126, + "step": 32613 + }, + { + "epoch": 2.75, + "learning_rate": 3.7367938000413497e-07, + "loss": 1.111, + "step": 32614 + }, + { + "epoch": 2.75, + "learning_rate": 3.73433258335435e-07, + "loss": 1.0742, + "step": 32615 + }, + { + "epoch": 2.75, + "learning_rate": 3.731872162045114e-07, + "loss": 1.1574, + "step": 32616 + }, + { + "epoch": 2.75, + "learning_rate": 3.7294125361339473e-07, + "loss": 1.1715, + "step": 32617 + }, + { + "epoch": 2.75, + "learning_rate": 3.7269537056411895e-07, + "loss": 1.1555, + "step": 32618 + }, + { + "epoch": 2.75, + "learning_rate": 3.7244956705871695e-07, + "loss": 1.1488, + "step": 32619 + }, + { + "epoch": 2.75, + "learning_rate": 3.7220384309921474e-07, + "loss": 1.095, + "step": 32620 + }, + { + "epoch": 2.75, + "learning_rate": 3.719581986876475e-07, + "loss": 1.085, + "step": 32621 + }, + { + "epoch": 2.75, + "learning_rate": 3.717126338260435e-07, + "loss": 1.1142, + "step": 32622 + }, + { + "epoch": 2.75, + "learning_rate": 3.7146714851642895e-07, + "loss": 1.0618, + "step": 32623 + }, + { + "epoch": 2.75, + "learning_rate": 3.712217427608367e-07, + "loss": 1.0766, + "step": 32624 + }, + { + "epoch": 2.75, + "learning_rate": 3.709764165612906e-07, + "loss": 1.1304, + "step": 32625 + }, + { + "epoch": 2.75, + "learning_rate": 3.7073116991981906e-07, + "loss": 1.1138, + "step": 32626 + }, + { + "epoch": 2.75, + "learning_rate": 3.7048600283844936e-07, + "loss": 1.2103, + "step": 32627 + }, + { + "epoch": 2.75, + "learning_rate": 3.702409153192066e-07, + "loss": 1.1198, + "step": 32628 + }, + { + "epoch": 2.75, + "learning_rate": 3.699959073641124e-07, + "loss": 0.9897, + "step": 32629 + }, + { + "epoch": 2.75, + "learning_rate": 3.697509789751974e-07, + "loss": 1.1805, + "step": 32630 + }, + { + "epoch": 2.75, + "learning_rate": 3.6950613015448e-07, + "loss": 1.1775, + "step": 32631 + }, + { + "epoch": 2.75, + "learning_rate": 3.6926136090398524e-07, + "loss": 1.1711, + "step": 32632 + }, + { + "epoch": 2.75, + "learning_rate": 3.690166712257359e-07, + "loss": 1.0461, + "step": 32633 + }, + { + "epoch": 2.75, + "learning_rate": 3.6877206112175267e-07, + "loss": 1.103, + "step": 32634 + }, + { + "epoch": 2.75, + "learning_rate": 3.685275305940572e-07, + "loss": 1.1298, + "step": 32635 + }, + { + "epoch": 2.75, + "learning_rate": 3.68283079644669e-07, + "loss": 1.1102, + "step": 32636 + }, + { + "epoch": 2.75, + "learning_rate": 3.680387082756087e-07, + "loss": 1.1225, + "step": 32637 + }, + { + "epoch": 2.75, + "learning_rate": 3.677944164888958e-07, + "loss": 1.0884, + "step": 32638 + }, + { + "epoch": 2.75, + "learning_rate": 3.6755020428654974e-07, + "loss": 1.0699, + "step": 32639 + }, + { + "epoch": 2.75, + "learning_rate": 3.673060716705834e-07, + "loss": 1.0661, + "step": 32640 + }, + { + "epoch": 2.75, + "learning_rate": 3.670620186430196e-07, + "loss": 1.1002, + "step": 32641 + }, + { + "epoch": 2.75, + "learning_rate": 3.6681804520587226e-07, + "loss": 1.0505, + "step": 32642 + }, + { + "epoch": 2.75, + "learning_rate": 3.665741513611565e-07, + "loss": 1.1639, + "step": 32643 + }, + { + "epoch": 2.75, + "learning_rate": 3.663303371108873e-07, + "loss": 1.0967, + "step": 32644 + }, + { + "epoch": 2.75, + "learning_rate": 3.660866024570808e-07, + "loss": 0.9952, + "step": 32645 + }, + { + "epoch": 2.75, + "learning_rate": 3.6584294740174995e-07, + "loss": 1.0367, + "step": 32646 + }, + { + "epoch": 2.75, + "learning_rate": 3.6559937194690865e-07, + "loss": 1.1642, + "step": 32647 + }, + { + "epoch": 2.75, + "learning_rate": 3.653558760945675e-07, + "loss": 1.0042, + "step": 32648 + }, + { + "epoch": 2.75, + "learning_rate": 3.651124598467404e-07, + "loss": 1.1212, + "step": 32649 + }, + { + "epoch": 2.75, + "learning_rate": 3.648691232054391e-07, + "loss": 1.1191, + "step": 32650 + }, + { + "epoch": 2.75, + "learning_rate": 3.646258661726698e-07, + "loss": 1.0719, + "step": 32651 + }, + { + "epoch": 2.75, + "learning_rate": 3.643826887504487e-07, + "loss": 1.0325, + "step": 32652 + }, + { + "epoch": 2.75, + "learning_rate": 3.6413959094077967e-07, + "loss": 1.1754, + "step": 32653 + }, + { + "epoch": 2.75, + "learning_rate": 3.6389657274567225e-07, + "loss": 1.1291, + "step": 32654 + }, + { + "epoch": 2.75, + "learning_rate": 3.636536341671371e-07, + "loss": 1.1161, + "step": 32655 + }, + { + "epoch": 2.75, + "learning_rate": 3.6341077520717807e-07, + "loss": 1.0666, + "step": 32656 + }, + { + "epoch": 2.75, + "learning_rate": 3.631679958678047e-07, + "loss": 1.1716, + "step": 32657 + }, + { + "epoch": 2.75, + "learning_rate": 3.62925296151021e-07, + "loss": 1.158, + "step": 32658 + }, + { + "epoch": 2.75, + "learning_rate": 3.6268267605883203e-07, + "loss": 1.1395, + "step": 32659 + }, + { + "epoch": 2.75, + "learning_rate": 3.6244013559324385e-07, + "loss": 1.2123, + "step": 32660 + }, + { + "epoch": 2.75, + "learning_rate": 3.6219767475626054e-07, + "loss": 1.0768, + "step": 32661 + }, + { + "epoch": 2.75, + "learning_rate": 3.619552935498827e-07, + "loss": 1.0926, + "step": 32662 + }, + { + "epoch": 2.75, + "learning_rate": 3.617129919761164e-07, + "loss": 1.1329, + "step": 32663 + }, + { + "epoch": 2.75, + "learning_rate": 3.614707700369613e-07, + "loss": 1.0712, + "step": 32664 + }, + { + "epoch": 2.75, + "learning_rate": 3.61228627734419e-07, + "loss": 1.0743, + "step": 32665 + }, + { + "epoch": 2.75, + "learning_rate": 3.6098656507049134e-07, + "loss": 1.1088, + "step": 32666 + }, + { + "epoch": 2.75, + "learning_rate": 3.6074458204717776e-07, + "loss": 1.1862, + "step": 32667 + }, + { + "epoch": 2.75, + "learning_rate": 3.605026786664767e-07, + "loss": 1.1007, + "step": 32668 + }, + { + "epoch": 2.75, + "learning_rate": 3.6026085493038877e-07, + "loss": 1.1773, + "step": 32669 + }, + { + "epoch": 2.75, + "learning_rate": 3.600191108409079e-07, + "loss": 1.1466, + "step": 32670 + }, + { + "epoch": 2.75, + "learning_rate": 3.597774464000359e-07, + "loss": 1.1324, + "step": 32671 + }, + { + "epoch": 2.75, + "learning_rate": 3.5953586160976994e-07, + "loss": 1.0907, + "step": 32672 + }, + { + "epoch": 2.75, + "learning_rate": 3.5929435647210076e-07, + "loss": 1.1471, + "step": 32673 + }, + { + "epoch": 2.75, + "learning_rate": 3.5905293098903004e-07, + "loss": 1.1917, + "step": 32674 + }, + { + "epoch": 2.75, + "learning_rate": 3.588115851625473e-07, + "loss": 1.0906, + "step": 32675 + }, + { + "epoch": 2.75, + "learning_rate": 3.585703189946499e-07, + "loss": 1.0747, + "step": 32676 + }, + { + "epoch": 2.75, + "learning_rate": 3.583291324873295e-07, + "loss": 1.1213, + "step": 32677 + }, + { + "epoch": 2.75, + "learning_rate": 3.5808802564258003e-07, + "loss": 1.1009, + "step": 32678 + }, + { + "epoch": 2.75, + "learning_rate": 3.578469984623922e-07, + "loss": 1.0637, + "step": 32679 + }, + { + "epoch": 2.75, + "learning_rate": 3.5760605094876e-07, + "loss": 1.1451, + "step": 32680 + }, + { + "epoch": 2.75, + "learning_rate": 3.573651831036706e-07, + "loss": 1.1181, + "step": 32681 + }, + { + "epoch": 2.75, + "learning_rate": 3.571243949291181e-07, + "loss": 1.0666, + "step": 32682 + }, + { + "epoch": 2.75, + "learning_rate": 3.568836864270897e-07, + "loss": 1.11, + "step": 32683 + }, + { + "epoch": 2.75, + "learning_rate": 3.566430575995716e-07, + "loss": 1.1553, + "step": 32684 + }, + { + "epoch": 2.75, + "learning_rate": 3.5640250844855893e-07, + "loss": 1.1176, + "step": 32685 + }, + { + "epoch": 2.75, + "learning_rate": 3.561620389760323e-07, + "loss": 1.0095, + "step": 32686 + }, + { + "epoch": 2.75, + "learning_rate": 3.559216491839823e-07, + "loss": 1.1604, + "step": 32687 + }, + { + "epoch": 2.75, + "learning_rate": 3.5568133907439406e-07, + "loss": 1.055, + "step": 32688 + }, + { + "epoch": 2.75, + "learning_rate": 3.554411086492537e-07, + "loss": 1.0948, + "step": 32689 + }, + { + "epoch": 2.75, + "learning_rate": 3.552009579105442e-07, + "loss": 1.0879, + "step": 32690 + }, + { + "epoch": 2.75, + "learning_rate": 3.549608868602528e-07, + "loss": 1.1299, + "step": 32691 + }, + { + "epoch": 2.75, + "learning_rate": 3.5472089550036005e-07, + "loss": 1.0709, + "step": 32692 + }, + { + "epoch": 2.75, + "learning_rate": 3.5448098383285113e-07, + "loss": 1.1349, + "step": 32693 + }, + { + "epoch": 2.75, + "learning_rate": 3.5424115185970665e-07, + "loss": 1.1763, + "step": 32694 + }, + { + "epoch": 2.75, + "learning_rate": 3.5400139958290834e-07, + "loss": 1.0534, + "step": 32695 + }, + { + "epoch": 2.75, + "learning_rate": 3.53761727004438e-07, + "loss": 1.1426, + "step": 32696 + }, + { + "epoch": 2.75, + "learning_rate": 3.535221341262751e-07, + "loss": 1.0572, + "step": 32697 + }, + { + "epoch": 2.75, + "learning_rate": 3.532826209504003e-07, + "loss": 1.1312, + "step": 32698 + }, + { + "epoch": 2.75, + "learning_rate": 3.5304318747879205e-07, + "loss": 1.1908, + "step": 32699 + }, + { + "epoch": 2.75, + "learning_rate": 3.5280383371342876e-07, + "loss": 1.0086, + "step": 32700 + }, + { + "epoch": 2.75, + "learning_rate": 3.5256455965628546e-07, + "loss": 1.1961, + "step": 32701 + }, + { + "epoch": 2.75, + "learning_rate": 3.523253653093439e-07, + "loss": 1.1393, + "step": 32702 + }, + { + "epoch": 2.75, + "learning_rate": 3.52086250674577e-07, + "loss": 1.1126, + "step": 32703 + }, + { + "epoch": 2.75, + "learning_rate": 3.51847215753961e-07, + "loss": 1.1604, + "step": 32704 + }, + { + "epoch": 2.75, + "learning_rate": 3.5160826054947085e-07, + "loss": 1.1025, + "step": 32705 + }, + { + "epoch": 2.75, + "learning_rate": 3.513693850630806e-07, + "loss": 1.0766, + "step": 32706 + }, + { + "epoch": 2.75, + "learning_rate": 3.5113058929676537e-07, + "loss": 1.0923, + "step": 32707 + }, + { + "epoch": 2.75, + "learning_rate": 3.508918732524969e-07, + "loss": 1.1668, + "step": 32708 + }, + { + "epoch": 2.75, + "learning_rate": 3.50653236932248e-07, + "loss": 1.1219, + "step": 32709 + }, + { + "epoch": 2.75, + "learning_rate": 3.504146803379904e-07, + "loss": 1.1913, + "step": 32710 + }, + { + "epoch": 2.75, + "learning_rate": 3.50176203471696e-07, + "loss": 1.1597, + "step": 32711 + }, + { + "epoch": 2.75, + "learning_rate": 3.499378063353309e-07, + "loss": 1.1536, + "step": 32712 + }, + { + "epoch": 2.75, + "learning_rate": 3.496994889308714e-07, + "loss": 1.0889, + "step": 32713 + }, + { + "epoch": 2.75, + "learning_rate": 3.494612512602824e-07, + "loss": 1.0619, + "step": 32714 + }, + { + "epoch": 2.75, + "learning_rate": 3.492230933255325e-07, + "loss": 1.1392, + "step": 32715 + }, + { + "epoch": 2.75, + "learning_rate": 3.4898501512859006e-07, + "loss": 1.1594, + "step": 32716 + }, + { + "epoch": 2.75, + "learning_rate": 3.487470166714224e-07, + "loss": 1.1173, + "step": 32717 + }, + { + "epoch": 2.75, + "learning_rate": 3.485090979559946e-07, + "loss": 1.1057, + "step": 32718 + }, + { + "epoch": 2.75, + "learning_rate": 3.4827125898427626e-07, + "loss": 1.0664, + "step": 32719 + }, + { + "epoch": 2.75, + "learning_rate": 3.4803349975822685e-07, + "loss": 1.0744, + "step": 32720 + }, + { + "epoch": 2.75, + "learning_rate": 3.4779582027981374e-07, + "loss": 1.0312, + "step": 32721 + }, + { + "epoch": 2.75, + "learning_rate": 3.4755822055100306e-07, + "loss": 1.0596, + "step": 32722 + }, + { + "epoch": 2.76, + "learning_rate": 3.4732070057375334e-07, + "loss": 1.0952, + "step": 32723 + }, + { + "epoch": 2.76, + "learning_rate": 3.4708326035003073e-07, + "loss": 1.0969, + "step": 32724 + }, + { + "epoch": 2.76, + "learning_rate": 3.468458998817947e-07, + "loss": 1.1277, + "step": 32725 + }, + { + "epoch": 2.76, + "learning_rate": 3.466086191710072e-07, + "loss": 1.1006, + "step": 32726 + }, + { + "epoch": 2.76, + "learning_rate": 3.463714182196298e-07, + "loss": 1.1036, + "step": 32727 + }, + { + "epoch": 2.76, + "learning_rate": 3.4613429702961996e-07, + "loss": 1.0887, + "step": 32728 + }, + { + "epoch": 2.76, + "learning_rate": 3.4589725560293943e-07, + "loss": 1.0593, + "step": 32729 + }, + { + "epoch": 2.76, + "learning_rate": 3.4566029394154546e-07, + "loss": 1.2077, + "step": 32730 + }, + { + "epoch": 2.76, + "learning_rate": 3.4542341204739427e-07, + "loss": 1.2356, + "step": 32731 + }, + { + "epoch": 2.76, + "learning_rate": 3.451866099224466e-07, + "loss": 1.1008, + "step": 32732 + }, + { + "epoch": 2.76, + "learning_rate": 3.4494988756865635e-07, + "loss": 1.0581, + "step": 32733 + }, + { + "epoch": 2.76, + "learning_rate": 3.4471324498797863e-07, + "loss": 1.1828, + "step": 32734 + }, + { + "epoch": 2.76, + "learning_rate": 3.4447668218237197e-07, + "loss": 1.0853, + "step": 32735 + }, + { + "epoch": 2.76, + "learning_rate": 3.4424019915378915e-07, + "loss": 1.1113, + "step": 32736 + }, + { + "epoch": 2.76, + "learning_rate": 3.4400379590418307e-07, + "loss": 1.1036, + "step": 32737 + }, + { + "epoch": 2.76, + "learning_rate": 3.437674724355089e-07, + "loss": 1.0211, + "step": 32738 + }, + { + "epoch": 2.76, + "learning_rate": 3.4353122874971835e-07, + "loss": 1.0687, + "step": 32739 + }, + { + "epoch": 2.76, + "learning_rate": 3.4329506484876206e-07, + "loss": 1.0815, + "step": 32740 + }, + { + "epoch": 2.76, + "learning_rate": 3.430589807345941e-07, + "loss": 1.0503, + "step": 32741 + }, + { + "epoch": 2.76, + "learning_rate": 3.4282297640916286e-07, + "loss": 1.0152, + "step": 32742 + }, + { + "epoch": 2.76, + "learning_rate": 3.425870518744179e-07, + "loss": 1.0888, + "step": 32743 + }, + { + "epoch": 2.76, + "learning_rate": 3.4235120713230986e-07, + "loss": 1.1665, + "step": 32744 + }, + { + "epoch": 2.76, + "learning_rate": 3.421154421847872e-07, + "loss": 1.1755, + "step": 32745 + }, + { + "epoch": 2.76, + "learning_rate": 3.4187975703379726e-07, + "loss": 1.1372, + "step": 32746 + }, + { + "epoch": 2.76, + "learning_rate": 3.416441516812885e-07, + "loss": 1.1066, + "step": 32747 + }, + { + "epoch": 2.76, + "learning_rate": 3.414086261292071e-07, + "loss": 1.153, + "step": 32748 + }, + { + "epoch": 2.76, + "learning_rate": 3.411731803794982e-07, + "loss": 1.1158, + "step": 32749 + }, + { + "epoch": 2.76, + "learning_rate": 3.4093781443410803e-07, + "loss": 1.1583, + "step": 32750 + }, + { + "epoch": 2.76, + "learning_rate": 3.407025282949794e-07, + "loss": 1.0938, + "step": 32751 + }, + { + "epoch": 2.76, + "learning_rate": 3.4046732196406087e-07, + "loss": 1.1015, + "step": 32752 + }, + { + "epoch": 2.76, + "learning_rate": 3.4023219544329077e-07, + "loss": 1.123, + "step": 32753 + }, + { + "epoch": 2.76, + "learning_rate": 3.399971487346132e-07, + "loss": 1.0299, + "step": 32754 + }, + { + "epoch": 2.76, + "learning_rate": 3.397621818399721e-07, + "loss": 1.1455, + "step": 32755 + }, + { + "epoch": 2.76, + "learning_rate": 3.3952729476130707e-07, + "loss": 1.1619, + "step": 32756 + }, + { + "epoch": 2.76, + "learning_rate": 3.3929248750055876e-07, + "loss": 0.9638, + "step": 32757 + }, + { + "epoch": 2.76, + "learning_rate": 3.3905776005966785e-07, + "loss": 1.1403, + "step": 32758 + }, + { + "epoch": 2.76, + "learning_rate": 3.38823112440575e-07, + "loss": 1.0322, + "step": 32759 + }, + { + "epoch": 2.76, + "learning_rate": 3.385885446452164e-07, + "loss": 1.0599, + "step": 32760 + }, + { + "epoch": 2.76, + "learning_rate": 3.3835405667553165e-07, + "loss": 1.0384, + "step": 32761 + }, + { + "epoch": 2.76, + "learning_rate": 3.38119648533457e-07, + "loss": 1.0798, + "step": 32762 + }, + { + "epoch": 2.76, + "learning_rate": 3.378853202209309e-07, + "loss": 1.1918, + "step": 32763 + }, + { + "epoch": 2.76, + "learning_rate": 3.3765107173988844e-07, + "loss": 1.2404, + "step": 32764 + }, + { + "epoch": 2.76, + "learning_rate": 3.374169030922647e-07, + "loss": 1.1078, + "step": 32765 + }, + { + "epoch": 2.76, + "learning_rate": 3.37182814279996e-07, + "loss": 1.0484, + "step": 32766 + }, + { + "epoch": 2.76, + "learning_rate": 3.3694880530501404e-07, + "loss": 1.0834, + "step": 32767 + }, + { + "epoch": 2.76, + "learning_rate": 3.36714876169254e-07, + "loss": 1.1287, + "step": 32768 + }, + { + "epoch": 2.76, + "learning_rate": 3.364810268746499e-07, + "loss": 1.1, + "step": 32769 + }, + { + "epoch": 2.76, + "learning_rate": 3.3624725742313013e-07, + "loss": 1.1246, + "step": 32770 + }, + { + "epoch": 2.76, + "learning_rate": 3.360135678166287e-07, + "loss": 1.079, + "step": 32771 + }, + { + "epoch": 2.76, + "learning_rate": 3.3577995805707754e-07, + "loss": 1.0633, + "step": 32772 + }, + { + "epoch": 2.76, + "learning_rate": 3.355464281464027e-07, + "loss": 1.1135, + "step": 32773 + }, + { + "epoch": 2.76, + "learning_rate": 3.353129780865394e-07, + "loss": 1.0548, + "step": 32774 + }, + { + "epoch": 2.76, + "learning_rate": 3.3507960787941164e-07, + "loss": 1.1471, + "step": 32775 + }, + { + "epoch": 2.76, + "learning_rate": 3.34846317526949e-07, + "loss": 1.0865, + "step": 32776 + }, + { + "epoch": 2.76, + "learning_rate": 3.3461310703107986e-07, + "loss": 1.1291, + "step": 32777 + }, + { + "epoch": 2.76, + "learning_rate": 3.3437997639373053e-07, + "loss": 1.0971, + "step": 32778 + }, + { + "epoch": 2.76, + "learning_rate": 3.3414692561682615e-07, + "loss": 1.1035, + "step": 32779 + }, + { + "epoch": 2.76, + "learning_rate": 3.3391395470229625e-07, + "loss": 1.0653, + "step": 32780 + }, + { + "epoch": 2.76, + "learning_rate": 3.336810636520593e-07, + "loss": 1.1371, + "step": 32781 + }, + { + "epoch": 2.76, + "learning_rate": 3.33448252468046e-07, + "loss": 1.2595, + "step": 32782 + }, + { + "epoch": 2.76, + "learning_rate": 3.3321552115217484e-07, + "loss": 1.1234, + "step": 32783 + }, + { + "epoch": 2.76, + "learning_rate": 3.3298286970637083e-07, + "loss": 1.0798, + "step": 32784 + }, + { + "epoch": 2.76, + "learning_rate": 3.327502981325581e-07, + "loss": 1.0459, + "step": 32785 + }, + { + "epoch": 2.76, + "learning_rate": 3.3251780643265506e-07, + "loss": 1.0945, + "step": 32786 + }, + { + "epoch": 2.76, + "learning_rate": 3.322853946085836e-07, + "loss": 1.206, + "step": 32787 + }, + { + "epoch": 2.76, + "learning_rate": 3.320530626622642e-07, + "loss": 1.114, + "step": 32788 + }, + { + "epoch": 2.76, + "learning_rate": 3.318208105956189e-07, + "loss": 1.0916, + "step": 32789 + }, + { + "epoch": 2.76, + "learning_rate": 3.315886384105604e-07, + "loss": 1.1421, + "step": 32790 + }, + { + "epoch": 2.76, + "learning_rate": 3.3135654610901403e-07, + "loss": 1.1461, + "step": 32791 + }, + { + "epoch": 2.76, + "learning_rate": 3.3112453369289253e-07, + "loss": 1.1274, + "step": 32792 + }, + { + "epoch": 2.76, + "learning_rate": 3.308926011641156e-07, + "loss": 1.1387, + "step": 32793 + }, + { + "epoch": 2.76, + "learning_rate": 3.306607485245972e-07, + "loss": 1.1438, + "step": 32794 + }, + { + "epoch": 2.76, + "learning_rate": 3.304289757762558e-07, + "loss": 1.0728, + "step": 32795 + }, + { + "epoch": 2.76, + "learning_rate": 3.301972829210032e-07, + "loss": 1.1364, + "step": 32796 + }, + { + "epoch": 2.76, + "learning_rate": 3.2996566996075675e-07, + "loss": 1.0958, + "step": 32797 + }, + { + "epoch": 2.76, + "learning_rate": 3.297341368974283e-07, + "loss": 1.1132, + "step": 32798 + }, + { + "epoch": 2.76, + "learning_rate": 3.295026837329307e-07, + "loss": 1.0875, + "step": 32799 + }, + { + "epoch": 2.76, + "learning_rate": 3.2927131046917914e-07, + "loss": 1.0512, + "step": 32800 + }, + { + "epoch": 2.76, + "learning_rate": 3.2904001710807986e-07, + "loss": 1.1383, + "step": 32801 + }, + { + "epoch": 2.76, + "learning_rate": 3.288088036515491e-07, + "loss": 1.0742, + "step": 32802 + }, + { + "epoch": 2.76, + "learning_rate": 3.285776701014953e-07, + "loss": 1.2628, + "step": 32803 + }, + { + "epoch": 2.76, + "learning_rate": 3.2834661645982703e-07, + "loss": 1.0848, + "step": 32804 + }, + { + "epoch": 2.76, + "learning_rate": 3.2811564272845487e-07, + "loss": 1.1068, + "step": 32805 + }, + { + "epoch": 2.76, + "learning_rate": 3.2788474890928623e-07, + "loss": 1.169, + "step": 32806 + }, + { + "epoch": 2.76, + "learning_rate": 3.276539350042296e-07, + "loss": 1.1084, + "step": 32807 + }, + { + "epoch": 2.76, + "learning_rate": 3.274232010151912e-07, + "loss": 1.064, + "step": 32808 + }, + { + "epoch": 2.76, + "learning_rate": 3.2719254694407844e-07, + "loss": 1.1406, + "step": 32809 + }, + { + "epoch": 2.76, + "learning_rate": 3.269619727927964e-07, + "loss": 1.0969, + "step": 32810 + }, + { + "epoch": 2.76, + "learning_rate": 3.2673147856325137e-07, + "loss": 1.0982, + "step": 32811 + }, + { + "epoch": 2.76, + "learning_rate": 3.265010642573441e-07, + "loss": 1.1699, + "step": 32812 + }, + { + "epoch": 2.76, + "learning_rate": 3.2627072987698405e-07, + "loss": 1.078, + "step": 32813 + }, + { + "epoch": 2.76, + "learning_rate": 3.260404754240698e-07, + "loss": 1.1656, + "step": 32814 + }, + { + "epoch": 2.76, + "learning_rate": 3.2581030090050424e-07, + "loss": 1.143, + "step": 32815 + }, + { + "epoch": 2.76, + "learning_rate": 3.2558020630819144e-07, + "loss": 1.0878, + "step": 32816 + }, + { + "epoch": 2.76, + "learning_rate": 3.2535019164903094e-07, + "loss": 1.0987, + "step": 32817 + }, + { + "epoch": 2.76, + "learning_rate": 3.251202569249234e-07, + "loss": 1.1132, + "step": 32818 + }, + { + "epoch": 2.76, + "learning_rate": 3.2489040213776855e-07, + "loss": 1.0886, + "step": 32819 + }, + { + "epoch": 2.76, + "learning_rate": 3.2466062728946477e-07, + "loss": 1.0754, + "step": 32820 + }, + { + "epoch": 2.76, + "learning_rate": 3.244309323819128e-07, + "loss": 1.2051, + "step": 32821 + }, + { + "epoch": 2.76, + "learning_rate": 3.242013174170089e-07, + "loss": 1.0818, + "step": 32822 + }, + { + "epoch": 2.76, + "learning_rate": 3.239717823966493e-07, + "loss": 1.0584, + "step": 32823 + }, + { + "epoch": 2.76, + "learning_rate": 3.2374232732273245e-07, + "loss": 1.1217, + "step": 32824 + }, + { + "epoch": 2.76, + "learning_rate": 3.235129521971536e-07, + "loss": 1.1114, + "step": 32825 + }, + { + "epoch": 2.76, + "learning_rate": 3.2328365702180676e-07, + "loss": 1.1202, + "step": 32826 + }, + { + "epoch": 2.76, + "learning_rate": 3.2305444179858704e-07, + "loss": 1.1127, + "step": 32827 + }, + { + "epoch": 2.76, + "learning_rate": 3.228253065293896e-07, + "loss": 1.1571, + "step": 32828 + }, + { + "epoch": 2.76, + "learning_rate": 3.2259625121610515e-07, + "loss": 1.1906, + "step": 32829 + }, + { + "epoch": 2.76, + "learning_rate": 3.2236727586063e-07, + "loss": 1.1752, + "step": 32830 + }, + { + "epoch": 2.76, + "learning_rate": 3.2213838046485147e-07, + "loss": 1.0735, + "step": 32831 + }, + { + "epoch": 2.76, + "learning_rate": 3.219095650306647e-07, + "loss": 1.102, + "step": 32832 + }, + { + "epoch": 2.76, + "learning_rate": 3.2168082955995717e-07, + "loss": 1.1618, + "step": 32833 + }, + { + "epoch": 2.76, + "learning_rate": 3.214521740546195e-07, + "loss": 1.1085, + "step": 32834 + }, + { + "epoch": 2.76, + "learning_rate": 3.212235985165435e-07, + "loss": 1.0688, + "step": 32835 + }, + { + "epoch": 2.76, + "learning_rate": 3.209951029476144e-07, + "loss": 1.0738, + "step": 32836 + }, + { + "epoch": 2.76, + "learning_rate": 3.207666873497217e-07, + "loss": 1.1778, + "step": 32837 + }, + { + "epoch": 2.76, + "learning_rate": 3.2053835172475176e-07, + "loss": 1.1835, + "step": 32838 + }, + { + "epoch": 2.76, + "learning_rate": 3.203100960745931e-07, + "loss": 1.153, + "step": 32839 + }, + { + "epoch": 2.76, + "learning_rate": 3.200819204011285e-07, + "loss": 1.0747, + "step": 32840 + }, + { + "epoch": 2.76, + "learning_rate": 3.198538247062477e-07, + "loss": 1.1867, + "step": 32841 + }, + { + "epoch": 2.77, + "learning_rate": 3.1962580899183026e-07, + "loss": 1.1772, + "step": 32842 + }, + { + "epoch": 2.77, + "learning_rate": 3.193978732597636e-07, + "loss": 1.1044, + "step": 32843 + }, + { + "epoch": 2.77, + "learning_rate": 3.1917001751192943e-07, + "loss": 1.1074, + "step": 32844 + }, + { + "epoch": 2.77, + "learning_rate": 3.1894224175021083e-07, + "loss": 1.1342, + "step": 32845 + }, + { + "epoch": 2.77, + "learning_rate": 3.1871454597649065e-07, + "loss": 1.1549, + "step": 32846 + }, + { + "epoch": 2.77, + "learning_rate": 3.1848693019264856e-07, + "loss": 1.1986, + "step": 32847 + }, + { + "epoch": 2.77, + "learning_rate": 3.182593944005652e-07, + "loss": 1.1769, + "step": 32848 + }, + { + "epoch": 2.77, + "learning_rate": 3.1803193860212246e-07, + "loss": 1.1002, + "step": 32849 + }, + { + "epoch": 2.77, + "learning_rate": 3.178045627991988e-07, + "loss": 1.1143, + "step": 32850 + }, + { + "epoch": 2.77, + "learning_rate": 3.1757726699367055e-07, + "loss": 1.081, + "step": 32851 + }, + { + "epoch": 2.77, + "learning_rate": 3.173500511874206e-07, + "loss": 1.1576, + "step": 32852 + }, + { + "epoch": 2.77, + "learning_rate": 3.1712291538232077e-07, + "loss": 1.0797, + "step": 32853 + }, + { + "epoch": 2.77, + "learning_rate": 3.168958595802518e-07, + "loss": 1.0775, + "step": 32854 + }, + { + "epoch": 2.77, + "learning_rate": 3.166688837830878e-07, + "loss": 1.0957, + "step": 32855 + }, + { + "epoch": 2.77, + "learning_rate": 3.164419879927039e-07, + "loss": 1.0555, + "step": 32856 + }, + { + "epoch": 2.77, + "learning_rate": 3.162151722109763e-07, + "loss": 1.0272, + "step": 32857 + }, + { + "epoch": 2.77, + "learning_rate": 3.1598843643977697e-07, + "loss": 1.1387, + "step": 32858 + }, + { + "epoch": 2.77, + "learning_rate": 3.1576178068098205e-07, + "loss": 1.0862, + "step": 32859 + }, + { + "epoch": 2.77, + "learning_rate": 3.155352049364613e-07, + "loss": 1.1419, + "step": 32860 + }, + { + "epoch": 2.77, + "learning_rate": 3.1530870920808975e-07, + "loss": 1.1208, + "step": 32861 + }, + { + "epoch": 2.77, + "learning_rate": 3.1508229349773487e-07, + "loss": 1.0855, + "step": 32862 + }, + { + "epoch": 2.77, + "learning_rate": 3.1485595780727184e-07, + "loss": 1.1086, + "step": 32863 + }, + { + "epoch": 2.77, + "learning_rate": 3.14629702138568e-07, + "loss": 1.1153, + "step": 32864 + }, + { + "epoch": 2.77, + "learning_rate": 3.1440352649349303e-07, + "loss": 1.0289, + "step": 32865 + }, + { + "epoch": 2.77, + "learning_rate": 3.141774308739154e-07, + "loss": 1.0812, + "step": 32866 + }, + { + "epoch": 2.77, + "learning_rate": 3.139514152817036e-07, + "loss": 1.1114, + "step": 32867 + }, + { + "epoch": 2.77, + "learning_rate": 3.1372547971872615e-07, + "loss": 1.2022, + "step": 32868 + }, + { + "epoch": 2.77, + "learning_rate": 3.1349962418684933e-07, + "loss": 1.0917, + "step": 32869 + }, + { + "epoch": 2.77, + "learning_rate": 3.132738486879361e-07, + "loss": 1.1644, + "step": 32870 + }, + { + "epoch": 2.77, + "learning_rate": 3.130481532238561e-07, + "loss": 1.0519, + "step": 32871 + }, + { + "epoch": 2.77, + "learning_rate": 3.128225377964733e-07, + "loss": 1.1764, + "step": 32872 + }, + { + "epoch": 2.77, + "learning_rate": 3.1259700240764854e-07, + "loss": 1.0737, + "step": 32873 + }, + { + "epoch": 2.77, + "learning_rate": 3.123715470592503e-07, + "loss": 1.1372, + "step": 32874 + }, + { + "epoch": 2.77, + "learning_rate": 3.121461717531382e-07, + "loss": 1.1189, + "step": 32875 + }, + { + "epoch": 2.77, + "learning_rate": 3.119208764911741e-07, + "loss": 1.1036, + "step": 32876 + }, + { + "epoch": 2.77, + "learning_rate": 3.1169566127522086e-07, + "loss": 1.0949, + "step": 32877 + }, + { + "epoch": 2.77, + "learning_rate": 3.1147052610713826e-07, + "loss": 1.144, + "step": 32878 + }, + { + "epoch": 2.77, + "learning_rate": 3.112454709887869e-07, + "loss": 1.2131, + "step": 32879 + }, + { + "epoch": 2.77, + "learning_rate": 3.1102049592202753e-07, + "loss": 1.1507, + "step": 32880 + }, + { + "epoch": 2.77, + "learning_rate": 3.107956009087154e-07, + "loss": 1.1529, + "step": 32881 + }, + { + "epoch": 2.77, + "learning_rate": 3.1057078595071343e-07, + "loss": 1.1528, + "step": 32882 + }, + { + "epoch": 2.77, + "learning_rate": 3.103460510498757e-07, + "loss": 1.0715, + "step": 32883 + }, + { + "epoch": 2.77, + "learning_rate": 3.1012139620805846e-07, + "loss": 1.0799, + "step": 32884 + }, + { + "epoch": 2.77, + "learning_rate": 3.098968214271214e-07, + "loss": 1.1426, + "step": 32885 + }, + { + "epoch": 2.77, + "learning_rate": 3.0967232670891746e-07, + "loss": 1.1183, + "step": 32886 + }, + { + "epoch": 2.77, + "learning_rate": 3.094479120553018e-07, + "loss": 1.0567, + "step": 32887 + }, + { + "epoch": 2.77, + "learning_rate": 3.092235774681285e-07, + "loss": 1.0398, + "step": 32888 + }, + { + "epoch": 2.77, + "learning_rate": 3.0899932294925273e-07, + "loss": 1.0797, + "step": 32889 + }, + { + "epoch": 2.77, + "learning_rate": 3.087751485005241e-07, + "loss": 1.0823, + "step": 32890 + }, + { + "epoch": 2.77, + "learning_rate": 3.0855105412379906e-07, + "loss": 1.1797, + "step": 32891 + }, + { + "epoch": 2.77, + "learning_rate": 3.083270398209259e-07, + "loss": 1.0028, + "step": 32892 + }, + { + "epoch": 2.77, + "learning_rate": 3.0810310559375665e-07, + "loss": 1.0311, + "step": 32893 + }, + { + "epoch": 2.77, + "learning_rate": 3.078792514441409e-07, + "loss": 1.1753, + "step": 32894 + }, + { + "epoch": 2.77, + "learning_rate": 3.076554773739282e-07, + "loss": 1.1774, + "step": 32895 + }, + { + "epoch": 2.77, + "learning_rate": 3.0743178338496827e-07, + "loss": 1.0898, + "step": 32896 + }, + { + "epoch": 2.77, + "learning_rate": 3.0720816947910956e-07, + "loss": 1.0461, + "step": 32897 + }, + { + "epoch": 2.77, + "learning_rate": 3.069846356581985e-07, + "loss": 1.0406, + "step": 32898 + }, + { + "epoch": 2.77, + "learning_rate": 3.067611819240823e-07, + "loss": 1.1165, + "step": 32899 + }, + { + "epoch": 2.77, + "learning_rate": 3.065378082786086e-07, + "loss": 1.1265, + "step": 32900 + }, + { + "epoch": 2.77, + "learning_rate": 3.0631451472362015e-07, + "loss": 1.1656, + "step": 32901 + }, + { + "epoch": 2.77, + "learning_rate": 3.0609130126096563e-07, + "loss": 1.1546, + "step": 32902 + }, + { + "epoch": 2.77, + "learning_rate": 3.0586816789248573e-07, + "loss": 1.132, + "step": 32903 + }, + { + "epoch": 2.77, + "learning_rate": 3.056451146200257e-07, + "loss": 1.1346, + "step": 32904 + }, + { + "epoch": 2.77, + "learning_rate": 3.0542214144542834e-07, + "loss": 1.1268, + "step": 32905 + }, + { + "epoch": 2.77, + "learning_rate": 3.051992483705368e-07, + "loss": 1.0344, + "step": 32906 + }, + { + "epoch": 2.77, + "learning_rate": 3.049764353971907e-07, + "loss": 1.0103, + "step": 32907 + }, + { + "epoch": 2.77, + "learning_rate": 3.047537025272318e-07, + "loss": 1.1075, + "step": 32908 + }, + { + "epoch": 2.77, + "learning_rate": 3.04531049762502e-07, + "loss": 1.1502, + "step": 32909 + }, + { + "epoch": 2.77, + "learning_rate": 3.043084771048388e-07, + "loss": 1.1185, + "step": 32910 + }, + { + "epoch": 2.77, + "learning_rate": 3.0408598455608286e-07, + "loss": 1.1039, + "step": 32911 + }, + { + "epoch": 2.77, + "learning_rate": 3.0386357211806936e-07, + "loss": 1.1048, + "step": 32912 + }, + { + "epoch": 2.77, + "learning_rate": 3.036412397926414e-07, + "loss": 1.1616, + "step": 32913 + }, + { + "epoch": 2.77, + "learning_rate": 3.0341898758163066e-07, + "loss": 1.0733, + "step": 32914 + }, + { + "epoch": 2.77, + "learning_rate": 3.0319681548687586e-07, + "loss": 1.1221, + "step": 32915 + }, + { + "epoch": 2.77, + "learning_rate": 3.0297472351021204e-07, + "loss": 1.0707, + "step": 32916 + }, + { + "epoch": 2.77, + "learning_rate": 3.027527116534756e-07, + "loss": 1.0748, + "step": 32917 + }, + { + "epoch": 2.77, + "learning_rate": 3.0253077991849846e-07, + "loss": 1.1784, + "step": 32918 + }, + { + "epoch": 2.77, + "learning_rate": 3.023089283071179e-07, + "loss": 1.1036, + "step": 32919 + }, + { + "epoch": 2.77, + "learning_rate": 3.020871568211614e-07, + "loss": 1.0479, + "step": 32920 + }, + { + "epoch": 2.77, + "learning_rate": 3.0186546546246644e-07, + "loss": 1.0787, + "step": 32921 + }, + { + "epoch": 2.77, + "learning_rate": 3.016438542328626e-07, + "loss": 0.9925, + "step": 32922 + }, + { + "epoch": 2.77, + "learning_rate": 3.014223231341806e-07, + "loss": 1.1204, + "step": 32923 + }, + { + "epoch": 2.77, + "learning_rate": 3.0120087216825246e-07, + "loss": 1.1168, + "step": 32924 + }, + { + "epoch": 2.77, + "learning_rate": 3.009795013369055e-07, + "loss": 1.2457, + "step": 32925 + }, + { + "epoch": 2.77, + "learning_rate": 3.007582106419704e-07, + "loss": 1.2128, + "step": 32926 + }, + { + "epoch": 2.77, + "learning_rate": 3.0053700008527476e-07, + "loss": 1.092, + "step": 32927 + }, + { + "epoch": 2.77, + "learning_rate": 3.00315869668647e-07, + "loss": 1.0839, + "step": 32928 + }, + { + "epoch": 2.77, + "learning_rate": 3.000948193939146e-07, + "loss": 1.1024, + "step": 32929 + }, + { + "epoch": 2.77, + "learning_rate": 2.9987384926290276e-07, + "loss": 1.095, + "step": 32930 + }, + { + "epoch": 2.77, + "learning_rate": 2.9965295927743664e-07, + "loss": 1.08, + "step": 32931 + }, + { + "epoch": 2.77, + "learning_rate": 2.9943214943934375e-07, + "loss": 1.1335, + "step": 32932 + }, + { + "epoch": 2.77, + "learning_rate": 2.992114197504459e-07, + "loss": 1.1152, + "step": 32933 + }, + { + "epoch": 2.77, + "learning_rate": 2.9899077021256715e-07, + "loss": 1.0315, + "step": 32934 + }, + { + "epoch": 2.77, + "learning_rate": 2.9877020082753394e-07, + "loss": 1.0913, + "step": 32935 + }, + { + "epoch": 2.77, + "learning_rate": 2.9854971159716473e-07, + "loss": 1.0874, + "step": 32936 + }, + { + "epoch": 2.77, + "learning_rate": 2.983293025232825e-07, + "loss": 1.0116, + "step": 32937 + }, + { + "epoch": 2.77, + "learning_rate": 2.9810897360770807e-07, + "loss": 1.1423, + "step": 32938 + }, + { + "epoch": 2.77, + "learning_rate": 2.978887248522644e-07, + "loss": 1.1492, + "step": 32939 + }, + { + "epoch": 2.77, + "learning_rate": 2.9766855625876554e-07, + "loss": 1.143, + "step": 32940 + }, + { + "epoch": 2.77, + "learning_rate": 2.974484678290379e-07, + "loss": 1.0816, + "step": 32941 + }, + { + "epoch": 2.77, + "learning_rate": 2.972284595648944e-07, + "loss": 1.03, + "step": 32942 + }, + { + "epoch": 2.77, + "learning_rate": 2.970085314681548e-07, + "loss": 1.1066, + "step": 32943 + }, + { + "epoch": 2.77, + "learning_rate": 2.967886835406364e-07, + "loss": 1.0611, + "step": 32944 + }, + { + "epoch": 2.77, + "learning_rate": 2.9656891578415446e-07, + "loss": 1.1723, + "step": 32945 + }, + { + "epoch": 2.77, + "learning_rate": 2.9634922820052646e-07, + "loss": 1.0921, + "step": 32946 + }, + { + "epoch": 2.77, + "learning_rate": 2.9612962079156647e-07, + "loss": 1.0653, + "step": 32947 + }, + { + "epoch": 2.77, + "learning_rate": 2.959100935590886e-07, + "loss": 1.1586, + "step": 32948 + }, + { + "epoch": 2.77, + "learning_rate": 2.9569064650490806e-07, + "loss": 1.1498, + "step": 32949 + }, + { + "epoch": 2.77, + "learning_rate": 2.9547127963083786e-07, + "loss": 1.17, + "step": 32950 + }, + { + "epoch": 2.77, + "learning_rate": 2.9525199293868766e-07, + "loss": 1.132, + "step": 32951 + }, + { + "epoch": 2.77, + "learning_rate": 2.950327864302738e-07, + "loss": 1.0416, + "step": 32952 + }, + { + "epoch": 2.77, + "learning_rate": 2.9481366010740364e-07, + "loss": 1.1414, + "step": 32953 + }, + { + "epoch": 2.77, + "learning_rate": 2.9459461397188915e-07, + "loss": 1.1211, + "step": 32954 + }, + { + "epoch": 2.77, + "learning_rate": 2.9437564802553996e-07, + "loss": 1.153, + "step": 32955 + }, + { + "epoch": 2.77, + "learning_rate": 2.9415676227016575e-07, + "loss": 1.1217, + "step": 32956 + }, + { + "epoch": 2.77, + "learning_rate": 2.9393795670757394e-07, + "loss": 1.1391, + "step": 32957 + }, + { + "epoch": 2.77, + "learning_rate": 2.9371923133957427e-07, + "loss": 1.1497, + "step": 32958 + }, + { + "epoch": 2.77, + "learning_rate": 2.9350058616797186e-07, + "loss": 1.2136, + "step": 32959 + }, + { + "epoch": 2.77, + "learning_rate": 2.932820211945742e-07, + "loss": 1.097, + "step": 32960 + }, + { + "epoch": 2.78, + "learning_rate": 2.930635364211876e-07, + "loss": 1.1027, + "step": 32961 + }, + { + "epoch": 2.78, + "learning_rate": 2.9284513184961507e-07, + "loss": 1.0801, + "step": 32962 + }, + { + "epoch": 2.78, + "learning_rate": 2.926268074816652e-07, + "loss": 1.2217, + "step": 32963 + }, + { + "epoch": 2.78, + "learning_rate": 2.924085633191387e-07, + "loss": 1.0555, + "step": 32964 + }, + { + "epoch": 2.78, + "learning_rate": 2.9219039936383865e-07, + "loss": 1.0384, + "step": 32965 + }, + { + "epoch": 2.78, + "learning_rate": 2.91972315617568e-07, + "loss": 1.107, + "step": 32966 + }, + { + "epoch": 2.78, + "learning_rate": 2.9175431208212977e-07, + "loss": 1.1003, + "step": 32967 + }, + { + "epoch": 2.78, + "learning_rate": 2.9153638875932475e-07, + "loss": 1.1956, + "step": 32968 + }, + { + "epoch": 2.78, + "learning_rate": 2.913185456509537e-07, + "loss": 1.1668, + "step": 32969 + }, + { + "epoch": 2.78, + "learning_rate": 2.9110078275881413e-07, + "loss": 1.0995, + "step": 32970 + }, + { + "epoch": 2.78, + "learning_rate": 2.908831000847079e-07, + "loss": 1.1804, + "step": 32971 + }, + { + "epoch": 2.78, + "learning_rate": 2.906654976304335e-07, + "loss": 1.1809, + "step": 32972 + }, + { + "epoch": 2.78, + "learning_rate": 2.9044797539778734e-07, + "loss": 1.0406, + "step": 32973 + }, + { + "epoch": 2.78, + "learning_rate": 2.902305333885691e-07, + "loss": 1.1158, + "step": 32974 + }, + { + "epoch": 2.78, + "learning_rate": 2.900131716045718e-07, + "loss": 1.1538, + "step": 32975 + }, + { + "epoch": 2.78, + "learning_rate": 2.897958900475939e-07, + "loss": 1.1139, + "step": 32976 + }, + { + "epoch": 2.78, + "learning_rate": 2.8957868871942964e-07, + "loss": 1.1186, + "step": 32977 + }, + { + "epoch": 2.78, + "learning_rate": 2.893615676218742e-07, + "loss": 1.0543, + "step": 32978 + }, + { + "epoch": 2.78, + "learning_rate": 2.8914452675672167e-07, + "loss": 1.1441, + "step": 32979 + }, + { + "epoch": 2.78, + "learning_rate": 2.8892756612576623e-07, + "loss": 1.0731, + "step": 32980 + }, + { + "epoch": 2.78, + "learning_rate": 2.887106857307964e-07, + "loss": 1.1452, + "step": 32981 + }, + { + "epoch": 2.78, + "learning_rate": 2.8849388557360745e-07, + "loss": 1.1547, + "step": 32982 + }, + { + "epoch": 2.78, + "learning_rate": 2.882771656559924e-07, + "loss": 1.1258, + "step": 32983 + }, + { + "epoch": 2.78, + "learning_rate": 2.8806052597973643e-07, + "loss": 1.1271, + "step": 32984 + }, + { + "epoch": 2.78, + "learning_rate": 2.8784396654663483e-07, + "loss": 1.1083, + "step": 32985 + }, + { + "epoch": 2.78, + "learning_rate": 2.87627487358475e-07, + "loss": 1.0462, + "step": 32986 + }, + { + "epoch": 2.78, + "learning_rate": 2.874110884170434e-07, + "loss": 1.0324, + "step": 32987 + }, + { + "epoch": 2.78, + "learning_rate": 2.8719476972413064e-07, + "loss": 1.0784, + "step": 32988 + }, + { + "epoch": 2.78, + "learning_rate": 2.869785312815232e-07, + "loss": 1.0845, + "step": 32989 + }, + { + "epoch": 2.78, + "learning_rate": 2.8676237309100743e-07, + "loss": 1.1068, + "step": 32990 + }, + { + "epoch": 2.78, + "learning_rate": 2.8654629515437074e-07, + "loss": 1.0822, + "step": 32991 + }, + { + "epoch": 2.78, + "learning_rate": 2.8633029747339503e-07, + "loss": 1.0764, + "step": 32992 + }, + { + "epoch": 2.78, + "learning_rate": 2.8611438004987004e-07, + "loss": 1.0754, + "step": 32993 + }, + { + "epoch": 2.78, + "learning_rate": 2.8589854288557537e-07, + "loss": 1.0383, + "step": 32994 + }, + { + "epoch": 2.78, + "learning_rate": 2.8568278598229415e-07, + "loss": 1.0793, + "step": 32995 + }, + { + "epoch": 2.78, + "learning_rate": 2.8546710934181376e-07, + "loss": 1.1286, + "step": 32996 + }, + { + "epoch": 2.78, + "learning_rate": 2.852515129659117e-07, + "loss": 1.1377, + "step": 32997 + }, + { + "epoch": 2.78, + "learning_rate": 2.8503599685637095e-07, + "loss": 1.0905, + "step": 32998 + }, + { + "epoch": 2.78, + "learning_rate": 2.8482056101497126e-07, + "loss": 1.1116, + "step": 32999 + }, + { + "epoch": 2.78, + "learning_rate": 2.846052054434956e-07, + "loss": 1.2395, + "step": 33000 + }, + { + "epoch": 2.78, + "learning_rate": 2.8438993014371807e-07, + "loss": 1.0869, + "step": 33001 + }, + { + "epoch": 2.78, + "learning_rate": 2.8417473511742286e-07, + "loss": 1.1355, + "step": 33002 + }, + { + "epoch": 2.78, + "learning_rate": 2.8395962036638416e-07, + "loss": 1.1355, + "step": 33003 + }, + { + "epoch": 2.78, + "learning_rate": 2.837445858923804e-07, + "loss": 1.2104, + "step": 33004 + }, + { + "epoch": 2.78, + "learning_rate": 2.8352963169718915e-07, + "loss": 1.1166, + "step": 33005 + }, + { + "epoch": 2.78, + "learning_rate": 2.8331475778258455e-07, + "loss": 1.1163, + "step": 33006 + }, + { + "epoch": 2.78, + "learning_rate": 2.83099964150344e-07, + "loss": 1.0345, + "step": 33007 + }, + { + "epoch": 2.78, + "learning_rate": 2.8288525080224173e-07, + "loss": 1.132, + "step": 33008 + }, + { + "epoch": 2.78, + "learning_rate": 2.8267061774005067e-07, + "loss": 1.0352, + "step": 33009 + }, + { + "epoch": 2.78, + "learning_rate": 2.82456064965545e-07, + "loss": 1.081, + "step": 33010 + }, + { + "epoch": 2.78, + "learning_rate": 2.822415924804989e-07, + "loss": 0.9843, + "step": 33011 + }, + { + "epoch": 2.78, + "learning_rate": 2.8202720028668084e-07, + "loss": 1.0423, + "step": 33012 + }, + { + "epoch": 2.78, + "learning_rate": 2.818128883858662e-07, + "loss": 1.1489, + "step": 33013 + }, + { + "epoch": 2.78, + "learning_rate": 2.8159865677982234e-07, + "loss": 1.1445, + "step": 33014 + }, + { + "epoch": 2.78, + "learning_rate": 2.8138450547032016e-07, + "loss": 1.0854, + "step": 33015 + }, + { + "epoch": 2.78, + "learning_rate": 2.811704344591304e-07, + "loss": 1.0787, + "step": 33016 + }, + { + "epoch": 2.78, + "learning_rate": 2.809564437480206e-07, + "loss": 1.1097, + "step": 33017 + }, + { + "epoch": 2.78, + "learning_rate": 2.807425333387592e-07, + "loss": 1.1439, + "step": 33018 + }, + { + "epoch": 2.78, + "learning_rate": 2.805287032331139e-07, + "loss": 1.0996, + "step": 33019 + }, + { + "epoch": 2.78, + "learning_rate": 2.8031495343284974e-07, + "loss": 1.0871, + "step": 33020 + }, + { + "epoch": 2.78, + "learning_rate": 2.8010128393973543e-07, + "loss": 1.109, + "step": 33021 + }, + { + "epoch": 2.78, + "learning_rate": 2.798876947555351e-07, + "loss": 1.0258, + "step": 33022 + }, + { + "epoch": 2.78, + "learning_rate": 2.7967418588201287e-07, + "loss": 1.1216, + "step": 33023 + }, + { + "epoch": 2.78, + "learning_rate": 2.7946075732093404e-07, + "loss": 1.1423, + "step": 33024 + }, + { + "epoch": 2.78, + "learning_rate": 2.792474090740616e-07, + "loss": 1.045, + "step": 33025 + }, + { + "epoch": 2.78, + "learning_rate": 2.7903414114315743e-07, + "loss": 1.1049, + "step": 33026 + }, + { + "epoch": 2.78, + "learning_rate": 2.788209535299835e-07, + "loss": 1.0472, + "step": 33027 + }, + { + "epoch": 2.78, + "learning_rate": 2.78607846236304e-07, + "loss": 1.1518, + "step": 33028 + }, + { + "epoch": 2.78, + "learning_rate": 2.7839481926387634e-07, + "loss": 1.171, + "step": 33029 + }, + { + "epoch": 2.78, + "learning_rate": 2.781818726144625e-07, + "loss": 1.1051, + "step": 33030 + }, + { + "epoch": 2.78, + "learning_rate": 2.7796900628982105e-07, + "loss": 1.2072, + "step": 33031 + }, + { + "epoch": 2.78, + "learning_rate": 2.7775622029171054e-07, + "loss": 1.1168, + "step": 33032 + }, + { + "epoch": 2.78, + "learning_rate": 2.775435146218919e-07, + "loss": 1.0112, + "step": 33033 + }, + { + "epoch": 2.78, + "learning_rate": 2.7733088928211804e-07, + "loss": 1.1476, + "step": 33034 + }, + { + "epoch": 2.78, + "learning_rate": 2.7711834427414985e-07, + "loss": 1.084, + "step": 33035 + }, + { + "epoch": 2.78, + "learning_rate": 2.7690587959974034e-07, + "loss": 1.1268, + "step": 33036 + }, + { + "epoch": 2.78, + "learning_rate": 2.766934952606459e-07, + "loss": 1.137, + "step": 33037 + }, + { + "epoch": 2.78, + "learning_rate": 2.7648119125862296e-07, + "loss": 1.1026, + "step": 33038 + }, + { + "epoch": 2.78, + "learning_rate": 2.7626896759542333e-07, + "loss": 1.1625, + "step": 33039 + }, + { + "epoch": 2.78, + "learning_rate": 2.760568242728012e-07, + "loss": 1.143, + "step": 33040 + }, + { + "epoch": 2.78, + "learning_rate": 2.758447612925119e-07, + "loss": 1.1387, + "step": 33041 + }, + { + "epoch": 2.78, + "learning_rate": 2.756327786563018e-07, + "loss": 1.0688, + "step": 33042 + }, + { + "epoch": 2.78, + "learning_rate": 2.754208763659294e-07, + "loss": 1.1199, + "step": 33043 + }, + { + "epoch": 2.78, + "learning_rate": 2.7520905442314007e-07, + "loss": 1.1692, + "step": 33044 + }, + { + "epoch": 2.78, + "learning_rate": 2.7499731282968455e-07, + "loss": 1.1307, + "step": 33045 + }, + { + "epoch": 2.78, + "learning_rate": 2.7478565158731707e-07, + "loss": 1.1087, + "step": 33046 + }, + { + "epoch": 2.78, + "learning_rate": 2.7457407069778066e-07, + "loss": 1.0959, + "step": 33047 + }, + { + "epoch": 2.78, + "learning_rate": 2.743625701628261e-07, + "loss": 1.2021, + "step": 33048 + }, + { + "epoch": 2.78, + "learning_rate": 2.7415114998420087e-07, + "loss": 1.1396, + "step": 33049 + }, + { + "epoch": 2.78, + "learning_rate": 2.7393981016365256e-07, + "loss": 1.2083, + "step": 33050 + }, + { + "epoch": 2.78, + "learning_rate": 2.7372855070292415e-07, + "loss": 1.0795, + "step": 33051 + }, + { + "epoch": 2.78, + "learning_rate": 2.7351737160376426e-07, + "loss": 1.0966, + "step": 33052 + }, + { + "epoch": 2.78, + "learning_rate": 2.73306272867917e-07, + "loss": 1.1046, + "step": 33053 + }, + { + "epoch": 2.78, + "learning_rate": 2.730952544971255e-07, + "loss": 1.0964, + "step": 33054 + }, + { + "epoch": 2.78, + "learning_rate": 2.7288431649313494e-07, + "loss": 1.1682, + "step": 33055 + }, + { + "epoch": 2.78, + "learning_rate": 2.7267345885768626e-07, + "loss": 1.1113, + "step": 33056 + }, + { + "epoch": 2.78, + "learning_rate": 2.7246268159252356e-07, + "loss": 1.0094, + "step": 33057 + }, + { + "epoch": 2.78, + "learning_rate": 2.722519846993865e-07, + "loss": 1.1723, + "step": 33058 + }, + { + "epoch": 2.78, + "learning_rate": 2.7204136818001714e-07, + "loss": 1.1041, + "step": 33059 + }, + { + "epoch": 2.78, + "learning_rate": 2.7183083203615626e-07, + "loss": 1.201, + "step": 33060 + }, + { + "epoch": 2.78, + "learning_rate": 2.716203762695424e-07, + "loss": 1.2072, + "step": 33061 + }, + { + "epoch": 2.78, + "learning_rate": 2.7141000088191204e-07, + "loss": 1.1149, + "step": 33062 + }, + { + "epoch": 2.78, + "learning_rate": 2.711997058750093e-07, + "loss": 1.131, + "step": 33063 + }, + { + "epoch": 2.78, + "learning_rate": 2.709894912505662e-07, + "loss": 1.0446, + "step": 33064 + }, + { + "epoch": 2.78, + "learning_rate": 2.707793570103212e-07, + "loss": 1.1076, + "step": 33065 + }, + { + "epoch": 2.78, + "learning_rate": 2.705693031560119e-07, + "loss": 1.0916, + "step": 33066 + }, + { + "epoch": 2.78, + "learning_rate": 2.7035932968937253e-07, + "loss": 1.0154, + "step": 33067 + }, + { + "epoch": 2.78, + "learning_rate": 2.701494366121382e-07, + "loss": 1.1213, + "step": 33068 + }, + { + "epoch": 2.78, + "learning_rate": 2.6993962392604323e-07, + "loss": 1.0909, + "step": 33069 + }, + { + "epoch": 2.78, + "learning_rate": 2.697298916328206e-07, + "loss": 1.1562, + "step": 33070 + }, + { + "epoch": 2.78, + "learning_rate": 2.6952023973420336e-07, + "loss": 1.2123, + "step": 33071 + }, + { + "epoch": 2.78, + "learning_rate": 2.6931066823192573e-07, + "loss": 1.108, + "step": 33072 + }, + { + "epoch": 2.78, + "learning_rate": 2.6910117712771634e-07, + "loss": 0.9966, + "step": 33073 + }, + { + "epoch": 2.78, + "learning_rate": 2.688917664233082e-07, + "loss": 1.1305, + "step": 33074 + }, + { + "epoch": 2.78, + "learning_rate": 2.6868243612042986e-07, + "loss": 1.1844, + "step": 33075 + }, + { + "epoch": 2.78, + "learning_rate": 2.684731862208123e-07, + "loss": 1.1391, + "step": 33076 + }, + { + "epoch": 2.78, + "learning_rate": 2.6826401672618295e-07, + "loss": 1.175, + "step": 33077 + }, + { + "epoch": 2.78, + "learning_rate": 2.6805492763827157e-07, + "loss": 1.1202, + "step": 33078 + }, + { + "epoch": 2.78, + "learning_rate": 2.678459189588045e-07, + "loss": 1.0415, + "step": 33079 + }, + { + "epoch": 2.79, + "learning_rate": 2.676369906895104e-07, + "loss": 1.0953, + "step": 33080 + }, + { + "epoch": 2.79, + "learning_rate": 2.674281428321113e-07, + "loss": 1.2019, + "step": 33081 + }, + { + "epoch": 2.79, + "learning_rate": 2.6721937538833786e-07, + "loss": 1.0832, + "step": 33082 + }, + { + "epoch": 2.79, + "learning_rate": 2.6701068835991326e-07, + "loss": 1.0603, + "step": 33083 + }, + { + "epoch": 2.79, + "learning_rate": 2.668020817485595e-07, + "loss": 1.1636, + "step": 33084 + }, + { + "epoch": 2.79, + "learning_rate": 2.6659355555600395e-07, + "loss": 1.0946, + "step": 33085 + }, + { + "epoch": 2.79, + "learning_rate": 2.6638510978396646e-07, + "loss": 1.0858, + "step": 33086 + }, + { + "epoch": 2.79, + "learning_rate": 2.661767444341701e-07, + "loss": 1.0265, + "step": 33087 + }, + { + "epoch": 2.79, + "learning_rate": 2.6596845950833784e-07, + "loss": 1.101, + "step": 33088 + }, + { + "epoch": 2.79, + "learning_rate": 2.6576025500818836e-07, + "loss": 1.073, + "step": 33089 + }, + { + "epoch": 2.79, + "learning_rate": 2.6555213093544364e-07, + "loss": 1.1351, + "step": 33090 + }, + { + "epoch": 2.79, + "learning_rate": 2.6534408729182336e-07, + "loss": 1.1695, + "step": 33091 + }, + { + "epoch": 2.79, + "learning_rate": 2.6513612407904286e-07, + "loss": 1.095, + "step": 33092 + }, + { + "epoch": 2.79, + "learning_rate": 2.6492824129882634e-07, + "loss": 1.162, + "step": 33093 + }, + { + "epoch": 2.79, + "learning_rate": 2.647204389528868e-07, + "loss": 1.0977, + "step": 33094 + }, + { + "epoch": 2.79, + "learning_rate": 2.645127170429418e-07, + "loss": 1.1087, + "step": 33095 + }, + { + "epoch": 2.79, + "learning_rate": 2.6430507557071104e-07, + "loss": 1.1404, + "step": 33096 + }, + { + "epoch": 2.79, + "learning_rate": 2.6409751453790655e-07, + "loss": 1.134, + "step": 33097 + }, + { + "epoch": 2.79, + "learning_rate": 2.6389003394624356e-07, + "loss": 1.1166, + "step": 33098 + }, + { + "epoch": 2.79, + "learning_rate": 2.6368263379743743e-07, + "loss": 1.0777, + "step": 33099 + }, + { + "epoch": 2.79, + "learning_rate": 2.634753140932023e-07, + "loss": 1.1118, + "step": 33100 + }, + { + "epoch": 2.79, + "learning_rate": 2.6326807483524786e-07, + "loss": 1.1358, + "step": 33101 + }, + { + "epoch": 2.79, + "learning_rate": 2.6306091602529173e-07, + "loss": 1.137, + "step": 33102 + }, + { + "epoch": 2.79, + "learning_rate": 2.6285383766504024e-07, + "loss": 1.1489, + "step": 33103 + }, + { + "epoch": 2.79, + "learning_rate": 2.626468397562076e-07, + "loss": 1.1262, + "step": 33104 + }, + { + "epoch": 2.79, + "learning_rate": 2.624399223005025e-07, + "loss": 1.1457, + "step": 33105 + }, + { + "epoch": 2.79, + "learning_rate": 2.6223308529963574e-07, + "loss": 1.1506, + "step": 33106 + }, + { + "epoch": 2.79, + "learning_rate": 2.6202632875531595e-07, + "loss": 1.0918, + "step": 33107 + }, + { + "epoch": 2.79, + "learning_rate": 2.618196526692507e-07, + "loss": 1.1459, + "step": 33108 + }, + { + "epoch": 2.79, + "learning_rate": 2.6161305704314743e-07, + "loss": 1.0856, + "step": 33109 + }, + { + "epoch": 2.79, + "learning_rate": 2.6140654187871485e-07, + "loss": 1.1027, + "step": 33110 + }, + { + "epoch": 2.79, + "learning_rate": 2.612001071776582e-07, + "loss": 1.0683, + "step": 33111 + }, + { + "epoch": 2.79, + "learning_rate": 2.6099375294168175e-07, + "loss": 1.1057, + "step": 33112 + }, + { + "epoch": 2.79, + "learning_rate": 2.6078747917249403e-07, + "loss": 1.0734, + "step": 33113 + }, + { + "epoch": 2.79, + "learning_rate": 2.60581285871796e-07, + "loss": 1.0638, + "step": 33114 + }, + { + "epoch": 2.79, + "learning_rate": 2.6037517304129176e-07, + "loss": 1.0902, + "step": 33115 + }, + { + "epoch": 2.79, + "learning_rate": 2.601691406826856e-07, + "loss": 1.0254, + "step": 33116 + }, + { + "epoch": 2.79, + "learning_rate": 2.599631887976783e-07, + "loss": 1.1606, + "step": 33117 + }, + { + "epoch": 2.79, + "learning_rate": 2.5975731738797307e-07, + "loss": 1.1229, + "step": 33118 + }, + { + "epoch": 2.79, + "learning_rate": 2.5955152645527173e-07, + "loss": 1.1132, + "step": 33119 + }, + { + "epoch": 2.79, + "learning_rate": 2.593458160012696e-07, + "loss": 1.0935, + "step": 33120 + }, + { + "epoch": 2.79, + "learning_rate": 2.591401860276721e-07, + "loss": 1.1564, + "step": 33121 + }, + { + "epoch": 2.79, + "learning_rate": 2.5893463653617555e-07, + "loss": 1.0887, + "step": 33122 + }, + { + "epoch": 2.79, + "learning_rate": 2.5872916752847755e-07, + "loss": 1.0699, + "step": 33123 + }, + { + "epoch": 2.79, + "learning_rate": 2.585237790062778e-07, + "loss": 1.1228, + "step": 33124 + }, + { + "epoch": 2.79, + "learning_rate": 2.5831847097127273e-07, + "loss": 1.1657, + "step": 33125 + }, + { + "epoch": 2.79, + "learning_rate": 2.581132434251565e-07, + "loss": 1.151, + "step": 33126 + }, + { + "epoch": 2.79, + "learning_rate": 2.579080963696279e-07, + "loss": 1.1119, + "step": 33127 + }, + { + "epoch": 2.79, + "learning_rate": 2.5770302980637983e-07, + "loss": 1.0874, + "step": 33128 + }, + { + "epoch": 2.79, + "learning_rate": 2.5749804373710773e-07, + "loss": 1.0993, + "step": 33129 + }, + { + "epoch": 2.79, + "learning_rate": 2.5729313816350467e-07, + "loss": 1.1681, + "step": 33130 + }, + { + "epoch": 2.79, + "learning_rate": 2.5708831308726255e-07, + "loss": 1.0339, + "step": 33131 + }, + { + "epoch": 2.79, + "learning_rate": 2.5688356851007565e-07, + "loss": 1.148, + "step": 33132 + }, + { + "epoch": 2.79, + "learning_rate": 2.566789044336371e-07, + "loss": 1.0871, + "step": 33133 + }, + { + "epoch": 2.79, + "learning_rate": 2.564743208596321e-07, + "loss": 1.1254, + "step": 33134 + }, + { + "epoch": 2.79, + "learning_rate": 2.5626981778975714e-07, + "loss": 1.0646, + "step": 33135 + }, + { + "epoch": 2.79, + "learning_rate": 2.560653952256986e-07, + "loss": 1.1213, + "step": 33136 + }, + { + "epoch": 2.79, + "learning_rate": 2.558610531691463e-07, + "loss": 1.1812, + "step": 33137 + }, + { + "epoch": 2.79, + "learning_rate": 2.556567916217878e-07, + "loss": 1.1018, + "step": 33138 + }, + { + "epoch": 2.79, + "learning_rate": 2.554526105853117e-07, + "loss": 1.0332, + "step": 33139 + }, + { + "epoch": 2.79, + "learning_rate": 2.552485100614055e-07, + "loss": 1.1961, + "step": 33140 + }, + { + "epoch": 2.79, + "learning_rate": 2.550444900517557e-07, + "loss": 1.072, + "step": 33141 + }, + { + "epoch": 2.79, + "learning_rate": 2.5484055055804427e-07, + "loss": 1.1713, + "step": 33142 + }, + { + "epoch": 2.79, + "learning_rate": 2.546366915819609e-07, + "loss": 1.0474, + "step": 33143 + }, + { + "epoch": 2.79, + "learning_rate": 2.544329131251888e-07, + "loss": 1.0785, + "step": 33144 + }, + { + "epoch": 2.79, + "learning_rate": 2.542292151894088e-07, + "loss": 1.0695, + "step": 33145 + }, + { + "epoch": 2.79, + "learning_rate": 2.540255977763073e-07, + "loss": 1.1375, + "step": 33146 + }, + { + "epoch": 2.79, + "learning_rate": 2.538220608875652e-07, + "loss": 1.1143, + "step": 33147 + }, + { + "epoch": 2.79, + "learning_rate": 2.5361860452486453e-07, + "loss": 1.1625, + "step": 33148 + }, + { + "epoch": 2.79, + "learning_rate": 2.534152286898861e-07, + "loss": 1.2055, + "step": 33149 + }, + { + "epoch": 2.79, + "learning_rate": 2.532119333843119e-07, + "loss": 1.1844, + "step": 33150 + }, + { + "epoch": 2.79, + "learning_rate": 2.5300871860981737e-07, + "loss": 1.1004, + "step": 33151 + }, + { + "epoch": 2.79, + "learning_rate": 2.5280558436808655e-07, + "loss": 1.1553, + "step": 33152 + }, + { + "epoch": 2.79, + "learning_rate": 2.5260253066079485e-07, + "loss": 1.1684, + "step": 33153 + }, + { + "epoch": 2.79, + "learning_rate": 2.5239955748961987e-07, + "loss": 1.1378, + "step": 33154 + }, + { + "epoch": 2.79, + "learning_rate": 2.5219666485623905e-07, + "loss": 1.071, + "step": 33155 + }, + { + "epoch": 2.79, + "learning_rate": 2.5199385276233004e-07, + "loss": 1.1521, + "step": 33156 + }, + { + "epoch": 2.79, + "learning_rate": 2.5179112120956696e-07, + "loss": 1.1764, + "step": 33157 + }, + { + "epoch": 2.79, + "learning_rate": 2.515884701996263e-07, + "loss": 1.0906, + "step": 33158 + }, + { + "epoch": 2.79, + "learning_rate": 2.5138589973418114e-07, + "loss": 1.1164, + "step": 33159 + }, + { + "epoch": 2.79, + "learning_rate": 2.5118340981490576e-07, + "loss": 1.1751, + "step": 33160 + }, + { + "epoch": 2.79, + "learning_rate": 2.5098100044347317e-07, + "loss": 1.1, + "step": 33161 + }, + { + "epoch": 2.79, + "learning_rate": 2.5077867162155433e-07, + "loss": 1.0651, + "step": 33162 + }, + { + "epoch": 2.79, + "learning_rate": 2.5057642335082454e-07, + "loss": 1.1966, + "step": 33163 + }, + { + "epoch": 2.79, + "learning_rate": 2.5037425563295137e-07, + "loss": 1.048, + "step": 33164 + }, + { + "epoch": 2.79, + "learning_rate": 2.501721684696057e-07, + "loss": 1.0735, + "step": 33165 + }, + { + "epoch": 2.79, + "learning_rate": 2.499701618624584e-07, + "loss": 1.1622, + "step": 33166 + }, + { + "epoch": 2.79, + "learning_rate": 2.497682358131781e-07, + "loss": 1.1862, + "step": 33167 + }, + { + "epoch": 2.79, + "learning_rate": 2.4956639032343356e-07, + "loss": 1.0984, + "step": 33168 + }, + { + "epoch": 2.79, + "learning_rate": 2.4936462539489227e-07, + "loss": 1.1403, + "step": 33169 + }, + { + "epoch": 2.79, + "learning_rate": 2.491629410292196e-07, + "loss": 1.069, + "step": 33170 + }, + { + "epoch": 2.79, + "learning_rate": 2.489613372280841e-07, + "loss": 0.9723, + "step": 33171 + }, + { + "epoch": 2.79, + "learning_rate": 2.487598139931513e-07, + "loss": 1.0819, + "step": 33172 + }, + { + "epoch": 2.79, + "learning_rate": 2.485583713260842e-07, + "loss": 1.1357, + "step": 33173 + }, + { + "epoch": 2.79, + "learning_rate": 2.4835700922855144e-07, + "loss": 1.1757, + "step": 33174 + }, + { + "epoch": 2.79, + "learning_rate": 2.4815572770221173e-07, + "loss": 1.0211, + "step": 33175 + }, + { + "epoch": 2.79, + "learning_rate": 2.4795452674873154e-07, + "loss": 1.1661, + "step": 33176 + }, + { + "epoch": 2.79, + "learning_rate": 2.4775340636977176e-07, + "loss": 1.0973, + "step": 33177 + }, + { + "epoch": 2.79, + "learning_rate": 2.4755236656699324e-07, + "loss": 1.1085, + "step": 33178 + }, + { + "epoch": 2.79, + "learning_rate": 2.473514073420602e-07, + "loss": 1.1266, + "step": 33179 + }, + { + "epoch": 2.79, + "learning_rate": 2.471505286966303e-07, + "loss": 1.1161, + "step": 33180 + }, + { + "epoch": 2.79, + "learning_rate": 2.4694973063236314e-07, + "loss": 1.2273, + "step": 33181 + }, + { + "epoch": 2.79, + "learning_rate": 2.467490131509198e-07, + "loss": 1.1408, + "step": 33182 + }, + { + "epoch": 2.79, + "learning_rate": 2.4654837625395777e-07, + "loss": 1.1653, + "step": 33183 + }, + { + "epoch": 2.79, + "learning_rate": 2.463478199431324e-07, + "loss": 1.1029, + "step": 33184 + }, + { + "epoch": 2.79, + "learning_rate": 2.4614734422010566e-07, + "loss": 1.0936, + "step": 33185 + }, + { + "epoch": 2.79, + "learning_rate": 2.459469490865296e-07, + "loss": 1.1101, + "step": 33186 + }, + { + "epoch": 2.79, + "learning_rate": 2.4574663454406287e-07, + "loss": 1.1538, + "step": 33187 + }, + { + "epoch": 2.79, + "learning_rate": 2.455464005943575e-07, + "loss": 1.1305, + "step": 33188 + }, + { + "epoch": 2.79, + "learning_rate": 2.453462472390711e-07, + "loss": 1.097, + "step": 33189 + }, + { + "epoch": 2.79, + "learning_rate": 2.451461744798545e-07, + "loss": 1.1358, + "step": 33190 + }, + { + "epoch": 2.79, + "learning_rate": 2.449461823183641e-07, + "loss": 1.1353, + "step": 33191 + }, + { + "epoch": 2.79, + "learning_rate": 2.4474627075624757e-07, + "loss": 1.0317, + "step": 33192 + }, + { + "epoch": 2.79, + "learning_rate": 2.4454643979516245e-07, + "loss": 1.0935, + "step": 33193 + }, + { + "epoch": 2.79, + "learning_rate": 2.443466894367552e-07, + "loss": 1.1324, + "step": 33194 + }, + { + "epoch": 2.79, + "learning_rate": 2.4414701968267675e-07, + "loss": 1.0666, + "step": 33195 + }, + { + "epoch": 2.79, + "learning_rate": 2.4394743053458013e-07, + "loss": 1.1879, + "step": 33196 + }, + { + "epoch": 2.79, + "learning_rate": 2.4374792199411077e-07, + "loss": 1.1089, + "step": 33197 + }, + { + "epoch": 2.79, + "learning_rate": 2.435484940629185e-07, + "loss": 1.1039, + "step": 33198 + }, + { + "epoch": 2.8, + "learning_rate": 2.4334914674265185e-07, + "loss": 1.1696, + "step": 33199 + }, + { + "epoch": 2.8, + "learning_rate": 2.431498800349574e-07, + "loss": 1.16, + "step": 33200 + }, + { + "epoch": 2.8, + "learning_rate": 2.429506939414794e-07, + "loss": 1.1416, + "step": 33201 + }, + { + "epoch": 2.8, + "learning_rate": 2.427515884638676e-07, + "loss": 1.0243, + "step": 33202 + }, + { + "epoch": 2.8, + "learning_rate": 2.425525636037629e-07, + "loss": 1.1464, + "step": 33203 + }, + { + "epoch": 2.8, + "learning_rate": 2.423536193628129e-07, + "loss": 1.1769, + "step": 33204 + }, + { + "epoch": 2.8, + "learning_rate": 2.4215475574265956e-07, + "loss": 1.0338, + "step": 33205 + }, + { + "epoch": 2.8, + "learning_rate": 2.419559727449472e-07, + "loss": 1.0687, + "step": 33206 + }, + { + "epoch": 2.8, + "learning_rate": 2.417572703713167e-07, + "loss": 1.0676, + "step": 33207 + }, + { + "epoch": 2.8, + "learning_rate": 2.415586486234112e-07, + "loss": 1.0972, + "step": 33208 + }, + { + "epoch": 2.8, + "learning_rate": 2.4136010750287044e-07, + "loss": 1.095, + "step": 33209 + }, + { + "epoch": 2.8, + "learning_rate": 2.411616470113365e-07, + "loss": 1.0544, + "step": 33210 + }, + { + "epoch": 2.8, + "learning_rate": 2.4096326715044915e-07, + "loss": 1.0528, + "step": 33211 + }, + { + "epoch": 2.8, + "learning_rate": 2.407649679218449e-07, + "loss": 0.9985, + "step": 33212 + }, + { + "epoch": 2.8, + "learning_rate": 2.405667493271646e-07, + "loss": 1.1078, + "step": 33213 + }, + { + "epoch": 2.8, + "learning_rate": 2.403686113680459e-07, + "loss": 1.1175, + "step": 33214 + }, + { + "epoch": 2.8, + "learning_rate": 2.4017055404612413e-07, + "loss": 1.1748, + "step": 33215 + }, + { + "epoch": 2.8, + "learning_rate": 2.399725773630368e-07, + "loss": 1.1289, + "step": 33216 + }, + { + "epoch": 2.8, + "learning_rate": 2.3977468132042046e-07, + "loss": 1.1088, + "step": 33217 + }, + { + "epoch": 2.8, + "learning_rate": 2.3957686591990826e-07, + "loss": 1.0385, + "step": 33218 + }, + { + "epoch": 2.8, + "learning_rate": 2.393791311631377e-07, + "loss": 1.0431, + "step": 33219 + }, + { + "epoch": 2.8, + "learning_rate": 2.391814770517376e-07, + "loss": 1.1064, + "step": 33220 + }, + { + "epoch": 2.8, + "learning_rate": 2.3898390358734536e-07, + "loss": 1.1485, + "step": 33221 + }, + { + "epoch": 2.8, + "learning_rate": 2.3878641077159316e-07, + "loss": 1.0624, + "step": 33222 + }, + { + "epoch": 2.8, + "learning_rate": 2.385889986061096e-07, + "loss": 1.0803, + "step": 33223 + }, + { + "epoch": 2.8, + "learning_rate": 2.3839166709253014e-07, + "loss": 1.122, + "step": 33224 + }, + { + "epoch": 2.8, + "learning_rate": 2.3819441623248118e-07, + "loss": 1.1127, + "step": 33225 + }, + { + "epoch": 2.8, + "learning_rate": 2.3799724602759477e-07, + "loss": 1.1102, + "step": 33226 + }, + { + "epoch": 2.8, + "learning_rate": 2.3780015647949962e-07, + "loss": 1.1007, + "step": 33227 + }, + { + "epoch": 2.8, + "learning_rate": 2.3760314758982328e-07, + "loss": 1.1478, + "step": 33228 + }, + { + "epoch": 2.8, + "learning_rate": 2.3740621936019338e-07, + "loss": 1.181, + "step": 33229 + }, + { + "epoch": 2.8, + "learning_rate": 2.372093717922397e-07, + "loss": 1.1416, + "step": 33230 + }, + { + "epoch": 2.8, + "learning_rate": 2.3701260488758425e-07, + "loss": 1.113, + "step": 33231 + }, + { + "epoch": 2.8, + "learning_rate": 2.3681591864785692e-07, + "loss": 1.2161, + "step": 33232 + }, + { + "epoch": 2.8, + "learning_rate": 2.366193130746819e-07, + "loss": 1.0576, + "step": 33233 + }, + { + "epoch": 2.8, + "learning_rate": 2.3642278816968013e-07, + "loss": 1.1185, + "step": 33234 + }, + { + "epoch": 2.8, + "learning_rate": 2.362263439344803e-07, + "loss": 1.1011, + "step": 33235 + }, + { + "epoch": 2.8, + "learning_rate": 2.3602998037070223e-07, + "loss": 1.1338, + "step": 33236 + }, + { + "epoch": 2.8, + "learning_rate": 2.3583369747996908e-07, + "loss": 1.0711, + "step": 33237 + }, + { + "epoch": 2.8, + "learning_rate": 2.35637495263904e-07, + "loss": 1.198, + "step": 33238 + }, + { + "epoch": 2.8, + "learning_rate": 2.354413737241257e-07, + "loss": 1.2619, + "step": 33239 + }, + { + "epoch": 2.8, + "learning_rate": 2.3524533286225614e-07, + "loss": 1.1474, + "step": 33240 + }, + { + "epoch": 2.8, + "learning_rate": 2.3504937267991634e-07, + "loss": 1.1617, + "step": 33241 + }, + { + "epoch": 2.8, + "learning_rate": 2.3485349317872163e-07, + "loss": 1.0601, + "step": 33242 + }, + { + "epoch": 2.8, + "learning_rate": 2.3465769436029405e-07, + "loss": 1.1895, + "step": 33243 + }, + { + "epoch": 2.8, + "learning_rate": 2.34461976226249e-07, + "loss": 1.121, + "step": 33244 + }, + { + "epoch": 2.8, + "learning_rate": 2.3426633877820403e-07, + "loss": 1.0988, + "step": 33245 + }, + { + "epoch": 2.8, + "learning_rate": 2.3407078201777677e-07, + "loss": 1.0685, + "step": 33246 + }, + { + "epoch": 2.8, + "learning_rate": 2.3387530594658257e-07, + "loss": 1.1051, + "step": 33247 + }, + { + "epoch": 2.8, + "learning_rate": 2.3367991056623573e-07, + "loss": 1.0845, + "step": 33248 + }, + { + "epoch": 2.8, + "learning_rate": 2.3348459587835047e-07, + "loss": 1.0945, + "step": 33249 + }, + { + "epoch": 2.8, + "learning_rate": 2.3328936188454333e-07, + "loss": 1.0953, + "step": 33250 + }, + { + "epoch": 2.8, + "learning_rate": 2.3309420858642295e-07, + "loss": 1.118, + "step": 33251 + }, + { + "epoch": 2.8, + "learning_rate": 2.3289913598560477e-07, + "loss": 1.0488, + "step": 33252 + }, + { + "epoch": 2.8, + "learning_rate": 2.3270414408369968e-07, + "loss": 1.1259, + "step": 33253 + }, + { + "epoch": 2.8, + "learning_rate": 2.3250923288231973e-07, + "loss": 1.1322, + "step": 33254 + }, + { + "epoch": 2.8, + "learning_rate": 2.3231440238307367e-07, + "loss": 1.0598, + "step": 33255 + }, + { + "epoch": 2.8, + "learning_rate": 2.321196525875724e-07, + "loss": 1.0682, + "step": 33256 + }, + { + "epoch": 2.8, + "learning_rate": 2.3192498349742354e-07, + "loss": 0.9835, + "step": 33257 + }, + { + "epoch": 2.8, + "learning_rate": 2.31730395114238e-07, + "loss": 1.131, + "step": 33258 + }, + { + "epoch": 2.8, + "learning_rate": 2.315358874396212e-07, + "loss": 1.1397, + "step": 33259 + }, + { + "epoch": 2.8, + "learning_rate": 2.313414604751818e-07, + "loss": 1.0896, + "step": 33260 + }, + { + "epoch": 2.8, + "learning_rate": 2.3114711422252634e-07, + "loss": 1.048, + "step": 33261 + }, + { + "epoch": 2.8, + "learning_rate": 2.3095284868325796e-07, + "loss": 1.0973, + "step": 33262 + }, + { + "epoch": 2.8, + "learning_rate": 2.307586638589865e-07, + "loss": 1.1116, + "step": 33263 + }, + { + "epoch": 2.8, + "learning_rate": 2.3056455975131175e-07, + "loss": 1.0529, + "step": 33264 + }, + { + "epoch": 2.8, + "learning_rate": 2.3037053636183914e-07, + "loss": 1.1368, + "step": 33265 + }, + { + "epoch": 2.8, + "learning_rate": 2.3017659369217184e-07, + "loss": 1.0983, + "step": 33266 + }, + { + "epoch": 2.8, + "learning_rate": 2.2998273174391183e-07, + "loss": 1.0777, + "step": 33267 + }, + { + "epoch": 2.8, + "learning_rate": 2.2978895051866236e-07, + "loss": 0.9947, + "step": 33268 + }, + { + "epoch": 2.8, + "learning_rate": 2.2959525001802319e-07, + "loss": 1.1225, + "step": 33269 + }, + { + "epoch": 2.8, + "learning_rate": 2.294016302435942e-07, + "loss": 1.1421, + "step": 33270 + }, + { + "epoch": 2.8, + "learning_rate": 2.2920809119697628e-07, + "loss": 1.1074, + "step": 33271 + }, + { + "epoch": 2.8, + "learning_rate": 2.2901463287976932e-07, + "loss": 1.1467, + "step": 33272 + }, + { + "epoch": 2.8, + "learning_rate": 2.2882125529356868e-07, + "loss": 1.0858, + "step": 33273 + }, + { + "epoch": 2.8, + "learning_rate": 2.286279584399753e-07, + "loss": 1.1166, + "step": 33274 + }, + { + "epoch": 2.8, + "learning_rate": 2.284347423205846e-07, + "loss": 1.1109, + "step": 33275 + }, + { + "epoch": 2.8, + "learning_rate": 2.2824160693699416e-07, + "loss": 1.0499, + "step": 33276 + }, + { + "epoch": 2.8, + "learning_rate": 2.2804855229079826e-07, + "loss": 1.0857, + "step": 33277 + }, + { + "epoch": 2.8, + "learning_rate": 2.278555783835923e-07, + "loss": 1.0619, + "step": 33278 + }, + { + "epoch": 2.8, + "learning_rate": 2.2766268521697055e-07, + "loss": 1.0601, + "step": 33279 + }, + { + "epoch": 2.8, + "learning_rate": 2.2746987279252952e-07, + "loss": 1.1001, + "step": 33280 + }, + { + "epoch": 2.8, + "learning_rate": 2.2727714111185683e-07, + "loss": 1.0791, + "step": 33281 + }, + { + "epoch": 2.8, + "learning_rate": 2.270844901765501e-07, + "loss": 1.1624, + "step": 33282 + }, + { + "epoch": 2.8, + "learning_rate": 2.2689191998819916e-07, + "loss": 1.1362, + "step": 33283 + }, + { + "epoch": 2.8, + "learning_rate": 2.2669943054839383e-07, + "loss": 1.0992, + "step": 33284 + }, + { + "epoch": 2.8, + "learning_rate": 2.265070218587273e-07, + "loss": 1.1163, + "step": 33285 + }, + { + "epoch": 2.8, + "learning_rate": 2.263146939207861e-07, + "loss": 1.1397, + "step": 33286 + }, + { + "epoch": 2.8, + "learning_rate": 2.2612244673616112e-07, + "loss": 1.1323, + "step": 33287 + }, + { + "epoch": 2.8, + "learning_rate": 2.2593028030644116e-07, + "loss": 1.089, + "step": 33288 + }, + { + "epoch": 2.8, + "learning_rate": 2.2573819463321268e-07, + "loss": 1.0822, + "step": 33289 + }, + { + "epoch": 2.8, + "learning_rate": 2.2554618971806442e-07, + "loss": 1.0841, + "step": 33290 + }, + { + "epoch": 2.8, + "learning_rate": 2.2535426556258176e-07, + "loss": 1.1955, + "step": 33291 + }, + { + "epoch": 2.8, + "learning_rate": 2.2516242216834905e-07, + "loss": 1.0812, + "step": 33292 + }, + { + "epoch": 2.8, + "learning_rate": 2.2497065953695384e-07, + "loss": 1.1686, + "step": 33293 + }, + { + "epoch": 2.8, + "learning_rate": 2.2477897766997937e-07, + "loss": 1.177, + "step": 33294 + }, + { + "epoch": 2.8, + "learning_rate": 2.2458737656900766e-07, + "loss": 1.1303, + "step": 33295 + }, + { + "epoch": 2.8, + "learning_rate": 2.2439585623562633e-07, + "loss": 1.11, + "step": 33296 + }, + { + "epoch": 2.8, + "learning_rate": 2.2420441667141413e-07, + "loss": 1.0888, + "step": 33297 + }, + { + "epoch": 2.8, + "learning_rate": 2.2401305787795312e-07, + "loss": 1.1243, + "step": 33298 + }, + { + "epoch": 2.8, + "learning_rate": 2.238217798568254e-07, + "loss": 1.192, + "step": 33299 + }, + { + "epoch": 2.8, + "learning_rate": 2.2363058260960968e-07, + "loss": 1.1556, + "step": 33300 + }, + { + "epoch": 2.8, + "learning_rate": 2.2343946613788803e-07, + "loss": 1.0764, + "step": 33301 + }, + { + "epoch": 2.8, + "learning_rate": 2.2324843044323917e-07, + "loss": 1.0386, + "step": 33302 + }, + { + "epoch": 2.8, + "learning_rate": 2.2305747552723967e-07, + "loss": 1.072, + "step": 33303 + }, + { + "epoch": 2.8, + "learning_rate": 2.2286660139146931e-07, + "loss": 1.1126, + "step": 33304 + }, + { + "epoch": 2.8, + "learning_rate": 2.2267580803750354e-07, + "loss": 1.1097, + "step": 33305 + }, + { + "epoch": 2.8, + "learning_rate": 2.224850954669189e-07, + "loss": 1.2027, + "step": 33306 + }, + { + "epoch": 2.8, + "learning_rate": 2.2229446368129182e-07, + "loss": 1.093, + "step": 33307 + }, + { + "epoch": 2.8, + "learning_rate": 2.2210391268219667e-07, + "loss": 1.144, + "step": 33308 + }, + { + "epoch": 2.8, + "learning_rate": 2.2191344247120993e-07, + "loss": 1.0562, + "step": 33309 + }, + { + "epoch": 2.8, + "learning_rate": 2.2172305304990262e-07, + "loss": 1.0778, + "step": 33310 + }, + { + "epoch": 2.8, + "learning_rate": 2.2153274441984895e-07, + "loss": 1.1991, + "step": 33311 + }, + { + "epoch": 2.8, + "learning_rate": 2.2134251658262107e-07, + "loss": 1.1433, + "step": 33312 + }, + { + "epoch": 2.8, + "learning_rate": 2.2115236953979214e-07, + "loss": 1.1503, + "step": 33313 + }, + { + "epoch": 2.8, + "learning_rate": 2.209623032929309e-07, + "loss": 1.1513, + "step": 33314 + }, + { + "epoch": 2.8, + "learning_rate": 2.207723178436083e-07, + "loss": 1.1426, + "step": 33315 + }, + { + "epoch": 2.8, + "learning_rate": 2.2058241319339536e-07, + "loss": 1.1116, + "step": 33316 + }, + { + "epoch": 2.81, + "learning_rate": 2.2039258934385964e-07, + "loss": 1.0698, + "step": 33317 + }, + { + "epoch": 2.81, + "learning_rate": 2.2020284629657106e-07, + "loss": 1.059, + "step": 33318 + }, + { + "epoch": 2.81, + "learning_rate": 2.2001318405309724e-07, + "loss": 1.0582, + "step": 33319 + }, + { + "epoch": 2.81, + "learning_rate": 2.1982360261500247e-07, + "loss": 1.083, + "step": 33320 + }, + { + "epoch": 2.81, + "learning_rate": 2.196341019838555e-07, + "loss": 1.1331, + "step": 33321 + }, + { + "epoch": 2.81, + "learning_rate": 2.1944468216122282e-07, + "loss": 1.1197, + "step": 33322 + }, + { + "epoch": 2.81, + "learning_rate": 2.1925534314866658e-07, + "loss": 1.1453, + "step": 33323 + }, + { + "epoch": 2.81, + "learning_rate": 2.1906608494775549e-07, + "loss": 1.1725, + "step": 33324 + }, + { + "epoch": 2.81, + "learning_rate": 2.188769075600494e-07, + "loss": 1.126, + "step": 33325 + }, + { + "epoch": 2.81, + "learning_rate": 2.1868781098711267e-07, + "loss": 1.1619, + "step": 33326 + }, + { + "epoch": 2.81, + "learning_rate": 2.184987952305073e-07, + "loss": 1.1076, + "step": 33327 + }, + { + "epoch": 2.81, + "learning_rate": 2.1830986029179546e-07, + "loss": 1.1143, + "step": 33328 + }, + { + "epoch": 2.81, + "learning_rate": 2.1812100617253806e-07, + "loss": 1.0579, + "step": 33329 + }, + { + "epoch": 2.81, + "learning_rate": 2.179322328742972e-07, + "loss": 1.1175, + "step": 33330 + }, + { + "epoch": 2.81, + "learning_rate": 2.1774354039862834e-07, + "loss": 1.1591, + "step": 33331 + }, + { + "epoch": 2.81, + "learning_rate": 2.175549287470946e-07, + "loss": 1.1698, + "step": 33332 + }, + { + "epoch": 2.81, + "learning_rate": 2.1736639792125257e-07, + "loss": 1.1602, + "step": 33333 + }, + { + "epoch": 2.81, + "learning_rate": 2.171779479226599e-07, + "loss": 1.0964, + "step": 33334 + }, + { + "epoch": 2.81, + "learning_rate": 2.169895787528742e-07, + "loss": 1.2176, + "step": 33335 + }, + { + "epoch": 2.81, + "learning_rate": 2.16801290413452e-07, + "loss": 1.132, + "step": 33336 + }, + { + "epoch": 2.81, + "learning_rate": 2.1661308290594762e-07, + "loss": 1.1085, + "step": 33337 + }, + { + "epoch": 2.81, + "learning_rate": 2.1642495623191763e-07, + "loss": 1.121, + "step": 33338 + }, + { + "epoch": 2.81, + "learning_rate": 2.1623691039291516e-07, + "loss": 1.2154, + "step": 33339 + }, + { + "epoch": 2.81, + "learning_rate": 2.160489453904946e-07, + "loss": 1.1278, + "step": 33340 + }, + { + "epoch": 2.81, + "learning_rate": 2.158610612262102e-07, + "loss": 1.1319, + "step": 33341 + }, + { + "epoch": 2.81, + "learning_rate": 2.1567325790161187e-07, + "loss": 1.2042, + "step": 33342 + }, + { + "epoch": 2.81, + "learning_rate": 2.1548553541825278e-07, + "loss": 1.1053, + "step": 33343 + }, + { + "epoch": 2.81, + "learning_rate": 2.1529789377768396e-07, + "loss": 1.1236, + "step": 33344 + }, + { + "epoch": 2.81, + "learning_rate": 2.1511033298145412e-07, + "loss": 1.1333, + "step": 33345 + }, + { + "epoch": 2.81, + "learning_rate": 2.149228530311165e-07, + "loss": 1.0951, + "step": 33346 + }, + { + "epoch": 2.81, + "learning_rate": 2.1473545392821647e-07, + "loss": 1.1438, + "step": 33347 + }, + { + "epoch": 2.81, + "learning_rate": 2.1454813567430399e-07, + "loss": 1.1482, + "step": 33348 + }, + { + "epoch": 2.81, + "learning_rate": 2.1436089827092555e-07, + "loss": 1.0899, + "step": 33349 + }, + { + "epoch": 2.81, + "learning_rate": 2.1417374171963102e-07, + "loss": 1.042, + "step": 33350 + }, + { + "epoch": 2.81, + "learning_rate": 2.1398666602196362e-07, + "loss": 1.0661, + "step": 33351 + }, + { + "epoch": 2.81, + "learning_rate": 2.137996711794721e-07, + "loss": 1.1507, + "step": 33352 + }, + { + "epoch": 2.81, + "learning_rate": 2.1361275719369633e-07, + "loss": 1.112, + "step": 33353 + }, + { + "epoch": 2.81, + "learning_rate": 2.1342592406618735e-07, + "loss": 1.0427, + "step": 33354 + }, + { + "epoch": 2.81, + "learning_rate": 2.1323917179848386e-07, + "loss": 1.1557, + "step": 33355 + }, + { + "epoch": 2.81, + "learning_rate": 2.1305250039213132e-07, + "loss": 1.1968, + "step": 33356 + }, + { + "epoch": 2.81, + "learning_rate": 2.1286590984867072e-07, + "loss": 1.0718, + "step": 33357 + }, + { + "epoch": 2.81, + "learning_rate": 2.1267940016964417e-07, + "loss": 1.154, + "step": 33358 + }, + { + "epoch": 2.81, + "learning_rate": 2.1249297135659264e-07, + "loss": 1.1462, + "step": 33359 + }, + { + "epoch": 2.81, + "learning_rate": 2.1230662341105602e-07, + "loss": 1.1082, + "step": 33360 + }, + { + "epoch": 2.81, + "learning_rate": 2.121203563345764e-07, + "loss": 1.1473, + "step": 33361 + }, + { + "epoch": 2.81, + "learning_rate": 2.1193417012868922e-07, + "loss": 1.074, + "step": 33362 + }, + { + "epoch": 2.81, + "learning_rate": 2.117480647949355e-07, + "loss": 1.1283, + "step": 33363 + }, + { + "epoch": 2.81, + "learning_rate": 2.1156204033485172e-07, + "loss": 1.097, + "step": 33364 + }, + { + "epoch": 2.81, + "learning_rate": 2.113760967499745e-07, + "loss": 1.0333, + "step": 33365 + }, + { + "epoch": 2.81, + "learning_rate": 2.1119023404184147e-07, + "loss": 1.131, + "step": 33366 + }, + { + "epoch": 2.81, + "learning_rate": 2.1100445221198696e-07, + "loss": 1.0226, + "step": 33367 + }, + { + "epoch": 2.81, + "learning_rate": 2.108187512619464e-07, + "loss": 1.0862, + "step": 33368 + }, + { + "epoch": 2.81, + "learning_rate": 2.1063313119325523e-07, + "loss": 1.0583, + "step": 33369 + }, + { + "epoch": 2.81, + "learning_rate": 2.1044759200744448e-07, + "loss": 1.0563, + "step": 33370 + }, + { + "epoch": 2.81, + "learning_rate": 2.1026213370604954e-07, + "loss": 1.1573, + "step": 33371 + }, + { + "epoch": 2.81, + "learning_rate": 2.1007675629060363e-07, + "loss": 1.0984, + "step": 33372 + }, + { + "epoch": 2.81, + "learning_rate": 2.0989145976263449e-07, + "loss": 1.1093, + "step": 33373 + }, + { + "epoch": 2.81, + "learning_rate": 2.0970624412367636e-07, + "loss": 1.0888, + "step": 33374 + }, + { + "epoch": 2.81, + "learning_rate": 2.0952110937525917e-07, + "loss": 1.0617, + "step": 33375 + }, + { + "epoch": 2.81, + "learning_rate": 2.0933605551891057e-07, + "loss": 1.1632, + "step": 33376 + }, + { + "epoch": 2.81, + "learning_rate": 2.0915108255616157e-07, + "loss": 1.0459, + "step": 33377 + }, + { + "epoch": 2.81, + "learning_rate": 2.0896619048853983e-07, + "loss": 1.1171, + "step": 33378 + }, + { + "epoch": 2.81, + "learning_rate": 2.0878137931757303e-07, + "loss": 1.1209, + "step": 33379 + }, + { + "epoch": 2.81, + "learning_rate": 2.085966490447877e-07, + "loss": 1.1639, + "step": 33380 + }, + { + "epoch": 2.81, + "learning_rate": 2.0841199967171044e-07, + "loss": 1.0693, + "step": 33381 + }, + { + "epoch": 2.81, + "learning_rate": 2.0822743119986667e-07, + "loss": 1.0908, + "step": 33382 + }, + { + "epoch": 2.81, + "learning_rate": 2.0804294363078293e-07, + "loss": 1.1974, + "step": 33383 + }, + { + "epoch": 2.81, + "learning_rate": 2.0785853696598024e-07, + "loss": 1.1161, + "step": 33384 + }, + { + "epoch": 2.81, + "learning_rate": 2.0767421120698627e-07, + "loss": 0.9804, + "step": 33385 + }, + { + "epoch": 2.81, + "learning_rate": 2.0748996635532092e-07, + "loss": 1.1335, + "step": 33386 + }, + { + "epoch": 2.81, + "learning_rate": 2.0730580241250741e-07, + "loss": 1.0439, + "step": 33387 + }, + { + "epoch": 2.81, + "learning_rate": 2.0712171938006787e-07, + "loss": 1.1361, + "step": 33388 + }, + { + "epoch": 2.81, + "learning_rate": 2.0693771725952216e-07, + "loss": 1.0364, + "step": 33389 + }, + { + "epoch": 2.81, + "learning_rate": 2.0675379605239132e-07, + "loss": 1.0133, + "step": 33390 + }, + { + "epoch": 2.81, + "learning_rate": 2.0656995576019635e-07, + "loss": 1.0698, + "step": 33391 + }, + { + "epoch": 2.81, + "learning_rate": 2.0638619638445267e-07, + "loss": 1.1512, + "step": 33392 + }, + { + "epoch": 2.81, + "learning_rate": 2.0620251792668245e-07, + "loss": 1.1248, + "step": 33393 + }, + { + "epoch": 2.81, + "learning_rate": 2.0601892038839998e-07, + "loss": 1.094, + "step": 33394 + }, + { + "epoch": 2.81, + "learning_rate": 2.05835403771123e-07, + "loss": 1.0733, + "step": 33395 + }, + { + "epoch": 2.81, + "learning_rate": 2.0565196807637022e-07, + "loss": 1.0737, + "step": 33396 + }, + { + "epoch": 2.81, + "learning_rate": 2.0546861330565493e-07, + "loss": 1.1975, + "step": 33397 + }, + { + "epoch": 2.81, + "learning_rate": 2.0528533946049257e-07, + "loss": 1.1205, + "step": 33398 + }, + { + "epoch": 2.81, + "learning_rate": 2.0510214654239747e-07, + "loss": 1.1555, + "step": 33399 + }, + { + "epoch": 2.81, + "learning_rate": 2.0491903455288286e-07, + "loss": 1.1192, + "step": 33400 + }, + { + "epoch": 2.81, + "learning_rate": 2.0473600349346202e-07, + "loss": 1.1105, + "step": 33401 + }, + { + "epoch": 2.81, + "learning_rate": 2.0455305336564813e-07, + "loss": 1.0623, + "step": 33402 + }, + { + "epoch": 2.81, + "learning_rate": 2.0437018417095e-07, + "loss": 1.1471, + "step": 33403 + }, + { + "epoch": 2.81, + "learning_rate": 2.04187395910882e-07, + "loss": 1.1093, + "step": 33404 + }, + { + "epoch": 2.81, + "learning_rate": 2.0400468858695287e-07, + "loss": 1.1506, + "step": 33405 + }, + { + "epoch": 2.81, + "learning_rate": 2.0382206220067147e-07, + "loss": 1.1768, + "step": 33406 + }, + { + "epoch": 2.81, + "learning_rate": 2.0363951675354765e-07, + "loss": 1.1123, + "step": 33407 + }, + { + "epoch": 2.81, + "learning_rate": 2.0345705224708912e-07, + "loss": 1.0041, + "step": 33408 + }, + { + "epoch": 2.81, + "learning_rate": 2.0327466868280354e-07, + "loss": 1.1427, + "step": 33409 + }, + { + "epoch": 2.81, + "learning_rate": 2.0309236606219862e-07, + "loss": 1.1128, + "step": 33410 + }, + { + "epoch": 2.81, + "learning_rate": 2.029101443867798e-07, + "loss": 1.0843, + "step": 33411 + }, + { + "epoch": 2.81, + "learning_rate": 2.0272800365805145e-07, + "loss": 0.954, + "step": 33412 + }, + { + "epoch": 2.81, + "learning_rate": 2.0254594387752235e-07, + "loss": 1.1282, + "step": 33413 + }, + { + "epoch": 2.81, + "learning_rate": 2.0236396504669243e-07, + "loss": 1.046, + "step": 33414 + }, + { + "epoch": 2.81, + "learning_rate": 2.0218206716706823e-07, + "loss": 1.0997, + "step": 33415 + }, + { + "epoch": 2.81, + "learning_rate": 2.020002502401508e-07, + "loss": 1.1108, + "step": 33416 + }, + { + "epoch": 2.81, + "learning_rate": 2.0181851426744337e-07, + "loss": 1.1538, + "step": 33417 + }, + { + "epoch": 2.81, + "learning_rate": 2.0163685925044696e-07, + "loss": 1.1824, + "step": 33418 + }, + { + "epoch": 2.81, + "learning_rate": 2.014552851906626e-07, + "loss": 1.1387, + "step": 33419 + }, + { + "epoch": 2.81, + "learning_rate": 2.012737920895902e-07, + "loss": 1.1645, + "step": 33420 + }, + { + "epoch": 2.81, + "learning_rate": 2.0109237994873078e-07, + "loss": 1.1163, + "step": 33421 + }, + { + "epoch": 2.81, + "learning_rate": 2.00911048769582e-07, + "loss": 1.1897, + "step": 33422 + }, + { + "epoch": 2.81, + "learning_rate": 2.007297985536405e-07, + "loss": 1.1514, + "step": 33423 + }, + { + "epoch": 2.81, + "learning_rate": 2.0054862930240837e-07, + "loss": 1.0148, + "step": 33424 + }, + { + "epoch": 2.81, + "learning_rate": 2.0036754101737777e-07, + "loss": 1.093, + "step": 33425 + }, + { + "epoch": 2.81, + "learning_rate": 2.0018653370004747e-07, + "loss": 1.208, + "step": 33426 + }, + { + "epoch": 2.81, + "learning_rate": 2.0000560735191188e-07, + "loss": 1.1696, + "step": 33427 + }, + { + "epoch": 2.81, + "learning_rate": 1.9982476197446753e-07, + "loss": 1.1303, + "step": 33428 + }, + { + "epoch": 2.81, + "learning_rate": 1.9964399756920662e-07, + "loss": 1.1295, + "step": 33429 + }, + { + "epoch": 2.81, + "learning_rate": 1.9946331413762344e-07, + "loss": 1.0731, + "step": 33430 + }, + { + "epoch": 2.81, + "learning_rate": 1.9928271168121016e-07, + "loss": 1.0384, + "step": 33431 + }, + { + "epoch": 2.81, + "learning_rate": 1.9910219020146115e-07, + "loss": 1.1179, + "step": 33432 + }, + { + "epoch": 2.81, + "learning_rate": 1.9892174969986632e-07, + "loss": 1.116, + "step": 33433 + }, + { + "epoch": 2.81, + "learning_rate": 1.987413901779156e-07, + "loss": 1.022, + "step": 33434 + }, + { + "epoch": 2.81, + "learning_rate": 1.9856111163710224e-07, + "loss": 1.1416, + "step": 33435 + }, + { + "epoch": 2.82, + "learning_rate": 1.983809140789128e-07, + "loss": 1.0529, + "step": 33436 + }, + { + "epoch": 2.82, + "learning_rate": 1.9820079750483722e-07, + "loss": 1.1473, + "step": 33437 + }, + { + "epoch": 2.82, + "learning_rate": 1.980207619163632e-07, + "loss": 1.1401, + "step": 33438 + }, + { + "epoch": 2.82, + "learning_rate": 1.9784080731497846e-07, + "loss": 1.1524, + "step": 33439 + }, + { + "epoch": 2.82, + "learning_rate": 1.9766093370217065e-07, + "loss": 1.1467, + "step": 33440 + }, + { + "epoch": 2.82, + "learning_rate": 1.974811410794264e-07, + "loss": 1.098, + "step": 33441 + }, + { + "epoch": 2.82, + "learning_rate": 1.9730142944822783e-07, + "loss": 1.0364, + "step": 33442 + }, + { + "epoch": 2.82, + "learning_rate": 1.9712179881006487e-07, + "loss": 1.0381, + "step": 33443 + }, + { + "epoch": 2.82, + "learning_rate": 1.9694224916641746e-07, + "loss": 1.0992, + "step": 33444 + }, + { + "epoch": 2.82, + "learning_rate": 1.9676278051876884e-07, + "loss": 1.1143, + "step": 33445 + }, + { + "epoch": 2.82, + "learning_rate": 1.9658339286860672e-07, + "loss": 1.0566, + "step": 33446 + }, + { + "epoch": 2.82, + "learning_rate": 1.964040862174088e-07, + "loss": 1.1321, + "step": 33447 + }, + { + "epoch": 2.82, + "learning_rate": 1.9622486056665723e-07, + "loss": 1.1393, + "step": 33448 + }, + { + "epoch": 2.82, + "learning_rate": 1.9604571591783416e-07, + "loss": 1.1372, + "step": 33449 + }, + { + "epoch": 2.82, + "learning_rate": 1.9586665227241842e-07, + "loss": 1.0912, + "step": 33450 + }, + { + "epoch": 2.82, + "learning_rate": 1.9568766963189101e-07, + "loss": 1.0869, + "step": 33451 + }, + { + "epoch": 2.82, + "learning_rate": 1.9550876799772967e-07, + "loss": 1.1656, + "step": 33452 + }, + { + "epoch": 2.82, + "learning_rate": 1.953299473714121e-07, + "loss": 1.198, + "step": 33453 + }, + { + "epoch": 2.82, + "learning_rate": 1.951512077544171e-07, + "loss": 1.1246, + "step": 33454 + }, + { + "epoch": 2.82, + "learning_rate": 1.9497254914822016e-07, + "loss": 1.0881, + "step": 33455 + }, + { + "epoch": 2.82, + "learning_rate": 1.947939715542979e-07, + "loss": 1.0575, + "step": 33456 + }, + { + "epoch": 2.82, + "learning_rate": 1.946154749741269e-07, + "loss": 1.0681, + "step": 33457 + }, + { + "epoch": 2.82, + "learning_rate": 1.9443705940918045e-07, + "loss": 0.9743, + "step": 33458 + }, + { + "epoch": 2.82, + "learning_rate": 1.942587248609329e-07, + "loss": 1.1503, + "step": 33459 + }, + { + "epoch": 2.82, + "learning_rate": 1.940804713308586e-07, + "loss": 1.1581, + "step": 33460 + }, + { + "epoch": 2.82, + "learning_rate": 1.9390229882042977e-07, + "loss": 1.0405, + "step": 33461 + }, + { + "epoch": 2.82, + "learning_rate": 1.937242073311174e-07, + "loss": 1.1334, + "step": 33462 + }, + { + "epoch": 2.82, + "learning_rate": 1.935461968643948e-07, + "loss": 1.1046, + "step": 33463 + }, + { + "epoch": 2.82, + "learning_rate": 1.9336826742173186e-07, + "loss": 1.1277, + "step": 33464 + }, + { + "epoch": 2.82, + "learning_rate": 1.9319041900459966e-07, + "loss": 1.1744, + "step": 33465 + }, + { + "epoch": 2.82, + "learning_rate": 1.9301265161446593e-07, + "loss": 1.0661, + "step": 33466 + }, + { + "epoch": 2.82, + "learning_rate": 1.9283496525280055e-07, + "loss": 1.0504, + "step": 33467 + }, + { + "epoch": 2.82, + "learning_rate": 1.9265735992107126e-07, + "loss": 1.0799, + "step": 33468 + }, + { + "epoch": 2.82, + "learning_rate": 1.924798356207458e-07, + "loss": 1.1856, + "step": 33469 + }, + { + "epoch": 2.82, + "learning_rate": 1.9230239235328962e-07, + "loss": 1.1942, + "step": 33470 + }, + { + "epoch": 2.82, + "learning_rate": 1.9212503012017048e-07, + "loss": 1.1157, + "step": 33471 + }, + { + "epoch": 2.82, + "learning_rate": 1.9194774892285495e-07, + "loss": 1.091, + "step": 33472 + }, + { + "epoch": 2.82, + "learning_rate": 1.917705487628041e-07, + "loss": 1.163, + "step": 33473 + }, + { + "epoch": 2.82, + "learning_rate": 1.9159342964148564e-07, + "loss": 1.0866, + "step": 33474 + }, + { + "epoch": 2.82, + "learning_rate": 1.914163915603595e-07, + "loss": 1.119, + "step": 33475 + }, + { + "epoch": 2.82, + "learning_rate": 1.9123943452089122e-07, + "loss": 1.1851, + "step": 33476 + }, + { + "epoch": 2.82, + "learning_rate": 1.910625585245418e-07, + "loss": 1.1172, + "step": 33477 + }, + { + "epoch": 2.82, + "learning_rate": 1.9088576357277344e-07, + "loss": 1.1088, + "step": 33478 + }, + { + "epoch": 2.82, + "learning_rate": 1.9070904966704605e-07, + "loss": 1.0797, + "step": 33479 + }, + { + "epoch": 2.82, + "learning_rate": 1.9053241680881961e-07, + "loss": 1.0725, + "step": 33480 + }, + { + "epoch": 2.82, + "learning_rate": 1.9035586499955295e-07, + "loss": 1.1261, + "step": 33481 + }, + { + "epoch": 2.82, + "learning_rate": 1.9017939424070598e-07, + "loss": 1.2503, + "step": 33482 + }, + { + "epoch": 2.82, + "learning_rate": 1.9000300453373753e-07, + "loss": 1.0311, + "step": 33483 + }, + { + "epoch": 2.82, + "learning_rate": 1.8982669588010204e-07, + "loss": 1.1677, + "step": 33484 + }, + { + "epoch": 2.82, + "learning_rate": 1.8965046828126056e-07, + "loss": 1.1608, + "step": 33485 + }, + { + "epoch": 2.82, + "learning_rate": 1.894743217386652e-07, + "loss": 1.1489, + "step": 33486 + }, + { + "epoch": 2.82, + "learning_rate": 1.8929825625377264e-07, + "loss": 1.1881, + "step": 33487 + }, + { + "epoch": 2.82, + "learning_rate": 1.8912227182803722e-07, + "loss": 1.2103, + "step": 33488 + }, + { + "epoch": 2.82, + "learning_rate": 1.8894636846291448e-07, + "loss": 1.1769, + "step": 33489 + }, + { + "epoch": 2.82, + "learning_rate": 1.8877054615985657e-07, + "loss": 1.1219, + "step": 33490 + }, + { + "epoch": 2.82, + "learning_rate": 1.885948049203168e-07, + "loss": 1.1577, + "step": 33491 + }, + { + "epoch": 2.82, + "learning_rate": 1.8841914474574507e-07, + "loss": 1.1264, + "step": 33492 + }, + { + "epoch": 2.82, + "learning_rate": 1.8824356563759694e-07, + "loss": 1.0352, + "step": 33493 + }, + { + "epoch": 2.82, + "learning_rate": 1.88068067597319e-07, + "loss": 1.0194, + "step": 33494 + }, + { + "epoch": 2.82, + "learning_rate": 1.878926506263623e-07, + "loss": 1.1082, + "step": 33495 + }, + { + "epoch": 2.82, + "learning_rate": 1.8771731472617793e-07, + "loss": 1.0172, + "step": 33496 + }, + { + "epoch": 2.82, + "learning_rate": 1.8754205989821362e-07, + "loss": 1.1065, + "step": 33497 + }, + { + "epoch": 2.82, + "learning_rate": 1.8736688614391706e-07, + "loss": 1.1697, + "step": 33498 + }, + { + "epoch": 2.82, + "learning_rate": 1.8719179346473494e-07, + "loss": 1.1062, + "step": 33499 + }, + { + "epoch": 2.82, + "learning_rate": 1.870167818621149e-07, + "loss": 1.1864, + "step": 33500 + }, + { + "epoch": 2.82, + "learning_rate": 1.8684185133750365e-07, + "loss": 1.0903, + "step": 33501 + }, + { + "epoch": 2.82, + "learning_rate": 1.8666700189234555e-07, + "loss": 1.1562, + "step": 33502 + }, + { + "epoch": 2.82, + "learning_rate": 1.8649223352808388e-07, + "loss": 1.0691, + "step": 33503 + }, + { + "epoch": 2.82, + "learning_rate": 1.863175462461664e-07, + "loss": 1.0285, + "step": 33504 + }, + { + "epoch": 2.82, + "learning_rate": 1.861429400480319e-07, + "loss": 1.1282, + "step": 33505 + }, + { + "epoch": 2.82, + "learning_rate": 1.8596841493512708e-07, + "loss": 1.1101, + "step": 33506 + }, + { + "epoch": 2.82, + "learning_rate": 1.8579397090889073e-07, + "loss": 1.1021, + "step": 33507 + }, + { + "epoch": 2.82, + "learning_rate": 1.8561960797076616e-07, + "loss": 1.0803, + "step": 33508 + }, + { + "epoch": 2.82, + "learning_rate": 1.8544532612219334e-07, + "loss": 1.1636, + "step": 33509 + }, + { + "epoch": 2.82, + "learning_rate": 1.8527112536461224e-07, + "loss": 1.1128, + "step": 33510 + }, + { + "epoch": 2.82, + "learning_rate": 1.8509700569946276e-07, + "loss": 1.1013, + "step": 33511 + }, + { + "epoch": 2.82, + "learning_rate": 1.8492296712818158e-07, + "loss": 1.103, + "step": 33512 + }, + { + "epoch": 2.82, + "learning_rate": 1.8474900965220977e-07, + "loss": 1.0435, + "step": 33513 + }, + { + "epoch": 2.82, + "learning_rate": 1.845751332729817e-07, + "loss": 1.0905, + "step": 33514 + }, + { + "epoch": 2.82, + "learning_rate": 1.8440133799193627e-07, + "loss": 1.1667, + "step": 33515 + }, + { + "epoch": 2.82, + "learning_rate": 1.8422762381050675e-07, + "loss": 1.1193, + "step": 33516 + }, + { + "epoch": 2.82, + "learning_rate": 1.8405399073013198e-07, + "loss": 1.1259, + "step": 33517 + }, + { + "epoch": 2.82, + "learning_rate": 1.8388043875224303e-07, + "loss": 1.1024, + "step": 33518 + }, + { + "epoch": 2.82, + "learning_rate": 1.8370696787827658e-07, + "loss": 1.1326, + "step": 33519 + }, + { + "epoch": 2.82, + "learning_rate": 1.8353357810966365e-07, + "loss": 1.0962, + "step": 33520 + }, + { + "epoch": 2.82, + "learning_rate": 1.833602694478387e-07, + "loss": 1.1218, + "step": 33521 + }, + { + "epoch": 2.82, + "learning_rate": 1.8318704189423386e-07, + "loss": 1.074, + "step": 33522 + }, + { + "epoch": 2.82, + "learning_rate": 1.8301389545027693e-07, + "loss": 1.0927, + "step": 33523 + }, + { + "epoch": 2.82, + "learning_rate": 1.8284083011740338e-07, + "loss": 1.1078, + "step": 33524 + }, + { + "epoch": 2.82, + "learning_rate": 1.826678458970399e-07, + "loss": 1.0834, + "step": 33525 + }, + { + "epoch": 2.82, + "learning_rate": 1.8249494279061753e-07, + "loss": 1.1102, + "step": 33526 + }, + { + "epoch": 2.82, + "learning_rate": 1.823221207995629e-07, + "loss": 1.1221, + "step": 33527 + }, + { + "epoch": 2.82, + "learning_rate": 1.821493799253049e-07, + "loss": 0.9856, + "step": 33528 + }, + { + "epoch": 2.82, + "learning_rate": 1.8197672016927126e-07, + "loss": 1.0429, + "step": 33529 + }, + { + "epoch": 2.82, + "learning_rate": 1.818041415328886e-07, + "loss": 1.0638, + "step": 33530 + }, + { + "epoch": 2.82, + "learning_rate": 1.816316440175814e-07, + "loss": 1.1418, + "step": 33531 + }, + { + "epoch": 2.82, + "learning_rate": 1.814592276247751e-07, + "loss": 1.0878, + "step": 33532 + }, + { + "epoch": 2.82, + "learning_rate": 1.8128689235589747e-07, + "loss": 1.1322, + "step": 33533 + }, + { + "epoch": 2.82, + "learning_rate": 1.8111463821236741e-07, + "loss": 1.0709, + "step": 33534 + }, + { + "epoch": 2.82, + "learning_rate": 1.8094246519561265e-07, + "loss": 1.151, + "step": 33535 + }, + { + "epoch": 2.82, + "learning_rate": 1.8077037330705315e-07, + "loss": 1.0871, + "step": 33536 + }, + { + "epoch": 2.82, + "learning_rate": 1.8059836254811113e-07, + "loss": 1.1825, + "step": 33537 + }, + { + "epoch": 2.82, + "learning_rate": 1.8042643292020768e-07, + "loss": 1.207, + "step": 33538 + }, + { + "epoch": 2.82, + "learning_rate": 1.8025458442476386e-07, + "loss": 1.1321, + "step": 33539 + }, + { + "epoch": 2.82, + "learning_rate": 1.8008281706319963e-07, + "loss": 1.0008, + "step": 33540 + }, + { + "epoch": 2.82, + "learning_rate": 1.7991113083693501e-07, + "loss": 1.0863, + "step": 33541 + }, + { + "epoch": 2.82, + "learning_rate": 1.7973952574738552e-07, + "loss": 1.1573, + "step": 33542 + }, + { + "epoch": 2.82, + "learning_rate": 1.7956800179597223e-07, + "loss": 1.0974, + "step": 33543 + }, + { + "epoch": 2.82, + "learning_rate": 1.7939655898411068e-07, + "loss": 1.1554, + "step": 33544 + }, + { + "epoch": 2.82, + "learning_rate": 1.7922519731321642e-07, + "loss": 1.0549, + "step": 33545 + }, + { + "epoch": 2.82, + "learning_rate": 1.7905391678470829e-07, + "loss": 1.1048, + "step": 33546 + }, + { + "epoch": 2.82, + "learning_rate": 1.7888271739999962e-07, + "loss": 1.1661, + "step": 33547 + }, + { + "epoch": 2.82, + "learning_rate": 1.7871159916050485e-07, + "loss": 1.0933, + "step": 33548 + }, + { + "epoch": 2.82, + "learning_rate": 1.7854056206763837e-07, + "loss": 1.0918, + "step": 33549 + }, + { + "epoch": 2.82, + "learning_rate": 1.783696061228124e-07, + "loss": 1.0485, + "step": 33550 + }, + { + "epoch": 2.82, + "learning_rate": 1.7819873132744025e-07, + "loss": 1.0033, + "step": 33551 + }, + { + "epoch": 2.82, + "learning_rate": 1.7802793768293413e-07, + "loss": 1.1685, + "step": 33552 + }, + { + "epoch": 2.82, + "learning_rate": 1.7785722519070404e-07, + "loss": 1.0848, + "step": 33553 + }, + { + "epoch": 2.82, + "learning_rate": 1.7768659385216213e-07, + "loss": 1.1384, + "step": 33554 + }, + { + "epoch": 2.83, + "learning_rate": 1.7751604366871622e-07, + "loss": 1.1161, + "step": 33555 + }, + { + "epoch": 2.83, + "learning_rate": 1.7734557464177627e-07, + "loss": 1.1103, + "step": 33556 + }, + { + "epoch": 2.83, + "learning_rate": 1.7717518677275002e-07, + "loss": 1.1317, + "step": 33557 + }, + { + "epoch": 2.83, + "learning_rate": 1.770048800630475e-07, + "loss": 1.0955, + "step": 33558 + }, + { + "epoch": 2.83, + "learning_rate": 1.7683465451407422e-07, + "loss": 1.1044, + "step": 33559 + }, + { + "epoch": 2.83, + "learning_rate": 1.7666451012723573e-07, + "loss": 1.1975, + "step": 33560 + }, + { + "epoch": 2.83, + "learning_rate": 1.764944469039409e-07, + "loss": 1.1696, + "step": 33561 + }, + { + "epoch": 2.83, + "learning_rate": 1.7632446484559085e-07, + "loss": 1.1406, + "step": 33562 + }, + { + "epoch": 2.83, + "learning_rate": 1.7615456395359333e-07, + "loss": 1.0826, + "step": 33563 + }, + { + "epoch": 2.83, + "learning_rate": 1.7598474422935052e-07, + "loss": 1.0975, + "step": 33564 + }, + { + "epoch": 2.83, + "learning_rate": 1.758150056742658e-07, + "loss": 1.0644, + "step": 33565 + }, + { + "epoch": 2.83, + "learning_rate": 1.7564534828974133e-07, + "loss": 1.2037, + "step": 33566 + }, + { + "epoch": 2.83, + "learning_rate": 1.7547577207717936e-07, + "loss": 1.0642, + "step": 33567 + }, + { + "epoch": 2.83, + "learning_rate": 1.7530627703798098e-07, + "loss": 1.1014, + "step": 33568 + }, + { + "epoch": 2.83, + "learning_rate": 1.7513686317354728e-07, + "loss": 1.2164, + "step": 33569 + }, + { + "epoch": 2.83, + "learning_rate": 1.7496753048527494e-07, + "loss": 1.0459, + "step": 33570 + }, + { + "epoch": 2.83, + "learning_rate": 1.7479827897456615e-07, + "loss": 1.1487, + "step": 33571 + }, + { + "epoch": 2.83, + "learning_rate": 1.7462910864281978e-07, + "loss": 1.193, + "step": 33572 + }, + { + "epoch": 2.83, + "learning_rate": 1.744600194914303e-07, + "loss": 1.1516, + "step": 33573 + }, + { + "epoch": 2.83, + "learning_rate": 1.7429101152179884e-07, + "loss": 1.1227, + "step": 33574 + }, + { + "epoch": 2.83, + "learning_rate": 1.7412208473531866e-07, + "loss": 1.1825, + "step": 33575 + }, + { + "epoch": 2.83, + "learning_rate": 1.7395323913338648e-07, + "loss": 1.1235, + "step": 33576 + }, + { + "epoch": 2.83, + "learning_rate": 1.737844747173967e-07, + "loss": 0.9901, + "step": 33577 + }, + { + "epoch": 2.83, + "learning_rate": 1.7361579148874597e-07, + "loss": 1.0854, + "step": 33578 + }, + { + "epoch": 2.83, + "learning_rate": 1.7344718944882434e-07, + "loss": 1.1544, + "step": 33579 + }, + { + "epoch": 2.83, + "learning_rate": 1.7327866859902952e-07, + "loss": 1.0932, + "step": 33580 + }, + { + "epoch": 2.83, + "learning_rate": 1.7311022894074825e-07, + "loss": 1.048, + "step": 33581 + }, + { + "epoch": 2.83, + "learning_rate": 1.7294187047537714e-07, + "loss": 1.1615, + "step": 33582 + }, + { + "epoch": 2.83, + "learning_rate": 1.727735932043062e-07, + "loss": 1.1523, + "step": 33583 + }, + { + "epoch": 2.83, + "learning_rate": 1.726053971289221e-07, + "loss": 1.0945, + "step": 33584 + }, + { + "epoch": 2.83, + "learning_rate": 1.724372822506204e-07, + "loss": 1.1384, + "step": 33585 + }, + { + "epoch": 2.83, + "learning_rate": 1.7226924857078553e-07, + "loss": 1.1701, + "step": 33586 + }, + { + "epoch": 2.83, + "learning_rate": 1.721012960908075e-07, + "loss": 1.1229, + "step": 33587 + }, + { + "epoch": 2.83, + "learning_rate": 1.7193342481207408e-07, + "loss": 1.1867, + "step": 33588 + }, + { + "epoch": 2.83, + "learning_rate": 1.7176563473597197e-07, + "loss": 1.0785, + "step": 33589 + }, + { + "epoch": 2.83, + "learning_rate": 1.7159792586388778e-07, + "loss": 0.9757, + "step": 33590 + }, + { + "epoch": 2.83, + "learning_rate": 1.7143029819720824e-07, + "loss": 1.0946, + "step": 33591 + }, + { + "epoch": 2.83, + "learning_rate": 1.7126275173731443e-07, + "loss": 1.1379, + "step": 33592 + }, + { + "epoch": 2.83, + "learning_rate": 1.7109528648559527e-07, + "loss": 1.2163, + "step": 33593 + }, + { + "epoch": 2.83, + "learning_rate": 1.7092790244343183e-07, + "loss": 1.1231, + "step": 33594 + }, + { + "epoch": 2.83, + "learning_rate": 1.7076059961220747e-07, + "loss": 1.134, + "step": 33595 + }, + { + "epoch": 2.83, + "learning_rate": 1.7059337799330555e-07, + "loss": 1.1011, + "step": 33596 + }, + { + "epoch": 2.83, + "learning_rate": 1.7042623758810716e-07, + "loss": 1.0266, + "step": 33597 + }, + { + "epoch": 2.83, + "learning_rate": 1.7025917839799233e-07, + "loss": 1.1685, + "step": 33598 + }, + { + "epoch": 2.83, + "learning_rate": 1.7009220042434217e-07, + "loss": 1.1355, + "step": 33599 + }, + { + "epoch": 2.83, + "learning_rate": 1.6992530366853666e-07, + "loss": 1.1973, + "step": 33600 + }, + { + "epoch": 2.83, + "learning_rate": 1.6975848813195363e-07, + "loss": 1.1078, + "step": 33601 + }, + { + "epoch": 2.83, + "learning_rate": 1.6959175381597305e-07, + "loss": 1.1321, + "step": 33602 + }, + { + "epoch": 2.83, + "learning_rate": 1.6942510072197048e-07, + "loss": 1.1265, + "step": 33603 + }, + { + "epoch": 2.83, + "learning_rate": 1.6925852885132487e-07, + "loss": 1.0236, + "step": 33604 + }, + { + "epoch": 2.83, + "learning_rate": 1.6909203820541065e-07, + "loss": 1.0582, + "step": 33605 + }, + { + "epoch": 2.83, + "learning_rate": 1.6892562878560558e-07, + "loss": 1.0294, + "step": 33606 + }, + { + "epoch": 2.83, + "learning_rate": 1.6875930059328193e-07, + "loss": 1.1059, + "step": 33607 + }, + { + "epoch": 2.83, + "learning_rate": 1.6859305362981525e-07, + "loss": 1.139, + "step": 33608 + }, + { + "epoch": 2.83, + "learning_rate": 1.6842688789658002e-07, + "loss": 1.1061, + "step": 33609 + }, + { + "epoch": 2.83, + "learning_rate": 1.6826080339494844e-07, + "loss": 1.0602, + "step": 33610 + }, + { + "epoch": 2.83, + "learning_rate": 1.6809480012629165e-07, + "loss": 1.1587, + "step": 33611 + }, + { + "epoch": 2.83, + "learning_rate": 1.67928878091983e-07, + "loss": 1.1462, + "step": 33612 + }, + { + "epoch": 2.83, + "learning_rate": 1.677630372933936e-07, + "loss": 1.1635, + "step": 33613 + }, + { + "epoch": 2.83, + "learning_rate": 1.6759727773189127e-07, + "loss": 1.1895, + "step": 33614 + }, + { + "epoch": 2.83, + "learning_rate": 1.6743159940884712e-07, + "loss": 1.1525, + "step": 33615 + }, + { + "epoch": 2.83, + "learning_rate": 1.6726600232563117e-07, + "loss": 1.0575, + "step": 33616 + }, + { + "epoch": 2.83, + "learning_rate": 1.6710048648360898e-07, + "loss": 1.0894, + "step": 33617 + }, + { + "epoch": 2.83, + "learning_rate": 1.6693505188414949e-07, + "loss": 1.0857, + "step": 33618 + }, + { + "epoch": 2.83, + "learning_rate": 1.6676969852862047e-07, + "loss": 1.0559, + "step": 33619 + }, + { + "epoch": 2.83, + "learning_rate": 1.6660442641838638e-07, + "loss": 1.1234, + "step": 33620 + }, + { + "epoch": 2.83, + "learning_rate": 1.664392355548139e-07, + "loss": 1.1168, + "step": 33621 + }, + { + "epoch": 2.83, + "learning_rate": 1.6627412593926862e-07, + "loss": 1.1438, + "step": 33622 + }, + { + "epoch": 2.83, + "learning_rate": 1.6610909757311165e-07, + "loss": 1.0321, + "step": 33623 + }, + { + "epoch": 2.83, + "learning_rate": 1.6594415045771083e-07, + "loss": 1.1144, + "step": 33624 + }, + { + "epoch": 2.83, + "learning_rate": 1.6577928459442505e-07, + "loss": 1.0956, + "step": 33625 + }, + { + "epoch": 2.83, + "learning_rate": 1.6561449998461876e-07, + "loss": 1.0831, + "step": 33626 + }, + { + "epoch": 2.83, + "learning_rate": 1.654497966296531e-07, + "loss": 1.0207, + "step": 33627 + }, + { + "epoch": 2.83, + "learning_rate": 1.652851745308881e-07, + "loss": 1.1107, + "step": 33628 + }, + { + "epoch": 2.83, + "learning_rate": 1.6512063368968488e-07, + "loss": 1.0699, + "step": 33629 + }, + { + "epoch": 2.83, + "learning_rate": 1.649561741074035e-07, + "loss": 1.3267, + "step": 33630 + }, + { + "epoch": 2.83, + "learning_rate": 1.6479179578540061e-07, + "loss": 1.0959, + "step": 33631 + }, + { + "epoch": 2.83, + "learning_rate": 1.6462749872503515e-07, + "loss": 1.1425, + "step": 33632 + }, + { + "epoch": 2.83, + "learning_rate": 1.64463282927666e-07, + "loss": 1.1289, + "step": 33633 + }, + { + "epoch": 2.83, + "learning_rate": 1.6429914839464879e-07, + "loss": 1.1502, + "step": 33634 + }, + { + "epoch": 2.83, + "learning_rate": 1.6413509512734016e-07, + "loss": 1.1859, + "step": 33635 + }, + { + "epoch": 2.83, + "learning_rate": 1.6397112312709462e-07, + "loss": 1.1011, + "step": 33636 + }, + { + "epoch": 2.83, + "learning_rate": 1.6380723239526774e-07, + "loss": 1.0847, + "step": 33637 + }, + { + "epoch": 2.83, + "learning_rate": 1.6364342293321289e-07, + "loss": 1.1277, + "step": 33638 + }, + { + "epoch": 2.83, + "learning_rate": 1.6347969474228563e-07, + "loss": 1.0918, + "step": 33639 + }, + { + "epoch": 2.83, + "learning_rate": 1.6331604782383604e-07, + "loss": 1.0905, + "step": 33640 + }, + { + "epoch": 2.83, + "learning_rate": 1.6315248217921852e-07, + "loss": 1.0444, + "step": 33641 + }, + { + "epoch": 2.83, + "learning_rate": 1.6298899780978205e-07, + "loss": 1.1219, + "step": 33642 + }, + { + "epoch": 2.83, + "learning_rate": 1.6282559471687997e-07, + "loss": 1.0664, + "step": 33643 + }, + { + "epoch": 2.83, + "learning_rate": 1.626622729018601e-07, + "loss": 1.1177, + "step": 33644 + }, + { + "epoch": 2.83, + "learning_rate": 1.6249903236607246e-07, + "loss": 1.0968, + "step": 33645 + }, + { + "epoch": 2.83, + "learning_rate": 1.6233587311086708e-07, + "loss": 1.1715, + "step": 33646 + }, + { + "epoch": 2.83, + "learning_rate": 1.6217279513759177e-07, + "loss": 1.1134, + "step": 33647 + }, + { + "epoch": 2.83, + "learning_rate": 1.6200979844759213e-07, + "loss": 1.1232, + "step": 33648 + }, + { + "epoch": 2.83, + "learning_rate": 1.6184688304221597e-07, + "loss": 1.105, + "step": 33649 + }, + { + "epoch": 2.83, + "learning_rate": 1.6168404892280997e-07, + "loss": 1.0697, + "step": 33650 + }, + { + "epoch": 2.83, + "learning_rate": 1.6152129609071977e-07, + "loss": 1.1696, + "step": 33651 + }, + { + "epoch": 2.83, + "learning_rate": 1.613586245472898e-07, + "loss": 1.1324, + "step": 33652 + }, + { + "epoch": 2.83, + "learning_rate": 1.6119603429386122e-07, + "loss": 1.1522, + "step": 33653 + }, + { + "epoch": 2.83, + "learning_rate": 1.6103352533178186e-07, + "loss": 1.1365, + "step": 33654 + }, + { + "epoch": 2.83, + "learning_rate": 1.6087109766239172e-07, + "loss": 1.1722, + "step": 33655 + }, + { + "epoch": 2.83, + "learning_rate": 1.607087512870331e-07, + "loss": 1.0996, + "step": 33656 + }, + { + "epoch": 2.83, + "learning_rate": 1.6054648620704826e-07, + "loss": 1.0787, + "step": 33657 + }, + { + "epoch": 2.83, + "learning_rate": 1.6038430242377723e-07, + "loss": 1.1509, + "step": 33658 + }, + { + "epoch": 2.83, + "learning_rate": 1.6022219993856004e-07, + "loss": 1.1783, + "step": 33659 + }, + { + "epoch": 2.83, + "learning_rate": 1.6006017875273672e-07, + "loss": 1.177, + "step": 33660 + }, + { + "epoch": 2.83, + "learning_rate": 1.59898238867644e-07, + "loss": 1.1686, + "step": 33661 + }, + { + "epoch": 2.83, + "learning_rate": 1.5973638028462303e-07, + "loss": 1.0885, + "step": 33662 + }, + { + "epoch": 2.83, + "learning_rate": 1.5957460300500937e-07, + "loss": 1.1178, + "step": 33663 + }, + { + "epoch": 2.83, + "learning_rate": 1.5941290703013757e-07, + "loss": 1.1763, + "step": 33664 + }, + { + "epoch": 2.83, + "learning_rate": 1.5925129236134762e-07, + "loss": 1.121, + "step": 33665 + }, + { + "epoch": 2.83, + "learning_rate": 1.5908975899997292e-07, + "loss": 1.1482, + "step": 33666 + }, + { + "epoch": 2.83, + "learning_rate": 1.5892830694734685e-07, + "loss": 1.0802, + "step": 33667 + }, + { + "epoch": 2.83, + "learning_rate": 1.5876693620480498e-07, + "loss": 1.1571, + "step": 33668 + }, + { + "epoch": 2.83, + "learning_rate": 1.5860564677368185e-07, + "loss": 1.1189, + "step": 33669 + }, + { + "epoch": 2.83, + "learning_rate": 1.5844443865530634e-07, + "loss": 1.1437, + "step": 33670 + }, + { + "epoch": 2.83, + "learning_rate": 1.5828331185101409e-07, + "loss": 1.0429, + "step": 33671 + }, + { + "epoch": 2.83, + "learning_rate": 1.5812226636213512e-07, + "loss": 1.08, + "step": 33672 + }, + { + "epoch": 2.83, + "learning_rate": 1.5796130218999728e-07, + "loss": 1.0565, + "step": 33673 + }, + { + "epoch": 2.84, + "learning_rate": 1.5780041933593616e-07, + "loss": 1.1167, + "step": 33674 + }, + { + "epoch": 2.84, + "learning_rate": 1.5763961780127624e-07, + "loss": 1.0755, + "step": 33675 + }, + { + "epoch": 2.84, + "learning_rate": 1.5747889758734757e-07, + "loss": 1.1135, + "step": 33676 + }, + { + "epoch": 2.84, + "learning_rate": 1.5731825869547913e-07, + "loss": 1.1156, + "step": 33677 + }, + { + "epoch": 2.84, + "learning_rate": 1.5715770112699645e-07, + "loss": 1.1163, + "step": 33678 + }, + { + "epoch": 2.84, + "learning_rate": 1.5699722488322744e-07, + "loss": 1.095, + "step": 33679 + }, + { + "epoch": 2.84, + "learning_rate": 1.5683682996549877e-07, + "loss": 1.056, + "step": 33680 + }, + { + "epoch": 2.84, + "learning_rate": 1.5667651637513158e-07, + "loss": 1.0742, + "step": 33681 + }, + { + "epoch": 2.84, + "learning_rate": 1.5651628411345487e-07, + "loss": 1.1724, + "step": 33682 + }, + { + "epoch": 2.84, + "learning_rate": 1.56356133181792e-07, + "loss": 1.2047, + "step": 33683 + }, + { + "epoch": 2.84, + "learning_rate": 1.56196063581463e-07, + "loss": 1.2064, + "step": 33684 + }, + { + "epoch": 2.84, + "learning_rate": 1.5603607531379462e-07, + "loss": 1.1312, + "step": 33685 + }, + { + "epoch": 2.84, + "learning_rate": 1.5587616838010467e-07, + "loss": 1.0768, + "step": 33686 + }, + { + "epoch": 2.84, + "learning_rate": 1.5571634278171767e-07, + "loss": 1.1309, + "step": 33687 + }, + { + "epoch": 2.84, + "learning_rate": 1.5555659851995252e-07, + "loss": 1.0268, + "step": 33688 + }, + { + "epoch": 2.84, + "learning_rate": 1.5539693559612935e-07, + "loss": 1.1485, + "step": 33689 + }, + { + "epoch": 2.84, + "learning_rate": 1.5523735401156704e-07, + "loss": 1.1737, + "step": 33690 + }, + { + "epoch": 2.84, + "learning_rate": 1.5507785376758454e-07, + "loss": 1.0567, + "step": 33691 + }, + { + "epoch": 2.84, + "learning_rate": 1.5491843486549975e-07, + "loss": 1.1219, + "step": 33692 + }, + { + "epoch": 2.84, + "learning_rate": 1.5475909730663042e-07, + "loss": 1.0669, + "step": 33693 + }, + { + "epoch": 2.84, + "learning_rate": 1.545998410922911e-07, + "loss": 1.0576, + "step": 33694 + }, + { + "epoch": 2.84, + "learning_rate": 1.544406662237985e-07, + "loss": 1.0903, + "step": 33695 + }, + { + "epoch": 2.84, + "learning_rate": 1.5428157270246935e-07, + "loss": 1.1585, + "step": 33696 + }, + { + "epoch": 2.84, + "learning_rate": 1.5412256052961704e-07, + "loss": 1.1383, + "step": 33697 + }, + { + "epoch": 2.84, + "learning_rate": 1.5396362970655388e-07, + "loss": 1.1744, + "step": 33698 + }, + { + "epoch": 2.84, + "learning_rate": 1.5380478023459545e-07, + "loss": 1.0433, + "step": 33699 + }, + { + "epoch": 2.84, + "learning_rate": 1.5364601211505292e-07, + "loss": 1.1108, + "step": 33700 + }, + { + "epoch": 2.84, + "learning_rate": 1.534873253492375e-07, + "loss": 1.0751, + "step": 33701 + }, + { + "epoch": 2.84, + "learning_rate": 1.5332871993846255e-07, + "loss": 1.1304, + "step": 33702 + }, + { + "epoch": 2.84, + "learning_rate": 1.5317019588403482e-07, + "loss": 1.119, + "step": 33703 + }, + { + "epoch": 2.84, + "learning_rate": 1.530117531872688e-07, + "loss": 1.0726, + "step": 33704 + }, + { + "epoch": 2.84, + "learning_rate": 1.5285339184947011e-07, + "loss": 1.0129, + "step": 33705 + }, + { + "epoch": 2.84, + "learning_rate": 1.5269511187194774e-07, + "loss": 1.0786, + "step": 33706 + }, + { + "epoch": 2.84, + "learning_rate": 1.525369132560106e-07, + "loss": 1.0701, + "step": 33707 + }, + { + "epoch": 2.84, + "learning_rate": 1.5237879600296434e-07, + "loss": 1.1071, + "step": 33708 + }, + { + "epoch": 2.84, + "learning_rate": 1.522207601141157e-07, + "loss": 1.1839, + "step": 33709 + }, + { + "epoch": 2.84, + "learning_rate": 1.5206280559077246e-07, + "loss": 1.1255, + "step": 33710 + }, + { + "epoch": 2.84, + "learning_rate": 1.5190493243423697e-07, + "loss": 1.1717, + "step": 33711 + }, + { + "epoch": 2.84, + "learning_rate": 1.5174714064581486e-07, + "loss": 1.2054, + "step": 33712 + }, + { + "epoch": 2.84, + "learning_rate": 1.5158943022681061e-07, + "loss": 1.1055, + "step": 33713 + }, + { + "epoch": 2.84, + "learning_rate": 1.5143180117852429e-07, + "loss": 1.179, + "step": 33714 + }, + { + "epoch": 2.84, + "learning_rate": 1.5127425350226266e-07, + "loss": 1.1153, + "step": 33715 + }, + { + "epoch": 2.84, + "learning_rate": 1.5111678719932354e-07, + "loss": 1.1261, + "step": 33716 + }, + { + "epoch": 2.84, + "learning_rate": 1.5095940227101036e-07, + "loss": 1.1015, + "step": 33717 + }, + { + "epoch": 2.84, + "learning_rate": 1.5080209871862205e-07, + "loss": 1.114, + "step": 33718 + }, + { + "epoch": 2.84, + "learning_rate": 1.5064487654346093e-07, + "loss": 1.078, + "step": 33719 + }, + { + "epoch": 2.84, + "learning_rate": 1.5048773574682152e-07, + "loss": 1.1564, + "step": 33720 + }, + { + "epoch": 2.84, + "learning_rate": 1.5033067633000498e-07, + "loss": 1.1408, + "step": 33721 + }, + { + "epoch": 2.84, + "learning_rate": 1.5017369829431027e-07, + "loss": 1.0903, + "step": 33722 + }, + { + "epoch": 2.84, + "learning_rate": 1.5001680164103195e-07, + "loss": 1.1514, + "step": 33723 + }, + { + "epoch": 2.84, + "learning_rate": 1.498599863714678e-07, + "loss": 1.1175, + "step": 33724 + }, + { + "epoch": 2.84, + "learning_rate": 1.4970325248691132e-07, + "loss": 1.0975, + "step": 33725 + }, + { + "epoch": 2.84, + "learning_rate": 1.4954659998866028e-07, + "loss": 1.069, + "step": 33726 + }, + { + "epoch": 2.84, + "learning_rate": 1.4939002887800813e-07, + "loss": 1.1068, + "step": 33727 + }, + { + "epoch": 2.84, + "learning_rate": 1.4923353915624716e-07, + "loss": 1.1652, + "step": 33728 + }, + { + "epoch": 2.84, + "learning_rate": 1.490771308246719e-07, + "loss": 1.104, + "step": 33729 + }, + { + "epoch": 2.84, + "learning_rate": 1.4892080388457463e-07, + "loss": 1.0697, + "step": 33730 + }, + { + "epoch": 2.84, + "learning_rate": 1.4876455833724435e-07, + "loss": 1.1412, + "step": 33731 + }, + { + "epoch": 2.84, + "learning_rate": 1.4860839418397556e-07, + "loss": 1.0453, + "step": 33732 + }, + { + "epoch": 2.84, + "learning_rate": 1.484523114260572e-07, + "loss": 1.0302, + "step": 33733 + }, + { + "epoch": 2.84, + "learning_rate": 1.4829631006477828e-07, + "loss": 1.1349, + "step": 33734 + }, + { + "epoch": 2.84, + "learning_rate": 1.481403901014289e-07, + "loss": 1.066, + "step": 33735 + }, + { + "epoch": 2.84, + "learning_rate": 1.4798455153729573e-07, + "loss": 1.1048, + "step": 33736 + }, + { + "epoch": 2.84, + "learning_rate": 1.478287943736678e-07, + "loss": 1.0459, + "step": 33737 + }, + { + "epoch": 2.84, + "learning_rate": 1.4767311861183186e-07, + "loss": 1.1259, + "step": 33738 + }, + { + "epoch": 2.84, + "learning_rate": 1.4751752425307243e-07, + "loss": 1.1034, + "step": 33739 + }, + { + "epoch": 2.84, + "learning_rate": 1.4736201129867732e-07, + "loss": 1.0846, + "step": 33740 + }, + { + "epoch": 2.84, + "learning_rate": 1.4720657974993114e-07, + "loss": 1.065, + "step": 33741 + }, + { + "epoch": 2.84, + "learning_rate": 1.4705122960811613e-07, + "loss": 1.0757, + "step": 33742 + }, + { + "epoch": 2.84, + "learning_rate": 1.4689596087451907e-07, + "loss": 1.1862, + "step": 33743 + }, + { + "epoch": 2.84, + "learning_rate": 1.4674077355041892e-07, + "loss": 0.9768, + "step": 33744 + }, + { + "epoch": 2.84, + "learning_rate": 1.4658566763710025e-07, + "loss": 1.1335, + "step": 33745 + }, + { + "epoch": 2.84, + "learning_rate": 1.4643064313584532e-07, + "loss": 1.1594, + "step": 33746 + }, + { + "epoch": 2.84, + "learning_rate": 1.4627570004793423e-07, + "loss": 1.0788, + "step": 33747 + }, + { + "epoch": 2.84, + "learning_rate": 1.4612083837464596e-07, + "loss": 1.1411, + "step": 33748 + }, + { + "epoch": 2.84, + "learning_rate": 1.4596605811726062e-07, + "loss": 1.1382, + "step": 33749 + }, + { + "epoch": 2.84, + "learning_rate": 1.4581135927705824e-07, + "loss": 1.1033, + "step": 33750 + }, + { + "epoch": 2.84, + "learning_rate": 1.4565674185531564e-07, + "loss": 1.1362, + "step": 33751 + }, + { + "epoch": 2.84, + "learning_rate": 1.4550220585331177e-07, + "loss": 1.1575, + "step": 33752 + }, + { + "epoch": 2.84, + "learning_rate": 1.4534775127232115e-07, + "loss": 1.0232, + "step": 33753 + }, + { + "epoch": 2.84, + "learning_rate": 1.4519337811362276e-07, + "loss": 1.1095, + "step": 33754 + }, + { + "epoch": 2.84, + "learning_rate": 1.4503908637848897e-07, + "loss": 1.1661, + "step": 33755 + }, + { + "epoch": 2.84, + "learning_rate": 1.4488487606819758e-07, + "loss": 1.1128, + "step": 33756 + }, + { + "epoch": 2.84, + "learning_rate": 1.447307471840209e-07, + "loss": 1.1268, + "step": 33757 + }, + { + "epoch": 2.84, + "learning_rate": 1.4457669972723242e-07, + "loss": 1.0581, + "step": 33758 + }, + { + "epoch": 2.84, + "learning_rate": 1.4442273369910553e-07, + "loss": 1.1417, + "step": 33759 + }, + { + "epoch": 2.84, + "learning_rate": 1.4426884910091256e-07, + "loss": 1.1236, + "step": 33760 + }, + { + "epoch": 2.84, + "learning_rate": 1.4411504593392468e-07, + "loss": 1.1199, + "step": 33761 + }, + { + "epoch": 2.84, + "learning_rate": 1.43961324199412e-07, + "loss": 1.1803, + "step": 33762 + }, + { + "epoch": 2.84, + "learning_rate": 1.4380768389864685e-07, + "loss": 1.1533, + "step": 33763 + }, + { + "epoch": 2.84, + "learning_rate": 1.436541250328949e-07, + "loss": 1.0763, + "step": 33764 + }, + { + "epoch": 2.84, + "learning_rate": 1.4350064760342841e-07, + "loss": 1.1526, + "step": 33765 + }, + { + "epoch": 2.84, + "learning_rate": 1.4334725161151308e-07, + "loss": 1.128, + "step": 33766 + }, + { + "epoch": 2.84, + "learning_rate": 1.431939370584179e-07, + "loss": 1.1676, + "step": 33767 + }, + { + "epoch": 2.84, + "learning_rate": 1.4304070394540846e-07, + "loss": 1.0799, + "step": 33768 + }, + { + "epoch": 2.84, + "learning_rate": 1.4288755227375273e-07, + "loss": 1.0328, + "step": 33769 + }, + { + "epoch": 2.84, + "learning_rate": 1.4273448204471295e-07, + "loss": 1.1175, + "step": 33770 + }, + { + "epoch": 2.84, + "learning_rate": 1.4258149325955595e-07, + "loss": 1.063, + "step": 33771 + }, + { + "epoch": 2.84, + "learning_rate": 1.4242858591954735e-07, + "loss": 1.0826, + "step": 33772 + }, + { + "epoch": 2.84, + "learning_rate": 1.4227576002594613e-07, + "loss": 1.0976, + "step": 33773 + }, + { + "epoch": 2.84, + "learning_rate": 1.4212301558001906e-07, + "loss": 1.103, + "step": 33774 + }, + { + "epoch": 2.84, + "learning_rate": 1.4197035258302517e-07, + "loss": 1.161, + "step": 33775 + }, + { + "epoch": 2.84, + "learning_rate": 1.4181777103622895e-07, + "loss": 1.0689, + "step": 33776 + }, + { + "epoch": 2.84, + "learning_rate": 1.4166527094088723e-07, + "loss": 1.0509, + "step": 33777 + }, + { + "epoch": 2.84, + "learning_rate": 1.4151285229826338e-07, + "loss": 1.0928, + "step": 33778 + }, + { + "epoch": 2.84, + "learning_rate": 1.4136051510961536e-07, + "loss": 1.1309, + "step": 33779 + }, + { + "epoch": 2.84, + "learning_rate": 1.412082593762032e-07, + "loss": 1.0988, + "step": 33780 + }, + { + "epoch": 2.84, + "learning_rate": 1.4105608509928148e-07, + "loss": 1.1228, + "step": 33781 + }, + { + "epoch": 2.84, + "learning_rate": 1.409039922801103e-07, + "loss": 1.0718, + "step": 33782 + }, + { + "epoch": 2.84, + "learning_rate": 1.4075198091994647e-07, + "loss": 1.0373, + "step": 33783 + }, + { + "epoch": 2.84, + "learning_rate": 1.4060005102004337e-07, + "loss": 0.9883, + "step": 33784 + }, + { + "epoch": 2.84, + "learning_rate": 1.4044820258166002e-07, + "loss": 1.1144, + "step": 33785 + }, + { + "epoch": 2.84, + "learning_rate": 1.4029643560604767e-07, + "loss": 1.031, + "step": 33786 + }, + { + "epoch": 2.84, + "learning_rate": 1.4014475009446305e-07, + "loss": 1.1354, + "step": 33787 + }, + { + "epoch": 2.84, + "learning_rate": 1.3999314604815629e-07, + "loss": 1.1931, + "step": 33788 + }, + { + "epoch": 2.84, + "learning_rate": 1.3984162346838304e-07, + "loss": 1.1292, + "step": 33789 + }, + { + "epoch": 2.84, + "learning_rate": 1.3969018235639343e-07, + "loss": 1.1345, + "step": 33790 + }, + { + "epoch": 2.84, + "learning_rate": 1.3953882271344088e-07, + "loss": 1.14, + "step": 33791 + }, + { + "epoch": 2.85, + "learning_rate": 1.3938754454077108e-07, + "loss": 1.1523, + "step": 33792 + }, + { + "epoch": 2.85, + "learning_rate": 1.3923634783963968e-07, + "loss": 1.0985, + "step": 33793 + }, + { + "epoch": 2.85, + "learning_rate": 1.3908523261129349e-07, + "loss": 1.125, + "step": 33794 + }, + { + "epoch": 2.85, + "learning_rate": 1.3893419885697923e-07, + "loss": 1.0829, + "step": 33795 + }, + { + "epoch": 2.85, + "learning_rate": 1.3878324657794818e-07, + "loss": 1.0784, + "step": 33796 + }, + { + "epoch": 2.85, + "learning_rate": 1.3863237577544485e-07, + "loss": 1.0509, + "step": 33797 + }, + { + "epoch": 2.85, + "learning_rate": 1.384815864507183e-07, + "loss": 1.1407, + "step": 33798 + }, + { + "epoch": 2.85, + "learning_rate": 1.3833087860501192e-07, + "loss": 1.0849, + "step": 33799 + }, + { + "epoch": 2.85, + "learning_rate": 1.3818025223957144e-07, + "loss": 1.1642, + "step": 33800 + }, + { + "epoch": 2.85, + "learning_rate": 1.3802970735564247e-07, + "loss": 1.0273, + "step": 33801 + }, + { + "epoch": 2.85, + "learning_rate": 1.3787924395446962e-07, + "loss": 1.1677, + "step": 33802 + }, + { + "epoch": 2.85, + "learning_rate": 1.3772886203729185e-07, + "loss": 1.1213, + "step": 33803 + }, + { + "epoch": 2.85, + "learning_rate": 1.375785616053571e-07, + "loss": 1.1462, + "step": 33804 + }, + { + "epoch": 2.85, + "learning_rate": 1.3742834265990434e-07, + "loss": 1.1464, + "step": 33805 + }, + { + "epoch": 2.85, + "learning_rate": 1.372782052021737e-07, + "loss": 1.0157, + "step": 33806 + }, + { + "epoch": 2.85, + "learning_rate": 1.3712814923340757e-07, + "loss": 0.9956, + "step": 33807 + }, + { + "epoch": 2.85, + "learning_rate": 1.3697817475484598e-07, + "loss": 1.0831, + "step": 33808 + }, + { + "epoch": 2.85, + "learning_rate": 1.368282817677269e-07, + "loss": 1.0848, + "step": 33809 + }, + { + "epoch": 2.85, + "learning_rate": 1.3667847027328818e-07, + "loss": 1.1085, + "step": 33810 + }, + { + "epoch": 2.85, + "learning_rate": 1.3652874027276997e-07, + "loss": 1.071, + "step": 33811 + }, + { + "epoch": 2.85, + "learning_rate": 1.3637909176740681e-07, + "loss": 1.1641, + "step": 33812 + }, + { + "epoch": 2.85, + "learning_rate": 1.3622952475843776e-07, + "loss": 1.1429, + "step": 33813 + }, + { + "epoch": 2.85, + "learning_rate": 1.3608003924709623e-07, + "loss": 1.0893, + "step": 33814 + }, + { + "epoch": 2.85, + "learning_rate": 1.3593063523461902e-07, + "loss": 1.1293, + "step": 33815 + }, + { + "epoch": 2.85, + "learning_rate": 1.3578131272223848e-07, + "loss": 1.187, + "step": 33816 + }, + { + "epoch": 2.85, + "learning_rate": 1.356320717111903e-07, + "loss": 1.1008, + "step": 33817 + }, + { + "epoch": 2.85, + "learning_rate": 1.3548291220270792e-07, + "loss": 1.1818, + "step": 33818 + }, + { + "epoch": 2.85, + "learning_rate": 1.3533383419802258e-07, + "loss": 1.1359, + "step": 33819 + }, + { + "epoch": 2.85, + "learning_rate": 1.3518483769836444e-07, + "loss": 1.0222, + "step": 33820 + }, + { + "epoch": 2.85, + "learning_rate": 1.3503592270496691e-07, + "loss": 1.1407, + "step": 33821 + }, + { + "epoch": 2.85, + "learning_rate": 1.3488708921906124e-07, + "loss": 1.111, + "step": 33822 + }, + { + "epoch": 2.85, + "learning_rate": 1.3473833724187423e-07, + "loss": 1.1725, + "step": 33823 + }, + { + "epoch": 2.85, + "learning_rate": 1.3458966677463714e-07, + "loss": 1.0761, + "step": 33824 + }, + { + "epoch": 2.85, + "learning_rate": 1.3444107781857785e-07, + "loss": 1.1001, + "step": 33825 + }, + { + "epoch": 2.85, + "learning_rate": 1.3429257037492317e-07, + "loss": 1.0132, + "step": 33826 + }, + { + "epoch": 2.85, + "learning_rate": 1.34144144444901e-07, + "loss": 1.1529, + "step": 33827 + }, + { + "epoch": 2.85, + "learning_rate": 1.33995800029737e-07, + "loss": 1.0643, + "step": 33828 + }, + { + "epoch": 2.85, + "learning_rate": 1.338475371306569e-07, + "loss": 1.1214, + "step": 33829 + }, + { + "epoch": 2.85, + "learning_rate": 1.3369935574888748e-07, + "loss": 1.135, + "step": 33830 + }, + { + "epoch": 2.85, + "learning_rate": 1.3355125588565e-07, + "loss": 1.1494, + "step": 33831 + }, + { + "epoch": 2.85, + "learning_rate": 1.334032375421701e-07, + "loss": 1.0734, + "step": 33832 + }, + { + "epoch": 2.85, + "learning_rate": 1.332553007196713e-07, + "loss": 1.1698, + "step": 33833 + }, + { + "epoch": 2.85, + "learning_rate": 1.331074454193726e-07, + "loss": 1.082, + "step": 33834 + }, + { + "epoch": 2.85, + "learning_rate": 1.329596716425008e-07, + "loss": 1.1576, + "step": 33835 + }, + { + "epoch": 2.85, + "learning_rate": 1.328119793902727e-07, + "loss": 1.2101, + "step": 33836 + }, + { + "epoch": 2.85, + "learning_rate": 1.3266436866391063e-07, + "loss": 1.0965, + "step": 33837 + }, + { + "epoch": 2.85, + "learning_rate": 1.3251683946463256e-07, + "loss": 1.1452, + "step": 33838 + }, + { + "epoch": 2.85, + "learning_rate": 1.323693917936586e-07, + "loss": 1.0388, + "step": 33839 + }, + { + "epoch": 2.85, + "learning_rate": 1.3222202565220665e-07, + "loss": 1.1769, + "step": 33840 + }, + { + "epoch": 2.85, + "learning_rate": 1.3207474104149576e-07, + "loss": 1.1508, + "step": 33841 + }, + { + "epoch": 2.85, + "learning_rate": 1.3192753796273937e-07, + "loss": 1.077, + "step": 33842 + }, + { + "epoch": 2.85, + "learning_rate": 1.3178041641715767e-07, + "loss": 1.1393, + "step": 33843 + }, + { + "epoch": 2.85, + "learning_rate": 1.3163337640596408e-07, + "loss": 1.2095, + "step": 33844 + }, + { + "epoch": 2.85, + "learning_rate": 1.314864179303721e-07, + "loss": 1.0584, + "step": 33845 + }, + { + "epoch": 2.85, + "learning_rate": 1.3133954099159962e-07, + "loss": 1.103, + "step": 33846 + }, + { + "epoch": 2.85, + "learning_rate": 1.3119274559085792e-07, + "loss": 1.1556, + "step": 33847 + }, + { + "epoch": 2.85, + "learning_rate": 1.3104603172936048e-07, + "loss": 1.2273, + "step": 33848 + }, + { + "epoch": 2.85, + "learning_rate": 1.3089939940831854e-07, + "loss": 1.1241, + "step": 33849 + }, + { + "epoch": 2.85, + "learning_rate": 1.3075284862894334e-07, + "loss": 1.1653, + "step": 33850 + }, + { + "epoch": 2.85, + "learning_rate": 1.3060637939244836e-07, + "loss": 1.0504, + "step": 33851 + }, + { + "epoch": 2.85, + "learning_rate": 1.3045999170004153e-07, + "loss": 1.1196, + "step": 33852 + }, + { + "epoch": 2.85, + "learning_rate": 1.303136855529319e-07, + "loss": 0.9837, + "step": 33853 + }, + { + "epoch": 2.85, + "learning_rate": 1.301674609523307e-07, + "loss": 1.1121, + "step": 33854 + }, + { + "epoch": 2.85, + "learning_rate": 1.3002131789944473e-07, + "loss": 1.1308, + "step": 33855 + }, + { + "epoch": 2.85, + "learning_rate": 1.2987525639548081e-07, + "loss": 1.0869, + "step": 33856 + }, + { + "epoch": 2.85, + "learning_rate": 1.297292764416458e-07, + "loss": 1.1552, + "step": 33857 + }, + { + "epoch": 2.85, + "learning_rate": 1.2958337803914756e-07, + "loss": 1.0817, + "step": 33858 + }, + { + "epoch": 2.85, + "learning_rate": 1.294375611891896e-07, + "loss": 1.0774, + "step": 33859 + }, + { + "epoch": 2.85, + "learning_rate": 1.2929182589297762e-07, + "loss": 1.1668, + "step": 33860 + }, + { + "epoch": 2.85, + "learning_rate": 1.291461721517151e-07, + "loss": 1.152, + "step": 33861 + }, + { + "epoch": 2.85, + "learning_rate": 1.2900059996660664e-07, + "loss": 1.1146, + "step": 33862 + }, + { + "epoch": 2.85, + "learning_rate": 1.288551093388557e-07, + "loss": 1.0525, + "step": 33863 + }, + { + "epoch": 2.85, + "learning_rate": 1.2870970026966023e-07, + "loss": 1.1217, + "step": 33864 + }, + { + "epoch": 2.85, + "learning_rate": 1.2856437276022704e-07, + "loss": 1.1068, + "step": 33865 + }, + { + "epoch": 2.85, + "learning_rate": 1.2841912681175295e-07, + "loss": 1.1592, + "step": 33866 + }, + { + "epoch": 2.85, + "learning_rate": 1.282739624254392e-07, + "loss": 1.1093, + "step": 33867 + }, + { + "epoch": 2.85, + "learning_rate": 1.2812887960248488e-07, + "loss": 1.1491, + "step": 33868 + }, + { + "epoch": 2.85, + "learning_rate": 1.279838783440912e-07, + "loss": 1.1356, + "step": 33869 + }, + { + "epoch": 2.85, + "learning_rate": 1.2783895865145168e-07, + "loss": 1.0917, + "step": 33870 + }, + { + "epoch": 2.85, + "learning_rate": 1.2769412052576758e-07, + "loss": 1.1994, + "step": 33871 + }, + { + "epoch": 2.85, + "learning_rate": 1.275493639682346e-07, + "loss": 1.1184, + "step": 33872 + }, + { + "epoch": 2.85, + "learning_rate": 1.274046889800462e-07, + "loss": 1.1437, + "step": 33873 + }, + { + "epoch": 2.85, + "learning_rate": 1.272600955624026e-07, + "loss": 1.1125, + "step": 33874 + }, + { + "epoch": 2.85, + "learning_rate": 1.271155837164939e-07, + "loss": 1.1829, + "step": 33875 + }, + { + "epoch": 2.85, + "learning_rate": 1.2697115344351696e-07, + "loss": 1.1294, + "step": 33876 + }, + { + "epoch": 2.85, + "learning_rate": 1.2682680474466412e-07, + "loss": 1.1121, + "step": 33877 + }, + { + "epoch": 2.85, + "learning_rate": 1.2668253762112782e-07, + "loss": 1.1024, + "step": 33878 + }, + { + "epoch": 2.85, + "learning_rate": 1.2653835207409926e-07, + "loss": 1.1425, + "step": 33879 + }, + { + "epoch": 2.85, + "learning_rate": 1.2639424810477307e-07, + "loss": 1.1095, + "step": 33880 + }, + { + "epoch": 2.85, + "learning_rate": 1.26250225714335e-07, + "loss": 1.0962, + "step": 33881 + }, + { + "epoch": 2.85, + "learning_rate": 1.2610628490397847e-07, + "loss": 1.0969, + "step": 33882 + }, + { + "epoch": 2.85, + "learning_rate": 1.259624256748926e-07, + "loss": 1.1492, + "step": 33883 + }, + { + "epoch": 2.85, + "learning_rate": 1.2581864802826416e-07, + "loss": 1.1071, + "step": 33884 + }, + { + "epoch": 2.85, + "learning_rate": 1.2567495196528335e-07, + "loss": 1.1536, + "step": 33885 + }, + { + "epoch": 2.85, + "learning_rate": 1.2553133748713587e-07, + "loss": 1.1601, + "step": 33886 + }, + { + "epoch": 2.85, + "learning_rate": 1.2538780459500855e-07, + "loss": 1.1434, + "step": 33887 + }, + { + "epoch": 2.85, + "learning_rate": 1.252443532900871e-07, + "loss": 1.1238, + "step": 33888 + }, + { + "epoch": 2.85, + "learning_rate": 1.2510098357355838e-07, + "loss": 1.2065, + "step": 33889 + }, + { + "epoch": 2.85, + "learning_rate": 1.2495769544660474e-07, + "loss": 1.078, + "step": 33890 + }, + { + "epoch": 2.85, + "learning_rate": 1.2481448891041191e-07, + "loss": 1.0882, + "step": 33891 + }, + { + "epoch": 2.85, + "learning_rate": 1.246713639661601e-07, + "loss": 1.1313, + "step": 33892 + }, + { + "epoch": 2.85, + "learning_rate": 1.2452832061503606e-07, + "loss": 1.0838, + "step": 33893 + }, + { + "epoch": 2.85, + "learning_rate": 1.243853588582189e-07, + "loss": 1.1171, + "step": 33894 + }, + { + "epoch": 2.85, + "learning_rate": 1.2424247869688987e-07, + "loss": 1.1085, + "step": 33895 + }, + { + "epoch": 2.85, + "learning_rate": 1.2409968013223138e-07, + "loss": 1.0688, + "step": 33896 + }, + { + "epoch": 2.85, + "learning_rate": 1.2395696316542138e-07, + "loss": 1.0982, + "step": 33897 + }, + { + "epoch": 2.85, + "learning_rate": 1.2381432779764002e-07, + "loss": 1.0687, + "step": 33898 + }, + { + "epoch": 2.85, + "learning_rate": 1.2367177403006525e-07, + "loss": 1.0797, + "step": 33899 + }, + { + "epoch": 2.85, + "learning_rate": 1.23529301863875e-07, + "loss": 1.1625, + "step": 33900 + }, + { + "epoch": 2.85, + "learning_rate": 1.2338691130024617e-07, + "loss": 1.1237, + "step": 33901 + }, + { + "epoch": 2.85, + "learning_rate": 1.2324460234035663e-07, + "loss": 1.1255, + "step": 33902 + }, + { + "epoch": 2.85, + "learning_rate": 1.2310237498537992e-07, + "loss": 1.0752, + "step": 33903 + }, + { + "epoch": 2.85, + "learning_rate": 1.2296022923649286e-07, + "loss": 1.1295, + "step": 33904 + }, + { + "epoch": 2.85, + "learning_rate": 1.228181650948701e-07, + "loss": 1.1091, + "step": 33905 + }, + { + "epoch": 2.85, + "learning_rate": 1.2267618256168402e-07, + "loss": 1.0801, + "step": 33906 + }, + { + "epoch": 2.85, + "learning_rate": 1.2253428163810922e-07, + "loss": 1.0643, + "step": 33907 + }, + { + "epoch": 2.85, + "learning_rate": 1.2239246232531588e-07, + "loss": 1.0909, + "step": 33908 + }, + { + "epoch": 2.85, + "learning_rate": 1.2225072462447862e-07, + "loss": 1.0821, + "step": 33909 + }, + { + "epoch": 2.85, + "learning_rate": 1.221090685367665e-07, + "loss": 1.1173, + "step": 33910 + }, + { + "epoch": 2.86, + "learning_rate": 1.2196749406335085e-07, + "loss": 1.06, + "step": 33911 + }, + { + "epoch": 2.86, + "learning_rate": 1.2182600120540067e-07, + "loss": 1.1673, + "step": 33912 + }, + { + "epoch": 2.86, + "learning_rate": 1.2168458996408727e-07, + "loss": 1.1008, + "step": 33913 + }, + { + "epoch": 2.86, + "learning_rate": 1.2154326034057528e-07, + "loss": 1.1045, + "step": 33914 + }, + { + "epoch": 2.86, + "learning_rate": 1.2140201233603598e-07, + "loss": 1.0838, + "step": 33915 + }, + { + "epoch": 2.86, + "learning_rate": 1.212608459516329e-07, + "loss": 1.1405, + "step": 33916 + }, + { + "epoch": 2.86, + "learning_rate": 1.2111976118853508e-07, + "loss": 1.1006, + "step": 33917 + }, + { + "epoch": 2.86, + "learning_rate": 1.2097875804790827e-07, + "loss": 1.124, + "step": 33918 + }, + { + "epoch": 2.86, + "learning_rate": 1.2083783653091596e-07, + "loss": 1.07, + "step": 33919 + }, + { + "epoch": 2.86, + "learning_rate": 1.2069699663872282e-07, + "loss": 1.0439, + "step": 33920 + }, + { + "epoch": 2.86, + "learning_rate": 1.205562383724934e-07, + "loss": 1.1286, + "step": 33921 + }, + { + "epoch": 2.86, + "learning_rate": 1.204155617333902e-07, + "loss": 1.1047, + "step": 33922 + }, + { + "epoch": 2.86, + "learning_rate": 1.2027496672257556e-07, + "loss": 1.1756, + "step": 33923 + }, + { + "epoch": 2.86, + "learning_rate": 1.2013445334121077e-07, + "loss": 1.1638, + "step": 33924 + }, + { + "epoch": 2.86, + "learning_rate": 1.1999402159045714e-07, + "loss": 1.1066, + "step": 33925 + }, + { + "epoch": 2.86, + "learning_rate": 1.1985367147147488e-07, + "loss": 1.0638, + "step": 33926 + }, + { + "epoch": 2.86, + "learning_rate": 1.197134029854241e-07, + "loss": 1.0801, + "step": 33927 + }, + { + "epoch": 2.86, + "learning_rate": 1.1957321613346285e-07, + "loss": 1.1288, + "step": 33928 + }, + { + "epoch": 2.86, + "learning_rate": 1.1943311091675015e-07, + "loss": 1.0976, + "step": 33929 + }, + { + "epoch": 2.86, + "learning_rate": 1.1929308733644395e-07, + "loss": 1.0907, + "step": 33930 + }, + { + "epoch": 2.86, + "learning_rate": 1.1915314539369782e-07, + "loss": 1.1693, + "step": 33931 + }, + { + "epoch": 2.86, + "learning_rate": 1.1901328508967302e-07, + "loss": 1.077, + "step": 33932 + }, + { + "epoch": 2.86, + "learning_rate": 1.1887350642552309e-07, + "loss": 1.1541, + "step": 33933 + }, + { + "epoch": 2.86, + "learning_rate": 1.1873380940240042e-07, + "loss": 1.1269, + "step": 33934 + }, + { + "epoch": 2.86, + "learning_rate": 1.1859419402146299e-07, + "loss": 1.044, + "step": 33935 + }, + { + "epoch": 2.86, + "learning_rate": 1.1845466028386321e-07, + "loss": 1.1201, + "step": 33936 + }, + { + "epoch": 2.86, + "learning_rate": 1.1831520819075237e-07, + "loss": 1.192, + "step": 33937 + }, + { + "epoch": 2.86, + "learning_rate": 1.18175837743284e-07, + "loss": 1.1827, + "step": 33938 + }, + { + "epoch": 2.86, + "learning_rate": 1.1803654894260941e-07, + "loss": 1.1435, + "step": 33939 + }, + { + "epoch": 2.86, + "learning_rate": 1.1789734178987988e-07, + "loss": 1.165, + "step": 33940 + }, + { + "epoch": 2.86, + "learning_rate": 1.1775821628624562e-07, + "loss": 1.1174, + "step": 33941 + }, + { + "epoch": 2.86, + "learning_rate": 1.1761917243285459e-07, + "loss": 1.1062, + "step": 33942 + }, + { + "epoch": 2.86, + "learning_rate": 1.174802102308581e-07, + "loss": 1.0366, + "step": 33943 + }, + { + "epoch": 2.86, + "learning_rate": 1.1734132968140188e-07, + "loss": 1.0555, + "step": 33944 + }, + { + "epoch": 2.86, + "learning_rate": 1.1720253078563393e-07, + "loss": 1.1035, + "step": 33945 + }, + { + "epoch": 2.86, + "learning_rate": 1.1706381354470331e-07, + "loss": 1.0454, + "step": 33946 + }, + { + "epoch": 2.86, + "learning_rate": 1.1692517795975355e-07, + "loss": 1.1245, + "step": 33947 + }, + { + "epoch": 2.86, + "learning_rate": 1.1678662403193041e-07, + "loss": 1.123, + "step": 33948 + }, + { + "epoch": 2.86, + "learning_rate": 1.1664815176237965e-07, + "loss": 1.0697, + "step": 33949 + }, + { + "epoch": 2.86, + "learning_rate": 1.1650976115224477e-07, + "loss": 1.0938, + "step": 33950 + }, + { + "epoch": 2.86, + "learning_rate": 1.1637145220267044e-07, + "loss": 1.1494, + "step": 33951 + }, + { + "epoch": 2.86, + "learning_rate": 1.1623322491479794e-07, + "loss": 1.1375, + "step": 33952 + }, + { + "epoch": 2.86, + "learning_rate": 1.160950792897686e-07, + "loss": 1.1163, + "step": 33953 + }, + { + "epoch": 2.86, + "learning_rate": 1.1595701532872594e-07, + "loss": 1.0958, + "step": 33954 + }, + { + "epoch": 2.86, + "learning_rate": 1.1581903303280905e-07, + "loss": 1.1726, + "step": 33955 + }, + { + "epoch": 2.86, + "learning_rate": 1.1568113240315926e-07, + "loss": 1.0998, + "step": 33956 + }, + { + "epoch": 2.86, + "learning_rate": 1.1554331344091452e-07, + "loss": 1.2003, + "step": 33957 + }, + { + "epoch": 2.86, + "learning_rate": 1.1540557614721503e-07, + "loss": 1.0559, + "step": 33958 + }, + { + "epoch": 2.86, + "learning_rate": 1.1526792052319769e-07, + "loss": 1.0804, + "step": 33959 + }, + { + "epoch": 2.86, + "learning_rate": 1.1513034657000155e-07, + "loss": 1.1532, + "step": 33960 + }, + { + "epoch": 2.86, + "learning_rate": 1.1499285428876016e-07, + "loss": 1.137, + "step": 33961 + }, + { + "epoch": 2.86, + "learning_rate": 1.1485544368061264e-07, + "loss": 1.137, + "step": 33962 + }, + { + "epoch": 2.86, + "learning_rate": 1.1471811474669358e-07, + "loss": 1.0549, + "step": 33963 + }, + { + "epoch": 2.86, + "learning_rate": 1.1458086748813546e-07, + "loss": 1.11, + "step": 33964 + }, + { + "epoch": 2.86, + "learning_rate": 1.1444370190607623e-07, + "loss": 1.0469, + "step": 33965 + }, + { + "epoch": 2.86, + "learning_rate": 1.14306618001645e-07, + "loss": 1.1252, + "step": 33966 + }, + { + "epoch": 2.86, + "learning_rate": 1.1416961577597751e-07, + "loss": 1.2106, + "step": 33967 + }, + { + "epoch": 2.86, + "learning_rate": 1.14032695230204e-07, + "loss": 1.086, + "step": 33968 + }, + { + "epoch": 2.86, + "learning_rate": 1.1389585636545575e-07, + "loss": 1.1964, + "step": 33969 + }, + { + "epoch": 2.86, + "learning_rate": 1.137590991828641e-07, + "loss": 1.0888, + "step": 33970 + }, + { + "epoch": 2.86, + "learning_rate": 1.1362242368355925e-07, + "loss": 1.091, + "step": 33971 + }, + { + "epoch": 2.86, + "learning_rate": 1.1348582986866919e-07, + "loss": 1.1382, + "step": 33972 + }, + { + "epoch": 2.86, + "learning_rate": 1.1334931773932412e-07, + "loss": 1.0963, + "step": 33973 + }, + { + "epoch": 2.86, + "learning_rate": 1.1321288729665203e-07, + "loss": 1.209, + "step": 33974 + }, + { + "epoch": 2.86, + "learning_rate": 1.1307653854177758e-07, + "loss": 1.0561, + "step": 33975 + }, + { + "epoch": 2.86, + "learning_rate": 1.1294027147582987e-07, + "loss": 1.1951, + "step": 33976 + }, + { + "epoch": 2.86, + "learning_rate": 1.1280408609993355e-07, + "loss": 1.1373, + "step": 33977 + }, + { + "epoch": 2.86, + "learning_rate": 1.1266798241521437e-07, + "loss": 1.1418, + "step": 33978 + }, + { + "epoch": 2.86, + "learning_rate": 1.125319604227959e-07, + "loss": 1.1297, + "step": 33979 + }, + { + "epoch": 2.86, + "learning_rate": 1.123960201238039e-07, + "loss": 1.1358, + "step": 33980 + }, + { + "epoch": 2.86, + "learning_rate": 1.122601615193597e-07, + "loss": 1.0639, + "step": 33981 + }, + { + "epoch": 2.86, + "learning_rate": 1.1212438461058572e-07, + "loss": 1.1941, + "step": 33982 + }, + { + "epoch": 2.86, + "learning_rate": 1.1198868939860663e-07, + "loss": 1.1843, + "step": 33983 + }, + { + "epoch": 2.86, + "learning_rate": 1.1185307588453931e-07, + "loss": 1.1748, + "step": 33984 + }, + { + "epoch": 2.86, + "learning_rate": 1.117175440695084e-07, + "loss": 1.0532, + "step": 33985 + }, + { + "epoch": 2.86, + "learning_rate": 1.1158209395463082e-07, + "loss": 1.0344, + "step": 33986 + }, + { + "epoch": 2.86, + "learning_rate": 1.1144672554102787e-07, + "loss": 1.0555, + "step": 33987 + }, + { + "epoch": 2.86, + "learning_rate": 1.1131143882981532e-07, + "loss": 1.0169, + "step": 33988 + }, + { + "epoch": 2.86, + "learning_rate": 1.111762338221134e-07, + "loss": 1.145, + "step": 33989 + }, + { + "epoch": 2.86, + "learning_rate": 1.1104111051903787e-07, + "loss": 1.0148, + "step": 33990 + }, + { + "epoch": 2.86, + "learning_rate": 1.1090606892170563e-07, + "loss": 1.0935, + "step": 33991 + }, + { + "epoch": 2.86, + "learning_rate": 1.1077110903123134e-07, + "loss": 1.1037, + "step": 33992 + }, + { + "epoch": 2.86, + "learning_rate": 1.1063623084873298e-07, + "loss": 1.1157, + "step": 33993 + }, + { + "epoch": 2.86, + "learning_rate": 1.105014343753219e-07, + "loss": 1.2001, + "step": 33994 + }, + { + "epoch": 2.86, + "learning_rate": 1.1036671961211276e-07, + "loss": 1.1126, + "step": 33995 + }, + { + "epoch": 2.86, + "learning_rate": 1.1023208656022022e-07, + "loss": 1.0703, + "step": 33996 + }, + { + "epoch": 2.86, + "learning_rate": 1.100975352207545e-07, + "loss": 1.128, + "step": 33997 + }, + { + "epoch": 2.86, + "learning_rate": 1.0996306559482805e-07, + "loss": 1.0699, + "step": 33998 + }, + { + "epoch": 2.86, + "learning_rate": 1.098286776835511e-07, + "loss": 1.0267, + "step": 33999 + }, + { + "epoch": 2.86, + "learning_rate": 1.0969437148803608e-07, + "loss": 1.1608, + "step": 34000 + }, + { + "epoch": 2.86, + "learning_rate": 1.0956014700939099e-07, + "loss": 1.1711, + "step": 34001 + }, + { + "epoch": 2.86, + "learning_rate": 1.0942600424872607e-07, + "loss": 1.0218, + "step": 34002 + }, + { + "epoch": 2.86, + "learning_rate": 1.092919432071482e-07, + "loss": 1.0955, + "step": 34003 + }, + { + "epoch": 2.86, + "learning_rate": 1.091579638857665e-07, + "loss": 1.0378, + "step": 34004 + }, + { + "epoch": 2.86, + "learning_rate": 1.0902406628568563e-07, + "loss": 1.2127, + "step": 34005 + }, + { + "epoch": 2.86, + "learning_rate": 1.0889025040801471e-07, + "loss": 1.0806, + "step": 34006 + }, + { + "epoch": 2.86, + "learning_rate": 1.087565162538573e-07, + "loss": 1.0641, + "step": 34007 + }, + { + "epoch": 2.86, + "learning_rate": 1.086228638243203e-07, + "loss": 1.0367, + "step": 34008 + }, + { + "epoch": 2.86, + "learning_rate": 1.0848929312050615e-07, + "loss": 1.1947, + "step": 34009 + }, + { + "epoch": 2.86, + "learning_rate": 1.0835580414351843e-07, + "loss": 1.154, + "step": 34010 + }, + { + "epoch": 2.86, + "learning_rate": 1.0822239689446178e-07, + "loss": 1.0842, + "step": 34011 + }, + { + "epoch": 2.86, + "learning_rate": 1.0808907137443759e-07, + "loss": 1.1515, + "step": 34012 + }, + { + "epoch": 2.86, + "learning_rate": 1.0795582758454715e-07, + "loss": 1.1787, + "step": 34013 + }, + { + "epoch": 2.86, + "learning_rate": 1.0782266552589072e-07, + "loss": 1.0732, + "step": 34014 + }, + { + "epoch": 2.86, + "learning_rate": 1.0768958519956963e-07, + "loss": 1.1155, + "step": 34015 + }, + { + "epoch": 2.86, + "learning_rate": 1.0755658660668411e-07, + "loss": 1.0711, + "step": 34016 + }, + { + "epoch": 2.86, + "learning_rate": 1.0742366974833107e-07, + "loss": 1.1138, + "step": 34017 + }, + { + "epoch": 2.86, + "learning_rate": 1.0729083462560963e-07, + "loss": 1.1391, + "step": 34018 + }, + { + "epoch": 2.86, + "learning_rate": 1.071580812396178e-07, + "loss": 1.0673, + "step": 34019 + }, + { + "epoch": 2.86, + "learning_rate": 1.0702540959145135e-07, + "loss": 1.1197, + "step": 34020 + }, + { + "epoch": 2.86, + "learning_rate": 1.0689281968220832e-07, + "loss": 1.1446, + "step": 34021 + }, + { + "epoch": 2.86, + "learning_rate": 1.0676031151298228e-07, + "loss": 1.0662, + "step": 34022 + }, + { + "epoch": 2.86, + "learning_rate": 1.0662788508486898e-07, + "loss": 1.1499, + "step": 34023 + }, + { + "epoch": 2.86, + "learning_rate": 1.0649554039896315e-07, + "loss": 1.0767, + "step": 34024 + }, + { + "epoch": 2.86, + "learning_rate": 1.06363277456355e-07, + "loss": 1.147, + "step": 34025 + }, + { + "epoch": 2.86, + "learning_rate": 1.0623109625814254e-07, + "loss": 1.0189, + "step": 34026 + }, + { + "epoch": 2.86, + "learning_rate": 1.060989968054138e-07, + "loss": 1.1454, + "step": 34027 + }, + { + "epoch": 2.86, + "learning_rate": 1.0596697909926235e-07, + "loss": 1.0935, + "step": 34028 + }, + { + "epoch": 2.86, + "learning_rate": 1.058350431407773e-07, + "loss": 1.0628, + "step": 34029 + }, + { + "epoch": 2.87, + "learning_rate": 1.0570318893105114e-07, + "loss": 1.0606, + "step": 34030 + }, + { + "epoch": 2.87, + "learning_rate": 1.0557141647116965e-07, + "loss": 1.1692, + "step": 34031 + }, + { + "epoch": 2.87, + "learning_rate": 1.0543972576222527e-07, + "loss": 1.1267, + "step": 34032 + }, + { + "epoch": 2.87, + "learning_rate": 1.0530811680530494e-07, + "loss": 1.0813, + "step": 34033 + }, + { + "epoch": 2.87, + "learning_rate": 1.0517658960149335e-07, + "loss": 1.1622, + "step": 34034 + }, + { + "epoch": 2.87, + "learning_rate": 1.0504514415188183e-07, + "loss": 1.1403, + "step": 34035 + }, + { + "epoch": 2.87, + "learning_rate": 1.0491378045755395e-07, + "loss": 1.0434, + "step": 34036 + }, + { + "epoch": 2.87, + "learning_rate": 1.0478249851959444e-07, + "loss": 1.2122, + "step": 34037 + }, + { + "epoch": 2.87, + "learning_rate": 1.0465129833908904e-07, + "loss": 1.0804, + "step": 34038 + }, + { + "epoch": 2.87, + "learning_rate": 1.0452017991712138e-07, + "loss": 1.1214, + "step": 34039 + }, + { + "epoch": 2.87, + "learning_rate": 1.04389143254775e-07, + "loss": 1.1639, + "step": 34040 + }, + { + "epoch": 2.87, + "learning_rate": 1.0425818835313239e-07, + "loss": 1.1643, + "step": 34041 + }, + { + "epoch": 2.87, + "learning_rate": 1.0412731521327601e-07, + "loss": 1.0961, + "step": 34042 + }, + { + "epoch": 2.87, + "learning_rate": 1.0399652383628723e-07, + "loss": 1.0313, + "step": 34043 + }, + { + "epoch": 2.87, + "learning_rate": 1.0386581422324515e-07, + "loss": 1.1215, + "step": 34044 + }, + { + "epoch": 2.87, + "learning_rate": 1.0373518637523117e-07, + "loss": 1.0976, + "step": 34045 + }, + { + "epoch": 2.87, + "learning_rate": 1.0360464029332551e-07, + "loss": 1.1842, + "step": 34046 + }, + { + "epoch": 2.87, + "learning_rate": 1.0347417597860398e-07, + "loss": 0.9921, + "step": 34047 + }, + { + "epoch": 2.87, + "learning_rate": 1.0334379343214685e-07, + "loss": 1.11, + "step": 34048 + }, + { + "epoch": 2.87, + "learning_rate": 1.03213492655031e-07, + "loss": 1.0498, + "step": 34049 + }, + { + "epoch": 2.87, + "learning_rate": 1.0308327364833226e-07, + "loss": 1.1615, + "step": 34050 + }, + { + "epoch": 2.87, + "learning_rate": 1.0295313641312644e-07, + "loss": 1.1311, + "step": 34051 + }, + { + "epoch": 2.87, + "learning_rate": 1.0282308095049043e-07, + "loss": 1.1317, + "step": 34052 + }, + { + "epoch": 2.87, + "learning_rate": 1.0269310726149561e-07, + "loss": 1.1385, + "step": 34053 + }, + { + "epoch": 2.87, + "learning_rate": 1.0256321534722002e-07, + "loss": 1.1975, + "step": 34054 + }, + { + "epoch": 2.87, + "learning_rate": 1.0243340520873391e-07, + "loss": 1.1227, + "step": 34055 + }, + { + "epoch": 2.87, + "learning_rate": 1.0230367684711084e-07, + "loss": 1.1023, + "step": 34056 + }, + { + "epoch": 2.87, + "learning_rate": 1.021740302634222e-07, + "loss": 1.1362, + "step": 34057 + }, + { + "epoch": 2.87, + "learning_rate": 1.0204446545873936e-07, + "loss": 1.1841, + "step": 34058 + }, + { + "epoch": 2.87, + "learning_rate": 1.0191498243413256e-07, + "loss": 1.035, + "step": 34059 + }, + { + "epoch": 2.87, + "learning_rate": 1.0178558119067316e-07, + "loss": 1.0891, + "step": 34060 + }, + { + "epoch": 2.87, + "learning_rate": 1.016562617294281e-07, + "loss": 1.1312, + "step": 34061 + }, + { + "epoch": 2.87, + "learning_rate": 1.0152702405146653e-07, + "loss": 1.1177, + "step": 34062 + }, + { + "epoch": 2.87, + "learning_rate": 1.0139786815785868e-07, + "loss": 1.201, + "step": 34063 + }, + { + "epoch": 2.87, + "learning_rate": 1.0126879404966705e-07, + "loss": 1.0463, + "step": 34064 + }, + { + "epoch": 2.87, + "learning_rate": 1.011398017279619e-07, + "loss": 1.1263, + "step": 34065 + }, + { + "epoch": 2.87, + "learning_rate": 1.0101089119380792e-07, + "loss": 1.103, + "step": 34066 + }, + { + "epoch": 2.87, + "learning_rate": 1.0088206244826981e-07, + "loss": 1.057, + "step": 34067 + }, + { + "epoch": 2.87, + "learning_rate": 1.007533154924123e-07, + "loss": 1.13, + "step": 34068 + }, + { + "epoch": 2.87, + "learning_rate": 1.0062465032729896e-07, + "loss": 1.1009, + "step": 34069 + }, + { + "epoch": 2.87, + "learning_rate": 1.0049606695399227e-07, + "loss": 1.086, + "step": 34070 + }, + { + "epoch": 2.87, + "learning_rate": 1.0036756537355585e-07, + "loss": 1.1334, + "step": 34071 + }, + { + "epoch": 2.87, + "learning_rate": 1.0023914558705106e-07, + "loss": 1.1545, + "step": 34072 + }, + { + "epoch": 2.87, + "learning_rate": 1.0011080759553926e-07, + "loss": 1.1143, + "step": 34073 + }, + { + "epoch": 2.87, + "learning_rate": 9.998255140007962e-08, + "loss": 1.1192, + "step": 34074 + }, + { + "epoch": 2.87, + "learning_rate": 9.98543770017324e-08, + "loss": 1.2383, + "step": 34075 + }, + { + "epoch": 2.87, + "learning_rate": 9.972628440155785e-08, + "loss": 1.1246, + "step": 34076 + }, + { + "epoch": 2.87, + "learning_rate": 9.959827360061181e-08, + "loss": 1.1767, + "step": 34077 + }, + { + "epoch": 2.87, + "learning_rate": 9.947034459995453e-08, + "loss": 1.0905, + "step": 34078 + }, + { + "epoch": 2.87, + "learning_rate": 9.934249740064073e-08, + "loss": 1.1032, + "step": 34079 + }, + { + "epoch": 2.87, + "learning_rate": 9.921473200372955e-08, + "loss": 1.2046, + "step": 34080 + }, + { + "epoch": 2.87, + "learning_rate": 9.90870484102735e-08, + "loss": 1.1354, + "step": 34081 + }, + { + "epoch": 2.87, + "learning_rate": 9.895944662132839e-08, + "loss": 1.0853, + "step": 34082 + }, + { + "epoch": 2.87, + "learning_rate": 9.883192663795115e-08, + "loss": 1.162, + "step": 34083 + }, + { + "epoch": 2.87, + "learning_rate": 9.870448846119096e-08, + "loss": 1.0898, + "step": 34084 + }, + { + "epoch": 2.87, + "learning_rate": 9.857713209210473e-08, + "loss": 1.1148, + "step": 34085 + }, + { + "epoch": 2.87, + "learning_rate": 9.844985753174274e-08, + "loss": 1.068, + "step": 34086 + }, + { + "epoch": 2.87, + "learning_rate": 9.832266478115749e-08, + "loss": 1.0256, + "step": 34087 + }, + { + "epoch": 2.87, + "learning_rate": 9.819555384139812e-08, + "loss": 1.1426, + "step": 34088 + }, + { + "epoch": 2.87, + "learning_rate": 9.806852471351603e-08, + "loss": 1.043, + "step": 34089 + }, + { + "epoch": 2.87, + "learning_rate": 9.794157739856147e-08, + "loss": 1.1101, + "step": 34090 + }, + { + "epoch": 2.87, + "learning_rate": 9.781471189758363e-08, + "loss": 1.1219, + "step": 34091 + }, + { + "epoch": 2.87, + "learning_rate": 9.768792821162831e-08, + "loss": 1.021, + "step": 34092 + }, + { + "epoch": 2.87, + "learning_rate": 9.75612263417447e-08, + "loss": 1.076, + "step": 34093 + }, + { + "epoch": 2.87, + "learning_rate": 9.743460628897971e-08, + "loss": 1.1636, + "step": 34094 + }, + { + "epoch": 2.87, + "learning_rate": 9.73080680543781e-08, + "loss": 1.1119, + "step": 34095 + }, + { + "epoch": 2.87, + "learning_rate": 9.718161163898898e-08, + "loss": 1.1881, + "step": 34096 + }, + { + "epoch": 2.87, + "learning_rate": 9.705523704385267e-08, + "loss": 1.0777, + "step": 34097 + }, + { + "epoch": 2.87, + "learning_rate": 9.69289442700161e-08, + "loss": 1.1037, + "step": 34098 + }, + { + "epoch": 2.87, + "learning_rate": 9.680273331852174e-08, + "loss": 1.0415, + "step": 34099 + }, + { + "epoch": 2.87, + "learning_rate": 9.667660419041214e-08, + "loss": 1.1502, + "step": 34100 + }, + { + "epoch": 2.87, + "learning_rate": 9.655055688673087e-08, + "loss": 1.0621, + "step": 34101 + }, + { + "epoch": 2.87, + "learning_rate": 9.642459140851824e-08, + "loss": 1.1938, + "step": 34102 + }, + { + "epoch": 2.87, + "learning_rate": 9.629870775681449e-08, + "loss": 1.0779, + "step": 34103 + }, + { + "epoch": 2.87, + "learning_rate": 9.617290593266104e-08, + "loss": 1.0635, + "step": 34104 + }, + { + "epoch": 2.87, + "learning_rate": 9.604718593709594e-08, + "loss": 1.0342, + "step": 34105 + }, + { + "epoch": 2.87, + "learning_rate": 9.592154777115947e-08, + "loss": 1.172, + "step": 34106 + }, + { + "epoch": 2.87, + "learning_rate": 9.579599143588858e-08, + "loss": 1.1489, + "step": 34107 + }, + { + "epoch": 2.87, + "learning_rate": 9.567051693232132e-08, + "loss": 1.1275, + "step": 34108 + }, + { + "epoch": 2.87, + "learning_rate": 9.554512426149354e-08, + "loss": 1.093, + "step": 34109 + }, + { + "epoch": 2.87, + "learning_rate": 9.541981342444217e-08, + "loss": 1.0812, + "step": 34110 + }, + { + "epoch": 2.87, + "learning_rate": 9.529458442220085e-08, + "loss": 1.1331, + "step": 34111 + }, + { + "epoch": 2.87, + "learning_rate": 9.516943725580762e-08, + "loss": 1.1434, + "step": 34112 + }, + { + "epoch": 2.87, + "learning_rate": 9.504437192629389e-08, + "loss": 1.1075, + "step": 34113 + }, + { + "epoch": 2.87, + "learning_rate": 9.491938843469217e-08, + "loss": 1.1108, + "step": 34114 + }, + { + "epoch": 2.87, + "learning_rate": 9.479448678203828e-08, + "loss": 1.1504, + "step": 34115 + }, + { + "epoch": 2.87, + "learning_rate": 9.466966696936141e-08, + "loss": 1.0991, + "step": 34116 + }, + { + "epoch": 2.87, + "learning_rate": 9.454492899769407e-08, + "loss": 1.128, + "step": 34117 + }, + { + "epoch": 2.87, + "learning_rate": 9.442027286806654e-08, + "loss": 1.1361, + "step": 34118 + }, + { + "epoch": 2.87, + "learning_rate": 9.429569858150911e-08, + "loss": 1.1318, + "step": 34119 + }, + { + "epoch": 2.87, + "learning_rate": 9.417120613905095e-08, + "loss": 1.1094, + "step": 34120 + }, + { + "epoch": 2.87, + "learning_rate": 9.404679554172014e-08, + "loss": 1.0912, + "step": 34121 + }, + { + "epoch": 2.87, + "learning_rate": 9.392246679054473e-08, + "loss": 1.0491, + "step": 34122 + }, + { + "epoch": 2.87, + "learning_rate": 9.37982198865528e-08, + "loss": 1.1924, + "step": 34123 + }, + { + "epoch": 2.87, + "learning_rate": 9.36740548307713e-08, + "loss": 1.1199, + "step": 34124 + }, + { + "epoch": 2.87, + "learning_rate": 9.354997162422386e-08, + "loss": 1.1057, + "step": 34125 + }, + { + "epoch": 2.87, + "learning_rate": 9.342597026793742e-08, + "loss": 1.2346, + "step": 34126 + }, + { + "epoch": 2.87, + "learning_rate": 9.330205076293563e-08, + "loss": 1.1337, + "step": 34127 + }, + { + "epoch": 2.87, + "learning_rate": 9.317821311024322e-08, + "loss": 1.1643, + "step": 34128 + }, + { + "epoch": 2.87, + "learning_rate": 9.30544573108838e-08, + "loss": 1.1458, + "step": 34129 + }, + { + "epoch": 2.87, + "learning_rate": 9.29307833658788e-08, + "loss": 1.0685, + "step": 34130 + }, + { + "epoch": 2.87, + "learning_rate": 9.28071912762496e-08, + "loss": 1.0784, + "step": 34131 + }, + { + "epoch": 2.87, + "learning_rate": 9.268368104301762e-08, + "loss": 1.0498, + "step": 34132 + }, + { + "epoch": 2.87, + "learning_rate": 9.256025266720537e-08, + "loss": 1.0143, + "step": 34133 + }, + { + "epoch": 2.87, + "learning_rate": 9.24369061498298e-08, + "loss": 1.125, + "step": 34134 + }, + { + "epoch": 2.87, + "learning_rate": 9.231364149191236e-08, + "loss": 1.1902, + "step": 34135 + }, + { + "epoch": 2.87, + "learning_rate": 9.219045869446997e-08, + "loss": 1.1269, + "step": 34136 + }, + { + "epoch": 2.87, + "learning_rate": 9.206735775852072e-08, + "loss": 1.0715, + "step": 34137 + }, + { + "epoch": 2.87, + "learning_rate": 9.194433868508268e-08, + "loss": 1.0981, + "step": 34138 + }, + { + "epoch": 2.87, + "learning_rate": 9.18214014751706e-08, + "loss": 1.1272, + "step": 34139 + }, + { + "epoch": 2.87, + "learning_rate": 9.169854612980145e-08, + "loss": 0.9526, + "step": 34140 + }, + { + "epoch": 2.87, + "learning_rate": 9.157577264998995e-08, + "loss": 1.1715, + "step": 34141 + }, + { + "epoch": 2.87, + "learning_rate": 9.145308103674976e-08, + "loss": 1.0858, + "step": 34142 + }, + { + "epoch": 2.87, + "learning_rate": 9.133047129109673e-08, + "loss": 1.1537, + "step": 34143 + }, + { + "epoch": 2.87, + "learning_rate": 9.120794341404226e-08, + "loss": 1.1254, + "step": 34144 + }, + { + "epoch": 2.87, + "learning_rate": 9.108549740659777e-08, + "loss": 1.0461, + "step": 34145 + }, + { + "epoch": 2.87, + "learning_rate": 9.096313326977802e-08, + "loss": 1.1087, + "step": 34146 + }, + { + "epoch": 2.87, + "learning_rate": 9.084085100459105e-08, + "loss": 1.0995, + "step": 34147 + }, + { + "epoch": 2.87, + "learning_rate": 9.071865061204832e-08, + "loss": 1.0241, + "step": 34148 + }, + { + "epoch": 2.88, + "learning_rate": 9.059653209316011e-08, + "loss": 1.183, + "step": 34149 + }, + { + "epoch": 2.88, + "learning_rate": 9.047449544893338e-08, + "loss": 1.1287, + "step": 34150 + }, + { + "epoch": 2.88, + "learning_rate": 9.035254068037957e-08, + "loss": 1.0442, + "step": 34151 + }, + { + "epoch": 2.88, + "learning_rate": 9.023066778850343e-08, + "loss": 1.1083, + "step": 34152 + }, + { + "epoch": 2.88, + "learning_rate": 9.010887677431301e-08, + "loss": 1.0626, + "step": 34153 + }, + { + "epoch": 2.88, + "learning_rate": 8.998716763881532e-08, + "loss": 1.1513, + "step": 34154 + }, + { + "epoch": 2.88, + "learning_rate": 8.98655403830151e-08, + "loss": 1.1149, + "step": 34155 + }, + { + "epoch": 2.88, + "learning_rate": 8.974399500791708e-08, + "loss": 1.0327, + "step": 34156 + }, + { + "epoch": 2.88, + "learning_rate": 8.962253151452605e-08, + "loss": 1.139, + "step": 34157 + }, + { + "epoch": 2.88, + "learning_rate": 8.950114990384562e-08, + "loss": 1.1305, + "step": 34158 + }, + { + "epoch": 2.88, + "learning_rate": 8.937985017687723e-08, + "loss": 1.0892, + "step": 34159 + }, + { + "epoch": 2.88, + "learning_rate": 8.925863233462561e-08, + "loss": 1.1302, + "step": 34160 + }, + { + "epoch": 2.88, + "learning_rate": 8.91374963780911e-08, + "loss": 1.1752, + "step": 34161 + }, + { + "epoch": 2.88, + "learning_rate": 8.9016442308274e-08, + "loss": 1.1742, + "step": 34162 + }, + { + "epoch": 2.88, + "learning_rate": 8.889547012617572e-08, + "loss": 1.0405, + "step": 34163 + }, + { + "epoch": 2.88, + "learning_rate": 8.877457983279325e-08, + "loss": 1.1146, + "step": 34164 + }, + { + "epoch": 2.88, + "learning_rate": 8.865377142912912e-08, + "loss": 1.1055, + "step": 34165 + }, + { + "epoch": 2.88, + "learning_rate": 8.853304491617809e-08, + "loss": 1.1443, + "step": 34166 + }, + { + "epoch": 2.88, + "learning_rate": 8.841240029493937e-08, + "loss": 1.0764, + "step": 34167 + }, + { + "epoch": 2.88, + "learning_rate": 8.829183756640991e-08, + "loss": 1.062, + "step": 34168 + }, + { + "epoch": 2.88, + "learning_rate": 8.817135673158561e-08, + "loss": 0.9858, + "step": 34169 + }, + { + "epoch": 2.88, + "learning_rate": 8.805095779146122e-08, + "loss": 1.0811, + "step": 34170 + }, + { + "epoch": 2.88, + "learning_rate": 8.793064074703261e-08, + "loss": 1.1634, + "step": 34171 + }, + { + "epoch": 2.88, + "learning_rate": 8.781040559929343e-08, + "loss": 1.0148, + "step": 34172 + }, + { + "epoch": 2.88, + "learning_rate": 8.769025234923622e-08, + "loss": 1.0852, + "step": 34173 + }, + { + "epoch": 2.88, + "learning_rate": 8.757018099785575e-08, + "loss": 1.1644, + "step": 34174 + }, + { + "epoch": 2.88, + "learning_rate": 8.74501915461412e-08, + "loss": 1.1225, + "step": 34175 + }, + { + "epoch": 2.88, + "learning_rate": 8.733028399508737e-08, + "loss": 1.1807, + "step": 34176 + }, + { + "epoch": 2.88, + "learning_rate": 8.721045834568232e-08, + "loss": 1.1173, + "step": 34177 + }, + { + "epoch": 2.88, + "learning_rate": 8.709071459891638e-08, + "loss": 1.0848, + "step": 34178 + }, + { + "epoch": 2.88, + "learning_rate": 8.697105275577989e-08, + "loss": 0.988, + "step": 34179 + }, + { + "epoch": 2.88, + "learning_rate": 8.685147281726202e-08, + "loss": 1.1017, + "step": 34180 + }, + { + "epoch": 2.88, + "learning_rate": 8.67319747843487e-08, + "loss": 1.1425, + "step": 34181 + }, + { + "epoch": 2.88, + "learning_rate": 8.661255865802798e-08, + "loss": 1.0887, + "step": 34182 + }, + { + "epoch": 2.88, + "learning_rate": 8.64932244392891e-08, + "loss": 1.1007, + "step": 34183 + }, + { + "epoch": 2.88, + "learning_rate": 8.637397212911346e-08, + "loss": 1.0643, + "step": 34184 + }, + { + "epoch": 2.88, + "learning_rate": 8.625480172849032e-08, + "loss": 1.2102, + "step": 34185 + }, + { + "epoch": 2.88, + "learning_rate": 8.61357132384033e-08, + "loss": 1.1115, + "step": 34186 + }, + { + "epoch": 2.88, + "learning_rate": 8.601670665983497e-08, + "loss": 0.9983, + "step": 34187 + }, + { + "epoch": 2.88, + "learning_rate": 8.589778199376897e-08, + "loss": 1.1049, + "step": 34188 + }, + { + "epoch": 2.88, + "learning_rate": 8.577893924118896e-08, + "loss": 1.0686, + "step": 34189 + }, + { + "epoch": 2.88, + "learning_rate": 8.566017840307639e-08, + "loss": 1.1272, + "step": 34190 + }, + { + "epoch": 2.88, + "learning_rate": 8.554149948041379e-08, + "loss": 1.1393, + "step": 34191 + }, + { + "epoch": 2.88, + "learning_rate": 8.542290247417817e-08, + "loss": 1.1375, + "step": 34192 + }, + { + "epoch": 2.88, + "learning_rate": 8.530438738535319e-08, + "loss": 1.1806, + "step": 34193 + }, + { + "epoch": 2.88, + "learning_rate": 8.518595421491583e-08, + "loss": 1.0465, + "step": 34194 + }, + { + "epoch": 2.88, + "learning_rate": 8.50676029638442e-08, + "loss": 1.0227, + "step": 34195 + }, + { + "epoch": 2.88, + "learning_rate": 8.494933363311863e-08, + "loss": 1.1373, + "step": 34196 + }, + { + "epoch": 2.88, + "learning_rate": 8.483114622371391e-08, + "loss": 1.1143, + "step": 34197 + }, + { + "epoch": 2.88, + "learning_rate": 8.471304073660813e-08, + "loss": 1.0953, + "step": 34198 + }, + { + "epoch": 2.88, + "learning_rate": 8.459501717277718e-08, + "loss": 1.1623, + "step": 34199 + }, + { + "epoch": 2.88, + "learning_rate": 8.447707553319472e-08, + "loss": 1.0409, + "step": 34200 + }, + { + "epoch": 2.88, + "learning_rate": 8.435921581883554e-08, + "loss": 1.0919, + "step": 34201 + }, + { + "epoch": 2.88, + "learning_rate": 8.424143803067553e-08, + "loss": 1.1425, + "step": 34202 + }, + { + "epoch": 2.88, + "learning_rate": 8.412374216968388e-08, + "loss": 1.1109, + "step": 34203 + }, + { + "epoch": 2.88, + "learning_rate": 8.400612823683763e-08, + "loss": 1.0294, + "step": 34204 + }, + { + "epoch": 2.88, + "learning_rate": 8.388859623310485e-08, + "loss": 1.0958, + "step": 34205 + }, + { + "epoch": 2.88, + "learning_rate": 8.377114615945814e-08, + "loss": 1.1474, + "step": 34206 + }, + { + "epoch": 2.88, + "learning_rate": 8.365377801686892e-08, + "loss": 1.1118, + "step": 34207 + }, + { + "epoch": 2.88, + "learning_rate": 8.35364918063053e-08, + "loss": 1.1115, + "step": 34208 + }, + { + "epoch": 2.88, + "learning_rate": 8.341928752873652e-08, + "loss": 1.103, + "step": 34209 + }, + { + "epoch": 2.88, + "learning_rate": 8.330216518513067e-08, + "loss": 1.073, + "step": 34210 + }, + { + "epoch": 2.88, + "learning_rate": 8.318512477645701e-08, + "loss": 1.1087, + "step": 34211 + }, + { + "epoch": 2.88, + "learning_rate": 8.306816630368142e-08, + "loss": 1.1567, + "step": 34212 + }, + { + "epoch": 2.88, + "learning_rate": 8.295128976777089e-08, + "loss": 1.12, + "step": 34213 + }, + { + "epoch": 2.88, + "learning_rate": 8.283449516969022e-08, + "loss": 1.1643, + "step": 34214 + }, + { + "epoch": 2.88, + "learning_rate": 8.27177825104053e-08, + "loss": 1.1635, + "step": 34215 + }, + { + "epoch": 2.88, + "learning_rate": 8.26011517908798e-08, + "loss": 1.1926, + "step": 34216 + }, + { + "epoch": 2.88, + "learning_rate": 8.248460301207628e-08, + "loss": 1.1222, + "step": 34217 + }, + { + "epoch": 2.88, + "learning_rate": 8.236813617496064e-08, + "loss": 1.1051, + "step": 34218 + }, + { + "epoch": 2.88, + "learning_rate": 8.225175128049322e-08, + "loss": 1.1299, + "step": 34219 + }, + { + "epoch": 2.88, + "learning_rate": 8.21354483296355e-08, + "loss": 1.1428, + "step": 34220 + }, + { + "epoch": 2.88, + "learning_rate": 8.201922732334888e-08, + "loss": 1.1171, + "step": 34221 + }, + { + "epoch": 2.88, + "learning_rate": 8.190308826259374e-08, + "loss": 1.1369, + "step": 34222 + }, + { + "epoch": 2.88, + "learning_rate": 8.178703114832931e-08, + "loss": 1.0911, + "step": 34223 + }, + { + "epoch": 2.88, + "learning_rate": 8.167105598151481e-08, + "loss": 1.03, + "step": 34224 + }, + { + "epoch": 2.88, + "learning_rate": 8.155516276310727e-08, + "loss": 1.0229, + "step": 34225 + }, + { + "epoch": 2.88, + "learning_rate": 8.143935149406701e-08, + "loss": 1.0815, + "step": 34226 + }, + { + "epoch": 2.88, + "learning_rate": 8.132362217534884e-08, + "loss": 1.1632, + "step": 34227 + }, + { + "epoch": 2.88, + "learning_rate": 8.120797480790865e-08, + "loss": 1.1948, + "step": 34228 + }, + { + "epoch": 2.88, + "learning_rate": 8.109240939270235e-08, + "loss": 1.2306, + "step": 34229 + }, + { + "epoch": 2.88, + "learning_rate": 8.097692593068585e-08, + "loss": 1.0931, + "step": 34230 + }, + { + "epoch": 2.88, + "learning_rate": 8.08615244228117e-08, + "loss": 1.1136, + "step": 34231 + }, + { + "epoch": 2.88, + "learning_rate": 8.074620487003471e-08, + "loss": 1.1191, + "step": 34232 + }, + { + "epoch": 2.88, + "learning_rate": 8.063096727330744e-08, + "loss": 1.0926, + "step": 34233 + }, + { + "epoch": 2.88, + "learning_rate": 8.051581163358135e-08, + "loss": 1.1567, + "step": 34234 + }, + { + "epoch": 2.88, + "learning_rate": 8.040073795180791e-08, + "loss": 1.0504, + "step": 34235 + }, + { + "epoch": 2.88, + "learning_rate": 8.028574622893859e-08, + "loss": 1.0306, + "step": 34236 + }, + { + "epoch": 2.88, + "learning_rate": 8.017083646592372e-08, + "loss": 1.098, + "step": 34237 + }, + { + "epoch": 2.88, + "learning_rate": 8.005600866371143e-08, + "loss": 1.1034, + "step": 34238 + }, + { + "epoch": 2.88, + "learning_rate": 7.994126282325099e-08, + "loss": 1.0754, + "step": 34239 + }, + { + "epoch": 2.88, + "learning_rate": 7.98265989454905e-08, + "loss": 1.0895, + "step": 34240 + }, + { + "epoch": 2.88, + "learning_rate": 7.971201703137921e-08, + "loss": 1.0647, + "step": 34241 + }, + { + "epoch": 2.88, + "learning_rate": 7.95975170818597e-08, + "loss": 1.0959, + "step": 34242 + }, + { + "epoch": 2.88, + "learning_rate": 7.948309909788232e-08, + "loss": 1.0621, + "step": 34243 + }, + { + "epoch": 2.88, + "learning_rate": 7.936876308038965e-08, + "loss": 1.1112, + "step": 34244 + }, + { + "epoch": 2.88, + "learning_rate": 7.925450903032761e-08, + "loss": 1.0987, + "step": 34245 + }, + { + "epoch": 2.88, + "learning_rate": 7.914033694863988e-08, + "loss": 1.0066, + "step": 34246 + }, + { + "epoch": 2.88, + "learning_rate": 7.902624683626903e-08, + "loss": 1.1193, + "step": 34247 + }, + { + "epoch": 2.88, + "learning_rate": 7.891223869415987e-08, + "loss": 1.0589, + "step": 34248 + }, + { + "epoch": 2.88, + "learning_rate": 7.879831252325165e-08, + "loss": 1.082, + "step": 34249 + }, + { + "epoch": 2.88, + "learning_rate": 7.868446832448695e-08, + "loss": 1.0863, + "step": 34250 + }, + { + "epoch": 2.88, + "learning_rate": 7.857070609880724e-08, + "loss": 1.159, + "step": 34251 + }, + { + "epoch": 2.88, + "learning_rate": 7.845702584715175e-08, + "loss": 1.0548, + "step": 34252 + }, + { + "epoch": 2.88, + "learning_rate": 7.834342757045865e-08, + "loss": 1.1024, + "step": 34253 + }, + { + "epoch": 2.88, + "learning_rate": 7.822991126966828e-08, + "loss": 1.0972, + "step": 34254 + }, + { + "epoch": 2.88, + "learning_rate": 7.811647694571766e-08, + "loss": 1.1226, + "step": 34255 + }, + { + "epoch": 2.88, + "learning_rate": 7.800312459954384e-08, + "loss": 0.9634, + "step": 34256 + }, + { + "epoch": 2.88, + "learning_rate": 7.788985423208384e-08, + "loss": 1.136, + "step": 34257 + }, + { + "epoch": 2.88, + "learning_rate": 7.777666584427246e-08, + "loss": 1.1173, + "step": 34258 + }, + { + "epoch": 2.88, + "learning_rate": 7.766355943704673e-08, + "loss": 1.1055, + "step": 34259 + }, + { + "epoch": 2.88, + "learning_rate": 7.755053501134035e-08, + "loss": 1.1363, + "step": 34260 + }, + { + "epoch": 2.88, + "learning_rate": 7.743759256808702e-08, + "loss": 1.1452, + "step": 34261 + }, + { + "epoch": 2.88, + "learning_rate": 7.732473210821933e-08, + "loss": 1.05, + "step": 34262 + }, + { + "epoch": 2.88, + "learning_rate": 7.721195363267098e-08, + "loss": 1.0648, + "step": 34263 + }, + { + "epoch": 2.88, + "learning_rate": 7.709925714237232e-08, + "loss": 1.0658, + "step": 34264 + }, + { + "epoch": 2.88, + "learning_rate": 7.698664263825595e-08, + "loss": 1.101, + "step": 34265 + }, + { + "epoch": 2.88, + "learning_rate": 7.687411012125223e-08, + "loss": 1.1342, + "step": 34266 + }, + { + "epoch": 2.88, + "learning_rate": 7.676165959228931e-08, + "loss": 1.1643, + "step": 34267 + }, + { + "epoch": 2.89, + "learning_rate": 7.664929105229868e-08, + "loss": 1.1395, + "step": 34268 + }, + { + "epoch": 2.89, + "learning_rate": 7.653700450220625e-08, + "loss": 1.0875, + "step": 34269 + }, + { + "epoch": 2.89, + "learning_rate": 7.642479994294128e-08, + "loss": 1.0199, + "step": 34270 + }, + { + "epoch": 2.89, + "learning_rate": 7.631267737543079e-08, + "loss": 1.1839, + "step": 34271 + }, + { + "epoch": 2.89, + "learning_rate": 7.620063680060074e-08, + "loss": 1.1324, + "step": 34272 + }, + { + "epoch": 2.89, + "learning_rate": 7.608867821937704e-08, + "loss": 1.1572, + "step": 34273 + }, + { + "epoch": 2.89, + "learning_rate": 7.597680163268562e-08, + "loss": 1.1502, + "step": 34274 + }, + { + "epoch": 2.89, + "learning_rate": 7.586500704144905e-08, + "loss": 1.1157, + "step": 34275 + }, + { + "epoch": 2.89, + "learning_rate": 7.575329444659219e-08, + "loss": 1.0765, + "step": 34276 + }, + { + "epoch": 2.89, + "learning_rate": 7.564166384903759e-08, + "loss": 1.0798, + "step": 34277 + }, + { + "epoch": 2.89, + "learning_rate": 7.553011524970788e-08, + "loss": 1.0611, + "step": 34278 + }, + { + "epoch": 2.89, + "learning_rate": 7.541864864952453e-08, + "loss": 1.1469, + "step": 34279 + }, + { + "epoch": 2.89, + "learning_rate": 7.530726404940903e-08, + "loss": 1.1804, + "step": 34280 + }, + { + "epoch": 2.89, + "learning_rate": 7.519596145028063e-08, + "loss": 1.1518, + "step": 34281 + }, + { + "epoch": 2.89, + "learning_rate": 7.508474085305973e-08, + "loss": 1.0502, + "step": 34282 + }, + { + "epoch": 2.89, + "learning_rate": 7.497360225866556e-08, + "loss": 1.0626, + "step": 34283 + }, + { + "epoch": 2.89, + "learning_rate": 7.486254566801521e-08, + "loss": 1.1492, + "step": 34284 + }, + { + "epoch": 2.89, + "learning_rate": 7.475157108202791e-08, + "loss": 1.1475, + "step": 34285 + }, + { + "epoch": 2.89, + "learning_rate": 7.464067850161849e-08, + "loss": 1.1161, + "step": 34286 + }, + { + "epoch": 2.89, + "learning_rate": 7.45298679277051e-08, + "loss": 1.1957, + "step": 34287 + }, + { + "epoch": 2.89, + "learning_rate": 7.441913936120149e-08, + "loss": 1.0992, + "step": 34288 + }, + { + "epoch": 2.89, + "learning_rate": 7.430849280302465e-08, + "loss": 1.121, + "step": 34289 + }, + { + "epoch": 2.89, + "learning_rate": 7.419792825408723e-08, + "loss": 1.0567, + "step": 34290 + }, + { + "epoch": 2.89, + "learning_rate": 7.40874457153029e-08, + "loss": 1.0318, + "step": 34291 + }, + { + "epoch": 2.89, + "learning_rate": 7.397704518758542e-08, + "loss": 1.0407, + "step": 34292 + }, + { + "epoch": 2.89, + "learning_rate": 7.386672667184624e-08, + "loss": 1.0786, + "step": 34293 + }, + { + "epoch": 2.89, + "learning_rate": 7.375649016899688e-08, + "loss": 1.1206, + "step": 34294 + }, + { + "epoch": 2.89, + "learning_rate": 7.364633567994661e-08, + "loss": 1.0968, + "step": 34295 + }, + { + "epoch": 2.89, + "learning_rate": 7.353626320560914e-08, + "loss": 1.1772, + "step": 34296 + }, + { + "epoch": 2.89, + "learning_rate": 7.342627274689152e-08, + "loss": 1.1537, + "step": 34297 + }, + { + "epoch": 2.89, + "learning_rate": 7.331636430470302e-08, + "loss": 1.1458, + "step": 34298 + }, + { + "epoch": 2.89, + "learning_rate": 7.320653787995069e-08, + "loss": 1.0849, + "step": 34299 + }, + { + "epoch": 2.89, + "learning_rate": 7.309679347354381e-08, + "loss": 1.0856, + "step": 34300 + }, + { + "epoch": 2.89, + "learning_rate": 7.298713108638832e-08, + "loss": 1.1207, + "step": 34301 + }, + { + "epoch": 2.89, + "learning_rate": 7.287755071939017e-08, + "loss": 1.0824, + "step": 34302 + }, + { + "epoch": 2.89, + "learning_rate": 7.276805237345419e-08, + "loss": 1.1726, + "step": 34303 + }, + { + "epoch": 2.89, + "learning_rate": 7.26586360494863e-08, + "loss": 1.1653, + "step": 34304 + }, + { + "epoch": 2.89, + "learning_rate": 7.254930174839026e-08, + "loss": 1.1544, + "step": 34305 + }, + { + "epoch": 2.89, + "learning_rate": 7.244004947106752e-08, + "loss": 1.0875, + "step": 34306 + }, + { + "epoch": 2.89, + "learning_rate": 7.233087921842408e-08, + "loss": 1.1521, + "step": 34307 + }, + { + "epoch": 2.89, + "learning_rate": 7.222179099135917e-08, + "loss": 1.1338, + "step": 34308 + }, + { + "epoch": 2.89, + "learning_rate": 7.211278479077543e-08, + "loss": 1.1799, + "step": 34309 + }, + { + "epoch": 2.89, + "learning_rate": 7.200386061757214e-08, + "loss": 1.0905, + "step": 34310 + }, + { + "epoch": 2.89, + "learning_rate": 7.18950184726519e-08, + "loss": 1.1249, + "step": 34311 + }, + { + "epoch": 2.89, + "learning_rate": 7.178625835691177e-08, + "loss": 1.0671, + "step": 34312 + }, + { + "epoch": 2.89, + "learning_rate": 7.167758027125105e-08, + "loss": 1.1929, + "step": 34313 + }, + { + "epoch": 2.89, + "learning_rate": 7.156898421656677e-08, + "loss": 1.098, + "step": 34314 + }, + { + "epoch": 2.89, + "learning_rate": 7.146047019375935e-08, + "loss": 1.1882, + "step": 34315 + }, + { + "epoch": 2.89, + "learning_rate": 7.135203820372138e-08, + "loss": 1.1727, + "step": 34316 + }, + { + "epoch": 2.89, + "learning_rate": 7.124368824734994e-08, + "loss": 1.1494, + "step": 34317 + }, + { + "epoch": 2.89, + "learning_rate": 7.11354203255421e-08, + "loss": 1.0694, + "step": 34318 + }, + { + "epoch": 2.89, + "learning_rate": 7.102723443919047e-08, + "loss": 1.1148, + "step": 34319 + }, + { + "epoch": 2.89, + "learning_rate": 7.091913058918876e-08, + "loss": 1.0185, + "step": 34320 + }, + { + "epoch": 2.89, + "learning_rate": 7.081110877643183e-08, + "loss": 1.1953, + "step": 34321 + }, + { + "epoch": 2.89, + "learning_rate": 7.070316900181118e-08, + "loss": 1.1144, + "step": 34322 + }, + { + "epoch": 2.89, + "learning_rate": 7.059531126621832e-08, + "loss": 1.1052, + "step": 34323 + }, + { + "epoch": 2.89, + "learning_rate": 7.048753557054588e-08, + "loss": 1.1677, + "step": 34324 + }, + { + "epoch": 2.89, + "learning_rate": 7.037984191568203e-08, + "loss": 1.06, + "step": 34325 + }, + { + "epoch": 2.89, + "learning_rate": 7.027223030251829e-08, + "loss": 1.0953, + "step": 34326 + }, + { + "epoch": 2.89, + "learning_rate": 7.016470073194392e-08, + "loss": 1.1046, + "step": 34327 + }, + { + "epoch": 2.89, + "learning_rate": 7.005725320484602e-08, + "loss": 1.1636, + "step": 34328 + }, + { + "epoch": 2.89, + "learning_rate": 6.994988772211275e-08, + "loss": 1.1351, + "step": 34329 + }, + { + "epoch": 2.89, + "learning_rate": 6.98426042846323e-08, + "loss": 1.1753, + "step": 34330 + }, + { + "epoch": 2.89, + "learning_rate": 6.973540289328951e-08, + "loss": 1.1548, + "step": 34331 + }, + { + "epoch": 2.89, + "learning_rate": 6.962828354897144e-08, + "loss": 1.1209, + "step": 34332 + }, + { + "epoch": 2.89, + "learning_rate": 6.952124625256296e-08, + "loss": 1.1072, + "step": 34333 + }, + { + "epoch": 2.89, + "learning_rate": 6.941429100494778e-08, + "loss": 1.1733, + "step": 34334 + }, + { + "epoch": 2.89, + "learning_rate": 6.930741780700966e-08, + "loss": 1.1156, + "step": 34335 + }, + { + "epoch": 2.89, + "learning_rate": 6.920062665963123e-08, + "loss": 1.1456, + "step": 34336 + }, + { + "epoch": 2.89, + "learning_rate": 6.909391756369732e-08, + "loss": 1.1489, + "step": 34337 + }, + { + "epoch": 2.89, + "learning_rate": 6.898729052008723e-08, + "loss": 1.1744, + "step": 34338 + }, + { + "epoch": 2.89, + "learning_rate": 6.888074552968138e-08, + "loss": 1.0925, + "step": 34339 + }, + { + "epoch": 2.89, + "learning_rate": 6.877428259336239e-08, + "loss": 1.1006, + "step": 34340 + }, + { + "epoch": 2.89, + "learning_rate": 6.866790171200843e-08, + "loss": 1.1704, + "step": 34341 + }, + { + "epoch": 2.89, + "learning_rate": 6.856160288649771e-08, + "loss": 1.1087, + "step": 34342 + }, + { + "epoch": 2.89, + "learning_rate": 6.845538611771063e-08, + "loss": 1.1295, + "step": 34343 + }, + { + "epoch": 2.89, + "learning_rate": 6.834925140652315e-08, + "loss": 1.1164, + "step": 34344 + }, + { + "epoch": 2.89, + "learning_rate": 6.824319875381235e-08, + "loss": 1.1016, + "step": 34345 + }, + { + "epoch": 2.89, + "learning_rate": 6.813722816045532e-08, + "loss": 1.1014, + "step": 34346 + }, + { + "epoch": 2.89, + "learning_rate": 6.803133962732689e-08, + "loss": 1.0655, + "step": 34347 + }, + { + "epoch": 2.89, + "learning_rate": 6.792553315530192e-08, + "loss": 1.0635, + "step": 34348 + }, + { + "epoch": 2.89, + "learning_rate": 6.78198087452553e-08, + "loss": 1.1061, + "step": 34349 + }, + { + "epoch": 2.89, + "learning_rate": 6.771416639805961e-08, + "loss": 1.1542, + "step": 34350 + }, + { + "epoch": 2.89, + "learning_rate": 6.760860611458864e-08, + "loss": 1.1463, + "step": 34351 + }, + { + "epoch": 2.89, + "learning_rate": 6.7503127895715e-08, + "loss": 1.0975, + "step": 34352 + }, + { + "epoch": 2.89, + "learning_rate": 6.739773174230801e-08, + "loss": 1.0731, + "step": 34353 + }, + { + "epoch": 2.89, + "learning_rate": 6.72924176552403e-08, + "loss": 1.0859, + "step": 34354 + }, + { + "epoch": 2.89, + "learning_rate": 6.718718563538119e-08, + "loss": 1.0551, + "step": 34355 + }, + { + "epoch": 2.89, + "learning_rate": 6.708203568360105e-08, + "loss": 1.117, + "step": 34356 + }, + { + "epoch": 2.89, + "learning_rate": 6.697696780076812e-08, + "loss": 1.1174, + "step": 34357 + }, + { + "epoch": 2.89, + "learning_rate": 6.687198198774947e-08, + "loss": 1.1898, + "step": 34358 + }, + { + "epoch": 2.89, + "learning_rate": 6.67670782454144e-08, + "loss": 1.0721, + "step": 34359 + }, + { + "epoch": 2.89, + "learning_rate": 6.666225657462777e-08, + "loss": 1.0626, + "step": 34360 + }, + { + "epoch": 2.89, + "learning_rate": 6.655751697625779e-08, + "loss": 1.1019, + "step": 34361 + }, + { + "epoch": 2.89, + "learning_rate": 6.645285945116819e-08, + "loss": 1.1034, + "step": 34362 + }, + { + "epoch": 2.89, + "learning_rate": 6.634828400022497e-08, + "loss": 1.1782, + "step": 34363 + }, + { + "epoch": 2.89, + "learning_rate": 6.624379062428965e-08, + "loss": 1.0999, + "step": 34364 + }, + { + "epoch": 2.89, + "learning_rate": 6.613937932422932e-08, + "loss": 1.1763, + "step": 34365 + }, + { + "epoch": 2.89, + "learning_rate": 6.603505010090328e-08, + "loss": 1.0742, + "step": 34366 + }, + { + "epoch": 2.89, + "learning_rate": 6.59308029551764e-08, + "loss": 1.1145, + "step": 34367 + }, + { + "epoch": 2.89, + "learning_rate": 6.582663788790688e-08, + "loss": 1.1237, + "step": 34368 + }, + { + "epoch": 2.89, + "learning_rate": 6.572255489995737e-08, + "loss": 1.1288, + "step": 34369 + }, + { + "epoch": 2.89, + "learning_rate": 6.56185539921883e-08, + "loss": 1.0334, + "step": 34370 + }, + { + "epoch": 2.89, + "learning_rate": 6.551463516545787e-08, + "loss": 1.0874, + "step": 34371 + }, + { + "epoch": 2.89, + "learning_rate": 6.541079842062425e-08, + "loss": 1.1636, + "step": 34372 + }, + { + "epoch": 2.89, + "learning_rate": 6.530704375854679e-08, + "loss": 1.0049, + "step": 34373 + }, + { + "epoch": 2.89, + "learning_rate": 6.520337118008258e-08, + "loss": 1.1413, + "step": 34374 + }, + { + "epoch": 2.89, + "learning_rate": 6.509978068608647e-08, + "loss": 1.1354, + "step": 34375 + }, + { + "epoch": 2.89, + "learning_rate": 6.499627227741667e-08, + "loss": 1.1464, + "step": 34376 + }, + { + "epoch": 2.89, + "learning_rate": 6.489284595492584e-08, + "loss": 1.0774, + "step": 34377 + }, + { + "epoch": 2.89, + "learning_rate": 6.478950171947107e-08, + "loss": 1.1359, + "step": 34378 + }, + { + "epoch": 2.89, + "learning_rate": 6.468623957190501e-08, + "loss": 1.0425, + "step": 34379 + }, + { + "epoch": 2.89, + "learning_rate": 6.45830595130803e-08, + "loss": 1.1047, + "step": 34380 + }, + { + "epoch": 2.89, + "learning_rate": 6.447996154385072e-08, + "loss": 1.1397, + "step": 34381 + }, + { + "epoch": 2.89, + "learning_rate": 6.437694566506669e-08, + "loss": 1.0388, + "step": 34382 + }, + { + "epoch": 2.89, + "learning_rate": 6.427401187758087e-08, + "loss": 1.129, + "step": 34383 + }, + { + "epoch": 2.89, + "learning_rate": 6.417116018224256e-08, + "loss": 1.145, + "step": 34384 + }, + { + "epoch": 2.89, + "learning_rate": 6.406839057990333e-08, + "loss": 1.1009, + "step": 34385 + }, + { + "epoch": 2.9, + "learning_rate": 6.396570307140915e-08, + "loss": 1.0119, + "step": 34386 + }, + { + "epoch": 2.9, + "learning_rate": 6.386309765761157e-08, + "loss": 1.1869, + "step": 34387 + }, + { + "epoch": 2.9, + "learning_rate": 6.376057433935656e-08, + "loss": 1.0845, + "step": 34388 + }, + { + "epoch": 2.9, + "learning_rate": 6.365813311749124e-08, + "loss": 1.117, + "step": 34389 + }, + { + "epoch": 2.9, + "learning_rate": 6.355577399286383e-08, + "loss": 1.1286, + "step": 34390 + }, + { + "epoch": 2.9, + "learning_rate": 6.345349696631809e-08, + "loss": 1.1599, + "step": 34391 + }, + { + "epoch": 2.9, + "learning_rate": 6.335130203869888e-08, + "loss": 1.1628, + "step": 34392 + }, + { + "epoch": 2.9, + "learning_rate": 6.324918921085221e-08, + "loss": 1.1374, + "step": 34393 + }, + { + "epoch": 2.9, + "learning_rate": 6.314715848361964e-08, + "loss": 1.0693, + "step": 34394 + }, + { + "epoch": 2.9, + "learning_rate": 6.304520985784601e-08, + "loss": 1.1357, + "step": 34395 + }, + { + "epoch": 2.9, + "learning_rate": 6.294334333437402e-08, + "loss": 1.1857, + "step": 34396 + }, + { + "epoch": 2.9, + "learning_rate": 6.284155891404298e-08, + "loss": 1.0435, + "step": 34397 + }, + { + "epoch": 2.9, + "learning_rate": 6.273985659769666e-08, + "loss": 1.1651, + "step": 34398 + }, + { + "epoch": 2.9, + "learning_rate": 6.263823638617328e-08, + "loss": 1.0504, + "step": 34399 + }, + { + "epoch": 2.9, + "learning_rate": 6.253669828031327e-08, + "loss": 1.0954, + "step": 34400 + }, + { + "epoch": 2.9, + "learning_rate": 6.243524228095488e-08, + "loss": 1.0498, + "step": 34401 + }, + { + "epoch": 2.9, + "learning_rate": 6.233386838893852e-08, + "loss": 1.1712, + "step": 34402 + }, + { + "epoch": 2.9, + "learning_rate": 6.223257660509797e-08, + "loss": 1.1639, + "step": 34403 + }, + { + "epoch": 2.9, + "learning_rate": 6.213136693027477e-08, + "loss": 1.0693, + "step": 34404 + }, + { + "epoch": 2.9, + "learning_rate": 6.203023936530051e-08, + "loss": 1.0811, + "step": 34405 + }, + { + "epoch": 2.9, + "learning_rate": 6.192919391101449e-08, + "loss": 1.09, + "step": 34406 + }, + { + "epoch": 2.9, + "learning_rate": 6.182823056824938e-08, + "loss": 1.1378, + "step": 34407 + }, + { + "epoch": 2.9, + "learning_rate": 6.172734933784009e-08, + "loss": 1.0754, + "step": 34408 + }, + { + "epoch": 2.9, + "learning_rate": 6.162655022062037e-08, + "loss": 1.1877, + "step": 34409 + }, + { + "epoch": 2.9, + "learning_rate": 6.15258332174229e-08, + "loss": 1.0654, + "step": 34410 + }, + { + "epoch": 2.9, + "learning_rate": 6.142519832907922e-08, + "loss": 1.0819, + "step": 34411 + }, + { + "epoch": 2.9, + "learning_rate": 6.132464555642093e-08, + "loss": 1.108, + "step": 34412 + }, + { + "epoch": 2.9, + "learning_rate": 6.122417490028066e-08, + "loss": 1.0978, + "step": 34413 + }, + { + "epoch": 2.9, + "learning_rate": 6.112378636148553e-08, + "loss": 1.1436, + "step": 34414 + }, + { + "epoch": 2.9, + "learning_rate": 6.102347994086822e-08, + "loss": 1.0788, + "step": 34415 + }, + { + "epoch": 2.9, + "learning_rate": 6.092325563925471e-08, + "loss": 1.151, + "step": 34416 + }, + { + "epoch": 2.9, + "learning_rate": 6.082311345747438e-08, + "loss": 1.0875, + "step": 34417 + }, + { + "epoch": 2.9, + "learning_rate": 6.072305339635431e-08, + "loss": 1.0311, + "step": 34418 + }, + { + "epoch": 2.9, + "learning_rate": 6.062307545672053e-08, + "loss": 1.0776, + "step": 34419 + }, + { + "epoch": 2.9, + "learning_rate": 6.052317963940124e-08, + "loss": 1.1427, + "step": 34420 + }, + { + "epoch": 2.9, + "learning_rate": 6.042336594521914e-08, + "loss": 1.1175, + "step": 34421 + }, + { + "epoch": 2.9, + "learning_rate": 6.032363437500133e-08, + "loss": 1.0336, + "step": 34422 + }, + { + "epoch": 2.9, + "learning_rate": 6.022398492957049e-08, + "loss": 1.1087, + "step": 34423 + }, + { + "epoch": 2.9, + "learning_rate": 6.012441760975041e-08, + "loss": 1.0039, + "step": 34424 + }, + { + "epoch": 2.9, + "learning_rate": 6.002493241636376e-08, + "loss": 1.1199, + "step": 34425 + }, + { + "epoch": 2.9, + "learning_rate": 5.99255293502321e-08, + "loss": 1.0966, + "step": 34426 + }, + { + "epoch": 2.9, + "learning_rate": 5.982620841217701e-08, + "loss": 1.1542, + "step": 34427 + }, + { + "epoch": 2.9, + "learning_rate": 5.972696960301894e-08, + "loss": 1.1631, + "step": 34428 + }, + { + "epoch": 2.9, + "learning_rate": 5.962781292357723e-08, + "loss": 1.2162, + "step": 34429 + }, + { + "epoch": 2.9, + "learning_rate": 5.9528738374672324e-08, + "loss": 1.1273, + "step": 34430 + }, + { + "epoch": 2.9, + "learning_rate": 5.942974595712247e-08, + "loss": 1.2007, + "step": 34431 + }, + { + "epoch": 2.9, + "learning_rate": 5.933083567174591e-08, + "loss": 1.1713, + "step": 34432 + }, + { + "epoch": 2.9, + "learning_rate": 5.9232007519358645e-08, + "loss": 1.0999, + "step": 34433 + }, + { + "epoch": 2.9, + "learning_rate": 5.913326150077781e-08, + "loss": 1.0842, + "step": 34434 + }, + { + "epoch": 2.9, + "learning_rate": 5.903459761682051e-08, + "loss": 1.108, + "step": 34435 + }, + { + "epoch": 2.9, + "learning_rate": 5.8936015868300556e-08, + "loss": 1.0867, + "step": 34436 + }, + { + "epoch": 2.9, + "learning_rate": 5.8837516256032846e-08, + "loss": 1.0405, + "step": 34437 + }, + { + "epoch": 2.9, + "learning_rate": 5.873909878083006e-08, + "loss": 1.2221, + "step": 34438 + }, + { + "epoch": 2.9, + "learning_rate": 5.864076344350711e-08, + "loss": 1.1604, + "step": 34439 + }, + { + "epoch": 2.9, + "learning_rate": 5.854251024487667e-08, + "loss": 1.1022, + "step": 34440 + }, + { + "epoch": 2.9, + "learning_rate": 5.844433918574921e-08, + "loss": 1.0942, + "step": 34441 + }, + { + "epoch": 2.9, + "learning_rate": 5.834625026693519e-08, + "loss": 1.0989, + "step": 34442 + }, + { + "epoch": 2.9, + "learning_rate": 5.824824348924729e-08, + "loss": 1.089, + "step": 34443 + }, + { + "epoch": 2.9, + "learning_rate": 5.815031885349487e-08, + "loss": 1.0307, + "step": 34444 + }, + { + "epoch": 2.9, + "learning_rate": 5.805247636048394e-08, + "loss": 1.0601, + "step": 34445 + }, + { + "epoch": 2.9, + "learning_rate": 5.79547160110272e-08, + "loss": 1.0697, + "step": 34446 + }, + { + "epoch": 2.9, + "learning_rate": 5.785703780593066e-08, + "loss": 1.0516, + "step": 34447 + }, + { + "epoch": 2.9, + "learning_rate": 5.7759441746000346e-08, + "loss": 1.0866, + "step": 34448 + }, + { + "epoch": 2.9, + "learning_rate": 5.7661927832043386e-08, + "loss": 1.0399, + "step": 34449 + }, + { + "epoch": 2.9, + "learning_rate": 5.756449606486581e-08, + "loss": 1.0582, + "step": 34450 + }, + { + "epoch": 2.9, + "learning_rate": 5.7467146445272516e-08, + "loss": 1.2145, + "step": 34451 + }, + { + "epoch": 2.9, + "learning_rate": 5.736987897406732e-08, + "loss": 1.16, + "step": 34452 + }, + { + "epoch": 2.9, + "learning_rate": 5.7272693652054014e-08, + "loss": 1.1206, + "step": 34453 + }, + { + "epoch": 2.9, + "learning_rate": 5.7175590480036405e-08, + "loss": 1.0747, + "step": 34454 + }, + { + "epoch": 2.9, + "learning_rate": 5.707856945881607e-08, + "loss": 1.0979, + "step": 34455 + }, + { + "epoch": 2.9, + "learning_rate": 5.698163058919459e-08, + "loss": 1.1803, + "step": 34456 + }, + { + "epoch": 2.9, + "learning_rate": 5.6884773871973554e-08, + "loss": 1.1033, + "step": 34457 + }, + { + "epoch": 2.9, + "learning_rate": 5.6787999307952316e-08, + "loss": 1.1012, + "step": 34458 + }, + { + "epoch": 2.9, + "learning_rate": 5.669130689793023e-08, + "loss": 1.096, + "step": 34459 + }, + { + "epoch": 2.9, + "learning_rate": 5.659469664270778e-08, + "loss": 1.149, + "step": 34460 + }, + { + "epoch": 2.9, + "learning_rate": 5.649816854308099e-08, + "loss": 1.1076, + "step": 34461 + }, + { + "epoch": 2.9, + "learning_rate": 5.6401722599850325e-08, + "loss": 1.2123, + "step": 34462 + }, + { + "epoch": 2.9, + "learning_rate": 5.630535881380961e-08, + "loss": 1.1992, + "step": 34463 + }, + { + "epoch": 2.9, + "learning_rate": 5.620907718575708e-08, + "loss": 1.0306, + "step": 34464 + }, + { + "epoch": 2.9, + "learning_rate": 5.611287771648766e-08, + "loss": 1.0116, + "step": 34465 + }, + { + "epoch": 2.9, + "learning_rate": 5.601676040679627e-08, + "loss": 1.1157, + "step": 34466 + }, + { + "epoch": 2.9, + "learning_rate": 5.592072525747672e-08, + "loss": 1.0692, + "step": 34467 + }, + { + "epoch": 2.9, + "learning_rate": 5.58247722693217e-08, + "loss": 1.0995, + "step": 34468 + }, + { + "epoch": 2.9, + "learning_rate": 5.572890144312615e-08, + "loss": 1.0826, + "step": 34469 + }, + { + "epoch": 2.9, + "learning_rate": 5.563311277968053e-08, + "loss": 1.0753, + "step": 34470 + }, + { + "epoch": 2.9, + "learning_rate": 5.553740627977644e-08, + "loss": 1.0773, + "step": 34471 + }, + { + "epoch": 2.9, + "learning_rate": 5.544178194420546e-08, + "loss": 1.0439, + "step": 34472 + }, + { + "epoch": 2.9, + "learning_rate": 5.534623977375697e-08, + "loss": 1.1202, + "step": 34473 + }, + { + "epoch": 2.9, + "learning_rate": 5.5250779769220334e-08, + "loss": 1.0902, + "step": 34474 + }, + { + "epoch": 2.9, + "learning_rate": 5.5155401931384914e-08, + "loss": 1.1223, + "step": 34475 + }, + { + "epoch": 2.9, + "learning_rate": 5.506010626103786e-08, + "loss": 1.0558, + "step": 34476 + }, + { + "epoch": 2.9, + "learning_rate": 5.4964892758967436e-08, + "loss": 1.0913, + "step": 34477 + }, + { + "epoch": 2.9, + "learning_rate": 5.4869761425959675e-08, + "loss": 1.0872, + "step": 34478 + }, + { + "epoch": 2.9, + "learning_rate": 5.477471226280062e-08, + "loss": 1.0974, + "step": 34479 + }, + { + "epoch": 2.9, + "learning_rate": 5.467974527027631e-08, + "loss": 1.1682, + "step": 34480 + }, + { + "epoch": 2.9, + "learning_rate": 5.458486044916944e-08, + "loss": 1.08, + "step": 34481 + }, + { + "epoch": 2.9, + "learning_rate": 5.449005780026717e-08, + "loss": 1.1005, + "step": 34482 + }, + { + "epoch": 2.9, + "learning_rate": 5.439533732434998e-08, + "loss": 1.0022, + "step": 34483 + }, + { + "epoch": 2.9, + "learning_rate": 5.43006990222017e-08, + "loss": 1.1022, + "step": 34484 + }, + { + "epoch": 2.9, + "learning_rate": 5.420614289460502e-08, + "loss": 1.1896, + "step": 34485 + }, + { + "epoch": 2.9, + "learning_rate": 5.411166894233821e-08, + "loss": 1.2204, + "step": 34486 + }, + { + "epoch": 2.9, + "learning_rate": 5.401727716618621e-08, + "loss": 1.0979, + "step": 34487 + }, + { + "epoch": 2.9, + "learning_rate": 5.392296756692505e-08, + "loss": 1.1148, + "step": 34488 + }, + { + "epoch": 2.9, + "learning_rate": 5.3828740145336343e-08, + "loss": 1.0814, + "step": 34489 + }, + { + "epoch": 2.9, + "learning_rate": 5.373459490219723e-08, + "loss": 1.0388, + "step": 34490 + }, + { + "epoch": 2.9, + "learning_rate": 5.364053183828599e-08, + "loss": 1.1004, + "step": 34491 + }, + { + "epoch": 2.9, + "learning_rate": 5.3546550954379774e-08, + "loss": 1.1553, + "step": 34492 + }, + { + "epoch": 2.9, + "learning_rate": 5.345265225125573e-08, + "loss": 1.0751, + "step": 34493 + }, + { + "epoch": 2.9, + "learning_rate": 5.3358835729688806e-08, + "loss": 1.0324, + "step": 34494 + }, + { + "epoch": 2.9, + "learning_rate": 5.326510139045504e-08, + "loss": 1.1187, + "step": 34495 + }, + { + "epoch": 2.9, + "learning_rate": 5.3171449234328264e-08, + "loss": 1.1176, + "step": 34496 + }, + { + "epoch": 2.9, + "learning_rate": 5.30778792620823e-08, + "loss": 1.1354, + "step": 34497 + }, + { + "epoch": 2.9, + "learning_rate": 5.298439147448986e-08, + "loss": 1.1713, + "step": 34498 + }, + { + "epoch": 2.9, + "learning_rate": 5.289098587232366e-08, + "loss": 1.1076, + "step": 34499 + }, + { + "epoch": 2.9, + "learning_rate": 5.279766245635642e-08, + "loss": 1.2007, + "step": 34500 + }, + { + "epoch": 2.9, + "learning_rate": 5.270442122735864e-08, + "loss": 1.1149, + "step": 34501 + }, + { + "epoch": 2.9, + "learning_rate": 5.261126218609969e-08, + "loss": 1.0979, + "step": 34502 + }, + { + "epoch": 2.9, + "learning_rate": 5.251818533335007e-08, + "loss": 1.1662, + "step": 34503 + }, + { + "epoch": 2.9, + "learning_rate": 5.2425190669879165e-08, + "loss": 1.1494, + "step": 34504 + }, + { + "epoch": 2.91, + "learning_rate": 5.2332278196455254e-08, + "loss": 1.1126, + "step": 34505 + }, + { + "epoch": 2.91, + "learning_rate": 5.22394479138455e-08, + "loss": 1.2059, + "step": 34506 + }, + { + "epoch": 2.91, + "learning_rate": 5.214669982281595e-08, + "loss": 1.1352, + "step": 34507 + }, + { + "epoch": 2.91, + "learning_rate": 5.205403392413599e-08, + "loss": 0.9933, + "step": 34508 + }, + { + "epoch": 2.91, + "learning_rate": 5.196145021856836e-08, + "loss": 1.1128, + "step": 34509 + }, + { + "epoch": 2.91, + "learning_rate": 5.1868948706877975e-08, + "loss": 1.0422, + "step": 34510 + }, + { + "epoch": 2.91, + "learning_rate": 5.177652938983202e-08, + "loss": 1.1528, + "step": 34511 + }, + { + "epoch": 2.91, + "learning_rate": 5.168419226819099e-08, + "loss": 1.1188, + "step": 34512 + }, + { + "epoch": 2.91, + "learning_rate": 5.159193734271983e-08, + "loss": 1.0903, + "step": 34513 + }, + { + "epoch": 2.91, + "learning_rate": 5.1499764614179046e-08, + "loss": 1.0298, + "step": 34514 + }, + { + "epoch": 2.91, + "learning_rate": 5.140767408333247e-08, + "loss": 1.1076, + "step": 34515 + }, + { + "epoch": 2.91, + "learning_rate": 5.131566575093838e-08, + "loss": 1.0814, + "step": 34516 + }, + { + "epoch": 2.91, + "learning_rate": 5.122373961775951e-08, + "loss": 1.0806, + "step": 34517 + }, + { + "epoch": 2.91, + "learning_rate": 5.113189568455301e-08, + "loss": 1.1606, + "step": 34518 + }, + { + "epoch": 2.91, + "learning_rate": 5.1040133952079404e-08, + "loss": 1.0837, + "step": 34519 + }, + { + "epoch": 2.91, + "learning_rate": 5.094845442109586e-08, + "loss": 1.2196, + "step": 34520 + }, + { + "epoch": 2.91, + "learning_rate": 5.085685709236066e-08, + "loss": 1.0756, + "step": 34521 + }, + { + "epoch": 2.91, + "learning_rate": 5.0765341966629855e-08, + "loss": 1.1235, + "step": 34522 + }, + { + "epoch": 2.91, + "learning_rate": 5.067390904466063e-08, + "loss": 1.0479, + "step": 34523 + }, + { + "epoch": 2.91, + "learning_rate": 5.0582558327206824e-08, + "loss": 1.1105, + "step": 34524 + }, + { + "epoch": 2.91, + "learning_rate": 5.0491289815024493e-08, + "loss": 1.0942, + "step": 34525 + }, + { + "epoch": 2.91, + "learning_rate": 5.0400103508868593e-08, + "loss": 1.0561, + "step": 34526 + }, + { + "epoch": 2.91, + "learning_rate": 5.0308999409489635e-08, + "loss": 1.1059, + "step": 34527 + }, + { + "epoch": 2.91, + "learning_rate": 5.021797751764257e-08, + "loss": 1.1539, + "step": 34528 + }, + { + "epoch": 2.91, + "learning_rate": 5.012703783407902e-08, + "loss": 1.1386, + "step": 34529 + }, + { + "epoch": 2.91, + "learning_rate": 5.00361803595506e-08, + "loss": 1.1568, + "step": 34530 + }, + { + "epoch": 2.91, + "learning_rate": 4.994540509480672e-08, + "loss": 1.0397, + "step": 34531 + }, + { + "epoch": 2.91, + "learning_rate": 4.9854712040599e-08, + "loss": 1.1818, + "step": 34532 + }, + { + "epoch": 2.91, + "learning_rate": 4.9764101197675716e-08, + "loss": 1.1465, + "step": 34533 + }, + { + "epoch": 2.91, + "learning_rate": 4.967357256678629e-08, + "loss": 1.1296, + "step": 34534 + }, + { + "epoch": 2.91, + "learning_rate": 4.9583126148678994e-08, + "loss": 1.0802, + "step": 34535 + }, + { + "epoch": 2.91, + "learning_rate": 4.94927619440988e-08, + "loss": 1.0181, + "step": 34536 + }, + { + "epoch": 2.91, + "learning_rate": 4.94024799537951e-08, + "loss": 1.1183, + "step": 34537 + }, + { + "epoch": 2.91, + "learning_rate": 4.9312280178512864e-08, + "loss": 1.089, + "step": 34538 + }, + { + "epoch": 2.91, + "learning_rate": 4.922216261899704e-08, + "loss": 1.0971, + "step": 34539 + }, + { + "epoch": 2.91, + "learning_rate": 4.913212727599148e-08, + "loss": 1.025, + "step": 34540 + }, + { + "epoch": 2.91, + "learning_rate": 4.904217415024226e-08, + "loss": 1.1374, + "step": 34541 + }, + { + "epoch": 2.91, + "learning_rate": 4.89523032424899e-08, + "loss": 1.0949, + "step": 34542 + }, + { + "epoch": 2.91, + "learning_rate": 4.886251455347935e-08, + "loss": 1.0849, + "step": 34543 + }, + { + "epoch": 2.91, + "learning_rate": 4.877280808395113e-08, + "loss": 1.1197, + "step": 34544 + }, + { + "epoch": 2.91, + "learning_rate": 4.868318383464687e-08, + "loss": 1.1562, + "step": 34545 + }, + { + "epoch": 2.91, + "learning_rate": 4.85936418063071e-08, + "loss": 1.1294, + "step": 34546 + }, + { + "epoch": 2.91, + "learning_rate": 4.850418199967122e-08, + "loss": 1.1314, + "step": 34547 + }, + { + "epoch": 2.91, + "learning_rate": 4.841480441547974e-08, + "loss": 1.1576, + "step": 34548 + }, + { + "epoch": 2.91, + "learning_rate": 4.8325509054468756e-08, + "loss": 1.0646, + "step": 34549 + }, + { + "epoch": 2.91, + "learning_rate": 4.8236295917377664e-08, + "loss": 1.0323, + "step": 34550 + }, + { + "epoch": 2.91, + "learning_rate": 4.814716500494365e-08, + "loss": 1.0973, + "step": 34551 + }, + { + "epoch": 2.91, + "learning_rate": 4.8058116317902804e-08, + "loss": 1.1576, + "step": 34552 + }, + { + "epoch": 2.91, + "learning_rate": 4.796914985699008e-08, + "loss": 1.1773, + "step": 34553 + }, + { + "epoch": 2.91, + "learning_rate": 4.788026562294157e-08, + "loss": 1.1955, + "step": 34554 + }, + { + "epoch": 2.91, + "learning_rate": 4.779146361649223e-08, + "loss": 1.131, + "step": 34555 + }, + { + "epoch": 2.91, + "learning_rate": 4.77027438383737e-08, + "loss": 1.096, + "step": 34556 + }, + { + "epoch": 2.91, + "learning_rate": 4.761410628932095e-08, + "loss": 1.0701, + "step": 34557 + }, + { + "epoch": 2.91, + "learning_rate": 4.7525550970065615e-08, + "loss": 1.1427, + "step": 34558 + }, + { + "epoch": 2.91, + "learning_rate": 4.7437077881339335e-08, + "loss": 1.0063, + "step": 34559 + }, + { + "epoch": 2.91, + "learning_rate": 4.734868702387263e-08, + "loss": 1.1479, + "step": 34560 + }, + { + "epoch": 2.91, + "learning_rate": 4.7260378398397145e-08, + "loss": 1.1226, + "step": 34561 + }, + { + "epoch": 2.91, + "learning_rate": 4.717215200564118e-08, + "loss": 1.0235, + "step": 34562 + }, + { + "epoch": 2.91, + "learning_rate": 4.708400784633416e-08, + "loss": 1.1449, + "step": 34563 + }, + { + "epoch": 2.91, + "learning_rate": 4.699594592120438e-08, + "loss": 1.123, + "step": 34564 + }, + { + "epoch": 2.91, + "learning_rate": 4.690796623098015e-08, + "loss": 1.1363, + "step": 34565 + }, + { + "epoch": 2.91, + "learning_rate": 4.682006877638756e-08, + "loss": 1.0975, + "step": 34566 + }, + { + "epoch": 2.91, + "learning_rate": 4.6732253558152694e-08, + "loss": 1.0373, + "step": 34567 + }, + { + "epoch": 2.91, + "learning_rate": 4.664452057700053e-08, + "loss": 1.1643, + "step": 34568 + }, + { + "epoch": 2.91, + "learning_rate": 4.6556869833658257e-08, + "loss": 1.0833, + "step": 34569 + }, + { + "epoch": 2.91, + "learning_rate": 4.646930132884753e-08, + "loss": 1.0416, + "step": 34570 + }, + { + "epoch": 2.91, + "learning_rate": 4.6381815063293315e-08, + "loss": 1.126, + "step": 34571 + }, + { + "epoch": 2.91, + "learning_rate": 4.629441103771837e-08, + "loss": 1.0658, + "step": 34572 + }, + { + "epoch": 2.91, + "learning_rate": 4.620708925284434e-08, + "loss": 1.078, + "step": 34573 + }, + { + "epoch": 2.91, + "learning_rate": 4.6119849709393985e-08, + "loss": 1.1333, + "step": 34574 + }, + { + "epoch": 2.91, + "learning_rate": 4.603269240808561e-08, + "loss": 1.1015, + "step": 34575 + }, + { + "epoch": 2.91, + "learning_rate": 4.594561734964198e-08, + "loss": 1.1252, + "step": 34576 + }, + { + "epoch": 2.91, + "learning_rate": 4.585862453478029e-08, + "loss": 1.1558, + "step": 34577 + }, + { + "epoch": 2.91, + "learning_rate": 4.5771713964221086e-08, + "loss": 1.0992, + "step": 34578 + }, + { + "epoch": 2.91, + "learning_rate": 4.5684885638681564e-08, + "loss": 1.1711, + "step": 34579 + }, + { + "epoch": 2.91, + "learning_rate": 4.5598139558878927e-08, + "loss": 1.0556, + "step": 34580 + }, + { + "epoch": 2.91, + "learning_rate": 4.5511475725531496e-08, + "loss": 1.1806, + "step": 34581 + }, + { + "epoch": 2.91, + "learning_rate": 4.5424894139353137e-08, + "loss": 1.1146, + "step": 34582 + }, + { + "epoch": 2.91, + "learning_rate": 4.533839480105995e-08, + "loss": 1.105, + "step": 34583 + }, + { + "epoch": 2.91, + "learning_rate": 4.525197771136691e-08, + "loss": 1.0768, + "step": 34584 + }, + { + "epoch": 2.91, + "learning_rate": 4.516564287098901e-08, + "loss": 1.0556, + "step": 34585 + }, + { + "epoch": 2.91, + "learning_rate": 4.5079390280636794e-08, + "loss": 1.1371, + "step": 34586 + }, + { + "epoch": 2.91, + "learning_rate": 4.4993219941026345e-08, + "loss": 1.1549, + "step": 34587 + }, + { + "epoch": 2.91, + "learning_rate": 4.4907131852867104e-08, + "loss": 1.0609, + "step": 34588 + }, + { + "epoch": 2.91, + "learning_rate": 4.482112601687072e-08, + "loss": 1.0642, + "step": 34589 + }, + { + "epoch": 2.91, + "learning_rate": 4.4735202433748845e-08, + "loss": 1.1825, + "step": 34590 + }, + { + "epoch": 2.91, + "learning_rate": 4.464936110421092e-08, + "loss": 1.0737, + "step": 34591 + }, + { + "epoch": 2.91, + "learning_rate": 4.456360202896637e-08, + "loss": 1.0066, + "step": 34592 + }, + { + "epoch": 2.91, + "learning_rate": 4.447792520872241e-08, + "loss": 1.0957, + "step": 34593 + }, + { + "epoch": 2.91, + "learning_rate": 4.4392330644188466e-08, + "loss": 1.1849, + "step": 34594 + }, + { + "epoch": 2.91, + "learning_rate": 4.430681833607175e-08, + "loss": 1.0783, + "step": 34595 + }, + { + "epoch": 2.91, + "learning_rate": 4.422138828507838e-08, + "loss": 1.1573, + "step": 34596 + }, + { + "epoch": 2.91, + "learning_rate": 4.413604049191334e-08, + "loss": 1.0842, + "step": 34597 + }, + { + "epoch": 2.91, + "learning_rate": 4.4050774957283824e-08, + "loss": 1.09, + "step": 34598 + }, + { + "epoch": 2.91, + "learning_rate": 4.3965591681891516e-08, + "loss": 1.1451, + "step": 34599 + }, + { + "epoch": 2.91, + "learning_rate": 4.388049066644251e-08, + "loss": 1.1725, + "step": 34600 + }, + { + "epoch": 2.91, + "learning_rate": 4.3795471911639575e-08, + "loss": 1.0724, + "step": 34601 + }, + { + "epoch": 2.91, + "learning_rate": 4.371053541818549e-08, + "loss": 1.1153, + "step": 34602 + }, + { + "epoch": 2.91, + "learning_rate": 4.3625681186779676e-08, + "loss": 1.1841, + "step": 34603 + }, + { + "epoch": 2.91, + "learning_rate": 4.354090921812715e-08, + "loss": 1.0666, + "step": 34604 + }, + { + "epoch": 2.91, + "learning_rate": 4.3456219512924e-08, + "loss": 1.1058, + "step": 34605 + }, + { + "epoch": 2.91, + "learning_rate": 4.337161207187302e-08, + "loss": 1.1483, + "step": 34606 + }, + { + "epoch": 2.91, + "learning_rate": 4.3287086895672514e-08, + "loss": 1.0798, + "step": 34607 + }, + { + "epoch": 2.91, + "learning_rate": 4.320264398501972e-08, + "loss": 1.2075, + "step": 34608 + }, + { + "epoch": 2.91, + "learning_rate": 4.3118283340614074e-08, + "loss": 1.0281, + "step": 34609 + }, + { + "epoch": 2.91, + "learning_rate": 4.303400496315169e-08, + "loss": 1.15, + "step": 34610 + }, + { + "epoch": 2.91, + "learning_rate": 4.294980885332867e-08, + "loss": 1.0956, + "step": 34611 + }, + { + "epoch": 2.91, + "learning_rate": 4.2865695011841124e-08, + "loss": 1.1874, + "step": 34612 + }, + { + "epoch": 2.91, + "learning_rate": 4.2781663439384055e-08, + "loss": 1.0788, + "step": 34613 + }, + { + "epoch": 2.91, + "learning_rate": 4.269771413665136e-08, + "loss": 1.1802, + "step": 34614 + }, + { + "epoch": 2.91, + "learning_rate": 4.261384710433691e-08, + "loss": 1.0393, + "step": 34615 + }, + { + "epoch": 2.91, + "learning_rate": 4.253006234313461e-08, + "loss": 1.0281, + "step": 34616 + }, + { + "epoch": 2.91, + "learning_rate": 4.2446359853735016e-08, + "loss": 1.1033, + "step": 34617 + }, + { + "epoch": 2.91, + "learning_rate": 4.2362739636829797e-08, + "loss": 1.0708, + "step": 34618 + }, + { + "epoch": 2.91, + "learning_rate": 4.2279201693110615e-08, + "loss": 1.1448, + "step": 34619 + }, + { + "epoch": 2.91, + "learning_rate": 4.2195746023268036e-08, + "loss": 1.1147, + "step": 34620 + }, + { + "epoch": 2.91, + "learning_rate": 4.2112372627991504e-08, + "loss": 1.147, + "step": 34621 + }, + { + "epoch": 2.91, + "learning_rate": 4.202908150796825e-08, + "loss": 1.071, + "step": 34622 + }, + { + "epoch": 2.91, + "learning_rate": 4.194587266388883e-08, + "loss": 1.0319, + "step": 34623 + }, + { + "epoch": 2.92, + "learning_rate": 4.186274609643937e-08, + "loss": 1.0764, + "step": 34624 + }, + { + "epoch": 2.92, + "learning_rate": 4.177970180630597e-08, + "loss": 1.0792, + "step": 34625 + }, + { + "epoch": 2.92, + "learning_rate": 4.1696739794176985e-08, + "loss": 1.1365, + "step": 34626 + }, + { + "epoch": 2.92, + "learning_rate": 4.161386006073631e-08, + "loss": 1.0574, + "step": 34627 + }, + { + "epoch": 2.92, + "learning_rate": 4.1531062606667837e-08, + "loss": 1.0943, + "step": 34628 + }, + { + "epoch": 2.92, + "learning_rate": 4.1448347432658794e-08, + "loss": 1.0255, + "step": 34629 + }, + { + "epoch": 2.92, + "learning_rate": 4.1365714539388645e-08, + "loss": 1.0786, + "step": 34630 + }, + { + "epoch": 2.92, + "learning_rate": 4.1283163927543504e-08, + "loss": 1.1285, + "step": 34631 + }, + { + "epoch": 2.92, + "learning_rate": 4.120069559780393e-08, + "loss": 1.1723, + "step": 34632 + }, + { + "epoch": 2.92, + "learning_rate": 4.11183095508505e-08, + "loss": 1.1083, + "step": 34633 + }, + { + "epoch": 2.92, + "learning_rate": 4.1036005787366e-08, + "loss": 1.1386, + "step": 34634 + }, + { + "epoch": 2.92, + "learning_rate": 4.095378430802877e-08, + "loss": 1.1599, + "step": 34635 + }, + { + "epoch": 2.92, + "learning_rate": 4.0871645113518263e-08, + "loss": 1.0709, + "step": 34636 + }, + { + "epoch": 2.92, + "learning_rate": 4.0789588204515064e-08, + "loss": 1.1739, + "step": 34637 + }, + { + "epoch": 2.92, + "learning_rate": 4.0707613581694174e-08, + "loss": 1.1625, + "step": 34638 + }, + { + "epoch": 2.92, + "learning_rate": 4.062572124573505e-08, + "loss": 1.122, + "step": 34639 + }, + { + "epoch": 2.92, + "learning_rate": 4.0543911197312716e-08, + "loss": 1.1043, + "step": 34640 + }, + { + "epoch": 2.92, + "learning_rate": 4.0462183437104395e-08, + "loss": 1.1564, + "step": 34641 + }, + { + "epoch": 2.92, + "learning_rate": 4.038053796578512e-08, + "loss": 1.1543, + "step": 34642 + }, + { + "epoch": 2.92, + "learning_rate": 4.029897478402989e-08, + "loss": 1.1842, + "step": 34643 + }, + { + "epoch": 2.92, + "learning_rate": 4.0217493892511506e-08, + "loss": 1.0543, + "step": 34644 + }, + { + "epoch": 2.92, + "learning_rate": 4.0136095291903876e-08, + "loss": 1.1051, + "step": 34645 + }, + { + "epoch": 2.92, + "learning_rate": 4.005477898287979e-08, + "loss": 1.1079, + "step": 34646 + }, + { + "epoch": 2.92, + "learning_rate": 3.9973544966109836e-08, + "loss": 1.131, + "step": 34647 + }, + { + "epoch": 2.92, + "learning_rate": 3.989239324226679e-08, + "loss": 1.0753, + "step": 34648 + }, + { + "epoch": 2.92, + "learning_rate": 3.981132381202013e-08, + "loss": 1.1244, + "step": 34649 + }, + { + "epoch": 2.92, + "learning_rate": 3.973033667604043e-08, + "loss": 1.181, + "step": 34650 + }, + { + "epoch": 2.92, + "learning_rate": 3.964943183499492e-08, + "loss": 1.0213, + "step": 34651 + }, + { + "epoch": 2.92, + "learning_rate": 3.95686092895553e-08, + "loss": 1.1019, + "step": 34652 + }, + { + "epoch": 2.92, + "learning_rate": 3.9487869040386594e-08, + "loss": 1.0829, + "step": 34653 + }, + { + "epoch": 2.92, + "learning_rate": 3.940721108815715e-08, + "loss": 1.1261, + "step": 34654 + }, + { + "epoch": 2.92, + "learning_rate": 3.93266354335331e-08, + "loss": 1.1225, + "step": 34655 + }, + { + "epoch": 2.92, + "learning_rate": 3.924614207718058e-08, + "loss": 1.1772, + "step": 34656 + }, + { + "epoch": 2.92, + "learning_rate": 3.916573101976351e-08, + "loss": 1.0909, + "step": 34657 + }, + { + "epoch": 2.92, + "learning_rate": 3.9085402261948014e-08, + "loss": 1.1023, + "step": 34658 + }, + { + "epoch": 2.92, + "learning_rate": 3.900515580439579e-08, + "loss": 0.9997, + "step": 34659 + }, + { + "epoch": 2.92, + "learning_rate": 3.892499164777075e-08, + "loss": 1.0021, + "step": 34660 + }, + { + "epoch": 2.92, + "learning_rate": 3.884490979273681e-08, + "loss": 1.0652, + "step": 34661 + }, + { + "epoch": 2.92, + "learning_rate": 3.876491023995344e-08, + "loss": 0.9884, + "step": 34662 + }, + { + "epoch": 2.92, + "learning_rate": 3.868499299008233e-08, + "loss": 1.1136, + "step": 34663 + }, + { + "epoch": 2.92, + "learning_rate": 3.860515804378295e-08, + "loss": 1.2169, + "step": 34664 + }, + { + "epoch": 2.92, + "learning_rate": 3.852540540171701e-08, + "loss": 1.1599, + "step": 34665 + }, + { + "epoch": 2.92, + "learning_rate": 3.8445735064541744e-08, + "loss": 1.1241, + "step": 34666 + }, + { + "epoch": 2.92, + "learning_rate": 3.836614703291663e-08, + "loss": 1.1674, + "step": 34667 + }, + { + "epoch": 2.92, + "learning_rate": 3.828664130749782e-08, + "loss": 1.0892, + "step": 34668 + }, + { + "epoch": 2.92, + "learning_rate": 3.820721788894366e-08, + "loss": 1.0237, + "step": 34669 + }, + { + "epoch": 2.92, + "learning_rate": 3.812787677790919e-08, + "loss": 1.0651, + "step": 34670 + }, + { + "epoch": 2.92, + "learning_rate": 3.804861797504944e-08, + "loss": 1.1266, + "step": 34671 + }, + { + "epoch": 2.92, + "learning_rate": 3.7969441481021665e-08, + "loss": 1.1622, + "step": 34672 + }, + { + "epoch": 2.92, + "learning_rate": 3.789034729647756e-08, + "loss": 1.1141, + "step": 34673 + }, + { + "epoch": 2.92, + "learning_rate": 3.781133542207327e-08, + "loss": 1.1362, + "step": 34674 + }, + { + "epoch": 2.92, + "learning_rate": 3.7732405858458274e-08, + "loss": 1.0523, + "step": 34675 + }, + { + "epoch": 2.92, + "learning_rate": 3.765355860628761e-08, + "loss": 1.0971, + "step": 34676 + }, + { + "epoch": 2.92, + "learning_rate": 3.757479366621075e-08, + "loss": 1.1446, + "step": 34677 + }, + { + "epoch": 2.92, + "learning_rate": 3.749611103888051e-08, + "loss": 1.0785, + "step": 34678 + }, + { + "epoch": 2.92, + "learning_rate": 3.741751072494415e-08, + "loss": 1.1824, + "step": 34679 + }, + { + "epoch": 2.92, + "learning_rate": 3.733899272505448e-08, + "loss": 1.1438, + "step": 34680 + }, + { + "epoch": 2.92, + "learning_rate": 3.7260557039857645e-08, + "loss": 0.9948, + "step": 34681 + }, + { + "epoch": 2.92, + "learning_rate": 3.7182203670002026e-08, + "loss": 1.0778, + "step": 34682 + }, + { + "epoch": 2.92, + "learning_rate": 3.7103932616135985e-08, + "loss": 1.0697, + "step": 34683 + }, + { + "epoch": 2.92, + "learning_rate": 3.7025743878905676e-08, + "loss": 1.1348, + "step": 34684 + }, + { + "epoch": 2.92, + "learning_rate": 3.694763745895835e-08, + "loss": 1.1135, + "step": 34685 + }, + { + "epoch": 2.92, + "learning_rate": 3.686961335693684e-08, + "loss": 1.2029, + "step": 34686 + }, + { + "epoch": 2.92, + "learning_rate": 3.6791671573488396e-08, + "loss": 1.0845, + "step": 34687 + }, + { + "epoch": 2.92, + "learning_rate": 3.671381210925473e-08, + "loss": 1.1469, + "step": 34688 + }, + { + "epoch": 2.92, + "learning_rate": 3.663603496488088e-08, + "loss": 1.1117, + "step": 34689 + }, + { + "epoch": 2.92, + "learning_rate": 3.6558340141008564e-08, + "loss": 1.0949, + "step": 34690 + }, + { + "epoch": 2.92, + "learning_rate": 3.648072763827948e-08, + "loss": 1.1736, + "step": 34691 + }, + { + "epoch": 2.92, + "learning_rate": 3.6403197457336446e-08, + "loss": 1.003, + "step": 34692 + }, + { + "epoch": 2.92, + "learning_rate": 3.632574959881785e-08, + "loss": 1.1093, + "step": 34693 + }, + { + "epoch": 2.92, + "learning_rate": 3.6248384063365395e-08, + "loss": 1.135, + "step": 34694 + }, + { + "epoch": 2.92, + "learning_rate": 3.617110085161746e-08, + "loss": 1.1266, + "step": 34695 + }, + { + "epoch": 2.92, + "learning_rate": 3.609389996421242e-08, + "loss": 1.0647, + "step": 34696 + }, + { + "epoch": 2.92, + "learning_rate": 3.601678140178866e-08, + "loss": 1.2271, + "step": 34697 + }, + { + "epoch": 2.92, + "learning_rate": 3.593974516498344e-08, + "loss": 1.1284, + "step": 34698 + }, + { + "epoch": 2.92, + "learning_rate": 3.5862791254432926e-08, + "loss": 1.0757, + "step": 34699 + }, + { + "epoch": 2.92, + "learning_rate": 3.5785919670772165e-08, + "loss": 1.1238, + "step": 34700 + }, + { + "epoch": 2.92, + "learning_rate": 3.5709130414637306e-08, + "loss": 1.1903, + "step": 34701 + }, + { + "epoch": 2.92, + "learning_rate": 3.5632423486663405e-08, + "loss": 1.1775, + "step": 34702 + }, + { + "epoch": 2.92, + "learning_rate": 3.555579888748217e-08, + "loss": 1.1218, + "step": 34703 + }, + { + "epoch": 2.92, + "learning_rate": 3.547925661772977e-08, + "loss": 1.19, + "step": 34704 + }, + { + "epoch": 2.92, + "learning_rate": 3.54027966780357e-08, + "loss": 1.1137, + "step": 34705 + }, + { + "epoch": 2.92, + "learning_rate": 3.532641906903278e-08, + "loss": 1.1672, + "step": 34706 + }, + { + "epoch": 2.92, + "learning_rate": 3.525012379135162e-08, + "loss": 1.1183, + "step": 34707 + }, + { + "epoch": 2.92, + "learning_rate": 3.517391084562394e-08, + "loss": 1.0952, + "step": 34708 + }, + { + "epoch": 2.92, + "learning_rate": 3.5097780232478116e-08, + "loss": 1.0782, + "step": 34709 + }, + { + "epoch": 2.92, + "learning_rate": 3.502173195254366e-08, + "loss": 1.1685, + "step": 34710 + }, + { + "epoch": 2.92, + "learning_rate": 3.4945766006448944e-08, + "loss": 1.067, + "step": 34711 + }, + { + "epoch": 2.92, + "learning_rate": 3.486988239482125e-08, + "loss": 1.1964, + "step": 34712 + }, + { + "epoch": 2.92, + "learning_rate": 3.479408111828897e-08, + "loss": 1.2268, + "step": 34713 + }, + { + "epoch": 2.92, + "learning_rate": 3.471836217747604e-08, + "loss": 1.1223, + "step": 34714 + }, + { + "epoch": 2.92, + "learning_rate": 3.464272557300974e-08, + "loss": 1.1166, + "step": 34715 + }, + { + "epoch": 2.92, + "learning_rate": 3.456717130551401e-08, + "loss": 1.1186, + "step": 34716 + }, + { + "epoch": 2.92, + "learning_rate": 3.449169937561392e-08, + "loss": 1.1379, + "step": 34717 + }, + { + "epoch": 2.92, + "learning_rate": 3.44163097839334e-08, + "loss": 1.1167, + "step": 34718 + }, + { + "epoch": 2.92, + "learning_rate": 3.434100253109418e-08, + "loss": 1.1753, + "step": 34719 + }, + { + "epoch": 2.92, + "learning_rate": 3.42657776177191e-08, + "loss": 1.0758, + "step": 34720 + }, + { + "epoch": 2.92, + "learning_rate": 3.4190635044429873e-08, + "loss": 1.1236, + "step": 34721 + }, + { + "epoch": 2.92, + "learning_rate": 3.4115574811846017e-08, + "loss": 1.1577, + "step": 34722 + }, + { + "epoch": 2.92, + "learning_rate": 3.4040596920590364e-08, + "loss": 1.0913, + "step": 34723 + }, + { + "epoch": 2.92, + "learning_rate": 3.396570137128019e-08, + "loss": 1.0515, + "step": 34724 + }, + { + "epoch": 2.92, + "learning_rate": 3.38908881645339e-08, + "loss": 1.0525, + "step": 34725 + }, + { + "epoch": 2.92, + "learning_rate": 3.38161573009721e-08, + "loss": 1.114, + "step": 34726 + }, + { + "epoch": 2.92, + "learning_rate": 3.374150878120985e-08, + "loss": 1.0886, + "step": 34727 + }, + { + "epoch": 2.92, + "learning_rate": 3.366694260586556e-08, + "loss": 1.1484, + "step": 34728 + }, + { + "epoch": 2.92, + "learning_rate": 3.359245877555428e-08, + "loss": 1.079, + "step": 34729 + }, + { + "epoch": 2.92, + "learning_rate": 3.3518057290892195e-08, + "loss": 1.1359, + "step": 34730 + }, + { + "epoch": 2.92, + "learning_rate": 3.344373815249324e-08, + "loss": 1.0902, + "step": 34731 + }, + { + "epoch": 2.92, + "learning_rate": 3.336950136097139e-08, + "loss": 1.1375, + "step": 34732 + }, + { + "epoch": 2.92, + "learning_rate": 3.329534691694059e-08, + "loss": 1.0901, + "step": 34733 + }, + { + "epoch": 2.92, + "learning_rate": 3.322127482101367e-08, + "loss": 1.1159, + "step": 34734 + }, + { + "epoch": 2.92, + "learning_rate": 3.314728507380349e-08, + "loss": 1.0915, + "step": 34735 + }, + { + "epoch": 2.92, + "learning_rate": 3.307337767591845e-08, + "loss": 1.1219, + "step": 34736 + }, + { + "epoch": 2.92, + "learning_rate": 3.299955262797139e-08, + "loss": 1.0703, + "step": 34737 + }, + { + "epoch": 2.92, + "learning_rate": 3.292580993057293e-08, + "loss": 1.0757, + "step": 34738 + }, + { + "epoch": 2.92, + "learning_rate": 3.2852149584330364e-08, + "loss": 1.1394, + "step": 34739 + }, + { + "epoch": 2.92, + "learning_rate": 3.27785715898532e-08, + "loss": 1.1722, + "step": 34740 + }, + { + "epoch": 2.92, + "learning_rate": 3.270507594774985e-08, + "loss": 1.1227, + "step": 34741 + }, + { + "epoch": 2.92, + "learning_rate": 3.26316626586265e-08, + "loss": 1.1849, + "step": 34742 + }, + { + "epoch": 2.93, + "learning_rate": 3.255833172309042e-08, + "loss": 1.0794, + "step": 34743 + }, + { + "epoch": 2.93, + "learning_rate": 3.2485083141747807e-08, + "loss": 1.081, + "step": 34744 + }, + { + "epoch": 2.93, + "learning_rate": 3.241191691520262e-08, + "loss": 1.1689, + "step": 34745 + }, + { + "epoch": 2.93, + "learning_rate": 3.2338833044062157e-08, + "loss": 1.0581, + "step": 34746 + }, + { + "epoch": 2.93, + "learning_rate": 3.226583152892593e-08, + "loss": 1.1685, + "step": 34747 + }, + { + "epoch": 2.93, + "learning_rate": 3.219291237040123e-08, + "loss": 1.075, + "step": 34748 + }, + { + "epoch": 2.93, + "learning_rate": 3.212007556908758e-08, + "loss": 1.1421, + "step": 34749 + }, + { + "epoch": 2.93, + "learning_rate": 3.204732112558895e-08, + "loss": 1.0499, + "step": 34750 + }, + { + "epoch": 2.93, + "learning_rate": 3.197464904050485e-08, + "loss": 1.1434, + "step": 34751 + }, + { + "epoch": 2.93, + "learning_rate": 3.190205931443702e-08, + "loss": 1.1227, + "step": 34752 + }, + { + "epoch": 2.93, + "learning_rate": 3.182955194798387e-08, + "loss": 1.1398, + "step": 34753 + }, + { + "epoch": 2.93, + "learning_rate": 3.175712694174604e-08, + "loss": 1.1391, + "step": 34754 + }, + { + "epoch": 2.93, + "learning_rate": 3.1684784296320824e-08, + "loss": 1.0961, + "step": 34755 + }, + { + "epoch": 2.93, + "learning_rate": 3.161252401230552e-08, + "loss": 1.0886, + "step": 34756 + }, + { + "epoch": 2.93, + "learning_rate": 3.1540346090298544e-08, + "loss": 1.06, + "step": 34757 + }, + { + "epoch": 2.93, + "learning_rate": 3.1468250530894974e-08, + "loss": 1.0681, + "step": 34758 + }, + { + "epoch": 2.93, + "learning_rate": 3.139623733469099e-08, + "loss": 1.135, + "step": 34759 + }, + { + "epoch": 2.93, + "learning_rate": 3.1324306502281685e-08, + "loss": 1.0721, + "step": 34760 + }, + { + "epoch": 2.93, + "learning_rate": 3.125245803426213e-08, + "loss": 1.1293, + "step": 34761 + }, + { + "epoch": 2.93, + "learning_rate": 3.118069193122408e-08, + "loss": 1.0728, + "step": 34762 + }, + { + "epoch": 2.93, + "learning_rate": 3.110900819376261e-08, + "loss": 1.0642, + "step": 34763 + }, + { + "epoch": 2.93, + "learning_rate": 3.103740682246836e-08, + "loss": 1.0838, + "step": 34764 + }, + { + "epoch": 2.93, + "learning_rate": 3.096588781793419e-08, + "loss": 1.0721, + "step": 34765 + }, + { + "epoch": 2.93, + "learning_rate": 3.089445118074963e-08, + "loss": 1.1361, + "step": 34766 + }, + { + "epoch": 2.93, + "learning_rate": 3.0823096911506423e-08, + "loss": 1.1102, + "step": 34767 + }, + { + "epoch": 2.93, + "learning_rate": 3.075182501079188e-08, + "loss": 1.0948, + "step": 34768 + }, + { + "epoch": 2.93, + "learning_rate": 3.0680635479197754e-08, + "loss": 1.1105, + "step": 34769 + }, + { + "epoch": 2.93, + "learning_rate": 3.060952831731023e-08, + "loss": 1.2012, + "step": 34770 + }, + { + "epoch": 2.93, + "learning_rate": 3.0538503525717745e-08, + "loss": 1.2244, + "step": 34771 + }, + { + "epoch": 2.93, + "learning_rate": 3.0467561105006483e-08, + "loss": 1.1081, + "step": 34772 + }, + { + "epoch": 2.93, + "learning_rate": 3.0396701055762645e-08, + "loss": 1.0428, + "step": 34773 + }, + { + "epoch": 2.93, + "learning_rate": 3.032592337857243e-08, + "loss": 1.167, + "step": 34774 + }, + { + "epoch": 2.93, + "learning_rate": 3.0255228074019816e-08, + "loss": 1.1812, + "step": 34775 + }, + { + "epoch": 2.93, + "learning_rate": 3.0184615142688776e-08, + "loss": 1.1232, + "step": 34776 + }, + { + "epoch": 2.93, + "learning_rate": 3.011408458516441e-08, + "loss": 1.0869, + "step": 34777 + }, + { + "epoch": 2.93, + "learning_rate": 3.0043636402027344e-08, + "loss": 1.1826, + "step": 34778 + }, + { + "epoch": 2.93, + "learning_rate": 2.997327059386046e-08, + "loss": 1.1146, + "step": 34779 + }, + { + "epoch": 2.93, + "learning_rate": 2.990298716124551e-08, + "loss": 1.196, + "step": 34780 + }, + { + "epoch": 2.93, + "learning_rate": 2.9832786104763144e-08, + "loss": 1.0459, + "step": 34781 + }, + { + "epoch": 2.93, + "learning_rate": 2.9762667424992898e-08, + "loss": 1.0981, + "step": 34782 + }, + { + "epoch": 2.93, + "learning_rate": 2.9692631122514304e-08, + "loss": 1.0475, + "step": 34783 + }, + { + "epoch": 2.93, + "learning_rate": 2.9622677197906902e-08, + "loss": 1.179, + "step": 34784 + }, + { + "epoch": 2.93, + "learning_rate": 2.9552805651746898e-08, + "loss": 1.1472, + "step": 34785 + }, + { + "epoch": 2.93, + "learning_rate": 2.9483016484612714e-08, + "loss": 1.1061, + "step": 34786 + }, + { + "epoch": 2.93, + "learning_rate": 2.941330969708167e-08, + "loss": 1.0524, + "step": 34787 + }, + { + "epoch": 2.93, + "learning_rate": 2.9343685289727752e-08, + "loss": 1.1957, + "step": 34788 + }, + { + "epoch": 2.93, + "learning_rate": 2.9274143263128274e-08, + "loss": 1.0539, + "step": 34789 + }, + { + "epoch": 2.93, + "learning_rate": 2.920468361785611e-08, + "loss": 1.1013, + "step": 34790 + }, + { + "epoch": 2.93, + "learning_rate": 2.913530635448636e-08, + "loss": 1.1777, + "step": 34791 + }, + { + "epoch": 2.93, + "learning_rate": 2.9066011473590782e-08, + "loss": 1.1495, + "step": 34792 + }, + { + "epoch": 2.93, + "learning_rate": 2.8996798975744477e-08, + "loss": 1.08, + "step": 34793 + }, + { + "epoch": 2.93, + "learning_rate": 2.892766886151699e-08, + "loss": 1.1355, + "step": 34794 + }, + { + "epoch": 2.93, + "learning_rate": 2.885862113148008e-08, + "loss": 1.1141, + "step": 34795 + }, + { + "epoch": 2.93, + "learning_rate": 2.8789655786204405e-08, + "loss": 1.1472, + "step": 34796 + }, + { + "epoch": 2.93, + "learning_rate": 2.8720772826259514e-08, + "loss": 1.1271, + "step": 34797 + }, + { + "epoch": 2.93, + "learning_rate": 2.8651972252216053e-08, + "loss": 1.1994, + "step": 34798 + }, + { + "epoch": 2.93, + "learning_rate": 2.858325406464024e-08, + "loss": 1.087, + "step": 34799 + }, + { + "epoch": 2.93, + "learning_rate": 2.8514618264101622e-08, + "loss": 1.0757, + "step": 34800 + }, + { + "epoch": 2.93, + "learning_rate": 2.84460648511653e-08, + "loss": 1.137, + "step": 34801 + }, + { + "epoch": 2.93, + "learning_rate": 2.8377593826400817e-08, + "loss": 1.0855, + "step": 34802 + }, + { + "epoch": 2.93, + "learning_rate": 2.8309205190369948e-08, + "loss": 1.0679, + "step": 34803 + }, + { + "epoch": 2.93, + "learning_rate": 2.8240898943641126e-08, + "loss": 1.0007, + "step": 34804 + }, + { + "epoch": 2.93, + "learning_rate": 2.8172675086776126e-08, + "loss": 1.1131, + "step": 34805 + }, + { + "epoch": 2.93, + "learning_rate": 2.8104533620341158e-08, + "loss": 1.0793, + "step": 34806 + }, + { + "epoch": 2.93, + "learning_rate": 2.8036474544896887e-08, + "loss": 1.1311, + "step": 34807 + }, + { + "epoch": 2.93, + "learning_rate": 2.7968497861007306e-08, + "loss": 1.1205, + "step": 34808 + }, + { + "epoch": 2.93, + "learning_rate": 2.7900603569233076e-08, + "loss": 1.0652, + "step": 34809 + }, + { + "epoch": 2.93, + "learning_rate": 2.7832791670135974e-08, + "loss": 1.1349, + "step": 34810 + }, + { + "epoch": 2.93, + "learning_rate": 2.7765062164275546e-08, + "loss": 1.1689, + "step": 34811 + }, + { + "epoch": 2.93, + "learning_rate": 2.7697415052211352e-08, + "loss": 1.2334, + "step": 34812 + }, + { + "epoch": 2.93, + "learning_rate": 2.762985033450405e-08, + "loss": 1.1098, + "step": 34813 + }, + { + "epoch": 2.93, + "learning_rate": 2.7562368011708752e-08, + "loss": 1.0841, + "step": 34814 + }, + { + "epoch": 2.93, + "learning_rate": 2.749496808438612e-08, + "loss": 1.1908, + "step": 34815 + }, + { + "epoch": 2.93, + "learning_rate": 2.7427650553091266e-08, + "loss": 1.1085, + "step": 34816 + }, + { + "epoch": 2.93, + "learning_rate": 2.7360415418380415e-08, + "loss": 1.0851, + "step": 34817 + }, + { + "epoch": 2.93, + "learning_rate": 2.7293262680809785e-08, + "loss": 1.1223, + "step": 34818 + }, + { + "epoch": 2.93, + "learning_rate": 2.722619234093338e-08, + "loss": 1.1455, + "step": 34819 + }, + { + "epoch": 2.93, + "learning_rate": 2.7159204399306304e-08, + "loss": 1.0629, + "step": 34820 + }, + { + "epoch": 2.93, + "learning_rate": 2.709229885648146e-08, + "loss": 1.0862, + "step": 34821 + }, + { + "epoch": 2.93, + "learning_rate": 2.7025475713012838e-08, + "loss": 1.0983, + "step": 34822 + }, + { + "epoch": 2.93, + "learning_rate": 2.6958734969450008e-08, + "loss": 1.0015, + "step": 34823 + }, + { + "epoch": 2.93, + "learning_rate": 2.6892076626346965e-08, + "loss": 1.1417, + "step": 34824 + }, + { + "epoch": 2.93, + "learning_rate": 2.6825500684252158e-08, + "loss": 1.0305, + "step": 34825 + }, + { + "epoch": 2.93, + "learning_rate": 2.6759007143718486e-08, + "loss": 1.098, + "step": 34826 + }, + { + "epoch": 2.93, + "learning_rate": 2.669259600529217e-08, + "loss": 1.1007, + "step": 34827 + }, + { + "epoch": 2.93, + "learning_rate": 2.6626267269524996e-08, + "loss": 1.147, + "step": 34828 + }, + { + "epoch": 2.93, + "learning_rate": 2.6560020936962082e-08, + "loss": 1.1, + "step": 34829 + }, + { + "epoch": 2.93, + "learning_rate": 2.649385700815299e-08, + "loss": 1.028, + "step": 34830 + }, + { + "epoch": 2.93, + "learning_rate": 2.6427775483642837e-08, + "loss": 1.0474, + "step": 34831 + }, + { + "epoch": 2.93, + "learning_rate": 2.6361776363978964e-08, + "loss": 1.1132, + "step": 34832 + }, + { + "epoch": 2.93, + "learning_rate": 2.629585964970649e-08, + "loss": 1.135, + "step": 34833 + }, + { + "epoch": 2.93, + "learning_rate": 2.6230025341368315e-08, + "loss": 1.2358, + "step": 34834 + }, + { + "epoch": 2.93, + "learning_rate": 2.6164273439511777e-08, + "loss": 1.1166, + "step": 34835 + }, + { + "epoch": 2.93, + "learning_rate": 2.6098603944676447e-08, + "loss": 1.0969, + "step": 34836 + }, + { + "epoch": 2.93, + "learning_rate": 2.6033016857406335e-08, + "loss": 1.1603, + "step": 34837 + }, + { + "epoch": 2.93, + "learning_rate": 2.5967512178244337e-08, + "loss": 1.1186, + "step": 34838 + }, + { + "epoch": 2.93, + "learning_rate": 2.5902089907730022e-08, + "loss": 1.119, + "step": 34839 + }, + { + "epoch": 2.93, + "learning_rate": 2.5836750046405178e-08, + "loss": 1.0234, + "step": 34840 + }, + { + "epoch": 2.93, + "learning_rate": 2.5771492594809377e-08, + "loss": 1.082, + "step": 34841 + }, + { + "epoch": 2.93, + "learning_rate": 2.5706317553482185e-08, + "loss": 1.1076, + "step": 34842 + }, + { + "epoch": 2.93, + "learning_rate": 2.564122492296095e-08, + "loss": 1.1119, + "step": 34843 + }, + { + "epoch": 2.93, + "learning_rate": 2.557621470378413e-08, + "loss": 1.1723, + "step": 34844 + }, + { + "epoch": 2.93, + "learning_rate": 2.5511286896489075e-08, + "loss": 1.0768, + "step": 34845 + }, + { + "epoch": 2.93, + "learning_rate": 2.5446441501612018e-08, + "loss": 1.1638, + "step": 34846 + }, + { + "epoch": 2.93, + "learning_rate": 2.538167851968809e-08, + "loss": 1.1135, + "step": 34847 + }, + { + "epoch": 2.93, + "learning_rate": 2.531699795125353e-08, + "loss": 1.0744, + "step": 34848 + }, + { + "epoch": 2.93, + "learning_rate": 2.5252399796842352e-08, + "loss": 1.1492, + "step": 34849 + }, + { + "epoch": 2.93, + "learning_rate": 2.5187884056988575e-08, + "loss": 1.0793, + "step": 34850 + }, + { + "epoch": 2.93, + "learning_rate": 2.5123450732223998e-08, + "loss": 1.1317, + "step": 34851 + }, + { + "epoch": 2.93, + "learning_rate": 2.5059099823081524e-08, + "loss": 1.1088, + "step": 34852 + }, + { + "epoch": 2.93, + "learning_rate": 2.499483133009406e-08, + "loss": 1.1007, + "step": 34853 + }, + { + "epoch": 2.93, + "learning_rate": 2.4930645253792296e-08, + "loss": 1.1172, + "step": 34854 + }, + { + "epoch": 2.93, + "learning_rate": 2.4866541594704697e-08, + "loss": 1.1061, + "step": 34855 + }, + { + "epoch": 2.93, + "learning_rate": 2.480252035336195e-08, + "loss": 1.0826, + "step": 34856 + }, + { + "epoch": 2.93, + "learning_rate": 2.4738581530294737e-08, + "loss": 1.2079, + "step": 34857 + }, + { + "epoch": 2.93, + "learning_rate": 2.46747251260282e-08, + "loss": 1.1001, + "step": 34858 + }, + { + "epoch": 2.93, + "learning_rate": 2.461095114109302e-08, + "loss": 1.0611, + "step": 34859 + }, + { + "epoch": 2.93, + "learning_rate": 2.4547259576014338e-08, + "loss": 1.1266, + "step": 34860 + }, + { + "epoch": 2.94, + "learning_rate": 2.4483650431318395e-08, + "loss": 1.1211, + "step": 34861 + }, + { + "epoch": 2.94, + "learning_rate": 2.442012370753033e-08, + "loss": 1.1096, + "step": 34862 + }, + { + "epoch": 2.94, + "learning_rate": 2.4356679405177497e-08, + "loss": 1.1157, + "step": 34863 + }, + { + "epoch": 2.94, + "learning_rate": 2.4293317524780592e-08, + "loss": 1.0809, + "step": 34864 + }, + { + "epoch": 2.94, + "learning_rate": 2.4230038066866966e-08, + "loss": 1.1513, + "step": 34865 + }, + { + "epoch": 2.94, + "learning_rate": 2.4166841031956213e-08, + "loss": 1.0781, + "step": 34866 + }, + { + "epoch": 2.94, + "learning_rate": 2.4103726420571238e-08, + "loss": 1.0531, + "step": 34867 + }, + { + "epoch": 2.94, + "learning_rate": 2.4040694233234964e-08, + "loss": 1.1315, + "step": 34868 + }, + { + "epoch": 2.94, + "learning_rate": 2.397774447046697e-08, + "loss": 1.0893, + "step": 34869 + }, + { + "epoch": 2.94, + "learning_rate": 2.391487713278684e-08, + "loss": 1.1584, + "step": 34870 + }, + { + "epoch": 2.94, + "learning_rate": 2.3852092220715273e-08, + "loss": 1.0273, + "step": 34871 + }, + { + "epoch": 2.94, + "learning_rate": 2.378938973477074e-08, + "loss": 1.0936, + "step": 34872 + }, + { + "epoch": 2.94, + "learning_rate": 2.3726769675470606e-08, + "loss": 1.1908, + "step": 34873 + }, + { + "epoch": 2.94, + "learning_rate": 2.3664232043333346e-08, + "loss": 1.0399, + "step": 34874 + }, + { + "epoch": 2.94, + "learning_rate": 2.3601776838875213e-08, + "loss": 1.126, + "step": 34875 + }, + { + "epoch": 2.94, + "learning_rate": 2.3539404062611348e-08, + "loss": 1.0952, + "step": 34876 + }, + { + "epoch": 2.94, + "learning_rate": 2.347711371505801e-08, + "loss": 1.1351, + "step": 34877 + }, + { + "epoch": 2.94, + "learning_rate": 2.3414905796729227e-08, + "loss": 1.1066, + "step": 34878 + }, + { + "epoch": 2.94, + "learning_rate": 2.3352780308140144e-08, + "loss": 1.0513, + "step": 34879 + }, + { + "epoch": 2.94, + "learning_rate": 2.3290737249802575e-08, + "loss": 1.1909, + "step": 34880 + }, + { + "epoch": 2.94, + "learning_rate": 2.3228776622230552e-08, + "loss": 1.105, + "step": 34881 + }, + { + "epoch": 2.94, + "learning_rate": 2.316689842593478e-08, + "loss": 1.1266, + "step": 34882 + }, + { + "epoch": 2.94, + "learning_rate": 2.310510266142818e-08, + "loss": 1.0388, + "step": 34883 + }, + { + "epoch": 2.94, + "learning_rate": 2.3043389329219235e-08, + "loss": 1.1148, + "step": 34884 + }, + { + "epoch": 2.94, + "learning_rate": 2.2981758429818646e-08, + "loss": 1.0889, + "step": 34885 + }, + { + "epoch": 2.94, + "learning_rate": 2.2920209963736005e-08, + "loss": 1.1108, + "step": 34886 + }, + { + "epoch": 2.94, + "learning_rate": 2.285874393148091e-08, + "loss": 1.0494, + "step": 34887 + }, + { + "epoch": 2.94, + "learning_rate": 2.2797360333558505e-08, + "loss": 1.0897, + "step": 34888 + }, + { + "epoch": 2.94, + "learning_rate": 2.2736059170478385e-08, + "loss": 1.0633, + "step": 34889 + }, + { + "epoch": 2.94, + "learning_rate": 2.2674840442745705e-08, + "loss": 1.1447, + "step": 34890 + }, + { + "epoch": 2.94, + "learning_rate": 2.2613704150865612e-08, + "loss": 1.0927, + "step": 34891 + }, + { + "epoch": 2.94, + "learning_rate": 2.255265029534548e-08, + "loss": 1.0919, + "step": 34892 + }, + { + "epoch": 2.94, + "learning_rate": 2.249167887668824e-08, + "loss": 1.0643, + "step": 34893 + }, + { + "epoch": 2.94, + "learning_rate": 2.243078989539793e-08, + "loss": 1.1378, + "step": 34894 + }, + { + "epoch": 2.94, + "learning_rate": 2.2369983351977486e-08, + "loss": 1.0339, + "step": 34895 + }, + { + "epoch": 2.94, + "learning_rate": 2.230925924692984e-08, + "loss": 1.0686, + "step": 34896 + }, + { + "epoch": 2.94, + "learning_rate": 2.2248617580756805e-08, + "loss": 1.1541, + "step": 34897 + }, + { + "epoch": 2.94, + "learning_rate": 2.2188058353957988e-08, + "loss": 1.0968, + "step": 34898 + }, + { + "epoch": 2.94, + "learning_rate": 2.2127581567035205e-08, + "loss": 1.1076, + "step": 34899 + }, + { + "epoch": 2.94, + "learning_rate": 2.206718722048806e-08, + "loss": 1.1611, + "step": 34900 + }, + { + "epoch": 2.94, + "learning_rate": 2.2006875314813936e-08, + "loss": 1.0944, + "step": 34901 + }, + { + "epoch": 2.94, + "learning_rate": 2.194664585051354e-08, + "loss": 1.1043, + "step": 34902 + }, + { + "epoch": 2.94, + "learning_rate": 2.1886498828083134e-08, + "loss": 1.0708, + "step": 34903 + }, + { + "epoch": 2.94, + "learning_rate": 2.1826434248020113e-08, + "loss": 1.0468, + "step": 34904 + }, + { + "epoch": 2.94, + "learning_rate": 2.1766452110819626e-08, + "loss": 1.1609, + "step": 34905 + }, + { + "epoch": 2.94, + "learning_rate": 2.1706552416979053e-08, + "loss": 1.1316, + "step": 34906 + }, + { + "epoch": 2.94, + "learning_rate": 2.1646735166991338e-08, + "loss": 1.1339, + "step": 34907 + }, + { + "epoch": 2.94, + "learning_rate": 2.158700036135275e-08, + "loss": 1.0535, + "step": 34908 + }, + { + "epoch": 2.94, + "learning_rate": 2.1527348000555114e-08, + "loss": 1.1795, + "step": 34909 + }, + { + "epoch": 2.94, + "learning_rate": 2.1467778085092483e-08, + "loss": 1.1523, + "step": 34910 + }, + { + "epoch": 2.94, + "learning_rate": 2.1408290615456683e-08, + "loss": 1.1532, + "step": 34911 + }, + { + "epoch": 2.94, + "learning_rate": 2.1348885592139546e-08, + "loss": 1.105, + "step": 34912 + }, + { + "epoch": 2.94, + "learning_rate": 2.1289563015630677e-08, + "loss": 1.0956, + "step": 34913 + }, + { + "epoch": 2.94, + "learning_rate": 2.1230322886421907e-08, + "loss": 1.1306, + "step": 34914 + }, + { + "epoch": 2.94, + "learning_rate": 2.117116520500173e-08, + "loss": 1.0758, + "step": 34915 + }, + { + "epoch": 2.94, + "learning_rate": 2.1112089971857542e-08, + "loss": 1.0816, + "step": 34916 + }, + { + "epoch": 2.94, + "learning_rate": 2.1053097187480054e-08, + "loss": 1.1112, + "step": 34917 + }, + { + "epoch": 2.94, + "learning_rate": 2.0994186852355546e-08, + "loss": 1.0842, + "step": 34918 + }, + { + "epoch": 2.94, + "learning_rate": 2.093535896697141e-08, + "loss": 1.159, + "step": 34919 + }, + { + "epoch": 2.94, + "learning_rate": 2.0876613531811695e-08, + "loss": 1.0733, + "step": 34920 + }, + { + "epoch": 2.94, + "learning_rate": 2.0817950547363797e-08, + "loss": 1.0413, + "step": 34921 + }, + { + "epoch": 2.94, + "learning_rate": 2.0759370014111767e-08, + "loss": 1.0715, + "step": 34922 + }, + { + "epoch": 2.94, + "learning_rate": 2.070087193253967e-08, + "loss": 1.13, + "step": 34923 + }, + { + "epoch": 2.94, + "learning_rate": 2.0642456303130444e-08, + "loss": 1.104, + "step": 34924 + }, + { + "epoch": 2.94, + "learning_rate": 2.058412312636704e-08, + "loss": 1.0694, + "step": 34925 + }, + { + "epoch": 2.94, + "learning_rate": 2.0525872402732406e-08, + "loss": 1.09, + "step": 34926 + }, + { + "epoch": 2.94, + "learning_rate": 2.0467704132705046e-08, + "loss": 1.0928, + "step": 34927 + }, + { + "epoch": 2.94, + "learning_rate": 2.040961831676902e-08, + "loss": 1.0987, + "step": 34928 + }, + { + "epoch": 2.94, + "learning_rate": 2.0351614955401722e-08, + "loss": 1.0819, + "step": 34929 + }, + { + "epoch": 2.94, + "learning_rate": 2.029369404908388e-08, + "loss": 1.2239, + "step": 34930 + }, + { + "epoch": 2.94, + "learning_rate": 2.0235855598292885e-08, + "loss": 1.1674, + "step": 34931 + }, + { + "epoch": 2.94, + "learning_rate": 2.017809960350725e-08, + "loss": 1.1183, + "step": 34932 + }, + { + "epoch": 2.94, + "learning_rate": 2.012042606520437e-08, + "loss": 1.1919, + "step": 34933 + }, + { + "epoch": 2.94, + "learning_rate": 2.0062834983860524e-08, + "loss": 1.1551, + "step": 34934 + }, + { + "epoch": 2.94, + "learning_rate": 2.0005326359950894e-08, + "loss": 1.1295, + "step": 34935 + }, + { + "epoch": 2.94, + "learning_rate": 1.9947900193951764e-08, + "loss": 1.0774, + "step": 34936 + }, + { + "epoch": 2.94, + "learning_rate": 1.9890556486337198e-08, + "loss": 1.1191, + "step": 34937 + }, + { + "epoch": 2.94, + "learning_rate": 1.983329523758126e-08, + "loss": 1.1322, + "step": 34938 + }, + { + "epoch": 2.94, + "learning_rate": 1.9776116448155804e-08, + "loss": 1.1619, + "step": 34939 + }, + { + "epoch": 2.94, + "learning_rate": 1.9719020118534882e-08, + "loss": 1.1529, + "step": 34940 + }, + { + "epoch": 2.94, + "learning_rate": 1.9662006249189234e-08, + "loss": 1.0664, + "step": 34941 + }, + { + "epoch": 2.94, + "learning_rate": 1.96050748405896e-08, + "loss": 1.1147, + "step": 34942 + }, + { + "epoch": 2.94, + "learning_rate": 1.9548225893207816e-08, + "loss": 1.157, + "step": 34943 + }, + { + "epoch": 2.94, + "learning_rate": 1.9491459407512405e-08, + "loss": 1.1061, + "step": 34944 + }, + { + "epoch": 2.94, + "learning_rate": 1.943477538397298e-08, + "loss": 1.0285, + "step": 34945 + }, + { + "epoch": 2.94, + "learning_rate": 1.9378173823058066e-08, + "loss": 1.1154, + "step": 34946 + }, + { + "epoch": 2.94, + "learning_rate": 1.9321654725233952e-08, + "loss": 1.195, + "step": 34947 + }, + { + "epoch": 2.94, + "learning_rate": 1.9265218090969152e-08, + "loss": 1.0963, + "step": 34948 + }, + { + "epoch": 2.94, + "learning_rate": 1.920886392072885e-08, + "loss": 1.0762, + "step": 34949 + }, + { + "epoch": 2.94, + "learning_rate": 1.9152592214979336e-08, + "loss": 1.1036, + "step": 34950 + }, + { + "epoch": 2.94, + "learning_rate": 1.90964029741858e-08, + "loss": 1.0403, + "step": 34951 + }, + { + "epoch": 2.94, + "learning_rate": 1.9040296198811205e-08, + "loss": 1.1525, + "step": 34952 + }, + { + "epoch": 2.94, + "learning_rate": 1.8984271889320726e-08, + "loss": 1.0637, + "step": 34953 + }, + { + "epoch": 2.94, + "learning_rate": 1.8928330046176225e-08, + "loss": 1.0473, + "step": 34954 + }, + { + "epoch": 2.94, + "learning_rate": 1.887247066984066e-08, + "loss": 1.1331, + "step": 34955 + }, + { + "epoch": 2.94, + "learning_rate": 1.8816693760774775e-08, + "loss": 1.1054, + "step": 34956 + }, + { + "epoch": 2.94, + "learning_rate": 1.876099931943931e-08, + "loss": 1.1079, + "step": 34957 + }, + { + "epoch": 2.94, + "learning_rate": 1.8705387346295013e-08, + "loss": 1.0659, + "step": 34958 + }, + { + "epoch": 2.94, + "learning_rate": 1.864985784180262e-08, + "loss": 1.1459, + "step": 34959 + }, + { + "epoch": 2.94, + "learning_rate": 1.8594410806418438e-08, + "loss": 1.0571, + "step": 34960 + }, + { + "epoch": 2.94, + "learning_rate": 1.853904624060099e-08, + "loss": 1.0264, + "step": 34961 + }, + { + "epoch": 2.94, + "learning_rate": 1.84837641448099e-08, + "loss": 1.099, + "step": 34962 + }, + { + "epoch": 2.94, + "learning_rate": 1.842856451949926e-08, + "loss": 1.0685, + "step": 34963 + }, + { + "epoch": 2.94, + "learning_rate": 1.8373447365126474e-08, + "loss": 1.1091, + "step": 34964 + }, + { + "epoch": 2.94, + "learning_rate": 1.8318412682147847e-08, + "loss": 1.1527, + "step": 34965 + }, + { + "epoch": 2.94, + "learning_rate": 1.8263460471016348e-08, + "loss": 1.0174, + "step": 34966 + }, + { + "epoch": 2.94, + "learning_rate": 1.820859073218606e-08, + "loss": 1.0595, + "step": 34967 + }, + { + "epoch": 2.94, + "learning_rate": 1.8153803466112174e-08, + "loss": 1.2181, + "step": 34968 + }, + { + "epoch": 2.94, + "learning_rate": 1.809909867324544e-08, + "loss": 1.0622, + "step": 34969 + }, + { + "epoch": 2.94, + "learning_rate": 1.804447635403883e-08, + "loss": 1.1367, + "step": 34970 + }, + { + "epoch": 2.94, + "learning_rate": 1.7989936508943095e-08, + "loss": 1.1227, + "step": 34971 + }, + { + "epoch": 2.94, + "learning_rate": 1.7935479138408984e-08, + "loss": 1.1536, + "step": 34972 + }, + { + "epoch": 2.94, + "learning_rate": 1.7881104242887247e-08, + "loss": 1.1271, + "step": 34973 + }, + { + "epoch": 2.94, + "learning_rate": 1.7826811822826417e-08, + "loss": 1.0852, + "step": 34974 + }, + { + "epoch": 2.94, + "learning_rate": 1.777260187867391e-08, + "loss": 1.0841, + "step": 34975 + }, + { + "epoch": 2.94, + "learning_rate": 1.771847441088048e-08, + "loss": 1.1307, + "step": 34976 + }, + { + "epoch": 2.94, + "learning_rate": 1.7664429419891328e-08, + "loss": 1.1205, + "step": 34977 + }, + { + "epoch": 2.94, + "learning_rate": 1.7610466906152756e-08, + "loss": 1.0823, + "step": 34978 + }, + { + "epoch": 2.94, + "learning_rate": 1.755658687011219e-08, + "loss": 1.179, + "step": 34979 + }, + { + "epoch": 2.95, + "learning_rate": 1.750278931221372e-08, + "loss": 1.005, + "step": 34980 + }, + { + "epoch": 2.95, + "learning_rate": 1.744907423290143e-08, + "loss": 1.1023, + "step": 34981 + }, + { + "epoch": 2.95, + "learning_rate": 1.7395441632620526e-08, + "loss": 1.1978, + "step": 34982 + }, + { + "epoch": 2.95, + "learning_rate": 1.7341891511812868e-08, + "loss": 1.025, + "step": 34983 + }, + { + "epoch": 2.95, + "learning_rate": 1.7288423870921445e-08, + "loss": 1.0693, + "step": 34984 + }, + { + "epoch": 2.95, + "learning_rate": 1.723503871038812e-08, + "loss": 1.1327, + "step": 34985 + }, + { + "epoch": 2.95, + "learning_rate": 1.7181736030653652e-08, + "loss": 1.1142, + "step": 34986 + }, + { + "epoch": 2.95, + "learning_rate": 1.7128515832157687e-08, + "loss": 1.008, + "step": 34987 + }, + { + "epoch": 2.95, + "learning_rate": 1.70753781153421e-08, + "loss": 1.0634, + "step": 34988 + }, + { + "epoch": 2.95, + "learning_rate": 1.7022322880644315e-08, + "loss": 1.0523, + "step": 34989 + }, + { + "epoch": 2.95, + "learning_rate": 1.696935012850287e-08, + "loss": 1.1275, + "step": 34990 + }, + { + "epoch": 2.95, + "learning_rate": 1.6916459859356305e-08, + "loss": 1.1684, + "step": 34991 + }, + { + "epoch": 2.95, + "learning_rate": 1.6863652073639823e-08, + "loss": 1.1093, + "step": 34992 + }, + { + "epoch": 2.95, + "learning_rate": 1.681092677179197e-08, + "loss": 1.1168, + "step": 34993 + }, + { + "epoch": 2.95, + "learning_rate": 1.6758283954246835e-08, + "loss": 1.0603, + "step": 34994 + }, + { + "epoch": 2.95, + "learning_rate": 1.670572362143963e-08, + "loss": 1.2411, + "step": 34995 + }, + { + "epoch": 2.95, + "learning_rate": 1.665324577380445e-08, + "loss": 1.1796, + "step": 34996 + }, + { + "epoch": 2.95, + "learning_rate": 1.6600850411775392e-08, + "loss": 1.1004, + "step": 34997 + }, + { + "epoch": 2.95, + "learning_rate": 1.6548537535786558e-08, + "loss": 1.0959, + "step": 34998 + }, + { + "epoch": 2.95, + "learning_rate": 1.6496307146267598e-08, + "loss": 1.1384, + "step": 34999 + }, + { + "epoch": 2.95, + "learning_rate": 1.64441592436515e-08, + "loss": 1.1684, + "step": 35000 + }, + { + "epoch": 2.95, + "learning_rate": 1.6392093828367926e-08, + "loss": 1.1002, + "step": 35001 + }, + { + "epoch": 2.95, + "learning_rate": 1.634011090084986e-08, + "loss": 1.0033, + "step": 35002 + }, + { + "epoch": 2.95, + "learning_rate": 1.6288210461523624e-08, + "loss": 1.1487, + "step": 35003 + }, + { + "epoch": 2.95, + "learning_rate": 1.6236392510819987e-08, + "loss": 1.0253, + "step": 35004 + }, + { + "epoch": 2.95, + "learning_rate": 1.6184657049166387e-08, + "loss": 1.1581, + "step": 35005 + }, + { + "epoch": 2.95, + "learning_rate": 1.6133004076990257e-08, + "loss": 1.1192, + "step": 35006 + }, + { + "epoch": 2.95, + "learning_rate": 1.608143359471903e-08, + "loss": 1.0969, + "step": 35007 + }, + { + "epoch": 2.95, + "learning_rate": 1.602994560277682e-08, + "loss": 1.0673, + "step": 35008 + }, + { + "epoch": 2.95, + "learning_rate": 1.5978540101592165e-08, + "loss": 1.1534, + "step": 35009 + }, + { + "epoch": 2.95, + "learning_rate": 1.5927217091586953e-08, + "loss": 1.062, + "step": 35010 + }, + { + "epoch": 2.95, + "learning_rate": 1.5875976573186403e-08, + "loss": 1.0236, + "step": 35011 + }, + { + "epoch": 2.95, + "learning_rate": 1.5824818546814612e-08, + "loss": 1.1472, + "step": 35012 + }, + { + "epoch": 2.95, + "learning_rate": 1.5773743012893473e-08, + "loss": 1.1912, + "step": 35013 + }, + { + "epoch": 2.95, + "learning_rate": 1.5722749971844864e-08, + "loss": 1.2061, + "step": 35014 + }, + { + "epoch": 2.95, + "learning_rate": 1.5671839424090674e-08, + "loss": 1.1615, + "step": 35015 + }, + { + "epoch": 2.95, + "learning_rate": 1.5621011370050564e-08, + "loss": 1.0828, + "step": 35016 + }, + { + "epoch": 2.95, + "learning_rate": 1.5570265810144204e-08, + "loss": 1.0971, + "step": 35017 + }, + { + "epoch": 2.95, + "learning_rate": 1.551960274479347e-08, + "loss": 1.0823, + "step": 35018 + }, + { + "epoch": 2.95, + "learning_rate": 1.5469022174413595e-08, + "loss": 1.081, + "step": 35019 + }, + { + "epoch": 2.95, + "learning_rate": 1.5418524099425347e-08, + "loss": 1.1985, + "step": 35020 + }, + { + "epoch": 2.95, + "learning_rate": 1.536810852024395e-08, + "loss": 1.0765, + "step": 35021 + }, + { + "epoch": 2.95, + "learning_rate": 1.5317775437287962e-08, + "loss": 1.1187, + "step": 35022 + }, + { + "epoch": 2.95, + "learning_rate": 1.5267524850970382e-08, + "loss": 1.0752, + "step": 35023 + }, + { + "epoch": 2.95, + "learning_rate": 1.5217356761709768e-08, + "loss": 1.0523, + "step": 35024 + }, + { + "epoch": 2.95, + "learning_rate": 1.5167271169916897e-08, + "loss": 1.0203, + "step": 35025 + }, + { + "epoch": 2.95, + "learning_rate": 1.5117268076009218e-08, + "loss": 1.1971, + "step": 35026 + }, + { + "epoch": 2.95, + "learning_rate": 1.506734748039751e-08, + "loss": 1.2027, + "step": 35027 + }, + { + "epoch": 2.95, + "learning_rate": 1.5017509383494777e-08, + "loss": 1.1096, + "step": 35028 + }, + { + "epoch": 2.95, + "learning_rate": 1.4967753785714025e-08, + "loss": 1.0353, + "step": 35029 + }, + { + "epoch": 2.95, + "learning_rate": 1.491808068746381e-08, + "loss": 1.1113, + "step": 35030 + }, + { + "epoch": 2.95, + "learning_rate": 1.486849008915714e-08, + "loss": 1.0981, + "step": 35031 + }, + { + "epoch": 2.95, + "learning_rate": 1.4818981991202575e-08, + "loss": 1.1701, + "step": 35032 + }, + { + "epoch": 2.95, + "learning_rate": 1.4769556394008678e-08, + "loss": 1.0992, + "step": 35033 + }, + { + "epoch": 2.95, + "learning_rate": 1.472021329798512e-08, + "loss": 1.2241, + "step": 35034 + }, + { + "epoch": 2.95, + "learning_rate": 1.4670952703538245e-08, + "loss": 1.15, + "step": 35035 + }, + { + "epoch": 2.95, + "learning_rate": 1.4621774611075501e-08, + "loss": 1.0918, + "step": 35036 + }, + { + "epoch": 2.95, + "learning_rate": 1.4572679021004344e-08, + "loss": 1.1593, + "step": 35037 + }, + { + "epoch": 2.95, + "learning_rate": 1.4523665933728893e-08, + "loss": 1.0542, + "step": 35038 + }, + { + "epoch": 2.95, + "learning_rate": 1.4474735349654379e-08, + "loss": 1.1104, + "step": 35039 + }, + { + "epoch": 2.95, + "learning_rate": 1.4425887269186033e-08, + "loss": 1.1133, + "step": 35040 + }, + { + "epoch": 2.95, + "learning_rate": 1.437712169272687e-08, + "loss": 1.0957, + "step": 35041 + }, + { + "epoch": 2.95, + "learning_rate": 1.4328438620678786e-08, + "loss": 1.1679, + "step": 35042 + }, + { + "epoch": 2.95, + "learning_rate": 1.4279838053445905e-08, + "loss": 1.0956, + "step": 35043 + }, + { + "epoch": 2.95, + "learning_rate": 1.4231319991427905e-08, + "loss": 1.1318, + "step": 35044 + }, + { + "epoch": 2.95, + "learning_rate": 1.41828844350278e-08, + "loss": 1.1664, + "step": 35045 + }, + { + "epoch": 2.95, + "learning_rate": 1.4134531384644156e-08, + "loss": 1.0683, + "step": 35046 + }, + { + "epoch": 2.95, + "learning_rate": 1.4086260840675548e-08, + "loss": 1.1445, + "step": 35047 + }, + { + "epoch": 2.95, + "learning_rate": 1.4038072803523872e-08, + "loss": 1.1764, + "step": 35048 + }, + { + "epoch": 2.95, + "learning_rate": 1.398996727358437e-08, + "loss": 1.0983, + "step": 35049 + }, + { + "epoch": 2.95, + "learning_rate": 1.3941944251256722e-08, + "loss": 1.1125, + "step": 35050 + }, + { + "epoch": 2.95, + "learning_rate": 1.3894003736936168e-08, + "loss": 1.144, + "step": 35051 + }, + { + "epoch": 2.95, + "learning_rate": 1.3846145731019057e-08, + "loss": 1.1753, + "step": 35052 + }, + { + "epoch": 2.95, + "learning_rate": 1.3798370233901737e-08, + "loss": 1.1276, + "step": 35053 + }, + { + "epoch": 2.95, + "learning_rate": 1.375067724597834e-08, + "loss": 1.1464, + "step": 35054 + }, + { + "epoch": 2.95, + "learning_rate": 1.3703066767641882e-08, + "loss": 1.08, + "step": 35055 + }, + { + "epoch": 2.95, + "learning_rate": 1.3655538799287604e-08, + "loss": 1.0455, + "step": 35056 + }, + { + "epoch": 2.95, + "learning_rate": 1.3608093341307416e-08, + "loss": 1.132, + "step": 35057 + }, + { + "epoch": 2.95, + "learning_rate": 1.3560730394092114e-08, + "loss": 1.0976, + "step": 35058 + }, + { + "epoch": 2.95, + "learning_rate": 1.3513449958034718e-08, + "loss": 1.1122, + "step": 35059 + }, + { + "epoch": 2.95, + "learning_rate": 1.3466252033526028e-08, + "loss": 1.0552, + "step": 35060 + }, + { + "epoch": 2.95, + "learning_rate": 1.341913662095351e-08, + "loss": 1.1908, + "step": 35061 + }, + { + "epoch": 2.95, + "learning_rate": 1.3372103720710183e-08, + "loss": 1.07, + "step": 35062 + }, + { + "epoch": 2.95, + "learning_rate": 1.3325153333181296e-08, + "loss": 1.2189, + "step": 35063 + }, + { + "epoch": 2.95, + "learning_rate": 1.3278285458756535e-08, + "loss": 0.9649, + "step": 35064 + }, + { + "epoch": 2.95, + "learning_rate": 1.3231500097823369e-08, + "loss": 1.1065, + "step": 35065 + }, + { + "epoch": 2.95, + "learning_rate": 1.3184797250767046e-08, + "loss": 1.0511, + "step": 35066 + }, + { + "epoch": 2.95, + "learning_rate": 1.3138176917973922e-08, + "loss": 1.0682, + "step": 35067 + }, + { + "epoch": 2.95, + "learning_rate": 1.3091639099829245e-08, + "loss": 1.1464, + "step": 35068 + }, + { + "epoch": 2.95, + "learning_rate": 1.3045183796718265e-08, + "loss": 1.0818, + "step": 35069 + }, + { + "epoch": 2.95, + "learning_rate": 1.2998811009024003e-08, + "loss": 1.0764, + "step": 35070 + }, + { + "epoch": 2.95, + "learning_rate": 1.2952520737130602e-08, + "loss": 1.0817, + "step": 35071 + }, + { + "epoch": 2.95, + "learning_rate": 1.2906312981418867e-08, + "loss": 1.1042, + "step": 35072 + }, + { + "epoch": 2.95, + "learning_rate": 1.2860187742271824e-08, + "loss": 1.1399, + "step": 35073 + }, + { + "epoch": 2.95, + "learning_rate": 1.2814145020069168e-08, + "loss": 1.1466, + "step": 35074 + }, + { + "epoch": 2.95, + "learning_rate": 1.276818481519393e-08, + "loss": 1.1687, + "step": 35075 + }, + { + "epoch": 2.95, + "learning_rate": 1.272230712802358e-08, + "loss": 1.1848, + "step": 35076 + }, + { + "epoch": 2.95, + "learning_rate": 1.2676511958937821e-08, + "loss": 1.0429, + "step": 35077 + }, + { + "epoch": 2.95, + "learning_rate": 1.2630799308314123e-08, + "loss": 1.0862, + "step": 35078 + }, + { + "epoch": 2.95, + "learning_rate": 1.2585169176532185e-08, + "loss": 1.1846, + "step": 35079 + }, + { + "epoch": 2.95, + "learning_rate": 1.2539621563968373e-08, + "loss": 1.1468, + "step": 35080 + }, + { + "epoch": 2.95, + "learning_rate": 1.2494156470997942e-08, + "loss": 1.1024, + "step": 35081 + }, + { + "epoch": 2.95, + "learning_rate": 1.2448773897997257e-08, + "loss": 1.19, + "step": 35082 + }, + { + "epoch": 2.95, + "learning_rate": 1.2403473845341574e-08, + "loss": 1.1433, + "step": 35083 + }, + { + "epoch": 2.95, + "learning_rate": 1.2358256313403927e-08, + "loss": 1.1062, + "step": 35084 + }, + { + "epoch": 2.95, + "learning_rate": 1.2313121302559571e-08, + "loss": 1.0685, + "step": 35085 + }, + { + "epoch": 2.95, + "learning_rate": 1.2268068813180433e-08, + "loss": 1.0463, + "step": 35086 + }, + { + "epoch": 2.95, + "learning_rate": 1.2223098845639548e-08, + "loss": 1.1172, + "step": 35087 + }, + { + "epoch": 2.95, + "learning_rate": 1.2178211400307727e-08, + "loss": 1.151, + "step": 35088 + }, + { + "epoch": 2.95, + "learning_rate": 1.2133406477555787e-08, + "loss": 1.147, + "step": 35089 + }, + { + "epoch": 2.95, + "learning_rate": 1.2088684077753433e-08, + "loss": 1.0944, + "step": 35090 + }, + { + "epoch": 2.95, + "learning_rate": 1.2044044201271477e-08, + "loss": 1.1217, + "step": 35091 + }, + { + "epoch": 2.95, + "learning_rate": 1.1999486848478515e-08, + "loss": 1.0248, + "step": 35092 + }, + { + "epoch": 2.95, + "learning_rate": 1.195501201974314e-08, + "loss": 1.1229, + "step": 35093 + }, + { + "epoch": 2.95, + "learning_rate": 1.1910619715430615e-08, + "loss": 1.173, + "step": 35094 + }, + { + "epoch": 2.95, + "learning_rate": 1.1866309935909536e-08, + "loss": 1.1864, + "step": 35095 + }, + { + "epoch": 2.95, + "learning_rate": 1.1822082681546277e-08, + "loss": 1.1573, + "step": 35096 + }, + { + "epoch": 2.95, + "learning_rate": 1.1777937952703877e-08, + "loss": 1.1587, + "step": 35097 + }, + { + "epoch": 2.95, + "learning_rate": 1.1733875749750933e-08, + "loss": 1.1643, + "step": 35098 + }, + { + "epoch": 2.96, + "learning_rate": 1.1689896073048268e-08, + "loss": 1.1407, + "step": 35099 + }, + { + "epoch": 2.96, + "learning_rate": 1.1645998922960034e-08, + "loss": 1.0236, + "step": 35100 + }, + { + "epoch": 2.96, + "learning_rate": 1.1602184299850382e-08, + "loss": 1.0575, + "step": 35101 + }, + { + "epoch": 2.96, + "learning_rate": 1.155845220407903e-08, + "loss": 1.1763, + "step": 35102 + }, + { + "epoch": 2.96, + "learning_rate": 1.1514802636009014e-08, + "loss": 1.0906, + "step": 35103 + }, + { + "epoch": 2.96, + "learning_rate": 1.1471235596000052e-08, + "loss": 1.145, + "step": 35104 + }, + { + "epoch": 2.96, + "learning_rate": 1.1427751084412964e-08, + "loss": 1.1245, + "step": 35105 + }, + { + "epoch": 2.96, + "learning_rate": 1.1384349101606352e-08, + "loss": 1.1422, + "step": 35106 + }, + { + "epoch": 2.96, + "learning_rate": 1.1341029647938817e-08, + "loss": 1.0414, + "step": 35107 + }, + { + "epoch": 2.96, + "learning_rate": 1.1297792723768964e-08, + "loss": 1.1364, + "step": 35108 + }, + { + "epoch": 2.96, + "learning_rate": 1.1254638329453171e-08, + "loss": 1.2539, + "step": 35109 + }, + { + "epoch": 2.96, + "learning_rate": 1.1211566465347823e-08, + "loss": 1.0549, + "step": 35110 + }, + { + "epoch": 2.96, + "learning_rate": 1.116857713181041e-08, + "loss": 1.212, + "step": 35111 + }, + { + "epoch": 2.96, + "learning_rate": 1.1125670329193982e-08, + "loss": 1.0471, + "step": 35112 + }, + { + "epoch": 2.96, + "learning_rate": 1.1082846057854924e-08, + "loss": 1.0311, + "step": 35113 + }, + { + "epoch": 2.96, + "learning_rate": 1.1040104318146283e-08, + "loss": 1.0472, + "step": 35114 + }, + { + "epoch": 2.96, + "learning_rate": 1.0997445110421113e-08, + "loss": 1.109, + "step": 35115 + }, + { + "epoch": 2.96, + "learning_rate": 1.0954868435031351e-08, + "loss": 1.1517, + "step": 35116 + }, + { + "epoch": 2.96, + "learning_rate": 1.0912374292330053e-08, + "loss": 1.073, + "step": 35117 + }, + { + "epoch": 2.96, + "learning_rate": 1.0869962682666935e-08, + "loss": 1.14, + "step": 35118 + }, + { + "epoch": 2.96, + "learning_rate": 1.0827633606393939e-08, + "loss": 1.2109, + "step": 35119 + }, + { + "epoch": 2.96, + "learning_rate": 1.0785387063859676e-08, + "loss": 1.0934, + "step": 35120 + }, + { + "epoch": 2.96, + "learning_rate": 1.0743223055412755e-08, + "loss": 1.1612, + "step": 35121 + }, + { + "epoch": 2.96, + "learning_rate": 1.0701141581402897e-08, + "loss": 1.0977, + "step": 35122 + }, + { + "epoch": 2.96, + "learning_rate": 1.06591426421776e-08, + "loss": 1.0796, + "step": 35123 + }, + { + "epoch": 2.96, + "learning_rate": 1.0617226238083256e-08, + "loss": 1.1322, + "step": 35124 + }, + { + "epoch": 2.96, + "learning_rate": 1.0575392369466254e-08, + "loss": 1.0881, + "step": 35125 + }, + { + "epoch": 2.96, + "learning_rate": 1.0533641036671871e-08, + "loss": 1.1385, + "step": 35126 + }, + { + "epoch": 2.96, + "learning_rate": 1.04919722400465e-08, + "loss": 1.0663, + "step": 35127 + }, + { + "epoch": 2.96, + "learning_rate": 1.0450385979933197e-08, + "loss": 1.1246, + "step": 35128 + }, + { + "epoch": 2.96, + "learning_rate": 1.0408882256676134e-08, + "loss": 1.1084, + "step": 35129 + }, + { + "epoch": 2.96, + "learning_rate": 1.0367461070617257e-08, + "loss": 1.068, + "step": 35130 + }, + { + "epoch": 2.96, + "learning_rate": 1.0326122422099628e-08, + "loss": 1.1205, + "step": 35131 + }, + { + "epoch": 2.96, + "learning_rate": 1.0284866311465191e-08, + "loss": 1.1792, + "step": 35132 + }, + { + "epoch": 2.96, + "learning_rate": 1.0243692739053679e-08, + "loss": 1.1365, + "step": 35133 + }, + { + "epoch": 2.96, + "learning_rate": 1.0202601705205928e-08, + "loss": 1.1178, + "step": 35134 + }, + { + "epoch": 2.96, + "learning_rate": 1.0161593210262777e-08, + "loss": 1.1037, + "step": 35135 + }, + { + "epoch": 2.96, + "learning_rate": 1.0120667254559513e-08, + "loss": 1.1573, + "step": 35136 + }, + { + "epoch": 2.96, + "learning_rate": 1.0079823838438085e-08, + "loss": 1.1308, + "step": 35137 + }, + { + "epoch": 2.96, + "learning_rate": 1.003906296223378e-08, + "loss": 1.1384, + "step": 35138 + }, + { + "epoch": 2.96, + "learning_rate": 9.998384626284107e-09, + "loss": 1.1345, + "step": 35139 + }, + { + "epoch": 2.96, + "learning_rate": 9.957788830925463e-09, + "loss": 1.0485, + "step": 35140 + }, + { + "epoch": 2.96, + "learning_rate": 9.917275576492026e-09, + "loss": 1.1442, + "step": 35141 + }, + { + "epoch": 2.96, + "learning_rate": 9.87684486331908e-09, + "loss": 0.9972, + "step": 35142 + }, + { + "epoch": 2.96, + "learning_rate": 9.836496691741914e-09, + "loss": 1.246, + "step": 35143 + }, + { + "epoch": 2.96, + "learning_rate": 9.796231062091378e-09, + "loss": 1.0724, + "step": 35144 + }, + { + "epoch": 2.96, + "learning_rate": 9.756047974702754e-09, + "loss": 1.0728, + "step": 35145 + }, + { + "epoch": 2.96, + "learning_rate": 9.715947429906891e-09, + "loss": 1.1264, + "step": 35146 + }, + { + "epoch": 2.96, + "learning_rate": 9.675929428035746e-09, + "loss": 1.0851, + "step": 35147 + }, + { + "epoch": 2.96, + "learning_rate": 9.635993969417946e-09, + "loss": 1.12, + "step": 35148 + }, + { + "epoch": 2.96, + "learning_rate": 9.596141054385443e-09, + "loss": 1.15, + "step": 35149 + }, + { + "epoch": 2.96, + "learning_rate": 9.55637068326798e-09, + "loss": 1.1173, + "step": 35150 + }, + { + "epoch": 2.96, + "learning_rate": 9.516682856391956e-09, + "loss": 1.178, + "step": 35151 + }, + { + "epoch": 2.96, + "learning_rate": 9.477077574087112e-09, + "loss": 1.015, + "step": 35152 + }, + { + "epoch": 2.96, + "learning_rate": 9.437554836679852e-09, + "loss": 1.1005, + "step": 35153 + }, + { + "epoch": 2.96, + "learning_rate": 9.39811464449658e-09, + "loss": 1.1176, + "step": 35154 + }, + { + "epoch": 2.96, + "learning_rate": 9.358756997863705e-09, + "loss": 1.1315, + "step": 35155 + }, + { + "epoch": 2.96, + "learning_rate": 9.31948189710541e-09, + "loss": 1.0631, + "step": 35156 + }, + { + "epoch": 2.96, + "learning_rate": 9.280289342548098e-09, + "loss": 1.1328, + "step": 35157 + }, + { + "epoch": 2.96, + "learning_rate": 9.24117933451374e-09, + "loss": 1.1289, + "step": 35158 + }, + { + "epoch": 2.96, + "learning_rate": 9.202151873327625e-09, + "loss": 1.1529, + "step": 35159 + }, + { + "epoch": 2.96, + "learning_rate": 9.1632069593095e-09, + "loss": 1.1519, + "step": 35160 + }, + { + "epoch": 2.96, + "learning_rate": 9.12434459278244e-09, + "loss": 1.1179, + "step": 35161 + }, + { + "epoch": 2.96, + "learning_rate": 9.085564774067302e-09, + "loss": 1.1217, + "step": 35162 + }, + { + "epoch": 2.96, + "learning_rate": 9.046867503486046e-09, + "loss": 1.1267, + "step": 35163 + }, + { + "epoch": 2.96, + "learning_rate": 9.008252781356198e-09, + "loss": 1.0973, + "step": 35164 + }, + { + "epoch": 2.96, + "learning_rate": 8.9697206079975e-09, + "loss": 1.1235, + "step": 35165 + }, + { + "epoch": 2.96, + "learning_rate": 8.93127098372859e-09, + "loss": 1.2202, + "step": 35166 + }, + { + "epoch": 2.96, + "learning_rate": 8.89290390886699e-09, + "loss": 1.112, + "step": 35167 + }, + { + "epoch": 2.96, + "learning_rate": 8.854619383730223e-09, + "loss": 1.1176, + "step": 35168 + }, + { + "epoch": 2.96, + "learning_rate": 8.816417408634704e-09, + "loss": 1.0474, + "step": 35169 + }, + { + "epoch": 2.96, + "learning_rate": 8.778297983894623e-09, + "loss": 1.1255, + "step": 35170 + }, + { + "epoch": 2.96, + "learning_rate": 8.740261109825287e-09, + "loss": 1.0018, + "step": 35171 + }, + { + "epoch": 2.96, + "learning_rate": 8.702306786743108e-09, + "loss": 1.1459, + "step": 35172 + }, + { + "epoch": 2.96, + "learning_rate": 8.66443501495895e-09, + "loss": 1.0782, + "step": 35173 + }, + { + "epoch": 2.96, + "learning_rate": 8.626645794787003e-09, + "loss": 1.1291, + "step": 35174 + }, + { + "epoch": 2.96, + "learning_rate": 8.588939126539242e-09, + "loss": 1.0505, + "step": 35175 + }, + { + "epoch": 2.96, + "learning_rate": 8.55131501052764e-09, + "loss": 1.0862, + "step": 35176 + }, + { + "epoch": 2.96, + "learning_rate": 8.513773447063056e-09, + "loss": 1.1905, + "step": 35177 + }, + { + "epoch": 2.96, + "learning_rate": 8.476314436455246e-09, + "loss": 1.1063, + "step": 35178 + }, + { + "epoch": 2.96, + "learning_rate": 8.438937979012852e-09, + "loss": 1.0992, + "step": 35179 + }, + { + "epoch": 2.96, + "learning_rate": 8.401644075046732e-09, + "loss": 1.0688, + "step": 35180 + }, + { + "epoch": 2.96, + "learning_rate": 8.364432724863314e-09, + "loss": 1.1494, + "step": 35181 + }, + { + "epoch": 2.96, + "learning_rate": 8.327303928771236e-09, + "loss": 1.1179, + "step": 35182 + }, + { + "epoch": 2.96, + "learning_rate": 8.290257687076919e-09, + "loss": 1.0796, + "step": 35183 + }, + { + "epoch": 2.96, + "learning_rate": 8.253294000085677e-09, + "loss": 1.1749, + "step": 35184 + }, + { + "epoch": 2.96, + "learning_rate": 8.216412868103929e-09, + "loss": 1.1535, + "step": 35185 + }, + { + "epoch": 2.96, + "learning_rate": 8.179614291434767e-09, + "loss": 1.1253, + "step": 35186 + }, + { + "epoch": 2.96, + "learning_rate": 8.142898270384614e-09, + "loss": 1.1457, + "step": 35187 + }, + { + "epoch": 2.96, + "learning_rate": 8.106264805255449e-09, + "loss": 1.1027, + "step": 35188 + }, + { + "epoch": 2.96, + "learning_rate": 8.069713896350362e-09, + "loss": 0.9862, + "step": 35189 + }, + { + "epoch": 2.96, + "learning_rate": 8.033245543970225e-09, + "loss": 1.069, + "step": 35190 + }, + { + "epoch": 2.96, + "learning_rate": 7.996859748418129e-09, + "loss": 1.1779, + "step": 35191 + }, + { + "epoch": 2.96, + "learning_rate": 7.960556509993833e-09, + "loss": 1.073, + "step": 35192 + }, + { + "epoch": 2.96, + "learning_rate": 7.924335828997099e-09, + "loss": 1.1061, + "step": 35193 + }, + { + "epoch": 2.96, + "learning_rate": 7.888197705726574e-09, + "loss": 1.1721, + "step": 35194 + }, + { + "epoch": 2.96, + "learning_rate": 7.852142140482023e-09, + "loss": 1.1008, + "step": 35195 + }, + { + "epoch": 2.96, + "learning_rate": 7.81616913356209e-09, + "loss": 1.0901, + "step": 35196 + }, + { + "epoch": 2.96, + "learning_rate": 7.780278685262099e-09, + "loss": 1.0896, + "step": 35197 + }, + { + "epoch": 2.96, + "learning_rate": 7.744470795878478e-09, + "loss": 1.08, + "step": 35198 + }, + { + "epoch": 2.96, + "learning_rate": 7.708745465708767e-09, + "loss": 1.1108, + "step": 35199 + }, + { + "epoch": 2.96, + "learning_rate": 7.673102695048285e-09, + "loss": 1.184, + "step": 35200 + }, + { + "epoch": 2.96, + "learning_rate": 7.63754248418902e-09, + "loss": 1.0889, + "step": 35201 + }, + { + "epoch": 2.96, + "learning_rate": 7.602064833427403e-09, + "loss": 1.1554, + "step": 35202 + }, + { + "epoch": 2.96, + "learning_rate": 7.56666974305542e-09, + "loss": 1.1137, + "step": 35203 + }, + { + "epoch": 2.96, + "learning_rate": 7.531357213365064e-09, + "loss": 1.0401, + "step": 35204 + }, + { + "epoch": 2.96, + "learning_rate": 7.496127244648322e-09, + "loss": 1.1782, + "step": 35205 + }, + { + "epoch": 2.96, + "learning_rate": 7.46097983719718e-09, + "loss": 1.0631, + "step": 35206 + }, + { + "epoch": 2.96, + "learning_rate": 7.425914991301408e-09, + "loss": 1.0919, + "step": 35207 + }, + { + "epoch": 2.96, + "learning_rate": 7.390932707250776e-09, + "loss": 1.05, + "step": 35208 + }, + { + "epoch": 2.96, + "learning_rate": 7.3560329853339384e-09, + "loss": 1.065, + "step": 35209 + }, + { + "epoch": 2.96, + "learning_rate": 7.3212158258406665e-09, + "loss": 1.1648, + "step": 35210 + }, + { + "epoch": 2.96, + "learning_rate": 7.286481229056286e-09, + "loss": 1.1756, + "step": 35211 + }, + { + "epoch": 2.96, + "learning_rate": 7.251829195269455e-09, + "loss": 1.1021, + "step": 35212 + }, + { + "epoch": 2.96, + "learning_rate": 7.217259724766612e-09, + "loss": 1.0103, + "step": 35213 + }, + { + "epoch": 2.96, + "learning_rate": 7.1827728178330835e-09, + "loss": 1.1485, + "step": 35214 + }, + { + "epoch": 2.96, + "learning_rate": 7.148368474753087e-09, + "loss": 1.0753, + "step": 35215 + }, + { + "epoch": 2.96, + "learning_rate": 7.114046695810839e-09, + "loss": 1.1368, + "step": 35216 + }, + { + "epoch": 2.96, + "learning_rate": 7.079807481291667e-09, + "loss": 0.9959, + "step": 35217 + }, + { + "epoch": 2.97, + "learning_rate": 7.045650831476458e-09, + "loss": 1.113, + "step": 35218 + }, + { + "epoch": 2.97, + "learning_rate": 7.011576746648319e-09, + "loss": 1.0718, + "step": 35219 + }, + { + "epoch": 2.97, + "learning_rate": 6.977585227088135e-09, + "loss": 1.1647, + "step": 35220 + }, + { + "epoch": 2.97, + "learning_rate": 6.943676273077904e-09, + "loss": 1.1153, + "step": 35221 + }, + { + "epoch": 2.97, + "learning_rate": 6.909849884897401e-09, + "loss": 1.179, + "step": 35222 + }, + { + "epoch": 2.97, + "learning_rate": 6.876106062825294e-09, + "loss": 1.0511, + "step": 35223 + }, + { + "epoch": 2.97, + "learning_rate": 6.842444807142468e-09, + "loss": 1.047, + "step": 35224 + }, + { + "epoch": 2.97, + "learning_rate": 6.808866118124258e-09, + "loss": 1.0749, + "step": 35225 + }, + { + "epoch": 2.97, + "learning_rate": 6.77536999604933e-09, + "loss": 1.1207, + "step": 35226 + }, + { + "epoch": 2.97, + "learning_rate": 6.741956441195241e-09, + "loss": 1.1195, + "step": 35227 + }, + { + "epoch": 2.97, + "learning_rate": 6.708625453837325e-09, + "loss": 1.1708, + "step": 35228 + }, + { + "epoch": 2.97, + "learning_rate": 6.675377034250918e-09, + "loss": 1.0457, + "step": 35229 + }, + { + "epoch": 2.97, + "learning_rate": 6.6422111827102455e-09, + "loss": 1.1084, + "step": 35230 + }, + { + "epoch": 2.97, + "learning_rate": 6.609127899490641e-09, + "loss": 1.1581, + "step": 35231 + }, + { + "epoch": 2.97, + "learning_rate": 6.5761271848641115e-09, + "loss": 1.1334, + "step": 35232 + }, + { + "epoch": 2.97, + "learning_rate": 6.5432090391037705e-09, + "loss": 1.1861, + "step": 35233 + }, + { + "epoch": 2.97, + "learning_rate": 6.510373462482733e-09, + "loss": 1.1809, + "step": 35234 + }, + { + "epoch": 2.97, + "learning_rate": 6.477620455270783e-09, + "loss": 1.1264, + "step": 35235 + }, + { + "epoch": 2.97, + "learning_rate": 6.444950017738816e-09, + "loss": 1.2099, + "step": 35236 + }, + { + "epoch": 2.97, + "learning_rate": 6.4123621501577246e-09, + "loss": 1.1007, + "step": 35237 + }, + { + "epoch": 2.97, + "learning_rate": 6.3798568527950745e-09, + "loss": 1.1772, + "step": 35238 + }, + { + "epoch": 2.97, + "learning_rate": 6.347434125921758e-09, + "loss": 1.1696, + "step": 35239 + }, + { + "epoch": 2.97, + "learning_rate": 6.315093969803121e-09, + "loss": 1.1072, + "step": 35240 + }, + { + "epoch": 2.97, + "learning_rate": 6.282836384707836e-09, + "loss": 1.0483, + "step": 35241 + }, + { + "epoch": 2.97, + "learning_rate": 6.250661370902356e-09, + "loss": 1.0942, + "step": 35242 + }, + { + "epoch": 2.97, + "learning_rate": 6.218568928652024e-09, + "loss": 1.2082, + "step": 35243 + }, + { + "epoch": 2.97, + "learning_rate": 6.1865590582221855e-09, + "loss": 1.148, + "step": 35244 + }, + { + "epoch": 2.97, + "learning_rate": 6.154631759878182e-09, + "loss": 1.1606, + "step": 35245 + }, + { + "epoch": 2.97, + "learning_rate": 6.122787033883137e-09, + "loss": 1.0801, + "step": 35246 + }, + { + "epoch": 2.97, + "learning_rate": 6.0910248804990634e-09, + "loss": 1.1804, + "step": 35247 + }, + { + "epoch": 2.97, + "learning_rate": 6.059345299990194e-09, + "loss": 1.2005, + "step": 35248 + }, + { + "epoch": 2.97, + "learning_rate": 6.02774829261743e-09, + "loss": 1.1305, + "step": 35249 + }, + { + "epoch": 2.97, + "learning_rate": 5.996233858640565e-09, + "loss": 1.1201, + "step": 35250 + }, + { + "epoch": 2.97, + "learning_rate": 5.964801998322722e-09, + "loss": 1.1085, + "step": 35251 + }, + { + "epoch": 2.97, + "learning_rate": 5.933452711921472e-09, + "loss": 1.0747, + "step": 35252 + }, + { + "epoch": 2.97, + "learning_rate": 5.902185999697718e-09, + "loss": 1.0095, + "step": 35253 + }, + { + "epoch": 2.97, + "learning_rate": 5.8710018619079216e-09, + "loss": 1.2074, + "step": 35254 + }, + { + "epoch": 2.97, + "learning_rate": 5.839900298809653e-09, + "loss": 1.1431, + "step": 35255 + }, + { + "epoch": 2.97, + "learning_rate": 5.808881310661596e-09, + "loss": 1.1989, + "step": 35256 + }, + { + "epoch": 2.97, + "learning_rate": 5.777944897719101e-09, + "loss": 1.1556, + "step": 35257 + }, + { + "epoch": 2.97, + "learning_rate": 5.74709106023752e-09, + "loss": 1.069, + "step": 35258 + }, + { + "epoch": 2.97, + "learning_rate": 5.716319798472203e-09, + "loss": 1.0879, + "step": 35259 + }, + { + "epoch": 2.97, + "learning_rate": 5.685631112677392e-09, + "loss": 1.1475, + "step": 35260 + }, + { + "epoch": 2.97, + "learning_rate": 5.6550250031062184e-09, + "loss": 1.2021, + "step": 35261 + }, + { + "epoch": 2.97, + "learning_rate": 5.624501470011812e-09, + "loss": 1.173, + "step": 35262 + }, + { + "epoch": 2.97, + "learning_rate": 5.5940605136461935e-09, + "loss": 1.0775, + "step": 35263 + }, + { + "epoch": 2.97, + "learning_rate": 5.563702134261384e-09, + "loss": 1.0797, + "step": 35264 + }, + { + "epoch": 2.97, + "learning_rate": 5.5334263321071834e-09, + "loss": 1.1136, + "step": 35265 + }, + { + "epoch": 2.97, + "learning_rate": 5.503233107435613e-09, + "loss": 1.119, + "step": 35266 + }, + { + "epoch": 2.97, + "learning_rate": 5.473122460494252e-09, + "loss": 1.046, + "step": 35267 + }, + { + "epoch": 2.97, + "learning_rate": 5.4430943915329e-09, + "loss": 1.0923, + "step": 35268 + }, + { + "epoch": 2.97, + "learning_rate": 5.4131489007991364e-09, + "loss": 1.1786, + "step": 35269 + }, + { + "epoch": 2.97, + "learning_rate": 5.383285988540543e-09, + "loss": 1.2361, + "step": 35270 + }, + { + "epoch": 2.97, + "learning_rate": 5.353505655004698e-09, + "loss": 1.1155, + "step": 35271 + }, + { + "epoch": 2.97, + "learning_rate": 5.323807900435851e-09, + "loss": 1.1411, + "step": 35272 + }, + { + "epoch": 2.97, + "learning_rate": 5.294192725080471e-09, + "loss": 1.0724, + "step": 35273 + }, + { + "epoch": 2.97, + "learning_rate": 5.264660129183918e-09, + "loss": 1.2203, + "step": 35274 + }, + { + "epoch": 2.97, + "learning_rate": 5.235210112989331e-09, + "loss": 1.1069, + "step": 35275 + }, + { + "epoch": 2.97, + "learning_rate": 5.205842676739847e-09, + "loss": 1.0056, + "step": 35276 + }, + { + "epoch": 2.97, + "learning_rate": 5.176557820678607e-09, + "loss": 1.0432, + "step": 35277 + }, + { + "epoch": 2.97, + "learning_rate": 5.147355545047639e-09, + "loss": 1.1444, + "step": 35278 + }, + { + "epoch": 2.97, + "learning_rate": 5.11823585008675e-09, + "loss": 1.1171, + "step": 35279 + }, + { + "epoch": 2.97, + "learning_rate": 5.08919873603908e-09, + "loss": 1.0248, + "step": 35280 + }, + { + "epoch": 2.97, + "learning_rate": 5.060244203143327e-09, + "loss": 1.2558, + "step": 35281 + }, + { + "epoch": 2.97, + "learning_rate": 5.031372251638189e-09, + "loss": 1.1794, + "step": 35282 + }, + { + "epoch": 2.97, + "learning_rate": 5.002582881762364e-09, + "loss": 1.0229, + "step": 35283 + }, + { + "epoch": 2.97, + "learning_rate": 4.973876093754548e-09, + "loss": 1.129, + "step": 35284 + }, + { + "epoch": 2.97, + "learning_rate": 4.945251887851221e-09, + "loss": 1.0699, + "step": 35285 + }, + { + "epoch": 2.97, + "learning_rate": 4.91671026428886e-09, + "loss": 1.1172, + "step": 35286 + }, + { + "epoch": 2.97, + "learning_rate": 4.888251223302831e-09, + "loss": 1.2004, + "step": 35287 + }, + { + "epoch": 2.97, + "learning_rate": 4.859874765129613e-09, + "loss": 1.0822, + "step": 35288 + }, + { + "epoch": 2.97, + "learning_rate": 4.831580890003462e-09, + "loss": 1.0591, + "step": 35289 + }, + { + "epoch": 2.97, + "learning_rate": 4.803369598157526e-09, + "loss": 1.0335, + "step": 35290 + }, + { + "epoch": 2.97, + "learning_rate": 4.77524088982384e-09, + "loss": 1.1423, + "step": 35291 + }, + { + "epoch": 2.97, + "learning_rate": 4.747194765236662e-09, + "loss": 1.1162, + "step": 35292 + }, + { + "epoch": 2.97, + "learning_rate": 4.719231224628029e-09, + "loss": 1.0967, + "step": 35293 + }, + { + "epoch": 2.97, + "learning_rate": 4.6913502682266465e-09, + "loss": 1.1044, + "step": 35294 + }, + { + "epoch": 2.97, + "learning_rate": 4.663551896264551e-09, + "loss": 1.0581, + "step": 35295 + }, + { + "epoch": 2.97, + "learning_rate": 4.635836108971559e-09, + "loss": 1.0409, + "step": 35296 + }, + { + "epoch": 2.97, + "learning_rate": 4.608202906576375e-09, + "loss": 1.066, + "step": 35297 + }, + { + "epoch": 2.97, + "learning_rate": 4.580652289306598e-09, + "loss": 1.1659, + "step": 35298 + }, + { + "epoch": 2.97, + "learning_rate": 4.55318425739093e-09, + "loss": 1.1423, + "step": 35299 + }, + { + "epoch": 2.97, + "learning_rate": 4.525798811054749e-09, + "loss": 1.1587, + "step": 35300 + }, + { + "epoch": 2.97, + "learning_rate": 4.49849595052676e-09, + "loss": 1.1332, + "step": 35301 + }, + { + "epoch": 2.97, + "learning_rate": 4.4712756760301175e-09, + "loss": 1.0445, + "step": 35302 + }, + { + "epoch": 2.97, + "learning_rate": 4.444137987791308e-09, + "loss": 1.2157, + "step": 35303 + }, + { + "epoch": 2.97, + "learning_rate": 4.417082886033485e-09, + "loss": 1.0741, + "step": 35304 + }, + { + "epoch": 2.97, + "learning_rate": 4.3901103709820256e-09, + "loss": 1.1199, + "step": 35305 + }, + { + "epoch": 2.97, + "learning_rate": 4.363220442856753e-09, + "loss": 1.109, + "step": 35306 + }, + { + "epoch": 2.97, + "learning_rate": 4.336413101883041e-09, + "loss": 1.1932, + "step": 35307 + }, + { + "epoch": 2.97, + "learning_rate": 4.309688348279606e-09, + "loss": 1.1284, + "step": 35308 + }, + { + "epoch": 2.97, + "learning_rate": 4.283046182269601e-09, + "loss": 1.0368, + "step": 35309 + }, + { + "epoch": 2.97, + "learning_rate": 4.256486604071741e-09, + "loss": 1.1843, + "step": 35310 + }, + { + "epoch": 2.97, + "learning_rate": 4.23000961390585e-09, + "loss": 1.0693, + "step": 35311 + }, + { + "epoch": 2.97, + "learning_rate": 4.203615211990641e-09, + "loss": 1.0043, + "step": 35312 + }, + { + "epoch": 2.97, + "learning_rate": 4.1773033985437195e-09, + "loss": 1.0701, + "step": 35313 + }, + { + "epoch": 2.97, + "learning_rate": 4.1510741737837975e-09, + "loss": 1.0855, + "step": 35314 + }, + { + "epoch": 2.97, + "learning_rate": 4.12492753792626e-09, + "loss": 1.1221, + "step": 35315 + }, + { + "epoch": 2.97, + "learning_rate": 4.0988634911875994e-09, + "loss": 1.267, + "step": 35316 + }, + { + "epoch": 2.97, + "learning_rate": 4.0728820337843085e-09, + "loss": 1.1248, + "step": 35317 + }, + { + "epoch": 2.97, + "learning_rate": 4.046983165928442e-09, + "loss": 1.1121, + "step": 35318 + }, + { + "epoch": 2.97, + "learning_rate": 4.021166887836491e-09, + "loss": 1.1374, + "step": 35319 + }, + { + "epoch": 2.97, + "learning_rate": 3.995433199720511e-09, + "loss": 1.1423, + "step": 35320 + }, + { + "epoch": 2.97, + "learning_rate": 3.969782101793662e-09, + "loss": 1.0946, + "step": 35321 + }, + { + "epoch": 2.97, + "learning_rate": 3.944213594266888e-09, + "loss": 1.1052, + "step": 35322 + }, + { + "epoch": 2.97, + "learning_rate": 3.918727677353351e-09, + "loss": 1.15, + "step": 35323 + }, + { + "epoch": 2.97, + "learning_rate": 3.8933243512617735e-09, + "loss": 1.1385, + "step": 35324 + }, + { + "epoch": 2.97, + "learning_rate": 3.868003616201987e-09, + "loss": 1.1233, + "step": 35325 + }, + { + "epoch": 2.97, + "learning_rate": 3.842765472384935e-09, + "loss": 1.0598, + "step": 35326 + }, + { + "epoch": 2.97, + "learning_rate": 3.817609920018228e-09, + "loss": 1.1089, + "step": 35327 + }, + { + "epoch": 2.97, + "learning_rate": 3.7925369593083686e-09, + "loss": 1.0827, + "step": 35328 + }, + { + "epoch": 2.97, + "learning_rate": 3.767546590465188e-09, + "loss": 1.1402, + "step": 35329 + }, + { + "epoch": 2.97, + "learning_rate": 3.742638813692967e-09, + "loss": 1.0814, + "step": 35330 + }, + { + "epoch": 2.97, + "learning_rate": 3.7178136291982083e-09, + "loss": 1.1243, + "step": 35331 + }, + { + "epoch": 2.97, + "learning_rate": 3.693071037186302e-09, + "loss": 1.1169, + "step": 35332 + }, + { + "epoch": 2.97, + "learning_rate": 3.66841103786153e-09, + "loss": 1.0848, + "step": 35333 + }, + { + "epoch": 2.97, + "learning_rate": 3.643833631428173e-09, + "loss": 1.1371, + "step": 35334 + }, + { + "epoch": 2.97, + "learning_rate": 3.6193388180882915e-09, + "loss": 1.0638, + "step": 35335 + }, + { + "epoch": 2.97, + "learning_rate": 3.5949265980439464e-09, + "loss": 1.0538, + "step": 35336 + }, + { + "epoch": 2.98, + "learning_rate": 3.570596971498308e-09, + "loss": 1.1409, + "step": 35337 + }, + { + "epoch": 2.98, + "learning_rate": 3.5463499386512167e-09, + "loss": 1.0609, + "step": 35338 + }, + { + "epoch": 2.98, + "learning_rate": 3.5221854997036233e-09, + "loss": 1.0829, + "step": 35339 + }, + { + "epoch": 2.98, + "learning_rate": 3.498103654855367e-09, + "loss": 1.1032, + "step": 35340 + }, + { + "epoch": 2.98, + "learning_rate": 3.4741044043051785e-09, + "loss": 1.0826, + "step": 35341 + }, + { + "epoch": 2.98, + "learning_rate": 3.4501877482506774e-09, + "loss": 1.1368, + "step": 35342 + }, + { + "epoch": 2.98, + "learning_rate": 3.4263536868894832e-09, + "loss": 1.1434, + "step": 35343 + }, + { + "epoch": 2.98, + "learning_rate": 3.402602220420326e-09, + "loss": 1.0496, + "step": 35344 + }, + { + "epoch": 2.98, + "learning_rate": 3.378933349037494e-09, + "loss": 1.1352, + "step": 35345 + }, + { + "epoch": 2.98, + "learning_rate": 3.3553470729374982e-09, + "loss": 1.0569, + "step": 35346 + }, + { + "epoch": 2.98, + "learning_rate": 3.3318433923146267e-09, + "loss": 1.0802, + "step": 35347 + }, + { + "epoch": 2.98, + "learning_rate": 3.3084223073642785e-09, + "loss": 1.1076, + "step": 35348 + }, + { + "epoch": 2.98, + "learning_rate": 3.2850838182785226e-09, + "loss": 1.1141, + "step": 35349 + }, + { + "epoch": 2.98, + "learning_rate": 3.261827925250538e-09, + "loss": 1.0761, + "step": 35350 + }, + { + "epoch": 2.98, + "learning_rate": 3.238654628472393e-09, + "loss": 1.0389, + "step": 35351 + }, + { + "epoch": 2.98, + "learning_rate": 3.215563928136156e-09, + "loss": 1.1908, + "step": 35352 + }, + { + "epoch": 2.98, + "learning_rate": 3.192555824432786e-09, + "loss": 1.0348, + "step": 35353 + }, + { + "epoch": 2.98, + "learning_rate": 3.1696303175521303e-09, + "loss": 1.0764, + "step": 35354 + }, + { + "epoch": 2.98, + "learning_rate": 3.1467874076829276e-09, + "loss": 1.0345, + "step": 35355 + }, + { + "epoch": 2.98, + "learning_rate": 3.1240270950139152e-09, + "loss": 1.0838, + "step": 35356 + }, + { + "epoch": 2.98, + "learning_rate": 3.101349379733831e-09, + "loss": 1.0916, + "step": 35357 + }, + { + "epoch": 2.98, + "learning_rate": 3.0787542620303035e-09, + "loss": 1.1655, + "step": 35358 + }, + { + "epoch": 2.98, + "learning_rate": 3.056241742088739e-09, + "loss": 1.1512, + "step": 35359 + }, + { + "epoch": 2.98, + "learning_rate": 3.0338118200967658e-09, + "loss": 1.146, + "step": 35360 + }, + { + "epoch": 2.98, + "learning_rate": 3.0114644962386808e-09, + "loss": 1.0696, + "step": 35361 + }, + { + "epoch": 2.98, + "learning_rate": 2.9891997706987807e-09, + "loss": 1.0542, + "step": 35362 + }, + { + "epoch": 2.98, + "learning_rate": 2.967017643662473e-09, + "loss": 1.1021, + "step": 35363 + }, + { + "epoch": 2.98, + "learning_rate": 2.9449181153118345e-09, + "loss": 1.1155, + "step": 35364 + }, + { + "epoch": 2.98, + "learning_rate": 2.9229011858289414e-09, + "loss": 1.0804, + "step": 35365 + }, + { + "epoch": 2.98, + "learning_rate": 2.9009668553969804e-09, + "loss": 1.1036, + "step": 35366 + }, + { + "epoch": 2.98, + "learning_rate": 2.8791151241969184e-09, + "loss": 1.0371, + "step": 35367 + }, + { + "epoch": 2.98, + "learning_rate": 2.8573459924097215e-09, + "loss": 1.0473, + "step": 35368 + }, + { + "epoch": 2.98, + "learning_rate": 2.835659460213025e-09, + "loss": 1.0965, + "step": 35369 + }, + { + "epoch": 2.98, + "learning_rate": 2.814055527788906e-09, + "loss": 1.1067, + "step": 35370 + }, + { + "epoch": 2.98, + "learning_rate": 2.7925341953138897e-09, + "loss": 1.1179, + "step": 35371 + }, + { + "epoch": 2.98, + "learning_rate": 2.771095462966722e-09, + "loss": 1.1523, + "step": 35372 + }, + { + "epoch": 2.98, + "learning_rate": 2.7497393309250387e-09, + "loss": 1.163, + "step": 35373 + }, + { + "epoch": 2.98, + "learning_rate": 2.7284657993631446e-09, + "loss": 1.0594, + "step": 35374 + }, + { + "epoch": 2.98, + "learning_rate": 2.7072748684586758e-09, + "loss": 1.1277, + "step": 35375 + }, + { + "epoch": 2.98, + "learning_rate": 2.6861665383870474e-09, + "loss": 1.145, + "step": 35376 + }, + { + "epoch": 2.98, + "learning_rate": 2.6651408093214537e-09, + "loss": 1.0215, + "step": 35377 + }, + { + "epoch": 2.98, + "learning_rate": 2.64419768143509e-09, + "loss": 1.1033, + "step": 35378 + }, + { + "epoch": 2.98, + "learning_rate": 2.6233371549033717e-09, + "loss": 1.0775, + "step": 35379 + }, + { + "epoch": 2.98, + "learning_rate": 2.602559229896162e-09, + "loss": 1.1126, + "step": 35380 + }, + { + "epoch": 2.98, + "learning_rate": 2.581863906586657e-09, + "loss": 1.0898, + "step": 35381 + }, + { + "epoch": 2.98, + "learning_rate": 2.5612511851447196e-09, + "loss": 1.1043, + "step": 35382 + }, + { + "epoch": 2.98, + "learning_rate": 2.540721065742435e-09, + "loss": 1.0401, + "step": 35383 + }, + { + "epoch": 2.98, + "learning_rate": 2.520273548547447e-09, + "loss": 1.0498, + "step": 35384 + }, + { + "epoch": 2.98, + "learning_rate": 2.49990863372962e-09, + "loss": 1.1394, + "step": 35385 + }, + { + "epoch": 2.98, + "learning_rate": 2.4796263214577067e-09, + "loss": 1.1445, + "step": 35386 + }, + { + "epoch": 2.98, + "learning_rate": 2.4594266118982414e-09, + "loss": 1.1323, + "step": 35387 + }, + { + "epoch": 2.98, + "learning_rate": 2.4393095052188675e-09, + "loss": 1.0564, + "step": 35388 + }, + { + "epoch": 2.98, + "learning_rate": 2.4192750015850088e-09, + "loss": 1.1783, + "step": 35389 + }, + { + "epoch": 2.98, + "learning_rate": 2.399323101163198e-09, + "loss": 1.0917, + "step": 35390 + }, + { + "epoch": 2.98, + "learning_rate": 2.3794538041177486e-09, + "loss": 1.0841, + "step": 35391 + }, + { + "epoch": 2.98, + "learning_rate": 2.3596671106129732e-09, + "loss": 1.1723, + "step": 35392 + }, + { + "epoch": 2.98, + "learning_rate": 2.339963020813185e-09, + "loss": 1.1567, + "step": 35393 + }, + { + "epoch": 2.98, + "learning_rate": 2.3203415348793667e-09, + "loss": 1.0736, + "step": 35394 + }, + { + "epoch": 2.98, + "learning_rate": 2.3008026529736107e-09, + "loss": 1.1525, + "step": 35395 + }, + { + "epoch": 2.98, + "learning_rate": 2.2813463752602292e-09, + "loss": 1.1002, + "step": 35396 + }, + { + "epoch": 2.98, + "learning_rate": 2.261972701896875e-09, + "loss": 1.0982, + "step": 35397 + }, + { + "epoch": 2.98, + "learning_rate": 2.24268163304453e-09, + "loss": 1.1626, + "step": 35398 + }, + { + "epoch": 2.98, + "learning_rate": 2.223473168863066e-09, + "loss": 1.1375, + "step": 35399 + }, + { + "epoch": 2.98, + "learning_rate": 2.2043473095112454e-09, + "loss": 1.1467, + "step": 35400 + }, + { + "epoch": 2.98, + "learning_rate": 2.1853040551467198e-09, + "loss": 1.1486, + "step": 35401 + }, + { + "epoch": 2.98, + "learning_rate": 2.1663434059271404e-09, + "loss": 1.0788, + "step": 35402 + }, + { + "epoch": 2.98, + "learning_rate": 2.147465362009049e-09, + "loss": 1.179, + "step": 35403 + }, + { + "epoch": 2.98, + "learning_rate": 2.128669923547877e-09, + "loss": 1.0753, + "step": 35404 + }, + { + "epoch": 2.98, + "learning_rate": 2.1099570906990554e-09, + "loss": 1.0886, + "step": 35405 + }, + { + "epoch": 2.98, + "learning_rate": 2.091326863618015e-09, + "loss": 1.0854, + "step": 35406 + }, + { + "epoch": 2.98, + "learning_rate": 2.0727792424590776e-09, + "loss": 1.0813, + "step": 35407 + }, + { + "epoch": 2.98, + "learning_rate": 2.054314227373233e-09, + "loss": 1.1527, + "step": 35408 + }, + { + "epoch": 2.98, + "learning_rate": 2.0359318185148024e-09, + "loss": 1.141, + "step": 35409 + }, + { + "epoch": 2.98, + "learning_rate": 2.0176320160347762e-09, + "loss": 1.0866, + "step": 35410 + }, + { + "epoch": 2.98, + "learning_rate": 1.999414820085255e-09, + "loss": 1.1381, + "step": 35411 + }, + { + "epoch": 2.98, + "learning_rate": 1.9812802308172284e-09, + "loss": 1.0197, + "step": 35412 + }, + { + "epoch": 2.98, + "learning_rate": 1.963228248378357e-09, + "loss": 1.0812, + "step": 35413 + }, + { + "epoch": 2.98, + "learning_rate": 1.945258872919631e-09, + "loss": 1.184, + "step": 35414 + }, + { + "epoch": 2.98, + "learning_rate": 1.9273721045887096e-09, + "loss": 1.0767, + "step": 35415 + }, + { + "epoch": 2.98, + "learning_rate": 1.909567943534363e-09, + "loss": 1.0772, + "step": 35416 + }, + { + "epoch": 2.98, + "learning_rate": 1.89184638990203e-09, + "loss": 1.1883, + "step": 35417 + }, + { + "epoch": 2.98, + "learning_rate": 1.874207443839371e-09, + "loss": 1.1839, + "step": 35418 + }, + { + "epoch": 2.98, + "learning_rate": 1.8566511054918247e-09, + "loss": 1.1165, + "step": 35419 + }, + { + "epoch": 2.98, + "learning_rate": 1.8391773750048304e-09, + "loss": 1.1458, + "step": 35420 + }, + { + "epoch": 2.98, + "learning_rate": 1.8217862525216068e-09, + "loss": 1.1733, + "step": 35421 + }, + { + "epoch": 2.98, + "learning_rate": 1.8044777381875933e-09, + "loss": 1.1037, + "step": 35422 + }, + { + "epoch": 2.98, + "learning_rate": 1.7872518321437882e-09, + "loss": 1.1756, + "step": 35423 + }, + { + "epoch": 2.98, + "learning_rate": 1.7701085345345204e-09, + "loss": 1.1714, + "step": 35424 + }, + { + "epoch": 2.98, + "learning_rate": 1.753047845499678e-09, + "loss": 1.1397, + "step": 35425 + }, + { + "epoch": 2.98, + "learning_rate": 1.7360697651813696e-09, + "loss": 1.0827, + "step": 35426 + }, + { + "epoch": 2.98, + "learning_rate": 1.7191742937194832e-09, + "loss": 1.1686, + "step": 35427 + }, + { + "epoch": 2.98, + "learning_rate": 1.702361431252797e-09, + "loss": 1.105, + "step": 35428 + }, + { + "epoch": 2.98, + "learning_rate": 1.6856311779223088e-09, + "loss": 1.1059, + "step": 35429 + }, + { + "epoch": 2.98, + "learning_rate": 1.6689835338645766e-09, + "loss": 1.1343, + "step": 35430 + }, + { + "epoch": 2.98, + "learning_rate": 1.6524184992172676e-09, + "loss": 1.0615, + "step": 35431 + }, + { + "epoch": 2.98, + "learning_rate": 1.6359360741180497e-09, + "loss": 1.1064, + "step": 35432 + }, + { + "epoch": 2.98, + "learning_rate": 1.61953625870237e-09, + "loss": 1.2071, + "step": 35433 + }, + { + "epoch": 2.98, + "learning_rate": 1.6032190531056757e-09, + "loss": 1.1365, + "step": 35434 + }, + { + "epoch": 2.98, + "learning_rate": 1.586984457463414e-09, + "loss": 1.2151, + "step": 35435 + }, + { + "epoch": 2.98, + "learning_rate": 1.570832471909922e-09, + "loss": 1.0852, + "step": 35436 + }, + { + "epoch": 2.98, + "learning_rate": 1.5547630965773163e-09, + "loss": 1.1653, + "step": 35437 + }, + { + "epoch": 2.98, + "learning_rate": 1.5387763315999337e-09, + "loss": 1.1081, + "step": 35438 + }, + { + "epoch": 2.98, + "learning_rate": 1.5228721771076705e-09, + "loss": 1.0797, + "step": 35439 + }, + { + "epoch": 2.98, + "learning_rate": 1.5070506332348633e-09, + "loss": 1.1285, + "step": 35440 + }, + { + "epoch": 2.98, + "learning_rate": 1.4913117001102983e-09, + "loss": 1.0844, + "step": 35441 + }, + { + "epoch": 2.98, + "learning_rate": 1.4756553778638716e-09, + "loss": 1.0258, + "step": 35442 + }, + { + "epoch": 2.98, + "learning_rate": 1.4600816666265893e-09, + "loss": 1.1406, + "step": 35443 + }, + { + "epoch": 2.98, + "learning_rate": 1.444590566525017e-09, + "loss": 1.105, + "step": 35444 + }, + { + "epoch": 2.98, + "learning_rate": 1.429182077689051e-09, + "loss": 1.0943, + "step": 35445 + }, + { + "epoch": 2.98, + "learning_rate": 1.4138562002441458e-09, + "loss": 1.1702, + "step": 35446 + }, + { + "epoch": 2.98, + "learning_rate": 1.3986129343179777e-09, + "loss": 1.068, + "step": 35447 + }, + { + "epoch": 2.98, + "learning_rate": 1.383452280037112e-09, + "loss": 1.0974, + "step": 35448 + }, + { + "epoch": 2.98, + "learning_rate": 1.3683742375258934e-09, + "loss": 1.1304, + "step": 35449 + }, + { + "epoch": 2.98, + "learning_rate": 1.3533788069097775e-09, + "loss": 1.0402, + "step": 35450 + }, + { + "epoch": 2.98, + "learning_rate": 1.3384659883108887e-09, + "loss": 1.1032, + "step": 35451 + }, + { + "epoch": 2.98, + "learning_rate": 1.3236357818546818e-09, + "loss": 1.1538, + "step": 35452 + }, + { + "epoch": 2.98, + "learning_rate": 1.3088881876621717e-09, + "loss": 1.1668, + "step": 35453 + }, + { + "epoch": 2.98, + "learning_rate": 1.2942232058554826e-09, + "loss": 1.0633, + "step": 35454 + }, + { + "epoch": 2.99, + "learning_rate": 1.279640836555629e-09, + "loss": 1.1806, + "step": 35455 + }, + { + "epoch": 2.99, + "learning_rate": 1.265141079883625e-09, + "loss": 1.1625, + "step": 35456 + }, + { + "epoch": 2.99, + "learning_rate": 1.2507239359582646e-09, + "loss": 1.1048, + "step": 35457 + }, + { + "epoch": 2.99, + "learning_rate": 1.236389404900562e-09, + "loss": 1.1152, + "step": 35458 + }, + { + "epoch": 2.99, + "learning_rate": 1.2221374868270907e-09, + "loss": 0.9919, + "step": 35459 + }, + { + "epoch": 2.99, + "learning_rate": 1.2079681818577549e-09, + "loss": 1.0655, + "step": 35460 + }, + { + "epoch": 2.99, + "learning_rate": 1.1938814901069073e-09, + "loss": 1.2088, + "step": 35461 + }, + { + "epoch": 2.99, + "learning_rate": 1.1798774116922317e-09, + "loss": 1.13, + "step": 35462 + }, + { + "epoch": 2.99, + "learning_rate": 1.1659559467291915e-09, + "loss": 1.0733, + "step": 35463 + }, + { + "epoch": 2.99, + "learning_rate": 1.1521170953343596e-09, + "loss": 1.1333, + "step": 35464 + }, + { + "epoch": 2.99, + "learning_rate": 1.138360857619869e-09, + "loss": 1.0874, + "step": 35465 + }, + { + "epoch": 2.99, + "learning_rate": 1.1246872337000724e-09, + "loss": 1.0973, + "step": 35466 + }, + { + "epoch": 2.99, + "learning_rate": 1.111096223689323e-09, + "loss": 1.1033, + "step": 35467 + }, + { + "epoch": 2.99, + "learning_rate": 1.0975878276975326e-09, + "loss": 1.0598, + "step": 35468 + }, + { + "epoch": 2.99, + "learning_rate": 1.0841620458379443e-09, + "loss": 1.0592, + "step": 35469 + }, + { + "epoch": 2.99, + "learning_rate": 1.0708188782204698e-09, + "loss": 1.0046, + "step": 35470 + }, + { + "epoch": 2.99, + "learning_rate": 1.0575583249572419e-09, + "loss": 1.1638, + "step": 35471 + }, + { + "epoch": 2.99, + "learning_rate": 1.044380386155952e-09, + "loss": 1.1036, + "step": 35472 + }, + { + "epoch": 2.99, + "learning_rate": 1.0312850619254022e-09, + "loss": 1.1115, + "step": 35473 + }, + { + "epoch": 2.99, + "learning_rate": 1.0182723523755045e-09, + "loss": 1.1741, + "step": 35474 + }, + { + "epoch": 2.99, + "learning_rate": 1.00534225761173e-09, + "loss": 1.2295, + "step": 35475 + }, + { + "epoch": 2.99, + "learning_rate": 9.924947777428807e-10, + "loss": 1.086, + "step": 35476 + }, + { + "epoch": 2.99, + "learning_rate": 9.797299128733172e-10, + "loss": 1.216, + "step": 35477 + }, + { + "epoch": 2.99, + "learning_rate": 9.670476631096215e-10, + "loss": 1.15, + "step": 35478 + }, + { + "epoch": 2.99, + "learning_rate": 9.544480285561542e-10, + "loss": 1.2009, + "step": 35479 + }, + { + "epoch": 2.99, + "learning_rate": 9.419310093172762e-10, + "loss": 1.0467, + "step": 35480 + }, + { + "epoch": 2.99, + "learning_rate": 9.294966054962384e-10, + "loss": 1.0805, + "step": 35481 + }, + { + "epoch": 2.99, + "learning_rate": 9.171448171962915e-10, + "loss": 1.1968, + "step": 35482 + }, + { + "epoch": 2.99, + "learning_rate": 9.048756445184659e-10, + "loss": 1.0984, + "step": 35483 + }, + { + "epoch": 2.99, + "learning_rate": 8.926890875649019e-10, + "loss": 1.0667, + "step": 35484 + }, + { + "epoch": 2.99, + "learning_rate": 8.805851464355198e-10, + "loss": 1.0949, + "step": 35485 + }, + { + "epoch": 2.99, + "learning_rate": 8.685638212313496e-10, + "loss": 1.0897, + "step": 35486 + }, + { + "epoch": 2.99, + "learning_rate": 8.566251120512014e-10, + "loss": 1.0859, + "step": 35487 + }, + { + "epoch": 2.99, + "learning_rate": 8.447690189938851e-10, + "loss": 1.1838, + "step": 35488 + }, + { + "epoch": 2.99, + "learning_rate": 8.329955421571001e-10, + "loss": 1.0806, + "step": 35489 + }, + { + "epoch": 2.99, + "learning_rate": 8.213046816385462e-10, + "loss": 1.1065, + "step": 35490 + }, + { + "epoch": 2.99, + "learning_rate": 8.096964375337024e-10, + "loss": 1.1311, + "step": 35491 + }, + { + "epoch": 2.99, + "learning_rate": 7.981708099402685e-10, + "loss": 1.1183, + "step": 35492 + }, + { + "epoch": 2.99, + "learning_rate": 7.867277989526134e-10, + "loss": 1.1125, + "step": 35493 + }, + { + "epoch": 2.99, + "learning_rate": 7.753674046639958e-10, + "loss": 1.1993, + "step": 35494 + }, + { + "epoch": 2.99, + "learning_rate": 7.64089627169895e-10, + "loss": 1.0832, + "step": 35495 + }, + { + "epoch": 2.99, + "learning_rate": 7.528944665635696e-10, + "loss": 1.1234, + "step": 35496 + }, + { + "epoch": 2.99, + "learning_rate": 7.417819229360579e-10, + "loss": 1.1366, + "step": 35497 + }, + { + "epoch": 2.99, + "learning_rate": 7.307519963817289e-10, + "loss": 1.0368, + "step": 35498 + }, + { + "epoch": 2.99, + "learning_rate": 7.198046869882902e-10, + "loss": 1.1003, + "step": 35499 + }, + { + "epoch": 2.99, + "learning_rate": 7.089399948490006e-10, + "loss": 1.0805, + "step": 35500 + }, + { + "epoch": 2.99, + "learning_rate": 6.981579200526778e-10, + "loss": 1.1243, + "step": 35501 + }, + { + "epoch": 2.99, + "learning_rate": 6.874584626881398e-10, + "loss": 1.0007, + "step": 35502 + }, + { + "epoch": 2.99, + "learning_rate": 6.768416228442043e-10, + "loss": 1.1527, + "step": 35503 + }, + { + "epoch": 2.99, + "learning_rate": 6.663074006074688e-10, + "loss": 1.0724, + "step": 35504 + }, + { + "epoch": 2.99, + "learning_rate": 6.558557960667511e-10, + "loss": 1.1559, + "step": 35505 + }, + { + "epoch": 2.99, + "learning_rate": 6.454868093075383e-10, + "loss": 1.1704, + "step": 35506 + }, + { + "epoch": 2.99, + "learning_rate": 6.352004404153178e-10, + "loss": 1.1218, + "step": 35507 + }, + { + "epoch": 2.99, + "learning_rate": 6.249966894755765e-10, + "loss": 1.0934, + "step": 35508 + }, + { + "epoch": 2.99, + "learning_rate": 6.148755565715813e-10, + "loss": 1.0881, + "step": 35509 + }, + { + "epoch": 2.99, + "learning_rate": 6.048370417888194e-10, + "loss": 1.0602, + "step": 35510 + }, + { + "epoch": 2.99, + "learning_rate": 5.94881145208337e-10, + "loss": 0.9904, + "step": 35511 + }, + { + "epoch": 2.99, + "learning_rate": 5.850078669134007e-10, + "loss": 1.0644, + "step": 35512 + }, + { + "epoch": 2.99, + "learning_rate": 5.75217206985057e-10, + "loss": 1.0431, + "step": 35513 + }, + { + "epoch": 2.99, + "learning_rate": 5.655091655054623e-10, + "loss": 1.0879, + "step": 35514 + }, + { + "epoch": 2.99, + "learning_rate": 5.558837425523322e-10, + "loss": 1.0772, + "step": 35515 + }, + { + "epoch": 2.99, + "learning_rate": 5.463409382078233e-10, + "loss": 1.039, + "step": 35516 + }, + { + "epoch": 2.99, + "learning_rate": 5.368807525496511e-10, + "loss": 1.1073, + "step": 35517 + }, + { + "epoch": 2.99, + "learning_rate": 5.275031856555312e-10, + "loss": 1.1333, + "step": 35518 + }, + { + "epoch": 2.99, + "learning_rate": 5.182082376031794e-10, + "loss": 1.0606, + "step": 35519 + }, + { + "epoch": 2.99, + "learning_rate": 5.089959084703111e-10, + "loss": 1.0673, + "step": 35520 + }, + { + "epoch": 2.99, + "learning_rate": 4.998661983324215e-10, + "loss": 1.1853, + "step": 35521 + }, + { + "epoch": 2.99, + "learning_rate": 4.908191072638957e-10, + "loss": 1.1479, + "step": 35522 + }, + { + "epoch": 2.99, + "learning_rate": 4.818546353413389e-10, + "loss": 1.1082, + "step": 35523 + }, + { + "epoch": 2.99, + "learning_rate": 4.72972782638026e-10, + "loss": 1.1752, + "step": 35524 + }, + { + "epoch": 2.99, + "learning_rate": 4.6417354922723144e-10, + "loss": 1.0409, + "step": 35525 + }, + { + "epoch": 2.99, + "learning_rate": 4.5545693518112e-10, + "loss": 1.1911, + "step": 35526 + }, + { + "epoch": 2.99, + "learning_rate": 4.468229405729663e-10, + "loss": 1.1346, + "step": 35527 + }, + { + "epoch": 2.99, + "learning_rate": 4.382715654727143e-10, + "loss": 1.1308, + "step": 35528 + }, + { + "epoch": 2.99, + "learning_rate": 4.2980280995252863e-10, + "loss": 1.0716, + "step": 35529 + }, + { + "epoch": 2.99, + "learning_rate": 4.214166740812431e-10, + "loss": 1.0595, + "step": 35530 + }, + { + "epoch": 2.99, + "learning_rate": 4.131131579288017e-10, + "loss": 1.1196, + "step": 35531 + }, + { + "epoch": 2.99, + "learning_rate": 4.0489226156292806e-10, + "loss": 1.0952, + "step": 35532 + }, + { + "epoch": 2.99, + "learning_rate": 3.96753985052456e-10, + "loss": 1.0728, + "step": 35533 + }, + { + "epoch": 2.99, + "learning_rate": 3.8869832846399893e-10, + "loss": 1.0746, + "step": 35534 + }, + { + "epoch": 2.99, + "learning_rate": 3.8072529186417015e-10, + "loss": 1.1667, + "step": 35535 + }, + { + "epoch": 2.99, + "learning_rate": 3.728348753195832e-10, + "loss": 1.1093, + "step": 35536 + }, + { + "epoch": 2.99, + "learning_rate": 3.6502707889463084e-10, + "loss": 1.1138, + "step": 35537 + }, + { + "epoch": 2.99, + "learning_rate": 3.5730190265481636e-10, + "loss": 1.0916, + "step": 35538 + }, + { + "epoch": 2.99, + "learning_rate": 3.496593466623122e-10, + "loss": 1.1088, + "step": 35539 + }, + { + "epoch": 2.99, + "learning_rate": 3.420994109815112e-10, + "loss": 1.1357, + "step": 35540 + }, + { + "epoch": 2.99, + "learning_rate": 3.34622095674586e-10, + "loss": 1.1098, + "step": 35541 + }, + { + "epoch": 2.99, + "learning_rate": 3.2722740080370907e-10, + "loss": 1.1345, + "step": 35542 + }, + { + "epoch": 2.99, + "learning_rate": 3.199153264299426e-10, + "loss": 1.2114, + "step": 35543 + }, + { + "epoch": 2.99, + "learning_rate": 3.1268587261212847e-10, + "loss": 1.0896, + "step": 35544 + }, + { + "epoch": 2.99, + "learning_rate": 3.055390394124391e-10, + "loss": 1.0669, + "step": 35545 + }, + { + "epoch": 2.99, + "learning_rate": 2.9847482688749597e-10, + "loss": 1.1106, + "step": 35546 + }, + { + "epoch": 2.99, + "learning_rate": 2.9149323509725104e-10, + "loss": 1.0484, + "step": 35547 + }, + { + "epoch": 2.99, + "learning_rate": 2.8459426409943593e-10, + "loss": 1.0848, + "step": 35548 + }, + { + "epoch": 2.99, + "learning_rate": 2.77777913950672e-10, + "loss": 1.0486, + "step": 35549 + }, + { + "epoch": 2.99, + "learning_rate": 2.710441847064704e-10, + "loss": 1.1342, + "step": 35550 + }, + { + "epoch": 2.99, + "learning_rate": 2.643930764245628e-10, + "loss": 1.0711, + "step": 35551 + }, + { + "epoch": 2.99, + "learning_rate": 2.578245891571296e-10, + "loss": 1.176, + "step": 35552 + }, + { + "epoch": 2.99, + "learning_rate": 2.513387229607922e-10, + "loss": 1.071, + "step": 35553 + }, + { + "epoch": 2.99, + "learning_rate": 2.449354778877311e-10, + "loss": 1.1846, + "step": 35554 + }, + { + "epoch": 2.99, + "learning_rate": 2.3861485399123694e-10, + "loss": 1.067, + "step": 35555 + }, + { + "epoch": 2.99, + "learning_rate": 2.3237685132460054e-10, + "loss": 1.1455, + "step": 35556 + }, + { + "epoch": 2.99, + "learning_rate": 2.2622146993778183e-10, + "loss": 1.1608, + "step": 35557 + }, + { + "epoch": 2.99, + "learning_rate": 2.201487098818511e-10, + "loss": 1.0622, + "step": 35558 + }, + { + "epoch": 2.99, + "learning_rate": 2.1415857120787865e-10, + "loss": 1.1925, + "step": 35559 + }, + { + "epoch": 2.99, + "learning_rate": 2.0825105396471424e-10, + "loss": 1.0875, + "step": 35560 + }, + { + "epoch": 2.99, + "learning_rate": 2.024261582012077e-10, + "loss": 1.1908, + "step": 35561 + }, + { + "epoch": 2.99, + "learning_rate": 1.9668388396509864e-10, + "loss": 1.0633, + "step": 35562 + }, + { + "epoch": 2.99, + "learning_rate": 1.9102423130523683e-10, + "loss": 1.1077, + "step": 35563 + }, + { + "epoch": 2.99, + "learning_rate": 1.8544720026714147e-10, + "loss": 1.1644, + "step": 35564 + }, + { + "epoch": 2.99, + "learning_rate": 1.7995279089744188e-10, + "loss": 1.0424, + "step": 35565 + }, + { + "epoch": 2.99, + "learning_rate": 1.7454100324165724e-10, + "loss": 1.0779, + "step": 35566 + }, + { + "epoch": 2.99, + "learning_rate": 1.692118373430862e-10, + "loss": 1.1084, + "step": 35567 + }, + { + "epoch": 2.99, + "learning_rate": 1.6396529324724798e-10, + "loss": 0.9681, + "step": 35568 + }, + { + "epoch": 2.99, + "learning_rate": 1.588013709974412e-10, + "loss": 1.2241, + "step": 35569 + }, + { + "epoch": 2.99, + "learning_rate": 1.5372007063585437e-10, + "loss": 1.1895, + "step": 35570 + }, + { + "epoch": 2.99, + "learning_rate": 1.4872139220467597e-10, + "loss": 1.0388, + "step": 35571 + }, + { + "epoch": 2.99, + "learning_rate": 1.4380533574498424e-10, + "loss": 1.105, + "step": 35572 + }, + { + "epoch": 2.99, + "learning_rate": 1.3897190129785742e-10, + "loss": 1.0473, + "step": 35573 + }, + { + "epoch": 3.0, + "learning_rate": 1.3422108890215335e-10, + "loss": 1.0483, + "step": 35574 + }, + { + "epoch": 3.0, + "learning_rate": 1.2955289859895025e-10, + "loss": 1.1032, + "step": 35575 + }, + { + "epoch": 3.0, + "learning_rate": 1.2496733042488551e-10, + "loss": 1.177, + "step": 35576 + }, + { + "epoch": 3.0, + "learning_rate": 1.204643844188169e-10, + "loss": 1.0588, + "step": 35577 + }, + { + "epoch": 3.0, + "learning_rate": 1.1604406061849205e-10, + "loss": 1.0622, + "step": 35578 + }, + { + "epoch": 3.0, + "learning_rate": 1.1170635905832783e-10, + "loss": 1.073, + "step": 35579 + }, + { + "epoch": 3.0, + "learning_rate": 1.0745127977718206e-10, + "loss": 1.0428, + "step": 35580 + }, + { + "epoch": 3.0, + "learning_rate": 1.0327882280836144e-10, + "loss": 1.1122, + "step": 35581 + }, + { + "epoch": 3.0, + "learning_rate": 9.918898818628286e-11, + "loss": 1.0913, + "step": 35582 + }, + { + "epoch": 3.0, + "learning_rate": 9.518177594425305e-11, + "loss": 1.0254, + "step": 35583 + }, + { + "epoch": 3.0, + "learning_rate": 9.12571861177991e-11, + "loss": 1.0791, + "step": 35584 + }, + { + "epoch": 3.0, + "learning_rate": 8.741521873689707e-11, + "loss": 1.0643, + "step": 35585 + }, + { + "epoch": 3.0, + "learning_rate": 8.365587383374341e-11, + "loss": 1.1693, + "step": 35586 + }, + { + "epoch": 3.0, + "learning_rate": 7.997915144053459e-11, + "loss": 1.1501, + "step": 35587 + }, + { + "epoch": 3.0, + "learning_rate": 7.638505158724663e-11, + "loss": 1.0791, + "step": 35588 + }, + { + "epoch": 3.0, + "learning_rate": 7.287357430274534e-11, + "loss": 1.0954, + "step": 35589 + }, + { + "epoch": 3.0, + "learning_rate": 6.944471961589649e-11, + "loss": 1.0384, + "step": 35590 + }, + { + "epoch": 3.0, + "learning_rate": 6.609848755667613e-11, + "loss": 1.1275, + "step": 35591 + }, + { + "epoch": 3.0, + "learning_rate": 6.283487815172961e-11, + "loss": 1.0438, + "step": 35592 + }, + { + "epoch": 3.0, + "learning_rate": 5.965389142770228e-11, + "loss": 1.1154, + "step": 35593 + }, + { + "epoch": 3.0, + "learning_rate": 5.6555527411239484e-11, + "loss": 1.1465, + "step": 35594 + }, + { + "epoch": 3.0, + "learning_rate": 5.353978612787636e-11, + "loss": 1.1072, + "step": 35595 + }, + { + "epoch": 3.0, + "learning_rate": 5.060666760203781e-11, + "loss": 1.1406, + "step": 35596 + }, + { + "epoch": 3.0, + "learning_rate": 4.775617185925896e-11, + "loss": 1.0909, + "step": 35597 + }, + { + "epoch": 3.0, + "learning_rate": 4.498829892174428e-11, + "loss": 1.0535, + "step": 35598 + }, + { + "epoch": 3.0, + "learning_rate": 4.230304881280845e-11, + "loss": 1.1622, + "step": 35599 + }, + { + "epoch": 3.0, + "learning_rate": 3.970042155576614e-11, + "loss": 1.1623, + "step": 35600 + }, + { + "epoch": 3.0, + "learning_rate": 3.718041717060139e-11, + "loss": 1.1098, + "step": 35601 + }, + { + "epoch": 3.0, + "learning_rate": 3.474303567840842e-11, + "loss": 1.1687, + "step": 35602 + }, + { + "epoch": 3.0, + "learning_rate": 3.238827709917125e-11, + "loss": 1.127, + "step": 35603 + }, + { + "epoch": 3.0, + "learning_rate": 3.011614145398412e-11, + "loss": 1.0404, + "step": 35604 + }, + { + "epoch": 3.0, + "learning_rate": 2.792662875950036e-11, + "loss": 1.1029, + "step": 35605 + }, + { + "epoch": 3.0, + "learning_rate": 2.581973903459378e-11, + "loss": 1.1393, + "step": 35606 + }, + { + "epoch": 3.0, + "learning_rate": 2.3795472297027945e-11, + "loss": 1.1715, + "step": 35607 + }, + { + "epoch": 3.0, + "learning_rate": 2.1853828563456192e-11, + "loss": 1.1589, + "step": 35608 + }, + { + "epoch": 3.0, + "learning_rate": 1.999480784942165e-11, + "loss": 1.0503, + "step": 35609 + }, + { + "epoch": 3.0, + "learning_rate": 1.821841017157766e-11, + "loss": 1.1022, + "step": 35610 + }, + { + "epoch": 3.0, + "learning_rate": 1.652463554213668e-11, + "loss": 1.1427, + "step": 35611 + }, + { + "epoch": 3.0, + "learning_rate": 1.491348397775205e-11, + "loss": 1.1476, + "step": 35612 + }, + { + "epoch": 3.0, + "learning_rate": 1.3384955489526007e-11, + "loss": 1.1442, + "step": 35613 + }, + { + "epoch": 3.0, + "learning_rate": 1.1939050091891446e-11, + "loss": 1.1045, + "step": 35614 + }, + { + "epoch": 3.0, + "learning_rate": 1.0575767795950597e-11, + "loss": 1.0762, + "step": 35615 + }, + { + "epoch": 3.0, + "learning_rate": 9.295108612805693e-12, + "loss": 1.1167, + "step": 35616 + }, + { + "epoch": 3.0, + "learning_rate": 8.09707255355896e-12, + "loss": 1.1345, + "step": 35617 + }, + { + "epoch": 3.0, + "learning_rate": 6.981659627092185e-12, + "loss": 1.0887, + "step": 35618 + }, + { + "epoch": 3.0, + "learning_rate": 5.9488698433973755e-12, + "loss": 1.1519, + "step": 35619 + }, + { + "epoch": 3.0, + "learning_rate": 4.998703211356315e-12, + "loss": 1.1015, + "step": 35620 + }, + { + "epoch": 3.0, + "learning_rate": 4.131159738740565e-12, + "loss": 1.1194, + "step": 35621 + }, + { + "epoch": 3.0, + "learning_rate": 3.346239432211462e-12, + "loss": 1.0935, + "step": 35622 + }, + { + "epoch": 3.0, + "learning_rate": 2.6439422984303463e-12, + "loss": 1.1247, + "step": 35623 + }, + { + "epoch": 3.0, + "learning_rate": 2.0242683429483322e-12, + "loss": 1.0189, + "step": 35624 + }, + { + "epoch": 3.0, + "learning_rate": 1.4872175713165349e-12, + "loss": 1.1108, + "step": 35625 + }, + { + "epoch": 3.0, + "learning_rate": 1.0327899879758462e-12, + "loss": 1.17, + "step": 35626 + }, + { + "epoch": 3.0, + "learning_rate": 6.609855962569356e-13, + "loss": 1.0914, + "step": 35627 + }, + { + "epoch": 3.0, + "learning_rate": 3.7180439949047184e-13, + "loss": 1.1643, + "step": 35628 + }, + { + "epoch": 3.0, + "learning_rate": 1.652464010071242e-13, + "loss": 1.1687, + "step": 35629 + }, + { + "epoch": 3.0, + "learning_rate": 4.131159969666954e-14, + "loss": 1.0934, + "step": 35630 + }, + { + "epoch": 3.0, + "learning_rate": 0.0, + "loss": 1.1182, + "step": 35631 + }, + { + "epoch": 3.0, + "step": 35631, + "total_flos": 8.751355214138704e+18, + "train_loss": 1.2137726634144466, + "train_runtime": 49531.3358, + "train_samples_per_second": 86.328, + "train_steps_per_second": 0.719 + } + ], + "logging_steps": 1.0, + "max_steps": 35631, + "num_input_tokens_seen": 0, + "num_train_epochs": 3, + "save_steps": 10000, + "total_flos": 8.751355214138704e+18, + "train_batch_size": 5, + "trial_name": null, + "trial_params": null +}